Sex-biased gene expression at homomorphic sex chromosomes in emus and its implication for sex chromosome evolution

Abstract

Sex chromosomes originate from autosomes. The accumulation of sexually antagonistic mutations on protosex chromosomes selects for a loss of recombination and sets in motion the evolutionary processes generating heteromorphic sex chromosomes. Recombination suppression and differentiation are generally viewed as the default path of sex chromosome evolution, and the occurrence of old, homomorphic sex chromosomes, such as those of ratite birds, has remained a mystery. Here, we analyze the genome and transcriptome of emu (Dromaius novaehollandiae) and confirm that most genes on the sex chromosome are shared between the Z and W. Surprisingly, however, levels of gene expression are generally sex-biased for all sex-linked genes relative to autosomes, including those in the pseudoautosomal region, and the male-bias increases after gonad formation. This expression bias suggests that the emu sex chromosomes have become masculinized, even in the absence of ZW differentiation. Thus, birds may have taken different evolutionary solutions to minimize the deleterious effects imposed by sexually antagonistic mutations: some lineages eliminate recombination along the protosex chromosomes to physically restrict sexually antagonistic alleles to one sex, whereas ratites evolved sex-biased expression to confine the product of a sexually antagonistic allele to the sex it benefits. This difference in conflict resolution may explain the preservation of recombining, homomorphic sex chromosomes in other lineages and illustrates the importance of sexually antagonistic mutations driving the evolution of sex chromosomes.

Sex chromosomes are derived from ordinary autosomes. Restriction of recombination and differentiation of sex chromosomes is often thought to be the ultimate fate of a pair of protosex chromosomes that acquired a sex-determining function (1–3). The evolutionary fuel driving the loss of recombination between nascent sex chromosomes is provided by sexually antagonistic mutations, that is, mutations that benefit one sex but are harmful to the other (4). If a sexually antagonistic mutation arises in the population on the recombining, nondifferentiated part of a protosex chromosome, selection for increased linkage of the sexually antagonistic allele to the sex determination locus can cause this mutation to be transmitted more often through the sex it benefits (5). Loss of recombination between the sex chromosomes allows the proto-X and proto-Y (or proto-Z and proto-W) to evolve independently, and also sets the stage for the population processes causing degeneration of the nonrecombining sex chromosome (6). Restriction of recombination across the entire length of the sex chromosomes and continuous gene decay on the Y results in heteromorphism between the X and Y (or Z and W). Sex chromosomes that show low levels of differentiation are generally considered to be evolutionarily young and only at the initial stages of this path; given time, they too are expected to differentiate.

However, some clades, such a birds or snakes, contain old, homologous sex chromosomes that have progressed to heteromorphism in some species but not others (7–10). The reasons for a lack of differentiation in some lineages are unclear, but several scenarios are possible. Clades might differ in the amount of sexually antagonistic selection operating in the genome. In particular, species experiencing more conflict between the sexes might be subject to higher rates of sexually antagonistic mutations and are thus expected to more rapidly eliminate recombination between their protosex chromosomes. Species experiencing lower levels of sexual antagonism, in contrast, lack the selective agent for recombination suppression and may maintain homomorphic sex chromosomes. Why different clades of birds or snakes would differ in the amount of sexual conflict, however, is unclear.

Recombination suppression is only one strategy for conflict resolution at loci undergoing opposite selection in males and females. In particular, the evolution of sex-biased gene expression is an alternative solution to eliminate the deleterious effects of a sexually antagonistic allele in the sex to which it is harmful (11–13). Thus, sexually antagonistic alleles might have accumulated at a similar rate in species with homo- and heteromorphic sex chromosomes, but different taxa may have used different strategies to resolve sexual conflict. Species could evolve close linkage between sexually antagonistic mutations and the sex determining region, eventually leading to the evolution of heteromorphic sex chromosomes. Alternatively, sex-biased expression might evolve at sexually antagonistic alleles along the protosex chromosomes, thus eliminating the selective pressure to abolish recombination on the nascent sex chromosomes (14). This evolutionary path would result in species with old, homomorphic sex chromosomes, but an excess of sex-biased expression at the recombining portion of their sex chromosomes.

Birds all have homologous sex chromosomes that formed about 120 Mya (15), similar in age to the mammalian sex chromosomes (about 165 My old, ref. 16). However, whereas all mammals and most bird lineages have highly differentiated sex chromosomes, in some groups of birds, such as ratites, the sex chromosomes remain homomorphic (17). The reason for this difference is unclear. Previous mapping studies have shown that the emu Z and W are largely homologous, containing only a small, differentiated segment (18–21). Here, we analyze male and female genomic sequences and transcriptomes of emus at different developmental time stages, in an attempt to identify the evolutionary forces causing these differences.

Results and Discussion

Levels of Differentiation Vary Along the Emu Sex Chromosomes.

Mapping studies have suggested that the emu Z and W chromosomes are largely homomorphic, with only a small, differentiated segment on the short arm of the Z chromosome (18–21). Genomic read coverage information and SNP patterns from male and female samples can be used to identify regions of sex chromosomes that are either fully or partly differentiated, or still recombining and homologous between the Z and W chromosome. Autosomal regions should show similar SNP densities in males and females and similar genomic read coverage in both sexes (i.e., a similar number of genomic reads obtained from male and female samples should map to autosomal fragments, normalized by the total number of reads per sample). Fully differentiated regions (which no longer have a W-linked homolog) are always hemizygous and thus lack SNPs in females but should have normal SNP densities in male samples. Such regions will also show only half as many genomic reads in females, compared with males. Partially differentiated regions will show an increased level of heterozygosity in females because, if W-derived reads still map to the Z, divergence between the W and the Z will be confounded with polymorphism. Such regions should also show reduced genomic coverage in females because only highly similar reads between the Z and the W will contribute to coverage (depending on the exact mapping parameters). Last, pseudoautosomal regions (PARs) are expected to have similar SNP densities and genomic coverage in males and females; i.e., they should be indistinguishable from the autosomes.

To identify differentiated regions along the emu sex chromosomes, we obtained DNA-seq and RNA-seq reads from male and female emu samples, which we assembled into genomic scaffolds and transcript sequences, respectively. We mapped these emu gene sequences along the chicken Z chromosome and used the gene content of the genomic scaffolds to estimate their location on the chicken Z (22) (Fig. 1). Although physical or genetic maps are unavailable for emus, the low rate of rearrangements observed between bird genomes, and reptiles in general, suggests that the location of the genes along the chicken genome may be a good proxy for their location on emu chromosomes (Fig. S1). For example, a large number of microsyntenic blocks are shared among chicken and lizard (23), and the gene content between the chicken and flycatcher Z chromosome is well conserved (24). Therefore, although a few rearrangements have probably occurred since the emu–chicken split, synteny at small scales can serve to group emu genes. Repeating our analysis of emu expression, but mapping of emu genes along Anolis chromosome 2 (which is homologous to the bird Z chromosome) to identify differentiated and pseudoautosomal regions, does not affect our conclusions (Figs. S1–S3).

Fig. 1.

Evolutionary strata on the emu sex chromosomes. (A) Female and male coverage along the Z chromosome (dots represent individual scaffolds, the lines a sliding window of 10 scaffolds). (B) Female/male SNP densities for emu transcripts, ordered along the chicken Z chromosome, using a sliding window of 10 genes. Putative differentiated regions are colored in yellow (Str0: stratum 0 and Str1: stratum 1), and pseudoautosomal regions are in green (PAR). The red vertical line indicates the location of dmrt1.

We mapped the female and male genomic reads separately to the genomic scaffolds and estimated male and female coverage of each scaffold from the resulting SAM alignment, after removal of reads with mismatches. The male and female SNP density of each gene was estimated by mapping RNA-seq reads back to the gene sequences with Bowtie and calling SNPs when two alleles were present at high frequency at a particular site. The genomic read coverage and levels of polymorphism in male and female emu were plotted along the chicken Z chromosome, using a sliding window of 10 scaffolds or genes (Fig. 1). Using this approach, we observed several candidate regions for differentiation on the emu Z; lists of genes found in each region are provided in Dataset S1. Specifically, we identify three candidate regions along the chicken Z that appear fully differentiated in emus: genomic read coverage for these regions is similar to autosomal regions in males, whereas females show only half the read coverage; further, few SNPs were observed within each female sample whereas the male samples showed levels of heterozygosity similar to other regions of the Z. These three regions (26471499–34336577, 50347741–55266322, and 62989377–69684220) presumably correspond to a single ancestral sex-determining region on the bird sex chromosomes, termed stratum 0, that has been broken into several smaller segments on the chicken Z by chromosomal rearrangements. An ancestral involvement of this genomic segment in sex determination in all birds is further corroborated by the location of dmrt1 [doublesex and mab-3 related transcription factor 1; presumably the ancestral sex-determining gene in birds (25)] within that segment, their adjacent location along Anolis chromosome 2 (Fig. 2), and a lack of W-homologous genes within this presumably ancient sex-linked region in chicken (Fig. 2 and see Dmrt1 and Evolutionary Strata in Birds). Another region, which we call emu stratum 1 (34336577–39235675), also showed reduced genomic coverage in the female sample, but highly increased heterozygosity in females relative to males (Fig. 1). This pattern suggests that this Z-linked region stopped recombining with the W relatively recently, i.e., sufficiently long ago for divergence between the Z and the W to build up (and thus genomic read coverage is reduced in females due to the more stringent mapping of genomic reads and higher divergence in intergenic and intron sequences), but recently enough for sequence similarity to remain high between the chromosomes (and detected as high SNP density between Z and W orthologs in females). The remaining regions along the sex chromosomes showed similar levels of genomic coverage and diversity in the male and female sample and likely correspond to the nondifferentiated, recombining portion of the sex chromosomes. We classified the regions adjacent to the Z-W differentiated regions as pseudoautosomal; here, genomic read coverage and SNP densities are indistinguishable from autosomal values (see Tables S1–S4 for comparisons of SNP densities). Mapping of emu genes to Anolis resulted in a very similar set of genes being classified as differentiated or pseudoautosomal, and the boundaries between the differentiated and pseudoautosomal segments often correspond to boundaries of large syntenic blocks between chicken and Anolis (Fig. 2).

Fig. 2.

(A) Synteny and evolutionary strata on the bird Z. Dot plot between Anolis chromosome 2 and the chicken Z chromosome. Dashed lines indicate the position of Z-linked genes in chicken that also have W-linked homologs; these genes have been used to infer the location of three evolutionary strata on the chicken Z (old strata I–III). Regions in yellow are fully differentiated on the emu Z, and the orange region is a more recently differentiated region. The yellow region supposedly corresponds to the ancestral sex-determining segment shared by all birds but contains no remaining W homologs in chicken and can thus not be dated based on Z-W divergence. We refer to this ancestral sex-linked region as stratum 0. The red line gives the location of dmrt1, the ancestral sex-determining gene in birds. (B) Schematic comparison of the chicken and emu Z chromosomes, with the emu differentiated (in yellow) and pseudoautosomal regions (green) colored on the chicken Z. DMRT1 has been mapped to the differentiated region of the emu Z chromosome using in situ hybridization.

The pseudoautosomal region (PAR) makes up most of the emu sex chromosomes (about 60% of genes according to our classification, versus 25% for stratum 0 and 5% for stratum 1), consistent with the observation that the emu Z and W are of similar size and appear largely undifferentiated under the microscope.

Male-Biased Expression in Pseudoautosomal and Differentiated Regions.

In chicken, the sex chromosomes are fully differentiated; i.e., the W is completely degenerated, and the Z is hemizygous in females (26–28). Transcripts derived from Z–linked genes in chicken are found, on average, at higher levels in males compared with females, supporting the notion that birds lack global mechanisms of dosage compensation (the equalizing of expression from the Z relative to the autosomes in the two sexes) (29). In the emu, most of the Z and W chromosome is pseudoautosomal; therefore, expression of most sex-linked genes in females relative to males (F/M) is expected to be similar to that of the autosomes, and only genes located in the differentiated region should show lower expression in females. Surprisingly, however, we find that genes located on the Z showed an overall reduction in F/M expression at both 15 d and 42 d (Fig. 3A); i.e., expression from the Z was 1.07 (15-d embryos) and 1.33 (42-d embryos) times as high in males compared with females. If degeneration of the W and lack of dosage compensation were driving this pattern in emu, a decreased female over male expression ratio should be largely confined to the differentiated regions (that is, stratum 0 and possibly emu stratum 1). Consistent with a lack of dosage compensation, both the 15-d and the 42-d samples showed reduced F/M expression at the fully differentiated region stratum 0. However, whereas the F/M expression ratio is indeed lowest at stratum 0, the pseudoautosomal region also shows evidence for reduced F/M expression in the 42-d sample (Fig. 3 B and C and Fig. S4), and we find a larger fraction of sex-biased genes in the PAR relative to autosomes (Tables S5–S8). Thus, gene expression from the emu Z is sex-biased, both in the differentiated region as well as in the PAR.

Fig. 3.

Patterns of gene expression in emus. Genes were assigned to different chromosomes according to their location in the chicken genome. (A) Log2(female/male FPKM) for the different emu chromosomes at 15 and 42 d. (B) Log2(female/male FPKM) along the Z chromosome using a sliding window of 10 genes at 15 d and 42 d. (C) Log2(male FPKM/female FPKM) at day 15 and 42 on the autosomes, the differentiated regions (stratum 0 and stratum 1), and the pseudoautosomal region (PAR) of the Z chromosome. The dashed lines indicate the median values for each sample.

Because genes in the PAR are diploid in both sexes, this pattern cannot result from simple gene dose differences. Instead, it suggests that male-biased genes accumulate on sex chromosomes, even in homologous regions that still recombine between the Z and the W. In particular, whereas a lack of dosage compensation should be detectable at any developmental stage, sex-biased expression of genes (other than the sex determination genes) should be more pronounced in older embryos, after the full development of gonads (30). Consistent with more sex-specific or sexually antagonistic selection operating at later stages in development, the variance in F/M expression among autosomal genes increased from day 15–42 (P value < 2.2e−16, F test), as did the fraction of sex-biased genes (7–11% using a twofold cutoff; Tables S5–S8). Further, only the 42-d sample shows male-biased expression on the PAR whereas, at 15 d, the F/M ratio is more similar to that of the autosomes (Fig. 3C). Specifically, the distribution of expression ratios of genes on the PAR is bimodal in the 42-d sample, with a second population of more strongly male-biased genes (peaking at M/F ∼1.35; Fig. 3C). Such a bimodal distribution is not found at autosomal genes, or genes within stratum 0, but is also absent at PAR-linked genes at the younger, sexually undifferentiated 15-d-old embryos. This pattern demonstrates that the excess of male-biased expression at PAR genes at 42 d is not caused by the inclusion of Z-linked genes that are differentiated on the W (which should be detected in both temporal samples) but instead is consistent with sex-specific expression of male-beneficial genes after sexual differentiation. Overall, male-biased expression at 42 d seems to be mostly due to a reduction in expression from the PAR in females, rather than an increase in males (Fig. S5). This finding is consistent with a scenario whereby the sex to which expression of a sexually antagonistic allele is detrimental reduces the deleterious consequences of such mutations through transcriptional down-regulation. The differences in F/M expression between early and late embryonic development are significantly greater in the PAR than those observed for the autosomes (P = 8.509e−09; Fig. S6) whereas the difference in F/M between the autosomes and stratum 0/stratum1 is not significant (but follows the same trend). In summary, we observe a distinct shift toward male-biased expression in the pseudoautosomal region during development, in contrast to the autosomes (where average sex bias does not change) or in stratum 0 (where genes are always male-biased, presumably due to a lack of dosage compensation). Stratum 1 contains only 26 genes but probably constitutes a mixture of genes where the W-linked ortholog is either expressed or degenerated. This mixture in genes could contribute to the apparent (but not significant) shift in the distribution of expression ratios in that region (Fig. 3C and Fig. S4).

Interestingly, we also observe a slight, yet significant, shift in the distribution of F/M expression ratios at PAR-linked genes at the 15-d sample toward female-biased expression (Fig. 3C and Fig. S4). This pattern suggests that female-biased genes accumulate in the PAR during early development, at the onset of sexual differentiation (day 15) whereas, at later stages, male-biased expression is more abundant (day 42). A shift in sex-biased gene expression over the life cycle has also been observed in chicken gonads, with the relative proportion of male-biased genes increasing over development (30).

Homomorphic vs. Heteromorphic Sex Chromosomes, and Sex-Biased Expression.

Why is the pseudoautosomal region different compared with the autosomes in emu, and how does this difference relate to sex chromosome evolution? Whereas pseudoautosomal regions are not confined to a single sex, even partial linkage to a sex determination locus can select for sexually antagonistic alleles to accumulate (14, 31, 32). For example, an excess of sexually antagonistic genes mapped to the pseudoautosomal region in the dioecious plant Silene latifolia (13, 33). Sexual conflict resulting from the accumulation of sexually antagonistic mutations within the PAR can be resolved either by restricting recombination between the sex chromosomes (i.e., restricting a sexually antagonistic mutation to a particular sex) or by down-regulating the expression of genes in the sex that they harm. In emu, male(female)-beneficial alleles might have accumulated at pseudoautosomal genes, and their deleterious effects to females (males) may have been resolved by down-regulating them in the sex they harm. This downregulation in turn would eliminate selective pressure for reduced recombination between the Z and W; additionally, most of the genes on the emu W chromosome are not subject to deleterious effects caused by loss of recombination. Sex-biased expression is viewed as the resolution of sexual antagonism at the molecular level, and our results suggest that the evolution of sex-biased expression could be a general mechanism to resolve conflicts occurring at sex chromosomes and could explain the stability of homomorphic, recombining sex chromosomes in some clades. The maintenance of homomorphic, nonrecombining sex chromosomes could also result from rare recombination in sex-reversed ZY male or XY females, as has occurred in tree frogs (34), but does not explain the lack of a restriction of recombination observed in ratites.

Why would different strategies to resolve sexual antagonism be adopted in different clades? As mentioned in the introduction, one factor likely to be paramount in driving divergence at nascent sex chromosomes is the extent of sexually antagonistic selection on males and females. Sexual selection is thought to be more prevalent in males than in females, as females (who usually invest more in the progeny) are more selective when choosing mates. Male-heterogametic clades, which can fully link male-beneficial/female-deleterious mutations to males by abolishing recombination on the Y, may therefore be more prone to suppressing recombination between the sex chromosomes; ZW species can achieve only partial linkage of these mutations to males by abolishing recombination and may obtain greater benefits from reducing their expression in females. The fact that both snakes and birds, the two major ZW clades that have been studied in great detail, show ancestral nondegenerated W chromosomes provides some support for this idea, but the generality of such an association needs to be confirmed on a larger scale. The great diversity of mating systems that occurs within birds can also lead to drastic differences in the occurrence of sexual antagonism: very little opportunity for sexual selection is expected in monogamous species with low frequencies of extrabond fertilization, whereas the opposite is true in polygamous/promiscuous species. Consistent with sexual antagonism differing between lineages, mating systems correlate with the extent of body sex-dimorphism in birds (35). It is therefore possible that differences in the strength of sexual selection may lead to different optimal strategies to resolve sexual antagonism in different lineages of birds. Whereas our current knowledge of the degeneration of the bird W chromosome is limited to a few species, the application of next-generation sequencing technologies to a variety of species with different mating systems will provide a unique opportunity to investigate the importance of sexual antagonism in the degeneration of W chromosomes and to disentangle it from other forces driving the evolution of sex chromosomes.

Dmrt1 and Evolutionary Strata in Birds.

Comparative analysis between Anolis and chicken, combined with patterns of differentiation along the emu Z, allows us to reconstruct, to some extend, the early evolutionary history of differentiation and strata formation on the bird Z chromosome. In chicken, dmrt1 has been identified as the master-switch sex-determining gene (25); dmrt1 is present on the Z, but absent on the W, and is believed to trigger sexual development in a dose-dependent fashion. dmrt1 has also been mapped to the differentiated region on the Z chromosome of ratite birds and is absent on the ratite W (21), further supporting its role as the master sex-determining switch gene in all birds.

In chicken, a limited number of W-linked genes with homologs on the Z have been identified and are remnants of the ancestral genes present on the autosome that formed the sex chromosome. The level of differentiation between Z and W homologs, however, is not uniform along the Z; instead, different gene pairs show discontinuous divergence levels, which are thought to represent different time points at which recombination ceased between the Z and W (36, 37). Genes with similar sequence divergence are grouped together to form “evolutionary strata,” and, in chicken, at least three such strata can be identified (36, 37). If dmrt1 is indeed the ancestral sex-determining gene that initiated sex chromosome evolution in the ancestor of all birds, it should be part of the oldest evolutionary stratum. However, the three evolutionary strata detected on the chicken Z are all reported to have formed after the ratite–chicken split (37), and dmrt1 is actually not found within the oldest stratum of chicken. Instead, it is located in the middle of stratum II, a region that ceased recombination only 50–70 Mya, long after the split of chicken and ratites. The placement of dmrt1 within a more recent addition to the nonrecombining region of the chicken sex chromosomes is not consistent with its putative role as the ancestral sex-determining gene shared by all birds; it is also counter to mapping studies showing Z-linkage of dmrt1 in ratites (which lack the three strata identified in chicken). The boundaries of stratum II, however, are defined only by a few Z-W gametologs, and a large region within that stratum (ranging from roughly 17 Mb to 39 Mb) completely lacks any W genes (36, 37); dmrt1 is located right in the middle of this region (at roughly 26 Mb on the chicken Z; Fig. 2). Thus, together these observations suggest that there may be an undetected, even older evolutionary stratum present on all bird sex chromosomes that contains the dmrt1 gene, but is lacking any W homologs (and thus could not be detected based on ZW divergence). Indeed, our genomic coverage analysis in emus suggests that part of the ancestral, differentiated fragment (stratum 0) on the bird sex chromosomes corresponds to position 26–33 Mb along the chicken Z, the region that contains the dmrt1 gene (Fig. 2). Thus, stratum 0 corresponds to the very initial differentiated fragment on the bird sex chromosomes and contains the master-switch sex-determining gene of birds. Additional genomic analysis of the sex chromosomes of other birds, and in particular other ratites, should allow us to further refine the evolutionary history of the avian sex chromosomes.

Materials and Methods

Sample Preparation.

Fertilized emu eggs were incubated at 37 °C for 15 and 42 d, and embryo heads (for 15-d embryos) and brains (42-d embryos) from sexed embryos were used for RNA extraction. DNA was extracted from the sexed 42 d embryos. Paired-end RNA-seq libraries (with a 200-bp fragment size) and DNA-seq libraries (with an 800-bp fragment size) were made following Illumina’s standard protocol and sequenced on the Illumina platform. The number of reads obtained for each sample is listed in Table S9.

Genomic Assembly and Read Coverage Estimation.

Female and male genomic reads were trimmed, pooled, and assembled using SOAPdenovo (http://soap.genomics.org.cn/soapdenovo.html). Female and male reads were mapped separately back to the genome assembly using Bowtie2 (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml), and SoapCov (http://soap.genomics.org.cn/soapaligner.html) was used to estimate the female and male coverage for each scaffold larger than 1 kb.

Mapping of Emu Genomic Scaffolds to the Chicken Z.

We mapped all known chicken gene sequences (www.ensembl.org) to the emu genomic assembly using blat (38) with a translated query and a translated database using a reciprocal best hit approach. Emu scaffolds were assigned chicken chromosome coordinates based on gene content.

Processing of the RNA-seq Reads and Transcriptome Assembly.

All 43.9 million RNA-seq reads were trimmed, pooled, and assembled with SOAPdenovo-Trans (http://soap.genomics.org.cn/SOAPdenovo-Trans.html), and only transcripts of at least 400 bp were kept for further analysis. Chicken protein sequences were downloaded from the Ensembl ftp server, and emu transcripts were mapped against the chicken protein sequences using blastx (39), and further scaffolded according to their position along the chicken protein sequences. Our resulting transcriptome consists of 10,102 genes, with a median length of 1,653 bp.

SNP Analysis.

RNA-seq reads from each individual sample were mapped separately to the assembled transcriptome using Bowtie (http://bowtie-bio.sourceforge.net/index.shtml) with default parameters. SNPs were called (within each sample) when two alleles were present at a frequency higher than 0.3, considering only sites covered by at least nine reads.

Identification of Pseudoautosomal and Differentiated Regions.

The female coverage for different genomic scaffolds along the Z chromosome is bimodal (Fig. S7) and was used to find a cutoff value to define differentiated regions. Sequential windows (window size of 10 scaffolds) with Log2(Female coverage) >1.2 were considered to be part of the PAR. Regions between these PAR segments were considered to be part of stratum 0 if they had female:male SNP density below 4, or stratum 1 if they had increased female SNP density (Table S10). A similar classification of genes was obtained if emu genes were mapped against the zebra finch Z chromosome (Fig. S8).

Expression Analysis.

RNA-seq reads from each individual were mapped separately to the assembled transcriptome using Bowtie2, and Cufflinks (http://cufflinks.cbcb.umd.edu/) was used to estimate expression levels [in fragments per kilobase of transcript per million mapped reads (FPKM)].