Excess of Nonsynonymous Polymorphism at Acph-1 in Different Gene Arrangements of Drosophila subobscura Free

Abstract

Nucleotide variation in the Acph-1 gene region was analyzed in a natural population of Drosophila subobscura from Bizerte (Tunisia). The lines studied differed in their gene arrangement for segment I of the O chromosome: 21 lines were O₃₊₄₊₈, 21 were O₃₊₄₊₂₃, and 3 were O₃₊₄. According to chromosomal phylogenies, O₃₊₄ is a central arrangement from which O₃₊₄₊₈ and O₃₊₄₊₂₃ originated. Strong genetic differentiation at Acph-1 was detected among the different arrangements, which is reflected in strong linkage disequilibrium between the variants at informative polymorphic sites and the type of arrangement. Estimates of silent nucleotide diversity are slightly lower within O₃₊₄₊₂₃ (π_silent = 0.0166) than within O₃₊₄₊₈ (π_silent = 0.0228) or O₃₊₄ (π_silent = 0.0234). In contrast, nonsynonymous nucleotide diversity estimates (around 0.1%) are similar in the three arrangements. Most nonsynonymous rare variants are singletons, which results in highly significant Tajima's neutrality tests within the young O₃₊₄₊₈ and O₃₊₄₊₂₃ arrangements. This test is not significant for nonsynonymous mutations within a large Spanish O₃₊₄ sample. In addition, a significant and marginally significant excess of nonsynonymous polymorphism was detected by the McDonald and Kreitman test within O₃₊₄₊₂₃ and O₃₊₄₊₈, respectively. This excess results in a rather high neutrality index (NI = 5.25) when both arrangements are jointly analyzed, in contrast to its value within the old O₃₊₄ arrangement (NI = 1.74). The pattern of variation at Acph-1 within the young arrangements is unusual for nuclear genes and has the same characteristics previously detected in most genes of the mitochondrial genome. Assuming that most nonsynonymous mutations at Acph-1 are under weak negative selection, a smaller effective size of the young arrangements relative to O₃₊₄ might explain the observed results. The relatively low frequency of O₃₊₄₊₈ and O₃₊₄₊₂₃ in the distribution area of D. subobscura, the more recent origin of these arrangements relative to O₃₊₄ and the suppression of recombination in heterokaryotypes might contribute to the relatively small effective size of the young arrangements. Therefore, present results indicate that the differences in effective size and recombination caused by chromosomal arrangements are modulating nonsynonymous variation at Acph-1.

Introduction

Deviations from the neutral model of molecular evolution (Kimura 1983) caused by a significant excess of nonsynonymous polymorphism have been extensively reported in the mitochondrial genome (mtDNA). Several organisms such as Drosophila, mice, marsh rats, and human present this excess of amino acid variation in mtDNA genes (Ballard and Kreitman 1994; Nachman, Boyer, and Aquadro 1994; Rand, Dorfsman, and Kann 1994; Nachman et al. 1996; Rand and Kann 1996; Hasegawa, Cao, and Yang 1998; Kennedy and Nachman 1998; Wise, Sraml, and Easteal 1998). Different selective factors have been proposed to explain this excess of nonsynonymous polymorphism. First, balancing selection maintaining protein variation or, alternatively, positive selection fluctuating over time or space might contribute to an excess of amino acid variation (Gillespie 1991). Second, mild negative selection acting on replacement mutations can cause these mutations to be maintained in the population at low frequency, but rarely to become fixed (Ohta 1992). Third, a relaxation of selection acting in the past might maintain previously nontolerated nonsynonymous polymorphism (Takahata 1993).

In contrast to mtDNA genes, only a few nuclear Drosophila genes exhibit a significant excess of nonsynonymous polymorphism. In Drosophila melanogaster, this excess has been detected at Gld (Hamblin and Aquadro 1997), Pgi (McDonald and Kreitman unpublished, cited in Moriyama and Powell 1996), Pgm (Verrelli and Eanes 2000), and bcd (Baines et al. 2002). Indeed, an excess of nonsynonymous fixed differences is a much more common result in this species (Moriyama and Powell 1996; Fay, Wyckoff, and Wu 2002). The paucity of nuclear genes showing an excess of nonsynonymous polymorphism led Weinreich and Rand (2000) to suggest that the lack of recombination in the mtDNA and the distribution of the selective effects of newly arisen mutations might explain the commonly detected excess of nonsynonymous polymorphism in mitochondrial genes.

Although most studies of nucleotide variation in Drosophila nuclear genes correspond to species of the melanogaster group, some surveys have been carried out in species of the obscura group. Only the analysis of nucleotide variation in the Acph-1 gene of D. subobscura has revealed an excess of nonsynonymous polymorphism (Navarro-Sabaté, Aguadé, and Segarra 1999a). In this case, however, the region under study maps at section 91C, very close to one of the breakpoints of the O₃ inversion. As a consequence, variation in the Acph-1 gene was analyzed separately for lines differing by their chromosomal arrangement for segment I (sections 91–99) of the O chromosome: O_st and O₃₊₄. The excess of nonsynonymous polymorphism was significant in the O_st arrangement, but not in O₃₊₄, which suggested an effect of chromosomal inversion polymorphism on amino acid variation, at least at Acph-1. In addition, assuming that most nonsynonymous mutations at Acph-1 are under weak negative selection, the contrasting results between O₃₊₄ and O_st are consistent with a much smaller effective size of O_st relative to O₃₊₄ (Navarro-Sabaté, Aguadé, and Segarra 1999a).

Chromosomal phylogeny in D. subobscura (Krimbas 1992) indicates that O₃₊₄ is a central arrangement from which several arrangements, such as O₃₊₄₊₈ and O₃₊₄₊₂₃, originated, giving rise to the O_[3+4] inversion complex. Therefore, it can be inferred unambiguously that these two arrangements are younger than O₃₊₄. The presence of such a complex system of gene arrangements in segment I of the O chromosome of D. subobscura offers the opportunity to further analyze the putative effect of chromosomal inversions on nonsynonymous variation. Indeed, differences in the effective size of these arrangements are expected as a result of differences in their frequency, and even in their age. The effective size of young arrangements might be strongly affected by the extreme bottleneck produced in their origin, which should cause a reduction in the long-term effective size of the arrangement.

Here, we have studied nucleotide polymorphism at Acph-1 in the young O₃₊₄₊₈ and O₃₊₄₊₂₃ arrangements. The quite low average frequency of these two arrangements in the distribution area of D. subobscura and their derived character suggest that their effective sizes should be lower than the effective size of the old O₃₊₄ arrangement. These differences in the effective size and the well-established role of inversions as suppressors of recombination might affect nonsynonymous variation if nonsynonymous mutations are under weak selection, as shown in the case of O_st. The results obtained support this prediction. A significant and marginally significant excess of nonsynonymous polymorphism is detected by the McDonald and Kreitman test (McDonald and Kreitman 1991) in the young arrangements O₃₊₄₊₂₃ and O₃₊₄₊₈, respectively. In addition, highly significant Tajima's D statistics (Tajima 1989) for nonsynonymous mutations were detected within both gene arrangements. Therefore, variation at Acph-1 within the young arrangements O₃₊₄₊₂₃ and O₃₊₄₊₈ presents the same pattern previously detected for most mitochondrial genes of D. melanogaster. According to these results, it can be inferred that differences in effective size and recombination caused by chromosomal arrangements are modulating nonsynonymous variation in a nuclear gene.

Materials and Methods

Fly Samples

Forty-five isochromosomal lines of D. subobscura were established from flies collected at Bizerte (Tunisia) in 1996. Most of these lines had been previously analyzed for variation at the rp49 gene region (Rozas et al. 1999). Lines differed in their arrangement for segment I of the O chromosome: 21 lines were O₃₊₄₊₈, 21 were O₃₊₄₊₂₃, and 3 were O₃₊₄. All O₃₊₄₊₂₃ lines presented an additional inversion (inversion 6) in segment II; therefore, their complete gene arrangement for the O chromosome was O_3+4+23+6.

In Situ Hybridization

Polytene chromosomes slides were obtained after dissection of the salivary glands from third-instar larvae heterozygous for different arrangements of segment I: O₃₊₄/O_st, O₃₊₄/O₃₊₄₊₂₃, O₃₊₄/O₃₊₄₊₈, and O₃₊₄₊₂₃/O₃₊₄₊₈. The DNA of the recombinant phage λDsubAcph-1.1 (Navarro-Sabaté, Aguadé, and Segarra 1999b) was labeled with biotin-16-dUTP and used as probe. Prehybridization, hybridization, and detection conditions were as described in Segarra and Aguadé (1992).

DNA Sequencing

Genomic DNA of the isochromosomal lines was purified according to Ashburner (1989). The complete Acph-1 gene region, approximately 2.2 kb long, was amplified by polymerase chain reaction (PCR). Amplification primers and PCR conditions were as described in Navarro-Sabaté, Aguadé, and Segarra (1999a). The PCR products were purified with QIAquick columns (Qiagen) and used as templates for sequencing with primers designed at approximately every 300 nucleotides. Both DNA strands were completely sequenced. The cycle sequencing kit from PerkinElmer was used to perform the sequencing reactions according to the manufacturer's instructions. The reaction products were run on an Abi Prism 377 automated DNA sequencer.

Complete sequences of the 45 Bizerte lines were multiply aligned manually with the SeqApp (version 1. 9) program (Gilbert 1993), and the sequence of line B2ST (Navarro-Sabaté, Aguadé, and Segarra 1999b) was used as the reference sequence. A multiple alignment including the 53 Spanish lines previously analyzed for the Acph-1 gene region (Navarro-Sabaté, Aguadé, and Segarra 1999a; European Molecular Biology Laboratory (EMBL) accession numbers AJ389424 to AJ389476) was also obtained. Alignments were edited with the MacClade program (Maddison and Maddison 1992) prior to analysis.

Data Analysis

Data analysis was performed with the DnaSP, version 3.4, program (Rozas and Rozas 1999). Nucleotide polymorphism was estimated as: S (the number of segregating sites), k (the average number of pairwise nucleotide differences), π (nucleotide diversity; Nei 1987) and 𝛉 (the heterozygosity per site expected in a population at mutation-drift equilibrium given the observed S value; Watterson 1975). Genetic differentiation between arrangements or between populations was estimated by the average number of nucleotide differences per site (d_xy; Nei 1987). The

statistic as proposed by Hudson, Boos, and Kaplan (1992) was used to determine whether arrangements or populations were significantly differentiated. The number of shared polymorphisms between arrangements (sites with the same nucleotide variants segregating in two different arrangements) expected by recurrent mutation was estimated assuming a hypergeometric distribution of mutations. Gene conversion tracts between arrangements were identified according to Betrán et al. (1997). Linkage disequilibrium between pairs of parsimonious informative sites (sites where the less frequent variant is present at least twice in the sample) was analyzed by the χ² test. The percentage of significant comparisons and the average r² (Hill and Robertson 1968) between pairs of informative sites were used as measures of linkage disequilibrium. The recombination parameter C (C = 2N_ec in Drosophila, where N_e is the effective size and c is the recombination rate between adjacent sites) was inferred according to Hudson (1987), and the minimum number of recombination events was inferred according to Hudson and Kaplan (1985).

Different neutrality tests (Tajima 1989; McDonald and Kreitman 1991) were performed to determine whether the level and pattern of nucleotide variation detected within gene arrangement conformed to neutral model expectations. Statistical significance of the Tajima's D observed values was obtained by coalescent simulation with no recombination (Hudson 1990) after 10,000 replicates. The probability for each test statistic was obtained as the frequency of replicates with a value lower than the observed statistic (one-tailed test).

The gene genealogy of the Acph-1 sequences was reconstructed by the Neighbor-Joining method (Saitou and Nei 1987) implemented in the Mega program, version 2.1 (Kumar et al. 2001). Genetic distances between sequences were estimated according to the Kimura two-parameter model (Kimura 1980) to correct for multiple hits. Sites with polymorphic indels were excluded from the analysis (complete deletion option). Bootstrap percentages for the different nodes were obtained after 500 replicates.

The sequences reported here have been deposited in the EMBL database library under accession numbers AJ563964 to AJ564008.

Results

In Situ Hybridization

The location of the Acph-1 gene in the inversion loop of the heterokaryotypes for the arrangements studied is shown in figure 1. In most nuclei of O₃₊₄/O_st heterokaryotypes, the hybridization signal appears in the asynaptic region of one of the inversion loops; this is in agreement with the location of Acph-1 inside and very close to one of the breakpoints of inversion 3. In O₃₊₄/O₃₊₄₊₂₃ and O₃₊₄₊₂₃/O₃₊₄₊₈ heterokaryotypes, the hybridization signal is detected inside the inversion loop and close to one of its ends, but in most nuclei the signal appears in the synaptic region of the loop. Finally, Acph-1 maps in a rather central position of the inversion loop present in O₃₊₄/O₃₊₄₊₈ heterokaryotypes.

Nucleotide Variation

The multiple alignment of the 45 sequences from the Bizerte sample had 2,169 sites. Nine polymorphic indels with one, two, or seven nucleotides were detected in noncoding regions. These polymorphic indels presented two or three variants segregating in the data set. All polymorphisms with three variants affected the length of runs of a single nucleotide. In addition, line TB21 presented in the third exon of the coding region a 16-nt long deletion that causes a frameshift mutation. After excluding all sites with alignment gaps, a total of 184 nucleotide polymorphisms were detected among the 2,135 sites of the data set (fig. 2). Three different nucleotides were segregating in six sites; thus the minimum number of mutations was 190.

Estimates of nucleotide polymorphism within each chromosomal arrangement are presented in table 1. These estimates were obtained independently for noncoding sites (flanking regions plus introns), silent sites (noncoding sites plus synonymous sites in the coding region), and synonymous and nonsynonymous sites of the coding region. Line TB21 with a frameshift mutation in the coding region was excluded from this analysis, because, for this line, the classification of sites in these different categories is ambiguous. For the O₃₊₄₊₈ lines from Bizerte, estimated nucleotide diversity was similar (both for all sites and for silent sites) to that present within the O₃₊₄ arrangement either from Bizerte or from Spain (π = 0.0123 and π_silent = 0.0219; Navarro-Sabaté, Aguadé, and Segarra 1999a). In contrast, variation within O₃₊₄₊₂₃ is slightly lower when all sites, silent sites, or only synonymous sites in the coding region are considered. In contrast, nucleotide diversity at nonsynonymous sites is similar in the three gene arrangements.

Linkage disequilibrium analysis between pairs of informative sites within each arrangement was also performed. In O₃₊₄₊₈, 162 of 1,431 comparisons (11%) were significant (P < 0.05) by the χ² test, but only 2% of the comparisons were significant when the Bonferroni procedure (Weir 1996) to correct for multiple tests was applied. In O₃₊₄₊₂₃ the extent of linkage disequilibrium was similar: the percentages of significant comparisons were 12% (99 of 820 comparisons), and only 1% with the Bonferroni correction. The lack of strong linkage disequilibrium is consistent with the rather high recombination detected within gene arrangements. Estimates of the recombination parameter (Hudson 1987) were C = 0.092 within O₃₊₄₊₈ and C = 0.029 within O₃₊₄₊₂₃. In contrast, the minimum number of recombination events (Hudson and Kaplan 1985) detected in the data set was 19 within O₃₊₄₊₈ and 14 within O₃₊₄₊₂₃.

Genetic Differentiation Between Arrangements

The d_xy and

estimates in the three pairwise comparisons between the O₃₊₄, O₃₊₄₊₈ and O₃₊₄₊₂₃ arrangements are shown in table 2. The three arrangements are significantly differentiated, although no fixed differences between them were detected. Otherwise, the O₃₊₄ lines from Bizerte are not genetically differentiated from the Spanish O₃₊₄ lines (P = 0.434). The number of silent polymorphisms shared between arrangements in Bizerte is much higher than that expected by recurrent mutation (table 2).

Genetic exchange between arrangements can explain the relatively high number of shared polymorphisms. Despite the lack of fixed differences and the presence of shared polymorphisms, genetic differentiation between arrangements prevails because of the relatively high number of almost fixed differences between them. For instance, in 17 of the 49 polymorphisms shared between O₃₊₄₊₈ and O₃₊₄₊₂₃, the most frequent variant differs between the two arrangements. Genetic differentiation between arrangements may also be detected by the analysis of linkage disequilibrium between nucleotide polymorphic sites and the type of chromosomal arrangement. Indeed, the presence of fixed or almost fixed variants between arrangements at a particular site should cause a strong association between the variants at this site and the type of chromosomal arrangement. This analysis was performed for all informative polymorphic sites and the chromosomal arrangement, in a sample including the O₃₊₄₊₂₃ and O₃₊₄₊₈ lines studied. The extent of linkage disequilibrium was estimated by the percentage of comparisons that were significant after applying the χ² test and by the average r² value. Despite the lack of fixed differences and the presence of shared polymorphisms between O₃₊₄₊₂₃ and O₃₊₄₊₈ (likely due to gene conversion), 28 of 69 comparisons (about 41%) were significant when all 21 lines of each arrangement O₃₊₄₊₂₃ and O₃₊₄₊₈ were considered (table 3). This rather high percentage contrasts with that around 10% detected within arrangement, which reflects that recombination is restricted in heterokaryotypes. Therefore, the maintenance of both chromosomal arrangements in the population contributes to an overall reduction of recombination in the Acph-1 gene region, as shown by the high percentage of significant associations between the alternative variants of the informative sites and the chromosomal arrangement.

The analysis of linkage disequilibrium between nucleotide polymorphic sites and gene arrangements was extended to the O_st and O₃₊₄ arrangements from the Spanish population (Navarro-Sabaté, Aguadé, and Segarra 1999a). As stated above, the rationale for this analysis is that high levels of linkage disequilibrium in joint samples of two arrangements (relative to linkage disequilibrium within each arrangement) indicate a strong genetic differentiation between them and thus a strong reduction of recombination in heterokaryotypes. For the results to be comparable, the analysis was conducted with samples where the two arrangements considered were equifrequent. The highest genetic differentiation was found in the samples including the O_st lines and the lowest differentiation in the sample with the O₃₊₄ and O₃₊₄₊₈ lines (table 3).

The extent of genetic differentiation between arrangements is also reflected in the gene genealogy reconstructed from variation at Acph-1. The D. subobscura lines group in two clear clusters (fig. 3). One cluster includes all O_st lines, and the other groups all lines of the O_[3+4] inversion complex. Bootstrap percentages supporting the O_st and O_[3+4] clusters are 51 and 70, respectively. Within the O_[3+4] cluster, some clustering of the lines according to their gene arrangement is observed, although the bootstrap support is in all cases very low. Thus, 19 of the 21 O₃₊₄₊₂₃ lines and 13 of the 21 O₃₊₄₊₈ lines group in unique clusters. In contrast, O₃₊₄ lines group mainly in two clusters: one that includes only O₃₊₄ lines (11 lines), and another that includes 17 O₃₊₄, 4 O₃₊₄₊₈ and 1 O₃₊₄₊₂₃ lines.

Replacement Polymorphism

The ACPH-1 preprotein has 447 residues in all lines except TB21. The frameshift mutation present in this line results in a truncated preprotein with 130 residues; thus this line can be considered a null allele for Acph-1. When this line is aligned in frame with the other lines, a single amino acid replacement (site 1537 in fig. 1) is detected, and no additional changes that disrupt the coding region (as nonsense mutations) are present. After excluding this null allele from the analysis, a total of 22 polymorphic residues were detected in the Bizerte sample: 2 within O₃₊₄, 11 within O₃₊₄₊₂₃, and 10 within O₃₊₄₊₈ (fig. 4). These amino acid polymorphisms result in three different protein haplotypes within O₃₊₄, 10 within O₃₊₄₊₂₃, and 10 within O₃₊₄₊₈. The G/V (glycine/valine) polymorphism at residue 387 is shared by the O₃₊₄₊₈ and O₃₊₄₊₂₃ arrangements. However, the presence of a V in TB167 (O₃₊₄₊₂₃) can be explained by a gene conversion tract detected in this line. In contrast, three different variants were detected in this residue within O₃₊₄₊₂₃. The rarest variant in all the polymorphic amino acid residues, except in residue 387, was present only once in the sample. According to the sequence of D. guanche (Navarro-Sabaté, Aguadé, and Segarra 1999b), it can be inferred that all nonsynonymous singleton variants are due to derived mutations.

Only three of the amino acid polymorphisms detected in Bizerte (at residues 205, 241, and 387) were previously detected in the Spanish sample (Navarro-Sabaté, Aguadé, and Segarra 1999a). Among them, the R/S (arginine/serine) polymorphism at residue 205 and the N/K (asparagine/lysine) polymorphism at residue 241 were identified as those responsible for the change in mobility between the Acph-1¹⁰⁰ and Acph-1^>100, and between the Acph-1¹⁰⁰ and Acph-1⁰⁵⁴ electromorphs, respectively, within O₃₊₄.

Neutrality Tests

The Tajima test (Tajima 1989) was used to contrast whether the pattern of variation within each gene arrangement is consistent with predictions of the neutral model. This test was performed independently for silent and for nonsynonymous mutations (table 4). The low number of O₃₊₄ lines from Bizerte prevented performance of this analysis within this chromosomal arrangement in this population. Thus, neutrality tests for this gene arrangement were performed with the Spanish random sample (Navarro-Sabaté, Aguadé, and Segarra 1999a). Even under the conservative assumption of no recombination, the test for nonsynonymous mutations is highly significant within the young arrangements: O₃₊₄₊₂₃ and O₃₊₄₊₈. In contrast, the test is not significant for nonsynonymous mutations within O₃₊₄, the oldest gene arrangement. None of the tests performed in the different arrangements was significant for silent mutations. The negative value of the significant Tajima's D statistic indicated a significant excess of low frequency (mostly singletons) nonsynonymous mutations in the data set. Although Tajima's D was also negative for silent sites within O₃₊₄₊₂₃ and O₃₊₄₊₈, the relative excess of low frequency silent mutations was not significant in these arrangements. In addition, the number of singleton and non-singleton polymorphisms differs significantly for nonsynonymous and synonymous mutations both within O₃₊₄₊₂₃ (χ² = 11.27, 1 df, P = 0.0008) and O₃₊₄₊₈ (χ² = 5.37, 1 df, P = 0.0205). The same test is not significant within O₃₊₄ (χ² = 0.13, 1 df, P = 0.7156). These results clearly indicate a different pattern of variation for silent and nonsynonymous mutations in the young arrangements (O₃₊₄₊₂₃ and O₃₊₄₊₈).

The McDonald and Kreitman (1991) test of neutrality (MK test) was also performed within each gene arrangement using the sequence of D. guanche (Navarro-Sabaté, Aguadé, and Segarra 1999b) in the interspecific comparison (table 5). This test is based on the comparison of the ratio of nonsynonymous to synonymous polymorphisms (P_nonsyn/P_syn) and the ratio of nonsynonymous to synonymous fixed differences between species (F_nonsyn/F_syn). Both ratios are expected to be equal under strict neutrality and putative deviations from this expectation can be determined by a G-test of independence. The test was significant (P = 0.0374) within O₃₊₄₊₂₃ and marginally significant (P = 0.0647) within O₃₊₄₊₈. In fact, the P_nonsyn/P_syn ratio within gene arrangements (0.28 and 0.24 within O₃₊₄₊₈ and O₃₊₄₊₂₃, respectively) was much higher than the F_nonsyn/F_syn ratio (0.06 in both comparisons). This deviation is also reflected in the neutrality index (NI; Rand and Kann 1996) estimated within O₃₊₄₊₈ and O₃₊₄₊₂₃ (table 5). This index is expected to be 1 under strict neutrality, and values higher than 1 indicate that the P_nonsyn/F_nonsyn ratio is higher than the P_syn/F_syn ratio.

The MK tests indicated a stronger deviation from predictions of the neutral model within the young arrangements (O₃₊₄₊₈ and O₃₊₄₊₂₃) than within O₃₊₄. In addition, neutrality indexes much higher than 1 were found within O₃₊₄₊₈ and O₃₊₄₊₂₃, which indicates that the deviation from neutrality is in the same direction in both young arrangements—i.e., an excess of nonsynonymous polymorphism (or an excess of synonymous fixed differences). As a consequence, the MK test is highly significant when all the lines with the young arrangements (O₃₊₄₊₈ and O₃₊₄₊₂₃) are jointly analyzed (table 5). In addition, the neutrality index equal to 5.25 for the young arrangements is clearly higher than that detected in the oldest arrangement O₃₊₄ (NI = 1.74). A G test of independence revealed a marginally significant difference between the numbers of synonymous and nonsynonymous polymorphisms in the young arrangements (O₃₊₄₊₈ and O₃₊₄₊₂₃) and in O₃₊₄ (G = 2.74, 1 df, P = 0.098). The same test revealed a significant difference between O_st (NI = 9.01) and O₃₊₄ (Navarro-Sabaté, Aguadé, and Segarra 1999a).

Discussion

Nucleotide Variation Within and Between Arrangements

Nucleotide variation at Acph-1 has been analyzed in four chromosomal arrangements segregating in natural populations of D. subobscura: O₃₊₄, O_st, O₃₊₄₊₈, and O₃₊₄₊₂₃. Because of the mutational process and the suppression of recombination in inversion heterokaryotypes, some genetic differentiation between arrangements is expected. However, differentiation may be counterbalanced by genetic exchange between arrangements. It has been proposed that the level of genetic exchange should be higher for markers located in central positions of the inversion loop than for markers located near the breakpoints (Rozas and Aguadé 1994; Navarro et al. 1997; Navarro, Barbadilla, and Ruiz 2000; Andolfatto, Depaulis, and Navarro 2001). Indeed, near breakpoints, genetic exchange should be possible only by gene conversion, whereas in the central region of the inversion loop not only gene conversion but also double crossover can contribute to the genetic exchange. The genetic exchange is considered to be eventually null at the inversion breakpoints themselves where chromosomes cannot establish synapsis properly (Wesley and Eanes 1994; Andolfatto, Wall, and Kreitman 1999; Cáceres et al. 1999; Cáceres, Puig, and Ruiz 2001).

Empirical data obtained for markers located at different positions relative to the breakpoints of, for instance, the In(3L)Payne inversion of D. melanogaster support the proposed differential genetic exchange along the inversion (Hasson and Eanes 1996). Acph-1 is located inside and rather close to one of the ends of the inversion loop formed in O₃₊₄₊₂₃/O₃₊₄ and O₃₊₄₊₂₃/O₃₊₄₊₈ heterokaryotypes. This is not the case for the O₃₊₄/O₃₊₄₊₈ heterokaryotypes, in which Acph-1 maps in a central position of the inversion loop (fig. 1). Consequently, genetic differentiation at Acph-1 is expected to be higher between O₃₊₄₊₂₃ relative to both O₃₊₄ and O₃₊₄₊₈ than between the two latter arrangements. Although the d_xy estimates are similar in the three comparisons between these arrangements, the data on linkage disequilibrium between nucleotide polymorphic sites and the type of arrangement (table 3) are consistent with this expectation. Linkage disequilibrium in the O₃₊₄ and O₃₊₄₊₈ joint sample (and thus genetic differentiation between O₃₊₄ and O₃₊₄₊₈) is lower than that detected in the O₃₊₄ and O₃₊₄₊₂₃ or in the O₃₊₄₊₂₃ and O₃₊₄₊₈ joint samples. In addition, the highest level of linkage disequilibrium was found in the samples including the O_st lines. This result was expected because Acph-1 maps very close to one of the breakpoints of inversion 3 and would thus be asynaptic in heterokaryotypes between O_st and any of the O_[3+4] arrangements (fig. 1 and results not shown). Therefore, the level of genetic exchange between O_st and the other chromosomal arrangements is expected to be considerably reduced at Acph-1.

Prevalence of genetic differentiation despite the presence of genetic exchange between arrangements is also reflected in the gene genealogy inferred from nucleotide variation at Acph-1. The clustering of lines into two main groups (O_st and O_[3+4]) is consistent with restricted genetic exchange between these chromosomal classes. In contrast, O_[3+4] lines do not completely cluster according to their gene arrangement, which is consistent with a higher level of genetic exchange within O_[3+4] than between O_st and O_[3+4]. However, genetic exchange within O_[3+4] has not been high enough to erase completely the original chromosomal relationships among lines, as reflected in the partial clustering of lines by gene arrangement. A similar result was reported earlier, according to variation at the rp49 gene region (Rozas et al. 1999) that is tightly linked to Acph-1.

Excess of Nonsynonymous Polymorphism

Under predictions of the neutral model, the ratio between the number of polymorphic variants within species and the number of fixed differences between species is expected to be similar for synonymous and nonsynonymous mutations. This prediction results in a neutrality index equal to 1. Neutrality indexes considerably higher than 1 have been detected in O₃₊₄₊₂₃ and O₃₊₄₊₈. Indeed, this index is 5.25 when both young arrangements are jointly analyzed. This deviation from strict neutrality is statistically significant, as revealed by the MK test (table 5). Also, a significant MK test associated with a neutrality index higher than 1 (NI = 9.01) was previously detected in the O_st arrangement (Navarro-Sabaté, Aguadé, and Segarra 1999a). In contrast, the MK test is not significant for the old O₃₊₄ arrangement. Neutrality indexes higher than 1 have been interpreted mainly as resulting from an excess of nonsynonymous polymorphism. Alternatively, they could be due to an excess of fixed synonymous differences, but this latter explanation should hold for all chromosomal arrangements and thus should not cause differences between arrangements. The contrasting results obtained by the MK test in O₃₊₄ versus O₃₊₄₊₂₃, O₃₊₄₊₈, and O_st indicate that different constraints act on nonsynonymous mutations in the two sets of arrangements.

The different constraints might be a priori attributable to differences in the strength of positive selection favoring or maintaining amino acid variation in the studied arrangements. In fact, an excess of replacement polymorphism is expected from positive selection fluctuating over time or space (Gillespie 1991). The excess of nonsynonymous polymorphism detected at Pgm in D. melanogaster (Verrelli and Eanes 2000) has been attributed to this kind of selection. Also, positive selection was proposed to explain the different ratio of replacement to silent polymorphism detected at fused-1 in standard X chromosomes and in chromosomes carrying a centric fusion between the X chromosome and the fourth chromosome of D. americana (Vieira, McAllister, and Charlesworth 2001). It does not seem likely, however, that positive selection can explain the results observed at Acph-1. All amino acid polymorphisms within O₃₊₄₊₂₃ and all but one within O₃₊₄₊₈ are singletons, which results in highly significant Tajima's D statistics for nonsynonymous variation (table 4). In addition, a significant negative Tajima's D was detected within O_st (Navarro-Sabaté, Aguadé, and Segarra 1999a and present data). If positive selection were maintaining nonsynonymous variation at Acph-1, then a higher proportion of nonsynonymous polymorphisms segregating at intermediate frequencies would be expected.

The excess of nonsynonymous polymorphism detected at Acph-1 can be more likely explained according to the nearly neutral model of molecular evolution (Ohta 1992), and assuming that most replacement mutations at Acph-1 are slightly deleterious. Apart from the variants responsible for the polymorphism at residues 205 and 241 of the ACPH-1 protein, all singleton nonsynonymous variants detected within O₃₊₄₊₈ and O₃₊₄₊₂₃ are derived and absent from the other arrangements. Therefore, it can be inferred that these variants appeared by mutation and were kept at low frequency within each arrangement by weak negative selection; i.e. nonsynonymous mutations are slightly deleterious. The highly significant negative Tajima's D statistic values for nonsynonymous mutations also support this argument. Tajima's D statistic is, however, highly sensitive to demographic factors such as population expansions and population subdivision. The presence of the different chromosomal arrangements clearly causes genetic substructuring at Acph-1, which in turn may affect Tajima's D statistic. Nevertheless, demographic factors are expected to affect different regions of the genome or different kinds of sites to a similar extent. The different pattern of variation detected for nonsynonymous and silent mutations is not consistent with demographic explanations.

The pattern of variation in the young O₃₊₄₊₂₃ and O₃₊₄₊₈ arrangements, and even in O_st, has the same signature as previously reported for most genes of the mitochondrial genome (mtDNA) in D. melanogaster and other organisms. This signature has been mainly explained according to the nearly neutral model and assuming that nonsynonymous mutations are weakly selected. Although theoretical models based on weak selection have recently been developed (McVean and Charlesworth 2000; Comeron and Kreitman 2002), the study of the joint effects of selection, drift, and recombination on weakly selected mutations requires further attention. However, as discussed by Weinreich and Rand (2000), two characteristics of the mitochondrial genome that may contribute to the detected pattern of variation are the lack of recombination and its lower effective size relative to the nuclear genome. Therefore, it is tempting to argue that these characteristics may also contribute to the pattern of variation detected at Acph-1 in the young O₃₊₄₊₂₃ and O₃₊₄₊₈ arrangements, and even in O_st. The lack of recombination in mtDNA might be related to the reduction in recombination at Acph-1 in heterokaryotypes. In addition, a reduction in recombination can be viewed as a reduction in effective size as a consequence of the Hill-Robertson effect (Hill and Robertson 1966). In contrast, the effective size of these arrangements might be low given their frequency and their young character. In any case, a small effective size results in a longer persistence of slightly deleterious mutations in the population (i.e., negative selection against slightly deleterious mutations is relaxed).

The frequencies of O₃₊₄₊₈ and O₃₊₄₊₂₃ in Bizerte are 0.58 and 0.31, respectively (Rozas et al. 1999). These frequencies are in the upper limit of the frequencies of these arrangements in the Palearctic region. O₃₊₄₊₂₃ has a very localized distribution and it is present at appreciable frequencies only around Tunisia and in England. O₃₊₄₊₈ has a wider distribution, but its frequency rarely is larger than 0.1 except in the north of Africa, where it reaches the highest frequency in Tunisia. Therefore, it can be inferred that the average frequencies of these arrangements in the whole distribution area of D. subobscura are lower than the frequencies in Bizerte, with the lowest average frequency for O₃₊₄₊₂₃. The frequency of O₃₊₄ is 0.79 in the Spanish sample, but only 0.11 in Bizerte (Rozas et al. 1999). The lack of genetic differentiation between both populations indicates that migration is high enough to homogenize the gene content within arrangement (see also Rozas et al. 1995). O₃₊₄ is an arrangement that reaches its highest frequencies (up to 0.9) in Iran and the Canary Islands (Krimbas 1992). Only in Tunisia and the north of Europe is its frequency lower than 0.1. In northwest Africa and the Mediterranean, frequencies from 0.2 to 0.7 are common. Therefore, the average frequency of O₃₊₄ in the distribution area of D. subobscura is expected to be slightly but not much lower than its frequency in Spain. The different average frequencies of these arrangements in the distribution area of D. subobscura allows us to predict that the effective size of O₃₊₄ is considerable higher than effective sizes of O₃₊₄₊₈ and O₃₊₄₊₂₃. In addition, these are the youngest arrangements, which may be why their long-term effective size was smaller than that of O₃₊₄. It has to be also considered that recombination at Acph-1 is only likely in homokaryotypes, which are expected to be rare for low-frequency arrangements. In this sense, the threefold difference in the recombination parameter between O₃₊₄₊₈ and O₃₊₄₊₂₃ (see Results) would be consistent with the expected frequencies of the O₃₊₄₊₈ and O₃₊₄₊₂₃ homokaryotypes in Bizerte (0.34 and 0.1, respectively).

All arrangements except O₃₊₄ exhibit an excess of nonsynonymous polymorphism at Acph-1. This excess should be also expected in other genes associated with the different arrangements. Previous results at rp49 do not support this expectation, because in this gene no nonsynonymous polymorphism was detected in any of the studied gene arrangements (Rozas et al. 1999). However, it has to be considered that the distribution of the selective coefficients acting against nonsynonymous mutations may differ among genes depending on the functional constraints of the encoded proteins. Constraints on the ACPH-1 protein would likely be relatively low, as indicated by the rather high nonsynonymous divergence at Acph-1 (Navarro-Sabaté, Aguadé, and Segarra 1999b) and the detection of one line (TB21) with a null allele. The low functional constraints on the ACPH-1 protein could have greatly contributed to the detected results. The effect of chromosomal polymorphism on nonsynonymous variation would thus vary with the level of constraints of the encoded protein, and such an effect might be negligible in genes subject to strong purifying selection, like rp49.

We thank J. Rozas for critical comments on a previous version of the manuscript, as well as D. M. Rand and the anonymous reviewers for their constructive comments. We also thank Serveis Científico-Tècnics, Universitat de Barcelona, for automated sequencing facilities. This work was supported by a predoctoral fellowship from Comissió Interdepartamental de Recerca i Innovació Tecnològica, Catalonia, Spain to À.N.-S. and by grants PB94-923 and PB97-918 from Comisión Interdepartamental de Ciencia y Tecnología, Spain, and 1995SGR-577, 1997SGR-59, and 1999SGR-25 from Comissió Interdepartamental de Recerca i Innovació Tecnològica, Catalonia, Spain, to M.A.

Literature Cited