Species-specific positive selection of the male-specific lethal complex that participates in dosage compensation in Drosophila

Abstract

In many taxa, males and females have unequal ratios of sex chromosomes to autosomes, which has resulted in the invention of diverse mechanisms to equilibrate gene expression between the sexes (dosage compensation). Failure to compensate for sex chromosome dosage results in male lethality in Drosophila. In Drosophila, a male-specific lethal (MSL) complex of proteins and noncoding RNAs binds to hundreds of sites on the single male X chromosome and up-regulates gene expression. Here we use population genetics of two closely related Drosophila species to show that adaptive evolution has occurred in all five protein-coding genes of the MSL complex. This positive selection is asymmetric between closely related species, with a very strong signature apparent in Drosophila melanogaster but not in Drosophila simulans. In particular, the MSL1 and MSL2 proteins have undergone dramatic positive selection in D. melanogaster, in domains previously shown to be responsible for their specific targeting to the X chromosome. This signature of positive selection at an essential protein–DNA interface of the complex is unexpected and suggests that X chromosomal MSL-binding DNA segments may themselves be changing rapidly. This highly asymmetric, rapid evolution of the MSL genes further suggests that misregulated dosage compensation may represent one of the underlying causes of male hybrid inviability in Drosophila, wherein the fate of hybrid males depends on which species' X chromosome is inherited.

Chromosomal aneuploidy is highly deleterious; deletions larger than 3% of the genome and duplications larger than 10% are not tolerated in Drosophila (1), presumably because an imbalance of expression levels of many genes is hard to accommodate in stoichiometric complexes involving many different proteins (2). In organisms with highly diverged sex chromosomes, there is frequently a difference in number of sex chromosomes versus autosomes in the heterogametic sex (XY or ZW). This difference requires “dosage compensation” strategies to equilibrate expression levels in both sexes. Recent evidence suggests that these strategies operate at two levels. A primary mechanism is to increase gene expression of the single X chromosome by 2-fold in the heterogametic sex, a strategy that appears to be universally conserved in animals (3, 4). However, different animal lineages have adopted diverse, secondary strategies to equilibrate gene expression in the two sexes (5). In mammals, this secondary modification involves the inactivation of one of the two female X chromosomes, whereas in Caenorhabditis elegans, it is achieved by 2-fold lower transcriptional output from both X chromosomes in hermaphrodites. Flies adopt a different strategy; they double the transcriptional output of the single male X chromosome in somatic cells (6, 7), which requires the targeting of a male-specific lethal complex (MSL) to the X chromosome but not to autosomes in Drosophila males (5).

In Drosophila melanogaster, the MSL complex consists of proteins encoded by five genes: male-specific lethal genes, msl1, msl2, and msl3, maleless (mle) and males absent on the first (mof) (Fig. 1A), as well as two noncoding RNAs (roX1 and roX2). MSL1 and MSL2 play a central role in the assembly of the MSL complex and targeting to the X chromosome (Fig. 1B). It is believed that this targeting of the MSL complex enables MOF to specifically acetylate lysine-16 on histone H4 tails, a histone modification correlated with active transcription (8–10). High-resolution mapping of MSL-binding sites has revealed a strong bias for the middle and 3′ ends of coding sequences, suggesting that any transcriptional up-regulation may involve increased elongation efficiency by RNA polymerase (11–14). Indeed, experiments in Saccharomyces cerevisiae have shown that MOF recruitment (using a Gal4 DNA-binding domain fusion) results in transcriptional up-regulation (5, 10, 13). However, it has been suggested that the absence of the rest of the MSL complex in these experiments complicates the exact implication of this result regarding X chromosome up-regulation in Drosophila (15).

Fig. 1.

MSL1 and MSL2 play a key role in assembly and targeting of the dosage compensation complex to the male X chromosome. (A) Five known protein components and two known RNAs comprising the MSL complex. The five MSL proteins are drawn to scale with known domains highlighted. MSL1 serves as a scaffold for the entire MSL complex. MSL1 binds to MSL2, and together they bind to the X chromosome. Amino acids 85–186 of MSL1 are necessary and sufficient for binding to amino acids 1–190 of MSL2. Together, amino acids 1–265 of MSL1 and amino acids 1–190 of MSL2 are sufficient for targeting to high-affinity binding sites on the male X chromosome (30, 33, 34). Targeting of MSL1 is abolished by deletion of amino acids 1–26 (32). MSL1 also binds to other components of the MSL complex, including MSL3 [which contains a chromobarrel domain (61) that binds RNA (62)] and MOF [which contains a chromobarrel domain (61), a zinc finger, and an acetyltransferase domain with specific activity for histone H4 (10, 63)]. MLE encodes ATPase and RNA/DNA helicase activities (64). (B) Schematic model of the assembly of the MSL complex onto the male X chromosome that highlights the central scaffolding role of MSL1 and MSL2.

Which DNA sequences target the MSL complex specifically to the X chromosome? There are ≈35–40 high-affinity sites on the X chromosome that are bound by this complex (16, 17). In total, there are estimated to be 700 separable regions where the MSL complex is bound as shown by chromatin immunoprecipitation experiments. These regions cover roughly 25% of the X chromosome and, presumably, include sites ranging in affinity (11, 12, 18). These regions occur mostly in coding sequences of genes (transposable elements were not included in the arrays used in the analyses) and may be enriched in GAGA motifs (11, 12, 19). There is some debate about the role of active transcription in attracting or maintaining the MSL complex (for review, see ref. 13), but active transcription alone cannot explain the strong bias for binding to X chromosomal DNA. Computational sequence analysis can identify some “higher-order” features on X chromosomal sequences that distinguish them from autosomal DNA, yet extensive efforts at identifying common sequence predictors of MSL-binding sites have yielded limited prediction power at best (12, 20). These findings have led to the suggestion that degenerate and multiple weak signals may contribute to targeting (12, 19, 21). Arguably, comparative genomics methodology has limited applicability to this problem because MSL-binding sites have not been mapped in divergent Drosophila species or even methodically in different D. melanogaster strains. In addition, genetic experiments have determined that any substantial segment of the X chromosome sequence appears to possess the ability to attract the MSL complex autonomously (22, 23). Although autosomal genes inserted onto the X chromosome will also frequently undergo dosage compensation, X chromosomal sequences are superior in their ability to recruit the MSL complex compared with autosomal sequences despite varying in their affinity for MSL recruitment (24). The questions remain: What is recruiting the MSL complex specifically to the male X, and why are these sequences so difficult to define?

We hypothesized that a MSL-binding site consensus is hard to define because these motifs might be evolutionarily labile. A selective pressure that prevented the stable coevolution of MSL proteins and DNA might have prevented the fixation of an optimal DNA sequence that could recruit the MSL complex. Such a scenario would preclude the identification of a consensus MSL-binding signature in the DNA. We explored this possibility by investigating the selective pressures shaping genes encoding MSL proteins as a “surrogate” to studying the MSL-binding sites themselves directly. We found strong evidence of positive selection acting on all five genes encoding protein components of the MSL complex. This finding is highly unexpected because MSL function is essential for male viability. We further found that the signature for rapid evolution is strikingly asymmetric, affecting D. melanogaster but (largely) not Drosophila simulans. Some of the strongest signatures of positive selection can be localized to the MSL domains responsible for X chromosomal targeting, suggesting that MSL-recruiting DNA segments may also have evolved rapidly in the D. melanogaster lineage, where all of the mapping studies have been done. Together, these findings also suggest that incompatibilities between MSL proteins and MSL-recruiting DNA elements on the D. melanogaster X chromosome may represent one of the underlying causes of male hybrid inviability in Drosophila.

Results

We sequenced all five protein-coding MSL complex genes from multiple strains of D. melanogaster and D. simulans, two species that diverged 2.5 million years ago. Summary statistics for polymorphisms seen in these genes are presented in Table 1. From these statistics, there is no evidence for a pattern of rare, singleton polymorphisms that might suggest recovery of polymorphisms after a recent adaptive sweep in most of the MSL genes. However, we see a Tajima's D value of −1.55 and a Fu and Li value of −2.58 (P < 0.05) in the mle gene of D. melanogaster, strongly supporting the possibility that a recent sweep has affected the polymorphism spectrum (Table 1) (25, 26). To investigate this possibility further, we compared all of the MSL genes in D. melanogaster by using a Hudson–Kreitman–Aguade (HKA) test, which examines whether interspecies divergence and intraspecies polymorphisms are correlated, as would be predicted under neutrality (27). We find a significant discordance in the polymorphism patterns between mle and three other MSL genes, msl2, msl3, and mof (Table 2). Thus, both the Fu and Li F* statistic and the HKA test results strongly implicate a recent adaptive sweep in the mle gene in D. melanogaster. None of the MSL genes shows a significantly discordant polymorphism pattern by the HKA test in D. simulans.

Table 1.

Summary statistics for polymorphisms in MSL complex genes

Gene	Location	Codon usage (ENC)	Nucleotide diversity	Tajima's D	Fu and Li F*	No. of strains	No. of bp
D. melanogaster
msl1	2L (36F11–37A1)	56.8	0.00279	−0.262	−0.819	14	3,180
msl2	2L (23F3)	51.8	0.00961	−0.215	−0.426†	11	2,374
msl3	3L (65E4)	52	0.00879	−0.0993	0.435	10	1,890
mof	X (5C5)	48	0.00468	−0.114	0.033	13	2,484
mle	2R (42A6)	53.1	0.00206	−1.55	−2.58‡	14	5,447
D. simulans
msl1		56.3	0.00691	−0.227	−0.658	14	3,195
msl2		50.7	0.00939	−0.0745	−0.340†	15	2,350§
msl3		52.2	0.01091	−0.0734	0.168	12	1,881
mof		47.1	0.00669	−0.549	0.0958	19	2,484
mle		53.5	0.00896	0.236	−0.0576	7	5,164

The Fu and Li F statistic was calculated by using either D. simulans or D. melanogaster as an outgroup species.

‡, P < 0.05.

^†Seventy-five nucleotides were excluded from analysis because of one region between D. melanogaster and D. simulans with ambiguity in the alignment.

^§Eighteen codons were excluded from analysis due to two polymorphic indels of 12 and 42 nucleotides within D. simulans.

Table 2.

HKA tests on the MSL complex genes in D. melanogaster

Gene	Intraspecific			Interspecies, total no. of differences*	msl2	msl3	mle	mof
	No. of segregating sites	Total no. of sites	Sample size†
msl1	29	3,168	14	162.63	0.117	0.029	0.742	0.117
msl2	61	2,245	11	150.51		0.560	0.034	0.982
msl3	48	1,880	10	86.14			0.003	0.545
mle	46	4,743	16	293.65				0.020
mof	37	2,484	13	112.36

P values for all pairwise comparisons between the D. melanogaster MSL complex genes are shown with significant deviations from neutral expectations highlighted in bold.

*D. simulans was used for the interspecies comparison.

^†No. of D. melanogaster strains sequenced.

We also tested for positive selection by using the McDonald–Kreitman (MK) test (Table 3). This test evaluates whether an excess number of replacement (amino acid altering) changes versus synonymous changes had been fixed between the two species compared with replacement and synonymous polymorphisms within each species (28). Under this test, we find that four of five MSL genes, msl1, msl2, msl3, mof, but not mle, show robust signatures of positive selection when compared across the whole gene. This finding is highly unusual for essential genes because a high degree of evolutionary constraint is expected to act to preserve function. Using Drosophila yakuba as an outgroup species, we can ascertain which lineage has been affected by positive selection by assigning the fixed changes to either the D. melanogaster or D. simulans lineages. We find that there is robust evidence for positive selection acting on the D. melanogaster lineage for four of five MSL genes (all except mle) by using the MK test. Interestingly, mle is the only gene that appears to have undergone a recent sweep in the D. melanogaster species (Tables 1 and 2) and has likely reduced our ability to detect positive selection by using the MK test on this gene. In contrast, we find that only one of five genes (mof) has been subjected to positive selection along the D. simulans lineage; even in this case, the intensity of positive selection is weaker than in the D. melanogaster lineage. Also, of the five MSL genes analyzed here, mof is the only one that is also expressed robustly in females (8), although the functional significance of MOF function in females is still unclear.

Table 3.

MK test for positive selection on MSL complex genes

MSL complex protein-encoding gene	Observed Sp	Observed Rp	Observed Sf	Observed Rf	G value	P value
msl1
Pooled	54	41	56	78	5.025	0.025
D. melanogaster	18	10	24	53	9.097	0.0026
D. simulans	36	32	23	22	0.036	0.85
msl2*
Pooled	83	40	51	63	12.47	0.00041
D. melanogaster	36	22	24	41	7.743	0.0054
D. simulans	47	18	22	18	3.201	0.074
msl3
Pooled	49	10	22	19	9.913	0.00164
D. melanogaster	24	3	11	15	13.23	0.00028
D. simulans	24	6	7	3	0.379	0.54
mof
Pooled	84	19	48	37	13.97	0.00019
D. melanogaster	30	7	25	23	7.86	0.0051
D. simulans	55	12	19	13	5.531	0.012
mle
Pooled	50	35	75	48	0.096	0.76
D. melanogaster	14	16	37	24	1.562	0.21
D. simulans	36	19	29	21	0.608	0.44

We compared the ratio of replacement and synonymous changes that were polymorphic within the species (Rp:Sp) with the ratio of replacement and synonymous changes that were found fixed between the species (Rf:Sf). If no alteration in selective regimes occurred during evolution, we expect these ratios to be statistically indistinguishable. However, an excess of fixed replacement changes is a clear indication of positive selection. Pooled changes refer to all the polymorphism and fixed changes using both lineages. However, by using an outgroup species (D. yakuba), we can also make the same comparison, specific either to the D. melanogaster or D. simulans lineage.

*A region of 75 nucleotides was excluded from all msl2 analysis because of ambiguity in the alignment between D. simulans and D. melanogaster(the region corresponds to nucleotides 904–978 relative to D. melanogaster).

The msl1 and msl2 genes are key players in targeting the MSL complex to the male X chromosome. Mutational analyses of each MSL gene have shown that MSL1 and MSL2 are capable of targeting “high-affinity” sites, independent of other known MSL components (29–31). Targeting requires an interaction between the N-terminal domains of MSL1 and MSL2 (Fig. 1) and is abolished by deletion of the first 26 amino acids of MSL1 (30, 32–34). Because these targeting domains have been roughly mapped, we next addressed whether positive selection had shaped these regions in particular, focusing our analyses only on the D. melanogaster lineage. If we parse the fixed or polymorphic changes that have taken place in MSL1 in the D. melanogaster lineage, we find that the N-terminal domain (amino acids 1–265), which is necessary and sufficient for both X chromosomal targeting and for interactions with MSL2 (30, 32–34), bears all of the hallmarks of positive selection (Rf:Sf::Rp:Sp = 30:6::5:8, G value 8.406, P < 0.005). In contrast, the remainder of the MSL1 protein [amino acids 266-1039, which includes interaction interfaces with both MOF and MSL3 (33)] shows no evidence for positive selection (Rf:Sf::Rp:Sp = 23:18::5:10, G value 2.229, P > 0.15).

Similarly, when we parse our fixed and polymorphic changes for the MSL2 protein, we find that the domain required for binding MSL1 and thereby, targeting to DNA (amino acids 1–190) (34) evolves under positive selection in D. melanogaster (Rf:Sf::Rp:Sp = 9:9::1:10, P < 0.025). In contrast to MSL1, however, even the remainder of the MSL2 protein, which consists of a central coiled-coil domain and a C-terminus of as yet-undefined function (Fig. 1), shows a robust signature of positive selection (Rf:Sf::Rp:Sp = 32:15::21:26, G value 5.203, P < 0.025).

Thus, the N-terminal domains of both MSL1 and MSL2 are hot spots for positive selection, despite the fact that these domains are essential for the MSL1–MSL2 interaction both with each other and to binding sites on the X chromosome. Our findings support the idea that the rules that guide X chromosomal sequence-directed DNA binding have been evolutionarily labile, at least in the D. melanogaster lineage.

Discussion

Rapid evolution of the MSL complex is at odds with the expectation that proteins so essential for male viability ought to be highly constrained and under purifying selection. Moreover, interactions between MSL proteins and their cognate DNA-binding sites should be especially well constrained because any mutations in one MSL component would have to be accommodated in the other MSL proteins and in the DNA target sites to retain the essential function of the complex. Indeed, it is reasonable to speculate that the selective force that drove the rapid evolution must have imposed a stringent selective cost, which would drive changes in the whole MSL complex.

Male-killing bacteria provide an example of just such a selective cost. For instance, Spiroplasma poulsonii specifically kill male D. melanogaster flies, as they are transmitted exclusively through females. Recent studies have directly implicated the presence of a functional MSL complex as a requirement for this male-specific killing by S. poulsonii (35). Under such a “genetic conflict” scenario, one could imagine bacterial proteins evolving to “detect” MSL components through direct binding, whereas MSL components could be under strong selective pressure to evolve away from this recognition. This “arms race” would result in changes in one or all of the MSL components because fixation of slightly deleterious mutations in the MSL complex would be preferred over bacteria-induced male lethality.

A second possible source driving positive selection of the MSL complex could be genetic conflict with retrotransposable elements. It has been suggested previously that LINE1 non-LTR retrotransposons may provide “landing sites” for dosage compensation (X inactivation) in mammalian X chromosomes (36). Under this second possibility, MSL binding to retrotransposons may be an important defense against them (37). Repeated specialization of the MSL complex to recognize retroelements may also result in some of these elements becoming preferred landing sites for the MSL complex, effectively altering the landscape of MSL binding to the X chromosome. It is important to note, however, that retrotransposons primarily mobilize in the germ line, whereas the MSL proteins discussed here are acting predominantly in somatic tissues.

Both of these conflict scenarios fit well with our finding of highly asymmetric positive selection because either the male-killing bacteria or retrotransposons may provide a lineage-specific selective pressure, not affecting even closely related species. It has been suggested (38) that because the relative stoichiometries of regulator proteins are so intricately linked to each other, rapid evolution of any one component driven by genetic conflict under any model [by male-killing bacteria or retrotransposons or even by “centromere drive” (39)] could inevitably trigger a “ripple effect of adaptation” in other MSL genes. Each such alteration would trigger a coevolutionary episode in which other target genes and regulator proteins would adjust to the changed landscape to ensure optimal function (38). Although it is unlikely that a single ripple event can explain the pervasive positive selection we have seen in multiple domains of all MSL proteins, selection on one member of a complex might bring along changes in other members at any number of domains if that produces an eventual fitness advantage by restoring optimal function.

One possible consequence of such rapid evolution is that MSL components may quickly become incompatible in different species. Such incompatibilities are thought to occur under a Dobzhansky–Muller (D–M) model wherein independently occurring allelic changes in different interacting components could manifest as negative epistasis in resulting hybrids (the simplest two-locus form is schematized in Fig. 2A). The possibility of this negative epistasis is greatly increased with accelerated change; genes responsible for postzygotic isolation are frequently subject to positive selection (40–42). Rapid evolution can also result in asymmetric postzygotic isolation (Fig. 2B). There are two formal possibilities for D–M incompatibilities arising from the positive selection of the genes encoding MSL1 and MSL2. The first possibility is that the two components could represent the MSL1 and MSL2 proteins themselves (Fig. 2C) because it is the protein–protein interaction surface between these two proteins that is a hot spot for positive selection. However, recognizing that these domains also determine the exclusive targeting to the X chromosome, the D–M incompatibility could exist between the MSL1–MSL2 proteins, and MSL-targeting sites on the X chromosome (Fig. 2D). Either of these incompatibilities would lead to compromised MSL function and thereby male inviability in interspecies hybrids.

Fig. 2.

Positive selection of the MSL complex and X chromosomal MSL-binding sites might result in hybrid incompatibility. (A) Two-locus D–M model for hybrid incompatibility between closely related species. Loci A and B interact in the ancestral species. During (reproductive or recombinational) isolation, there is a neutral fixation of the a and b alleles in the two populations, which is tolerated because the new alleles (a and b) are still compatible with the old alleles (B and A, respectively). However, this fixation results in hybrid incompatibility because of negative epistatic interactions between the new a and b alleles. This model can explain the onset of incompatibilities even under neutral evolution (56). (B) In the case of positive selection (bold arrows) driving the interaction of the A and B loci, only one lineage may evolve to the new a and b alleles, resulting in incompatibility with the other lineage, which still preserves the ancestral A and B alleles. (C) MSL1 and MSL2 could represent the A and B loci in the D–M model, with the positive selection (bold arrows) at their interaction interface resulting in hybrid inviability. Under this model, male hybrids containing either the D. melanogaster or D. simulans X chromosomes would be inviable because the protein composition is expected to be the same in both cases. (D) Model for hybrid incompatibility with MSL1–MSL2 and the X chromosomal MSL-binding sites, as A and B loci, respectively. Positive selection (bold arrows) in D. melanogaster has resulted in rapid evolution of the MSL1–MSL2 genes and (we infer) the X chromosomal MSL-binding sites. In male hybrids, D. simulans MSL1 and MSL2 are unable to recognize “newly evolved” MSL-binding sites on the D. melanogaster X chromosome resulting in mislocalization of the MSL complex in hybrids with a D. melanogaster X chromosome (44). However, hybrids with a D. simulans X chromosome localize the MSL complex normally (45) because the D. melanogaster MSL1 and MSL2 proteins retain an ancestral DNA-binding ability. (E) Known male hybrid incompatibility in D. melanogaster crosses to D. simulans. Male inviability occurs when a D. melanogaster X chromosome is combined with a hybrid autosomal background (43).

Intriguingly, male hybrids have different outcomes in a cross between D. melanogaster and D. simulans, depending on which X chromosome is inherited in the hybrid males (Fig. 2E). Hybrid males that inherit the D. simulans X chromosome are viable. However, hybrid males that inherit the D. melanogaster X chromosome suffer larval lethality, dying at a developmental stage similar to that of pure-species D. melanogaster males that have mutated MSL components (43, 44). Because the msl1 and msl2 genes reside on autosomes (Table 1), both hybrids should acquire both D. simulans and D. melanogaster versions of both genes. The viability of the D. simulans X-bearing hybrid males implies that an incompatibility between MSL1 and MSL2 proteins (Fig. 2B) is not likely to be causal for hybrid male inviability. Instead, our findings suggest that although the MSL1–MSL2 interaction is not severely affected in hybrids, hybrid inviability may result from negative epistasis between the D. melanogaster X chromosome and D. simulans MSL components (Fig. 2C). Because most of the MSL positive selection has occurred along the D. melanogaster lineage, D. simulans MSL proteins may not have the requisite changes for the correct targeting to the D. melanogaster X. In support of this idea, a recent study has found that MSL components do not correctly target in D. melanogaster X-bearing hybrid males (44). In contrast, MSL localization and function are known to be normal in D. simulans X-bearing hybrid males (44, 45).

The genetic dissection of the determinants of postzygotic isolation in Drosophila has been greatly aided by the discovery of hybrid rescue genes, so called because mutations in these genes restore hybrid viability. (MSL components are not expected to be hybrid rescue genes because compromised MSL function would be strongly deleterious.) Hybrid males that inherit a D. melanogaster X chromosome and are otherwise inviable, can be rescued by a naturally occurring mutation in the Lhr (lethal hybrid rescue) gene (43). A recent study firmly establishes that the Lhr gene has a heterochromatic localization (41). This study is especially noteworthy because several heterochromatin proteins and remodeling factors have been directly implicated in transcriptional regulation of the male X chromosome (46–49). Indeed, it is possible that other hybrid incompatibility factors mapped in this cross may arise from defects in dosage compensation. For instance, the intriguing finding that nuclear pore complex proteins cause hybrid male lethality (40) could be viewed in light of recent findings connecting them to the MSL proteins (50). It is likely that a balance of “negative” dosage regulators and “positive” MSL proteins is required to achieve the correct level of transcription in Drosophila hybrid males (51). Such nonadditive expression phenotypes have been observed in hybrids of D. melanogaster and D. simulans (52) with an apparent overabundance of misregulated genes on the X chromosome (53).

The generality of Haldane's rule, wherein it is more likely that the heterogametic sex will be inviable, has led to several attractive theories about how hybrid inviability could represent a breakdown in dosage compensation in hybrids (54, 55). Indeed, it has been clear for quite some time that the X chromosome plays a disproportionate role in hybrid incompatibilities, referred to as the “large-X” effect (56). Until recently, it has not been clear that D–M incompatibilities could arise in such systems because they are so essential for function and thereby predicted to evolve under a high degree of constraint. However, our present analysis on MSL complex genes suggests that such genes can and do evolve rapidly, which implies that even genes that participate in essential chromatin functions such as dosage compensation (as described here), chromosome segregation (57), and defining origins of DNA replication (58) are not immune from being called to participate in genetic conflict and adaptation. Indeed, D–M incompatibilities arising because of rapid evolution of these essential protein–DNA interactions are more likely to result in hybrid inviability and sterility rather than incompatibilities between two proteins that carry out a nonessential role in either species.

Methods

All Drosophila strains were obtained from the Species Stock Center (Tucson, AZ) except for the African isofemale lines that were a gift from Daven Presgraves (University of Rochester, Rochester, NY). Genomic DNA was prepared as described previously (57). Genes were amplified by using PCR Supermix High Fidelity (Invitrogen, Carlsbad, CA) and primers based on D. melanogaster genomic sequence. Most PCR products were sequenced directly except in the case of mle for some D. simulans strains. When direct sequencing of PCR products was not possible because of a low yield of PCR products, these products were cloned by using Topo-TA vectors (Invitrogen), and sequencing was done on at least three separate colonies. ClustalX (59) was used to obtain multiple alignments, which were subsequently hand-edited with the amino acid sequence as a guide. The DNASP software package (60) was used to perform several tests for positive selection, including the Tajima's D (25) and Fu and Li tests (26), as well as the MK (28) and the HKA (27) tests.

Abbreviations

D–M model

Dobzhansky–Muller model

HKA test

Hudson–Kreitman–Aguade test

MK test

McDonald–Kreitman test

MLE

maleless

MOF

males absent on the first

MSL

male-specific lethal.