RESEARCH ARTICLE

Inter-island but not intra-island divergence among populationsof sea oats, Uniola paniculata L. (Poaceae)

Cara L. Gormally • J. L. Hamrick •

Lisa A. Donovan

Received: 29 May 2012 / Accepted: 18 December 2012 / Published online: 28 December 2012

� Springer Science+Business Media Dordrecht 2012

Abstract Understanding the underlying causes of phe-

notypic trait variation among populations is important for

informing conservation decisions. This knowledge can be

used to determine whether locality matters when sourcing

populations for habitat restoration. Uniola paniculata is a

federally protected coastal dune grass native to the south-

eastern Atlantic and the Gulf coasts of the USA that is

often used to stabilize restored dune habitats. This study

uses neutral genetic markers (allozymes) and a greenhouse

common garden study to determine the relative contribu-

tions of neutral evolutionary processes and natural selec-

tion to patterns of phenotypic variation among natural

populations of U. paniculata. Seeds were sourced from

foredune and backdune populations spanning shoreline-to-

landward environmental gradients on each of four Georgia

barrier islands. Based on previous work, we expected to

find evidence of divergent selection among populations

located on the shoreline-to-landward environmental gradi-

ent. However, differences among islands, rather than intra-

island habitat differences, drive divergent selection on

aboveground and total biomass. The lack of evidence for

divergent selection across the shoreline-to-landward gra-

dient suggests that previously documented intra-island trait

variation is likely due to phenotypic plasticity. Our findings

have implications for conservation and restoration efforts

involving U. paniculata, as there is evidence for divergent

selection among populations located on neighboring

islands.

Keywords FST � Allozymes � QST � Population

differentiation � Uniola paniculata � Coastal dunes

Introduction

Determining mechanisms that underlie trait variation among

populations is particularly critical for informing restoration

and conservation decisions, e.g., choosing plant material for

transplanting, and maintaining genetic diversity and adaptive

potential (Lynch 1996; Sgro et al. 2011). The source of

transplanted plant material may have a profound effect on the

success of transplant establishment, due to the adaptation of

populations to local environmental conditions (Rice et al.

1997; McKay et al. 2005 and see a meta-analysis by Hereford

(2009) for an in-depth discussion of local adaptation and fit-

ness). However, multiple evolutionary and ecological pro-

cesses may produce trait variation, including local adaption

due to divergent selection among heterogeneous environ-

mental conditions (Turesson 1922; Clausen et al. 1948; Her-

eford 2009), phenotypic plasticity (Via and Lande 1985;

Schlichting and Pigliucci 1998), historical genetic structure

(Batista et al. 2004), genetic drift, and barriers to gene flow

(Briggs and Walters 2001). Consequently, identifying the

principal cause driving trait variation is challenging, since

multiple processes may produce similar outcomes.

A standard way of unraveling these multiple hypotheses

is to compare patterns of quantitative trait variation with

neutral genetic variation (Spitze 1993; Merila and Crno-

krak 2001; Whitlock 2008; Edelaar and Bjorklund 2011).

The genetic basis of variation in quantitative traits can be

C. L. Gormally � J. L. Hamrick � L. A. Donovan

Department of Plant Biology, University of Georgia,

2502 Miller Plant Sciences, 30602 Athens, Georgia

C. L. Gormally (&)

School of Biology, Georgia Institute of Technology,

310 Ferst Drive, 30332 Atlanta, Georgia

e-mail: cara.gormally@biology.gatech.edu

123

Conserv Genet (2013) 14:185–193

DOI 10.1007/s10592-012-0441-z

assessed using common garden experiments under condi-

tions that minimize environmentally-induced trait varia-

tion. Neutral genetic variation can be measured with

genetic markers (Merila and Crnokrak 2001) and, Wright’s

FST can be used to examine the distribution of neutral

genetic variation among populations (Wright 1965). A

similar measure for quantitative traits, QST, can be directly

compared to FST (Spitze 1993). Variation among popula-

tions for neutral markers (FST) results from drift, mutation,

and gene flow, but not selection. In contrast, variation

among populations for quantitative traits (QST) is poten-

tially due to all evolutionary factors, including selection

(Workman and Niswander 1970; McKay and Latta 2002).

There are three possible outcomes resulting from the

comparison of the two analyses (Merila and Crnokrak

2001):

(1) If QST is not significantly different from FST, trait

differentiation is due to genetic drift and gene flow;

(2) If QST is significantly greater than FST, selection on

the trait is divergent across the sampled range of

populations;

(3) If QST is significantly less than FST, selection is

uniform across the sampled range of populations.

Coastal dunes are spatially and temporally heterogeneous

habitats. Along a localized environmental gradient spanning

the dynamic embryonic foredunes to the more stabilized

dunes situated further inland, environmental conditions can

differ at distances as small as 0.5 m. This may drive changes

in plant community composition and species’ distributions

(Cowles 1899; Oosting and Billings 1942; Boyce 1954;

Wagner 1964; Doing 1985; Barbour et al. 1999; Maun and

Perumal 1999; Smith and Smith 2001). Additionally, dunes

are often detrimentally impacted by human activities,

including dune destruction due to construction, changes in

on- and off-shore sand movement due to hard protection

structures (e.g. jetties, seawalls, groins), changes in plant

species distributions resulting from human trampling, and

changes to dune form and plant community composition as a

result of planting exotics, ostensibly to stabilize dunes

(Wiedemann and Pickart 1996; Acosta et al. 2000; Davis

and Fitzgerald 2004). Since coastal dunes are dynamic and

frequently disturbed, understanding the underlying genetic

responses to heterogeneous environmental factors may

inform management strategies for the long-term persistence

of these populations (Lynch 1996; Franks 2009).

We estimated neutral genetic variation using allozymes

and examined trait variation in Uniola paniculata L. (Poa-

ceae), a dominant perennial coastal dune grass on primary

dunes. Uniola paniculata is legally protected because it sta-

bilizes dune habitats and is frequently used in dune restoration

projects. Its natural history as a rhizomatous perennial grass

capable of both sexual and clonal reproduction makes it

representative of many dune plant species (Duncan and

Duncan 1987). Uniola paniculata ranges from Virginia to the

Bahamas along the southeastern Atlantic coast and along the

Gulf coast to Veracruz, Mexico; across one dune system, its

habitat spans a localized environmental gradient from fored-

unes to dunes situated farther inland (Wagner 1964). At

geographic scales, there are only weak regional patterns of

genetic structure among populations of U. paniculata (Franks

et al. 2004; Subudhi et al. 2005).

We designed this study expecting that divergent selec-

tion pressures along the shoreline-to-landward environ-

mental gradient would be responsible for documented

patterns of phenotypic variation among populations that

corresponded with the underlying habitat variation along

the dune gradient (Gormally and Donovan 2010). In a

reciprocal transplant experiment, we found no support of

local adaptation to dune microhabitats, but that test for

local adaptation was compromised by the loss of foredune

plots to storms prior to harvest (Gormally and Donovan

2011). Other research findings suggested that localized

microhabitat differences might drive trait variation among

U. paniculata populations, as fine-scale clonal structure

and diversity have been shown to vary widely even within

relatively small areas (Franks et al. 2004; Bush and Stelato

2007). Here, comparative studies of quantitative genetic

trait variation and neutral allozyme variation, considered in

light of these previous results, are used to determine

whether phenotypic trait variation among U. paniculata

populations is due to phenotypic plasticity, divergent

selection, or genetic drift.

Materials and methods

Seed collection and study sites

Our study focused on four barrier islands on the Georgia

coast: Cabretta, Nannygoat Beach on Sapelo Island, Sea

Island, and Jekyll Island (Table 1). Cabretta, the northern-

most island, is located approximately 0.3 km north of

Sapelo Island, which is located about 18 km north of Sea

Island. Sea Island is located about 2 km north of Jekyll

Island. Sapelo is the largest of the four islands at nearly

6,900 ha, spanning approximately 17 km in length and

4 km at its widest point. Cabretta and Sea Island were

formed 4,000–5,000 years ago during the Holocene with

the rise of sea level following the last glacial maximum.

The Sapelo and Jekyll beaches sampled were also formed

during the Holocene, although other parts of these islands

were formed during the Pleistocene (35,000–40,000 years

ago, before the last glacial maximum).

In November 2006, we collected panicles from two

populations, dynamic foredunes and stabilized backdunes on

186 Conserv Genet (2013) 14:185–193

123

each of the four islands. Foredune populations were defined

as plants growing in the 10 m closest to the shoreline and

backdune populations were defined as the farthest inland

edge of the species’ range, based on previous characteriza-

tion of this species’ habitat (Gormally and Donovan 2010,

2011). In each population, panicles were collected from 20

individuals located at least 3 m apart to increase the likeli-

hood that seeds from unique genetic individuals were sam-

pled following the protocol from Franks et al. (2004). Seeds

were stored at room temperature prior to use.

Common garden experiment

From our field collection, we planned a randomized com-

plete block design. Since many panicles contained few or no

seeds, several populations represented in the study contain

fewer families and individuals within families (Table 1). In

January 2008, 1,624 seeds were removed from panicles,

scarified by nicking the seed coat with a razor blade, and

germinated in Petri dishes in a growth chamber (95 �F for

7 h/65 �F for 17 h), following the protocol described by

Seneca (1972). Germination rates ranged from 38 to 57 %

by population, with 47 % of all seeds germinating. Low seed

production and low germination rates are typical for this

species (Wagner 1964; Seneca 1972). Seedlings were

transplanted into 2.5 9 10 inches (40 inch3 capacity) Dee-

pots (Steuwe and Sons) and grown under ambient conditions

in the University of Georgia Plant Biology greenhouses.

We measured 10 quantitative traits. Since U. paniculata is

a grass, we were limited in our choice of quantitative traits.

We chose traits expected to be ecologically important to the

species, and which might affect plant success depending on

variation in localized abiotic conditions, e.g., sand burial.

These traits, describing plant morphology and phenology,

were measured either during plant growth or at harvest in

November 2008. Plant traits measured during growth inclu-

ded time to germination (days), seedling height (cm), which

was measured in March 2008, and adult stem diameter (mm),

which was measured in May 2008. At harvest in November

2008, we measured average root diameter (mm) of the three

largest roots, number of ramets, number of rhizomes, and

number of roots. We then dried the harvested plants to

constant weight, and measured aboveground biomass (g),

belowground biomass (g), and total biomass (g). Because U.

paniculata reproduces both sexually and clonally, we mea-

sured number of ramets, since reproductive strategy might

differ by habitat. We measured biomass partitioning (above-

and below-ground biomass; number of ramets and number of

rhizomes) because dune species tend to respond to sand

burial by increasing size through shifts in resource allocation

(e.g., shifting resources from roots to aboveground biomass)

(Zhang and Maun 1992; Martinez and Moreno-Casasola

1996; Brown 1997; Perumal and Maun 2006).

Allozyme gel electrophoresis

A total of 117 seedlings were crushed for protein extraction

using liquid nitrogen, sea sand, and the extraction buffer of

Wendel and Parks (1982): 53 from backdune populations

and 64 from foredune populations. Sample sizes were

limited as many panicles contained few or no seeds and

germination rates tended to be low. Extracted proteins were

run on 9.5 % starch gels and fourteen enzymes were

examined: aspartate amino transferase (Aat), diaphorase

(Dia), glutamate dehydrogenase (Gdh), malic enzyme

(Me), and triose-phosphate isomerase (Tpi) on buffer sys-

tem 8-; menadione reductase (Mnr), phosphoglucoisomer-

ase (Pgi), and phosphoglucomutase (Pgm) on buffer system

6; 6-phosphogluconate dehydrogenase (Pgd), F-1,6-

diphosphate (F1,6), and malate dehydrogenase (Mdh) on

buffer system 4; and adenylate kinase (Ak), isocitrate

dehydrogenase (Idh), and shikimate dehydrogenase (Skdh)

on buffer system 11 (Soltis et al. (1983) for all buffers and

stains except: Aat, Ak, Dia, and Mnr (Cheliak and Pitel

1984). Following the established protocols cited above,

gels were stained with enzyme-specific stains and scored

for allozyme banding patterns, with interpretation guided

by Wendel and Weeden (1989) and Kephart (1990).

Statistical analyses

For each quantitative trait, variance components were used

to estimate QST following Spitze (1993), i.e., QST = rb2/

(2rw2 ? rb

2) The population variance component was used

Table 1 Populations sampled are presented with their GPS coordinates and number of individuals sampled, including the number of half-sibling

families sampled in parentheses

Island Population abbreviation Latitude Longitude N (families)

Backdune Foredune

Cabretta Island Cab B/Cab F 31�250 81�140 189 (17) 194 (20)

Nannygoat Beach, Sapelo Island Sap B/Sap F 31�230 81�150 200 (18) 201 (16)

Sea Island Sea B/Sea F 31�100 81�200 209 (16) 203 (17)

Jekyll Island Jek B/Jek F 31�040 81�240 197 (18) 231 (18)

Conserv Genet (2013) 14:185–193 187

123

as the estimate of among population variance (rb2) and four

times the family within-population variance component

was used as the estimate of within-population variance

(rw2 ) to account for the half-sibling (i.e., open pollinated)

design (Lynch and Walsh 1998). For diploids, genetic

variation among populations is proportional to two times

FST, thus the ‘‘2’’ in the denominator. For each quantitative

trait, variance was partitioned into among population,

among families within populations, and error components

(PROC MIXED, SAS 9.2, Cary, NC). For each QST value,

its standard error was estimated as the standard deviation of

QST estimates calculated from 1,000 iterations boot-

strapped over all samples. The model structure described

above was used to test three comparisons: (1) comparison

of individual populations, accounting for both island and

habitat type (n = 8 populations); (2) comparison of

foredune and backdune habitats, with populations pooled

across islands (n = 2 populations); and (3) comparison of

islands, with populations pooled across habitats within

each island (n = 4 populations). We present the results of

the latter two comparisons to explicate our primary find-

ings from the comparison of individual populations, with

the caveat that the latter two models may have resulted in

artificially inflated or deflated estimates, depending upon

the manner in which populations were pooled.

For each quantitative trait, narrow-sense heritabilities (hn2)

were estimated for each population using family and residual

variance components from a separate random model

following Falconer and Mackay (1996), i.e., hn2 = 4rhs

2 /

(rhs2 ? re

2) (PROC MIXED, SAS 9.2, Cary, NC). To account

for the half-sibling design, hn2 was calculated as four times the

within family variance component divided by the sum of the

family and residual variance components (Falconer and

Mackay 1996).

FST was calculated following Wright (1965) with the

program GDA (Weir 1996; Lewis and Zaykin 2001) and

bootstrapped confidence intervals were estimated over

1,000 iterations, to account for sampling error, as recom-

mended by Whitlock (2008). Values of FST were calculated

to test three comparisons: (1) comparison of individual

populations, accounting for both island and habitat type

(n = 8 populations); (2) comparison among foredune and

backdune habitats, with populations pooled across islands

(n = 2 populations); and (3) comparison of islands, with

populations pooled across habitats within each island

(n = 4 populations).

The following genetic diversity parameters were esti-

mated for each locus and for the three comparisons of

populations described above, as well as pooled across all

populations, using the program GDA (Lewis and Zaykin

2001): percentage of polymorphic loci (P), mean number

of alleles per locus (A), mean number of alleles per poly-

morphic locus (Ap), and observed (Ho) and expected

heterozygosity (He). Subscript p and s are used to designate

population-level and pooled effects, respectively.

Whitlock (2008) suggests that in addition to the standard

comparison of the average FST value to QST values for

particular traits, QST values may be compared to the fre-

quency distributions of FST across all loci to infer if popu-

lation differentiation in quantitative variation differs from

neutral expectations (and see Eroukhmanoff et al. 2009). We

could not utilize this approach because we had fewer than

the recommended number of loci (N [ 50) (Whitlock

2008). However, the distribution of FST is fairly well pre-

dicted by the mean value of FST when values are less than

0.1 (Whitlock 2008). Our calculated FST value fell below

0.1, indicating that using the mean FST to compare to spe-

cific QST values is reasonable for evaluating causes of trait

differentiation. We compared the mean FST to specific QST

values, determining significant differences by non-overlap-

ping bootstrapped error bars.

Since the sample sizes of the Cabretta populations

included in our allozyme study were quite small, we cal-

culated FST and genetic diversity parameters both with and

without the CAB populations. The inclusion of the CAB

populations produced little difference in calculated FST

values (0.035 excluding CAB populations versus 0.032

including CAB populations) so we chose to include these

populations.

Results

These eight populations maintained relatively high levels of

allozyme variation. Sixteen of the 23 allozyme loci analyzed

were polymorphic in multiple populations (Table 2a–c). The

most variable loci were Pgm-1 and Tpi-3, with 4 and 3

alleles each, respectively. For the 16 polymorphic loci, a

total of 35 alleles were found among the individuals sam-

pled. Seven loci (Aat, Ak-1, F16, Mdh-1, Pgi-2, Pgi-3, and

Skdh) were monomorphic in all populations. Across pooled

populations, there was an average of 2.23 alleles per poly-

morphic locus (APs) and genetic diversity (Hes) was esti-

mated to be 0.158 (Table 2a). Within populations there was

an average of 2.06 alleles per polymorphic locus (APp) and

mean genetic diversity (Hep) was estimated to be 0.147

(Table 2a).

There were also moderate levels of additive genetic

variation within these populations. Trait heritabilities ran-

ged from 0.114 for number of ramets, to 0.393 for time to

germination (Table 3). The estimate of heritability for

number of ramets in the Sapelo backdune population was

excluded because it was greater than 1.0, which is theo-

retically impossible. By utilizing results from Franks et al.

(2004) to inform our sampling strategy for seed collection,

we had hoped to avoid collecting seeds from ramets

188 Conserv Genet (2013) 14:185–193

123

belonging to the same clone. However, if clonality occur-

red on a larger scale than expected, inadvertent sampling

within genets may have occurred. The estimate of herita-

bility for number of ramets in the Sapelo backdune popu-

lation was less than 1 when estimated using a full-sibling

design, indicating perhaps that significant genetic related-

ness among maternal plants may have existed.

The values of QST varied greatly among the 10 traits that

were measured. Estimates of QST varied from 0.005 for

time to germination to 0.116 for aboveground biomass at

harvest, with a mean QST of 0.055 (Fig. 1a; Table 3). FST

was estimated to be 0.032 among all eight populations. FST

values for differences among islands and between habitats

within islands were 0.026 and 0.005, respectively. The QST

values for all traits, except two (above ground biomass and

total biomass), fell within the confidence intervals of the

estimated mean value of FST for the comparison among all

populations (Fig. 1a). For the habitat comparison (foredune

versus backdune), only the QST for mean root diameter was

significantly higher than the mean FST (Fig. 1b). For the

among island comparisons, the number of roots, above-

ground biomass, and total biomass are significantly dif-

ferent from the mean FST (Fig. 1c). Consequently, results

from the comparisons of islands and habitats indicate that

selection among populations likely differs due to differ-

ences among islands, rather than due to variation in abiotic

conditions along the localized environmental gradient

(Fig. 1).

Discussion

Many comparisons of quantitative genetic variation and

neutral marker variation report QST values that far exceed FST

(Merila and Crnokrak 2001; Leinonen et al. 2008). This is

interpreted as divergent selection resulting in different opti-

mal phenotypes for populations in different habitats. We

expected to find evidence for divergent selection to be stron-

gest among habitats within islands, i.e. across the shoreline-to-

landward environmental gradient, based on previous results

for U. paniculata (Franks et al. 2004; Bush and Stelato 2007;

Gormally and Donovan 2010; Gormally and Donovan 2011)

as well as given the general expectation of adaptive differ-

entiation across environmental gradients (McKay et al. 2005).

However, our analyses indicated that differentiation was

greater among populations located on different islands

(number of roots, aboveground biomass, and total biomass;

Fig. 1c) than among populations located across the shoreline-

to-landward environmental gradient on the primary dunes

(mean root diameter; Fig. 1b). Finally, in our comparison of

all eight populations, we found evidence of divergent selec-

tion for only 2 of the 10 traits measured: aboveground biomass

and total biomass, traits that are most likely correlated.

For the majority of traits, QST values were not signifi-

cantly different from FST, indicating similar structuring of

quantitative trait variation and neutral genetic variation

among populations. For these traits, there is no evidence to

suggest that divergent selection plays a stronger role in

shaping differences among populations than genetic drift.

Our findings of similar values of FST and QST for many traits

and low amounts of among-population genetic structure are

consistent with results for other plant species with geo-

graphically or ecologically narrow distributions (Waldmann

and Andersson 1998; Widen et al. 2002; Jorgensen et al.

2006; Badri et al. 2008; Helsen et al. 2009). Species with life

history traits similar to U. paniculata have GST values

similar to our FST estimate, e.g., long-lived outcrossing

perennials (0.094) and long-lived perennials with wind-dis-

persed seeds (0.086) (Hamrick and Godt 1996). Our FST

estimate is lower than allozyme-based estimates of GST for

Poaceae (0.284) (Hamrick and Godt 1996) and for other

outcrossing grasses (0.112) (Godt and Hamrick 1998), as

Table 2 Genetic diversity parameters derived from 23 allozyme loci

in U. paniculata presented for each model comparison as described in

the ‘‘Materials and methods’’ (a) island and habitat comparison,

(b) habitat comparison, (c) island comparison

Population N Pp (%) Ap APp Hep Hop

(a)Island and habitat comparison

CabB 2 30.4 1.30 2.00 0.188 0.261

CabF 8 47.8 1.48 2.00 0.143 0.174

JekB 13 56.5 1.61 2.08 0.179 0.197

JekF 22 56.5 1.61 2.08 0.169 0.184

SapB 20 47.8 1.52 2.09 0.152 0.196

SapF 10 39.1 1.44 2.11 0.132 0.204

SeaB 18 43.5 1.44 2.00 0.163 0.225

SeaF 24 39.1 1.44 2.11 0.137 0.199

Mean 14.63 45.1 1.48 2.06 0.147 0.205

Pooled 117 56.5 1.83 2.23 0.158 0.199

(b) Habitat comparison

Backdune 53 56.5 1.74 2.154 0.163 0.208

Foredune 64 56.5 1.74 2.231 0.153 0.192

Mean 117 56.5 1.74 2.192 0.158 0.200

(c) Island comparison

Cab 10 47.8 1.52 2.091 0.147 0.191

Jek 35 65.2 1.74 2.133 0.175 0.189

Sap 29 47.8 1.52 2.091 0.144 0.198

Sea 43 47.8 1.57 2.182 0.149 0.210

Mean 117 52.2 1.59 2.124 0.154 0.197

Sample size (N), percentage of polymorphic loci (Pp), number of

alleles per locus (Ap), mean number of alleles per polymorphic locus

(APp), expected heterozygosity (Hep) and observed heterozygosity

(Hop) are presented

Conserv Genet (2013) 14:185–193 189

123

well as estimates for U. paniculata reported by Franks et al.

(2004). However, Franks et al. (2004) sampled populations

across the species’ geographic range, as did an AFLP-based

survey which reported higher levels of genetic diversity

(Subudhi et al. 2005). Our FST estimate is probably lower, in

part, due to the limited geographic area that was sampled

(i.e., higher potential rates of gene flow). Our genetic

diversity parameter measurements, including Pp, and Hop,

are similar to those reported by Franks et al. (2004). How-

ever, Hop was always greater than Hep in our surveyed

populations, while Franks et al. (2004) reported equal mean

values of Hop and Hep, though measures of Hop exceeded Hep

in some populations (including the Sapelo Island popula-

tion). Observations of excess heterozygosity in populations

of perennial plants are relatively commonplace (Mitton

1997) and there is also some indication that heterozygotes

are favored in heterogeneous environments (Mitton 1997).

Our data, however, are not sufficient to demonstrate whether

this explanation is applicable to the observed excesses of

heterozygosity in the populations sampled in this study. The

lack of evidence for intra-island divergent selection (fored-

une vs. backdune populations) for nine of the ten traits

measured suggests that previously observed phenotypic

variation across the shoreline-to-landward environmental

dune gradient likely results from phenotypic plasticity

(Gormally and Donovan 2010). Because U. paniculata is a

relatively long-lived perennial clonal grass, an individual

genotype may experience and respond to the entire spectrum

of environmental gradient conditions as dunes build and

erode over its lifetime. The ability to respond plastically to

temporal environmental heterogeneity likely has been a

favorable strategy contributing to the long-term persistence

of these populations (Schlichting and Pigliucci 1998).

In addition to plasticity, sufficient genetic variation

affects the ability of populations to evolve in response to

environmental change (Frankham et al. 2002; Jump et al.

2008; Kramer and Havens 2009). This may be particularly

important for U. paniculata because it is expected to

experience further fragmentation and reduced population

size due to man-made disturbances. Our measures of her-

itability indicate that these populations have the genetic

capacity to respond to changes in environmental condi-

tions. It should be noted that our heritability estimates may

include non-additive genetic components such as environ-

mental maternal effects, since plants used in this study

were grown from field-collected seeds. Additionally, the

Fig. 1 Point estimates and bootstrapped 95 % confidence intervals for

quantitative traits measured and estimated FST. The values represent

the proportion of total heritable variance that is partitioned among

populations. The solid line indicates the estimated value of FST. The

dashed lines indicate the upper and lower bounds of the confidence

interval around the estimated value of FST. a–c Values estimated for

each model as described in the ‘‘Materials and methods’’

b

190 Conserv Genet (2013) 14:185–193

123

paternity of our field-collected seeds is unknown, so it is

possible that some families contain a mixture of half and

full siblings, resulting in both heritability and QST values

that do not reflect expected genetic relatedness.

In recent years, questions have been raised about the

QST–FST comparison (Miller et al. 2008; Pujol et al. 2008;

Santure and Wang 2009). Assumptions of FST and QST

under neutrality have been called into question (Miller et al.

2008), and inbreeding and dominance have been shown, in

theory, to affect the comparison (Santure and Wang 2009).

Our study utilized neutral markers, allozymes, rather than

hypervariable genetic markers such as microsatellites, so as

to not impact the validity of the comparison of FST and QST

(Edelaar and Bjorklund 2011; Edelaar et al. 2011). Further,

we utilized a common garden in our research design, so that

the comparison did not confound additive genetic diver-

gence among populations with environmental effects (Pujol

et al. 2008). In particular, questions have been raised con-

cerning the ability to not only compare mean estimates of

each parameter, but to compare the parameters’ distributions

themselves, particularly for FST values C0.1 (Whitlock

2008). While the number of loci sampled (N = 23) pre-

vented us from using this approach, our FST value is B0.1

(FST = 0.032). We also used 95 % confidence intervals

which are more conservative tests of significance than dif-

ference testing at a = 0.05, since there are methodological

difficulties with the construction of confidence intervals on

variance component ratios for QST which are still under

development and debate (Bonnin et al. 1996; O’Hara and

Merila 2005; Whitlock 2008). Though there are valid con-

cerns about QST–FST comparisons, it remains a valuable

technique for making evolutionary inferences concerning

which traits are undergoing divergent selection.

The spatial scale of genetic differentiation among popula-

tions can inform restoration decisions by suggesting whether

sourcing transplants locally is important for restoration success

(Frankham et al. 2002; McKay et al. 2005). Genotype-envi-

ronment matching, resulting from local adaptation or historical

genetic structure, can profoundly influence the success of some

restoration efforts (Rice et al. 1997). However, recent work

highlights several benefits of genetic translocation, including

alleviating threats of population extinction and inbreeding

resulting from small population size, spurring the probability

of adaptation in the context of climate change (Sgro et al.

2011), and introducing individuals that are not locally sourced

but are genetically diverse to increase long-term adaptive

potential for restored populations (Broadhurst et al. 2008). Our

findings, combined with results from prior studies, suggest that

that U. paniculata individuals are successful across the range

of environmental conditions occurring across a single dune

system (Gormally and Donovan 2010; Gormally and Donovan

2011). This is promising, since coastal dunes undergoing res-

toration often experience substantial habitat loss of the shore-

line-to-landward environmental gradient due to beach erosion

or subsidence (Feagin et al. 2005).

Since natural selection appears to be divergent amongst

closely neighboring islands, conservation efforts should use

care in seed or ramet sourcing for populations located on

different dune systems. While the populations analyzed were

all from Holocene beaches with similar geomorphologies,

there may be considerable inter-island variation in dune

topography and the frequency and extent of beach and dune

erosion. These differences may result from island orientation

relative to prevailing winds, affecting overwash patterns and

sand accretion and erosion, as well as local beach and dune

sediment budgets (Stallins and Parker 2003). Differences in

quantitative traits, particularly the ones documented here—

aboveground biomass and total biomass—may impact the

successful translocation of individuals from one environ-

ment to another (Frankham et al. 2002). To address inter-

island genotype translocation, an important next step

would be to conduct inter-island reciprocal transplants of

Table 3 Traits measured

Trait N hn2 rb

2 rw2 QST U L

Time to germination 761 0.393 (0.410) 0.03752 4.144 0.004507 0.014 -0.005

Seedling height 668 0.214 (0.452) 0.18 0.9104 0.089964 0.117 0.063

Adult stem diameter 643 0.150 (0.431) 0.000054 0.0044 0.006039 0.016 -0.004

Number of ramets 644 0.114 (0.445) 0.1164 0.9592 0.057205 0.077 0.037

Number of rhizomes 644 0.249 (0.535) 0.000822 0.0326 0.012462 0.026 -0.001

Number of roots 644 0.359 (0.535) 4.4136 25.5244 0.079578 0.108 0.051

Root diameter 642 0.153 (0.493) 0.000435 0.00586 0.035788 0.050 0.021

Aboveground biomass 643 0.200 (0.541) 1.1652 4.4484 0.115802 0.138 0.093

Belowground biomass 643 0.166 (0.566) 0.04568 0.4304 0.050393 0.069 0.032

Total biomass 643 0.200 (0.541) 1.4893 6.5832 0.101619 0.122 0.082

Sample size (N), mean narrow-sense heritabilities (hn2) and standard errors in parentheses, among-(rb

2) and within-(rw2 ) population variances, and

QST values, with upper (U) land lower (L) limits of the bootstrap 95 % confidence intervals for the QST statistics are presented

Conserv Genet (2013) 14:185–193 191

123

populations to understand whether divergent selection

among populations drives local adaptation, as well as to

determine the environmental conditions associated with

selection favoring certain phenotypes. Together, compara-

tive studies of neutral genetic variation and quantitative trait

variation help to reveal the complex interactions occurring

within and among populations and provide genetic infor-

mation that can be used to guide plant restoration efforts.

Acknowledgments The authors wish to thank Scott Gevaert for

assistance with seed collections and critical feedback for early drafts

of this manuscript, Cecile Deen for assistance with allozymes, Beau

Brouillette for assistance with statistical analyses, and the University

of Georgia Plant Biology Greenhouse staff for assistance in germi-

nating and keeping the plants alive. Research at the Sapelo Island

National Estuarine Research Reserve (SINERR) was facilitated by the

research reserve coordinator, Dorset Hurley. Tom Patrick at the

Georgia Department of Natural Resources Wildlife Resources Center

provided invaluable help in the permitting process for seed collection.

We thank the National Estuarine Research Reserve System (National

Oceanic and Atmospheric Administration) for a graduate fellowship

that supported this work (NAO7NOS42-00039), as well as the

Georgia Sea Grant College Program (NA04OAR4170033) and the

Georgia Botanical Society for financial support. This is contribution

number 993 from the University of Georgia Marine Institute.

References

Acosta A, Blasi C, Stanisci A (2000) Spatial connectivity and

boundary processes in coastal dune vegetation in the Circeo

National Park, Central Italy. J Veg Sci 11:149–154

Badri M, Zitoun A, Soula S, Ilahi H, Huguet T, Aouani ME (2008) Low

levels of quantitative and molecular genetic differentiation among

natural populations of Medicago ciliaris Kroch. (Fabaceae)

of different Tunisian eco-graphical origin. Conserv Genet 9:

1509–1520

Barbour MG, Burk JH et al (1999) Terrestrial plant ecology.

Benjamin Cummings, Menlo Park

Batista F, Bouza N, Gonzalez-Perez MA, Caujape-Castells J, Sosa PA

(2004) Genetic variation within and between populations of two

endangered endemic species of the laurel forest from the Canary

Islands, Myrica rivas-martinezii (Myricaceae) and Sideritisdiscolor (Lamiaceae). Aust J Bot 52:471–480

Bonnin I, Prosperi JM, Olivieri I (1996) Genetic markers and

quantitative genetic variation in Medicago truncatula (Legumi-

nosae): a comparative analysis of population structure. Genetics

143:1795–1805

Boyce SG (1954) The salt spray community. Ecol Monogr 24(1):

29–67

Briggs D, Walters SM (2001) Plant variation and evolution, 3rd edn.

Cambridge University Press, New York

Broadhurst LM, Lowe A, Coates DJ, Cunningham SA, McDonald M,

Vesk PA, Yates C (2008) Seed supply for broadscale restoration:

maximizing evolutionary potential. Evol Appl 1:587–597

Brown JF (1997) Effects of experimental burial on survival, growth,

and resource allocation of three species of dune plants. J Ecol

85:151–158

Bush SP, Stelato ME (2007) Clonal diversity in differently aged

patches of the dune grass Uniola paniculata. Southeast Nat 6(2):

359–364

Cheliak WH, Pitel JA (1984) Techniques for starch gel electropho-

resis of enzymes from forest tree species. Petawawa National

Forestry Institute, Ontario

Clausen JD, Keck D, Heisey WM (1948) Experimental studies on the

nature of species. III. Environmental responses of climatic races

of Achillea. Carnegie Institution of Washington Publication 520

Cowles HC (1899) The ecological relations of the vegetation of the

sand dunes of Lake Michigan. Bot Gaz 27:95–391

Davis RA Jr, Fitzgerald DM (2004) Beaches and coasts. Blackwell,

Malden

Doing H (1985) Coastal fore-dune zonation and succession in various

parts of the world. Vegetatio 61(1):65–75

Duncan WH, Duncan MB (1987) The Smithsonian guide to seaside

plants of the Gulf and Atlantic coasts. Smithsonian Press,

Washington

Edelaar P, Bjorklund M (2011) If FST does not measure neutral

genetic differentiation, then comparing it with QST is misleading.

Or is it? Mol Ecol 20(9):1805–1812

Edelaar P, Burraco P, Gomez-Mestre I (2011) Comparisons between

QST and FST: how wrong have we been? Mol Ecol 20(23):

4830–4839

Eroukhmanoff E, Hargeby A, Svensson EI (2009) Rapid adaptive

divergence between ecotypes of an aquatic isopod inferred from

FST–QST analysis. Mol Ecol 18:4912–4923

Falconer DS, Mackay TFC (1996) Introduction to quantitative

genetics. Prentice Hall, New York

Feagin RA, Sherman DJ, Grant WE (2005) Coastal erosion, global

sea-level rise, and the loss of sand dune plant habitats. Front Ecol

Environ 3(7):359–364

Frankham R, Ballou JD, Briscoe DA (2002) Introduction to conser-

vation genetics. Cambridge University Press, New York

Franks SJ (2009) Genetics, evolution, and conservation of island

plants. J Plant Biol 53:1–9

Franks SJ, Richards CL, Gonzales E, Cousins JE, Hamrick JL (2004)

Multi-scale genetic analysis of Uniola paniculata (Poaceae): a

coastal species with a linear, fragmented distribution. Am J Bot

91:1345–1351

Godt MJW, Hamrick JL (1998) Allozyme diversity in the grasses. In:

Cheplick GP (ed) Population biology of the grasses. Cambridge

University Press, Cambridge, pp 11–29

Gormally CL, Donovan LA (2010) Responses of Uniola paniculataL. (Poaceae), an essential dune-building grass, to complex

changing environmental gradients on the coastal dunes. Estuar

Coasts 33(5):1237

Gormally CL, Donovan LA (2011) No evidence of local adaptation in

Uniola paniculata. Southeast Nat 10(4):751–760

Hamrick JL, Godt MJW (1996) Effects of life history traits on genetic

diversity in plant species. Philos Trans Royal Soc Lond B 351:

1291–1298

Helsen P, Verdyck P, Tye A, Van Dongen S (2009) Low levels of

genetic differentiation between Opuntia echios varieties on

Santa Cruz (Galapagos). Plant Syst Evol 279:1–10

Hereford J (2009) A quantitative survey of local adaptation and

fitness trade-offs. Am Nat 173(5):579–588

Jorgensen TH, Richardson DS, Andersson S (2006) Comparative

analyses of population structure in two subspecies of Nigelladegenii: evidence for diversifying selection on pollen-color

dimorphisms. Evolution 60(3):518–528

Jump AS, Marchant R, Penuelas J (2008) Environmental change and the

option value of genetic diversity. Trends Plant Sci 14(1):51–58

Kephart SR (1990) Starch gel electrophoresis of plant isozymes: a

comparative analysis of techniques. Am J Bot 77(5):693–712

Kramer AT, Havens K (2009) Plant conservation genetics in a

changing world. Trends Plant Sci 14(11):599–607

192 Conserv Genet (2013) 14:185–193

123

Leinonen T, O’Hara RB, Cano JM, Merila J (2008) Comparative

studies of quantitative trait and neutral marker divergence: a

meta-analysis. J Evol Biol 21:1–17

Lewis PO, Zaykin D (2001) Genetic data analysis: computer program

for the analysis of allelic data. Version 1.0 (d16c). Free program

distributed by the authors over the internet from http://lewis.eeb.

uconn.edu/lewishome/software.html

Lynch M (1996) A quantitative-genetic perspective on conservation

issues. In: Avise JC, Hamrick JL (eds) Conservation genetics:

case histories from nature. Chapman and Hall, New York,

pp 471–501

Lynch M, Walsh B (1998) Genetics and analysis of quantitative traits,

2nd edn. Sinauer Associates Inc., Sunderland

Martinez ML, Moreno-Casasola P (1996) Effects of burial by sand on

seedling growth and survival in six tropical sand dune species

from the Gulf of Mexico. J Coastal Res 12(2):406–419

Maun MA, Perumal J (1999) Zonation of vegetation on lacustrine

coastal dunes: effects of burial by sand. Ecol Lett 2:14–18

McKay JK, Latta RG (2002) Adaptive population divergence: markers,

QTL and traits. Trends Ecol Evol 17:285–291

McKay JK, Christian CE, Harrison S, Rice KJ (2005) How local is

local? A review of practical and conceptual issues in the genetics

of restoration. Restor Ecol 13:432–440

Merila J, Crnokrak P (2001) Comparison of genetic differentiation at

marker loci and quantitative traits. J Evol Biol 14:892–903

Miller JR, Wood BP, Hamilton MB (2008) FST and QST under

neutrality. Genetics 180:1023–1037

Mitton JB (1997) Selection in natural populations. Oxford University

Press, New York

O’Hara RB, Merila J (2005) Bias and precision in QST estimates:

problems and some solutions. Genetics 171:1331–1339

Oosting HJ, Billings WD (1942) Factors affecting vegetational

zonation on coastal dunes. Ecology 23:131–142

Perumal VJ, Maun MA (2006) Ecophysiological response of dune

species to experimental burial under field and controlled

conditions. Plant Ecol 184:89–104

Pujol B, Wilson AJ, Ross RIC, Pannell JR (2008) Are QST–FST

comparisons for natural populations meaningful? Mol Ecol 17:

4782–4785

Rice SLW, Montalvo AM, Buchmann SL, Cory C, Handel SN,

Nabhan GP, Robichaux RH (1997) Restoration biology: a

population biology perspective. Restor Ecol 5:277–290

Santure AW, Wang J (2009) The joint effects of selection and

dominance on the QST–FST contrast. Genetics 181:259–276

Schlichting CD, Pigliucci M (1998) Phenotypic evolution: a reaction

norm perspective. Sinauer Associates Inc., Sunderland

Seneca ED (1972) Germination and seedling response of Atlantic and

Gulf coast populations of Uniola paniculata. Am J Bot 59(3):

290–296

Sgro CM, Lowe AJ, Hoffmann AA (2011) Building evolutionary

resilience for conserving biodiversity under climate change. Evol

Appl 4:326–337

Smith RL, Smith TM (eds) (2001) Ecology and field biology.

Benjamin Cummings, New York

Soltis DE, Haufler CH, Darrow DC, Gastony GJ (1983) Starch gel

electrophoresis of ferns: a compilation of grinding buffer, gel,

and electrode buffers, and staining schedules. Am Fern J 73:9–27

Spitze K (1993) Population structure in Daphnia obtusa: quantitative

genetic and allozymic variation. Genetics 135:367–374

Stallins JA, Parker AJ (2003) The influence of complex systems

interactions on barrier island dune vegetation pattern and process.

Ann Assoc Am Geogr 93(1):13–29

Subudhi PK, Parami NP, Harrison SA, Materne MD, Murphy JP, Nash

D (2005) An AFLP-based survey of genetic diversity among

accessions of sea oats (Uniola paniculata, Poaceae) from the

southeastern Atlantic and Gulf coast states of the United States.

Theor Appl Genet 111:1632–1641

Turesson G (1922) The genotypical response of the plant species to

habitat. Hereditas 3:211–350

Via S, Lande R (1985) Genotype-environment interaction and the

evolution of phenotypic plasticity. Evolution 39(3):505–522

Wagner RH (1964) The ecology of Uniola paniculata L. in the dune-

strand habitat of North Carolina. Ecol Monogr 34:79–96

Waldmann P, Andersson S (1998) Comparison of quantitative genetic

variation and allozyme diversity within and between populations

of Scabiosa canescens and S. columbaria. Heredity 81:79–86

Weir BS (1996) Genetic data analysis II. Sinauer Associates Inc.,

Sunderland

Wendel JF, Parks CR (1982) Genetic control of isozyme variation in

Camellia japonica L. J Hered 73(3):197–204

Wendel JF, Weeden NF (1989) Visualization and interpretation of

plant isozymes. In: Soltis DE (ed) Isozymes in plant biology.

Dioscorides Press, Portland

Whitlock MC (2008) Evolutionary inference from QST. Mol Ecol

17:1885–1896

Widen BS, Andersson G, Rao Y, Widen M (2002) Population

divergence of genetic (co)variance matrices in a subdivided plant

species, Brassica cretica. J Evol Biol 15:961–970

Wiedemann AM, Pickart A (1996) The Ammophila problem on the

Northwest Coast of North America. Landsc Urban Plan 34:287–299

Workman PL, Niswander JD (1970) Population studies on south-

western Indian tribes. II. Local genetic differentiation in the

Papago. Am J Hum Genet 22:24–49

Wright S (1965) The interpretation of population structure by

F-statistics with special regard to the systems of mating.

Evolution 19:395–420

Zhang J, Maun MA (1992) Effects of burial in sand on the growth and

reproduction of Cakile edentula. Ecography 15(3):296–302

Conserv Genet (2013) 14:185–193 193

123