Male sterility in Drosophila mojavensis × D. arizonae hybrids is not caused by gene movements in the Y chromosome

Abstract

Y chromosomes play important roles in sex determination, male fertility, and, in some cases, hybrid male sterility. Nearly 40 years ago, Zouros and co-workers found that Drosophila mojavensis males carrying a Y chromosome from the sibling species D. arizonae are sterile, with fertility restored by chromosome 4 from D. arizonae. While these findings could result from gene interactions, they may also suggest the movement of an essential male fertility gene from the Y to chromosome 4 in the D. arizonae lineage. To test this, we analyzed the Y-linked gene content of D. mojavensis, D. arizonae, and the closely related D. navojoa. We identified 5 previously unreported Y-linked genes, bringing the total to 14. These genes fit the pattern observed in other Drosophila species: they originated from autosomal, testis-specific genes that duplicated onto the Y chromosome. All Y-linked genes are shared between D. arizonae and D. mojavensis, ruling out the movement of a protein-coding gene as the cause of sterility. However, we found a huge difference in the copy number of GI26128, an HP1 gene family member, with ∼1,400 Y-linked copies in D. arizonae but only 6 in D. mojavensis, making it a candidate to explain the hybrid male sterility.

Introduction

Although the central role of the Drosophila melanogaster Y chromosome in male fertility has been known since the dawn of genetics (Bridges, 1916), knowledge about its gene content is still incomplete. To date, molecular genetics and genomic investigations have identified 14 Y-linked protein-encoding genes in D. melanogaster (Carvalho et al., 2000, 2001, 2015; Gepner & Hays, 1993; Krsticevic et al., 2010; Mahajan & Bachtrog, 2017; Vibranovski et al., 2008). Six of these genes are essential for male fertility, corresponding to the “fertility factors” identified in earlier formal genetic studies (Brosseau, 1960; Hazelrigg et al., 1982; Kennison, 1981), while the remaining genes also likely contribute to male fertility, as has been shown for the PRY gene (Hafezi et al., 2020). Investigations on other Drosophila Y chromosomes revealed several shared characteristics: (i) They are gene-poor despite being physically large; (ii) their genes arise from duplications of autosomal genes, and more rarely from the X chromosome, usually followed by the loss of the original copy; (iii) the vast majority of those genes were testis-specific prior to duplication; (iv) there is no clear evidence of ancestral X-Y homology; and (v) the rate of gene gains is ∼10-fold higher than the rate of gene loss (Carvalho & Clark, 2013; Koerich et al., 2008; Ricchio et al., 2021). The last two characteristics are sharply discrepant from canonical Y chromosomes (e.g., mammals), in which X-Y homology is very clear and evolution primarily occurs through gene loss (Hughes & Rozen, 2012). Drosophila Y chromosomes are also unusual because they are not involved in sex determination; instead, sex determination is based on the number of X chromosomes or the X/autosomes ratio (Bridges, 1916; Cline, 1993; Erickson & Quintero, 2007).  Drosophila Y chromosomes are entirely heterochromatic, which largely explains the slow progress in their study (Carvalho et al., 2009). The high content of repetitive DNA, typical of heterochromatic regions, frequently leads to gaps and fragmentation in the genome assembly; as a result, the pieces of the Y chromosome are not identified as such, and end up in a set of unmapped scaffolds (sometimes referred to as “chromosome U,” for unmapped), along with other heterochromatic regions and misassembled sequences. Furthermore, as a result of assembly fragmentation, exons of the same gene are frequently scattered across different scaffolds, hampering gene annotation efforts (Carvalho et al., 2000). These problems were addressed by specific computational methods that allow the identification of Y-linked scaffolds and handle the annotation of fragmented genes (Carvalho & Clark, 2013; Hall et al., 2013). Larracuente and co-workers recently applied long-read sequencing to produce near chromosome-level assemblies of Y chromosomes in four Drosophila species (Chang et al., 2022).

The species trio D. mojavensis, D. arizonae, and D. navojoa form the mojavensis cluster. Their relationships are well known, based on morphological studies (Richmond et al., 2012; Ruiz et al., 1990), cytogenetics (Ruiz et al., 1990), molecular markers (Machado et al., 2007; Oliveira et al., 2012), hybridization (Reed & Markow, 2004; Ruiz et al., 1990), and, more recently, genome sequencing (Sanchez-Flores et al., 2016). All evidence indicates that D. arizonae and D. mojavensis are more closely related to each other than to D. navojoa. Interestingly, males carrying the D. arizonae Y chromosome in a D. mojavensis background are sterile, whereas the reciprocal genotype is fertile (Figure 1A; Vigneault & Zouros, 1986). Furthermore, fertility is restored by a single chromosome 4 from D. arizonae (Figure 1B; Pantazidis & Zouros, 1988). Finally, introgression and mapping experiments showed that the fourth chromosome effect can be attributed to a chromosomal region named sperm motility factor (SMF), which seems to segregate as a Mendelian factor (Figure 1C; Pantazidis et al., 1993).

Figure 1.

Fertility of D. mojavensis and D. arizonae hybrids. Hybrids with the genetic background of D. mojavensis and the Y chromosome of D. arizonae are sterile (panel A). Hybrid fertility is restored by a single copy of chromosome 4 of D. arizonae (panel B) or by the SMF region (from chromosome 4) of D. arizonae (panel C). Note that SMF was mapped to a region corresponding to ∼1/3 of chromosome 4 (Pantazidis et al., 1993) and may be composed of more than one gene. Modified from Pantazidis et al. (1993).

Two hypotheses can explain these observations. The first is that sterility results from Bateson–Dobzhansky–Muller epistatic interactions between genes in chromosomes Y and 4, which is the traditional explanation for hybrid sterility (Brideau et al., 2006; Pantazidis & Zouros, 1988). Alternatively, sterility could arise from the translocation of a gene essential to male fertility from the Y chromosome to chromosome 4 in D. arizonae after the split of the species (Figure 2). Under this hypothesis, hybrid males with genotype X^mojY^ari 4^moj/4^moj would lack any copy of the translocated gene and would be sterile. It is worth mentioning that a similar situation has been found in certain hybrids between D. melanogaster and D. simulans, although in this case, the gene moved between autosomes (Masly et al., 2006). Note also that gene movements involving the sex chromosomes are more prone to generate hybrid sterility as their effects can manifest in F1 hybrid males, whereas autosomal movements typically require F2 segregation to reveal incompatibilities (Ricchio et al., 2021).

Figure 2.

Male sterility in hybrids between D. mojavensis and D. arizonae as a result of gene movement. An ancestrally Y-linked gene essential for male fertility (represented by a black square) would have moved to chromosome 4 in D. arizonae after the divergence from D. mojavensis. Hybrids carrying the Y chromosome of D. arizonae and homozygous for chromosome 4 of D. mojavensis (X^moj/Y^ari 4^moj/4^moj) would be sterile because they have no copy of the fertility gene. Fertility would be restored by a single copy of the D. arizonae chromosome 4, or by the region containing the gene (“SMF”; see Figure 1, panels B and C). The reverse scenario could also explain the sterility: If a fertility gene ancestrally located on chromosome 4 moved to the Y chromosome in the D. mojavensis lineage, the X^moj/Y^ari 4^moj/4^moj would again lack any copy of the gene.

Here, we tested the “gene movement hypothesis” by thoroughly identifying Y-linked protein-coding genes in D. mojavensis, D. arizonae, and D. navojoa. Besides this motivation, full identification of Y-linked genes has been made in only a few Drosophila species: D. melanogaster (Carvalho et al., 2000, 2001, 2015; Krsticevic et al., 2010; Mahajan & Bachtrog, 2017; Vibranovski et al., 2008), the three species of the simulans clade (Chang et al., 2022), D. virilis (Carvalho & Clark, 2013), and D. willistoni (Ricchio et al., 2021), and hence has intrinsic interest.

Materials and methods

Genome assemblies

All three male genomes (D. mojavensis, D. arizonae, and the outgroup D. navojoa) have been sequenced before (Clark et al., 2007; Sanchez-Flores et al., 2016) and have reference gene annotations maintained by NCBI. So, in principle, the test of the gene movement hypothesis would entail a simple comparison of the Y-linked gene content in the two focal species, using the outgroup D. navojoa to ascertain the ancestral state. However, Drosophila Y chromosomes are notoriously tricky to assemble and annotate, so we re-sequenced with Nanopore the two focal species (D. mojavensis and D. arizonae) due to the limitations of previous assemblies (they used Sanger and Illumina technologies, respectively). We then did a thorough annotation of the Y-linked protein-coding genes in the three species. As described in the Results section, most new genes we found are missing in the current reference annotation of these three species.

For D. mojavensis, two long-read assemblies are available, but they are not useful for identifying Y-linked genes. Miller et al. (2018) sequenced only females. Jaworski et al. (2020) used mixed sexes, but we found only a few incomplete Y-linked genes in this assembly (kl-2, kl-3, kl-5, and CG11719). We used two assemblies based on the inbred strain 15081-1352.22: the reference Sanger assembly (GCA_018153725.1) from Clark et al. (2007) and a new Nanopore assembly (GCA_046254515.1) using PromethION 10.4.1 flow cells, which have ∼1% error rate (the Nanopore assembly of D. mojavensis section in the Supplementary Methods). We also sequenced with Illumina the brown (br) strain originally used by Zouros and co-workers, in order to molecularly identify the br marker, and to attempt to identify the SMF gene (see the Illumina assembly of D. mojavensis brown strain section in the Supplementary Methods). Unfortunately, the br strain was lost years ago in the stock centers and individual labs. (We sequenced an ethanol-preserved sample, kindly donated by Dr. William Etges, from the University of Arkansas.)

We used two different assemblies for D. arizonae. Oliveira et al. (2024) sequenced the strain 15081-1271.17 using Nanopore R9.4.1 flow cells and Flye 2.8 for assembly; we reassembled their raw reads using the latest Flye version (2.9.3-b1797), which solved some misassemblies in Y-linked sequences. The assembly GCA_001654025.1 (Sanchez-Flores et al., 2016) used strain 15081-1271.41 and Illumina sequencing technology. Finally, for D. navojoa, we used the assembly GCA_001654015.2 (Vanderlinde et al., 2019), which used strain navoj_Jal97 and Illumina technology.

Identification of Y-linked scaffolds

We used the Y chromosome Genome Scan method (YGS; Carvalho & Clark, 2013) to identify the Y-linked scaffolds. This method compares female reads (Supplementary Table S1) with a genome assembly containing male DNA (i.e., either males or mixed sexes). The X-linked and autosomal sequences are expected to display extensive similarity to the female reads, whereas Y-linked sequences remain largely unmatched. YGS uses only single-copy sequences since repetitive sequences (which are abundant in the Y chromosome) are usually shared between the Y and the female reads. This comparison is done by first decomposing the assembled genome and the female reads into k-mers, using k-mers of size 15–18. To increase accuracy, YGS optionally incorporates male reads: k-mers present in the assembled genome but absent from the male reads are deemed assembly errors and excluded from the analysis (Carvalho & Clark, 2013). YGS has been successfully used in humans and several Diptera (Carvalho & Clark, 2013; Ricchio et al., 2021; Vigoder et al., 2021). The scripts used to perform YGS are shown in the Supplementary Material.

Annotation of Y-linked genes

The YGS method identifies Y-linked scaffolds. In order to annotate the protein-coding genes in these Y-linked sequences, we did BLAST searches against Diptera protein and transposable element databases. These searches yielded a set of candidate genes, which we validated with RNA-seq data (Supplementary Table S1). Due to the peculiarities of the Y chromosome, we adapted the standard gene annotation procedures. In particular, exons of Y-linked genes are frequently scattered across different scaffolds due to difficulty in assembling the long and repeat-rich introns (Carvalho et al., 2000). As described in the Results section, we found that the Illumina assemblers are generally ineffective in assembling multicopy genes. To solve both the gaps and the fragmentation, we used transcriptome assemblies. As all known Drosophila Y-linked genes have testis-restricted expression (Ricchio et al., 2021; Vibranovski et al., 2008), we used testis transcriptome assemblies for D. mojavensis and D. arizonae, or adult male whole-body transcriptomes when testis libraries were unavailable, as in the case of D. navojoa. For the gene GI18320, which has both autosomal and Y-linked copies, we investigated the sequence divergence by estimating dN/dS ratios (a detailed description of the methods used for this analysis is provided in the Supplementary Methods section: Molecular evolutionary analysis (dN/dS) of the GI18320 gene in the repleta group).

Like other heterochromatic regions, Y chromosomes usually contain a fair amount of pseudogenes (Carvalho & Clark, 2013; Carvalho et al., 2015; Smith et al., 2007; Tobler et al., 2017). We used two criteria for classifying a candidate gene as functional: the presence of an intact coding sequence (CDS) and RNA expression; candidates lacking either were classified as pseudogenes. Y-linkage of all candidate genes was confirmed by PCR using male and female DNA; in D. mojavensis, we tested all four subspecies (PCR primers are listed in Supplementary Table S2).

Timeline of gene acquisitions by the Y chromosome

We determined the phylogenetic point at which the Y chromosome acquired its genes by examining species progressively more distant from the mojavensis cluster, within the repleta group, and extending out to D. virilis. We performed blastN searches, or TblastN (in more distant species), using D. mojavensis Y-linked genes as queries and the available genomes of D. buzzatii, D. repleta, D. hydei, D. nannoptera, and D. virilis as databases. The genome of D. aldrichi was too fragmented for use. Gene location was determined through synteny analysis; if a gene was found on a large scaffold with other genes, and this location was shared with other species, we concluded that the gene was not Y-linked. Conversely, hits on small scaffolds lacking other genes suggested Y-linkage. In cases where only female genomes were available, Y-linkage was confirmed by the gene’s absence in the female genome and its presence in the male transcriptomes.

Gene expression

We measured gene expression of the Y-linked genes and their homologs using the RSEM program (Li & Dewey, 2011) and RNA-seq data from various tissues available for the mojavensis cluster and D. virilis (Supplementary Table S1). We included D. virilis in our analysis to determine the ancestral expression pattern before a given gene moved to the Y chromosome. RSEM requires a reference transcriptome, but Y-linked genes are commonly mis-annotated. Hence, we prepared custom reference transcriptomes using the FlyBase annotation release 1.3 as the starting point for D. mojavensis and D. virilis and the NCBI annotation for D. arizonae and D. navojoa (downloaded on May 19 2023, Supplementary Table S1). We removed from these databases all sequences corresponding to the 14 annotated Y-linked genes (the nine previously known, plus those identified in this paper) using the cd-hit-est-2d program (Li & Godzik, 2006). Finally, we added the 14 Y-linked genes that were correctly annotated to our databases and ran the RSEM program. The final reference transcriptomes are available at https://doi.org/10.6084/m9.figshare.28098287.v1.

Copy number estimation

Even with long-read sequencing, repetitive genes can be collapsed during assembly, resulting in underestimation of copy number. In order to obtain an assembly independent estimate of copy number, we did a blastN search of the coding sequence (CDS) of the Y-linked genes against databases of Illumina reads of the corresponding species (Supplementary Table S1) and extracted from the tabular-formatted output (m8 format) the coverage of each base of the gene CDS. We then computed the median coverage for each gene. For male samples, the median coverage of each gene was converted to copy number by normalizing it against the median coverage of a set of known single-copy Y-linked genes (kl-2, kl-3, kl-5, etc.). When female reads from the same strain are available, we repeated the above procedures in order to check for the presence of copies outside the Y chromosome and estimate their number. In this case, the copy number was estimated by dividing the median coverage of the target gene by the median coverage of a set of 20 single-copy X-linked genes. When non-Y copies were detected, the final estimate of Y-linked copies was obtained by subtracting the number of non-Y copies from the initial estimate of male copies.

As a control for this method, we performed the same analyses using 20 single-copy X-linked genes and 20 single-copy autosomal genes. The resulting autosome-to-X-median coverage ratio (A median/X median) was ∼2 for male samples and ∼1 for female samples, as expected. We further validated the method using Y-linked genes with up to 10 known copies to assess its accuracy (Supplementary Table S3).

Results

Identification of Y-linked sequences in the mojavensis cluster

The three species were assembled with different sequencing technologies and assemblers, leading to very different scaffold counts (Table 1). Despite these differences, the YGS method successfully identified Y-linked sequences across all assemblies, producing the typical bimodal distribution: Y-linked sequences formed a peak on the right (with a large amount of sequence unmatched by female reads), while the X and autosomal sequences formed a peak on the left (Figure 3 and Supplementary Figure S1). The red dots in the first column (panels A, D, and G) mark scaffolds previously known to be Y-linked genes in the mojavensis cluster (Carvalho & Clark, 2013; Dupim et al., 2018; Koerich et al., 2008), serving as positive controls (note that all known Y-linked genes are located in the right peak). The small scaffolds in the intermediate region of the graphs (between the two peaks) usually correspond to misassembled repetitive regions and other assembly errors (Carvalho & Clark, 2013).

Table 1.

Basic statistics of the assemblies used in this study.

	D. mojavensis Nanopore	D. mojavensis Sanger	D. arizonae Nanopore	D. arizonae Illumina	D. navojoa Illumina
Total number of scaffolds	817	6,841	321	5,133	13,813
Genome size (Mbp)	181.3	193.8	187.8	141.1	147.3
Complete BUSCOs (%)	98.6	98.1	98.1	93.4	97.4
Number of Y-linked scaffoldsa	461	2,401	134	377	1,107
Sum (Mbp) of Y-linked scaffoldsa	34.4	21.1	19.1	0.68	1.01

^aCut-off for Y-linkage: 50% unmatched single-copy k-mers.

Figure 3.

Identification of Y-linked scaffolds in the mojavensis cluster with the YGS method. First row (panels A–C), D. mojavensis Nanopore assembly; second row (panels D–F), D. arizonae Nanopore assembly; third row (panels G–I), D. navojoa Illumina assembly (Vanderlinde et al., 2019). The X-axis in all graphs is the evidence of Y-linkage (% unmatched single-copy k-mers); scaffolds with a high proportion of sequence unmatched by female reads are Y-linked. Column 1 (panels A, D, and G): Each dot represents a scaffold, and the red dots mark the previously known Y-linked sequences (Carvalho & Clark, 2013; Dupim et al., 2018; Koerich et al., 2008). Column 2 (panels B, E, and H) shows the scaffold distribution as the number of scaffolds, whereas column 3 (panels C, F, and I) shows the scaffold distribution as the sequence amount. The Y-linked scaffolds are usually small, while numerous (column 2; panels B, E, and H); they account for a small amount of sequence (column 3; panels C, F, and I). Note that the bulk of the sequence can be reliably classified as Y-linked (peak on the right-hand) or X/autosomal (left-hand peak). Non-informative scaffolds (less than 20 single-copy k-mers) were excluded from the analysis (see text). The raw YGS results are available at https://doi.org/10.6084/m9.figshare.28098287.v1.

The Y-linked sequences usually appear as numerous small scaffolds due to assembly fragmentation, which might give the misleading impression that a large portion of the assembled genome is Y-linked (second column, Figure 3B, E, H and Supplementary Figure S1). In contrast, the distribution based on sequence length more accurately reflects the proportion of Y-linked sequences relative to the autosomes (third column, Figure 3C, F, I, and Supplementary Figure S1). A cut-off of “50% unmatched single-copy k-mers” indicates that the D. mojavenis Y chromosome comprises 34.4 Mbp distributed across 461 scaffolds, whereas the X and autosomes account for 162 Mbp in 419 scaffolds. A more stringent cut-off of 80% unmatched k-mers yields fairly similar results: 28.1 Mbp of Y-linked sequence and 168.3 Mbp for X and autosomes. Table 1 shows the results for all species.

While comparing the amounts of Y-linked sequences among species, it is important to keep in mind that they are heavily influenced by the sequencing technology: the Nanopore assembly of D. arizonae contains 19.06 Mbp of Y-linked sequences, whereas the corresponding value for the Illumina assembly is 0.68 Mbp (see the Supplementary Discussion for more details). Although the sequenced strains were different, this discrepancy is likely due to the Y chromosome’s huge content of repetitive sequences and their superior representation achieved with long-read assemblies (Nanopore and PacBio) compared to Illumina short-reads. Similar results were reported for D. willistoni (Ricchio et al., 2021). This bias most likely explains the small amount of Y-linked sequences we found in D. navojoa, sequenced using Illumina technology.

Identification of Y-linked protein-coding genes

We searched for protein-coding genes among the scaffolds identified as Y-linked in the three species of the mojavensis cluster. The two focal species (D. mojavensis and D. arizonae) were independently annotated by two experienced annotators (T.V. and F.U., with a final pass by A.B.C.), and no disagreement was found. We found all nine genes already known to be Y-linked in the three species (Carvalho & Clark, 2013; Dupim et al., 2018; Koerich et al., 2008), confirming the effectiveness of the YGS method. These known Y-linked genes are kl-2, kl-3, kl-5, ORY, Ppr-Y, JY-alpha, GJ19835, CG11719, and CG2964. As usually seen in Drosophila Y-linked genes, all have testis-restricted expression (Supplementary Table S4).

Additionally, we identified five previously unknown Y-linked genes, which we refer to as “new genes”. Only one of these (Tes123) is present in the reference gene annotation of the D. mojavensis genome (Supplementary Table S5), which is maintained by NCBI (file: https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/018/153/725/GCF_018153725.1_ASM1815372v1/GCF_018153725.1_ASM1815372v1_rna.fna.gz, downloaded on November 21, 2022). All five genes had been at least partially annotated when D. mojavensis was first sequenced (Clark et al., 2007) or in subsequent studies (Yang et al., 2018). In D. arizonae and D. navojoa, only one of these five genes (GI11242) was partially annotated in the NCBI reference annotation; the remaining four genes are missing completely. Since all previous annotation efforts were automated, these genes had not been manually curated or identified as Y-linked before (Supplementary Table S5). Furthermore, most of the original annotations were incomplete due to fragmentation and gaps in the Y chromosome assemblies. We kept the original gene names, even if only a small portion of their coding sequence had been previously identified. We improved the annotation of most of these Y-linked genes, especially in D. arizonae and D. navojoa. We also made improvements to some previously known Y-linked genes in D. mojavensis; for instance, the first exon of the kl-3 gene had not been annotated, which we did now using RNAseq data. We confirmed the identities of the genes by phylogenetic and synteny analyses (Supplementary Figures S2–S6). The five newly found Y-linked genes are GI11242, GI26128, GI18320, Tes123, and GI25647 (Table 2). In addition, we identified pseudogenes and instances of misannotation, as described in the Supplementary Results (sections Identification of Y-linked pseudogenes and Misannotated genes) and summarized in Supplementary Table S6.

Table 2.

The Y-linked genes of the mojavensis cluster found in this work.

Gene	moja	aria	nava	vira	Ancestral expressionb	Predicted functionc
GI11242	1 Y	2 Y	1 Y	1 A	Testis-specific	Sugar transporter
GI26128	6 Yd 1 Xe	1429 Y 150 X 3 Af	603 Y 3 Xe	0	Testis-specific	HP1 family; chromo shadow domain
GI18320	4 Y 1 Ae	2 Y 1 A	1 A	1 A	Testis-specific	Unknown function
Tes123	260 Y	249 Y	78 Y	1 A	Testis-specific	Protein kinase, catalytic
GI25647	74 Y	18 Y	0	1 A	Testis-specific	Transcription initiation factor IIF, beta subunit

^aNumber of gene copies counted in the assembled genomes or, for copy number greater than 10, estimated by coverage analysis of Illumina genomic reads (due to gross errors in the assembly of repetitive genes). See Supplementary Table S9 for data on strain variation or D. mojavensis and D. arizonae.

^bGene expression of the D. virilis orthologs, which are autosomal (GJ15101, GJ20974, LOC6634084, GJ22467, and GJ21374, respectively). Expression is deemed testis-specific when expression in other tissues is absent or very weak.

^cData from InterPro and CDD/NCBI.

^dTwo potentially functional copies, and four pseudogenes.

^ePseudogene.

^fMuller elemente B, chromosome 3.

Movement of protein-coding genes cannot explain male sterility in D. mojavensis × D. arizonae hybrids

We found no presence–absence difference among the 14 Y-linked genes in D. mojavensis and D. arizonae. Hence, gene movements cannot explain the hybrid sterility reported by Zouros and co-workers (Pantazidis & Zouros, 1988; Pantazidis et al., 1993). Comparative genomics confirmed that all 14 genes moved to the Y chromosome before the divergence of D. mojavensis and D. arizonae (Figure 4). This also explains why all four D. mojavensis subspecies share the same Y-linked gene content. In D. navojoa, we found 12 Y-linked genes, the same as its sister species, except for GI18320 and GI25647; the first is autosomal, and the second is absent from the D. navojoa genome.

Figure 4.

Gene gains and losses in the mojavensis cluster Y chromosome. Gene location (Y-linkage vs. autosomal/X-linkage) was determined by YGS data, synteny, and BWA analysis. Arrows represent the direction of the movements (gene gains by the Y chromosome, red arrows; gene losses due to movement to autosome, solid blue arrows; genes lost from the genome, dashed arrows). The ten genes on the basal branch were already known to be Y-linked; the loss of the PRY gene due to a movement to an autosome was also known (Carvalho & Clark, 2013; Dupim et al., 2018; Koerich et al., 2008). Genes were labeled with the name of the D. melanogaster orthologs. In the absence of an ortholog in D. melanogaster, we used the name of D. virilis (GJ19835) or D. mojavensis (GI26128, GI25647, and Tes123) orthologs. Phylogenetic relationships were taken from Oliveira et al. (2012).

The five newly found genes are typical Drosophila Y-linked genes

The GI11242 gene encodes a sugar transporter. It is Y-linked in most species of the repleta group analyzed here (Supplementary Table S7), indicating that the translocation occurred in the ancestor of the repleta group (Supplementary Figure S3). In contrast, it remains autosomal or X-linked in the closely related groups mesophragmatica and nannoptera. Its orthologs in D. melanogaster (CG12783) and D. virilis (GJ15101) have testis-specific expression (Supplementary Table S8; Thurmond et al., 2019). It has two copies in D. arizonae and is single-copy in the two other species.

The GI26128 gene is Y-linked in all three species of the mojavensis cluster and seems to have originated from a duplication of the autosomal gene HP1Mcsd in their common ancestor. It likely represents a divergent member of the HP1 family (Supplementary Results section GI26128 is a divergent member of the HP1 family; Supplementary Figure S4) and exhibits unusual features for a Drosophila Y-linked gene: It has a huge copy number on the Y chromosome in D. arizonae (∼1,400 copies) and D. navojoa (∼600 copies), along with ∼150 additional X-linked copies in D. arizonae (Supplementary Table S9). These characteristics are strikingly similar to the Stellate—Suppressor of Stellate genes of D. melanogaster (Kogan et al., 2000). Furthermore, GI26128 has a huge difference in copy number between D. arizonae and D. mojavensis, ∼1,400 vs. 6 (Supplementary Table S9), which may explain the role of the Y chromosome in the sterility of their hybrids.

The GI25647 gene seems to be completely absent from the D. navojoa genome, as we could not find it even in the raw reads (Supplementary Table S9). This gene loss may correlate with its movement to the Y chromosome since both events occurred roughly around the same phylogenetic interval (after the split from D. buzzatii). GI25647 and its orthologs are present only in the Drosophila subgenus, where they probably originated from a duplication of the TfIIFbeta gene (Supplementary Figure S4). TfIIFbeta (Transcription factor II F beta) is present in all Drosophila species examined; it is part of the transcription factor II F complex and is involved in mRNA transcription by RNA polymerase II. As expected from its function, it is expressed in all tissues and stages (Thurmond et al., 2019). On the other hand, its paralog GI25647 had testis-restricted expression before moving to the Y chromosome (Supplementary Table S8). Interestingly, in species outside the mojavensis cluster, GI25647 is located in the immediate vicinity of HP1Mcsd, which probably gave rise to the Y-linked GI26128 gene (Supplementary Figure S4). Note also that GI25647 and GI26128 were acquired in the same branch of the phylogeny (Figure 4). These findings strongly suggest that the Y chromosome acquired the two genes in a single event. GI25647 is another multi-copy gene (Table 2).

The Tes123 gene is found only in species from the virilis-repleta branch of the Drosophila subgenus, having moved to the Y chromosome after the divergence between an ancestor of the mojavensis cluster and D. buzzatii (Figure 4). In other species, Tes123 is located on Muller element C, adjacent to its paralog SRPK gene, likely originating through a tandem duplication (Supplementary Figure S5). SRPK is present across all Drosophila species analyzed. Interestingly, a member of the SRPK gene family was also found to be Y-linked in the D. simulans clade (Chang et al., 2022), indicating convergent evolution of Y-linked gene content. Both Tes123 and SRPK contain a Serine/Threonine protein kinase domain. Although Tes123 was initially annotated as a seminal fluid protein (Wagstaff & Begun, 2005), we found that it is highly expressed in testes of both the mojavensis cluster and D. virilis (Supplementary Table S8), whereas SRPK is expressed in various tissues and life stages (Thurmond et al., 2019). Tes123 is a multi-copy gene in all three species (Table 2).

The GI18320 gene is the ortholog of the D. melanogaster gene CG31226 (Supplementary Figure S6), a testis-expressed gene with unknown function. It was acquired by the Y chromosome in the ancestor of D. mojavensis and D. arizonae, being autosomal in all other species, including D. navojoa. Thus, it is a very young Y-linked gene (age between 1.5 and 5.8 Myr old, based on the divergence times among the three species of the mojavensis cluster; Sanchez-Flores et al., 2016). As frequently happens with young Y-linked genes (Carvalho et al., 2015; Ricchio et al., 2021), we can find its remnants at the original autosomal location (Figure 5). Interestingly, in D. mojavensis, the autosomal copy became a pseudogene, with premature stop codons, loss of initial methionine (Supplementary Figure S7), and no RNAseq expression signal (Supplementary Table S10). In contrast, the autosomal copy in D. arizonae has all the signs of a functional gene (Supplementary Figure S7), including abundant transcription (Supplementary Table S10). Molecular evolutionary analysis revealed that GI18320 is evolving under purifying selection (dN/dS ratio << 1) in most branches, including the Y-linked ones, implying gene functionality in these branches (Figure 6; Supplementary Figure S8 and Tables S12–S14). Indeed, we can formally reject the hypothesis that the GI18320 Y-linked copies of D. mojavensis or D. arizonae are evolving neutrally (dN/dS ratio = 1; p ≤ .03 in both cases; likelihood ratio test). Both Y-linked copies seem to be evolving at similar rates (p > .99; line 6 of Supplementary Tables S12 and S13).

Figure 5.

Fate of gene GI18320 in the mojavensis cluster. The Y-linked copy of GI18320 remained functional in D. mojavensis and D. arizonae. The autosomal copy became a pseudogene in D. mojavensis, since it lacks gene expression and has mutations that disrupt the coding sequence (GI18320Ψ). In D. arizonae, the autosomal copy (GI18320A) seems to be functional: It is expressed, and its coding sequence is intact. In D. navojoa the gene is in the ancestral position (Muller B) and has no Y copy (see Supplementary Figure S6 for more details).

Figure 6.

Molecular evolution of the GI18320 gene in the repleta group. The non-synonymous to synonymous substitutions ratio (dN/dS ratio), shown above each branch, was estimated with HyPhy (Kosakovsky Pond et al., 2005), as detailed in the Supplementary Material. See Supplementary Table S12 for the dN and dS values in each branch. Branches with no substitutions were labeled as “0/0.”

Consistent with the previously noted signs of pseudogenization (premature stop codons, no mRNA expression), the molecular evolutionary analysis indicates that the autosomal copy of GI18320 in D. mojavensis is a pseudogene: It accumulated numerous synonymous and non-synonymous substitutions, with a dN/dS ratio close to 1 (p > .8 for the hypothesis dN_mojA = dS_mojA; likelihood ratio test). The situation of D. arizonae is more ambiguous: its autosomal copy has two non-synonymous substitutions and zero synonymous substitutions. A likelihood ratio test shows that this is not significantly different from the hypothesis dN_ariA = dS_ariA (p > .3), suggesting it may also be a pseudogene. However, taking into consideration that it has gene expression and an intact Open Reading Frame (ORF), its elevated dN/dS ratio is also suggestive of positive selection. Indeed, we found strong statistical support for a difference in evolutionary rates between the autosomal copies in D. mojavensis (which is a pseudogene) and D. arizonae (p = .001 for the hypothesis dN_mojA = dN_ariA and dS_mojA = dS_ariA). Hence, in two closely related species, the same duplicated gene (the GI18320 autosomal copy) followed opposite evolutionary trajectories: pseudogeneization in D. mojavensis and functionality (possibly under positive selection) in D. arizonae. Ricchio et al. (2021) found a similar situation in D. willistoni. While we cannot rule out a fully deterministic explanation for these cases, they suggest that chance plays a role in the fate of these duplicated genes.

Searching for the SMF gene

The full account of this section is presented in the Supplementary Material and Supplementary Results; here, we present a brief summary.

Zouros and co-workers (Pantazidis & Zouros, 1988; Pantazidis et al., 1993; Vigneault & Zouros, 1986) reported that chromosome 4 of D. arizonae restores the male fertility in hybrids between D. arizonae and D. mojavensis, mapping the effect of SMF to the interval between markers Pgm and br (Figure 1). We attempted to physically delimit this region and found that it encompasses ∼23 Mbp, nearly the entire chromosome 4 (section Physical delimitation of the SMF region in the Supplementary Results, Supplementary Table S15, Supplementary Figures S9, and S10), containing thousands of genes. This large size is consistent with the original genetic mapping, which placed the markers 30 centiMorgans apart (Pantazidis et al., 1993).

The huge size of the SMF region precludes the investigation of minor sequence changes as possible SMF candidates (there would be too many), so we focused on detecting broad structural differences (section Identification of candidate genes in the SMF region in the Supplementary Results). We aligned the SMF region of both species and identified differential regions totaling ∼137 kb in D. mojavensis (62 regions) and ∼213 kb in D. arizonae (120 regions). The D. mojavensis-specific regions contained four annotated genes; three were lowly expressed, multicopy sequences likely representing repeats. The remaining gene, LOC6586363, resembles D. melanogaster Jon65Aiv (a gut protease, part of a large gene family) and shows strong gut expression and weak testis expression. However, its presence/absence is polymorphic in D. arizonae strains, making it an unlikely SMF candidate. The D. arizonae-specific regions contained a testis-expressed gene (LOC108614613, similar to D. melanogaster CG6602), but the presence/absence of this gene is polymorphic in D. mojavensis, and the original br strain used by Zouros and co-workers contains a full copy, which excludes it as an SMF candidate.

These analyses revealed no compelling candidates for the SMF gene, suggesting that finer genetic mapping will be required to identify it.

Discussion

Identification of all Y-linked protein-coding genes in the mojavensis cluster

Previous investigations identified seven Y-linked genes in D. arizonae and D. navojoa (Dupim et al., 2018) and nine in D. mojavensis (Carvalho & Clark, 2013; Koerich et al., 2008). These results were obtained by searching for genes already known to be Y-linked in the reference species D. melanogaster and D. virilis. While simple and effective, this approach misses genes not Y-linked in the reference species. In this paper, we used a de novo approach that identifies Y-linked genes irrespective of their Y-linkage in other species (Carvalho & Clark, 2013). We found five previously unknown Y-linked genes, which, as expected, were acquired by the Y chromosome after the split between the ancestors of the mojavensis cluster and D. virilis. These five genes fit the known pattern of Drosophila Y-linked genes: they originated from DNA-based duplications of autosomal or X-linked genes with pre-existing testis-specific expression (Table 2; Supplementary Table S8). However, three of them have a large number of copies (above 15, and up to ∼1,400), which is not commonly observed.

Movement of protein-coding genes cannot explain male sterility in D. mojavensis × D. arizonae hybrids

The original motivation for the present study was to investigate the hypothesis that a gene movement involving the Y chromosome causes male sterility in D. mojavensis × D. arizonae hybrids. Namely, if a movement in the D. mojavensis lineage (after the split from D. arizonae) repositioned a male fertility gene from chromosome 4 to the Y chromosome, hybrid males with genotype X^moj/Y^ari 4^moj/4^moj would have no copy of this gene and be sterile. Alternatively, a gene movement in the D. arizonae lineage in the opposite direction (from the Y chromosome to chromosome 4) would yield the same outcome. Both scenarios would explain the results of Zouros and co-workers (Pantazidis & Zouros, 1988; Pantazidis et al., 1993; Vigneault & Zouros, 1986).

However, we found no evidence of a Y-linked gene repositioned between D. mojavensis and D. arizonae. This suggests either that the culprit is a protein-coding gene that escaped detection, that it belongs to a class of genes not investigated here (e.g., non-coding RNA genes), or that gene movements are not the cause of the hybrid sterility. The first possibility seems unlikely, as our thorough annotation detected all previously known protein-coding Y-linked genes. However, it is still possible that a poorly conserved small ORF or de novo gene (Chanut-Delalande & Zanet, 2024; Couso & Patraquim, 2017) might have escaped detection. Regarding the second explanation, several non-coding RNA genes are essential for male fertility in D. melanogaster (Wen et al., 2016). We are currently investigating Y-linked non-coding RNA genes, but our provisional conclusion is that gene movements are not the cause of male sterility in D. mojavensis × D. arizonae hybrids. It is likely caused by gene interactions (Brideau et al., 2006), as originally proposed by Zouros and co-workers (Pantazidis & Zouros, 1988; Pantazidis et al., 1993; Vigneault & Zouros, 1986). Identifying the interacting gene in chromosome 4 (SMF) would be a promising approach, but as we noted before, this would require a much finer mapping of this gene.

The copy number of the GI26128 gene is a candidate for the cause of hybrid male sterility

We found three large gene families on the Y of the mojavensis cluster: GI26128, Tes123, and GI25647. As shown in Table 2, one of them, the GI26128 gene, has six copies in the Y chromosome of D. mojavensis and ∼1,400 copies in D. arizonae (range in three strains: 1,400–2,300; Supplementary Table S9). Although less dramatic, the GI25647 gene also has differences in copy number: 74 vs. 18, respectively. Such considerable differences in gene dosage in testis-expressed genes can easily cause male sterility. In particular, GI26128 is a member of the HP1 gene family, whose members play important roles in heterochromatin formation and transcriptional regulation (Schoelz et al., 2021), and are also involved with meiotic drive (Helleu et al., 2016). Brideau et al. (2006) and Greil et al. (2007) found that HP1 protein interacts with the product of Lethal hybrid rescue. This gene causes hybrid lethality in D. simulans × D. melanogaster hybrids. Unfortunately, large-scale protein interaction networks have been studied only in D. melanogaster (Tang et al., 2023), and this species does not have GI26128 or GI25647 orthologs. Understanding the interaction network of these genes might help elucidate their possible role in this male hybrid sterility case and perhaps the SMF gene’s identity (Pantazidis et al., 1993). Note that the above copy-number estimations are based on Illumina coverage, and should be viewed as approximations (Supplementary Table S3).

The GI26128 genes of D. arizonae have striking similarities to the Ste and Su(Ste) genes of D. melanogaster

Drosophila arizonae has a very large number of Y-linked copies of the GI26128 gene (∼1,400), a large number of X-linked copies (∼150), and three copies in a Muller B contig (Table 2). The presence of numerous X- and Y-linked copies is reminiscent of the Stellate/Suppressor of Stellate system of D. melanogaster. The multi-copy X-linked Stellate gene is abundantly transcribed, but its mRNA is degraded by antisense transcripts (and small RNAs) generated by the multi-copy Y-linked gene Suppressor of Stellate [Su(Ste)]. In testes, most small RNAs are produced by Su(Ste) gene (Aravin et al., 2001), and the lack of Y-linked copies results in complete male sterility. Given these similarities, we looked for antisense transcripts of GI26128, and we indeed found them. Furthermore, as shown in Supplementary Table S11, the Y chromosome is responsible for 82% of the sense transcription (and the X for the remaining 18%), whereas the antisense transcription comes from a Muller B contig and the X (75% and 25%, respectively). Finally, we found that the most abundant small RNAs from D. arizonae testis match GI26128 (Attilio Pane and A. Bernardo Carvalho, in preparation). The Stellate and Suppressor of Stellate genes are believed to be a relict meiotic driver and a relict suppressor of drive, respectively (Hurst, 1992; Meng & Yamashita, 2025). Note that in D. arizonae, the driver would be the Y-linked copies, and the X and autosomal copies would be the suppressors. This location of the putative suppressors is compatible with the theoretical expectation (and empirical observation) that the spread of a sex-linked meiotic driver elicits the evolution of suppressors of the drive in the other sex chromosome and in the autosomes (e.g., Carvalho et al., 1997, 1998; Courret et al., 2019). It is also suggestive that GI26128 belongs to the HP1 gene family, whose members are involved in meiotic drive: in the Paris sex-ratio (SR) system of D. simulans, one of the X-linked meiotic drivers is HP1D2, a young and rapidly evolving member of the HP1 gene family (Helleu et al., 2016). Meiotic drive has long been suspected to be a cause of hybrid sterility (Frank, 1991; Hurst & Pomiankowski, 1991; Searle & Villena, 2024), and the D. arizonae/D. mojavensis may provide a very interesting system to investigate these questions.

Convergence of gene content between D. simulans and the mojavensis cluster Y chromosomes

Y-linked gene families have been reported before in other Drosophila species (Chang et al., 2022; Ellison & Bachtrog, 2019; Krsticevic et al., 2010; Lyckegaard & Clark, 1989). Strikingly, Chang et al. (2022) found, in the D. simulans clade, a Y-linked gene family (Lhk) derived from the SRPK gene, as the progenitor of the Tes123 gene. They proposed that the duplication to the Y may have resolved intralocus sexual antagonism, as SRPK is essential for both male and female reproduction (Loh et al., 2012). However, the situation we found is somewhat different: Tes123 arose from a duplication of SRPK and acquired testis-specific expression before moving to the Y chromosome (Supplementary Table S8), which, in principle, would have solved the putative conflict. Only later (in the ancestor of the mojavensis cluster) did Tes123 move to the Y and get amplified. This suggests that the movement to the Y and amplification may have been driven not by an intra-locus conflict but by some other evolutionary force, such as meiotic drive (Ellison & Bachtrog, 2019; Hurst, 1992). SRPK encodes a protein involved in splicing Y-linked genes with gigantic introns (Fingerhut et al., 2024). Therefore, if Tes123 maintains this same function, multiple Y-linked copies could help splice these giant introns.

It is interesting to note that all known Y-linked large gene families, such as Su(Ste), Mst77Y, Lhk, GI26128, Tes123, and GI25647, have been acquired recently by the Y chromosome (∼5.8 Myr, at most). In contrast, several single-copy genes are shared between species that diverged 40–60 Myr ago (Koerich et al., 2008). This pattern strongly suggests that these large Y-linked gene families are evolutionarily short-lived, which is compatible with the idea that they are part of a transient “arms race,” potentially mediated by meiotic drive (Ellison & Bachtrog, 2019; Hurst, 1992).

Concluding remarks

Drosophila Y chromosomes participate in important biological phenomena (Carvalho et al., 2009; Chippindale & Rice, 2001; Kaiser & Bachtrog, 2010; Lemos et al., 2008; Montchamp-Moreau et al., 2001; Stoltenberg & Hirsch, 1997). In the mojavensis cluster, they also play a critical role in reproductive isolation (Pantazidis & Zouros, 1988; Pantazidis et al., 1993). Although our initial hypothesis—that a simple gene movement could explain the sterility of X^moj/Y^ari 4^moj/4^moj males—was not supported, we found genetic differences between the Y chromosomes of D. arizonae and D. mojavensis that are promising avenues. Furthermore, the tantalizing similarities between GI26128 and the Ste/Su(Ste) open many interesting avenues of research, such as the possible existence of a relict Y-linked meiotic drive system and the connection between meiotic drive and hybrid sterility. We hope our results stimulate further research on the role of the Y chromosome in this interesting case of hybrid sterility.

Data availability

Fly strains are available at the National Drosophila Species Stock Center. DNA sequencing data, RNA-seq, and the genome files were deposited by the original researchers in SRA/NCBI. Scripts used for the analyses in this study are included in the Supplementary Material and/or deposited in Figshare, available at https://doi.org/10.6084/m9.figshare.28098287.v1.

Author contributions

T.V. and A.B.C. conceived the project. All authors contributed to the sequencing and assembly of the genomes and transcriptomes. T.V., F.U., G.G., and A.B.C. performed the analyses. A.B.C., C.M.A.C., C.V., and T.A.M. funded the project. T.V. and A.B.C. wrote the first version of the manuscript, with contributions from F.U. and T.A.M. All authors reviewed and corrected the manuscript.

Funding

This research was funded by FAPERJ—Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro, grant number CNE2018 (A.B.C.); CNPq—Conselho Nacional de Desenvolvimento Científico e Tecnológico, grant INCT-EM (A.B.C.); Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brasil (CAPES)—Finance Code 001 (F.U.), Wellcome Trust, grant 207486/Z/17/Z (A.B.C.); São Paulo Research Foundation (FAPESP)/Université de Lyon 1 (UdL—IDEXLyon) Joint Call to C.M.A.C [2020/06238-2] and to C.V.; ANR ANR-20-CE02-0015 (C.V.); and CONACyT Fronteras FC-2604/2016 (T.A.M.).

Conflict of interest

The authors declare no conflict of interest.