Drosophila subobscura from Madeira
Populations of D. subobscura from Madeira, like those from the Canary Islands, are nearly monomorphic for the O3+4 chromosomal arrangement (Prevosti, 1971; Larruga et al., 1983). This is in contrast with the rich chromosomal polymorphism that most other populations harbour for this chromosome (Krimbas, 1992). Furthermore, the absence in the islands of arrangements present in high frequency in nearby continental populations (e.g. O3+4+8 in North-western Africa and O3+4+7 in the Atlantic coast of the Iberian peninsula) does not seem to favour the hypothesis of recent migration from the continent. Otherwise, selection against the establishment of these other arrangements in the islands should be very strong given their high frequency in those continental populations (see Khadem et al., 1998).
According to present results, the level of nucleotide variation in the rp49 gene region is similar in O3+4 lines from Madeira and the continent. Additionally, O3+4 lines from these locations are not genetically differentiated for that region. Some differentiation had been, however, detected in a previous restriction-map survey (Khadem et al., 1998). The discrepancy between the results of these studies might be partly due to the relatively low number of nucleotides sampled in the previous study and therefore to the high stochastic variance associated with nucleotide variation estimates.
Available data for chromosomal, mtDNA and rp49 variation in D. subobscura populations from Madeira and the continent do not support the hypothesis that extant populations from Madeira are the descendents of a single colonization event occurring soon after the origin of O3+4 (Khadem et al., 1998). Two different scenarios would be compatible with available data: (i) a rather recent and massive colonization of Madeira by continental D. subobscura, and (ii) multiple colonization events from the continent. In both scenarios, selection would have precluded the establishment in Madeira of chromosomal arrangements other than O3+4 (see Khadem et al., 1998). Although we cannot ascertain which scenario most likely reflects the origin of extant D. subobscura populations in Madeira, the similar level of nucleotide variation detected within O3+4 in insular and continental populations allows us to assert that the colonization of Madeira was not associated with a strong founder event.
Comparison of DNA variation between D. madeirensis and D. subobscura
The rp49 region is the first nuclear region whose variation has been analysed in a natural population from the endemic species D. madeirensis. Endemic species inhabiting rather small islands are expected to have a lower effective population size than closely related species with a worldwide distribution. Therefore, under the strict neutral model (Kimura, 1983), a lower level of nucleotide variation would be expected in endemic insular species than in mainland species. Comparison of nucleotide variation at the rp49 gene region between D. madeirensis and D. subobscura is not in agreement with that prediction. In fact, the estimated nucleotide diversity in D. madeirensis, which is monomorphic for the O3 arrangement, was similar to that estimated for O3+4 and slightly higher than for the Ost chromosomal arrangement (Table 1). The high level of nucleotide variation in D. madeirensis might be explained if ancestral populations of this species were much larger than current populations and the species had suffered a reduction in population size. This reduction could be associated with the colonization and expansion of D. subobscura in Madeira after the origin of D. madeirensis; it could also be associated with destruction of the natural habitat of D. madeirensis (laurisilva forest) occurred during the last 400 years (Doria, 1945; Frutuoso, 1979). In any case, ancestral populations of this insular species would probably be not as large as continental populations of D. subobscura. Nevertheless, the level of variation in D. madeirensis clearly indicates that the origin of this species was not associated with a strong founder event. In fact, D. madeirensis and D. subobscura exhibit a comparable level of nucleotide variation in the rp49 region and they also present a similar frequency spectrum, as measured by Tajima’s D and Fu and Li’s D and F statistics.
Present data on nucleotide variation in D. madeirensis and D. subobscura have been compared to those available for other closely related Drosophila species pairs with one insular representative. Two such pairs can be found in the triad formed by the cosmopolitan species D. simulans and the endemic species D. sechellia and D. mauritiana (Hey & Kliman, 1993; Kliman & Hey, 1993). These authors analysed nucleotide variation at the period, zeste and yolk protein 2 gene regions and found that D. simulans and D. mauritiana showed comparable levels of variation. In contrast, D. sechellia exhibited a much lower nucleotide variation than D. simulans. Our results showing an a priori unexpected high level of variation in the insular species (D. madeirensis) are quite similar to those reported for the D. simulans and D. mauritiana pair. Nevertheless, the expected positive correlation between heterozygosity and effective population size has been, and still is, a controversial issue in population genetics (Lewontin, 1974; Maynard Smith & Haigh, 1974; Gillespie, 1999, 2000). According to the strict neutral mutation model, the expected heterozygosity is a function of the effective population size (Kimura, 1983). Under several selection models, however, heterozygosity is nearly independent of the population size (Gillespie, 1999). For example, under some deleterious mutation models, heterozygosity is rather insensitive to the population size, and is thus mainly a function of the mutation rate (Gillespie, 1999). This insensitivity is also predicted by the pseudohitchhiking model (Gillespie, 2000) that considers the effect of advantageous mutations on the dynamics of neutral variation at a closely linked locus. Under this model, heterozygosity can even decrease with increasing population size.
Some shared polymorphisms between species were detected both in the sequence comparison of the period locus between D. simulans and D. mauritiana, and of the rp49 region between D. madeirensis and D. subobscura. In both cases, the data were compatible with the isolation model, where shared polymorphisms are due to common ancestry. However, shared polymorphisms between closely related species could also be due to the introgression resulting from rare hybridization between species. In fact, it has been recently proposed that, for regions not involved in reproductive isolation, introgressive hybridization might be more important than previously thought (Ting et al., 2000). If that were the case for the rp49 region, some of the observed shared variants could have originated by mutation in one of the descendent lineages after the split of the species, and they would have entered the second species gene pool by introgression. Although some hybrids between D. madeirensis and D. subobscura have been detected in recent collections (N. Khadem, unpublished results), this observation is not a proof of introgression. In addition, for a neutrally evolving gene, the relatively high number of fixed differences observed between species seems unlikely under the introgression scenario.
As in the species studied, the rp49 gene is located in a region affected by chromosomal inversions, there is an additional difficulty for explaining the presence of shared polymorphisms both if they are due to common ancestry or to introgressive hybridization. In fact, the unique origin of an inversion would a priori preclude the existence of shared polymorphisms between the ancestral and derived chromosomal arrangements. In our case, D. madeirensis is monomorphic for the ancestral O3 arrangement that went extinct in the D. subobscura lineage, while extant populations of D. subobscura segregate for the O3+4 and Ost arrangements that originated from O3. Therefore, gene transfer between arrangements, either by double crossing over or gene conversion, would be required to explain the presence of shared polymorphisms between species (Rozas & Aguadé, 1994; Rozas et al., 1999). In this case, the extent of shared polymorphism might be lower at the rp49 region than in other regions not affected by chromosomal inversions. Only analysis of multiple regions will allow establishing the effect of chromosomal polymorphism on the number of shared polymorphisms between D. madeirensis and D. subobscura. Additionally, the comparative analysis of nucleotide variation in multiple regions of D. subobscura and D. madeirensis will improve our understanding of the speciation process in Madeira.