Abstract
An ∼1.3-kb region including the rp49 gene plus its 5′ and 3′ flanking regions was sequenced in 24 lines of Drosophila simulans (10 from Spain and 14 from Mozambique). Fifty-four nucleotide and 8 length polymorphisms were detected. All nucleotide polymorphisms were silent: 52 in noncoding regions and 2 at synonymous sites in the coding region. Estimated silent nucleotide diversity was similar in both populations (π= 0.016, for the total sample). Nucleotide variation revealed an unusual haplotype structure showing a subset of 11 sequences with a single polymorphism. This haplotype was present at intermediate frequencies in both the European and the African samples. The presence of such a major haplotype in a highly recombining region is incompatible with the neutral equilibrium model. This haplotype structure in both a derived and a putatively ancestral population can be most parsimoniously explained by positive selection. As the rate of recombination in the rp49 region is high, the target of selection should be close to or within the region studied.
DROSOPHILA simulans, like D. melanogaster, is a cosmopolitan human commensal that originated in tropical Africa ∼2.5 mya (Lachaiseet al. 1988; Powell 1997). Populations of both species from that area could be considered ancestral populations and, therefore, neutral variation in those populations would be expected to be at mutation-drift equilibrium. On the other hand, populations from other areas would be derived populations and their variation might or might not be at equilibrium. In fact, there would have been ample room for adaptive evolution, i.e., for the action of natural selection, during the dispersal of these species from tropical to temperate regions. Additionally, derived populations of these species might still reflect the possible founder events associated with the out-of-Africa expansion of these species.
Both demographic events and natural selection acting in a particular genomic region can have similar effects on the pattern of nucleotide variation in that region. However, population history has a genome-wide effect and should affect, therefore, all regions of the genome. In contrast, both directional and balancing selection are locus specific and affect neutral variation only at loci tightly linked to the locus under selection. In this sense, the level of both within-population variation and between-population differentiation is generally lower for morphological characters, allozymes, and mitochondrial DNA variation in D. simulans than in D. melanogaster (as reviewed in Singh and Long 1992); this might reflect a more recent expansion of the distribution area of D. simulans (but see Ballardet al. 1996 for mtDNA).
Initial surveys of nucleotide sequence variation in nuclear genes of D. simulans generally analyzed few sequences sampled from different populations (see Begun and Whitley 2000 for references). An unusual haplotype structure was detected in some of these surveys (Begun and Aquadro 1995; Eaneset al. 1996; Hassonet al. 1998; Labateet al. 1999), suggesting the presence of at least two old lineages in this species.
Hamblin and Veuille (1999) have studied variation in two X-linked gene regions (vermilion and G6pd) in several African and non-African populations of D. simulans. A strong haplotype substructure was detected both in West African and in all non-African populations studied. The haplotype structure observed in the vermilion region departed from predictions of the neutral theory for stationary populations in the non-African, but not in Central African populations. This departure from neutrality was considered evidence of a bottleneck in their rather recent foundation (Hamblin and Veuille 1999). The authors also suggested that the nonequilibrium haplotype distributions might be compatible with ancient population subdivision and recent admixture in populations of D. simulans ancestral to the European and American populations.
The paucity in the number of haplotypes and/or in haplotype diversity detected in three out of the four loci analyzed in non-African populations by Hamblin and Veuille (1999) was considered to support a genome-wide phenomenon and, therefore, the population admixture hypothesis. In vermilion, however, the region surveyed was not randomly chosen, but its choice was based on the previous knowledge that it presented two divergent haplotypes (Begun and Aquadro 1995). Indeed, in the North American sample sequenced by Begun and Aquadro (1995), the haplotype test gave different results when applied to the complete region than when applied to the particular region chosen by Hamblin and Veuille (1999). To draw any general conclusion, it is important, therefore, to survey nucleotide variation in African and non-African samples of D. simulans in a larger number of randomly chosen regions.
We have analyzed variation in an ∼1.3-kb region encompassing the rp49 gene (named RpL32 in the FlyBase Drosophila database; http://flybase.bio.indiana.edu) in a European and a Southeast African population of D. simulans. This gene is located in band 99D of D. simulans and encodes ribosomal protein 49 (ribosomal protein L32 in FlyBase). Similarly to the vermilion region surveyed by Hamblin and Veuille (1999), recombination in this autosomal region is expected to be high (Kliman and Hey 1993a; Trueet al. 1996). Extensive surveys of variation in the homologous region of D. subobscura in relation to chromosomal polymorphism (Rozas and Aguadé 1993, 1994; Rozaset al. 1999) indicate that, at least in this species, the rp49 region is a neutrally evolving region with normal levels of nucleotide polymorphism. Surprisingly, our initial survey of nucleotide variation in a European population of D. simulans revealed an unusual haplotype structure showing one rather common haplotype with zero variation even though the complete sample had a normal level of variation. This motivated extending the survey to a putatively ancestral population of that species in an effort to discern between historical and selective explanations.
MATERIALS AND METHODS
Fly samples: Twenty-four lines randomly sampled from two natural populations of D. simulans were studied: 10 lines from Montblanc, Tarragona, Spain (SimS lines) and 14 from Maputo, Mozambique (SimMz lines). The European and African samples were collected in September 1993 and in August 1997, respectively. We obtained highly inbred lines after 10 generations of sibmating. We also used 1 line of D. melanogaster (line M66), which was collected in Montemayor, Córdoba, Spain in March 1990 and was subsequently made isochromosomal for the third chromosome by the standard series of crosses with the TM6/MKRS balancer stock.
DNA extraction, PCR amplification, and DNA sequencing: Genomic DNA was extracted using a modification of protocol 48 from Ashburner (1989). An ∼1.4-kb fragment, which included the rp49 gene (402 bp of coding region and a small intron of 59 bp) and its 5′ and 3′ flanking regions, was amplified by PCR (Saikiet al. 1988) using oligonucleotides designed on the published sequence of D. melanogaster (O'Connell and Rosbash 1984; GenBank accession no. X00848) . Several oligonucleotides, designed at intervals of ∼300 nucleotides, were used as primers for sequencing. The amplified fragments were cyclesequenced and separated on a Perkin-Elmer (Norwalk, CT) ABI PRISM 377 automated DNA sequencer following the manufacturer's instructions. For each line, the DNA was sequenced on both strands. The nucleotide sequences are available from the EMBL nucleotide sequence database under accession nos. Y13939 (D. melanogaster) and AJ309023–AJ309046 (D. simulans).
Data analysis: Nucleotide sequences were assembled using the SeqEd version 1.0.3 program (Applied Biosystems, Inc., Foster City, CA), multiply aligned using the Clustal W program (Thompsonet al. 1994), and edited with the MacClade version 3.06 program (Maddison and Maddison 1992). Phylogenetic analysis was performed with genetic distances corrected according to the Jukes and Cantor model (Jukes and Cantor 1969) using the neighbor-joining algorithm (Saitou and Nei 1987) implemented in the MEGA version 2 (Kumaret al. 2000) program. The analysis was conducted using the rp49 nucleotide sequence of D. melanogaster (line M66) as the outgroup; the bootstrap values were based on 1000 replicates.
The DnaSP version 3.50 software (Rozas and Rozas 1999) was used to estimate population genetic parameters and genetic distances and also to perform different neutrality tests. The confidence intervals (and the P values) of several test statistics were obtained by Monte Carlo simulations based on the coalescent process for a neutral infinite-sites model and assuming a large and constant population size (Kingman 1982a,b; Hudson 1983, 1990). The simulations were carried out either by assuming a value of θ (θ= 4Nu, where N is the effective population size and u is the per gene mutation rate; Watterson 1975) or by fixing the number of segregating sites. As in both cases the simulations yielded similar results, we present only results based on the coalescent conditional on the number of segregating sites. The simulations were performed assuming either no intragenic recombination or intermediate levels of recombination (Hudson 1983, 1990). Each computer simulation was based on 10,000 (for no recombination) or 1000 (for intermediate levels of recombination) independent replicates. The empirical distribution of the corresponding statistic was thus generated and used to determine the confidence intervals.
The recombination parameter C (in Drosophila C = 2Nc, where N is the effective population size and c is the recombination rate per generation between the most distant sites) was estimated using the methods of Hudson and Kaplan (1985) and of Hudson (1987). The first method is based on RM or the minimum number of recombination events in the sample; estimates of RM were used to estimate C by coalescent simulations. The Hudson (1987) method is based on the variance of the number of differences between pairs of sequences; in that case, the estimate of C can be obtained numerically. We also estimated the minimum value of C compatible with the observed value of RM (CL); thus, CL is an underestimate of the true C value. The CL value was estimated as the lowest value of C for which the right tail (5%) of the RM distribution contains values equal to or higher than the observed value of RM.
An estimate of C based on the estimates of c (Trueet al. 1996) in the rp49 region, or CM, was also obtained (CM = 49.2). This estimate was obtained following Andolfatto and Przeworski (2000) and considering that (1) the rp49 and the Tpi regions (located in cytological bands 99D and 99E, respectively) have the same recombination rates (i.e., c = 0.92 × 10–8), (2) in D. simulans N = 2 × 106, and (3) the average length of the rp49 region is 1337 bp.
The overall genetic association between polymorphic sites was measured by the ZnS statistic (Kelly 1997), which is the average of r2 (Hill and Robertson 1968) over all pairwise comparisons,
The effect of intragenic recombination on nucleotide variation was studied by analyzing the levels of linkage disequilibrium between polymorphic sites in relation to the physical distance. A new test statistic, ZZ, that is defined as
RESULTS
DNA sequence variation: The rp49 gene plus its 5′ and 3′ flanking regions were sequenced in 24 lines of D. simulans (10 from Europe and 14 from Africa) and in one line of D. melanogaster. In D. simulans, a total of 54 polymorphic nucleotide sites (corresponding to 56 mutations) were identified over the 1292 bp examined (excluding all sites with alignment gaps). Polymorphisms at sites 626 and 713 (exon 2) were synonymous, whereas the rest were in noncoding regions (Figure 1). Eight insertion/deletion polymorphisms (ranging from 1 to 29 bp in length) were also detected in noncoding regions. Estimates of nucleotide variation are shown in Table 1.
Ten of the 24 rp49 sequences surveyed were identical (for both nucleotide and insertion/deletion changes); there was an additional sequence (line SimMz7) that differed from this common haplotype by a single nucleotide substitution (Figure 1). These 11 sequences were designated as L#a. Most other sequences were designated as L#b. Two sequences (SimS13 and SimMz39) probably originated by recombination between the two divergent major haplotypes (L#a and L#b). Lines with the L#a haplotype were found both in Montblanc (six lines) and in Maputo (five lines); the frequency of this haplotype did not differ significantly between both populations (Fisher's exact test, P = 0.41).
Estimates of nucleotide divergence between populations (dxy = 0.0118, and da = 0.000) and FST values (FST = 0.020) were consistent with weak population subdivision. The methods of Roff and Bentzen (1989) and of Hudson et al. (1992) were used for detecting genetic differentiation between populations (considering either all sequences or only L#b sequences). None of the tests performed detected any significant differentiation between populations (results not shown). This lack of genetic differentiation between populations is reflected in the neighbor-joining tree, where European and African sequences are interspersed (Figure 2).
Intragenic recombination and linkage disequilibrium: We tested the effect of intragenic recombination on nucleotide sequence variation (Table 2). The estimated ZZ values were significantly positive, suggesting that in this region intragenic recombination has played an important role in shuffling nucleotide variation among DNA sequences. Estimates of the recombination parameter C obtained by the methods of Hudson and Kaplan (1985) and of Hudson (1987) are shown in Table 2. The discrepancy between both estimates could be due to the particular structure of genetic variation found at the rp49 gene region (see below). This structure would also cause the CL values to be conservative; indeed, larger CL values were obtained when a single L#a sequence was considered in the computer simulations (16.9 and 2.7 for Montblanc and Maputo, respectively).
The significance of the pairwise associations between polymorphic sites, or linkage disequilibrium, was established by the chi-square test. In the total sample, 332 out of 1326 pairwise comparisons showed a significant association; 86 of these comparisons remained significant after applying the Bonferroni procedure. No significant overall association between polymorphic sites was detected by using the ZnS statistic (Kelly 1997) without recombination or introducing conservative recombination estimates (C = CL; results not shown).
Neutrality tests: We tested whether the observed pattern of nucleotide variation is compatible with that expected under neutrality. We applied several tests that compare different estimates of θ either using only intraspecific data (Tajima 1989) or using intraspecific data and sequence information of another species (the outgroup) to determine the polarity of mutations (Fu and Li 1993; Fay and Wu 2000). All these tests failed to reject the neutral equilibrium model (Table 1). The HKA test (Hudsonet al. 1987) was conducted to assess whether levels of polymorphism and divergence were correlated. We compared polymorphism (in D. simulans) and divergence (between D. simulans and D. melanogaster) in the rp49 region (present results) and in the vermilion region in samples from North Carolina and the Congo (Begun and Aquadro 1995). None of the HKA tests showed a significant deviation from neutral predictions (results not shown).
Nucleotide polymorphism and divergence in the rp49 region
—Nucleotide polymorphism at the rp49 gene region in D. simulans. Nucleotide numbering is according to the rp49 sequence of the M66 line of D. melanogaster. Nucleotides identical to the first sequence are indicated by a dot. For length polymorphisms, the nucleotide position refers to the first site affected, and a dash indicates absence of the corresponding length variant. d, deletion; i, insertion; d#, deletion of #bp; i#, insertion of #bp. Polymorphic sites 471–479 correspond to the intron of the rp49 gene (I), and polymorphic sites 626 and 713 to exon 2 (E2) of the rp49 gene. SimS, D. simulans from Montblanc; SimMz, D. simulans from Maputo; MelM66, D. melanogaster line M66. A continuous line indicates a deletion affecting more than one polymorphic site. Shaded blocks indicate L#b sequence information. Rec, recombinant sequence. The last row gives the nucleotide information of the outgroup species D. melanogaster (putative ancestral state) for all polymorphic sites in D. simulans.
We tested by coalescent simulations whether the large number of identical sequences found in the sample was compatible with the neutral equilibrium model (see Hudsonet al. 1994). We also investigated whether the presence of such a major haplotype was compatible with the equilibrium neutral model by analyzing the distribution of the number of haplotypes and of haplotype diversity (Ewens 1972; Strobeck 1987; Depaulis and Veuille 1998). According to the neutral model, both the number of haplotypes (and the number of identical sequences) and the haplotype diversity are a function of the sample size and of θ. The analyses were performed under conservative assumptions (under no recombination and using the conservative CL estimate of the recombination parameter) and also under a more realistic assumption (using the CM estimate that is based on the comparison of the physical and genetic maps). The analyses showed a significant (or nearly significant) excess of identical sequences and a significant (or nearly significant) reduction in the number of haplotypes and in the haplotype diversity values (Table 3). Values of Fu's FS statistic (Fu 1997), which is related to Strobeck's S statistic, pointed in the same direction (results not shown).
DISCUSSION
In D. simulans, as in D. melanogaster, the rp49 gene is located at band 99D where recombination is rather high (Kliman and Hey 1993a; Trueet al. 1996; Andolfatto and Przeworski 2000; this article). The relatively high RM values and the significant ZZ estimates obtained from our data would support the conclusion that recombination in this region is not reduced. Furthermore, D. simulans is monomorphic at the chromosomal level and thus, in this species, no local decrease of recombination is expected as a consequence of chromosomal polymorphism.
Estimates of the recombination parameter
Haplotype distribution tests
—Neighbor-joining tree of the rp49 gene region sequences of D. simulans. Bootstrap values >90% are shown on the tree. D. mel, D. melanogaster (M66 line); Rec, recombinant sequence. The open and solid circles indicate lines from Montblanc and Maputo, respectively. (A) Tree built considering only nucleotide substitutions information. (B) Tree built using information on both nucleotide substitutions and length variants. Each indel was treated as a single mutational event.
The estimated silent nucleotide variation in the rp49 region (θ= 0.016) was lower than estimates for other regions that were also located on the 3R chromosomal arm (the average θ for 19 genes was 0.035; Begun and Whitley 2000). However, most silent variation at the rp49 region corresponds to variation at noncoding sites, while estimates in Begun and Whitley (2000) are based on synonymous variation. Our data conform, therefore, to the general observation of lower variation in noncoding flanking regions than at synonymous sites of coding regions (Moriyama and Powell 1996).
Haplotype substructure and demographic factors: Nucleotide variation at the rp49 region in the two populations of D. simulans stands out because it is highly structured. Both the European and African samples present the same haplotype at intermediate frequency. They also share other minor haplotypes (Figures 1 and 2) and, in fact, no significant genetic differentiation was detected between populations.
In the subsample of lines that constitute the major haplotype in the rp49 region (11 lines designated as L#a), there was a single polymorphism and its rarest variant was present in only one line. Forty-five polymorphisms segregated, however, in L#b lines (n = 11). There were nine fixed differences between L#a and L#b lines (Figure 1). The presence of such a major and divergent haplotype (L#a lines) at the rp49 region is incompatible with the neutral equilibrium model, even in the absence of recombination.
Variation at the vermilion and rp49 regions departed from neutral expectations in a similar way. There are, however, important differences between both sets of results. First, in the vermilion region, only non-African populations showed a significant reduction in haplotype number and/or haplotype diversity. Second, in that region there was no major haplotype common to all populations surveyed. Both observations in the vermilion region are compatible with an important founder effect in the origin of these derived populations. The haplotype structure detected in other regions surveyed in samples from non-African populations might also be the result of founder events (Hamblin and Veuille 1999; Labateet al. 1999; Andolfatto and Kreitman 2000; Duvernell and Eanes 2000). It has been also argued that the haplotype structure detected in most surveys of non-African populations could be the result of population subdivision and recent admixture.
The presence in the rp49 region of the same major haplotype in an ancient and in a recently established population cannot be easily explained by founder events. It could be argued, however, that African populations of D. simulans were genetically differentiated (Hamblin and Veuille 1999) and that the Maputo population was a recently established population. Even in that case, it would be rather unlikely that the same haplotype (haplotype L#a) was present at relatively high frequency in both populations (Montblanc and Maputo). Variation at the Acp26Aa region in lines from the same collections does not show such a haplotype structure (M. Aguadé, unpublished results) suggesting that the observed haplotype structure in the rp49 region is not the result of a genome-wide phenomenon. On the other hand, for the rp49 region only one subset of lines is depleted of variation. If the pattern observed were due to population subdivision and recent admixture, one of the subpopulations should have been nearly monomorphic for this region. This is rather unlikely since the level of silent nucleotide diversity in regions with normal rates of recombination can be quite important, even in species with small effective population sizes such as D. mauritiana and D. madeirensis (e.g., Hey and Kliman 1993; Kliman and Hey 1993b; Khademet al. 2001; see also Gillespie 1999, 2000).
Haplotype structure and selective causes: In Drosophila, only a few surveys of DNA sequence variation in regions with normal (or high) levels of recombination have revealed a high proportion of sequences with zero (or nearly zero) variation in fragments longer than 1.3 kb. The first such pattern was detected in the Sod region of D. melanogaster (Hudson et al. 1994, 1997), and it was considered to reflect the hitchhiking effect of an advantageous mutation that was increasing in frequency. However, unlike in this study, only North American and European populations had been surveyed, which would always leave room for historical explanations. Similarly, a survey of DNA variation at the runt region of D. simulans in lines from different North American populations revealed that most lines (six out of eight) were identical or nearly identical (Labateet al. 1999).
Selection on favorable mutations can remove nucleotide variation at linked sites, causing a selective sweep or hitchhiking effect (Maynard Smith and Haigh 1974; Kaplanet al. 1989; Barton 1998; Fay and Wu 2000; Gillespie 2000; Kim and Stephan 2000). Positive selection has thus been proposed to explain the pattern of nucleotide variation found at the Sod region in two populations of D. melanogaster (Hudson et al. 1994, 1997). The rp49 data of D. simulans exhibits a pattern of nucleotide variation similar to that found for the Sod locus, i.e., two sets of highly diverged sequences in a region with normal levels of recombination. Moreover, we have found the same haplotype structure not only in Europe, but also in a population from the putative ancestral distribution area. It is thus unlikely that the unexpected pattern of variation found at the rp49 region was due to some founder event associated with the colonization of Europe by D. simulans. Therefore, positive selection would most parsimoniously explain the pattern of variation observed at this region. Although we did not detect the excess of high frequency-derived variants expected immediately after a selective sweep (Fay and Wu 2000; Kim and Stephan 2000), eight of the nine mutations fixed in L#a sequences (relative to L#b sequences) were derived (see Figure 1), which would support the selective hypothesis. There are, however, several selective scenarios that could explain the presence of a major haplotype with low variation: (1) the selected haplotype could be in its transient phase either to fixation or to an equilibrium frequency, (2) the pattern could reflect a very recently established balanced polymorphism, or (3) the advantageous mutation could have attained fixation, but the rp49 region could be relatively far away from the advantageous mutation.
Because in D. simulans the rp49 gene is located in a genomic region with high recombination, the fragment affected by the proposed selective sweep should be short (Kaplanet al. 1989; Stephanet al. 1992). In D. melanogaster, the genomic region encompassing the rp49 gene presents a high density of coding regions: 12 coding regions have been identified in a 20-kb fragment spanning the rp49 gene (FlyBase database). The conserved synteny between D. melanogaster and D. simulans allows prediction of a similar density in the latter species and, thus, any of these coding regions (or some regulatory regions) could have been the target of selection.
Time of hitchhiking: The time back to the hitchhiking event can be inferred from the amount of nucleotide variation present in the hitchhiked haplotype. For this inference, we need to know (1) the neutral mutation rate for the rp49 region and (2) the expected topology of the gene genealogy. Assuming that silent substitutions (both at noncoding and synonymous sites) are neutral, the neutral mutation rate for the rp49 region can be estimated from the estimated silent nucleotide divergence between D. simulans and D. melanogaster (K silent = 0.072; Table 1). Assuming that the split of the D. melanogaster and D. simulans lineages occurred 2.5 mya (Lachaiseet al. 1988; Powell 1997), the silent mutation rate would be 1.4 × 10–8 per nucleotide and per year and 1.4 × 10–5 on a per sequence basis (as the rp49 region includes 987 silent sites; Table 1). Several authors have shown (e.g., Slatkin and Hudson 1991) that after a selective sweep the gene genealogy is star-like, i.e., a genealogy compressed at the internal nodes. If mutations are Poisson distributed, the expected number of mutations on the genealogy is μE(T), where T is the total length of the branches in the genealogy (in years), and μ is the mutation rate per sequence per year (Hudson 1990, Equation 1). In the rp49 region, only one mutation (a singleton variant) was detected (site 59 in SimMz7 line) among all L#a sequences (n = 11). Assuming a star genealogy for our sample, T would be 11*t (where t is the time back to the hitchhiking event) and, consequently, t would be ∼6500 years. Thus, the proposed selective sweep would have occurred very recently. The lack of length variation in L#a sequences (Figures 1 and 2) would also support the conclusion that the hitchhiking event was rather recent.
Although hitchhiking would most consistently explain the pattern of variation observed in the rp49 region, we have not definitively ruled out historical explanations. Indeed, the detection of some haplotype structure in other surveyed regions pointed to historical explanations. Only a multilocus approach using large population samples might allow discarding the admixture hypothesis. Also, analysis of variation across contiguous regions of the genome might be used to detect the differential signature of natural selection (Nurminskyet al. 2001). If the haplotype structure detected at the rp49 region were due to hitchhiking, it would decay and eventually disappear at some distance of this region. Analysis of variation in genomic regions located at increasing distances from the rp49 gene would, thus, allow contrasting of the selective hypothesis and, if confirmed, it would also allow delimiting the target of selection.
Acknowledgments
We thank Gema Blasco and David Salguero for technical support and Serveis Científico-Tècnics, Universitat de Barcelona, for automated sequencing facilities. We are especially grateful to A. Barbal and C. Arribas for collecting flies in Maputo, and S. O. Kolokotronis for his collaboration in the project. We also thank C. Segarra for valuable comments on the manuscript. G.B. and M.G. were supported by the Erasmus program between Université Denis Diderot, Paris, France, and Universitat de Barcelona, Spain. This work was supported by grants PB97-0918 from Comisión Interdepartamental de Ciencia y Tecnología, Spain and 1999SGR-25 from Comissió Interdepartamental de Recerca i Innovació Tecnològica, Catalonia, Spain, to M.A.
Footnotes
-
Communicating editor: W. Stephan
- Received January 13, 2001.
- Accepted April 11, 2001.
- Copyright © 2001 by the Genetics Society of America
