Testing Founder Effect Speciation: Divergence Population
Genetics of the Spoonbills Platalea regia and Pl. minor
(Threskiornithidae, Aves)
Carol K. L. Yeung, ,1 Pi-Wen Tsai, ,2 R. Terry Chesser,3 Rong-Chien Lin,1 Cheng-Te Yao,4
Xiu-Hua Tian,5 and Shou-Hsien Li*,1
1
Department of Life Science, National Taiwan Normal University, Taipei, Taiwan, R.O.C.
Department of Mathematics, National Taiwan Normal University, Taipei, Taiwan, R.O.C.
3
U.S. Geological Survey, Patuxent Wildlife Research Center, National Museum of Natural History, Smithsonian Institution,
Washington, DC
4
Middle Altitude Experimental Station, Endemic Species Research Institute, Chi-chi, Nantou, Taiwan, R.O.C.
5
College of Wildlife Resources, Northeast Forestry University, Harbin, Heilongjiang, China.
These authors contributed equally to this work.
*Corresponding author: E-mail: [email protected].
Associate editor: Scott Edwards
2
Although founder effect speciation has been a popular theoretical model for the speciation of geographically isolated taxa,
its empirical importance has remained difficult to evaluate due to the intractability of past demography, which in
a founder effect speciation scenario would involve a speciational bottleneck in the emergent species and the complete
cessation of gene flow following divergence. Using regression-weighted approximate Bayesian computation, we tested the
validity of these two fundamental conditions of founder effect speciation in a pair of sister species with disjunct
distributions: the royal spoonbill Platalea regia in Australasia and the black-faced spoonbill Pl. minor in eastern Asia. When
compared with genetic polymorphism observed at 20 nuclear loci in the two species, simulations showed that the founder
effect speciation model had an extremely low posterior probability (1.55 108) of producing the extant genetic pattern.
In contrast, speciation models that allowed for postdivergence gene flow were much more probable (posterior
probabilities were 0.37 and 0.50 for the bottleneck with gene flow and the gene flow models, respectively) and
postdivergence gene flow persisted for a considerable period of time (more than 80% of the divergence history in both
models) following initial divergence (median 5 197,000 generations, 95% credible interval [CI]: 50,000478,000, for the
bottleneck with gene flow model; and 186,000 generations, 95% CI: 45,000477,000, for the gene flow model).
Furthermore, the estimated population size reduction in Pl. regia to 7,000 individuals (median, 95% CI: 48712,000,
according to the bottleneck with gene flow model) was unlikely to have been severe enough to be considered a bottleneck.
Therefore, these results do not support founder effect speciation in Pl. regia but indicate instead that the divergence
between Pl. regia and Pl. minor was probably driven by selection despite continuous gene flow. In this light, we discuss the
potential importance of evolutionarily labile traits with significant fitness consequences, such as migratory behavior and
habitat preference, in facilitating divergence of the spoonbills.
Key words: founder effect speciation, spoonbills, Platalea, postdivergence gene flow, approximate Bayesian computation.
Introduction
Founder effect speciation (or speciation by founder principle, Mayr 1942) has been a popular hypothesis for the origin of species featuring marked geographic isolation (e.g.,
island speciation, reviewed in Provine 1989 and Coyne and
Orr 2004) or those that disperse rarely but survive well in
new environments (e.g., Paulay 2002). In a classic founder
effect speciation scenario as conceived by Mayr, a characteristically small number of individuals are isolated
from the parental population. Increased genetic drift
as a result of the extreme bottleneck and isolation then
accelerates the formation of novel allelic combinations
adapted to the new environment (a process termed
‘‘genetic revolution’’; Mayr 1954). Finally and incidentally,
reproductive isolation from the parental population arises
as a by-product of the divergence of the genomes. Founder
speciation is essentially a genetic model, anchored by the
assumptions of small founder population size and complete cessation of gene flow due to spatial isolation following the founding event (adhering to Mayr’s description in
Mayr 1942, 1954, 1963, 1970, 1976). The model is often
regarded as a variant of peripatric speciation in which
a few individuals isolated from a nearby parental population form a new species as the result of intense drift, novel
selective pressure, and the lack of gene flow (Mayr 1954,
1982; Coyne and Orr 2004; Gavrilets 2004).
© The Author 2010. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. All rights reserved. For permissions, please
e-mail: [email protected]
Mol. Biol. Evol. 28(1):473–482. 2011 doi:10.1093/molbev/msq210
Advance Access publication August 12, 2010
473
Research article
Abstract
Yeung et al. · doi:10.1093/molbev/msq210
The plausibility of founder effect speciation can be examined by closely monitoring the consequences of founding events. However, this has been attempted in only a few
studies with results ambiguously supportive of the model.
In organisms with short life spans, such as the fruit fly
(Drosophila spp.), reproductive isolation had a chance of
arising only after multiple bottlenecks in the laboratory
(e.g., as reviewed in Moya et al. 1995). In the wild, colonization of Daphne Major of the Galapagos Islands by merely
five large ground finches Geospiza magnirostris resulted in
abrupt morphological divergence after 9 years (Grant et al.
2001); however, this phenomenon has been attributed to
the joint effect of chance and selection on genetic variation
introduced through recurrent immigration and thus does
not support founder effect speciation sensu stricto. Alternatively, founder effect speciation has been inferred via
phylogenetic comparative methods (reviewed in Losos
and Glor 2003), which are based on the assumption that
geographic attributes such as range size and degree of range
overlap are reliable correlates of demographic attributes
such as population size and degree of gene flow. Such studies have yielded results in support of (in birds, Friesen and
Anderson 1997; Barraclough and Vogler 2000) or against
(e.g., in birds, Chesser and Zink 1994; in felids, Mattern
and McLennan 2000) the prevalence of peripatric speciation, which is similar to founder effect speciation except
that it is characterized by shorter geographic distance between the ancestral and derived species. However, extant
species range, which is the key evolutionary feature in these
studies, is highly labile and therefore an unreliable trait (reviewed in Losos and Glor 2003). Alternative methods, such
as demographic genetic methods, are needed to validate
various speciation models independent of extant species
range.
Recent advances in evolutionary genetics have enabled
the inference of historical demographic attributes during
divergence. Contrary to expectations based on founder effect speciation, it has been shown that effective population
size (Ne) at incipient speciation for some oceanic island
species may have been quite large (e.g., Darwin’s finches
on the Galápagos Archipelago, Vincek et al. 1997; the silvereye species complex Zosterops lateralis in Australia and
southwest Pacific islands, Clegg et al. 2002; the auklets Aethia spp. Walsh et al. 2005) relative to that expected for
a bottlenecked population (e.g., on the order of several
hundred individuals or fewer; Wright 1931, Mayr 1942).
Nevertheless, historical bottlenecks could not be rejected
in some cases (e.g., beetle Orphraella bilineata, Knowles
et al. 1999; Z. lateralis population on Norfolk Island, Clegg
et al. 2002; zebra finch Taeniopygia guttata, Balakrishnan
and Edwards, 2009).
The isolation with migration (IM) model has also been
widely applied to estimate historical gene flow between
divergent species over their divergence history (e.g., in
Drosophila fruit flies, Machado et al. 2002; in human Homo
sapiens, Hey 2005; in chimpanzees Pan troglodytes, Pa. t.
versus, and Pa. paniscus, Won and Hey 2005; in Icterus orioles, Kondo et al. 2008; in cave salamanders Gyrinophilus
474
MBE
spp., Niemiller et al. 2008; in Solanum wild tomatoes,
Stadler et al. 2008). However, this method cannot answer
the question of whether gene flow ceases immediately following initial divergence between taxa, as postulated by
founder effect speciation. The recently developed techniques of approximate Bayesian computation (ABC; Tavare
et al. 1997; Pritchard et al. 1999), which allow flexible model
construction to include variables such as the size of the
speciational bottleneck and time at which postdivergence
gene flow ceases, potentially allow the contentious issue of
founder effect speciation to be resolved empirically.
In this study, we conducted genetic analyses to investigate the demography during divergence of a pair of closely
related but disjunctly distributed spoonbills: the royal
spoonbill Pl. regia of Australasia and the black-faced spoonbill Pl. minor of eastern Asia. A recent phylogenetic analysis
of the genus Platalea (Chesser et al. 2010) provided strong
evidence that these species are sister taxa (94–98% bootstrap support, 0.92 posterior probability) and that they
speciated rather recently (differing in mitochondrial
DNA by less than 1%). These species bear a strong morphological resemblance to the Eurasian spoonbill Pl. leucorodia,
which was found to be sister to the regia–minor clade
(100% bootstrap, 1.00 posterior probability; Chesser
et al. 2010). Given the geographical isolation of Pl. regia
relative to the east Asian and Palearctic distributions of
Pl. minor and Pl. leucorodia, respectively (fig. 1), our phylogenetic hypothesis of spoonbills suggests that Pl. regia
represents a potential case of founder effect speciation
as the result of long-distance dispersal of an Asian–Palearctic
ancestor to Australasia.
We used variation at 20 independent nuclear loci
(11,943 bp) and the regression-weighted ABC (Beaumont
et al. 2002) to infer the probabilities of four alternative demographic models for the divergence between Pl. regia and
Pl. minor, focusing on the potential for founder effect speciation in Pl. regia. Each model differed with respect to the
two fundamental assumptions of founder effect speciation:
the null or founder effect speciation model incorporated
a speciational bottleneck but no postdivergence gene flow,
the isolation model included neither a speciational bottleneck nor postdivergence gene flow, the bottleneck with
gene flow model included both a speciational bottleneck
and postdivergence gene flow, and the gene flow model
incorporated no speciational bottleneck but did include
postdivergence gene flow (fig. 2). Modeling of a potential
speciational bottleneck was limited to Pl. regia, whereas reciprocal gene flow between Pl. regia and Pl. minor following
initial divergence was modeled as postdivergence gene
flow.
For the favored model(s), demographic parameters including long-term Ne of each of the two species and their
common ancestor, divergence time, the amount and time
of cessation of postdivergence gene flow (if present), and
the magnitude and duration of the speciational bottleneck
in Pl. regia (if present) were estimated via coalescent simulation and the ABC (Beaumont et al. 2002). Founder effect
speciation would be rejected if its posterior probability was
Testing Founder Effect Speciation · doi:10.1093/molbev/msq210
MBE
FIG. 1 Distribution map of the closely related Pl. regia, Pl. minor, and Pl. leucorodia (modified from Hancock et al. 1992). The arrows indicate
general direction of migration from breeding to wintering grounds.
significantly lower than that of the other three models. Our
results suggested that a founder effect speciation scenario
was highly unlikely for Pl. regia and instead indicated pro-
longed postdivergence gene flow between Pl. regia and its
sister species Pl. minor. This finding accentuates the importance of selection despite gene flow in the diversification of
FIG. 2 Demographic models of the divergence between Pl. regia and its sister species Pl. minor. NPr and NPm, long-term equilibrium effective
population size of Pl. regia and Pl. minor, respectively; NA, effective population size of the ancestral species; T1, divergence time in number of
generations ago; MPrPm and MPmPr, asymmetrical gene flow from Pl. regia to Pl. minor and from Pl. minor to Pl. regia; T2 and T3, time of
cessation of gene flow from Pl. regia to Pl. minor and from Pl. minor to Pl. regia and speciational bottleneck as a proportion B of NPr that lasted
until Tb generations ago.
475
MBE
Yeung et al. · doi:10.1093/molbev/msq210
these birds. The speciational scenario compatible with the
deduced historical demography may involve changes in
evolutionary labile traits such as migratory behavior and
habitat preference.
Materials and Methods
Sampling and DNA Preparation
Blood, feathers, muscle, or liver tissues were sampled for the
focal species pair Pl. regia (N 5 10) and Pl. minor (N 5 21),
as well as single individuals of Pl. leucorodia, yellow-billed
spoonbill Pl. flavipes, and roseate spoonbill Pl ajaja, which
were used as outgroup taxa in defining mutations in the
ABC analysis. Sample details are provided in supplementary
table S1, Supplementary Material online. Hundred percent
ethanol was added to fresh tissue samples for storage in the
field and the sample ethanol mixtures were later transferred to a 80 °C freezer for long-term storage. Genomic
DNA was extracted using a modified LiCl method (Gemmell
and Akiyama 1996).
Nuclear DNA Amplification, Sequencing, and
Phasing
All samples were genotyped at 17 nuclear autosomal loci
and three Z-linked loci (locus details and primer sequences
are provided in supplementary table S2, Supplementary
Material online). Primers for these genes were derived from
a vinous-throated parrotbill Paradoxornis webbianus
spleen cDNA library and were identified/annotated by
blasting against the chicken genome (Yeung CKL and Li
SH, unpublished data); two of the three Z-linked loci were
published by Backström et al. (2006). A touchdown polymerase chain reaction (PCR) scheme was employed to amplify these fragments in reaction volumes of 12 ll
containing 50–200 ng of template DNA, 0.2 lM of each
primer, 0.5 mM deoxynucleotide triphosphates, 10 mM
Tris–HCl, pH 9.0, 50 mM KCl, 0.4 U Taq DNA polymerase
(GE Biosciences), and 1.5 mM MgCl2. The PCR thermoprofile included denaturing at 94 °C for 2 min, followed by 10
cycles of 94 °C for 30 s, 30 s at 60 °C ramping down to 50 °C
at 1 °C per cycle, and then 72 °C for 1 min 30 s, and 30
cycles in which the annealing temperature remained at
50 °C, then a final extension at 72 °C for 7 min. Sequencing
reactions were performed using the DYEnamic ET Dye terminator cycle Sequencing Kit for MegaBACE (GE Biosciences) and electrophoresed on the MegaBACE 1000
autosequencer (GE Biosciences). Sequences were proofread
and assembled with the aid of the software SEQUENCHER
(V4.7, Gene Codes Corporation). Indels, despite containing
evolutionary information (e.g., Simmons and Ochoterena
2000), were removed from the data set because the coalescent simulation used (see below) could not model indel
mutation. To determine individual haplotypes, sequences
of nuclear loci were phased into different haplotypes for
each individual using the software PHASE (Stephens
et al. 2001; Stephens and Scheet 2005), which implements
likelihood reconstruction of haplotypes from population
data. For each locus, 20,000 iterations were performed
476
while thinning at every 20 steps, and the first 100 samples
were discarded as burn-in; individuals whose genotype at
a given locus could not be phased with probability of 60%
were excluded for that locus; one sample was excluded for
locus PPP1R12A, ARF1, ZNF148, and ABCA1. To ascertain
the number of chromosomes sampled at Z-linked loci,
we sex-typed individuals that were typed at the three
Z-linked loci using primers published in Hornfeldt et al.
(2000) and PCR conditions in Yeung et al. (2006).
Analyses of Nuclear Polymorphism
Genetic diversity, recombination, linkage, and neutrality
tests: For Pl. regia and Pl. minor, we calculated genetic diversity indices such as the number of segregating sites (s)
and haplotypes (h), nucleotide diversity (mean pairwise
number of nucleotide differences per gene, p), and Watterson’s theta (hW) using DnaSP 4.9.1 (Rozas et al. 2003)
at each of the 20 nuclear loci. In addition, we used the
program LDhat2.1 (McVean et al. 2002) to estimate the recombination parameter q (q 5 4Ner, where r is the recombination rate) that would be incorporated into the
subsequent coalescent simulations for each locus in the
two species. Interspecific summary statistics, including
the number of polymorphic sites shared between the
two species (Ss), unique to each species (SPr for Pl. regia
and SPm for Pl. minor), fixed in either species (Sf) as defined
in Wakeley and Hey (1997), and total number of polymorphic sites across the two species (St), which is the sum of
the former three classes of polymorphism, were also calculated using DnaSP. Following Becquet and Przeworski
(2007), a mutation was defined as ancestral or derived
by comparing with sequences of Pl. leucorodia, Pl. flavipes,
and Pl. ajaja, where available, using parsimony criteria
(Maddison DR and Maddison PW 2000) within a phylogenetic framework. We also calculated the index of population
differentiation Fst (Hudson et al. 1992, equation 3) using
DnaSP. Within each species, Hardy–Weinberg equilibrium
(HWE) of each polymorphic locus and gametic disequilibrium between pairs of polymorphic loci (excluding female
genotypes at Z-linked loci) were tested using the program
GENEPOP (Raymond and Rousset 1995); the Benjamini–
Yekutieli correction (Benjamini and Yekutieli 2001) was applied
to control false discovery rate in multiple tests with a value set
as 0.05.
For each locus, the average number of nucleotide differences between the two focal species was calculated using
MEGA 4 (Tamura et al. 2007); this information was then
used in a multilocus Hudson–Kreitman–Aguadé (HKA;
Hudson et al. 1987) test for molecular neutrality across
all loci, as implemented in the program HKA (available from
Jody Hey at http://lifesci.rutgers.edu/;heylab/HeylabSoftware
.htm). Two other tests of neutrality, the D test of Tajima
(Tajima 1989, equation 38) and the D test of Fu and Li (Fu
and Li 1993, equation 32), were also conducted using
DnaSP.
Approximate Bayesian inference of speciational demography: Parameters considered in the four demographic
MBE
Testing Founder Effect Speciation · doi:10.1093/molbev/msq210
models (fig. 2) included the long-term Ne of each of the two
species (NPr for Pl. regia and NPm for Pl. minor) and of the
ancestral species (NA), divergence time in number of generations ago (T1), reciprocal postdivergence gene flow between the two species in the bottleneck with gene flow
model, and the gene flow model (MPrPm from Pl. regia to
Pl. minor and MPmPr from Pl. minor to Pl. regia; M 5 4
Nem, where m is the proportion of migrants each generation) until some time ago (T2 for that going from Pl. regia to
Pl. minor and T3 for that the other way around in units of
generations) and size of founder population of Pl. regia as
a proportion B of NPr until Tb generations ago (founder effect speciation model and isolation model). To select the
model that best explains the genetic polymorphism observed in the two species, 1,000,000 multilocus genetic data
sets were simulated using the program MSNSAM (Hudson
2002; Ross-Ibarra et al. 2008), given parameter values randomly drawn from uniform prior distributions while assuming the observed sample size, fragment length, and
mutation rate for each locus. Each equilibrium Ne (NPr,
NPm, and NA) was a uniformly distributed random variable
between 100 and 100,000 scaled by the locus-specific mutation rate (l) to yield the h (4 Nel for autosomal loci and 3
Nel for Z-linked loci) value required by the program. The
mutation rate of each locus was deduced as follows
(supplementary table S3, Supplementary Material online):
divergence, in net Tamura–Nei distance, of Pl. regia versus
Pl. flavipes and Pl. minor versus Pl. flavipes was averaged
and compared with average divergence at the mitochondrial gene CYTB (0.0849; Chesser et al. 2010) to obtain a ratio;
mutation rate per locus per generation was then obtained
by multiplying this ratio by the Ciconiiformes CYTB molecular clock of 1.05 10 8 substitution per site per year
(Weir and Schluter 2008), locus length, and a generation
time of 10.7 years (deduced for Pl. minor, Yeung et al.
2006). M was a random variable between 0.0001 and 500.
T1 was a uniformly distributed random variable between
50 and 500,000 generations. T2, T3, and Tb were taken to
be a fraction of T1. All time estimates were scaled to be
in units of 4 NPr generations as required by the program.
B was a uniformly distributed random variable fraction of
NPr. Priors for the recombination rate q were set as uniformly distributed between the lowest and the highest value
estimated in the two focal species at each locus.
For each simulated data set, 14 summary statistics were
calculated using a personally developed program written in
C/Cþþ (available from authors upon request): s, h, p, and
hW for each species as well as St, Ss, Sf, SPr, SPm, and Fst. The
average of each summary statistic over all nuclear loci was
normalized and compared with the observed multilocus
average of each summary statistic to obtain the Euclidean
distance according to which summary statistics among the
0.005 percentile closest to the observed statistics (i.e., tolerance 5 0.005) were retained and weighted by the Epanechnikov kernel (Beaumont et al. 2002). Euclidean
distances were calculated and regression procedures were
performed using modified R codes provided by Beaumont
(http://www.rubic.rdg.ac.uk/;mab/stuff/). We estimated
the posterior probabilities of models by treating each
model as a categorical variable in a weighted multinomial
logistic regression procedure (Beaumont 2008). Bayes factor was calculated for each model by dividing the posterior
probability by that of the null model, the founder effect
speciation model; a ratio .3 was considered indicative
of a statistically significant difference (Jeffreys 1961). For
the model(s) with the highest posterior probability, we simulated an additional 4,000,000 data sets and estimated demographic parameters by adopting the ABC method
described in Beaumont et al. (2002): demographic parameters were tangent-transformed to ensure that the posterior density fell within the range of the prior distribution
(suggested by Hamilton et al. 2005); the Epanechnikov kernel and local linear regression smoothing method was used
to calculate the posterior density from the retained parameter values (tolerance 5 0.005). The median of the posterior distribution, which has been suggested to be associated
with lower mean square error (Li et al. 2010), was used as
the best estimate of each parameter.
To evaluate the goodness of fit of the demographic parameters to the observed genetic data, we used the approach suggested by Thornton and Andolfatto (2006):
we randomly resampled 10,000 values from the 95% highest probability density of each demographic parameter of
the favored model(s) and then used them to perform coalescent simulations for which the multilocus mean of the
same set of summary statistics, as used previously in the
ABC procedure, were calculated. The 95% credible interval
(CI) of the simulated summary statistics was compared
with the observed summary statistics.
Results
Genetic diversity, Recombination, Linkage, and
Neutrality Test in the Focal Species
Sequences at 20 nuclear loci for Pl. regia, Pl. minor, Pl. leucorodia, Pl. flavipes, and Pl. ajaja individuals were produced
and characterized in Pl. regia and Pl. minor as shown in
supplementary table S4, Supplementary Material online
(GenBank accession numbers of the original sequences
are provided; processed sequences, which are trimmed
and without indels, can be obtained from the authors upon
request). Although there were roughly twice as many samples of Pl. minor as Pl. regia, levels of genetic diversity were
similar between the two species. None of the loci showed
significant linkage disequilibrium with another or deviated
consistently and significantly from HWE across species. The
HKA test, the D test of Tajima, and the D test of Fu and Li
indicated that none of the loci significantly deviated from
the expectation of neutrality (supplementary table S4,
Supplementary Material online).
Speciational Demography
The model posterior probability was extremely low for
the null model, the founder effect speciation model
(1.55 10 8), low for the isolation model (0.13), and highest for the two models characterized by postdivergence
477
MBE
Yeung et al. · doi:10.1093/molbev/msq210
Table 1. Posterior Median and 95% CI (rounded to the closest thousand for numbers larger than 1,000) for Demographic Parameters in the
Bottleneck with Gene Flow Model and the Gene Flow Model.
Bottleneck with gene
Flow model
Gene flow model
NPr
12,000
7,000
44,000
13,000
9,000
19,000
NPm
12,000
9,000
17,000
12,000
8,000
17,000
NA
42,000
2,000
97,000
38,000
2,000
96,000
T1
197,000
50,000
478,000
186,000
45,000
477,000
T2
33,000
16,000
254,000
33,000
17,000
274,000
T3
36,000
19,000
283,000
32,000
15,000
256,000
MPrPm
262
13
488
258
13
488
MPmPr
237
12
486
243
12
486
Tb
95,000
4,000
401,000
—
—
—
B
0.55
0.04
0.98
—
—
—
NOTE.—Posterior median and 5% lower bound and 95% upper bound are indicated in the first, second, and third rows right of the model name, respectively. NPr and NPm,
long-term equilibrium effective population size of Pl. regia and Pl. minor, respectively; NA, population size of the ancestral species; T1, divergence time in number of
generations ago; T2 and T3, time of cessation of gene flow from Pl. regia to Pl. minor and from Pl. minor to Pl. regia; MPrPm and MPmPr, asymmetrical gene flow from Pl. regia
to Pl. minor and from Pl. minor to Pl. regia; speciational bottleneck in Pl. regia as a proportion B of NPr that lasted until Tb generations ago.
gene flow (bottleneck with gene flow model: 0.50; gene flow
model: 0.37). The difference in model probabilities was significant between the null model and the three alternative
models (Bayes factor 5 8.39 106, 23.87 106, and 32.26 106 when compared with the isolation model, the bottleneck with gene flow model, and the gene flow model,
respectively) and between the isolation model and the gene
flow model (Bayes factor 5 3.85), but not between the isolation model and the gene flow model (Bayes factor 5 2.85)
or between the bottleneck with gene flow model and the
gene flow model (Bayes factor 5 1.35). We therefore simulated an additional 4,000,000 data sets and estimated posterior densities of demographic parameters only for the
bottleneck with gene flow model and the gene flow model.
All parameter estimates and posterior density distributions
were highly congruent between the two models (table 1
and fig. 3). NPr, NPm, T1, T2, and T3 were well resolved with
completely bell-shaped distributions (fig. 3A and C). On the
other hand, NA, B, Tb, MPrPm, and MPmPr were less clearly resolved with distributions characterized by moderately sloped
nonzero peaks (fig. 3B–D).
The observed multilocus mean and variance of each of
the 14 summary statistics fell within the 95% CI of the predictive posterior distribution under the bottleneck with
gene flow model and the gene flow model (table 2), indicating that the demographic parameters inferred from
these models could produce genetic polymorphism similar
to that observed in contemporary Pl. regia and Pl. minor
species.
Our results showed that even if there had been a population size reduction in the emergent Pl. regia, the founder
population size of 480–12,000 individuals (a conservative
estimate obtained by multiplying the distribution median
NPr 5 12,000 by the 95% CI in the bottleneck with gene
flow model) was nevertheless rather large for a founder
population (fig. 3). More intriguingly, gene flow persisted
for a considerably long period following the divergence
of Pl. regia and Pl. minor and ceased only relatively recently.
Cessation of gene flow, obtained by multiplying the distribution median in the bottleneck with gene flow and gene
flow models by a generation time of 10.7 years, was estimated to have occurred 351,281–354,834 years ago from
Pl. regia to Pl. minor and 337,810–384,537 years ago from
Pl. minor to Pl. regia.
478
Discussion
Nullifying Founder Effect Speciation: Speciation
with Gene Flow
The most prominent finding of this study is that the postdivergence gene flow between Pl. regia and Pl. minor
spanned more than 80% of the divergence history. This
clearly violates the assumption of founder effect speciation
that gene flow ceases completely following initial divergence. A speciational bottleneck in the emergent Pl. regia
was also shown to be unlikely. The lower resolution of the
bottleneck proportion B may be partly attributed to postdivergence gene flow between two diverging lineages,
which continuously introduced genetic variation to the
emergent Pl. regia, thereby masking the effect of drift even
if there had been population size reduction. Nevertheless,
the most conservative estimate of the size of the bottlenecked population was a minimum of 480 individuals
and possibly much larger, which barely supported founder
effect speciation.
The four demographic models constructed here can be
viewed as more complex variants of the IM model (Nielsen
and Wakeley 2001). Because the results of ABC are thought
to approximate those obtained via full likelihood analysis
(Becquet and Przeworski 2007), the effects of violating assumptions of the IM model in full likelihood analyses or
ABC could potentially be used to discuss the consequences
of model violation in this study. One of the major assumptions here is that the ancestral species consisted of a single
panmictic population. It might be argued that postdivergence gene flow detected in this study could be the artifact
of ancestral structure or early parapatry followed by divergence in isolation. However, Becquet and Przeworski (2009)
have shown that IM model–based analyses via both ABC
and full likelihood would not detect postdivergence gene
flow when these assumptions were violated. Postdivergence gene flow is also unlikely to be confounded with contemporary gene flow because the two species are disjunctly
distributed and there is no observable dispersal to each
other’s range; our data indicate that gene flow ceased before present time.
Potential geographic structure within extant species also
would have little effect on the results. It is possible that
a third closely related species, in this case Pl. leucorodia,
Testing Founder Effect Speciation · doi:10.1093/molbev/msq210
MBE
FIG. 3 Posterior distributions of demographic parameters in the bottleneck with gene flow model and the gene flow model, indicated by gray
and black lines, respectively. (A) Long-term effective population size NPr and NPm for Pl. regia and Pl. minor, respectively. (B) Effective
population size of the ancestral species NA. (C) Divergence time T1 in number of generations ago, time of cessation of gene flow from Pl. regia
to Pl. minor T2 and that the other way around T3, and time of speciational bottleneck Tb generations ago. (D) Gene flow from Pl. regia to Pl.
minor MPrPm and from Pl. minor to Pl. regia, MPmPr. (E) Speciational bottleneck in Pl. regia as a proportion B of NPr.
had maintained gene flow with the focal species pair
throughout their divergence or that potential geographic
variation, particularly of the less well-sampled species
Pl. regia, is missing from the data set. However, recent
simulation studies (Strasburg and Rieseberg 2010) showed
that the IM model is quite robust to violation of the assumptions of gene flow solely between the focal species
pair and population admixture within species. Therefore,
479
MBE
Yeung et al. · doi:10.1093/molbev/msq210
Table 2. Predictive Posterior Summary Statistics for Pl. regia and Pl. minor based on 10,000 Coalescent Simulations under the Demographic
Parameters Inferred from the Two Models Permitting Postdivergence Gene Flow.
Mean
Platalea regia
Pl. minor
Variance
Pl. regia
Pl. minor
s
h
uW
p
SPr or SPm
St
1.20–3.40
1.35–4.25
2.75
1.50–4.00
1.60–4.60
3.05
2.05–3.45
2.10–3.65
2.95
2.30–4.10
2.35–4.15
3.40
0.29–0.96
0.38–1.23
0.81
0.31–0.95
0.38–1.10
0.72
0.32–1.04
0.38–1.36
0.81
0.33–1.04
0.36–1.27
0.93
0
0–2.45
1.80
0
0–2.45
2.10
1.75–4.40
3.50–8.40
5.05
1.32–12.37
1.61–21.00
4.51
1.78–14.21
1.92–22.13
8.47
0.88–4.03
0.96–4.27
1.31
1.25–6.04
1.27–6.06
3.83
0.14–1.12
0.14–1.85
0.39
0.13–0.97
0.12–1.4
0.46
0.12–1.62
0.15–2.82
0.41
0.12–1.58
0.14–2.75
1.12
0
0–8.34
3.96
0
0–8.34
5.46
2.13–16.09
0–8.34
5.46
Ss
Sf
Fst
1.0–3.0
0–1.60
0.95
0
0–2.45
0.20
0
0–2.45
0.47
1.04–10.64
0–7.67
1.52
0
0–8.34
0.48
0
0–8.34
0.07
NOTE.—The 95% CI of mulitilocus (20 loci) mean and variance of summary statistics simulated under the bottleneck with gene flow model are indicated in the first row
right of the species’ name and those under the gene flow model are listed on the second row; observed values of each summary statistics are provided on the third row.
Summary statistics evaluated include the number of segregating sites (s), number of haplotypes (h), Watterson’s h (hW), mean number of pairwise nucleotide difference per
gene (p), and number of mutations polymorphic in either species (SPr and SPm for Pl. regia and Pl. minor, respectively). Total number of polymorphic sites (St), number of
polymorphic sites shared between species (Ss), number of mutations fixed in either species (Sf), and Fst between species are listed in the rows for Pl. regia.
whether or not our samples included geographic variation
would have little effect on the results.
Recent research based on the IM model has uncovered
many incidences of postdivergence gene flow (e.g.,
Machado et al. 2002; Hey 2005; Won and Hey 2005; Kondo
et al. 2008; Niemiller et al. 2008; Stadler et al. 2008; Li et al.
2010), and our results further suggest that postdivergence
gene flow can last for a protracted period of time. The
most salient implication of prolonged postdivergence
gene flow between Pl. regia and Pl. minor is that most
of the speciation process (around 80% of the divergence
history) might not have been allopatric; rather, selection
(e.g., sexual or differential adaptive selection) may have
been more important than physical isolation in facilitating divergence. Selection-driven differentiation in the face
of gene flow has been supported by studies identifying
ecologically based divergent selection that facilitates reproductive isolation between closely related taxa (e.g.,
as reviewed in Rundle and Nosil 2005; Hendry et al.
2007). In these cases, differentiation is initiated at traits
that may lead to further reduction in gene flow, depending on their degree of association with reproductive isolation. Such a process is even more conceivable at the
genic level with the ‘‘porous genome’’ analogy (Wu
2001): gene flow between two populations ceases earliest
at loci where selection is strongest, whereas the rest of the
genome freely introgresses; as genomic regions of differential adaptation expand and develop linkage through coadaptation with one another, gene flow between the two
genomes is reduced until the mechanism of reproductive
isolation arises to prevent further genetic exchange. Owing to the growing availability of genomic data, this concept has been corroborated empirically, for example, by
the higher divergence level observed at genes of putative
adaptive value (e.g., candidate genes, coding, or regulatory
sequences) relative to that at presumably neutral loci (e.g.,
noncoding sequences, Osada and Wu 2005) or by variable
480
degree of isolation across different genomic regions (e.g.,
Turner et al. 2005).
Speciational Hypotheses Compatible with
Demographic Histories of Pl. regia and Pl. minor
Prolonged postdivergence gene flow suggests that for most
of their divergence history, Pl. regia and Pl. minor bred in
sympatry, at least partially. This calls for alternative speciational scenarios compatible with divergence with gene
flow. One such possibility is the evolution of a novel migratory route that led some individuals to winter in or near
Australasia, such that despite occasional interbreeding, reproductive isolation between populations on different
sides of the migratory divide could have developed by assortative, possibly allochronic mating (e.g., in the blackcap
Sylvia atricapilla, Bearhop et al. 2005), and by differential
selection pressure in the different winter quarters. Such
variation in migratory behavior may be analogous to that
observed in contemporary spoonbill populations. For example, in Pl. leucorodia, there are populations with distinctive
migratory routes as well as resident subspecies (Pl. leucorodia
leucorodia, Pl. l. balsaci, and Pl. l. archeri, fig. 1). The ubiquity
of evolutionary lability (reviewed in Pulido 2007) and significant fitness consequences of migratory behavior in birds
make it an intriguing candidate agent of avian speciation
(Bearhop et al. 2005).
Additionally, and nonexclusively, the species could have
diverged through exploitation of novel coastal habitats by
Pl. minor. Although both coastal and inland populations
exist in Pl. regia, all known Pl. minor populations occur exclusively in coastal habitats. Usage of coastal habitats may
confer advantages such as relative climatic stability (e.g.,
Welker 2000) or reduced exposure to parasites (Figuerola
1999) and may promote spatiotemporal divergence in wintering and breeding ecology, including migratory behavior.
In fact, coastal and inland habitats may be so inherently
Testing Founder Effect Speciation · doi:10.1093/molbev/msq210
distinct that population differentiation in many organisms
has been reported along the coastal-inland gradient (e.g., in
fruticose lichen Ramalina menziesii, Boucher and Nash
1990; in three-spined stickleback Gasterosteus aculeatus,
Raeymaekers et al. 2007; in the yellow monkeyflower Mimulus guttatus, Lowry et al. 2008), although these species are
less mobile than the spoonbills. Moreover, some of these
studies have also shown that differentiation takes place in
the face of considerable gene flow (e.g., in terrestrial garter
snake Thamnophis elegans, Manier et al. 2007).
Conclusions
Our study shows that speciation models involving protracted periods of postdivergence gene flow are much
more probable explanations for the divergence between
Pl. regia and Pl. minor than is founder effect speciation.
As an increasing number of studies reveal postdivergence
gene flow, future speciation studies should focus on mechanisms that underlie adaptive differentiation in the face of
gene flow. Migratory birds, which are the most mobile of
terrestrial vertebrates and therefore least limited by geographic barriers, may yield many insights into such a speciation process.
Supplementary Material
Supplementary tables S1–S4 are available at Molecular Biology and Evolution online (http://www.mbe.oxfordjournals
.org/).
Acknowledgments
We thank the American Museum of Natural History of the
United States and the Commonwealth Scientific and Industrial Research Organization of Australia for providing
samples. Use of trade, product, or firm names does not imply endorsement by the U.S. Government. We thank
D. Chang for assistance in writing codes for calculating
summary statistics from the simulated data. We are also
grateful for the valuable help and comments by M. Beaumont and technical support by P. Sweet and Y.-T. Lin. This
work was supported by grants to S.-H.L. from the National
Science Council, Taiwan, R.O.C.
References
Backström N, Brandström M, Gustafsson L, Qvarnström A, Cheng H,
Ellegren H. 2006. Genetic mapping in a natural population of
collared flycatchers (Ficedula albicollis): conserved synteny but
gene order rearrangements on the avian Z chromosome.
Genetics 174:337–386.
Balakrishnan CN, Edwards SV. 2009. Nucleotide variation, linkage
disequilibrium and founder-faciliated speciation in wild populations of the zebra finch (Taeniopygia guttata). Genetics
181:645–660.
Barraclough TG, Vogler AP. 2000. Detecting the geographical
pattern of speciation from species-level phylogenies. Am Nat.
155:419–434.
Bearhop S, Fiedler W, Furness RW, Votier SC, Waldron S, Newton J,
Gabriel JB, Berthold P, Farnsworth K. 2005. Assortative mating as
a mechanism for rapid evolution of a migratory divide. Science
310:502–504.
MBE
Beaumont MA. 2008. Joint determination of topology, divergence
time and immigration in population tree. In: Matsumura S,
Forster P, editors. Simulation, genetics and human prehistory.
Cambridge: University of Cambridge. p. 135–154.
Beaumont MA, Zhang W, Balding DJ. 2002. Approximate Bayesian
computation in population genetics. Genetics 162:2025–2035.
Becquet C, Przeworski M. 2007. A new approach to estimate
parameters of speciation models with application to apes.
Genome Res. 17:1505–1519.
Bequet C, Przeworski M. 2009. Learning about modes of speciation
by computational approaches. Evolution 63:2547–2562.
Benjamini Y, Yekutieli D. 2001. The control of the false discovery
rate under dependency. Ann Stat. 29:1165–1188.
Boucher VL, Nash TH III. 1990. Growth patterns in Ramalina
menziesii in California: coastal vs. inland populations. Bryologist
93:295–302.
Chesser RT, Yeung CKL, Yao CT, Tian XH, Li SH. 2010. Molecular
phylogeny of the spoonbills (Aves: Threskiornithidae) based on
mitochondrial DNA. Zootaxa 2603:53–60.
Chesser RT, Zink RM. 1994. Modes of speciation in birds: a test of
Lynch’s method. Evolution 48:490–497.
Clegg SM, Degnan SM, Kikkawa J, Moritz C, Estoup A, Owens IPF.
2002. Genetic consequences of sequential founder events by an
island-colonizing bird. Proc Natl Acad Sci. 99:8127–8132.
Coyne JA, Orr HA. 2004. Speciation. Sunderland (MA): Sinauer
Associates.
Figuerola J. 1999. Effects of salinity on rates of infestation of
waterbirds by haematozoa. Ecography 22:681–685.
Friesen VL, Anderson DJ. 1997. Phylogeny and evolution of the
Sulidae (Aves: pelecaniformes): a test of alternative modes of
speciation. Mol Phylogenet Evol. 7:252–260.
Fu YX, Li WH. 1993. Statistical tests of neutrality of mutations.
Genetics 133:693–709.
Gavrilets S. 2004. Fitness landscapes and the origin of species.
Princeton (NJ): Princeton University Press.
Gemmell NJ, Akiyama S. 1996. An efficient method for the
extraction of DNA from vertebrate tissues. Trends Genet.
12:338–339.
Grant PR, Grant BR, Peetren K. 2001. A population founded by
a single pair of individuals: establishment, expansion, and
evolution. Genetica 112-113:359–382.
Hamilton G, Stoneking M, Excoffier L. 2005. Molecular analysis
reveals tighter social regulation of immigration in patrilocal
populations than in matrilocal populations. Proc Natl Acad Sci.
102:7476–7480.
Hancock JA, Kushlan JA, Kahl MP. 1992. Storks, ibises and spoonbills
of the world. London: Academic Press.
Hendry AP, Nosil P, Rieseberg LH. 2007. The speed of ecological
speciation. Funct Ecol. 21:455–464.
Hey J. 2005. On the number of New World founders: a population
genetic portrait of the peopling of the Americas. PLoS Biol.
3:e193.
Hornfeldt B, Hipkiss T, Fridolfsson A-K, Eklund U, Ellegren H. 2000.
Sex ratio and fledging success of supplementary-fed Tengmalm’s
owl broods. Mol Ecol. 9:187–192.
Hudson R. 2002. Generating samples under a Wright-Fisher neutral
model of genetic variation. Bioinformatics 18:337–338.
Hudson RR, Kreitman M, Aguadé M. 1987. A test of neutral molecular
evolution based on nucleotide data. Genetics 116:153–159.
Hudson RR, Slatkin M, Maddison WP. 1992. Estimation of levels of
gene flow from DNA sequence data. Genetics 132:583–589.
Jeffreys H. 1961. The theory of probability, 3rd ed. Oxford: Oxford
University Press.
Knowles LL, Futuyma DJ, Eanes WF, Rannala B. 1999. Insight into
speciation from historical demography in the phytophagous
beetle genus Orphraella. Evolution 53:1846–1856.
481
Yeung et al. · doi:10.1093/molbev/msq210
Kondo B, Peters JL, Rosensteel BB, Omland KE. 2008. Coalescent
analyses of multiple loci support a new route to speciation in
birds. Evolution 62:1182–1191.
Li JW, Yeung CKL, Tsai PW, et al. (11 co-authors). 2010. Rejecting
strictly allopatric speciation on a continental island: prolonged
post-divergence gene flow between Taiwan (Leucodioptron
taewanus, Passeriformes Timaliidae) and Chinese (L. canorum
canorum) hwameis. Mol Ecol. 19:494–507.
Losos JB, Glor RE. 2003. Phylogenetic comparative methods and the
geography of speciation. Trends Ecol Evol. 18:220–227.
Lowry DB, Rockwood RC, Willis JH. 2008. Ecological reproductive
isolation of coast and inland races of Mimulus guttatus. Evolution
62:2196–2214.
Machado CA, Kliman RM, Markert JA, Hey J. 2002. Inferring the
history of speciation from multilocus DNA sequence data: the
case of Drosophila pseudoobscura and close relatives. Mol Biol
Evol. 19:472–488.
Maddison DR, Maddison PW. 2000. MacClade 4: analysis of
phylogeny and character evolution. Sunderland (MA): Sinauer
Associates.
Manier MK, Seyler CM, Arnold SJ. 2007. Adaptive divergence within
and between ecotypes of the terrestrial garter snake, Thamnophis elegans, assessed with FST- QST comparisons. J Evol Biol.
20:1705–1719.
Mattern MY, McLennan DA. 2000. Phylogeny and speciation of
felids. Cladistics 16:232–253.
Mayr E. 1942. Systematics and the origin of species. New York:
Columbia University Press.
Mayr E. 1954 Changes in genetic environment and evolution. In:
Evolution as a process. Huxley JS, Hardy AC, Ford EB, editors.
London: Allen & Unwin. p. 156–180.
Mayr E. 1963. Animal species and evolution. Cambridge (MA):
Harvard University Press.
Mayr E. 1970. Populations, species and evolution. Cambridge (MA):
Harvard University Press.
Mayr E. 1976. Evolution and the diversity of life: selected essays.
Cambridge (MA): Harvard University Press.
Mayr E. 1982. Speciation and macroevolution. Evolution
36:1119–1132.
McVean G, Awadalla P, Fearnhead P. 2002. A coalescent-based
method for detecting and estimating recombination from gene
sequences. Genetics 160:1231–1241.
Moya A, Galiana A, Ayala FJ. 1995. Founder-effect speciation theory:
failure of experimental corroboration. Proc Natl Acad Sci.
92:3983–3986.
Nielsen R, Wakeley J. 2001. Distinguishing migration from isolation:
a Markov chain Monte Carlo approach. Genetics 158:885–896.
Niemiller ML, Fitzpatrick BM, Miller BT. 2008. Recent divergence
with gene flow in Tennessee cave salamanders (Plethodontidae:
gyrinophilus) inferred from gene genealogies. Mol Ecol.
17:2258–2275.
Osada N, Wu C-I. 2005. Inferring the mode of speciation from
genomic data. Genetics 169:259–264.
Paulay G. 2002. Diversification in the tropical Pacific: comparisons
between marine and terrestrial systems and the importance of
founder speciation. Integr Comp Biol. 42:922–934.
Pulido F. 2007. The genetics and evolution of avian migration.
Bioscience 57:165–174.
Pritchard JK, Seielstad MT, Perez-Lezaun A, Feldman MW. 1999.
Population growth of human Y chromosomes: a study of
Y chromosome microsatellites. Mol Biol Evol. 16:1791–1798.
Provine W. 1989. Founder effects and genetic revolutions in
microevolution and speciation. A historical perspective. In:
Gidding LV, Kaneshiro KY, Anderson WW, editors. Genetics,
speciation and the founder principle. New York: Oxford
University Press. p. 43–76.
482
MBE
Raeymaekers J, Van Houdt J, Larmuseau M, Geldof S, Volckaert F.
2007. Divergent selection as revealed by PST and QTL-based
FST in three-spined stickleback (Gasterosteus aculeatus)
populations along a coastal-inland gradient. Mol Ecol.
16:891–905.
Raymond M, Rousset F. 1995. GENEPOP (version 1.2): population
genetics software for exact tests and ecumenicism. J Hered.
86:248–249.
Ross-Ibarra J, Wright S, Foxe J, Kawabe A, DeRose-Wilson L. 2008.
Patterns of polymorphism and demographic history in natural
populations of Arabidopsis lyrata. PLoS ONE. 3:e2411.
Rozas J, Sánchez-DelBarrio JC, Messegyer X, Rozas R. 2003. DnaSP,
DNA polymorphism analyses by the coalescent and other
methods. Bioinformatics 19:2496–2497.
Rundle HD, Nosil P. 2005. Ecological speciation. Ecol Lett. 8:336–352.
Simmons MP, Ochoterena H. 2000. Gaps as characters in sequencebased phylogenetic analysis. Syst Biol. 49:369–381.
Stadler T, Arunyawat U, Stephan W. 2008. Population genetics of
speciation in two closely related wild tomatoes. (Solanum
Section Lycopersicon). Genetics 178:339–350.
Stephens M, Scheet P. 2005. Accounting for decay of linkage
disequilibrium in haplotype inference and missing-data imputation. Am J Hum Genet. 76:449–462.
Stephens M, Smith N, Donnelly P. 2001. A new statistical method
for haplotype reconstruction from population data. Am J Hum
Genet. 68:978–989.
Strasburg JL, Rieseberg LH. 2010. How robust are ‘‘Isolation with
Migration’’ analyses to violation of the IM model? A simulation
study. Mol. Biol. Evol. 27:297–310.
Tajima F. 1989. Statistical method for testing the neutral mutation
hypothesis by DNA polymorphism. Genetics 123:585–595.
Tamura K, Dudley J, Nei M, Kumar S. 2007. MEGA4: Molecular
Evolutionary Genetics analysis (MEGA) software. version 430.
Mol. Biol. Evol. 24:1596–1599.
Tavare S, Balding DJ, Griffiths RC, Donnelly P. 1997. Inferring
coalescence times from DNA sequence data. Genetics
145:505–518.
Thornton K, Andolfatto P. 2006. Approximate Bayesian
inference reveals evidence for a recent, severe bottleneck in
a Netherlands population of Drosophila melanogaster. Genetics
172:1607–1619.
Turner TL, Hahn MW, Nuzhdin SV. 2005. Genomic islands of
speciation in Anopheles gambiae. PLoS Biol. 3:e285.
Vincek V, O’Huigin C, Satta Y, Takahata N, Boag PT, Grant PR,
Grant BR, Klein J. 1997. How large was the founding population
of Darwin’s finches? Proc R Soc B Biol Sci. 264:111–118.
Wakeley J, Hey J. 1997. Estimating ancestral population parameters.
Genetics 145:847–855.
Walsh HE, Jones IL, Friesen VL. 2005. A test of founder effect
speciation using multiple loci in the auklets (Aethia spp.).
Genetics 171:1885–1894.
Weir JW, Schluter D. 2008. Calibrating the avian molecular clock.
Mol Ecol. 17:2321–2328.
Welker JM. 2000. Isotopic (d18O) characteristics of weekly precipitation collected across the USA: an initial analysis with
application to water source studies. Hydrol Process. 14:1449–1464.
Won Y-J, Hey J. 2005. Divergence population genetics of
chimpanzees. Mol Biol Evol. 22:297–307.
Wright S. 1931. Evolution in Mendelian populations. Genetics
16:97–159.
Wu C-I. 2001. The genic view of the process of speciation. J Evol Biol.
14:851–865.
Yeung CK-L, Yao C-T, Hsu Y-C, Wang J-P, Li S- H. 2006. Assessment
of the historical population size of an endangered bird, the
black-faced spoonbill (Platalea minor) by analysis of mitochondrial DNA diversity. Anim Conserv. 9:1–10.