Demographic Expansion

The observation of a chromosome-wide excess of singletons in this study is in close agreement with the results reported by GLINKA et al. (2003). Since we found no evidence for selection by various neutrality tests applied, this chromosome-wide pattern may be due to demographic processes. STAIJCH and HAHN (2005) have recently shown that also admixed population samples lead to a negative Tajima’s D. However there is strong evidence for sexual isolation between D. melanogaster from Zimbabwe and those from other geographic locations (WU et al. 1995). In addition, evidence of admixture of the X chromosome by alleles of non-African ancestry of the rural D. melanogaster population in Zimbabwe is missing (KAUER et al. 2003). Therefore, the overall excess of singletons clearly shows that the ancestral population has recently been expanding. This size expansion cannot be due to the colonization of new habitats, since we found no sign of adaptive processes (see also GLINKA et

43

X-Linked Variation of African D. melanogaster

climatic changes in the past 20,000 years, which has been a key determinant of the distribution of animal and plant species around the world (e.g., HEWITT2000).

During this time, the Earth’s climate has undergone a transition from glacial to interglacial conditions resulting in large biotic responses including migrations of individual taxa and rearrangements of vegetation (WEB III and BARTLEIN 1992). The last glacial maximum (LGM) in the late Pleistocene (18 to 21 kya) was dry and arid, leading to a reduction of rain forests in favor to an extension of deserts and a mosaic of savannas and open forests on the African continent (DE VIVO and CARMIGNOTTO 2004). After the LGM, however, the climate changed substantially towards warmer and moisture conditions and rains returned 12 kya in the Holocene (GROVE 1993). In East Africa, the Holocene climatic optimum (HCO) occurred between 12,000– 10,000 to 4,000–3,000 years, in which forests began to expand (MALEY 1993) and dense savanna was covering most of that region (DE VIVO and CARMIGNOTTO 2004). Assuming that D. melanogaster was a forest-dwelling species during that time, stable forest habitats in Central Africa (DE VIVO and CARMIGNOTTO 2004; LACHAISE

and SILVAIN 2004) could have served as a refuge during the LGM, from which the

ancestral population expanded during the HCO. In contrast to this hypothesis, the reduction of forest during the LGM could also explain the wild-to-domestic habitat shift in D. melanogaster, since some of the hunter-gatherers (HG) were already sedentary (LACHAISE and SILVAIN 2004). However, any sign of a recent expansion is missing for HG’s (EXCOFFIER and SCHNEIDER 1999). Furthermore, human populations in Africa show signals of Pleistocene expansions at around 70 kya (EXCOFFIER and

SCHNEIDER 1999). This time estimate is substantially different from the estimate of our

study. Therefore, one can postulate that D. melanogaster expanded its range as a wild forest-dwelling species since the time when forests extended their ranges (see above), fitting well the estimated time of expansion of the ancestral population from Zimbabwe. Besides the compatibility of our sequencing data to an expansion model with various parameters, a constant size model fits equally well. However, the growth phase is unlikely to have started earlier than 30 kya. The inclusion of additional parameters (i.e., patterns of segregating sites; WAKELEY and HEY 1997) may result in sharper estimates and conditioning the likelihoods on the entire information in the data would reduce the variability of the estimates (GRIFFITHS and TAVARÉ 1994). In addition, since recombination is known to reduce the variance of the distributions of S and K (HUDSON 1990; WALL 1999), allowing recombination within each fragment might have led to sharper estimates as well.

44 X-Linked Variation of African D. melanogaster

However, the evidence of a recent size expansion of the ancestral D. melanogaster

population fits well with the observed levels of LD in our study. Since LD is primarily governed by recombination (i.e., recombination erodes LD over time; e.g., BROWN

et al. 2004), the observed deficit in LD can be explained by more recombination in polymorphism data than expected under the equilibrium model (PRZEWORSKI and WALL 2001). Thus, an overestimate of the population recombination rate, R (i.e., estimated by 2N_ec; see MATERIALSANDMETHODS), of the X chromosome (PRZEWORSKIet al. 2001) can result by a violation of the assumption of a constant population size (i.e., higher effective population size N_e) or by recombination events beside crossing-over. Gene conversion is likely to play an important role in breaking down allelic associations over short distances (FRISSE et al. 2001) and high levels of gene conversion were reported from the fourth chromosome of D. melanogaster, leading to lower than expected levels of LD (JENSEN et al. 2002). In contrast, ANDOLFATTO and WALL (2003) reported an excess of LD in the Zimbabwean D. melanogaster population only if this population is close to the mutation-drift equilibrium. The discrepancy with our study may be explained by the underlying assumptions of a constant mutation rate and/or constant population size (ANDOLFATTO and WALL 2003). In conclusion, both population size expansion and a sufficiently high rate of gene conversion may have led to the observed deficit in LD of this study.