Sex-Dependent Gene Expression and Evolution of the Drosophila Transcriptome

See allHide authors and affiliations

Science  13 Jun 2003:
Vol. 300, Issue 5626, pp. 1742-1745
DOI: 10.1126/science.1085881


Comparison of the gene-expression profiles between adults of Drosophila melanogaster and Drosophila simulans has uncovered the evolution of genes that exhibit sex-dependent regulation. Approximately half the genes showed differences in expression between the species, and among these, ∼83% involved a gain, loss, increase, decrease, or reversal of sex-biased expression. Most of the interspecific differences in messenger RNA abundance affect male-biased genes. Genes that differ in expression between the species showed functional clustering only if they were sex-biased. Our results suggest that sex-dependent selection may drive changes in expression of many of the most rapidly evolving genes in the Drosophila transcriptome.

Sexual dimorphism is ubiquitous among higher eukaryotes. Differential selection pressure between the sexes has been postulated to explain the substantial between-sex differences observed in morphology, physiology, and behavior, indicating the existence of different optimal sex-dependent phenotypes (1). Studies of gene expression during the life cycle of Drosophila melanogaster have found that, for sexually mature males and females, a substantial fraction of the Drosophila transcriptome displays sex-dependent regulation (24). Increasing evidence suggests that molecular mechanisms associated with sex and reproduction change substantially faster between species than those more narrowly restricted to survival (5, 6). New data also suggest that some of the interspecific changes that are driven by differential selection between the sexes have a regulatory origin (7, 8). However, the evolutionary pattern of differences in gene expression between the sexes on a genomic scale is presently unknown.

We performed competitive hybridizations with cDNA microarrays (fig. S1) (9) to identify genome-wide regulatory differences in sex-biased genes between D. melanogaster and D. simulans. These morphologically nearly identical species belong to the melanogaster subgroup of the subgenus Sophophora and diverged ∼2.5 million years ago (10, 11). Our results are based on the 30 hybridizations outlined in fig. S1, which were performed with either cDNA to assay differences in transcript abundance or else genomic DNA as controls (9). The microarrays contained 4776 coding sequences amplified from cDNA clones (9). The hybridizations with genomic DNA were performed to detect coding sequences whose apparent transcript abundance might be affected by sequence divergence or by changes in gene-copy number. The species differ in an estimated 3.8% of nucleotides at the DNA sequence level (12) and in copy number of some transposable elements (13) and a few multicopy genes (14). Across the six interspecific DNA hybridizations, genomic DNA from D. melanogaster showed an average of 4.2% greater hybridization than genomic DNA from D. simulans, in good agreement with the estimated sequence divergence. The distribution of hybridization intensities across coding sequences was essentially gaussian (15) with only a few outliers identified, mostly as transposable elements such as the retrotransposon springer or multicopy genes such as Stellate (14, 16). Apart from these exceptional sequences, the differences in genomic hybridization are well within the limit of detection of significant differences in gene expression with our level of replication. Accordingly, no correction for sequence divergence between the species was required for the estimates of transcript abundance.

The cDNA hybridization data were analyzed by a Bayesian method (17) that yielded an estimated mean and 95% credible interval of the relative level of expression of each gene in each sex of each species (table S1). Genes were classified as differentially expressed between sexes within species or for the same sex between species if their 95% credible intervals failed to overlap (Fig. 1). The main categories into which the 4776 coding sequences were classified are shown in Table 1. Comparison with the reported pattern of expression in D. melanogaster was used to validate our classification (9). Random permutations of the data provided an estimated false-positive rate of 0.03%; hence, no adjustment was made for multiple tests.

Fig. 1.

Illustrative examples of gene classification according to the relative levels of expression for the four sex-by-species combinations compared. (A) qtc, a male-biased gene expressed to the same extent in both species. (B) PGRP-SC1b, a non–sex-biased gene with greater expression in D. simulans. (C) BcDNA:LD09936, a male-biased gene expressed with a greater bias in D. simulans. (D) CG12200, a female-biased gene overexpressed in D. melanogaster relative to D. simulans. The 95% credible intervals associated with the estimated mean expression levels that were obtained by a Bayesian approach (17) are compared between sexes within D. melanogaster and D. simulans (below the graphs) and for the same sex between species (to the right). Inequality signs denote statistically significant differences. Expression levels appear as a fold-change relative to the node(s), with the lowest level of expression set to 1 (17). m, D. melanogaster; s, D. simulans.

Table 1.

Classification of genes by their pattern of evolution and sex bias. A Bayesian method was used to determine the 95% credible intervals of the mean expression level for all genes in the hybridizations in fig. S1 (17). Those intervals were then compared between sexes within species and for the same sex between species. A difference in gene expression was regarded as significant if the 95% credible intervals failed to overlap. In Bayesian analysis, the 95% credible interval is the analog of the 95% confidence interval in conventional frequentist statistics. Sixty-seven different categories of gene expression are possible among the four comparisons (fig. S2 and table S1).

Class Number % of total
By pattern of evolution
No difference in expression in either sex between species and no difference in sex bias between species 2493 52.2
Sex-independent difference in expression between species 380 8.0
Increase or decrease in sex-biased expression between species 952 19.9
Gain, loss, or reversal of sex-biased expression between species 951 19.9
Total 4776
By sex bias
Total female-enriched 2031 42.5
    Female-biased in both species 1507 31.6
    Female-biased in D. melanogaster only 327 6.8
    Female-biased in D. simulans only 197 4.1
Total male-enriched 1318 27.6
    Male-biased in both species 911 19.1
    Male-biased in D. melanogaster only 174 3.6
    Male-biased in D. simulans only 233 4.9

More than half of the genes show a sex bias in expression; many of these genes are known or expected to be expressed in reproductive organs (3, 4). Among the 2493 genes that show no detectable change in level of expression since the divergence of D. melanogaster and D. simulans, 57.5% show sex-biased expression (Fig. 1A). This is a somewhat greater fraction of sex-biased genes than reported previously (2, 3), which we attribute largely to our level of replication.

Approximately half of the genes (2283 out of 4776) have evolved a difference in the level of expression between the species. This is far greater than the number of genes that show significant differences in expression among strains within a species (2). Indeed, pairwise comparisons among eight strains of D. melanogaster analyzed with the same microarrays and methods reported here yielded an average of only 677 genes with significant differential expression (18).

Among the genes with differential expression between D. melanogaster and D. simulans, only a small number (380 out of 2283, or 16.6%) show parallel differences in expression in both sexes (Fig. 1B). Instead, the majority of genes (1903 out of 2283, or 83.4%) exhibit an evolutionary pattern that is sex-specific. In particular, 952 genes retain the same sex bias in D. melanogaster and D. simulans but have evolved different levels of expression between males or females (Fig. 1C). For the remaining 951 genes, the evolutionary change entails the gain, loss, or reversal of sex-biased expression (Fig. 1D). Among the 20 genes that show reversal of sex bias, six illustrate the most extreme discordance, in which the sex that shows the greatest level of expression in one species shows the smallest level of expression in the other.

For genes that show a significant difference in expression between the sexes in both species, male-biased and female-biased genes show different patterns in relation to the magnitude of the bias in expression (table S2). For sex-biased genes with mean expression between the sexes differing by a factor ≤2, no significant sex bias in gene expression was detected [G test, adjusted G value (Gadj) = 2.4, df = 1, not significant]; for genes differing by a factor >2 but ≤4, there is a tendency for female overexpression (Gadj = 73.6, df = 1, P < 9.5 × 1018). However, for genes differing by a factor >4, there is a pronounced excess of genes with male overexpression (Gadj = 49.1, df = 1, P < 2.4 × 1012). Although genes with male-biased expression are underrepresented in the cDNA library (19) that was used to construct our microarray, this seems unlikely to account for the pattern reported here.

Male-biased genes also show greater expression divergence between species than either female-biased genes or non–sex-biased genes. Fig. 2 compares the divergence of expression level for genes that are male-biased, female-biased, or non–sex-biased in their expression. Across all genes, the mean divergences of the distributions of all three classes are significantly different from each other, with the greatest mean divergence among the male-biased genes and the smallest among the female-biased genes (male- versus female-biased: Student's t test, t = 8.1, df = 1863, P < 1.3 × 1015; male- versus non–sex-biased: t = 3.6, df = 1549, P < 2.7 × 104; female- versus non–sex-biased: t = –6.1, df = 2800, P < 1.5 × 109).

Fig. 2.

Differential divergence in the level of expression among male-biased, female-biased, and non–sex-biased genes. Genes were ranked on a logarithmic scale by their difference in expression level between D. melanogaster and D. simulans. Blue, male-biased genes; red, female-biased genes; green, non–sex-biased genes. The plot includes 425 genes of each class, chosen so that the origin of the plot would correspond to genes whose divergence is essentially equal. Divergence is calculated as the coefficient of variation of the mean expression values for the sex that shows the expression bias in D. melanogaster and D. simulans. For non–sex-biased genes, the divergence is calculated as the average of the male and female coefficients of variation. Male-biased genes show a larger difference between D. melanogaster and D. simulans than female-biased genes or non–sex-biased genes.

Because of their potentially opposite fitness effects in males and females, genes with sex-limited expression are candidates for the occurrence of sexually antagonistic mutations or mutations that have an epistatic interaction with antagonistic loci (20). Theory predicts an asymmetrical distribution of sex-biased genes between the X chromosome and autosomes (21). Among genes with a significant male bias in expression, we found a 32% deficit of genes on the X chromosome of D. melanogaster and D. simulans as compared to the autosomes (Gadj = 19.9, df = 1; P < 0.0001). A deficiency of male-biased genes on the D. melanogaster X chromosome has been reported previously (4). The data presented here confirm the generality of this pattern in other Drosophila species and also highlight a significant overrepresentation of female-biased genes in expression on the X chromosome. We find a 22% excess of female-biased genes on the X chromosome (Gadj = 13.5, df = 1, P < 0.001), a pattern not previously detected except in microarray experiments that directly compare testes and ovaries (4).

We looked for biologically coherent patterns of functional divergence between D. melanogaster and D. simulans to assess nonrandom changes in gene expression during the evolution of the Drosophila transcriptome. For this purpose, we searched for significant overrepresentation of particular molecular functions, biological processes, cellular components, or regulatory pathways in the subsets of genes differentially expressed between species (9, 22). Genes that show differential expression between these species present a pattern of functional clustering that is primarily sex-by-species–specific but not necessarily related to germline function. Among genes that show no sex bias in expression in either species, functional clustering is undetectable under statistically conservative criteria that correct for multiple tests (tables S3 and S4). In contrast, sex-biased genes that are differentially expressed between D. melanogaster and D. simulans exhibit diverse patterns of functional and pathway enrichment (tables S5 to S8). Especially notable is the presence of genes that are potentially involved in mating behavior (23). Males of D. simulans show overexpression of genes involved in the phototransduction cascade (P = 0.009), whose products are all localized to the rhabdomere (P = 5.7 × 104). Greater male-biased expression of genes involved in phototransduction in D. simulans may reflect the fact that visual stimuli are more important for mating in D. simulans than in D. melanogaster (24), as D. simulans does not mate efficiently in the dark (25). In D. melanogaster males, two genes related to olfaction are upregulated (uncorrected P = 0.019). These newly evolved sex-interspecific expression differences closely mirror the epicuticular hydrocarbons that play a key role in the chemical communication during courtship. D. melanogaster, unlike D. simulans, displays a hydrocarbon profile that varies markedly between the sexes (26).

Overall, our results highlight profound changes in sex-dependent gene expression that have taken place in the Drosophila transcriptome over 2 to 3 million years. The more rapid divergence in gene expression found here for males parallels the observation that male-specific morphological features evolve more rapidly than those of females (27, 28). Random genetic drift and/or relaxed selection for sex-dependent gene expression could result in large, yet nearly neutral, interspecific differences in gene expression. However, the finding that reversals of sex-dependent expression between species are rare, whereas other kinds of changes are common, argues against this hypothesis. A purely neutral model of transcriptome divergence is also difficult to reconcile with the substantial excess of sex-biased genes among those that have evolved differences in expression, as well as with the underlying nonrandom pattern of functional divergence between D. melanogaster and D. simulans in sex-biased genes. The patterns of change are more consistent with the action of sex-dependent selection both within and between species. Sex-dependent selection may result from antagonistic fitness effects between the sexes, by sexual selection (1), or both together (29). Divergent, sex-dependent selection patterns between species could account for the interspecific differences reported here.

The nonrandom chromosomal distribution of sex-biased genes in Drosophila contrasts sharply with that found in mammals, where the X chromosome shows an excess of genes expressed in the male germ line (30). The Drosophila pattern is also at odds with that expected from the theory of sexually antagonistic alleles. In theory, X-linked mutations that favor males at the expense of females will be more likely to fix if they are recessive, whereas X-linked mutations that favor females at the expense of males are more likely to fix if partially dominant (21). Part of the discrepancy may reflect a transposition bias, for example, the preferential movement of reverse-transcribed copies of genes expressed late in spermatogenesis from the X chromosome to the autosomes (31). Alternatively, in contrast with loss-of-function mutations that tend to be nearly recessive (32), sex-limited and sexually antagonistic mutations might consist largely of gain-of-function mutations with partial dominance (33), in which case the pattern reported here would be expected.

Since the divergence between D. melanogaster and D. simulans, significant changes in adult gene expression have evolved in approximately half of the transcriptome, whereas the other half has retained the ancestral pattern of expression. The observations that 83% of the interspecific changes in gene expression are sex-dependent and that divergence in expression levels is greater in males suggest that sex-dependent selection is a major force driving the recent evolution of the Drosophila expression profile. The rapid accumulation of sex-related changes in biologically coherent functional categories and the contrasting chromosomal locations of male-biased and female-biased genes lend additional support to this view. Further experiments should help clarify whether a substantial fraction of the interspecific changes in gene expression are coordinately regulated by a limited number of genes or have evolved those differences largely independently from one another.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

Tables S1 to S10

References and Notes

References and Notes

Stay Connected to Science

Navigate This Article