Introduction

Speciation occurs through the evolution of reproductive isolation (Dobzhansky 1937; Coyne and Orr 2004). Reproductive isolation is said to be achieved when there are barriers that prevents two species from producing fit hybrid offspring. Among sympatric Drosophila species, prezygotic reproductive isolation (sexual isolation) evolves faster than post-zygotic isolation (hybrid incompatibility), whereas among allopatric Drosophila species, prezygotic reproductive isolation and intrinsic post-zygotic isolation evolves roughly at the same rate (Coyne and Orr 1989, 1997). Between recently evolved species, intrinsic post-zygotic isolation manifests in the form of hybrid male sterility (HMS), hybrid female sterility (HFS) and hybrid inviability (HI). Bateson, Dobzhansky and Muller independently proposed a model to explain the evolution of hybrid incompatibility that result in intrinsic post-zygotic isolation (Bateson 1909; Dobzhansky 1937; Muller 1942). Hybrid incompatibility involves a negative epistatic interaction between genes from two different species. When two species diverge from one another, they accumulate genetic substitutions that function normally within their genomic background but, can cause disruption of gametogenesis or development when brought together in a hybrid (Coyne and Orr 2004, Dobzhansky 1937).

Genes expressed in male reproductive tract are known to evolve rapidly among closely related species. In Drosophila a pattern of faster evolution of male-specific genes relative to female and non-reproductive genes has been shown both in terms of coding sequence divergence and loss and gain of genes among distantly related species (Haerty et al. 2007). Drosophila spermatogenesis is a multistep process where a single progenitor germline stem cell undergoes a sequential division and morphological changes to become a mature motile sperm. Spermatogenesis in Drosophila can be broadly divided into four stages: germline establishment, mitotic proliferation of germline cells, spermatid formation through meiotic division and spermatid differentiation or spermiogenesis (Fuller 1998; Wakimoto et al. 2004; White-Cooper and Bausek 2010). Given the rapid divergence of male-specific genes, it is not surprising that the common outcome of crosses between closely related species is hybrid male sterility (Haldane 1922).

Cytological studies of many Drosophila interspecies sterile hybrid males showed the disruption of spermatogenesis at the stage of spermatid differentiation. Spermatogenesis proceeds normally until the meiotic division and spermatid formation and encounter problems in the post-meiotic stages. These problems include the lack of synchrony in spermatid development and failure of spermatid individualization wherein the interconnected spermatid bundles fail to differentiate and mature into a motile sperm (Dobzhansky 1937). In the sterile male hybrids of D. simulans clade (D. simulans, D. mauritiana, and D. sechellia), it was observed that spermatogenesis arrest occurs both before and after the onset of meiosis, depending on the species pair used in the interspecies cross (Kulathinal and Singh 1998; Lachaise et al. 1986).

Several genes have been identified that play major role in spermatogenesis of Drosophila. Bag of marbles (bam), benign gonial cell neoplasma (bgcn), roughex (rux) are involved in regulating the early stage of spermatogenesis, bam and bgcn limit the number of mitotic divisions facilitating the mitosis to meiosis transition. Always early (aly), achintya (achi), cookie monster (comr), and spermatocyte arrest (sa) which are known as the meiotic arrest genes are involved in regulating the progression into meiosis and the initiation of spermiogenesis (Fuller 1998; Jiang and White-Cooper 2003; White-Cooper 2010). Finally, don juan (dj), JYalpha, Mst84Dc and Mst98Ca are involved in the maturation of round spermatid into elongated spermatid during spermiogenesis (Michalak and Noor 2004; Moehring et al. 2007; Santel et al. 1997).

The Drosophila nasuta-subgroup of immigrans species offers an excellent case to study the role of rapid genetic divergence in bringing about hybrid incompatibility. It is an young cluster of species with the history of multiple speciation events in a short period resulting in morphologically very similar species with varying degrees of post-zygotic isolation (Wilson et al. 1969) and extensive chromosomal polymorphism (Hatsumi et al. 1988; Suzuki et al. 1990).The nasuta-subgroup consist of a dozen closely related species or subspecies that are widely distributed across South-East Asia (Kitagawa et al. 1982; Wilson et al. 1969), and the crosses between many species of this subgroup produce sterile and sometimes fertile offspring. Females of nasuta-subgroup of Drosophila are morphologically indistinguishable, whereas males can be classified into three groups based on the markings on the frons and thorax (Kitagawa et al. 1982; Wilson et al. 1969). The first category includes D. nasuta, D. albomicans, D. kepulauana and D. kohkoa, which show a continuous silvery patch on their frons and dark band on their thorax. Second category comprises of D. pulaua, D. sulfurigaster sulfurigaster, D. sulfurigaster bilimbata, D. sulfurigaster albostrigata, and D. sulfurigaster neonasuta, which have whitish patch along the edges of their compound eyes. The third category includes D. pallidifrons, Taxon-F, I and J, which has reduced white patch. D. niveifrons is an exception with an X-shaped silvery patch on their forehead. The earliest known member of the subgroup D. niveifrons emerged about 3.5 million years ago (Mya), later the D. pulaua, D. s. sulfurigaster, D. kohkoa and Taxon- F diverged from D. niveifrons about 2.5 Mya. Rest of the species emerged between 0.7 and 1 Mya (Yu et al. 1999).

The study of pattern of rapid divergence of genes involved in hybrid incompatibilities is potentially an important way of understanding the mechanisms of speciation. Most of the studies are conducted in species of D. melanogaster and other sibling species. Investigations in recently diverged species has great potential in identifying underlying evolutionary forces that drive the process of speciation in the early stages of speciation. The nasuta-subgroup of Drosophila which comprises of many young species provides an excellent model to understand the role of these evolutionary processes in the process of speciation. The nasuta-subgroup comprises of many species which show symptoms of post-zygotic reproductive isolation such as hybrid inviability, hybrid male and female sterility (Kitagawa et al. 1982; Wilson et al. 1969; Nirmala and Krishnamurthy 1973). The availability of genome sequences of these species provides an opportunity to understand the role of rapid divergence in bringing about hybrid incompatibilities in young species group of Drosophila.

The main goal of our study is to characterize the nucleotide divergence of key spermatogenesis genes likely to have been involved in the hybrid incompatibility interaction resulting in the sterility across the species of nasuta-subgroup of Drosophila. We compare the interspecies polymorphism and divergence between species that result in an inviable hybrid and also performed gene-wide, codon-wide and lineage-specific selection analysis in the phylogenetic framework.

Materials and Methods

Ortholog Search

We downloaded whole genome raw sequences of 35 strains of 11 Drosophila species of nasuta-subgroup (Table S1) from NCBI-SRA (Mai et al. 2020; Mohanty and Khanna, 2017). Genome assemblies were built using UniCycler (Wick et al. 2017), which assembled the sequences into scaffolds. We used both paired end and single data and selected normal bridging mode which allows moderate contig size and moderate misassembly rate. We excluded contigs that are shorter than 100 base pairs in the final assembly. The amino acid sequences of individual spermatogenesis genes of Drosophila melanogaster amino acid sequences were acquired from Flybase (Marygold et al. 2013). These sequences were used in a tBLASTn (Gerts et al. 2006) search against the D. albomicans assembled genomes with a liberal cut off of E = 0.1 in order to ensure the detection of divergent orthologs. The best BLAST hit scaffold for each gene was taken and 3 kb upstream and downstream of the homologous region was extracted. We included 3 kb upstream and downstream sequences to ensure that the open reading frame is not missed during the gene prediction. The extracted DNA sequence was used for gene prediction using AUGUSTUS webserver (Stanke and Morgenstern 2005), a Generalized Hidden Markov Model (GHMM) based gene prediction tool which predicts the gene structure, coding sequences and amino acid sequences of the respective gene. We selected D. melanogaster as the organism of reference since that was the only species from Drosophila group available on AUGUSTUS. Gene orthology was confirmed by reciprocal best BLAST hit approach by blasting the predicted amino acid sequence against the annotated Drosophila albomicans protein database. We also checked for the presence of conserved protein domains present in the respective genes using NCBI-Common conserved Domain Database (CDD) (Last accessed:16/07/2021) to confirm the orthology. We identified the upstream and downstream genes that were annotated to further support the correct orthology. Rest of the genomes were annotated using D. albomicans amino acid sequences as query.

Sequence Alignment and Phylogenetic Inference

Codon-based phylogenetic analysis require the accurate alignment of ortholog sequences. We used the predicted coding sequences to build alignments using MUSCLE (Edgar 2004) and CLUSTAL Omega (Sievers et al. 2011). We translated the coding sequences into aminoacid sequences using Seaview v5.0.4 (Gouy et al. 2010) aminoacid translator and built the alignments. The aligned amino acid sequences were reverse translated and the resulting codon alignments were used in phylogenetic and selection inference.

Since many selection analyses we performed require the phylogenetic trees of the ortholog sequences, we built maximum-likelihood and Bayesian phylogenetic trees using IQtree (Trifinopoulos et al. 2016) (Last accessed: 16/07/2021) and MrBayes (Ronquist et al. 2012), respectively. Maximum-likelihood trees were built by selecting GTR nucleotide substitution model and node support was evaluated with 1000 bootstrapping replicates. Bayesian tree was constructed by running two independent runs, each with four chains (one cold and three heated). The analysis was run for 2 million generation saving every 1000th tree. The runs were terminated when the split frequency reached the value less than 0.001. 25% of the trees were discarded as burn-in while summarizing the trees. The summarized trees were edited and rendered using Figtree v1.4.4 (Rambaut 2010).

Divergence Analysis

To detect the nature and strength of selection acting on the spermatogenesis genes, we employed a combination of evolutionary analysis. These analyses estimate the ratio of non-synonymous to synonymous substitution rate (dN/dS = ω) across the genes, codons and lineages. When there is no selection acting on a given gene/codon/lineage, both non-synonymous and synonymous substitutions are expected to become fixed with the same probability (ω = 1). In the presence of selection, selection advantage can increase the fixation probability of non-synonymous substitution (ω > 1, positive selection), or decrease it due to selection constrains (ω < 1, negative or purifying selection).

Several methods have been developed for detecting signature of positive selection based on the ratio of dN/dS at different levels such as whole alignment (gene-wide), branch specific, codon based and a combination of these. Gene-wide selection analysis were performed to detect the signature of selection using the alignments of all the species, without making any assumption about foreground branches. First, we employed codeml implemented in Phylogenetic Analysis by Maximum Likelihood v4 package (PAML) (Yang 2007). We compared a null model (M7) in which ω is assumed to be beta-distributed among sites and a selection model (M8), in which codons are allowed to have an extra category of positively selected sites with ω > 1. The significance of this test was validated using likelihood ratio test. A set of gene-wide selection tests were also employed using HyPhy (Kosakovsky Pond et al. 2005) webserver (Last accessed: 16/07/2021). We used BUSTED (Branch-site Unrestricted Statistical Test for Episodic Diversification) (Murrell et al. 2015), which specifically tests whether a gene has experienced positive selection in at least one site or one of the branches of a given phylogeny. BSR (Branch-site Random effects likelihood test) (Kosakovsky Pond et al. 2011) was used to test for episodic diversifying selection. Finally, aBSREL test (adaptive Branch-Site Random Effects Likelihood) which is a improved version of branch-site model, was used to test if positive selection has occurred in a proportion of branches.

Codon-based selection analysis was performed by employing maximum-likelihood methods implemented in CODEML of PAML v.4. CODEML estimates the ratio of non-synonymous to synonymous substitutions (ω) under various models allowing ω to vary among sites (site models) and branches (branch models) and a combination of both (branch-site models). A likelihood ratio test was performed in all the tests by comparing the null model against an alternative model. The test statistic 2Δl = 2(l1 − l2) where l1 and l2 are the likelihood values of null and alternative models, respectively, was calculated. The twice the difference between two likelihood values was compared with the chi-square distribution with degree of freedom to be the difference between number parameters. The Bayesian Empirical Bayes (BEB) approach was employed to identify positively selected sited by calculating the posterior probabilities of a particular site belongs to the class of sites under positive selection where sites with greater posterior probability (≥ 95%) were considered to be under strong positive selection. Additionally, we performed mixed effects model of evolution (MEME) (Murrell et al. 2012) and Fast Unconstrained Bayesian Approximation for inferring selection (FUBAR) (Murrell et al. 2013) tests available in HyPhy webserver (Last accessed: 16/07/2021). MEME employs a mixed-effects maximum likelihood approach to test the hypothesis that individual sites have been subject to episodic positive selection or diversifying selection. FUBAR uses a Bayesian approach to infer synonymous (dN) and synonymous (dS) substitution rates on a per site basis for a given alignment and phylogeny.

We applied branch-site models for both frontal sheen complex and orbital sheen complex. The hybridization among species of frontal and orbital sheen complex often produces fertile hybrids, whereas hybridization between species from different complex often produce only sterile males and some combination of species produce both sterile males and females. Upon marking the branch of interest (foreground branch), the alternative hypothesis assigns some sites in the foreground branch to be under positive selection, whereas null hypothesis does not. The likelihood ratios of each model were compared and the significance and sites under positive selection were identified as stated above. Additionally, we also performed BSR and aBSREL tests described above to detect lineage-specific positive selection.

Polymorphism Analysis

We employed McDonals-Kreitman test (MK test) (McDonald and Kreitman 1991) using DnaSP v.6 (Librado and Rozas 2009) to detect the signature of recent selection by comparing the ω ratios within species with those between species. This test takes advantage of the intraspecific variation, where the ω ratios within species are expected to be equal to the ratios between species under neutral scenarios. We compared the combinations of species that produce fertile hybrids, sterile males, and both sterile males and females and compared their divergence rates. FDR correction was performed to account for multiple comparisons across genes for individual tests (Benjamini and Hochberg 1995).

Protein Domain Identification and Protein Modelling/Functional Assessment

We identified protein domains using Common Conserved domain Database (CDD) and Pfam 33.1 (El-Gebali et al. 2019), (Last accessed: 16/07/2021). The protein models bam, bgcn, aly, comr and dj were built using D. melanogaster protein structures as reference. We mapped all the sites with significant positive selection on to the three-dimensional protein structure using PyMOL (Schrodinger, LLC, 2015). We looked for the presence of human orthologs of all the genes analysed using DIOPT v8.0 integrated in Flybase (Marygold et al. 2013) (Last accessed: 16/07/2021). Human (Accession number: NP_001332905.1) and mice (Accession number: NP_001156485.1) ortholog sequences of bgcn were extracted from NCBI-Nucleotide databse. Aminoacid sequences of bgcn from D. melanogaster, D. albomicans, Mus musculus and Homo sapiens were aligned using MUSCLE and sites under positive selection were mapped on to the conserved domains between the orthologs.

Results

Identification of Orthologous Spermatogenesis Genes in nasuta-Subgroup of Drosophila

We assembled a total of 38 genomes of species belonging to nasuta-subgroup of Drosophila. Amino acid sequences of 10 spermatogenesis genes of D. melenogaster were extracted from Flybase. These amino acid sequences were used as query in tBLASTn search against 6 D. albomicans genomes we assembled since it is the only species which has complete annotated genome available. We predicted the CDS and respective amino acid sequences of each gene from D. albomicans (See methods). We performed reciprocal blast search using NCBI-Blast program to reassure the right orthology. Further D. albomicans sequences were used as query to annotate rest of the orthologs. We were able to extract a total of 331 orthologs from 35 genomes (Supplementary Table 1). A homolog of Mst98Ca was found upon the Blast search, we included the homolog in the analysis and named it Mst98Ca-like homolog. We excluded D. nevifrons and D. immigrans orthologs due to the high divergence of nucleotide sequences to avoid the problems associated with saturation of synonymous sites when comparing the diverged species. All the sequences are available in at figshare (https://figshare.com/s/ad586ca0f83d86871a45).

Phylogenetic Inference

The Bayesian phylogenetic trees (Figs. 1 and 2) constructed for individual genes were consistent with the species tree constructed by (Mai et al. 2020). D. nasuta, D. albomicans and D. kepulauana formed a single clade (nasuta subclade/frontal sheen complex). D. pulaua, D. s. sulfurigaster, D. s. bilimbata, D. s. albostrigata, D. s. neonasuta formed a seperate clade (sulfurigaster subclade/orbital sheen complex), whereas Taxon-F, the only species formed clade branching from the root of the tree. Node support values for all the major nodes was significant.

Fig. 1
figure 1

Bayesian phylogeny of bgcn infered using nucleotide sequences of nasuta-subgroup species. Node support for each major clade is indicated. Position of amino acid sites under positive selection are shown next to the individual species

Full size image
Fig. 2
figure 2

Bayesian phylogeny of comr infered using nucleotide sequences of nasuta-subgroup species. Node support for each major clade is indicated. Position of amino acid sites under positive selection are shown next to the individual species

Full size image

Rapid Divergence of Spermatogenesis Genes and Weaker Selective Constrain

We employed free ratio (M0) model of PAML to estimate the global ω (dN/dS) for all the spermatogenesis genes. The global ω estimates were similar for both the alignment methods used but varied significantly for each gene analysed (Supplementary Fig. 1). bam showed the highest ω (0.62) followed by aly (0.37), dj (0.34) and comr (0.30). 8 out of 11 spermatogenesis genes analysed showed higher ω than the reported median ω for spermatogenesis genes (0.10) in D. melanogaster subgroup (Haerty et al. 2007). Mst98Ca (0.009) had the least ω estimate followed by sa (0.066) and JYalpha (0.042). We employed the Fast, Unconstrained Bayesian Approximation (FUBAR) analysis, which detects sites evolving through purifying and diversifying selection. The strongest constrain was observed for Mst98Ca with 43.55% (Supplementary Fig. 2) of the codons evolving under negative purifying selection. The selective constrains was the weakest for early-stage spermatogenesis genes such as Bam (2.3%), Bgcn (4.19%) and Rux (2.47%).

Evidence of Gene-Wide and Codon-Based Positive Selection

All the genes except Mst98Ca showed signature of positive selection in at least one of the gene-wide selection analyses (Table 1). Likelihood ratio test (LRT) of codeml favoured the M8 selection model for all the genes except rux, achi, Mst98Ca and sa. BUSTED detected positive selection at all the genes except for aly, bam, Mst98Ca and Mst98Ca-like. Both BSR and aBSREL showed positive selection for at least one of the lineages in the gene tree of all the genes except bam and Mst98Ca.

Table 1 Gene-wide tests for positive selection and McDonald-Kreitman test
Full size table

We investigated the nature of natural selection influencing the spermatogenesis genes in the codon level by employing maximum-likelihood models implemented in PAML (see methods). We employed two pairs of models M1a vs M2a and M7 vs M8 and only considered the sites with significant positive selection (posterior probability ≥ 90%) inferred by M7 vs M8 comparison. The Byes Empirical Bayes (BEB) implemented in M8 identified 8 out of 11 genes with significant positively selected sites (Table 2). Early spermatogenesis genes such as bam and bgcn showed significant positive selection (PP ≥ 90%) at 3 and 8 sites, respectively. rux did not show any sites under positive selection. Spermatocyte arrest class genes such as aly and comr showed significant positive selection at 7 and 9 sites, respectively (Table 2). BEB identified one site with positive selection in sa but it was insignificant with PP less than 90%, whereas in achi, there were no sites under positive selection. dj and JYalpha showed 4 and 1 sites under positive selection (Table 2) among the genes involved in late spermatogenesis. Mst98Ca and Mst98Ca-like did not show any positive selection acting on any of the codons.

Table 2 Likelihood ratio test statistic for site models (M7 vs. M8)
Full size table

Additionally, we analysed the codon alignments for signature of positive selection using MEME and FUBAR (Supplementary Table 2). Among early spermatogenesis genes, MEME identified 7 and 2 sites under significant (P ≤ 0.05) diversifying selection for bgcn and rux, respectively. bam did not show signature of diversifying selection at any sites. Among spermatocyte arrest class genes aly, comr and sa each showed 2,7 and one site under significant diversifying selection, whereas achi did not show diversifying selection. Among late spermatogenesis genes, only Dj and JYalpha both showed 2 sites each under significant diversifying selection. FUBAR identified 7 sites for bam, 13 sites for bgcn, and 3 sites for rux under positive selection with posterior probability ≥ 90 which is considered significant. aly, comr and sa showed 6, 11 and one sites under significant positive selection. Finally, dj, JYalpha and Mst98Ca showed 5, 2 and one site each under significant positive selection.

Test for Lineage-Specific Positive Selection

To investigate whether the signature of positive selection observed in gene-wide and codon-based selection analysis is due to the effect of single lineage, we applied branch-site and branch-specific models to infer positive selection. The phylogeny of nasuta-subgroup species splits into frontal sheen complex (FSC) and orbital sheen complex (OSC), we performed branch-site tests considering one of the lineages as foreground and the other as background branch (described in methods). Upon performing likelihood ratio test, we found that the signature of lineage-specific positive selection was insignificant for all the genes analysed (Table 3). Although insignificant, BEB identified sites under positive selection for 6 genes we analysed. bgcn showed positive selection in the branch leading to FSC with 10 sites identified by BEB (Table 3). Interestingly, rux and achi which did not show any sites positive selection in the site models of codeml showed although insignificant, some sites under positive selection in the branch-site test. BEB picked one and two sites, respectively, for FSC and OBS for rux and 3 sites in FSC for achi (Table 3). aly showed one site in the branch leading to FSC and comr and dj showed one and two sites, respectively, in the branch leading to OSC (Table 3).

Table 3 Likelihood ratio test statistic for branch-site tetst
Full size table

Additionally, we employed aBSREL test to detect selection acting on a proportion of sites in individual lineages. All the genes except bam and Mst98Ca showed signature of positive selection in at least one of the branches in the phylogeny (Supplementary Fig. 3).

Selection Inference Using Pattern of Polymorphism and Divergence

Four (achi, aly, comr and Mst98Ca-like) of the eleven genes analysed showed the significant departure from neutrality in at least one of the hybridizing pair compared in MK test (Table 1). The departure from the neutrality observed at these four genes was due to both the excess of non-synonymous differences between species and excess of synonymous polymorphisms. Except for achi, there other genes (aly, comr and Mst98Ca-like) showed departure only in the comparison between species pair that result in sterile hybrids. There was pattern of increased synonymous and non-synonymous polymorphism in the comparison between the species producing fertile hybrids, whereas among the species that produce sterile hybrids, there was excess of between species divergence (Supplementary Table 14).

Discussion

Drosophila has long been used as a model to understand the mechanisms of speciation such as pattern of genetic diversification and identifying the genes involved in hybrid incompatibilities (Orr 1993). Most of these studies have been conducted in D. melanogaster subgroup (Bayes and Malik 2009; Brideau et al. 2006; Phadnis and Orr 2009; Presgraves 2003) where molecular mechanism of hybrid incompatibilities is understood in crosses between many sibling species. D. melanogaster and its sibling species have accumulated many such hybrid incompatibilities (Masly and Presgraves 2007; Presgraves 2003). Investigating a much younger subgroup potentially helps in understanding molecular mechanisms and evolutionary forces acting at the early stage of speciation process.

The nasuta-subgroup which diverged only about 3.5 MYA, with its pronounced difference in pre and post-zygotic reproductive isolation provides an excellent model to understand the process of speciation. Many species in the subgroup can produce viable, fertile and sterile offspring upon crossing between other members of the species complex (Kitagawa et al. 1982; Spieth et al. 1969; Wilson et al. 1969). Rapid divergence and has been established as one of the evolutionary forces capable of bringing about such incompatible interactions between closely related species. Our analysis of key spermatogenesis genes provide evidence for possible role of rapid divergence in bringing about hybrid incompatibilities.

We annotated a total of 331 orthologs of key spermatogenesis genes which are involved in the key stages of early, mid and late spermatogenesis process. We employed robust selection analysis to infer the mode and strength of Darwinian selection acting on these genes. Our study shows a pattern of high sequences divergence for five of eleven genes analysed between closely related hybridizing species of nasuta-subgroup of Drosophila. Such pattern of rapid divergence is expected for sex and reproduction-related genes (Haerty et al. 2007), but it is inconsistent considering the selection constraints on germline stem cell regulatory genes such as bam and bgcn. However, evidence from previous population genetic studies of spermatogenesis genes with major role in germline stem cell (GSC) regulation (Bauer DuMont et al. 2007; Choi and Aquadro 2014; Civetta et al. 2006) suggest that many genes with role in stem cell regulation evolve adaptively.

We analysed three genes bam, bgcn and rux, with key role in early stage of spermatogenesis. bam and bgcn are two genetically interacting genes which are regulators of gametogenesis in both the sexes (Lavoie et al. 1999). In females, the proper functioning of bam and bgcn is essential for the initiation of cytoblast differentiation. In addition, bam and bgcn are also involved in the assembly of endoplasmic reticulum-like fusome. In males, bam and bgcn are required for the switch from spermatogonial program of mitotic divisions to the spermatocyte differentiation (Fuller 1998; Schulz et al. 2004). rux is an essential cell cycle regulator in Drosophila, which has been shown to down-regulate CyclinA-dependent activity during G1 phase and is also responsible for temporary G1 arrest. Considering the role of bam, bgcn and rux in regulating the developmental witches during gamatogenesis, one might expect them to evolve under high selective constrains. However, evidence for rapid amino acid evolution of bam and bgcn has been documented in D. melanogaster and D. simulans clade (Bauer DuMont et al. 2007; Civetta et al. 2006) and in melanogaster subgroup for rux (Avedisov et al. 2001; Llopart and Comeron 2008).

Our codon-based analysis revealed that bam and bgcn are evolving under a strong positive selection, whereas rux only showed such signature in individual branches (FSC and OSC). One of the three positively selected site in bam is situated in the predicted nuclease domain (SI Fig.) and other two (Proline and Serine) on the PEST domain which is rich in proline (P), glutamic acid (E), serine (S) and threonine (T). PEST motif has been associated protein that are unstable and rapidly degraded by proteases (Rogers et al. 1986). The cytoplasmic form of Bam transiently expressed and it starts to accumulate at the cytoblast differentiation and disappears after completion of four rounds of mitosis (McKearin and Ohlstein 1995; Szakmary et al. 2005). Despite the significant positive selection detected in the M7 vs M8 comparison, branch-site models and polymorphism analysis failed to identify any sites under positive selection. This could be due to the fact that rapid divergence is common among genes that transiently expressed (Cutter and Ward 2005).

The predicted bgcn protein domain architecture of D. albomicans consists of 1325 amino acids (Fig. 3). bgcn is predicted to have helicase core module, an ankyrin repeat domain (ARD) inserted between the two helicase core domains and containing a pair of ankyrin repeats domains and two C-terminal extensions such as helicase-associated 2 (HA2) and oligonucleotide binding (OB) domains. 3 of 12 sites under significant positive selection are found in HA2 and OB domains, respectively. MEIOC and YTHDC2 are proposed to be the mammalian homolog of bam and bgcn known to play a role in the stem cell transition from mitotic to meiotic division. Ketu (keen to exit meiosis leaving testes under-populated) is a non-synonymous mutation in ythdc2 (Morohashi et al. 2011; Stoilov et al. 2002) and the for ketu mutation homozygotes are both male and female sterile in mice. Most insects’ lineages have YTHDC2 orthologs with full architecture including YTH domain. However, the orthologs in Drosophila lack the YTH domain (Supplementary Fig. 5) suggesting the loss of YTH domain in Last Common Ancestor (LCA). Multiple sequence alignment of Human, mice and three Drosophila species (Supplementary Fig. 5) showed many sites under positive selection are distributed among highly conserved sites.

Fig. 3
figure 3

Nucleotide divergence in early spermatogenesis gene bgcn among species of nasuta-subgroup of Drosophila. A Representation of bgcn protein showing predicted domains. Sites with significant signature of positive selection are shown in red and magentha. B Predicted three-dimensional model of bgcn protein (PDB of the template: 6up4.1.A). Amino acid sites under positive selection are highlighted (BEB posterior probability ≥ 90: Red, BEB posterior probability < 90: Green) (Color figure online)

Full size image

rux is a dose-dependent regulator of second meiotic division during spermatogenesis, in the absence of rux function, germ cells execute meiosis I and II, but then undergo and additional division as haploid cells. High expression of rux has been shown to result in failure to execute meiosis II (Lifschytz and Meyer 1977). Although, M7 vs M8 comparison of did not show any significance for positive selection MEME and FUBAR identified 2 and 3 sites, respectively. Branch-site model inferred weak positive selection on one and two sites in frontal and orbital sheen complex, respectively. Rux has a nuclear localization signal (NLS) domain which spans between 253 and 276 amino acids. The sites 256 falls in the NLS domain of the rux protein. NLS domain of rux shows high divergence in melanogaster subgroup. Mutants of rux show sterility in males but not in females, hence it has been proposed to be male-biased gene with role in spermatogenesis shows rapid divergence potentially driven by post-copulatory sexual selection/sexual conflict (Ellegren and Parsch 2007).

There is an existing hypothesis that germline genes coevolve with pathogens infecting the germline can result in elevated non-synonymous substation rate in bam and bgcn (Bauer DuMont et al. 2007). Wolbachia and Spiroplasma are two maternally inherited bacterial endosymbionts known to infect some Drosophila species (Mateos et al. 2006; Watts et al. 2009). Extensive divergence of bam due to Wolbachia infection between D. melanogaster and D. simulans affects bam function in females but has no apparent effect in males (Flores et al. 2015). Wolbachia infection can have both beneficial and deleterious effect on the fitness of Drosophila by increasing resistance viral infection and reducing the fecundity and life-span of the infected individuals, respectively (Chrostek et al. 2013). Maintaining the balance between both the beneficial and deleterious effects could potentially contribute to an ‘arms race’ between GSC regulatory genes and endosymbionts (Bauer DuMont et al. 2007). D. ananassae has been infected with Wolbachia for longer than D. melanogaster and despite which bam and bgcn did not show any signature of positive selection (Choi and Aquadro 2014). Considering that D. nasuta and D. albomicans are free from Wolbachia infection (Ravikumar et al. 2011), we can rule out the possibility of divergence of GSC genes to be driven by endosymbiont infection.

In Drosophila spermatogenesis, most transcription ceases during the entry into meiotic divisions. Therefore, the genes encoding proteins required for spermatid differentiation are transcribed in primary spermatocytes but translationally repressed until the appropriate time later in gamete development (Fuller 1998; White-Cooper and Bausek 2010). More than 2000 testis-specific transcripts are synthesized in primary spermatocyte (Doggett et al. 2011; White-Cooper 2010). Transcription in primary spermatocyte depends on a group of genes together named ‘‘meiotic arrest’’ genes (Ayyar et al. 2003; Jiang and White-Cooper 2003; Wang and Mann 2003; White-Cooper 2000, 1998). Broadly there are two meiotic arrest genes: aly-class (aly, comr, tomb, topi and achi/vis) and can-class (can, mia, nht, rye and sa). Among the aly-class genes, aly encodes the Drosophila homologue of C. elegans synMuvB gene lin-9 (Beitel et al. 2000; White-Cooper 2000). comr encodes a novel protein of unknown function (Jiang and White-Cooper 2003). achintya/vismay (achi/vis) and matotopetli (topi) encode sequence-specific DNA-binding proteins (Ayyar et al. 2003; Perezgasga et al. 2004; Wang and Mann 2003). The can-class genes encode the testis-specific TBP-associated factors (tTAFs), suggesting that their products form a testis-specific TFIID complex in primary spermatocytes (Hiller et al. 2001, 2004). We found evidence of rapid divergence at two of the four meiotic arrest genes analysed in the current study. aly and comr had seven and nine sites under positive selection with PP ≥ 90% in M7 vs M8 comparison of codeml (Table 2). The predicted D. albomicans comr protein has 891 amino acids and eight of the nine positively selected sites identified by BEB are mapped onto a single domain with unknown function (Supplementary Fig. 4). comr and achi also showed deviation from neutrality in polymorphism analysis. D. melanogaster comr predicted protein has an acidic domain in the C terminus of the protein (amino acids 518–570), and a predicted nuclear localisation sequence (NLS) (amino acids 583–589). In addition, a region that may represent a very divergent PB1 domain (amino acids 348–431). PB1 domains have been shown to mediate protein–protein interactions (Ito, 2001; Ponting et al. 2002).

After four rounds of mitotic divisions, the Drosophila germ cells enter meiotic prophase. After rapid meiotic divisions sperm morphogenesis takes places. During morphogenesis the chromatin undergoes condensation and the nuclei acquire needle-like shape. During this stage, the two mitochondrial derivatives elongate along the entire length of the axoneme to form the flagellum simultaneously in all 64 spermatids of one cyst. Throughout this process, the germ cells remain interconnected via cytoplasmic bridges. Finally, spermatids become individualized and stored as motile sperm. In Drosophila spermatogenesis, transcriptional activity ceases after the meiotic divisions while translation proceeds. Hence, many mRNAs are translationally repressed during meiotic prophase and translationally activated during sperm morphogenesis making translational control is a crucial feature of spermatogenesis (Schafer et al. 1990). The genes such as don juan (Santel et al. 1997) and Drosophila gene family Mst(3)CGP (Gigliotti et al. 1997; Kuhn et al. 1988; Schafer et al. 1990) are known to express during spermatogenesis and encode translationally repressed mRNA. dj encodes a protein of 29 kDa with structural similarities to histone H1 and it is localized in haploid nuclei during chromatin condensation and nuclear shaping. It can also be detected in the mitochondrial derivatives of the flagellum (Santel et al. 1998). Of four spermatid differentiation genes analysed in the current study, dj showed signature of rapid divergence in all the tests employed (Table 1). One and two positively selected sites are present on the two predicted domains of Dj (Fig. 4). JYAlpha encodes the alpha subunit of Na+ and K+ adenosine triphosphatase (Na + /K + ATPase), a transmembrane protein involved in ion exchange (Blanco and Mercer, 1998). One of the four mammalian isoforms of the Na + /K + ATPase alpha subunit, a4, is expressed exclusively in testes and is essential for sperm motility (Woo et al. 2000). JYAlpha is located on the fourth chromosome of D. melanogaster but on the third chromosome of D. simulans. Because of this transposition event of JYAlpha, a fraction of hybrids completely lacks JYAlpha and are sterile. The coding region of JYAlpha shows no signs of divergence by positive natural selection between D. melanogaster and D. simulans making a special case of reproductive isolation without sequence evolution. Contrast to this, our analysis showed signature of rapid divergence and positive selection (3 sites) in both gene-wide and codon-based analysis performed (Tables 1 and 2).

Fig. 4
figure 4

Nucleotide divergence in late spermatogenesis gene dj among species of nasuta-subgroup of Drosophila. A Representation of domain architecture of dj protein. dj has a NDUF V3 domain and a large domain with multiple subdomains of unknown function. Sites with significant signature of positive selection inferred from BEB are shown in red, sites identified by MEME and FUBAR are shown in magenta and blue, respectively. B Three-dimensional model of dj protein subunit covering the NDUF V3 domain (PDB of the template: 6a70.1.B). Amino acid sites under positive selection are highlighted (BEB posterior probability ≥ 90: Red, BEB posterior probability < 90: Green) (Color figure online)

Full size image

Hybrid incompatibilities such as inviability or sterility result from failed interactions between the genomes of parental species in F1 hybrids. Sterility of heterogametic sex is one of the most frequent result of crosses between closely related species (Haldane 1922). In Drosophila genus, the males being the heterogametic sex, the males show the sterility phenotypes. Several recent studies have suggested that disruptions in gene expression may be one source for sterility phenotypes (Hoekstra and Coyne 2007; Ortiz-Barrientos et al. 2006; Ranz and Machado 2006). The fact that sperm development is disrupted in Drosophila interspecies sterile hybrids, combined with the knowledge of spermatogenesis gene function in Drosophila melanogaster, has recently led to a series of studies comparing patterns of spermatogenesis gene expression in fertile parental species and sterile hybrids. The studies suggest that, more post-meiotic (spermiogenesis) than meiotic and pre-meiotic genes have been found to be significantly under expressed in sterile hybrids compared to parental species (Catron and Noor 2008; Michalak and Noor 2003, 2004; Moehring et al. 2007). Genome-wide miss-expression comparisons of D. simulans, D. mauritiana and their sterile male progeny found don juan, Mst84Dc and Mst98Ca, the three spermatid differentiation genes to be consistently down regulated in sterile hybrids (Michalak and Noor 2003; Moehring et al. 2007). Consistent with our analysis proving rapid divergence of don juan, and homolog of Mst98Ca, abnormalities such as fused sperm tails have been observed in crosses between some strains of D. nasuta and D. albomicans (Zhang et al. 2015). The same study showed Mst98Ca mapping on to on of the one of the hybrid male sterility QTL.

A typical speciation genetics study starts with studying the divergent reproductive traits between two species. Numerous such studies have identified genes that are rapidly diverging between closely related species, but these genes cannot be qualified as ‘speciation genes’ considering the possibility of genetic divergence after the speciation event. Nevertheless, two common pattern that have emerged from so far speciation genetic. The first is the ‘faster male’ evolution where HMS evolves at a rate an order of magnitude higher than HFS and HI (Tao et al. 2003; Tao and Hartl 2003). Second is the “large X” evolution in which HMS genes are enriched on the X chromosomes (Masly and Presgraves, 2007; Tao and Hartl 2003; White et al. 2012).

The above two patterns are better explained by the “conflict theory” where genomic divergence is driven by selfish genes, prominently by sex ratio distortion (SRD), also called sex chromosome meiotic drive (Frank 1991; Hurst and Pomiankowski 1991; Meiklejohn and Tao 2010). Meiotic drive is generally harmful to a genome since it breaches Mendelian ratio by gaining more than 50% transmission while quenching its homolog’s share in the gene pool of next generation. Thus, suppressors to silence the distorter are under strong selection to evolve and make the meiotic drive cryptic (Hartl 1975). When an SRD arises on the X chromosome, counter evolution on the Y and the autosomes is anticipated, hence, SRD operates as a perpetual dynamo for genome evolution and bouts of this distortion-suppression process eventually lead to speciation (Meiklejohn and Tao 2010). D. albomicans has been shown to have a SRD in a hybridization between D. albomicans (Okinawa) females and D. nasuta (India) males. The F1 males from this cross produce female-biased offspring. The driver was found to be located on the neo-X chromosome of D. albomicans, along with a drive suppressor, while D. nasuta was found to be suppressor-free (Yang et al. 2004). The same study also reported sterility in hybrid F1 and F2 males probably due to an interaction between the 3rd and Y chromosomes of D. nasuta and the autosomes of D. albomicans.

Combining the observed patterns such as ‘faster-sex’, ‘faster-male’, ‘large-X effect’ and ‘conflict theory’, our study proposes that the rapid evolution of spermatogenesis genes involved at the key stages of the process is by-product of the combination of these forces acting together in the whole of nasuta-subgroup. Despite the evolutionary constrains and no history of endosymbiont infection, GSC genes such as bam and bgcn showed higher divergence mediated by Darwinian positive selection. The hybrid male sterile phenotypes observed in the crossed between the species of nasuta-subgroup are consistent with the observed rapid divergence of late spermatogenesis genes such as dj and Mst98Ca. An extended investigation involving studying the specific stages of spermatogenesis arrest in the interspecies crosses would help in enhancing our understanding of intrinsic post-zygotic reproductive isolation in this subgroup. Comprehensive molecular population genetic analysis of more spermatogenesis loci would help in confirm the lineages or species-specific effect of positive selection and its role in hybrid male sterility phenotypes. Our study is the first attempt of understanding the genetic basis of post-zygotic reproductive isolation in nasuta-subgroup of Drosophila and lays a foundation for future exploration in the subgroup. Further detailed investigations using the genetic manipulation studies will enrich our understanding of the potential role of rapid divergence in bringing about hybrid male sterility.

Conclusions

In this study we have examined the molecular evolution of candidate genes with key role in various stages of spermatogenesis in species of nasuta-subgroup of Drosophila. We found evidence of rapid divergence at two early spermatogenesis genes, bam and bgcn. Another cell cycle regulator rux only showed lineage-specific positive selection in frontal sheen complex of the subgroup. We also observed signature of rapid divergence at dj and Mst98Ca, the key genes involved spermatid individualization. Our observations are consistent with the presence of Mst98Ca at one of the HMS QTL and of sperm-tail abnormality phenotype observed in the hybrids of D. nasuta and D. albomicans.