Embed Size (px)
Citation preview
copy
New Phytologist
(2003)
161
235ndash252
wwwnewphytologistorg
235
Research
Blackwell Publishing Ltd
Convergent evolution of flower polymorphism in
Narcissus
(Amaryllidaceae)
Rociacuteo Peacuterez Pablo Vargas and Juan Arroyo
Departamento de Biologiacutea Vegetal y Ecologiacutea Universidad de Sevilla Apartado 1095 E-41080 Seville Spain and Royal Botanic Garden CSIC Madrid Spain
Summary
bull
We present new results on morphological variation for style polymorphism andperianth features in the seven species of
Narcissus
sect Apodanthi and in
N pallid-ulus
(sect Cyclaminei) which exhibits a wide array of flower conditions heterostylystyle dimorphism of variable reciprocity and style monomorphismbull There is a significant association between perianth morphology and style polymor-phism in the studied species Among significant flower features the two heterostylousspecies show pendulous flowers Most style-dimorphic species have patent flowerswhereas monomorphic and one style-dimorphic species display erect flowersbull Interpretation of a chloroplast (
trn
L-
trn
F) phylogeny of species representing mostsections in the genus leads us to conclude that the two heterostylous species haveindependent origins supporting previous hypothesis of convergence of heterostylyin
Narcissus
Remarkable resemblance of flower features between both species indi-cates that pollinator activity might have driven flower convergencebull A second chloroplast (
trn
T-
trn
L) phylogeny in sect Apodanthi helps elucidatesome evolutionary transitions in moulding heterostyly Stylar dimorphism appears toprecede distyly and style monomorphism Population studies on style dimorphicand monomorphic species of
Narcissus
are congruent with a pollinator-mediatedprocess also involved in a shift to monomorphism
Key words
chloroplast sequences evolutionary convergence heterostylyMediterranean morphometry plantndashpollinator interactions phylogeny
copy
New Phytologist
(2003)
161
235ndash252
Author for correspondence
Juan Arroyo
Tel +34 954557058
Fax +34 954557059
Email arroyouses
Received
30 August 2003
Accepted
16 October 2003
doi 101046j1469-8137200300955x
Introduction
Heterostyly is a genetically based discrete style polymorphismoriginated independently multiple times as indicated by itsoccurrence in at least 28 disparate families of angiosperms(Barrett
et al
2000) This polymorphism includes within thesame population distyly (two morphs) and tristyly (threemorphs) In both cases flower morphs are characterised by areciprocal positioning of stigmas and stamens in height(reciprocal herkogamy) In its typical form morphs are eitherlong-styled (L-morph) mid-styled (M-morph) or short-styled (S-morph) in tristylous species and L- and S-styled indistylous species Although recognition of style polymorphismand its evolutionary implications have been long discussed(Ornduff 1992) there is not yet a general agreement about itsorigin and evolution Several hypotheses on evolution ofheterostyly involve different scenarios and selective forces
which necessarily imply alternative initial and intermediatesteps (Charlesworth amp Charlesworth 1979 Lloyd amp Webb1992ab Richards 1998) Experimental approaches to testpredictions of evolutionary models build up a body ofknowledge in evolution of flower polymorphism (Kohn ampBarrett 1992 Stone amp Thomson 1994 Lau amp Bosque2003 Thompson
et al
2003) Comparative studies in thesame plant group with presumable intermediate stages offlower polymorphism are absolutely necessary to exploreevolutionary pathways to heterostyly (Lloyd amp Webb 1992a)
Lloyd amp Webbrsquos model (Lloyd amp Webb 1992ab) has gen-erated acceptance subject to experimental observational andcomparative evaluation (Faivre 2000 Nishihiro
et al
2000Faivre amp McDade 2001 Lau amp Bosque 2003) This modelassumes that the main selective force behind heterostyly is thepromotion of outcrossing through efficient pollen transferbetween flower morphs Lloyd amp Webbrsquos model (Lloyd amp
wwwnewphytologistorg
copy
New Phytologist
(2003)
161
235ndash252
Research236
Webb 1992a) emphasizes the role of a pollination environ-ment rather than the existence of a particular genetic incom-patibility system as considered instead by the Charlesworthamp Charlesworthrsquos model (Charlesworth amp Charlesworth 1979)Lloyd amp Webb (1992ab) hypothesized that the ancestral con-dition is an approach-herkogamous flower then a style-lengthdimorphic flower (stylar dimorphism) and finally a reciprocal-herkogamous position of sex organs (strict heterostyly)Accordingly shifts between these stages would be driven bychanges to more efficient pollinators in transferring pollenbetween morphs There are more cases supporting Lloyd ampWebbrsquos model (eg
Anchusa
Lithodora
Narcissus
) than thosefitting into Charlesworth amp Charlesworthrsquos model (1979)(species in the tribe
Staticeae
of Plumbaginaceae Lloyd ampWebb 1992a) Additionally based on taxonomic (nonphylo-genetic) information Lloyd amp Webb (1992a) assumed thatmost of nonheterostylous taxa within a particular heterosty-lous group are mostly approach herkogamous and interpretthe rarity of style dimorphism as a result of its instability forsome unknown reason However occurrence of style dimor-phism in
Narcissus
has been reported in many species (Barrett
et al
1996)Variation in flower characteristics of
Narcissus
appears tobe useful for testing models of heterostyly (Lloyd
et al
1990Barrett
et al
1996 Arroyo amp Barrett 2000 Arroyo 2002)This highly diverse Mediterranean genus includes speciesshowing nonherkogamous monomorphism in which thestigma is located at the same height than anthers (eg
Nserotinus
) approach-herkogamous monomorphism wherestigma has a higher position than anthers (eg all species ofsections Bulbocodii and Pseudonarcissi) style dimorphism inwhich populations include approach-herkogamous flowers(L-styled morphs) and reverse-herkogamous flowers (S-morph)with no anther reciprocity (eg some species of sections TazettaJonquilla and Apodanthi) distyly with L- and S-styled morphsand anther reciprocity (
N albimarginatus
) and tristyly withL- M- and S-styled morphs and anther reciprocity (the
Ntriandrus
complex) Testing evolutionary patterns behindsuch an array of flower polymorphism within the genus hasremained elusive without any phylogenetic analysis Hencewe have addressed phylogenetic reconstructions from neutralchloroplast sequences (
trn
L-F
trn
T-L) of some representativespecies of
Narcissus
which may help shed further light onevolution of stylar polymorphism As an initial step we havefocused on a taxonomic group within the genus sect Apo-danthi which exhibits a wide range of polymorphism variation(except tristyly and approach herkogamous monomorphism)through its seven species two monomorphic four style dimorphicand one distylous For the sake of comparison
N pallidulus
(
N triandrus
complex) a species in sect Cyclaminei display-ing heterostyly (tristyly) has been included in the analyses
There are three specific aims in this study First to ascertainthe kind of stylar polymorphism in populations of a presum-ably natural group section Apodanthi which preliminary
observations suggested as variable (Arroyo 2002) Second toinvestigate phylogenetic relationships and evolutionary tran-sitions of style polymorphism between the only two knowncases of heterostyly (
N albimarginatus
and
N triandrus
com-plex) and among all species in sect Apodanthi by means ofanalysis of chloroplast sequences And third to test whethershifts in style-polymorphism conditions are related to peri-anth features and pollinators as it was explicitly suggested byArroyo amp Barrett (2000) for
N albimarginatus
Materials and Methods
Study taxa
The long history of cultivation and naturalisation in
Narcissus
coupled with extensive hybridisation has prevented recognitionof a stable taxonomic treatment (Mathew 2002) In factthere is no taxonomic monograph for the entire genus Wehave followed the most recent comprehensive treatmentlisting analytically sections and species (Dorda amp Fernaacutendez-Casas 1989) This taxonomic treatment considers six speciesin our study group (sect Apodanthi)
N rupicola
N scaberulus
N calcicola
N cuatrecasasii
N marvieri
and
N watieri
Weinclude one more species (
N albimarginatus
) describedafterwards (Muller-Doblies amp Muller-Doblies 1989 see alsoArroyo amp Barrett 2000)
Although sect Apodanthi displays a wide range of flowermorphology it has been considered one of the most clearlyrecognised and taxonomically stable groups within the genus(Fernandes 1967 1975 Dorda amp Fernaacutendez-Casas 1989)Morphometrical analyses and pollination surveys were per-formed in sect Apodanthi to investigate relationships betweenthe degree of perianth variation among species and stylepolymorphisms which include the occurrence of dimorphicheterostyly (distyly) in only one species of the genus (
Nalbimarginatus
) One more species (
N pallidulus
) was alsoincluded in the study to represent the other group (the
N tri-andrus
complex
N triandrus
N lusitanicus
and
N pallidulus
)of
Narcissus
displaying heterostyly (Barrett
et al
1997) and tocontrast variation in perianth features and polymorphismwith respect to species of sect Apodanthi All the species arebulbous geophytes occurring in mountains at variable eleva-tions and blossom in late winter to spring The geographicrange of the species their typical habitats and elevation andpopulation locations are shown in Table 1 A comprehensivemap of the geographic range of species of sect Apodanthi mayalso be found in Arroyo (2002)
Population sampling
Flower morphometry
We selected two widely separated popu-lations from each species to represent within-species variationin flower morphology (see Table 1) In each population wecollected one recently opened flower per plant typically the
copy
New Phytologist
(2003)
161
235ndash252
wwwnewphytologistorg
Research 237
Tabl
e 1
Spec
ies
of
Nar
ciss
us
sam
pled
for
flow
er (
peria
nth
and
sex
orga
ns)
mor
phom
etry
()
Pop
ulat
ions
whe
re in
sect
cen
suse
s w
ere
done
Spec
ies
Popu
latio
nC
oord
inat
esH
abita
tEl
evat
ion
Flow
ers
colle
cted
N p
allid
ulus
1 Sp
ain
Jaeacute
n S
egur
a N
aval
caba
llo38
deg
20
prime
N 2
deg
37
prime
WPi
ne f
ores
t on
lim
esto
ne13
5012
2
N p
allid
ulus
2 Sp
ain
Bad
ajoz
Cal
era
Sie
rra
de T
entu
dia
38
deg
5
prime
N 6
deg
20
prime
WD
ecid
uous
oak
for
est
on s
hale
900
63
N a
lbim
argi
natu
s
()
3 M
oroc
co C
haou
en J
bel B
ouha
chem
35
deg
15
prime
N 5
deg
26
prime
WC
edar
for
est
on s
ands
tone
1500
200
N a
lbim
argi
natu
s
4 M
oroc
co C
haou
en J
bel K
elti
35
deg
21
prime
N 5
deg
17
prime
WO
ak f
ores
t on
lim
esto
ne15
2020
6
N c
uatr
ecas
asii
()
5 Sp
ain
Caacuted
iz G
raza
lem
a P
uert
o de
l Boy
ar36
deg
46
prime
N 5
deg
24
prime
WR
ock
fissu
res
on li
mes
tone
1100
100
N c
uatr
ecas
asii
6 Sp
ain
Jaeacute
n C
azor
la L
as C
orre
huel
as37
deg
59
prime
N 2
deg
54
prime
WM
ixed
pin
e fo
rest
on
limes
tone
1615
100
N c
alci
cola
7 Po
rtug
al L
eiriacutea
Ser
ra d
e Si
coacute39
deg
55
prime
N 8
deg
32
prime
WR
ock
fissu
res
on li
mes
tone
550
186
N c
alci
cola
8 Po
rtug
al L
eiriacutea
Ser
ra d
e St
o A
nton
io39
deg
32
prime
N 8
deg
44
prime
WR
ock
fissu
res
on li
mes
tone
714
212
N s
cabe
rulu
s
9 Po
rtug
al V
iseu
Oliv
eira
do
Con
de40
deg
26
prime
N 7
deg
57
prime
WSc
rubl
and
on g
rani
te25
923
2
N s
cabe
rulu
s
10 P
ortu
gal
Gua
rda
Erv
edal
40
deg
26
prime
N 7
deg54prime
WSc
rubl
and
on g
rani
te33
310
9N
rup
icol
a11
Spa
in J
aeacuten
Ald
eaqu
emad
a38
deg23prime
N 3
deg24prime
WC
liff
fissu
res
on g
rani
te10
0078
N r
upic
ola
()
12 S
pain
Mad
rid N
avac
erra
da B
ola
del M
undo
40deg4
7prime N
3deg5
9prime W
Alp
ine
scru
blan
d on
gra
nite
2257
154
N m
arvi
eri
13 M
oroc
co M
arra
kesh
Zer
etke
n31
deg27prime
N 7
deg26prime
WTr
ee-l
ine
ceda
r fo
rest
on
shal
e22
5051
N m
arvi
eri
14 M
oroc
co T
aza
Taz
zeka
34deg5
prime N 4
deg11prime
WD
ense
ced
ar f
ores
t an
d sc
rubl
and
on s
hale
1837
50N
wat
ieri
()
15 M
oroc
co M
arra
kesh
Ouk
aim
eden
31deg1
3prime N
7deg5
1prime W
Soil
pock
ets
on a
lpin
e m
eado
ws
on s
hale
2474
50N
wat
ieri
16 M
oroc
co M
arra
kesh
Tiz
irsquonrsquoT
est
31deg5
prime N 8
deg5prime W
Hol
m o
ak f
ores
t on
lim
esto
ne20
5040
earliest in the inflorescence At least 40 flowers wereimmediately preserved in ethanol 70 and kept refrigerateduntil measurement Given that species present vegetativemultiplication through division of bulbs an effort was made toavoid repeated sampling of ramets by collecting flowers inplants at least 2 m apart (Arroyo et al 2002) From eachpopulation we obtained additional flowers with the sameprocedure in order to have larger samples for estimatingmorph ratio in populations (see Table 1 for sample sizes)
Phylogenetic study Individuals and populations were selectedto represent the morphological diversity of the genus thegeographical breadth of the species and flower polymorphismThe seven species of Narcissus sect Apodanthi the four ofNarcissus sect Cyclaminei and seven representatives of the10 remaining sections were sampled (a single individualper population) and sequenced for the noncoding regiontrnL(UAA)-trnF(GAA) (Table 2) As a result we obtained andanalyzed 27 sequences of Narcissus Outgroup species (GalanthusLapiedra Pancratium) were chosen as suggested in a previousphylogeny of Amaryllidaceae based on rbcL and trnL-trnFsequences (Meerow et al 1999) Additionally a second samplefocused on the seven species of Narcissus sect Apodanthi wasaddressed to sequence and analyze the chloroplast noncodingregion trnT(UGU)-trnL(UAA) Narcissus bulbocodium andN papyraceus were used as outgroup species based on thetrnL-F phylogeny of Narcissus herein presented
Flower morphometry
Measurements Typically a Narcissus species has a flower tubeon the inferior ovary to which the six stamen filaments areattached two stamen whorls six tepals and above all acorona Each stamen whorl has three stamens reaching tworespective anther levels in sect Apodanthi and N triandruscomplex (Fig 1) Preserved flowers were slit longitudinallyfrom just above the ovary to the tube mouth The apex of theovary was the baseline for all measurements which wereperformed by digital calipers to the nearest 01 mm Measure-ments were (Fig 1) flower width tepal length corona widthcorona height tube length tube mouth width style lengthupper stamen height lower stamen height and flower angle tostalk Stamen height was measured up to the insertion of thefilament into the anther avoiding variation caused by anthersize Anthers display up to threefold change in length duringtheir fast ripening (R Peacuterez personal observation)
Given that all studied species have two whorls of stamensdifferent in length it is difficult to quantify the level of sexorgan reciprocity in dimorphic species which usually haveone stamen level in distylous species On functional grounds(ie proficiency of cross pollination between equivalent heightof sex organs) we have computed the degree of reciprocity asthe difference between means of style height in each morphand the closest stamen whorl of the alternative morph It
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research238
Tabl
e 2
Acc
essi
ons
for
the
chlo
ropl
ast
sequ
ence
stu
dy o
f N
arci
ssus
inc
ludi
ng p
lant
iden
tifica
tion
follo
wed
by
popu
latio
n nu
mbe
r lo
calit
y of
wild
pop
ulat
ions
nam
e of
pla
nt c
olle
ctor
s an
d G
enBa
nk a
cces
sion
num
bers
for
trn
T-L
and
trnL
-F s
eque
nces
Taxo
nLo
calit
y
Gen
Bank
ac
cess
ion
num
ber
trnT
-L
Gen
Bank
ac
cess
ion
num
ber
trnL
-F
sect
Apo
dant
hi A
Fer
nand
esN
cal
cico
la 1
Men
doccedila
Port
ugal
Ser
ra S
to A
nton
io A
Ham
pe amp
B G
arrid
oFo
rthc
omin
gFo
rthc
omin
gN
cal
cico
la 2
Men
doccedila
Port
ugal
Ser
ra S
ico
A H
ampe
amp B
Gar
rido
ndashFo
rthc
omin
gN
cua
trec
asas
ii 1
Fern
Cas
as e
t al
Sp
ain
Jaeacute
n S
ierr
a de
Caz
orla
J A
rroy
oFo
rthc
omin
gFo
rthc
omin
gN
cua
trec
asas
ii 2
Fern
Cas
as e
t al
Sp
ain
Caacuted
iz S
ierr
a de
Gra
zale
ma
J A
rroy
ondash
Fort
hcom
ing
N m
arvi
eri 1
Jah
and
amp M
aire
Mor
occo
Zer
ekte
n J
Arr
oyo
amp A
Ham
peFo
rthc
omin
gFo
rthc
omin
gN
mar
vier
i 2 J
ahan
d amp
Mai
reM
oroc
co T
azek
a J
Arr
oyo
amp A
Ter
rab
ndashFo
rthc
omin
gN
rup
icol
a 1
Leacuteon
-Duf
our
Spai
n C
uenc
a G
rado
de
Pico
P V
arga
s et
al
ndashFo
rthc
omin
gN
rup
icol
a 2
Leacuteon
-Duf
our
Spai
n M
adrid
Sie
rra
de la
Cab
rera
P V
arga
s et
al
Fort
hcom
ing
Fort
hcom
ing
N r
upic
ola
3 Leacute
on-D
ufou
rSp
ain
Cue
nca
Gra
do d
e Pi
co P
Var
gas
et a
lndash
Fort
hcom
ing
N s
cabe
rulu
s 1
Hen
riqu
esPo
rtug
al E
rved
al A
Ham
pe amp
B G
arrid
oFo
rthc
omin
gFo
rthc
omin
gN
sca
beru
lus
2 H
enri
ques
Port
ugal
Erv
edal
A H
ampe
amp B
Gar
rido
ndashFo
rthc
omin
gN
sca
beru
lus
3 H
enri
ques
Port
ugal
Oliv
eira
do
Con
de A
Ham
pe amp
B G
arrid
ondash
Fort
hcom
ing
N w
atie
ri M
aire
Mor
occo
Ouk
aim
eden
J A
rroy
o amp
A H
ampe
Fort
hcom
ing
Fort
hcom
ing
N a
lbim
argi
natu
s 1
U M
uumllle
r-D
oblie
s amp
D M
uumllle
r-D
oblie
sM
oroc
co B
ouha
chem
J A
rroy
oFo
rthc
omin
gFo
rthc
omin
gN
alb
imar
gina
tus
2 U
Muumll
ler-
Dob
lies
amp D
Muumll
ler-
Dob
lies
Mor
occo
Kel
ti F
Oje
dandash
Fort
hcom
ing
sect
Cyc
lam
inei
DC
N
cyc
lam
ineu
s D
C
Spai
n P
onte
vedr
a S
Cas
trov
iejo
616
2ndash
Fort
hcom
ing
N l
usit
anic
us D
orda
amp F
ern
Cas
asPo
rtug
al P
ena
Cov
a de
Oliv
eira
A B
arra
AB2
585
ndashFo
rthc
omin
gN
pal
lidul
us 1
Gra
ells
Spai
n M
adrid
Ald
ea d
el F
resn
o P
Var
gas
amp O
Goacutem
ezndash
Fort
hcom
ing
N p
allid
ulus
2 G
rael
lsSp
ain
Jaeacute
n A
ldea
quem
ada
R P
eacuterez
ndashFo
rthc
omin
gN
tri
andr
us L
Sp
ain
Zam
ora
Mah
ide
B H
ernaacute
ndez
ndashFo
rthc
omin
gse
ct D
ubii
Fern
Cas
asN
tor
tifo
lius
Fern
Cas
asSp
ain
Alm
eriacutea
Riacuteo
de
Agu
as A
Bar
randash
Fort
hcom
ing
sect
Jon
quill
ae D
C
N j
onqu
illa
LSp
ain
Sev
illa
Sie
rra
Nor
te R
Peacuter
ez amp
R P
eacuterez
de
Guz
maacuten
ndashFo
rthc
omin
gse
ct S
erot
ini P
arl
N s
erot
inus
Loe
fl e
x L
Spai
n S
evill
a D
os H
erm
anas
C A
ndreacute
s amp
Z D
iacuteaz
ndashFo
rthc
omin
gse
ct T
azet
tae
DC
N
taz
etta
L
Gen
Bank
ndashA
J232
562
sect
Bul
boco
dii D
C
N b
ulbo
codi
um L
Sp
ain
Hue
lva
Car
taya
J C
astil
loFo
rthc
omin
gFo
rthc
omin
gse
ct N
arci
ssus
L
N p
oeti
cus
LSp
ain
Leacuter
ida
Val
le d
e A
raacuten
X P
icoacute
ndashFo
rthc
omin
gse
ct P
seud
onar
ciss
i DC
N
pse
udon
arci
ssus
L
Spai
n Aacute
vila
Sie
rra
de G
redo
s R
Peacuter
ez amp
J A
rroy
ondash
Fort
hcom
ing
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 239
means that in one dimorphic or distylous Narcissus speciesthe long style (L-morph) has its reciprocal male organ in upperstamens of S-morph Accordingly S-style is the reciprocal tolower stamens of L-morph (Fig 1) Flowers of the tristylousNarcissus pallidulus does not pose special problems as thiscondition usually involves two stamen whorls and the com-parison of reciprocal heights is straightforward as originallydescribed by Darwin (1877) in Lythrum salicaria We aver-aged the absolute values of this difference across the morphsof each population in order to compare dimorphic and tri-morphic species A perfectly reciprocal population wouldtherefore rate a value of zero In order to control for effects offlower size in the extent of reciprocity we also divided the rec-iprocity of a population by its mean flower diameter
Statistical analyses All morphometrical variables were checkedfor normality and homoscedasticity and transformed whennecessary Only flower tube length across populations andspecies needed to be log transformed Data were averaged formorphs and populations and compared accordingly Studentt-tests and one-way were used for univariate compar-isons of flower features between morphs for dimorphic andtrimorphic species respectively In the rare cases wherevariances were not homogeneous (Levenersquos test) we applieda t-test with separate variance estimates After fordifference of means posthoc Tukey HSD test for unequalsample sizes was applied for multiple comparison betweenpairs When performing multiple tests of one hypothesisDunn-Sidak correction of sequential Bonferroni adjustementof α levels was applied (Sokal amp Rohlf 1995) Reciprocity(between-morph stigma-anther separation) was not comparedbetween morphs given that it was obtained as a differencebetween morph means hence error estimates of the differencewere not available Multivariate analyses were performed tocheck the overall pattern of morphological variation of flower
perianth An ordination by Principal Component Analysis(PCA) was performed to draw the morphological relationshipsbetween populations and species In multivariate analysispopulation means were used rather than individual data inorder to avoid pseudoreplication An analysis of the intra-population variation (Faivre amp McDade 2001) is intendedto be published in a subsequent paper
To analyse the relationships between perianth morphologyand sex organ polymorphism we applied multiple regressionanalyses Independent variables were the scores of the populationmeans for each of three first PCA axes We used these perianthpseudovariables rather than raw univariate values because axesaccounted for a high variance of perianth morphology andbecause axes are by definition orthogonal overcoming the seriousproblem of multicollinearity in multiple regression ( James ampMcCulloch 1990 Philippi 1993) Two regression modelswere adjusted with respective dependent variables related toreciprocity of stigma-anther separation of each population(i) the mean between morphs of absolute values for differencebetween style height and equivalent stamens whorl height and(ii) the values of (i) divided by flower diameter The latter var-iable was considered to control for potential allometric effectof flower size on the magnitude of departures from stigma-anther reciprocity (see above) Values for monomorphic popu-lations were scored as the highest obtained value of departurefrom exact reciprocity Ordination analysis was performedwith PCORD package (McCune amp Mefford 1999) whereasunivariate statistics and multiple regression models were com-puted with STATISTICA software (Statsoft 1997)
Pollinators
Data were taken from populations of four Apodanthi spe-cies some in common with those of the study of flowermorphometry (Table 1) We performed systematic censuses
Fig 1 Long-styled (L) and short-styled (S) flowers of a typical style-dimorphic Narcissus species (N cuatrecasasii) Numbers correspond to the measurements taken in each flower sampled (1) flower width (2) tepal length (3) corona width (4) corona height (5) tube length (6) tube width (7) style length (8) upper stamen height (9) lower stamen height (10) angle of flower to stalk Measurements were accordingly modified for tristylous distylous and monomorphic species
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research240
of insects visiting the flowers and recorded positive dataproviding that they enter into the tube and do contact any sexorgan Species subject to study were N rupicola (time effortof 12 h including some nocturnal periods) N cuatrecasasii(12 h) N albimarginatus (9 h) and N watieri (3 h) Observationtiming and area depend on density of flowers in everypopulation In sparse patches we made several censuses of15 min observing all insects visiting flowers in an area ofapprox 100 m2 We also made static observations in densepatches on all flowers in small areas (approx 10 m2)
Phylogenetic studies
PCR amplification and sequencing Total genomic DNA wasextracted from silica-dried material collected in the field andthe living collection of the Royal Botanic Garden of MadridSpain DNA extractions were performed using the DNeasyPlant Mini Kit (QIAGEN Inc Chatsworth CA USA) ForPolymerase Chain Reaction (PCR) amplifications we used aPerkin-Elmer (Foster City CA USA) PCR System 9700 thermalcycler primers trna and trnb for trnT(UGU)-trnL(UAA)trne and trn f for trnL(UAA)-trnF(GAA) and PCR conditionsfollowing Taberlet et al (1991) 51degC of annealing temperaturefor the trnT-L and 50degC for the trnL-F plus 2-min elongationtime Amplified products were cleaned using spin filter columns(PCR Clean-up kit MoBio Laboratories Solana Beach CAUSA) following the protocols provided by the manufacturerCleaned products were then directly sequenced using dyeterminators (Big Dye Terminator vs 20 Applied BiosystemsLittle Chalfont UK) and run into polyacrylamideelectrophoresis gels (7) using an Applied Biosystems Prismmodel 3700 automated sequencer Primers trna trnb trneand trn f were also used for cycle sequencing of the trnT-trnLand trnL-trnF spacers under the following conditions 95degCfor 2 min followed by 25 cycles of 95degC for 10 s 50degC for 5s and 60degC for 4 min Sequence data were placed in a contigfile and edited using the program Seqed (Applied BiosystemsFoster City CA USA) The limits of the trnT-trnL and trnL-trnF regions were determined by comparison with otherAmaryllidaceae sequences available in the GenBank
Sequence analyses Three different sequence matrices wereused to perform the phylogenetic analyses trnL-F matrixof Narcissus trnT-L matrix of Narcissus sect Apodanthiand a combined matrix of trnL-F and trnT-L for only thespecies of Narcissus sect Apodanthi as ingroup Alignmentsof the 30 trnL-F and the nine trnT-L sequences wereobtained using the program Clustal X 162b (httpevolutiongeneticswashingtoneduphylipsoftwareetc1html)with further manual adjustments Two phylogenetic analyseswere performed by using two primary approaches to treeconstruction tree-searching (parsimony) and algorithmic(Neighbor-Joining) Parsimony-based analyses were conductedusing Fitch parsimony (as implemented in Swofford
1999) with equal weighting of all characters and oftransitionstransversions Heuristic searches were replicated100 times with random taxon-addition sequences TreeBisection-Reconnection (TBR) branch swapping and withthe options MULPARS and STEEPEST DESCENT ineffect Relative support for clades identified by parsimonyanalysis was assessed by bootstrapping with the heuristicsearch strategy as indicated above In addition phylogeneticreconstructions of sequences were also performed in using the Kimura 2-parameter distance model (Kimura 1980)and the Neighbor-Joining method (Saitou amp Nei 1987)Indels in the noncoding regions trnT-L and trnL-F werecoded as appended characters following the logic of Kelchner(2000) and Simmons amp Ochoterena (2000) As a resultadditional analyses were performed when coding parsimony-informative indels affecting the trnL-F and trnT-L matricesCharacter evolution of flower polymorphism was investigatedusing MacClade 35 (Maddison amp Maddison 1992) Fourstates of flower polymorphism (approach herkogamy non-herkogamous monomorphism style dimorphism and hetero-styly) were traced and mapped on the resulting trnL-F andtrnT-L phylogenies of the consensus tree of all maximum-parsimony trees We explored both accelerated transformation(ACCTRAN) and delayed transformation (DELTRAN)optimizations with equal weighting of all characters andtransitions among all states equally probable
Results
Style polymorphism
Average values of sex organ position for the seven species ofsect Apodanthi and for N pallidulus are shown in Tables 3 4and 5 and comparative diagrams in Fig 2 Species in sectionApodanthi have a remarkable variation in style polymorphismdistyly in N albimarginatus style dimorphism in four species(N calcicola N cuatrecasasii N rupicola and N scaberulus)and monomorphism in N marvieri and N watieri (Fig 2Table 3) Narcissus albimarginatus shows two style-lengthmorphs with approximate reciprocal positioning of anthersand style Two patterns were found for dimorphic speciesthree species (N calcicola N scaberulus N cuatrecasasii ) presentthe L-morph with the stigma almost at an equivalent level ofthat of upper anthers whereas in N rupicola the stigma of theL-morph is positioned between the two anther levels The latterspecies displayed the highest values of stigma-anther separation(Table 4) that is the lowest reciprocity among dimorphicspecies Monomorphic species present a position of sex organssimilar to that in L-morph of dimorphic N rupicola (Fig 2)
Narcissus pallidulus is the only trimorphic species in ourstudy with mostly reciprocal positioning of stigmas andstamen whorls Although equivalent positions are not equal inall cases sequence height of sex organs corresponds to typicaltristylous species
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 241
At the population level morphs are in general unequallyrepresented L-morph is the most frequent (Tables 4 and 5) inall species and populations with the exception of one popu-lation of N cuatrecasasii and one of N rupicola which showsimilar frequency of morphs (isoplethy)
Perianth variation
Values of perianth variables are averaged over populationsand morphs (Table 6) Univariate morphometrical analyses offlower tubes tepals coronas and flower angle of all studiedspecies and populations revealed clearly absence of significantdifferences between morphs (82 between-morph comparisonsresults not shown) There were only slightly significant dif-ferences in tristylous N pallidulus at Cazorla for corona height(P = 001) and in distylous N albimarginatus at Bouhachemfor perianth width (P = 005)
Principal component analysis is shown in Fig 3 The firstsecond and third axes accounted for 616 233 and 94of the variance although only Axes 1 and 2 have eigenvalueshigher than 10 and are hence further discussed These twoaxes depict quite clearly the relationships between species andpopulations in perianth morphology At the negative extremeof Axis 1 we find together N pallidulus and N albimarginatusand at the positive extreme of this axis N marvieri Nwatieri and N rupicola form a second group Axis 2 separatesat its positive end the Iberian endemics N cuatrecasasii N cal-cicola and N scaberulus Eigenvectors of perianth variablesreflect trends of morphological variation The highest valuesfor Axis 1 are flower angle to stalk (04519) width of flowertube (minus04410) and corona height (minus04217) Largest eigen-vectors for Axis 2 correspond to corona width (minus06470)tepal length (minus04362) and flower tube length (minus04087)These results indicate that populations of two species (Npallidulus and N albimarginatus) at the left in the ordinationplot are characterised by small angles (pendulous flowers)wide flower tubes and long coronas (Fig 4) A second groupof species (N rupicola N watieri N marvieri ) display pre-dominantly erect flowers narrow and very long tubes andfunnel-shaped coronas On the top of the ordination plotthree species (N cuatrecasasii N calcicola N scaberulus) arecharacterised by patent flowers cup-shaped coronas smalltepals and moderate flower-tube length
The interrelationship between perianth variables and stylarpolymorphism is evaluated by means of multiple regressionanalyses We calculated two measurements of reciprocity inthe position of sex organs considering or not flower sizeeffects Both regression equations fitted quite similarly withscores of populations along PCA axes on perianth variablesAs a result we may conclude that there is a significant rela-tionship between perianth morphology and sex organ posi-tion across populations and species with 72ndash80 of thevariance being explained by the regression models (Table 7)Positioning of reciprocal stigma and anthers (heterostyly) cor-responds with the group of pendulous flowers Style dimorphicspecies (N calcicola N cuatrecasasii N scaberulus see Table 4)with medium values of reciprocity have patent flowers amongother features (see above) Monomorphism and less reciprocalstyle dimorphism in N rupicola (Table 4) correspond to theerect flowered group
Pollinators
As typically occurs in early blooming species it was difficultto observe pollinators in four species of Narcissus sectApodanthi (0ndash467 insect visits per hour) Twelve hours ofobservations in N cuatrecasasii ( Jaeacuten Cazorla range) revealedrarity of pollinators although medium sized solitary bees(Anthophora sp) were consistently spotted (10 visits) probingnectar and small unidentified bees (8 visits) reaching pollenof the upper anthers In the same area Anthophora bees visitedflowers of Helleborus foetidus but always in different boutsCasual observations made on populations at the oppositeextreme of the geographic range of N cuatrecasasii (CaacutedizGrazalema range) allow recognising Anthophora as the mostcommon visitor By censuses in N rupicola at the core of itsgeographic range (Avila Puerto de Mijares Madrid Puertode Navacerrada) we observed only two visits by the mothMacroglossum stellatarum one by the hoverfly Eristalis tenaxand one by an unidentified small bee Flies and small beestried to enter the flower tube but they only contacted upperanthers and L-stigma because of its narrowness and shorttongues Infrequent pollinator visits is not the result of lownumber of insects as they visited profusely another co-occurring flowers of some genera (Gagea Ranunculus SaxifragaArmeria Cytisus Erysimum) Some nocturnal censuses revealed
Table 3 Height of stamens and stigmas in monomorphic species of Narcissus sect Apodanthi Values are means plusmn 1 se in mm
Species Population Sample size Style heightUpper anther height
Lower anther height
Stigma-Anther separation
N marvieri Zeretken 51 2069 plusmn 030 2255 plusmn 032 1852 plusmn 035 186N marvieri Tazzeka 50 1001 plusmn 015 1234 plusmn 018 865 plusmn 014 233N watieri Oukaimeden 50 2092 plusmn 028 2205 plusmn 029 1815 plusmn 023 113N watieri TizirsquonrsquoTest 26 1736 plusmn 030 1927 plusmn 040 1595 plusmn 041 191
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research242
Tabl
e 4
Hei
ght
of s
tam
ens
and
stig
mas
in d
imor
phic
spe
cies
of
Nar
ciss
us s
ect
Apo
dant
hi L
lon
g-st
yled
mor
ph S
sho
rt-s
tyle
d m
orph
Val
ues
are
mea
ns plusmn
1 s
e in
mm
For
est
imat
ing
stig
ma-
anth
er s
epar
atio
n w
e co
nsid
er o
nly
the
equi
vale
nt w
horl
to th
e st
igm
a of
the
alte
rnat
e m
orph
s to
be
the
reci
proc
al (L
stig
ma
vs u
pper
sta
men
who
rl of
S fl
ower
S s
tigm
a vs
low
er s
tam
en w
horl
of L
flow
er s
ee F
ig 1
) A
ppro
pria
te t
-tes
ts a
re g
iven
for
the
diff
eren
ces
betw
een
mea
ns o
f th
ese
leng
ths
(P
lt 0
05
P
lt 0
01
P lt
00
01)
Spec
ies
Popu
latio
n
Sam
ple
size
L S
Styl
e he
ight
U
pper
ant
her
heig
ht
Low
er a
nthe
r he
ight
St
igm
a-A
nthe
r se
para
tion
Mor
ph r
atio
(1
)L
SL
SL
SL
S
N a
lbim
argi
natu
sBo
uhac
hem
69 5
824
40
plusmn 3
8
9
77 plusmn
17
015
97
plusmn 2
14
24
54
plusmn 2
899
96 plusmn
17
8
21
80
plusmn 2
550
14
019
065
0
35N
alb
imar
gina
tus
Kel
ti53
45
243
7 plusmn
032
936
plusmn 0
19
148
9 plusmn
028
216
7 plusmn
036
849
plusmn 0
23
188
1 plusmn
041
270
0
870
66
034
N c
uatr
ecas
asii
Gra
zale
ma
66 3
415
08
plusmn 0
20
8
01 plusmn
01
713
79
plusmn 0
18 n
s 14
36
plusmn 0
269
05 plusmn
01
6
12
03
plusmn 0
270
72
104
065
0
35N
cua
trec
asas
iiC
azor
la42
58
146
0 plusmn
020
690
plusmn 0
12
131
8 plusmn
018
13
95
plusmn 0
158
25 plusmn
01
4
10
74
plusmn 0
110
65
162
042
0
58N
cal
cico
laSi
coacute50
43
163
6 plusmn
034
819
plusmn 0
18
144
6 plusmn
018
152
5 plusmn
029
925
plusmn 0
21
113
2 plusmn
018
112
1
380
69
031
N c
alci
cola
Sto
Ant
onio
51 4
416
49
plusmn 0
29
8
06 plusmn
02
715
30
plusmn 0
23 n
s 15
37
plusmn 0
229
44 plusmn
02
1
11
68
plusmn 0
231
11
106
076
0
24N
sca
beru
lus
Oliv
eira
51 3
115
12
plusmn 0
24
7
72 plusmn
01
714
56
plusmn 0
15 n
s 15
05
plusmn 0
279
21 plusmn
01
6
11
50
plusmn 0
240
071
490
86
014
N s
cabe
rulu
sEr
veda
l50
14
147
7 plusmn
029
778
plusmn 0
28
148
5 plusmn
021
ns
149
9 plusmn
065
986
plusmn 0
15
11
71
plusmn 0
370
22
208
077
0
23N
rup
icol
aA
ldea
quem
ada
42 3
614
94
plusmn 0
50
10
43
plusmn 0
2216
08
plusmn 0
27
18
19
plusmn 0
26 1
109
plusmn 0
25
139
8 plusmn
021
325
0
660
54
046
N r
upic
ola
Bola
del
Mun
do53
35
131
9 plusmn
024
816
plusmn 0
14
154
5 plusmn
023
ns
163
1 plusmn
029
105
6 plusmn
018
126
6 plusmn
025
312
2
400
72
028
(1)
Sam
ple
size
use
d fo
r m
orph
rat
io is
the
tot
al n
umbe
r of
flow
ers
colle
cted
(se
e Ta
ble
1)
occurrence of some moths around the flowers of N rupicolabut they were not observed to come into contact with flowersIn N watieri insect visits were much more abundant After a3-hour census we observed long-tongued visitors (diurnalmoths six visits pierid butterflies three visits) and short-tongued hoverflies (five visits) Despite considerable time(9 h) spent on N albimarginatus we were not able to spot anypollinator
Chloroplast sequence variation
Length of trnL-F sequences in Narcissus ranges between337 bp (population 1 of Narcissus pallidulus) and 365 bp(population 2 of Narcissus marvieri ) This sequence variationis caused by seven indels of 1ndash19 bp Number of variableparsimony-informative characters is of 1711 across species ofNarcissus The highest pairwise divergence (Kimura-2-parametermodel) of species sequences (34) was found betweenaccessions of N bulbocodium and N cyclamineus Sequencedivergence neither was found between species of sectApodanthi (excluding N calcicola and N scaberulus) nor wasobserved between two of the three species of the N triandruscomplex (N pallidulus and N lusitanicus) Kimura-2-parameter distances between accessions of the same speciesdid not rendered either divergence
In Narcissus sect Apodanthi length variation of trnT-Lsequences ranges between 827 bp (N marvieri ) and 966 bp(N calcicola) Ten gaps in the trnT-L sequence matrix from1 bp to 130 bp are responsible for great sequence variationAn unusual indel of 130 bp is shared by N rupicola N marv-ieri and N watieri The number of variableparsimony-informative characters is 1911 in Narcissus sect ApodanthiThe highest pairwise divergence (Kimura-2-parameter model)between sequence species (14) was found between acces-sions of N calcicola and N cuatrecasasii whereas the lowest(021) was between N calcicola and N scaberulus
Phylogenetic relationships
Maximum parsimony (MP) using the 27 trnL-F sequencesof Narcissus and Galanthus Lapiedra and Leucojum as theoutgroup yielded 1529 trees of 78 steps (CI 095 and RI093 including uninformative characters) (results not shown)A congruent somewhat more resolved topology was retrievedin the semistrict consensus tree of 13 225 trees (101 stepsCI 085 and RI 085) when coding the seven indels asadditional characters however new branches do not displaysignificant values (bootstrap below 50) (Fig 5) From bothanalyses (with and without indel coding) a basal polytomy isdepicted in the semistrict consensus tree of which two well-supported clades are formed by the accessions of five speciesof sect Apodanthi (clade 1) and six species including the threeof the N triandrus complex (clade 2) Relatively long branchesin the Neighbor-Joining (NJ) trees using both alignments
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg
copy
New Phytologist
(2003)
161
235ndash252
Research236
Webb 1992a) emphasizes the role of a pollination environ-ment rather than the existence of a particular genetic incom-patibility system as considered instead by the Charlesworthamp Charlesworthrsquos model (Charlesworth amp Charlesworth 1979)Lloyd amp Webb (1992ab) hypothesized that the ancestral con-dition is an approach-herkogamous flower then a style-lengthdimorphic flower (stylar dimorphism) and finally a reciprocal-herkogamous position of sex organs (strict heterostyly)Accordingly shifts between these stages would be driven bychanges to more efficient pollinators in transferring pollenbetween morphs There are more cases supporting Lloyd ampWebbrsquos model (eg
Anchusa
Lithodora
Narcissus
) than thosefitting into Charlesworth amp Charlesworthrsquos model (1979)(species in the tribe
Staticeae
of Plumbaginaceae Lloyd ampWebb 1992a) Additionally based on taxonomic (nonphylo-genetic) information Lloyd amp Webb (1992a) assumed thatmost of nonheterostylous taxa within a particular heterosty-lous group are mostly approach herkogamous and interpretthe rarity of style dimorphism as a result of its instability forsome unknown reason However occurrence of style dimor-phism in
Narcissus
has been reported in many species (Barrett
et al
1996)Variation in flower characteristics of
Narcissus
appears tobe useful for testing models of heterostyly (Lloyd
et al
1990Barrett
et al
1996 Arroyo amp Barrett 2000 Arroyo 2002)This highly diverse Mediterranean genus includes speciesshowing nonherkogamous monomorphism in which thestigma is located at the same height than anthers (eg
Nserotinus
) approach-herkogamous monomorphism wherestigma has a higher position than anthers (eg all species ofsections Bulbocodii and Pseudonarcissi) style dimorphism inwhich populations include approach-herkogamous flowers(L-styled morphs) and reverse-herkogamous flowers (S-morph)with no anther reciprocity (eg some species of sections TazettaJonquilla and Apodanthi) distyly with L- and S-styled morphsand anther reciprocity (
N albimarginatus
) and tristyly withL- M- and S-styled morphs and anther reciprocity (the
Ntriandrus
complex) Testing evolutionary patterns behindsuch an array of flower polymorphism within the genus hasremained elusive without any phylogenetic analysis Hencewe have addressed phylogenetic reconstructions from neutralchloroplast sequences (
trn
L-F
trn
T-L) of some representativespecies of
Narcissus
which may help shed further light onevolution of stylar polymorphism As an initial step we havefocused on a taxonomic group within the genus sect Apo-danthi which exhibits a wide range of polymorphism variation(except tristyly and approach herkogamous monomorphism)through its seven species two monomorphic four style dimorphicand one distylous For the sake of comparison
N pallidulus
(
N triandrus
complex) a species in sect Cyclaminei display-ing heterostyly (tristyly) has been included in the analyses
There are three specific aims in this study First to ascertainthe kind of stylar polymorphism in populations of a presum-ably natural group section Apodanthi which preliminary
observations suggested as variable (Arroyo 2002) Second toinvestigate phylogenetic relationships and evolutionary tran-sitions of style polymorphism between the only two knowncases of heterostyly (
N albimarginatus
and
N triandrus
com-plex) and among all species in sect Apodanthi by means ofanalysis of chloroplast sequences And third to test whethershifts in style-polymorphism conditions are related to peri-anth features and pollinators as it was explicitly suggested byArroyo amp Barrett (2000) for
N albimarginatus
Materials and Methods
Study taxa
The long history of cultivation and naturalisation in
Narcissus
coupled with extensive hybridisation has prevented recognitionof a stable taxonomic treatment (Mathew 2002) In factthere is no taxonomic monograph for the entire genus Wehave followed the most recent comprehensive treatmentlisting analytically sections and species (Dorda amp Fernaacutendez-Casas 1989) This taxonomic treatment considers six speciesin our study group (sect Apodanthi)
N rupicola
N scaberulus
N calcicola
N cuatrecasasii
N marvieri
and
N watieri
Weinclude one more species (
N albimarginatus
) describedafterwards (Muller-Doblies amp Muller-Doblies 1989 see alsoArroyo amp Barrett 2000)
Although sect Apodanthi displays a wide range of flowermorphology it has been considered one of the most clearlyrecognised and taxonomically stable groups within the genus(Fernandes 1967 1975 Dorda amp Fernaacutendez-Casas 1989)Morphometrical analyses and pollination surveys were per-formed in sect Apodanthi to investigate relationships betweenthe degree of perianth variation among species and stylepolymorphisms which include the occurrence of dimorphicheterostyly (distyly) in only one species of the genus (
Nalbimarginatus
) One more species (
N pallidulus
) was alsoincluded in the study to represent the other group (the
N tri-andrus
complex
N triandrus
N lusitanicus
and
N pallidulus
)of
Narcissus
displaying heterostyly (Barrett
et al
1997) and tocontrast variation in perianth features and polymorphismwith respect to species of sect Apodanthi All the species arebulbous geophytes occurring in mountains at variable eleva-tions and blossom in late winter to spring The geographicrange of the species their typical habitats and elevation andpopulation locations are shown in Table 1 A comprehensivemap of the geographic range of species of sect Apodanthi mayalso be found in Arroyo (2002)
Population sampling
Flower morphometry
We selected two widely separated popu-lations from each species to represent within-species variationin flower morphology (see Table 1) In each population wecollected one recently opened flower per plant typically the
copy
New Phytologist
(2003)
161
235ndash252
wwwnewphytologistorg
Research 237
Tabl
e 1
Spec
ies
of
Nar
ciss
us
sam
pled
for
flow
er (
peria
nth
and
sex
orga
ns)
mor
phom
etry
()
Pop
ulat
ions
whe
re in
sect
cen
suse
s w
ere
done
Spec
ies
Popu
latio
nC
oord
inat
esH
abita
tEl
evat
ion
Flow
ers
colle
cted
N p
allid
ulus
1 Sp
ain
Jaeacute
n S
egur
a N
aval
caba
llo38
deg
20
prime
N 2
deg
37
prime
WPi
ne f
ores
t on
lim
esto
ne13
5012
2
N p
allid
ulus
2 Sp
ain
Bad
ajoz
Cal
era
Sie
rra
de T
entu
dia
38
deg
5
prime
N 6
deg
20
prime
WD
ecid
uous
oak
for
est
on s
hale
900
63
N a
lbim
argi
natu
s
()
3 M
oroc
co C
haou
en J
bel B
ouha
chem
35
deg
15
prime
N 5
deg
26
prime
WC
edar
for
est
on s
ands
tone
1500
200
N a
lbim
argi
natu
s
4 M
oroc
co C
haou
en J
bel K
elti
35
deg
21
prime
N 5
deg
17
prime
WO
ak f
ores
t on
lim
esto
ne15
2020
6
N c
uatr
ecas
asii
()
5 Sp
ain
Caacuted
iz G
raza
lem
a P
uert
o de
l Boy
ar36
deg
46
prime
N 5
deg
24
prime
WR
ock
fissu
res
on li
mes
tone
1100
100
N c
uatr
ecas
asii
6 Sp
ain
Jaeacute
n C
azor
la L
as C
orre
huel
as37
deg
59
prime
N 2
deg
54
prime
WM
ixed
pin
e fo
rest
on
limes
tone
1615
100
N c
alci
cola
7 Po
rtug
al L
eiriacutea
Ser
ra d
e Si
coacute39
deg
55
prime
N 8
deg
32
prime
WR
ock
fissu
res
on li
mes
tone
550
186
N c
alci
cola
8 Po
rtug
al L
eiriacutea
Ser
ra d
e St
o A
nton
io39
deg
32
prime
N 8
deg
44
prime
WR
ock
fissu
res
on li
mes
tone
714
212
N s
cabe
rulu
s
9 Po
rtug
al V
iseu
Oliv
eira
do
Con
de40
deg
26
prime
N 7
deg
57
prime
WSc
rubl
and
on g
rani
te25
923
2
N s
cabe
rulu
s
10 P
ortu
gal
Gua
rda
Erv
edal
40
deg
26
prime
N 7
deg54prime
WSc
rubl
and
on g
rani
te33
310
9N
rup
icol
a11
Spa
in J
aeacuten
Ald
eaqu
emad
a38
deg23prime
N 3
deg24prime
WC
liff
fissu
res
on g
rani
te10
0078
N r
upic
ola
()
12 S
pain
Mad
rid N
avac
erra
da B
ola
del M
undo
40deg4
7prime N
3deg5
9prime W
Alp
ine
scru
blan
d on
gra
nite
2257
154
N m
arvi
eri
13 M
oroc
co M
arra
kesh
Zer
etke
n31
deg27prime
N 7
deg26prime
WTr
ee-l
ine
ceda
r fo
rest
on
shal
e22
5051
N m
arvi
eri
14 M
oroc
co T
aza
Taz
zeka
34deg5
prime N 4
deg11prime
WD
ense
ced
ar f
ores
t an
d sc
rubl
and
on s
hale
1837
50N
wat
ieri
()
15 M
oroc
co M
arra
kesh
Ouk
aim
eden
31deg1
3prime N
7deg5
1prime W
Soil
pock
ets
on a
lpin
e m
eado
ws
on s
hale
2474
50N
wat
ieri
16 M
oroc
co M
arra
kesh
Tiz
irsquonrsquoT
est
31deg5
prime N 8
deg5prime W
Hol
m o
ak f
ores
t on
lim
esto
ne20
5040
earliest in the inflorescence At least 40 flowers wereimmediately preserved in ethanol 70 and kept refrigerateduntil measurement Given that species present vegetativemultiplication through division of bulbs an effort was made toavoid repeated sampling of ramets by collecting flowers inplants at least 2 m apart (Arroyo et al 2002) From eachpopulation we obtained additional flowers with the sameprocedure in order to have larger samples for estimatingmorph ratio in populations (see Table 1 for sample sizes)
Phylogenetic study Individuals and populations were selectedto represent the morphological diversity of the genus thegeographical breadth of the species and flower polymorphismThe seven species of Narcissus sect Apodanthi the four ofNarcissus sect Cyclaminei and seven representatives of the10 remaining sections were sampled (a single individualper population) and sequenced for the noncoding regiontrnL(UAA)-trnF(GAA) (Table 2) As a result we obtained andanalyzed 27 sequences of Narcissus Outgroup species (GalanthusLapiedra Pancratium) were chosen as suggested in a previousphylogeny of Amaryllidaceae based on rbcL and trnL-trnFsequences (Meerow et al 1999) Additionally a second samplefocused on the seven species of Narcissus sect Apodanthi wasaddressed to sequence and analyze the chloroplast noncodingregion trnT(UGU)-trnL(UAA) Narcissus bulbocodium andN papyraceus were used as outgroup species based on thetrnL-F phylogeny of Narcissus herein presented
Flower morphometry
Measurements Typically a Narcissus species has a flower tubeon the inferior ovary to which the six stamen filaments areattached two stamen whorls six tepals and above all acorona Each stamen whorl has three stamens reaching tworespective anther levels in sect Apodanthi and N triandruscomplex (Fig 1) Preserved flowers were slit longitudinallyfrom just above the ovary to the tube mouth The apex of theovary was the baseline for all measurements which wereperformed by digital calipers to the nearest 01 mm Measure-ments were (Fig 1) flower width tepal length corona widthcorona height tube length tube mouth width style lengthupper stamen height lower stamen height and flower angle tostalk Stamen height was measured up to the insertion of thefilament into the anther avoiding variation caused by anthersize Anthers display up to threefold change in length duringtheir fast ripening (R Peacuterez personal observation)
Given that all studied species have two whorls of stamensdifferent in length it is difficult to quantify the level of sexorgan reciprocity in dimorphic species which usually haveone stamen level in distylous species On functional grounds(ie proficiency of cross pollination between equivalent heightof sex organs) we have computed the degree of reciprocity asthe difference between means of style height in each morphand the closest stamen whorl of the alternative morph It
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research238
Tabl
e 2
Acc
essi
ons
for
the
chlo
ropl
ast
sequ
ence
stu
dy o
f N
arci
ssus
inc
ludi
ng p
lant
iden
tifica
tion
follo
wed
by
popu
latio
n nu
mbe
r lo
calit
y of
wild
pop
ulat
ions
nam
e of
pla
nt c
olle
ctor
s an
d G
enBa
nk a
cces
sion
num
bers
for
trn
T-L
and
trnL
-F s
eque
nces
Taxo
nLo
calit
y
Gen
Bank
ac
cess
ion
num
ber
trnT
-L
Gen
Bank
ac
cess
ion
num
ber
trnL
-F
sect
Apo
dant
hi A
Fer
nand
esN
cal
cico
la 1
Men
doccedila
Port
ugal
Ser
ra S
to A
nton
io A
Ham
pe amp
B G
arrid
oFo
rthc
omin
gFo
rthc
omin
gN
cal
cico
la 2
Men
doccedila
Port
ugal
Ser
ra S
ico
A H
ampe
amp B
Gar
rido
ndashFo
rthc
omin
gN
cua
trec
asas
ii 1
Fern
Cas
as e
t al
Sp
ain
Jaeacute
n S
ierr
a de
Caz
orla
J A
rroy
oFo
rthc
omin
gFo
rthc
omin
gN
cua
trec
asas
ii 2
Fern
Cas
as e
t al
Sp
ain
Caacuted
iz S
ierr
a de
Gra
zale
ma
J A
rroy
ondash
Fort
hcom
ing
N m
arvi
eri 1
Jah
and
amp M
aire
Mor
occo
Zer
ekte
n J
Arr
oyo
amp A
Ham
peFo
rthc
omin
gFo
rthc
omin
gN
mar
vier
i 2 J
ahan
d amp
Mai
reM
oroc
co T
azek
a J
Arr
oyo
amp A
Ter
rab
ndashFo
rthc
omin
gN
rup
icol
a 1
Leacuteon
-Duf
our
Spai
n C
uenc
a G
rado
de
Pico
P V
arga
s et
al
ndashFo
rthc
omin
gN
rup
icol
a 2
Leacuteon
-Duf
our
Spai
n M
adrid
Sie
rra
de la
Cab
rera
P V
arga
s et
al
Fort
hcom
ing
Fort
hcom
ing
N r
upic
ola
3 Leacute
on-D
ufou
rSp
ain
Cue
nca
Gra
do d
e Pi
co P
Var
gas
et a
lndash
Fort
hcom
ing
N s
cabe
rulu
s 1
Hen
riqu
esPo
rtug
al E
rved
al A
Ham
pe amp
B G
arrid
oFo
rthc
omin
gFo
rthc
omin
gN
sca
beru
lus
2 H
enri
ques
Port
ugal
Erv
edal
A H
ampe
amp B
Gar
rido
ndashFo
rthc
omin
gN
sca
beru
lus
3 H
enri
ques
Port
ugal
Oliv
eira
do
Con
de A
Ham
pe amp
B G
arrid
ondash
Fort
hcom
ing
N w
atie
ri M
aire
Mor
occo
Ouk
aim
eden
J A
rroy
o amp
A H
ampe
Fort
hcom
ing
Fort
hcom
ing
N a
lbim
argi
natu
s 1
U M
uumllle
r-D
oblie
s amp
D M
uumllle
r-D
oblie
sM
oroc
co B
ouha
chem
J A
rroy
oFo
rthc
omin
gFo
rthc
omin
gN
alb
imar
gina
tus
2 U
Muumll
ler-
Dob
lies
amp D
Muumll
ler-
Dob
lies
Mor
occo
Kel
ti F
Oje
dandash
Fort
hcom
ing
sect
Cyc
lam
inei
DC
N
cyc
lam
ineu
s D
C
Spai
n P
onte
vedr
a S
Cas
trov
iejo
616
2ndash
Fort
hcom
ing
N l
usit
anic
us D
orda
amp F
ern
Cas
asPo
rtug
al P
ena
Cov
a de
Oliv
eira
A B
arra
AB2
585
ndashFo
rthc
omin
gN
pal
lidul
us 1
Gra
ells
Spai
n M
adrid
Ald
ea d
el F
resn
o P
Var
gas
amp O
Goacutem
ezndash
Fort
hcom
ing
N p
allid
ulus
2 G
rael
lsSp
ain
Jaeacute
n A
ldea
quem
ada
R P
eacuterez
ndashFo
rthc
omin
gN
tri
andr
us L
Sp
ain
Zam
ora
Mah
ide
B H
ernaacute
ndez
ndashFo
rthc
omin
gse
ct D
ubii
Fern
Cas
asN
tor
tifo
lius
Fern
Cas
asSp
ain
Alm
eriacutea
Riacuteo
de
Agu
as A
Bar
randash
Fort
hcom
ing
sect
Jon
quill
ae D
C
N j
onqu
illa
LSp
ain
Sev
illa
Sie
rra
Nor
te R
Peacuter
ez amp
R P
eacuterez
de
Guz
maacuten
ndashFo
rthc
omin
gse
ct S
erot
ini P
arl
N s
erot
inus
Loe
fl e
x L
Spai
n S
evill
a D
os H
erm
anas
C A
ndreacute
s amp
Z D
iacuteaz
ndashFo
rthc
omin
gse
ct T
azet
tae
DC
N
taz
etta
L
Gen
Bank
ndashA
J232
562
sect
Bul
boco
dii D
C
N b
ulbo
codi
um L
Sp
ain
Hue
lva
Car
taya
J C
astil
loFo
rthc
omin
gFo
rthc
omin
gse
ct N
arci
ssus
L
N p
oeti
cus
LSp
ain
Leacuter
ida
Val
le d
e A
raacuten
X P
icoacute
ndashFo
rthc
omin
gse
ct P
seud
onar
ciss
i DC
N
pse
udon
arci
ssus
L
Spai
n Aacute
vila
Sie
rra
de G
redo
s R
Peacuter
ez amp
J A
rroy
ondash
Fort
hcom
ing
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 239
means that in one dimorphic or distylous Narcissus speciesthe long style (L-morph) has its reciprocal male organ in upperstamens of S-morph Accordingly S-style is the reciprocal tolower stamens of L-morph (Fig 1) Flowers of the tristylousNarcissus pallidulus does not pose special problems as thiscondition usually involves two stamen whorls and the com-parison of reciprocal heights is straightforward as originallydescribed by Darwin (1877) in Lythrum salicaria We aver-aged the absolute values of this difference across the morphsof each population in order to compare dimorphic and tri-morphic species A perfectly reciprocal population wouldtherefore rate a value of zero In order to control for effects offlower size in the extent of reciprocity we also divided the rec-iprocity of a population by its mean flower diameter
Statistical analyses All morphometrical variables were checkedfor normality and homoscedasticity and transformed whennecessary Only flower tube length across populations andspecies needed to be log transformed Data were averaged formorphs and populations and compared accordingly Studentt-tests and one-way were used for univariate compar-isons of flower features between morphs for dimorphic andtrimorphic species respectively In the rare cases wherevariances were not homogeneous (Levenersquos test) we applieda t-test with separate variance estimates After fordifference of means posthoc Tukey HSD test for unequalsample sizes was applied for multiple comparison betweenpairs When performing multiple tests of one hypothesisDunn-Sidak correction of sequential Bonferroni adjustementof α levels was applied (Sokal amp Rohlf 1995) Reciprocity(between-morph stigma-anther separation) was not comparedbetween morphs given that it was obtained as a differencebetween morph means hence error estimates of the differencewere not available Multivariate analyses were performed tocheck the overall pattern of morphological variation of flower
perianth An ordination by Principal Component Analysis(PCA) was performed to draw the morphological relationshipsbetween populations and species In multivariate analysispopulation means were used rather than individual data inorder to avoid pseudoreplication An analysis of the intra-population variation (Faivre amp McDade 2001) is intendedto be published in a subsequent paper
To analyse the relationships between perianth morphologyand sex organ polymorphism we applied multiple regressionanalyses Independent variables were the scores of the populationmeans for each of three first PCA axes We used these perianthpseudovariables rather than raw univariate values because axesaccounted for a high variance of perianth morphology andbecause axes are by definition orthogonal overcoming the seriousproblem of multicollinearity in multiple regression ( James ampMcCulloch 1990 Philippi 1993) Two regression modelswere adjusted with respective dependent variables related toreciprocity of stigma-anther separation of each population(i) the mean between morphs of absolute values for differencebetween style height and equivalent stamens whorl height and(ii) the values of (i) divided by flower diameter The latter var-iable was considered to control for potential allometric effectof flower size on the magnitude of departures from stigma-anther reciprocity (see above) Values for monomorphic popu-lations were scored as the highest obtained value of departurefrom exact reciprocity Ordination analysis was performedwith PCORD package (McCune amp Mefford 1999) whereasunivariate statistics and multiple regression models were com-puted with STATISTICA software (Statsoft 1997)
Pollinators
Data were taken from populations of four Apodanthi spe-cies some in common with those of the study of flowermorphometry (Table 1) We performed systematic censuses
Fig 1 Long-styled (L) and short-styled (S) flowers of a typical style-dimorphic Narcissus species (N cuatrecasasii) Numbers correspond to the measurements taken in each flower sampled (1) flower width (2) tepal length (3) corona width (4) corona height (5) tube length (6) tube width (7) style length (8) upper stamen height (9) lower stamen height (10) angle of flower to stalk Measurements were accordingly modified for tristylous distylous and monomorphic species
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research240
of insects visiting the flowers and recorded positive dataproviding that they enter into the tube and do contact any sexorgan Species subject to study were N rupicola (time effortof 12 h including some nocturnal periods) N cuatrecasasii(12 h) N albimarginatus (9 h) and N watieri (3 h) Observationtiming and area depend on density of flowers in everypopulation In sparse patches we made several censuses of15 min observing all insects visiting flowers in an area ofapprox 100 m2 We also made static observations in densepatches on all flowers in small areas (approx 10 m2)
Phylogenetic studies
PCR amplification and sequencing Total genomic DNA wasextracted from silica-dried material collected in the field andthe living collection of the Royal Botanic Garden of MadridSpain DNA extractions were performed using the DNeasyPlant Mini Kit (QIAGEN Inc Chatsworth CA USA) ForPolymerase Chain Reaction (PCR) amplifications we used aPerkin-Elmer (Foster City CA USA) PCR System 9700 thermalcycler primers trna and trnb for trnT(UGU)-trnL(UAA)trne and trn f for trnL(UAA)-trnF(GAA) and PCR conditionsfollowing Taberlet et al (1991) 51degC of annealing temperaturefor the trnT-L and 50degC for the trnL-F plus 2-min elongationtime Amplified products were cleaned using spin filter columns(PCR Clean-up kit MoBio Laboratories Solana Beach CAUSA) following the protocols provided by the manufacturerCleaned products were then directly sequenced using dyeterminators (Big Dye Terminator vs 20 Applied BiosystemsLittle Chalfont UK) and run into polyacrylamideelectrophoresis gels (7) using an Applied Biosystems Prismmodel 3700 automated sequencer Primers trna trnb trneand trn f were also used for cycle sequencing of the trnT-trnLand trnL-trnF spacers under the following conditions 95degCfor 2 min followed by 25 cycles of 95degC for 10 s 50degC for 5s and 60degC for 4 min Sequence data were placed in a contigfile and edited using the program Seqed (Applied BiosystemsFoster City CA USA) The limits of the trnT-trnL and trnL-trnF regions were determined by comparison with otherAmaryllidaceae sequences available in the GenBank
Sequence analyses Three different sequence matrices wereused to perform the phylogenetic analyses trnL-F matrixof Narcissus trnT-L matrix of Narcissus sect Apodanthiand a combined matrix of trnL-F and trnT-L for only thespecies of Narcissus sect Apodanthi as ingroup Alignmentsof the 30 trnL-F and the nine trnT-L sequences wereobtained using the program Clustal X 162b (httpevolutiongeneticswashingtoneduphylipsoftwareetc1html)with further manual adjustments Two phylogenetic analyseswere performed by using two primary approaches to treeconstruction tree-searching (parsimony) and algorithmic(Neighbor-Joining) Parsimony-based analyses were conductedusing Fitch parsimony (as implemented in Swofford
1999) with equal weighting of all characters and oftransitionstransversions Heuristic searches were replicated100 times with random taxon-addition sequences TreeBisection-Reconnection (TBR) branch swapping and withthe options MULPARS and STEEPEST DESCENT ineffect Relative support for clades identified by parsimonyanalysis was assessed by bootstrapping with the heuristicsearch strategy as indicated above In addition phylogeneticreconstructions of sequences were also performed in using the Kimura 2-parameter distance model (Kimura 1980)and the Neighbor-Joining method (Saitou amp Nei 1987)Indels in the noncoding regions trnT-L and trnL-F werecoded as appended characters following the logic of Kelchner(2000) and Simmons amp Ochoterena (2000) As a resultadditional analyses were performed when coding parsimony-informative indels affecting the trnL-F and trnT-L matricesCharacter evolution of flower polymorphism was investigatedusing MacClade 35 (Maddison amp Maddison 1992) Fourstates of flower polymorphism (approach herkogamy non-herkogamous monomorphism style dimorphism and hetero-styly) were traced and mapped on the resulting trnL-F andtrnT-L phylogenies of the consensus tree of all maximum-parsimony trees We explored both accelerated transformation(ACCTRAN) and delayed transformation (DELTRAN)optimizations with equal weighting of all characters andtransitions among all states equally probable
Results
Style polymorphism
Average values of sex organ position for the seven species ofsect Apodanthi and for N pallidulus are shown in Tables 3 4and 5 and comparative diagrams in Fig 2 Species in sectionApodanthi have a remarkable variation in style polymorphismdistyly in N albimarginatus style dimorphism in four species(N calcicola N cuatrecasasii N rupicola and N scaberulus)and monomorphism in N marvieri and N watieri (Fig 2Table 3) Narcissus albimarginatus shows two style-lengthmorphs with approximate reciprocal positioning of anthersand style Two patterns were found for dimorphic speciesthree species (N calcicola N scaberulus N cuatrecasasii ) presentthe L-morph with the stigma almost at an equivalent level ofthat of upper anthers whereas in N rupicola the stigma of theL-morph is positioned between the two anther levels The latterspecies displayed the highest values of stigma-anther separation(Table 4) that is the lowest reciprocity among dimorphicspecies Monomorphic species present a position of sex organssimilar to that in L-morph of dimorphic N rupicola (Fig 2)
Narcissus pallidulus is the only trimorphic species in ourstudy with mostly reciprocal positioning of stigmas andstamen whorls Although equivalent positions are not equal inall cases sequence height of sex organs corresponds to typicaltristylous species
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 241
At the population level morphs are in general unequallyrepresented L-morph is the most frequent (Tables 4 and 5) inall species and populations with the exception of one popu-lation of N cuatrecasasii and one of N rupicola which showsimilar frequency of morphs (isoplethy)
Perianth variation
Values of perianth variables are averaged over populationsand morphs (Table 6) Univariate morphometrical analyses offlower tubes tepals coronas and flower angle of all studiedspecies and populations revealed clearly absence of significantdifferences between morphs (82 between-morph comparisonsresults not shown) There were only slightly significant dif-ferences in tristylous N pallidulus at Cazorla for corona height(P = 001) and in distylous N albimarginatus at Bouhachemfor perianth width (P = 005)
Principal component analysis is shown in Fig 3 The firstsecond and third axes accounted for 616 233 and 94of the variance although only Axes 1 and 2 have eigenvalueshigher than 10 and are hence further discussed These twoaxes depict quite clearly the relationships between species andpopulations in perianth morphology At the negative extremeof Axis 1 we find together N pallidulus and N albimarginatusand at the positive extreme of this axis N marvieri Nwatieri and N rupicola form a second group Axis 2 separatesat its positive end the Iberian endemics N cuatrecasasii N cal-cicola and N scaberulus Eigenvectors of perianth variablesreflect trends of morphological variation The highest valuesfor Axis 1 are flower angle to stalk (04519) width of flowertube (minus04410) and corona height (minus04217) Largest eigen-vectors for Axis 2 correspond to corona width (minus06470)tepal length (minus04362) and flower tube length (minus04087)These results indicate that populations of two species (Npallidulus and N albimarginatus) at the left in the ordinationplot are characterised by small angles (pendulous flowers)wide flower tubes and long coronas (Fig 4) A second groupof species (N rupicola N watieri N marvieri ) display pre-dominantly erect flowers narrow and very long tubes andfunnel-shaped coronas On the top of the ordination plotthree species (N cuatrecasasii N calcicola N scaberulus) arecharacterised by patent flowers cup-shaped coronas smalltepals and moderate flower-tube length
The interrelationship between perianth variables and stylarpolymorphism is evaluated by means of multiple regressionanalyses We calculated two measurements of reciprocity inthe position of sex organs considering or not flower sizeeffects Both regression equations fitted quite similarly withscores of populations along PCA axes on perianth variablesAs a result we may conclude that there is a significant rela-tionship between perianth morphology and sex organ posi-tion across populations and species with 72ndash80 of thevariance being explained by the regression models (Table 7)Positioning of reciprocal stigma and anthers (heterostyly) cor-responds with the group of pendulous flowers Style dimorphicspecies (N calcicola N cuatrecasasii N scaberulus see Table 4)with medium values of reciprocity have patent flowers amongother features (see above) Monomorphism and less reciprocalstyle dimorphism in N rupicola (Table 4) correspond to theerect flowered group
Pollinators
As typically occurs in early blooming species it was difficultto observe pollinators in four species of Narcissus sectApodanthi (0ndash467 insect visits per hour) Twelve hours ofobservations in N cuatrecasasii ( Jaeacuten Cazorla range) revealedrarity of pollinators although medium sized solitary bees(Anthophora sp) were consistently spotted (10 visits) probingnectar and small unidentified bees (8 visits) reaching pollenof the upper anthers In the same area Anthophora bees visitedflowers of Helleborus foetidus but always in different boutsCasual observations made on populations at the oppositeextreme of the geographic range of N cuatrecasasii (CaacutedizGrazalema range) allow recognising Anthophora as the mostcommon visitor By censuses in N rupicola at the core of itsgeographic range (Avila Puerto de Mijares Madrid Puertode Navacerrada) we observed only two visits by the mothMacroglossum stellatarum one by the hoverfly Eristalis tenaxand one by an unidentified small bee Flies and small beestried to enter the flower tube but they only contacted upperanthers and L-stigma because of its narrowness and shorttongues Infrequent pollinator visits is not the result of lownumber of insects as they visited profusely another co-occurring flowers of some genera (Gagea Ranunculus SaxifragaArmeria Cytisus Erysimum) Some nocturnal censuses revealed
Table 3 Height of stamens and stigmas in monomorphic species of Narcissus sect Apodanthi Values are means plusmn 1 se in mm
Species Population Sample size Style heightUpper anther height
Lower anther height
Stigma-Anther separation
N marvieri Zeretken 51 2069 plusmn 030 2255 plusmn 032 1852 plusmn 035 186N marvieri Tazzeka 50 1001 plusmn 015 1234 plusmn 018 865 plusmn 014 233N watieri Oukaimeden 50 2092 plusmn 028 2205 plusmn 029 1815 plusmn 023 113N watieri TizirsquonrsquoTest 26 1736 plusmn 030 1927 plusmn 040 1595 plusmn 041 191
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research242
Tabl
e 4
Hei
ght
of s
tam
ens
and
stig
mas
in d
imor
phic
spe
cies
of
Nar
ciss
us s
ect
Apo
dant
hi L
lon
g-st
yled
mor
ph S
sho
rt-s
tyle
d m
orph
Val
ues
are
mea
ns plusmn
1 s
e in
mm
For
est
imat
ing
stig
ma-
anth
er s
epar
atio
n w
e co
nsid
er o
nly
the
equi
vale
nt w
horl
to th
e st
igm
a of
the
alte
rnat
e m
orph
s to
be
the
reci
proc
al (L
stig
ma
vs u
pper
sta
men
who
rl of
S fl
ower
S s
tigm
a vs
low
er s
tam
en w
horl
of L
flow
er s
ee F
ig 1
) A
ppro
pria
te t
-tes
ts a
re g
iven
for
the
diff
eren
ces
betw
een
mea
ns o
f th
ese
leng
ths
(P
lt 0
05
P
lt 0
01
P lt
00
01)
Spec
ies
Popu
latio
n
Sam
ple
size
L S
Styl
e he
ight
U
pper
ant
her
heig
ht
Low
er a
nthe
r he
ight
St
igm
a-A
nthe
r se
para
tion
Mor
ph r
atio
(1
)L
SL
SL
SL
S
N a
lbim
argi
natu
sBo
uhac
hem
69 5
824
40
plusmn 3
8
9
77 plusmn
17
015
97
plusmn 2
14
24
54
plusmn 2
899
96 plusmn
17
8
21
80
plusmn 2
550
14
019
065
0
35N
alb
imar
gina
tus
Kel
ti53
45
243
7 plusmn
032
936
plusmn 0
19
148
9 plusmn
028
216
7 plusmn
036
849
plusmn 0
23
188
1 plusmn
041
270
0
870
66
034
N c
uatr
ecas
asii
Gra
zale
ma
66 3
415
08
plusmn 0
20
8
01 plusmn
01
713
79
plusmn 0
18 n
s 14
36
plusmn 0
269
05 plusmn
01
6
12
03
plusmn 0
270
72
104
065
0
35N
cua
trec
asas
iiC
azor
la42
58
146
0 plusmn
020
690
plusmn 0
12
131
8 plusmn
018
13
95
plusmn 0
158
25 plusmn
01
4
10
74
plusmn 0
110
65
162
042
0
58N
cal
cico
laSi
coacute50
43
163
6 plusmn
034
819
plusmn 0
18
144
6 plusmn
018
152
5 plusmn
029
925
plusmn 0
21
113
2 plusmn
018
112
1
380
69
031
N c
alci
cola
Sto
Ant
onio
51 4
416
49
plusmn 0
29
8
06 plusmn
02
715
30
plusmn 0
23 n
s 15
37
plusmn 0
229
44 plusmn
02
1
11
68
plusmn 0
231
11
106
076
0
24N
sca
beru
lus
Oliv
eira
51 3
115
12
plusmn 0
24
7
72 plusmn
01
714
56
plusmn 0
15 n
s 15
05
plusmn 0
279
21 plusmn
01
6
11
50
plusmn 0
240
071
490
86
014
N s
cabe
rulu
sEr
veda
l50
14
147
7 plusmn
029
778
plusmn 0
28
148
5 plusmn
021
ns
149
9 plusmn
065
986
plusmn 0
15
11
71
plusmn 0
370
22
208
077
0
23N
rup
icol
aA
ldea
quem
ada
42 3
614
94
plusmn 0
50
10
43
plusmn 0
2216
08
plusmn 0
27
18
19
plusmn 0
26 1
109
plusmn 0
25
139
8 plusmn
021
325
0
660
54
046
N r
upic
ola
Bola
del
Mun
do53
35
131
9 plusmn
024
816
plusmn 0
14
154
5 plusmn
023
ns
163
1 plusmn
029
105
6 plusmn
018
126
6 plusmn
025
312
2
400
72
028
(1)
Sam
ple
size
use
d fo
r m
orph
rat
io is
the
tot
al n
umbe
r of
flow
ers
colle
cted
(se
e Ta
ble
1)
occurrence of some moths around the flowers of N rupicolabut they were not observed to come into contact with flowersIn N watieri insect visits were much more abundant After a3-hour census we observed long-tongued visitors (diurnalmoths six visits pierid butterflies three visits) and short-tongued hoverflies (five visits) Despite considerable time(9 h) spent on N albimarginatus we were not able to spot anypollinator
Chloroplast sequence variation
Length of trnL-F sequences in Narcissus ranges between337 bp (population 1 of Narcissus pallidulus) and 365 bp(population 2 of Narcissus marvieri ) This sequence variationis caused by seven indels of 1ndash19 bp Number of variableparsimony-informative characters is of 1711 across species ofNarcissus The highest pairwise divergence (Kimura-2-parametermodel) of species sequences (34) was found betweenaccessions of N bulbocodium and N cyclamineus Sequencedivergence neither was found between species of sectApodanthi (excluding N calcicola and N scaberulus) nor wasobserved between two of the three species of the N triandruscomplex (N pallidulus and N lusitanicus) Kimura-2-parameter distances between accessions of the same speciesdid not rendered either divergence
In Narcissus sect Apodanthi length variation of trnT-Lsequences ranges between 827 bp (N marvieri ) and 966 bp(N calcicola) Ten gaps in the trnT-L sequence matrix from1 bp to 130 bp are responsible for great sequence variationAn unusual indel of 130 bp is shared by N rupicola N marv-ieri and N watieri The number of variableparsimony-informative characters is 1911 in Narcissus sect ApodanthiThe highest pairwise divergence (Kimura-2-parameter model)between sequence species (14) was found between acces-sions of N calcicola and N cuatrecasasii whereas the lowest(021) was between N calcicola and N scaberulus
Phylogenetic relationships
Maximum parsimony (MP) using the 27 trnL-F sequencesof Narcissus and Galanthus Lapiedra and Leucojum as theoutgroup yielded 1529 trees of 78 steps (CI 095 and RI093 including uninformative characters) (results not shown)A congruent somewhat more resolved topology was retrievedin the semistrict consensus tree of 13 225 trees (101 stepsCI 085 and RI 085) when coding the seven indels asadditional characters however new branches do not displaysignificant values (bootstrap below 50) (Fig 5) From bothanalyses (with and without indel coding) a basal polytomy isdepicted in the semistrict consensus tree of which two well-supported clades are formed by the accessions of five speciesof sect Apodanthi (clade 1) and six species including the threeof the N triandrus complex (clade 2) Relatively long branchesin the Neighbor-Joining (NJ) trees using both alignments
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
copy
New Phytologist
(2003)
161
235ndash252
wwwnewphytologistorg
Research 237
Tabl
e 1
Spec
ies
of
Nar
ciss
us
sam
pled
for
flow
er (
peria
nth
and
sex
orga
ns)
mor
phom
etry
()
Pop
ulat
ions
whe
re in
sect
cen
suse
s w
ere
done
Spec
ies
Popu
latio
nC
oord
inat
esH
abita
tEl
evat
ion
Flow
ers
colle
cted
N p
allid
ulus
1 Sp
ain
Jaeacute
n S
egur
a N
aval
caba
llo38
deg
20
prime
N 2
deg
37
prime
WPi
ne f
ores
t on
lim
esto
ne13
5012
2
N p
allid
ulus
2 Sp
ain
Bad
ajoz
Cal
era
Sie
rra
de T
entu
dia
38
deg
5
prime
N 6
deg
20
prime
WD
ecid
uous
oak
for
est
on s
hale
900
63
N a
lbim
argi
natu
s
()
3 M
oroc
co C
haou
en J
bel B
ouha
chem
35
deg
15
prime
N 5
deg
26
prime
WC
edar
for
est
on s
ands
tone
1500
200
N a
lbim
argi
natu
s
4 M
oroc
co C
haou
en J
bel K
elti
35
deg
21
prime
N 5
deg
17
prime
WO
ak f
ores
t on
lim
esto
ne15
2020
6
N c
uatr
ecas
asii
()
5 Sp
ain
Caacuted
iz G
raza
lem
a P
uert
o de
l Boy
ar36
deg
46
prime
N 5
deg
24
prime
WR
ock
fissu
res
on li
mes
tone
1100
100
N c
uatr
ecas
asii
6 Sp
ain
Jaeacute
n C
azor
la L
as C
orre
huel
as37
deg
59
prime
N 2
deg
54
prime
WM
ixed
pin
e fo
rest
on
limes
tone
1615
100
N c
alci
cola
7 Po
rtug
al L
eiriacutea
Ser
ra d
e Si
coacute39
deg
55
prime
N 8
deg
32
prime
WR
ock
fissu
res
on li
mes
tone
550
186
N c
alci
cola
8 Po
rtug
al L
eiriacutea
Ser
ra d
e St
o A
nton
io39
deg
32
prime
N 8
deg
44
prime
WR
ock
fissu
res
on li
mes
tone
714
212
N s
cabe
rulu
s
9 Po
rtug
al V
iseu
Oliv
eira
do
Con
de40
deg
26
prime
N 7
deg
57
prime
WSc
rubl
and
on g
rani
te25
923
2
N s
cabe
rulu
s
10 P
ortu
gal
Gua
rda
Erv
edal
40
deg
26
prime
N 7
deg54prime
WSc
rubl
and
on g
rani
te33
310
9N
rup
icol
a11
Spa
in J
aeacuten
Ald
eaqu
emad
a38
deg23prime
N 3
deg24prime
WC
liff
fissu
res
on g
rani
te10
0078
N r
upic
ola
()
12 S
pain
Mad
rid N
avac
erra
da B
ola
del M
undo
40deg4
7prime N
3deg5
9prime W
Alp
ine
scru
blan
d on
gra
nite
2257
154
N m
arvi
eri
13 M
oroc
co M
arra
kesh
Zer
etke
n31
deg27prime
N 7
deg26prime
WTr
ee-l
ine
ceda
r fo
rest
on
shal
e22
5051
N m
arvi
eri
14 M
oroc
co T
aza
Taz
zeka
34deg5
prime N 4
deg11prime
WD
ense
ced
ar f
ores
t an
d sc
rubl
and
on s
hale
1837
50N
wat
ieri
()
15 M
oroc
co M
arra
kesh
Ouk
aim
eden
31deg1
3prime N
7deg5
1prime W
Soil
pock
ets
on a
lpin
e m
eado
ws
on s
hale
2474
50N
wat
ieri
16 M
oroc
co M
arra
kesh
Tiz
irsquonrsquoT
est
31deg5
prime N 8
deg5prime W
Hol
m o
ak f
ores
t on
lim
esto
ne20
5040
earliest in the inflorescence At least 40 flowers wereimmediately preserved in ethanol 70 and kept refrigerateduntil measurement Given that species present vegetativemultiplication through division of bulbs an effort was made toavoid repeated sampling of ramets by collecting flowers inplants at least 2 m apart (Arroyo et al 2002) From eachpopulation we obtained additional flowers with the sameprocedure in order to have larger samples for estimatingmorph ratio in populations (see Table 1 for sample sizes)
Phylogenetic study Individuals and populations were selectedto represent the morphological diversity of the genus thegeographical breadth of the species and flower polymorphismThe seven species of Narcissus sect Apodanthi the four ofNarcissus sect Cyclaminei and seven representatives of the10 remaining sections were sampled (a single individualper population) and sequenced for the noncoding regiontrnL(UAA)-trnF(GAA) (Table 2) As a result we obtained andanalyzed 27 sequences of Narcissus Outgroup species (GalanthusLapiedra Pancratium) were chosen as suggested in a previousphylogeny of Amaryllidaceae based on rbcL and trnL-trnFsequences (Meerow et al 1999) Additionally a second samplefocused on the seven species of Narcissus sect Apodanthi wasaddressed to sequence and analyze the chloroplast noncodingregion trnT(UGU)-trnL(UAA) Narcissus bulbocodium andN papyraceus were used as outgroup species based on thetrnL-F phylogeny of Narcissus herein presented
Flower morphometry
Measurements Typically a Narcissus species has a flower tubeon the inferior ovary to which the six stamen filaments areattached two stamen whorls six tepals and above all acorona Each stamen whorl has three stamens reaching tworespective anther levels in sect Apodanthi and N triandruscomplex (Fig 1) Preserved flowers were slit longitudinallyfrom just above the ovary to the tube mouth The apex of theovary was the baseline for all measurements which wereperformed by digital calipers to the nearest 01 mm Measure-ments were (Fig 1) flower width tepal length corona widthcorona height tube length tube mouth width style lengthupper stamen height lower stamen height and flower angle tostalk Stamen height was measured up to the insertion of thefilament into the anther avoiding variation caused by anthersize Anthers display up to threefold change in length duringtheir fast ripening (R Peacuterez personal observation)
Given that all studied species have two whorls of stamensdifferent in length it is difficult to quantify the level of sexorgan reciprocity in dimorphic species which usually haveone stamen level in distylous species On functional grounds(ie proficiency of cross pollination between equivalent heightof sex organs) we have computed the degree of reciprocity asthe difference between means of style height in each morphand the closest stamen whorl of the alternative morph It
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research238
Tabl
e 2
Acc
essi
ons
for
the
chlo
ropl
ast
sequ
ence
stu
dy o
f N
arci
ssus
inc
ludi
ng p
lant
iden
tifica
tion
follo
wed
by
popu
latio
n nu
mbe
r lo
calit
y of
wild
pop
ulat
ions
nam
e of
pla
nt c
olle
ctor
s an
d G
enBa
nk a
cces
sion
num
bers
for
trn
T-L
and
trnL
-F s
eque
nces
Taxo
nLo
calit
y
Gen
Bank
ac
cess
ion
num
ber
trnT
-L
Gen
Bank
ac
cess
ion
num
ber
trnL
-F
sect
Apo
dant
hi A
Fer
nand
esN
cal
cico
la 1
Men
doccedila
Port
ugal
Ser
ra S
to A
nton
io A
Ham
pe amp
B G
arrid
oFo
rthc
omin
gFo
rthc
omin
gN
cal
cico
la 2
Men
doccedila
Port
ugal
Ser
ra S
ico
A H
ampe
amp B
Gar
rido
ndashFo
rthc
omin
gN
cua
trec
asas
ii 1
Fern
Cas
as e
t al
Sp
ain
Jaeacute
n S
ierr
a de
Caz
orla
J A
rroy
oFo
rthc
omin
gFo
rthc
omin
gN
cua
trec
asas
ii 2
Fern
Cas
as e
t al
Sp
ain
Caacuted
iz S
ierr
a de
Gra
zale
ma
J A
rroy
ondash
Fort
hcom
ing
N m
arvi
eri 1
Jah
and
amp M
aire
Mor
occo
Zer
ekte
n J
Arr
oyo
amp A
Ham
peFo
rthc
omin
gFo
rthc
omin
gN
mar
vier
i 2 J
ahan
d amp
Mai
reM
oroc
co T
azek
a J
Arr
oyo
amp A
Ter
rab
ndashFo
rthc
omin
gN
rup
icol
a 1
Leacuteon
-Duf
our
Spai
n C
uenc
a G
rado
de
Pico
P V
arga
s et
al
ndashFo
rthc
omin
gN
rup
icol
a 2
Leacuteon
-Duf
our
Spai
n M
adrid
Sie
rra
de la
Cab
rera
P V
arga
s et
al
Fort
hcom
ing
Fort
hcom
ing
N r
upic
ola
3 Leacute
on-D
ufou
rSp
ain
Cue
nca
Gra
do d
e Pi
co P
Var
gas
et a
lndash
Fort
hcom
ing
N s
cabe
rulu
s 1
Hen
riqu
esPo
rtug
al E
rved
al A
Ham
pe amp
B G
arrid
oFo
rthc
omin
gFo
rthc
omin
gN
sca
beru
lus
2 H
enri
ques
Port
ugal
Erv
edal
A H
ampe
amp B
Gar
rido
ndashFo
rthc
omin
gN
sca
beru
lus
3 H
enri
ques
Port
ugal
Oliv
eira
do
Con
de A
Ham
pe amp
B G
arrid
ondash
Fort
hcom
ing
N w
atie
ri M
aire
Mor
occo
Ouk
aim
eden
J A
rroy
o amp
A H
ampe
Fort
hcom
ing
Fort
hcom
ing
N a
lbim
argi
natu
s 1
U M
uumllle
r-D
oblie
s amp
D M
uumllle
r-D
oblie
sM
oroc
co B
ouha
chem
J A
rroy
oFo
rthc
omin
gFo
rthc
omin
gN
alb
imar
gina
tus
2 U
Muumll
ler-
Dob
lies
amp D
Muumll
ler-
Dob
lies
Mor
occo
Kel
ti F
Oje
dandash
Fort
hcom
ing
sect
Cyc
lam
inei
DC
N
cyc
lam
ineu
s D
C
Spai
n P
onte
vedr
a S
Cas
trov
iejo
616
2ndash
Fort
hcom
ing
N l
usit
anic
us D
orda
amp F
ern
Cas
asPo
rtug
al P
ena
Cov
a de
Oliv
eira
A B
arra
AB2
585
ndashFo
rthc
omin
gN
pal
lidul
us 1
Gra
ells
Spai
n M
adrid
Ald
ea d
el F
resn
o P
Var
gas
amp O
Goacutem
ezndash
Fort
hcom
ing
N p
allid
ulus
2 G
rael
lsSp
ain
Jaeacute
n A
ldea
quem
ada
R P
eacuterez
ndashFo
rthc
omin
gN
tri
andr
us L
Sp
ain
Zam
ora
Mah
ide
B H
ernaacute
ndez
ndashFo
rthc
omin
gse
ct D
ubii
Fern
Cas
asN
tor
tifo
lius
Fern
Cas
asSp
ain
Alm
eriacutea
Riacuteo
de
Agu
as A
Bar
randash
Fort
hcom
ing
sect
Jon
quill
ae D
C
N j
onqu
illa
LSp
ain
Sev
illa
Sie
rra
Nor
te R
Peacuter
ez amp
R P
eacuterez
de
Guz
maacuten
ndashFo
rthc
omin
gse
ct S
erot
ini P
arl
N s
erot
inus
Loe
fl e
x L
Spai
n S
evill
a D
os H
erm
anas
C A
ndreacute
s amp
Z D
iacuteaz
ndashFo
rthc
omin
gse
ct T
azet
tae
DC
N
taz
etta
L
Gen
Bank
ndashA
J232
562
sect
Bul
boco
dii D
C
N b
ulbo
codi
um L
Sp
ain
Hue
lva
Car
taya
J C
astil
loFo
rthc
omin
gFo
rthc
omin
gse
ct N
arci
ssus
L
N p
oeti
cus
LSp
ain
Leacuter
ida
Val
le d
e A
raacuten
X P
icoacute
ndashFo
rthc
omin
gse
ct P
seud
onar
ciss
i DC
N
pse
udon
arci
ssus
L
Spai
n Aacute
vila
Sie
rra
de G
redo
s R
Peacuter
ez amp
J A
rroy
ondash
Fort
hcom
ing
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 239
means that in one dimorphic or distylous Narcissus speciesthe long style (L-morph) has its reciprocal male organ in upperstamens of S-morph Accordingly S-style is the reciprocal tolower stamens of L-morph (Fig 1) Flowers of the tristylousNarcissus pallidulus does not pose special problems as thiscondition usually involves two stamen whorls and the com-parison of reciprocal heights is straightforward as originallydescribed by Darwin (1877) in Lythrum salicaria We aver-aged the absolute values of this difference across the morphsof each population in order to compare dimorphic and tri-morphic species A perfectly reciprocal population wouldtherefore rate a value of zero In order to control for effects offlower size in the extent of reciprocity we also divided the rec-iprocity of a population by its mean flower diameter
Statistical analyses All morphometrical variables were checkedfor normality and homoscedasticity and transformed whennecessary Only flower tube length across populations andspecies needed to be log transformed Data were averaged formorphs and populations and compared accordingly Studentt-tests and one-way were used for univariate compar-isons of flower features between morphs for dimorphic andtrimorphic species respectively In the rare cases wherevariances were not homogeneous (Levenersquos test) we applieda t-test with separate variance estimates After fordifference of means posthoc Tukey HSD test for unequalsample sizes was applied for multiple comparison betweenpairs When performing multiple tests of one hypothesisDunn-Sidak correction of sequential Bonferroni adjustementof α levels was applied (Sokal amp Rohlf 1995) Reciprocity(between-morph stigma-anther separation) was not comparedbetween morphs given that it was obtained as a differencebetween morph means hence error estimates of the differencewere not available Multivariate analyses were performed tocheck the overall pattern of morphological variation of flower
perianth An ordination by Principal Component Analysis(PCA) was performed to draw the morphological relationshipsbetween populations and species In multivariate analysispopulation means were used rather than individual data inorder to avoid pseudoreplication An analysis of the intra-population variation (Faivre amp McDade 2001) is intendedto be published in a subsequent paper
To analyse the relationships between perianth morphologyand sex organ polymorphism we applied multiple regressionanalyses Independent variables were the scores of the populationmeans for each of three first PCA axes We used these perianthpseudovariables rather than raw univariate values because axesaccounted for a high variance of perianth morphology andbecause axes are by definition orthogonal overcoming the seriousproblem of multicollinearity in multiple regression ( James ampMcCulloch 1990 Philippi 1993) Two regression modelswere adjusted with respective dependent variables related toreciprocity of stigma-anther separation of each population(i) the mean between morphs of absolute values for differencebetween style height and equivalent stamens whorl height and(ii) the values of (i) divided by flower diameter The latter var-iable was considered to control for potential allometric effectof flower size on the magnitude of departures from stigma-anther reciprocity (see above) Values for monomorphic popu-lations were scored as the highest obtained value of departurefrom exact reciprocity Ordination analysis was performedwith PCORD package (McCune amp Mefford 1999) whereasunivariate statistics and multiple regression models were com-puted with STATISTICA software (Statsoft 1997)
Pollinators
Data were taken from populations of four Apodanthi spe-cies some in common with those of the study of flowermorphometry (Table 1) We performed systematic censuses
Fig 1 Long-styled (L) and short-styled (S) flowers of a typical style-dimorphic Narcissus species (N cuatrecasasii) Numbers correspond to the measurements taken in each flower sampled (1) flower width (2) tepal length (3) corona width (4) corona height (5) tube length (6) tube width (7) style length (8) upper stamen height (9) lower stamen height (10) angle of flower to stalk Measurements were accordingly modified for tristylous distylous and monomorphic species
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research240
of insects visiting the flowers and recorded positive dataproviding that they enter into the tube and do contact any sexorgan Species subject to study were N rupicola (time effortof 12 h including some nocturnal periods) N cuatrecasasii(12 h) N albimarginatus (9 h) and N watieri (3 h) Observationtiming and area depend on density of flowers in everypopulation In sparse patches we made several censuses of15 min observing all insects visiting flowers in an area ofapprox 100 m2 We also made static observations in densepatches on all flowers in small areas (approx 10 m2)
Phylogenetic studies
PCR amplification and sequencing Total genomic DNA wasextracted from silica-dried material collected in the field andthe living collection of the Royal Botanic Garden of MadridSpain DNA extractions were performed using the DNeasyPlant Mini Kit (QIAGEN Inc Chatsworth CA USA) ForPolymerase Chain Reaction (PCR) amplifications we used aPerkin-Elmer (Foster City CA USA) PCR System 9700 thermalcycler primers trna and trnb for trnT(UGU)-trnL(UAA)trne and trn f for trnL(UAA)-trnF(GAA) and PCR conditionsfollowing Taberlet et al (1991) 51degC of annealing temperaturefor the trnT-L and 50degC for the trnL-F plus 2-min elongationtime Amplified products were cleaned using spin filter columns(PCR Clean-up kit MoBio Laboratories Solana Beach CAUSA) following the protocols provided by the manufacturerCleaned products were then directly sequenced using dyeterminators (Big Dye Terminator vs 20 Applied BiosystemsLittle Chalfont UK) and run into polyacrylamideelectrophoresis gels (7) using an Applied Biosystems Prismmodel 3700 automated sequencer Primers trna trnb trneand trn f were also used for cycle sequencing of the trnT-trnLand trnL-trnF spacers under the following conditions 95degCfor 2 min followed by 25 cycles of 95degC for 10 s 50degC for 5s and 60degC for 4 min Sequence data were placed in a contigfile and edited using the program Seqed (Applied BiosystemsFoster City CA USA) The limits of the trnT-trnL and trnL-trnF regions were determined by comparison with otherAmaryllidaceae sequences available in the GenBank
Sequence analyses Three different sequence matrices wereused to perform the phylogenetic analyses trnL-F matrixof Narcissus trnT-L matrix of Narcissus sect Apodanthiand a combined matrix of trnL-F and trnT-L for only thespecies of Narcissus sect Apodanthi as ingroup Alignmentsof the 30 trnL-F and the nine trnT-L sequences wereobtained using the program Clustal X 162b (httpevolutiongeneticswashingtoneduphylipsoftwareetc1html)with further manual adjustments Two phylogenetic analyseswere performed by using two primary approaches to treeconstruction tree-searching (parsimony) and algorithmic(Neighbor-Joining) Parsimony-based analyses were conductedusing Fitch parsimony (as implemented in Swofford
1999) with equal weighting of all characters and oftransitionstransversions Heuristic searches were replicated100 times with random taxon-addition sequences TreeBisection-Reconnection (TBR) branch swapping and withthe options MULPARS and STEEPEST DESCENT ineffect Relative support for clades identified by parsimonyanalysis was assessed by bootstrapping with the heuristicsearch strategy as indicated above In addition phylogeneticreconstructions of sequences were also performed in using the Kimura 2-parameter distance model (Kimura 1980)and the Neighbor-Joining method (Saitou amp Nei 1987)Indels in the noncoding regions trnT-L and trnL-F werecoded as appended characters following the logic of Kelchner(2000) and Simmons amp Ochoterena (2000) As a resultadditional analyses were performed when coding parsimony-informative indels affecting the trnL-F and trnT-L matricesCharacter evolution of flower polymorphism was investigatedusing MacClade 35 (Maddison amp Maddison 1992) Fourstates of flower polymorphism (approach herkogamy non-herkogamous monomorphism style dimorphism and hetero-styly) were traced and mapped on the resulting trnL-F andtrnT-L phylogenies of the consensus tree of all maximum-parsimony trees We explored both accelerated transformation(ACCTRAN) and delayed transformation (DELTRAN)optimizations with equal weighting of all characters andtransitions among all states equally probable
Results
Style polymorphism
Average values of sex organ position for the seven species ofsect Apodanthi and for N pallidulus are shown in Tables 3 4and 5 and comparative diagrams in Fig 2 Species in sectionApodanthi have a remarkable variation in style polymorphismdistyly in N albimarginatus style dimorphism in four species(N calcicola N cuatrecasasii N rupicola and N scaberulus)and monomorphism in N marvieri and N watieri (Fig 2Table 3) Narcissus albimarginatus shows two style-lengthmorphs with approximate reciprocal positioning of anthersand style Two patterns were found for dimorphic speciesthree species (N calcicola N scaberulus N cuatrecasasii ) presentthe L-morph with the stigma almost at an equivalent level ofthat of upper anthers whereas in N rupicola the stigma of theL-morph is positioned between the two anther levels The latterspecies displayed the highest values of stigma-anther separation(Table 4) that is the lowest reciprocity among dimorphicspecies Monomorphic species present a position of sex organssimilar to that in L-morph of dimorphic N rupicola (Fig 2)
Narcissus pallidulus is the only trimorphic species in ourstudy with mostly reciprocal positioning of stigmas andstamen whorls Although equivalent positions are not equal inall cases sequence height of sex organs corresponds to typicaltristylous species
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 241
At the population level morphs are in general unequallyrepresented L-morph is the most frequent (Tables 4 and 5) inall species and populations with the exception of one popu-lation of N cuatrecasasii and one of N rupicola which showsimilar frequency of morphs (isoplethy)
Perianth variation
Values of perianth variables are averaged over populationsand morphs (Table 6) Univariate morphometrical analyses offlower tubes tepals coronas and flower angle of all studiedspecies and populations revealed clearly absence of significantdifferences between morphs (82 between-morph comparisonsresults not shown) There were only slightly significant dif-ferences in tristylous N pallidulus at Cazorla for corona height(P = 001) and in distylous N albimarginatus at Bouhachemfor perianth width (P = 005)
Principal component analysis is shown in Fig 3 The firstsecond and third axes accounted for 616 233 and 94of the variance although only Axes 1 and 2 have eigenvalueshigher than 10 and are hence further discussed These twoaxes depict quite clearly the relationships between species andpopulations in perianth morphology At the negative extremeof Axis 1 we find together N pallidulus and N albimarginatusand at the positive extreme of this axis N marvieri Nwatieri and N rupicola form a second group Axis 2 separatesat its positive end the Iberian endemics N cuatrecasasii N cal-cicola and N scaberulus Eigenvectors of perianth variablesreflect trends of morphological variation The highest valuesfor Axis 1 are flower angle to stalk (04519) width of flowertube (minus04410) and corona height (minus04217) Largest eigen-vectors for Axis 2 correspond to corona width (minus06470)tepal length (minus04362) and flower tube length (minus04087)These results indicate that populations of two species (Npallidulus and N albimarginatus) at the left in the ordinationplot are characterised by small angles (pendulous flowers)wide flower tubes and long coronas (Fig 4) A second groupof species (N rupicola N watieri N marvieri ) display pre-dominantly erect flowers narrow and very long tubes andfunnel-shaped coronas On the top of the ordination plotthree species (N cuatrecasasii N calcicola N scaberulus) arecharacterised by patent flowers cup-shaped coronas smalltepals and moderate flower-tube length
The interrelationship between perianth variables and stylarpolymorphism is evaluated by means of multiple regressionanalyses We calculated two measurements of reciprocity inthe position of sex organs considering or not flower sizeeffects Both regression equations fitted quite similarly withscores of populations along PCA axes on perianth variablesAs a result we may conclude that there is a significant rela-tionship between perianth morphology and sex organ posi-tion across populations and species with 72ndash80 of thevariance being explained by the regression models (Table 7)Positioning of reciprocal stigma and anthers (heterostyly) cor-responds with the group of pendulous flowers Style dimorphicspecies (N calcicola N cuatrecasasii N scaberulus see Table 4)with medium values of reciprocity have patent flowers amongother features (see above) Monomorphism and less reciprocalstyle dimorphism in N rupicola (Table 4) correspond to theerect flowered group
Pollinators
As typically occurs in early blooming species it was difficultto observe pollinators in four species of Narcissus sectApodanthi (0ndash467 insect visits per hour) Twelve hours ofobservations in N cuatrecasasii ( Jaeacuten Cazorla range) revealedrarity of pollinators although medium sized solitary bees(Anthophora sp) were consistently spotted (10 visits) probingnectar and small unidentified bees (8 visits) reaching pollenof the upper anthers In the same area Anthophora bees visitedflowers of Helleborus foetidus but always in different boutsCasual observations made on populations at the oppositeextreme of the geographic range of N cuatrecasasii (CaacutedizGrazalema range) allow recognising Anthophora as the mostcommon visitor By censuses in N rupicola at the core of itsgeographic range (Avila Puerto de Mijares Madrid Puertode Navacerrada) we observed only two visits by the mothMacroglossum stellatarum one by the hoverfly Eristalis tenaxand one by an unidentified small bee Flies and small beestried to enter the flower tube but they only contacted upperanthers and L-stigma because of its narrowness and shorttongues Infrequent pollinator visits is not the result of lownumber of insects as they visited profusely another co-occurring flowers of some genera (Gagea Ranunculus SaxifragaArmeria Cytisus Erysimum) Some nocturnal censuses revealed
Table 3 Height of stamens and stigmas in monomorphic species of Narcissus sect Apodanthi Values are means plusmn 1 se in mm
Species Population Sample size Style heightUpper anther height
Lower anther height
Stigma-Anther separation
N marvieri Zeretken 51 2069 plusmn 030 2255 plusmn 032 1852 plusmn 035 186N marvieri Tazzeka 50 1001 plusmn 015 1234 plusmn 018 865 plusmn 014 233N watieri Oukaimeden 50 2092 plusmn 028 2205 plusmn 029 1815 plusmn 023 113N watieri TizirsquonrsquoTest 26 1736 plusmn 030 1927 plusmn 040 1595 plusmn 041 191
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research242
Tabl
e 4
Hei
ght
of s
tam
ens
and
stig
mas
in d
imor
phic
spe
cies
of
Nar
ciss
us s
ect
Apo
dant
hi L
lon
g-st
yled
mor
ph S
sho
rt-s
tyle
d m
orph
Val
ues
are
mea
ns plusmn
1 s
e in
mm
For
est
imat
ing
stig
ma-
anth
er s
epar
atio
n w
e co
nsid
er o
nly
the
equi
vale
nt w
horl
to th
e st
igm
a of
the
alte
rnat
e m
orph
s to
be
the
reci
proc
al (L
stig
ma
vs u
pper
sta
men
who
rl of
S fl
ower
S s
tigm
a vs
low
er s
tam
en w
horl
of L
flow
er s
ee F
ig 1
) A
ppro
pria
te t
-tes
ts a
re g
iven
for
the
diff
eren
ces
betw
een
mea
ns o
f th
ese
leng
ths
(P
lt 0
05
P
lt 0
01
P lt
00
01)
Spec
ies
Popu
latio
n
Sam
ple
size
L S
Styl
e he
ight
U
pper
ant
her
heig
ht
Low
er a
nthe
r he
ight
St
igm
a-A
nthe
r se
para
tion
Mor
ph r
atio
(1
)L
SL
SL
SL
S
N a
lbim
argi
natu
sBo
uhac
hem
69 5
824
40
plusmn 3
8
9
77 plusmn
17
015
97
plusmn 2
14
24
54
plusmn 2
899
96 plusmn
17
8
21
80
plusmn 2
550
14
019
065
0
35N
alb
imar
gina
tus
Kel
ti53
45
243
7 plusmn
032
936
plusmn 0
19
148
9 plusmn
028
216
7 plusmn
036
849
plusmn 0
23
188
1 plusmn
041
270
0
870
66
034
N c
uatr
ecas
asii
Gra
zale
ma
66 3
415
08
plusmn 0
20
8
01 plusmn
01
713
79
plusmn 0
18 n
s 14
36
plusmn 0
269
05 plusmn
01
6
12
03
plusmn 0
270
72
104
065
0
35N
cua
trec
asas
iiC
azor
la42
58
146
0 plusmn
020
690
plusmn 0
12
131
8 plusmn
018
13
95
plusmn 0
158
25 plusmn
01
4
10
74
plusmn 0
110
65
162
042
0
58N
cal
cico
laSi
coacute50
43
163
6 plusmn
034
819
plusmn 0
18
144
6 plusmn
018
152
5 plusmn
029
925
plusmn 0
21
113
2 plusmn
018
112
1
380
69
031
N c
alci
cola
Sto
Ant
onio
51 4
416
49
plusmn 0
29
8
06 plusmn
02
715
30
plusmn 0
23 n
s 15
37
plusmn 0
229
44 plusmn
02
1
11
68
plusmn 0
231
11
106
076
0
24N
sca
beru
lus
Oliv
eira
51 3
115
12
plusmn 0
24
7
72 plusmn
01
714
56
plusmn 0
15 n
s 15
05
plusmn 0
279
21 plusmn
01
6
11
50
plusmn 0
240
071
490
86
014
N s
cabe
rulu
sEr
veda
l50
14
147
7 plusmn
029
778
plusmn 0
28
148
5 plusmn
021
ns
149
9 plusmn
065
986
plusmn 0
15
11
71
plusmn 0
370
22
208
077
0
23N
rup
icol
aA
ldea
quem
ada
42 3
614
94
plusmn 0
50
10
43
plusmn 0
2216
08
plusmn 0
27
18
19
plusmn 0
26 1
109
plusmn 0
25
139
8 plusmn
021
325
0
660
54
046
N r
upic
ola
Bola
del
Mun
do53
35
131
9 plusmn
024
816
plusmn 0
14
154
5 plusmn
023
ns
163
1 plusmn
029
105
6 plusmn
018
126
6 plusmn
025
312
2
400
72
028
(1)
Sam
ple
size
use
d fo
r m
orph
rat
io is
the
tot
al n
umbe
r of
flow
ers
colle
cted
(se
e Ta
ble
1)
occurrence of some moths around the flowers of N rupicolabut they were not observed to come into contact with flowersIn N watieri insect visits were much more abundant After a3-hour census we observed long-tongued visitors (diurnalmoths six visits pierid butterflies three visits) and short-tongued hoverflies (five visits) Despite considerable time(9 h) spent on N albimarginatus we were not able to spot anypollinator
Chloroplast sequence variation
Length of trnL-F sequences in Narcissus ranges between337 bp (population 1 of Narcissus pallidulus) and 365 bp(population 2 of Narcissus marvieri ) This sequence variationis caused by seven indels of 1ndash19 bp Number of variableparsimony-informative characters is of 1711 across species ofNarcissus The highest pairwise divergence (Kimura-2-parametermodel) of species sequences (34) was found betweenaccessions of N bulbocodium and N cyclamineus Sequencedivergence neither was found between species of sectApodanthi (excluding N calcicola and N scaberulus) nor wasobserved between two of the three species of the N triandruscomplex (N pallidulus and N lusitanicus) Kimura-2-parameter distances between accessions of the same speciesdid not rendered either divergence
In Narcissus sect Apodanthi length variation of trnT-Lsequences ranges between 827 bp (N marvieri ) and 966 bp(N calcicola) Ten gaps in the trnT-L sequence matrix from1 bp to 130 bp are responsible for great sequence variationAn unusual indel of 130 bp is shared by N rupicola N marv-ieri and N watieri The number of variableparsimony-informative characters is 1911 in Narcissus sect ApodanthiThe highest pairwise divergence (Kimura-2-parameter model)between sequence species (14) was found between acces-sions of N calcicola and N cuatrecasasii whereas the lowest(021) was between N calcicola and N scaberulus
Phylogenetic relationships
Maximum parsimony (MP) using the 27 trnL-F sequencesof Narcissus and Galanthus Lapiedra and Leucojum as theoutgroup yielded 1529 trees of 78 steps (CI 095 and RI093 including uninformative characters) (results not shown)A congruent somewhat more resolved topology was retrievedin the semistrict consensus tree of 13 225 trees (101 stepsCI 085 and RI 085) when coding the seven indels asadditional characters however new branches do not displaysignificant values (bootstrap below 50) (Fig 5) From bothanalyses (with and without indel coding) a basal polytomy isdepicted in the semistrict consensus tree of which two well-supported clades are formed by the accessions of five speciesof sect Apodanthi (clade 1) and six species including the threeof the N triandrus complex (clade 2) Relatively long branchesin the Neighbor-Joining (NJ) trees using both alignments
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research238
Tabl
e 2
Acc
essi
ons
for
the
chlo
ropl
ast
sequ
ence
stu
dy o
f N
arci
ssus
inc
ludi
ng p
lant
iden
tifica
tion
follo
wed
by
popu
latio
n nu
mbe
r lo
calit
y of
wild
pop
ulat
ions
nam
e of
pla
nt c
olle
ctor
s an
d G
enBa
nk a
cces
sion
num
bers
for
trn
T-L
and
trnL
-F s
eque
nces
Taxo
nLo
calit
y
Gen
Bank
ac
cess
ion
num
ber
trnT
-L
Gen
Bank
ac
cess
ion
num
ber
trnL
-F
sect
Apo
dant
hi A
Fer
nand
esN
cal
cico
la 1
Men
doccedila
Port
ugal
Ser
ra S
to A
nton
io A
Ham
pe amp
B G
arrid
oFo
rthc
omin
gFo
rthc
omin
gN
cal
cico
la 2
Men
doccedila
Port
ugal
Ser
ra S
ico
A H
ampe
amp B
Gar
rido
ndashFo
rthc
omin
gN
cua
trec
asas
ii 1
Fern
Cas
as e
t al
Sp
ain
Jaeacute
n S
ierr
a de
Caz
orla
J A
rroy
oFo
rthc
omin
gFo
rthc
omin
gN
cua
trec
asas
ii 2
Fern
Cas
as e
t al
Sp
ain
Caacuted
iz S
ierr
a de
Gra
zale
ma
J A
rroy
ondash
Fort
hcom
ing
N m
arvi
eri 1
Jah
and
amp M
aire
Mor
occo
Zer
ekte
n J
Arr
oyo
amp A
Ham
peFo
rthc
omin
gFo
rthc
omin
gN
mar
vier
i 2 J
ahan
d amp
Mai
reM
oroc
co T
azek
a J
Arr
oyo
amp A
Ter
rab
ndashFo
rthc
omin
gN
rup
icol
a 1
Leacuteon
-Duf
our
Spai
n C
uenc
a G
rado
de
Pico
P V
arga
s et
al
ndashFo
rthc
omin
gN
rup
icol
a 2
Leacuteon
-Duf
our
Spai
n M
adrid
Sie
rra
de la
Cab
rera
P V
arga
s et
al
Fort
hcom
ing
Fort
hcom
ing
N r
upic
ola
3 Leacute
on-D
ufou
rSp
ain
Cue
nca
Gra
do d
e Pi
co P
Var
gas
et a
lndash
Fort
hcom
ing
N s
cabe
rulu
s 1
Hen
riqu
esPo
rtug
al E
rved
al A
Ham
pe amp
B G
arrid
oFo
rthc
omin
gFo
rthc
omin
gN
sca
beru
lus
2 H
enri
ques
Port
ugal
Erv
edal
A H
ampe
amp B
Gar
rido
ndashFo
rthc
omin
gN
sca
beru
lus
3 H
enri
ques
Port
ugal
Oliv
eira
do
Con
de A
Ham
pe amp
B G
arrid
ondash
Fort
hcom
ing
N w
atie
ri M
aire
Mor
occo
Ouk
aim
eden
J A
rroy
o amp
A H
ampe
Fort
hcom
ing
Fort
hcom
ing
N a
lbim
argi
natu
s 1
U M
uumllle
r-D
oblie
s amp
D M
uumllle
r-D
oblie
sM
oroc
co B
ouha
chem
J A
rroy
oFo
rthc
omin
gFo
rthc
omin
gN
alb
imar
gina
tus
2 U
Muumll
ler-
Dob
lies
amp D
Muumll
ler-
Dob
lies
Mor
occo
Kel
ti F
Oje
dandash
Fort
hcom
ing
sect
Cyc
lam
inei
DC
N
cyc
lam
ineu
s D
C
Spai
n P
onte
vedr
a S
Cas
trov
iejo
616
2ndash
Fort
hcom
ing
N l
usit
anic
us D
orda
amp F
ern
Cas
asPo
rtug
al P
ena
Cov
a de
Oliv
eira
A B
arra
AB2
585
ndashFo
rthc
omin
gN
pal
lidul
us 1
Gra
ells
Spai
n M
adrid
Ald
ea d
el F
resn
o P
Var
gas
amp O
Goacutem
ezndash
Fort
hcom
ing
N p
allid
ulus
2 G
rael
lsSp
ain
Jaeacute
n A
ldea
quem
ada
R P
eacuterez
ndashFo
rthc
omin
gN
tri
andr
us L
Sp
ain
Zam
ora
Mah
ide
B H
ernaacute
ndez
ndashFo
rthc
omin
gse
ct D
ubii
Fern
Cas
asN
tor
tifo
lius
Fern
Cas
asSp
ain
Alm
eriacutea
Riacuteo
de
Agu
as A
Bar
randash
Fort
hcom
ing
sect
Jon
quill
ae D
C
N j
onqu
illa
LSp
ain
Sev
illa
Sie
rra
Nor
te R
Peacuter
ez amp
R P
eacuterez
de
Guz
maacuten
ndashFo
rthc
omin
gse
ct S
erot
ini P
arl
N s
erot
inus
Loe
fl e
x L
Spai
n S
evill
a D
os H
erm
anas
C A
ndreacute
s amp
Z D
iacuteaz
ndashFo
rthc
omin
gse
ct T
azet
tae
DC
N
taz
etta
L
Gen
Bank
ndashA
J232
562
sect
Bul
boco
dii D
C
N b
ulbo
codi
um L
Sp
ain
Hue
lva
Car
taya
J C
astil
loFo
rthc
omin
gFo
rthc
omin
gse
ct N
arci
ssus
L
N p
oeti
cus
LSp
ain
Leacuter
ida
Val
le d
e A
raacuten
X P
icoacute
ndashFo
rthc
omin
gse
ct P
seud
onar
ciss
i DC
N
pse
udon
arci
ssus
L
Spai
n Aacute
vila
Sie
rra
de G
redo
s R
Peacuter
ez amp
J A
rroy
ondash
Fort
hcom
ing
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 239
means that in one dimorphic or distylous Narcissus speciesthe long style (L-morph) has its reciprocal male organ in upperstamens of S-morph Accordingly S-style is the reciprocal tolower stamens of L-morph (Fig 1) Flowers of the tristylousNarcissus pallidulus does not pose special problems as thiscondition usually involves two stamen whorls and the com-parison of reciprocal heights is straightforward as originallydescribed by Darwin (1877) in Lythrum salicaria We aver-aged the absolute values of this difference across the morphsof each population in order to compare dimorphic and tri-morphic species A perfectly reciprocal population wouldtherefore rate a value of zero In order to control for effects offlower size in the extent of reciprocity we also divided the rec-iprocity of a population by its mean flower diameter
Statistical analyses All morphometrical variables were checkedfor normality and homoscedasticity and transformed whennecessary Only flower tube length across populations andspecies needed to be log transformed Data were averaged formorphs and populations and compared accordingly Studentt-tests and one-way were used for univariate compar-isons of flower features between morphs for dimorphic andtrimorphic species respectively In the rare cases wherevariances were not homogeneous (Levenersquos test) we applieda t-test with separate variance estimates After fordifference of means posthoc Tukey HSD test for unequalsample sizes was applied for multiple comparison betweenpairs When performing multiple tests of one hypothesisDunn-Sidak correction of sequential Bonferroni adjustementof α levels was applied (Sokal amp Rohlf 1995) Reciprocity(between-morph stigma-anther separation) was not comparedbetween morphs given that it was obtained as a differencebetween morph means hence error estimates of the differencewere not available Multivariate analyses were performed tocheck the overall pattern of morphological variation of flower
perianth An ordination by Principal Component Analysis(PCA) was performed to draw the morphological relationshipsbetween populations and species In multivariate analysispopulation means were used rather than individual data inorder to avoid pseudoreplication An analysis of the intra-population variation (Faivre amp McDade 2001) is intendedto be published in a subsequent paper
To analyse the relationships between perianth morphologyand sex organ polymorphism we applied multiple regressionanalyses Independent variables were the scores of the populationmeans for each of three first PCA axes We used these perianthpseudovariables rather than raw univariate values because axesaccounted for a high variance of perianth morphology andbecause axes are by definition orthogonal overcoming the seriousproblem of multicollinearity in multiple regression ( James ampMcCulloch 1990 Philippi 1993) Two regression modelswere adjusted with respective dependent variables related toreciprocity of stigma-anther separation of each population(i) the mean between morphs of absolute values for differencebetween style height and equivalent stamens whorl height and(ii) the values of (i) divided by flower diameter The latter var-iable was considered to control for potential allometric effectof flower size on the magnitude of departures from stigma-anther reciprocity (see above) Values for monomorphic popu-lations were scored as the highest obtained value of departurefrom exact reciprocity Ordination analysis was performedwith PCORD package (McCune amp Mefford 1999) whereasunivariate statistics and multiple regression models were com-puted with STATISTICA software (Statsoft 1997)
Pollinators
Data were taken from populations of four Apodanthi spe-cies some in common with those of the study of flowermorphometry (Table 1) We performed systematic censuses
Fig 1 Long-styled (L) and short-styled (S) flowers of a typical style-dimorphic Narcissus species (N cuatrecasasii) Numbers correspond to the measurements taken in each flower sampled (1) flower width (2) tepal length (3) corona width (4) corona height (5) tube length (6) tube width (7) style length (8) upper stamen height (9) lower stamen height (10) angle of flower to stalk Measurements were accordingly modified for tristylous distylous and monomorphic species
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research240
of insects visiting the flowers and recorded positive dataproviding that they enter into the tube and do contact any sexorgan Species subject to study were N rupicola (time effortof 12 h including some nocturnal periods) N cuatrecasasii(12 h) N albimarginatus (9 h) and N watieri (3 h) Observationtiming and area depend on density of flowers in everypopulation In sparse patches we made several censuses of15 min observing all insects visiting flowers in an area ofapprox 100 m2 We also made static observations in densepatches on all flowers in small areas (approx 10 m2)
Phylogenetic studies
PCR amplification and sequencing Total genomic DNA wasextracted from silica-dried material collected in the field andthe living collection of the Royal Botanic Garden of MadridSpain DNA extractions were performed using the DNeasyPlant Mini Kit (QIAGEN Inc Chatsworth CA USA) ForPolymerase Chain Reaction (PCR) amplifications we used aPerkin-Elmer (Foster City CA USA) PCR System 9700 thermalcycler primers trna and trnb for trnT(UGU)-trnL(UAA)trne and trn f for trnL(UAA)-trnF(GAA) and PCR conditionsfollowing Taberlet et al (1991) 51degC of annealing temperaturefor the trnT-L and 50degC for the trnL-F plus 2-min elongationtime Amplified products were cleaned using spin filter columns(PCR Clean-up kit MoBio Laboratories Solana Beach CAUSA) following the protocols provided by the manufacturerCleaned products were then directly sequenced using dyeterminators (Big Dye Terminator vs 20 Applied BiosystemsLittle Chalfont UK) and run into polyacrylamideelectrophoresis gels (7) using an Applied Biosystems Prismmodel 3700 automated sequencer Primers trna trnb trneand trn f were also used for cycle sequencing of the trnT-trnLand trnL-trnF spacers under the following conditions 95degCfor 2 min followed by 25 cycles of 95degC for 10 s 50degC for 5s and 60degC for 4 min Sequence data were placed in a contigfile and edited using the program Seqed (Applied BiosystemsFoster City CA USA) The limits of the trnT-trnL and trnL-trnF regions were determined by comparison with otherAmaryllidaceae sequences available in the GenBank
Sequence analyses Three different sequence matrices wereused to perform the phylogenetic analyses trnL-F matrixof Narcissus trnT-L matrix of Narcissus sect Apodanthiand a combined matrix of trnL-F and trnT-L for only thespecies of Narcissus sect Apodanthi as ingroup Alignmentsof the 30 trnL-F and the nine trnT-L sequences wereobtained using the program Clustal X 162b (httpevolutiongeneticswashingtoneduphylipsoftwareetc1html)with further manual adjustments Two phylogenetic analyseswere performed by using two primary approaches to treeconstruction tree-searching (parsimony) and algorithmic(Neighbor-Joining) Parsimony-based analyses were conductedusing Fitch parsimony (as implemented in Swofford
1999) with equal weighting of all characters and oftransitionstransversions Heuristic searches were replicated100 times with random taxon-addition sequences TreeBisection-Reconnection (TBR) branch swapping and withthe options MULPARS and STEEPEST DESCENT ineffect Relative support for clades identified by parsimonyanalysis was assessed by bootstrapping with the heuristicsearch strategy as indicated above In addition phylogeneticreconstructions of sequences were also performed in using the Kimura 2-parameter distance model (Kimura 1980)and the Neighbor-Joining method (Saitou amp Nei 1987)Indels in the noncoding regions trnT-L and trnL-F werecoded as appended characters following the logic of Kelchner(2000) and Simmons amp Ochoterena (2000) As a resultadditional analyses were performed when coding parsimony-informative indels affecting the trnL-F and trnT-L matricesCharacter evolution of flower polymorphism was investigatedusing MacClade 35 (Maddison amp Maddison 1992) Fourstates of flower polymorphism (approach herkogamy non-herkogamous monomorphism style dimorphism and hetero-styly) were traced and mapped on the resulting trnL-F andtrnT-L phylogenies of the consensus tree of all maximum-parsimony trees We explored both accelerated transformation(ACCTRAN) and delayed transformation (DELTRAN)optimizations with equal weighting of all characters andtransitions among all states equally probable
Results
Style polymorphism
Average values of sex organ position for the seven species ofsect Apodanthi and for N pallidulus are shown in Tables 3 4and 5 and comparative diagrams in Fig 2 Species in sectionApodanthi have a remarkable variation in style polymorphismdistyly in N albimarginatus style dimorphism in four species(N calcicola N cuatrecasasii N rupicola and N scaberulus)and monomorphism in N marvieri and N watieri (Fig 2Table 3) Narcissus albimarginatus shows two style-lengthmorphs with approximate reciprocal positioning of anthersand style Two patterns were found for dimorphic speciesthree species (N calcicola N scaberulus N cuatrecasasii ) presentthe L-morph with the stigma almost at an equivalent level ofthat of upper anthers whereas in N rupicola the stigma of theL-morph is positioned between the two anther levels The latterspecies displayed the highest values of stigma-anther separation(Table 4) that is the lowest reciprocity among dimorphicspecies Monomorphic species present a position of sex organssimilar to that in L-morph of dimorphic N rupicola (Fig 2)
Narcissus pallidulus is the only trimorphic species in ourstudy with mostly reciprocal positioning of stigmas andstamen whorls Although equivalent positions are not equal inall cases sequence height of sex organs corresponds to typicaltristylous species
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 241
At the population level morphs are in general unequallyrepresented L-morph is the most frequent (Tables 4 and 5) inall species and populations with the exception of one popu-lation of N cuatrecasasii and one of N rupicola which showsimilar frequency of morphs (isoplethy)
Perianth variation
Values of perianth variables are averaged over populationsand morphs (Table 6) Univariate morphometrical analyses offlower tubes tepals coronas and flower angle of all studiedspecies and populations revealed clearly absence of significantdifferences between morphs (82 between-morph comparisonsresults not shown) There were only slightly significant dif-ferences in tristylous N pallidulus at Cazorla for corona height(P = 001) and in distylous N albimarginatus at Bouhachemfor perianth width (P = 005)
Principal component analysis is shown in Fig 3 The firstsecond and third axes accounted for 616 233 and 94of the variance although only Axes 1 and 2 have eigenvalueshigher than 10 and are hence further discussed These twoaxes depict quite clearly the relationships between species andpopulations in perianth morphology At the negative extremeof Axis 1 we find together N pallidulus and N albimarginatusand at the positive extreme of this axis N marvieri Nwatieri and N rupicola form a second group Axis 2 separatesat its positive end the Iberian endemics N cuatrecasasii N cal-cicola and N scaberulus Eigenvectors of perianth variablesreflect trends of morphological variation The highest valuesfor Axis 1 are flower angle to stalk (04519) width of flowertube (minus04410) and corona height (minus04217) Largest eigen-vectors for Axis 2 correspond to corona width (minus06470)tepal length (minus04362) and flower tube length (minus04087)These results indicate that populations of two species (Npallidulus and N albimarginatus) at the left in the ordinationplot are characterised by small angles (pendulous flowers)wide flower tubes and long coronas (Fig 4) A second groupof species (N rupicola N watieri N marvieri ) display pre-dominantly erect flowers narrow and very long tubes andfunnel-shaped coronas On the top of the ordination plotthree species (N cuatrecasasii N calcicola N scaberulus) arecharacterised by patent flowers cup-shaped coronas smalltepals and moderate flower-tube length
The interrelationship between perianth variables and stylarpolymorphism is evaluated by means of multiple regressionanalyses We calculated two measurements of reciprocity inthe position of sex organs considering or not flower sizeeffects Both regression equations fitted quite similarly withscores of populations along PCA axes on perianth variablesAs a result we may conclude that there is a significant rela-tionship between perianth morphology and sex organ posi-tion across populations and species with 72ndash80 of thevariance being explained by the regression models (Table 7)Positioning of reciprocal stigma and anthers (heterostyly) cor-responds with the group of pendulous flowers Style dimorphicspecies (N calcicola N cuatrecasasii N scaberulus see Table 4)with medium values of reciprocity have patent flowers amongother features (see above) Monomorphism and less reciprocalstyle dimorphism in N rupicola (Table 4) correspond to theerect flowered group
Pollinators
As typically occurs in early blooming species it was difficultto observe pollinators in four species of Narcissus sectApodanthi (0ndash467 insect visits per hour) Twelve hours ofobservations in N cuatrecasasii ( Jaeacuten Cazorla range) revealedrarity of pollinators although medium sized solitary bees(Anthophora sp) were consistently spotted (10 visits) probingnectar and small unidentified bees (8 visits) reaching pollenof the upper anthers In the same area Anthophora bees visitedflowers of Helleborus foetidus but always in different boutsCasual observations made on populations at the oppositeextreme of the geographic range of N cuatrecasasii (CaacutedizGrazalema range) allow recognising Anthophora as the mostcommon visitor By censuses in N rupicola at the core of itsgeographic range (Avila Puerto de Mijares Madrid Puertode Navacerrada) we observed only two visits by the mothMacroglossum stellatarum one by the hoverfly Eristalis tenaxand one by an unidentified small bee Flies and small beestried to enter the flower tube but they only contacted upperanthers and L-stigma because of its narrowness and shorttongues Infrequent pollinator visits is not the result of lownumber of insects as they visited profusely another co-occurring flowers of some genera (Gagea Ranunculus SaxifragaArmeria Cytisus Erysimum) Some nocturnal censuses revealed
Table 3 Height of stamens and stigmas in monomorphic species of Narcissus sect Apodanthi Values are means plusmn 1 se in mm
Species Population Sample size Style heightUpper anther height
Lower anther height
Stigma-Anther separation
N marvieri Zeretken 51 2069 plusmn 030 2255 plusmn 032 1852 plusmn 035 186N marvieri Tazzeka 50 1001 plusmn 015 1234 plusmn 018 865 plusmn 014 233N watieri Oukaimeden 50 2092 plusmn 028 2205 plusmn 029 1815 plusmn 023 113N watieri TizirsquonrsquoTest 26 1736 plusmn 030 1927 plusmn 040 1595 plusmn 041 191
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research242
Tabl
e 4
Hei
ght
of s
tam
ens
and
stig
mas
in d
imor
phic
spe
cies
of
Nar
ciss
us s
ect
Apo
dant
hi L
lon
g-st
yled
mor
ph S
sho
rt-s
tyle
d m
orph
Val
ues
are
mea
ns plusmn
1 s
e in
mm
For
est
imat
ing
stig
ma-
anth
er s
epar
atio
n w
e co
nsid
er o
nly
the
equi
vale
nt w
horl
to th
e st
igm
a of
the
alte
rnat
e m
orph
s to
be
the
reci
proc
al (L
stig
ma
vs u
pper
sta
men
who
rl of
S fl
ower
S s
tigm
a vs
low
er s
tam
en w
horl
of L
flow
er s
ee F
ig 1
) A
ppro
pria
te t
-tes
ts a
re g
iven
for
the
diff
eren
ces
betw
een
mea
ns o
f th
ese
leng
ths
(P
lt 0
05
P
lt 0
01
P lt
00
01)
Spec
ies
Popu
latio
n
Sam
ple
size
L S
Styl
e he
ight
U
pper
ant
her
heig
ht
Low
er a
nthe
r he
ight
St
igm
a-A
nthe
r se
para
tion
Mor
ph r
atio
(1
)L
SL
SL
SL
S
N a
lbim
argi
natu
sBo
uhac
hem
69 5
824
40
plusmn 3
8
9
77 plusmn
17
015
97
plusmn 2
14
24
54
plusmn 2
899
96 plusmn
17
8
21
80
plusmn 2
550
14
019
065
0
35N
alb
imar
gina
tus
Kel
ti53
45
243
7 plusmn
032
936
plusmn 0
19
148
9 plusmn
028
216
7 plusmn
036
849
plusmn 0
23
188
1 plusmn
041
270
0
870
66
034
N c
uatr
ecas
asii
Gra
zale
ma
66 3
415
08
plusmn 0
20
8
01 plusmn
01
713
79
plusmn 0
18 n
s 14
36
plusmn 0
269
05 plusmn
01
6
12
03
plusmn 0
270
72
104
065
0
35N
cua
trec
asas
iiC
azor
la42
58
146
0 plusmn
020
690
plusmn 0
12
131
8 plusmn
018
13
95
plusmn 0
158
25 plusmn
01
4
10
74
plusmn 0
110
65
162
042
0
58N
cal
cico
laSi
coacute50
43
163
6 plusmn
034
819
plusmn 0
18
144
6 plusmn
018
152
5 plusmn
029
925
plusmn 0
21
113
2 plusmn
018
112
1
380
69
031
N c
alci
cola
Sto
Ant
onio
51 4
416
49
plusmn 0
29
8
06 plusmn
02
715
30
plusmn 0
23 n
s 15
37
plusmn 0
229
44 plusmn
02
1
11
68
plusmn 0
231
11
106
076
0
24N
sca
beru
lus
Oliv
eira
51 3
115
12
plusmn 0
24
7
72 plusmn
01
714
56
plusmn 0
15 n
s 15
05
plusmn 0
279
21 plusmn
01
6
11
50
plusmn 0
240
071
490
86
014
N s
cabe
rulu
sEr
veda
l50
14
147
7 plusmn
029
778
plusmn 0
28
148
5 plusmn
021
ns
149
9 plusmn
065
986
plusmn 0
15
11
71
plusmn 0
370
22
208
077
0
23N
rup
icol
aA
ldea
quem
ada
42 3
614
94
plusmn 0
50
10
43
plusmn 0
2216
08
plusmn 0
27
18
19
plusmn 0
26 1
109
plusmn 0
25
139
8 plusmn
021
325
0
660
54
046
N r
upic
ola
Bola
del
Mun
do53
35
131
9 plusmn
024
816
plusmn 0
14
154
5 plusmn
023
ns
163
1 plusmn
029
105
6 plusmn
018
126
6 plusmn
025
312
2
400
72
028
(1)
Sam
ple
size
use
d fo
r m
orph
rat
io is
the
tot
al n
umbe
r of
flow
ers
colle
cted
(se
e Ta
ble
1)
occurrence of some moths around the flowers of N rupicolabut they were not observed to come into contact with flowersIn N watieri insect visits were much more abundant After a3-hour census we observed long-tongued visitors (diurnalmoths six visits pierid butterflies three visits) and short-tongued hoverflies (five visits) Despite considerable time(9 h) spent on N albimarginatus we were not able to spot anypollinator
Chloroplast sequence variation
Length of trnL-F sequences in Narcissus ranges between337 bp (population 1 of Narcissus pallidulus) and 365 bp(population 2 of Narcissus marvieri ) This sequence variationis caused by seven indels of 1ndash19 bp Number of variableparsimony-informative characters is of 1711 across species ofNarcissus The highest pairwise divergence (Kimura-2-parametermodel) of species sequences (34) was found betweenaccessions of N bulbocodium and N cyclamineus Sequencedivergence neither was found between species of sectApodanthi (excluding N calcicola and N scaberulus) nor wasobserved between two of the three species of the N triandruscomplex (N pallidulus and N lusitanicus) Kimura-2-parameter distances between accessions of the same speciesdid not rendered either divergence
In Narcissus sect Apodanthi length variation of trnT-Lsequences ranges between 827 bp (N marvieri ) and 966 bp(N calcicola) Ten gaps in the trnT-L sequence matrix from1 bp to 130 bp are responsible for great sequence variationAn unusual indel of 130 bp is shared by N rupicola N marv-ieri and N watieri The number of variableparsimony-informative characters is 1911 in Narcissus sect ApodanthiThe highest pairwise divergence (Kimura-2-parameter model)between sequence species (14) was found between acces-sions of N calcicola and N cuatrecasasii whereas the lowest(021) was between N calcicola and N scaberulus
Phylogenetic relationships
Maximum parsimony (MP) using the 27 trnL-F sequencesof Narcissus and Galanthus Lapiedra and Leucojum as theoutgroup yielded 1529 trees of 78 steps (CI 095 and RI093 including uninformative characters) (results not shown)A congruent somewhat more resolved topology was retrievedin the semistrict consensus tree of 13 225 trees (101 stepsCI 085 and RI 085) when coding the seven indels asadditional characters however new branches do not displaysignificant values (bootstrap below 50) (Fig 5) From bothanalyses (with and without indel coding) a basal polytomy isdepicted in the semistrict consensus tree of which two well-supported clades are formed by the accessions of five speciesof sect Apodanthi (clade 1) and six species including the threeof the N triandrus complex (clade 2) Relatively long branchesin the Neighbor-Joining (NJ) trees using both alignments
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 239
means that in one dimorphic or distylous Narcissus speciesthe long style (L-morph) has its reciprocal male organ in upperstamens of S-morph Accordingly S-style is the reciprocal tolower stamens of L-morph (Fig 1) Flowers of the tristylousNarcissus pallidulus does not pose special problems as thiscondition usually involves two stamen whorls and the com-parison of reciprocal heights is straightforward as originallydescribed by Darwin (1877) in Lythrum salicaria We aver-aged the absolute values of this difference across the morphsof each population in order to compare dimorphic and tri-morphic species A perfectly reciprocal population wouldtherefore rate a value of zero In order to control for effects offlower size in the extent of reciprocity we also divided the rec-iprocity of a population by its mean flower diameter
Statistical analyses All morphometrical variables were checkedfor normality and homoscedasticity and transformed whennecessary Only flower tube length across populations andspecies needed to be log transformed Data were averaged formorphs and populations and compared accordingly Studentt-tests and one-way were used for univariate compar-isons of flower features between morphs for dimorphic andtrimorphic species respectively In the rare cases wherevariances were not homogeneous (Levenersquos test) we applieda t-test with separate variance estimates After fordifference of means posthoc Tukey HSD test for unequalsample sizes was applied for multiple comparison betweenpairs When performing multiple tests of one hypothesisDunn-Sidak correction of sequential Bonferroni adjustementof α levels was applied (Sokal amp Rohlf 1995) Reciprocity(between-morph stigma-anther separation) was not comparedbetween morphs given that it was obtained as a differencebetween morph means hence error estimates of the differencewere not available Multivariate analyses were performed tocheck the overall pattern of morphological variation of flower
perianth An ordination by Principal Component Analysis(PCA) was performed to draw the morphological relationshipsbetween populations and species In multivariate analysispopulation means were used rather than individual data inorder to avoid pseudoreplication An analysis of the intra-population variation (Faivre amp McDade 2001) is intendedto be published in a subsequent paper
To analyse the relationships between perianth morphologyand sex organ polymorphism we applied multiple regressionanalyses Independent variables were the scores of the populationmeans for each of three first PCA axes We used these perianthpseudovariables rather than raw univariate values because axesaccounted for a high variance of perianth morphology andbecause axes are by definition orthogonal overcoming the seriousproblem of multicollinearity in multiple regression ( James ampMcCulloch 1990 Philippi 1993) Two regression modelswere adjusted with respective dependent variables related toreciprocity of stigma-anther separation of each population(i) the mean between morphs of absolute values for differencebetween style height and equivalent stamens whorl height and(ii) the values of (i) divided by flower diameter The latter var-iable was considered to control for potential allometric effectof flower size on the magnitude of departures from stigma-anther reciprocity (see above) Values for monomorphic popu-lations were scored as the highest obtained value of departurefrom exact reciprocity Ordination analysis was performedwith PCORD package (McCune amp Mefford 1999) whereasunivariate statistics and multiple regression models were com-puted with STATISTICA software (Statsoft 1997)
Pollinators
Data were taken from populations of four Apodanthi spe-cies some in common with those of the study of flowermorphometry (Table 1) We performed systematic censuses
Fig 1 Long-styled (L) and short-styled (S) flowers of a typical style-dimorphic Narcissus species (N cuatrecasasii) Numbers correspond to the measurements taken in each flower sampled (1) flower width (2) tepal length (3) corona width (4) corona height (5) tube length (6) tube width (7) style length (8) upper stamen height (9) lower stamen height (10) angle of flower to stalk Measurements were accordingly modified for tristylous distylous and monomorphic species
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research240
of insects visiting the flowers and recorded positive dataproviding that they enter into the tube and do contact any sexorgan Species subject to study were N rupicola (time effortof 12 h including some nocturnal periods) N cuatrecasasii(12 h) N albimarginatus (9 h) and N watieri (3 h) Observationtiming and area depend on density of flowers in everypopulation In sparse patches we made several censuses of15 min observing all insects visiting flowers in an area ofapprox 100 m2 We also made static observations in densepatches on all flowers in small areas (approx 10 m2)
Phylogenetic studies
PCR amplification and sequencing Total genomic DNA wasextracted from silica-dried material collected in the field andthe living collection of the Royal Botanic Garden of MadridSpain DNA extractions were performed using the DNeasyPlant Mini Kit (QIAGEN Inc Chatsworth CA USA) ForPolymerase Chain Reaction (PCR) amplifications we used aPerkin-Elmer (Foster City CA USA) PCR System 9700 thermalcycler primers trna and trnb for trnT(UGU)-trnL(UAA)trne and trn f for trnL(UAA)-trnF(GAA) and PCR conditionsfollowing Taberlet et al (1991) 51degC of annealing temperaturefor the trnT-L and 50degC for the trnL-F plus 2-min elongationtime Amplified products were cleaned using spin filter columns(PCR Clean-up kit MoBio Laboratories Solana Beach CAUSA) following the protocols provided by the manufacturerCleaned products were then directly sequenced using dyeterminators (Big Dye Terminator vs 20 Applied BiosystemsLittle Chalfont UK) and run into polyacrylamideelectrophoresis gels (7) using an Applied Biosystems Prismmodel 3700 automated sequencer Primers trna trnb trneand trn f were also used for cycle sequencing of the trnT-trnLand trnL-trnF spacers under the following conditions 95degCfor 2 min followed by 25 cycles of 95degC for 10 s 50degC for 5s and 60degC for 4 min Sequence data were placed in a contigfile and edited using the program Seqed (Applied BiosystemsFoster City CA USA) The limits of the trnT-trnL and trnL-trnF regions were determined by comparison with otherAmaryllidaceae sequences available in the GenBank
Sequence analyses Three different sequence matrices wereused to perform the phylogenetic analyses trnL-F matrixof Narcissus trnT-L matrix of Narcissus sect Apodanthiand a combined matrix of trnL-F and trnT-L for only thespecies of Narcissus sect Apodanthi as ingroup Alignmentsof the 30 trnL-F and the nine trnT-L sequences wereobtained using the program Clustal X 162b (httpevolutiongeneticswashingtoneduphylipsoftwareetc1html)with further manual adjustments Two phylogenetic analyseswere performed by using two primary approaches to treeconstruction tree-searching (parsimony) and algorithmic(Neighbor-Joining) Parsimony-based analyses were conductedusing Fitch parsimony (as implemented in Swofford
1999) with equal weighting of all characters and oftransitionstransversions Heuristic searches were replicated100 times with random taxon-addition sequences TreeBisection-Reconnection (TBR) branch swapping and withthe options MULPARS and STEEPEST DESCENT ineffect Relative support for clades identified by parsimonyanalysis was assessed by bootstrapping with the heuristicsearch strategy as indicated above In addition phylogeneticreconstructions of sequences were also performed in using the Kimura 2-parameter distance model (Kimura 1980)and the Neighbor-Joining method (Saitou amp Nei 1987)Indels in the noncoding regions trnT-L and trnL-F werecoded as appended characters following the logic of Kelchner(2000) and Simmons amp Ochoterena (2000) As a resultadditional analyses were performed when coding parsimony-informative indels affecting the trnL-F and trnT-L matricesCharacter evolution of flower polymorphism was investigatedusing MacClade 35 (Maddison amp Maddison 1992) Fourstates of flower polymorphism (approach herkogamy non-herkogamous monomorphism style dimorphism and hetero-styly) were traced and mapped on the resulting trnL-F andtrnT-L phylogenies of the consensus tree of all maximum-parsimony trees We explored both accelerated transformation(ACCTRAN) and delayed transformation (DELTRAN)optimizations with equal weighting of all characters andtransitions among all states equally probable
Results
Style polymorphism
Average values of sex organ position for the seven species ofsect Apodanthi and for N pallidulus are shown in Tables 3 4and 5 and comparative diagrams in Fig 2 Species in sectionApodanthi have a remarkable variation in style polymorphismdistyly in N albimarginatus style dimorphism in four species(N calcicola N cuatrecasasii N rupicola and N scaberulus)and monomorphism in N marvieri and N watieri (Fig 2Table 3) Narcissus albimarginatus shows two style-lengthmorphs with approximate reciprocal positioning of anthersand style Two patterns were found for dimorphic speciesthree species (N calcicola N scaberulus N cuatrecasasii ) presentthe L-morph with the stigma almost at an equivalent level ofthat of upper anthers whereas in N rupicola the stigma of theL-morph is positioned between the two anther levels The latterspecies displayed the highest values of stigma-anther separation(Table 4) that is the lowest reciprocity among dimorphicspecies Monomorphic species present a position of sex organssimilar to that in L-morph of dimorphic N rupicola (Fig 2)
Narcissus pallidulus is the only trimorphic species in ourstudy with mostly reciprocal positioning of stigmas andstamen whorls Although equivalent positions are not equal inall cases sequence height of sex organs corresponds to typicaltristylous species
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 241
At the population level morphs are in general unequallyrepresented L-morph is the most frequent (Tables 4 and 5) inall species and populations with the exception of one popu-lation of N cuatrecasasii and one of N rupicola which showsimilar frequency of morphs (isoplethy)
Perianth variation
Values of perianth variables are averaged over populationsand morphs (Table 6) Univariate morphometrical analyses offlower tubes tepals coronas and flower angle of all studiedspecies and populations revealed clearly absence of significantdifferences between morphs (82 between-morph comparisonsresults not shown) There were only slightly significant dif-ferences in tristylous N pallidulus at Cazorla for corona height(P = 001) and in distylous N albimarginatus at Bouhachemfor perianth width (P = 005)
Principal component analysis is shown in Fig 3 The firstsecond and third axes accounted for 616 233 and 94of the variance although only Axes 1 and 2 have eigenvalueshigher than 10 and are hence further discussed These twoaxes depict quite clearly the relationships between species andpopulations in perianth morphology At the negative extremeof Axis 1 we find together N pallidulus and N albimarginatusand at the positive extreme of this axis N marvieri Nwatieri and N rupicola form a second group Axis 2 separatesat its positive end the Iberian endemics N cuatrecasasii N cal-cicola and N scaberulus Eigenvectors of perianth variablesreflect trends of morphological variation The highest valuesfor Axis 1 are flower angle to stalk (04519) width of flowertube (minus04410) and corona height (minus04217) Largest eigen-vectors for Axis 2 correspond to corona width (minus06470)tepal length (minus04362) and flower tube length (minus04087)These results indicate that populations of two species (Npallidulus and N albimarginatus) at the left in the ordinationplot are characterised by small angles (pendulous flowers)wide flower tubes and long coronas (Fig 4) A second groupof species (N rupicola N watieri N marvieri ) display pre-dominantly erect flowers narrow and very long tubes andfunnel-shaped coronas On the top of the ordination plotthree species (N cuatrecasasii N calcicola N scaberulus) arecharacterised by patent flowers cup-shaped coronas smalltepals and moderate flower-tube length
The interrelationship between perianth variables and stylarpolymorphism is evaluated by means of multiple regressionanalyses We calculated two measurements of reciprocity inthe position of sex organs considering or not flower sizeeffects Both regression equations fitted quite similarly withscores of populations along PCA axes on perianth variablesAs a result we may conclude that there is a significant rela-tionship between perianth morphology and sex organ posi-tion across populations and species with 72ndash80 of thevariance being explained by the regression models (Table 7)Positioning of reciprocal stigma and anthers (heterostyly) cor-responds with the group of pendulous flowers Style dimorphicspecies (N calcicola N cuatrecasasii N scaberulus see Table 4)with medium values of reciprocity have patent flowers amongother features (see above) Monomorphism and less reciprocalstyle dimorphism in N rupicola (Table 4) correspond to theerect flowered group
Pollinators
As typically occurs in early blooming species it was difficultto observe pollinators in four species of Narcissus sectApodanthi (0ndash467 insect visits per hour) Twelve hours ofobservations in N cuatrecasasii ( Jaeacuten Cazorla range) revealedrarity of pollinators although medium sized solitary bees(Anthophora sp) were consistently spotted (10 visits) probingnectar and small unidentified bees (8 visits) reaching pollenof the upper anthers In the same area Anthophora bees visitedflowers of Helleborus foetidus but always in different boutsCasual observations made on populations at the oppositeextreme of the geographic range of N cuatrecasasii (CaacutedizGrazalema range) allow recognising Anthophora as the mostcommon visitor By censuses in N rupicola at the core of itsgeographic range (Avila Puerto de Mijares Madrid Puertode Navacerrada) we observed only two visits by the mothMacroglossum stellatarum one by the hoverfly Eristalis tenaxand one by an unidentified small bee Flies and small beestried to enter the flower tube but they only contacted upperanthers and L-stigma because of its narrowness and shorttongues Infrequent pollinator visits is not the result of lownumber of insects as they visited profusely another co-occurring flowers of some genera (Gagea Ranunculus SaxifragaArmeria Cytisus Erysimum) Some nocturnal censuses revealed
Table 3 Height of stamens and stigmas in monomorphic species of Narcissus sect Apodanthi Values are means plusmn 1 se in mm
Species Population Sample size Style heightUpper anther height
Lower anther height
Stigma-Anther separation
N marvieri Zeretken 51 2069 plusmn 030 2255 plusmn 032 1852 plusmn 035 186N marvieri Tazzeka 50 1001 plusmn 015 1234 plusmn 018 865 plusmn 014 233N watieri Oukaimeden 50 2092 plusmn 028 2205 plusmn 029 1815 plusmn 023 113N watieri TizirsquonrsquoTest 26 1736 plusmn 030 1927 plusmn 040 1595 plusmn 041 191
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research242
Tabl
e 4
Hei
ght
of s
tam
ens
and
stig
mas
in d
imor
phic
spe
cies
of
Nar
ciss
us s
ect
Apo
dant
hi L
lon
g-st
yled
mor
ph S
sho
rt-s
tyle
d m
orph
Val
ues
are
mea
ns plusmn
1 s
e in
mm
For
est
imat
ing
stig
ma-
anth
er s
epar
atio
n w
e co
nsid
er o
nly
the
equi
vale
nt w
horl
to th
e st
igm
a of
the
alte
rnat
e m
orph
s to
be
the
reci
proc
al (L
stig
ma
vs u
pper
sta
men
who
rl of
S fl
ower
S s
tigm
a vs
low
er s
tam
en w
horl
of L
flow
er s
ee F
ig 1
) A
ppro
pria
te t
-tes
ts a
re g
iven
for
the
diff
eren
ces
betw
een
mea
ns o
f th
ese
leng
ths
(P
lt 0
05
P
lt 0
01
P lt
00
01)
Spec
ies
Popu
latio
n
Sam
ple
size
L S
Styl
e he
ight
U
pper
ant
her
heig
ht
Low
er a
nthe
r he
ight
St
igm
a-A
nthe
r se
para
tion
Mor
ph r
atio
(1
)L
SL
SL
SL
S
N a
lbim
argi
natu
sBo
uhac
hem
69 5
824
40
plusmn 3
8
9
77 plusmn
17
015
97
plusmn 2
14
24
54
plusmn 2
899
96 plusmn
17
8
21
80
plusmn 2
550
14
019
065
0
35N
alb
imar
gina
tus
Kel
ti53
45
243
7 plusmn
032
936
plusmn 0
19
148
9 plusmn
028
216
7 plusmn
036
849
plusmn 0
23
188
1 plusmn
041
270
0
870
66
034
N c
uatr
ecas
asii
Gra
zale
ma
66 3
415
08
plusmn 0
20
8
01 plusmn
01
713
79
plusmn 0
18 n
s 14
36
plusmn 0
269
05 plusmn
01
6
12
03
plusmn 0
270
72
104
065
0
35N
cua
trec
asas
iiC
azor
la42
58
146
0 plusmn
020
690
plusmn 0
12
131
8 plusmn
018
13
95
plusmn 0
158
25 plusmn
01
4
10
74
plusmn 0
110
65
162
042
0
58N
cal
cico
laSi
coacute50
43
163
6 plusmn
034
819
plusmn 0
18
144
6 plusmn
018
152
5 plusmn
029
925
plusmn 0
21
113
2 plusmn
018
112
1
380
69
031
N c
alci
cola
Sto
Ant
onio
51 4
416
49
plusmn 0
29
8
06 plusmn
02
715
30
plusmn 0
23 n
s 15
37
plusmn 0
229
44 plusmn
02
1
11
68
plusmn 0
231
11
106
076
0
24N
sca
beru
lus
Oliv
eira
51 3
115
12
plusmn 0
24
7
72 plusmn
01
714
56
plusmn 0
15 n
s 15
05
plusmn 0
279
21 plusmn
01
6
11
50
plusmn 0
240
071
490
86
014
N s
cabe
rulu
sEr
veda
l50
14
147
7 plusmn
029
778
plusmn 0
28
148
5 plusmn
021
ns
149
9 plusmn
065
986
plusmn 0
15
11
71
plusmn 0
370
22
208
077
0
23N
rup
icol
aA
ldea
quem
ada
42 3
614
94
plusmn 0
50
10
43
plusmn 0
2216
08
plusmn 0
27
18
19
plusmn 0
26 1
109
plusmn 0
25
139
8 plusmn
021
325
0
660
54
046
N r
upic
ola
Bola
del
Mun
do53
35
131
9 plusmn
024
816
plusmn 0
14
154
5 plusmn
023
ns
163
1 plusmn
029
105
6 plusmn
018
126
6 plusmn
025
312
2
400
72
028
(1)
Sam
ple
size
use
d fo
r m
orph
rat
io is
the
tot
al n
umbe
r of
flow
ers
colle
cted
(se
e Ta
ble
1)
occurrence of some moths around the flowers of N rupicolabut they were not observed to come into contact with flowersIn N watieri insect visits were much more abundant After a3-hour census we observed long-tongued visitors (diurnalmoths six visits pierid butterflies three visits) and short-tongued hoverflies (five visits) Despite considerable time(9 h) spent on N albimarginatus we were not able to spot anypollinator
Chloroplast sequence variation
Length of trnL-F sequences in Narcissus ranges between337 bp (population 1 of Narcissus pallidulus) and 365 bp(population 2 of Narcissus marvieri ) This sequence variationis caused by seven indels of 1ndash19 bp Number of variableparsimony-informative characters is of 1711 across species ofNarcissus The highest pairwise divergence (Kimura-2-parametermodel) of species sequences (34) was found betweenaccessions of N bulbocodium and N cyclamineus Sequencedivergence neither was found between species of sectApodanthi (excluding N calcicola and N scaberulus) nor wasobserved between two of the three species of the N triandruscomplex (N pallidulus and N lusitanicus) Kimura-2-parameter distances between accessions of the same speciesdid not rendered either divergence
In Narcissus sect Apodanthi length variation of trnT-Lsequences ranges between 827 bp (N marvieri ) and 966 bp(N calcicola) Ten gaps in the trnT-L sequence matrix from1 bp to 130 bp are responsible for great sequence variationAn unusual indel of 130 bp is shared by N rupicola N marv-ieri and N watieri The number of variableparsimony-informative characters is 1911 in Narcissus sect ApodanthiThe highest pairwise divergence (Kimura-2-parameter model)between sequence species (14) was found between acces-sions of N calcicola and N cuatrecasasii whereas the lowest(021) was between N calcicola and N scaberulus
Phylogenetic relationships
Maximum parsimony (MP) using the 27 trnL-F sequencesof Narcissus and Galanthus Lapiedra and Leucojum as theoutgroup yielded 1529 trees of 78 steps (CI 095 and RI093 including uninformative characters) (results not shown)A congruent somewhat more resolved topology was retrievedin the semistrict consensus tree of 13 225 trees (101 stepsCI 085 and RI 085) when coding the seven indels asadditional characters however new branches do not displaysignificant values (bootstrap below 50) (Fig 5) From bothanalyses (with and without indel coding) a basal polytomy isdepicted in the semistrict consensus tree of which two well-supported clades are formed by the accessions of five speciesof sect Apodanthi (clade 1) and six species including the threeof the N triandrus complex (clade 2) Relatively long branchesin the Neighbor-Joining (NJ) trees using both alignments
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research240
of insects visiting the flowers and recorded positive dataproviding that they enter into the tube and do contact any sexorgan Species subject to study were N rupicola (time effortof 12 h including some nocturnal periods) N cuatrecasasii(12 h) N albimarginatus (9 h) and N watieri (3 h) Observationtiming and area depend on density of flowers in everypopulation In sparse patches we made several censuses of15 min observing all insects visiting flowers in an area ofapprox 100 m2 We also made static observations in densepatches on all flowers in small areas (approx 10 m2)
Phylogenetic studies
PCR amplification and sequencing Total genomic DNA wasextracted from silica-dried material collected in the field andthe living collection of the Royal Botanic Garden of MadridSpain DNA extractions were performed using the DNeasyPlant Mini Kit (QIAGEN Inc Chatsworth CA USA) ForPolymerase Chain Reaction (PCR) amplifications we used aPerkin-Elmer (Foster City CA USA) PCR System 9700 thermalcycler primers trna and trnb for trnT(UGU)-trnL(UAA)trne and trn f for trnL(UAA)-trnF(GAA) and PCR conditionsfollowing Taberlet et al (1991) 51degC of annealing temperaturefor the trnT-L and 50degC for the trnL-F plus 2-min elongationtime Amplified products were cleaned using spin filter columns(PCR Clean-up kit MoBio Laboratories Solana Beach CAUSA) following the protocols provided by the manufacturerCleaned products were then directly sequenced using dyeterminators (Big Dye Terminator vs 20 Applied BiosystemsLittle Chalfont UK) and run into polyacrylamideelectrophoresis gels (7) using an Applied Biosystems Prismmodel 3700 automated sequencer Primers trna trnb trneand trn f were also used for cycle sequencing of the trnT-trnLand trnL-trnF spacers under the following conditions 95degCfor 2 min followed by 25 cycles of 95degC for 10 s 50degC for 5s and 60degC for 4 min Sequence data were placed in a contigfile and edited using the program Seqed (Applied BiosystemsFoster City CA USA) The limits of the trnT-trnL and trnL-trnF regions were determined by comparison with otherAmaryllidaceae sequences available in the GenBank
Sequence analyses Three different sequence matrices wereused to perform the phylogenetic analyses trnL-F matrixof Narcissus trnT-L matrix of Narcissus sect Apodanthiand a combined matrix of trnL-F and trnT-L for only thespecies of Narcissus sect Apodanthi as ingroup Alignmentsof the 30 trnL-F and the nine trnT-L sequences wereobtained using the program Clustal X 162b (httpevolutiongeneticswashingtoneduphylipsoftwareetc1html)with further manual adjustments Two phylogenetic analyseswere performed by using two primary approaches to treeconstruction tree-searching (parsimony) and algorithmic(Neighbor-Joining) Parsimony-based analyses were conductedusing Fitch parsimony (as implemented in Swofford
1999) with equal weighting of all characters and oftransitionstransversions Heuristic searches were replicated100 times with random taxon-addition sequences TreeBisection-Reconnection (TBR) branch swapping and withthe options MULPARS and STEEPEST DESCENT ineffect Relative support for clades identified by parsimonyanalysis was assessed by bootstrapping with the heuristicsearch strategy as indicated above In addition phylogeneticreconstructions of sequences were also performed in using the Kimura 2-parameter distance model (Kimura 1980)and the Neighbor-Joining method (Saitou amp Nei 1987)Indels in the noncoding regions trnT-L and trnL-F werecoded as appended characters following the logic of Kelchner(2000) and Simmons amp Ochoterena (2000) As a resultadditional analyses were performed when coding parsimony-informative indels affecting the trnL-F and trnT-L matricesCharacter evolution of flower polymorphism was investigatedusing MacClade 35 (Maddison amp Maddison 1992) Fourstates of flower polymorphism (approach herkogamy non-herkogamous monomorphism style dimorphism and hetero-styly) were traced and mapped on the resulting trnL-F andtrnT-L phylogenies of the consensus tree of all maximum-parsimony trees We explored both accelerated transformation(ACCTRAN) and delayed transformation (DELTRAN)optimizations with equal weighting of all characters andtransitions among all states equally probable
Results
Style polymorphism
Average values of sex organ position for the seven species ofsect Apodanthi and for N pallidulus are shown in Tables 3 4and 5 and comparative diagrams in Fig 2 Species in sectionApodanthi have a remarkable variation in style polymorphismdistyly in N albimarginatus style dimorphism in four species(N calcicola N cuatrecasasii N rupicola and N scaberulus)and monomorphism in N marvieri and N watieri (Fig 2Table 3) Narcissus albimarginatus shows two style-lengthmorphs with approximate reciprocal positioning of anthersand style Two patterns were found for dimorphic speciesthree species (N calcicola N scaberulus N cuatrecasasii ) presentthe L-morph with the stigma almost at an equivalent level ofthat of upper anthers whereas in N rupicola the stigma of theL-morph is positioned between the two anther levels The latterspecies displayed the highest values of stigma-anther separation(Table 4) that is the lowest reciprocity among dimorphicspecies Monomorphic species present a position of sex organssimilar to that in L-morph of dimorphic N rupicola (Fig 2)
Narcissus pallidulus is the only trimorphic species in ourstudy with mostly reciprocal positioning of stigmas andstamen whorls Although equivalent positions are not equal inall cases sequence height of sex organs corresponds to typicaltristylous species
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 241
At the population level morphs are in general unequallyrepresented L-morph is the most frequent (Tables 4 and 5) inall species and populations with the exception of one popu-lation of N cuatrecasasii and one of N rupicola which showsimilar frequency of morphs (isoplethy)
Perianth variation
Values of perianth variables are averaged over populationsand morphs (Table 6) Univariate morphometrical analyses offlower tubes tepals coronas and flower angle of all studiedspecies and populations revealed clearly absence of significantdifferences between morphs (82 between-morph comparisonsresults not shown) There were only slightly significant dif-ferences in tristylous N pallidulus at Cazorla for corona height(P = 001) and in distylous N albimarginatus at Bouhachemfor perianth width (P = 005)
Principal component analysis is shown in Fig 3 The firstsecond and third axes accounted for 616 233 and 94of the variance although only Axes 1 and 2 have eigenvalueshigher than 10 and are hence further discussed These twoaxes depict quite clearly the relationships between species andpopulations in perianth morphology At the negative extremeof Axis 1 we find together N pallidulus and N albimarginatusand at the positive extreme of this axis N marvieri Nwatieri and N rupicola form a second group Axis 2 separatesat its positive end the Iberian endemics N cuatrecasasii N cal-cicola and N scaberulus Eigenvectors of perianth variablesreflect trends of morphological variation The highest valuesfor Axis 1 are flower angle to stalk (04519) width of flowertube (minus04410) and corona height (minus04217) Largest eigen-vectors for Axis 2 correspond to corona width (minus06470)tepal length (minus04362) and flower tube length (minus04087)These results indicate that populations of two species (Npallidulus and N albimarginatus) at the left in the ordinationplot are characterised by small angles (pendulous flowers)wide flower tubes and long coronas (Fig 4) A second groupof species (N rupicola N watieri N marvieri ) display pre-dominantly erect flowers narrow and very long tubes andfunnel-shaped coronas On the top of the ordination plotthree species (N cuatrecasasii N calcicola N scaberulus) arecharacterised by patent flowers cup-shaped coronas smalltepals and moderate flower-tube length
The interrelationship between perianth variables and stylarpolymorphism is evaluated by means of multiple regressionanalyses We calculated two measurements of reciprocity inthe position of sex organs considering or not flower sizeeffects Both regression equations fitted quite similarly withscores of populations along PCA axes on perianth variablesAs a result we may conclude that there is a significant rela-tionship between perianth morphology and sex organ posi-tion across populations and species with 72ndash80 of thevariance being explained by the regression models (Table 7)Positioning of reciprocal stigma and anthers (heterostyly) cor-responds with the group of pendulous flowers Style dimorphicspecies (N calcicola N cuatrecasasii N scaberulus see Table 4)with medium values of reciprocity have patent flowers amongother features (see above) Monomorphism and less reciprocalstyle dimorphism in N rupicola (Table 4) correspond to theerect flowered group
Pollinators
As typically occurs in early blooming species it was difficultto observe pollinators in four species of Narcissus sectApodanthi (0ndash467 insect visits per hour) Twelve hours ofobservations in N cuatrecasasii ( Jaeacuten Cazorla range) revealedrarity of pollinators although medium sized solitary bees(Anthophora sp) were consistently spotted (10 visits) probingnectar and small unidentified bees (8 visits) reaching pollenof the upper anthers In the same area Anthophora bees visitedflowers of Helleborus foetidus but always in different boutsCasual observations made on populations at the oppositeextreme of the geographic range of N cuatrecasasii (CaacutedizGrazalema range) allow recognising Anthophora as the mostcommon visitor By censuses in N rupicola at the core of itsgeographic range (Avila Puerto de Mijares Madrid Puertode Navacerrada) we observed only two visits by the mothMacroglossum stellatarum one by the hoverfly Eristalis tenaxand one by an unidentified small bee Flies and small beestried to enter the flower tube but they only contacted upperanthers and L-stigma because of its narrowness and shorttongues Infrequent pollinator visits is not the result of lownumber of insects as they visited profusely another co-occurring flowers of some genera (Gagea Ranunculus SaxifragaArmeria Cytisus Erysimum) Some nocturnal censuses revealed
Table 3 Height of stamens and stigmas in monomorphic species of Narcissus sect Apodanthi Values are means plusmn 1 se in mm
Species Population Sample size Style heightUpper anther height
Lower anther height
Stigma-Anther separation
N marvieri Zeretken 51 2069 plusmn 030 2255 plusmn 032 1852 plusmn 035 186N marvieri Tazzeka 50 1001 plusmn 015 1234 plusmn 018 865 plusmn 014 233N watieri Oukaimeden 50 2092 plusmn 028 2205 plusmn 029 1815 plusmn 023 113N watieri TizirsquonrsquoTest 26 1736 plusmn 030 1927 plusmn 040 1595 plusmn 041 191
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research242
Tabl
e 4
Hei
ght
of s
tam
ens
and
stig
mas
in d
imor
phic
spe
cies
of
Nar
ciss
us s
ect
Apo
dant
hi L
lon
g-st
yled
mor
ph S
sho
rt-s
tyle
d m
orph
Val
ues
are
mea
ns plusmn
1 s
e in
mm
For
est
imat
ing
stig
ma-
anth
er s
epar
atio
n w
e co
nsid
er o
nly
the
equi
vale
nt w
horl
to th
e st
igm
a of
the
alte
rnat
e m
orph
s to
be
the
reci
proc
al (L
stig
ma
vs u
pper
sta
men
who
rl of
S fl
ower
S s
tigm
a vs
low
er s
tam
en w
horl
of L
flow
er s
ee F
ig 1
) A
ppro
pria
te t
-tes
ts a
re g
iven
for
the
diff
eren
ces
betw
een
mea
ns o
f th
ese
leng
ths
(P
lt 0
05
P
lt 0
01
P lt
00
01)
Spec
ies
Popu
latio
n
Sam
ple
size
L S
Styl
e he
ight
U
pper
ant
her
heig
ht
Low
er a
nthe
r he
ight
St
igm
a-A
nthe
r se
para
tion
Mor
ph r
atio
(1
)L
SL
SL
SL
S
N a
lbim
argi
natu
sBo
uhac
hem
69 5
824
40
plusmn 3
8
9
77 plusmn
17
015
97
plusmn 2
14
24
54
plusmn 2
899
96 plusmn
17
8
21
80
plusmn 2
550
14
019
065
0
35N
alb
imar
gina
tus
Kel
ti53
45
243
7 plusmn
032
936
plusmn 0
19
148
9 plusmn
028
216
7 plusmn
036
849
plusmn 0
23
188
1 plusmn
041
270
0
870
66
034
N c
uatr
ecas
asii
Gra
zale
ma
66 3
415
08
plusmn 0
20
8
01 plusmn
01
713
79
plusmn 0
18 n
s 14
36
plusmn 0
269
05 plusmn
01
6
12
03
plusmn 0
270
72
104
065
0
35N
cua
trec
asas
iiC
azor
la42
58
146
0 plusmn
020
690
plusmn 0
12
131
8 plusmn
018
13
95
plusmn 0
158
25 plusmn
01
4
10
74
plusmn 0
110
65
162
042
0
58N
cal
cico
laSi
coacute50
43
163
6 plusmn
034
819
plusmn 0
18
144
6 plusmn
018
152
5 plusmn
029
925
plusmn 0
21
113
2 plusmn
018
112
1
380
69
031
N c
alci
cola
Sto
Ant
onio
51 4
416
49
plusmn 0
29
8
06 plusmn
02
715
30
plusmn 0
23 n
s 15
37
plusmn 0
229
44 plusmn
02
1
11
68
plusmn 0
231
11
106
076
0
24N
sca
beru
lus
Oliv
eira
51 3
115
12
plusmn 0
24
7
72 plusmn
01
714
56
plusmn 0
15 n
s 15
05
plusmn 0
279
21 plusmn
01
6
11
50
plusmn 0
240
071
490
86
014
N s
cabe
rulu
sEr
veda
l50
14
147
7 plusmn
029
778
plusmn 0
28
148
5 plusmn
021
ns
149
9 plusmn
065
986
plusmn 0
15
11
71
plusmn 0
370
22
208
077
0
23N
rup
icol
aA
ldea
quem
ada
42 3
614
94
plusmn 0
50
10
43
plusmn 0
2216
08
plusmn 0
27
18
19
plusmn 0
26 1
109
plusmn 0
25
139
8 plusmn
021
325
0
660
54
046
N r
upic
ola
Bola
del
Mun
do53
35
131
9 plusmn
024
816
plusmn 0
14
154
5 plusmn
023
ns
163
1 plusmn
029
105
6 plusmn
018
126
6 plusmn
025
312
2
400
72
028
(1)
Sam
ple
size
use
d fo
r m
orph
rat
io is
the
tot
al n
umbe
r of
flow
ers
colle
cted
(se
e Ta
ble
1)
occurrence of some moths around the flowers of N rupicolabut they were not observed to come into contact with flowersIn N watieri insect visits were much more abundant After a3-hour census we observed long-tongued visitors (diurnalmoths six visits pierid butterflies three visits) and short-tongued hoverflies (five visits) Despite considerable time(9 h) spent on N albimarginatus we were not able to spot anypollinator
Chloroplast sequence variation
Length of trnL-F sequences in Narcissus ranges between337 bp (population 1 of Narcissus pallidulus) and 365 bp(population 2 of Narcissus marvieri ) This sequence variationis caused by seven indels of 1ndash19 bp Number of variableparsimony-informative characters is of 1711 across species ofNarcissus The highest pairwise divergence (Kimura-2-parametermodel) of species sequences (34) was found betweenaccessions of N bulbocodium and N cyclamineus Sequencedivergence neither was found between species of sectApodanthi (excluding N calcicola and N scaberulus) nor wasobserved between two of the three species of the N triandruscomplex (N pallidulus and N lusitanicus) Kimura-2-parameter distances between accessions of the same speciesdid not rendered either divergence
In Narcissus sect Apodanthi length variation of trnT-Lsequences ranges between 827 bp (N marvieri ) and 966 bp(N calcicola) Ten gaps in the trnT-L sequence matrix from1 bp to 130 bp are responsible for great sequence variationAn unusual indel of 130 bp is shared by N rupicola N marv-ieri and N watieri The number of variableparsimony-informative characters is 1911 in Narcissus sect ApodanthiThe highest pairwise divergence (Kimura-2-parameter model)between sequence species (14) was found between acces-sions of N calcicola and N cuatrecasasii whereas the lowest(021) was between N calcicola and N scaberulus
Phylogenetic relationships
Maximum parsimony (MP) using the 27 trnL-F sequencesof Narcissus and Galanthus Lapiedra and Leucojum as theoutgroup yielded 1529 trees of 78 steps (CI 095 and RI093 including uninformative characters) (results not shown)A congruent somewhat more resolved topology was retrievedin the semistrict consensus tree of 13 225 trees (101 stepsCI 085 and RI 085) when coding the seven indels asadditional characters however new branches do not displaysignificant values (bootstrap below 50) (Fig 5) From bothanalyses (with and without indel coding) a basal polytomy isdepicted in the semistrict consensus tree of which two well-supported clades are formed by the accessions of five speciesof sect Apodanthi (clade 1) and six species including the threeof the N triandrus complex (clade 2) Relatively long branchesin the Neighbor-Joining (NJ) trees using both alignments
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 241
At the population level morphs are in general unequallyrepresented L-morph is the most frequent (Tables 4 and 5) inall species and populations with the exception of one popu-lation of N cuatrecasasii and one of N rupicola which showsimilar frequency of morphs (isoplethy)
Perianth variation
Values of perianth variables are averaged over populationsand morphs (Table 6) Univariate morphometrical analyses offlower tubes tepals coronas and flower angle of all studiedspecies and populations revealed clearly absence of significantdifferences between morphs (82 between-morph comparisonsresults not shown) There were only slightly significant dif-ferences in tristylous N pallidulus at Cazorla for corona height(P = 001) and in distylous N albimarginatus at Bouhachemfor perianth width (P = 005)
Principal component analysis is shown in Fig 3 The firstsecond and third axes accounted for 616 233 and 94of the variance although only Axes 1 and 2 have eigenvalueshigher than 10 and are hence further discussed These twoaxes depict quite clearly the relationships between species andpopulations in perianth morphology At the negative extremeof Axis 1 we find together N pallidulus and N albimarginatusand at the positive extreme of this axis N marvieri Nwatieri and N rupicola form a second group Axis 2 separatesat its positive end the Iberian endemics N cuatrecasasii N cal-cicola and N scaberulus Eigenvectors of perianth variablesreflect trends of morphological variation The highest valuesfor Axis 1 are flower angle to stalk (04519) width of flowertube (minus04410) and corona height (minus04217) Largest eigen-vectors for Axis 2 correspond to corona width (minus06470)tepal length (minus04362) and flower tube length (minus04087)These results indicate that populations of two species (Npallidulus and N albimarginatus) at the left in the ordinationplot are characterised by small angles (pendulous flowers)wide flower tubes and long coronas (Fig 4) A second groupof species (N rupicola N watieri N marvieri ) display pre-dominantly erect flowers narrow and very long tubes andfunnel-shaped coronas On the top of the ordination plotthree species (N cuatrecasasii N calcicola N scaberulus) arecharacterised by patent flowers cup-shaped coronas smalltepals and moderate flower-tube length
The interrelationship between perianth variables and stylarpolymorphism is evaluated by means of multiple regressionanalyses We calculated two measurements of reciprocity inthe position of sex organs considering or not flower sizeeffects Both regression equations fitted quite similarly withscores of populations along PCA axes on perianth variablesAs a result we may conclude that there is a significant rela-tionship between perianth morphology and sex organ posi-tion across populations and species with 72ndash80 of thevariance being explained by the regression models (Table 7)Positioning of reciprocal stigma and anthers (heterostyly) cor-responds with the group of pendulous flowers Style dimorphicspecies (N calcicola N cuatrecasasii N scaberulus see Table 4)with medium values of reciprocity have patent flowers amongother features (see above) Monomorphism and less reciprocalstyle dimorphism in N rupicola (Table 4) correspond to theerect flowered group
Pollinators
As typically occurs in early blooming species it was difficultto observe pollinators in four species of Narcissus sectApodanthi (0ndash467 insect visits per hour) Twelve hours ofobservations in N cuatrecasasii ( Jaeacuten Cazorla range) revealedrarity of pollinators although medium sized solitary bees(Anthophora sp) were consistently spotted (10 visits) probingnectar and small unidentified bees (8 visits) reaching pollenof the upper anthers In the same area Anthophora bees visitedflowers of Helleborus foetidus but always in different boutsCasual observations made on populations at the oppositeextreme of the geographic range of N cuatrecasasii (CaacutedizGrazalema range) allow recognising Anthophora as the mostcommon visitor By censuses in N rupicola at the core of itsgeographic range (Avila Puerto de Mijares Madrid Puertode Navacerrada) we observed only two visits by the mothMacroglossum stellatarum one by the hoverfly Eristalis tenaxand one by an unidentified small bee Flies and small beestried to enter the flower tube but they only contacted upperanthers and L-stigma because of its narrowness and shorttongues Infrequent pollinator visits is not the result of lownumber of insects as they visited profusely another co-occurring flowers of some genera (Gagea Ranunculus SaxifragaArmeria Cytisus Erysimum) Some nocturnal censuses revealed
Table 3 Height of stamens and stigmas in monomorphic species of Narcissus sect Apodanthi Values are means plusmn 1 se in mm
Species Population Sample size Style heightUpper anther height
Lower anther height
Stigma-Anther separation
N marvieri Zeretken 51 2069 plusmn 030 2255 plusmn 032 1852 plusmn 035 186N marvieri Tazzeka 50 1001 plusmn 015 1234 plusmn 018 865 plusmn 014 233N watieri Oukaimeden 50 2092 plusmn 028 2205 plusmn 029 1815 plusmn 023 113N watieri TizirsquonrsquoTest 26 1736 plusmn 030 1927 plusmn 040 1595 plusmn 041 191
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research242
Tabl
e 4
Hei
ght
of s
tam
ens
and
stig
mas
in d
imor
phic
spe
cies
of
Nar
ciss
us s
ect
Apo
dant
hi L
lon
g-st
yled
mor
ph S
sho
rt-s
tyle
d m
orph
Val
ues
are
mea
ns plusmn
1 s
e in
mm
For
est
imat
ing
stig
ma-
anth
er s
epar
atio
n w
e co
nsid
er o
nly
the
equi
vale
nt w
horl
to th
e st
igm
a of
the
alte
rnat
e m
orph
s to
be
the
reci
proc
al (L
stig
ma
vs u
pper
sta
men
who
rl of
S fl
ower
S s
tigm
a vs
low
er s
tam
en w
horl
of L
flow
er s
ee F
ig 1
) A
ppro
pria
te t
-tes
ts a
re g
iven
for
the
diff
eren
ces
betw
een
mea
ns o
f th
ese
leng
ths
(P
lt 0
05
P
lt 0
01
P lt
00
01)
Spec
ies
Popu
latio
n
Sam
ple
size
L S
Styl
e he
ight
U
pper
ant
her
heig
ht
Low
er a
nthe
r he
ight
St
igm
a-A
nthe
r se
para
tion
Mor
ph r
atio
(1
)L
SL
SL
SL
S
N a
lbim
argi
natu
sBo
uhac
hem
69 5
824
40
plusmn 3
8
9
77 plusmn
17
015
97
plusmn 2
14
24
54
plusmn 2
899
96 plusmn
17
8
21
80
plusmn 2
550
14
019
065
0
35N
alb
imar
gina
tus
Kel
ti53
45
243
7 plusmn
032
936
plusmn 0
19
148
9 plusmn
028
216
7 plusmn
036
849
plusmn 0
23
188
1 plusmn
041
270
0
870
66
034
N c
uatr
ecas
asii
Gra
zale
ma
66 3
415
08
plusmn 0
20
8
01 plusmn
01
713
79
plusmn 0
18 n
s 14
36
plusmn 0
269
05 plusmn
01
6
12
03
plusmn 0
270
72
104
065
0
35N
cua
trec
asas
iiC
azor
la42
58
146
0 plusmn
020
690
plusmn 0
12
131
8 plusmn
018
13
95
plusmn 0
158
25 plusmn
01
4
10
74
plusmn 0
110
65
162
042
0
58N
cal
cico
laSi
coacute50
43
163
6 plusmn
034
819
plusmn 0
18
144
6 plusmn
018
152
5 plusmn
029
925
plusmn 0
21
113
2 plusmn
018
112
1
380
69
031
N c
alci
cola
Sto
Ant
onio
51 4
416
49
plusmn 0
29
8
06 plusmn
02
715
30
plusmn 0
23 n
s 15
37
plusmn 0
229
44 plusmn
02
1
11
68
plusmn 0
231
11
106
076
0
24N
sca
beru
lus
Oliv
eira
51 3
115
12
plusmn 0
24
7
72 plusmn
01
714
56
plusmn 0
15 n
s 15
05
plusmn 0
279
21 plusmn
01
6
11
50
plusmn 0
240
071
490
86
014
N s
cabe
rulu
sEr
veda
l50
14
147
7 plusmn
029
778
plusmn 0
28
148
5 plusmn
021
ns
149
9 plusmn
065
986
plusmn 0
15
11
71
plusmn 0
370
22
208
077
0
23N
rup
icol
aA
ldea
quem
ada
42 3
614
94
plusmn 0
50
10
43
plusmn 0
2216
08
plusmn 0
27
18
19
plusmn 0
26 1
109
plusmn 0
25
139
8 plusmn
021
325
0
660
54
046
N r
upic
ola
Bola
del
Mun
do53
35
131
9 plusmn
024
816
plusmn 0
14
154
5 plusmn
023
ns
163
1 plusmn
029
105
6 plusmn
018
126
6 plusmn
025
312
2
400
72
028
(1)
Sam
ple
size
use
d fo
r m
orph
rat
io is
the
tot
al n
umbe
r of
flow
ers
colle
cted
(se
e Ta
ble
1)
occurrence of some moths around the flowers of N rupicolabut they were not observed to come into contact with flowersIn N watieri insect visits were much more abundant After a3-hour census we observed long-tongued visitors (diurnalmoths six visits pierid butterflies three visits) and short-tongued hoverflies (five visits) Despite considerable time(9 h) spent on N albimarginatus we were not able to spot anypollinator
Chloroplast sequence variation
Length of trnL-F sequences in Narcissus ranges between337 bp (population 1 of Narcissus pallidulus) and 365 bp(population 2 of Narcissus marvieri ) This sequence variationis caused by seven indels of 1ndash19 bp Number of variableparsimony-informative characters is of 1711 across species ofNarcissus The highest pairwise divergence (Kimura-2-parametermodel) of species sequences (34) was found betweenaccessions of N bulbocodium and N cyclamineus Sequencedivergence neither was found between species of sectApodanthi (excluding N calcicola and N scaberulus) nor wasobserved between two of the three species of the N triandruscomplex (N pallidulus and N lusitanicus) Kimura-2-parameter distances between accessions of the same speciesdid not rendered either divergence
In Narcissus sect Apodanthi length variation of trnT-Lsequences ranges between 827 bp (N marvieri ) and 966 bp(N calcicola) Ten gaps in the trnT-L sequence matrix from1 bp to 130 bp are responsible for great sequence variationAn unusual indel of 130 bp is shared by N rupicola N marv-ieri and N watieri The number of variableparsimony-informative characters is 1911 in Narcissus sect ApodanthiThe highest pairwise divergence (Kimura-2-parameter model)between sequence species (14) was found between acces-sions of N calcicola and N cuatrecasasii whereas the lowest(021) was between N calcicola and N scaberulus
Phylogenetic relationships
Maximum parsimony (MP) using the 27 trnL-F sequencesof Narcissus and Galanthus Lapiedra and Leucojum as theoutgroup yielded 1529 trees of 78 steps (CI 095 and RI093 including uninformative characters) (results not shown)A congruent somewhat more resolved topology was retrievedin the semistrict consensus tree of 13 225 trees (101 stepsCI 085 and RI 085) when coding the seven indels asadditional characters however new branches do not displaysignificant values (bootstrap below 50) (Fig 5) From bothanalyses (with and without indel coding) a basal polytomy isdepicted in the semistrict consensus tree of which two well-supported clades are formed by the accessions of five speciesof sect Apodanthi (clade 1) and six species including the threeof the N triandrus complex (clade 2) Relatively long branchesin the Neighbor-Joining (NJ) trees using both alignments
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research242
Tabl
e 4
Hei
ght
of s
tam
ens
and
stig
mas
in d
imor
phic
spe
cies
of
Nar
ciss
us s
ect
Apo
dant
hi L
lon
g-st
yled
mor
ph S
sho
rt-s
tyle
d m
orph
Val
ues
are
mea
ns plusmn
1 s
e in
mm
For
est
imat
ing
stig
ma-
anth
er s
epar
atio
n w
e co
nsid
er o
nly
the
equi
vale
nt w
horl
to th
e st
igm
a of
the
alte
rnat
e m
orph
s to
be
the
reci
proc
al (L
stig
ma
vs u
pper
sta
men
who
rl of
S fl
ower
S s
tigm
a vs
low
er s
tam
en w
horl
of L
flow
er s
ee F
ig 1
) A
ppro
pria
te t
-tes
ts a
re g
iven
for
the
diff
eren
ces
betw
een
mea
ns o
f th
ese
leng
ths
(P
lt 0
05
P
lt 0
01
P lt
00
01)
Spec
ies
Popu
latio
n
Sam
ple
size
L S
Styl
e he
ight
U
pper
ant
her
heig
ht
Low
er a
nthe
r he
ight
St
igm
a-A
nthe
r se
para
tion
Mor
ph r
atio
(1
)L
SL
SL
SL
S
N a
lbim
argi
natu
sBo
uhac
hem
69 5
824
40
plusmn 3
8
9
77 plusmn
17
015
97
plusmn 2
14
24
54
plusmn 2
899
96 plusmn
17
8
21
80
plusmn 2
550
14
019
065
0
35N
alb
imar
gina
tus
Kel
ti53
45
243
7 plusmn
032
936
plusmn 0
19
148
9 plusmn
028
216
7 plusmn
036
849
plusmn 0
23
188
1 plusmn
041
270
0
870
66
034
N c
uatr
ecas
asii
Gra
zale
ma
66 3
415
08
plusmn 0
20
8
01 plusmn
01
713
79
plusmn 0
18 n
s 14
36
plusmn 0
269
05 plusmn
01
6
12
03
plusmn 0
270
72
104
065
0
35N
cua
trec
asas
iiC
azor
la42
58
146
0 plusmn
020
690
plusmn 0
12
131
8 plusmn
018
13
95
plusmn 0
158
25 plusmn
01
4
10
74
plusmn 0
110
65
162
042
0
58N
cal
cico
laSi
coacute50
43
163
6 plusmn
034
819
plusmn 0
18
144
6 plusmn
018
152
5 plusmn
029
925
plusmn 0
21
113
2 plusmn
018
112
1
380
69
031
N c
alci
cola
Sto
Ant
onio
51 4
416
49
plusmn 0
29
8
06 plusmn
02
715
30
plusmn 0
23 n
s 15
37
plusmn 0
229
44 plusmn
02
1
11
68
plusmn 0
231
11
106
076
0
24N
sca
beru
lus
Oliv
eira
51 3
115
12
plusmn 0
24
7
72 plusmn
01
714
56
plusmn 0
15 n
s 15
05
plusmn 0
279
21 plusmn
01
6
11
50
plusmn 0
240
071
490
86
014
N s
cabe
rulu
sEr
veda
l50
14
147
7 plusmn
029
778
plusmn 0
28
148
5 plusmn
021
ns
149
9 plusmn
065
986
plusmn 0
15
11
71
plusmn 0
370
22
208
077
0
23N
rup
icol
aA
ldea
quem
ada
42 3
614
94
plusmn 0
50
10
43
plusmn 0
2216
08
plusmn 0
27
18
19
plusmn 0
26 1
109
plusmn 0
25
139
8 plusmn
021
325
0
660
54
046
N r
upic
ola
Bola
del
Mun
do53
35
131
9 plusmn
024
816
plusmn 0
14
154
5 plusmn
023
ns
163
1 plusmn
029
105
6 plusmn
018
126
6 plusmn
025
312
2
400
72
028
(1)
Sam
ple
size
use
d fo
r m
orph
rat
io is
the
tot
al n
umbe
r of
flow
ers
colle
cted
(se
e Ta
ble
1)
occurrence of some moths around the flowers of N rupicolabut they were not observed to come into contact with flowersIn N watieri insect visits were much more abundant After a3-hour census we observed long-tongued visitors (diurnalmoths six visits pierid butterflies three visits) and short-tongued hoverflies (five visits) Despite considerable time(9 h) spent on N albimarginatus we were not able to spot anypollinator
Chloroplast sequence variation
Length of trnL-F sequences in Narcissus ranges between337 bp (population 1 of Narcissus pallidulus) and 365 bp(population 2 of Narcissus marvieri ) This sequence variationis caused by seven indels of 1ndash19 bp Number of variableparsimony-informative characters is of 1711 across species ofNarcissus The highest pairwise divergence (Kimura-2-parametermodel) of species sequences (34) was found betweenaccessions of N bulbocodium and N cyclamineus Sequencedivergence neither was found between species of sectApodanthi (excluding N calcicola and N scaberulus) nor wasobserved between two of the three species of the N triandruscomplex (N pallidulus and N lusitanicus) Kimura-2-parameter distances between accessions of the same speciesdid not rendered either divergence
In Narcissus sect Apodanthi length variation of trnT-Lsequences ranges between 827 bp (N marvieri ) and 966 bp(N calcicola) Ten gaps in the trnT-L sequence matrix from1 bp to 130 bp are responsible for great sequence variationAn unusual indel of 130 bp is shared by N rupicola N marv-ieri and N watieri The number of variableparsimony-informative characters is 1911 in Narcissus sect ApodanthiThe highest pairwise divergence (Kimura-2-parameter model)between sequence species (14) was found between acces-sions of N calcicola and N cuatrecasasii whereas the lowest(021) was between N calcicola and N scaberulus
Phylogenetic relationships
Maximum parsimony (MP) using the 27 trnL-F sequencesof Narcissus and Galanthus Lapiedra and Leucojum as theoutgroup yielded 1529 trees of 78 steps (CI 095 and RI093 including uninformative characters) (results not shown)A congruent somewhat more resolved topology was retrievedin the semistrict consensus tree of 13 225 trees (101 stepsCI 085 and RI 085) when coding the seven indels asadditional characters however new branches do not displaysignificant values (bootstrap below 50) (Fig 5) From bothanalyses (with and without indel coding) a basal polytomy isdepicted in the semistrict consensus tree of which two well-supported clades are formed by the accessions of five speciesof sect Apodanthi (clade 1) and six species including the threeof the N triandrus complex (clade 2) Relatively long branchesin the Neighbor-Joining (NJ) trees using both alignments
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 243
Tabl
e 5
Hei
ght
of s
tam
ens
and
stig
mas
in t
rimor
phic
Nar
ciss
us p
allid
ulus
For
cal
cula
tion
of s
tigm
a-an
ther
sep
arat
ion
see
Tabl
e 4
L l
ong-
styl
ed m
orph
M m
id-s
tyle
d m
orph
S s
hort
-sty
led
mor
ph V
alue
s ar
e m
eans
plusmn 1
se
in m
m A
ppro
pria
te p
osth
oc c
ompa
rison
s of
mea
ns (
Tuke
y H
SD t
est
for
uneq
ual s
ampl
e si
zes)
aft
er o
ne-w
ay A
NO
VAs
are
give
n (
P lt
00
5
P lt
00
1
P
lt 0
001
) Sp
ecie
sPo
pula
tion
Stig
ma
heig
ht
Upp
er a
nthe
r he
ight
Lo
wer
ant
her h
eigh
t M
orph
rat
io
LM
SL
MS
LM
SL
MS
N p
allid
ulus
Segu
ra24
64
plusmn 0
3414
65
plusmn 0
318
68 plusmn
02
821
06
plusmn 0
3921
69
plusmn 0
2522
30
plusmn 0
5010
54
plusmn 0
1810
07
plusmn 0
1813
99
plusmn 0
410
40
042
0
18L
ndash
ndash
nsns
_ns
M
ndashndash
ndashndash
nsndash
ndashns
Sndash
ndashndash
ndashndash
ndashndash
ndashndash
Stig
ma-
Ant
her
Sepa
ratio
n L
Stig
ma-
Ant
her
Sepa
ratio
n M
Stig
ma-
Ant
her
Sepa
ratio
n S
295
234
066
186
139
N p
allid
ulus
Cal
era
258
9 plusmn
052
151
5 plusmn
035
903
plusmn 0
30
204
6 plusmn
049
207
7 plusmn
039
220
5 plusmn
042
108
2 plusmn
028
102
4 plusmn
040
140
7 plusmn
044
042
0
41
017
Lndash
ndashns
nsndash
ns
Mndash
ndash
ndash
ndashns
ndashndash
S
ndashndash
ndashndash
ndashndash
ndashndash
ndashSt
igm
a-A
nthe
r Se
para
tion
LSt
igm
a-A
nthe
r Se
para
tion
MSt
igm
a-A
nthe
r Se
para
tion
S
512
384
108
179
121
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research244
(results not shown) were congruent with the major well-supported clades of the MP trees Apart from three groupsof species retrieved from the parsimony-based analysis aforth grouping formed by the accessions of N scaberulus andN calcicola is recognised in the NJ tree (results not shown)Low molecular variation across trnL-F sequences and tree
resolution prevent from inferring whether species of sectApodanthi form a monophyletic group Our trnL-Fphylogeny (Fig 5) indicates that heterostylous species (Nalbimarginatus N lusitanicus N pallidulus and N triandrus)form part of two independent clades (clade 1 and clade 2)Optimization of character states for flower polymorphism
Fig 2 Sex organ position of species of Narcissus sect Apodanthi and N pallidulus (sect Cyclaminei) Numbers in parentheses correspond to populations in Table 1 For simplicity only two out of three stamens in each whorl are represented Drawings are based on mean values (mm) included in Tables 3ndash5
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 245
using MacClade reconstructions reveals occurrence ofheterostyly at least twice as a result of two independentevents (results not shown) Low number of nucleotidesubstitutions in trnL-F sequences within species of clade 2also prevent from inferring whether heterostyly in the Ntriandrus complex is the result of one or additionalevolutionary processes
MP analyses of trnT-L sequences of sect Apodanthi andtwo outgroup species yielded congruent results in bothanalyses with and without coding indels A most resolvedsemistrict consensus tree was retrieved when performing equalweighting of the 10 indels in which a basal-most position ofN papyraceus is followed by a clade of Narcissus sect Apodan-thi species resolved biphyletically into two subclades (Fig 6)Two outgroup species are however an insufficient number toinfer monophyly of sect Apodanthi Close relationshipbetween N scaberulus and N calcicola is evidenced by a well-defined trnT-L subclade (100 bootstrap) A second subcladeconsists of the distylous species N albimarginatus togetherwith N cuatrecasasii N rupicola N watieri and N marvieri
(98 bootstrap) Within this subclade basal position of Nalbimarginatus N cuatrecasasii and a pectinate branch ofthree species (76 bootstrap) is observed in which N rupi-cola is sister to N waitieri and N marvieri (77 bootstrap)Character reconstruction of flower polymorphism in sectApodanthi was traced onto the trnT-L tree (Fig 6) Mac-Clade reconstructions help interpret that distyly may be arelatively primitive character in sect Apodanthi whereasstyle monomorphism is most derived
Discussion
Our results clearly show that Narcissus sect Apodanthi is veryvariable both in terms of stylar conditions and in terms ofperianth features Moreover there is a consistent pattern ofconcordance between perianth features herein described andstyle polymorphism formerly outlined (Arroyo 2002)Interestingly morphological correlates of distyly only foundin sect Apodanthi are also described in the other reported caseof heterostyly in Narcissus (Barrett et al 1997)
Table 6 Perianth variables of species of Narcissus sect Apodanthi and N pallidulus (see Fig 1 for variable names)
Species Flower width Tepal length Corona width Corona height Tube length Tube width Flower angle
N pallidulus 3048 plusmn 023 1478 plusmn 013 868 plusmn 010 835 plusmn 009 1355 plusmn 009 454 plusmn 005 4638 plusmn 163N albimarginatus 3828 plusmn 032 1867 plusmn 017 740 plusmn 009 711 plusmn 007 1257 plusmn 014 474 plusmn 004 5415 plusmn 229N cuatrecasasii 2455 plusmn 019 1057 plusmn 009 420 plusmn 004 681 plusmn 008 1392 plusmn 008 464 plusmn 004 10265 plusmn 165N calcicola 1875 plusmn 014 795 plusmn 010 733 plusmn 007 566 plusmn 005 1392 plusmn 011 420 plusmn 006 11141 plusmn 130N scaberulus 1423 plusmn 014 587 plusmn 008 591 plusmn 007 419 plusmn 005 1449 plusmn 010 363 plusmn 005 12017 plusmn 167N rupicola 2319 plusmn 021 1101 plusmn 011 887 plusmn 013 519 plusmn 007 1899 plusmn 019 358 plusmn 004 15997 plusmn 120N marvieri 2127 plusmn 025 975 plusmn 016 974 plusmn 016 601 plusmn 012 1912 plusmn 053 332 plusmn 006 14683 plusmn 204N watieri 2319 plusmn 035 1070 plusmn 021 799 plusmn 019 374 plusmn 009 2272 plusmn 022 324 plusmn 005 16822 plusmn 161
For all species data are averaged over populations and in polymorphic species data are averaged over morphs Values are means plusmn 1 se in mm See text for statistical differences
Fig 3 Principal component analysis (PCA) of perianth variables of Narcissus sect Apodanthi and N pallidulus Species names are abbreviated by their initials and numbers correspond to populations in Table 1
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research246
Fig 4 Photographs of five Narcissus species displaying contrasted flower morphologies pendulous flowers (a and b) patent to erect flowers (c d and e) cup-shaped coronas (a b and c) and funnel-shaped coronal (d and e) (a) N pallidulus (b) N albimarginatus (c) N cuatrecasasii (d) N rupicola (e) N watieri
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 247
Flower morphology
Variation in sex organ position Description of distyly forthe first time in Narcissus based on a limited sample (Arroyoamp Barrett 2000) is herein confirmed with extensive data froma second population of N albimarginatus (Peacuterez et al unpub-
lished) The two populations studied of N pallidulus have atypical tristylous condition as it was previously documentedin detail (Barrett et al 1997 as N triandrus sl) and servedefficiently to contrast distyly In these two heterostylous speciessex organs show approximately reciprocal position Perfectreciprocity is seldom accomplished by typical heterostylous
Table 7 Results of multiple regression analyses of perianth features as resumed by PCA axes on stigma-anther reciprocity considering the possible effect of flower size (see details about measurements in the text)
Independent variables (perianth)
Dependent variables (sex organ reciprocity)
Size uncorrected stigma-anther separation Size corrected stigma-anther separation
Intercept 00771 17497PCA axis I 00126 01700PCA axis II minus00135 minus04616PCA axis III minus00145 nsR2 0804 0724F312 16452 10512
Fig 5 Semistrict consensus tree of 13 225 most-parsimonious trees of 101 steps from the analysis of the trnL-F sequences (with indel coding) of Narcissus and outgroup genera (CI 085 RI 085 including uninformative characters) Galanthus Lapiedra and Pancratium served as the outgroup Bootstrap values above 50 are shown above tree branches The three main clades are discussed in the text Numbersletters after species names refer to numbering in Table 2 as GeneBank (GB) accessions Heterostylous species are marked in bold Sectional classification of Narcissus is marked on the right margin
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research248
species (Lloyd et al 1990 Faivre amp McDade 2001)although it is required to have at least a reciprocity in thesequence of presentation of sex organs to pollinators (Lloyd ampWebb 1992ab)
The majority of species in sect Apodanthi (N cuatrecasasiiN calcicola N scaberulus N rupicola) has style dimorphismand deviated reciprocity However this deviation is variable inthis species group We had the opportunity to check sex organreciprocity in a continuous range of variation in the wholespecies sample Among the dimorphic species N cuatrecasasiiyielded the lowest anther-stigma separation and N rupicolathe highest That is the lowest and the highest deviation fromreciprocity respectively Style dimorphism has been shown tobe very frequent in Narcissus (Barrett et al 1996) and remainsone of the unsolved challenges to the Lloyd amp Webbrsquos(1992ab) model for the evolution of heterostyly The reportedvariability in style dimorphism may explain why this condi-tion is maintained in Narcissus At least in some cases theremay be enough disassortative mating as to maintain stylemorphs even without a reciprocal positioning
Only two species show monomorphism in sect Apodanthithe Moroccan endemics N marvieri and N watieri Interest-
ingly the position of the sex organs is very similar to that ofthe L-morph found in the Iberian N rupicola (Fig 2) Obvi-ously we cannot rule out existence of dimorphism in undis-covered populations of both species as both dimorphic andmonomorphic populations occur sometimes within the samespecies (N papyraceus Arroyo et al 2002 N tazetta Arroyoamp Dafni 1995 N dubius Baker et al 2000a N jonquilla J Arroyo unpublished)
We have found an L-biased morph ratio in most populationsof dimorphic and distylous species as described in a formerspecies survey (Barrett et al 1996) Given the occurrence ofintramorph compatibility in most of these species (Dulberger1964 Barrett et al 1996 Sage et al 1999 Baker et al 2000bArroyo et al 2002) it is interpreted that L-biased morphratios are a consequence of differential level of assortative-disassortative mating between morphs in dimorphic speciesThese ratios also depend on population size (Arroyo et al 2002Baker et al 2000a) and the level of sex organ reciprocity Allthese factors might explain why widely distributed species (Ncuatrecasasii N rupicola) display strong differences in morphratio between populations (Table 4) which larger surveyshave made clearer (Arroyo 2002 R Peacuterez unpublished)
Fig 6 Semistrict consensus of the three most-parsimonious trees of 90 steps from analysis of trnT-L sequences (with indel coding) of Narcissus sect Apodanthi and outgroup species (CI 096 RI 092 including uninformative characters) onto which style polymorphism has been mapped after implementing ACCTRAN optimization of MacClade (Maddison amp Maddison 1992) Narcissus bulbocodium ssp nivalis and N papyraceus were used as the outgroup Bootstrap values above 50 are also shown beside tree branches See Table 2 for accession numbers
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 249
Morph ratio in distylous N albimarginatus is also L-biasedin the two populations surveyed which contrasts with theformer isoplethy reported by Arroyo amp Barrett (2000) It islikely a higher level of assortative mating for the L-morph thanformerly suggested A breeding system program across speciesin sect Apodanthi is under study and may shed further lighton this respect At least N scaberulus N calcicola and N cuat-recasassi seem to present self-incompatibility and intramorphcompatibility (unpublished data)
Perianth features and pollinators Only two perianth featuresdisplayed significant differences between morphs within popu-lations of the only two heterostylous species studied Npallidulus (corona height) and N albimarginatus (flower width)These differences have been seldom interpreted as a mechanismto compensate asymmetric pollen flow (Ganders 1979) Alter-natively there is reliable evidence that reflects allometriceffects associated with different patterns of flower developmentof morphs (Dulberger 1992 Richards amp Barrett 1992 Faivre2000)
Morphometrical analysis of perianth features clearly groupstogether heterostylous species (N pallidulus and N albima-rginatus) despite they belong both to different taxonomicsections (Dorda amp Fernaacutendez-Casas 1989 Muller-Doblies ampMuller-Doblies 1989) and unrelated lineages (Fig 5) Thesespecies look indeed similar (Fig 4) Morphological resem-blance was already discussed by Arroyo amp Barrett (2000) whopointed out a role played by pollinators on driving flowersimilarity PCA and the multiple regression analyses resultedin significant perianth correlates of style polymorphism(Table 7) This relationship is unaffected by flower size becausea regression model accounting for allometric effects displayedsimilar results (Table 7) All associated traits are of importancefor pollinator activity from attraction (flower size) to insecthandling of flowers (flower tube and angle corona size andshape) Heterostyly in Narcissus is associated with pendulousflowers short and wide perianth tubes and large cup-shapedcoronas The two heterostylous species share additional traitssuch as similar fragrances at least to humans reflexed tepals(Fig 4) and flower number per inflorescence (1ndash3) being thelatter trait however very variable in Narcissus (Worley et al2000 see also Fernandes 1975 for a evolutionary hypo-thesis) The remaining Apodanthi species have a contrastingflower morphology albeit variable Interestingly two sets ofmorphological features parallel two stylar conditions styledimorphism (N calcicola N scaberulus N cuatrecasasii ) withpatent flowers variable perianth-tube length and cup-shapedcoronas and monomorphism (N watieri N marvieri ) witherect flowers narrow and long perianth tubes and funnel shapedcoronas Both morphological types present patent tepals Nar-cissus rupicola displays an intermediate situation in whichperianth features are similar to those of monomorphicspecies and but sex organ position to that of style dimorphicspecies
Large bees actively work in pendulous and patent flowersIn fact Anthophora and Bombus species have been reported aspollinators of the heterostylous species N triandrus and Npallidulus (Barrett et al 1997 J Arroyo personal observations)where they feed on both nectar and pollen whilst handlinglarge cup-shaped coronas Similarly within the variable genusIxia (Iridaceae) species of cup-shaped flowers are pollinatedby Anthophora species (Goldblatt et al 2000) Unfortunatelywe do not have reliable data on pollinators of N albimargin-atus Taking into account relative flower similarity betweenN pallidulus and N cuatrecasasii (Fig 4) with respect to Nalbimarginatus we hypothesize that Anthophora spp could actas pollinators in the tree species (Barrett et al 1997 the presentstudy) These solitary endothermic bees are able to fly evenunder relatively cold weather conditions (Batra 1994 Stone1994) and are able to face harsh winter conditions when Nalbimarginatus blooms It has been shown that bees arecomparatively more proficient for disassortative pollen transfera necessary condition for proper functioning of reciprocalherkogamy (Stone 1996) Anthophora bees have been foundto pollinate the heterostylous Jasminum fruticans of similarflower size shape and colour in Iberia (Guitiaacuten et al 1998)One of the most striking similarities between the two hetero-stylous species concerns pendulous flowers Flower angle isvariable in Narcissus (Blanchard 1990) even within a singlespecies This feature limits pollinator types handling flowersand in fact it has been shown to vary in populations withdifferent pollinators (Arroyo amp Dafni 1995)
We hypothesize that low reciprocal style dimorphic (Nrupicola) and monomorphic species of Narcissus which have thenarrowest and longest flower tubes are mostly pollinated bylong-tongued pollinators as it has been found in other similarspecies (Arroyo et al 2002 Thompson et al 2003) Despitescarcity of actual observations of insects visiting flowers thereis some evidence that they do occur In the population fromMadrid of N rupicola we found a 94 of fruit set in the sub-sequent fruiting season It is probably a result of night pollina-tion by moths The only white-flowered species N watierishows a variety of pollinators including butterflies It isremarkable the similarity of monomorphic species and the L-morph of N rupicola (Fig 2) The promotion of assortativemating between L plants by long-tongued lepidopterans(Stone 1996) may fix the L morph A similar process ofpollinator-mediated shift has been suggested to promote Lmonomorphism in populations of N tazetta (Arroyo amp Dafni1995) N papyraceus (Arroyo et al 2002) and perhaps Ndubius (Baker et al 2000a) which are white-flowered specieswith similar flower traits and array of pollinators
Evolution of flower polymorphisms in Narcissus
Phylogenetic analyses of 27 trnL-F sequences representingmuch of the diversity of Narcissus yielded low resolution(Fig 5) because of a limited number of variableparsimony-
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research250
informative characters (1711) Some conclusions are howeverinferred from analysis of phylogenetic relationships the threespecies of the N triandrus complex coupled with three morespecies of sect Cyclaminei Pseudonarcissi and Narcissus(clade 2) form a natural group a well-defined independentlineage includes five of the seven species of sect Apodanthi(clade 1) the remaining two species of sect Apodanthi (Nscaberulus N calcicola) and the representatives of sectionsTazetta Jonquilla Serotini and Dubii are unresolved at abasal-most position These results indicate that sect CyclamineiPseudonarcissi and Narcissus are closely related based on thischloroplast marker Circumscription of sect Apodanthiincluding N scaberulus and N calcicola together with Nwatieri N marvieri N rupicola N albimarginatus and Ncuatrecasasii needs further investigation
Phylogenies based on molecular markers have been rarelyused to reconstruct character evolution of flower featuresrelated to heterostyly (but see Graham amp Barrett 1995 andKohn et al 1996 for Pontederiaceae) Evolutionary pathwaysof reproductive features concerning gynodioecy in Lycium(Miller amp Venable 2003) and selfing in Collinsia (Armbrusteret al 2002) have been already investigated in detail by usingnuclear sequences In our study phylogenetic reconstructionsof chloroplast sequences suggest that heterostyly has takenplace at least twice in the course of the evolution of NarcissusTwo well-defined lineages including heterostylous species andhigh sequence divergence between them (average of 196 inpairwise distance estimates) lead us to interpret two independ-ent evolutionary histories in acquisition of distyly (N albima-rginatus) and tristyly (N triandrus complex) Althoughlimited resolution is depicted in the clade 2 of the trnL-F phy-logeny (Fig 5) low molecular divergence (lt 031) betweenspecies of the N triandrus complex (N lusitanicus N pallid-ulus N triandrus) indicates close relatedness The occurrenceof two independent lineages with heterostylous species sup-ports a prediction (Arroyo amp Barrett 2000) of independentacquisition of heterostyly in Narcissus The role of pollinatorsin acquisition of heterostyly remains an open question How-ever a significant flower similarity of both heterostylous spe-cies supports this hypothesis Multiple origins of heterostylywithin a single genus is not surprising because independentorigins have been already documented at the familial level(Barrett 1992) To our knowledge this is the first report ofconvergence of heterostyly at the generic level Unfortunatelybasal polytomies in the trnL-F phylogeny prevents from infer-ring whether ancestral condition to heterostyly is style dimor-phism as proposed by Lloyd amp Webb (1992ab)
The trnT-L phylogeny of section Apodanthi shows a claderesolution (Fig 6) in which an evolutionary pattern of stylepolymorphism can be inferred First split of an early lineageof N scaberulus-N calcicola endemic to Portugal and theremaining species of sect Apodanthi is suggested These twoPortuguese species have a style dimorphism condition incommon with some reciprocity in lower anthers and a similar
perianth Therefore stylar dimorphism appears to be anancestral condition within sect Apodanthi as predicted byLloyd amp Webbrsquos (1992ab) model of heterostyly Although theparsimony-based (cladistic) reconstruction gives limited reso-lution (Fig 6) to pinpoint a sister species to the heterostylousN albimarginatus a distance-based phylogeny (NJ) clearlyjoins N albimarginatus and N cuatrecasii result in congru-ence with similarity of perianth morphology and stylar con-dition between both species The remaning three Apodanthispecies form a well-defined lineage with a pectinate topologygiving strong support for an evolutionary pattern in whichacquisition of nonherkogamous monomorphism (N marvieriand N watieri ) may have occurred from an ancestor with astyle-dimorphic condition (as in N rupicola) The loss ofstylar polymorphism to monomorphism has been shown tooccur in other taxonomic groups (Kohn et al 1996) Withthe available phylogenetic information it is not possible todetermine shifts from style dimorphism to monomorphism inother sections in Narcissus However this may be tendency atleast in some cases Similarity between L-morph of N rupicolaand flowers of the two monomorphic species together with thefrequent loss of the S-morph in other dimorphic species ofsect Tazetta (Arroyo amp Dafni 1995 Arroyo et al 2002)support this view A similar pollinator array in N marvierito that in sect Tazetta suggests that this shift may be mediatedby insects Style polymorphism shift from style dimorphismto monomorphism and the role of pollinators should beexplored in other sections of Narcissus
Phylogenetic reconstructions coupled with patterns offlower similarity both in stylar conditions and perianth fea-tures across unrelated lineages support the role of pollinatorsin moulding style polymorphim in Narcissus Our results leadus to conclude that multiple convergence events have occurredin the course of evolution of Narcissus particularly to build upheterostyly (distyly and tristyly) and monomorphism throughintermediate insect-mediated stages as proposed by theLloyd ampWebbrsquos (1992a) model
Acknowledgements
The authors thanks Spencer Barrett John Thompson FernandoOjeda Adeline Cesaro Paco Rodriacuteguez and the colleagues inthe EVOCA seminar group for discussions and suggestionsduring the study Fernando Ojeda Spencer Barrett SeanGraham Joseacute M Gomez and two anonymous reviewers madevery useful comments on the manuscript Adeline Cesarohelped design Fig 2 Many colleagues helped with fieldsampling and flower measurements particularly ArndtHampe Begontildea Garrido Jesuacutes Lera Jara Cano and AixaRivero Joseacute L Medina and Laura Fernaacutendez helped in insectcensuses Mohammed Ater Abdelmalek Benabid MustafaLamrani and Javier Fernaacutendez Casas Abdelardo providedinformation on Moroccan populations or gave strong logisticsupport Aparicio kindly provided the photograph of
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
copy New Phytologist (2003) 161 235ndash252 wwwnewphytologistorg
Research 251
Narcissus watieri Colleagues of the Molecular Systematicslab at the Royal Botanic Garden of Madrid provided a friendlyatmosphere and helped with protocols and techniques for labwork This study is part of fulfilment of a PhD degree of RocioPeacuterez at the University of Seville Rociacuteo Peacuterez is funded by apredoctoral fellowship of the Spanish Ministry of EducationCulture and Sports This study was also supported by theSpanish Ministry of Science and Technology (PB98-1144REN2001 4738 E and BOS2003-07924-CO2-01)
References
Armbruster WS Mulder CPH Baldwin BG Kalisz S Wessa B Nute H 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae sl) American Journal of Botany 89 37ndash49
Arroyo J 2002 Narcissus la evolucioacuten de los polimorfismos florales y la conservacioacuten maacutes allaacute de las lsquolistas rojasrsquo Revista Chilena de Historia Natural 75 39ndash55
Arroyo J Barrett SCH 2000 Discovery of distyly in Narcissus (Amaryllidaceae) American Journal of Botany 87 748ndash751
Arroyo J Barrett SCH Hidalgo R Cole WW 2002 Evolutionary maintenance of stygma-height dimorphism in Narcissus papyracens (Amaryllidaceae) American Journal of Botany 89 1242ndash1249
Arroyo J Dafni A 1995 Variations in habitat season flower traits and pollinators in dimorphic Narcissus tazetta L (Amaryllidaceae) in Israel New Phytologist 129 135ndash146
Baker AM Thompson JD Barrett SCH 2000a Evolution and maintenance of stigma-height dimorphism in Narcissus I Floral variation and style-morph ratios Heredity 84 502ndash513
Baker AM Thompson JD Barrett SCH 2000b Evolution and maintenance of stigma-height dimorphism in Narcissus II Fitness comparisons between style morphs Heredity 84 514ndash524
Barrett SCH 1992 Evolution and function of heterostyly Berlin Germany Springer-Verlag
Barrett SCH Cole WW Arroyo J Cruzan MB Lloyd DG 1997 Sexual polymorphism in Narcissus triandrus (Amaryllidaceae) Is this species tristylous Heredity 78 135ndash145
Barrett SCH Jesson LK Baker AM 2000 The evolution and function of stylar polymorphisms in flowering plants Annals of Botany 85 253ndash265
Barrett SCH Lloyd DG Arroyo J 1996 Stylar polymorphisms and the evolution of heterostyly in Narcissus (Amaryllidaceae) In Lloyd DG Barrett SCH eds Floral biology studies on floral evolution in animal-pollinated plants New York USA Chapman amp Hall 339ndash376
Batra SWT 1994 Anthophora pilipes villosula SM (Hymenoptera Anthophoridae) a manageable japanese bee that visits bluberries and apples during cool rainy spring weather Proceedings of the Entomological Society of Washington 96 98ndash119
Blanchard JW 1990 Narcissus a guide to wild daffodils Surrey UK Alpine Garden Society
Charlesworth D Charlesworth B 1979 A model for the evolution of distyly American Naturalist 114 467ndash498
Darwin C 1877 The Different forms of flowers on plants of the same species London UK John Murray
Dorda E Fernaacutendez-Casas J 1989 Estudios morfoloacutegicos en el geacutenero Narcissus L Anatomiacutea de hoja y escapo III Fontqueria 27 103ndash162
Dulberger R 1964 Flower dimorphism and self-incompatibility in Narcissus tazetta L Evolution 18 361ndash363
Dulberger R 1992 Floral polymorphisms and their functional significance in the heterostylous syndrome In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 41ndash84
Faivre AE 2000 Ontogenetic differences in heterostylous plants and implications for development from herkogamous ancestor Evolution 54 847ndash858
Faivre AE McDade LA 2001 Population level variation in the expression of heterostyly in three species of Rubiaceae does reciprocal placement of anthers and stigmas characterize heterostyly American Journal of Botany 188 841ndash853
Fernandes A 1967 Contribution agrave la connaisance de la biosystematique de quelques espegravecies de genre Narcissus Portugaliae Acta Biologica (B) 9 1ndash44
Fernandes A 1975 LrsquoEvolution chez le genre Narcissus L Anales del Instituto Botaacutenico Cavanilles 32 843ndash872
Ganders FR 1979 The biology of heterostyly New Zealand Journal of Botany 17 607ndash635
Goldblatt P Bernhardt P Manning JC 2000 Adaptative radiation of pollination mechanisms in Ixia (Iridaceae Crocoideae) Annals of the Missouri Botanical Garden 87 564ndash577
Graham SW Barrett SCH 1995 Phylogenetic systematics of Pontederiales Implications for breeding-system evolution In Rudall PJ Cribb PJ Cutler DF Humphries CJ eds Monocotyledons systematics and evolution Kew UK Royal Botanic Gardens 415ndash451
Guitiaacuten J Guitiaacuten P Medrano M 1998 Floral biology of the distylous Mediterranean shrub Jasminum fruticans (Oleaceae) Nordic Journal of Botany 18 195ndash201
James FC McCulloch CE 1990 Multivariate-analysis in ecology and systematics ndash Panacea or Pandora Box Annual Review of Ecology and Systematics 21 129ndash166
Kelchner SA 2000 The evolution of non-coding chloroplast DNA and its application in plant systematics Annals of the Missouri Botanical Garden 87 482ndash498
Kimura M 1980 A simple method for estimating evolutionary rate of base substitution through comparative studies of nucleotide sequences Journal of Molecular Evolution 16 11ndash120
Kohn JR Barrett SCH 1992 Experimental studies on the functional significance of heterostyly Evolution 46 43ndash55
Kohn JR Graham SW Morton B Doyle JJ Barrett SCH 1996 Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation Evolution 50 1454ndash1469
Lau P Bosque C 2003 Pollen flow in the distylous Palicourea fendleri (Rubiaceae) an experimental test of the disassortative pollen flow hypothesis Oecologia 135 593ndash600
Lloyd DG Webb CJ 1992a The evolution of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 151ndash178
Lloyd DG Webb CJ 1992b The selection of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 179ndash207
Lloyd DG Webb CJ Dulberger R 1990 Heterostyly in species of Narcissus (Amaryllidaceae) Hugonia (Linaceae) and other disputed cases Plant Systematics and Evolution 172 215ndash227
Maddison WP Maddison DR 1992 MacClade Version 3 Analysis of Phylogeny and Character Evolution Sunderland MA USA Sinauer Associates Inc
Mathew B 2002 Clasification of the genus Narcissus In Hanks GR ed Narcissus and Daffodil London UK Taylor amp Francis 30ndash52
McCune B Mefford MJ 1999 PC-ORD Multivariate Analysis of Ecological Data Version 4 Gleneden Beach OR USA MjM Software Design
Meerow AW Fay MF Guy CL Li Q-B Zaman FQ Chase M 1999 Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data American Journal of Botany 86 1325ndash1345
Miller JS Venable DL 2003 Floral morphometrics and the evolution of sexual dimorphism in Lycium (Solanaceae) Evolution 57 74ndash86
Muller-Doblies D Muller-Doblies U 1989 Narcissus albimarginatus plate 1986 In The flowering plants of Africa Vol 50 Part 2 Cape Town South Africa Creda Press
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
wwwnewphytologistorg copy New Phytologist (2003) 161 235ndash252
Research252
Nishihiro J Washitani I Thomson JD Thomson BA 2000 Patterns and consequences of stigma height variation in a natural population of distylous plant Primula sieboldii Functional Ecology 14 502ndash512
Ornduff R 1992 Historical perspective on heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 31ndash39
Philippi TE 1993 Multiple regression Herbivory In Scheiner SM Gurevitch J eds Design and analysis of ecological experiments New York USA Chapman amp Hall 183ndash210
Richards AJ 1998 Lethal linkage and its role in the evolution of plant breeding systems In Owens SJ Rudall PJ eds Reproductive biology Kew UK Royal Botanic Gardens
Richards JH Barrett SCH 1992 The development of heterostyly In Barrett SCH ed Evolution and function of heterostyly Berlin Germany Springer-Verlag 85ndash124
Sage TL Strumas F Cole WW Barrett SCH 1999 Differential ovule development following self- and cross-pollination the basis of self-sterility in Narcissus triandrus (Amaryllidaceae) American Journal of Botany 86 855ndash870
Saitou N Nei M 1987 The Neighbor-Joining method a new method for reconstructing phylogentic trees Molecular Biology Evolution 4 406ndash425
Simmons MP Ochoterena H 2000 Gaps as character in sequence-based phylogenetic analysis Systematic Biology 49 369ndash381
Sokal RR Rohlf FJ 1995 Biometry 3rd ed New York NY USA FreemanStatSoft 1997 STATISTICA for Windows Version 51 Tulsa OK USA
StatSoft IncStone GN 1994 Patterns of evolution of warm-up rates and body
temperatures in flight in solitary bees of the genus Anthophora Functional Ecology 8 324ndash335
Stone JL 1996 Components of pollination effectiveness in Psychotria surrensis a tropical distylous shrub Oecologia 107 504ndash512
Stone JL Thomson JD 1994 The evolution of distyly pollen transfer in artificial flowers Evolution 48 1595ndash1606
Swofford DL 1999 PAUP Phylogenetic Analysis Using Parsimony Version 40 Champaign IL USA Illinois Natural History Survey
Taberlet P Gielly L Pautou G Bouvet J 1991 Universal primers for amplification of three non-coding regions of choroplast DNA Plant Molecular Biology 17 1105ndash1109
Thompson JD Barrett SCH Baker AM 2003 Frequency-dependent variation in reproductive success in Narcissus implications for the maintance of stigma-height dimorphism Proceedings of the Royal Society of London B 270 949ndash953
Worley AC Baker AM Thompson JD Barrett SCH 2000 Floral display in Narcissus variation in flower size and number at the species population and individual levels International Journal of Plant Science 161 69ndash79
About New Phytologist
bull New Phytologist is owned by a non-profit-making charitable trust dedicated to the promotion of plant science facilitating projectsfrom symposia to open access for our Tansley reviews Complete information is available at wwwnewphytologistorg
bull Regular papers Letters Research reviews Rapid reports and Methods papers are encouraged We are committed to rapidprocessing from online submission through to publication lsquoas-readyrsquo via OnlineEarly ndash average first decisions are just 5ndash6 weeksEssential colour costs are free and we provide 25 offprints as well as a PDF (ie an electronic version) for each article
bull For online summaries and ToC alerts go to the website and click on lsquoJournal onlinersquo You can take out a personal subscription tothe journal for a fraction of the institutional price Rates start at pound108 in Europe$193 in the USA amp Canada for the online edition(click on lsquoSubscribersquo at the website)
bull If you have any questions do get in touch with Central Office (newphytollancasteracuk tel +44 1524 592918) or for a localcontact in North America the USA Office (newphytolornlgov tel 865 576 5261)
