Present patterns of genetic variation within species havebeen influenced by many factors. Of major importance isthe influence of the last period of glaciation and subse-quent migration of taxa from glacial refugia. Ice agesoccur at regular intervals of 100 000 years with warminterglacial periods lasting 1520 000 years as a result ofinstabilities in the earths climate caused by theMilankovitch cycles (Bennett 1990). Many tree speciescommon in northern Europe today survived these glacialperiods in small, low-density populations in refugia in themountains of southern Europe (Bennett et al. 1991).
The slower rate of evolution of chloroplast DNA(cpDNA) compared to nuclear DNA in plants (Wolfe et al.1987) has limited its use in population studies at theintraspecific level (Palmer 1987). Several papers have beenpublished describing varying levels of intraspecificcpDNA variation in a wide range of plant species(reviewed in Soltis et al. 1992). cpDNA is maternally inher-ited in the majority of flowering plants and therefore pro-vides a seed-specific marker. In species where seed flow ismuch less than pollen flow, it is predicted that organelle
genes (chloroplast and mitochondrial) will be highly struc-tured when compared to nuclear genes (Petit et al. 1993).
Recent studies on European and North American treespecies have shown that refugial areas and postglacialmigration routes can be identified using DNA markers.Fossil pollen maps for European deciduous oaks indicaterefugia in southern Spain, southern Italy and the BalkanPeninsula (Huntley & Birks 1983; Bennett et al. 1991). Twoseparate studies based on cpDNA variation have con-firmed the existence of these three refugia (Dumolin-Lapgue et al. 1997; Ferris et al. 1998) and also identifyareas of northern Europe colonized by oaks from eachrefugium. Similar studies have also been carried out onFagus sylvatica in Europe (Demesure et al. 1996) andLiriodendron tulipifera in North America (Sewell et al. 1996).
Alnus glutinosa (black alder) is a wind-pollinated, self-incompatible tree species of riparian and water-loggedhabitats (McVean 1953). It is common in Europe and theMediterranean, and extends as far as the mountains ofTurkey and North Africa. Seed dispersal is most effectiveby water, but seeds may also be dispersed by wind up to30 m from the parent tree (McVean 1953). Huntley & Birks(1983) provide evidence for at least three glacial refugiafor Alnus based on fossil pollen data, including Corsica,the Carpathian Mountains and southwestern Russia, and
Molecular Ecology (1998) 7, 11511161
1998 Blackwell Science Ltd
Chloroplast DNA phylogeography of Alnus glutinosa(L.) Gaertn.
R. ANDREW KING and COLIN FERRISDepartment of Biology, University of Leicester, Leicester, LE1 7RH, UK
Traditionally, information on the postglacial history of plant species has been gainedfrom the analysis of fossil pollen data. More recently, surveys of present patterns ofgenetic variation have given valuable insights into species phylogeography. The genusAlnus, based on fossil data, is known to have had at least four glacial refugia. A survey ofchloroplast DNA (cpDNA) diversity in populations of black alder (A. glutinosa) wasundertaken in order to gain more insight into its postglacial history. This revealed a highdegree of structuring of 13 cpDNA haplotypes on a European scale which indicated thatmost of northern and central Europe was colonized from a refuge in the CarpathianMountains. Based on the distribution of two common cpDNA haplotypes, colonizationroutes from this refuge can be determined. The locations of other previously identifiedrefugia are confirmed and two formerly unconfirmed refugial areas for alder (southernSpain and Turkey) are proposed.
Keywords: Betulaceae, black alder, PCRRFLP, polymorphism, postglacial history, refugia
Received 5 December 1997; revision received 17 February 1998; accepted 2 March 1998
Correspondence: R. A. King. Fax: +44 (0) 1162522791; E-mail:firstname.lastname@example.org
the Bay of Biscay region. To this may be added refugia insouthern Italy and Greece (Bennett et al. 1991). Alderpollen has also been found in late glacial deposits fromsouthwest Turkey (Van Zeist et al. 1975) and northern Iran(Van Zeist & Bottema 1977).
A previous study of isozyme variation in A. glutinosa(Prat et al. 1992) demonstrated strong differentiationbetween populations that was attributed to both ecologi-cal and historical events affecting population evolution.This contrasts with the results of Bousquet et al. (1990)who found very little population differentiation in theNorth American A. sinuata and A. crispa.
The extent to which the different putative refugia havecontributed to the present European distribution of alderis unknown and thus it was decided to study the chloro-plast DNA phylogeography of A. glutinosa using aPCRRFLP (polymerase chain reactionrestriction frag-ment length polymorphism) approach. Using theseresults, it should be possible to identify glacial refugia andrelate observed patterns of cpDNA variation to possiblepostglacial migration routes.
Materials and methods
Two sources of Alnus glutinosa were used in the study. Seedmaterial was obtained from the International Alder SeedBank at Geraardsbergen, Belgium. Seeds were germinatedon damp compost and harvested for DNA extraction when46-weeks old. Alternatively, fresh leaf material was col-lected in the field and either snap-frozen in liquid nitrogenor dried over silica gel (Chase & Hills 1991). Where possi-ble, a minimum of three nonadjacent trees per populationwere sampled. The total sample consisted of 217 individualtrees from 101 populations covering the entire naturalrange of the species within Europe. Details of site locationand sample size per site are given in Appendix 1.
DNA was extracted from frozen or dried material usingthe CTAB method of Doyle & Doyle (1990) with the addi-tion of fine sand to aid grinding of the leaf material andsubstituting PVPP (polyvinyl polypyrrolidone) in placeof PVP. For seedling material, a miniprep modification ofthe CTAB method was used. Leaf material (0.1 g) wasground with fine sand in liquid nitrogen, added to 500 m Lof 2 CTAB in a 1.5 mL centrifuge tube and incubated at60 C for 3060 min. A volume of 500 m L of 24:1 chloro-form:iso-amyl alcohol was added, the tubes were mixedand spun at 13 000 rpm for 10 min in a bench-top cen-trifuge. The top layer was then pipetted into a clean tubeto which 250 m L of isopropanol was added. The tubes
were rocked gently to aid precipitation of the DNA. TheDNA was washed in an ethanol wash buffer (76%ethanol, 10 mM ammonium acetate), air dried at roomtemperature for a few minutes and then dissolved in50100 m L of TE (Tris-EDTA buffer, pH 8).
cpDNA was amplified using the universal primers ofTaberlet et al. (1991) and Demesure et al. (1995) (Table 1).Reactions were carried out in a total volume of 25 m L con-sisting of 17.55 m L of double-distilled water, 2.5 m L of 10PCR buffer (Bioline), 1.25 m L of dNTPs (2 mM), 1 m L ofMgCl2 (50 mM), 0.5 m L of each of the forward and reverseprimers (10 m M), 1 unit of BIOTAQ DNA polymerase(Bioline) and 1.5 m L of genomic DNA. PCR amplificationswere performed in a DNA thermal cycler (Perkin ElmerCetus). An initial 5 min denaturation at 94 C was fol-lowed by 30 cycles of 94 C for 30 s, annealing at 5462 Cfor 30 s and extension at 72 C for from 90 s to 3 mins.Annealing temperature was dependent upon primersused and extension time was dependent on the length ofthe PCR product (Table 1). Reactions were given a final10 min extension time at 72 C.
PCR product (10 m L) was restricted overnight with eitherone or two restriction enzymes following the methods ofFerris et al. (1993, 1995). Initially, each of the nine PCRproducts from six individuals of A. glutinosa weredigested with six 4-bp cutting and two 6-bp cuttingrestriction enzymes (AluI, CfoI, EcoRI, HaeIII, HindIII,HinfI, MboI and RsaI; Gibco BRL). Restriction digests wererun on either 6% polyacrylamide or 1.6% agarose gels andvisualized by staining with ethidium bromide(0.5 m g/mL). Polymorphisms were scored visually andnumbered in order of decreasing molecular weight.Bands were scored as presence (1) vs. absence (0) of theband. In order to determine the nature of each mutation,amplicons were digested with several restriction enzymecombinations. Size variation was assumed when similarpatterns were observed with different enzymes.
Each polymorphism was scored as an unordered multi-state character and subjected to phylogenetic analysisusing the heuristic search option of PAUP version 3.1.(Swofford 1993). The number of mutational differencesbetween haplotypes was calculated and analysed usingM I N S P N E T (Excoffier & Smouse 1994) to produce aminimum-spanning tree of haplotypes found. This proce-dure is used to connect points, in this case haplotypes, by
1152 R. A KING AND C. FERRIS
1998 Blackwell Science Ltd, Molecular Ecology, 7, 11511162
direct links having the smallest possible total length (Prim1957). Minimum-spanning networks are alternatives toWagner parsimony trees, but better convey the connec-tions between haplotypes (Excoffier & Smouse 1994).
The level of population subdivision for a cytoplasmi-cally inherited genome using unordered alleles (GSTc) wascalculated following the method of Pons & Petit (1995)using the computer program H A P L O I D I V. NSTc, the levelof population subdivision for ordered alleles was calcu-lated using the program H A P L O N S T (Pons & Petit 1996).
A ratio of seed to pollen flow was calculated using theequation:
(1/GSTb 1) 2 (1/GSTc 1)(pollen flow/seed flow =
were GSTb is the level of population subdivision based onbiparentally (nuclear) inherited genomes. For this study,GSTb is taken from Prat et al. (1992) This is a modificationof the equation of Ennos (1994) with the substitution ofGST values for FST values.
An initial screen of nine pairs of universal chloroplastprimers with eight restriction enzymes revealed variationin six Alnus glutinosa fragments (Table 1). Only thoseprimerenzyme combinations that gave easily scorablevariation were used in the full survey. Four of these
primers pairs proved sufficient to identify all haplotypesfound. Numerous mutations were detected (Table 2,Fig. 1). In all, a total of 13 cpDNA haplotypes were foundand these are described in Table 3.
The geographical distribution of these haplotypes ishighly structured (Fig. 2). Southeast Europe is a major areafor cpDNA variation with a total of seven haplotypes beingfound in an area covering Bulgaria, Greece, Turkey,Georgia and the Ukraine. Southern Europe possesses a fur-ther three haplotypes (one each in southern Italy, Corsicaand Spain). Most of central and northern European popula-tions are comprised of one or the other of two commonhaplotypes, with a third rare haplotype restricted to a sin-gle population in central Norway. The area aroundHungary and northern Croatia is polymorphic.
Due to the low number of variable and phylogeneti-cally informative mutations and the presence of homo-plasy in the data set a reliable phylogeny of haplotypescould not be found using parsimony analysis. Using PAUP(Swofford 1993), a total of 638 trees of length 17(C.I. = 0.842) were found. Haplotype relatedness was rep-resented using the M I N S P N E T (Excoffier & Smouse 1994)program (Fig. 3) which clearly indicates four groupings.The seven southeast European haplotypes (Fig. 2) are splitinto two groups, one containing haplotypes A, B, C and D,the other L, M and N. The Corsican (K) and Spanish (J)haplotypes form their own grouping, as do the remainingEuropean haplotypes (E, F, G and H).
PHYLOGEOGRAPHY OF BLACK ALDER 1153
1998 Blackwell Science Ltd, Molecular Ecology, 7, 11511162
Table 1 Details of primers used in this study
Annealing Extension In thisPrimers Code temperature time Variable study Reference
trnH [tRNA-His (GUG)] - A 62 C 2 min Yes Yes Demesure et al. (1995)trnK1 [tRNA-Lys (UUU) 3 exon]
trnC [tRNA-Cys (GCA)] - B 58 C 3 min Yes Yes Demesure et al. (1995)trnD [tRNA-Asp (GUC)]
trnD [tRNA-Asp (GUC)] - C 54 C 2 min Yes No Demesure et al. (1995)trnT [tRNA-Thr (GGU)]
psbC [psII 44 kd protein] - D 57 C 2 min No No Demesure et al. (1995)trnS [tRNA-Ser (GGA)]
trnS [tRNA-Ser (UGA)] - E 62 C 2 min Yes Yes Demesure et al. (1995)trnfM [tRNA-fMet (CAU)]
trnM [tRNA-Met (CAU)] - G 59 C 3 min No No Demesure et al. (1995)rbcL [RuBisCo large subunit]
trnK1 [tRNA-Lys (UUU) 3 exon] - K 53 C 3 min Yes Yes Demesure et al. (1995)trnK2 [tRNA-Lys (UUU) 5 exon]
trnS [tRNA-Ser (GGA)] - S 57 C 2 min No No Demesure et al. (1995)trnT [tRNA-Thr (UGU)]
trnT [tRNA-Thr (UGU)] - T 55 C 90 s Yes No Taberlet et al. (1991)trnL2 [tRNA-Leu (UAA) 3 exon]
A total of 217 individual trees from 101 populations wasanalysed. Only populations where three or more individ-uals could be obtained were used in the H A P L O I D I V andH A P L O N S T analyses. As single plants represented popu-lations from Istanbul, Kvam and southern Italy their rep-resentative haplotypes were omitted from the analysis. Ofthe 43 populations used, seven were polymorphic. Thelevel of population subdivision within A. glutinosa washigh, GSTc = 0.866 (hs = 0.103; ht = 0.773). For the NSTc anal-ysis, a distance matrix derived from the pairwise numberof mutational differences between haplotypes was used.Again, the level of population subdivision was high,NSTc = 0.905 (vs = 0.190: vt = 2.00).
Combining GSTc with the value of subdivision for nuclearmarkers (GSTb = 0.204; Prat et al. 1992) gives effective geneflow via pollen 23 times greater than that via seed.
A general knowledge of the postglacial history of alder isknown from fossil pollen analysis. Restricted to south-eastern Europe at 13 000 BP, Alnus glutinosa migrated
northward and westward from this refugial area, reach-ing western Europe by 10 000 BP (Huntley & Birks 1983),and arrived in Fennoscandia around 8500 BP (Tallantire1974). Colonization of Britain took place about 8000 BP(Birks 1989). No data are available on the postglacial his-tory of alder in Spain or Turkey. We undertook this pre-sent study to compare molecular phylogenetic data withwhat is known from fossil analysis and to gain furtherinsights into the postglacial history of black alder. Bycombining both approaches we get a very good picture ofthe glacial/postglacial history of species.
The geographical distribution of the 13 alder haplotypesis highly structured (Fig. 2). From the present study, the dis-tribution of cpDNA haplotypes confirms that most ofnorthern Europe was colonized from a refuge in the area ofthe Carpathian Mountains (present-day Hungary andRomania). The initial westward migration identified byHuntley & Birks (1983) apparently involved haplotype G,while the migration into northern Europe andFennoscandia comprised haplotype F. The presence of hap-lotype F in c...