Genetic delineation of local provenance defines seed collection zones along a climate gradient
AoB PLANTS www.aobplants.oxfordjournals.org
Downloaded from https://academic.oup.com/aobpla/article
Genetic delineation of local provenance defines seed collection zones along a climate gradient
Kristina M. Hufford 1 2
Erik J. Veneklaas 2
Hans Lambers 2
Siegfried L. Krauss 0 2
Associate Editor: S´ılvia Castro
0 Kings Park and Botanic Garden, Botanic Gardens and Parks Authority , West Perth, WA 6005 , Australia
1 Department of Ecosystem Science and Management, University of Wyoming , Laramie, WY 82071 , USA
2 School of Plant Biology, The University of Western Australia , Crawley, WA 6009 , Australia
Efforts to re-establish native plant species should consider intraspecific variation if we are to restore genetic diversity and evolutionary potential. Data describing spatial genetic structure and the scale of adaptive differentiation are needed for restoration seed sourcing. Genetically defined provenance zones provide species-specific guidelines for the distance within which seed transfer likely maintains levels of genetic diversity and conserves locally adapted traits. While a growing number of studies incorporate genetic marker data in delineation of local provenance, they often fail to distinguish the impacts of neutral and non-neutral variation. We analysed population genetic structure for 134 amplified fragment length polymorphism (AFLP) markers in Stylidium hispidum (Stylidiaceae) along a north - south transect of the species' range with the goal to estimate the distance at which significant genetic differences occur among source and recipient populations in restoration. In addition, we tested AFLP markers for signatures of selection, and examined the relationship of neutral and putatively selected markers with climate variables. Estimates of population genetic structure revealed significant levels of differentiation (FPT ¼ 0.23) and suggested a global provenance distance of 45 km for pairwise comparisons of 16 populations. Of the 134 markers, 13 exhibited evidence of diversifying selection (FPT ¼ 0.52). Using data for precipitation and thermal gradients, we compared genetic, geographic and environmental distance for subsets of neutral and selected markers. Strong isolation by distance was detected in all cases, but positive correlations with climate variables were present only for markers with signatures of selection. We address findings in light of defining local provenance in ecological restoration.
AFLP; BayeScan; ecological restoration; southwestern Australia; spatial genetic structure; Stylidium hispidum
Ecological restoration is often conducted with limited
consideration of genetic diversity or the environmental factors
that influence intraspecific variation
(Rice and Emery 2003;
Bischoff et al. 2010; Byrne et al. 2011)
introductions of propagules can result in genetic bottlenecks
if seeds are collected from a limited number of sources.
Alternatively, wide mixing of provenances may result in
negative consequences due to lower fitness of introduced
plants, or outbreeding depression as a result of
crosspollination among differently adapted genotypes
and Mazer 2003)
. Efforts to re-establish native species
should target evolutionary potential as well as ecological
processes, and restoration programmes can benefit from
knowledge of spatial genetic structure and the scale of
adaptive differentiation in focal species
(Broadhurst et al.
2008; Williams et al. 2014)
Data for intraspecific variation in adaptive traits are
difficult to obtain for most native, non-commercial species,
especially at the scale of large restoration programmes
(Kawecki and Ebert 2004; Savolainen et al. 2013)
. As a
result, seed sourcing guidelines for restoration are often
limited to general ‘rules of thumb’ to conserve genetic
diversity and match environmental conditions between
donor and restoration sites
(Knapp and Rice 1994; Lesica
and Allendorf 1999; McKay et al. 2005)
. A growing number
of studies aim to improve these guidelines through the
use of genetic data to determine the scale of local
adaptation and delineate species-specific seed provenance
(e.g. Stingemore and Krauss 2013; Bower et al.
2014; Dillon et al. 2014)
. Marker-delineated provenance
zones describe the radius within which seed transfer is
predicted to limit the risk for population fitness by
defining the distance beyond which significant genetic
divergence among populations occurs
(Krauss and Koch 2004)
While there is considerable evidence for the association
of genetic diversity and fitness
(McKay and Latta 2002;
Reed and Frankham 2003)
, the correspondence between
molecular markers and adaptive differentiation is unclear
(Edmands 2002; Frankham et al. 2011)
genetic divergence with field survival and breeding studies
can test the efficacy of marker-delineated provenance
zones (Hufford et al. 2012), but these studies are time
consuming and largely unavailable for most species. An
alternative approach is to identify markers with the
signature of diversifying selection
(Foll and Gaggiotti 2008;
Fischer et al. 2011; Funk et al. 2012)
. Comparisons can
then be made between subsets of neutral and candidate
selected markers for the delineation of provenance
distance, and also to describe the scale of intraspecific
adaptation for environmental variables that drive natural
(Krauss et al. 2013; Stingemore and Krauss
2013; Hamlin and Arnold 2015)
Previously, we examined the consequences of
withinpopulation, short- and long-distance crosses (at scales of
100 m, 10 km and 100 km, respectively) for early fitness
of the plant species Stylidium hispidum, endemic to
(Hufford et al. 2012)
. We found evidence
for both inbreeding and outbreeding depression among F1
progeny, supporting an intermediate optimal outcrossing
distance in this species
(Waser 1993; Schierup and Christiansen
. At the same time, we compared genetic structure and
patterns of gene flow within and among the four populations
included in the cross-pollination study. Significant genetic
differentiation among populations correlated with an increased
risk of outbreeding depression, suggesting that parental
divergence corresponded to fitness of intraspecific hybrid progeny
(Pekkala et al. 2012). Further characterization of spatial
genetic structure representative of the species’ range is useful to
provide greater resolution for the estimate of an optimal
distance to minimize population divergence among seed
sources in reintroduction programmes.
In this study, we examined molecular marker
differentiation among 16 populations of S. hispidum, including
4 populations represented in the original study of hybrid
(Hufford et al. 2012)
. We analysed genetic diversity
and population structure along a north–south transect of
the species’ range for 134 amplified fragment length
polymorphism (AFLP) markers. Our goal was to estimate the
distance at which significant differences are likely among
potential source and recipient populations in restoration.
We also conducted surveys for markers displaying
signatures of selection, and examined the relationship of neutral
and putatively selected markers with relevant climate
variables. As a result, we were able to (i) estimate seed transfer
distance within which genetic divergence is low and
population fitness less likely to be affected by maladaptation and
outbreeding depression and (ii) identify both putative
markers and environmental variables linked to fitness
differences in populations across the species’ range. We discuss
our findings in light of seed sourcing for reintroduction of
this species as well as implications for the definition of
‘local’ provenance in ecological restoration.
Location and study species
Southwestern Australia is a global biodiversity hotspot
with .8000 recognized vascular plant species, of which
nearly half are endemic to the region
(Myers et al. 2000;
Hopper and Gioia 2004)
. Vegetation in the Southwest
Australian Floristic Region (SWAFR) is adapted to highly
weathered and severely nutrient-impoverished soils within an
ancient landscape unaffected by glaciation and large
tectonic disturbances (Lambers et al. 2014). The climate is
Mediterranean and annual rainfall primarily occurs during
winter months, with a range of 500– 1400 mm in native
Eucalyptus marginata (jarrah) forest. Records indicate that
the region has experienced a 17 % decline in precipitation
between 1975 and 2011, and the increasing severity of
drought reflects higher temperatures as well as rainfall
(Nicholls 2004; Standish et al. 2015)
The family Stylidiaceae includes .240 species that
occur mainly in Australia, New Zealand and Southeast
(Erickson 1958; Wagstaff and Wege 2002)
. A majority
of those taxa are found in southwestern Australia, a
region identified as the centre of triggerplant evolution
(James 1979; Coates et al. 2003)
. Stylidium hispidum (or
white butterfly triggerplant) is endemic to the SWAFR
and can be found in the jarrah forest understorey along
the Darling Scarp east of the Swan Coastal Plain
Australian Herbarium 1998)
. Plants are herbaceous
perennials with a rosette growth form and produce one or
more scapes with flowering racemes in spring. Flowers
are protandrous and have fused styles and filaments,
forming a column that is triggered by insects, resulting
in pollen transfer. Early chromosome research
determined that S. hispidum is diploid
(n ¼ 14; James 1979)
and evidence supports obligate outcrossing
and James 1991; Hufford et al. 2012)
Collection sites and sampling
We sampled leaf and bud tissue from 16 sites and a
minimum of 30 plants per site along a north – south transect
from Julimar Conservation Park to Dwellingup National
Forest (Fig. 1). This transect spanned a latitudinal
gradient of 160 km, representing much of the species’ range.
Four of the 16 sites included populations for which we
have data describing both inbreeding and outbreeding
depression as a result of short- and long-distance crosses
in this species
(Table 1; Hufford et al. 2012)
of S. hispidum at these sites were found in a patchy
distribution on lateritic soils and consisted of 300 or
more plants. Tissue for genetic analyses was collected
in spring and stored at 280 8C prior to DNA extraction.
Plants within each population were sampled an average
of 10 m apart to avoid genotyping related individuals.
Genomic DNA was extracted according to the methods of
Wagner et al. (1987)
Byrne et al. (2001)
amplification of AFLP markers followed
Vos et al. (1995)
(Hufford et al. 2012)
. Two primer
combinations, EcoRI-AGG/MseI-CTG and EcoRI-ACC/MseI-CTG,
produced distinct bands and were selected for analysis using a
Beckman CEQ 8000 Genetic Analyser. DNA fingerprints
were scored manually with Beckman fragment analysis
software and error rates were calculated at ≤3 % by
comparison of one or more duplicate samples for each
genotyping run. Fragment analyses were performed for 518
individuals from the 16 sites representing an average of
32.4 individuals per site (Table 1). Amplification of DNA
was not successful for all individuals and resulted in a
sample size ,30 for two populations. Data files consisted of
the presence or absence of AFLP bands were prepared for
analyses with functions available in GenAlEx 6.5 and
(Ehrich 2006; Peakall and Smouse 2006)
Summary statistics for AFLP data were calculated using
GenAlEx and included measures of the proportion of
polymorphic loci (PLP) and unbiased estimates of expected
heterozygosity (He). We conducted regression analysis
to test the relationship between population size and
estimates of genetic diversity. We also recorded the number
of private markers, as well as locally common markers
found in ,50 % of populations. Differentiation among
and within populations (calculated as FPT, an analogue
of FST) was estimated using an analysis of molecular
variance (AMOVA) implemented in GenAlEx and based on
9999 permutations. We examined the relative genetic
dissimilarity among sites using non-metric
multidimensional scaling (MDS) ordination in PRIMER 6.1.13 software
(Clarke and Gorley 2006)
for a matrix of pairwise FPT
values calculated in GenAlEx.
Bayesian methods allowed investigation of the number
of significant genetic clusters represented in the dataset
without prior assumptions of population number.
Assignment of individuals to population clusters was conducted
using methods implemented in STRUCTURE 2.3.3
software with the recessive alleles option for dominant
(Pritchard et al. 2000; Falush et al. 2007)
We ran 10 iterations with a burn-in period of 150 000
and 300 000 Markov chain Monte Carlo (MCMC) cycles
(University of Oslo Bioportal; Kumar et al. 2009) using
the default model that infers a and assumes admixture
and correlated allele frequencies. The most likely number
of clusters represented by the data was determined using
the method described in
Evanno et al. (2005)
calculates DK as the second-order rate of change of the log
probability of the data. In cases where population
structure is hierarchical, the method of
Evanno et al. (2005)
only detects significant clusters at the highest level of
the hierarchy. Accordingly, we ran STRUCTURE for subsets
of the data based on K clusters identified in the first run
and repeated analyses for each new subset until the
number of clusters was either K ¼ 1 or very small
(Coulon et al.
. Values of DK were calculated using STRUCTURE
(Earl and vonHoldt 2012)
subsequently ran CLUMPP
(Jakobsson and Rosenberg 2007)
combine results for the 10 runs at each K, and results
were visualized using DISTRUCT software
Pairwise population dissimilarities were investigated
via ANOSIM, a non-parametric, multivariate test similar
to analysis of variance that calculates R statistics using
(Clarke 1993; Chapman and
. In ANOSIM, values of each pairwise R statistic
are compared with a global test statistic to determine
whether populations are significantly differentiated from
one another. We tested the significance of pairwise R values
for a genetic distance matrix representing the 518
individuals and 16 sites using PRIMER and 9999 permutations.
Pairwise R values were subsequently matched to
geographic distance among sites and visualized graphically to
determine (i) the minimum distance representing
significant genetic divergence for S. hispidum and (ii) the global
provenance distance inferred from the intercept of the
global R value with the line of best fit
(Krauss et al. 2013;
Stingemore and Krauss 2013)
. ANOSIM analyses were
then conducted for the pairwise matrix of linearized FPT
values among the 16 sites to test the significance of clusters
derived in Bayesian STRUCTURE analysis.
We tested the assumption that the AFLP dataset
represented neutral markers using an approach available in
BayeScan V.2.1 software
(Foll and Gaggiotti 2008)
identifies markers with unusually high or low levels of
genetic differentiation as outliers that have signatures of
diversifying or balancing selection, respectively. Specifically,
selection is inferred at an AFLP marker if the marker-specific
estimates of FSTare needed in addition to population-specific
estimates to explain observed patterns of differentiation in
(Fischer et al. 2011)
. The analysis was performed
with 20 pilot runs and a 50 000 step burn-in followed by
50 000 iterations and a thinning interval of 10 for the set
of polymorphic AFLP markers. Only polymorphic loci were
included in the analysis. Outliers were identified at the 1 %
significance level, which corresponds to a Bayes factor
threshold of ‘decisive’ evidence for selection relative to the
(log10 of posterior odds .2; Jeffreys 1961)
The false discovery rate (FDR) calculated the expected
proportion of false positives for statistically significant results
(Foll and Gaggiotti 2008)
Environmental data and Mantel analyses
To characterize environmental differences, BIOCLIM
variables were obtained for each of the 16 sites by
extrapolating climate data to the GPS coordinates for each
population using DIVA-GIS software
(Hijmans et al. 2001, 2005)
The BIOCLIM dataset includes 19 variables that describe
monthly temperature and precipitation patterns for a
spatial resolution of 1 km2 (http://www.worldclim.org/). The
sampled area spanned much of the known S. hispidum
range, which occurs primarily in the high rainfall zone,
along a north– south precipitation gradient from 700 –
To avoid redundancy in environmental data, we first
removed variables with high levels of correlation where
|r| . 0.8 and subsequently conducted principal component
analysis (PCA) in JMP 9.0 software. Factor loadings resulting
from Varimax rotation were examined to determine the
variables with the greatest contribution to the variance in
(King and Jackson 1999; Graham 2003)
those variables were added to the reduced dataset. Prior
to subsequent analyses, data were log10(x + 1)
transformed to improve normality and reduce
heteroscedasticity. Dissimilarity matrices of Euclidean distances were
calculated among normalized climate variables using
PRIMER software. A matrix of geographic distances
among sites was generated from GPS coordinates with
the SoDA package in R software and also log10 transformed
(R Development Core Team 2014)
. Multidimensional scaling
ordination was conducted for a similarity matrix of
environmental variables among sites.
Correlations among the 16 sites for measures of
genetic, environmental and geographic distance were
calculated using Mantel and partial Mantel tests in R software
with functions in the ‘vegan’ package
(Oksanen et al.
. Mantel statistics were estimated using Pearson’s
method and 9999 permutations. Spatial structuring of
environmental variables can inflate associations unless
the effects of geographic distance are removed. Thus,
partial Mantel tests were also conducted to determine
the strength of the correlation between two distance
matrices after removing the effect of a third matrix
(Reynolds and Houle 2003)
. Comparisons of genetic and
environmental or geographic distance were made for
the full dataset of 134 markers as well as subsets of
markers with signatures of diversifying selection, or markers
representing neutral genetic variation.
All but 2 of the 134 AFLP markers were polymorphic and
levels of gene diversity were relatively high for the 16
populations with a PLP range from 63 to 91 % (Table 1).
Regression analysis detected a significant correlation
between sampled population size and per cent
polymorphism (R2 ¼ 0.40, P ¼ 0.008) that was not detected
for He. Locally common markers were identified for each
of the 16 populations, but private markers were not
present. The AMOVA partitioned 23 % of the total genetic
variation among sites and 77 % of the total genetic
variation within sites (FPT ¼ 0.23; P , 0.0001; Table 2).
STRUCTURE analysis of the full dataset assigned
individuals to two clusters between northern and southern
sites along a border defined by John Forrest National
Park (JFP1 and JFP2) and Bungendore Nature Reserve
(BG) (Fig. 2). Little admixture was apparent between the
two regions. We conducted one or two additional rounds
of analysis for each northern or southern cluster. The first
STRUCTURE run for the northern sites identified a
separate genetic cluster at John Forrest National Park, while the
final run separated remaining sites from JCP1 and
provided evidence for considerable admixture. Additional
runs for the southern region detected four or six
population clusters consistent with isolation by distance. Overall,
outcomes of nested analyses confirmed the presence of
hierarchical population structure with evidence for 6 or
10 distinct population clusters among the 2 regions and
16 sampled sites (Fig. 2).
Results of ordination analyses reflected the outcomes
of Bayesian clustering methods, and indicated a
distribution of populations consistent with their geographic
distance (Fig. 3). ANOSIM test statistics for the matrix of
FPT values identified significant pairwise differentiation
among the two northern clusters (JFP and JCP/AV sites;
P , 0.05), the relatively isolated collection at BG and
southern sites. However, ANOSIM did not detect
significant differences among six of the southern populations.
Instead, populations at Serpentine National Park (SERP1
and SERP2) were significantly differentiated from nearby
populations in the northern Dwellingup forest (including
CPC, Scarp and TOR), and both of these clusters differed
from the population (YS) furthest south. ANOSIM,
therefore, supported six genetically distinct population clusters
similar to the second tier of Bayesian cluster analyses,
although population clusters reflected some differences
between the two methods.
ANOSIM of genetic distance among the 518 individuals
and 16 sites generated a global R statistic of 0.687 with a
significance level of P , 0.0001. Of the 120 pairwise
combinations, 76 sites were significantly differentiated
from one another. Investigation of the geographic
separation among sites that had significant pairwise
comparisons indicated that the shortest distance between any
two sites that were significantly genetically differentiated
was 12.7 km (WD and YS in the southern region). All
remaining significant pairwise comparisons occurred
between sites 22.6 km or more apart, suggesting a
minimum patch size of 13 – 23 km
(Krauss and Koch 2004)
We inferred a global provenance distance of 45 km
represented by the point where the global R value intersected
the line of best fit (Fig. 4). When considering the distance
class from 45 to 67 km, 44 % of pairwise combinations
resulted in an R test statistic ,0.687, and so were not
significantly different. All sites located at distances .67 km
apart were significantly differentiated from one another.
BayeScan identified 13 outliers that exceeded the 1 %
threshold for selection (posterior odds ¼ 100; Fig. 5) with
a FDR of 0.001. All outliers had higher than expected FST
values, indicating evidence for diversifying selection, and
9 of the 13 markers were retained in BayeScan analyses
when posterior odds were set at 1000. Analysis of
molecular variance of the 13 candidate markers reflected
greater levels of differentiation (FPT ¼ 0.52; P , 0.0001)
but similar levels of polymorphism (81.7 %) relative to
the full dataset (Table 2). The remaining 121 markers fit
a model for neutral variation, and AMOVA represented
significantly lower levels of differentiation among sites
when the 13 selected markers were no longer included
in the AFLP dataset (FPT ¼ 0.18; P , 0.0001).
We ran the ANOSIM analysis a second time using the
genetic distance matrix representing the 13 markers
with signatures of selection among all individuals for
the 16 sites. In this case, 71 of the 120 comparisons
were significantly differentiated based on the global R
statistic (0.737, P ¼ 0.0001), and we noted one pair of
S. hispidum populations that represented significant
genetic differentiation at a very short distance (YS and YRLP
located 0.8 km apart). However, all but two significant
comparisons (including WD and YS, 12.7 km apart) once
again occurred among sites separated by 23 km or
greater, the global provenance distance was 45 km
and pairs of sites .73 km apart were significantly
different in all cases. In effect, the subset of genetic data for
outliers replicated results for the full AFLP dataset, but
pointed to the potential for sites to differ in adaptive traits
at shorter distances than those identified by use of
principally neutral marker datasets.
Environmental data and Mantel analyses
The PCA of climate variables described three factors that
explained 96.1 % of the variation in the data. We selected
four variables that contributed significantly to factor
loadings and had low levels of intercorrelation
et al. 2010; Hamlin and Arnold 2015)
. These variables
included the mean monthly temperature range, the
mean temperature of the driest quarter (or 13-week
period), seasonality of precipitation (defined as the standard
deviation of weekly rainfall estimates divided by the
mean) and the sum of annual precipitation.
Multidimensional scaling analysis resulted in two primary clusters of
northern and southern sites, separated by the two
populations at John Forrest National Park. Upon examination,
the 16 sites represented warmer and drier conditions in
the north relative to southern locations, and greater
variation in annual temperatures and rainfall at northern or
southern margins of the species’ range.
121 Neutral markers
Mantel tests indicated a strong correlation in all cases
between geographic distance and genetic distance
represented by matrices of linearized FPT values (P , 0.001;
Table 3). Simple tests also supported strong correlations
between genetic distance and the four climate variables.
However, partial Mantel tests for the full 134 marker AFLP
dataset and subset of 121 neutral markers only
supported a relationship between genetic and geographic
distance. This provided evidence for strong spatial
autocorrelation of gene diversity among sites. In contrast,
the 13 candidate markers were strongly correlated with
climate variables, both in simple tests and also when
effects of geographic distance were removed in partial
Mantel comparisons (P ¼ 0.002). At the same time,
outliers continued to exhibit a strong correlation with
geographic distance when partial Mantel tests removed the
effects of climate (Table 3). We subsequently divided
climate variables into two subsets reflecting variation in
mean temperature or precipitation, and discovered that
the relationship between the candidate markers and
climate variation was solely the result of correlation with
the two variables for annual precipitation (Mantel’s r ¼
0.325, P ¼ 0.005). The effects of spatial autocorrelation
remained strongly significant for outliers in all cases.
Genetic differentiation and seed sourcing
We observed genetic isolation by distance among
populations of S. hispidum along a north – south transect of the
species’ range. This pattern is consistent with prior results
characterizing genetic structure among 4 of the 16
sampled populations, including evidence of strong
differentiation between northern and southern sites
et al. 2012)
. In our earlier study, significant population
genetic divergence at a distance of 100 km
corresponded to an approximate 6- to 10-fold increased risk
of outbreeding depression in intraspecific hybrid progeny
at early life stages. Consequently, levels of genetic
differentiation reported here provide support for the
rangewide application of a provenance zone that corresponds
to an optimal outcrossing distance by which outbreeding
(as well as inbreeding) depression may be minimized
when sourcing seed for ecological restoration of this
(Lynch 1991; Waser 1993)
We noted a sharp disjunction in genetic clustering
between northern and southern sites at John Forrest
National Park and Bungendore Nature Reserve. The
distance between those sites was the second largest span
(32.7 km) between any two adjacent populations along
the sampled transect, and may indicate a limit for pollen
dispersal. The largest span between any two adjacent
populations occurred between John Forrest National
Park and the northernmost sites (34.8 km; JFP2 and
AV2), and this distance was also reflected in results of
MDS and Bayesian cluster analysis. Evidence of
hierarchical structure suggests that populations are nested within
northern and southern regions, and defined by limits of
gene flow as well as biotic and abiotic factors that drive
(Evanno et al. 2005)
. Examination of the
four climate variables included in analyses detected
significant differentiation along a north – south boundary
similar to results for genetic markers, and supported
regional as well as clinal patterns of environmental
variation. This pattern was noted previously for both genetic
and climate variation representing the range of Banksia
menziesii in southwestern Australia
(Krauss et al. 2013)
Results to describe seed sourcing distance were
supported by data from both marker and breeding studies
of S. hispidum
(Hufford et al. 2012)
. Comparisons of the
relationship between pairwise FPT values and geographic
distance determined that significant genetic
differentiation may occur at a range as small as 13 – 23 km for
the 16 sampled populations. Similarly, intraspecific hybrid
progeny exhibited improved fitness when populations
were 3 – 10 km apart relative to within-population or
longdistance (111 – 124 km) crosses
(Hufford et al. 2012)
addition, regression of pairwise R values and geographic
distance identified a global provenance distance between
any two populations of 45 km. The combined studies
suggest a minimum patch size for S. hispidum with an
average radius no greater than 23 km. These data provide
a range-wide, quantitative estimate to assist seed
sourcing in restoration, and greatly improve upon general
‘rules of thumb’
(Krauss et al. 2013; Stingemore and
Molecular marker studies often fail to distinguish
neutral and non-neutral variation and, therefore, can only
provide indirect evidence for adaptation with limited
(McKay et al. 2005)
. The identification of
candidate markers and their comparison with neutral
marker data, however, is directly relevant for
(Kirk and Freeland 2011; Funk
et al. 2012)
. Knowledge of the scale of adaptive variation
improves the odds of restoring locally adapted traits and,
ultimately, species-level evolutionary potential
and Mazer 2003)
. Analysis of putative selected markers
in S. hispidum confirmed a minimum global provenance
distance of 45 km for sampled sites. These markers
correlated strongly with precipitation variables, suggesting
that the calculated seed provenance zone corresponds
to the scale of adaptive differentiation for climate drivers.
Further testing in the field is warranted, however, to
determine whether seed transfer within this distance
will maintain population fitness.
Approximately 90 % of the AFLP markers characterized in
S. hispidum were consistent with hypotheses of neutral
(Reed and Frankham 2001)
markers correlated strongly with geographic distance but
were not associated with sampled environmental
variables. In contrast, the subset of 13 markers with signatures
of selection was highly correlated with environmental as
well as geographic distance in partial Mantel tests. This
difference supports the hypothesis that climate variation,
as well as spatial autocorrelation, drives locally adapted
genetic differentiation in this species. We detected
significant associations with precipitation but not temperature.
Fitzpatrick et al. (2008)
found that altered precipitation
regimes are likely to strongly impact species’ distributions
in southwestern Australia. Given the significance of rainfall
patterns for population genetic differentiation in this
species, future restoration of S. hispidum may need to draw
more heavily from northern populations adapted to
drought conditions. In this case, unless evidence supports
translocation over longer distances, seed sourcing should
maintain local provenance while selecting plant material
from drier, northern climates. It is likely that rainfall is
not the only driver of adaptive genetic differentiation in
this species, and the detection of strong associations
between genetic and environmental variation will depend
on the variables selected for comparison.
Defining local provenance
The use of local provenance remains a subject of debate,
and composite or admixture collections have been
argued to avoid genetically depauperate sources near
the restoration site, and also to maximize evolutionary
potential in altered environments
(Broadhurst et al.
2008; Breed et al. 2013)
. In most cases, suitably diverse
collections from local provenance zones will meet these
objectives, and conserve locally adapted traits as well
as maintain genetic variation (e.g. Stevens et al. 2015).
When necessary, the provenance zone may expand if
practitioners note high levels of environmental
disturbance, small size of remnant populations and evidence
that locally adapted genotypes are no longer best suited
for restoration sites
(Rice and Emery 2003; Breed et al.
. Even in these conditions, data would not support
transfer of seeds at distances .67 km for this species,
the upper limit beyond which all populations were
significantly genetically differentiated.
Provenance zones define a collection radius, but do not
describe measures to conserve diversity. We noted a
strong positive association of sample size and genetic
diversity measured as per cent polymorphism among all
sites. Therefore, general rules for seed collection
representing multiple individuals and populations would still
(McKay et al. 2005; Leimu et al. 2006)
variation correlates with fitness in many species
, and seed collections should target
locally common alleles to maintain regional variation
(Marshall and Brown 1975)
. These collections would
depend on prior knowledge of spatial genetic
differentiation. In addition, the potential for significant divergence
at short distances (e.g. 0.8 km apart in this study) in
analyses of selected markers strengthens the argument for
habitat matching when combining seeds from multiple
(Krauss and Koch 2004)
The debate concerning the definition and efficacy of
‘local’ seed provenance zones will likely continue,
particularly in light of changing climate conditions
Broadhurst et al. 2008; Sgr o` et al. 2011; Breed et al. 2013;
Havens et al. 2015; Prober et al. 2015)
. Our method
defining local provenance as the threshold at which
geographic distance corresponds to statistically significant
genetic distance is promising, and contributes to
quantitative rather than qualitative guidelines for ecological
(Krauss et al. 2013)
. Moreover, this analysis may
meet restoration requirements for a range of relatively
pristine to highly degraded sites through identification
of the distances at which 50 – 100 % of populations of
target species are genetically differentiated. In highly
fragmented landscapes, the risk of reintroduction of sources
from long distances can be weighed against the
likelihood of population genetic divergence
(Byrne et al.
2011; Breed et al. 2013)
, and corresponding risks of
outbreeding depression. Thus, knowledge of population
structure and historical patterns of gene flow will remain
a critical component of the restoration practitioner’s
‘toolbox’ and, when combined with data for selected
markers, may shed light on the factors that define
species’ distributions and the limits of adaptation.
Sources of Funding
This study was supported by an Australian Research
Council Linkage Grant (LP0669757) and industry partners
Alcoa World Alumina of Australia and BHP Billiton Worsley
Alumina Pty Ltd.
Contributions by the Authors
K.M.H., S.L.K., H.L. and E.J.V. conceived and designed the
study. K.M.H. and S.L.K. conducted field collections.
K.M.H. conducted laboratory and data analyses and
wrote the manuscript with S.L.K., H.L. and E.J.V.
Conflict of Interest Statement
We thank J. Koch, S. Vlahos, D. Coates, J. Wege, J. Anthony,
L. Cockram and personnel at Kings Park and Botanic
Garden and the University of Western Australia for their
assistance with field collections and laboratory studies.
This research was carried out with permission of the
Department of Parks and Wildlife, Western Australia
Bischoff A , Steinger T , Mu¨ ller -Scha¨ rer H. 2010 . The importance of plant provenance and genotypic diversity of seed material used for ecological restoration . Restoration Ecology 18 : 338 - 348 .
Bower AD , St. Clair JB , Erickson V. 2014 . Generalized provisional seed zones for native plants . Ecological Applications 24 : 913 - 919 .
Breed MF , Stead MG , Ottewell KM , Gardner MG , Lowe AJ . 2013 . Which provenance and where? Seed sourcing strategies for revegetation in a changing environment . Conservation Genetics 14 : 1 - 10 .
Broadhurst LM , Lowe A , Coates DJ , Cunningham SA , Mcdonald N , Vesk PA , Yates C. 2008 . Seed supply for broadscale restoration: maximizing evolutionary potential . Evolutionary Applications 1 : 587 - 597 .
Burbidge AH , James SH . 1991 . Postzygotic seed abortion in the genetic system of Stylidium (Angiospermae: Stylidiaceae) . Journal of Heredity 82 : 319 - 328 .
Byrne M , Macdonald B , Francki M. 2001 . Incorporation of sodium sulfite into extraction protocol minimizes degradation of Acacia DNA . Biotechniques 30 : 742 - 748 .
Byrne M , Stone L , Millar MA . 2011 . Assessing genetic risk in revegetation . Journal of Applied Ecology 48 : 1365 - 1373 .
Chapman MG , Underwood AJ . 1999 . Ecological patterns in multivariate assemblages: information and interpretation of negative values in ANOSIM tests . Marine Ecology Progress Series 180 : 257 - 265 .
Clarke KR . 1993 . Non-parametric multivariate analyses of changes in community structure . Australian Journal of Ecology 18 : 117 - 143 .
Clarke KR , Gorley RN. 2006 . PRIMER v6, user manual/tutorial . Plymouth: Primer-E Ltd .
Coates DJ , Carstairs S , Hamley VL . 2003 . Evolutionary patterns and genetic structure in localized and widespread species in the Stylidium caricifolium complex (Stylidiaceae) . American Journal of Botany 90 : 997 - 1008 .
Coulon A , Fitzpatrick JW , Bowman R , Stith BM , Makarewich CA , Stenzler LM , Lovette IJ . 2008 . Congruent population structure inferred from dispersal behaviour and intensive genetic surveys of the threatened Florida scrub-jay (Aphelocoma coerulescens) . Molecular Ecology 17 : 1685 - 1701 .
Dillon S , McEvoy R , Baldwin DS , Rees GN , Parsons Y , Southerton S. 2014 . Characterisation of adaptive genetic diversity in environmentally contrasted populations of Eucalyptus camaldulensis Dehnh. (River Red Gum) . PLoS ONE 9 : e103515 .
Earl DA , VonHoldt BM . 2012 . STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method . Conservation Genetics Resources 4 : 359 - 361 .
Edmands S. 2002 . Does parental divergence predict reproductive compatibility? Trends in Ecology and Evolution 17 : 520 - 527 .
Ehrich D. 2006 . AFLPdat: a collection of R functions for convenient handling of AFLP data . Molecular Ecology Notes 6 : 603 - 604 .
Erickson R. 1958 . Triggerplants. Perth, Western Australia: Paterson Brokensha Pty.
Evanno G , Regnaut S , Goudet J. 2005 . Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study . Molecular Ecology 14 : 2611 - 2620 .
Falush D , Stephens M , Pritchard JK . 2007 . Inference of population structure using multilocus genotype data: dominant markers and null alleles . Molecular Ecology Notes 7 : 574 - 578 .
Fischer MC , Foll M , Excoffier L , Heckel G. 2011 . Enhanced AFLP genome scans detect local adaptation in high-altitude populations of a small rodent (Microtus arvalis) . Molecular Ecology 20 : 1450 - 1462 .
Fitzpatrick MC , Gove AD , Sanders NJ , Dunn RR . 2008 . Climate change, plant migration, and range collapse in a global biodiversity hotspot: the Banksia (Proteaceae) of Western Australia . Global Change Biology 14 : 1337 - 1352 .
Foll M , Gaggiotti O. 2008 . A genome-scan method to identify selected loci appropriate for both dominant and codominant markers: a Bayesian perspective . Genetics 180 : 977 - 993 .
Frankham R , Ballou JD , Eldridge MD , Lacy RC , Ralls K , Dudash M , Fenster CB . 2011 . Predicting the probability of outbreeding depression . Conservation Biology 25 : 465 - 475 .
Funk WC , McKay JK , Hohenlohe PA , Allendorf FW . 2012 . Harnessing genomics for delineating conservation units . Trends in Ecology and Evolution 27 : 489 - 496 .
Graham MH . 2003 . Confronting multicollinearity in ecological multiple regression . Ecology 84 : 2809 - 2815 .
Hamlin JAP , Arnold ML . 2015 . Neutral and selective processes drive population differentiation for Iris hexagona . Journal of Heredity 106 : 628 - 636 .
Havens K , Vitt P , Still S , Kramer AT , Fant JB , Schatz K. 2015 . Seed sourcing for restoration in an era of climate change . Natural Areas Journal 35 : 122 - 133 .
Hijmans RJ , Guarino L , Cruz M , Rojas E. 2001 . Computer tools for spatial analysis of plant genetic resources data: 1. DIVA-GIS . Plant Genetic Resources Newsletter 127 : 15 - 19 .
Hijmans RJ , Cameron SE , Parra JL , Jones PG , Jarvis A. 2005 . Very high resolution interpolated climate surfaces for global land areas . International Journal of Climatology 25 : 1965 - 1978 .
Hopper SD , Gioia P. 2004 . The southwest Australian floristic region: evolution and conservation of a global hot spot of biodiversity . Annual Review of Ecology, Evolution, and Systematics 35 : 623 - 650 .
Hufford KM , Mazer SJ . 2003 . Plant ecotypes: genetic differentiation in the age of ecological restoration . Trends in Ecology and Evolution 18 : 147 - 155 .
Hufford KM , Krauss SL , Veneklaas EJ . 2012 . Inbreeding and outbreeding depression in Stylidium hispidum: implications for mixing seed sources for ecological restoration . Ecology and Evolution 2 : 2262 - 2273 .
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 .
James SH . 1979 . Chromosome numbers and genetic systems in the trigger plants of Western Australia (Stylidium; Stylidiaceae) . Australian Journal of Botany 27 : 17 - 25 .
Jeffreys H. 1961 . Theory of probability . Oxford, UK: Oxford University Press.
Kawecki TJ , Ebert D. 2004 . Conceptual issues in local adaptation . Ecology Letters 7 : 1225 - 1241 .
King JR , Jackson DR . 1999 . Variable selection in large environmental data sets using principal components analysis . Environmetrics 10 : 67 - 77 .
Kirk H , Freeland JR . 2011 . Applications and implications of neutral versus non-neutral markers in molecular ecology . International Journal of Molecular Sciences 12 : 3966 - 3988 .
Knapp EE , Rice KJ . 1994 . Starting from seed: genetic issues in using native grasses for restoration . Restoration & Management Notes 12 : 40 - 45 .
Krauss SL , Koch JM . 2004 . Methodological insights: rapid genetic delineation of provenance for plant community restoration . Journal of Applied Ecology 41 : 1162 - 1173 .
Krauss SL , Sinclair EA , Bussell JD , Hobbs RJ . 2013 . An ecological genetic delineation of local seed-source provenance for ecological restoration . Ecology and Evolution 3 : 2138 - 2149 .
Kumar S , Skjaeveland A ˚ , Orr RJ , Enger P , Ruden T , Mevik BH , Burki F , Botnen A , Shalchian-Tabrizi K. 2009 . AIR: a batch-oriented web program package for construction of supermatrices ready for phylogenomic analyses . BMC Bioinformatics 10 : 357 .
Lambers H , Shane MW , Laliberte´ E, Swarts ND , Teste FP , Zemunik G. 2014 . Plant mineral nutrition . In: Lambers H, ed. Plant life on the sandplains in Southwest Australia, a global biodiversity hotspot . Crawley: UWA Publishing , 101 - 127 .
Leimu R , Mutikainen P , Koricheva J , Fischer M. 2006 . How general are positive relationships between plant population size, fitness and genetic variation? Journal of Ecology 94 : 942 - 952 .
Lesica P , Allendorf FW . 1999 . Ecological genetics and the restoration of plant communities: mix or match? Restoration Ecology 7 : 42 - 50 .
Lynch M. 1991 . The genetic interpretation of inbreeding depression and outbreeding depression . Evolution 45 : 622 - 629 .
Manel S , Poncet BN , Legendre P , Gugerli F , Holderegger R. 2010 . Common factors drive adaptive genetic variation at different spatial scales in Arabis alpina . Molecular Ecology 19 : 3824 - 3835 .
Marshall DR , Brown AHD . 1975 . Optimum sampling strategies in genetic conservation . In: Frankel OH , Hawkes JG , eds. Crop genetic resources for today and tomorrow . London: Cambridge University Press, 53 - 80 .
McKay JK , Latta RG . 2002 . Adaptive population divergence: markers, QTL and traits . Trends in Ecology and Evolution 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 . Restoration Ecology 13 : 432 - 440 .
Myers N , Mittermeier RA , Mittermeier CG , Da Fonseca GAB , Kent J. 2000 . Biodiversity hotspots for conservation priorities . Nature 403 : 853 - 858 .
Nicholls N. 2004 . The changing nature of Australian droughts . Climatic Change 63 : 323 - 336 .
Oksanen J , Blanchet FG , Kindt R , Legendre P , Minchin PR , O'hara RB , Simpson GL , Solymos P , Stevens MHH , Wagner H. 2013 . Vegan: community ecology package . R package version 2 . 0 - 10 . http:// cran.r-project.org/web/packages/vegan/index.html ( 21 March 2015 ).
Peakall R , Smouse PE . 2006 . GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research . Molecular Ecology Notes 6 : 288 - 295 .
Pekkala N , Knott KE , Kotiaho JS , Nissinen K , Puurtinen M. 2012 . The benefits of interpopulation hybridization diminish with increasing divergence of small populations . Journal of Evolutionary Biology 25 : 2181 - 2193 .
Pritchard JK , Stephens M , Donnelly P. 2000 . Inference of population structure using multilocus genotype data . Genetics 155 : 945 - 959 .
Prober SM , Byrne M , McLean EH , Steane DA , Potts BM , Vaillancourt RE , Stock WD . 2015 . Climate-adjusted provenancing: a strategy for climate-resilient ecological restoration . Frontiers in Ecology and Evolution 3 : 65 .
R Development Core Team . 2014 . R: a language and environment for statistical computing . Vienna, Austria: R Foundation for Statistical Computing.
Reed DH , Frankham R. 2001 . How closely correlated are molecular and quantitative measures of genetic variation? A meta-analysis . Evolution 55 : 1095 - 1103 .
Reed DH , Frankham R. 2003 . Correlation between fitness and genetic diversity . Conservation Biology 17 : 230 - 237 .
Reynolds CE , Houle G. 2003 . Mantel and partial mantel tests suggest some factors that may control the local distribution of Aster laurentianus at Iˆles de la Madeleine, Que´bec. Plant Ecology 164 : 19 - 27 .
Rice KJ , Emery NC . 2003 . Managing microevolution: restoration in the face of global change . Frontiers in Ecology and the Environment 1 : 469 - 478 .
Rosenberg NA . 2004 . Distruct: a program for the graphical display of population structure . Molecular Ecology Notes 4 : 137 - 138 .
Savolainen O , Lascoux M , Merila¨ J. 2013 . Ecological genomics of local adaptation . Nature Reviews Genetics 14 : 807 - 820 .
Schierup MH , Christiansen FB . 1996 . Inbreeding depression and outbreeding depression in plants . Heredity 77 : 461 - 468 .
Sgro` CM , Lowe AJ , Hoffmann AA . 2011 . Building evolutionary resilience for conserving biodiversity under climate change . Evolutionary Applications 4 : 326 - 337 .
Standish RJ , Daws MI , Gove AD , Didham RK , Grigg AH , Koch JM , Hobbs RJ . 2015 . Long-term data suggest jarrah-forest establishment at restored mine sites is resistant to climate variability . Journal of Ecology 103 : 78 - 89 .
Stevens MI , Clarke AC , Clarkson FM , Goshorn M , Gemmill CEC . 2015 . Are current ecological restoration practices capturing natural levels of genetic diversity? A New Zealand case study using AFLP and ISSR data from mahoe (Melicytus ramiflorus) . New Zealand Journal of Ecology 39 : 190 - 197 .
Stingemore JA , Krauss SL . 2013 . Genetic delineation of local provenance in Persoonia longifolia: implications for seed sourcing for ecological restoration . Restoration Ecology 21 : 49 - 57 .
Vos P , Hogers R , Bleeker M , Reijans M , Van de Lee T , Hornes M , Friters A , Pot J , Paleman J , Kuiper M , Zabeau M. 1995 . AFLP: a new technique for DNA fingerprinting . Nucleic Acids Research 23 : 4407 - 4414 .
Wagner DB , Furnier GR , Saghai-Maroof MA , Williams SM , Dancik BP , Allard RW . 1987 . Chloroplast DNA polymorphisms in lodgepole and jack pines and their hybrids . Proceedings of the National Academy of Sciences of the USA 84 : 2097 - 2100 .
Wagstaff SJ , Wege J. 2002 . Patterns of diversification in New Zealand Stylidiaceae . American Journal of Botany 89 : 865 - 874 .
Waser NM . 1993 . Population structure, optimal outbreeding and assortative mating in angiosperms . In: Thornhill NM, ed. The natural history of inbreeding and outbreeding: theoretical and empirical perspectives . Chicago: University of Chicago Press, 173 - 199 .
Western Australian Herbarium . 1998 . FloraBase-the Western Australian Flora . Perth, Western Australia: Department of Environment and Conservation. https://florabase.dpaw.wa.gov. au/ (10 June 2015 ).
Williams AV , Nevill PG , Krauss SL . 2014 . Next generation restoration genetics: applications and opportunities . Trends in Plant Science 19 : 529 - 537 .