Received Date : 06-Nov-2012

Accepted Article
Revised Date : 25-Jul-2013

Accepted Date : 14-Aug-2013

Article type : Original Article

Title: Low plant density enhances gene dispersal in the Amazonian understory herb

Heliconia acuminata

Authors: Marina Corrêa Côrtes1,8, María Uriarte1, Maristerra R. Lemes2,3, Rogério Gribel2,3,

W. John Kress4, Peter E. Smouse5, Emilio M. Bruna6,7,8

Addresses:

1. Department of Ecology, Evolution and Environmental Biology, Columbia University,

New York, NY, USA.

2. Laboratório de Genética e Biologia Reprodutiva de Plantas, Instituto Nacional de

Pesquisas da Amazônia, Manaus, AM, Brazil.

3. Instituto de Pesquisas Jardim Botânico do Rio de Janeiro, Rio de Janeiro, RJ, Brazil.

4. Department of Botany, National Museum of Natural History, MRC-166, Smithsonian

Institution, Washington DC, USA.

5. Department of Ecology, Evolution and Natural Resources, Rutgers University, New

Brunswick, NJ, USA.

This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process, which may
lead to differences between this version and the Version of Record. Please cite this article
as doi: 10.1111/mec.12495
This article is protected by copyright. All rights reserved.
 6. Department of Wildlife Ecology and Conservation, University of Florida, Gainesville,
Accepted Article
FL, USA.

7. Center for Latin American Studies, University of Florida, Gainesville, FL, USA.

8. Biological Dynamics of Forest Fragments Project, Instituto Nacional de Pesquisas da

Amazônia and Smithsonian Tropical Research Institute, Manaus, AM, Brazil

Keywords: Hummingbird, manakin, pollination, reproductive dominance, seed dispersal,

thrush

Corresponding author: Marina Corrêa Côrtes ([email protected])

Phone: (917) 749 2090; Fax: (212) 854 8188

Running title: Gene dispersal in Heliconia

Abstract

In theory, conservation genetics predicts that forest fragmentation will reduce

gene dispersal, but in practice, genetic and ecological processes are also dependent on

other population characteristics. We used Bayesian genetic analyses to characterize

parentage and propagule dispersal in Heliconia acuminata L. C. Richard (Heliconiaceae), a

common Amazonian understory plant that is pollinated and dispersed by birds. We

studied these processes in two continuous forest sites and three 1-ha fragments in Brazil’s

Biological Dynamics of Forest Fragments Project. These sites showed variation in the

This article is protected by copyright. All rights reserved.

 density of H. acuminata. Ten microsatellite markers were used to genotype flowering
Accepted Article
adults and seedling recruits and to quantify realized pollen and seed dispersal distances,

immigration of propagules from outside populations, and reproductive dominance among

parents. We tested whether gene dispersal is more dependent on fragmentation or

density of reproductive plants. Low plant densities were associated with elevated

immigration rates and greater propagule dispersal distances. Reproductive dominance

among inside-plot parents was higher for low density than for high density populations.

Elevated local flower and fruit availability is probably leading to spatially more proximal

bird foraging and propagule dispersal in areas with high density of reproductive plants.

Nevertheless, genetic diversity, inbreeding coefficients and fine-scale spatial genetic

structure were similar across populations, despite differences in gene dispersal. This result

may indicate that the opposing processes of longer dispersal events in low-density

populations versus higher diversity of contributing parents in high-density populations

balance the resulting genetic outcomes and prevent genetic erosion in small populations

and fragments.

Introduction

Tropical deforestation and habitat fragmentation are proceeding at unprecedented

rates (FAO 2011), with ecological and genetic consequences for plant populations (Aguilar

et al. 2008; DiBattista 2008; Laurance et al. 2002). Forest fragmentation is thought to

disrupt gene flow by reducing dispersal distances and immigration of propagules among

populations (Ouborg et al. 2006; Young et al. 1996), enhancing fine-scale spatial genetic

This article is protected by copyright. All rights reserved.

 structure and reducing effective population sizes within populations (Wang et al. 2011;
Accepted Article
Young et al. 1996). Low effective population sizes can change plant mating patterns,

increasing genetic drift and inbreeding (Aguilar et al. 2008; Eckert et al. 2010). Effective

population size is reduced by high reproductive dominance among adults, increasing the

level of correlated parentage (Breed et al. 2012a; Robledo-Arnuncio et al. 2004) and

source-biased limitation (García & Grivet 2011; Jordano & Godoy 2002), resulting in fewer

individuals successfully contributing genes to the next generation (Moran & Clark 2012b;

Young & Pickup 2010).

Over the long term, these processes are predicted to decrease population genetic

diversity within a given population and increase genetic divergence among isolated

patches (Aguilar et al. 2008; DiBattista 2008; Vranckx et al. 2012). Empirical studies,

however, have often found extensive gene flow via pollen (Dick et al. 2003; Lander et al.

2010; White et al. 2002) or seed (Bacles et al. 2006; Kamm et al. 2009) in forest fragments.

Such conflicting evidence, known as the “paradox of forest fragmentation genetics”

(Kramer et al. 2008), is hardly surprising, given that neither the plants themselves nor

their animal propagule vectors exhibit uniform responses to habitat fragmentation (Hobbs

& Yates 2003; Watling & Donnelly 2006).

Including the ecological characteristics of plant and animal populations in studies

of fragmentation genetics should help to elucidate the mechanisms that determine

propagule flow. Although the variation of plant abundance across fragmented populations

This article is protected by copyright. All rights reserved.

 is regarded as an important factor impacting levels of inbreeding and gene flow, it has
Accepted Article
seldom been explicitly related to the genetic consequences of disturbance (Honnay &

Jacquemyn 2007). Low population density can affect mating and pollen dispersal, by

limiting pollinator and seed dispersal visitation and increasing inbreeding rates (Eckert et

al. 2010; Ghazoul 2005). Conversely, pollen dispersal distances may increase in low-

density populations, by virtue of high mobility of pollinators in search of food sources

(Byrne et al. 2007; Llorens et al. 2012).

Plant population-level characteristics that influence pollen and seed dispersal, such

as adult density or flower and seed production, are often affected by landscape

modification (Eckert et al. 2010; Herrera et al. 2011; Kolb 2008). In that context, an

increasing number of studies have conducted paternity or maternity analysis, while taking

into consideration the effects of changing plant and animal abundance or behavior across

sites (Byrne et al. 2007; Dick et al. 2003; García et al. 2009; Lander et al. 2010). Yet,

studies of animal-vectored propagule flow across fragmented landscapes are still

uncommon (but see Kamm et al. 2009).

The direct assessment of pollen and seed dispersal permits more realistic inference

about the evolutionary consequences of fragmentation for natural plant populations

(Bacles & Jump 2011; Sork & Smouse 2006) and provides insights into processes that drive

contemporary gene flow, rather than relying on indirect comparisons of extant genetic

variation, as a means of assessing historical patterns of gene flow (Meagher 2010; Oddou-

This article is protected by copyright. All rights reserved.

 Muratorio & Klein 2008). Highly polymorphic molecular markers (e.g., microsatellites),
Accepted Article
coupled with parentage analyses, can provide an accurate assessment of contemporary

gene flow (Ashley 2010). Researchers using highly polymorphic markers for parentage

analysis can employ exclusion methods, in which candidate adult plants that do not share

alleles with the offspring are eliminated as parental candidates for that particular seedling.

Alternatively, they assign the most likely single or pair of parents, based on log-likelihood

ratios (Jones & Ardren 2003; Jones et al. 2010). These classic approaches, however, can

provide poor results if polymorphism is insufficient or if the presence of null alleles or

genotype mistyping errors are not taken into account (Chybicki & Burczyk 2010; Jones et

al. 2010). Moreover, researchers are often less interested in the parentage allocation per

se than in population-level processes such as seed and pollen dispersal distances (Moran

& Clark 2011; Moran & Clark 2012a; Oddou-Muratorio & Klein 2008). In contrast, full

probability models can jointly estimate population-level parameters and parentage, while

incorporating both genetic and ecological data, such as spatial location and reproductive

status of individual plants (Burczyk et al. 2006; Hadfield et al. 2006; Moran & Clark 2011).

Here, we use a hierarchical Bayesian approach to quantify the contributions of

pollen and seed movement to gene dispersal in Heliconia acuminata L. C. Richard

(Heliconiaceae), an understory herb pollinated by hummingbirds (Bruna et al. 2004),

whose seed is dispersed by manakins and thrushes (Uriarte et al. 2011). The species is

native to central Amazonia and the Guyanas and has been the subject of a long-term

demographic study in an experimentally fragmented landscape (Bruna & Kress 2002;

This article is protected by copyright. All rights reserved.

 Gascon & Bierregaard 2001). Using this model system, we ask whether fragmentation and
Accepted Article
population density influence: (a) realized pollen and seed dispersal distances, (b)

immigration of propagules from outside populations, and (c) reproductive dominance

among parents.

We tested two contrasting hypotheses. First, assuming the traditional prediction

that forest fragmentation interrupts movement of animals and plant propagules, we

hypothesized that forest fragments would experience less gene dispersal than continuous

forest sites. Alternatively, if animal movement is driven by the availability of food

resources and if dispersers are able to move freely across the landscape, we hypothesized

that the local density of reproductive plants would exert a greater influence on gene

dispersal. To better understand the resulting patterns of gene dispersal, we also

characterized the genetic diversity, inbreeding and fine-scale spatial genetic structure of

seedlings and reproductive plants. We expected that reduced gene dispersal, low

immigration rates and high reproductive dominance would increase spatial aggregation

and mating of related plants. Ultimately, these patterns should result in stronger fine-

scale spatial genetic structure, higher inbreeding coefficients and decrease genetic

diversity.

This article is protected by copyright. All rights reserved.

 Materials and Methods
Accepted Article
Study site and system

The study was conducted at the Biological Dynamics of Forest Fragments Project

(BDFFP), located 70 km north of Manaus, Brazil (2o 30’ S, 60’ W, Data S1, Supporting

information). The BDFFP is a 1000 km2 landscape comprised of forest fragment reserves,

ranging in size from 1 - 100 ha, and continuous forest sites. The fragments, experimentally

created for scientific studies, were isolated from 1980 - 1984 by clear-cutting the trees

surrounding the patches and, in some cases, burning the felled trees (Gascon &

Bierregaard 2001). Subsequently, secondary forests have colonized and developed in the

intervening clearcuts (Mesquita et al. 2001). Studies comparing bird capture rates before

and after the isolation of the BDFFP’s fragments suggest that spatial structure of the

landscape is likely to affect the abundance and movement of birds, including pollinators

and seed dispersers of Heliconia, and that these effects can be expected to vary spatially

and temporally (Stouffer & Bierregaard 1995a, b; Stouffer & Bierregaard 1996).

Heliconia acuminata has been the subject of a comprehensive demographic study

since 1998 (Bruna 2003; Bruna & Kress 2002). Thirteen 0.5 ha plots (50 x 100 m) were

established in continuous forest (N = 6 sites) and fragments (N = 7 fragments); all H.

acuminata individuals in these plots were tagged, mapped, and censused annually. In the

present study we used two plots in continuous forest and three in 1 - ha fragments. The

approximate distance between fragment edges to the nearest forest patch is 100 m (Data

S1, Supporting information).

This article is protected by copyright. All rights reserved.

 Heliconia acuminata has a scattered distribution in the forest understory (Bruna &
Accepted Article
Ribeiro 2005) and exhibits limited vegetative spread (E.M. Bruna and W.J. Kress,

unpublished data). It is one of the most abundant understory plants at this site (E. Bruna &

W.J. Kress, personal observation), though its local density can range from approximately

200-1600 plants/hectare (Bruna 2003). Plants produce 20-25 flowers per inflorescence,

with individual flowers opening on successive days and for one day each. This reduces the

probability of intra-inflorescence pollen transport by the pollinating hummingbirds

(Dobkin 1984, 1987).

Heliconia acuminata is monoecious and functionally self-incompatible – in

experiments designed to asses self-compatibility, autogamy was observed only in a small

proportion of flowers in which self-pollen was manually placed on stigmas (16%). In

contrast, in treatments where inflorescences were bagged and flowers were not

manipulated no fruits were produced (E. M. Bruna, unpubl. data). The primary pollinators

of H. acuminata are the hermit hummingbirds Phaethornis superciliosus and P. bourcieri,

which “trapline” from one inflorescence to the next, rather than establishing and

defending a territory. They persist in both primary and secondary forests (Stouffer &

Bierregaard 1995a). Evidence suggests that they may forage over large distances and

move through a variety of habitats (Stouffer & Bierregaard 1995a), although no detailed

information about their movement patterns is available. Visitation rates by pollinators are

low (median = 0.182 visits/hour per plant), with hummingbirds failing to visit 28% of the

observed plants (Bruna et al. 2004). These low visitation rates, coupled with the results of

This article is protected by copyright. All rights reserved.

 hand-pollination studies, suggest fruits resulting from self-pollination are extremely
Accepted Article
unlikely.

The primary dispersers of the seeds at this site are the white-collar thrush (Turdus

albicollis), the thrush-like-manakin (Schiffornis turdinus), and several species of manakins

(Pipra erythrocephala, P. pipra, Lepidothrix serena, Corapipo gutturalis). Manakins

disperse seeds an average of 19 m from maternal plants at this site, while thrushes have

an average dispersal distance of 24 m (Uriarte et al. 2011). About 90% of ripe fruits were

removed, and that rate did not vary across the landscape, nor was it affected by forest

fragmentation or neighborhood density of reproductive plants (Uriarte et al. 2011).

For parentage analysis, we collected samples of leaves from mapped seedlings

instead of seeds, because we were interested in the realized pollen and seed dispersal,

which is the ultimate result of both successful mating and seed deposition (Meagher &

Thompson 1987). Here, we define gene dispersal as the combined movement of pollen

and seeds that successfully transitions to seedlings and therefore changes the spatial

distribution of genes in the population of interest. In 1999, we collected leaf samples of all

reproductive plants (potential parents) and all seedlings in the five 0.5 ha plots. In 2009,

we resampled the same plots, collecting leaf tissues of new seedlings that had recruited

since 2000 and new reproductive plants that had flowered since 2000. Seedlings ranged in

age from a few months (if recruited in 1999 or 2009) to about nine years (if recruited in

2000). In 2009, we also recollected 123 surviving plants that had been genotyped from the

1999 collections to confirm that genotyping was consistent, independent of sample age,

This article is protected by copyright. All rights reserved.

 storage and DNA isolation methods. To increase the likelihood of determining the
Accepted Article
potential parents of seedlings inside the 0.5 ha plot, we also mapped and collected leaf

tissue from all adults with inflorescences of current or past reproduction in a 20 m buffer

around each plot (no seedlings were sampled in this buffer zone). Because plants in the

buffer zones were not part of the long term demographic census, we relied on the

observation of old inflorescences as a measure of current and past reproduction.

Inflorescences can remain attached to the plant for more than a year, so it is relatively

easy to identify potential reproductive individuals (E. Bruna and P. Rubim, personal

observations). The inclusion of the 20 m buffer increased the sampling area for

reproductive plants from 0.5 ha to 1.26 ha.

Leaf tissue was either frozen in liquid nitrogen or dried in silica gel and then stored

at −80°C. Total genomic DNA was manually extracted, using a modified CTAB extraction

method (Ferreira & Grattapaglia 1998) or automatically, by using a AutoGenprep 965

robot (AutoGen, Inc). Ten nuclear microsatellite markers that had been previously

developed for H. acuminata were used to genotype adults and recruits; the PCR protocols

and genotyping procedures were described in Côrtes et al. (2009). Genotyping error rates

resulting from mistyping and drop-out were calculated by regenotyping 23% of the

individuals. Across loci, the mistyping rate was 2.9% (1.4 – 5.1% per locus), and the drop-

out rate was 2.8% (range 0.9 – 6.6% per locus). These errors, although relatively low on a

per locus basis, could result in erroneous parentage assignments. We incorporated these

rates in the Bayesian model to account for genotyping uncertainty.

This article is protected by copyright. All rights reserved.

 Gene dispersal model
Accepted Article
We use the Bayesian approach developed by Moran & Clark (2011) to estimate

pedigree and realized pollen and seed dispersal. One advantage of this model is that it

permits the inclusion of prior information and multiple sources of uncertainty associated

with genotyping and specific ecological processes, which results in more realistic

parameter estimates (Jones et al. 2010; Moran & Clark 2011). A second advantage is that

it incorporates the contribution of plants located outside the sampled area, so that

immigration is also used to model the dispersal kernel. Parentage analysis of monoecious

species usually assumes that the nearest assigned parent is the mother (Bacles et al.

2006), whereas the current approach assumes that both maternity and paternity are

assigned with uncertainty, given separate pollen and seed dispersal kernels (Moran &

Clark 2011; Moran & Clark 2012a, see Data S2, Supporting information). The pedigree and

pollen and seed dispersal parameters are jointly estimated, based on offspring and adult

genotypes, two types of genotyping error, distances between plants, and plant phenology

(Data S2, Supporting information), as follows (Moran & Clark 2011):

(Eqn. 1)

where P is the pedigree; up and us are the pollen and seed dispersal parameters; GO is the

observed genotype of all individuals for locus l; d is the pairwise distance between

This article is protected by copyright. All rights reserved.

 individuals; c and f are the weight factors represented by the number of flowers (i.e.,
Accepted Article
pollen production of paternal plant i’) and number of seeds (i.e., fecundity of maternal

plant i), respectively; r is the plant-seedling temporal compatibility, indicating whether a

seedling k recruited after mother i flowered (1 or 0); s is the flowering synchronization to

assure that plants are able to mate by indicating whether flowering of plant i’ and i occur

in the same year (1 or 0); e1 and e2 are the mistyping and dropout errors of locus l; and

p(up) and p(us) are the priors related to the dispersal parameters. Selfing was not allowed

in the model, so the same plant cannot be simultaneously the mother and father of the

same seedling.

Flower production (c) was measured as the total number of flowers each individual

plant produced over the study period, and is the product of the number of inflorescences

and the average number of flowers per inflorescence. Fecundity (f) was calculated as the

product of the maturation rate from flower to ripe fruits (from Uriarte et al. 2011) and the

number of seeds per fruits. The maturation rates used for calculating the number of seeds

(fecundity) were 0.15 for CF1, 0.08 for CF2 and 0.5 for F1, F2 and F3 (Uriarte et al. 2011,

M.T.B. da Silva, unpublished data). Each fruit produces two seeds on average (1.9 ± 0.02

seeds/fruit, mean ± SE, n=873 fruits, E. Bruna unpublished data).

The distance kernel for both pollen and seed dispersal is given by the 2D - t

function (Clark et al. 1999), which takes the form:

(Eqn. 2)

This article is protected by copyright. All rights reserved.

 where parameters are as in Eqn. 1 and separate u’s are estimated for pollen and seeds.
Accepted Article
We chose the 2D - t function, instead of other commonly used functional forms, because it

allows for higher probabilities of both short and long distance dispersal, relative to a

normal distribution (Clark et al. 1999; Moran & Clark 2012a). To compare pollination and

seed dispersal distances, we used the mode, rather than the mean, to characterize the

most frequent dispersal events (Clark et al. 1999).

Pedigree and other parameters in Eqns. 1-2 were estimated using a Gibbs sampler,

using parentage probabilities and ecological data. In the resulting pedigree, each seedling

is assigned to the pair of parents that presented the highest proportional allocation in the

50,000 simulations conducted within the model. Many times, plants within the plot are

not ecologically and genetically likely to be a seedling’s parent. In this case, the seedling is

more likely assigned to a hypothetical plant (located outside the 1.26 ha sampled plot),

which conveys the rate of immigration. It is possible that immigration rates were

overestimated, because some reproductive plants died before they could be genotyped

(Table 1). Model implementation follows the code proposed by Moran and Clark (2011).

Implementation and information on the effects of different priors and density of

hypothetical parents on posteriors are provided in the Supporting information (Data S2).

Reproductive dominance

Reproductive dominance was investigated using the pedigree recovered from the

gene dispersal model and considering only the seedlings that had at least one parent

This article is protected by copyright. All rights reserved.

 identified within the 1.26 ha plot. Reproductive dominance is a measure of the genetic
Accepted Article
contribution of reproductive plants to the seedlings in the population, either via pollen or

seeds. It was calculated using the probability of parental identity (PPaI) metric (Data S3,

Supporting information). PPaI is analogous to the probability of paternal identity (Smouse

& Robledo-Arnuncio 2005) and maternal identity (Grivet et al. 2005), and measures the

probability that two offspring randomly sampled from a population share the genotype of

either a father or mother. PPaI was estimated using a variation of the unbiased r -

estimator R0 (Data S3, Supporting information), and ranges between 0 (seedlings do not

share any parental genotype) and 1 (seedlings share genotypes of both parents).

Confidence intervals were calculated by extracting PPaI values of 1000 bootstrapped

samples of the actual sample.

Genetic diversity, inbreeding and fine-scale spatial genetic structure

To evaluate if differences in gene dispersal metrics were reflected in the genetic

make-up of the populations, we also analyzed the genetic diversity, inbreeding coefficient,

and fine-scale spatial genetic structure. Genetic diversity of seedlings and reproductive

plants of each population was characterized by the unbiased expected heterozygosity

(UHe) and average number of alleles per locus (Na). These metrics were calculated using

GenAlEx (Peakall & Smouse 2006). The inbreeding coefficient and fine-scale spatial genetic

structure were characterized using the Loiselle kinship estimator (Loiselle et al. 1995)

using SPAGEDi (Hardy & Vekemans 2002). The inbreeding coefficient (Fis) was measured as

the intra-individual kinship coefficient. Because H. acuminata is effectively non-selfing, the

This article is protected by copyright. All rights reserved.

 coefficient represents the effect of bi-parental inbreeding. The fine-scale spatial genetic
Accepted Article
structure was quantified by the sp-statistic, which measures the rate of decay of pairwise

kinship with the logarithm of the distance between individuals. The sp –statistic is

calculated as -bF/(1 - F1), where bF is the slope of the regression between pairwise kinship

coefficients and distance, and F1 is the average pairwise kinship coefficient between

neighbors (defined here as individuals within 10 m-radius for seedlings and 15 m for

adults, chosen to optimize the proportion of participating individuals in the interval)

(Vekemans & Hardy 2004). Significance of the inbreeding coefficients was tested by

permuting genes among all individuals 2000 times, whereas significance of sp-statistic was

tested by permuting the spatial location of individuals 2000 times to obtain a frequency

distribution of the regression slope (bF) under the null hypothesis that kinship is not

correlated with distance. To evaluate whether plant density would influence the

estimated parameters and to compare these values across populations, we estimated a

confidence interval around the observed inbreeding and sp-statistic by running the

analysis on 2000 bootstrapped samples of size equal the number of seedlings and adults in

the least dense population.

Results

Propagule dispersal distances and immigration.

Absolute number and density of flowering plants was 5 to 14 times greater in CF1

than for the other areas (Table 1), although the number of flowering plants varied across

years (Data S4, Supporting information). Plant density was the dominant factor associated

This article is protected by copyright. All rights reserved.

 with gene dispersal. Regardless of fragmentation status (forest fragments vs. continuous
Accepted Article
forest), realized pollen and seed dispersal distances were greater in low density

populations than in the dense continuous forest population CF1 (Fig. 1). Modal distances

were almost four-fold shorter in CF1 (16 and 14 m for pollination and seed dispersal,

respectively) than for the low-density populations (average range 37 – 64 m), leading to

more restricted dispersal in CF1 (Fig. 1).

Immigration of both pollen and seeds was also different among plots (χ2 = 163.85,

df= 4, p-value < 0.001), presenting higher rates in low-density populations. Immigration

rate was highest for CF2, with only one parental pair assignment within the sampled plot,

and lowest in CF1, with only 2% of the seedlings generated from parent pairs located

outside the plot (Table 2). Fragments experienced intermediate rates of propagule

immigration, with 13-23% of the seedlings with both parents located outside plots (Table

2).

Reproductive dominance

On average, 70% (range 62-91%) of the reproductive plants contributed genes (via

pollen or seed) to seedlings inside the plot, with the exception of F1, in which more than

90% of the reproductive plants contributed genes (Table 2). For the seedlings that had

either the father or mother inside the plots, the probabilities that they shared a parent

(PPaI-values) were always smaller than 9%, indicating that multiple plants contributed to

the genotypes of seedlings, with weak dominance of few reproductive plants (R0, Fig. 2).

This article is protected by copyright. All rights reserved.

 CF2 exhibited the highest reproductive dominance, although presenting very large
Accepted Article
confidence intervals around the value of PPaI. CF1 was the population with the most even

genetic contribution of adults to seedlings (R0 = 0.0068, Fig. 2), with PPaI an order of

magnitude smaller than that for CF2 (R0 = 0.0905, Fig. 2). The populations in fragments

presented similar PPaI values , with overlapping confidence intervals (average R0 = 0.0307,

Fig. 2).

Genetic diversity, fine-scale spatial genetic structure, and inbreeding

Expected heterozygosity and numbers of alleles were consistently similar across

populations and between seedlings and reproductive adults, with a total heterozygosity

average of 0.673 and 7.9 alleles per locus (Table 3). Fine-scale spatial genetic structure

was significant but weak in all populations of seedlings (ranging from 0.0025 to 0.0142)

(Table 3). In fact, autocorrelograms of average pairwise kinship plotted for each distance

interval showed that in most cases, kinship values are within the confidence interval

envelopes and that only CF1, F2 and F3 presented significant positive kinship values at

short distances (Fig. 3). For reproductive adults, only CF1 and F3 yielded significant sp-

statistics (0.0041 and 0.0125, respectively) (Table 3), with positive average kinship values

in the first distance interval (< 15 m) (Fig. 3). Sp-values, however, did not vary significantly

across populations, as indicated by the overlapping confidence intervals (Data S5, Fig. S5,

Supporting information).

This article is protected by copyright. All rights reserved.

 Inbreeding coefficients of seedlings and of adults were significantly different from
Accepted Article
random mating in all populations, with CF1 presenting low inbreeding coefficients of 0.038

and 0.025, compared to a range of 0.051–0.131 in the other populations (Table 3). Despite

differences in the values of inbreeding and fine-scale spatial structure, the confidence

intervals of bootstrapped samples overlapped, indicating that values are not significantly

different among populations (Data S5, Fig. S5, Supporting information).

Discussion

We examined gene dispersal across populations using a comprehensive approach

that jointly exploits genetic and ecological data – including data on fecundity and

phenology – while incorporating multiple sources of uncertainty and the contribution of

outside parents via immigration of propagules. We found that populations with low

density of reproductive plants presented higher realized propagule dispersal distances,

immigration, and reproductive dominance than did the high-density population.

Therefore, our results corroborate the hypothesis that propagule dispersal is more

strongly associated with the density of reproductive plants than with the potential effects

of fragmentation on pollinators and seed dispersers. We found that genetic diversity,

inbreeding and fine-scale spatial genetic structure were not significantly different across

the landscape, which may indicate that populations can sustain similar levels of genetic

patterns through distinct processes, with longer dispersal events in low density

populations and higher diversity of contributing parents in high density populations.

This article is protected by copyright. All rights reserved.

 Fragmentation, plant density and gene dispersal
Accepted Article
Counter to the prediction that forest fragmentation disrupts pollination and seed

dispersal, we found gene dispersal to be enhanced in populations with low density of

reproductive plants. This means that the detrimental effects of fragmentation on genetic

processes are not applicable in all situations. For instance, the isolation of fragments may

not always hinder animal and propagule movement (Hadley & Betts 2011; Kramer et al.

2008). In fact, the abundance of pollinators at BDFFP did not change after landscape

fragmentation (Stouffer & Bierregaard 1995a), and frugivorous birds regularly visited small

forest fragments, as the cleared matrix experienced secondary succession (Stouffer &

Bierregaard 2007). High gene flow via pollen movement beyond boundaries of isolated

forest fragments has been recorded elsewhere for animal-pollinated plants (Aldrich &

Hamrick 1998; Dick et al. 2003; Kamm et al. 2009; Lander et al. 2010; Nason & Hamrick

1997; White et al. 2002). In contrast, gene flow via seed dispersal by animals beyond

fragment boundaries has received scant attention, and results to date are inconsistent.

For instance, Hanson et al. (2007) assigned 14 out of 23 seed endocarps of Dipteryx

panamensis to mothers outside fragments, demonstrating that bat-mediated dispersal can

connect isolated patches. Conversely, parentage analysis of Araucaria angustifolia in Brazil

showed that seed immigration into forest fragment was absent, possibly the result of

dispersal by gravity and limited dispersal by secondary dispersers (Bittencourt & Sebbenn

2007).

This article is protected by copyright. All rights reserved.

 Propagule dispersal distances in Heliconia acuminata exhibited a stronger
Accepted Article
association with density of reproductive plants than with fragmentation per se. Average

pollen and seed dispersal distances for the low-density continuous forest (CF2) were more

similar to those of fragments than to that of the high-density continuous forest site (CF1).

Both theoretical and empirical studies predict that pollinators will spend more time

visiting flowers within the same plant or forage on the nearest neighbor when plant

density is low, ultimately resulting in shorter pollination distances (see Ghazoul (2005) and

references therein). Paternity analyses across fragmented landscapes, however, have

found extensive gene flow in populations with reduced plant density (Breed et al. 2012b;

Llorens et al. 2012; Stacy et al. 1996). As appears to be the case in our study, this is

frequently found to be due to more localized foraging in dense patches. Our seed

dispersal findings also corroborate results from a few seed dispersal studies showing that

increasing plant aggregation and abundance of fleshy fruits decreases seed dispersal

distances, as birds concentrate foraging in areas of higher fruit density (Herrera et al.

2011; Morales & Carlo 2006).

In sites with sparsely distributed reproductive plants, birds must travel farther and

cover larger areas to meet their energetic requirements (Hadley & Betts 2011; Khamcha et

al. 2012). This pattern is particularly marked for specialist pollinators (Ahmed et al. 2009)

and frugivorous birds with narrow dietary preferences (Kwit et al. 2004). The abundance

of hummingbirds in the study site is higher between January and April, the period when H.

acuminata is flowering (Stouffer & Bierregaard 1996), suggesting that the birds are

This article is protected by copyright. All rights reserved.

 tracking critical nutritional resources over time. Experiments in captivity and lab analyses
Accepted Article
have also shown that the high lipid content of the fruits of H. acuminata make them a

preferred food resource for manakins (S. Hashimoto, unpublished data), suggesting that

frugivorous birds may also track fruiting across the landscape.

Immigration and reproductive dominance

Both immigration and reproductive dominance also exhibited a stronger

association with plant density than fragmentation. The population with the lowest plant

density (CF2) had the highest reproductive dominance, with 60% of all reproductive plants

contributing to the genetic pool and two individuals contributing more than 30% to the

seedling genotypes. Immigration of propagules was also highest for this plot, with 62% of

the seedlings originating from parents outside the plot. This high immigration rate is likely

to increase effective population size of the recipient population and dilute overall

reproductive dominance of the population. At the other extreme, the population with the

highest density of reproductive plants (CF1) had the most diverse array of parents

contributing to the seedling genetic pool.

Despite these differences in reproductive dominance among populations, the

values of PPaI were generally low. This indicates that contributions of reproductive plants

to seedling genotypes were relatively even, relative to other systems (e.g., Aldrich &

Hamrick 1998; Sezen et al. 2005). The low reproductive dominance observed in our study

may result from the large number of flowering plants (in CF1), flowering asynchrony, and

This article is protected by copyright. All rights reserved.

 high seed removal rates, relative to highly fecund tree species that have been the focus of
Accepted Article
previous studies.

Genetic diversity, inbreeding and fine-scale spatial genetic structure

The significant inbreeding coefficient of seedlings and reproductive plants across

populations indicates that mating between genetically related plants does occur in

Heliconia acuminata, which can help generate the significant fine-scale spatial genetic

structure of seedlings in all populations. Although the fine-scale spatial genetic structure

did not differ between reproductive plants and seedlings, most of the populations

exhibited non-significant (or extremely low) sp-values for adult plants. It is possible that

there is a trend towards the attenuation of spatial genetic structure with increasing life

stage, which may occur due to demographic thinning and density-dependent mortality

(Chung et al. 2003; Zhou & Chen 2010) or to the spatial pattern of flowering (Hirao & Kudo

2008). Comparisons across populations, however, show that neither inbreeding levels nor

spatial genetic structure are significantly different. Given the consistently similar genetic

diversity of seedlings and reproductive plants across populations, it is possible that genetic

diversity is maintained across this landscape by different processes: the immigration of

propagules into fragments and low-density populations, and the high diversity of parental

contribution in denser populations.

This article is protected by copyright. All rights reserved.

 Caveats
Accepted Article
In our discussion, we have largely ignored myriad ecological and genetic processes

that take place between pollen deposition on floral stigmas and seedling establishment. It

is possible that post-dispersal processes may re-structure the spatial structure of

seedlings. For instance, spatially structured mortality, due to Janzen-Connell effects, could

reduce the number of offspring close to maternal plants (Choo et al. 2012; Isagi et al.

2007; Steinitz et al. 2011). If survival of plants is plot and microsite dependent, sampling

seedlings could have exacerbated differences in dispersal distances among populations. In

the study site, the proportion of dead seedlings did vary across populations (48% in CF1,

33% in CF2, 51% in F1, 39% in F2 and 36% in F3 of all seedlings recruited between 1999

and 2008 were dead in 2009, χ2 = 12.68, df = 4, P = 0.0129, N = 1046). Fine-scale spatial

genetic structure, however, did not vary much between cohorts or among populations,

suggesting that plant mortality at later stages is unlikely to be spatially structured.

Moreover, the BDFFP landscape is surrounded by large expanses of primary forest

with fragment boundaries distant only 100 m from continuous forests. Given more

substantial habitat isolation, manakins and thrushes might not move to other forested

patches. If trapped within a fragment, these birds might eventually disappear from the

system, leading to the genetic erosion that we do not see here. Nevertheless, our study

shows that the secondary growth on cleared land does not impede the movement of both

seed dispersers and pollinators across this landscape. A considerable portion of global

tropical forest cover consists of forest regrowth, following logging, agricultural

This article is protected by copyright. All rights reserved.

 abandonment, or conversion to agroforests (FAO 2011), making our findings relevant to
Accepted Article
tropical forests elsewhere.

Conclusions

Our results show that gene dispersal across a heterogeneous, historically

fragmented landscape, is more related to density of flower and fruit resources than to

fragmentation per se. Forest fragmentation, however, can further enhance gene dispersal

by reducing the abundance of plants within patches (Bruna 1999, 2002; Uriarte et al.

2010), enforcing movement among forested areas. Our study shows that continuous

forest sites can have striking variation in plant abundance, which translates into divergent

propagule dispersal outcomes.

Plant population dynamics and persistence in fragmented landscapes can be

assessed using methods that allow contemporary and spatially explicit evaluation of

ongoing genetic processes. We suggest that future studies of contemporary gene flow

should take into consideration plant and dispersal vector features, which vary across

changing landscapes. Reformulating a new set of predictions within conservation genetics

will require the contribution of additional studies to draw a more representative picture of

how interactions between landscape configuration and organismal traits influence the

processes of pollination and seed dispersal.

This article is protected by copyright. All rights reserved.

 Acknowledgments
Accepted Article
We thank Dustin Rubenstein and three anonymous reviewers for valuable

comments on the manuscript. We would like to thank Jeffrey Hunt, Gabriel Johnson, Ida

Lopez and David Erickson for assistance in the development of the molecular markers and

Lee Weigt for facilitating our work in the Museum Support Center, Smithsonian

Institution. We are grateful for the help provided by Carla Sardelli and Carolina Medeiros

at LabGen, INPA, and Osmaildo Ferreira da Silva for assistance in the field. We thank Emily

Moran for sharing the code and helping with the model interpretation and Lora Murphy

for help in adapting the code to our system. We also thank BDFFP and INPA for their

logistical support. Financial support was provided by the US National Science Foundation

(award DEB-0614339 to MU, DEB-0614149, DEB-0309819, DBI-0109226, and INT 98-06351

to EB) and the Smithsonian Institution. MRL and RG acknowledge research fellowships

from CNPq/Brazil. This is publication number --- in the BDFFP Technical Series.

References

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. Molecular Ecology 17, 5177-5188.

Ahmed S, Compton SG, Butlin RK, Gilmartin PM (2009) Wind-borne insects mediate

directional pollen transfer between desert fig trees 160 kilometers apart.

Proceedings of the National Academy of Sciences 106, 20342-20347.

This article is protected by copyright. All rights reserved.

 Aldrich PR, Hamrick JL (1998) Reproductive dominance of pasture trees in a fragmented
Accepted Article
tropical forest mosaic. Science 281, 103-105.

Ashley MV (2010) Plant parentage, pollination, and dispersal: How DNA microsatellites

have altered the landscape. Critical Reviews in Plant Sciences 29, 148-161.

Bacles CFE, Jump AS (2011) Taking a tree's perspective on forest fragmentation genetics.

Trends in Plant Science 16, 13-18.

Bacles CFE, Lowe AJ, Ennos RA (2006) Effective seed dispersal across a fragmented

landscape. Science 311, 628-628.

Bittencourt JVM, Sebbenn AM (2007) Patterns of pollen and seed dispersal in a small,

fragmented population of the wind-pollinated tree Araucaria angustifolia in

southern Brazil. Heredity 99, 580-591.

Breed MF, Marklund MHK, Ottewell KM, et al. (2012a) Pollen diversity matters: revealing

the neglected effect of pollen diversity on fitness in fragmented landscapes.

Molecular Ecology 21, 5955-5968.

Breed MF, Ottewell KM, Gardner MG, et al. (2012b) Mating system and early viability

resistance to habitat fragmentation in a bird-pollinated eucalypt. Heredity.

Bruna EM (1999) Biodiversity - Seed germination in rainforest fragments. Nature (London)

402, 139-139.

Bruna EM (2002) Effects of forest fragmentation on Heliconia acuminata seedling

recruitment in central Amazonia. Oecologia 132, 235-243.

Bruna EM (2003) Are plant populations in fragmented habitats recruitment limited? Tests

with an Amazonian herb. Ecology 84, 932-947.

This article is protected by copyright. All rights reserved.

 Bruna EM, Kress WJ (2002) Habitat fragmentation and the demographic structure of an
Accepted Article
Amazonian understory herb (Heliconia acuminata). Conservation Biology 16, 1256-

1266.

Bruna EM, Kress WJ, Marques F, Silva OF (2004) Heliconia acuminata reproductive success

is independent of local floral density. Acta amazonica 34, 467.

Bruna EM, Ribeiro MBN (2005) Regeneration and population structure of Heliconia

acuminata in Amazonian secondary forests with contrasting land-use histories.

Journal of Tropical Ecology 21, 127-131.

Burczyk J, Adams WT, Birkes DS, Chybicki IJ (2006) Using genetic markers to directly

estimate gene flow and reproductive success parameters in plants on the basis of

naturally regenerated seedlings. Genetics 173, 363-372.

Byrne M, Elliott CP, Yates C, Coates DJ (2007) Extensive pollen dispersal in a bird-

pollinated shrub, Calothamnus quadrifidus, in a fragmented landscape. Molecular

Ecology 16, 1303-1314.

Choo J, Juenger TE, Simpson BB (2012) Consequences of frugivore-mediated seed

dispersal for the spatial and genetic structures of a neotropical palm. Molecular

Ecology 21, 1019-1031.

Chung MY, Epperson BK, Chung MG (2003) Genetic structure of age classes in Camellia

japonica (Theaceae). Evolution 57, 62-73.

Chybicki IJ, Burczyk J (2010) NM plus : software implementing parentage-based models for

estimating gene dispersal and mating patterns in plants. Molecular Ecology

Resources 10, 1071-1075.

This article is protected by copyright. All rights reserved.

 Clark JS, Silman M, Kern R, Macklin E, HilleRisLambers J (1999) Seed dispersal near and far:
Accepted Article
patterns across temperate and tropical forests. Ecology 80, 1475-1494.

Côrtes MC, Gowda V, Kress WJ, Bruna EM, Uriarte M (2009) Characterization of 10

microsatellite markers for the understorey Amazonian herb Heliconia acuminata.

Molecular Ecology Resources 9, 1261-1264.

DiBattista JD (2008) Patterns of genetic variation in anthropogenically impacted

populations. Conservation Genetics 9, 141-156.

Dick CW, Etchelecu G, Austerlitz F (2003) Pollen dispersal of tropical trees (Dinizia excelsa:

Fabaceae) by native insects and African honeybees in pristine and fragmented

Amazonian rainforest. Molecular Ecology 12, 753-764.

Dobkin DS (1984) Flowering patterns of long-lived Heliconia inflorescences: implications

for visiting and resident nectarivores. Oecologia 64, 245-254.

Dobkin DS (1987) Synchronous flower abscission in plants pollinated by hermit

hummingbirds and the evolution of one-day flowers. Biotropica 19, 90-93.

Eckert CG, Kalisz S, Geber MA, et al. (2010) Plant mating systems in a changing world.

Trends in Ecology & Evolution 25, 35-43.

FAO (2011) State of the world's forests. Food and Agriculture Organization, Rome.

Ferreira ME, Grattapaglia D (1998) Introdução ao uso de marcadores moleculares em

análise genética de plantas (ed. EMBRAPA). EMBRAPA, Brasília, DF.

García C, Grivet D (2011) Molecular insights into seed dispersal mutualisms driving plant

population recruitment. Acta Oecologica 37, 632-640.

This article is protected by copyright. All rights reserved.

 García C, Jordano P, Arroyo JM, Godoy JA (2009) Maternal genetic correlations in the seed
Accepted Article
rain: effects of frugivore activity in heterogeneous landscapes. Journal of Ecology

97, 1424-1435.

Gascon C, Bierregaard RO, Jr. (2001) The Biological Dynamics of Forest Fragments Project:

the study site, experimental design, and research activity. In: Lessons from

Amazonia: the ecology and conservation of a fragmented forest (eds. Bierregaard

RO, Jr., Gascon C, Lovejoy TE, Mesquita RCG), pp. 31-46. Yale University Press New

Haven, Connecticut.

Ghazoul J (2005) Pollen and seed dispersal among dispersed plants. Biological Reviews 80,

413-443.

Grivet D, Smouse P, Sork V (2005) A novel approach to an old problem: tracking dispersed

seeds. Molecular Ecology 14, 3585-3595.

Hadfield JD, Richardson DS, Burke T (2006) Towards unbiased parentage assignment:

combining genetic, behavioural and spatial data in a Bayesian framework.

Molecular Ecology 15, 3715-3730.

Hadley AS, Betts MG (2011) The effects of landscape fragmentation on pollination

dynamics: absence of evidence not evidence of absence. Biological Reviews 87,

526-544.

Hanson T, Brunsfeld S, Finegan B, Waits L (2007) Conventional and genetic measures of

seed dispersal for Dipteryx panamensis (Fabaceae) in continuous and fragmented

Costa Rican rain forest. Journal of Tropical Ecology 23, 635-642.

This article is protected by copyright. All rights reserved.

 Hardy OJ, Vekemans X (2002) SPAGEDi: a versatile computer program to analyse spatial
Accepted Article
genetic structure at the individual or population levels. Molecular Ecology Notes 2,

618-620.

Herrera JM, Morales JM, Garcia D (2011) Differential effects of fruit availability and habitat

cover for frugivore-mediated seed dispersal in a heterogeneous landscape. Journal

of Ecology 99, 1100-1107.

Hirao AS, Kudo G (2008) The effect of segregation of flowering time on fine-scale spatial

genetic structure in an alpine-snowbed herb Primula cuneifolia. Heredity 100, 424-

430.

Hobbs RJ, Yates CJ (2003) Impacts of ecosystem fragmentation on plant populations:

generalising the idiosyncratic. Australian Journal of Botany 51, 471-488.

Honnay O, Jacquemyn H (2007) Susceptibility of common and rare plant species to the

genetic consequences of habitat fragmentation. Conservation Biology 21, 823-831.

Isagi Y, Saito D, Kawaguchi H, Tateno R, Watanabe S (2007) Effective pollen dispersal is

enhanced by the genetic structure of an Aesculus turbinata population. Journal of

Ecology 95, 983-990.

Jones AG, Ardren WR (2003) Methods of parentage analysis in natural populations.

Molecular Ecology 12, 2511-2523.

Jones AG, Small CM, Paczolt KA, Ratterman NL (2010) A practical guide to methods of

parentage analysis. Molecular Ecology Resources 10, 6-30.

Jordano P, Godoy JA (2002) Frugivore-generated seed shadows: a landscape view of

demographic and genetic effects CAB International.

This article is protected by copyright. All rights reserved.

 Kamm U, Rotach P, Gugerli F, et al. (2009) Frequent long-distance gene flow in a rare
Accepted Article
temperate forest tree (Sorbus domestica) at the landscape scale. Heredity 103,

476-482.

Khamcha D, Savini T, Brockleman WY, Chimchome V, Gale GA (2012) Influence of food

availability and distribution on the movement patterns of a forest avian frugivore,

the puff-throated bulbul (Alophoixus pallidus). Journal of Tropical Ecology 28, 1-9.

Kolb A (2008) Habitat fragmentation reduces plant fitness by disturbing pollination and

modifying response to herbivory. Biological Conservation 141, 2540-2549.

Kramer AT, Ison JL, Ashley MV, Howe HF (2008) The paradox of forest fragmentation

genetics. Conservation Biology 22, 878-885.

Kwit C, Levey DJ, Greenberg CH, et al. (2004) Fruit abundance and local distribution of

wintering hermit thrushes (Catharus guttatus) and yellow-rumped warblers

(Dendroica coronata) in South Carolina. Auk 121, 46-57.

Lander TA, Boshier DH, Harris SA (2010) Fragmented but not isolated: contribution of

single trees, small patches and long-distance pollen flow to genetic connectivity for

Gomortega keule, an endangered Chilean tree. Biological Conservation 143, 2583-

2590.

Laurance WF, Lovejoy TE, Vasconcelos HL, et al. (2002) Ecosystem decay of Amazonian

forest fragments: A 22-year investigation. Conservation Biology 16, 605-618.

Llorens TM, Byrne M, Yates CJ, Nistelberger HM, Coates DJ (2012) Evaluating the influence

of different aspects of habitat fragmentation on mating patterns and pollen

This article is protected by copyright. All rights reserved.

 dispersal in the bird-pollinated Banksia sphaerocarpa var. caesia. Molecular
Accepted Article
Ecology 21, 314-328.

Loiselle BA, Sork VL, Nason J, Graham C (1995) Spatial genetic-structure of a tropical

understory shrub, Psychotria officinalis (Rubiaceae) American Journal of Botany 82,

1420-1425.

Meagher TR (2010) A fragmented future for the forest flora. Heredity 105, 163-164.

Meagher TR, Thompson E (1987) Analysis of parentage for naturally established seedlings

of Chamaelirium luteum (Liliaceae) Ecology 68, 803-812.

Mesquita RCG, Ickes K, Ganade G, Williamson GB (2001) Alternative successional

pathways in the Amazon Basin. Journal of Ecology 89, 528-537.

Morales JM, Carlo TA (2006) The effects of plant distribution and frugivore density on the

scale and shape of dispersal kernels. Ecology 87, 1489-1496.

Moran EV, Clark JS (2011) Estimating seed and pollen movement in a monoecious plant: a

hierarchical Bayesian approach integrating genetic and ecological data. Molecular

Ecology 20, 1248-1262.

Moran EV, Clark JS (2012a) Between-site differences in the scale of dispersal and gene

flow in red oak. Plos One 7.

Moran EV, Clark JS (2012b) Causes and consequences of unequal seedling production in

forest trees: a case study in red oaks. Ecology 93, 1082-1094.

Nason JD, Hamrick JL (1997) Reproductive and genetic consequences of forest

fragmentation: Two case studies of neotropical canopy trees. Journal of Heredity

88, 264-276.

This article is protected by copyright. All rights reserved.

 Oddou-Muratorio S, Klein EK (2008) Comparing direct vs. indirect estimates of gene flow
Accepted Article
within a population of a scattered tree species. Molecular Ecology 17, 2743-2754.

Ouborg NJ, Vergeer P, Mix C (2006) The rough edges of the conservation genetics

paradigm for plants. Journal of Ecology 94, 1233-1248.

Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic

software for teaching and research. Molecular Ecology Notes 6, 288-295.

Robledo-Arnuncio JJ, Alia R, Gil L (2004) Increased selfing and correlated paternity in a

small population of a predominantly outcrossing conifer, Pinus sylvestris.

Molecular Ecology 13, 2567-2577.

Sezen UU, Chazdon RL, Holsinger KE (2005) Genetic consequences of tropical second-

growth forest regeneration. Science 307, 891-891.

Smouse PE, Robledo-Arnuncio JJ (2005) Measuring the genetic structure of the pollen pool

as the probability of paternal identity. Heredity 94, 640-649.

Sork VL, Smouse PE (2006) Genetic analysis of landscape connectivity in tree populations.

Landscape ecology 21, 821-836.

Stacy EA, Hamrick JL, Nason JD, et al. (1996) Pollen dispersal in low-density populations of

three neotropical tree species. American Naturalist 148, 275-298.

Steinitz O, Troupin D, Vendramin GG, Nathan R (2011) Genetic evidence for a Janzen–

Connell recruitment pattern in reproductive offspring of Pinus halepensis trees.

Molecular Ecology 20, 4152-4164.

Stouffer PC, Bierregaard RO (1995a) Effects of forest fragmentation on understory

hummingbirds in Amazonian Brazil Conservation Biology 9, 1085-1094.

This article is protected by copyright. All rights reserved.

 Stouffer PC, Bierregaard RO (1995b) Use of Amazonian forest fragments by understory
Accepted Article
insectivorous birds Ecology 76, 2429-2445.

Stouffer PC, Bierregaard RO (2007) Recovery potential of understory bird communities in

Amazonian rainforest fragments. Revista Brasileira De Ornitologia 15, 219-229.

Stouffer PC, Bierregaard RO, Jr. (1996) Forest fragmentation and seasonal patterns of

hummingbird abundance in Amazonian Brazil. Ararajuba 4, 9-14.

Uriarte M, Anciães M, da Silva MTB, et al. (2011) Disentangling the drivers of reduced

long-distance seed dispersal by birds in an experimentally fragmented landscape.

Ecology 92, 924-937.

Uriarte M, Bruna EM, Rubim P, Anciaes M, Jonckheere I (2010) Effects of forest

fragmentation on the seedling recruitment of a tropical herb: assessing seed vs.

safe-site limitation. Ecology 91, 1317-1328.

Vekemans X, Hardy OJ (2004) New insights from fine-scale spatial genetic structure

analyses in plant populations. Molecular Ecology 13, 921-935.

Vranckx G, Jacquemyn H, Muys B, Honnay O (2012) Meta-Analysis of Susceptibility of

Woody Plants to Loss of Genetic Diversity through Habitat Fragmentation.

Conservation Biology 26, 228-237.

Wang R, Compton SG, Chen XY (2011) Fragmentation can increase spatial genetic

structure without decreasing pollen-mediated gene flow in a wind-pollinated tree.

Molecular Ecology 20, 4421-4432.

Watling JI, Donnelly MA (2006) Fragments as islands: a synthesis of faunal responses to

habitat patchiness. Conservation Biology 20, 1016-1025.

This article is protected by copyright. All rights reserved.

 White GM, Boshier DH, Powell W (2002) Increased pollen flow counteracts fragmentation
Accepted Article
in a tropical dry forest: An example from Swietenia humilis Zuccarini. Proceedings

of the National Academy of Sciences of the United States of America 99, 2038-

2042.

Young A, Boyle T, Brown T (1996) The population genetic consequences of habitat

fragmentation for plants. Trends in Ecology & Evolution 11, 413-418.

Young AG, Pickup M (2010) Low S-allele numbers limit mate availability, reduce seed set

and skew fitness in small populations of a self-incompatible plant. Journal of

Applied Ecology 47, 541-548.

Zhou HP, Chen J (2010) Spatial genetic structure in an understorey dioecious fig species:

the roles of seed rain, seed and pollen-mediated gene flow, and local selection.

Journal of Ecology 98, 1168-1177.

Data accessibility

Original code of the gene dispersal model can be found on Emily Moran’s website:

https://sites.google.com/site/emilyvmoran/

Sample locations, microsatellite data, and modified code deposited in the Dryad

repository: doi: 10.5061/dryad.b003f.

Author contributions

M.C.C., W.J.K. and E.B. collected the data in the field; M.C.C. performed the research and

wrote the paper; M.C.C., M.U. and P.S. contributed to the statistical analyses; W.J.K.,

This article is protected by copyright. All rights reserved.

 M.R.L. and R.G. contributed reagents and lab material; all authors contributed to writing
Accepted Article
the paper.

Supporting information

Data S1. Study site

Data S2. Implementation of the Bayesian model

Data S3. Probability of parental identity (PPaI)

Data S4. Flowering and seedling phenology

Data S5. Inbreeding coefficient and fine-scale spatial genetic structure

Table 1- Characteristics of populations of Heliconia acuminata in the study site: total

number of reproductive plants (number of dead plants not sampled in parentheses),

percentage (%) and density of flowering plants, and number of sampled seedlings (the

total number of seedlings recruited is indicated in parentheses).

Total No Total
% Total Average of
of density of No of
flowering total plant
Popula Reserve flowering flowering seedlings
(1999-2009) density (1999-
tion number plants plants (1999-
(yearly 2009) (yearly
(1999- (plants per 2009)
range) range)
2
2009) m)

CF1 1501 285 (3) 15 (0.4 - 0.0226 732 (544 - 374 (681)

This article is protected by copyright. All rights reserved.

 12.3) 837)
Accepted Article
None 122 (110 -

CF2 (Dimona) 20 (5) 5 (0 - 5.6) 0.0016 132) 52 (80)

3114 227 (203 -

F1 44 (2) 7 (0 - 3.4) 0.0035 249) 118 (172)

2107 221 (208 -

F2 50 (0) 9 (0 - 7.4) 0.0040 232) 73 (112)

2108 16 (2.4 - 185 (146 -

F3 60 (3) 15.0) 0.0047 219) 83 (127)

Table 2. Total number and percentage of seedlings with parents located inside and outside

the 1.26 ha plot, and number and percentage of reproductive plants that were assigned as

either pollen (fathers) or seed donors (mothers).

Parent pair Parent pair Father Contributing

Mother outside
inside plot outside plot outside plants

CF1 276 (74%) 9 (2%) 8 (2%) 81 (22%) 212 (74%)

CF2 1 (2%) 32 (62%) 10 (19%) 9 (17%) 13 (62%)

F1 26 (22%) 38 (32%) 20 (17%) 34 (29%) 40 (91%)

This article is protected by copyright. All rights reserved.

 F2 23 (31%) 18 (25%) 17 (23%) 15 (21%) 35 (70%)
Accepted Article
F3 47 (57%) 11 (13%) 18 (22%) 7 (8%) 46 (76%)

Table 3. Unbiased expected heterozygosity (He), mean number of alleles per locus (Na),

inbreeding coefficient (Fis), and fine-scale spatial genetic structure (sp-value) of

reproductive plants and seedlings of Heliconia acuminata.

Seedlings Reproductive plants

p- p- p- p-
UHe Na Fis sp UHe Na Fis sp
value value value value

0.62 9. 0.62 9.
0.03 <0.00 0.002 <0.00 0.02 0.004
CF1
4 2 8 1 5 1 1 1 5 1
0.029 0.000

0.69 7. 0.69 7. -
0.08 <0.00 0.014 <0.00 0.13
CF2 0.014
4 3 9 1 2 1 3 0 1
0.001 6 0.068

0.66 7. 0.68 6. -
0.08 <0.00 0.005 0.08
F1 0.006 0.000
1 4 6 1 1 7 7 5
0.002 5 0.920

0.70 8. 0.70 8.
0.11 <0.00 0.012 <0.00 0.10 0.004
F2
9 2 7 1 2 1 0 0 2 3
0.000 0.122

0.69 7. 0.70 8.
0.05 0.013 <0.00 0.06 0.012
F3 0.008 0.010 0.002
8 8 1 1 1 1 2 2 5

This article is protected by copyright. All rights reserved.

 Figure captions:
Accepted Article
Figure 1. Realized dispersal of pollen and seeds. Main graph: Modeled kernel using the

2Dt-function, given the posterior mean of pollen dispersal (up) and seed dispersal (us)

parameters. Scale of y-axis is different for pollen and seed dispersal, with seed

dispersal reaching higher limits, by virtue of being more distance-restricted. The y-axis

is broken to permit clear visualization of all kernels because probability of pollen and

seed dispersal at short distances in CF1 was higher than in other populations. The X-

axis was truncated at 140 m. Graph inset: Average modal distance of pollen and seed

dispersal and associated 95% support intervals obtained from 50,000 simulations for

each population of Heliconia acuminata. Non-overlapping intervals indicate that

modal distances are significantly different.

Figure 2. Number of seedlings each plant fathered or mothered (main graph);

proportional genetic contribution of individual reproductive plants (graph inset) via

either pollen or seeds to the next generations of seedlings and R0 (PPaI - values) of

Heliconia acuminata in each population. The y-axis represents the number of seedlings

or proportional contribution (sum across reproductive plants is equal to one) and the x-

axis represents each reproductive plant that contributed genes. The curve represents the

decreasing ranking of the plants given their contribution. Flat curves indicate even

contribution, whereas steep curves represent uneven genetic contribution.

Figure 3. Autocorrelogram of pairwise kinship against distance (meters) of seedlings and

reproductive plants across the five populations of Heliconia acuminata at the BDFFP.

This article is protected by copyright. All rights reserved.

Accepted Article

This article is protected by copyright. All rights reserved.

Accepted Article

This article is protected by copyright. All rights reserved.