Dosage compensation and demasculinization of X chromosomes in Drosophila

SUMMARY

The X chromosome of Drosophila shows a deficiency of genes with male-biased expression [1–4], while mammalian X chromosomes are enriched for both spermatogenesis genes expressed pre-meiosis and multi-copy testis genes [5, 6]. Meiotic X inactivation and sexual antagonism can only partly account for these patterns. Here, we provide evidence that dosage compensation (DC) in Drosophila may contribute substantially to the depletion of male genes on the X. To equalize expression between X-linked and autosomal genes in the two sexes, male Drosophila hyper-transcribe their single X, while female mammals silence one of their two X chromosomes. We combine fine-scale mapping-data of dosage compensated regions in D. melanogaster with genome-wide expression profiles in males and females and show that most male-biased genes on the Drosophila X chromosome are located outside dosage compensated regions. Additionally, X-linked genes that have newly acquired male-biased expression in the D. melanogaster lineage are less likely to be dosage compensated, and parental X-linked genes that gave rise to an autosomal male-biased retrocopy are more likely to be found within compensated regions. This suggests that DC in Drosophila contributes to the observed demasculinization of X chromosomes in Drosophila, both by limiting the emergence of male-biased expression patterns of existing X genes, and by contributing to gene trafficking of male-genes off the X.

RESULTS & DISCUSSION

In many animals with separate sexes, a large fraction of the genome shows sex-biased gene expression [7]. In Drosophila, for example, between 30–60% of the transcriptome is expressed differently in males and females (sex-biased genes) [4, 8, 9]. Genes with sex-biased expression often show a non-random genomic distribution, with male-biased genes being significantly depleted on the Drosophila X chromosome [1, 2, 4].

Two main models have been proposed to explain the observed deficiency of male-biased genes on the X [10, 11, 12]. The X of several taxa is transcriptionally inactivated early in spermatogenesis (meiotic X inactivation; [11, 12]), implying that the X is a disfavored location for genes required during spermatogenesis, and X-linked copies of such genes will be selected against in favor of autosomal copies [1, 3]. While X inactivation certainly plays a role to explain the depletion of testis-biased X-linked genes, it cannot account for the observed deficiency of X-linked male-biased genes in somatic tissues [2, 7, 13]. Sexually antagonistic selection can also contribute to observed patterns of X-chromosomal gene content [7, 13, 14]. Specifically, mutations that are beneficial to one sex while deleterious to the other (sexually antagonistic mutations) can accumulate differently on sex chromosomes [10]. Due to hemizygosity of the X in Drosophila males, recessive male-beneficial mutations will accumulate more rapidly on the X relative to autosomes while dominant female-beneficial mutations are expected to fix more easily on the X due to its female-biased transmission [10]. If sexually antagonistic mutations are more often dominant, this could explain the deficiency of male-biased genes on the X of Drosophila [1].

Here, we provide evidence for a third force contributing to the depletion of male-biased genes [15]: hyper-transcription of the single X chromosome in Drosophila males (i.e. dosage compensation [16, 17]). To equalize expression levels between autosomal and X-linked genes, Drosophila males recruit an RNA/protein complex (termed the MSL complex) to their single X chromosome, which induces acetylation of histone H4 [17, 18]. This results in a global change of chromatin structure, facilitating increased rates of transcription of X-linked genes in males [19]. Recent high-resolution mapping experiments support a two-step model of MSL recruitment to the X chromosome of Drosophila [20–23] (see Figure 1A). The MSL complex is thought to first target over 100 chromatin entry sites (termed high-affinity sites; HAS) containing specific MSL recognition elements on the X in males [20]. After this initial, sequence-specific recognition step, local spreading from entry sites in cis along the chromosome leads to MSL-binding to the majority of active genes on the X [21, 22, 24–27].

Figure 1.

A. The two-step model of MSL targeting to the X chromosome. I. The MSL complex (red circles) targets specific high affinity chromatin entry sites (HAS) on the X-chromosome in a sequence-dependent manner. HAS represent a subset of the MSL-bound regions in wildtype Drosophila that also recruit the MSL complex under more stringent conditions (such as when inserted into an autosome, or when integral subunits of the MSL complex are missing). II. After initial targeting, the MSL-complex spreads along in cis from the entry sites (shown by blue arrows), and predominantly binds to the 3′ end of actively transcribed genes. MSL-binding causes acetylation at histone H4 and results in a global change of the chromatin structure, facilitating a two-fold transcriptional up-regulation of X-linked genes in males (green line). B. Models of sex-biased expression versus dosage compensation in Drosophila. I. Direct interference of the dosage compensation machinery with male-biased expression. Binding of the dosage compensation complex, changes in chromatin structure, and global hyper-transcription of X-linked genes may interfere with subsequent transcriptional modifications and up-regulation of X genes in males (green circles). Genes further away from a HAS (or those not bound by the MSL complex) are more likely to be up-regulated in males. II. Indirect effects of dosage compensation on sex-biased gene expression. Genes further from a HAS will less likely be compensated in males, resulting in female-biased expression for genes further away from a HAS.

The mechanism of dosage compensation in Drosophila males could contribute to the observed deficiency of male-biased X-linked genes through direct or indirect effects. First, mechanistic or functional constraint could actively limit further up-regulation of already hyper-transcribed X-linked genes in males [15]. Specifically, the modified chromatin structure of the X may directly interfere with subsequent transcriptional modification of X-linked genes in males [28], or transcription rates may have reached an upper limit on the hyperactive X [15]. Second, the Drosophila X chromosome may undergo incomplete dosage compensation, with many genes on the haploid male X showing little or no up-regulation. Lack of dosage compensation would indirectly result in a deficiency of genes with male-biased expression on the X relative to autosomes. The two-step model of dosage compensation makes several predictions that allow us to distinguish between these two hypotheses (Figure 1B). If the deficiency of male-biased genes in Drosophila is a consequence of the dosage compensation machinery directly interfering with further up-regulation of X-linked genes, genes with male-biased expression should mainly be found outside dosage compensated regions. In particular, 1) genes with male-biased expression should, on average, be further away from the nearest HAS; 2) genes further away from a HAS should exhibit more male-biased gene expression; 3) genes with male-biased gene expression should be dosage compensated less frequently (less likely be bound by the MSL complex) than genes with unbiased or female-biased expression. Alternatively, if the depletion of male-biased genes indirectly results from incomplete dosage compensation, male-biased genes should mainly reside in dosage compensated regions, and we would expect to see opposite patterns (i.e. male-biased genes should be closer to the nearest HAS; genes closer to a HAS should be more male-biased in expression; and genes with male-biased expression should more often be MSL-bound). Finally, if dosage compensation has no influence on patterns of male-biased expression, sex-biased genes should be distributed randomly with regards to HAS and MSL-bound regions.

The 2-step model of dosage compensation

To test these predictions, we combine fine-scale mapping data of HAS and MSL-bound regions in D. melanogaster [20, 22, 23] with genome-wide data on sex-biased gene expression [29]. Recent high-resolution ChIP-seq analysis has identified approximately 150 HAS along the D. melanogaster X [20, 23], and 53% (1132 out of 2142) of all X-linked genes studied were clearly bound by the MSL complex [21, 22, 24]. If the HAS indeed function as chromatin entry sites from which the MSL complex spreads, we expect that genes that are closer to such an entry site are more likely to be dosage compensated. We find that genes bound by the MSL complex in wildtype Drosophila are significantly closer to a HAS (median distance 12507-bp) than unbound genes (median distance 49311-bp; Wilcoxon two-sample test, p<0.0001; Figure 2A). In addition, reduced X expression has been demonstrated in male-like tissue culture cells following RNAi knockdown of MSL2 [30], and genes bound by the MSL complex are more likely to be down-regulated during RNAi treatment (defined by a 1.4-fold decrease or more) than those not bound (24% of bound genes are significantly down-regulated, relative to 14% of genes not bound by MSL; p<0.01). Thus, these data support the two-step model of MSL targeting to the X chromosome, with MSL binding directly resulting in up-regulation of X-linked genes in D. melanogaster males.

Figure 2.

A. Distance to nearest HAS for X-linked genes. X-linked genes are categorized according to their MSL-binding profile [20], expression bias [29] or patterns of retroposition [31]. Genes targeted by the MSL complex (bound) are significantly closer to a HAS than genes not bound (unbound). Genes with male-biased expression are significantly further away from a HAS compared to female- and un-biased genes. Parental genes that have a retrocopy in the genome are significantly closer to a HAS than genes not having a duplicate retrocopy. B. Sex-biased expression versus distance to HAS. Plot of gene expression sex ratio (log₂ female/male ratio) for each X-linked gene in D. melanogaster against distance to its closest HAS. The lines are regression highlighting the trends: Red, female-biased genes; blue, male-biased genes; grey, unbiased genes. The black line describes the trend of all data. C. Fraction of dosage compensated genes. X-linked genes are categorized according to their expression bias [1], their change in sex-biased gene expression on the D. melanogaster lineage [29], or patterns of retroposition [31]. Genes with male-biased expression are significantly less likely to be bound by the MSL complex. Parental genes with a retrocopy are significantly more likely to be MSL-bound than genes not having a retrocopy.

Dosage compensation and sex-biased expression

To test for a link between dosage compensation and sex-biased expression, we categorized X-linked genes as male-, female- and un-biased using expression profiles from gonadectomized flies [1]. All analyses were repeated using expression profiles from whole adult flies or gonads to classify sex-biased expression, with similar results obtained (see Supplementary Materials 1). Patterns of MSL-binding provide evidence that dosage compensation indeed is shaping sex-biased gene expression on the Drosophila X chromosome (Figure 2). Specifically, genes with male-biased expression are significantly further away from HAS than female-biased or unbiased genes (p=5.406×10⁻⁷ and p=4.462×10⁻¹⁰, respectively, Wilcoxon two-sample tests, Figure 2A). Median distance to the closest HAS is 18359-bp for female-biased genes and 17453-bp for unbiased genes, but 46543-bp for genes showing male-biased expression (Figure 2A). Additionally, we observe a significant positive correlation between the magnitude of male-biased expression (measured as the log₂ male:female expression ratio) of individual X-linked genes and their distance to the closest HAS (Figure 2B, Kendall’s τ=0.109, p=3.46 × 10⁻⁸). Thus, contrary to incomplete dosage compensation indirectly resulting in a deficiency of male-biased genes, this data suggests that it is easier to achieve male-biased expression away from compensated regions. Finally, if X-linked genes are classified as bound by the MSL complex or not bound, we observe a significant deficiency of genes with male-biased expression that are targeted by MSL (Figure 2C, χ²⁼26.7, d.f.=2, p=1.58×10⁻⁶). Taken together, we find compelling support that dosage compensation is influencing patterns of sex-biased expression. Importantly, we find no evidence that a simple lack of dosage compensation is indirectly causing the observed deficiency of male-biased genes on the X. Instead, all our findings are consistent with dosage compensation actively limiting or interfering with the evolution of male-biased gene expression at the X chromosome of Drosophila.

Dosage compensation and demasculinisation

X chromosomes of Drosophila are depleted of genes with male-biased expression [1, 2]. If dosage compensation contributes to the deficiency of male-biased genes [1–4], we expect that X-linked genes not bound by the MSL complex contain a similar proportion of male-biased genes as autosomes, while MSL-bound genes are disproportionally deficient for genes with male-biased expression. Figure 3A plots the distribution of sex-biased genes across the different chromosomes, with X-linked genes further classified as either MSL-bound or not. Genes with female-biased expression are randomly distributed across the genome, independent of their MSL-binding pattern (χ²⁼6.9, d.f.=4, p >0.05). As shown previously, male-biased genes are underrepresented on the X relative to autosomes (χ²⁼52.2, d.f.=1, p<0.0001). Strikingly, the observed deficiency appears to be almost entirely driven by a lack of MSL-bound male-biased genes. We find a highly significant under-representation of male-biased genes on the X that are targeted by the MSL complex (χ²⁼81.0; d.f.=1, p<0.0001; Figure 3A), but not for male-biased genes not bound by MSL (χ²⁼1.0; d.f.=1, p >0.05; Figure 3A). Specifically, ~8.1% of MSL bound genes are male-biased, but 15.1% of autosomal genes are classified as male-biased, while X-linked genes not bound by the MSL complex show a proportion of male-biased genes (17.1%) similar to autosomes (see Figure 3A). Additionally, if global hyper-transcription of the X in males limits further up-regulation of individual genes, we not only expect to find fewer male-biased genes, but also the magnitude of sex-biased expression to be less for male-biased genes on the X relative to autosomal male-biased genes [15]. Indeed, absolute values of sex-biased expression ratios are similar for X-linked and autosomal female-biased genes (median log₂ male:female expression ratio −0.496 vs. −0.498, Wilcoxon two-sample test, p >0.5, Figure 3B). However, sex-biased gene expression is significantly lower for male-biased genes on the X relative to autosomes (median log₂ male:female expression ratio 0.324 vs. 0.420, Wilcoxon two-sample test, p<0.001, Figure 3B). Thus, these results suggest that dosage compensation plays a significant role in explaining the demasculinization of the X chromosome in Drosophila. Consistent with the dosage compensation machinery limiting further sex-specific modifications of patterns of gene expression, genes with equal expression in the two sexes are more likely to be bound by the MSL complex than sex-biased genes (see Figure S4).

Figure 3.

A. Fraction of sex-biased genes across Drosophila chromosomes. The percentages of genes with male-biased (blue) or female-biased (red) expression are shown. X-linked genes are divided into those bound by the MSL complex (X_b) or not bound (X_u). Male-biased genes bound by the MSL complex are significantly underrepresented on the X. B. Sex-biased expression ratios across Drosophila chromosomes. Comparison of the extent of male and female biased gene expression ratios for each chromosome in D. melanogaster. Sex-biased gene expression is significantly lower for male-biased genes on the X relative to autosomes. The absolute value of the log₂ expression ratio is plotted on the y-axis. Bold horizontal bars are the median value, the box is the inter-quartile range, and the whiskers is the 95% confidence interval.

Dosage compensation and turnover in sex-biased gene expression

Sex-biased gene expression changes accumulate over time, with male-biased expression displaying a higher rate of turnover in the genus Drosophila than female-biased expression [4, 8, 9]. Extensive categorical changes in sex-bias class were reported between D. melanogaster and D. simulans (~12% of their orthologues show a categorical change in sex-biased expression), mostly between genes showing non-sex-biased expression and genes displaying modest sex-biased expression [29]. If dosage compensation is interfering with the evolution of de novo male-biased gene expression of existing genes on the X, we expect genes that have acquired male-biased expression less often to be bound by the MSL complex. To identify genes that have changed sex-biased expression patterns in the D. melanogaster lineage, we compared sex-biased gene expression profiles for orthologous genes from D. melanogaster, D. simulans and D. yakuba [29]. Patterns of changes in sex-biased gene expression are consistent with dosage compensation limiting the acquisition of male-biased expression of X-linked genes. In particular, 989 genes are classified as male-biased in D. melanogaster but not D. simulans or D. yakuba (11% of which are X-linked), and 278 genes are classified as female-biased specifically in D. melanogaster (13% of which are X-linked). However, while 50% of newly female-biased X-linked genes are MSL-bound (mirroring the chromosome-average pattern of MSL binding), only 37% of newly male-biased genes are bound by the MSL complex (χ²=5.1; d.f.=1, p<0.02; Figure 3C). Thus, the significant deficiency of newly evolved male-biased genes in D. melanogaster bound by MSL provides independent evidence that dosage compensation is interfering with the evolution of male-biased gene expression on the X.

Dosage compensation and gene trafficking

Comparative genomic studies in Drosophila have uncovered an excess of retrogenes that originate from the X chromosome and retropose to the autosomes, where they evolve male-biased expression [3, 31]. If dosage compensation is contributing to this exodus of genes by preventing the evolution of male-biased expression at their X chromosomal location, we expect parental X-linked copies of retroposed genes to reside closer to a HAS, on average, and more likely be bound by the MSL complex relative to X-genes that have not produced a duplicate copy. Using whole-genome data from multiple Drosophila species, over 90 retroposition events that gave rise to candidate functional genes in D. melanogaster were recently identified, a third of which originated from parental genes located on the X [31]. Consistent with dosage compensation interfering with acquiring male-biased expression at these genes at their X location, parental genes reside significantly closer to a HAS than non-parental genes (Figure 2A). Median distance to the closest HAS is 10294-bp for parental genes with a retrotransposed duplicate, but 26544-bp for genes without a retrocopy (Wilcoxon two-sample test, p<0.001, Figure 2A). Additionally, parental genes that gave rise to a retroposed gene are significantly more likely to be MSL bound than non-parental genes (70% of parental X-linked genes are MSL-bound, compared to 51% genome-average, χ²=1211.4, d.f.=1, p <2×10⁻¹⁶, Figure 2C). Thus, patterns of retroposition provide further evidence that dosage compensation is limiting the evolution of male-biased expression, and suggests that dosage compensation significantly contributes to observed trafficking of male-biased genes off the X chromosome.

Conclusion

We find compelling evidence that dosage compensation influences patterns of sex-biased expression in Drosophila, and contributes to movement of male-biased genes off the X. Our analysis suggests that the deficiency of male-biased genes on the Drosophila X does not simply reflect a lack of dosage compensation at some genes but instead can partly be accounted for by dosage compensation directly interfering with further up-regulation of MSL-bound, already hyper-transcribed X-linked genes in males. The X chromosome in male Drosophila is encumbered by the MSL complex and its chromatin structure is modified globally, which may limit subsequent transcription factor binding or chromatin remodeling, and thus inhibit further transcriptional activation. Indeed, direct interference between chromatin remodeling complexes and the dosage compensation machinery has been reported in Drosophila [28]. Additionally, male-biased gene expression originates mainly by increasing transcription of non-biased genes in males (rather than down-regulation in females [15]), and higher expression levels may be harder to achieve on an already hyper-transcribed chromosome. High-expression male-biased genes are located less often on the X than low-expression male-biased genes. This is expected if limits in rates of transcription prevent the accumulation of male-biased genes on the X, since such limitations are less likely to affect genes that are transcribed only at low levels [15].

Not all organisms show a deficiency of male-biased genes on the X. In particular, mammalian X chromosomes are enriched for single-copy spermatogenesis genes that are expressed pre-meiosis [5, 6], and multi-copy testis genes showing post-meiotic expression [32]. This difference in X-chromosomal gene content between taxa could result from fundamental differences in the mechanisms of dosage compensation [13, 33, 34]. Dosage compensation in mammals is achieved by first doubling global expression levels of the X in both sexes [35], followed by inactivation of one X in females [36]. The chromatin structure of the active X in mammals and baseline transcription rates of X-linked genes thus appear the same between the sexes (even though they might differ from average autosomal rates of transcription), therefore imposing no male-specific restrictions on the evolution of sex-biased expression patterns. Thus, the difference in X-chromosomal gene content between Drosophila and mammals – with a deficiency vs. an accumulation of male-biased genes – may be understood in light of their vastly different dosage compensation mechanisms.

EXPERIMENTAL PROCEDURES

Data

FlyBase Drosophila melanogaster release 5.5 was used to cross-reference the different data sets employed in this study. We used the physical location of each of the 150 HAS identified in ref. [20] to calculate the distance from the 3′ end (the MSL complex preferentially binds to the 3′ end of coding genes; [20, 22]) to the nearest HAS of each X-linked gene obtained from FlyBase. Note that these 150 high affinity sites were originally identified in msl3 mutant male embryos, and confirmed in MSL3-TAP male cell lines [20]. Using different experimental approaches and Drosophila lines to identify high affinity sites (either by reducing levels of MSL-spreading factors by RNAi against MOF, MLE, or MSL3 or by lowering levels of crosslinking to reveal sites of more intimate contact of MSL proteins with DNA), ref. [23] identified a highly overlapping set of HAS in D. melanogaster. In particular, 90 of the 130 X-chromosomal HAS identified in ref. [23] perfectly overlap with the 150 HAS identified by ref. [20]. We used the high-resolution ChIP-seq mapping data from ref. [20] to classify X-linked genes as bound or unbound by the MSL complex; 1132 X genes of release 5.5 were classified as MSL-bound, and 1010 were classified as unbound [20]. Sex-biased gene expression data for D. melanogaster were extracted from published global expression profiles of gonadectomized flies using species-specific microarrays [1]. The D. melanogaster expression data and platform descriptions can be found at NCBI’s Gene Expression Omnibus accession numbers GPL4629. We classified genes showing male-biased, female-biased or unbiased expression using the same criteria as in the database Sebida [37]. Briefly, gene expression data were input to BAGEL [38] to obtain p-values and permutations of the data were input to BAGEL to obtain a 5% FDR cutoff. According to these criteria, 135 D. melanogaster X-linked genes were classified as male-biased, 144 were female-biased and 864 unbiased. In addition, we also performed our analysis on sex-biased gene expression versus dosage compensation with genes classified as male-, female- and un-biased from whole adult flies and gonads [1]. The results of this analysis are shown in the Supplementary Materials. To identify changes in patterns of sex-biased expression in the D. melanogaster lineage, we used global species-specific expression profiles from ref. [29]. X-linked parental genes that gave rise to candidate functional retrogenes in D. melanogaster were taken from ref. [31], which identified over 90 retroposition events using whole genome data for multiple Drosophila species; 31 events involved parental genes located on the X. We also performed Gene Ontology categorization versus MSL-binding patterns (Table S1) or sex-biased expression (Table S2). Datasets were integrated in R based on FBgn number. All the integrated data used are available as Supplementary Data.

Statistical analysis

To test for a difference in distance to the closest HAS for genes categorized according to their MSL-binding profile, their expression bias or patterns of retrotransposition, Wilcoxon tests were used. We performed non-parametric Kendall tau rank correlations between distance to closest HAS and log₂ male:female expression ratio for three groupings of X-linked expression data: male-biased, female-biased, and unbiased genes. A chi square test was used to evaluate independence between MSL-binding and expression bias or retroposition patterns, by comparing the observed number of genes showing binding for a given condition (i.e. male-, female- unbiased or parental gene, non-parental gene) with the expected number based on the fraction of X-linked genes showing MSL binding. To test for heterogeneity in the chromosomal distribution of male-biased and female-biased transcripts, chi square statistics were calculated, by comparing the observed number of genes showing differential expression (male- or female- biased) on each chromosome with the expected number based on the fraction of the genome contained on each arm. X-linked genes were further categorizes as bound by the MSL complex or unbound. To compare the distributions of sex-biased expression ratios between autosomal genes and X-linked genes, X-linked genes bound and not bound by the MSL complex, Wilcoxon tests were used. Bonferroni corrected p-values are used to account for multiple testing.

HIGHLIGHTS.

• Increased transcription of X chromosome in Drosophila males to equalize X-autosome expression (dosage compensation)

• Genes with male-biased expression are depleted on the Drosophila X chromosome

• Dosage compensation interferes with acquisition of male-biased expression of X-linked genes

• Dosage compensation contributes to gene trafficking of male-genes off the X