Genetic equilibrium despite habitat ... - Wiley Online Library

5 downloads 0 Views 242KB Size Report
*Laboratory of Animal Ecology, Department of Biology, University of Antwerp, ... of Kenya, Nairobi, Kenya; ‡Department of Zoology, University of Oxford, Edward.
Molecular Ecology (2004) 13, 1409–1421

doi: 10.1111/j.1365-294X.2004.02175.x

Genetic equilibrium despite habitat fragmentation in an Afrotropical bird

Blackwell Publishing, Ltd.

P E T E R G A L B U S E R A ,* M W A N G I G I T H I R U ,†,‡ L U C L E N S § and E R I K M A T T H Y S E N * *Laboratory of Animal Ecology, Department of Biology, University of Antwerp, Universiteitsplein 1, B-2610 Wilrijk, Belgium; †Department of Ornithology, National Museums of Kenya, Nairobi, Kenya; ‡Department of Zoology, University of Oxford, Edward Grey Institute of Field Ornithology, Oxford, UK; §Laboratory of Terrestrial Ecology, Department of Biology, Ghent University, Ghent, Belgium

Abstract We examined the effects of habitat fragmentation of the white-starred robin Pogonocichla stellata metapopulation in the Taita Hills archipelago, a hotspot for biodiversity which was fragmented ~40 years ago. Using seven microsatellite markers, we analysed the robin’s genetic structure and tested for equilibrium between migration and drift (testing the probability of decreased dispersal) as well as between mutation and drift (test for recent reduction in effective population size, i.e. bottlenecks). This metapopulation was found to retain relatively high levels of genetic variability (HE between 0.63 and 0.71) and to be in migration–drift equilibrium, suggesting that increased isolation between fragments did not have much effect on the dispersal between them. Furthermore, this equilibrium test greatly enhanced the interpretation of parameters (e.g. FST) assumed to have reached an equilibrium value. In contrast to previous findings on the related and sympatric Taita thrush Turdus helleri (which is critically endangered), there were no indications for recent bottlenecks in any of the robin subpopulations. This difference can be attributed to the higher dispersal capacity of the robin compared with the thrush (deduced from both the genetic and capture–recapture data). Our results stress the importance of sustained dispersal for species conservation. Keywords: avian ecology, equilibrium, habitat fragmentation, microsatellite, mobility Received 27 October 2003; revision received 9 February 2004; accepted 9 February 2004

Introduction Habitat fragmentation threatens the survival of many species as small and isolated populations are more prone to extinction (Frankham 1998; Saccheri et al. 1998). The impact of habitat fragmentation, i.e. the reduction of large, continuous habitats to small, isolated remnants, on the demographic and genetic structure of natural populations depends on both landscape aspects (level of habitat isolation, degradation and surrounding land uses) and species characteristics (natural population size and density, dispersal power and stress tolerance) (Matthysen et al. 1995; Newton 1998; Bohonak 1999; Desrochers et al. 1999). Often, decreased interfragment dispersal cannot compensate for rising levels of genetic drift as subpopulations become Correspondence: Peter Galbusera. Tel.: +32 3820 2262; Fax: +32 3820 2271; E-mail: [email protected] © 2004 Blackwell Publishing Ltd

smaller and more isolated (e.g. O’Ryan et al. 1998; Slatkin 1987; Uimaniemi et al. 2000). The Taita Hills, a severely fragmented tropical rainforest and biodiversity hotspot in southeast Kenya (Myers et al. 2000), are divided into three distinct mountain isolates (massifs): Sagala Hill (SA), Mbololo Hill (MB) and Dabida Hill (DA). Geographically, DA is separated from MB by a large valley (~ 11 km between the nearest fragments), whereas SA is separated from both DA and MB by the Voi River and 34 km of dry plains. This archipelago is isolated from other highlands by over 80 km of semiarid plains in any direction (Lovett 1985). The indigenous cloud forests on these hills have suffered substantial loss and degradation since the early 1960s (Beentje 1987). At present, the forest persists in a scatter of 12 patches (Fig. 1) — one (Sagala) is located on SA, two (Mbololo and Ronge) on MB and nine (Ngangao, Chawia, Fururu, Ndiwenyi, Mwachora, Macha, Yale, Vuria and Kichuchenyi) on DA — embedded in a

1410 P . G A L B U S E R A E T A L .

Fig. 1 Map of the Taita Hills showing the geographical locations of Dabida and Mbololo Hill. Sagala is situated 34 km to the southeast of these fragments. Arrows indicate dispersing individuals as detected with CMR-methods (except for one, numbered 167, which is a genetically misassigned individual). Individuals between brackets were sighted only temporarily in another fragment. A single individual dispersed from DA to SA (SA has not been drawn). Circles group those fragments that form subpopulations according to the genetic analysis.

mosaic of human settlements, small-holder cultivation plots and plantations of exotic trees (Brooks et al. 1998). Slightly over 400 ha of original forest are retained in these patches: Mbololo (179 ha), Ngangao (136 ha), Chawia (94 ha), while the rest (nine) are tiny remnants (1– 4 ha, one of 8 ha). An analysis of patch size, biomass, stem density, canopy cover, shrub density, stratification and extent of herbaceous ground cover (Wilder et al. 2000), when compared with earlier data (Beentje 1987), showed that Mbololo suffered the least disturbance and forest loss. Ngangao suffered intermediate levels, whereas Chawia and the small patches were most heavily impacted, e.g. with between 95 and 99% deforestation in Sagala and Vuria, respectively (Lens et al. 1999a). Recent studies in the Taita Hills, revealed rising levels of fluctuating asymmetry, mortality and male-biased skew in sex-ratio with increasing levels of habitat disturbance in the critically endangered Taita thrush Turdus helleri (Lens et al. 1998, 1999b, 2002a; Lens & Van Dongen 1999). Fragmentation clearly also affected the genetic variability of the smallest of the three subpopulations of this species (Galbusera et al. 2000a). There remains a pressing need to pin down the mechanisms that shape the genetic structure of such small populations at the landscape level by comparison between different species. This necessitates a

metapopulation-level study focusing on a species that persists throughout the landscape. To achieve this objective, we studied the effects of habitat fragmentation on the white-starred robin Pogonocichla stellata in the Taita Hills. This robin species occurs in montane forests of eastern to southern Africa (Keith et al. 1992). Currently, nine races are recognized, six of which occur in Kenya and northern Tanzania. Sub-species helleri is confined to the Taita Hills and Mt. Kasigau (Moreau 1951; Keith et al. 1992; Zimmerman et al. 1996) where it is relatively common in patches of indigenous forest (Githiru 2002). It is forest dependent (Bennun et al. 1996), and forages at all forest levels, most frequently in the undergrowth and at ant trails (Oatley 1982; Keith et al. 1992). It is territorial and showed homing ability and site fidelity after translocations in south-central Africa (Dowsett 1985; Dowsett & DowsettLemaire 1986). The isolated nature of the Taita Hills (it is surrounded by plains) allows the sampling of a complete metapopulation (in the broad sense of the term; migration occurs between the subpopulations, which do not necessarily experience regular extinction) with little chance of individuals dispersing to or from other populations. We sampled and analysed between 11 and 30 individuals of the whitestarred robin from each of 11 forest fragments in the Taita Hills, using 7 polymorphic microsatellite DNA markers. These genetic data allowed us to verify whether genetic diversity and population structure has been affected by habitat fragmentation in this commonly occurring species. The methods to study the effect of habitat fragmentation on the genetic structure of populations and on dispersal between them, are sometimes controversial because the most commonly inferred parameter (FST) often has not reached its equilibrium value due to the very disturbed nature of the populations (Hedrick 1999; Goodman et al. 2001). Consequently, the underlying specific rates of dispersal (and drift) might be wrongly estimated with a bias towards the previous conditions (Whitlock & McCauley 1999). Hence, it is important to verify whether an equilibrium has been reached. If one knows the time since fragmentation, it might be possible to do so by comparing this time to the expected required time for FST to reach equilibrium. However, the time it takes for FST to reach equilibrium depends on the migration rate and effective population size (Whitlock & McCauley 1999). Unfortunately, to estimate these parameters one often has to assume equilibrium. Recently, a method has become available that tests exactly whether an equilibrium between gene flow and drift has been reached (Ciofi et al. 1999; Beaumont 2001; Goodman et al. 2001; Palo et al. 2001). This Bayesian method explores the most probable demographic and genealogical histories consistent with the sample of chromosomes typed (Ciofi et al. 1999; Beaumont 2001). Its main assumption is a negligible © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

F R A G M E N T A T I O N A N D P O P U L A T I O N E Q U I L I B R I U M 1411 mutation rate. The gene flow model assumes a mutation rate much smaller than the migration rate and the drift model assumes the reciprocal of the mutation rate is much larger than the divergence time. This seems reasonable for our data set because capture–recapture data suggest relatively high dispersal rates and the fragmentation occurred relatively recently. In this study we apply this method to test for equilibrium and hence to ascertain that the estimated dispersal estimates (based on equilibrium) are not biased due to this factor. However, as assumptions other than those regarding equilibrium (no selection, equal population sizes, random migration … ) have to be made to obtain fully unbiased FST estimates (Whitlock & McCauley 1999), we also estimated current dispersal using so-called direct methods. Whereas dispersal estimates obtained by FST or the coalescent method, which is better suited to obtain pairwise dispersal estimates (Beerli & Felsenstein 1999), reflect effective dispersal over many generations (e.g. Monaghan et al. 2001), direct estimates, assessed by assignment tests (Cornuet et al. 1999; Pritchard et al. 2000; Waser et al. 2002), give a more instantaneous picture and include dispersal events not necessarily resulting in the successful reproduction of the individuals in their new population (Greenwood & Harvey 1982; Crochet 1996). Direct methods have been described as inadequate (e.g. Koenig et al. 1996) because they are more dependent on the proportion of the total population sampled. Yet, even indirect methods are not without problems (e.g. Bossart & Prowell 1998; Whitlock & McCauley 1999). Bossart & Prowell (1998) conclude that direct assessment of movement remains the most valid approach. However, application of both direct and indirect methods, because they possess complementary resolution power and concern different time and geographical scales, will improve our understanding of dispersal in natural populations (Crochet 1996; Neigel 2002; Wang & Whitlock 2003). Because the underlying mutation rates of the genetic markers and magnitude of change in effective population size were unknown, the presence or absence of mutation– drift equilibrium could not be predicted a priori. Therefore, assessment of this equilibrium was achieved by estimating effective population sizes based on mutation rates as estimated in other (bird) species (Ellegren et al. 1995; Saino et al. 1997), and by comparing observed with expected heterozygosity (Cornuet & Luikart 1996).

Materials and methods Fieldwork Robins were trapped in all patches except Kichuchenyi where it is absent. Trapping effort was considerable given the high percentage of recaptures, which allowed us to estimate the probability of dispersal (Lens et al. 2002b). © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

According to plumage, the percentage of juveniles sampled, across all fragments, was 30%. Upon first capture, all robins were ringed, colour-banded, aged, measured and blood sampled. All genotyped individuals were sexed by examining the highly conserved W-chromosome linked gene CHDW (Griffiths et al. 1998); Lens et al. (1999a, 1998) describe the methodology in greater detail.

Genotyping Blood samples were collected by brachial vein puncture and stored in 95% ethanol or DMSO. DNA was obtained either by boiling in a 5% Chelex solution (BioRad) after incubation for 90 min at 55 °C in the presence of 100 µg proteinase K (ethanol storage) (Walsh et al. 1991), or by a normal phenol–chloroform extraction also in the presence of 100 µg proteinase K (DMSO storage). Polymerase chain reaction (PCR) amplification was achieved in a 10 µL reaction volume containing ~100 ng DNA, 1× buffer (75 mm Tris– HCl pH 9.0, 20 mm (NH4)2SO4, 0.01% Tween 20), 0.5 U Taq DNA polymerase (Eurogentec), 200 µm dNTPs (Gibco), 1.0–3.0 mm MgCl2 (see Appendix) and 250–500 nm of seven highly variable microsatellite DNA markers (Galbusera et al. 2000b). These markers were originally developed for the following passerines: Parus atricapillus (Pat14; Otter et al. 1998), Chiroxiphia linearis (Ltmr6; McDonald & Potts 1994), Plocepasser mahali (WBSW2 and WBSW9; McRae & Amos 1999), Malurus cyaneus (Mcyµ4; Double et al. 1997) and Geospiza fortis (Gf5B and Gf6; Petren 1998). Starting with the original PCR conditions, variable MgCl2 and template DNA concentrations and annealing temperatures were tested in a gradient PCR device (PC-960G Gradient Thermal Cycler, Labortechnic). Optimal reaction conditions and PCR product size ranges are summarized in the Appendix. Genotypes were scored on a 6% acrylamide gel in an automatic sequencer (ALF express, Pharmacia Biotech).

Analyses Partitioning of genetic variation and estimation of effective population size, Ne. Genetic variation was explored at both fragment and isolate levels. Allelic richness (and a test for significant differences between groups, using 1000 permutations) was calculated with fstat (Goudet 1995), whereas allele frequencies, observed and expected heterozygosity, mean number of alleles per locus and genetic differentiation among the isolates were calculated using genepop Version 3.1d (Raymond & Rousset 1995a). For each pair of isolates, we estimated pair-wise FST using parameter θ (Weir & Cockerham 1984), and used a Markov Chain method to calculate significance levels using 10 000 permutations in an exact test for population differentiation (Guo & Thompson 1992; Raymond & Rousset 1995b). Approximate confidence limits (CI) for F ST were obtained as 2.5 and

1412 P . G A L B U S E R A E T A L . 97.5% quartiles after bootstrap sampling over all loci for 1000 times (GDA, Lewis & Zaykin 1999). Isolation-by-distance was explored by comparing the genetic (FST-based) and geographical distances using a Mantel’s test, and the genetic structure illustrated by principal component analysis (PCA) in pcagen (http://www.unil.ch/izea/softwares/pcagen.html). Although the commonly accepted microsatellite mutation rate for a wide range of animal taxa is 5 × 10 −4 (Ellegren et al. 1995), there is some evidence that birds might have higher mutation rates. For instance, based on 1000 meiotic events in three Hirundo rustica microsatellite loci, Saino et al. (1997) calculated mutation rates of 1 × 10 −3, 5 × 10 −3 and 36 × 10 −3, respectively. Thus, we chose the intermediate rate (µ = 0.001) to estimate effective population size, Ne, in migrate (Beerli & Felsenstein 1999). Finally, future levels of genetic diversity were predicted in geneloss (England & Osler 2001), based on current allelic frequencies and Ne-values. Estimation of gene flow. Based on the confirmed three-isolate structure (see Results), the numbers of current migrants and/or descendants from migrants were then estimated using assignment tests in structure version 1.0 (Waser & Strobeck 1998; Davies et al. 1999; Pritchard et al. 2000). We chose a burn-in period of 30 000 iterations and to collect data for 106 iterations (admixture model). Based on Bayesian statistics and the Markov chain Monte Carlo (MCMC) method, individuals are assigned to population clusters in such a way that each population cluster is in Hardy–Weinberg and linkage equilibrium. The number of ‘misassigned’ individuals (i.e. those that are assigned to a population other than the one in which they were captured) reflects the relative (probably not in absolute terms, see Introduction) dispersal rate between each pair of isolates (Haig et al. 1997; Waser et al. 2002). Variation in dispersal rates between sexes was tested in fstat (Goudet et al. 2002), using the assignment indices (AI) pooled for males and females (e.g. see Favre et al. 1997; Dallimer et al. 2002). Given that the average AI value for a population was 0, individuals with negative AI values are less likely than average to belong in that population and vice versa for positive AI values (Dallimer et al. 2002). We then estimated the levels of gene flow using the coalescence theory in migrate (Beerli & Felsenstein 1999). We ran 10 short chains (10 000 trees) and 3 long chains (100 000 trees) under the infinite allele model (IAM). This analysis was carried out at the level of the isolates (3) because the time required to obtain results with large numbers of samples (11 fragments) was too long (113, instead of 15, parameters have to be estimated). Unidirectional gene flow estimates were calculated by running this program 10 times (the estimates from each run were used as starting values for the next run). Unlike traditional FST-methods, coalescent methods compute directional gene

flow taking differences in subpopulation sizes into consideration (Luikart & England 1999; Beerli & Felsenstein 2001; Neigel 2002). Equilibrium testing. We tested for migration–drift equilibrium with the program 2mod (http://www.rubic.rdg.ac.uk/ ∼mab/software.html; Beaumont 2001). This program identifies populations subjected to genetic drift and gene flow as opposed to those under genetic drift only, through estimation of the relative likelihood of the two models using a MCMC procedure (Ciofi et al. 1999; Goodman et al. 2001; Palo et al. 2001). We ran the program for 105 iterations with a burn-in of 104. Next, the extent of the interaction between drift and gene flow was assessed using parameters F (the probability that two genes shared a common ancestor within a population) and M (the number of migrants per generation), in the gene flow model where M = (1 − F)/(4F) (Ciofi et al. 1999; Beaumont personal communication). Posterior distributions and the 90% highest probability density (HPD) limits were estimated from the simulated points by local density estimation using the program locfit (Loader 1996). Finally, we used bottleneck (Cornuet & Luikart 1996) to test whether the level of heterozygosity derived from the observed allele frequencies per forest fragment and isolate differed from the heterozygosity expected under mutation–drift equilibrium. As the mutation model underlying the microsatellite markers was unknown, we analysed the data under three different model assumptions: the IAM, the two-phase model (TPM) and the stepwise mutation model (SMM) (Jarne & Lagoda 1996).

Results Hardy–Weinberg and linkage equilibrium No linkage disequilibrium was detected between any pair of loci (all P > 0.05; Raymond & Rousset 1995a). All locus–forest fragment combinations were in Hardy– Weinberg equilibrium, except for the following combinations (significant after Bonferroni correction, Rice 1989): LTMR6 in Ronge, WB9 in Macha and Vuria, GF5B in Mbololo, Fururu and Vuria, and Pat14 in Yale (all heterozygosity deficits). No substructuring could be detected within each of the 11 forest fragments, excluding spatial Wahlund effect as an explanation of heterozygosity deficit. Likewise, a temporal Wahlund effect, i.e. subunits of individuals reproducing at different times (Morand et al. 2002), seems unlikely as FIS values (not shown) were much higher than FST values. However, subunits in Dabida bred at slightly different times over the 6-month breeding period (October to March; Githiru, unpublished). Such heterozygosity deficit might also be caused by inbreeding in some forest fragments. If so, a deficit would be expected for all loci, which was not the case in our study. Alternatively, the use of cross-specific © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

F R A G M E N T A T I O N A N D P O P U L A T I O N E Q U I L I B R I U M 1413 markers might have caused the lack of amplification of some alleles (null alleles). Hence, we attempted to detect the presence of null alleles by analysing pedigrees. We analysed five nests with two offspring each (in total, 10 ‘tests’) for the seven microsatellite markers. Many (Mendelian) inconsistencies were observed but most of these could be attributed to extra-pair matings (both ‘parents’ and offspring are heterozygous and one allele of the offspring is not found in these ‘parents’). However, two cases of null alleles were apparent, one for WB9 and another for GF5B. Mother and one offspring were homozygous for different alleles, suggesting that a null allele is present. Both cases were detected in the same individuals from a nest sampled in Chawia; giving a (crude) null allele frequency estimate of 10% (1/10) for WB9 and GF5B. We also calculated the frequency of null alleles according to the method of Brookfield (1996): r = (HE − HO)/ 1 + HE. Across fragments, these estimated frequencies ranged between −0.12 and 0.04, −0.04 and 0.25, −0.07 and 0.05, −0.07 and 0.12, −0.04 and 0.07, 0.02 and 0.17, −0.06 and 0.13 for locus Mcyµ4, LTMR6, GF6, WB9, WB2, GF5B and Pat14, respectively. It is noticeable that for Chawia the highest frequencies are estimated for WB9 and GF5B (11 and 9%, respectively). This seems in agreement with the pedigree data. In conclusion, the heterozygote-deficits observed are likely partially due to the presence of null alleles and we should be cautious about the results when including the loci where deficits were observed for various fragments (GF5B). Hence, we compared some of these results with the results obtained when taking into account the null alleles or when excluding locus GF5B. No noticeable differences were observed, however, when: (i) whichrun was used to assign individuals (with or without null alleles), (ii) 2mod was run (with or without null alleles), and (iii) PCA was performed (with or without GF5B). When locus WB9 was also dropped from the bottleneck analysis, however, the heterozygosity deficiency in Vuria was no longer significant.

Population structure In order to estimate the levels of gene flow, it is essential to establish that different populations are indeed genetically discrete. To check this, we estimated the number of subpopulations (K) in our total sample assuming no substructuring a priori (i.e. without effect of sampling locality; option ‘USE POPINFO = 0’) using structure (Pritchard et al. 2000), and found a population substructuring (K = 4) to be the most probable (Table 1). However, these values for K are intended as a guide, and should not override views based on inspection of the results (Pritchard, personal communication). As one cluster contained only one fragment (Yale) and the other three clusters correspond largely to the three isolates (DA, MB & SA) we propose a structure based © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

Table 1 Inference of K, the number of subpopulations, for the whitestarred robin using structure (106 interations, burn-in 30 000) K

Ln Pr(X|K)

Pr(K|X)

1 2 3 4 5 6 7

−7563 −7489 −7446 −7420 −7444 −7468 −7495

0 0 0 0.999 0 0 0

on these three isolates. This structure was also evident when correlation in allele frequencies was allowed (option ‘FREQSCORR = 1’). Also, no significant further structuring (K = 1) was found within the DA isolate when this isolate was analysed separately (as suggested in the documentation to the software). However, there was significant pairwise genetic differentiation (FST) between Ngangao and three other fragments: Fururu, Macha and Ndiwenyi. Overall, F-statistics revealed significant differentiation both at fragment and isolate levels (FST = 0.0179, P < 0.001, 95% CI: 0.0114–0.0257 and FST = 0.0343, P < 0.001, 95% CI: 0.0184–0.0495, respectively). Note that at the level of isolates, Mbololo Hill clusters fragments Mbololo and Ronge (pairwise FST for Mbololo vs. Ronge = 0.005; not significantly different from 0). Pairwise analyses for the isolates showed relatively high θ-values all which were significant at the 0.01-level after Bonferroni correction (Rice 1989): DA-MB 0.034, DA-SA 0.025 and MB-SA 0.051. This significant differentiation of the isolates was illustrated by the PCA (Fig. 2a) where Mbololo and Ronge were separated on the first axis (37% inertia; P = 0.010) and Sagala by the second axis (18% inertia; P = 0.080). The third axis separating the DA fragments was not statistically significant (Fig. 2b: 10% inertia; P = 1.00), however, it showed that Ngangao was relatively differentiated from the rest. Lastly, there was a significant correlation between the geographical distance and FST-values (Mantel test; P = 0.042) for the entire study area, but no such correlation was evident among the forest fragments within DA by themselves.

Current and predicted genetic diversity and effective population size The mean number of alleles per locus (MNA) averaged 7.8 ± 0.29 per forest fragment (Table 2). Mwachora, with the smallest sample size, also had the lowest MNAvalue, followed by Sagala, Mbololo and Ronge. Excluding Mwachora, Ngangao and Macha had lowest values within Dabida. The expected heterozygosity HE ranged from 0.64 (Sagala) to 0.71 (Chawia). Again Macha and Ngangao had intermediate values, whereas Sagala, Mbololo and Ronge

1414 P . G A L B U S E R A E T A L . Table 2 Estimates of allelic diversity and heterozygosity both per forest fragment and per isolate. N = number of individuals genotyped, MNA = mean number of alleles per locus, HO = observed heterozygosity, HE = expected heterozygosity. Allelic richness (AR) of fragments and subpopulations is based on 11 and 31 individuals, respectively; different letters per column (AR or HE) indicate significant differences, one-sided P < 0.05 (1000 permutations in fstat). Italic numbers in brackets are predicted estimates (geneloss results) in five generations per subpopulation Isolate/fragment

N

MNA

AR

HO

HE

Dabida Chawia (CH) Ngangao (NG) Fururu (FU) Macha (MA) Ndiwenyi (ND) Vuria (VU) Yale (YA) Mwachora (MW) Mbololo Hill Mbololo (MB) Ronge (RO) Sagala (SA)

228 30 30 32 30 33 32 30 11 61 31 30 31

11.6 (10.5) 8.7 7.4 8.4 7.9 8.9 9.1 8.3 6.1 8.3 (6.9) 7.3 7.3 6.7 (5.3)

6.3a 6.5 5.7 6.3 6.0 6.5 6.5 6.3 6.1 5.6b 5.6 5.6 5.3b

0.63 0.65 0.60 0.62 0.60 0.66 0.63 0.62 0.61 0.58 0.57 0.58 0.60

0.70a (0.69) 0.71 0.67 0.68 0.66 0.70 0.69 0.69 0.65 0.65b (0.63) 0.64 0.65 0.64b (0.60)

Estimating dispersal Fig. 2 Principal component analysis (PCA) based on genetic data (seven microsatellite markers) of white-starred robins showing separation (differentiation) of the different fragments (Ro, Ronge; Mb, Mbololo; Sa, Sagalla; Ng, Ngangao; Ch, Chawia; Nd, Ndiwenyi; Fu, Fururu; Vu, Vuria; Mw, Mwachora; Ma, Macha; Ya, Yale; Ng, Ngangao). Respectively, 37, 18 and 10% of the variation between fragments is explained by the first, second and third principal component. Fragments were clearly separated according to the two first axes (Fig. 2a) but less so by axis one and axis three (Fig. 2b).

had lowest values. Both HE and allelic richness were significantly lower (P < 0.01) in the least diverse populations (Sagala, Mbololo and Ronge) compared with the most diverse ones (Chawia, Fururu, Ndiwenyi, Yale and Vuria). When we grouped forest fragments into their isolates (i.e. by subpopulation) and reanalysed the genetic variability, DA retained the highest HE and MNA (Table 2). The difference in genetic diversity (and allelic richness) between DA and the other isolates was significant, but MB and SA did not differ (Table 2). Estimates of Ne (and its 95% CI) were 408 (392–412) individuals for DA, 60 (56 – 64) for MB and 39 (36 –43) for SA (migrate results). Simulations of future genetic diversity (MNA and HE) for each isolate indicated that, proportionally, MNA dropped faster than HE, and increasingly so in the smallest subpopulation (SA) (Table 2).

‘Direct’ gene flow: we probabilistically assigned individuals (in structure; option ‘USE POPINFO = 1’) to their subpopulation of origin based on the three-subpopulation structure (corresponding to the isolates: Fig. 1). Six individuals had a relatively high probability (P > 0.5) to be migrants or offspring of migrants: one individual (#12) might be an offspring of a migrant from SA to DA; four individuals (#c43, #318, #27 and #167) might have migrated from MB to DA, and one individual (#53) from DA to MB. However, a lack of sufficient power in our analysis due to the relatively small number of markers used (see Pritchard et al. 2000) meant that only one of these individuals (#167) had such a low probability (P < 0.05) of being assigned to its subpopulation of origin that it could be confidently classified as a migrant (from MB to Ndwenyi; Fig. 3). The mean assignment index (AI) was actually lower than average for females (−0.019) and higher than average for males (0.008), suggesting female-biased dispersal overall. This difference approached significance (P = 0.06; 1000 randomizations). Finally, direct capture–recapture data yielded nine dispersers: a juvenile female, three adult females and five adult males, with all (except one adult male that moved from DA to SA) dispersing within DA. ‘Indirect’ gene flow. The unidirectional estimates of number of migrants (Nm) between subpopulations are given in Table 3. The highest mean dispersal rates were found © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

F R A G M E N T A T I O N A N D P O P U L A T I O N E Q U I L I B R I U M 1415 Table 3 Comparison of ‘historical’ vs. ‘current’ gene flow. Upper value, unidirectional estimates of gene flow (Nm, number of dispersers per generation) between the three white-starred robin subpopulations calculated using migrate (between brackets: 95% CI from profile likelihoods). Lower value, in italics, the number of individuals that had a posterior probability > 0.5 of coming from another isolate, as estimated by structure (between brackets: the number of misassigned individuals with a probability < 0.95 to belong to its isolate of capture, see text)

Subpopulation

To Sagala

From Sagala



Dabida Mbololo Fig. 3 Summary of the assignment test assuming three subpopulations. Each point shows the mean estimated ancestry, based on individual genotypes and the estimated allele frequencies per subpopulation, for an individual in the samples. Position on the axes refers to the genetic affiliation of each individual, symbols refer to their site of first capture: Mbololo Hill (open circles), Sagala (open squares), Dabida (filled diamonds). Numbered individuals are suspected as migrants or of migrant ancestry (only individual 167 had such a low probability (P < 0.05) to be assigned to its subpopulation of origin that it could be strictly labelled as a disperser; see text for details).

1.9 (1.7–2.1) 0 (0)* 0.7 (0.6–0.8) 0 (0)

Dabida

Mbololo

3.1 (2.8–3.4) 1 (0) —

0.4 (0.4 – 0.5) 0 (0) 6.3 (6.0 – 6.6) 1(0)

4.8 (4.4–5.1) 4 (1)

*Only one individual has been observed (CMR-method) migrating between isolates (from Dabida to Sagala).

between DA and MB (5.6), then between DA and SA (2.5), and lowest rates between MB and SA (0.6).

Equilibrium testing Migration–drift equilibrium: the gene flow–drift equilibrium model was 1000 times more likely than the pure drift model, implying that gene frequencies were determined by a balance between genetic drift and immigration. The probability (F) that two alleles were identical by descent was at least twice as high in Sagala, Ronge and Mbololo (mode-F was 0.069, 0.059 and 0.057, respectively) compared with the Dabida forest fragments (with F ranging from 0.001 in Vuria to 0.023 in Ngangao). Density curves for F and 90% HPD limits for F and M (2mod) are given in Fig. 4 and Table 4, respectively. This denotes a larger immigrant-effect relative to drift in DA compared with the other isolates, as demonstrated by greater estimates of number of migrants per generation in this isolate (mode-M = 11–239 in DA, and M = 4, 4 and 3 for RO, MB and SA). These values agree well with the migrate estimates for Sagala (2.6 immigrants) and Mbololo Hill (6.7 immigrants), but not for Dabida (7.9 immigrants). Almost all the DA immigrants suggested by 2mod probably originate from within Dabida. © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

Fig. 4 Density of F-values conditional of the gene flow/drift model (2mod) per fragment for the White-starred Robin (see Table 2 for abbreviations).

Mutation–drift equilibrium. We did not detect any recent (or severe) bottleneck: under the IAM, all allele frequency distributions were L-shaped, while neither the sign test nor the Wilcoxon test detected more heterozygosity than expected under mutation–drift equilibrium. Assuming the SMM and TPM (95% SMM) there were no significant deviations from expectations in all fragments, except Vuria, which showed a significant deficiency of heterozygosity (sign test: P = 0.002; one-tailed for H-deficiency Wilcoxon test P = 0.0039). This effect was not apparent when loci GF5B and WB9 were dropped from the analysis, however.

1416 P . G A L B U S E R A E T A L .

Location

log(ratio)

Mode F

LL-F

UL-F

Mode M

LL-M

UL-M

VU ND FU MW CH YA MA NG MB RO SA

1.6190 1.5521 1.3876 1.4932 1.5503 1.5687 1.5026 1.3720 1.3478 1.4133 1. 1995

0.0010 0.0012 0.0024 0.0047 0.0065 0.0097 0.0154 0.0226 0.0566 0.0594 0.0689

0.0001 0.0000 0.0001 0.0001 0.0003 0.0018 0.0079 0.0126 0.0347 0.0430 0.0447

0.0062 0.0076 0.0116 0.0239 0.0145 0.0182 0.0304 0.0424 0.0873 0.0948 0.1001

239 205 103 53 38 26 16 11 4 4 3

40 33 21 10 17 13 8 6 3 2 2

4 902 50 000 2 907 3 378 746 137 32 20 7 6 5

Discussion Genetic diversity and differentiation We found significant differences in genetic variability within each subpopulation (isolate), plus significant genetic differentiation between subpopulations. The large number of patches within DA potentially generated the higher genetic variability and effective population size in this isolate compared with MB and SA. Indeed, the probability that two alleles were identical by descent was twice as high in MB and SA compared with DA, indicating that most exchange of individuals occurred within DA. This is also supported by capture–recapture data, with seven of eight known dispersers being within DA. Thus, although DA and MB were comparable in total size (~200 ha), DA had greater genetic diversity. Whereas the history of the area might have had some additional effect (if DA originally consisted of a larger area than the other isolates), this finding advocates for several (even if some are small) fragments for conservation purposes, as long as ample exchange of individuals occurs between them (Burkey 1989; Nunney & Campbell 1993). Both Mantel’s test and the PCA demonstrated an isolationby-distance effect, in which the genetic structure was influenced by isolate juxtaposition and the topographical features separating them. In a recent study, naturally occurring barriers have also been shown to affect genetic structure of an avian species (black grouse) with similarly high dispersal capacities (Caizergues et al. 2003). In addition, our data suggested likely additional substructuring within DA with Ngangao on one side and the other fragments on the other. Indeed, Ngangao was significantly differentiated from three of the small fragments, and was distinctly separated on the third PCA-axis, although not significantly so because of the effects of other fragments lying in the middle. The location of a town in between Ngangao and the other fragments (see map in Brooks et al. 1998) might have presented a barrier to (direct) movement.

Table 4 Mode and 90% HPD limits for F (the probability that two genes shared a common ancestor within a population) and M (the number of migrants per generation): log(ratio) = log(density mode/density HPD limits); LL-F = 90% HPD lower limit F; UL-F = 90% HPD upper limit F; LL-M = 90% HPD lower limit M; UL-M = 90% HPD upper limit M

A similar explanation was invoked to explain the pattern of seed dispersal by the silvery-cheeked hornbill, and the resultant distribution of the Maesopsis eminii tree in these forest fragments (Githiru 2000).

Gene flow and migration–drift equilibrium Based on previous capture–recapture data for this species (Lens et al. 1999a, 2002b), we expected rates of dispersal between subpopulations to be sufficiently high to counter the effect of random genetic drift. Based on our genetic data and the Bayesian method, we found strong indications that the robin population was indeed in migration–drift equilibrium, suggesting limited impact of fragmentation on dispersal. A comparison of estimates on current and historical gene flow largely confirmed this finding. The current and historical numbers of migrants between subpopulations were estimated to be 0–1 and 0.6–5.6, respectively. It is likely that we lacked sufficient power to definitely label all individuals (except one) as migrants or having migrant ancestry due to the small number of markers (Pritchard et al. 2000). Hence, we probably underestimated the current dispersal (only one individual had a probability < 0.05 of belonging to its population of origin). Furthermore, the (absolute) estimates of the assignment method (and also the capture–recapture method) are dependent on the proportion of the total population sampled (additional samples could yield additional dispersers). Using capture–recapture we found, in addition to seven dispersers within Dabida, one individual to be dispersing from Dabida to Sagala but no dispersers between Dabida and Mbololo, whereas estimates of gene flow based on the assignment test were the highest between the latter subpopulations. These ‘omissions’ using capture–recapture methods were probably related to sampling after dispersal, a common feature in species with juvenile dispersal (e.g. Van Treuren et al. 1999). Moreover, methods based on direct observations are often limited in geographical scope, frequently missing many such long-distance dispersal events (Barrowclough © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

F R A G M E N T A T I O N A N D P O P U L A T I O N E Q U I L I B R I U M 1417 1978; Crochet 1996; Gauthreaux 1996). Low dispersal between SA and both DA and MB could be attributed to the presence of the Voi River and extensive plains between them, in contrast to only a valley/ridge between MB and DA. Extra effects of fragment juxtaposition on dispersal will become clear once a digitized GIS map (under preparation) is available as various features can then be associated with enhancing or restricting movements. So, the discrepancies between current and historical gene flow (0 –1 vs. 0.6 – 5.6) can probably be explained by an underestimation of current gene flow. Still, there was some congruence between relative estimates of current and historical gene flow (number of migrants); with both methods, the highest exchange was observed between DA and MB. We also found a genetic structure that could be explained by the distance between all forest fragments (isolation-by-distance; see Slatkin 1993). This all suggests that the inference of migration–drift equilibrium, through the Bayesian method, was made correctly because rates and patterns in gene flow have not been drastically altered, despite severe fragmentation. However, we can not exclude the possibility that there is a slight decrease in gene flow (slightly lower current vs. historical estimates) which remained undetected using the Bayesian method, because this method can only contrast ‘dispersal’ vs. ‘no dispersal at all’. Furthermore, the estimated small population sizes and moderate gene flow rates do not exclude that a ‘new’ (after fragmentation) equilibrium might have been reached. Using the formula (time for FST to reach halfway from an old value to a new equilibrium = ln(1/2)/ln[(1 − m)2(1 – 1/2N)]) of Whitlock (1992), the expected number of generations to go half way to equilibrium is ~ 4 –14. This estimate encompasses the time since fragmentation (~ 12 generations). So we cannot exclude the possibility that the migration rates underlying a ‘previous’ (before fragmentation) equilibrium were different (higher), but recall that equilibrium is assumed to estimate m and Ne (implying circular reasoning).

Mutation-drift equilibrium The subpopulations seemed to be in mutation–drift equilibrium; no recent bottlenecks were detected. This was not surprising because strong bottlenecks are only expected in subpopulations that have been drastically reduced to effective population sizes much lower than the least value we obtained (39 in SA) (Pimm et al. 1989; Gilligan et al. 1997; Spencer et al. 2000). Furthermore, immigrants, even a few, can restore genetic diversity in as quickly as three generations (Keller et al. 2001). However, in Vuria, we found less heterozygosity than expected under the SSM at mutation– drift equilibrium. This might indicate a recent expansion of the population (if closed). Given the history of the fragment, this is highly unlikely. The deficiency might be © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

caused by null alleles because excluding locus GF5B or WB9 (known to deviate from Hardy–Weinberg equilibrium in Vuria) from the analysis did alter the results. Dropping loci from the analysis also reduces the power, however. An alternative explanation is a recent influx of rare alleles from genetically distinct populations (Cornuet & Luikart 1996). This explanation is supported by the 2mod results (Vuria has the highest immigration rates). It also implies that Vuria used to be genetically distinct from the other fragments in Dabida, which is not unprobable, given its relatively isolated location. It is not clear whether the local individuals are reproductively successful or whether Vuria should be considered as a ‘sink’ fragment. It might signify a ‘sink’, plausible if most individuals in the original population were extirpated when it underwent a 99% reduction, and the current population consists of recent immigrants with different alleles and allele frequencies. These immigrants could quickly increase the number of rare alleles without substantially affecting the heterozygosity, mimicking an increase (or hiding a decrease) of the population size (Cornuet & Luikart 1996).

Conclusions Overall, the robin metapopulation revealed high genetic diversity (i.e. allelic richness and heterozygosity) and was evidently at both migration- and mutation–drift equilibrium. This suggests that habitat loss and fragmentation have not significantly disrupted dispersal or drastically reduced effective population size (numbers of breeders). However, it remains difficult to be completely sure that the population has not been affected during or immediately following this fragmentation, but we are confident that the dispersal estimates reflect ongoing gene flow. The simulations on future diversity, especially for Sagala, warn of fairly swift reductions in genetic diversity through drift as effective population sizes decline and landscape permeability is lost through increasing habitat loss, patch isolation and intensification of land use in the surrounding matrix. Although this is a ‘worst-case’ scenario, because it is based on the assumption of closed populations, female-biased dispersal if associated with high mortality, will severely decrease the number of females making such a scenario realistic. Currently, however, it seems that any loss of alleles though drift (and deleterious mutations) is successfully countered by dispersal. Sustained dispersal is thus key to defusing the impact(s) of habitat fragmentation on genetic diversity, and consequently forestalling any negative impact on population structure (see also Andrianarimisa et al. 2000). For practical purposes, that dispersal is such an important factor means that focusing conservation only on the habitat patches themselves is insufficient to maintain all subpopulations (Opdam & Wiens 2002). Finally, higher dispersal capacity yielded higher genetic diversity for

1418 P . G A L B U S E R A E T A L . the robin compared with the Taita thrush. This boosted the effective population sizes and helped avert bottlenecks such as that found for the thrush subpopulation in Chawia (Galbusera et al. 2000a). This implies that inability to traverse this landscape matrix and successfully establish themselves in the new populations is at least one of the difficulties facing the thrush populations in these fragments. The impact of habitat fragmentation on population viability has been shown to vary with interspecific differences in life history straits such as levels of mobility, even for relatively mobile species such as butterflies and birds (Desrochers et al. 1999; Baillie et al. 2000; Thomas 2000). Also in the Taita Hills, 45% of the variation in patch occupancy among eight rainforest bird species was explained by varying dispersal rates (Lens et al. 2002b). Our results demonstrate the usefulness of collecting population genetic data for sympatric species with varying dispersal capacities, in order to gain a more complete understanding and make better predictions of how habitat fragmentation impacts on bird species.

Acknowledgements We thank K. Otter, H. Bickle and T. Burke for sharing unpublished sequences of the primer set Pat14. Tine Schenck and Conny Wouters performed most of the microsatellite genotyping. J. Barnes, R. Barnes, T. Brooks, D. Gitau, T. Imboma, C. Jackson, J. Kageche, P. Kariuki, J. Lindsell and C. Wilder provided assistance in the field, and B. Bytebier provided logistic support. Subject editor Dallas and two anonymous reviewers greatly improved the manuscript. Fieldwork was funded by a postdoctoral fellowship (to LL) and research grant G0258.01 (to EM) of the Fund for Scientific Research — Flanders, by a Rhodes International scholarship (to MG), and by Flemish Interuniversity Council project 02/6/7-338-607 (to W.N. Verheyen and EM).

References Andrianarimisa A, Bachmann L, Ganzhorn JU, Goodman SM, Tomiuk J (2000) Effects of forest fragmentation on genetic variation in endemic understory forest birds in central Madagascar. Journal für Ornithologie, 141, 152 –159. Baillie SR, Sutherland WJ, Freeman SN, Gregory RD, Paradis E (2000) Consequences of large-scale processes for the conservation of bird populations. Journal of Applied Ecology, 37, 88–102. Barrowclough GF (1978) Sampling bias in dispersal studies based on finite area. Bird-Banding, 49, 333 – 341. Beaumont MA (2001) Conservation genetics. In Handbook of Statistical Genetics (eds Balding DJ, Cannings C), pp. 779 – 812. Wiley, Chichester. Beentje HJ (1987) An ecological and floristic study of the forests of the Taita Hill, Kenya. Utafiti, 1, 23 – 66. Beerli P, Felsenstein J (1999) Maximum-likelihood estimation of migration rates and population numbers of two populations using a coalescent approach. Genetics, 152, 763 –773. Beerli P, Felsenstein J (2001) Maximum likelihood estimation of a migration matrix and effective population sizes in n subpopulations by using a coalescent approach. Proceedings of the National Academy of Sciences of the USA, 98, 4563 – 4568.

Bennun L, Dranzoa C, Pomeroy D (1996) The forest birds of Kenya and Uganda. Journal of East African Natural History, 85, 23–48. Bohonak AJ (1999) Dispersal, gene flow, and population structure. Quarterly Review of Biology, 74, 21–45. Bossart JL, Prowell DP (1998) Genetic estimates of population structure and gene flow: limitations, lessons and new directions. Trends in Ecology and Evolution, 13, 202–206. Brookfield JFY (1996) A simple new method for estimating null allele frequency from heterozygote deficiency. Molecular Ecology, 5, 453–455. Brooks T, Lens L, Barnes J, Barnes R, Kihuria JK, Wilder C (1998) The conservation status of the forest birds of the Taita Hills, Kenya. Bird Conservation International, 8, 119–139. Burkey TV (1989) Extinction in nature reserves: the effect of fragmentation and the importance of migration between reserve fragments. Oikos, 55, 75–81. Caizergues A, Rätti O, Hella P, Rotelli L, Ellison L, Rasplus J-Y (2003) Population genetic structure of male black grouse (Tetrao tetrix L.) in fragmented vs. continuous landscapes. Molecular Ecology, 12, 2297–2305. Ciofi C, Beaumont MA, Swingland IR, Bruford MW (1999) Genetic divergence and units for conservation in the Komodo dragon Varanus komodoensis. Proceedings of the Royal Society of London B, 266, 2269–2274. Cornuet JM, Luikart G (1996) Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics, 144, 2001–2014. Cornuet JM, Piry S, Luikart G, Estoup A, Solignac M (1999) New methods employing multilocus genotypes to select or exclude populations as origins of individuals. Genetics, 153, 1989 – 2000. Crochet P-A (1996) Can measures of gene flow help to evaluate bird dispersal. Acta Oecologica, 17, 459–474. Dallimer M, Blackburn C, Jones PJ, Pemberton JM (2002) Genetic evidence for male biased dispersal in the red-billed quelea Quelea quelea. Molecular Ecology, 11, 529–533. Davies N, Villablanca FX, Roderick GK (1999) Determining the source of individuals: multilocus genotyping in nonequilibrium population genetics. Trends in Ecology and Evolution, 14, 17–21. Desrochers A, Hannon SJ, Bélisle M, St. Clair CC (1999) Movement of songbirds in fragmented forests: can we ‘scale-up’ from behaviour to explain occupancy patterns in the landscape? In Proceedings Of the 22nd International Ornithological Congress Johannesberg (eds Adams NJ, Slotow RH), pp. 2447–2464. BirdLife South Africa, Durban. Double MC, Dawson D, Burke T, Cockburn A (1997) Finding the fathers in the least faithful bird: a microsatellite-based genotyping system for the superb fairy-wren Malurus cyaneus. Molecular Ecology, 6, 691–693. Dowsett RJ (1985) Site fidelity and survival rates of some montane forest birds in Malawi, south-central Africa. Biotropica, 17, 145 – 154. Dowsett RJ, Dowsett-Lemaire F (1986) Homing ability and territorial replacement in some forest birds in south-central Africa. Ostrich, 57, 25–31. Ellegren H, Lifjeld JT, Slagsvold T, Primmer CR (1995) Handicapped males and extrapair paternity in pied flycatchers: a study using microsatellite markers. Molecular Ecology, 4, 739 – 744. England PR, Osler GHR (2001) geneloss: a computer program for simulating the effects of population bottlenecks on genetic diversity. Molecular Ecology Notes, 1, 111–113. © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

F R A G M E N T A T I O N A N D P O P U L A T I O N E Q U I L I B R I U M 1419 Favre L, Balloux F, Goudet J, Perrin N (1997) Female-biased dispersal in the monogamous mammal Crocidura russula: evidence from field data and microsatellite patterns. Proceedings of the Royal Society of London B, 264, 127–132. Frankham R (1998) Inbreeding and extinction: island populations. Conservation Biology, 12, 665 – 675. Galbusera P, Lens L, Waiyaki E, Schenck T, Matthysen E (2000a) Genetic variability and gene flow in the globally, critically endangered Taita thrush. Conservation Genetics, 1, 45 – 55. Galbusera P, Van Dongen S, Matthysen E (2000b) Cross-species amplification of microsatellite primers in passerine birds. Conservation Genetics, 2, 163 –168. Gauthreaux SA (1996) Bird migration: methodologies and major research trajectories (1945 –95). Condor, 98, 442– 453. Gilligan DM, Woodworth LM, Montgomery ME, Briscoe DA, Frankham R (1997) Is mutation accumulation a threat to the survival of endangered populations. Conservation Biology, 11, 1235–1241. Githiru M (2000) Avian frugivory and seed dispersal in some of the Taita Hills forest fragments. MSc Thesis, Kenyatta University. Githiru M (2002) Endemic forest birds of the Taita Hills: using a model species to understand the effect of forest fragmentation on small populations. PhD Thesis, University of Oxford. Goodman SJ, Tamate HB, Wilson R et al. (2001) Bottlenecks, drift and differentiation: the population structure and demographic history of sika deer (Cervus nippon) in the Japanese archipelago. Molecular Ecology, 10, 1357–1370. Goudet J (1995) fstat (vers. 1.2): a computer program to calculate F-statistics. Journal of Heredity, 86, 485 – 486. Goudet J, Perrin N, Waser P (2002) Tests for sex-biased dispersal using bi-parentally inherited genetic markers. Molecular Ecology, 11, 1103–1114. Greenwood PJ, Harvey PH (1982) The natal and breeding dispersal of birds. Annual Review of Ecology and Systematics, 13, 1–21. Griffiths R, Double MCY, Orr K, Dawson RJG (1998) A DNA test to sex most birds. Molecular Ecology, 7, 1071–1075. Guo SW, Thompson EA (1992) Performing the exact test of Hardy–Weinberg proportion for multiple alleles. Biometrics, 48, 361–372. Haig SM, Gratto-Trevor CL, Mullins TD, Colwell MA (1997) Population identification of western hemisphere shorebirds throughout the annual cycle. Molecular Ecology, 6, 412– 427. Hedrick PW (1999) Perspective: highly variable loci and their interpretation on evolution and conservation. Evolution, 53, 313–318. Jarne P, Lagoda PJL (1996) Microsatellites, from molecules to populations and back. Trends in Ecology and Evolution, 11, 424– 429. Keith S, Urban EK, Fry CH (1992) The Birds of Africa. Academic Press, London. Keller LF, Jeffrey KJ, Arcese P et al. (2001) Immigration and the ephemerality of a natural population bottleneck: evidence frrom molecular markers. Proceedings of the Royal Society of London B, Biological Science, 268, 1387–1394. Koenig WD, Van Vuren D, Hooge PN (1996) Detectability, philopatry, and the distribution of dispersal distances in vertebrates. Trends in of Ecology and Evolution, 11, 514 – 517. Lens L, Adriaensen F, Matthysen E (1999a) Dispersal studies in recently and historically fragmented forests: A comparison between Kenya and Belgium. In Proceedings of the 22nd International Ornithological Congess, Durban (eds Adams N, Slotow R), pp. 2480–2491. University of Natal, Durban. © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

Lens L, Galbusera P, Brooks T, Waiyaki E, Schenck T (1998) Highly skewed sex ratios in the critically endangered Taita thrush as revealed by CHD genes. Biodiversity and Conservation, 7, 869 – 873. Lens L, Van Dongen S (1999) Evidence for organism-wide asymmetry in five bird species of a fragmented afrotropical forest. Proceedings of the Royal Society of London B, 266, 1055–1060. Lens L, Van Dongen S, Matthysen E (2002a) Fluctuating asymmetry as an early warning system in the critically endangered Taita thrush. Conservation Biology, 16, 479–487. Lens L, Van Dongen S, Norris K, Githiru M, Matthysen E (2002b) Avian persistence in fragmented rainforest. Science, 298, 1236 – 1238. Lens L, Van Dongen S, Wilder CM, Brooks TM, Matthysen E (1999b) Fluctuating asymmetry increases with habitat disturbance in seven bird species of a fragmented afrotropical forest. Proceedings of the Royal Society of London B, 266, 1241–1246. Lewis PO, Zaykin D (1999) Genetic Data Analysis: Computer Program for the Analysis of Allelic Data, Version 1.0 (d12). Free program distributed by the authors available at: http://chee.unm.edu/ gda/. Loader CR (1996) Local likelihood density estimation. Annals of Statistics, 24, 1602–1618. Lovett J (1985) Moist forests of Eastern Tanzania. Swara, 8, 8 –9. Luikart G, England PR (1999) Statistical analysis of microsatellite DNA data. Trends in Ecology and Evolution, 14, 253–256. Matthysen E, Lens L, Van Dongen S et al. (1995) Diverse effects of forest fragmentation on a number of animal species. Belgian Journal of Zoology, 125, 175–183. McDonald DB, Potts W (1994) Cooperative display and relatedness among males in a lek mating bird. Science, 266, 1030 – 1032. McRae SB, Amos W (1999) Characterization of hypervariable microsatellites in the cooperatively breeding white-browed sparrow weaver Plocepasser mahali. Molecular Ecology, 8, 895 – 906. Monaghan MT, Spaak P, Robinson CT, Ward JV (2001) Genetic differentiation of Baetis alpinus Picted (Ephemeroptera: Baetidae) in fragmented alpine streams. Heredity, 86, 395–403. Morand M-E, Brachet S, Rossignol P, Dufour J, Frascaria-Lacoste N (2002) A generalized heterozygote deficiency assessed with microsatellites in French common ash populations. Molecular Ecology, 11, 377–386. Moreau RE (1951) Geographical variation and plumage sequence in Pogonocichla. Ibis, 93, 383–401. Myers N, Mittermeier RA, Mittermeier CG, da Fonseca AB, Kent J (2000) Biodiversity hotspots for conservation priorities. Nature, 403, 853–858. Neigel JE (2002) Is FST obsolete? Conservation Genetics, 3, 167–173. Newton I (1998) Population Limitation in Birds. Academic Press, San Diego, CA. Nunney L, Campbell KA (1993) Assessing minimum viable population size: demography meets population genetics. Trends in of Ecology and Evolution, 8, 234–239. O’Ryan C, Harley EH, Bruford MW, Beaumont M, Wayne RK, Cherry MI (1998) Microsatellite analysis of genetic diversity in fragmented South African buffalo populations. Animal Conservation, 1, 85–94. Oatley TB (1982) The starred robin in Natal, Part 2: annual cycles and feeding ecology. Ostrich, 53, 193–205. Opdam P, Wiens JA (2002) Fragmentation, habitat loss and landscape management. In Conserving Bird Biodiversity: General

1420 P . G A L B U S E R A E T A L . Principles and their Application (eds Norris K, Pain DJ), pp. 202– 223. Cambridge University Press, Cambridge. Otter K, Ratcliffe L, Michaud D, Boag PT (1998) Do female blackcapped chickadees prefer high-ranking males as extra-pair partners? Bahaviour Ecology Socioabiology, 43, 25 – 36. Palo JU, Makinen HS, Helle E, Stenman O, Vainola R (2001) Microsatellite variation in ringed seals (Phoca hispida): genetic structure and history of the Baltic Sea population. Heredity, 86, 609–617. Petren K (1998) Microsatellite primers from Geospiza fortis and cross-species amplification in Darwin’s finches. Molecular Ecology, 7, 1771–1788. Pimm SL, Gittleman JL, McCracken GF, Gilpin M (1989) Plausible alternatives to bottlenecks to explain reduced genetic diversity. Trends in of Ecology and Evolution, 4, 176 –178. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics, 155, 945–959. Raymond M, Rousset F (1995a) genepop (Version 1.2): population genetics software for exact tests and ecumenicism. Journal of Heredity, 86, 248–249. Raymond M, Rousset F (1995b) An exact test for population differentiation. Evolution, 49, 1280 –1283. Rice WR (1989) Analyzing tables of statistical tests. Evolution, 43, 223–225. Saccheri I, Kuussaari M, Kankare M, Vikman P, Fortelius W, Hanski I (1998) Inbreeding and extinction in a butterfly metapopulation. Nature, 392, 491– 494. Saino N, Primmer CR, Ellegren H, Moller A (1997) An experimental study of paternity and tail ornamentation in the barn swallow (Hirundo rustica). Evolution, 51, 562 – 570. Slatkin M (1987) Gene flow and the geographic structure of natural populations. Science, 236, 787–792. Slatkin M (1993) Isolation by distance in equilibrium and nonequilibrium populations. Evolution, 47, 264 –279. Spencer CC, Neigel JE, Leberg PL (2000) Experimental evaluation of the usefulness of microsatellite DNA for detecting demographic bottlenecks. Molecular Ecology, 9, 1517–1528. Thomas CD (2000) Dispersal and extinction in fragmented landscapes. Proceedings of the Royal Society of London B, 267, 139–145. Uimaniemi L, Orell M, Mönkkönen M, Huhta E, Jokimäki J, Lumme J (2000) Genetic diversity in the Siberian jay Perisoreus

infaustus in fragmented old-growth forests of Fennoscandia. Ecography, 23, 669–677. Van Treuren R, Bijlsma R, Tinbergen JM, Heg D, Van de Zande L (1999) Genetic analysis of the population structure of socially organized oystercatchers (Haematopus ostralegus) using microsatellites. Molecular Ecology, 8, 181–187. Walsh PS, Metzger DA, Higuchi R (1991) Chelex 100 as a medium for simple extraction of DNA for PCR-based typing from forensic material. Biotechniques, 10, 506–513. Wang J, Whitlock MC (2003) Estimating effective population size and migration rates from genetic samples over space and time. Genetics, 163, 429–446. Waser PM, Strobeck C (1998) Genetic signatures of interpopulation dispersal. Trends in Ecology and Evolution, 13, 43–44. Waser PM, Strobeck C, Paetkau D (2002) Estimating interpopulation dispersal rates. In Carnivore Conservation (eds Gittleman JL, Funk SM, Macdonald D, Wayne RK), Cambridge University Press, Cambridge. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution, 38, 1358–1370. Whitlock MC (1992) Temporal fluctuations in demographic parameters and the genetic variance among populations. Evolution, 46, 608–616. Whitlock MC, McCauley DE (1999) Indirect measures of gene flow and migration: FST not equal to 1/(4Nm + 1). Heredity, 82, 117– 125. Wilder CM, Brooks TM, Lens L (2000) Vegetation structure and composition of the Taita Hills forests. Journal of the East African Natural History Society, 87, 181–187. Zimmerman DJ, Turner DA, Pearson DJ (1996) Birds of Kenya and Northern Tanzania. Russell Freedman Books, Halfway House, South Africa.

Peter Galbusera uses genetic tools to study dispersal, population structure and demography of animals. Mwangi Githiru is studying bird ecology and behaviour. Luc Lens is interested in the ecology of animals (mainly birds). Erik Matthysen conducts and supervises research in the Laboratory of Animal Ecology on behavioural, ecological and evolutionary aspects of animal dispersal and population structure.

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421

F R A G M E N T A T I O N A N D P O P U L A T I O N E Q U I L I B R I U M 1421

Appendix Product size range, number of alleles and optimal reaction condition per microsatellite primer set Locus

PCR product size (bp)

No. of alleles

Annealing temp. (°C)

MgCl2 conc. (mm)

Pat14 Mcyµ4 Gf5B Ltmr6 Gf6 WBSW2 WBSW9

143 –173 132 –160 194 –218 190 –198 136 –154 125 –131 102 –122

24 15 19 3 11 4 9

50 55 57 54 56 54 54

1.5 1.0 1.5 2.0 2.5 1.5 1.5

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1409–1421