Introduction

Improving our understanding of the process of speciation is a central problem in biology. Speciation requires that reproductive barriers arise to impede free gene flow among nascent species. Among sexually reproducing organisms, interspecies hybrid male sterility is a commonly observed postzygotic isolation barrier. Species which are separated by partial reproductive barriers, such as those that produce only sterile male hybrids, are ideal candidates for identifying changes associated with speciation. Evolutionary geneticists have used a myriad of approaches to establish associations between speciation phenotypes and gene variants1. During the last ten years, a series of studies have used genome-wide (microarrays) or gene-specific (qRT-PCR) approaches to compare levels of gene expression in sterile hybrids to fertile parental species. These studies have reported significant misexpression, particularly under expression, in sterile Drosophila hybrids relative to parental species for genes of spermatogenesis2,3,4,5,6. In particular, microarray studies have found an overrepresentation of genes that act on the final steps of sperm individualization and maturation (i.e. spermiogenesis)2,4. Such observations, given that during spermatogenesis the majority of transcripts accumulate premeiotically, have lent support for the sterility hypothesis which suggests that down regulation of postmeiotic spermatogenesis genes in sterile hybrids is a causative factor of hybrid male sterility3.

The use of species pairs in which male hybrids are sterile regardless of the direction of the cross cannot distinguish whether misregulation of gene expression is a condition linked to sterility or a byproduct of incompatibilities between divergent regulatory elements brought together in a hybrid genome (i.e. fast male regulatory divergence). A way around this problem has been the use of backcrosses to generate both fertile and sterile partial hybrids. Studies using this approach have found support for misregulation linked to sterility3,7 as well as fast male regulatory divergence8,9. Another approach is to use species pairs that produce unidirectional sterility to compare gene expression between F1 hybrids that are sterile or fertile. Using this method, we have recently shown support for both the sterility and the fast male hypotheses10. However, a surprising result was that out of the 13 spermatogenesis genes targeted, none displayed misexpression in hybrids of the D. p. pseudoobscura (D. p. pseudoobscura) and D. p. bogotana (D. p. bogotana) cross.

Here we use RNA sequencing in an effort to detect genome-wide differences in regulation associated with sterility in hybrids between D. p. pseudoobscura and D. p. bogotana, where only hybrids produced by D. p. bogotana females are sterile (i.e. unidirectional sterility). The two species are geographically separated, with D. p. pseudoobscura found across North America and D. p. bogotana restricted to Colombia in South America. The species have diverged about 0.2–0.25 Myr ago, with estimates of genome-wide intergenic sequence divergence per 500 kb ranging approximately between 0.002 and 0.0211,12.

We found a larger proportion of misregulated genes associated with the sterile than the fertile hybrid condition. Most genes uniquely misregulated in the sterile hybrids were under or over expressed (i.e. transgressive) relative to expression in the parental species, with similar proportions of both over and under expressed genes. In contrast, a larger proportion of genes in the fertile hybrid were expressed additively. An analysis of allele-specific expression differences between species revealed that divergence in regulation is dominated by cis-only and cis-trans effects driving allelic expression in the same direction. Allelic expression in both hybrids was overall positively correlated, but a subset of genes displayed significant changes in expression ratios between hybrids. These genes showed within-chromosome distribution bias and we highlight a potentially interesting region within the third chromosome. Finally, we did not find significant differences in the proportion of sperm/spermatogenesis genes uniquely misregulated in sterile and fertile hybrids, but we observed an overrepresentation of putative proteases. Interestingly, three such peptidases (GA21772, GA24794, GA24796) and one peptidase inhibitor (GA15722) were located within a previously mapped hybrid male sterility QTL13.

Results

Differential gene expression between fertile and sterile conditions

We generated over 400 million individual reads from all samples combined. The raw data has been deposited at the Sequences Read Archive (SRA; GenBank) under accession number SRA205377. Slightly more than 310 million reads were mapped to the available D. p. pseudoobscura reference genome (Table 1). This has the risk of mapping bias against D. p. bogotana derived reads, due to sequence divergence between the species. Three lines of evidence showed that this is unlikely. Firstly, D. p. bogotana had a similar percentage of total RNA sequence reads mapped to the reference genome (72.8% for D. p. pseudoobscura vs. 74.0% D. p. bogotana) (Table 1). Secondly, if the sequence reads were preferentially mapped to the D. p. pseudoobscura reference genome, there would be an overrepresentation of genes with higher expression in D. p. pseudoobscura than in D. p. bogotana, which was not observed (6,839 and 7,075 gene IDs with higher expression in either D. p. pseudoobscura or D. p. bogotana, respectively) and this pattern held when we only included genes that were significantly differentially expressed between the two parental species (1,014 vs. 1,025) (Supplementary Table S1). Finally, the SNP analysis also shows no bias towards one parental species with more gene IDs with higher expression (5,296 vs. 4,808) (binomial test, P = 1.0) (Supplementary Table S2).

Table 1 Sequencing depth per condition, before and after adaptor trimming.
Full size table

Differential gene expression assays were conducted on gene IDs with a minimum of 10 mapped reads in at least one sample (13,928 gene IDs, Supplementary Table S1). Seventy seven percent of gene IDs showed no pairwise differences in gene expression (Supplementary Table S1). Of the remaining 23% of genes showing at least one significant pairwise difference, 2,039 (14.6%) were differentially expressed between the two parental species. Comparison of expression levels between hybrids and parental species showed significantly more gene misregulation in sterile (325) than fertile hybrids (220) (Z = 6.36; P < 0.001). The sterile hybrids also showed a higher proportion of transgressive misregulation (over and under expression) (Z = 12.7; P < 0.001) and lower proportion of additive expression (Z = 3.1; P = 0.002) than the fertile hybrids relative to parental species (Fig. 1). In the context of hybrid male sterility, misregulation unique to each hybrid is informative for the purpose of identifying potential sterility-linked functional groups, pathways, or modes of misregulation. There were significantly more gene IDs uniquely misexpressed in the sterile hybrid (164) than in the fertile hybrid (64) (Z = 9.37; P < 0.001). The fertile hybrid had a significantly larger proportion of additive genes (34) than the sterile hybrid (17) (Z = 3.37; P < 0.001) while the sterile hybrid showed significantly more genes (147) with uniquely transgressive expression than the fertile hybrids (30) (Z = 12.44; P < 0.001) (Supplementary Table S3a).

Figure 1
figure 1

Genes with significant differences in expression between parental species and hybrids.

H = hybrids; P = D. p. pseudoobscura; B = D. p. bogotana. Fertile and sterile hybrid data are represented with green and red circles respectively. Log2 ratio values are capped at 10. The number of genes is shown in each quadrant and partitioned into fertile and sterile hybrids respectively. The upper right and lower left quadrants show significantly over and under expressed genes (transgressive) with the two other quadrants indicating additive effects.

Full size image

We identified inferred biological and molecular functions of genes uniquely misregulated in sterile hybrids. There was an overrepresentation of the biological process ‘proteolysis’ (13 out of 44 genes) (Fig. 2) as well as the molecular function ‘peptidase/ peptidase inhibitor’ (18 out of 57) relative to other classes (Supplementary Table S3b). The large number of proteases among misregulated genes might be a condition linked to the hybrid nature of the genome. However, fertile hybrids had only 1 out of 30 (3.3%) misregulated genes with an identifiable peptidase domain compared to 20 out of 87 (23.0%) in the sterile hybrid (Supplementary Tables S3b,c). This is a significant overrepresentation of genes with a peptidase domain in sterile relative to fertile hybrids (Z = 2.42; P = 0.016). Using D. melanogaster orthologs of the genes uniquely misregulated in sterile and fertile hybrids, we searched for those with a role in spermatogenesis or that have been identified as expressed in the sperm proteome. We found the same proportion of sperm/spermatogenesis genes (12%) among uniquely misregulated genes in sterile and fertile hybrids (Z = 0.022; P = 0.984) (Supplementary Table S4).

Figure 2
figure 2

Proportion of uniquely misregulated genes per biological process in the sterile hybrid (A) and the genome-wide expected proportions of each biological process (B).

Both proteases and cell adhesive genes were proportionally enriched.

Full size image

Chromosomal distribution of misregulated genes in sterile hybrids

The chromosomal distribution of uniquely misregulated genes between fertile and sterile hybrids showed no bias (Table 2). A previous study identified four X-linked, one second and one third chromosome QTLs of both small and large effects on binary (none vs progeny produced) hybrid male sterility between D. p. pseudoobscura and D. p. bogotana13. We looked for genes uniquely misexpressed in the sterile hybrid that mapped within these QTLs. We identified one gene within a small effect X-chromosome QTL, four genes within the large effect second chromosome QTL and fourteen genes within the small effect third chromosome QTL (Table 3). We checked these genes for the presence of domains that might indicate involvement in reproduction-related biological processes or molecular functions (Table 3). Two processes which appear more than once are particularly interesting. Two genes (GA17404 and GA20821) have cell adhesion domains. Cell adhesion was proportionally enriched among misregulated genes in sterile hybrids (Fig. 2) and the D. melanogaster orthologs of GA17404 and GA20821 have high to very high levels of expression in the male accessory glands (http://flybase.org/). Accessory gland expressed proteins are known to trigger a variety of physiological responses in females after mating and the presence of cell adhesion domains is interesting in the context of possible male × female interactions. Three genes are peptidases and one is a peptidase inhibitor (Table 3). This class, as mentioned earlier, is overrepresented in sterile relative to fertile male hybrids. The fact that these four genes mapped to regions previously identified as significant in terms of hybrid male sterility further suggests a major role of previously unsuspected peptidases contributing to reproductive isolation between D. p. pseudoobscura and D. p. bogotana.

Table 2 Chromosomal distribution of uniquely misexpressed genes in sterile and fertile hybrids.
Full size table
Table 3 Genes uniquely misexpressed in the sterile hybrid within previously mapped hybrid male sterility loci (Phadnis et al. 2011).
Full size table

A previous study identified Overdrive (Ovd) as a major sterility gene for this species pair14. The D. p. bogotana X-linked allele in a hybrid genome background causes hybrid sterility, though the presence of other species-specific alleles at a number of other loci is also necessary13. This gene was not misexpressed in sterile or fertile hybrids in this study, though it was significantly differentially expressed between the two parental species.

The contributions of cis and trans regulatory differences to divergence between species

Using SNP information, we assigned RNA sequence reads in hybrids to their parent of origin for 6,530 autosomal genes and 3,574 X-linked genes. In order to determine patterns of cis and trans regulatory divergence between species, we compared relative expression between parental species to relative allelic expression in the hybrids. For autosomal genes, we assayed expression in phenotypically normal (fertile) hybrids but, given that the X chromosome is hemizygous in males, we used information from both hybrids for X-linked genes. For autosomal genes, we found a significant deviation from equal proportions of cis (0.55) and trans effects (0.45) (binomial test, P < 0.001). Cis-only (21%) and cis + trans (17%) effects were overrepresented (Supplementary Table S2a) (Fig. 3A). For X-linked genes, we found a higher proportion of X-only (both cis- and X-linked trans-effects) followed by X-linked & autosomal effects. Though this could indicate a pattern similar to autosomal genes of divergence driven by cis (X-linked) effects, a definitive determination cannot be made by this method, as any diverged regulatory factors on the X cannot be assigned as cis-only or trans-only. (Supplementary Table S2b) (Fig. 3B).

Figure 3
figure 3

Identification of regulatory differences between species showing autosomal (A) and X-linked genes (B).

The genes are further divided based on whether they showed evidence of significant divergence between species driven by loci that act in cis (Cis-only) or trans (Trans-only). Genes with significant evidence of both cis- and trans-regulatory differences were subdivided into “Cis + Trans” when differentially expressed genes between species favored expression of the same allele in the hybrids, “Cis × Trans” when differentially expressed genes favored expression of opposite alleles and “Compensatory” for genes with no significant expression differences between species despite evidence for both cis- and trans-regulatory divergence. The term “Ambiguous” refers to genes with significant expression divergence between species but no significant evidence of cis- or trans-regulatory differences. For X-linked genes the classification is similar except that we could only identify effects as “X-only (Cis & Trans)”, meaning that the differences in expression between species is similar between hybrids despite the shared hybrid nature of their autosomes or “Trans-only (Autosomal)”, meaning that the differences in expression between species was not detected between hybrids due to the fact that one of the hybrids autosomes matches the species origin of the X-chromosome. “X-linked & Autosomal” refers to X-linked genes showing differential expression between species driven by a combination of both X-chromosome and autosomal regulatory elements (see Table 4).

Full size image

In order to detect changes in allelic expression patterns in sterile relative to fertile hybrids, we compared the ratios of allelic expression across autosomal genes (Supplementary Table S2a). We found an overall positive correlation between hybrids (r = 0.798; P < 0.001) indicative of significant genome-wide conservation of allelic expression patterns, with a subset of genes having allelic ratios that significantly differed between hybrids (24% at q < 0.5% and 17% at q < 0.1%). Particularly striking are ten genes with the largest reversal in allelic expression between the two hybrids, with eight genes favoring the allele matching the X-chromosome genotype (Fig. 4). Seven of these eight have a mode of evolutionary divergence involving trans effects (i.e. compensatory and cis + trans).

Figure 4
figure 4

The ratio of parental allelic expression for each autosomal gene in fertile and sterile hybrids.

Genes with the largest reversal in allelic expression between the two hybrids are indicated with a red circle.

Full size image

It is possible that large differences in allelic expression ratios between hybrids contribute to impaired fertility. We explored the intra- and inter-chromosomal distribution of sixty five genes in the top 1% of genes with the largest fold-change in allelic ratio between hybrids (at least a nine-fold change in allelic expression ratio) as well as their possible biological function. We detected no overrepresentation among these genes in terms of their GO biological process / molecular function, or the presence of particular polypeptide domains (Supplementary Table S3d). The proportion of these genes per million base pairs did not differ among the second, third and fourth chromosomes (0.87, 0.90 and 0.73 respectively) (Z = 1.56; P = 0.119 for third and fourth chromosomes comparisons). However, the distribution within the two fully assembled chromosomes (i.e. second and third) is not random (Fig. 5). Of the ten genes with the largest reversal in allelic expression between the two hybrids (Fig. 4), five genes are on the third chromosome, three of which are clustered within a 104 Kb chromosomal region (Fig. 5).

Figure 5
figure 5

Absolute differences in allelic expression ratio between sterile and fertile hybrids.

Scatter plot of second (A) and third (B) chromosome allelic ratio differences per chromosomal position. Filled black circles are the 1% of genes with the highest ratio differences. Genes with the largest reversal in allelic expression between the two hybrids are shown in red.

Full size image

Discussion

We have used RNA sequencing to map genome-wide differences in male reproductive tract expression between two closely related species of Drosophila and their hybrids. We chose the D. p. pseudoobscura and D. p. bogotana pair as they produce unidirectional hybrid male sterility10, thus allowing us to identify changes unique to the hybrid sterile condition.

Among genes that were misregulated only in the hybrids, four findings are of particular relevance: First, we found more uniquely misregulated genes in the sterile than the fertile condition, with a large proportion of genes displaying levels of expression outside the range of their parental species (i.e. transgressive expression). The deviation from additive effects among genes misregulated only in sterile hybrids is indicative of defective gene regulation and this non-additive pattern is common among male-biased genes15,16,17. We found no evidence of an overrepresentation of down-regulation, as has been common for sterile hybrid males of other species of Drosophila4,16,18,19. The lack of overrepresented under-expressed genes cannot be explained by differences in divergence time between species pairs compared, as both species pairs within the simulans clade and the species pair we studied have diverged for approximately 0.25 Myr, while the D. santomea × D. yakuba pair have diverged for 0.4 Myr11,20,21. Instead, differences in the proportion of over- and under-expressed genes might be explained by the severity of sterility. Sterile male hybrids between species of Drosophila previously analyzed are at best unable to produce individualized sperm22,23, whereas sterile males between D. p. pseudoobscura and D. p. bogotana produce amotile individualized sperm cells10. Thus, it is possible to speculate that over expression is more frequent in sterile hybrids with a less severe sterility condition. In fact, Llopart19 found that the sterile D. yakuba × D. santomea hybrids that produced spermatids but no sperm have more overexpressed genes than the reciprocal hybrid with normal testis but no spermatids.

Second, we found no overrepresentation of sperm or spermatogenesis genes among those uniquely misregulated in the sterile hybrids, contrary to previous studies2,4. This is rather unexpected given that male reproductive genes in general and spermatogenesis genes in particular, have been shown to rapidly diverge at both coding and regulatory regions between species of Drosophila2,24,25,26,27,28,29. Once again, the lack of an overrepresentation of misregulated sperm/spermatogenesis genes might reflect the lack of spermatogenic developmental arrest in the D. p. bogotana × D. p. pseudoobscura sterile hybrid. This raises the question as to whether any other gene classes are uniquely misregulated in these sterile hybrids.

This brings us to our third point, our finding of significant overrepresentation of proteases among uniquely misexpressed genes in sterile hybrids. Misregulation of male reproductively expressed proteases has not been previously identified as a possible contributor to hybrid male sterility in Drosophila. The result is intriguing given recent studies on the reproductive role of male accessory gland proteases and protease inhibitors, as well as proteolytically active genes within the female reproductive tract of D. melanogaster30. Proteases are abundant throughout the male reproductive system, as well as within female reproductive tracts31,32 and while specific functions are not always known, several with fertility effects have been identified. The seminal fluid protease ‘seminase’ plays a key role in post-mating success, by causing ‘sex peptide’ to localize to sperm and by activating other proteins resulting in the gradual release of sperm from female storage, ovulation and stimulation of egg laying33. Proteases also seem to be required for the acquisition of sperm motility in some species34,35. Male reproductive proteolytic proteins show evidence of rapid evolution; in a study comparing D. mojavensis and D. melanogaster, many genes showed species-specific proteolytic/protease inhibitor functions in the seminal fluid36. Similarly, the misexpressed proteases in our sterile hybrids may be specific to the D. pseudoobscura lineage - though one of our misexpressed genes, GA14907, is orthologous to a known sperm protein (S-Lap 5) in D. melanogaster, which is upregulated during meiotic and post-meiotic stages of sperm development37. The misexpression of proteolytic proteins in hybrids could contribute to sterility via tissue damage, especially in the case of overexpression. Of the 17 uniquely misexpressed proteases in the sterile hybrid, 13 were overexpressed and 3 of the 4 protease inhibitors were underexpressed. Misexpression of proteases can also contribute to sterility through disruption of important pathways in the male reproductive system, or improper post-mating interactions with female proteins30,33. We propose that male reproductive tract proteases and inhibitors might have evolved under different species-specific selective pressures between D. p. pseudoobscura and D. p. bogotana, in concert with their respective species-specific female tract proteolytic/inhibitor genes, resulting in hybrid regulatory incompatibilities that contribute to sterility. Interestingly, a group of four genes with peptidase/ peptidase inhibitor domains were within previously mapped hybrid male sterility loci13. These genes should serve as future candidates for gene-phenotype association studies.

Fourthly, we do not find an overrepresentation of uniquely misregulated X-linked genes in the sterile compared to the fertile hybrid and the proportion of misregulated genes was slightly lower in the X chromosome than the autosomes. This indicates that gene-specific misregulatory effects have not accumulated on this chromosome per se. However the previously identified ‘large X effect’ explanation for hybrid male sterility in the D. p. pseudoobscura × D. p. bogotana cross13,38 could be a consequence of X-linked trans-regulatory elements contributing towards male sterility. Ovd, the only identified hybrid male sterility gene in crosses between D. p. pseudoobscura and D. p. bogotana, is an X-linked gene that contains a DNA binding (MADF) domain and has seven fixed non-synonymous differences between the two species14. These fixed protein differences together with the presence of a DNA binding domain and our finding of differential expression between parental species suggests that Ovd may contribute to sterility by acting as a transcription factor that causes the misregulation of downstream genes. A potentially disproportional effect of X-linked trans regulatory gene divergence is also hinted at by the fact that eight out of ten autosomal genes with the largest reversal in allelic expression between hybrids favored the allele matching the X-chromosome genotype.

Our results also provide a first glance at genome-wide regulatory divergence between these two species. We found a preponderance of cis rather than trans divergence. This is in line with several studies showing a high proportion of interspecies cis regulatory divergence39,40,41. The dominant pattern of cis-only regulatory divergence between the two D. pseudoobscura species is particularly interesting, as cis-only changes affect mostly the expression of genes at terminal steps of regulatory networks or genes not involved in large interactive networks42. A study comparing trans and cis regulatory changes within D. pseudoobscura using females had found more trans than cis mutations within species43. This is expected as there are more genome-wide trans than cis mutational targets, but because trans acting mutations can affect the expression of many genes, selection would tend to eliminate the accumulation of such mutations over time. This form of purifying selection eliminating trans mutations would be detectable among closely-related species, like the species pair in our study. We also detect a higher genome-wide proportion of cis + trans (16.6%) than cis x trans antagonistic (8%) divergence. This could be explained by directional selection between species favoring alleles with additive cis-only and agonistic cis + trans effects17,18,44. The paucity of uniquely misregulated transgressively expressed genes in sterile hybrids also aligns with the paucity of cis x trans antagonistic regulatory divergence (8%). Finally, despite a similar pattern of allelic expression between sterile and fertile hybrids, some individual genes show large differences in expression. One particular region of the third chromosome has three closely mapped genes with very high fold changes in allelic ratios between hybrids. One gene has a spindle associated domain and another a proteasome complex domain. These domains have been associated with sperm differentiation function45,46 and although it is premature to speculate, the chromosomal proximity between genes, the presence of potential sperm differentiation genes among them and the large change in allelic expression ratio between hybrids highlights these genes as interesting targets for future gene function analysis in these species.

Methods

Species selection and stock maintenance

Stocks used for this study were obtained from the UCSD Drosophila Stock Centre (https://stockcenter.ucsd.edu/): D. p. pseudoobscura (14011–0121.139) and D. p. bogotana (14011-0121.175). Flies were maintained on cornmeal–molasses–yeast–agar (CMYA) medium at constant temperature (24 °C) on a 12 hour light–dark cycle. Virgin females were collected post-eclosion and flies were mass crossed in bottles containing CMYA medium. Reciprocal crosses were used to create hybrids.

Sample Preparation and Sequencing

Total RNA was extracted from reproductive tracts (testes, accessory glands and ejaculatory bulb) using the Qiagen RNeasy Plus kit. Two biological replicates were obtained from parental species and hybrids, each sample consisting of 30 to 40 male reproductive tracts. RNA samples were tested for quality using an Agilent Bioanalyzer and quantified using a nanodrop (Thermo Scientific). RNA samples were sent to the McGill University and Génome Québec Innovation Centre for library preparation and sequencing (http://gqinnovationcenter.com/). Briefly, cDNA libraries were prepared using Illumina’s TruSeq Stranded mRNA sample preparation kit from 250 ng of total RNA, followed by 100 bp paired-end sequencing. All eight samples were run multiplexed on a single lane of an Illumina HiSeq2000 machine.

Mapping and Differential Expression Analysis

After sequencing, reads were adaptor-trimmed using the Trimmomatic program47. Any paired reads in which either the forward or reverse read was shorter than 50 bp after trimming were discarded. The Tuxedo tool suite was used for gene expression analysis, following steps previously outlined but excluding the final CummeRbund step48. Tools were run via the Galaxy platform (https://usegalaxy.org/). Because there is no sequenced genome for D. p. bogotana, reads from all samples were aligned to version 3.1 of the D. p. pseudoobscura genome from Flybase (http://flybase.org/) using the short-read gap-junction aligner TopHat48. A maximum of 8 mismatches to the genome were allowed during mapping (Supplementary information).

Aligned reads were assigned to a gene of origin by Cufflinks, using the D. p. pseudoobscura V3.1 annotation as a guide for transcript assembly. The annotation was filtered to include only genes annotated as ‘Flybase’. Pair-wise differential expression testing between each parental species and reciprocal hybrid was conducted using Cuffdiff. Cuffdiff reports expression levels as fragments per kilobase of exon per million mapped reads (FPKM), which controls for gene length and per-sample sequencing depth. Cuffdiff conducts significance testing using a beta negative binomial distribution, At least one sample with ten mapped reads was required for significance testing, with FDR correction set at 0.05. We used the FDR corrected pairwise comparisons from CuffDiff to identify genes differentially expressed between the two parental species and hybrids. Genes differentially expressed in the hybrids were classified as additive or transgressive (i.e. genes with expression higher or lower than the parental species).

Genes misexpressed in hybrids were assigned functional classifications using Gene Ontology biological processes, molecular function and identification of polypeptide domains (Interpro) within Flybase (http://flybase.org).

Allele-Specific Gene Expression

The relative contribution of cis and trans interspecies divergence on gene expression was inferred using species-specific SNPs and relative allelic expression in the F1 fertile hybrid18,39. We utilized expression data from the fertile F1 male hybrids to avoid condition-specific (sterility) effects in assaying overall regulatory divergence.

SNPs identified between the two parental species using Naïve variant caller and Variant annotator49 were considered fixed if each parental species was represented by a different single allele, with at least 3 reads supporting each parent. RNA-seq reads in the hybrid samples were assigned to a parent of origin based on the identity of the allele at fixed SNP positions in each parent. Counts of all reads containing fixed SNPs mapping to a given gene ID were summed and any gene IDs with at least 20 supporting reads in the two parental samples combined were retained. Counts were normalized to reflect differences in sequencing depth between samples. Any samples with zero reads mapping were adjusted to one in order to allow statistical testing. Relative contributions of mapped reads were calculated and significant differences in expression between parents (binomial exact test), between alleles in the hybrid (binomial exact test) and between the ratio of parental read counts to counts of each parental allele in the hybrid (Fisher’s exact test), done as in McManus et al.18. FDR corrected q-values were used for all three tests (significance q < 0.5%).

For X-linked genes we examined regulatory divergence by comparing SNP counts in the fertile and sterile hybrids. If regulatory factors on the X-chromosome (either cis or trans) are responsible for parental species expression divergence, then each hybrid will experience maternal-species levels of gene expression and the ratio of parental expression will equal that of the ratio between the two hybrids. If autosomally derived trans factors alone are responsible for misexpression, then the sterile and fertile hybrid should express the X-linked gene in equal amounts and the parental and hybrid ratios will differ (Table 4).

Table 4 Types of regulatory divergence indicated by patterns of allelic expression of (a) autosomal genes and (b) X-linked genes.
Full size table

Finally, we compared changes in ratios of autosomal allelic expression in the hybrids using both correlations as well as Fisher’s exact tests in search of genome-wide patterns of differential allelic expression. We also examined the chromosomal distribution as well as potential function of the 1% of genes with the highest fold-change in allelic expression ratio between hybrids.

Additional Information

How to cite this article: Gomes, S. and Civetta, A. Hybrid male sterility and genome-wide misexpression of male reproductive proteases. Sci. Rep. 5, 11976; doi: 10.1038/srep11976 (2015).