Skip to main content
NCBI home page
Genetics logoLink to Genetics
. 2021 Nov 17;220(1):iyab207. doi: 10.1093/genetics/iyab207

Population biology of accessory gland-expressed de novo genes in Drosophila melanogaster

Julie M Cridland 1,✉, Alex C Majane 1, Li Zhao 2, David J Begun 1
Editor: M Hahn
PMCID: PMC8733444  PMID: 34791207

Abstract

Early work on de novo gene discovery in Drosophila was consistent with the idea that many such genes have male-biased patterns of expression, including a large number expressed in the testis. However, there has been little formal analysis of variation in the abundance and properties of de novo genes expressed in different tissues. Here, we investigate the population biology of recently evolved de novo genes expressed in the Drosophila melanogaster accessory gland, a somatic male tissue that plays an important role in male and female fertility and the post mating response of females, using the same collection of inbred lines used previously to identify testis-expressed de novo genes, thus allowing for direct cross tissue comparisons of these genes in two tissues of male reproduction. Using RNA-seq data, we identify candidate de novo genes located in annotated intergenic and intronic sequence and determine the properties of these genes including chromosomal location, expression, abundance, and coding capacity. Generally, we find major differences between the tissues in terms of gene abundance and expression, though other properties such as transcript length and chromosomal distribution are more similar. We also explore differences between regulatory mechanisms of de novo genes in the two tissues and how such differences may interact with selection to produce differences in D. melanogaster de novo genes expressed in the two tissues.

Keywords: Drosophila, evolution, RNA-seq, de novo genes, accessory gland, testis

Introduction

The especially rapid divergence of male-limited behavioral and morphological phenotypes in many animal lineages, presumed to be a consequence of various forms of sexual selection, is mirrored in the genome, most conspicuously in the portion functioning specifically in male-specific reproductive tissues. This rapid divergence of genes exhibiting male-biased or male-specific expression applies to several evolutionary phenomena observed in Drosophila, including protein sequence evolution (e.g., Coulthart and Singh 1988; Swanson et al. 2001; Wagstaff and Begun 2005; Haerty et al. 2007), the evolution of canonical gene duplications (e.g., Wagstaff and Begun, 2005; Mikhaylova et al. 2008; Belote and Zhong 2009; Sorourian et al. 2014), gene expression divergence (e.g., Meiklejohn et al. 2003; Zhang et al. 2007), and the origination of genetic novelties, such as retrogene duplications (Long and Langley 1993; Betrán et al. 2002) and de novo genes (e.g., Begun et al. 2006; Levine et al. 2006), the last of which is the focus of this report.

We define de novo genes here as DNA sequences producing derived transcripts, coding or noncoding, that are independent of mature ancestral transcripts and located in ancestrally intergenic or intronic DNA. Such genes have been found in several taxa, including Drosophila (Begun et al. 2006, 2007; Levine et al. 2006; Zhou et al. 2008; Zhao et al. 2014), rodents (Heinen et al. 2009; Murphy and McLysaght 2012; Neme and Tautz 2013; Casola 2018), primates (Knowles and McLysaght 2009), plants (Zhang et al. 2019), and fungi (Cai et al. 2008; Li et al. 2010; Carvunis et al. 2012; Vakirlis et al. 2018). Nevertheless, the abundance (Casola 2018), persistence times (Palmieri et al. 2014), and functions of de novo genes (e.g., Cai et al. 2008; Heinen et al. 2009; Li et al. 2010) remain unclear. While the identification of de novo genes could be viewed operationally as an annotation problem, it is challenging for several reasons, some of which may derive from the properties of de novo genes themselves (e.g., low expression levels; Zhao et al. 2014) and others of which derive from the fact that identification of de novo genes relies upon marshaling evidence in support of gene absence in orthologous DNA of nonfocal species (reviewed in Van Oss and Carvunis 2019).

The first experimental investigation of de novo gene evolution (Begun et al. 2006) took place in the context of the accessory gland (AG), which produces among other molecules, secreted proteins that are transferred to the female along with sperm during mating. These molecules are required for fertility, mediate a number of female postmating physiological responses, and may also influence female sperm storage and sperm competition (reviewed in Wilson et al. 2017; Wigby et al. 2020). In that work, early Drosophila genome assemblies from the melanogaster subgroup (Begun et al. 2007; Clark et al. 2007) and AG-derived expressed sequence tags from cDNA libraries were used to reveal evidence of several small, AG-expressed genes in Drosophila yakuba or Drosophila erecta that appeared to be unexpressed in related species, but for which orthologous, syntenic sequence could be identified (Begun et al. 2006). Those genes were hypothesized to have recently originated de novo from ancestral intergenic DNA in D. yakuba, D. erecta, or their common ancestor.

An early annotation-based phylogenetic investigation of de novo gene evolution in the melanogaster subgroup (Levine et al. 2006) found that de novo genes often exhibit testis-biased or testis-specific expression, thereby providing the first clue that D. melanogaster de novo genes may be biased toward male reproductive functions. A similar pattern was observed in the obscura group of Drosophila (Palmieri et al. 2014). To investigate the youngest class of de novo genes, Zhao et al. (2014) carried out a detailed population level investigation of testis transcriptomes in a sample of six D. melanogaster inbred genotypes, which detected 106 putative de novo genes that had fixed since the split from common ancestor with Drosophila simulans, and 142 that were segregating in D. melanogaster (Zhao et al. 2014). Many of the polymorphic testis-expressed de novo genes occurred at intermediate or high frequency, and population genetic evidence suggested that these genes had been influenced by directional selection. Nevertheless, because that study investigated only the testis it could not speak to the question of whether the abundance and population biology properties of de novo genes in the testis are typical or atypical. Thus, while the Drosophila data point overall to a role for de novo genes in the evolution of genetic novelty in both testis and AG transcriptomes, comparison of the abundance and properties of de novo genes expressed in these two organs awaits more thorough investigation of the D. melanogaster AG.

Several functional or evolutionary attributes of AG function could facilitate the origin and spread of AG-expressed de novo genes. First, AG-specific proteins tend to be small (e.g., Findlay et al. 2009). If de novo genes are protein-coding and often originate from ancestrally noncoding DNA carrying latent open reading frames (ORFs), shorter de novo genes would be more common than longer ones simply because ceteris paribus, shorter latent ORFs are more abundant than longer latent ORFs in noncoding eukaryotic DNA. Second, while the majority of secreted seminal fluid proteins require a signal peptide, the protein sequence constraints for signal peptides are fairly lax (Nielsen et al. 1997), which might also lead to high origination rates of novel proteins competent for secretion (Begun et al. 2006). Third, while many protein functional domains are widely shared among seminal fluid proteins, several Drosophila seminal fluid proteins have no known functional domains (Findlay et al. 2008). Genes coding for such proteins may be more likely than many functional gene classes to have atypical structures or functions, and thus, more likely to arise de novo. Finally, regardless of whether most de novo evolved genes are coding vs noncoding (see below), strong selection favoring novelty in male–male or male–female interactions could facilitate the spread of de novo AG-expressed genes. Thus, while the fixation rate of de novo genes will always depend on the cellular processes underlying the expression of noncoding or nongenic DNA (Begun et al. 2006), the breadth of properties of novel proteins or RNAs capable of functioning in a particular tissue or cell type, and the strength of selection acting on evolutionary novelties in a given tissue or cell type, the investigation of how these factors may interact to influence de novo gene origination and fixation is still in its infancy.

One possibility is that the phenomena promoting the spread of Drosophila de novo genes are relatively homogeneous across tissues relating to male-specific reproductive functions. In that case, we would expect roughly similar contributions of de novo genes to testis and AG transcriptomes, and roughly similar dynamics (selected or neutral) of de novo gene spread. Alternatively, differences between the tissues could reflect the strength and/or nature of selection. For example, greater fixation rates of AG-expressed de novo genes might indicate stronger selection on novelty associated with male–male or male–female postcopulatory phenotypes, while greater fixation rates of testis-expressed de novo genes could be indicative of greater selection favoring novelty in germline phenomena, perhaps due to genomic conflicts associated with sex-ratio X chromosomes (Levine et al. 2006) or transposable elements. Differences between the regulatory environments of AG vs testis cells could affect the rate at which novel transcripts originate. Finally, functional differences between cells and tissues in the testis vs AG could lead to differences in the universe of novel proteins or RNAs that are not strongly deleterious and thus could be exposed to positive selection in different cellular milieus. Here, we begin addressing some of these questions through an investigation of young AG-expressed de novo genes that originated in D. melanogaster since the split from its sibling species, D. simulans.

Materials and methods

Fly strains, datasets, and sequencing

Most of the data used here are described in detail in Cridland et al. (2020). Briefly, we dissected AG + anterior ejaculatory duct (referred to throughout as AG) of 2-day-old virgin males from six highly inbred D. melanogaster strains established by the Drosophila Genetic Reference Panel (DGRP; Mackay et al. 2012): RAL-304, RAL-307, RAL-357, RAL-360, RAL-399, and RAL-517. We carried out an allelic imbalance assay using F1 flies generated from crosses among DGRP lines (RAL-307 male × RAL-304 female, RAL-357 male × RAL-399 female, RAL-360 male × RAL-517 female). Two inbred strains of D. simulans were used. One strain, w501, was the strain used for the D. simulans reference sequence (Begun et al. 2007); a second strain, Lara10, was established from flies collected September 2011 in Homestead, FL, and was sib-mated for 10 generations in our laboratory (Zhao et al. 2014). We used the D. yakuba reference sequence strain, Tai18E2 (Begun et al. 2007) as our second outgroup for most analyses (existing data from a third outgroup, Drosophila ananassae strain 14021-0371.13, was also used for some analyses—Yang et al. 2018). All flies were reared on standard cornmeal medium at 25°C under a 12:12 light/dark cycle. RNA was extracted using Trizol (Invitrogen). RNA-seq libraries were made and sequenced as described in Cridland et al. (2020), resulting in paired-end, 100 bp reads.

De novo transcriptome assemblies

Trinity (Grabherr et al. 2011; v2.11) was used to create de novo transcriptome assemblies for each species, using both individual strains and pooled data across strains from each species (Supplementary Table S1). Based on the k-mer distribution generated by Jellyfish V1.1.5, a k-mer of 25 was used for assembly. We also used Trinity using default parameters to carry out de novo transcriptome assemblies for eight tissues and both sexes for D. yakuba and D. ananassae (Yang et al. 2018).

Criteria for calling D. melanogaster AG-expressed de novo genes

We used BLAST (v. 2.10.1+, Altschul et al. 1990) to screen all transcripts from our D. melanogaster AG de novo transcriptome assemblies that were greater than 300 bp long for matches to known genes or transcripts from D. melanogaster (v. 6.34), D. simulans (v. 2.02), and D. yakuba (v. 1.05). The files used for screening were fasta files from each species containing records for CDS, exon, 5′UTR, miRNA, miscRNA, ncRNA, pseudogene, transcript, transposon, tRNA, and 3′UTR (downloaded from www.flybase.org, July 23, 2020; Thurmond et al. 2019). To consider a transcript, a match we required 80% identity over at least 100 bp. We also separately aligned transcripts to D. melanogaster introns to identify potential intronic de novo gene candidates. Transcripts that matched only intronic or intergenic sequences were retained for further analysis. To further reduce the likelihood of erroneously inferring de novo gene status for ancestral genes unannotated in D. melanogaster we also screened our D. melanogaster AG transcripts against Trinity-generated de novo transcriptome assemblies from our outgroup AG RNA-seq data, as well as against Trinity-generated de novo assemblies derived from D. yakuba and D. ananassae libraries from eight different tissues (Yang et al. 2018, files downloaded from SRA January 2021). Thus, all D. melanogaster transcripts matching existing D. melanogaster or outgroup gene annotations, or any outgroup transcripts we assembled, were removed from further consideration.

To reduce the likelihood of mistaking an unannotated D. melanogaster exon of an ancestral gene for a de novo gene we required each candidate, intergenic or intronic, to be at least 500 bp from any annotated exon boundary. We then generated a GTF file for this set of candidates, including all transcripts of each candidate de novo gene, combined these new records with the D. melanogaster v6.34 GTF file, and estimated TPMs for each of the six inbred RAL lines separately. Candidates with a TPM ≥ 1 in one or more RAL lines were retained.

To confirm that the remaining de novo gene candidates reside in orthologous DNA in all three main species we performed a microsynteny analysis by identifying the nearest neighbor genes of each candidate and identified their orthologs in D. simulans and D. yakuba. The location of these orthologs in the outgroup genomes was used to confirm that the candidate gene plus 5 kb upstream and downstream of the transcript start and stop aligned to the syntenic region. Most synteny analyses were carried out using a perl script to compare the positions of the candidate to the syntenic region. The remainder were checked manually, largely due to small deletions in the outgroup(s) in the syntenic region that resulted in the perl script flagging the candidate for a manual check.

The final list of D. melanogaster candidates have the following attributes: they do not overlap existing exon annotations in D. melanogaster, D. simulans, or D. yakuba and are at least 500 bp from known exons, they are expressed at TPM ≥ 1 in at least one inbred D. melanogaster genotype, they reside in orthologous regions of all three species, and they exhibit no evidence of expression in any tissue of any of the three outgroup species (D. simulans, D. yakuba, or D. ananassae). Notably, this approach is expected to be more conservative than the one previously used for our analysis of the testis (Zhao et al. 2014), as we impose no minimum expression level on outgroup TPM estimates and we include a substantial amount of new outgroup transcriptome data. Segregating genes are defined as those for which at least one line expresses at TPM ≥ 1 and at least one expresses at TPM < 1 (Cridland et al. 2020). “Fixed” genes are those expressed at TPM ≥ 1 in all six Raleigh inbred lines. Once a final list of de novo genes was generated, we estimated TPMs in the three Raleigh F1s.

Sequence alignments, variant calling, and allelic imbalance

Our methods closely follow Cridland et al. (2020) which generally follows McManus et al. (2010). Briefly, parental RAL TPM estimates and corresponding estimates from their F1s were used to partition variation into cis and trans effects. We used a fold-change cutoff of 1.25 to call differences in (1) expression between RAL parents, (2) between parent-specific estimates in hybrids, and (3) between the observed overall F1 expression and the expected F1 expression assuming additivity. To reduce the influence of noise on inferring cis- and trans-effects we restricted the analysis to genes for which at least one parent expressed at TPM ≥ 1 and the other parent expressed at TPM < 0.2. We further required at least 10 unique fragments from the F1s to include the observation. Genes were categorized as exhibiting cis- or trans-effects as described in Cridland et al. (2020).

Coding potential and signal peptide prediction

We used the Coding Potential Assessment Tool (CPAT; Wang et al. 2013) to estimate the probability that a transcript was derived from a protein-coding gene vs noncoding gene, generating sets of the top five ORFs per transcript. We used SignalP 5.0 (Armenteros et al. 2019) to determine the probability of a signal sequence for each of the most likely ORFs predicted by CPAT.

Ancestral AG-biased genes

To identify ancestral AG-biased genes, we used the male data for several tissues from FlyAtlas 2 (Leader et al. 2018). We defined AG-biased genes as those that: (1) had FPKM ≥ 1 in the AG, (2) exhibited the highest expression in the AG relative to other male tissues, and (3) exhibited strong AG bias, with estimated tau (Yanai et al. 2005) ≥ 0.9.

Physical distribution of de novo genes

We investigated the physical distribution of de novo genes at several scales, ranging from entire chromosome arms to regions of a few kilobases. To determine whether de novo genes tend to be co-localized with ancestral AG-biased genes we segmented the genome into 500- and 100-kb nonoverlapping windows and asked whether de novo genes are more likely than expected to fall in windows harboring ancestral AG-biased genes. Some de novo genes are tandemly located, defined here as being adjacent and ≤10 kb apart. To ascertain whether such de novo genes tend to show correlated expression across genotypes, we compared the frequency with which genes in clusters tend to be expressed or not expressed together across genotypes. We first converted the TPMs of each de novo gene in a cluster to 0 (if TPM < 1) or 1 (if TPM ≥ 1). We then permuted the expression of each gene over the six RAL lines to generate 1000 sets of de novo genes with expression randomized over lines but with the number of lines expressing each de novo gene preserved. We then calculated an index for each gene cluster that summarized the number of RAL lines that had the same expression value, 0 for not expressed or 1 for expressed, for all the genes within a given cluster. The mean for this index was then calculated across all de novo gene clusters for observed and permuted data. To further examine if expressed genes in a cluster were more likely to be expressed in the same RAL line(s) than would be expected, we calculated a second index where we determined the number of RAL lines per cluster for which all de novo genes were expressed and then compared the mean across all clusters to the distribution of means from the permuted data.

Comparison to testis de novo genes

To compare AG de novo genes to previously identified testis-expressed de novo genes from the same inbred lines (Zhao et al. 2014) we applied the methods described above to the previously reported data from the testis (Zhao et al. 2014), which enables direct comparison of the putative de novo genes expressed in the two tissues. We used BLAST to compare the testis candidates to D. simulans and D. yakuba annotations to transcript assemblies made from our D. simulans and D. yakuba AG + ejaculatory duct libraries, and to outgroup de novo transcript assemblies made from RNA-seq data from several D. yakuba and D. ananassae tissues (Yang et al. 2018), as described above. To enable direct comparisons of cis- and trans- effects on de novo gene expression for testis and AG, we subjected the set of previously identified testis-expressed de novo gene candidates to the same pipeline described above used for candidate AG-expressed de novo genes.

Results

Basic attributes of AG-expressed de novo genes

We identified a total of 133 candidate de novo genes (49 intergenic and 84 intronic; Supplementary Table S2) in the six DGRP strains, of which 131 were segregating and two (both intronic) were “fixed” (expressed at TPM >1 in all six Raleigh inbred lines); 99 genes were expressed in only one strain, while 34 were expressed in more than one strain. For convenience, we sometimes refer to these genes as “de novo genes” rather than “candidate de novo genes,” despite the absence of evidence for genic function. While ancestral genes nested in introns of annotated genes are strongly biased (71.3%) toward being on the opposite strand in Drosophila (Lee and Chang 2013), multiexonic candidate de novo genes located in introns of ancestral genes were roughly equally likely to be on the same strand vs opposite strand (n = 10 and 9, respectively), which is significantly different from ancestral genes (binomial probability, P = 0.025). The mean (median) number of AG-expressed de novo genes per line across the six inbred lines was 33.5 (32.5). The greatest and least number of candidates were expressed in line RAL 517 (54 expressed genes) and RAL 307 (21 expressed genes), respectively. We observed no significant correlation between the mean TPM of ancestral AG-biased genes for each line and the number of candidate de novo genes it expressed.

To compare the number of AG-expressed de novo genes to that observed in the testis we first reassessed the candidate genes reported in Zhao et al. (2014), all of which were intergenic as constrained by the filtering process used. Of the 106 fixed and 142 segregating genes reported in Zhao et al. (2014) we discovered apparently homologous transcripts, defined as BLAST matches to outgroup databases of ≥80% over ≥100 bp in one or more outgroups, for 53 fixed and 27 segregating candidates (Supplementary Table S3). Therefore, for the same strains used here for the AG, our current conservative estimate for fixed and segregating testis-expressed de novo genes are 53 and 115, respectively. Comparing the number of intergenic de novo genes expressed in the two tissues relative to the total number of annotated genes expressed in those tissues at TPM ≥ 1 reveals greater than twofold more testis-expressed than AG-expressed de novo genes. We observed a median/mean of 10/12 intergenic AG-expressed genes per strain, while the median number of intergenic testis-expressed de novo genes expressed per strain (in the same six RAL strains) was 99. Testis-expressed de novo genes were more likely to be expressed in more than one RAL strain (70% of candidates) than AG de novo genes (26% of candidates; Fisher’s exact test; P = 7.6 × e−15). Overall then, it seems safe to conclude that intergenic de novo genes make a substantially smaller contribution to AG transcriptome complexity than to testis transcriptome complexity.

The mean length of the longest transcript for AG intergenic candidate genes, 701 bp, was shorter than the mean for intergenic testis-expressed de novo genes (935 bp, Zhao et al. 2014; t-test, P = 0.001). Nineteen AG-expressed genes were associated with a transcript > 1000 bp; the longest observed transcript was 2214 bp. There was no significant difference in the length of the longest transcript per gene for intergenic vs intronic candidates (Wilcoxon test; P = 0.52). Most genes (106/133, 80%) were single-exon; the maximum exon-number for any transcript was three. While the majority (94%) of intron splice junctions were canonical GT/AG, this proportion is significantly smaller than that observed for ancestral genes (binomial; P < 0.001; Crosby et al. 2015). Twenty-eight genes (including six single-exon genes) exhibited multiple transcripts, while the maximum number of transcripts for any gene was four. The maximum number of transcripts/gene and exons/gene were both positively, though weakly correlated with expression levels (mean of expressing RAL line TPM estimates; P = 3.5 × e−4 and P = 9.4 × e−3, respectively), similar to Zhao et al. (2014).

As expected, the expression of many candidate genes (TPMs: expressed mean 2.67, median 1.39) is low relative to that of ancestrally expressed genes (TPMs: mean 106, median 6.2). However, Supplementary Figure S1, which shows the distribution of max TPM across strains for all candidates, reveals that a considerable number of genes show relatively high maximum across-strain TPMs. For example, 18 genes exhibited a maximum TPM ≥ 5, with the most highly expressed gene expressed at maximum TPM = 53 (though this gene is only expressed in two strains). Genes expressed at a higher level (mean of nonzero TPM estimates) tended to be expressed in more lines (Kruskal–Wallis test, P = 1.6 × e−5). Comparing singleton genes (expressed in only one genotype) vs nonsingleton genes, we find that nonsingletons have a greater median longest transcript (696 vs 485, P = 1.36 × e−6) and greater median maximum TPM (3.26 vs 1.25, P = 4.46 × e−8). Thus, the overall picture is consistent with the literature—candidate de novo genes tend to be short, simple, and lowly expressed, some candidate genes, nevertheless, exhibit high expression and multiple transcripts, and there is a general trend for longer transcripts and greater expression to be associated with genes expressed in more genotypes (Zhao et al. 2014).

Coding potential

Using CPAT with default settings based on a D. melanogaster training set (Wang et al. 2013) we identified the top five ORFs for each of the 170 de novo transcripts (corresponding to 133 genes) and for each ORF determined the coding vs noncoding likelihood. An ORF was identified by CPAT for 165 of 170 transcripts. This analysis revealed that 98% of genes (131 of 133) and 99% of transcripts (163/165) were predicted to be noncoding. Both predicted coding transcripts were X-linked and expressed in more than one RAL strain. To investigate whether the unusually short transcript length of de novo gene candidates relative to the lengths of ancestral annotated protein-coding genes used to train CPAT was potentially biasing this conclusion, we used CPAT to categorize all annotated D. melanogaster protein-coding transcripts as either coding or noncoding, and then binned these annotated protein-coding genes by transcript length in 100-bp increments up to 2 kb (Supplementary Figure S2). Assuming that annotated protein-coding status in Flybase is correct, this analysis revealed that very short protein-coding genes are more likely than other protein-coding genes to be predicted by CPAT to be noncoding. To determine whether this possible bias could influence our conclusion about de novo gene coding probability, we used a resampling procedure. We made random draws without replacement from annotated D. melanogaster transcripts to generate sets of 165 transcripts with the same length distribution (based on length bins in 100-bp increments) as the observed de novo gene candidates, and repeated this 1000 times. We then compared the predicted coding probability of the highest scoring ORF for each annotated transcript to that of each de novo gene transcript. In none of the 1000 permutations did we observe a fraction of coding genes as small or smaller than the observed value, with the most extreme permutation still exhibiting 112/165 transcripts as compared to only 2/165 observed. On average 130/165 transcripts from these draws exhibited a coding probability score ≥0.39 (the cutoff for calling a D. melanogaster gene as coding), yielding a binomial test P < 1 × e−7. This suggests that the finding that a substantial proportion of candidate de novo genes is likely noncoding is probably not attributable solely to artifacts or biases associated with CPAT.

The five most likely ORFs predicted for each transcript by CPAT were also used to determine the probability that the corresponding predicted proteins harbored a signal sequence (SignalP 5.0; Armenteros et al. 2019). Only six were predicted to be secreted; none of the predicted proteins showed similarity to known proteins. Thus, if these genes are in fact protein coding, it seems unlikely for most that their products are transferred to females during mating. Using male data from FlyAtlas 2 and a cutoff of tau > 0.9 (Yanai et al. 2005) revealed 538 ancestral strongly AG-biased genes. Of these, 324 were identified as coding based on the FlyBase annotation, 239 of which (73.8%) contained predicted signal sequences (Armenteros et al. 2019). In contrast, 209 ancestral AG-biased genes were annotated as noncoding in FlyBase. To investigate the possibility that some of these putatively noncoding genes are actually coding genes producing secreted proteins we used CPAT to identify associated ORFs. For those genes for which CPAT identified at least one ORF, we used SignalP to determine whether the most likely ORF contained a predicted signal sequence. Of the 209 AG-biased genes annotated as noncoding, 38 (18%) contained predicted signal sequences (Supplementary Tables S4 and S5). It is quite plausible that these noncoding genes are misannotated and in reality code for seminal fluid proteins. In addition, ORFs of 12 AG-biased ncRNAs match FlyBase polypeptide sequences (Supplementary Tables S4 and S5), including three of the 38 with predicted signal sequences. Thus, we speculate that 47 of the 209 AG-biased genes annotated as noncoding are likely to be coding. Nevertheless, the proportion of candidate de novo genes predicted to be noncoding (0.98) is much greater than the proportion of ancestral AG-biased genes likely to be noncoding (162/538 = 0.30; binomial P = 1.45 × e−67). Thus, AG-expressed de novo genes appear much more likely to be noncoding relative to ancestral AG-biased genes.

Physical distribution of candidate genes at different scales

To investigate the genomic distribution of de novo genes we first asked whether the proportion of genes on each chromosome arm differs from that observed for all ancestral AG-biased genes. Consistent with previous reports for male-biased genes in general (Sturgill et al. 2007) and seminal fluid protein genes and AG-biased genes specifically (Findlay et al. 2008; Meisel et al. 2012), ancestral AG-biased genes are underrepresented on the X (n = 33) relative to the major autosomes (n = 504, or 126 per major autosomal arm; Fisher’s exact test P = 3.2 × e−10). Across autosomes, arm 2L (n = 159 genes) is significantly enriched for these genes, also consistent with the literature (Findlay et al. 2008; Table 1). Candidate de novo genes exhibit similar chromosomal patterns; relatively few are located on the X, while 2L harbors the greatest number. There is no significant deficit of AG-expressed X-linked de novo genes relative to the expected value based on the fraction of all annotated genes that are X-linked. However, direct comparison of de novo genes and ancestral AG-biased genes on the X chromosome vs autosomes reveals that the X/A ratio for de novo genes (16/117 = 0.14) is about twice that of ancestral genes (33/505 = 0.065; Fisher’s exact test; P = 0.025). Thus, whatever processes lead to the strong autosomal bias of ancestral AG-biased genes (Meisel et al. 2012) are weaker for de novo genes. Because these de novo genes are polymorphic and expressed at a low level, reduced expression constraints associated with X-linkage and dosage compensation (Meisel et al. 2012) might contribute to this pattern. We then determined the X/A ratio for testis-expressed de novo genes (Zhao et al. 2014) and ancestral testis-biased genes (tau > 0.9 and highest male expression in the testis in male FlyAtlas 2 data). For testis-expressed de novo genes the X/A ratio (14/154 = 0.09) is smaller than the ratio for ancestral testis-biased genes (411/2580 = 0.16; Fisher’s exact test, P = 0.048). Thus, while the two classes of de novo genes, AG- and testis-expressed, are very similar in their X/A distributions (Fisher’s exact test; P = 0.33), deviations from comparable ancestral-biased genes X/A distributions are very different, with testis-biased candidate de novo genes showing X underrepresentation and AG-biased candidate genes showing X overrepresentation. Interestingly, while both testis- and AG-biased ancestral genes are underrepresented on the X, the effect is roughly twofold greater for AG-biased genes (cf. Meisel et al. 2012).

Table 1.

AG de novo candidate genes

Chromosome AG de novo genes De novo enrichment AG-biased genes AG-biased enrichment Total genes
2L 43 (32.3%) 7.14E–04 159 (29.6%) 3.32E−08 3559 (20%)
2R 20 (15%) 1.33E–01 92 (17.1%) 2.18E–02 3673 (20.6%)
3L 15 (11.3%) 1.58E–02 119 (22.1%) 7.41E–02 3501 (19.6%)
3R 37 (27.8%) 3.09E–01 134 (24.9%) 2.71E–01 4262 (23.9%)
X 16 (12%) 3.97E–01 33 (6.1%) 5.70E−11 2706 (15.2%)
4 2 (1.5%) 2.15E–01 1 (0.2%) 8.62E–01 116 (0.6%)
Open in a new tab

To investigate within chromosome-arm heterogeneity and the possible connection between the locations of AG-expressed de novo genes and ancestral AG-biased genes, we segmented each chromosome arm into 500-kb windows and then asked whether windows harboring an ancestral AG-biased gene were also more likely to harbor a de novo gene. We found that consistently across all arms (genome-wide Fisher’s exact test; P = 2.6 × e−8), de novo genes were much more likely to reside in windows containing an ancestral AG-biased gene, with enrichments ranging from about two to fivefold (Table 2). This pattern remained when we considered only ancestral AG-biased genes annotated as noncoding (Supplementary Table S6). Thus, there is strong evidence of chromosomal domains of correlated expression for de novo genes and ancestral AG-biased genes, consistent with previous studies of correlated patterns of gene expression along Drosophila chromosomes (Spellman and Rubin 2002; Boutanaev et al. 2002; Parisi et al. 2004).

Table 2.

AG-biased vs de novo genes in 500-kb windows

Chromosome AG-biased genes absent
 AG-biased genes present
 AG de novo genes/AG-biased genes absent (%) AG de novo genes/AG-biased genes present (%) Fisher’s exact test
De novo genes absent De novo genes present De novo genes absent De novo genes present
All 104 19 84 73 5.45 46.50 2.60E−08
2L 9 1 19 19 10.00 50.00 3.10E−02
2R 19 4 15 13 17.39 46.43 3.90E−02
3L 19 1 26 11 5.00 29.73 4.10E−02
3R 22 6 15 22 21.43 59.46 2.60E−03
X 26 6 9 7 18.75 43.75 9.00E−02
4 1 1 0 1 50.00 100.00 1.00E+00
Y 8 0 0 0 NA NA NA
Open in a new tab

At a smaller scale, we observed 14 clusters (six in intergenic regions, eight in intronic regions) of either two or three adjacent de novo genes exhibiting less than 10 kb between genes. To investigate whether these very tightly linked de novo genes tend to be expressed in a correlated manner across genotypes (i.e., if the first gene in a cluster is expressed in a line it is likely the second gene is also expressed in that line) we compared the observed expression of clustered de novo genes to that expected under the null hypothesis that the de novo genes in each cluster are independently expressed. We found that for the mean cluster, 4.86 out of six lines exhibited consistent expression of genes in that cluster—that is, across lines either all genes in the cluster were expressed or all were unexpressed. This was substantially more consistent than the permuted data, which exhibited a mean consistency of 3.57 out of 6 (z-score 6.4, P 9.9 × e−11). Focusing just on the correlation for expression (omitting nonexpression as an observation), we found that across all clusters, 0.86 out of six lines on average express all genes in a cluster, which was substantially higher than the correlation in the permuted data, 0.27 out of six lines (z-score 6.1, P 4.9 × e−10). These data support the notion that the regulatory processes underlying the expression of de novo genes in the AG are spatially heterogeneous on multiple scales, ranging from a few kilobases to entire chromosome arms, and contribute to the physical distribution of candidate de novo genes across chromosome arms.

To compare the genomic distribution of AG-expressed de novo genes to testis-expressed de novo genes we repeated the windowing analysis for the testis-expressed de novo genes identified in Zhao et al. (2014; modified as described above) and strongly testis-biased ancestral genes (as described above; Supplementary Table S7). Because the number of testis-biased genes on the major chromosome arms is much higher than the number of AG-biased genes, 3049 vs 538, we compared the tissues using 100-kb windows so that for both tissues we had a sufficient number of windows in each of two categories (with one or more ancestral tissue-biased genes and with zero ancestral tissue-biased genes) for a reasonably powered analysis. At the whole genome level, we see a consistent pattern for both tissues—windows that contain de novo genes tend also to harbor ancestral genes exhibiting that tissue bias (Fisher’s exact test; P = 1.1 × e−4 for AG; P = 3.9 × e−7 for testis). The degree of this enrichment at the 100 kb scale differed between tissues for the X chromosome, however, with a much smaller percentage of windows with testis-biased genes also containing testis de novo genes (6.4%) compared to the AG (18%), even though the total number of windows with testis-biased genes is substantially greater. Thus, it appears that whatever local regulatory phenomena are facilitating or driving de novo gene expression on the X chromosome, these effects are weaker for the testis than the AG, consistent with the greater underrepresentation of X-linked testis-expressed de novo genes.

Finally, to investigate the possible correlation between AG- and testis-expressed genes we first compared the number of ancestral testis-biased genes in windows with vs without ancestral AG-biased genes. We found that windows with AG-biased genes had significantly more testis-biased genes (mean = 3.2) than windows without AG-biased genes (mean = 2; Wilcoxon rank sum test; P = 1.4 × e−13), which supports the notion that our previous observation of correlated expression in the testis and AG (Cridland et al. 2020) could be explained in part by correlated chromosomal gene locations. Similarly, we found that windows with AG-expressed de novo genes contained more testis-expressed de novo genes (mean = 0.3) than windows without AG-expressed de novo genes (mean = 0.1; Wilcoxon rank sum test; P = 1.6 × e−6), suggesting that the de novo gene origination process is physically correlated across the genome for these two tissues.

Regulatory mechanism

We investigated the mechanisms of de novo gene expression using allelic imbalance experiments following Cridland et al. (2020). We had sufficient numbers of observations for only 13 genes. Of these, nine (69%) exhibit both cis and trans effects, three (23%) exhibit only trans effects, and one (7.6%) exhibits only cis effects (Table 3 and Supplementary Table S8). To compare these patterns to those of testis-expressed de novo genes, we reanalyzed 47 segregating testis-expressed de novo genes (Zhao et al. 2014) using the same pipeline and found that 30 (64%) exhibited both cis- and trans-effects, three (6%) exhibited only trans-effects, and 14 (30%) exhibited only cis-effects (Supplementary Table S8). While the proportion of cis-only testis-expressed de novo genes is smaller here than reported in Zhao et al. (2014) as a result of methodological differences, the important point is that for the candidate de novo genes subjected to the same allelic imbalance analysis, those expressed in the AG exhibit dramatically less cis-only regulation compared to those expressed in the testis (Fisher’s Exact test; P = 1.6 × e−09).

Table 3.

Regulation of de novo genes

Regulatory mechanism AG Testis
Cis 1 14
Cis and trans 9 30
Trans 3 3
Open in a new tab

Expression in different cell types

We used AG single-nucleus RNA-seq data from RAL 517 (Majane et al. 2021) to investigate expression of de novo genes in the three major cell types of our dissected bulk tissue: main cells, secondary cells, and ejaculatory duct cells. Of the 13 candidate genes called as expressed (TPM > 1) in RAL 517, 10 were also called as expressed in single-nucleus data. This independent validation of our de novo gene candidates in a different experiment using different technology suggests that our approach for identifying such genes is robust. Imposing a much lower cutoff of TPM > 0.1 to categorize a de novo gene as expressed in bulk transcriptome data from RAL 517 results in 86 expressed genes, of which 39 were also expressed in the single-nucleus data, strongly suggesting that the TPM > 1 criterion is quite conservative. The conservative nature of de novo gene TPM criterion would be especially pronounced for genes that tend to be expressed at a higher level in secondary or ejaculatory duct cells, as those cells are considerably less abundant than main cells in bulk tissue used in the experiment (Majane et al. 2021). It also suggests that we may have underestimated the proportion of RAL strains expressing a given de novo gene candidate.

Majane et al. (2021) used the single-nucleus data to identify marker de novo genes—those that exhibited biased expression across the three major cell types. Of the 43 total de novo gene candidates identified here that were defined as expressed in the single-cell data from RAL 517 (Majane et al. 2021), five were classified as marker genes, all of which showed ejaculatory duct biased expression. This represents a significant enrichment of de novo genes expressed at a high level in this cell type relative to the other two types (hypergeometric, P < 0.0005). Interestingly, four of the five ejaculatory duct marker de novo genes are nonsingletons, which represents a significant enrichment of higher frequency genes (hypergeometric, P = 0.016), though the small sample size precludes strong conclusions about this pattern. Majane et al. (2021) reported that ejaculatory duct-cell transcriptomes tend to evolve more quickly than main cell and secondary cell transcriptomes. Our finding that expression-biased de novo in the AG tends to be ejaculatory duct-biased is consistent with the notion that this cell type may be prone to higher rates of transcriptome turnover compared to main and secondary cells.

Discussion

Our comparisons of de novo gene candidates expressed in the AG and testis revealed several differences that illuminate variation in the processes underlying their origin and evolution. While there were minor differences between the starting material and analyses used in our investigation here of the AG and our previous investigation of the testis, such methodological differences cannot plausibly explain the stark differences between the two tissues in the abundance, chromosomal distributions, regulatory mechanisms, and population frequencies of de novo genes.

The AG expresses many fewer intergenic de novo genes than the testis, often at lower levels. Moreover, most AG-expressed de novo genes are expressed in only one genotype, and only two were fixed in our sample. The testis, in contrast, expresses many more intermediate and high frequency/fixed genes (Zhao et al. 2014), most of which are germline expressed (Witt et al. 2019). Thus, at the population level, the contribution of de novo genes to the somatic male reproductive tissue transcriptome appears much smaller than their contribution to the germline transcriptome.

There are also apparent differences between the two tissues in the way de novo gene candidates are regulated. While strictly cis-acting variation is common for testis-expressed de novo genes, the AG tends to exhibit more complex regulatory variation, including a more substantial trans-acting component. The selective spread or removal of a de novo gene would be more efficient if its expression resulted from a novel, tightly linked cis-regulatory element that co-opted existing trans-acting regulatory factors, as appears to be the case for testis (Zhao et al. 2014), compared to a situation where the underlying genetics of novel expression is more complex, as appears to be the case for the AG. This difference may contribute to the lower fixation rate for AG- vs testis-biased de novo genes and the relative paucity of singleton testis-expressed de novo genes. Larger samples from both tissues would shed light on their potential differences in the relative proportion of rarely expressed de novo genes.

The physical distribution of de novo gene candidates expressed in the AG or testis show some similarities, but also important differences. Both types of de novo genes tend to be found in chromosomal regions harboring ancestral genes exhibiting the same tissue-biased expression, and the locations of testis- and AG-expressed ancestral and de novo genes are correlated. Furthermore, very tightly linked AG-expressed de novo genes, which show no evidence of origination by duplication, show strongly correlated expression patterns across genotypes, also supporting a physically correlated origination process. The fact that ancient genes and young de novo genes show correlated physical distributions for both testis and AG suggests that regulatory phenomena underlying the origination process and/or the selective retention of such genes during evolution play a role in generating the distribution. However, this effect is diminished for X-linked testis-expressed de novo genes, which could contribute to the lower (though not significantly so) X/A ratio for testis- than for AG-expressed de novo genes. Both testis- and AG-expressed de novo genes are underrepresented on the X chromosome, as expected for strongly male-biased genes (Sturgill et al. 2007; Meisel et al. 2012). However, relative to the X/A distributions for their corresponding ancestral tissue-biased genes, AG-expressed de novo genes are roughly twice as likely to be X-linked, while testis-expressed de novo genes are roughly twice as likely to be autosomal. The disparate deviations of young de novo gene X/A distributions from that of their ancestral counterparts could result from effects of heterogeneous origin processes and/or heterogeneous selective processes. For example, X-chromosome inactivation in the male germline (Lifschytz and Lindsley 1972; Kemkemer et al. 2011; Landeen et al. 2016; Mahadevaraju et al. 2021) could lead to reduced birth-rates for X-linked testis-expressed de novo genes. Alternatively, assuming the very strong autosomal enrichment of ancestral AG-biased genes is shaped by selection, the relative enrichment of X-linkage for AG-expressed de novo genes is consistent with a weaker selective effect on their chromosomal distribution, consistent with their very young age. The increased proportion of noncanonical splice junctions in these genes also supports the view that for at least some, their properties have not been optimized by natural selection. The observation that ancestral and de novo testis- and AG-biased genes reside in shared chromosomal domains of expression yet differ in the importance of cis-acting regulatory variation is suggestive of underlying hetergeneous processes, but the relative contributions of differences in origination processes vs selective effects are difficult to discern with existing data.

The general question of how frequently de novo genes originate as coding vs noncoding is of great importance, but currently unresolved (e.g., Ruiz-Orera and Alba 2019), and cannot be addressed using bioinformatic approaches that start with the premise that these genes are coding. Our computational analyses suggest that most young AG-expressed de novo genes are noncoding, though direct inferences from proteomic or ribo-profiling data would be required to put this conclusion on firmer ground. Similarly, the testis-expressed de novo candidates discussed here are also predicted by CPAT to be noncoding. Even if the small ORFs associated with these AG-expressed transcripts are translated, the vast majority of potential predicted proteins do not carry signal sequences. Thus, if several of the novelties described here have functions, they are unlikely to be directly related to processes requiring conventional protein secretion, though it remains possible that noncoding RNAs derived from de novo genes are transferred to females during mating (Bono et al. 2011; Ahmed-Braimah et al. 2021). Nevertheless, it follows from their properties that if AG-expressed de novo gene products have biological functions, they are likely biased toward those occurring inside the gland.

The complex regulatory variation influencing expression of de novo genes in the AG, the small sample of genotypes investigated here, the fact that many candidate genes are expressed in only one or a few genotypes, and the likelihood that most candidates are predicted to be noncoding, all conspire to compromise population genetic investigation into the possible influence of selection on these sequences. For example, approaches that seek evidence for protein functional constraint would have no value for noncoding genes. Moreover, the observation that the majority of candidates, which are predicted to be noncoding, are not associated with homologous ORFs in D. simulans (not shown) would not speak to the possible existence of a D. melanogaster-specific ORF, and in any case would also be entirely consistent with the hypothesis that a gene was D. melanogaster specific and noncoding. Similarly, the complex regulation and expression in one or a few genotypes for most candidates make it challenging to seek evidence of hitchhiking effects. Under the premise that trans-acting variants are likely to be more strongly deleterious than cis-acting variants, the influence of trans-acting variation on de novo gene expression in the AG would be consistent with the apparently low frequency of these genes in the population. However, without information on the population genetics of the trans-acting variants themselves, this idea cannot be evaluated. Thus, it seems to us that the question of the influence of selection on these candidate genes, while vitally important, cannot be adequately addressed with existing data.

Nevertheless, the relationships between gene frequency, size and expression may provide some information about evolutionary mechanisms. If AG-expressed de novo genes were on average deleterious, either due to the cost of transcription and/or translation, or because of deleterious interactions of their products (RNAs or proteins) in the cell, then we might expect singletons to be longer and expressed at a higher level than nonsingletons (Zhao et al. 2014), as deleterious genes should be overrepresented in the singleton class. Instead, we observe that singletons are shorter and expressed at lower levels than nonsingletons. This pattern provides no support for the idea that the genes described here are, on average, deleterious or strictly neutral (in which case there would be no expectation of heterogeneity of their attributes across frequency classes). The observed differences between singleton and nonsingleton genes, which is similar to that observed for testis-expressed de novo genes, would be consistent with an influence of positive selection (Zhao et al. 2014). However, the observation that very few AG-expressed de novo genes occur at high frequency does not support the simple hypothesis that simple directional selection plays an important role in their dynamics, as in this case, several would have fixed. Moreover, their overrepresentation on the X chromosome relative to ancestral AG-biased genes does not support the idea that directional selection is playing an important role. This does not imply, however, that these genes are uninfluenced by positive selection, as various forms of balancing selection acting on AG function could generate substantial polymorphism but relatively low fixation rates (e.g., Hughes 1997; Brisson 2018). That the molecular biology of the AG also appears to be highly variable in terms of amino acid polymorphism (Coulthart and Singh 1987; Begun et al. 2001), presence/absence of intact gene copies (Begun and Lindfors 1995), and gene expression or lack thereof in the organ (Cridland et al. 2020), is consistent with this possibility. Testing models of drift or selection acting on novel AG variation would be greatly facilitated by detailed information on the genetic variants underlying the gain of genes (Zhao et al. 2014) or gene expression (Cridland et al. 2020).

Data availability

Raw sequence data for all experiments are available from SRA, PRJNA575046, PRJNA210329. Supplemental Material available online at figshare: https://doi.org/10.25386/genetics.16934209.

Acknowledgments

We thank the anonymous reviewers for their comments, and A. Kopp and L. Blair for sharing D. yakuba sequence data. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Funding

This work was supported by the National Institute of Health, NIH R01GM110258 and NIH R35GM134930 awarded to D.J.B.

Conflicts of interest

The authors declare that there is no conflict of interest.

Literature cited

  1. Ahmed-Braimah YH, Wolfner MF, Clark AG.  2021. Differences in transcriptional responses between conspecific and heterospecific matings in Drosophila. Mol Biol Evol. 38:986–989. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ.  1990. Basic local alignment search tool. J Mol Biol. 215:403–410. [DOI] [PubMed] [Google Scholar]
  3. Armenteros JJA, Tsirigos KD, Sonderby CK, Petersen TN, Winther O, et al.  2019. SignalP5.0 improves signal peptide predictions using deep neural networks. Nat Biotechnol. 37:420–423. [DOI] [PubMed] [Google Scholar]
  4. Begun DJ, Whitley P, Todd BL, Waldrip-Dail HM, Clark AG.  2001. Molecular population genetics of male accessory gland proteins in Drosophila. Genetics. 156:1879–1888. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Begun DJ, Lindfors HA.  2005. Rapid evolution of genomic Acp complement in the melanogaster subgroup of Drosophila. Mol Biol Evol. 22:2010–2021. [DOI] [PubMed] [Google Scholar]
  6. Begun DJ, Lindfors HA, Thompson ME, Holloway AK.  2006. Recently evolved genes identified from Drosophila yakuba and D. erecta accessory gland expressed sequence tags. Genetics. 172:1675–1681. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Begun DJ, Lindfors HA, Kern AD, Jones CD.  2007. Evidence for de novo evolution of testis-expressed genes in the Drosophila yakuba/Drosophila erecta clade. Genetics. 176:1131–1137. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Belote JM, Zhong L.  2009. Duplicated proteasome subunit genes in Drosophila and their roles in spermatogenesis. Heredity (Edinb). 103:23–31. [DOI] [PubMed] [Google Scholar]
  9. Betrán E, Thornton K, Long M.  2002. Retroposed new genes out of the X in Drosophila. Genome Res. 12:1854–1959. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Bono JM, Matzkin LM, Kelleher E, Markow TA.  2011. Postmating transcriptional changes in reproductive tract of con- and heterospecifically mated Drosophila mojavensis females. Proc Natl Acad Sci U S A. 108:7878–7883. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Boutanaev AM, Kalmykova AI, Shevelyov YY, Nurminsky DI.  2002. Large clusters of co-expressed genes in the Drosophila genome. Nature. 420:666–669. [DOI] [PubMed] [Google Scholar]
  12. Brisson D.  2018. Negative frequency-dependent selection is frequently confounding. Front Ecol Evol. 6:10 doi:10.3389/fevo.2018.00010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cai J, Zhao R, Jiang H, Wang W.  2008. De novo origination of a new protein-coding gene in Saccharomyces cerevisiae. Genetics. 179: 487–496. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Casola C.  2018. From de novo to “de nono”: he majority of novel protein-coding genes identified with phylostratigraphy are old genes or recent duplicates. Genome Biol Evol. 10:2906–2918. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Carvunis A-R, Rolland T, Wapinski I, Calderwood MA, Yildirim MA, et al.  2012. Proto-genes and de novo gene birth. Nature. 487:370–374. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Clark AG, Eisen MB, Smith DR, Bergman CM, Oliver B, et al.  2007. Evolution of genes and genomes on the Drosophila phylogeny. Nature. 450:203–218. [DOI] [PubMed] [Google Scholar]
  17. Coulthart MB, Singh RS.  1987. Differing amounts of genetic polymorphism in testes and male accessory glands of Drosophila melanogaster and Drosophila simulans. Biochem Genet. 26:153–164. [DOI] [PubMed] [Google Scholar]
  18. Coulthart MB, Singh RS.  1988. High level of divergence of male reproductive tract proteins between Drosophila melanogaster and its sibling species, D. simulans. Mol Biol Evol. 5:182–191. [DOI] [PubMed] [Google Scholar]
  19. Crosby MA, Gramates LS, Dos Santos G, Matthews BB, St Pierre SE, et al. ; FlyBase Consortium. 2015. Gene model annotations for Drosophila melanogaster: The rule-benders. G3 (Bethesda). 5:1737–1749. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Cridland JM, Majane AJ, Sheehy HK, Begun DJ.  2020. Polymorphism and divergence of novel gene expression patterns in Drosophila melanogaster. Genetics. 216:79–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Findlay GD, Yi X, MacCoss MJ, Swanson WJ.  2008. Proteomics reveals novel Drosophila seminal proteins transferred during mating. PLoS Biol. 6:e178. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Findlay GD, MacCoss MJ, Swanson WJ.  2009. Proteomic discovery of previously unannotated, rapidly evolving seminal fluid genes in Drosophila. Genome Res. 19:886–896. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Grabherr MG, Haas BJ, Yassour M, Levin JZ, Thompson DA, et al.  2011. Trinity: reconstructing a full length transcriptome without a genome from RNA-seq data. Nat Biotechnol. 29:644–652. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Haerty W, Jagadeeshan S, Kulathinal RJ, Wong A, Ram KR, et al.  2007. Evolution in the fast lane: rapidly evolving sex-related genes in Drosophila. Genetics. 177:1321–1335. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Heinen TJAJ, Staubach F, Haming D, Tautz D.  2009. Emergence of a new gene from an intergenic region. Curr Biol. 19:1527–1531. [DOI] [PubMed] [Google Scholar]
  26. Hughes KA.  1997. Quantitative genetics of sperm precedence in Drosophila melanogaster. Genetics. 145:139–151. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kemkemer C, Hense W, Parsch J.  2011. Fine-scale analysis of X-chromosome inactivation in the male germline of Drosophila melanogaster. Mol Biol E. 28:1561–1563. [DOI] [PubMed] [Google Scholar]
  28. Knowles DG, McLysaght A.  2009. Recent de novo origin of human protein-coding genes. Genome Res. 19:1752–1759. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Landeen EL, Muirhead CA, Wright L, Meiklejohn CD, Presgraves DC.  2016. Sex chromosome-wide transcriptional suppression and compensatory cis-regulatory evolution mediate gene expression in the Drosophila male germline. PLoS Biol. 14:e1002499. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Leader DP, Krause SA, Pandit A, Davies SA, Dow JAT.  2018. FlyAtlas 2: a new version of the Drosophila melanogaster expression atlas with RNA-Seq, miRNA-Seq and sex-specific data. Nucleic Acids Res. 46:D809–D815. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Lee YCG, Chang H-H.  2013. The evolution and functional significance of nested gene structures in Drosophila melanogaster. Genome Biol Evol. 5:1978–1985. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Levine MT, Jones CD, Kern AD, Lindfors HA, Begun DJ.  2006. Novel genes derived from non-coding DNA in Drosophila melanogaster are frequently X-linked and show testis-biased expression. Proc Natl Acad Sci U S A. 103:9935–9939. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Li D, Dong Y, Jiang Y, Jiang H, Cai J, et al.  2010. A de novo originated gene depresses budding yeast mating pathway and is repressed by the protein encoded by its antisense strand. Cell Res. 20:408–420. [DOI] [PubMed] [Google Scholar]
  34. Lifschytz E, Lindsley DL.  1972. The role of X-chromosome inactivation during spermatogenesis. Proc Natl Acad Sci U S A. 69:182–186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Long M, Langley CH.  1993. Natural selection and the origin of Jingwei, a chimeric processed functional gene in Drosophila. Science. 260:91–95. [DOI] [PubMed] [Google Scholar]
  36. Mackay TFC, Richards S, Stone EA, Barbadilla A, Ayroles JF, et al.  2012. The Drosophila melanogaster genetic reference panel. Nature. 482:173–178. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Mahadevaraju S, Fear JM, Akeju M, Galletta BJ, Pinheiro MMLS, et al.  2021. Dynamic sex chromosome expression in Drosophila male germ cells. Nat Commun. 12:1–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Majane AC, Cridland JM, Begun DJ.  2021. Single-nucleus transcriptomes reveal function and evolutionary properties of cell types in the Drosophila accessory gland. Genetics. iyab213. doi:10.1093/genetics/iyab213. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. McManus JC, Coolon JD, Duff MO, Eipper-Mains J, Graveley BR, et al.  2010. Regulatory divergence in Drosophila revealed by mRNA-seq. Genome Res. 20:816–825. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Meiklejohn CD, Parsch J, Ranz JM, Hartl DL.  2003. Rapid evolution of male-biased gene expression in Drosophila. Proc Natl Acad Sci U S A. 100:9894–9899. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Meisel RP, Malone JH, Clark AG.  2012. Disentangling the relationship between sex-biased expression and X-linkage. Genome Res. 22:1255–1265. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Mikhaylova LM, Nguyen K, Nurminsky DI.  2008. Analysis of the Drosophila melanogaster testes transcriptome reveals coordinate regulation of paralogous genes. Genetics. 179:305–315. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Murphy DN, McLysaght A.  2012. De novo origin of protein-coding genes in murine rodents. PLoS One. 7:e48650. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Neme R, Tautz D.  2013. Phylogenetic patterns of emergence of new genes support a model of frequent de novo evolution. BMC Genomics. 14:117. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Nielsen H, Engelbrecht J, Brunak S, Von Heijne G.  1997. Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites. Protein Eng. 10:1–6. [DOI] [PubMed] [Google Scholar]
  46. Palmieri N, Kosiol C, Schlötterer C.  2014. The life cycle of Drosophila orphan genes. eLife. 2014;3:e01311. doi:10.7554/eLife.01311.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Parisi M, Nuttall R, Edwards P, Minor J, Naiman D, et al.  2004. A survey of ovary-, testis-, and soma-biased gene expression in the Drosophila melanogaster adult. Genome Biol. 5:R40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Ruiz-Orera J, Alba MM.  2019. Translation of small open reading frames: roles in regulation and evolutionary innovation. Trends Genet. 35:186–198. [DOI] [PubMed] [Google Scholar]
  49. Sorourian M, Kunte MM, Domingues S, Gallach M, Özdil F, et al.  2014. Relocation facilitates the acquisition of short cis-regulatory regions that drive the expression of retrogenes during spermatogenesis in Drosophila. Mol Biol E. 31:2170–2180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Spellman PT, Rubin GM.  2002. Evidence for large domains of similarly expressed genes in the Drosophila genome. J Biol. 1:5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Sturgill D, Zhang Y, Parisi M, Oliver B.  2007. Demasculinization of X chromosomes in the Drosophila genus. Nature. 450:238–237. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Swanson WJ, Clark AG, Waldrip-Dail HM, Wolfner MF, Aquadro CF.  2001. Evolutionary EST analysis identifies rapidly evolving male reproductive proteins in Drosophila. Proc Natl Acad Sci U S A. 98: 7375–7379. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Thurmond J, Goodman JL, Strelets VB, Attrill H, Gramates LS, et al. ; FlyBase Consortium. 2019. FlyBase 2.0: the next generation. Nucleic Acids Res. 47:D759–D765. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Van Oss SB, Carvunis A-R.  2019. De novo gene birth. PLoS Genet. 15:e1008160. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Vakirlis N, Hebert AS, Opulente DA, Achaz G, Hittinger CT, et al.  2018. A molecular portrait of de novo genes in yeasts. Mol Biol Evol. 35:631–664. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Wagstaff BJ, Begun DJ.  2005. Molecular population genetics of accessory gland protein genes and testis-expressed genes in Drosophila mojavensis and D. arizonae. Genetics. 171:1083–1101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Wang L, Park HJ, Dasari S, Wang S, Kocher J-P, et al.  2013. CPAT: Coding-Potential Assessment Tool using an alignment free logistic regression model. Nucleic Acids Res. 41:e74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Wigby S, Brown NC, Allen SE, Misra S, Sitnik JL, et al.  2020. The Drosophila seminal proteome and its role in postcopulatory sexual selection. Philos Trans R Soc B. 7;375:20200072. doi:10.1098/rstb.2020.0072. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Wilson C, Leiblich A, Goberdhan DCI, Hamdy F.  2017. The Drosophila accessory gland as a model for prostate cancer and other pathologies. In Pick L., editor. Fly Models of Human Disease. Cambridge: Academic Press. p. 339–375. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Witt E, Benjamin S, Svetec N, Zhao L.  2019. Testis single-cell RNA-seq reveals the dynamics of de novo gene transcription and germline mutation bias in Drosophila. eLife. 8:e47138.doi:10.7554/eLife.47138. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Yanai I, Benjamin H, Shmoish M, Chalifa-Caspi V, Shklar M, et al.  2005. Genome-wide midrange transcription profiles reveal expression level relationships in human tissue specification. Bioinformatics. 21:650–659. [DOI] [PubMed] [Google Scholar]
  62. Yang H, Jaime M, Polihronakis M, Kanegawa K, Markow T, et al.  2018. Re-annotation of eight Drosophila genomes. Life Sci Alliance. 1:e201800156. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Zhang L, Ren Y, Yang T, Li G, Chen J, et al.  2019. Rapid evolution of protein diversity by de novo origination in Oryza. Nat Ecol Evol. 3:679–690. [DOI] [PubMed] [Google Scholar]
  64. Zhang Y, Sturgill D, Parisi M, Kumar S, Oliver B.  2007. Constraint and turnover in sex-biased gene expression in the genus Drosophila. Nature. 450:233–237. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Zhao L, Saelao P, Jones CD, Begun DJ.  2014. Origin and spread of de novo genes in Drosophila melanogaster populations. Science. 343:769–772. [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Zhou Q, Zhang G, Zhang Y, Xu S, Zhao R, et al.  2008. On the origin of new genes in Drosophila. Genome Res. 18:1446–1455. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Raw sequence data for all experiments are available from SRA, PRJNA575046, PRJNA210329. Supplemental Material available online at figshare: https://doi.org/10.25386/genetics.16934209.

Articles from Genetics are provided here courtesy of Oxford University Press

RESOURCES