Delayed genetic effects of habitat fragmentation on the ecologically

advertisement
Conserv Genet (2009) 10:1281–1297
DOI 10.1007/s10592-008-9707-x
RESEARCH ARTICLE
Delayed genetic effects of habitat fragmentation on the
ecologically specialized Florida sand skink (Plestiodon reynoldsi)
Jonathan Q. Richmond Æ Duncan T. Reid Æ
Kyle G. Ashton Æ Kelly R. Zamudio
Received: 21 February 2008 / Accepted: 7 September 2008 / Published online: 24 September 2008
Ó Springer Science+Business Media B.V. 2008
Abstract Populations rarely show immediate genetic
responses to habitat fragmentation, even in taxa that possess
suites of traits known to increase their vulnerability to
extinction. Thus conservation geneticists must consider the
time scale over which contemporary evolutionary processes
operate to accurately portray the effects of habitat isolation.
Here, we examine the genetic impacts of fragmentation on
the Florida sand skink Plestiodon reynoldsi, a sand swimming lizard that is highly adapted to the upland scrub
habitat of central Florida. We studied fragments located on
the southern Lake Wales Ridge, where human activity in the
latter half of the 20th century has modified the natural
patchiness of the landscape. Based on a relaxed molecular
clock method, we estimate that sand skinks have persisted
in this region for approximately 1.5 million years and that
the time frame of human disturbance is equivalent to fewer
than 30 skink generations. Using genotypes from eight
microsatellite loci, we screened for molecular signatures of
this disturbance by assessing congruence between population structure, as inferred from spatially-informed Bayesian
assignment tests, and the current geography of scrub fragments. We also tested for potential intrapopulation genetic
effects of inbreeding in isolated populations by comparing
J. Q. Richmond (&) D. T. Reid K. R. Zamudio
Department of Ecology and Evolutionary Biology,
Cornell University, Ithaca, NY 14850, USA
e-mail: jqr2@cornell.edu
D. T. Reid
National Institutes of Health, 50 South Drive, MSC 8004,
Bethesda, MD 20892-8004, USA
K. G. Ashton
Archbold Biological Station, P.O. Box 2057, Lake Placid,
FL 33862, USA
the average pairwise relatedness of individuals within
fragments of different areas and isolation. Our results
indicate that although some patches show a higher degree of
relatedness than expected under random mating, the genetic
effects of recent isolation are not evident in this part of the
species’ range. We argue that this result is an artefact of a
time-lag in the response to disturbance, and that speciestypical demographic features may explain the genetic
inertia observed in these populations.
Keywords Habitat fragmentation Skink Ecological specialization Genetic inertia Plestiodon reynoldsi
Introduction
Studies of anthropogenic habitat loss and fragmentation
have contributed substantial information on patterns of
species vulnerability (Fahrig 2003; reviewed in Ewers and
Didham 2006). However, because most anthropogenic
fragmentation is recent on evolutionary time scales (Watson
2002; Stockwell et al. 2003; Watson 2003; Sumner et al.
2004), investigations of its effects can be biased by timelags in species responses to disturbance (Tilman et al. 1994;
Cowlishaw 1999). These delays occur as populations
establish new genetic and community-level equilibria, and
not all species respond at the same rate. Instead, the process
is driven by a variety of spatial demographic conditions that
influence the new equilibrium, including effective population size, the degree of reduction in habitat connectivity,
generation time, and the strength of natural selection (Endler
1986; Wiens 1997; Brooks et al. 1999). Thus the evolutionary fate of recently isolated populations may not be
apparent unless investigators carefully consider the time
123
1282
frames over which genetic and community-level responses
are most likely to occur for a given species.
As a general guideline, genetic variability decreases at a
constant rate of 1/(2Ne) per generation in a closed population (where Ne is the effective population size, Wright
1978). Thus smaller, isolated populations lose genetic variability more quickly than larger and more continuously
distributed populations (Young and Clarke 2000). While
this relationship provides a rough estimate of response
times, other trait-mediated differences can significantly
influence the rate, trajectory and consistency of evolution.
For example, species with a high degree of ecological
specialization and restricted dispersal are particularly vulnerable to disturbance (Simberloff 1986; McKinney 1997;
Davies et al. 2004) and interactions between these traits and
spatial landscape features may further accelerate the rate at
which species respond. Alternatively, species that live in
naturally patchy environments with small population sizes
may experience greater inertia in their response times than
species that require large contiguous habitat. These and
other factors indicate that simple deterministic equations or
generalized principles are of limited use for predicting how
organisms will respond to disturbance (Stockwell et al.
2003; Caro 2007). Instead, our understanding of the process
is perhaps best pursued through case-specific studies at
different temporal and spatial scales.
In this paper, we investigate the effects of recent
anthropogenic fragmentation on an ecological specialist
distributed within a network of habitat patches. The sand
skink Plestiodon reynoldsi (formerly in the genus Neoseps:
Brandley et al. 2005) is a small lizard endemic to upland
scrub and sandhill habitats of central Florida that is federally listed as a threatened species by the United States
Fish and Wildlife Service (52 FR 42658-42662). The
species is a burrower that spends most time below ground;
individuals have countersunk jaws, a wedge-shaped snout,
reduced eyes, and vestigial limbs that facilitate sand
swimming 1–8 cm beneath the ground surface (Telford
1959; Christman 1992). This eel-like form of locomotion
requires well-drained, loose substrates with large particle
size (Collazos 1998; Gianopulos 2001; Christman 2005);
thus, scrub patches with altered soil densities such as those
in matrix habitats or at fragment edges can create lowpermeability barriers to dispersal (Stamps et al. 1987).
Sand skinks may also rely on successional vegetative
changes caused by natural disturbance (e.g. feeding habits
of Gopher tortoises and periodic fires, Christman 1992) that
maintain interconnected patches of bare, sandy microhabitats (Webb 1990; Myers 1991). These characteristics
suggest a potentially heightened sensitivity to anthropogenic fragmentation, and therefore we predicted that
declines in genetic diversity would be evident over relatively short time scales following habitat disturbance.
123
Conserv Genet (2009) 10:1281–1297
Plestiodon reynoldsi inhabits three peninsular ridges in
central Florida: the Lake Wales, Mount Dora, and the
Winter Haven Ridges (Telford 1959; Christman 1992).
These narrow ridges (\16 km wide in most areas) are
vestiges of ancient shorelines that cover approximately
240 km of central Florida (Webb 1990) and host one of the
highest proportions of endemic fauna in North America
(Myers 1990; Wood 1996). Topography, fluctuating sea
levels, and natural fires have historically maintained the
natural patchiness of the landscape (Webb 1990); however,
fragments have become gradually more isolated in the
latter half of the 20th century due to rapid urban and
agricultural development, fire suppression, and vegetative
overgrowth (Myers 1991; McCoy and Mushinsky 1992;
Greenberg et al. 1994; Hokit et al. 1999; Branch and Hokit
2000; Hall et al. 2002; Meshaka and Layne 2002). Over
85% of the original upland habitat has disappeared since
the 1940s (McCoy et al. 1999; Weekley et al. 2007). The
human population in Florida is expected to double from
approximately 18–36 million people over the next
50 years, thus an additional 7 million acres of undeveloped
land will likely be converted to urban use (Zwick and Carr
2006). Therefore, habitat loss, degradation, and fragmentation will continue to pose problems for the endemic taxa
within this already threatened landscape.
In addition to increased mortality caused by habitat
modification, P. reynoldsi may also suffer indirect losses in
genetic variability resulting from shrinking habitat area and
changes in connectivity of scrub patches within ridges. If
fragmentation causes a significant reduction in local Ne and
migration rates (m), increased genetic drift in isolated
populations will redistribute genetic diversity such that
lower levels of variability occur within versus among
populations (Wright 1931). This can lead to the accumulation of deleterious alleles and ultimately decreased
population fitness (Lande 1995). However, one important
question for the conservation of this species remains openended: At what geographic and temporal scales are the
adverse genetic effects of habitat fragmentation detectable
in extant populations?
We report on a fine-scale genetic survey of P. reynoldsi
populations in the southernmost part of the range to test
whether genetic partitioning is detectable among recently
isolated, local populations. We combined information from
mitochondrial DNA sequences, microsatellite markers, and
spatial landscape features to (1) estimate the age of haplotype lineages within the three ridges, (2) calculate
standard indices of genetic diversity and identify natural
genetic demes in our focal populations, and (3) test whether habitat area, degree of isolation, or a combination
thereof are significant predictors of genetic diversity within
populations. Our results show that although sand skinks
possess ‘high-vulnerability’ traits, fragmentation within the
Conserv Genet (2009) 10:1281–1297
past 60 years has had limited impact on genetic diversity of
populations in this part of the species’ range. We argue that
this result may be biased by genetic inertia arising from
species-typical demographic features and that insufficient
time has lapsed to reveal the full genetic consequences of
this disturbance.
Methods
Population sampling and microsatellite genotyping
Sand skink samples were collected from 11 fragments in
the southernmost tracts of scrub and sandhill habitat in the
Lake Wales Ridge, Highlands County, Florida (Fig. 1).
Scrubs in this region include undisturbed mosaics as well
as formerly connected remnants that are now highly isolated (Lohrer 1993). The fragments vary in degree of
isolation, area of suitable soils, and the quality and type of
intervening matrix (Fig. 1). Skinks were captured in pit-fall
trap arrays (40–56 traps/site) between May 2002 and
September 2002. Latitude and longitude coordinates of
each trap were recorded using a global positioning system
(Appendix 1). Small tail clips were taken from each individual prior to release on site, and tissues were stored in
100% ethanol.
We extracted whole genomic DNA using standard lysis
buffer and proteinase K digestion followed by phenolchloroform purification protocols (Sambrook and Russell
2001). Eight microsatellite loci were amplified for 179
individuals (10–32 individuals/site; Table 2) following
previously published protocols (Reid et al. 2004). Amplified products of non-overlapping size ranges with different
50 -fluorescent labels were multiplexed and electrophoresed
on a 5% polyacrylamide gel on an ABI 377 DNA
sequencer (PE Biosystems, Foster City, CA). Fragment
sizes were determined with a ROX-500 standard using
GENESCAN version 3.1 and GENOTYPER version 2.1 (PE
Biosystems, Foster City, CA).
Relaxed molecular clock estimates of divergence times
The effects of fragmentation depend in part on the historical equilibrium of populations prior to the onset of
disturbance and the number of generations that populations
have been isolated due to reduced habitat connectivity
(Ovaskainen and Hanski 2004). We re-analyzed cytochrome b sequences from a previous phylogeographic
analysis of Plestiodon reynoldsi (Branch et al. 2003) to
verify that our sampled populations were members of the
same evolutionary lineage and to estimate a divergence
date for the clade to which our populations belong. To
place the time scale for human disturbance within this
1283
deeper phylogenetic history of P. reynoldsi populations,
we used a Bayesian relaxed clock method (Drummond
et al. 2006) as implemented in BEAST 1.4.7 (Drummond
and Rambaut 2007) to estimate the time to most recent
common ancestry for sand skink lineages occurring on the
three major ridges of Central Florida, and the age of
divergence from its sister taxon the mole skinks (Plestiodon egregius species complex: Brandley et al. 2005). Our
analyses included 53 haplotypes for P. reynoldsi and 15
haplotypes from different members of the P. egregius
complex (GenBank accession #’s; AF470632–46; [Branch
et al. 2003]).
The relaxed clock model was calibrated using the
cytochrome b substitution rates of two small-bodied lizards
(Tarentola delalandii: 8.25 9 10-3 s/s/my, [Gubitz et al.
2000]; Anolis occulatus: 7.15 9 10-3 s/s/my, [Malhotra
and Thorpe 2000]), given that fossil data were unavailable
for P. reynoldsi. Uncertainty was incorporated into these
calibrations by allowing each branch in the phylogeny to
have independent rates drawn from a lognormal prior distribution (mean = 8.25 9 10-3 or 7.15 9 10-3 in separate
analyses). We implemented the following priors for substitution and site rate heterogeneity models: GTR ? C
substitution model (based on the Akaike Information Criterion in Mr. Modeltest [Nylander 2004]); substitution
rates = Jeffery’s prior; a shape parameter = exponential
(1.0), four rate categories; default values for all scale
operators. BEAST 1.4.7 also requires one to specify a tree
prior for modelling changes in population size through
time; thus, we selected the Bayesian Skyline plot because
of its minimal assumptions about demographic history, and
compared these results with analyses that assumed constant
population size through time. We considered congruent
inferences (i.e. overlap in the 95% highest posterior density
for age estimates) under the different mutation rates and
coalescence priors as an indication of robust divergence
estimates.
We performed identical, duplicate analyses for 1 9 107
steps, and sampled the Markov chain every 1000th generation. We used TRACER v1.4 (http://beast.bio.ed.ac.uk/) to
test for stationarity of model parameters, to verify adequate
samples sizes, and to determine an appropriate amount of
burn-in. Duplicate analyses were combined in to a single
file if the parameter estimates converged to similar values.
To summarize trees from the posterior distribution, we
constructed a maximum clade credibility tree using
TREEANNOTATER v1.4.7, which is included in the BEAST
software package.
Genetic diversity and number of genetic groups
Reid et al. (2004) detected a deficiency of heterozygotes at
locus Nr60.34 in 10 of our 11 sampled populations,
123
1284
Conserv Genet (2009) 10:1281–1297
Fig. 1 Plestiodon reynoldsi
sampling localities on the south
Lake Wales Ridge. Insert shows
the location of Lake Wales
Ridge in central Florida.
Locality codes are as follows
(listed from north to south):
HHP = Highlands Hammock
State Park; HPE = Highland
Park Estates; LJW = Lake
June-in-Winter; LPS = Lake
Placid Scrub; SSr55/SSr67/
SSr91/SSr99 = Archbold
Biological Station sites;
SH = Sandhill (also within
Archbold); GLD = Gould
Road; HNR = Hendrie Ranch.
Sample sizes for each locality
are in parenthesis
N
HHP (10)
home polygon
Josephine Creek
pre-disturbance xeric soils
extant xeric soils
Lake Wales Ridge
0
1
2
3
4
5
HPE (12)
Kilometers
LJW (11)
LPS (10)
SSr67 (32)
SSr99 (11)
SH (13)
Lake Wales Ridge
Study area
SSr55 (25)
0
100
GLD (20)
200
Kilometers
SSr91 (24)
HNR (11)
suggesting the presence of a null allele (Reid et al. 2004).
To assess the effects of including Nr60.34 in our full
analysis, we performed Spearman correlation tests on FST
values, allelic diversity, heterozygosity, and homozygosity
estimates generated from preliminary analyses with and
without this locus. For all comparisons, correlation coefficients were positive and significant (rs C 0.94), indicating
that the heterozygote deficiency at this locus would have
limited or no effects on our results. Thus, we retained
Nr60.34 in all final analyses.
We used several standard measures of genetic diversity
to explore the effects of recent fragmentation, as significant
loss of diversity will be evident in certain indices before
others following disturbance. For example, in the short
term, reduced dispersal is expected to have a greater effect
on estimates of allelic diversity than on heterozygosity, a
pattern observed in fragmented populations of the Cunningham’s skink Egernia cunninghami (Stow and Briscoe
2005). We calculated mean expected heterozygosity (He),
observed heterozygosity (Ho), allelic richness (AR), and
private allelic richness (pAR) for each microsatellite locus
in all populations. To account for uneven sample sizes
across populations, we used a rarefaction procedure
123
implemented in HP-RARE 1.0 (Kalinowski 2005) to estimate
of AR and pAR. Fisher’s exact test (Fisher 1922) was used
to estimate the pairwise probability of linkage disequilibrium in GENEPOP 3.4 (Raymond and Rousset 1995). We
tested for deviation from Hardy-Weinberg proportions
using a Monte-Carlo approximation of Fisher’s exact test
and a Bonferroni correction for multiple comparisons
implemented in ARLEQUIN 2.0 (Schneider et al. 2000).
To test whether genetic diversity among populations can
be explained by natural or anthropogenic barriers, we used
an analysis of molecular variance (AMOVA) to examine
the distribution of variation among fragments. For the
AMOVA between naturally fragmented habitats, two
groups were compared: Highlands Hammock State Park
(HHP; Fig. 1) versus all remaining populations. We treated
HHP as a singleton because it is isolated from all other
fragments to the south by Josephine Creek, a natural barrier
for other scrub taxa (Deyrup 1996; Clark et al. 1999). To
test for effects of anthropogenic fragmentation, we performed a second analysis by grouping sites within the
Archbold Biological Station and comparing them to all
other fragments. We grouped the Archbold sites because
they are separated only by sandy fire lanes that do not act
Conserv Genet (2009) 10:1281–1297
as complete dispersal barriers (K.G. Ashton, pers. obs.),
whereas the remaining sites were treated individually
because of paved roads or other human-altered habitat. In
both analyses, the proportion of variation was partitioned at
three levels: the individual, the population, and the habitat
fragment (either natural or anthropogenic). The AMOVAs
were performed in ARLEQUIN v.2.0 and significance was
tested using permutation analyses at the different hierarchical levels (Excoffier 2003).
To identify genetic groups, we used a spatially explicit
Bayesian assignment method that implements hidden
Markov random fields (HMRF) as a prior on group membership in TESS 1.1 (http://www-timc.imag.fr/Olivier.
Francois/software.html). The HMRF model assumes that
individuals in continuously distributed populations are
more likely to share group membership with their geographically proximate neighbours than with distant ones
(François et al. 2006; Chen et al. 2007). Population structure is inferred using a Markov Chain Monte Carlo
(MCMC) procedure that detects clinal discontinuities in the
allele frequencies without assuming predefined genetic
groups; analyses are informed by spatial priors derived
from the XY geographic coordinates of the sampling
localities. The degree to which neighbours cluster into the
same group is specified by an interaction parameter w;
values of w [ 0.7 represent a high level of spatial interaction, whereas w B 0.4 represent little or no interaction.
Fixed values of w = 0.5–0.7 generally perform well with
datasets of about 10 populations and are appropriate for
observing several clusters (François et al. 2006). We used
this approach because sand skinks are non-randomly distributed in the landscape and dispersal is limited
(Gianopulos 2001; Christman 2005). Thus, individuals
born in close proximity are more likely to exchange genes
than individuals born far away. Incorporating this spatial
dependency into the assignment tests is therefore likely to
improve model fit and increase the accuracy of parameter
estimates (Gelfand et al. 2005; Latimer et al. 2006), particularly if the sampled populations show weak genetic
differentiation (Chen et al. 2007).
The HMRF model assumes that spatial coordinates are
derived from the birth place of sampled individuals.
Because birth localities were not available, we generated
random coordinates within the limits prescribed by the
distribution of pitfall traps in each fragment, and used these
points as a proxy. Both individual assignment (IA) and
admixture models (AD) were explored in TESS. For the first
series of analyses, we plotted likelihood scores for
K = 1 - 12 (n ? 1 sampling localities) to determine the
value of K that corresponds to the point at which the
likelihood curve plateaus. This threshold was then used as
an indicator of the number of demes present in our sample.
TESS also provides an estimate of K using a statistical
1285
technique known as regularization. The procedure involves
a series of runs beginning with Kmax = 1 (where Kmax is
user specified) and increasing this value until the estimated
K reaches some stable value lower than Kmax. Ideally, the K
value(s) inferred from the likelihood score plots and regularization should be similar. We considered overlap in
these estimates as support for the particular number of
genetic groups in our sample.
After establishing a preferred K, we performed 100 runs
using IA and AD models by setting Kmax = Kpreferred. The
interaction parameter w and the admixture parameter were
fixed at 0.6 and 0.5, respectively. Likelihood plots showed
that removal of 10,000 samples provided ample burn-in,
and 50,000 samples were retained from the posterior distribution. We also performed a series of runs with different
values of the w parameter (w = 0.0, 0.1, 0.2, 0.3, 0.6, 0.9,
and 1.5) to assess the spatial dependence of K and the
group assignment of individuals. If limited information on
genetic structure is available in the microsatellite data, we
expected our results to have greater sensitivity to increased
values of w. Because independent runs can produce different permutations of the group labels we used CLUMPP 1.1
(Jakobsson and Rosenberg 2007) to align the membership
coefficient matrices from 10 runs with the best likelihood
scores (Greedy algorithm, 250 random input orders for
three separate runs). The CLUMPP output consists of the
same matrices permuted so that all replicates are as closely
matched as possible, and the results can be visualized in a
graphics program.
To provide an independent estimate of K, we used a
second MCMC sampling strategy that treats the group
assignment of individuals and the number of groups as
random variables following a Dirichlet process prior
(Antoniak 1974; Pella and Masuda 2006; Huelsenbeck and
Andolfatto 2007). Under this prior, the probability that two
randomly chosen individuals (i and j) fall within the same
group is f ðzi ¼ zj jaÞ ¼ 1=ð1 þ aÞ;where a is a concentration parameter that determines the degree to which
individuals cluster together. Small values of a indicate a
high probability of finding individuals in the same group,
whereas large values indicate a tendency for individuals to
cluster separately. The estimate of K and the assignment of
individuals is then conditional on a and the number of
individuals sampled (Huelsenbeck and Andolfatto 2007).
These analyses were performed in STRUCTURAMA (www.
structurama.org) using an uninformative gamma hyperprior
on a (MacEachern and Muller 1998). Appropriate gamma
priors could not be specified a priori, therefore we explored
a range of values covering several major shapes of the
distribution (shape and scale parameters were, respectively:
1/2, 3/1, 3/2, 3/3, and 6/3). We assumed that robust
estimates of K, as well as the estimate of a that returns K,
should be relatively insensitive to the gamma hyperprior.
123
1286
Markov chains were run for 100,000 generations, sampled
every 100 generations with a 10% burn-in.
Effects of isolation on patterns of genetic structure
We used several approaches to investigate how the spatial
distribution of scrub fragments influences patterns of population differentiation. First, we tested for genetic isolation
by distance (IBD) by performing a regression of pairwise
FST values on the geographic distance (with and without log
transformation) separating sampling localities. Pairwise
FST values (Wright 1965) among all populations were
estimated in FSTAT 2.9.3 (Goudet 1995) and significance of
IBD was assessed using a Mantel matrix correspondence
test (MCT: Mantel 1967) in the software package ZT (10,000
randomizations: Bonnet and Van de Peer 2002). We used
FST over other measures of genetic distance (e.g. Slatkin’s
RST) because it is the most conservative approach when the
number of loci is fewer than 20 and the number of samples
is fewer than 50 (Gaggiotti et al. 1999).
We tested whether isolation by distance is driven by the
geographic separation of anthropogenically isolated fragments, or by the historical distribution of scrub and sandhill
patches within a more continuous landscape. Global Information System (GIS) habitat layers for the Lake Wales
Ridge were used to calculate pairwise distances between
sampled fragments at two time scales, one representing the
historical landscape and one representing the current
landscape. Layers were composites built from USGS topographic maps that included county soil surveys, plant
community and land use coverage, and previously published
maps of habitat distribution; details on their construction are
outlined in Weekley et al. (2007). The historical ‘pre-disturbance’ layer was first assembled to identify areas of
contiguous habitat with appropriate soil types that were
present prior to human settlement. The pre-disturbance map
of habitat availability is necessarily an oversimplification,
because it does not capture various ecological factors such
as fire regimes and differences in microhabitats; however, it
is a good approximation of the distribution of scrub habitat
before human settlement of this region (Weekley et al.
2007). To map the distribution of current scrub habitat, we
then pared the pre-disturbance layer to build an ‘extant’
layer using aerial images and site knowledge, removing
human-altered areas such as paved roads, railroad tracks,
housing developments, and citrus groves. Area polygon
shape files were then converted to point files, and pairwise
geographic distances were calculated from the closest points
along the edges of fragments. If two trap sites belonged to
the same extant or pre-disturbance area, the distance
between them was zero in the matrix.
We used simple MCTs to test the null hypothesis of no
association between genetic differentiation and the degree
123
Conserv Genet (2009) 10:1281–1297
of isolation among fragments. The first test compared
matrices of pairwise population FST values with minimum
geographic distances separating extant fragments, and a
second compared FST with minimum distances separating
historical fragments. We expanded on these analyses by
performing partial MCTs (Smouse et al. 1996) to test for
significant relationships between the FST matrix and the
two geographic distance matrices simultaneously. In these
analyses, partial regressions were performed between two
matrices (e.g. genetic and extant distances) while controlling for the effect of the third matrix (e.g. historical
distance). Partial matrix correspondence tests allow one to
detect spurious correlations that might arise in the simple
pairwise MCTs. Because partial MCTs can be misleading
when the dependent variables are spatially autocorrelated
(Raufaste and Rousset 2001; Castellano and Balleto 2002),
we used these analyses primarily as a means for verifying
the results of the pairwise MCTs.
Effects of habitat area on genetic variability
We tested for recent bottleneck events possibly associated
with restriction in habitat availability and fragmentation
using the program BOTTLENECK v1.2.2 (Cornuet and Luikart
1996). We used two different models of microsatellite
evolution: the strict stepwise mutation model (SMM) and
the two-phase model in which 30% of the mutations consisted of changes by more than a single repeat unit.
Probability values of heterozygosity excess or deficit were
estimated for each population by comparison with a null
distribution based on 1000 iterations.
We also investigated how fragment area affects genetic
variability within populations by estimating a coefficient of
relatedness R (Queller and Goodnight 1989) for individuals
at each sampling locality in GENALEX v.6 (Peakall and
Smouse 2006). Low genetic variability within a population
can result from inbreeding or the existence of closely
related individuals, resulting in an elevated coefficient of
relatedness. Thus, the relationship between intra-population relatedness and fragment size should be significant if
smaller habitat area decreases the amount of genetic
diversity through reduced numbers of breeding individuals.
To test for significant differences among our sampled
populations, we permuted genotypes from all populations
999 times and derived upper and lower 95% confidence
intervals (CI) for the expected range of R, based on all
populations. These intervals represent the range of R
expected if random mating occurs across all populations. In
a second test, we excluded populations HHP, LJW, and
HPE, and permuted genotypes from the remaining populations that were historically connected by scrub habitat
(Fig. 1). The 95% CIs in this case represent the range of R
expected if random mating occurs across these historically
Conserv Genet (2009) 10:1281–1297
connected populations only, and eliminates possible bias
associated with including the most distant and naturally
isolated populations. Population R values that fall above
the upper bound of the 95% CI indicate that reproductive
skew, inbreeding, or drift are increasing relatedness,
despite potential gene flow among some localities.
The pre-disturbance and extant GIS habitat layers were
used to calculate the minimum area of suitable habitat at the
two different time scales for each sampling locality. For
both layers, we tested the null hypothesis of no association
between fragment area and the three measures of withinpopulation diversity (R, He, AR, and pAR). If the size of
current habitat fragments has affected genetic diversity,
higher R values are predicted for populations occurring in
the smallest and most isolated fragments. These same populations are predicted to have lower AR and lower He
compared to populations inhabiting larger fragments. Of
these, AR is more sensitive to recent declines in population
size than heterozygosity (Garza and Williamson 2001; Stow
and Briscoe 2005). Spearman rank-order and linear regression analyses were conducted in SPSS 11.0 (SPSS Inc.) with
and without log transformation of the area data and with
arcsine transformed values for He (Archie 1985). We also
conducted an analysis of variance to test for differences in
He, AR, and pAR between sites in the relatively undisturbed
Archbold Biological Station and all other sites in our study
area that are separated by high-contrast barriers.
Results
Bayesian relaxed clock estimates of divergence times
Our Bayesian relaxed clock estimates of divergence reveal
a deep split between P. reynoldsi and their sister taxon, the
mole skinks (Table 1). The mean estimate under the different coalescence models and mutation rates suggests a
divergence time over 30 million years ago; however, the
wide 95% credible intervals indicate a substantial amount
of uncertainty around those estimates. Nonetheless, the
lower bounds of the credible intervals provide evidence of
an ancient split between the two species (minimum values
range from 16.03 to 20.00 mya; Table 1). Current P. reynoldsi haplotype lineages form a well-supported clade
(posterior probability [PP] = 1.00) that diverged from a
common ancestor roughly 3.2 mya. Haplotypes fall within
regional clades corresponding to northern, central and
southern regions of the Lake Wales Ridge, although the
relationships among these clades are ambiguous (i.e.
PP \ 0.90; Fig. 2). Support values for the central and
southern clades are strong (PP = 1.00 for both), and based
on the more conservative mutation rate, these lineages may
be the oldest of the three regional clades (*1.1 and
1287
1.5 million years old, respectively), while haplotypes from
the North clade diverged from a common ancestor
approximately 750,000 years ago. The inferred divergence
times were consistent under the different coalescent models
and substitution rates (Table 1). All populations from
Highland County belong to the southern clade, and therefore have shared a common evolutionary history for at least
one million years.
Genetic diversity and number of genetic groups
The average number of alleles per locus ranged from 8.25 to
16.63 within populations and from 8.73 to 14.18 across loci
for all populations (Table 2). Allelic richness (AR) showed
limited variation when corrected for the smallest sample
size, ranging from 5.13 to 6.24 within populations and 5.19–
6.79 across loci. The expected heterozygosity (He) per locus
across all sites ranged from 0.68 to 0.99 with an average of
0.90 across all sites, and Ho ranged from 0.20 to 1.00 with
an average of 0.88. Differences in average observed heterozygosity among fragments approached significance
(ANOVA, df = 10, F = 1.90, P = 0.058) and correcting
for multiple comparisons showed that HHP (Fig. 1) had
slightly lower heterozygosity than the remaining populations. Allelic richness (AR) was significantly different
among two groups of fragments: those within the Archbold
Biological Station had higher values (mean AR = 6.01 ±
0.09) than more northern fragments separated by natural
barriers or urban development (mean AR = 5.60 ± 0.12;
ANOVA, df = 10, F = 8.03, P \ 0.01). Private allelic
richness (pAR) was also significantly different between
these same groups of sites (pAR = 0.55 ± 0.04 versus
0.76 ± 0.05; ANOVA, df = 10, F = 10.22, P \ 0.01).
Results of the AMOVA showed that only 1.1% of the
total molecular variation is explained by the anthropogenically isolated fragments, suggesting a substantial level
of admixture among populations inhabiting those sites. In
contrast, differences between HHP and the remaining
fragments south of Josephine Creek account for nearly
4.0% of the variation. Lower average heterozygosity and
greater genetic differentiation between HHP and the
remaining localities suggests that the historical barrier
currently contributes more to the genetic structuring than
anthropogenic barriers.
We explored HMRF models that implemented both
admixture and individual assignment methods in TESS.
Likelihood score plots did not have a clear mode for either
approach, but both tended to plateau at Kmax values above
six. Likewise, the regularization method implemented in
TESS consistently yielded six clusters as Kmax was increased
from 7 to 12. These results were confirmed by our estimates of K using the Dirichlet process prior, where six
groups (range = 6.13–6.69) were consistently identified
123
1288
Conserv Genet (2009) 10:1281–1297
Table 1 Age estimates for sand skink populations on the southern Lake Wales Ridge using Bayesian skyline and constant population size
coalescence models
Clade
Plre
Plre|Pleg
Skyline R1 (mya)
Skyline R2
Constant R1
Constant R2
3.74 (2.13–5.28)
3.21 (2.05–4.57)
3.82 (2.40–5.49)
3.27 (2.01–4.66)
35.00 (18.91–54.53)
30.46 (16.03–46.61)
36.71 (20.00–56.13)
31.77 (17.88–48.65)
Lake Wales North
0.83 (0.35–1.51)
0.72 (0.29–1.29)
0.93 (0.32–1.64)
0.80 (0.26–1.45)
Lake Wales Central
1.69 (0.89–2.59)
1.47 (0.82–2.26)
1.75 (0.98–2.70)
1.50 (0.88–2.30)
Lake Wales South
1.23 (0.53–2.13)
1.07 (0.49–1.81)
1.43 (0.65–2.28)
1.20 (0.54–1.94)
Reported values are the time to most recent common ancestry for the species and regions specified (means followed by 95% credible intervals).
Substitution rate (R1) = 7.15 9 10-3 substitutions/site/million years; rate (R2) = 8.25 9 10-3 s/s/my. Plre = P. reynoldsi, Plre|Pleg = split
between P. reynoldsi and the P. egregius species complex
Fig. 2 Maximum clade
credibility tree for P. reynoldsi
mtDNA haplotypes using
Bayesian skyline coalescent and
GTR ? C nucleotide
substitution models (similar
results were obtained for the
constant population size
coalescent model). Numbers on
the branches indicate posterior
probabilities; we report values
only for those clades with strong
statistical support (i.e. posterior
probabilities [ 0.90).
Haplotypes cluster by region
within the Lake Wales Ridge,
but the relationships among and
within these clades are weakly
supported (i.e. posterior
probabilities \ 0.90).
Haplotype identities follow
Branch et al. (2003)
under a range of plausible gamma hyperpriors. Thus, we
considered six as the best estimate for the number of demes
within our sampled populations.
123
We performed a final series of 50 independent runs
setting Kmax equal to 6 (w = 0.6, 50000 posterior samples
each) and used the 10 best scoring runs as input for CLUMPP
Conserv Genet (2009) 10:1281–1297
1289
Table 2 Population genetic variation of 11 Plestiodon reynoldsi sampling localities at eight microsatellite loci
Locality
N
Nr60.11
Nr60.2
Nr60.34
Nr60.5
Nr52.11
Nr52.2
Nr52.4
Nr52.7
Mean total
HHP
10
0.90/0.89
0.89/0.82
0.56/0.68
0.70/0.89
0.9/0.94
0.90/0.88
0.78/0.80
0.90/0.85
0.82/0.84
5.56/0.19
4.74/1.08
3.89/0.79
5.56/2.57
6.61/0.86
5.50/0.21
4.38/0.10
4.84/0.02
5.13/0.73
0.83/0.92
0.92/0.88
0.50/0.93
0.67/0.76
0.91/0.94
0.83/0.92
1.00/0.89
0.42/0.86
0.76/0.89
6.16/0.64
5.28/0.83
6.24/1.00
4.45/0.42
6.51/1.61
6.15/0.18
5.63/0.57
5.07/0.06
5.68/0.66
0.90/0.84
0.91/0.88
0.25/0.88
1.00/0.96
1.00/0.95
0.82/0.89
0.90/0.89
0.55/0.83
0.79/0.89
4.86/0.02
5.38/0.45
5.46/1.59
6.97/1.73
6.84/1.28
5.86/0.50
5.56/0.91
4.52/0.09
5.68/0.82
0.90/0.93
1.00/0.92
0.22/0.90
0.60/0.87
1.00/0.97
1.00/0.92
0.89/0.84
0.90/0.91
0.81/0.91
6.23/0.74
5.97/1.02
5.76/1.29
5.60/0.65
7.15/1.08
6.01/0.81
4.81/0.78
5.76/0.09
5.91/0.81
0.91/0.91
0.91/0.85
0.38/0.92
0.82/0.92
1.00/0.95
1.00/0.92
1.00/0.92
1.00/0.85
0.88/0.91
HPE
12
LJW
11
LPS
10
SSr99
11
5.98/0.45
5.09/0.07
6.03/1.08
6.13/0.57
6.86/1.14
6.07/0.80
6.00/0.75
4.93/0.03
5.89/0.61
SSr67
32
0.84/0.89
5.75/0.41
0.90/0.88
5.69/0.49
0.37/0.92
6.18/0.51
0.97/0.95
6.79/1.61
1.00/0.94
6.58/0.72
0.93/0.93
6.28/0.27
0.71/0.89
5.65/0.15
0.84/0.84
4.84/0.01
0.82/0.91
5.97/0.52
SH
13
0.83/0.91
1.00/0.92
0.40/0.91
0.83/0.92
1.00/0.95
0.92/0.96
1.00/0.89
0.82/0.86
0.85/0.92
5.96/0.16
6.14/0.33
5.91/0.64
6.16/1.45
6.67/0.69
6.86/0.12
5.71/0.11
5.11/0.05
6.06/0.44
0.96/0.93
0.92/0.87
0.17/0.94
0.83/0.96
0.86/0.95
0.95/0.92
0.82/0.90
0.72/0.89
0.78/0.92
6.33/0.34
5.47/0.55
6.46/0.42
7.01/1.80
6.68/0.44
6.28/1.07
5.85/0.47
5.62/0.17
6.21/0.66
1.00/0.88
0.39/0.87
0.95/0.94
1.00/0.99
0.93/0.95
0.94/0.92
0.84/0.89
0.86/0.92
SSr55
25
GLD
20
0.82/0.92
6.15/0.44
5.70/0.78
5.32/0.56
6.63/0.65
7.69/0.58
6.74/1.02
6.10/0.61
5.56/0.10
6.24/0.59
SSr91
24
0.62/0.93
0.87/0.93
0.20/0.93
0.79/0.88
0.85/0.92
0.94/0.91
0.81/0.90
0.83/0.86
0.74/0.91
6.25/0.67
6.34/1.25
6.26/0.97
5.66/0.85
6.12/0.65
6.08/0.47
5.73/0.41
5.21/0.10
5.96/0.67
HNR
11
0.89/0.90
0.89/0.87
0.50/0.90
0.91/0.90
1.00/0.96
1.00/0.91
1.00/0.87
0.90/0.90
0.89/0.90
5.64/0.09
5.48/0.44
5.99/1.12
5.82/0.35
7.00/0.45
5.71/0.24
5.05/0.04
5.66/0.26
5.79/0.37
N = number of individuals genotyped. The top line for each locality/locus indicates observed (Ho)/expected (He) heterozygosity—the bottom
line indicates the allelic richness (AR)/private allelic richness (pAR). Heterozygosities in bold indicate nonconformity to Hardy–Weinberg
expectations. Locality acronyms follow those in Appendix 1
1.1. Admixture plots revealed little visual resolution of
population structure (not shown), a result consistent with
our AMOVA analyses of anthropogenically isolated populations. In analyses where group assignment was
constrained, individuals tended to share cluster membership with their closest geographic neighbours (Fig. 3).
Adjustments to the spatial interaction parameter w had
marked effects on teasing out membership assignments, a
result that is expected when populations are weakly differentiated. When w = 0, a setting that eliminates spatial
information from the model, individual assignments were
ambiguous with the exception of those from HHP (Fig. 3a).
At non-zero values of w, our clustering results were consistent; assuming w = 0.3 (relatively weak spatial
interaction) individuals were assigned to groups that were
essentially the same as those obtained for models with
moderate to strong spatial interaction (w C 0.6; Fig. 3b).
Thus, the spatial prior compensates for the weak genetic
signal in the dataset, but does not alter assignments when
the model is spatially-informed.
The tessellation plot representing moderate spatial
interaction (w = 0.6) demonstrates patterns of spatial
associations that are consistent with the more traditional
distance-based analyses of our data (Fig. 3b). First, we see
the signature of historical isolation between more northern
isolated populations and the more densely clustered populations in the southern Lake Wales Ridge. Not
surprisingly, HHP forms its own deme, reflecting historical
isolation of this scrub parcel at the northern end of our
sampling range. Further south, LJW and HPE are also
members of a single genetic deme. The remaining populations cluster into four demes that are not explained solely
by geography (Fig. 1). Populations LPS, SSr67 and SSr99
form one deme, with SSr67 and SSr99 currently separated
only by sandy fire lanes inside the Archbold Biological
Station. The SH, GLD, and HNR populations occur in the
southernmost fragments that were once included in the
same historical tract of scrub as the Archbold populations,
but now belong to a separate deme. Finally, skinks from
SSr55 and SSr91 each form single demes, although a few
individuals cluster into groups from immediately adjacent
sites. The SSr55 and SSr91 populations exist in relatively
small fragments, and one of them (SSr55) showed a signal
of population bottleneck (assuming TPM mutation model);
however, neither showed strong evidence of inbreeding or
deviation from migration-drift equilibrium. In summary,
123
1290
Fig. 3 Genetic assignment
plots generated in TESS
(Kmax = 6). Black dots
represent the birth places of
sampled individuals and each
colour represents a unique
genetic group. The Dirichlet
tiling is defined by the spatial
coordinates; two sites are
neighbours if they share a
common edge, and this
information is used by the
HMRF model as a prior on
group membership. (a) Analysis
using a non-informative spatial
prior (i.e. w = 0.0). (b)
Analysis using a moderately
informative spatial prior (i.e.
w = 0.6; a similar pattern is
observed with w = 0.3). Note
that the ability to assign group
membership improves with
higher values of the spatial
interaction parameter w,
indicating limited population
structure in the genetic data
alone
Conserv Genet (2009) 10:1281–1297
A
B
N
HHP
HPE
LJW
LPS
SSr67
SH
GLD
SSr55
SSr91
HNR
the distribution of genetic diversity in this subdivided
landscape is not explained by proximity of fragments
alone; rather, it reflects the historical and current distribution of habitat and population connectivity, as well as
possible demographic changes (either recent or historical)
within individual populations.
Effects of isolation on patterns of genetic structure
Multilocus estimates of FST across all populations indicated
significant genetic differentiation (FST = 0.03; P \ 0.05).
Pairwise FST values across loci ranged from 0.00 to 0.08
(Table 2), with an average of 0.04 between population
pairs. The lowest FST values were between populations in
geographically adjacent fragments (GLD and SSr55;
Fig. 1), despite their current separation by two paved
highways and citrus groves. The highest FST was obtained
for the HNR and HHP populations, the two populations at
opposite ends of the sampling range. Comparing pairwise
FST values to the distances between collecting sites shows a
clear pattern of isolation by distance using a Mantel test
(P = 0.00001, r = 0.5853; Fig. 4). These results suggest a
moderate level of differentiation among sand skink populations and a pattern of increasing genetic isolation as a
function of geographic distance.
123
The MCTs also revealed significant results for matrix
comparisons of pairwise FST values and minimum distances
separating the edges of pre-disturbance (P = 0.0039,
r = 0.6833) and extant areas (P = 0.0003, r = 0.7559).
However, partial MCTs showed that a significant correlation between FST and geographic distance was maintained
only when controlling for isolation between historically
available habitats (P = 0.0018, r2 = 0.5340) and not when
Fig. 4 Isolation by distance inferred from regression between
pairwise population FST values and geographic distances among
sampling sites. FST values were calculated between all pairs of
populations and distances were estimated between the centres of
minimum convex polygons for pitfall arrays at each site
Conserv Genet (2009) 10:1281–1297
1291
controlling for currently available habitats (P = 0.0719,
r2 = 0.3331). This result indicates that once the effect of
extant isolation is ‘removed’ from the analysis, the influence of historical isolation is no longer significant. Thus,
variation in genetic differentiation among sand skink
populations is more highly associated with minimum
distances among contemporary scrub fragments (extant
distance matrix) than with minimum distances among
historical fragments.
Effects of habitat area on genetic variability
We detected some differences between the two-phase
model (TPM) and the single stepwise mutation (SMM)
model in our BOTTLENECK analyses, but the overall pattern
for sand skink populations was a signal of temporal stability. Most sites either showed no significant deviation
from mutation-drift equilibrium, suggesting that recent
bottlenecks have not reduced population sizes, even in the
smallest scrub fragments (Table 3). A single population
(LPS) showed a consistent signal of population bottleneck
across both mutation models. This population occupies a
scrub fragment intermediate in size and is clustered with
other populations in a fairly large tract of preserved scrub
habitat (Fig. 1). Populations SSr55 and HHP showed evidence of a population bottleneck and population expansion,
respectively. However, those two populations showed
conflicting evidence depending on the mutation model
used.
Table 3 Demographic inference of changes in sand skink population
sizes using BOTTLENECK v. 1.2.2 (Cornuet and Luikart 1996)
TPM
SMM
Population
Direction
Probability
Direction
Probability
HHP
HPE
No deviation
No deviation
0.94531
0.84375
Deficiency
No deviation
0.00977
0.38281
LJW
No deviation
0.64063
No deviation
0.54688
LPS
Excess
0.00195
Excess
0.01367
SSr99
No deviation
0.19531
No deviation
0.94531
SSr67
No deviation
0.38281
No deviation
0.31250
SH
No deviation
0.64063
No deviation
0.46094
SSr55
Excess
0.02734
No deviation
0.52734
GLD
No deviation
0.93750
No deviation
0.37500
SSr91
No deviation
0.00195
No deviation
0.74219
HNR
No deviation
0.31250
No deviation
0.94531
We estimated the probability of deviations from migration-drift
equilibrium measured either as heterozygosity excess (implying a
bottleneck), heterozygote deficiency (implying population expansion), or no deviation (no significant excess or deficiency, P [ 0.05).
We used two possible mutation models appropriate for microsatellites: the Two Phase Model (TPM) that assumes 70% single step and
30% multi step mutations, and the Strict Stepwise Mutation Model
(SMM)
Our parallel analyses of relatedness within populations
showed similar patterns independent of whether the simulation included the three northern and historically isolated
populations (Fig. 5). Relatedness, R, ranged from 0.015 to
0.176 including all populations and from 0.008 to 0.085 for
the subset of historically connected populations (Fig. 5). In
the analysis including all populations, the northernmost
fragment HHP had the highest mean coefficient of relatedness at 0.176 (Fig. 5). In both analyses, the lowest
coefficient was observed at one of the Archbold Biological
Station sites, SSr55. Four localities showed significantly
higher relatedness that expected when compared to the
null model of panmictic breeding across the sampled range:
HHP, HPE, LJW, and SSR67. This last population SSR67
still showed significantly increased relatedness when
compared only to other southern Lake Wales Ridge
populations.
Slight negative correlations were detected between
relatedness (R; across all populations and for southern
populations only) and fragment area, indicating a tendency
for higher relatedness among individuals in smaller fragments, but neither result was significant. Spearman rank
order and linear regression analyses revealed slight positive
relationships between fragment area and average He and AR
for both pre-disturbance and extant layers, suggesting that
larger habitat areas harbour more genetic diversity than
smaller areas; however, these relationships were also not
statistically significant. Finally, a non-significant, but
negative correlation was revealed for pAR and pre-disturbance/extant areas. Thus, each index of genetic diversity
varies in the direction predicted by reduced habitat area;
however, sand skink populations occurring in the smallest
habitat remnants have evidently not experienced greater
losses of genetic variability than populations in larger
fragments.
Discussion
Despite the species-typical characteristics that are expected
to promote population differentiation, namely patchy distribution, high habitat specificity, and restricted dispersal
capacity (Didham et al. 1998; Foufopoulos and Ives 1999;
Tscharntke et al. 2002; Davies et al. 2004; Ewers and
Didham 2006), we found limited evidence that sand skink
population genetics have been impacted by recent humanmediated fragmentation. Our analyses of genetic diversity,
estimates of relatedness within scrub patches, and tests of
the effects of temporal variability in the amount suitable
habitat, suggest that some populations diverged before the
onset of major human disturbance, either due to restricted
movement over large geographic areas or to natural historical barriers. The remaining populations that now
123
1292
Conserv Genet (2009) 10:1281–1297
Fig. 5 Estimates of within population relatedness, R, among individuals of Plestiodon reynoldsi. Numbers at the top of each column
indicate fragment area in hectares; open circles = mean population R
with 95% bootstrap intervals); dark gray horizontal lines = range of
expected values under random mating across all sampled populations;
light gray horizontal lines = range of expected values under random
mating across all historically connected populations; light gray box
indicates northern populations that were historically disconnected
from populations in the southernmost scrub habitat patches. Observed
R estimates with 95% intervals outside of the gray lines are
significantly different than the expected values assuming random
mating
occupy a highly fragmented landscape, but once inhabited
nearly continuous scrub habitat, still carry the genetic
signature of this historical connectivity. This finding is
consistent with other studies of taxa subjected to similar
forms of disturbance in the past 50–100 years (Swei et al.
2003; Galbusera et al. 2004; McLoughlin et al. 2004;
Sumner et al. 2004; Driscoll and Hardy 2005). Here, we
propose three possible scenarios to explain the observed
data and discuss the implications of our findings for future
management of the species.
Wales Ridge are persisting below an extinction threshold
defined by the number, size, and spatial distribution of
fragments (Sarre et al. 1995; Hanski and Ovaskainen
2002).
Data from mark-recapture and habitat use studies indicate that current migration among isolated scrub fragments
is likely not sufficient to explain the genetic patterns we
observed in our study. Although locomotor performance
through different substrates has yet to be rigorously
examined for this species (but see Lee 1969; Collazos
1998; Gianopulos 2001; Christman 2005), the available
data show that high contrast boundaries created by
roads, citrus groves, and other urban development
almost certainly constitute low-permeability barriers for
a sand-swimming lizard. Therefore, it is unlikely that
historical levels of migration are currently maintained
among anthropogenically isolated populations. Long-term
suppression of natural fires has also led to vegetative
overgrowth within upland communities, increasing root
densities, canopy cover, and disturbing the natural deposition of wind-blown sand (Myers 1991; Hokit et al. 1999;
Hall et al. 2002; Meshaka and Layne 2002). These factors
likely reduce suitable microhabitat area for sand skinks and
may also limit their dispersal ability. Similarly, altered
habitat at fragment edges can impede movement if individuals avoid the periphery of scrub patches, thereby
amplifying the degree of isolation. Edge avoidance has
been detected in P. reynoldsi and is correlated with soil
compaction and other environmental variables (Gianopulos
2001). At some sites, Gianopulos (2001) found that sand
skinks avoided areas up to 50 m inside the edge of small
patches. However, this pattern was not detected at one site
that was part of a larger scrub remnant, suggesting that
habitat area is also related to edge avoidance behavior.
Thus qualitative features of fragment edges and matrix
environments likely represent significant barriers to gene
Scenario 1: contemporary migration prevents
inbreeding in fragmented populations
One possible explanation for the lack of congruence
between population structuring, genetic diversity, and the
current distribution and size of scrub fragments is that
habitat fragmentation does not restrict sand skink migration
as expected based on species ecology. In field markrecapture studies, individual movement distances ranged
from 35 to 140 m, and distances varied substantially by
location and time of year (Gianopulos 2001; Penney 2001).
Thus, sand skinks have the potential to cover considerable
ground in the appropriate soils, and a surprisingly small
amount of immigration can markedly abate the loss of
genetic variation (Lacy 1987). This movement ability may
be most relevant for fragmented populations within the
Archbold Biological Station or other minimally disturbed
areas, where qualitative habitat barriers (e.g. sandy fire
lanes) do not completely restrict dispersal. Even outside of
protected areas, upland scrub may be in a transitional state
between being able to support or not support a viable
metapopulation of sand skinks. If the landscape has
maintained enough of its original integrity to permit historical metapopulation dynamics, it is theoretically
possible that sand skink populations in the southern Lake
123
Conserv Genet (2009) 10:1281–1297
exchange that are only beginning to affect population
genetic structuring. Future field studies investigating
matrix and edge permeability would provide critical insight
on the ability of these lizards to traverse suboptimal habitat, and the levels of disturbance that can be tolerated
without significantly reducing gene exchange within scrub
networks.
Scenario 2: high population densities prevent loss
of genetic diversity through drift
A second possibility is that despite recent habitat fragmentation, population sizes still exceed a density threshold that
prevents loss of genetic diversity even within small fragments. Our data lend some support to this hypothesis. While
several of our study populations showed higher degrees of
relatedness than expected under random mating, we did not
find any correlation between increased intra-population
relatedness or genetic diversity and the amount of suitable
habitat within a given fragment. Thus, sand skinks currently
inhabiting smaller scrub patches are not significantly
more depauparate in genetic variation than those in larger
patches. A few field studies have shown that densities
vary considerably among sites, ranging from 0.025 to
0.148 individuals/m2 in Orange and Osceola Counties
(Sutton 1996; Collazos 1998) to approximately 0.015 individuals/m2 in Polk and Highlands counties (Christman
2005). At the SSr67 site, density was conservatively estimated at 0.028 individuals/m2 based only on pitfall trap
captures (K.G. Ashton, unpub. data; 2002). It is likely that
Archbold Biological Station alone has thousands of sand
skinks because they are present on every rosemary bald
(*100) as well as in intervening undisturbed sandhill and
scrubby flatwoods. Thus, local abundances may be sufficient
enough to allow random mating within most fragments,
preventing losses of genetic diversity through drift.
Our results on sand skink genetic variability may parallel the findings of vertebrate richness surveys in upland
habitats, where small scrub parcels were shown to support
at least as many taxa as large parcels (McCoy and Mushinsky 1994). Although ecological specialists with poor
dispersal ability are expected to go extinct first when
habitat area decreases (Steffan-Dewenter and Tscharntke
2000; Tscharntke et al. 2002; Davies et al. 2004), landscapes experiencing habitat loss and fragmentation often
show a transient excess of rare species in the short term
(Hanski and Ovaskainen 2002). In upland scrub tracts,
McCoy and Mushinsky (1994) found that rare taxa were
more common than expected in small to medium tracts,
and less common than expected in large tracts. This excess
is often a tell-tale sign of extinction debt, where threshold
conditions for survival are no longer met for certain species
but populations have yet to go extinct because of the
1293
delayed response to disturbance (Andrén 1994; Tilman
et al. 1994; Fahrig 2002; Hanski and Ovaskainen 2002).
Species richness in habitat islands (such as fragments) is
very different than genetic diversity within populations;
however, recent empirical analyses show that the two are
often correlated because migration-drift equilibrium within
species is parallel to the processes of isolation and
extinction that shape the diversity of isolated communities
(Vellend 2003). Therefore, it is possible that inertia exists
in the genetic response of sand skink populations, and as in
the patterns of species richness, we have yet to see the
effects of this reduction in population size. This hypothesis
could be tested with simulation methods that include
effects of population size, longevity, and mating patterns to
predict the expected lag in genetic erosion under different
scenarios of population connectivity.
The lack of a strong correlation between indices of
genetic diversity and habitat area, and the apparent historical demographic stability even in the smallest
fragments, could also reflect a limitation of our ability to
detect important microhabitat boundaries for sand skinks.
We know from demographic studies that abundance varies
even within undisturbed scrub and is determined by several
features. Densities tend to be highest in areas of loose,
large particle sand, low soil moisture, low soil temperature,
and reduced understory vegetation (Collazos 1998; McCoy
et al. 1999; Gianopulos 2001; Christman 2005). Thus,
preferred patches may be more localized than expected
based on the total amount of seemingly available habitat,
and our conception of suitable area may be too coarse to
detect incipient inbreeding.
Scenario 3: life history and demography interact
to delay responses to fragmentation
The third possible explanation for our results is an extension of the population size effects discussed in Scenario 2;
large population densities combined with other demographic characteristics may buffer changes in genetic
diversity such that contemporary evolution in sand skinks
is simply too slow to register the effects of recent fragmentation. Human modification of upland scrub habitat in
central Florida began roughly 60 years ago (McCoy et al.
1999) a very small fraction of the time that sand skinks
have evolved in this region (Table 1). Divergence estimates based on the relaxed molecular clock indicate that
the split between P. reynoldsi and its sister taxon the
P. egregius species group is surprisingly old and that the
sand skink’s adaptive phenotype apparently evolved late in
the existence of this lineage. The time to most recent
common ancestry for all P. reynoldsi haplotypes is slightly
over 3 million years, a time frame that coincides well with
the proposed late Pliocene origin of scrub habitat and the
123
1294
general aridification of Florida (Webb 1990). Thus, it
seems likely that the loose soils within these relict shorelines facilitated the evolution of this lizard’s unique
morphology and fossorial habits. The length of time that
sand skinks have persisted in naturally fragmented scrubs
within the central Florida ridges suggests that they are
adapted to survive in small and patchily distributed areas,
which could in turn contribute to slow response times to
fragmentation. As the size and spatial configurations of
scrub mosaics progress beyond extinction thresholds,
declines in genetic diversity may occur more rapidly as this
adaptive life history trait becomes less effective at buffering against this process. At the same time, the longer
these populations delay their response to disturbance, the
more they accumulate the so-called extinction debt (Tilman
et al. 1994). This debt is usually large in species that persist
close to their extinction threshold following habitat loss, as
might be the case with sand skinks, because the metapopulation time lag is especially long in such species
(Hanski and Ovaskainen 2002).
While no generic method exists to predict the timing of a
species’ response to habitat isolation, effective population
size (Ne) and generation time are two factors that affect
the rate of loss in genetic variability (Wright 1978). Slow
responses are expected for populations with large Ne
and longer generation times, both of which are observed
in sand skinks. Using a minimum generation time of
1.58–1.92 years and a lifespan of 8–10 years (Ashton 2005;
Meneken et al. 2005), we estimate that current populations
are in their 30–38th generation following the onset of
human disturbance. At face value, this would seem sufficient for the accumulation of significant genetic differences,
and simulation studies show that heterozygosity at a single
locus can be completely lost in as few as 100 generations if
population sizes are small (*125 individuals: Lacy 1987).
However, a recent study on the Australian scincid lizard
Gnypetoscincus queenslandiae showed that even with
substantially reduced population densities (0.006–0.014
individuals 9 generation/m2), declines in genetic diversity
within fragments were negligible 50 to 80 years following
habitat isolation (or nine to 12 generations: Sumner et al.
2004). If certain life history traits allow G. queenslandiae to
persist at low populations densities without significant loss
of genetic diversity, it is reasonable to assume that the
substantially higher densities estimated for P. reynoldsi
may have similar if not greater safeguarding effects.
Our estimate of 30–38 generations is likely an overestimate of the number of generations affected by
fragmentation, given the potentially long lifespan and nonannual reproductive mode of sand skinks. Modification of
Florida scrub has been gradually ongoing since the mid1940s, making it difficult to pinpoint when the effects of
human disturbance were significant for P. reynoldsi in this
123
Conserv Genet (2009) 10:1281–1297
part of the range. Additionally, land clearing may lead to
dispersal of refugee individuals rather than mass mortalities in the immediate aftermath, further extending the time
lag of genetic responses following the initial disturbance
(Driscoll and Hardy 2005). Without accurate estimates of
Ne, it is also impossible to tell whether 30–38 generations
exceeds a half-life for which we would expect to see
marked genetic effects. Thus, we cannot exclude the possibility that limited genetic differentiation among recently
isolated scrubs may simply be an artefact of a time-lag in
the response to disturbance.
The hypothesis of delayed response is further supported
by the observation that much of the existing genetic
structure within P. reynoldsi is observed among natural
historic barriers over larger geographic scales. Phylogenetic analyses of mtDNA sequences collected from
throughout the species range (Branch et al. 2003) show that
haplotypes cluster mainly by ridge region, with no appreciable differentiation within ridges. In the microsatellite
dataset, individuals from HHP had the highest level of
within-population relatedness, and results of the AMOVA
show that most of the variation in the data is explained by
allelic differences in this population. This result is best
explained by the natural dispersal barrier, Josephine Creek,
which separates HHP from all other populations to the
south. This barrier is also detectable in the population
structuring of other vertebrates and insects in Highlands
County (Deyrup 1996; Clark et al. 1999). Combined with
the fact that many extant fragments belong to the same
genetic deme, these patterns suggest that reduced connectivity among fragments within regions is too recent to erase
signatures of historical gene flow.
Summary
Sand skinks have evolved in naturally fragmented scrub
habitats, have long generation times, and are locally
abundant in upland scrub and sandhill habitat. These factors have likely contributed to a time-lag in the genetic
response to recent human activity. The importance of
connectivity among sand skink populations is unknown
because gene flow among naturally fragmented patches has
not been studied. Thus, although sand skinks have attributes that predict rapid responses to habitat disturbance
(ecological specialization, patchy habitat distribution, and
restricted dispersal) anthropogenic fragmentation on the
south Lake Wales Ridge is apparently too recent to produce
major shifts in local genetic diversity.
Because our study populations do not show significant
declines in genetic diversity, conservation measures should
be enacted now while fragmented populations remain
relatively admixed (Driscoll and Hardy 2005). Given the
Conserv Genet (2009) 10:1281–1297
1295
length of time that sand skinks have existed in this narrow
stretch of restricted habitat, we agree with earlier proposals
(e.g. McCoy and Mushinsky 1994) that even small reserves
with effective corridors (maintained via controlled burning
or vegetation thinning) may suffice to preserve this unique
species for many years. Future studies should address
barriers to dispersal imposed by matrix habitats, including
movement ability through different substrates, as connectivity among scrub patches is probably important for
population dynamics in this species (Szacki 1999; Berry
et al. 2005). Better sampling for population estimates of Ne
in habitat fragments of different sizes would also provide a
clearer picture of how habitat loss affects local genetic
diversity, and allow for more precise estimates of the time
to new genetic equilibria following disturbance. Finally,
our results underscore the importance of considering the
time scale over which contemporary evolutionary processes operate, and that management practice based on
genetic data must pay close attention to the ways in which
demography influences evolutionary response patterns.
Acknowledgements We thank A. Knipps, B. Branciforte, B.
Meneken, J. Zipser, and volunteers from the Earthwatch Institute for
help with fieldwork, S. Bogdanowicz for help with microsatellite
development, and R. Pickert for assistance with GIS landscape
modelling. H. Mushinsky provided constructive comments on the
manuscript. We also thank H. Swain for providing support with field
efforts and landscape modelling at the Archbold Biological Station.
O. François provided valuable advice concerning the HMRF models.
R. Bukowski facilitated the use of computer resources at the Computational Biology Service Unit (Cornell University) that receives
partial funding from Microsoft. This study was funded by research
grants from: Archbold Biological Station, the Florida Fish and
Wildlife Conservation Commission, and the Earthwatch Institute
(KGA); the Cornell Hughes Scholars Program, Sigma Xi Grants in
Aid of Research, Einhorn Discovery Grant, and the Cornell Undergraduate Board (DTR); and the National Science Foundation (DEB
9907798) and Cornell College of Arts and Sciences (KZ).
Appendix 1
Populations and sample sizes of Plestiodon reynoldsi
included in our study. Locality coordinates are reported for
the centre of the minimum convex polygon formed by
pitfall traps within a site.
Locality Name
N
Lat
Long
HHP
Highlands Hammock State Park
10 27.4283 -81.5192
HPE
Highland Park Estates
12 27.3348 -81.3452
LJW
Lake June-in-Winter
11 27.3121 -81.4201
LPS
Lake Placid Scrub
10 27.2124 -81.3772
SSr99
Archbold Biological Station
SSr99
11 27.2015 -81.3559
Appendix continued
Locality Name
N
Lat
Long
SSr67
Archbold Biological Station
SSr67
32 27.2007 -81.3558
SH
Sandhill
13 27.1858 -81.3408
SSr55
Archbold Biological Station
SSr55
25 27.1407 -81.3552
GLD
SSr91
Gould Road
Archbold Biological Station
SSr91
20 27.1317 -81.3251
24 27.1233 -81.3621
HNR
Hendrie Ranch
11 27.0931 -81.3164
References
Andrén H (1994) Effects of habitat fragmentation on birds and
mammals in landscapes with different proportions of suitable
habitat: a review. Oikos 71:355–366. doi:10.2307/3545823
Antoniak CE (1974) Mixtures of Dirichlet processes with applications
to non-parametric problems. Ann Stat 2:1152–1174. doi:
10.1214/aos/1176342871
Archie JW (1985) Statistical analysis of heterozygosity data: independent sample comparisons. Evol Int J Org Evol 39:623–637.
doi:10.2307/2408657
Ashton KG (2005) Life history of a fossorial lizard, Neoseps reynoldsi. J Herpetol 39:389–395. doi:10.1670/148-04A.1
Berry O, Tocher MD, Gleeson DM, Sarre SD (2005) Effect of vegetation
matrix on animal dispersal: genetic evidence from a study of
endangered skinks. Conserv Biol 19:855–864. doi:10.1111/
j.1523-1739.2005.00161.x
Bonnet E, Van de Peer Y (2002) zt: a software tool for simple and
partial Mantel tests. J Stat Softw 7:1–12
Branch LC, Hokit DG (2000) A comparison of scrub herpetofauna on
two central Florida sand ridges. Fla Sci 63:108–117
Branch LC, Clark AM, Moler P, Bowen BW (2003) Fragmented
landscapes, habitat specificity, and conservation genetics of three
lizards in Florida scrub. Conserv Genet 4:199–212. doi:
10.1023/A:1023398908793
Brandley MC, Schmitz A, Reeder TW (2005) Partitioned Bayesian
analysis, partition choice, and the phylogenetic relationships of
scincid lizards. Syst Biol 54:373–390. doi:10.1080/1063515059
0946808
Brooks TM, Pimm SL, Oyugi JO (1999) Time lag between deforestation and bird extinction in tropical forest fragments. Conserv
Biol 13:1140–1150. doi:10.1046/j.1523-1739.1999.98341.x
Caro T (2007) Behavior and conservation: a bridge too far? Trends
Ecol Evol 22:394–400. doi:10.1016/j.tree.2007.06.003
Castellano S, Balleto E (2002) Is the partial Mantel test inadequate?
Evol Int J Org Evol 56:1871–1873
Chen C, Durand E, Forbes F, Francois O (2007) Bayesian clustering
algorithms ascertaining spatial population structure: a new
computer program and a comparison study. Mol Ecol Notes
7:747–756. doi:10.1111/j.1471-8286.2007.01769.x
Christman SP (1992) Threatened: sand skink, Neoseps reynoldsi
(Stejneger). In: Moler PE (ed) Rare and endangered biota of
Florida. Volume III amphibians and reptiles. University of
Florida Press, Gainesville
Christman SP (2005) Densities of Neoseps reynoldsi on the Lake
Wales Ridge. Final report submitted to U.S. Fish and Wildlife
Service, Vero Beach, FL
123
1296
Clark AM, Bowen BW, Branch LC (1999) Effects of natural habitat
fragmentation on an endemic scrub lizard (Sceloporus woodi):
an historical perspective based on a mitochondrial DNA gene
genealogy. Mol Ecol 8:1093–1104. doi:10.1046/j.1365-294x.
1999.00653.x
Collazos A (1998) Microhabitat selection in Neoseps reynoldsi, the
Florida sand swimming skink. Master’s Thesis, University of
South Florida
Cornuet JM, Luikart G (1996) Description and power analysis of two
tests for detecting recent population bottlenecks from allele
frequency data. Genetics 144:2001–2014
Cowlishaw G (1999) Predicting the pattern of decline of African
primate diversity: an extinction debt from historical deforestation. Conserv Biol 13:1183–1193. doi:10.1046/j.1523-1739.
1999.98433.x
Davies KF, Margules CR, Lawrence JF (2004) A synergistic effect
puts rare, specialized species at greater risk of extinction.
Ecology 85:265–271. doi:10.1890/03-0110
Deyrup M (1996) Two new grasshoppers from relict uplands of
Florida (Orthoptera: Acrididae). Trans Am Entomol Soc
122:199–211
Didham RK, Hammond PM, Lawton JH, Eggleton P, Stork NE
(1998) Beetle species responses to tropical forest fragmentation.
Ecol Monogr 68:295–323
Driscoll DA, Hardy CM (2005) Dispersal and phylogeography of the
agamid lizard Amphibolurus nobbi in fragmented and continuous habitat. Mol Ecol 14:1613–1629. doi:10.1111/j.1365-294X.
2005.02509.x
Drummond AJ, Ho SYW, Phillips MJ, Rambaut A (2006) Relaxed
phylogenetics and dating with confidence. PLoS Biol 4:699–710.
doi:10.1371/journal.pbio.0040088
Drummond AJ, Rambaut A (2007) BEAST: bayesian evolutionary
analysis by sampling trees. BMC Evol Biol 7:214. doi:
10.1186/1471-2148-7-214
Endler JA (1986) Natural selection in the wild. Princeton University
Press, Princeton
Ewers RM, Didham RK (2006) Confounding factors in the detection
of species responses to habitat fragmentation. Biol Rev Camb
Philos Soc 81:117–142. doi:10.1017/S1464793105006949
Excoffier L (2003) Analysis of population subdivision. In: Balding
DJ, Bishop M, Cannings C (eds) Handbook of statistical
genetics. John Wiley and Sons, Ltd., New York, pp 713–750
Fahrig L (2002) Effect of habitat fragmentation on the extinction
threshold: a synthesis. Ecol Appl 12:346–353
Fahrig L (2003) Effects of habitat fragmentation on biodiversity.
Annu Rev Ecol Evol Syst 34:487–515. doi:10.1146/annurev.
ecolsys.34.011802.132419
Fisher RA (1922) On the interpretation of chi-squared from contingency tables, and the calculation of P. J R Stat Soc [Ser A]
85:87–94. doi:10.2307/2340521
Foufopoulos J, Ives AR (1999) Reptile extinctions on land-bridge
islands: life-history attributes and vulnerability to extinction. Am
Nat 153:1–25. doi:10.1086/303149
François O, Ancelet S, Guillot G (2006) Bayesian clustering using
hidden Markov random fields in spatial population genetics.
Genetics 174:805–816. doi:10.1534/genetics.106.059923
Gaggiotti OE, Lange O, Rassmann K, Gliddon C (1999) A
comparison of two indirect methods for estimating average
levels of gene flow using microsatellite data. Mol Ecol 8:1513–
1520. doi:10.1046/j.1365-294x.1999.00730.x
Galbusera P, Githiru M, Lens L, Matthysen E (2004) Genetic
equilibrium despite habitat fragmentation in an Afrotropical bird.
Mol Ecol 13:1409–1421. doi:10.1111/j.1365-294X.2004.02175.x
Garza JC, Williamson EG (2001) Detection of reduction in population
size using data from microsatellite loci. Mol Ecol 10:305–318.
doi:10.1046/j.1365-294x.2001.01190.x
123
Conserv Genet (2009) 10:1281–1297
Gelfand AE, Schmidt AM, Wu S, Silander JA, Latimer A, Rebelo AG
(2005) Modelling species diversity through species level hierarchical modelling. J Roy Stat Soc C-App 54:1–20
Gianopulos KD (2001) Response of the threatened sand skink
(Neoseps reynoldsi) and other herpetofaunal species to burning
and clearcutting in the Florida sand pine scrub habitat. Master’s
Thesis, University of South Florida
Goudet J (1995) FSTAT (Version 1.2): a computer program to
calculate F statistics. J Hered 86:485–486
Greenberg CH, Neary DG, Harris LD (1994) Effect of high-intensity
wildfire and silvicultural treatments on reptile communities in
sand-pine scrub. Conserv Biol 8:1047–1057. doi:10.1046/j.15231739.1994.08041047.x
Hall JM, Gillespie TW, Richardson D, Reader S (2002) Fragmentation of Florida scrub in an urban landscape. Urban Ecosyst
6:143–255. doi:10.1023/B:UECO.0000004825.51640.8b
Hanski I, Ovaskainen O (2002) Extinction debt at extinction
threshold. Conserv Biol 16:666–673. doi:10.1046/j.1523-1739.
2002.00342.x
Hokit DG, Stith BM, Branch LC (1999) Effects of landscape structure
in Florida scrub: a population perspective. Ecol Appl 9:124–134.
doi:10.1890/1051-0761(1999)009[0124:EOLSIF]2.0.CO;2
Huelsenbeck JP, Andolfatto P (2007) Inference of population
structure under a Dirichlet process model. Genetics 175:1787–
1802. doi:10.1534/genetics.106.061317
Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster matching
and permutation program for dealing with label switching and
multimodality in analysis of population structure. Bioinformatics
23:1801–1806. doi:10.1093/bioinformatics/btm233
Kalinowski ST (2005) HP-Rare: a computer program for performing
rarefaction on measures of allelic diversity. Mol Ecol Notes
5:187–189. doi:10.1111/j.1471-8286.2004.00845.x
Lacy RC (1987) Loss of genetic diversity from managed populations:
interacting effects of drift, mutation, immigration, selection, and
population subdivision. Conserv Biol 1:143–158. doi:10.1111/
j.1523-1739.1987.tb00023.x
Lande R (1995) Mutation and conservation. Conserv Biol 9:782–791.
doi:10.1046/j.1523-1739.1995.09040782.x
Latimer A, Wu S, Gelfand AE, Silander JAJ (2006) Building
statistical models to analyze species distributions. Ecol Appl
16:33–50. doi:10.1890/04-0609
Lee DS (1969) Moisture toleration: a possible key to dispersal ability
in three fossorial lizards. Bull Md Herpetol Soc 5:53–56
Lohrer FE (1993) Archbold biological station biennial report 1991–
1992. Archbold Biological Station, Lake Placid, Florida
MacEachern SN, Muller P (1998) Estimating mixture of Dirichlet
process models. J Comput Graph Statist 7:223–238. doi:10.2307/
1390815
Mantel N (1967) The detection of disease clustering and a generalized
regression approach. Cancer Res 27:209–220
McCoy ED, Mushinsky HR (1992) Rarity of organisms in the sand
pine scrub habitat of Florida. Conserv Biol 6:537–548. doi:
10.1046/j.1523-1739.1992.06040537.x
McCoy ED, Mushinsky HR (1994) Effects of fragmentation on the
richness of vertebrates in the Florida scrub habitat. Ecology
75:446–457. doi:10.2307/1939548
McCoy ED, Sutton PE, Mushinsky HR (1999) The role of guesswork
in conserving the threatened sand skink. Conserv Biol 13:190–
194. doi:10.1046/j.1523-1739.1999.97394.x
McKinney ML (1997) Extinction vulnerability and selectivity:
combining ecological and paleontological views. Annu Rev
Ecol Syst 28:495–516. doi:10.1146/annurev.ecolsys.28.1.495
McLoughlin PD, Paetkau D, Duda M, Boutin S (2004) Genetic
diversity and relatedness of boreal caribou populations in
western Canada. Biol Conserv 118:593–598. doi:10.1016/
j.biocon.2003.10.008
Conserv Genet (2009) 10:1281–1297
Meneken BM, Knipps ACS, Layne JN, Ashton KG (2005) Neoseps reynoldsi. Longevity. Herpetol Rev 37:164–165
Meshaka JWE, Layne JN (2002) Herpetofauna of a long-unburned
sandhill habitat in south-central Florida. Fla Sci 65:35–50
Myers RL (1990) Scrub and high pine. University of Central Florida
Press, Orlando
Myers RL (1991) Scrub and high pine. In: Myers RL, Ewel JJ (eds)
Ecosystems of Florida. University of Central Florida Press,
Orlando
Ovaskainen O, Hanski I (2004) Metapopulation dynamics in highly
fragmented landscapes. In: Hanski I, Ovaskainen O (eds)
Ecology, genetics, and evolution of metapopulations. Elsevier
Academic Press, San Diego, pp 73–103
Peakall R, Smouse PE (2006) Genalex 6: genetic analysis in Excel.
Population genetic software for teaching and research. Mol Ecol
Notes 6:288–295. doi:10.1111/j.1471-8286.2005.01155.x
Pella J, Masuda M (2006) The Gibbs and split-merge sampler for
population mixture analysis from genetic data with incomplete
baselines. Can J Fish Aquat Sci 63:576–596. doi:10.1139/
f05-224
Penney KM (2001) Factors affecting translocation success and
estimates of dispersal and movement patterns of the sand skink
Neoseps reynoldsi on restored scrub. M.S. Thesis, University of
South Florida, Tampa, FL
Queller DC, Goodnight KF (1989) Estimating relatedness using
genetic markers. Evol Int J Org Evol 43:258–275. doi:10.2307/
2409206
Raufaste N, Rousset F (2001) Are partial Mantel tests adequate? Evol
Int J Org Evol 55:1703–1705
Raymond M, Rousset F (1995) GENEPOP (Version 1.2): population
genetics software for exact tests and ecumenicism. J Hered
86:248
Reid DT, Ashton KG, Zamudio KR (2004) Characterization of
microsatellite markers in the threatened sand skink (Neoseps reynoldsi). Mol Ecol Notes 4:691–693. doi:10.1111/j.14718286.2004.00788.x
Sambrook J, Russell DW (2001) Molecular cloning: a laboratory
manual, 3rd edn. Cold Springs Harbor Laboratory Press, Cold
Springs Harbor
Sarre SD, Smith GT, Meyers JA (1995) Persistence of two species of
gecko (Oedura reticulata and Gehyra variegata) in remnant
habitat. Biol Conserv 71:25–33. doi:10.1016/0006-3207(94)
00017-K
Schneider S, Roessli D, Excoffier L (2000) ARLEQUIN, version
2.000: a software for population genetics data analysis. Genetic
and Biometry Laboratory, University of Geneva, Geneva
Simberloff D (1986) The proximate causes of extinction. In: Raup
DM, Jablonski D (eds) Patterns and processes in the history of
life. Springer-Verlag, Berlin, pp 259–276
Smouse PE, Long JC, Sokal RR (1996) Multiple regression and
correlation extensions of the Mantel test of matrix correspondence. Syst Zool 35:627–632. doi:10.2307/2413122
Stamps JA, Buechner M, Krishnan VV (1987) The effects of edge
permeability and habitat geometry on emigration from patches of
habitat. Am Nat 129:533–552. doi:10.1086/284656
Steffan-Dewenter I, Tscharntke T (2000) Butterfly community
structure in fragmented habitats. Ecol Lett 3:449–456. doi:
10.1111/j.1461-0248.2000.00175.x
Stockwell CA, Hendry AP, Kinnison MT (2003) Contemporary
evolution meets conservation biology. Trends Ecol Evol 18:94–
101. doi:10.1016/S0169-5347(02)00044-7
1297
Stow AJ, Briscoe DA (2005) Impact of habitat fragmentation on
allelic diversity at microsatellite loci in Cunningham’s skink
(Egernia cunninghami); a preliminary study. Conserv Genet
6:455–459. doi:10.1007/s10592-005-4976-0
Sumner J, Jessop T, Paetkau D, Moritz C (2004) Limited effect of
anthropogenic habitat fragmentation on molecular diversity in a
rain forest skink, Gnypetoscincus queenslandiae. Mol Ecol
13:259–269. doi:10.1046/j.1365-294X.2003.02056.x
Sutton PE (1996) A mark-recapture study of the Florida sand skink
Neoseps reynoldsi and a comparison of sand skink sampling
methods. Master’s Thesis, University of South Florida
Swei A, Brylski PV, Spencer WD, Dodd SC, Patton JL (2003)
Hierarchical genetic structure in fragmented populations of the little
pocket mouse (Perognathus longimembris) in Southern California.
Conserv Genet 4:501–514. doi:10.1023/A:1024768831808
Szacki J (1999) Spatially structured populations: how much do they
match the classic metapopulation concept? Landscape Ecol
14:369–379. doi:10.1023/A:1008058208370
Telford SR (1959) A study of the sand skink, Eumeces reynoldsi
Stejneger. Copeia 2:100–119
Tilman D, May RM, Lehman CL, Nowak MA (1994) Habitat
destruction and the extinction debt. Nature 371:65–66. doi:
10.1038/371065a0
Tscharntke T, Steffan-Dewenter I, Kruess A, Thies C (2002) Characteristics of insect populations on habitat fragments: a mini review.
Ecol Res 17:229–239. doi:10.1046/j.1440-1703.2002.00482.x
Vellend M (2003) Island biogeography of genes and species. Am Nat
162:358–365. doi:10.1086/377189
Watson DM (2002) A conceptual framework for studying species
composition in fragments, islands and other patchy ecosystems.
J Biogeogr 29:823–834. doi:10.1046/j.1365-2699.2002.00726.x
Watson DM (2003) Long-term consequences of habitat fragmentation—highland birds in Oaxaca, Mexico. Biol Conserv 111:283–
303. doi:10.1016/S0006-3207(02)00271-9
Webb SD (1990) Historical biogeography. In: Myers RL, Ewel JJ
(eds) Ecosystems of Florida. University of Central Florida Press,
Orlando, pp 70–102
Weekley CW, Menges ES, Pickert RL (2007) An ecological map of
Florida’s Lake Wales Ridge: a new boundary delineation and an
assessment of post-Columbian habitat loss. Fla Sci 71:45–64
Wiens JA (1997) Metapopulation dynamics and landscape ecology.
In: Hanski IA, Gilpin ME (eds) Metapopulation biology.
Ecology, genetics and evolution. Academic Press, San Diego,
pp 32–60
Wood D (1996) Official lists of Florida’s endangered species,
threatened species and species of special concern. Florida Game
and Fresh Water Fish Commission, Tallahassee, Florida, USA
Wright S (1931) Evolution in Mendelian populations. Genetics
16:97–158
Wright S (1965) The interpretation of population structure by
F-statistics with special regard to systems of mating. Evol Int J
Org Evol 19:395–420. doi:10.2307/2406450
Wright S (1978) Evolution and the genetics of populations. University
of Chicago Press, Chicago
Young AC, Clarke GM (2000) Genetics, demography and viability of
fragmented populations. University Press, Cambridge
Zwick PD, Carr MH (2006) Florida 2060. A population distribution
scenario for the state of Florida. http://www.1000friends
offlorida.org/PUBS/2060/Florida-2060-Report-Final.pdf.
123
Download