DISCUSSION

Variability of microsatellites in D. simulans: To assess levels of microsatellite variability in D. simulans, we conducted a survey of PCR product lengths at 34 loci. For 16 of these microsatellites, which were identified from D. melanogaster sequences in GenBank (Wetterstrand 1997), the expected heterozygosity and variance in repeat number detected in our survey of D. simulans are significantly less than the variability found in a similar survey of the homologous microsatellite loci in D. melanogaster (see Figures 2 and 3; paired sign test, P = 0.021 for heterozygosity and P = 0.021 for variance in repeat number; D. melanogaster data from Wetterstrand 1997). In contrast, D. melanogaster and D. simulans expected heterozygosities and average variance in repeat number are not significantly different from each other (Mann Whitney U = 45, P = 0.064, and U = 53, P = 0.150, respectively) when assaying the same types of microsatellites (e.g., dinucleotide repeats) in the species in which they were identified (i.e., when one compares D. simulans-derived dinucleotide microsatellites surveyed in D. simulans to D. melanogaster-derived dinucleotide microsatellites surveyed in D. melanogaster levels of variability are similar; we compared only the D. simulans data from Zimbabwe, Ecuador, and Florida to the D. melanogaster data from Zimbabwe, Ecuador, and Florida). This is consistent with the ascertainment bias hypothesis regarding the choice of microsatellite loci to be surveyed (Ellegrenet al. 1997). A separate investigation of this issue in D. simulans and D. melanogaster will be reported elsewhere (Hutteret al. 1998).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

—Microsatellite variability in D. melanogaster from Zimbabwe, Ecuador, Florida, China, and Australia (data from Wetterstrand 1997). (A) Average expected heterozygosity per locus. (B) Average variance in repeat number. (C) Average number of alleles per locus. (D) Private alleles. Numbers in parentheses are the average values for nine dinucleotides only.

Patterns of locus-specific variability: A recent study of microsatellite variation in natural populations of D. melanogaster shows that microsatellites may be influenced by the effects of selection acting on linked sites (Schlöttereret al. 1997), leading to a reduction in variation at the microsatellite locus. We have seen this pattern as well in our surveys of microsatellite variation in D. simulans. Two microsatellite markers (DROFASI and DROSEV1) are completely fixed for one allele size for our Florida sample, while remaining variable in other populations. In addition, one locus (DMPROSPER) is fixed in the Ecuador population and highly variable in three other populations (heterozygosities at the locus are 0.508, 0.337, and 0.268, for Zimbabwe, California, and Florida, respectively). Adaptive substitutions closely linked to the microsatellite loci may be responsible for these patterns. Variability is especially low near the achaete-scute region of D. simulans where no variation is seen at microsatellite DROACS2 across all four populations. This microsatellite is in the 5′ flanking region of the scute gene, and the reduced level of microsatellite variation is consistent with the reduced nucleotide sequence variation found in this region near the tip of the X chromosome (Begun and Aquadro 1991). Further analysis of the region surrounding these microsatellites is needed to determine whether adaptive fixation (e.g., Begun and Aquadro 1991) or background selection (Charlesworthet al. 1993) account for these low levels of variation.

Population structuring and genetic differentiation: Surveys of D. melanogaster-derived and D. simulans-derived microsatellites in D. simulans yield roughly equivalent F_ST between the Zimbabwe, Ecuador, Florida, and California population samples. F_ST is moderate among the four D. simulans populations overall (mean = 0.136), and not significantly different (Table 4; Wilcoxon rank sum test, P = 0.256) from the F_ST found for a similar microsatellite survey in D. melanogaster (mean = 0.105). Our calculations of F_ST in D. simulans do not differ significantly between D. melanogaster-derived and D. simulans-derived microsatellites, while average heterozygosity differs by 36% between the two surveys overall.

Unlike similar surveys of allozymes, our surveys of microsatellites in D. simulans (this study) and in D. melanogaster (Wetterstrand 1997) do not show significant differences between the species for their overall level of differentiation. As a comparison to the spatial structure detected for allozymes, in Table 4 we have summarized the average F_ST values for the microsatellite markers on each chromosome. Differences in the levels of population structure would not be expected between D. simulans chromosomes a priori because inversion polymorphisms are not common for the species (Lemeunier and Aulard 1992). However, the X and second chromosomes are the main chromosomes where allozyme differentiation in D. simulans is markedly lower than for D. melanogaster (Singh 1989; Table 3 from Choudhary and Singh 1987). Microsatellite F_ST values for any one of the chromosomes, or for all loci pooled together are not significantly different between the species (Wilcoxon rank sum test; see Table 4).

Partitioning of variation among D. simulans populations: Levels of variability vary widely among populations of D. simulans. Having fewer alleles per locus and fewer private alleles in our New World D. simulans samples suggests that founder effects may have occurred during the spread of D. simulans populations into the New World (Lachaiseet al. 1988).

A decrease in the average number of alleles per locus with less of an effect on average heterozygosity is consistent with theoretical and experimental studies of population bottlenecks (Maruyama and Fuerst 1985; Leberg 1992). Maruyama and Fuerst (1985) predict that a population just having undergone a single bottleneck event should have an apparent deficiency in the number of alleles compared to its heterozygosity. Leberg (1992) demonstrated empirically that a reduction in the number of alleles per locus is a more sensitive indicator of a postbottleneck population than is the multiple-locus heterozygosity. The deficit in the number of alleles is mainly expected due to the absence of alleles expected to appear only once or twice in a sample and could be formulated as an excess in heterozygosity, given the number of alleles observed. However, when the population bottleneck is followed by rapid population growth (Maruyama and Fuerst 1985) an apparent excess in the number of alleles (or deficiency of heterozygosity) should be observed.

To further investigate whether founder effects may have occurred during the colonization of New World habitats by D. simulans, we used the program BOTTLENECK (Cornuet and Luikart 1996; Piryet al. 1996) to implement three tests for an excess of heterozygosity in our D. simulans data when compared to the number of alleles observed. This is essentially the equivalent of testing for a deficiency in the number of alleles observed, given the heterozygosity expected from microsatellite allele frequencies. The results of these tests in Table 5 are consistent with a historically recent reduction in size for New World populations of D. simulans under the infinite alleles model, but results do not show a significant deviation from the expectation in these populations under the stepwise mutation model. We could interpret this to mean that the microsatellites in our survey follow the stepwise mutation model in New World populations and that there has been a rapid expansion of the Zimbabwe population (or some other reason for an excess of rare alleles such as recent migration, or that the alleles are being maintained at low frequencies because they are slightly deleterious). However, given the distribution of private alleles among D. simulans populations and our knowledge of the biogeography the species subgroup, we are led to favor the interpretation that the data may more closely fit the infinite alleles model and that founder effects occurred during colonization of New World habitats by D. simulans.

Also interesting are the average F_ST values for the four D. simulans populations surveyed as summarized in Table 3. The level of differentiation found between Ecuador and Florida is highest among pairwise comparisons of differentiation including comparisons between Zimbabwe and Ecuador or Zimbabwe and either North American population. One logical way to interpret these data is to infer that the Ecuador population was founded earlier than the Florida and California populations as a separate occurrence rather than as a sequential founding of South, then North America.

Implications of our study for understanding the demographies of D. simulans and D. melanogaster: Several aspects of the microsatellite, allozyme, and nucleotide variation suggest complex historical differences between these species. The primary goal of this effort was to understand the levels of population subdivision within D. simulans. A related motivation for surveying microsatellites was to assess species-effective population size, as this is also useful in interpreting the demography of this species. In Table 6, we have summarized the results of several comparisons of variability and differentiation for D. simulans and D. melanogaster. Summary tables of this sort are difficult to compile due the peculiarities of each study, and it has not been altogether possible to limit the comparisons to surveys which included samples from geographically similar locations in each species. Viewed in a qualitative sense, however, some meaningful conclusions can be drawn.

Assuming that microsatellite alleles are selectively neutral or nearly neutral, one would expect microsatellite variability to be greater in a species with a larger effective population size. Surveys of nucleotide polymorphisms have suggested that the population size of D. simulans has remained large for a significant period of time, and the inferred effective population size of this species being on average two- to threefold higher than for D. melanogaster (Aquadroet al. 1988; Aquadro 1992; Begun and Aquadro 1995; Moriyama and Powell 1996).

We have compared the levels of microsatellites variability for D. simulans and D. melanogaster using the measures of both heterozygosity and variance in repeat number. For the most equitable comparison permitted by our analysis (a comparison of just dinucleotide repeats surveyed in the species from which they were identified, assaying similar populations), we do not see a significant difference in the level of variability between these species (see Table 6). There is small (though nonsignificant) trend for variability to be higher in D. melanogaster than in D. simulans. There is also a significant difference in the average number of alleles per locus, with D. melanogaster having more alleles on average than D. simulans. However, much of this difference results from the deficit of alleles among New World populations D. simulans as described above. While we do not yet know the significance of these findings, they seem inconsistent with a larger effective population size in D. simulans relative to D. melanogaster, equal mutation rates in the species, and/or strict neutrality of microsatellites. One possibility is a lower mutation rate in D. simulans; however, additional studies are needed to determine this directly. Also, it is worth noting that if we compare just African populations of the two species, there is no difference in the level of nucleotide variability (π) in the introns of the vermilion gene (see Table 6). This is interesting because the Zimbabwe is the only population from our sample predicted to be at migration-drift equilibrium. Among the New World populations of D. simulans, variability seems to be higher than D. melanogaster in single nucleotide polymorphisms yet lower for microsatellite variability. The reasons for this discrepancy in the direction of the difference in these populations for nucleotide variability and microsatellite remains to be explored. But there may be differences in the rate of the introduction or loss of new microsatellite or single nucleotide polymorphisms under the varying demographic parameters that are suggested from the current analysis in D. simulans. An important test will be to determine the impact of regional rates of recombination with respect to the variability at these microsatellites in D. simulans. The distribution of private alleles among populations of D. simulans is most striking in comparison to that for D. melanogaster. In contrast to our finding that private alleles for D. simulans are scarce outside of Africa (Figure 2), private alleles in D. melanogaster populations are distributed more uniformly among populations (Wetterstrand 1997; see Figure 3). This suggests D. simulans colonized its current species range more recently than did D. melanogaster. Based on the assumption that variation within newly colonized areas represents a subset of variation segregating in more ancestral populations, and that each unique allele arises by an independent mutation, we estimate that D. simulans populations have had much less time than even D. melanogaster for mutations (private alleles) to arise since colonization of the New World. This hypothesis is consistent with theoretical and empirical studies of population bottlenecks and is supported by few rare alleles among New World populations of D. simulans. If such a process has also occurred in D. melanogaster, it appears that enough time has elapsed to allow recovery of rare variants among D. melanogaster New World populations. However, if the mutation in D. simulans were lower as suggested above, recovery of private alleles after a founding event would be slower.

Studies of allozyme, morphological, and behavioral traits have revealed less differentiation in D. simulans compared to D. melanogaster, and allozyme clines are much less pronounced in D. simulans than in D. melanogaster. Allozyme differentiation data for geographically similar populations from each species are summarized in Table 6. There is a significant difference in the level of allozyme differentiation but not for microsatellites. Given that microsatellites might be expected to evolve closer to neutrality, these data may indicate that a sufficient amount of time has not passed for selection to become apparent in D. simulans. However, these patterns could also result from lack of genetic heterogeneity within which such clines are ultimately selected. Colonization of much of the current D. simulans species range may have occurred from a source lacking spatial substructure, or simply may have occurred too recently for the type of selection that has modified the genetic architecture of D. melanogaster to be effective in D. simulans. To test between these hypotheses, and further understand the contribution of genetic differentiation to the colonization history of each species, it will be necessary to analyze genetic variation in populations that are believed to be more closely linked to founding D. simulans and D. melanogaster populations and to look for evidence of genetic differentiation among them. More extensive sampling is needed especially in D. simulans to truly get a picture of what differentiation has occurred within this species and to understand the changes that may have occurred during worldwide colonization. Our results do suggest, however, that there have been some significant historical differences in the demography of each species.