RESULTS

Nucleotide polymorphism and recombination: Figure 1 shows a total of 69 polymorphic sites in a sample of 27 sequences of the tin and bap genes. There are five length polymorphisms: four in intron I of tin and one in exon II of bap. Some relevant statistics are given in Table 1. The π value for the full sequence is 0.0056, which is within the range of values observed in highly recombining gene regions in D. melanogaster (Moriyama and Powell 1996; Powell 1997). The silent π value is 0.0063 for tin but significantly higher (by coalescent simulation), 0.0140, for bap. The level of synonymous variation is 0.0052 in the tin coding region but three times higher, 0.0167, in the bap coding region (the difference is highly significant by coalescent simulation even without recombination, P = 0.01). The difference is less pronounced for nonsynonymous variation, which is 0.0011 in the tin gene and two times higher, 0.0021, in the bap gene (P > 0.05). Silent variability is three times higher in exon II (π= 0.0224) than in exon I (π= 0.0073) of bap; the difference is statistically significant by coalescent simulation, even without recombination (P = 0.01). The elevated level of nucleotide variability within the bap exon II cannot be accounted for by relaxation of functional constraints. This is because the bap exon II contains a functionally important homeodomain (183 bp at position 3136–3318, our coordinates), which is conserved within a wide phylogenetic scale (Duboule 1994). There are four polymorphic sites within this bap homeodomain region (Figure 1, positions 3192, 3228, 3246, and 3255), with π= 0.0051, which is slightly lower than that for the whole bap exon II (π= 0.0063).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

—DNA polymorphism in the tin and bap genes of 27 strains of Drosophila melanogaster. The lines are numbered consecutively according to their genetic similarity. The numbers above the top sequence represent the position of segregating sites and the start of a deletion or insertion. Nucleotides are numbered from the beginning of our sequence, position 155034 for the tin and 165505 for the bap gene in Adams et al. (2000). The coding regions of the genes are underlined below the top reference sequence. Amino acid replacement polymorphisms are marked with asterisks. Dots indicate the same nucleotide as the reference sequence. A hyphen represents deleted nucleotides. ▴, a deletion; †,the absence of a deletion. ▴1, a 2-bp deletion of TT (position 252–253); ▴2, a 3-bp deletion of TCC (position 1007–1009); ▴3, a 101-bp deletion of TACCAAAATCTACTAAAT CGAATCAGTGTACATATATGATATAAAACTAATTATTATACCGGTTAACTATCTAAATTGCAATGATTTAAGAAGGTATCTGT (position 1058–1158); ▴4, a 9-bp deletion of TCCTGCTTT (position 1842–1850). ▾, a 12-bp insertion of GAGCGGAGAGCG (position 3716–3727); ‡, the absence of an insertion. The coordinates for the functional regions of the genes are: 1–767 (tin, 5′-flanking region), 768–953 (tin, exon I), 954–1858 (tin, intron I), 1859–2428 (tin, exon II), 2429–2492 (bap,5′-flanking region), 2493–2936 (bap, exon I), 2937–3057 (bap, intron), 3058–3774 (bap, exon II), and 3775–3830 (bap, 3′flanking region). sec, D. sechellia; sim, D. simulans.

The silent variability of bap (π= 0.0140) is close to that of one of the most polymorphic genes of D. melanogaster, Est-6 (π = 0.0156), in the same North American population (Balakirevet al. 2002; Balakirev and Ayala 2003a). The silent variability of bap exon II is very close to that of x03A8;Est-6 (π= 0.0253), which is almost twice as variable as Est-6 (Balakirevet al. 2003; Balakirev and Ayala 2004). The nucleotide variability detected in the tin gene is in the range of values observed in many other regulatory genes evolving under negative selection (Moriyama and Powell 1996; Purugganan 2000; Rileyet al. 2003). The differences in the level of variability between tin and bap could be due to positive selection reflecting different degrees of functional constraint (see below).

The level of divergence, K, between D. melanogaster and D. simulans is similar for tin and bap (Table 1). The K/π ratio is, however, substantially higher for tin in both the coding and noncoding regions: K_cod/π_cod is 10.67 for tin, but 5.29 for bap and K_ncod/π_ncod is 9.28 for tin, but 3.85 for bap. These differences reflect that the pressure of purifying selection is higher for tin than for bap. Silent divergence is very similar in exon I (K = 0.0988) and exon II (K = 0.1012) of bap, despite the fact that the level of silent variability is significantly different in these coding regions, as noted above. The level of divergence between D. melanogaster and D. sechellia is very close to the level of divergence between D. melanogaster and D. simulans (Table 1); the only difference concerns nonsynonymous divergence, which is 1.4 and 1.7 times higher between D. melanogaster and D. sechellia than between D. melanogaster and D. simulans for the tin and bap genes, respectively.

The method of Hudson and Kaplan (1985) reveals a minimum of nine recombination events in the whole region analyzed. The minimum number of recombination events is six for bap but two for tin. There is a large difference (33 times) in the recombination estimator (ρ, McVean et al. 2002) for tin (ρ= 0.0008) and bap (ρ= 0.0264) (Table 2). For the chromosomal region 93DE the laboratory estimate of recombination rate (C_lab; Comeronet al. 1999) based on the physical and genetic maps of D. melanogaster is 0.0744. A substantial difference between the laboratory-based and sequence-based estimates of recombination has been explained by a recent bottleneck (for review, see Wall 2001). The large difference between the sequence-based estimates of recombination rate for the closely linked tin and bap genes (7.1 kb apart) is inconsistent with the demographic explanation of the difference between the laboratory-based and sequence-based estimates of recombination rate (Andolfatto and Przeworski 2000; Wall 2001).

The method of Sawyer (1989, 1999) detects gene conversion events within both tin (P = 0.0097) and bap (P = 0.0029). The number of significant fragments is 7 for tin (fragment length varies from 1149 to 1581 bp, average 1396 bp) but 49 for bap (fragment length varies from 323 to 1135 bp, average 665 bp). Intragenic conversion involves two tin regions (coordinates 751–1899 and 319–1899) in 6 sequences; for bap, intragenic conversion involves eight regions (coordinates 1–763, 1–1135, 250–763, 592–1135, 592–1303, 592–1402, 996–1318, and 996–1402) in 23 sequences. Thus, intragenic gene conversion is more pronounced within bap than within tin. Intergenic gene conversion is not detected, which is likely due to the low nucleotide similarity between tin and bap (42.7%), which is insufficient to satisfy the homology requirements for efficient intergenic conversion. The recombination machinery is sensitive even to a single-nucleotide mismatch; individual nucleotide substitutions have been shown to affect recombination in many organisms (for review, see Balakirev and Ayala 2003c).

Haplotype structure: We have previously described the presence of two sets of haplotypes for the β-esterase gene cluster of D. melanogaster, localized on the left arm of the third chromosome (Balakirev et al. 1999, 2002, 2003). There is also strong haplotype structure for the tin and bap genes (Figures 1, 2, 3). For the tin gene, there are two sets of haplotypes, each completely associated with one of two deletions, 3 bp and 101 bp, within intron I (Figure 1, ▴2 and ▴3) and almost completely associated with the replacement polymorphism at position 1884 (Figure 1). The ▴2 deletion exists also in D. simulans and D. sechellia (Figure 1), which suggests that it is the ancestral condition, whereas the ▴3 deletion has appeared after the split of D. melanogaster from the other two species. Strong haplotype structure is also observed for the bap gene (Figures 1 and 3). There is a set of 13 sequences, 12 of them identical to each other plus one (no. 10) that differs by a nonsynonymous substitution, and a second set of fairly heterogeneous sequences (nos. 14–27). The difference between the two tin haplotype sets (1–18 vs. 19–27) is highly significant (K_st^* = 0.4692; K_st^*_0.999 = 0.1580, P < 0.001 by the permutation test of Hudsonet al. 1992). However, the level of variability is not significantly different between the two tin haplotype sets (π= 0.009 vs. π= 0.0019).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

—Neighbor-joining tree of the tin haplotypes of D. melanogaster, based on Kimura’s two-parameter distance. The numbers at the nodes are bootstrap percentage of probability values based on 10,000 replications.

The homogeneous haplotype set of bap (strains 1–13, Figures 1 and 3) has scant variability (π= 0.0001). The heterogeneous set (strains 14–27) is significantly different from the first set (K_st^* = 0.2852; K_st^*_0.999 = 0.1002, P < 0.001 by the permutation test) and has significantly greater variability (π= 0.0089; P < 0.0001). The distribution of mutations in the bap gene is highly asymmetrical in the two haplotype sets: the heterogeneous set is 74 times more variable than the homogeneous haplotype set.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

—Neighbor-joining tree of the bap haplotypes of D. melanogaster based on Kimura’s two-parameter distance. The numbers at the nodes are bootstrap percentage of probability values based on 10,000 replications.

Sliding-window analysis: Figure 4 shows sliding-window plots of the distribution of nucleotide polymorphism in D. melanogaster (Figure 4A), divergence between this species and D. simulans (Figure 4B), polymorphism-to-divergence ratio (Figure 4C), and linkage disequilibrium (Figure 4D). There are two distinct peaks of nucleotide variability in tin (Figure 4A), in the 5′-flanking region (midpoint coordinates 271–280) and intron (midpoint coordinates 992–1027). These peaks coincide with the peaks of divergence (Figure 4B) and linkage disequilibrium (Figure 4D). The highest peak of the polymorphism-to-divergence ratio within the tin gene, however, does not coincide with the peaks of variability and divergence (it is centered on the tin exon I, Figure 4C). There is also a peak of variability in the bap gene (Figure 4A) in the intron region (midpoint coordinates 2969–3003). This peak is not accompanied by a peak of divergence (Figure 4B); rather, there is an obvious decrease of divergence in the intron region of bap, which produces a peak in the silent divergence-to-polymorphism ratio (Figure 4C, midpoint coordinates 2700–2900). The peak in the bap intron coincides with a peak of linkage disequilibrium (Figure 4D). The highest peak of the silent divergence-to-polymorphism ratio is centered on the bap exon II (Figure 4C, midpoint coordinate 3200). The minimal values of polymorphism-to-divergence ratio within the bap gene are at the beginning of exon I and centered on the two replacement polymorphic sites (Figures 4C and 5, positions 2676 and 2677). The lowest and highest polymorphism-to-divergence ratios are accompanied by the largest maximum and average sliding G values of the McDonald’s (1996, 1998) tests (Figure 6).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

—Sliding-window plots of (A) nucleotide diversity (measured by π), (B) silent nucleotide diversity (thin line) and divergence (K, thick line), (C) polymorphism-to-divergence ratio, and (D) linkage disequilibrium (measured by D) along the tin and bap genes of D. melanogaster. K is the average number of nucleotide substitutions per site between D. melanogaster and D. simulans. Window sizes are 100 nucleotides with 1-nucleotide increments for A, 50 nucleotides with 1-nucleotide increments for B, and 200 nucleotides with 50-nucleotide increments for D. Window size is 10 variable substitutions for the sliding-window plot of polymorphism-to-divergence ratio (C). A schematic of the region investigated is displayed at bottom. Exons are indicated by boxes; the introns and the 5′- and 3′-flanking regions are shown by thin lines.

The variants within the intron sequences of both genes segregate as single-haplotype blocks (Figure 1). Correspondingly there are obvious peaks of nucleotide variability and linkage disequilibrium in the introns (Figure 4, A and D). For the tin gene the intron peak of nucleotide variability also corresponds to the high peak of divergence but for the bap gene there is an opposite tendency: the peak of variability corresponds to the lowest level of the divergence (Figure 4, A and B). A similar tendency is observed for the bap exon II: the level of variability and divergence is close in this region (Figure 4B). Thus for the tin gene the amount of divergence between species is in accordance with the amount of intraspecific polymorphism but for the bap gene the pattern is different: a decrease of the divergence and a parallel increase of polymorphism in the intron and exon II regions that could be explained by the influence of positive selection (see below).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

—Sliding-window plot of synonymous polymorphism-to-divergence ratio in the tin and bap genes of D. melanogaster. A vertical line indicates the separation between tin and bap. D. simulans is used as an outgroup species. Window size is 10 variable substitutions.

We have measured heterogeneity in the distribution of polymorphic sites along the sequence and discordance between the level of within-melanogaster polymorphism and the melanogaster-simulans (or melanogastersechellia) divergence by means of Goss and Lewontin’s (1996) and McDonald’s (1996, 1998) statistics and have assessed their significance by Monte Carlo simulations of the coalescent model incorporating recombination (McDonald 1996, 1998; Table 3). On the basis of 10,000 simulations, with the recombination parameters varying from 1 to 64, the tests are not significant for the tin gene. However, the tests are significant for the bap gene (Table 3). There are two areas within the bap gene with the largest average and maximum sliding G values (Figure 6, A and B). The first (and most pronounced) area is located at the beginning of bap exon I and coincides with a region of low polymorphism-to-divergence ratio (Figure 5). The second area is located in bap exon II and coincides with a region of high polymorphism-to-divergence ratio. The region of low polymorphism-to-divergence ratio is centered on the two replacement substitutions (positions 2676 and 2677). The region of high polymorphism-to-divergence ratio is localized within exon II but is not centered on the replacement polymorphism (Figures 1 and 4C). An area of low polymorphism could result from a selective sweep whereas high polymorphism could result from balancing selection (McDonald 1996, 1998). Previously, we have shown that both types of selection are involved in the evolution of the Est-6 gene of D. melanogaster (Balakirevet al. 2002). We suggest that both types of selection are involved within the bap gene. To examine this suggestion one would analyze separately the polymorphism-to-divergence ratio in the regions with low and high polymorphism (that roughly correspond to exon I and exon II of bap), using the McDonald and Kreitman (1991) and/or the HKA test (Hudsonet al. 1987). However, both tests are hardly applicable in this case because there are only two silent polymorphic sites within exon I of bap. Thus, we cannot contrast the pattern of polymorphism-to-divergence in both exons. However, for bap exon II the HKA test reveals a higher polymorphism-to-divergence ratio in comparison with the complete tin gene (χ² = 4.484, P = 0.034 if D. simulans is used as an outgroup and χ² = 4.813, P = 0.028 if D. sechellia is used as an outgroup), which is in accordance with the possible action of positive selection on this region.

Figure 6.

—Sliding-window plot of the largest average sliding G value (A) and the largest maximum sliding G value (B) for the bap gene (synonymous variability). Window size is 13 variable substitutions for A and 15 for B.

Linkage disequilibrium: For the whole region there are 1378 pairwise comparisons and 514 (37.30%) of them are significant (Figure 7). With the Bonferroni correction, 156 (11.32%) remain significant. There is strong linkage disequilibrium within the tin gene: 83.69% (272 out of 325) pairwise comparisons are significant (55.69%, 181 with the Bonferroni correction). The significant associations are due mostly to polymorphic sites located within the tin noncoding regions (5′-flanking region and intron); only three exonic polymorphic sites (positions 1884, 2245, and 2284) are involved in significant associations (Figure 7). Within the bap gene 48.43% pairwise comparisons (170 out of 351) show statistically significant linkage disequilibrium (9.69%, 34 with the Bonferroni correction). The distribution of significant associations within bap is not homogeneous: more than half of the sites involved in significant associations are located within exon II. Between tin and bap genes only 5.22% (72 out of 1378) of tests are significant (Figure 7, shaded areas); none is significant with the Bonferroni correction.

Within both the tin and bap genes there are strong associations between intronic sites (Figure 7). Clusters of significant linkage disequilibrium that occur predominantly in introns have been repeatedly observed in Drosophila (Miyashita and Langley 1988; Miyashitaet al. 1993; Schaeffer and Miller 1993). Kirby et al. (1995) showed that the linkage disequilibria clustered within the introns of the Adh locus of D. pseudoobscura are caused by epistatic selection maintaining the secondary structures of precursor mRNA. The mechanism underlying the action of epistatic selection is based on a model of compensatory fitness interactions (Kimura 1985), which suggests that mutations occurring in RNA helices are individually deleterious but become neutral in appropriate combinations. Stephan (1996) has shown that the rate of compensatory evolution substantially decays over a distance of 100 nucleotides, which is in agreement with our results. Indeed, the distance between the intronic sites is <100 nucleotides for both tin (positions 975, 982, 1010, 1011, 1016, 1019, 1022, 1037, and 1045) and bap (positions 2954, 2999, 3001, 3005, 3011, and 3019). Thus the evolution of the intronic sequences of the tin and bap genes may be subjected to secondary structure constraints.

Figure 7.

—Fisher’s exact test of nonrandom associations between pairs of tin and bap polymorphisms. Singleton mutations are excluded from the analysis. Each box in the matrix represents the comparison of two polymorphic sites. The location of the segregating sites on the tin and bap genes is shown on the diagonal, which indicates the position of the 5′-flanking, coding, and 3′-flanking regions. The intergenic associations are shaded boxes. The associations that remain significant after Bonferroni correction are solid boxes. ‡, P < 0.001; †, 0.001 < P < 0.01; ○, 0.01 < P < 0.05.

We have analyzed the relationship between linkage disequilibrium (LD; measured as D′) and physical distance between sites by the method of McVean et al. (2002), with the significance of Pearson’s correlation coefficient estimated by 10,000 permutations. There is a significant decline in LD with increasing distance for both tin (Pearson’s correlation coefficient is -0.1432; P = 0.0070) and bap (-0.2189; P = 0.0009).

Tests of neutrality: Kelly’s (1997) Z_nS test (Table 4; based on linkage disequilibrium between segregating sites) detects significant deviations from neutrality for the entire region with recombination 0.0072 (the value of recombination obtained by the method of McVean et al. 2002, Table 2). Wall’s (1999) B and Q tests are significant even without recombination. The areas of significant values of Kelly’s and Wall’s statistics coincide with peaks of linkage disequilibrium within both tin and bap (not shown). For tin the tests are significant with a lower level of recombination than for bap. For instance, the Z_nS statistic obtained for tin is significant (P = 0.01) with recombination rate C = 0.0072, while bap requires C = 0.020 (Table 4). Overall the tests are significant for the entire region as well as for each gene separately, with a recombination rate much lower than the laboratory estimate (C_lab = 0.0744) based on the physical and genetic maps of D. melanogaster (J. M. Comeron, personal communication; Comeronet al. 1999). We suggest that the significance of the tests could reflect the action of selection combined with the demographic history of D. melanogaster, which originated from Africa and migrated in relatively recent times to the rest of the world. The higher test statistics values for tin may reflect the specific character (rare recombination) of the evolution of this gene, as well as the demographic history of D. melanogaster.

There are two sets of divergent haplotypes for the tin and bap genes (Figure 1). It is appropriate to use the haplotype test (Hudsonet al. 1994) in this case to see whether directional selection has increased the frequency of some haplotypes. For tin, there are a total of 37 polymorphic sites and a subset of 18 sequences with 13 sites (Figure 1, strains 1–18). The haplotype test is not significant (P = 0.103) even with a laboratory estimate of recombination equal to 0.0744 (see above). A total of 32 polymorphic sites are in the sample of 27 bap sequences, but the set of homogeneous strains (1–13) includes just one polymorphic site (Figures 1 and 3). The probability of this configuration, obtained by the haplotype test, is 0.002, even without recombination. The region of amino acid substitution between species (Figure 1) at the beginning of bap exon I may be a likely candidate for a selective sweep.

We have also used the neutrality tests of Depaulis and Veuille (1998) to analyze the haplotype distribution. The tests are not significant for the tin gene (as in the Hudsonet al. 1994 and McDonald 1996, 1998 tests). The tests are significant for the bap gene when applied to the homogeneous and heterogeneous sets of sequences separately (the haplotype groups are the same as for the Hudsonet al. 1994 test). Particularly, for the homogeneous set of sequences, the observed haplotype diversity is 0.167, while the expected haplotype diversity is significantly higher (0.640, P < 0.05), which is compatible with the hypothesis of directional selection. For the heterogeneous set of sequences, this test reveals a significant excess of variability: the observed haplotype diversity is 0.952, but the expected haplotype diversity is 0.850 (P < 0.05), which is compatible with the hypothesis of diversifying selection. We have also applied the Depaulis and Veuille (1998) tests for different functional parts of the tin and bap genes. There is no deviation from neutral expectation for any partition (5′-flanking region, intron, and exon II) of the tin gene (exon I is excluded from these separate analyses because it is only 186 bp). For the bap gene, the tests are significant for exon I and exon II separately, but this significance is due to opposite patterns. For exon I, the test reveals a significant excess of different haplotypes: the observed number is 7, but only 4.9 haplotypes are expected (P < 0.05). For exon II, the test reveals a significant deficit of haplotypes: the observed diversity is 0.604, while the expected haplotype diversity is significantly higher (0.820, P < 0.05). These additional tests corroborate our suggestion that the bap gene is under the complex influence of positive selection (see above).