Ecology Letters, (2012)
doi: 10.1111/j.1461-0248.2012.01752.x
LETTER
Shifts in reproductive assurance strategies and inbreeding costs
associated with habitat fragmentation in Central American
mahogany
Martin F. Breed,1,2 Michael G.
Gardner,1,3 Kym M. Ottewell,1,4
Carlos M. Navarro5 and Andrew J.
Lowe1,6*
Abstract
The influence of habitat fragmentation on mating patterns and progeny fitness in trees is critical for
understanding the long-term impact of contemporary landscape change on the sustainability of biodiversity. We
examined the relationship between mating patterns, using microsatellites, and fitness of progeny, in a common
garden trial, for the insect-pollinated big-leaf mahogany, Swietenia macrophylla King, sourced from forests and
isolated trees in 16 populations across Central America. As expected, isolated trees had disrupted mating
patterns and reduced fitness. However, for dry provenances, fitness was negatively related to correlated
paternity, while for mesic provenances, fitness was correlated positively with outcrossing rate and negatively
with correlated paternity. Poorer performance of mesic provenances is likely because of reduced effective
pollen donor density due to poorer environmental suitability and greater disturbance history. Our results
demonstrate a differential shift in reproductive assurance and inbreeding costs in mahogany, driven by
exploitation history and contemporary landscape context.
Keywords
Common garden experiment, global change, habitat fragmentation, inbreeding depression, logging, mating
system.
Ecology Letters (2012)
INTRODUCTION
Forests form key global ecosystems that humans have utilised for
millennia, but many have experienced unsustainable exploitation
resulting in disrupted ecosystem processes (e.g. pollination; Fig. 1).
Effects of human disturbance (e.g. clearing, selective logging) on tree
mating patterns have been well studied (Eckert et al. 2010), including
numerous neotropical examples (Lowe et al. 2005; Ward et al. 2005).
Both self-compatible and self-incompatible species are expected to
experience fitness declines through reduced mate availability with
disturbance (Fig. 1, response a). Additionally, disturbances typically
manifest as increased selfing for self-compatible species (Fig. 1,
response b), and as increased reproduction between related parents
for self-compatible and self-incompatible species (Fig. 1, response c;
hereafter biparental inbreeding), both leading to fitness reductions due
to inbreeding depression (Szulkin et al. 2010). Regardless of mating
system, inbreeding depression is more commonly expressed in more
stressful environments (Fox & Reed 2010) and is expected to become
more severe as environment-dependent stress increases due to global
change (Beaumont et al. 2011). For species within these altered
1
Australian Centre for Evolutionary Biology and Biodiversity (ACEBB) and
habitats, new shifts in reproductive assurance and fitness costs
associated with inbreeding may be observed (Herlihy & Eckert 2002;
Kalisz et al. 2004).
In agricultural landscapes, the practice of retaining pasture trees
produces a network of spatially isolated trees. Mating pattern
disruption is expected to be most severe in these artificially lowdensity tree systems, particularly for animal-pollinated species
(Ottewell et al. 2009; Breed et al. 2011). Indeed, several authors have
reported dramatic fitness declines for trees in highly isolated contexts
and have generally attributed these declines to inbreeding depression
(Murawski & Hamrick 1991; Gribel et al. 1999). However, many
studies investigating fitness impacts of tree isolation due to habitat
disturbance lack sufficient sampling breadth to overcome potentially
confounding population-specific effects. Additionally, these studies
rarely consider mating system responses other than selfing, often
neglecting correlated paternity and biparental inbreeding, thus limiting
conclusions on the process driving changes in fitness.
Habitat disturbance is also expected to reduce the diversity of
pollen received into tree canopies (Fig. 1, response d; Cascante et al.
2002; Fuchs et al. 2003). Consequently, benefits of pollen diversity (i.e.
5
Universidad Nacional, Instituto de Investigaciones y Servicios Forestales,
School of Earth and Environmental Sciences, University of Adelaide, North
86-3000 Heredia, Costa Rica
Terrace, Adelaide, South Australia 5005
6
2
E-mail: martin.breed@adelaide.edu.au
Environment and Natural Resources, Hackney Road, SA 5005, Australia
3
School of Biological Sciences, Flinders University, Bedford Park, Adelaide,
*Correspondence: E-mail: andrew.lowe@adelaide.edu.au
State Herbarium of South Australia, Science Resource Centre, Department of
South Australia
4
Department of Ecology and Evolutionary Biology, Tulane University, Saint
Charles Avenue, New Orleans, Louisiana 70118
Re-use of this article is permitted in accordance with the Terms and Conditions
set out at http://wileyonlinelibrary.com/onlineopen#OnlineOpen_Terms
2012 Blackwell Publishing Ltd/CNRS
2 Martin F. Breed et al.
Letter
(a)
(b)
(c)
(d)
Figure 1 Possible mating system responses of trees to habitat disturbance, including (a) insufficient pollen ⁄ pollinators; (b) increased selfing; (c) increased biparental inbreeding;
(d) decreased pollen diversity. Genetically similar trees are represented by colours. Pollination is indicated by lines that show failed pollination by coloured dotted lines followed
by a red cross (·); reduced pollination by coloured dotted lines; normal pollination by solid coloured lines; a relative increase in pollination by coloured lines with a green plus
(+). Responses (b, increased selfing), (c, increased biparental inbreeding) and (d, decreased pollen diversity) are not mutually exclusive for hermaphroditic and self-compatible
species.
acquisition of Ôgood genesÕ associated with the augmentation of
genetic diversity; Yasui 1998) are expected to decline in disturbed
areas. The routinely estimated mating system parameter, correlated
paternity (rp), indirectly measures pollen diversity and is inversely
related to the effective number of pollen donors in sampled pollen
clouds (Sork & Smouse 2006). Thus, correlated paternity is expected
to increase in human-altered landscapes, as fewer pollen donors are
present in the landscape and, as predicted by optimal foraging theory
(Charnov 1976), animal pollinators stay longer in more isolated trees
and move between neighbouring trees. However, surprisingly few
studies have assessed the fitness consequences of correlated paternity
changes resulting from habitat disturbance. Where correlated paternity
and fitness have been assessed together (e.g. Rocha & Aguilar 2001;
Cascante et al. 2002; Fuchs et al. 2003), fitness assessments have been
undertaken over periods of time that were insufficient to rule out
alternative explanations (e.g. maternal effects; Charlesworth 1988;
Donohue 2009).
Despite some well-documented examples of changes in mating
patterns resulting from tree isolation, there is still a scarcity of
empirical data investigating the relative importance of individual
mating patterns on fitness (i.e. selfing, biparental inbreeding,
correlated paternity). The few studies that have conducted assessments of multiple mating pattern parameters together with fitness
have assessed only a single population and have relied on comparisons
of study group means (e.g. isolated vs. continuous forest contexts)
rather than family means (i.e. means of offspring from a single mother
tree). By deriving mating system parameters for families rather than
2012 Blackwell Publishing Ltd/CNRS
groups of families, variance increases due to reduced sample sizes, but
this approach can be used to derive statistical relationships between
individual mating system parameters and fitness, rather than relying on
post hoc comparisons of mating system values and fitness means. The
larger number of comparisons within and between groups balances
the increased variance observed for family, compared with population,
level values. This approach can also be used to assess mating system
changes in populations across the range of a species.
Due to its long history of timber exploitation in Central America,
genetic consequences of habitat disturbance have been studied
extensively for mahogany species in the genus Swietenia–a neotropical,
small-insect pollinated (e.g. moths, bees), monoecious tree (Gillies
et al. 1999; White et al. 2002; Lowe et al. 2003; Novick et al. 2003).
Studies of spatial genetic structure of Swietenia suggest that pollen flow
is limited because of small-insect pollination (Lowe et al. 2003).
Consequently, Swietenia species may be susceptible to genetic drift
following habitat disturbance. However, Swietenia species should be
partially genetically buffered from habitat disturbance effects by
strong outcrossing (Lemes et al. 2007), which may reduce a loss of
genetic diversity by inbreeding avoidance. In addition, Navarro &
Hernandez (2004) and Navarro et al. (2011) have demonstrated strong
fitness impacts due to disturbance for Swietenia macrophylla King. These
authors noted strong region-specific responses, best described when
grouping trees into dry and mesic provenance types. On the basis of
these observations, the authors speculate that the adaptive status of
S. macrophylla was either the result of recent adaptive shifts or a shared
evolutionary history under similar environmental conditions. How-
Letter
ever, without an analysis of genetic data, which can shed light on the
mating system dynamics of populations, Navarro & Hernandez (2004)
and Navarro et al. (2011) were not able to resolve the processes
driving variation in fitness.
Here, we report an analysis of genetic variation in S. macrophylla
progeny sourced from forest and isolated tree contexts in 16 Central
America populations, spanning seven countries, that had previously
been raised over a five-year period in a common garden experiment
(reported by Navarro & Hernandez 2004 and Navarro et al. 2011).
Rearing progeny to full maturity (> 20 years) would be ideal to best
understand genetic and mating system impacts on fitness in this
long-lived tree. However, 5 years is longer than is usually reported in
these types of studies and should allow an examination of
differences that probably affect growth in later life (beyond maternal
effects). We predict that increased isolation will associate with
increased selfing, increased biparental inbreeding (although no
change or decreased biparental inbreeding is expected if isolation
exceeds genetic neighbourhood size) and increased correlated
paternity across the study populations (summarised in Fig. 1). We
then predict that greater inbreeding and ⁄ or correlated paternity
should negatively impact progeny growth (as an indicator of
progeny fitness; Yasui 1998; Szulkin et al. 2010; Fig. 1). However,
the region-specific growth responses demonstrated by Navarro &
Hernandez (2004) and Navarro et al. (2011) suggest that factors
governing progeny growth may vary across the study region. Thus,
we predict that regional-level variation in environmental (e.g.
latitude; Colautti et al. 2009), population (e.g. effective density)
and ⁄ or exploitation history factors may drive variation in mating
system parameters and growth.
We demonstrate that increased selfing and correlated paternity both
negatively impact progeny growth, but that variation in these mating
system parameters impacts growth differentially in rainfall-determined
seed provenances (i.e. mesic or dry provenances). Using these data, we
attempt to identify the mechanisms underlying progeny changes in
growth to habitat disturbance, implementing a rigorous analytical
framework that should overcome previous methodological shortfalls.
Finally, we explore implications of habitat disturbance-driven shifts in
inbreeding costs and reproductive assurance for tree species in a
changing world.
Reproductive assurance shifts in mahogany 3
MATERIALS AND METHODS
Collection methods
Open-pollinated progeny used for field trials were collected from 16
natural populations across seven countries throughout the range of
S. macrophylla in Central America (Fig. 2). In each population, mature
trees (> 20 years of age) had seeds sampled according to methods
described by Navarro et al. (2002), where trees were sampled along
transects with a minimum distance of 100 m between trees. The
number of trees sampled per population varied according to
population size and accessibility. Approximately 20 fruits were
collected per tree, each containing approximately 50 viable seeds.
Approximately five seeds per fruit were germinated and reared for
planting into the common garden experiment (details below). Mother
trees were classified as being in an isolated context when no
conspecifics were observed within 500 m, or as being from a forest
context when conspecifics were present within 500 m and were
located in either large remnant forests or forest patches (Navarro et al.
2002, 2011). To account for confounding population effects (e.g.
adaptation), populations were grouped into high- or low-rainfall
provenances following Navarro & Hernandez (2004) and Navarro
et al. (2011). These provenances had comparable densities of
trees estimated from data from Gillies et al. (1999; mean density:
mesic provenances = 1.28 trees ha)1; dry provenances = 1.23 trees
ha)1).
Common garden experiment
To observe variation in progeny growth, approximately nine individuals from each of 71 mother trees were germinated and planted in a
common garden experiment at Los Chiles, Alajuela, Costa Rica (Fig. 2)
and measured over a five-year period. Los Chiles has tropical and mesic
climate with approximately 2500 mm of mean annual rainfall and a
mean annual temperature of 24 C. A generalised randomised block
design (Addelman 1969) was applied, with rows spaced every 3 m and
seedlings spaced 3 m within rows. The height of the ith seedling from
the bth block (hib) and circumference at ground level (cib = 2prib)
were estimated and used to estimate seedling conical volume,
Figure 2 Study area map showing common garden experiment
location (·), sample population locations from mesic (grey filled
circles) and dry (black filled circles) provenances. Provenancecontext sample sizes shown in parentheses (forest and isolated
trees). Inset Table shows mean (mm ± standard error) annual
rainfall of provenance-context groups. Inset map shows location
of study region, highlighting study area (box).
2012 Blackwell Publishing Ltd/CNRS
4 Martin F. Breed et al.
âib = 1 ⁄ 3prib2hib. The difference between the mean volume of the bth
block (âb) and the mean across all blocks (a ) was used to adjust âib to
account for variation among blocks by adjustedâib = âib ) (âb ) a), and
was averaged within families (adjustedâib), hereafter ÔgrowthÕ. Using
growth as a proxy for progeny fitness has limitations, as established
seedlings may have experienced strong selection at pre- or early postzygotic stages. In this study, germination rates were not observed to be
different across progeny classes, but detailed data were not available for
further analysis.
Letter
Observed family multilocus heterozygosity (H^ j) was estimated
following Szulkin et al. (2010) and scaled to between 0 and 1 to account
for missing data by H^ j = RHij ⁄ nj, where Hij is progeny multilocus
heterozygosity for the ith individual in the jth family, nj is the number of
progeny in the jth family, and Hij = Rhik ⁄ nk where h is a heterozygote for
the ith individual at successfully genotyped kth locus and nk is the
number of successfully genotyped loci. Observed adult multilocus
heterozygosity (Hj) was also estimated by Hj = Rhjk ⁄ nk. Samples that
failed to genotype at four or more of the seven loci (n = 1) were
excluded from analyses (samples genotyped at five loci n = 17).
Genetic analysis
DNA was extracted from leaves collected from seedlings using the
Macherey-Nagel Nucleospin Plant II Kit at the Australian Genome
Research Facility (AGRF, Adelaide, Australia). Each progeny was
genotyped using seven dinucleotide microsatellite loci (SM01, SM22,
SM31, SM32, SM40, SM46, SM51), described in Lemes et al. (2002).
An adjusted PCR protocol (compared to Lemes et al. 2002) was used
by amplifying each locus separately, using 0.25 lM final concentration
of BSA and HotMasterTM Taq (Eppendorf, Hamburg, Germany).
PCR products were pooled and LIZ500 size standard was added
before separation on an AB3730 genetic analyser with a 36 cm
capillary array (Applied Biosystems, Foster City, MA, USA). Alleles
were sized using GeneMapper software (Applied Biosystems) and
double-checked manually.
Null alleles result in false homozygotes and are expected to reduce
estimates of observed heterozygosity, increase selfing, biparental
inbreeding and correlated paternity. While null alleles have not
previously been reported at these microsatellite loci (Lemes et al. 2002,
2007; Novick et al. 2003), we assessed the sensitivity of our results to
the presence of null alleles. We estimated the frequency of null alleles
at each locus in both wet and dry provenances using maternal
genotypes in MICRO-CHECKER (Oosterhout et al. 2004). We
detected significant levels of null alleles at two loci in both
provenances (33% and 24% for SM46 and 10% and 14% for SM51
in mesic and dry provenances respectively). For loci with significant
null alleles, we used MICRO-CHECKER to adjust datasets for the
presence of these null alleles. We then compared FIS estimates
(disparity between expected and observed heterozygosity) for
observed and null allele-adjusted datasets in both provenances for
the loci that had significant null alleles in FSTAT (Goudet 1995).
Additionally, we wished to determine the frequency of provenancespecific null alleles required to give comparable genetic results across
provenances (i.e. to match mating system and observed heterozygosity
in both dry and mesic provenances). To do this, we randomly inflated
null allele frequencies for offspring data and re-ran mating system and
heterozygosity analyses (see Appendix S1 for a full description of null
allele methods and results).
Mating system and genetic diversity estimation
The mating system parameters – multilocus outcrossing rate (tm),
biparental inbreeding (tm ) ts) and correlated paternity (rp) – were
estimated in MLTR (Ritland 2002). To calculate parameter variance
for groups of families (i.e. provenances, isolation contexts, populations), families were bootstrapped 1000 times (families are groups of
offspring from a known mother tree). To calculate parameter variance
for families, individuals within families were bootstrapped 1000 times.
2012 Blackwell Publishing Ltd/CNRS
Statistical analysis
For each provenance, we used Gaussian general linear models (GLMs)
in a maximum likelihood, multi-model inference framework in R
v.2.12.1 (R Project for Statistical Computing, http://www.r-project.
org; Burnham & Andersen 2002) to test for hypothesised relationships
between family-level genetic predictors (tm, tm ) ts and rp) and growth.
Additionally, we bootstrapped the regression slopes 10 000 times to
assess the validity of these relationships. As mother tree isolation
should impact mating systems rather than offspring fitness directly, we
did not consider isolation as a statistical factor. To account for
potentially confounding environmental variation across populations
(Colautti et al. 2009), population latitude, mean annual rainfall and
mean annual temperature data were integrated into the GLMs. We
conducted a principal component analysis to summarise environmental differences between populations, as considerable correlations were
present among these environmental variables (see Appendix S2 for
intervariable correlations and principal component analysis information). The first principal component explained 52.3% of variation
among variables and this component was implemented as a
continuous predictor variable in the GLMs. We relied on AkaikeÕs
Information Criterion, corrected for small sample sizes (AICc), for
model selection (Burnham & Andersen 2002).
Heterozygosity-fitness correlations (HFCs) were investigated for
each provenance after Szulkin et al. (2010), by initially regressing mean
family observed heterozygosity (H^ j) with growth (xj = adjustedâib)
and estimating the slope (bxj,H^ j) and variance explained (r2xj,H^ j) by
this relationship. As a correlation between heterozygosity and fitness
does not indicate how much variation is explained by inbreeding, the
inbreeding load (bxj,f ) and variance in fitness explained by inbreeding
(r2xj,f ) were also estimated after Szulkin et al. (2010). HFCs rely on
correlations between observed heterozygosity at genotyped markers
and heterozygosity at functional loci (i.e. correlation due to identity
disequilibrium) and therefore the interlocus heterozygosity correlation
for each provenance (g2) was estimated in RMES (David et al. 2007),
where a significant correlation indicates the presence of identity
disequilibrium. HFCs also rely on variation in inbreeding, thus the
inbreeding estimate, f, was derived from the selfing rate, s, where
s = 1 ) tm, and then f = s ⁄ (2 ) s) (David et al. 2007).
RESULTS
Variation in mating patterns and genetic diversity
Mating system analysis showed that S. macrophylla across Central
America was primarily outcrossed and that families from dry
provenances were almost entirely outcrossed (Table 1). Mesic
Letter
Reproductive assurance shifts in mahogany 5
Table 1 Fitness, genetic diversity, and mating system summary data for Swietenia macrophylla samples across Central America from contrasting provenances and landscape
contexts (nfamily, total number of families (i.e. mother trees) per group; nprogeny, total number of progeny across families per group; growth, mean block adjusted growth; Hj,
mean observed multilocus heterozygosity of adults; H^ j, mean observed multilocus heterozygosity of progeny; tm, multilocus outcrossing rate; tm ) ts, biparental inbreeding
estimate; rp, multilocus correlated paternity; standard deviations in parentheses; 95% confidence interval homogeneous subgroups indicated by ÔaÕ, ÔbÕand ÔcÕ )
Group
nfamily, nprogeny
Growth (m3)
Hj
H^ j
tm
tm ) ts
rp
Central America
Provenance
Mesic
Dry
Landscape context
Forest
Isolated
Provenance and landscape
Mesic forest
Mesic isolated
Dry forest
Dry isolated
71 611
0.056 (0.012)
0.585 (0.24)
0.590 (0.104)
0.968 (0.010)
0.173 (0.020)
0.208 (0.026)
36 294
35 317
0.053 (0.013)a
0.059 (0.010)a
0.520 (0.224)a
0.705 (0.162)b
0.593 (0.139)a
0.632 (0.056)a
0.938 (0.018)a
0.992 (0.033)b
0.198 (0.031)a
0.171 (0.038)a
0.310 (0.045)a
0.170 (0.030)b
47 407
24 204
context
24 194
12 100
23 213
12 104
0.060 (0.010)a
0.048 (0.012)b
0.588 (0.224)a
0.579 (0.255)a
0.607 (0.096)a
0.556 (0.112)a
0.991 (0.020)a
0.925 (0.026)b
0.129 (0.026)a
0.250 (0.032)b
0.163 (0.024)a
0.341 (0.061)b
0.059
0.043
0.062
0.054
(0.011)a
(0.012)b
(0.010)a
(0.010)a b
0.560
0.440
0.687
0.740
(0.214)a b
(0.232)b
(0.160)a
(0.167)a
provenances exhibited greater correlated paternity than dry provenances. Biparental inbreeding did not differ between provenances.
Across all Central American populations, isolated trees had
markedly disrupted mating patterns with reduced outcrossing rates,
increased biparental inbreeding and increased correlated paternity,
compared with progeny sampled from a forest context (Table 1).
However, the effect was not uniform among provenances. Isolated
mesic provenances experienced the most severe disruption to mating
patterns, with reduced outcrossing, increased biparental inbreeding
and increased correlated paternity. Although, for dry provenances,
levels of outcrossing remained unchanged, biparental inbreeding and
correlated paternity were elevated in isolated provenances compared
with progeny from a forest context. Heterozygosity of adults and
offspring was greater in dry provenances than mesic provenances.
Heterozygosity of mesic provenance progeny and adults declined
with isolation, but both progeny and adults of dry provenances
displayed limited change in heterozygosity with isolation.
We detected significant null alleles at two microsatellite loci in both
provenances (SM46: 33 and 24%, SM51: 10 and 14%, in mesic and dry
provenances respectively), but original and null allele adjusted FIS
estimates were similar, thus the empirical data were used for
subsequent analyses. Considerable increases in null alleles (e.g. 15%
at all loci) were required to give comparable mating patterns and
genetic diversity estimates in both provenances (see Appendix S1 for
details of null allele results).
0.604
0.503
0.611
0.610
(0.125)a
(0.115)a
(0.055)a
(0.082)a
0.987
0.845
0.992
0.992
(0.042)a
(0.041)b
(0.054)a
(0.103)a
0.148
0.314
0.149
0.208
(0.052)a
(0.027)b
(0.055)a
(0.086)a
0.254
0.445
0.153
0.278
(0.052)a
(0.086)b
(0.026)c
(0.084)a
tions (Table 2). Overall, increasing correlated paternity had a negative
effect on growth and its effect was equivalent to that of outcrossing rate.
Neither biparental inbreeding nor environmental differences showed a
relationship with growth when considered in single parameter models.
For mesic provenances, outcrossing rate and correlated paternity
explained most variation in progeny growth, even when controlling
for environmental differences among populations (Table 2; Fig. 3).
Biparental inbreeding was also negatively correlated with growth, but
had lower influence than outcrossing rate and correlated paternity. For
dry provenances, only the model that included variation in correlated
paternity fitted the growth data better than the null model.
Outcrossing rate, biparental inbreeding and environmental differences
showed no relationship with growth in dry provenances.
As family-level estimates of mating system parameters have higher
levels of variance than mating system parameters estimated for groups
of families, we ran correlations between population mating system
estimates and fitness for each provenance (see Appendix S3 for full
details). Additionally, we bootstrapped the regression slopes of the
family-level analyses. These analyses all supported our original familylevel analyses. Each correlation between population-level mating
system parameters and growth was in the same direction as our familylevel analyses (Appendix S3). Additionally, the 2.5 and 97.5 percentiles
of the bootstrapped slope distributions also confirmed the family-level
trends (Appendix S3).
Correlations between heterozygosity and fitness
Variation in fitness traits
A total of 611 progeny from 71 families sampled from 16 populations
were observed over the five-year common garden experiment. Mesic
provenance progeny were significantly smaller than dry provenance
progeny (Table 1). Progeny from trees in isolated contexts were
significantly smaller than forest context trees and this difference was
more pronounced in mesic provenance progeny than dry provenance
progeny.
Correlations between mating system parameters and fitness
Analysed across all Central American populations, increasing outcrossing rate had a strong positive effect on progeny growth, even when
controlling for variation in environmental differences among popula-
Progeny from dry provenances exhibited higher heterozygosity than
those from mesic provenances. However, progeny from mesic
provenances had greater variation in heterozygosity and displayed
greater inbreeding than dry provenances (Table 1, 3). Heterozygosity
was positively correlated with growth in both provenances, but this
relationship was much stronger in mesic provenances. Interlocus
correlation of heterozygosity (i.e. identity disequilibrium), as measured
by g2, was significant for mesic provenances, but not dry. Consequently,
the relationship between inbreeding and fitness (using progeny growth
as the quantified variable, r2xj,f ) and inbreeding load (bxj,f ) was only
estimated for mesic provenances, where inbreeding explained 31.7% of
variation in fitness and had an inbreeding load of) 0.042 m3. This
inbreeding load translates to an average change in fitness after one
generation of selfing (i.e. f = 1 ⁄ 2) of ) 0.042 m3 · 1 ⁄ 2 = ) 0.021 m3.
2012 Blackwell Publishing Ltd/CNRS
6 Martin F. Breed et al.
Letter
Table 2 General linear models of relationships among genetic and environmental predictors and response variable ÔgrowthÕ, a fitness proxy of Swietenia macrophylla. Analyses
conducted for both isolated and forest landscape context samples grouped to include all samples, samples from only mesic and dry provenances (% DE, percent deviance
explained by model i; wAIC, AIC weights shows the relative likelihood of model i; DAICc, difference between model AIC and minimum AIC in the set of models; AICc,
AkaikeÕs Information Criterion corrected for small samples sizes; k, number of parameters in the given model; ß, unstandardised regression slopes and standard errors for each
predictor variable in models with DAICc < 4; tm, outcrossing rate; tm ) ts, biparental inbreeding; rp, correlated paternity; PCENV, first component of principal component
analysis of environmental variables; 1, null model)
Model
% DE
wAIC
DAICc
All families
growth tm
growth rp
growth tm + PCENV
18.00
16.92
23.57
0.43
0.27
0.18
0.00
0.93
1.75
growth tm ) ts
growth rp + PCENV
growth tm ) ts + PCENV
growth PCENV
growth 1
Mesic provenance
growth tm
growth rp
growth tm + PCENV
12.93
20.32
18.31
13.39
0.00
0.05
0.04
0.02
0.01
0.00
4.26
4.71
6.48
8.31
11.97
22.91
18.88
33.46
0.46
0.18
0.14
growth tm ) ts
growth PCENV
growth rp + PCENV
growth tm ) ts + PCENV
growth 1
Dry provenance
growth rp
growth 1
growth PCENV
growth rp + PCENV
growth tm ) ts
growth tm
growth tm ) ts + PCENV
growth tm + PCENV
15.57
22.99
28.51
27.72
0.00
9.92
0.00
5.89
9.92
2.13
0.42
6.07
6.04
AICc
k
) 436.70
) 435.77
) 434.95
3
3
6
432.44
431.98
430.22
428.39
424.72
3
6
6
5
2
0.00
1.83
2.39
) 215.11
) 213.27
) 212.72
3
3
6
0.09
0.04
0.04
0.03
0.01
3.27
4.92
4.98
5.37
7.11
)
)
)
)
)
211.83
210.19
210.13
209.74
208.00
3
5
6
6
2
0.50
0.25
0.12
0.09
0.02
0.01
0.01
0.01
0.00
1.41
2.90
3.51
6.46
7.63
9.10
9.11
)
)
)
)
)
)
)
)
223.17
221.76
220.27
219.66
216.71
215.54
214.07
214.06
3
2
3
3
5
6
6
6
DISCUSSION
We demonstrate that across 16 Central American populations of big-leaf
mahogany, S. macrophylla, a globally threatened species of paramount
ecological and economic importance, variation in mating system and
progeny growth (used as an indicator of fitness) was significantly
impacted by the level of habitat disturbance. Interestingly, populations
from mesic regions exhibited greater impact than those from dry regions,
indicating the importance of regional or cross-population comparisons
when assessing the fitness and mating impacts of habitat disturbance, a
factor rarely considered, but often speculated on (Lowe et al. 2005;
Aguilar et al. 2008; Eckert et al. 2010; Szulkin et al. 2010). Additionally, by
tracking growth of progeny over 5 years and by examining individual
family (i.e. groups of offspring from a known mother tree) rather than the
mean of groups of families, we demonstrate that progeny growth was
reduced not only by increased selfing (i.e. inbreeding depression
component due to the inbreeding load; Szulkin et al. 2010) but also by
increased correlated paternity (i.e. via pollen diversity effects; Yasui
1998) – a factor seldom integrated into these types of studies.
Context-dependent effects of disturbance on mating patterns and
genetic diversity
Our data support the wealth of previous studies that document the
negative effects of habitat disturbance by increasing inbreeding (Lowe
2012 Blackwell Publishing Ltd/CNRS
)
)
)
)
)
ß (m3)
0.052 (0.013)
) 0.024 (0.007)
tm = 0.042 (0.014);
PCENV = 0.003 (0.001)
0.048 (0.015)
) 0.029 (0.010)
tm = 0.035 (0.015);
PCENV = 0.014 (0.006)
) 0.034 (0.014)
) 0.016 (0.008)
et al. 2005; Eckert et al. 2010), reducing genetic diversity (Lowe et al.
2005; Aguilar et al. 2008) and increasing correlated paternity (Lowe
et al. 2005). Interestingly, the effect of disturbance on mating patterns
in our study was provenance-specific. In particular, dry provenances
had similarly high levels of progeny heterozygosity and outcrossing in
both forest and isolated tree contexts (£ 0.1% difference). In contrast,
mesic provenances demonstrated reduced outcrossing rate and
progeny heterozygosity in isolated compared with forest context trees
(14 and 10% reduction respectively). In addition, both provenances
experienced increased correlated paternity and biparental inbreeding
with isolation, but to a much greater extent in mesic compared with
dry provenances (rp = 0.19 and 0.13 increase, respectively;
tm ) ts = 0.17 and 0.06 increase respectively).
Most studies examining the genetic effects of disturbance in tree
species have studied only one or a few populations per species (Ward
et al. 2005; Eckert et al. 2010). As such, the stark difference between
provenances (i.e. groups of ecologically similar populations) found
here highlights the importance of considering intraspecific variation
when making either species-wide conclusions or generalities about the
effects of disturbance. Upon consideration of multiple populations
within a species, which ideally cross known environmental and ⁄ or
genetic breaks to account for variation in evolutionary history (e.g. by
rainfall provenance in this case), more confident extrapolations to
species-wide trends can be made. Intraspecific variation in mating
systems has been previously discussed (Barrett 2003), but little
Letter
Reproductive assurance shifts in mahogany 7
0.1
Mesic provenances
Dry provenances
Growth (m3)
0.08
0.06
0.04
0.02
0
0
0.2
0.6
0.8
1 0
0.2
0.4
0.6
0.8
1
0.8
1
Outcrossing rate
0.08
Growth (m3)
Figure 3 Scatterplots showing relationships between family-level
genetic parameters and growth for both provenances. Growth is
shown on the y-axis and genetic parameter values are shown on
the x-axis. Mesic provenance family data are indicated by greyfilled squares, dry provenance family data indicated by black-filled
squares. Linear trend lines between genetic parameters and growth
shown for relationships where DAICc < 4 (DAICc values presented in Table 2), with grey lines for mesic provenances and
black lines for dry provenances.
0.4
Heterozygosity
0.1
0.06
0.04
0.02
0
0
0.2
0.4
0.6
0.8
1 0
0.2
Correlated paternity
0.4
0.6
Biparental in breeding
Table 3 HFC comparisons following Szulkin et al. (2010) for both mesic and dry provenances of Swietenia macrophylla (H^ j and r2(H^ j), heterozygosity mean and variance,
respectively; f, inbreeding estimate derived from the MLTR selfing rate where f = s ⁄ (2 ) s); g2, interlocus heterozygosity correlation inferred from RMES (David et al. 2007);
r2xj,H^ j, the variation in fitness explained by heterozygosity; bxj,H^ j, regression slope of fitness-heterozygosity regression; r2xj,f, variation in fitness explained by inbreeding; bxj,f,
regression slope of fitness-inbreeding, the inbreeding load; variance parameters in parentheses; g2 values followed by Ô*Õ or ÔNSÕ indicate significant and non-significant interlocus
heterozygosity correlation respectively)
Provenance
H^ j
r2(H^ j)
f
bxj,H^ j
r2xj,H^ j
g2
r2xj,f
bxj,f
Mesic
Dry
0.593
0.632
0.139
0.056
0.032
0.004
0.047
0.020
0.209
0.016
0.033*
) 0.012NS
0.317
) 0.042
attention has been applied to contemporary landscape change as a
driver for this change. Environmental and genetic breaks may not
necessarily correlate as genetic breaks may, for example, follow latitude
and environmental breaks may follow longitude (Colautti et al. 2009).
Thus, efforts to control for population-specific effects are dependent
upon a priori knowledge about what kind of break is important. Here,
quantitative genetic data were used to inform provenance delineation
from Navarro & Hernandez (2004) and Navarro et al. (2011). Further
population genetic structure data would offer greater insight into
population effects and both types of data are important considerations
when attempting to overcome environmental and genetic effects on
fitness. We analysed relative progeny growth among families within
provenances (each provenance was analysed separately and environmental data were included in the statistical models to control for
environmental effects); therefore, population differences were unlikely
to be confounding genetic trends within each provenance (e.g. local
adaptation or poor suitability of dry or wet provenances to the
experimental conditions).
There are a number of explanations for why we observe a greater
reduction in the effective density of pollen donors in mesic compared
with dry provenances. First, wetter conditions are less favourable than
dry to S. macrophylla survival (Lamb 1966; Holdridge et al. 1971; Gillies
et al. 1999). Second, mesic provenances have experienced greater
human disturbance. Gillies et al. (1999) presented logging intensity
data and, using these data, the average logging intensity was higher for
mesic provenances in our study than dry (logging intensity:
mesic = 0.42; dry = 0.29), but did not increase isolation of mature
trees (mean density: mesic provenances = 1.28 trees ha)1; dry provenances = 1.23 trees ha)1). These factors will serve to reduce the
effective density of mahogany trees in mesic compared with dry
habitats and drive greater genetic impacts in the former compared
with the latter. Indeed, mesic provenances exhibit lower genetic
diversity (adult heterozygosity: mesic = 0.520; dry = 0.705; progeny
heterozygosity: mesic = 0.593; dry = 0.632) and increased inbreeding
(selfing: mesic = 6.2%; dry = 0.8%; biparental inbreeding: mesic = 0.198; dry = 0.171) compared with dry provenances. Reduced
population genetic diversity is also expected to lead to a decline in
genetic diversity of pollen clouds, which will be reflected as higher
correlated paternity values (correlated paternity: mesic = 0.310;
dry = 0.170). Third, it is plausible that mesic provenances have a
different pollinator community dominated by less mobile pollinators,
resulting in stronger mating pattern shifts with disturbance. If less
mobile pollinators were present in mesic provenances, we would
expect elevated levels of selfing in these provenances as less mobile
pollinators would spend more time in individual canopies, increasing
selfing rates (Ottewell et al. 2009) and correlated paternity in isolated
provenances (an effect we observe). While less mobile pollinators in
mesic provenances may be contributing to the observed mating
patterns, there is currently little information on the relative pollinator
abundance and behaviour in these different provenances. These
pollinator effects would synergistically emphasise the effect of small
population sizes (as outlined above) rather than counter them. Further
studies on pollinator dynamics in these regions is recommended as
follow-up work.
Independent of the mechanism, higher correlated paternity and
reduced effective pollen donor density are expected to lead to a shift
in reproductive assurance strategies and costs of inbreeding, as
occasional selfing would be favoured over reproductive failure (as
2012 Blackwell Publishing Ltd/CNRS
8 Martin F. Breed et al.
long as selfing was allowed), even if inbreeding depression was
significant (Herlihy & Eckert 2002; Barrett 2003; Kalisz et al. 2004).
However, ephemeral changes to pollen clouds are unlikely to result in
strong directional selection for increased mixed mating, although
directional or balancing selection for increased mixed mating may
indeed occur in chronically fragmented landscapes due to mate
limitations (Kennedy & Elle 2008; Winn et al. 2011). Levels of selfing
observed in isolated mesic provenances in our study (selfing
rate = 15%) are approximately twice that observed by Lemes et al.
(2007) in a study of a logged Brazilian population of S. macrophylla
(selfing rate = 7%). Lack of selfing in dry provenances from Central
America is likely due to higher effective pollen donor density, resulting
in preferential outcrossing and reproduction both being assured
without requiring an increase in selfing.
Mating system-fitness and heterozygosity-fitness relationships
We demonstrate that family-level reductions in outcrossing rate and
increases in correlated paternity have strong fitness costs (at least at the
stage of older saplings, which were used as a surrogate to measure
fitness in this study) for isolated mahogany trees and are provenancedependent. Families from mesic provenances experienced significant
selfing and greater variance in heterozygosity and, accordingly, we
observed a significant interlocus correlation of heterozygosity
(g2 = 0.033) and a correlation between inbreeding and fitness (Szulkin
et al. 2010). In contrast, families from dry provenances were almost
completely outcrossed and had low variance in heterozygosity,
therefore were not expected to express an interlocus correlation of
heterozygosity or a relationship between inbreeding and fitness.
Variation in pollen cloud diversity, as measured by correlated paternity,
appears to be a major factor reducing fitness in dry provenances.
Further work needs to be undertaken to assess the potential for fitness
effects to be expressed at life stages other than progeny growth (e.g.
pollination and germination). However, the ability for saplings to
quickly capture a site is a significant advantage for this pioneer species
and is a life stage that selection pressures are expected to be strong.
The few studies where correlated paternity has been examined in
conjunction with fitness warrant discussion here, but represent data
from single populations and fitness measurements taken over short
time-frames. Cascante et al. (2002) reported that Samanea saman trees
occurring in low density tended to have higher correlated paternity
than trees in high densities. This difference in mating pattern was
associated with poorer progeny vigour (measured 45 days postgermination). Fuchs et al. (2003) found that isolated Pachira quinata
trees had higher levels of correlated paternity than trees in continuous
forest. These isolated trees had lower flower-to-fruit set conversion
rates, which were hypothesised to be due to limited compatible pollen
received by isolated trees. In contrast, Rocha & Aguilar (2001) found
that isolated Enterolobium cyclocarpum trees had lower correlated
paternity than high-density trees, yet progeny from isolated trees
generally had lower vigour. The authors suggested that this result
might be due to fewer opportunities for selective abortions (i.e. less
pollen competition) in pasture trees because less pollen was received.
The difference in fitness could equally be explained by early-acting
maternal effects, particularly as seed from forest trees were significantly larger than those from pastures (Charlesworth 1988).
Greater pollen diversity should facilitate the acquisition of more
Ôgood genesÕ within a progeny array and is expected to be driven
2012 Blackwell Publishing Ltd/CNRS
Letter
through pollen competition by providing a forum to remove
individuals carrying deleterious recessive alleles (Charlesworth 1988;
Yasui 1998; Armbruster & Gobeille 2004). The findings of our study
fit this hypothesis, where, in dry provenances, greater pollen
competition (i.e. lower correlated paternity) potentially mitigates
fitness impacts of fragmentation. In contrast, mesic provenances
probably experienced far less pollen competition, as evidenced by high
correlated paternity, and exhibited a strong relationship between
inbreeding and fitness. Without examining pollen tube growth rates,
we cannot exclude the possibility that these inbreeding-fitness trends
were only the result of greater inbreeding in wet provenances.
CONCLUSIONS AND FUTURE DIRECTIONS
We link variation in family-level mating system patterns with fitness
observed over 5 years for S. macrophylla across 16 Central American
populations. Our results illustrate the first case of dramatic intraspecific variation in mating system and fitness responses to habitat
disturbance, as well as shifts in reproductive assurance and inbreeding
costs in different portions of the range of a species. Consequently, we
strongly recommend caution when extrapolating limited mating
system data to species-wide conclusions.
These results also have important ecological and applied implications (Fig. 1). Mahogany seed sourced from disturbed landscapes
throughout Central America is likely to be lower quality than seed
sourced from intact forest, leading to poorer outcomes for agroforestry and revegetation projects. This work also highlights the need to
protect remnant forest resources and develop appropriate provenance
sourcing strategies for restoration plantings (Broadhurst et al. 2008;
Sgrò et al. 2011).
ACKNOWLEDGEMENTS
The study was funded by the European Commission SEEDSOURCE
project (contract number 003708) awarded to AJL, ARC Linkage
project (LP110200805) awarded to AJL, and MFB was supported by
two National Climate Change Adaptation Research Facility travel
grants. Field sampling and the common garden experiment were
conducted by Centro Agronómico Tropical de Investigación y
Enseñanza (CATIE), Costa Rica. The authors thank Bert Harris for
statistical assistance, Patrice David for discussions on HFCs, and four
anonymous reviewers for their valuable comments and suggestions.
STATEMENT OF AUTHORSHIP
AJL and CMN designed the study, CMN collected field data and
samples, MGG, CMN and MFB generated genetic data, MFB, AJL
and KMO performed analyses. MFB wrote the first draft of the
manuscript and all authors contributed substantially to revisions. The
authors declare no conflicts of interest.
REFERENCES
Addelman, S. (1969). The generalized randomized block design. Am. Stat., 23, 35–
36.
Aguilar, R., Quesada, M., Ashworth, L., Herrerias-Diego, Y. & Lobo, J. (2008).
Genetic consequences of habitat fragmentation in plant populations: susceptible
signals in plant traits and methodological approaches. Mol. Ecol., 17, 5177–5188.
Letter
Armbruster, W.S. & Gobeille, R.D. (2004). Does pollen competition reduce the
cost of inbreeding? Am. J. Bot., 91, 1939–1943.
Barrett, S.C.H. (2003). Mating strategies in flowering plants: the outcrossing-selfing
paradigm and beyond. Phil. Trans. R. Soc. B., 358, 991–1004.
Beaumont, L.J., Pitman, A., Perkins, S., Zimmermann, N.E., Yoccoz, N.G. &
Thuiller, W. (2011). Impacts of climate change on the worldÕs most exceptional
ecoregions. Proc. Natl. Acad. Sci. USA, 108, 2306–2311.
Breed, M.F., Ottewell, K.M., Gardner, M.G. & Lowe, A.J. (2011). Clarifying climate
change adaptation responses for scattered trees in modified landscapes. J. Appl.
Ecol., 48, 637–641.
Broadhurst, L.M., Lowe, A., Coates, D.J., Cunningham, S.A., McDonald, M., Vesk,
P.A. et al. (2008). Seed supply for broadscale restoration: maximizing evolutionary potential. Evol. Appl., 1, 587–597.
Burnham, K.P. & Andersen, D.R. (2002). Model Selection and Multimodel Inference. 2nd
edn. Springer, New York.
Cascante, A., Quesada, M., Lobo, J.J. & Fuchs, E.A. (2002). Effects of dry tropical
forest fragmentation on the reproductive success and genetic structure of the tree
Samanea saman. Conserv. Biol., 16, 137–147.
Charlesworth, D. (1988). Evidence for pollen competition in plants and its relationship to progeny fitness: a comment. Am. Nat., 132, 298–302.
Charnov, E.L. (1976). Optimal foraging, the marginal value theorem. Theor. Popul.
Biol., 9, 129–136.
Colautti, R.I., Maron, J.L. & Barrett, S.C.H. (2009). Common garden comparisons
of native and introduced plant populations: latitudinal clines can obscure evolutionary inferences. Evol. Appl., 2, 187–199.
David, P., Pujol, B., Viard, F., Castella, V. & Goudet, J. (2007). Reliable selfing rate
estimates from imperfect population genetic data. Mol. Ecol., 16, 2474–2487.
Donohue, K. (2009). Completing the cycle: maternal effects as the missing link in
plant life histories. Phil. Trans. R. Soc. B., 364, 1059–1074.
Eckert, C.G., Kalisz, S., Geber, M.A., Sargent, R., Elle, E., Cheptou, P.-O. et al.
(2010). Plant mating systems in a changing world. Trends Ecol. Evol., 25, 35–43.
Fox, C.W. & Reed, D.H. (2010). Inbreeding depression increases with environmental stress: an experimental study and meta-analysis. Evolution, 65, 246–258.
Fuchs, E., Lobo, J. & Quesada, M. (2003). Effects of forest fragmentation and
flowering phenology on the reproductive success and mating patterns of the
tropical dry forest tree Pachira quinata. Conserv. Biol., 17, 149–157.
Gillies, A.C.M., Navarro, C., Lowe, A.J., Newton, A.C., Hernandez, M., Wilson, J.
et al. (1999). Genetic diversity in Mesoamerican populations of mahogany (Swietenia macrophylla), assessed using RAPDs. Heredity, 83, 722–732.
Goudet, J. (1995). FSTAT Version 1.2: a computer program to calculate F-statistics.
J. Hered., 86, 485–486.
Gribel, R., Gibbs, P.E. & Queiróz, A.L. (1999). Flowering phenology and pollination biology of Ceiba pentandra (Bombacaceae) in Central Amazonia. J. Trop.
Ecol., 15, 247–263.
Herlihy, C.R. & Eckert, C.G. (2002). Genetic cost of reproductive assurance in a
self-fertilizing plant. Nature, 416, 320–323.
Holdridge, L., Grenke, W., Hatheway, W., Liang, T. & Tosi, J. (1971). Forest
Environments in Tropical Life Zones: A Pilot Study. Pergamon Press, Oxford.
Kalisz, S., Vogler, D.W. & Hanley, K.M. (2004). Context-dependent autonomous selffertilization yields reproductive assurance and mixed mating. Nature, 430, 884–887.
Kennedy, B. & Elle, E. (2008). The reproductive assurance benefit of selfing:
importance of flower size and population size. Oecologia, 155, 469–477.
Lamb, F.B. (1966). Mahogany of Tropical America: Its Ecology and Management. The
University of Michigan Press, Ann Arbor.
Lemes, M.R., Brondani, R.P.V. & Grattapaglia, D. (2002). Multiplexed systems of
microsatellite markers for genetic analysis of mahogany, Swietenia macrophylla King
(Meliaceae), a threatened Neotropical timber species. J. Hered., 93, 287–291.
Lemes, M.R., Grattapaglia, D., Grogan, J., Proctor, J. & Gribel, R. (2007). Flexible
mating system in a logged population of Swietenia macrophylla King (Meliaceae):
implications for the management of a threatened neotropical tree species. Plant
Ecol., 192, 169–179.
Lowe, A.J., Jourde, B., Breyne, P., Colpaert, N., Navarro, C., Wilson, J. et al. (2003).
Fine-scale genetic structure and gene flow within Costa Rican populations of
mahogany (Swietenia macrophylla). Heredity, 90, 268–275.
Lowe, A.J., Boshier, D., Ward, M., Bacles, C.F.E. & Navarro, C. (2005). Genetic
resource impacts of habitat loss and degradation; reconciling empirical evidence
and predicted theory for neotropical trees. Heredity, 95, 255–273.
Reproductive assurance shifts in mahogany 9
Murawski, D.A. & Hamrick, J.L. (1991). The effect of the density of flowering
individuals on the mating systems of nine tropical tree species. Heredity, 67, 167–
174.
Navarro, C. & Hernandez, G. (2004). Progeny test analysis and population differentiation of Mesoamerican mahogany (Swietenia macrophylla). Agronomia Costarricaense, 28, 37–51.
Navarro, C., Wilson, J., Gillies, A. & Hernandez, M. (2002). A new Mesoamerican
collection of big-leaf mahogany. In: Big Leaf Mahogany: Genetics, Ecology, and
Management (eds Lugo, A, Figueroa, J & Alayón, M). Springer, New York, pp.
103–114.
Navarro, C., Cavers, S. & Lowe, A. (2011). Seed sourcing recommendations for
forest restoration: impacts of tree isolation on progeny performance of cedar and
mahogany in the neotropics. In: Restoring Degraded Landscapes with Native Species in
Latin America (eds Montagnini, F & Finney, C). Nova Science Publishers, New
York, pp. 51–61.
Novick, R.R., Dick, C.W., Lemes, M.R., Navarro, C., Caccone, A. & Bermingham,
E. (2003). Genetic structure of Mesoamerican populations of big-leaf mahogany
(Swietenia macrophylla) inferred from microsatellite analysis. Mol. Ecol., 12, 2885–
2893.
Oosterhout, C.V., Hutchinson, W.F., Wills, D.P.M. & Shipley, P. (2004). MICROCHECKER: software for identifying and correcting genotyping errors in microsatellite data. Mol. Ecol. Notes, 4, 535–538.
Ottewell, K.M., Donnellan, S.C., Lowe, A.J. & Paton, D.C. (2009). Predicting
reproductive success of insect- versus bird-pollinated scattered trees in agricultural landscapes. Biol. Conserv., 142, 888–898.
Ritland, K. (2002). Extensions of models for the estimation of mating systems using
n independent loci. Heredity, 88, 221–228.
Rocha, O.J. & Aguilar, G. (2001). Reproductive biology of the dry forest tree
Enterolobium cyclocarpum (guanacaste) in Costa Rica: a comparison between trees
left in pastures and trees in continuous forest. Am. J. Bot., 88, 1607–1614.
Sgrò, C.M., Lowe, A.J. & Hoffmann, A.A. (2011). Building evolutionary resilience
for conserving biodiversity under climate change. Evol. Appl., 4, 326–337.
Sork, V.L. & Smouse, P.E. (2006). Genetic analysis of landscape connectivity in tree
populations. Landsc. Ecol., 21, 821–836.
Szulkin, M., Bierne, N. & David, P. (2010). Heterozygosity-fitness correlations:
a time for reappraisal. Evolution, 64, 1202–1217.
Ward, M., Dick, C.W., Gribel, R. & Lowe, A.J. (2005). To self, or not to self...
a review of outcrossing and pollen-mediated gene flow in neotropical trees.
Heredity, 95, 246–254.
White, G.M., Boshier, D.H. & Powell, W. (2002). Increased pollen flow counteracts
fragmentation in a tropical dry forest: an example from Swietenia humilis Zuccarini.
Proc. Natl. Acad. Sci. USA, 99, 2038–2042.
Winn, A.A., Elle, E., Kalisz, S., Cheptou, P.-O., Eckert, C.G., Goodwillie, C. et al.
(2011). Analysis of inbreeding depression in mixed-mating plants provides evidence for selective interference and stable mixed mating. Evolution, 65, 3339–
3359.
Yasui, Y. (1998). The Ôgenetic benefitsÕ of female multiple mating reconsidered.
Trends Ecol. Evol., 13, 246–250.
SUPPORTING INFORMATION
Additional Supporting Information may be downloaded via the online
version of this article at Wiley Online Library (www.ecologyletters.com).
As a service to our authors and readers, this journal provides
supporting information supplied by the authors. Such materials are
peer-reviewed and may be re-organised for online delivery, but are not
copy edited or typeset. Technical support issues arising from
supporting information (other than missing files) should be addressed
to the authors.
Editor, David Hooper
Manuscript received 9 December 2011
First decision made 14 January 2012
Manuscript accepted 2 February 2012
2012 Blackwell Publishing Ltd/CNRS