Pollinating fig waSPS: genetic consequences of island recolonization

Authors


Monika Zavodna, Department of Plant Population Biology, Netherlands Institute of Ecology (NIOO-KNAW), PO Box 40, 6666 ZG Heteren, The Netherlands.
Tel.: +31-26-4791259; fax: +31-26-4723227;
e-mail: m.zavodna@nioo.knaw.nl; z_monika@hotmail.com

Abstract

The levels of genetic diversity and gene flow may influence the long-term persistence of populations. Using microsatellite markers, we investigated genetic diversity and genetic differentiation in island (Krakatau archipelago, Indonesia) and mainland (Java and Sumatra, Indonesia) populations of Liporrhopalum tentacularis and Ceratosolen bisulcatus, the fig wasp pollinators of two dioecious Ficus (fig tree) species. Genetic diversity in Krakatau archipelago populations was similar to that found on the mainland. Population differentiation between mainland coastal sites and the Krakatau islands was weak in both wasp species, indicating that the intervening 40 km across open sea may not be a barrier for wasp gene flow (dispersal) and colonization of the islands. Surprisingly, mainland populations of the fig waSPS may be more genetically isolated than the islands, as gene flow between populations on the Javan mainland differed between the two wasp species. Contrasting growth forms and relative ‘immunity’ to the effects of deforestation in their host fig trees may account for these differences.

Introduction

The mutualistic interaction between fig plants (Ficus spp.) and their pollinating waSPS (Agaonidae) is one of the most fascinating plant–insect interactions. Seed production in Ficus species is completely dependent on pollination by minute waSPS (1–2 mm) and the fig waSPS, in turn, depend on the fig trees for the completion of their life cycle. Generally, a specific wasp species pollinates a specific fig species. Since (i) waSPS are short-lived insects (Kjellberg et al., 1988), (ii) receptivity of figs is restricted to a short period of time (Khadari et al., 1995), and (iii) densities of fig trees in tropical forests are often low, dispersal of pollen-carrying female fig waSPS may be one of the most critical processes in the life cycles of both fig plants and waSPS.

Only a few studies have looked at the dispersal capacity and behaviour of fig waSPS. It is believed that long-distance wasp dispersal is mainly passive, floating on air currents (Ware & Compton, 1994a,b), followed by an active flight stage triggered by volatiles emitted by receptive figs (Ware et al., 1993; Compton, 2002). Nason et al. (1998) investigated the breeding structure of seven monoecious fig species in Panama using allozyme markers and estimated that pollen (and so pollinator) dispersal took place over distances of 5.8–14.2 km. The study of Shanahan et al. (2001) revealed that pollinating fig waSPS were capable of long-distance dispersal enabling them to colonize Long island, which lies 55 km off the coast of the mainland of Papua New Guinea. Recently, Harrison (2003) investigated fig wasp dispersal by trapping insects at various heights in an isolated fragment (ca. 4500 ha) of Bornean rain forest. His study indicated that pollinators of monoecious fig species might reach distances of 30 km or more, whereas dispersal of pollinators of dioecious figs may be more limited. It was suggested that the relative importance of active and passive dispersal varies among species, and may be related to the density of host figs and their canopy position.

In the present study, we focused on populations of Liporrhopalum tentacularis and Ceratosolen bisulcatus, the pollinators of two dioecious fig species (Ficus montana and Ficus septica, respectively), on the Krakatau archipelago and the mainlands of Java and Sumatra in Indonesia (Fig. 1). The sterilizing volcanic eruption in 1883, the isolation by sea (40 km to the nearest mainland), and the absence of direct human activity make the Krakatau islands an ideal natural study area. The (re) colonization process has been relatively well monitored and documented (Compton et al., 1988; Whittaker et al., 1989; Compton et al., 1994; Thornton et al., 1996; Whittaker et al., 2000; Parrish, 2002; Thornton et al., 2002). Thirteen years after the volcanic eruption Ficus species were among the first trees to recolonize Krakatau islands. Ficus septica, the host of C. bisulcatus, was documented for the first time in 1897 on Sertung and in 1930 was common on all other Krakatau islands (Docters van Leeuwen, 1936). Ficus montana, the host of L. tentacularis, was recorded for the first time in 1908 on Rakata. Since then it has spread on Rakata (Docters van Leeuwen, 1936), but has not been documented on the other Krakatau islands. Fig waSPS were first recorded on the Krakatau islands in 1919 (Dammerman, 1948), with C. bisulcatus first collected in 1920 (Dammerman, 1948). These reports suggest that fig waSPS were rare or absent on the Krakatau islands for about 20 years after the Ficus species had recolonized the islands then increased in numbers over the following 10 years. This coincided with the spread of Ficus species (Docters van Leeuwen, 1936), indicating that recolonization of the Krakatau islands by fig pollinators allowed Ficus species to produce seeds and expand their local populations. The relatively long initial absence of fig waSPS on the Krakatau islands may, however, suggest that this plant–insect relationship might be vulnerable to habitat fragmentation and disturbance.

Figure 1.

Map of the Sunda Strait region in Indonesia, the Krakatau islands and collection sites (filled spots). In August 1883, the three Krakatau islands (Panjang, Rakata and Sertung) were completely destroyed and sterilized by volcanic eruption. Since then the islands, which are about equidistant (40 km) from Java and Sumatra, have been recolonized by many animal and plant species and are nowadays covered by a dense secondary rainforest. The fourth island, Anak Krakatau, was formed due to renew volcanic activity in the middle of three existing islands and emerged above the sea in 1930. The vegetated area of Anak Krakatau is largely restricted to the eastern coastal strip.

Using recently developed highly polymorphic codominant microsatellite markers (Zavodna et al., 2002 and this study) we investigated, for the first time, the genetic diversity and population structure of pollinating fig waSPS associated with two dioecious fig species. The host plants of the pollinators L. tentacularis and C. bisulcatus occur in similar habitats, but differ in their growth-form. Compton et al. (2000) suggested that fig waSPS might actively fly up above the canopy irrespective of growth form of their host plants and then use the wind to disperse. Ficus septica, the host plant of C. bisulcatus, is a medium-sized tree, whereas F. montana, the host of L. tentacularis, is a small shrub. Hence, one would expect that L. tentacularis might encounter more difficulties on its way to the top of the canopy compared to C. bisulcatus. The potential differences in dispersal capabilities of two wasp species may shape the genetic structure of their populations.

Genetic diversity of the Krakatau island wasp populations was compared with that on the surrounding ‘mainlands’ of Java and Sumatra. In general, island populations have lower levels of genetic variation than corresponding mainland populations. Such a pattern may develop, when few founders colonized islands and gene flow (dispersal) from the mainland is restricted (Frankham, 1997). Therefore, due to the additive effects of colonization (founder event) and isolation by sea (fragmentation) we predicted reduced genetic diversity in Krakatau island wasp populations when compared to mainland populations. An effect of limited dispersal between populations is the reduction of gene flow between populations, which would lead to increased genetic differentiation amongst populations. As the life cycle of the studied pollinating fig waSPS is about 6–8 weeks (M. Zavodna, personal observation), several hundreds of wasp generations have passed since the initial recolonization of the Krakatau islands. This led us to assume that populations may have reached drift/migration equilibrium. We therefore used population differentiation to estimate gene flow, and so wasp dispersal, between populations. By comparing the two wasp species conclusions regarding possibilities for generalizations of the results and the relative importance of fig species characteristics will be drawn.

Material and methods

Study species

Liporrhopalum tentacularis Grandi (Agaoninae: Agaonidae) is the pollinator of F. montana Blume (subgenus: Ficus, section: Sycidium), which is a small, shrubby (up to 2 m tall), vegetatively spreading tree occurring in wet lowland forest in South-East Asia. The second species, C. bisulcatus Mayr (Agaoninae: Agaonidae) is a pollinating fig wasp of F. septica Burm f. (subgenus: Ficus, section: Sycocarpus), which is a medium-sized (up to 15 m tall) tree widespread in South-East Asia and Australia (Queensland). It occurs in lowland forest or secondary growth and open country, often near rivers and road verges.

As typical hymenopterans, both wasp species have a haplodiploid sex determination system, with females that are diploid (and develop from fertilized eggs) and males that are haploid (and develop from unfertilized eggs). Moreover, sexual size dimorphism is apparent, as males are smaller (ca 0.8 mm L. tentacularis and ca 1.7 mm C. bisulcatus), wingless, with shorter antennae, and do not disperse, while females are bigger (ca 1.5 mm L. tentacularis and ca 2.0 mm C. bisulcatus), have wings, long antennae, and disperse (Wiebes, 1994).

Both host plant species are gynodioecious (functionally dioecious) fig species, with separate hermaphrodite (functionally male) and female plants that produce pollen-carrying waSPS and seeds, respectively. Consequently, the waSPS were collected from ‘male’ figs only.

Collection sites and sampling

In the dry season between April and June 2001, both L. tentacularis and C. bisulcatus waSPS were collected in Indonesia – on Rakata island (06°09′S, 105°27′E; Krakatau archipelago), coastlines in the Sunda Strait region of Java (06°19′S, 105°50′E; Carita) and Sumatra (05°44′S, 105°35′E; Kalianda) and inland Java (06°34′S, 106°45′E; Bogor). Additionally, C. bisulcatus was also collected on Anak (06°06′S, 105°26′E; Krakatau archipelago) and on the island Sebesi (05°57′S, 105°30′E), which lies about halfway between Krakatau and Sumatra in the Sunda Strait region. All collection sites are shown in Fig. 1.

The adult host ‘male’ trees producing C. bisulcatus and L. tentacularis were often found in low densities. Hence, the area of collection sites varied from 1 to 10 ha. One or two mature male figs (D phase sensuGalil & Eisikowitch, 1968) per tree were collected and placed separately into small containers with mesh lids, just before pollinator emergence. The following day any emerged female waSPS were collected and preserved in 96% ethanol. Afterwards, the figs were dissected and inspected for males and any further females, which were also collected and preserved in 96% ethanol until microsatellite genotyping was performed.

Due to their biology and mating system, pollinating fig waSPS developing within a single fig are anticipated to be mostly siblings (Zavodna et al., 2002). Furthermore, in population genetic studies using microsatellite markers, diploid organisms are more informative than haploids. Therefore, one female wasp per collected fig was used for genotyping in this study.

DNA methods

Total genomic DNA was extracted from single waSPS using the PUREGENE® DNA extraction kit (Gentra). For L. tentacularis, seven recently developed microsatellite markers were used (Zavodna et al., 2002). About 8–22 female waSPS (one female wasp per fig) per collection site were analysed. The amplification reactions and polymerase chain reaction (PCR) protocols were performed as described in Zavodna et al. (2002). Fluorescently labelled PCR products were analysed on an ABI 3700 sequencer (Applied Biosystems, Warrington, UK) and scored using Genotyper software (Applied Biosystems, Warrington, UK).

For C. bisulcatus, microsatellite markers have been developed and are reported here for the first time. Briefly, di-, tri-, and tetranucleotide repeat enriched libraries of C. bisulcatus’ genomic DNA were constructed by a selective hybridization procedure (Karagyozov et al., 1993), using the method described by Zavodna et al. (2002). Genomic DNA from 250 individuals of C. bisulcatus was used for genomic library construction. Out of 563 positive clones, 80 were sequenced using PRISM dGTP Big Dye Terminator Ready Reaction Kit (Applied Biosystems, Warrington, UK). Primer pairs for 10 microsatellite repeats were designed using the software package LASERGENE (DNA star) and tested as described in Zavodna et al. (2002). Out of 10 tested markers, seven were polymorphic in C. bisulcatus. The three loci produced unambiguously scoreable amplification products and were used in this population genetic study (Table 1). Additionally, microsatellite markers developed for L. tentacularis were also tested on C. bisulcatus individuals. Two of them (LT3-11 and LT4-21A, see Zavodna et al., 2002 for primer sequences) were found to be polymorphic in C. bisulcatus, and therefore these two additional markers were also used for C. bisulcatus genotyping. About 5–17 female waSPS (one female wasp per fig) per collection site were analysed. The amplification reactions were performed as described earlier (Zavodna et al., 2002) with PCR protocols as shown in Table 1. Fluorescently labeled PCR products were analysed on an ABI 3700 sequencer (Applied Biosystems, Warrington, UK) and scored using Genotyper software (Applied Biosystems, Warrington, UK).

Table 1.  Characterization of three microsatellite markers in C. bisulcatus.
Microsatellite markerPrimer sequence forward, reverse, 5′–3′Amplification conditions*Repeat†Expected product length (observed range)EMBL nucleotide sequence database accession no.
  1. *CB60/32: 1 cycle 3 min at 94 °C, 32 cycles (15 s at 94 °C, 45 s at 60 °C, 2 min at 72 °C), 20 min at 72 °C. CB60/35: 1 cycle 3 min at 94 °C, 35 cycles (15 s at 94 °C, 45 s at 60 °C, 2 min at 72 °C), 20 min at 72 °C.

  2. †The number after hyphen denotes a number of mismatches from the perfect repeat.

CB1-02CTCCCGCCGACATTCCTAATCAC
GAGCTAAAGAGGGGCACCGAGTG
CB60/32(CT)29178 (128–182)AJ849708
CB4-15ATATAAAGGCAACACGAGAAGC
CGTCTAAGGGTATATTGTCACTGA
CB60/35(TGC)11176 (162–189)AJ849709
CB3-61CTGGGATCCCTCGCAGATT
ATTCTCCAGCTGATTGATGATTTC
CB60/35(AAC)26−5(GGC)8−2256 (242–260)AJ849710

Data analysis

For each fig wasp species, within-population genetic variation was assessed for each collection site using descriptive population genetic parameters such as total number of alleles per microsatellite locus, number of private alleles (alleles present in one population only), allelic richness (an estimate of the number of alleles per population standardized using the rarefaction method) (El Mousadik & Petit, 1996), observed and expected heterozygozity, and linkage disequilibrium between pairs of loci. These parameters were estimated using population genetic analysis software POPGENE Version 1.3.2 (available at http://www.ualberta.ca/fyeh/index.htm) and FSTAT Version 2.9.3.2 (Goudet, 1995) (available at http://www2.unil.ch/izea/softwares/fstat.html). Hardy–Weinberg genotypic equilibrium (HWE) was tested for each locus in all populations of both wasp species using population genetic analysis software TFPGA Version 1.3 (available at http://bioweb.usu.edu/mpmbio/). Using the statistical software STATISTICA Version 6 (StatSoft®), the effect of collection site on the allelic richness was tested with one-way anova (Analysis of variance). The assumptions of homogeneity of variances and normality were checked using Levene's and Kolmogorov–Smirnov tests, respectively (Leven's: L. tentacularisF3,24 = 0.21, P = n.s.; C. bisulcatusF5,24 =1.79, P = n.s.; Kolmogorov–Smirnov: L. tentacularisd =0.14, P = n.s.; C. bisulcatusd = 0.09, P = n.s.).

Weir & Cockerham (1984) estimates of genetic differentiation between collection sites (FST) and inbreeding (FIS) were calculated using population genetic analysis software FSTAT Version 2.9.3.2 (Goudet, 1995). Averages are given with either standard errors (SE) or confidence interval (CI). To test significance of genetic differentiation between collection sites the Fisher's exact test and a Markov chain algorithm (10 batches and 2000 permutations per batch) were performed with TFPGA Version 1.3. The FSTvalues are used as a comparative estimate of gene flow. Qualitative estimates of gene flow can be then determined (Hartl & Clark, 1997): when FST is less than 0.05 gene flow is high, gene flow is more restricted at FST values between 0.05 and 0.15, highly restricted between 0.15 and 0.25 and extremely restricted, when FST is higher than 0.25. Isolation-by-distance was tested for each species by examining the relationship between geographic distance and genetic differentiation. The significance of the association between FST/(1–FST) and the natural logarithm of distance (Rousset, 1997) was assessed using the Mantel test (Mantel, 1967) with 10 000 permutations. The test was performed with the software FSTAT Version 2.9.3.2 (Goudet, 1995).

Results

Genetic diversity

Microsatellite marker characteristics for the four L. tentacularis collection sites are shown in Table 2. The total number of alleles amplified from all individuals (n = 73) ranged from 4 to 29 per locus with a mean (±SE) of 12.9 ± 3.2 alleles per locus. Total numbers of private alleles of 3, 6, 4 and 26 were observed in Sumatra, Rakata, Java coastal, and Java inland sites, respectively. Since the mean number of alleles is dependent on sample size, allelic richness per locus and collection site was also calculated (Table 2). The mean allelic richness did not differ significantly across the collection sites (anova: F3,24 = 0.96, P = n.s.; Table 3). About 3 of 78 locus-pair tests indicated significant deviations from linkage equilibrium (P < 0.05). Observed heterozygozities per locus and over all loci in each collection site were lower than expected, suggesting the presence of inbreeding (Tables 2 and 3). Indeed, the estimates of inbreeding showed high values for each population, with the highest FIS value on Rakata (0.578) and the lowest on Sumatra (0.297) (Table 3). The estimated total FIS over all loci was 0.507 (95% CI 0.439–0.565) indicating a substantial overall shortage of heterozygotes. Consequently, most of the loci in each collection site were not in HWE (all seven loci in Java inland and coast populations, six loci in the Rakata population and two loci in the Sumatra population, P < 0.05).

Table 2.  Microsatellite marker characteristics per each collection site of L. tentacularis.
 LT3-08LT3-11LT4-15LT4-16LT4-21ALT4-21BLT4-27
  1. NA:total number of alleles; NP: number of private alleles; AR: allelic richness; HO: observed heterozygozity; HE: expected heterozygozity. n represents number of analysed wasp individuals.

Sumatra (n = 8)
 NA23531410
 NP0000003
 AR23531410
 HO0.500.380.630.250.250.63
 HE0.530.570.730.240.690.90
Rakata island (n = 21)
 NA351052714
 NP0101012
 AR24842510
 HO0.140.520.450.140.190.290.22
 HE0.510.660.880.480.370.740.93
Java coast (n = 22)
 NA24973614
 NP0021001
 AR24752510
 HO0.140.360.500.320.140.450.42
 HE0.490.630.870.690.210.770.93
Java inland (n = 22)
 NA66111241113
 NP3325156
 AR5578389
 HO0.500.500.320.550.180.330.36
 HE0.780.780.790.850.630.880.89
Total (n = 73)
 NA68151541329
 AR45873712
 HO0.290.450.440.350.150.350.37
 HE0.680.770.880.760.560.800.95
Table 3.  Mean allelic richness (AR), observed (HO) and expected (HE) heterozygosity and estimates of inbreeding (FIS) coefficient over all loci for each of the collection sites of L. tentacularis and C. bisulcatus. n represents number of analysed wasp individuals.
 L. tentacularisC. bisulcatus
nARHOHEFISnARHOHEFIS
  1. *L. tentacularis was absent in these collection sites.

Sumatra84.000.380.490.30152.950.510.570.15
Sebesi*****123.020.410.640.41
Anak Krakatau*****112.800.350.530.39
Rakata215.010.280.640.58173.110.480.620.26
Java coast225.080.330.640.50143.040.450.610.31
Java inland226.390.390.780.5252.900.390.490.32

Ceratosolen bisulcatus waSPS were collected at six locations. Microsatellite marker characteristics for each collection site are shown in Table 4. The total number of alleles amplified in all combined individuals (n = 74) ranged from 5 to 17 per locus with a mean of 10.4 ± 2.0 alleles per locus. Total numbers of private alleles were 2, 3, 3, 4, 6, and 1, in Sumatra, Sebesi, Anak Krakatau, Rakata, Java coastal, and Java inland sites, respectively. The differences in allelic richness between collection sites were not statistically significant (anova: F5,24 = 0.07, P = n.s.; Table 3). About 2 of 59 locus-pair tests showed significant deviations from linkage equilibrium (P < 0.05). Also, in this wasp species the observed heterozygosities per locus and over all loci in each population were again lower than expected, suggesting presence of inbreeding (Tables 3 and 4). Calculated estimates of inbreeding showed the highest FIS on Sebesi (0.407) and the lowest on Sumatra (0.153) (Table 3). The estimated total FIS over all loci was 0.292 (95% CI 0.197–0.412). Consequently, at some collection sites, some of the loci were found not to be in HWE (four loci in the Sebesi population, three loci in the Sumatra and Rakata populations, and two loci in Anak and Java coastal populations, P < 0.05). No significant deviations from HWE were found at the Java inland site, which was perhaps due to the small sample size at this location.

Table 4.  Microsatellite marker characteristics per each collection site of C. bisulcatus.
 CB1-02CB4-15CB3-61LT3-11LT4-21A
  1. NA: total number of alleles;NP: number of private alleles; AR: allelic richness; HO: observed heterozygozity; HE: expected heterozygozity. n represents number of analysed wasp individuals.

Sumatra (n = 15)
 NA88533
 NP00101
 AR43322
 HO0.730.530.530.090.64
 HE0.860.670.650.260.54
Sebesi island (n = 12)
 NA85534
 NP10101
 AR33332
 HO0.420.830.420.000.36
 HE0.700.700.740.660.54
Anak island (n = 11)
 NA77242
 NP10020
 AR44232
 HO0.500.270.450.500.00
 HE0.790.830.370.530.26
Rakata island (n = 17)
 NA107553
 NP11110
 AR43332
 HO0.590.760.530.180.33
 HE0.810.710.670.630.40
Java coast (n = 14)
 NA126462
 NP20040
 AR43242
 HO0.710.570.290.570.08
 HE0.810.620.470.790.49
Java inland (n = 5)
 NA56322
 NP10000
 AR45222
 HO0.600.600.200.330.20
 HE0.800.910.510.330.20
Total (n = 74)
 NA17108125
 AR43332
 HO0.600.610.430.260.31
 HE0.800.720.600.580.46

Genetic population differentiation

The estimated total FST value over all loci in L. tentacularis was 0.132 (95% CI 0.062–0.233). The Fisher's exact tests performed on the pair-wise comparisons indicated significant population differentiation between all the four collection sites (Table 5a). Pair-wise FST values ranged from 0.008 to 0.233. However, it is important to note that population differentiation between mainland coastal sites and Rakata (Krakatau archipelago) was weak (FST < 0.015; Table 5a). There was a significant evidence of isolation by distance (Mantel test, r2 = 0.85, P < 0.05) among L. tentacularis populations.

Table 5.  Estimates of genetic differentiation (FST) between collection sites of (a) L. tentacularis and (b) C. bisulcatus.
 SumatraRakataJava coastJava inland  
(a) L. tentacularis
 Sumatra      
 Rakata0.014***     
 Java coast0.010**0.008***    
 Java inland0.233***0.176***0.191***   
 SumatraSebesiAnak KrakatauRakataJava coastJava inland
  1. Significance of population differentiation: *P < 0.05; **P < 0.01; ***P < 0.001.

(b) C. bisulcatus
 Sumatra      
 Sebesi0.084***     
 Anak Krakatau0.0110.052***    
 Rakata0.010*0.004–0.009   
 Java coast0.027*0.041***0.011*0.003*  
 Java inland0.0110.026–0.040–0.0190.012 

Pair-wise comparisons of genetic differentiation FST between collection sites of C. bisulcatus are shown in Table 5b. The estimated total FST value over all loci was low at 0.019 (95% CI 0.005–0.039), but the Fisher's exact test performed on pair-wise comparisons revealed significant differentiation between some of the collection sites (Table 5b). No significant correlation was observed between genetic differentiation and geographic distance (Mantel test, r2 = 0.02, P = n.s.) among C. bisulcatus populations.

Discussion

Neither L. tentacularis nor C. bisulcatus showed reduced genetic variation on the recently recolonized Krakatau islands when compared to the surrounding mainland populations on Java and Sumatra. The mean allelic richness per locus on the Krakatau collection sites was similar to that on the mainland sites in both pollinating fig wasp species, implying that there was no strong founder effect. Studies on the effects of founder events on genetic structure have been carried out in natural populations in the context of metapopulation theory (Hanski & Gilpin, 1991), e.g. in Silene dioica (Giles & Goudet, 1997). The effect of extinction/recolonization events on the genetic structure of the populations is dependent on population dynamics and determined by the number of founders (the strength of the founder event), the rate of population growth and immigration after initial founding (Nei et al., 1975; Ingvarsson, 1997). In general, founding events are associated with a reduction of effective population sizes (local bottleneck events) leading to decreased genetic diversity within and increased genetic differentiation among local populations (Giles & Goudet, 1997; Ingvarsson et al., 1997). However, consistent with our results, some studies did not observe reduced genetic diversity in recolonized populations (e.g. Frankham, 1997; Helenurm, 2001; Barber et al., 2002; Clegg et al., 2002; Tremetsberger et al., 2003).

Our observations that genetic variation in the Krakatau islands populations is comparable to that on the mainland can therefore be explained by various scenarios: (1) after initial colonization the Krakatau island populations have been exposed to high immigration rates from the mainland, indicating that the islands are by no means isolated for fig waSPS and/or (2) the Krakatau island populations were founded by a few highly diverse colonists that subsequently experienced a high rate of population growth, maintaining genetic diversity.

How likely are these scenarios? First, the fig waSPS have been shown to be above-canopy flyers taking advantage of faster-moving air to disperse for long distances (Compton et al., 2000; Harrison, 2003). The flowering of the figs is asynchronous, with usually several figs per tree maturing at the same time and tens (L. tentacularis) to hundreds (C. bisulcatus) of fig waSPS emerging from a single fig (M. Zavodna, personal observation). Therefore, large numbers of fig waSPS may be released from the figs at a given time and given the strong air currents in the Sunda Strait (Thornton et al., 2002) a high immigration rate (gene flow) of fig waSPS is possible. Second, if ‘male’ receptive figs are available locally at the time of wasp release, the high production of waSPS per fig tree and their relatively short-life cycle may result in a strong population growth. It is thus likely that the observed genetic variation in the Krakatau islands is maintained by both wasp immigration and population growth. In 1992 Compton et al. (1994) found a reduced pollination rate of 81% for F. septica fruiting trees (n = 80) on Anak Krakatau compared to the other islands of the Krakatau archipelago, indicating some extent of pollinator shortage. They suggested that although the F. septica population on Anak Krakatau was probably too small to maintain a resident C. bisulcatus population, pollinator dispersal from the other islands might not have been limiting. To our knowledge, other dispersal data for L. tentacularis and C. bisulcatus are lacking. The study of Serrato et al. (2004) on Mexican Ficus populations also suggested that the maintenance and persistence of the fig–fig wasp mutualism might depend on long-distance dispersal by pollinating waSPS (see also Nason et al., 1996).

In both investigated pollinating fig wasp species, we have observed deficits of heterozygotes relative to HWE expectations in all loci across all populations. This could result from our sampling regime (the Wahlund effect) or high levels of inbreeding rather than the presence of null alleles. The Wahlund effect, however, is unlikely as the deficiency of the heterozygotes was observed in all locations irrespective of the size of the collection area and distance between neighbouring trees. In contrast, the pollinator's mating system, with the occurrence of local mate competition, predicts inbreeding in pollinating fig waSPS (Herre, 1985,1987; Herre et al., 1997). Accordingly, estimated inbreeding levels were high across all populations of both wasp species, with L. tentacularis wasp populations being more inbred than those of C. bisulcatus. Similarly, high inbreeding levels were shown in other genetic studies of fig wasp populations (Greeff, 2002; Molbo et al., 2004). Inbreeding levels reflect the breeding structure of pollinating fig waSPS and are related to the number of female waSPS ovipositing (foundresses) in a fig fruit. Therefore, the difference in inbreeding level that was found between the two wasp species in this study is expected to result from different average numbers of foundresses per fruit. Small fig fruits tend to be host to fewer foundresses (or even a single foundress) than larger fruits (Herre, 1989; Anstett et al., 1996). Consequently, mating of pollinating waSPS within a small fig fruit is more likely to be between siblings, whereas mating within larger fruits is likely to involve more nonsiblings. The host fig fruits of C. bisulcatus are approximately 2.5-fold larger in diameter than the host fig fruits of L. tentacularis (M. Zavodna, unpublished data). Field observations of Moore (2001) and kinship analysis of broods (Zavodna, 2004) revealed that 90% of the L. tentacularis broods have up to three foundresses contributing to a brood, with 50% of the broods being single foundress broods. To our knowledge the number of C. bisulcatus foundresses ovipositing in a fig has not been investigated.

Our genetic analyses revealed different patterns of genetic differentiation and gene flow between populations in the two pollinating fig wasp species. The estimates of L. tentacularis gene flow determined from FST values revealed high gene flow between Rakata (Krakatau archipelago) and coastal sites in Sumatra and Java (FST < 0.015). This indicates that distances of over 40 km across open sea (which is approximately the distance between Rakata and the mainlands of either Java or Sumatra) may not be limiting for wasp dispersal. However, gene flow between the Javan inland site and all other collection sites was found to be highly restricted (FST > 0.17). This may not be surprising in the case of the most distant sites sampled (Sumatra and Java inland), but genetic differentiation within Java was similarly high. The isolation-by-distance correlation between populations of L. tentacualris was linked to Java inland site (data not shown). Therefore, the divergence of the Java inland site cannot be explained by isolation-by-distance alone, and other geographic barriers, historical differentiation or reproductive isolation may have played a role (Bossart & Pashley Prowell, 1998).

In comparison to L. tentacularis, the genetic differentiation between all collection sites of C. bisulcatus waSPS was lower (FST < 0.08). This indicated high to moderate gene flow between all collection sites, suggesting that long-distance dispersal (over 40 km) is common in C. bisulcatus. Accordingly, there was no correlation between geographic distance and genetic differentiation in this species. As far as we know, there have been only two other studies (Yokoyama, 2003; Molbo et al., 2004), which have investigated the genetic structure of pollinating fig wasp populations. The study of Yokoyama (2003) using mtDNA loci suggested restricted gene flow in pollinators of two dioecious Ficus species between two of the Ogasawara islands, which are about 50 km apart; however, genetic differentiation among these populations based on RAPD markers was low. The genetic analyses by Molbo et al. (2004) using microsatellite markers in four pollinators of monoecious Ficus species did not reveal significant genetic differentiation within any of the wasp species at a geographical scale of 10–20 km, which is consistent with our results.

It is clear that the dispersal rates of both wasp species in this study are high across open sea between mainland coastal sites and the islands, and sufficient for successful colonization of the Krakatau islands. However, the differences in the two wasp species’ dispersal that are found within the mainland of Java imply that further generalizations cannot be made. These findings support Harrison's (2003) suggestion that dispersal rates vary among species and may be related to the density of host figs and their canopy position. First, in contrast to C. bisulcatus, L. tentacularis might encounter more difficulties on its way to the top of the canopy from the forest understory host plant. High mortalities of fig waSPS during transit from tree to tree have been demonstrated (Bronstein, 1989; Herre, 1989). Second, F. septica is a weedy and widespread Ficus species, which commonly occurs along roads, whereas F. montana is a more habitat-specialized species occurring in lowland forests (Zavodna, personal observation). Ongoing habitat fragmentation on Java due to human activity has been documented (Whitten et al., 1996) and it is likely that its effect on the fig plants and their pollinators may vary. Third, the fact that fig waSPS use the wind to passively carry them over long distances to reach their new host plant (Compton et al., 2000), implies that the force of the wind and its direction will greatly affect pollinating fig wasp dispersal. Wind in the Sunda Strait is known to be strong, favouring movement from Sumatra to Krakatau rather than the reverse (Thornton et al., 2002). Canopy position of the host Ficus species and habitat fragmentation may not affect pollinating fig wasp dispersal if climatic conditions are favourable (e.g. strong winds), as is demonstrated by high gene flow in both wasp species in the Sunda Strait region. On the other hand, canopy position of the host Ficus species and their ‘immunity’ to the effects of deforestation and habitat fragmentation may affect pollinating fig wasp dispersal if climatic conditions are not favourable, which may explain the observed differences in gene flow in the wasp species across the mainland of Java.

The present study indicates that the populations of these fig wasp species on the Krakatau islands do not lack genetic variation and that high gene flow (long-distance dispersal) in the waSPS may facilitate the maintenance of the viable local fig–fig wasp populations on the islands. However, population genetic analyses of additional samples and collection sites, for example from inland Sumatra and Java, is warranted to improve our understanding of the maintenance of fig–fig wasp mutualism, especially in their mainland populations, which may turn out to be more genetically isolated than the islands.

Acknowledgments

This research was financially supported by the ‘Biodiversity in Disturbed Ecosystems’ priority program of the Netherlands Organization for Scientific Research (NWO; no. 895.100.018). We thank the Indonesian Institute of Sciences (LIPI) and the Indonesian National Parks (PHPA) for permission to conduct fieldwork in Indonesia. We gratefully acknowledge the assistance of Herbarium Bogoriense while working in Indonesia. We wish to thank Tracey Parrish, Drude Molbo, Allen Herre, Jamie Moore and George Weiblen for helpful discussions and advices when preparing fieldwork, Dirk Visser for excellent assistance with fieldwork in Indonesia, Wendy van't Westende and Danny Esselink for technical advice on development and analyses of microsatellite markers, and GreenomicsTM for very prompt assistance with electrophoreses. Discussions with Rene Smulders, comments of three anonymous reviewers and editing by Steve Compton improved the previous version of this manuscript. This is NIOO-KNAW publication number 3577.

Ancillary