Conservation of Regional Variation in Sex-Specific Sex Chromosome Regulation

Abstract

Regional variation in sex-specific gene regulation has been observed across sex chromosomes in a range of animals and is often a function of sex chromosome age. The avian Z chromosome exhibits substantial regional variation in sex-specific regulation, where older regions show elevated levels of male-biased expression. Distinct sex-specific regulation also has been observed across the male hypermethylated (MHM) region, which has been suggested to be a region of nascent dosage compensation. Intriguingly, MHM region regulatory features have not been observed in distantly related avian species despite the hypothesis that it is situated within the oldest region of the avian Z chromosome and is therefore orthologous across most birds. This situation contrasts with the conservation of other aspects of regional variation in gene expression observed on the avian sex chromosomes but could be the result of sampling bias. We sampled taxa across the Galloanserae, an avian clade spanning 90 million years, to test whether regional variation in sex-specific gene regulation across the Z chromosome is conserved. We show that the MHM region is conserved across a large portion of the avian phylogeny, together with other sex-specific regulatory features of the avian Z chromosome. Our results from multiple lines of evidence suggest that the sex-specific expression pattern of the MHM region is not consistent with nascent dosage compensation.

SEX chromosomes exhibit extensive sex-specific gene regulation in many animals, largely resulting from two independent forces. First, the reduced gene content on the sex-limited Y or W chromosome observed in many species (Bachtrog 2013) leads to a reduction in gene dose in the heterogametic sex for X- or Z-linked genes. Because gene expression is often a function of gene dose, this dosage effect (Zhang et al. 2013) means that many X- or Z-linked loci show differences in gene expression between males and females. Variation in gene dose is often thought to be detrimental, and different mechanisms of dosage compensation have evolved to address the differences in sex chromosome dose in some organisms (Mank 2013). Dosage compensation mechanisms in several organisms show extensive regional variation, with some regions exhibiting less complete compensation than other regions on the therian (Carrel and Willard 2005) and stickleback (Schultheiß et al. 2015) X chromosomes.

Second, independent of gene dose differences, sex chromosomes are also inherited unequally between the sexes, leading to distinct evolutionary forces relative to the autosomes (Rice 1984; Charlesworth et al. 1987). In particular, conflicting selection on males and females results in unequal sex-specific selection pressures acting on the sex chromosomes. This, in turn, influences rates of sequence and expression evolution (Vicoso and Charlesworth 2009; Meisel and Connallon 2013). Older sex chromosome regions, where recombination between the Z-W orthologs was halted earlier, have been exposed to these selection pressures for a greater period of time. This cumulative nature of sex-specific selection can lead to regional differences in sex-specific gene regulation based on regional age (Wright et al. 2012; Wang et al. 2014).

The avian Z chromosome, which is conserved across all extant birds, is particularly interesting in the study of regional differences in sex-specific gene regulation. All avian species examined thus far show incomplete dosage compensation (Ellegren et al. 2007; Itoh et al. 2007; Wolf and Bryk 2011; Wright et al. 2012; Moghadam et al. 2013; Uebbing et al. 2013; Wang et al. 2014). As a result, most of the Z chromosome exhibits male-biased expression (Ellegren et al. 2007). However, there is extensive variation in the degree of male bias in expression across the chromosome. Previous work in the chicken identified the male hypermethylated (MHM) region, located between 25 and 35 Mb on the Z chromosome, as a region with extreme sex-specific regulatory effects (Melamed and Arnold 2007). In addition to hypermethylation in males, which decreases average expression in this sex, the MHM locus encodes a female-specific noncoding RNA (ncRNA) that accumulates around the site of transcription in female chickens (Teranishi et al. 2001). There is also evidence for female-specific acetylation of histones in the chromatin surrounding this area, which is also typically associated with increased gene expression in females (Bisoni et al. 2005).

The regulatory landscape of the MHM region is interesting for several reasons. The pronounced deficit of male-biased genes, together with the role of ncRNAs, such as Xist, in dosage compensation mechanisms in other species (Deng and Meller 2006; Rens et al. 2010), leads to the suggestion that the MHM region might represent a region of nascent dosage compensation (Melamed and Arnold 2007). Given that it is situated within the oldest portion of the avian sex chromosomes (Zhou et al. 2014), it may be that selection for dosage compensation has simply had longer to act on this region, in a similar way to the more complete dosage compensation on the older regions of the therian X chromosome (Carrel and Willard 2005). In contrast, on the basis of coexpression patterns, others have suggested that the MHM locus is a regulator of DMRT1 (Yang et al. 2010b; Caetano et al. 2014), the major avian sex-determining gene (Smith et al. 2009), which is located nearby on the Z chromosome. Therefore, the distinct sex-specific regulatory pattern of genes in this region instead may reflect an important role in sex-specific fitness and sexual differentiation.

Furthermore, older regions of the Z chromosome show elevated levels of male-biased expression in the chicken (Wright et al. 2012). Although the regions of the sex chromosomes stopped recombining independently across the avian phylogeny (Zhou et al. 2014), the degree of male bias is also associated with sex chromosome age in distantly related avian species (Wang et al. 2014), suggesting that the effect is broadly conserved. This is in sharp contrast to the MHM region, where sex-specific regulation has been observed only in the chicken (Teranishi et al. 2001; Melamed and Arnold 2007). Studies in zebra finches, flycatchers, crows, emus, and ostriches (Itoh et al. 2010; Wolf and Bryk 2011; Uebbing et al. 2013; Wang et al. 2014) failed to recover regional variation in male-biased gene regulation in this portion of the Z chromosome. Therefore, it has yet to be established whether this region is a chicken-specific regulatory feature of the Z chromosome or is more broadly conserved.

Here we assessed regional variation in gene regulation across the avian Z chromosome in six species spanning 90 million years. Our results support the growing body of evidence that incomplete compensation is a central feature of avian sex chromosome evolution (Mank 2013) and that broadly consistent patterns of masculinization in gene expression accumulate over time across the Z chromosome. More important, we show that the MHM region is conserved across a wide range of avian evolutionary history. Furthermore, our analysis suggests that the MHM region is not a region of nascent dosage compensation.

Materials and Methods

Transcriptome assembly

We previously assembled adult gonad and spleen transcriptomes across six avian species. Detailed methods for the assembly are described elsewhere (Harrison et al. 2015; Wright et al. 2015), and Illumina reads have been deposited in the NIH Short Read Archive (PRJNA271731). Briefly, we obtained RNA sequencing (RNA-seq) data from the adult spleen and gonad of captive populations of Anas platyrhynchos (mallard duck), Anser cygnoides (swan goose), Numida meleagris (helmeted guineafowl), Pavo cristatus (Indian peafowl), Meleagris gallopavo (wild turkey), and Phasianus colchicus (common pheasant) at the start of their first breeding season. RNA was sequenced on an Illumina HiSeq 2000 at the Wellcome Trust Centre for Human Genetics, University of Oxford, resulting in, on average, 26 million 100-bp paired-end reads per sample. Data were quality assessed using FastQC v0.10.1 and trimmed with Trimmomatic v0.22 (Lohse et al. 2012). The Trinity method (Grabherr et al. 2011) was used to assemble de novo transcriptomes for each species separately, and expression levels of genes were obtained using RSEM v1.1.21 (Li and Dewey 2011). The isoform with the highest expression level in each Trinity contig cluster was selected for further analysis, and ribosomal RNA (rRNA) was removed. We also imposed a minimum expression threshold of 2 FPKM (fragments per kilobase of exons per million mapped reads) in at least half of any of the tissues from either sex (Supporting Information, Table S1).

Identification of reciprocal orthologs and chromosomal location

To maximize our ability to characterize variation in gene regulation across the Z chromosome and male MHM region, located at 25–35 Mb (Teranishi et al. 2001; Melamed and Arnold 2007), we identified pairwise reciprocal orthologs between each species and Gallus gallus (chicken). Full details of this approach are described in Harrison et al. (2015). Briefly, chicken cDNA was obtained from Ensembl v73 (Galgal4/GCA_000002315.2) (Flicek et al. 2013), the longest isoforms were extracted, and orthologs were identified using BLASTN v2.2.27+ (Altschul et al. 1990) with a minimum percentage identify of 30% and an E-value cutoff of 1 × 10⁻¹⁰. Reciprocal orthologs were identified using the highest BLAST score. Because avian genomes exhibit an unusual degree of stability (Stiglec et al. 2007) and the Z chromosome is highly conserved across birds (Skinner et al. 2009; Vicoso et al. 2013), chromosomal location was assigned from chicken.

Quantifying regulatory variation across the avian Z chromosome

Reads from each sample were separately mapped back to the Trinity contigs using RSEM v1.1.21 (Li and Dewey 2011). Read counts were extracted from RSEM for each tissue and species separately and normalized using the trimmed mean of M-values (TMM) in edgeR (Robinson et al. 2010). A minimum expression filter of 2 RPKM (reads per kilobase of exons per million mapped reads) in at least half of either sex was applied separately to each tissue. Because our filtering criteria permit sex-limited contigs, we added a small integer (1) to all RPKM values to allow log₂ transformation of genes with zero expression.

A moving average of sex-biased expression (log₂ male − log₂ female RPKM) was calculated across the Z chromosome for each species using previously optimized parameters (Mank and Ellegren 2009). Specifically, a moving average with a window of 30 genes and a shift of 1 gene was calculated using the running function from the gtools package v.3.4.2 in R (R Core Team 2014). Confidence intervals for median expression of genes located on the Z chromosome and within the MHM region (at 25–35 Mb) (Melamed and Arnold 2007) were calculated using bootstrapping with 1000 repetitions. Wilcoxon tests were used in R to specifically test whether median expression of the MHM region is significantly different from the whole Z chromosome.

In order to identify the genes driving the distinct expression pattern of the MHM region, the most strongly female-biased genes in the MHM region were removed sequentially until there was no significant difference in median sex bias (log₂ male − log₂ female RPKM) between the MHM region and the entire Z chromosome. This was conducted using a one-tailed Wilcox test. Moving averages were recalculated using this filtered data set.

Sequence evolution of genes in the MHM region

For each species, each contig was BLASTed against the orthologous chicken protein sequence using BLASTX v2.2.30+ (Altschul et al. 1990) with a minimum percentage identify of 30% and an E-value cutoff of 1 × 10⁻¹⁰. Coding frames were extracted using BLAST outputs, and contigs with no valid protein-coding sequence were excluded.

We aligned coding sequences with orthologous chicken sequences using PRANK v140603 (Loytynoja and Goldman 2005). Genes were removed if the aligned regions were fewer than 33 amino acids in length. We used the one-ratio model test (model = 0, NSsites = 0) in the CODEML package in PAML v4.8 (Yang 2007) to identify the contribution of purifying selection to coding sequence evolution of Z-linked genes. For each pairwise comparison, we compared the one-ratio model where ω is fixed to equal 1 (expected d_N/d_S under neutral evolution) to the model where ω is estimated, specifying the following pairwise gene tree (focal Galloanserae species, chicken). P-values were calculated using the resulting likelihood values with one degree of freedom and corrected for multiple tests using the qvalue function in R (R Core Team 2014). The proportions of genes evolving with a significant contribution of purifying selection were compared between the Z chromosome and the MHM region using chi-squared tests. Finally, we calculated median d_N/d_S for Z-linked and MHM region genes, and 95% confidence intervals (CIs) were calculated using bootstrapping with 1000 replicates. Genes were removed from the analysis if d_S > 2, thereby excluding genes subject to mutational saturation (Axelsson et al. 2008).

Functional gene analysis

We tested whether the MHM region is enriched for gene function terms compared to the whole Z chromosome using GOrilla (Eden et al. 2007; Eden et al. 2009). Mouse reciprocal orthologs were identified using BioMart (Ensembl v80) for chicken genes. The target list was comprised of MHM-linked orthologs and the background list of Z-linked orthologs. P-values were corrected for multiple tests with the Benjamini-Hochberg method (Benjamini and Hochberg 1995).

Synteny of avian Z chromosome

We assessed synteny of the Z chromosome using the method described in Dean et al. (2015). Proteomes were obtained from Ensembl v73 and v79 for chicken and Anolis carolinensis, respectively; the longest isoforms were extracted, and orthologs were identified using reciprocal BLASTP v2.2.27+ (Altschul et al. 1990) with an E-value cutoff of 1 × 10⁻¹⁰. Synteny between chicken and Anolis was quantified using MCScanX (Wang et al. 2012) with default values. MC ScanX returns collinear blocks that represent syntenic regions where gene order is preserved.

Data availability

Illumina reads used in this study are deposited in the NIH Short Read Archive (PRJNA271731)

Results

We previously assembled adult gonad and spleen transcriptomes across six species within the Galloanserae lineage (Harrison et al. 2015; Wright et al. 2015). The Galloanserae, which include both the Galliformes (landfowl) and Anseriformes (waterfowl), form a monophyletic clade that originated 90 million years ago (Van Tuinen and Hedges 2001), making it one of the oldest within Aves (Figure 1). Major genomic rearrangements are infrequent in the avian genome (Stiglec et al. 2007), potentially resulting from a lack of active transposons (Toups et al. 2011), and previous work has revealed that Z-chromosome synteny is highly conserved across the Galloanserae (Skinner et al. 2009). Therefore, we used the chicken to identify reciprocal pairwise orthologs for each of our study species, with which we assigned chromosomal location. This resulted in sequence and expression data for, on average, 421 Z-linked and 7634 autosomal orthologs in the spleen and 497 Z-linked and 9083 autosomal orthologs in the gonad for each species.

Figure 1.

Taxonomic distribution of MHM region. Shown are the evolutionary relationships and relative branch lengths (from birdtree.org) of taxa in which presence (purple) or absence (black) of the MHM region has been examined. Studies used a combination of expression, methylation, and sequence orthology data. We found that sex-specific regulatory variation in the MHM region is conserved across the Galloanserae (species names in this study are highlighted in gray), and this, together with preliminary evidence in the plover (highlighted in light purple), suggests that the MHM regulatory pattern was lost secondarily in the passeriform birds. ¹Itoh et al. (2007); ²Wang et al. (2014); ³Uebbing et al. (2013); ⁴Wolf and Bryk (2011); ⁵Moghadam et al. (2013); ⁶this study; ⁷Melamed and Arnold (2007); ⁸Mank and Ellegren (2009); ⁹Teranishi et al. (2011); ¹⁰Itoh et al. (2011).

Our results are consistent with previous work demonstrating incomplete dosage compensation in many avian species (Ellegren et al. 2007; Itoh et al. 2007; Naurin et al. 2011; Wolf and Bryk 2011; Wright et al. 2012; Moghadam et al. 2013; Wang et al. 2014). Z-chromosome expression is male biased in both the gonad and spleen across all of our species (Figure 2 and Figure S1). Specifically, median female expression of the Z chromosome is significantly lower than female autosomal expression and male Z-chromosome expression (Table S2 and Table S3). Furthermore, the proportion of Z-linked dosage-compensated genes (defined as log₂ male:log₂ female RPKM ratio ≤ 0.5 and ≥ −0.5) is higher in the spleen relative to the gonad (Table 1), consistent with previous findings that gene-by-gene (i.e., local) mechanisms of dosage compensation are most effective in somatic tissue (Mank and Ellegren 2009). Additionally, we found a much greater proportion of Z-linked sex-biased genes (defined as P < 0.05 and log₂ male:log₂ female RPKM ratio ≤ −1 or ≥ 1) in the gonad relative to the spleen (Table 1). Our results therefore contribute to the increasing body of evidence that incomplete dosage compensation is a central tenet of avian sex chromosome evolution (Mank 2013).

Figure 2.

Density plots of male:female expression ratio. Density plots of log₂ male:log₂ female RPKM ratio for genes expressed in the adult gonad (A) and spleen (B) for each species calculated using kernel density estimation. Density plots for autosomal genes are shown in purple, and median autosomal log₂ male:log₂ female expression ratios are shown by a solid purple line. Density plots for Z-linked genes and median expression ratios are shown in green. For all species, the autosomal density plot is centered near zero (shown by a black dotted line), indicating that autosomal gene expression is typically unbiased. In contrast, the density plots of Z-linked genes are shifted to the right, indicating that Z-linked genes are typically male biased. The width of the plot indicates the variation in sex-biased gene regulation, which is greater in the gonad than in the spleen.

Table 1. Proportion of dosage-compensated genes on the Z chromosome.

Species	Spleen			Gonad
	No. of Z-linked genes	Proportion compensated	Proportion sex biased	No. of Z-linked genes	Proportion compensated	Proportion sex-biased
Goose	421	0.710	0.017	494	0.277	0.480
Duck	408	0.706	0.017	470	0.215	0.568
Guineafowl	431	0.548	0.019	500	0.264	0.498
Peafowl	430	0.519	0.030	507	0.239	0.556
Turkey	421	0.589	0.033	504	0.260	0.512
Pheasant	414	0.558	0.063	504	0.234	0.530

Chromosomal location is based on synteny with chicken reciprocal orthologs. Dosage compensation is defined as log₂ male:log₂ female RPKM ratio ≤ 0.5 and ≥ −0.5. Sex-biased genes were defined as female biased (P < 0.05 and log₂ male:log₂ female RPKM ratio ≤ −1) and male biased (P < 0.05 and log₂ male:log₂ female RPKM ratio ≥ 1).

Regional variation in Z-chromosome masculinization

To compare regional sex-specific regulation across our study species, we plotted a moving average of sex-biased expression across the Z chromosome (Figure 3 and Figure S2). The Z chromosome is present most often in males and therefore is subject to increased masculinizing selection. However, the avian Z and W chromosomes diverged in a stepwise process, and there is a considerable difference in the age of these strata (Wright et al. 2012, 2014; Wang et al. 2014; Zhou et al. 2014). Given the difference in age across these regions, we expected cumulative exposure to masculinizing selection to differ across the Z chromosome (Wright et al. 2012).

Figure 3.

Moving average of male-biased expression across the Z chromosome. Moving averages of log₂ male:log₂ female RPKM ratio for genes expressed in the adult spleen for each of the six Galloanserae species. Positional information on the Z chromosome was taken from chicken reciprocal orthologs. Moving averages were calculated using consecutive windows of 30 genes, moving along the Z chromosome one gene at a time. The two black vertical dashed lines depict the boundaries of the MHM region defined by Melamed and Arnold (2007).

Recombination between the Z and W ceased independently in Galliformes and Anseriformes for the youngest region of the sex chromosomes (Wright et al. 2014), and we can therefore divide the Z into the old conserved region (regions corresponding to 43–80 Mb on the chicken Z chromosome) and young independently evolving portion (regions corresponding to 0–43 Mb on the chicken Z chromosome). We excluded the MHM region from this analysis. We found that the Z chromosome is masculinized in both the spleen and gonad, where male-biased expression is greater in the older regions of the Z chromosome across all species. This difference is significant in the spleen across all species with the exception of the turkey and significant in the turkey and pheasant gonad (Table 2).

Table 2. Masculinization of the avian Z chromosome.

Species	Spleen log2 expression		Gonad log2 expression
	Median male bias of old Z region (95% CI)	Median male bias of young Z region (95% CI)	Median male bias of old Z region (95% CI)	Median male bias of young Z region (95% CI)
Goose	0.409 (0.380–0.472)	0.243 (0.214–0.289)	0.578 (0.371–0.709)	0.320 (0.132–0.603)
		P < 0.001		P = 0.130
Duck	0.412 (0.352–0.450)	0.292 (0.229–0.340)	0.578 (0.324–0.819)	0.355 (0.159–0.645)
		P < 0.001		P = 0.134
Guineafowl	0.524 (0.463–0.575)	0.425 (0.380–0.491)	0.713 (0.446–0.921)	0.517 (0.260–0.766)
		P = 0.009		P = 0.061
Peafowl	0.520 (0.480–0.559)	0.464 (0.430–0.498)	0.686 (0.487–0.931)	0.569 (0.338–0.814)
		P = 0.024		P = 0.142
Turkey	0.461 (0.374–0.524)	0.396 (0.323–0.454)	0.774 (0.567–1.018)	0.563 (0.348–0.780)
		P = 0.082		P = 0.017
Pheasant	0.497 (0.402–0.552)	0.398 (0.345–0.466)	0.867 (0.608–1.021)	0.478 (0.289–0.710)
		P = 0.034		P = 0.003

The 95% confidence intervals (CIs) were calculated by bootstrapping with 1000 repetitions. Significant differences in log₂ median expression between the old and young regions of the Z chromosome are shown in bold and were assessed using one-tailed Wilcox P-values. Male bias is defined as log₂ male:log₂ female RPKM ratio.

Conservation of sex-specific regulation in the MHM region

We uncovered the characteristic valley of male-biased expression associated with the MHM region, defined by previous studies (Melamed and Arnold 2007; Mank and Ellegren 2009) and located at 25–35 Mb on the chicken Z chromosome (Figure 3), in all six species. Specifically, in the spleen, we observed a pronounced valley of male-biased expression, resulting from a deficit of male-biased genes and an excess of strongly female-biased genes (Figure 3). In fact, male bias is significantly lower in this region compared to the whole Z chromosome in all species (Table 3). Our results indicate that the MHM region is not a chicken-specific feature of the Z chromosome but is conserved at least within the 90-million-year-old Galloanserae (Figure 1).

Table 3. Gene regulation of the MHM region.

Species	Spleen		Gonad
	Z chromosome log2 median male bias (95% CI)	MHM region log2 median male bias (95% CI)	Z chromosome log2 median male bias (95% CI)	MHM region log2 median male bias (95% CI)
Goose	0.319 (0.287–0.351)	0.268 (0.101–0.378)	0.457 (0.285–0.617)	0.301 (−0.202–0.897)
		P = 0.038		P = 0.513
Duck	0.325 (0.290–0.357)	0.221 (0.096–0.266)	0.485 (0.313–0.654)	0.486 (-0.305–1.023)
		P = 0.007		P = 0.504
Guineafowl	0.457 (0.408–0.491)	0.314 (0.221–0.427)	0.585 (0.435–0.771)	0.483 (-0.221–0.924)
		P = 0.001		P = 0.182
Peafowl	0.491 (0.457–0.510)	0.458 (0.293–0.502)	0.663 (0.476–0.821)	0.607 (0.222–1.148)
		P = 0.049		P = 0.573
Turkey	0.396 (0.353–0.442)	0.214 (0.064–0.290)	0.695 (0.541–0.824)	0.791 (0.133–1.111)
		P = 0.001		P = 0.579
Pheasant	0.419 (0.372–0.466)	0.221 (−0.026–0.418)	0.665 (0.476–0.768)	0.477 (0.032–0.960)
		P = 0.002		P = 0.288

95% confidence intervals were calculating by bootstrapping with 1000 repetitions. Significant differences in log₂ median expression between the MHM and Z chromosomes are shown in bold and were assessed using one-tailed Wilcox P-values. Male-bias is defined as log₂ male: log₂ female RPKM ratio.

In contrast, in the adult gonad, we did not observe the characteristic MHM region male-biased expression valley (Figure S2 and Figure S3). Male bias is also not significantly lower in this region than in the whole Z chromosome in any of the six Galloanserae species (Table 3), consistent with a previous study showing that the MHM region is less pronounced in the adult gonad (Mank and Ellegren 2009). Sex-specific regulation of gonadal expression is considerably more variable than the spleen (Figure 2), and it is possible that this may mask any distinct gene regulation pattern in the MHM region. However, our findings highlight the variable and tissue-biased nature of gene regulation in this portion of the Z chromosome.

Is the MHM region a region of dosage compensation?

The MHM region has been proposed previously to represent regional dosage compensation on the avian Z chromosome because of the high concentration of female-biased genes (Melamed and Arnold 2007). Here we used Ohno’s theory of dosage compensation as a framework to explicitly test the status of dosage compensation in this region (Ohno 1967). Ohno’s theory predicts that hemizygosity of the Z chromosome selects for hypertranscription of Z-linked genes in females to compensate for the single Z chromosome. This balances expression between the sexes and between the Z chromosome and autosomes.

Although Ohno’s theory suggests that female expression of the MHM region should be higher than uncompensated regions of the Z chromosome, we found no significant difference in female expression between the MHM region and the whole Z chromosome in any species (Table 4). In contrast, in the male spleen, we observed significantly lower expression of the MHM region compared to the whole Z chromosome in four of six species. Male downregulation is not consistent with Ohno’s theory of dosage compensation, particularly without hypertranscription in females (Ohno 1967). In the gonad, we observe no significant difference in male or female expression between the MHM region and Z chromosome (Table S4), consistent with our previous findings that male-specific regulation in the MHM region is not pronounced in adult gonadal tissue. We conclude that together the deficit of male-biased genes and enrichment of strongly female-biased genes in this portion of the Z chromosome is not likely to represent regional dosage compensation as defined by Ohno.

Table 4. The MHM region is downregulated in males.

Species	Z chromosome log2 median expression		MHM region log2 median expression
	Female (95% CI)	Male (95% CI)	Female (95% CI)	Male (95% CI)
Goose	3.545 (3.395–3.752)	3.820 (3.708–4.065)	3.146 (2.692–3.996)	3.434 (2.760–3.855)
			P = 0.134	P = 0.036
Duck	3.636 (3.489–3.824)	3.982 (3.845–4.149)	3.336 (2.753–3.642)	3.586 (3.174–3.974)
			P = 0.098	P = 0.024
Guineafowl	3.306 (3.177–3.435)	3.771 (3.647–3.886)	2.965 (2.698–3.619)	3.074 (2.561–3.745)
			P = 0.102	P = 0.007
Peafowl	3.377 (3.178–3.523)	3.789 (3.612–4.008)	3.390 (2.964–3.809)	3.629 (3.095–4.028)
			P = 0.861	P = 0.441
Turkey	3.114 (2.908–3.331)	3.509 (3.376–3.722)	2.664 (2.445–2.943)	2.903 (2.493–3.286)
			P = 0.120	P = 0.007
Pheasant	3.142 (3.013–3.309)	3.508 (3.372–3.781)	3.075 (2.575–3.423)	3.015 (2.728–3.717)
			P = 0.621	P = 0.094

The 95% confidence intervals (CIs) were calculated by bootstrapping with 1000 repetitions. Significant differences in log₂ median expression between the MHM region and Z chromosome are shown in bold and were assessed using one-tailed Wilcox P-values. Expression values are from spleen samples.

Architecture of the MHM region

We used an iterative approach to identify the genes underlying the characteristic valley of male-biased expression in the spleen MHM region. For each species, we sequentially removed the most strongly female-biased MHM genes until there was no significant difference in sex bias (log₂ male:log₂ female RPKM ratio) between the MHM region and Z chromosome as a whole.

We found that a very low proportion of Z-linked genes (<0.160 in all species) drives the distinct gene-regulatory pattern we observed (Table S5). Removing these few genes not only restores male bias in this region to the Z-chromosome median (Figure 4) but also eliminates any significant difference between male expression of this region and the Z chromosome (Figure 5). In contrast, female expression of the MHM region remains statistically indistinguishable from the Z chromosome, reinforcing our earlier finding that the MHM region is downregulated in males (Table S5).

Figure 4.

Sex-specific gene regulation in the MHM region. A deficit of male-biased genes and a pronounced valley in the moving averages of log₂ male:log₂ female expression ratio (solid black lines) are characteristic of the MHM region. A limited number of genes drive the distinct expression pattern of the MHM region (listed in Table S6), and dotted gray lines show recalculated moving averages after the exclusion of these female-biased genes. Expression values are from the adult spleen. Physical position of chicken DMRT1 (ENSGALG00000010160) and the locus encoding the MHM region ncRNA are indicated. The dashed vertical lines depict unbiased expression.

Figure 5.

MHM region is downregulated in males. Male expression is shown in blue and female expression in red. Solid lines show moving averages of log₂ RPKM calculated using consecutive windows of 30 genes, moving along the Z chromosome one gene at a time. Dotted lines show recalculated moving averages after the exclusion of the female-biased genes listed in Table S6. Solid horizontal lines depict median Z-chromosome expression. Expression values are from the adult spleen, and the asterisk indicates that there was a significant difference in expression between the MHM region and the entire Z chromosome before these female-biased genes were excluded.

Together we identified 16 genes that are female biased in one or more of our species and contribute to the MHM region valley of male-specific regulation (Table S6). Specifically, they cluster between 25 and 28 and 31 and 34 Mb on the chicken Z chromosome. Of these 16 genes, two have been identified previously as strongly female biased in the chicken (RFLB and ENSGALG00000018479) (Mank and Ellegren 2009; Nätt et al. 2014), and half underlie the MHM regulatory valley in two or more of our species. Taken together, our results indicate a common architecture of the MHM region regulatory landscape.

Sequence evolution of genes in the MHM region

We used the CODEML package in PAML to identify the contribution of purifying selection acting on Z-linked genes and genes located in the MHM region (at 25–35 Mb). We found that the proportion of MHM genes evolving with a significant contribution of purifying selection was consistently higher than the Z chromosome as a whole, and this difference was significant in three of the six species Table 5. Of these, all the female-biased genes we identified as underlying the distinct sex-specific regulation of the MHM region are evolving under purifying selection, with the exception of one gene in the pheasant (Table S7).

Table 5. Proportion of genes evolving under purifying selection.

Species	MHM region	Z chromosome	Chi-squared statistic	P-value
Goose	1.000	0.882	6.186	0.013
Duck	1.000	0.874	7.254	0.007
Guineafowl	0.959	0.861	3.803	0.051
Peafowl	0.940	0.818	4.754	0.029
Turkey	0.904	0.833	1.767	0.184
Pheasant	0.923	0.841	2.481	0.115

Significant differences in the proportion of genes evolving under purifying selection between the MHM region and Z chromosome are shown in bold.

Functional gene analysis

Of the 16 MHM genes we identified (Table S6), one is known to play a role in sexual differentiation. Previous work has shown that VLDL encodes a receptor that is important for the transport of plasma lipoproteins, yolk synthesis, and avian ovarian follicle development (Barber et al. 1991; Wang et al. 2013; Hu et al. 2014).

Given the strong female-biased expression of these MHM genes in contrast to the male bias of the Z chromosome, we conducted a GOrilla analysis (Eden et al. 2007, 2009) to explicitly test whether there was an enrichment of gene-function terms in this region. We found no significantly enriched gene ontology terms for the 16 MHM genes when compared to the Z chromosome. There was significant enrichment in regulatory-region nucleic acid binding [GO:0001067, P < 0.001, false discovery rate (FDR) q-value = 0.271] for the whole MHM region (at 25–35Mb) when compared to the Z chromosome; however, this was nonsignificant after multiple-testing correction.

Synteny of the avian Z chromosome

The chicken sex chromosomes are syntenic with Anolis chromosome 2 (Alfoeldi et al. 2011; Vicoso et al. 2013). We used MCScanX (Wang et al. 2012) to investigate the fine-scale synteny of the MHM region and surrounding area. Consistent with prior work (Wang et al. 2014), we found evidence of intrachromosomal rearrangement within this region, with evidence of five colinear blocks, syntenic regions where gene order is preserved, spanning approximately 50 Mb on Anolis chromosome 2 (Figure 6). One of these colinear blocks, only 3.3 Mb in length, encompasses nine of the MHM genes responsible for the characteristic MHM gene regulation we identified (Table S6), together with the putative avian sex-determining gene DMRT1.

Figure 6.

Synteny between the chicken Z chromosome and Anolis chromosome 2. Synteny between chicken and Anolis was quantified using MC ScanX. Consistent with previous work, we found that the chicken Z chromosome (GG Z) and MHM region (shown in light gray from 25–35 Mb) are both syntenic with Anolis chromosome 2 (AC 2). We found that the MHM region is comprised of five collinear blocks (shown in blue and green), syntenic regions where gene order is preserved, spanning approximately 50 Mb on Anolis chromosome 2. One of these collinear blocks (shown in blue) encompasses nine of the MHM region genes together with the avian sex-determining gene DMRT1 (in red) (Table S6).

Discussion

Regional variation in sex-specific gene regulation has been observed across the chicken sex chromosomes (Melamed and Arnold 2007; Mank and Ellegren 2009; Wright et al. 2012; Wang et al. 2014). However, the extent to which certain features of regional variation are species specific has not been clear. More importantly, it was not clear whether some of this regional variation was related to dosage compensation. Here we present a comparative analysis of Z-chromosome gene regulation across six species to comprehensively assess conservation of regional differences in sex-specific gene regulation.

Variation in dosage compensation has been documented across the mammalian sex chromosomes (Carrel and Willard 2005), where X inactivation is less complete in the younger portions of the human X chromosome [but see Yang et al. (2010a)]. The MHM region of the chicken Z chromosome exhibits an excess of female-biased genes (Mank and Ellegren 2009), and given the pronounced male bias of the Z chromosome as a whole, it has been proposed as a region of dosage compensation in birds (Melamed and Arnold 2007; Melamed et al. 2009). However, subsequent work on other avian species found no evidence of regional MHM region regulatory variation in the ratites (Wang et al. 2014) or passerines (Itoh et al. 2010; Wolf and Bryk 2011; Uebbing et al. 2013), raising questions about whether the MHM region was confined to the chicken lineage. We found that the MHM region is not a chicken-specific regulatory feature of the Z chromosome but conserved across one of the oldest avian clades, the Galloanserae (Figure 1). Our results are supported by other lines of evidence, including the presence of MHM ncRNA within the turkey genome (Itoh et al. 2011) and hypermethylation across the MHM region in several Galliform species (Teranishi et al. 2001). Further work is required to establish whether the MHM region is present outside the Galloanserae, but preliminary evidence in the plover (Moghadam et al. 2013) may suggest that the regulatory variation in this region is a prominent feature of Z-chromosome evolution across a large portion of the avian phylogeny. Given that the plover lies outside the Galloanserae, this could indicate either that the MHM region regulatory pattern was lost secondarily in the passeriform birds or that it has evolved convergently.

Our results are consistent with previous findings (Mank 2013) that birds lack a complete chromosome-wide mechanism of dosage compensation but do not support a role of the MHM region as an area of nascent dosage compensation (Melamed and Arnold 2007). The classic model of dosage-compensation evolution (Ohno 1967) predicts that dosage-compensation mechanisms arise as a consequence of selection to balance gene dose of the single X or Z chromosome in the heterogametic sex with the two copies of interacting loci on the autosomes. In female heterogametic species such as the birds assessed here, this predicts increased expression in females on the Z chromosome. However, instead of the expected upregulation in females, we observed downregulation of the MHM region in males across all six of our species. We therefore conclude that the distinct gene regulation in the MHM region does not represent dosage compensation as defined by Ohno. However, without knowing the ancestral expression of genes in the MHM region, it is difficult to concretely reject female upregulation if the ancestral baseline expression of the MHM region was lower than the Z chromosome as a whole.

Furthermore, even if dosage compensation evolves by another mechanism and not via the specific set of regulatory steps defined by Ohno, we would still expect equal expression between males and females for most of the genes. Instead, inconsistent with dosage compensation, we found that a small proportion of female-biased Z-linked genes is responsible for the characteristic MHM region valley of male-biased expression. All these genes were located on the ancestral avian proto-sex chromosome, and half are located on the same conserved syntenic block as the major sex-determining gene DMRT1 (Smith et al. 2009). Interestingly, recent work suggests that the MHM region is in the oldest stratum of the avian Z chromosome (Vicoso et al. 2013; Zhou et al. 2014), and the pronounced female-bias expression we observed, together with previous work (Barber et al. 1991; Wang et al. 2013; Hu et al. 2014), may indicate that some of these genes play a role in female fecundity.

Our failure to observe pronounced sex-specific MHM region regulation in the gonad relative to the rest of the Z chromosome might initially appear inconsistent with a role of the MHM region in female reproduction. However, there are many strongly female-biased genes in the MHM gonad, but gonadal expression is typically more variable than somatic tissue expression and may mask distinct gene regulation patterns in this region. Therefore, further empirical work is required to verify the role of the MHM region in sexual differentiation and female reproduction.

Previous work has revealed that male-biased expression accumulates on the chicken Z chromosome over time (Wright et al. 2012), likely a result of cumulative exposure to male-specific selection. Consistent with previous work in the zebra finch and ratites (Wang et al. 2014), we found that male-biased expression is greatest in the older regions of the Z chromosome across all species. Together our findings highlight the complex nature of selective forces driving variation in gene regulation across the avian Z chromosome.

Concluding Remarks

We present a comprehensive comparative analysis of regional variation in sex-specific gene regulation across the avian Z chromosome. We found that male bias accumulates on the avian Z chromosome, where older regions are the most sex biased. Additionally, we showed that the MHM region is not chicken specific and is conserved across the Galloanserae. Our evidence concordantly suggests that this region does not represent regional dosage compensation on the avian sex chromosomes. Instead, the MHM region may play an important role in sexual differentiation and female fecundity.