Data and method reproducibility across species

Given the temporal distribution of Drosophila spermatogenesis along the testis [39], dissections based on published methods [11] (Fig 1A) were undertaken to isolate cells from the main stages of spermatogenesis. This technique relies on cellular morphology as a guide to isolate regions enriched for three major phases of D. melanogaster spermatogenesis: mitosis, meiosis, and post-meiosis (Fig 1A).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Stage-Specific transcriptional profiling during Drosophila spermatogenesis.

A) Main cell types involved in Drosophila spermatogenesis and their spatial localization within the testis. B) Phylogenetic tree of selected Drosophila species analyzed in this study, accompanied by whole testis images highlighting regions enriched with mitotic (blue), meiotic (green), and post-meiotic (orange) cells. Chromosome insets depict the karyotypes for each species, marking autosomes (gray), X chromosome (red), Neo-X chromosome (dark red), and dot chromosome (cyan). Notably, D. willistoni exhibits two chromosomal fusions involving the X and dot chromosomes with autosomes. C) Heatmap displaying k-means clustering of gene expression (Z-scores) across different stages of D. melanogaster spermatogenesis: mitosis (Mit: blue), meiosis (Mei: green), and post-meiosis (PM: orange). Gray indicates non-unique profiles with broad expressions. D) Comparison of transcriptome cluster proportions between microarray data [11] and RNA-seq data from this study across four species.

https://doi.org/10.1371/journal.pgen.1011511.g001

Cell enrichment isolation for RNA sequencing was conducted across four species within the Drosophila genus: D. melanogaster, D. simulans, which are from the melanogaster subgroup; D. willistoni, a member of the willistoni subgroup within the Sophophora subgenus and considered a cousin group, features a neo-X chromosome [40,41]; and lastly, D. mojavensis, serving as an outgroup to the Sophophora subgenus (Fig 1B). Since the technique was initially standardized in D. melanogaster [11], we adapted it to the other species under scrutiny to ensure the appropriate section of each enriched region (See Materials and Methods and S1 Fig).

Importantly, our stage-enriched D. melanogaster transcriptome from RNA-seq strongly correlates with previous data obtained using microarrays (Pearson correlation > 0.85; S2 Fig). Additionally, high reproducibility was achieved across biological replicates for all spermatogenesis stages (Pearson correlation > 90%; S3 Fig).

To facilitate interspecies comparisons, we needed to ensure that the adaptations applied to the dissection method would accurately capture equivalent spermatogenesis stages across species. Therefore, orthologs following stringent criteria (1,913 out of 6,110 genes) were selected: (1) genes with only one annotated transcript (enabling 1:1 comparisons), (2) genes on the same Müller element, and (3) excluding those located on the Y chromosome and the D element (which functions as a Neo-X chromosome in D. willistoni and an autosome in other species, potentially biasing chromosome-specific analyses); we also removed their orthologs from the corresponding autosomes in the other species to ensure consistency.

Markedly, the PCA analysis in S4B Fig examines inter- and intra-species variation in Drosophila gene expression during spermatogenesis. It reveals that the variance between spermatogenesis stages (PC1, 35.2%) is greater than the variance between species within the same stage (PC2, 14.9%), indicating that gene expression profiles are more similar across species at the same stage than across different stages. Additionally, the selected gene set exhibited a consistent high correlation across species and stages (Pearson correlation > 0.62, S5 Fig and S5 Table). Finally, in alignment with a prior investigation for Drosophila microarray data [11,38], our findings unveiled a total of 12 distinct expression profile clusters across all transcriptomes (Fig 1C). Notably, the proportions of these clusters within each species demonstrated substantial similarity across transcriptomes (Fig 1D). Altogether, these results highlight the reproducibility and inter-species applicability of the dissection method originally established in D. melanogaster by [11], despite potential variations in cell proportions between species and expression detection methods.

MSCI across the Drosophila genus

To examine the MSCI-induced downregulation of X-linked genes relative to autosomes, we performed a comparative analysis across all annotated gene products, rather than restricting the analysis to the subset of orthologous genes, focusing on the proportion of gene products differentially expressed between meiosis and mitosis.

A signal consistent with MSCI is characterized by a higher proportion of X-linked genes being downregulated during meiosis compared to autosomal genes. Importantly, even in an ideal scenario in which where all technical limitations of measuring mRNA during spermatogenesis are absent and complete cell purification is achieved, we would not expect 100% of X-linked genes to be downregulated. This is because the proportion is calculated based on the total number of annotated coding sequences for each chromosome type, and not all X-linked genes are expressed during spermatogenesis at any given stage. Moreover, while our method enriches for mitotic and meiotic cells, it does not yield pure populations composed exclusively of a single spermatogenic cell type [11]. It is likely that our mitotic samples include some early spermatocytes, where MSCI may already be occurring, but at a lower proportion than in our meiotic samples, which may still contain other cell types in addition to cells actively undergoing MSCI. Thus, a key indicator of MSCI is a statistically significant overrepresentation of downregulated genes on the X chromosome during meiosis relative to mitosis.

In line with prior microarray findings in D. melanogaster, our RNA-seq data revealed a higher proportion of downregulated coding sequences on the X chromosome during meiosis compared to autosomes across all species, consistent to what is expected for MSCI in Drosophila (Fig 2B). Conversely, as an indirect effect of MSCI, autosomes showed a significantly higher proportion of upregulated gene products during meiosis compared to the X chromosome across all species (Fig 2B). Altogether, these results indicate that MSCI is a robust and evolutionarily conserved phenomenon, not restricted to D. melanogaster or any particular lineage.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Chromosomal distribution of coding genes down and upregulated in Drosophila male meiosis.

A) Schematization of meiotic profiles. Down: genes downregulated in meiosis in relative to mitosis; Up: genes upregulated in meiosis relative to mitosis. B) Chromosomal proportion of down (top) and upregulated (bottom) transcripts in D. melanogaster, D. simulans, D. willistoni, and D. mojavensis, respectively. Significant proportion differences (Chi-square test with Yates correction) are indicated by *, **, and ***, indicating p values ≤ 0.05, ≤ 0.01, and ≤0.001, respectively. Autosomes, Neo-X, X and Dot chromosomes are indicated by the color legend by Aut, NeoX, ChrX and Dot, respectively.

https://doi.org/10.1371/journal.pgen.1011511.g002

Interestingly, the Neo-X chromosome of D. willistoni exhibits patterns similar to those of the X chromosome (Fig 2B). Originating from a fusion and diverging from the melanogaster group around 32 million years ago [40], D. willistoni’s Neo-X shares key features with the X chromosome, such as demasculinization, indicated by a lower proportion of genes preferentially expressed in males [30,33,42]. These results indicate that Neo-X regulation mirrors that of the X chromosome, highlighting the widespread and enduring impact of MSCI.

To provide a broader perspective on chromosomal expression beyond individual gene-level analyses, we calculated the X: autosome (X:AA) expression ratio based on the median expression of all annotated gene products per chromosome. This metric offers a chromosome-wide view of expression dynamics across different stages of spermatogenesis. However, due to the same technical limitations discussed previously, we do not expect X:AA ratios to fall below 0.5 as a signal of MSCI consistency. A recent study [43,44] also showed that using overall median expression can artificially inflate X:AA estimates. To address this, we applied the filter-by-fraction method, which compares X and autosomal genes at matched expression thresholds (see Methods). Using this approach across four Drosophila species, we observed a consistent reduction in X:AA ratios during meiosis compared to mitosis (S7A Fig), with the effect becoming more pronounced under stricter expression thresholds. Interestingly, although meiosis is the peak of transcriptional activity during Drosophila spermatogenesis [45], our findings reveal a reduction in the expression of X-linked and neo-X genes during meiosis compared to mitosis—even among highly expressed genes.

A layer of complexity to Drosophila sex chromosome evolution is the fourth chromosome, also known as the dot chromosome. Evidence suggests that the dot originated from an ancestral X chromosome and later reverted to an autosome [46,47]. Although the dot chromosome shows a general depletion of upregulated genes during meiosis, it does not exhibit significant downregulation typically associated with inactivation in any species, including D. melanogaster, where single-cell studies have reported an overall reduction in expression [12]. Its significantly lower transcript counts, about 18 times fewer than other chromosomes (Fig 2B), might limit statistical power to detect robust differences. Altogether, it might suggest that its MSCI signature, a remnant of the dot chromosome’s X chromosome origin, may be weaker or less conserved across the genus.

In addition, through the X-to-autosome ratio results (S7 Fig), we provide an overview of post-meiotic regulation across the four species. While D. melanogaster exhibits a possible signature of post-meiotic sex chromosome repression, this pattern is not consistently observed in the other species. These findings should be interpreted with caution, given technical and biological limitations of later spermatogenic stages, such as generally low expression levels [48] and the potential higher persistence of mRNAs synthesized at earlier stages [22].

MSCI affects long non-coding gene expression

To investigate whether the MSCI-consistent pattern extends to Drosophila lncRNAs, an area not yet explored in the literature, we focused our analysis on RNA-seq data from Drosophila melanogaster due to its more comprehensive genome annotation [49]. The assessment revealed significant enrichment of downregulated genes during meiosis on the X and dot chromosomes, with autosomes exhibiting a higher proportion of upregulated genes (Fig 3B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. D. melanogaster chromosomal proportions of down and upregulated transcripts of lncRNAs.

A) Schematization of meiotic profiles. B) Chromosomal proportion of Down and Upregulated transcripts. C) Detailed chromosomal distribution of expression patterns of lncRNAs. Regions in red represent upregulated genes, while blue represents the opposite pattern. Black segments represent genomic regions with no annotated lncRNAs. Significant proportion differences (Chi-square test with Yates correction) are indicated by *, **, and ***, indicating p values ≤ 0.05, ≤ 0.01, and ≤ 0.001, respectively. The lncRNAs were analyzed only in Drosophila melanogaster due to its extensive genome annotation. Autosomes, X and Dot chromosomes are indicated by the color legend by Aut, ChrX and Dot, respectively.

https://doi.org/10.1371/journal.pgen.1011511.g003

In mammals, somatic X chromosome inactivation is driven by the long non-coding RNA Xist, which initiates silencing by spreading across the X chromosome from the X inactivation center (XIC) [50,51]. Inspired by this spatial regulation mechanism, we sought to understand the distribution of downregulated long noncoding genes on the Drosophila X chromosome. To achieve this, we mapped the chromosomal coordinates of each lncRNA, revealing that those differentially expressed during meiosis are uniformly distributed along the X chromosome (Fig 3C).

Additionally, our RNA-seq results reveal that long noncoding gene expression is predominantly post-meiotic, with 197 transcripts highly expressed during mitosis, 137 during meiosis, and 846 post-meiosis (S1 Table). This observation supports previous studies indicating that most lncRNAs expressed in the testes likely serve post-meiotic functions [52,53].

Evolutionary age-dependent gene expression dynamics in Spermatogenesis

If MSCI is a widespread evolutionary effect, an intriguing question arises about the fate of genes that have recently originated on the X chromosome. In D. willistoni, MSCI appears to be as robust on newly incorporated X-linked regions as it is on the Drosophila genus X chromosome (Fig 2B), suggesting a whole-chromosome response. However, it would also be valuable to explore how MSCI impacts individual genes that arise on an established X chromosome, as this could clarify whether MSCI can act only on new chromosomal blocks or can also target genes individually based on their genomic context. To answer this question, we examined genes that recently emerged on the X chromosome of D. melanogaster [35], comprehensive data on new genes are not available for the other species). Fig 4 offers a thorough overview of the dynamics of gene expression changes across the stages of spermatogenesis, elucidating the influence of age on these intricate processes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Gene expression dynamics by gene age and chromosomal location.

A) Average transcript expression (log₂ normalized) across stages of spermatogenesis for newly evolved genes (blue) and older genes (salmon) located on the X chromosome, autosomes, and dot chromosome, denoted by circles, squares, and triangles, respectively. Error bars indicate the standard error of the mean (SEM). Notably, no newly evolved genes were detected on the dot chromosome. B) Diagrams illustrating haploid and diploid expression dynamics for new genes on the X chromosome and autosomes. The diagrams depict cell phenotypes based on gene expression relative to chromosomal location and dominance effects: turquoise indicates X-linked allele expression (X^a), red denotes autosomal recessive allele expression (A^a), and gray represents autosomal dominant allele expression (A^A). mRNA expression levels are color-coded accordingly. Note that meiosis involves a diploid phase until the completion of meiosis I, followed by a haploid phase. The top rectangles highlight, for each phase, the relative selective advantage of chromosomal location for an adaptive recessive mutation in a newly evolved gene and, consequently, the expected relative expression.

https://doi.org/10.1371/journal.pgen.1011511.g004

In the context of D. melanogaster, genes that emerged before the split of the Sophophora and Drosophila subgenera (i.e., old genes) show a progressive decrease in X-linked expression throughout spermatogenesis. This decline, also observed for dot-linked genes, is particularly pronounced from mitosis to meiosis (Fig 4A, Wilcoxon: p-value < 2.2e-16), coinciding with the stage where X chromosome inactivation occurs. Notably, this phenomenon is absent from the expression of autosomes (Fig 4A) and persists, at least to some extent, in the post-meiotic stages.

In contrast, newly evolved X genes, D. melanogaster genes that originated after the split between Drosophila and Sophophora subgenus, exhibit a markedly higher expression during both the mitotic and meiotic phases (Wilcoxon: p-value ≤ 0.009), surpassing the mean expression of autosomal genes in these phases (Fig 4A). Therefore, it is clear that MSCI mechanisms do not influence newly emerged genes, implying that MSCI might take longer to affect individual new sequences or may be incapable of extending to new individual regions once it is established.

Our three-phase spermatogenic profiling enables a detailed examination of gene expression dynamics not only in mitosis and meiosis but also in the post-meiotic phase. Across various organisms, including mammals, plants, and Drosophila, new genes, primarily located on autosomes, tend to show heightened expression in the later stages of the male germline [38,54,55]. This pattern is linked to the advantages of haploid selection in the early evolution of recently emerged genes, as recessive adaptive mutations can be directly exposed to natural selection during this stage [38] (Fig 4B). Previous studies have also shown that higher expression in later phases of spermatogenesis in Drosophila is associated with increased levels of positive selection acting on new genes, suggesting that post-meiotic expression may be a key driver of their adaptive evolution [38]. Our RNA-seq analysis provides at least partial support for these findings, showing that, on average, newly evolved autosomal genes display a gradual increase in expression throughout spermatogenesis (Fig 4A), in agreement with the previously described expression pattern.

A novel insight, however, comes in the post-meiotic phase, where we observe a convergence in mean expression levels between newly evolved autosomal and X-linked genes (Fig 4A, Wilcoxon: p-value = 0.68). This pattern suggests a broad selective advantage for new genes in haploid expression: while the hemizygous state of X-linked genes offers this benefit consistently across all germline phases, autosomal genes experience this advantage exclusively in the post-meiotic phase (Fig 4B).

Importantly, the fact that new X-linked genes already exhibit elevated expression during mitosis, before the onset of MSCI and also prior to any post-meiotic transcript accumulation, indicates that their higher expression levels cannot be attributed solely to increased transcript stability, reduced translation, or the absence of post-transcriptional regulation. Rather, these differences likely reflect genuinely elevated transcriptional activity.

Fly stocks and sample preparation

By incorporating species with diverse evolutionary relationships and karyotypes, along with maintaining comparable testis morphology, the study aims to uncover evolutionary patterns in sex chromosome expression regulation while controlling for genetic and morphological variations. The Drosophila testis, exhibiting variations in size across the studied species, consistently maintains a characteristic shape among the selected ones: a long, coiled tube, presenting a thin and elongated structure [69]. The similarity in the tubular shape among these species reflects commonalities in how spermatogenic developmental cells are arranged sequentially, extending from the apical to the distal end (Fig 1A). Drosophila spermatogenesis starts in the apical end, which is enriched with stem hub cells, spermatogonial stem cells, and interconnected spermatogonia, corresponding to the mitotic cells [39]. The spermatogonial cells undergo four rounds of mitotic divisions, and after completing the premeiotic S-phase, they undergo enlargement and extensive transcriptional changes, generating cysts with 16 primary spermatocytes. At this point, they are concentrated in the proximal region and their volume increases by 20–25 times, followed by cysts with 32 secondary spermatocytes. The distal region is enriched with post-meiotic cells, such as 64 round spermatid cysts, elongated spermatids, and spermatozoa [39,70]. Samples of the different phases of spermatogenesis (mitosis, meiosis, and post-meiosis) were prepared following the methods outlined in [11] for D. melanogaster, with minor adjustments for the other species regarding their cell distribution (see S1 Fig). Briefly, we dissected testes (without seminal vesicles) in PBS from virgin males aged 6–10 days after eclosion, using 0.25 mm diameter insect pins. Fifty to one hundred dissections per sample were isolated and stored in RNAlater at -20 °C until extraction.

RNA isolation, RNA-seq library generation, and quality control

Total RNA was extracted using PicoPure RNA Isolation Kit (Arcturus). Until the sequencing and quality control, RNA samples were stored in RNAstable (Biomatrica). RNA hydration, quality control tests, library preparation, and sequencing were done at the Genomics Facility at the University of Chicago. Quality control of the samples was carried out on a Bioanalyzer 2100 instrument (Agilent). Stranded multiplexed libraries were sequenced on the HiSeq Illumina 4000 platform. The D. melanogaster libraries were sequenced using a 50 bp single-end protocol, while the remaining species were sequenced using 100 bp paired-end. Each combination of species and spermatogenesis phase was done with at least three biological replicates and 30 million fragments per replicate.

Expression analyses and Data Quality Control

In all four cases, we included coding (CDS) and non-coding sequences (lncRNA). For D. melanogaster, we used the reference genome Release 6.34 (ref FlyBase, May 2020). For the other three species, (D. willistoni, D. mojavensis, and D. simulans), we started from NCBI RefSeq releases 101, 101, and 103, respectively (NCBI currently serves as the primary platform for genome and annotation updates for these species). Notably, for the Y chromosome, the original sequences from the designated releases were substituted with sequences identified by [71] for D. mojavensis, and [72]for D. willistoni. To refine the dataset, CD-HIT (version: 4.8.1) was employed to eliminate redundant sequences, following the methodology by [73] and [74].

FastQ reads were aligned to the corresponding reference transcriptome with the RSEM pipeline [75] using bowtie2 [76] as the aligner. Estimated counts were imported to R with Tximport [77] for differential expression analysis with DeSeq2 [78], which employs paired Wald tests with false discovery rate (FDR) correction [79]. Pearson correlations were estimated using log2 transformations (S3 Fig). To identify different spermatogenic expression profiles, K-means clustering was applied to all transcriptomes, including the previous microarray-generated database for D. melanogaster [11].

Chromosomes’ locations were collected from the reference annotations for D. melanogaster and D. simulans. For D. willistoni and D. mojavensis, scaffolds were mapped to chromosomes using genome coordinates according to [80] (S5 and S6 Tables). The ancestral dot chromosome was fused to the Muller E element in the D. willistoni lineage [80,81]. For this species, we used the mapping information provided by [80], in which genes proximal to CG17119-PA (E), were considered as dot genes (scaffold scf2_1100000004943. See S6 Fig). Ortholog relationships among the genes of the four species were retrieved from FlyBase (2020).

List of new genes identification

The dataset of new genes in Drosophila was obtained from the GenTree platform (http://gentree.ioz.ac.cn/), [82]. These genes are categorized into six age groups based on their age [35]. The youngest group is exclusive to Drosophila melanogaster and is labeled as branch 6, with subsequent branches denoted accordingly. We grouped branches 1–6, corresponding to new genes (less than 62 million years old), and Branch 0, which comprises orthologs found both in the Sophophora and Drosophila subgenus, forming the old genes group (S1 Table).

Calculation of the expression ratio of X-linked to autosomal genes

To calculate the expression ratio of X-linked to autosomal genes (X:AA), we applied the filter-by-fraction strategy described by [43]. This method selects an identical proportion of highly expressed genes from the X (or Z) chromosomes and the autosomes to produce an unbiased estimate of the X:AA ratio in bulk RNA-seq data. For each TPM threshold (>0, > 2, and >10), we computed the fraction of X-linked and autosomal genes exceeding the threshold and used the smaller of the two fractions to extract the top expressed genes from both categories. For D. willistoni, we extended this approach to include the neo-X chromosome, selecting the smallest fraction among X-linked, neo-X, and autosomal genes.

The X:AA ratio was then calculated by dividing the median expression of the top X-linked genes by the median expression of the top autosomal genes [44]. Autosomal genes were defined as those located on Muller elements B, C, D, and E in D. melanogaster, D. simulans, and D. mojavensis, and on elements B, C, and E in D. willistoni, where Muller D was classified as part of the neoX.