Generation of D. simulans dcr-2 deletion mutants marked by white⁺

We recently reported deletion alleles of core RNAi factors [dcr-2 and ago2] in D. simulans [14]. Although these mutants are viable, they are completely male sterile, and therefore cannot be maintained as stable stocks. This presents a technical challenge since D. simulans lacks balancer chromosomes and the 3xP3:DsRed marker was not fully reliable to distinguish heterozygotes from homozygotes. Therefore, preparation of pure homozygous material required extensive genotyping of small batches of dissected flies prior to combining samples for RNA isolation. Moreover, in initial RNA-seq analyses, genotyped samples were still prone to contamination. As a further complication, due to extremely high expression of many accessory gland transcripts [18], we noticed that even minute quantities of contaminating accessory gland could produce large biases in gene expression between libraries. Since D. simulans mutants were generated in a white mutant background (w[XD1]), their testis was colorless and thus more challenging to visualize compared to a white+ background, where the testis is bright yellow. For these reasons, the preparation of suitable quantities of D. simulans RNAi mutant testis was not straightforward.

To address these issues, we used CRISPR/Cas9 to generate multiple founders of a new D. simulans Δdcr-2 null allele, where most of its coding region is replaced with mini-white⁺ (w⁺) (Fig 1A and 1B). We anticipated selecting homozygotes with deeper eye color, as is typical in D. melanogaster; however, this was also not fully reliable due to the red eyes of these alleles. Instead, by crossing dcr-2 alleles marked by 3xP3:DsRed and w⁺ (Fig 1C–1E), we could select trans-heterozygotes carrying both dominant markers. Although DsRed⁺ eyes cannot be effectively scored in a w⁺ background, it is still possible to score DsRed⁺ ocelli (Fig 1F–1F’). Our independent Δdcr-2[w⁺] alleles were viable but specifically sterile in males, exhibited severe spermatogenesis defects, and failed to complement their corresponding DsRed alleles. We therefore used the trans-heterozygotes for subsequent analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Functional annotation of hpRNAs using small RNA and RNA-seq data dcr-2 mutants.

(A) dcr-2 location, sgRNAs, homology donor arms with white[+] marker used for mutant selection. (B) PCR genotyping and validation of dcr-2[w+] CRISPR mutant showing absence of dcr-2 amplicon; ago2 amplicon was used as control. (C) D. simulans w[XD1], a white mutant background used for CRISPR. (D) dcr-2[DsRed] deletion allele exhibits red fluorescent eyes. (E) dcr-2[w+] deletion allele exhibits pigmented eyes. (F) dcr-2[w+/DsRed] trans-heterozygous mutants. DsRed+ eyes are not very visible in a w+ background (F), but their fluorescence can be identified in the ocelli (F’), which are primitive light sensing organs. (G) Schematic of Dcr-2 processing of primary hpRNA (pri-hpRNA) transcripts into 21–22 nt siRNAs. pri-hpRNAs can be detected using RNA-seq, and the corresponding siRNAs derived from the hairpin can be identified via small RNA-seq. (H) hpRNA1 illustrates the behavior of hpRNA-derived testis RNA products in wildtype and dcr-2 mutants. Normalized RNA-seq and spike-in normalized sRNA-seq tracks of the hpRNA shown in w[XD1] and dcr-2. The top two tracks show RNA-seq of primary hpRNA transcript is increased in dcr-2 mutants due to upregulation of hpRNA primary transcript. The middle tracks show that the 5’ and 3’ ends of pri-hpRNA1 are defined by 5’-seq and 3’-seq data, respectively. The bottom two tracks show that hpRNA1-derived small RNAs are biased to 21 nt and mostly eliminated in dcr-2 mutant.

https://doi.org/10.1371/journal.pgen.1010787.g001

Signature features of hpRNA loci in short/long RNAs from wildtype and dcr-2 mutants

In contrast to D. melanogaster RNAi mutants, which are viable and sub-fertile [19], deletion of core RNAi factors in D. simulans result in complete male sterility [14]. This is due at least in part to the requirements of Nmy and Tmy, which are de novo hpRNAs that silence incipient X chromosome sex ratio distorters (Dox and MDox) in the male germline [14]. To assess the impact of RNAi loss more globally, and to compare D. melanogaster and D. simulans in greater detail, we generated biological replicates of small RNA and total RNA sequencing data from testis of dcr-2 heterozygotes and mutants in D. melanogaster, and w[XD1] vs. dcr-2 mutants in D. simulans. In control, we expect that primary hpRNA (pri-hpRNA) transcripts are cleaved by Dcr-2 into 21–22 nucleotide (nt) siRNAs, while bona fide siRNAs should be lost in dcr-2 mutants concomitant with accumulation of their progenitor mRNAs (Figs 1G and S1). Together, this combination of datasets permits functional categorization of genuine hpRNAs with high specificity, as illustrated by hpRNA1 (Fig 1H) and others (S2 Fig).

We plotted differential expression of D. melanogaster and D. simulans RNA-seq data in MA plots (Fig 2A–2D), which display the log-ratio of dcr-2 vs. wild type controls of RNA-expression (M-axis) and the average gene expression in the RNA-seq data (A-axis). Strikingly, on the genomewide scale, hpRNA precursors were amongst the highest-upregulated transcripts in dcr-2 mutants (Fig 2). We initially assessed this using the set of D. melanogaster hpRNAs, all of which are conserved in D. simulans [19]. The known hpRNAs dominate the highest-upregulated transcripts in dcr-2 mutants of both species. For example, Dmel-hp-mir-997-1 was the 2nd-highest elevated locus genome-wide, and 7/9 hpRNAs were in the top 20 upregulated loci in D. melanogaster (Fig 2A and 2B). We similarly observed that pri-hpRNAs of D. simulans orthologs of D. melanogaster hpRNAs were strongly elevated in dcr-2 mutants (Fig 2C and 2D). These effects were specific, since primary miRNA transcripts were largely unaffected in dcr-2 mutants of either species (Fig 2B and 2D). This was expected, as miRNAs are processed instead by Dcr-1. One exception was mir-985, whose primary transcript was elevated in mutants of D. melanogaster dcr-2 (Fig 2A and 2B), but not in D. simulans dcr-2 (Fig 2C and 2D). The reason for this discrepancy is unknown, but a potential explanation is that transcription of mir-985 is elevated in D. melanogaster as a secondary effect that is not shared in D. simulans. We also note elevated novel hpRNAs in D. simulans than in D. melanogaster (Fig 2C and 2D; two-tailed Fisher’s exact test P = 0.0010). Finally, hpRNA-derived siRNAs were strongly depleted in dcr-2 mutants, including from all previously classified D. melanogaster hpRNAs and their D. simulans orthologs (Fig 2E and 2F).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Comparative hpRNA annotation using functional genomic data.

(A) MA-plot comparing D. melanogaster wild-type and dcr-2 testis RNA-seq data. Red dots denote hpRNAs conserved with D. simulans, including one newly-recognized locus hp-426. (B) Primary hpRNA transcripts are all elevated in D. melanogaster dcr-2 testis, unlike primary miRNA transcripts; mir-985 is an exception. (C) MA-plot comparing D. simulans control w[XD1] and dcr-2 testis RNA-seq data. Red dots mark hpRNAs conserved with D. melanogaster, while blue dots indicate de novo hpRNAs in D. simulans. (D). Comparison of primary hpRNA and primary miRNA transcripts in D. simulans. All pri-miRNA transcripts are unchanged (including mir-985), while all pri-hpRNA transcripts are elevated in both conserved and de novo hpRNAs. Compared to D. melanogaster, D. simulans has an elevated number of novel hpRNAs (two-tailed Fisher’s exact test, p = 0.0010). (E-G) Expression of hpRNA-derived small RNAs in D. melanogaster (E) and D. simulans (F-G). Note that reads from the multicopy tandem hp-CG4068 repeats are condensed into a single hpRNA locus, to avoid over-tabulation of hpRNA numbers. All hpRNA-siRNAs are decreased in dcr-2 mutants, in both species and regardless of the hpRNA age. (H) Phylogenetic analyses of the presence/absence of orthologs to newly-annotated D. simulans hpRNAs in the simulans clade sister species (D. sechellia and D. mauritiana), compared with D. melanogaster and the close outgroup species D. yakuba. All of the newly-annotated D. simulans hpRNAs were born within the simulans clade ancestor, or within D. simulans itself.

https://doi.org/10.1371/journal.pgen.1010787.g002

Inspection of the local genomic regions of known hpRNAs revealed provocative differences between D. simulans and D. melanogaster in the vicinity of hpRNA clusters, and provided additional evidence for rapid flux in hpRNA loci. The largest set of dispersed hpRNA loci in D. melanogaster are members of the hp-pncr009 family, for which 3 separate hpRNAs (hp-pncr009, hp-CR32207, and hp-CR32205) are interspersed with 9 protein-coding target genes of the 825-Oak family [19]. We have theorized that transcription across pairs of divergently-oriented 825-Oak family loci might beget pncr009 hpRNAs. Interestingly, in the short evolutionary distance that separates D. melanogaster and D. simulans, we identify two additional pncr009-class hpRNAs in D. simulans (S3A Fig). To facilitate intuitive connection of these hpRNAs to target genes in the 825-Oak family, we named these novel D. simulans hpRNAs as hp-Oak1 and hp-Oak2 (Fig 2C and 2D).

We also documented evolutionary flux in the tandem hpRNA repeats of the hp-CG4068 cluster. Although the copy number was potentially in question from prior short-read genome assemblies, the recent availability of simulans-clade PacBio genomes [20] demonstrates radical copy number of the hp-CG4068 cluster. There are 20 tandem copies in D. melanogaster but only 9 tandem copies in D. simulans (S3B Fig), as well as 14 copies in D. mauritiana, and 10 copies in D. sechellia. We note that the available data do not rule out that their copy numbers might not also be variable within a species. In any case, there is high evolutionary divergence in the copy number of hpRNAs located in both genomically linked copies (hp-pncr009 cluster) as well as in tandem copies produced from a common transcript (hp-CG4068 cluster). Such dynamics are much greater than observed for canonical miRNAs, which only occasionally exhibit similar changes amongst these species [21, 22].

Unidirectional expansion of hpRNAs in D. simulans compared to D. melanogaster

With these validations of known hpRNAs in hand, we undertook more systematic hpRNA annotations as a basis of more comprehensive evolutionary comparisons. In particular, we sought highly structured loci that produce ~21 nt-biased, Dcr-2-dependent small RNAs, that also accumulate primary transcripts in dcr-2 mutants (Fig 1H). However, we did not absolutely require pri-hpRNA changes in dcr-2, for several reasons. First, pri-hpRNA transcripts might not be stable and/or might be subject to other RNA decay pathways. Taking the canonical miRNA pathway as an analogy, not all pri-miRNA transcripts accumulate as unprocessed full-length mRNAs in mutants of nuclear miRNA processing factor Drosha, and RNase III enzyme. Second, it was conceivable that some loci are processed earlier in development to yield stable small RNAs, but are not substantially transcribed in adult testis. Finally, transcription of some pri-hpRNA loci might be decreased in RNAi mutants.

In D. melanogaster, we previously annotated hpRNAs from nearly 300 small RNA libraries of diverse developmental, tissue and cell origins, yielding only nine confident hpRNAs [19]. Given that testis is the predominant location of hpRNA-siRNA accumulation, we interrogated our new genetically paired testis data for evidence of additional Dcr-2-dependent inverted repeat loci. However, beyond previously known hpRNAs, we only recovered a single new hpRNA, hp426 (Fig 2A, 2B, and 2E), which produces siRNAs in D. simulans. Thus, we did not identify any D. melanogaster-specific hpRNAs.

A very different picture emerged from analysis of D. simulans. Although we had far fewer small RNA libraries to annotate from, especially of testis datasets, we found numerous novel hpRNAs (S1 Table). These annotations were of high confidence, as nearly all of them exhibited reciprocal behavior of primary transcripts and mature siRNAs, when comparing control and dcr-2 testis libraries (Fig 2C, 2D and 2G). In fact, most novel pri-hpRNA loci resided amongst the most-upregulated transcripts across D. simulans dcr-2 testis (Fig 2C). A majority of these accumulated discrete spliced RNAs. However, depending on the locus, transcript coverage was often non-uniform. This was particularly the case within highly-duplexed portions of foldback structures (S4 Fig), evidently indicating that library construction was compromised within strongly double-stranded transcript regions. Coverage was also an issue at transcript termini, which is generally the case with typical RNA-seq protocols [23].

We therefore employed two more approaches to help annotate pri-hpRNA transcripts in control and dcr-2 mutant testis: analysis of 5’ ends from low-input RNA [24] and 3’-end sequencing to determine polyadenylation sites [25]. As illustrated in Fig 1I, 5’-seq and 3’-seq directly visualize the beginnings and ends of pri-hpRNA transcripts, and indicate that hpRNA progenitors are mRNAs. We note that some hpRNA loci are exact genomic copies that remain valid upon scrutiny of the largely contiguous PacBio D. simulans assembly [20], a phenomenon we return to later in this study. However, even when conservatively counting identical hpRNA copies within a tandem cluster as a single locus, there are 23 distinct, de novo hpRNA transcription units in D. simulans, which can be assigned to 15 families that are not simply genomic copies (Fig 2C, 2D, 2G, and 2H). As described later, some of these families can still be traced as sharing evolutionary heritage (as is the case for hp-Nmy/hp-Tmy, which are related in sequence but have separable functions).

From head-to-head comparison of D. melanogaster and D. simulans alone, we cannot distinguish if D. simulans gained hpRNAs, or if D. melanogaster lost them. Therefore, we analyze the distribution of our newly annotated D. simulans hpRNAs in the sister simulans-clade species D. sechellia and D. mauritiana, as well as D. yakuba as a close outgroup species. None of the novel hpRNAs identified in D. simulans have orthologs in D. melanogaster and D. yakuba, while most of them exist in the syntenic positions of the D. sechellia and/or D. mauritiana genomes (Fig 2H). These analyses clearly support that the vast majority of these hpRNAs were born in the simulans-clade ancestor, and two in D. simulans itself. Overall, the radical and asymmetric expansion of hpRNAs in the simulans-clade lineage, compared to D. melanogaster, strongly suggests that the RNAi pathway has been deployed adaptively in these sister species, putatively to address emerging regulatory situations such as intragenomic conflicts.

Target network of D. simulans-specific hpRNAs

We next investigated targets of novel hpRNAs annotated in D. simulans. Although many of these hpRNAs are shared in other simulans-clade species, we will subsequently refer to D. simulans-specific targets since we define these with respect to functional derepression in RNA-seq data (and not solely on the basis of hpRNA/target homology). In our previous work in D. melanogaster, we observed that hpRNAs typically exhibit substantial complementarity to one or a few target genes, ranging from an individual siRNA to extended regions that encompass multiple siRNAs [19, 26]. On this basis, we proposed that hpRNAs typically derive from their targets [19], analogous to plant miRNAs [27]. This set the stage that it seemed plausible, if not likely, that de novo hpRNAs in D. simulans might also have overt complementary targets. Indeed, we were able to identify compelling targets with antisense matching to most newly-recognized D. simulans-specific hpRNAs (S1 Table). In the following sections, we describe notable insights from specific aspects of the D. simulans hpRNA target network.

Unexpected complexity in the D. simulans hpRNA network related to Dox family loci

Multiple X-linked Dox family genes, including two newly-recognized members (PDox1 and PDox2), share an HMG-box domain that is derived from protamine and are targeted by newly-emerged hpRNAs (Nmy/Tmy class) [14–16]. Now, with expanded D. simulans hpRNA-target maps based on functional genomics, we reveal additional, de novo innovations within the Dox/hpRNA regulatory network.

We recently identified a sub-lineage of Dox-related loci that lack the HMG-box [15]. Via synteny comparisons with D. melanogaster, we inferred these to derive originally from fusion of a protamine-like copy in between CG8664 and forked loci in an ancestor of simulans clade species, termed “original Dox” (ODox, Fig 3A). Our evolutionary tracing supports that ODox spawned the contemporary Dox family genes PDox, UDox, MDox and Dox across simulans clade Drosophilids [15]. Perhaps confusingly, then, the ODox locus in contemporary simulans-clade species retains segments of CG8664 and the 5’ UTR of protamine, but has lost its HMG-box (Fig 3B). ODox subsequently duplicated and mobilized to yield the related ODox2 locus, which shares predicted domains with ODox and lacks an HMG-box, but also has divergent sequence material (Fig 3B). ODox and ODox2 loci are proximal to centromere on the X chromosome at ~16Mb on the D. simulans long-read assembly [20], while the contemporary amplification of Dox family genes occurred at a distal genomic window of ~9-10Mb (Fig 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Innovations in the Dox-related hpRNA network.

(A) Schematic of the evolution of Dox family genes, and the contemporary state of inferred, ancestral fusion of Protamine and CG8664 (Prot-CG8446 fusion) that birthed the Dox family genes. Segments 1, 2 and 3 in the inset box (inferred ODox fusion), shows genomic regions corresponding to their sequence of origin. Segment 1 (blue) is derived from CG8664, segment 2 (green) is derived from the 5’ UTR of Protamine, and segment 3 (orange) contains Protamine coding sequence, including the HMG-box domain. Within segment 1, is a fragment of DNAREP1 TE (turquoise). While contemporary Dox family genes Dox, MDox, and PDox share this segment structure, we infer that the ancestral "ODox" locus lost the HMG segment in contemporary D. simulans (B). In contemporary D. simulans, the ancestral ODox fusion is an hpRNA (hp-ODox1) and its duplication and insertion at CG5004/GD17329 hosts another hpRNA (hp-ODox2). (C) hp-ODox1 is a discrete hpRNA locus, revealed by upregulation of its primary hairpin transcript in dcr-2 RNA-seq, presence of Dcr-2-dependent siRNAs, and demarcated by 5’-seq and 3’-seq. The inverted repeat arms of hp-ODox1 bear homology to Tapas (pink), and other regions of hp-ODox1 bear TE sequences. (D) hp-ODox2 bears all the genomic signatures of an hpRNA, and its inverted repeat arms are homologous to BS2/Jockey TE. pri-hp-ODox2 also bears homology to Krimper, although this sequence is not part of the inverted repeat and does not yield siRNAs. (E) Sequence alignment between hp-ODox and tapas. Examples of siRNAs with fully complementarity to tapas are boxed in green. (F) Antisense complementarity between BS2/Jockey TE and hp-ODox2 with examplar siRNAs highlighted in green.

https://doi.org/10.1371/journal.pgen.1010787.g003

As HMG-box domains seem relevant to the distorting functions of Dox family factors [15, 16], we wondered about impacts of Dox-superfamily loci lacking HMG-box domains. Unexpectedly, we realized that both D. simulans ODox and ODox2 contain inverted repeats that bear functional hpRNA signatures, i.e., they generate Dcr-2-dependent small RNAs and their primary transcripts are upregulated in dcr-2 mutants (Figs 2, 3C, and 3D). Accordingly, we renamed these loci hp-ODox1 and hp-ODox2. Detailed analysis reveals further unanticipated features of their domain content. In particular, the inverted repeat at ODox bears ~200 bp homology to exon 2 of Tapas/GD11509 and generates siRNAs with antisense complementarity to the parental Tapas/GD11509 on chr2R (Fig 3E). In addition, hp-ODox1 transcript contains a mix of repetitive sequences (Fig 3C).

Unexpectedly, the hairpin at hp-ODox2 is not homologous to the hairpin in hp-ODox1, despite their shared lineage. Instead, the hp-ODox2 inverted repeat contains sequence from the BS2/Jockey transposable element (TE), generating siRNAs with antisense complementarity to BS2/Jockey (Fig 3F). In addition, other regions of the pri-hp-ODox2 transcript bear other repetitive sequences as well as fragments from Krimper. However, the inverted repeat does not include Krimper sequence, and Krimper-targeting siRNAs were not observed.

Overall, we uncover additional innovations of the Dox-family lineage, which extend beyond HMG-box/protamine domains. In particular, derivatives of an ancestral "ODox" gene now generate hpRNA-siRNAs in contemporary D. simulans, and may be engaged in distinct genetic conflicts with connections to TE biology (Fig 3).

Innovation of D. simulans hpRNAs that target other HMG-box loci

In addition to Dox family genes targeted by hpRNAs, we also find evidence that another class of HMG-box domain containing genes (testis HMG) is also targeted by novel hpRNAs in D. simulans. The Dox family genes belong to a distinct "MST-HMG-box" subclass [28], owing to their relatedness to Drosophila protamines [15, 16]. tHMG is not formally included in the MST-HMG-box family, but is nevertheless also testis-specific and expressed at highest levels during the histone-to-protamine transition [29]. We note evidence for rapid evolution of testis HMG-box loci, since protamine is locally duplicated in D. melanogaster (MST35Ba/Bb), but exists as a single copy in simulans clade species [15, 28]. Similarly, tHMG is locally duplicated in D. melanogaster (tHMG1 and tHMG2), but bears a single copy in the syntenic region of D. simulans (Fig 4A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Novel D. simulans hpRNAs related to testis HMG-box (tHMG) loci.

(A) Schematic of tHMG loci in D. melanogaster and D. simulans, and the origins of hpRNA suppressors and targets encompassing multiple novel X-linked tHMG-derived loci. tHMG is locally duplicated in D. melanogaster, but the syntenic location of D. simulans bears a single gene. However, D. simulans X chromosome bears multiple de novo hpRNAs with homology to tHMG, a single copy locus and another genomic cluster containing 13 tandem hpRNAs. (B) Genomic details of the hp-tHMG-solo locus, which is a spliced transcript with defined 5’ and 3’ ends, which generates abundant small RNAs from its duplex arm. Note that the 5’ region of hp-tHMG-solo has been locally duplicated (red asterisk). (C) Genomic details of the hp-tHMG-cluster locus. This region bears 13 nearly identical hpRNAs, although the nature of the primary transcript(s) is not clearly evident, although 6 cluster units show evidence for 5’-seq and 3’-seq (detailed in S6 Fig). RNA-seq data across the tHMG-cluster locus appears depleted in the hpRNA duplex regions. Small RNA tracks show Dcr-2-dependent siRNAs from each cluster member. (D, E) Examples of antisense complementarity between tHMG genes and hp-tHMG-siRNAs, including autosomal tHMG1 and hp-tHMG-solo, and tHMG-Xhet1 and tHMG cluster loci.

https://doi.org/10.1371/journal.pgen.1010787.g004

Our functional profiling of D. simulans RNAi mutant testis allows us to discern further evolutionary dynamics of HMG-box-related hpRNAs (Fig 4A). First, we find that D. simulans contains a locus related to autosomal D. melanogaster tHMG2, which has mobilized to the X chromosome (Fig 4A). Detailed inspection shows that this locus is actually an hpRNA, as it generates Dcr-2-dependent small RNAs, and is associated with a spliced transcript that is derepressed in dcr-2 mutants (Fig 4B). As is the case with certain other strong inverted repeat loci, the RNA-seq signal is poorly represented in the duplex arms of the hairpin, which resides near the 5’ end of the primary transcript. However, the presence of capped and polyadenylated species provide experimental evidence for its termini. Of note, there is a local tandem duplication of the left arm of the hairpin, including within 5’-end mapping data (Fig 4B), and hints that its genesis as an hpRNA involved local duplications of a transposed sequence.

There are further complexities. We observe recent amplification of tHMG2 in the heterochromatin boundary of chromosome X (Xhet), associated with numerous tandem hpRNAs within ~27kb (hp-tHMG-cluster) (Fig 4C). Based on the local duplication of tHMG2, and hpRNA secondary structure, we classify the 13 hp-tHMG-cluster hpRNAs into 3 subcategories (S5 Fig). The hp-tHMG-cluster also harbors three paralogs of tHMG2 that are not part of inverted repeats. Of these, two paralogs appear to be full-length copies (88aa, compared to D. simulans tHMG2 ortholog at 91aa), while a truncated paralog within this cluster encodes for only 24 aa (Figs 4C and S5).

At present, the nature of the primary transcript for the hp-tHMG cluster is uncertain, as near perfect complementarity of inverted repeats results in depletion of RNA-seq signal within tHMG hpRNA copies. We document 13 local inverted repeats (hpRNA copies) within the hp-tHMG-cluster. Of these, 6 have flanking 5’-end and 3’-seq signals, along with RNA-seq mapping between the hairpin arms that is upregulated in dcr-2 mutants. While it is unclear how many individual transcription units exist with this hpRNA cluster, it is clear that the solo and clustered tHMG hairpin loci (located on opposite ends of the D. simulans X chromosome) are evolutionarily related, to each other and to the parental autosomal copy of tHMG (Figs 4C and S5). Indeed, we identify abundant siRNAs generated from hp-tHMG-solo and the hp-tHMG-cluster that exhibit perfect complementarily to autosomal tHMG1 and/or novel paralogs of tHMG2 on Xhet (Fig 4D).

We analyzed the phylogenetic relationships amongst the general family of testis-restricted HMG-box factors in D. melanogaster and D. simulans (S6 Fig), which emphasizes that tHMG factors are an outgroup to the MST-HMG-box family [28], of which protamine/Dox family loci comprise a distinct subfamily. Interestingly, members of all these groups seem to be involved in sex chromosome conflicts [15–17, 30]. Altogether, these data emphasize greater co-evolutionary arms races between hpRNAs and their HMG-box targets than currently recognized.

Molecular delineation of the stages of hpRNA emergence and evolution

With broader evidence that hpRNAs generally target specific genes, or groups of related loci, we sought broader perspective on evolution of hpRNA regulatory networks. Although all D. melanogaster hpRNAs emerged relatively recently [19], it is instructive to note that all D. melanogaster hpRNAs are conserved in D. simulans (Fig 2E and 2F). Therefore, we may consider D. melanogaster hpRNAs to be relatively older, compared to the numerous de novo D. simulans hpRNAs we identify in this study (Fig 2C and 2G). In principle, then, this collection of extremely young D. simulans hpRNAs may illuminate the earliest evolutionary stages of hpRNA birth.

An ongoing conundrum concerns "where" hpRNAs come from, especially as the functionally validated hpRNA-target relationships comprise examples where the hpRNA is genomically distant from its target [14, 19, 26, 31]. This could be due to mobilization of the hpRNA locus itself, or to derivation of the hpRNA via retrogene insertion [9, 11]. In the latter case, acquisition of a promoter may be an issue. We note that formation of D. simulans hp88 occurred with the 3’ region of an existing apparent non-coding locus CR43306 (Fig 5A), suggesting that hpRNAs could take advantage of pre-existing transcription units for their expression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Genomic features of novel D. simulans hpRNAs inform early stages in their birth.

(A) Dsim-hp88 emerged within the 3’ UTR of non-coding RNA CR43306, which is syntenic between D. melanogaster and D. simulans. The inverted repeat arms bear homology to de novo paralogs on the X chromosome GD15542 (3 copies expressed, but there are 13 other paralogous sequences of this novel gene on the X). GD15542 is upregulated in dcr-2 mutant indicating functional suppression via RNAi. hpRNA-target alignments for both Dsim-hp58 and Dsim-hp88 are provided with highlighted green box showing fully complementary siRNAs to their targets, CG17385 and GD15542, respectively. (B) Dsim-hp58 is a de novo hairpin that was born one gene away from its progenitor gene CG17385. It bears an inverted repeat fragment of exon 2 of CG17385. Similar to pri-hp58, CG17385 is upregulated in dcr-2 mutant testis RNA-seq data, indicating its functional suppression via the RNAi pathway.

https://doi.org/10.1371/journal.pgen.1010787.g005

Our catalog of de novo hpRNAs includes many other hpRNA loci that are genomically distant from their targets (S1 Table). However, the precedent of the interweaved locations of multiple related hpRNAs and targets of the hp-pncr009/825-Oak families in D. melanogaster suggested that some hpRNAs might be born from a genomic location close to their target. We now find several examples of this. For example, D. simulans hp58 is a newly-emerged hpRNA that is adjacent to its pre-existing target gene GD25683/CG17385 (Fig 5B). Similarly, we identify a novel hpRNA located at D. simulans ballchen (GD18116), created by a partial duplication of the 5’ region adjacent to the conserved ballchen gene annotation. This is reminiscent of the partial 5’ duplication of the hp-tHMG-solo locus (Fig 4).

We also emphasize that hpRNA evolution frequently involves emergence of multiple copies. Beyond the described examples of expanding hp-pncr009 family clustered loci and hp-CG4068 tandem hpRNAs, we discovered novel amplified hpRNAs in D. simulans. These include tandem duplicates (as in the novel tHMG-related hpRNA cluster), local genomic duplicates of independent transcription units (as in de novo hp-pncr009 family members, S3B Fig), or genomically dispersed copies (as in the case of four copies of the hp70 family, S1 Table). Finally, we highlight D. simulans hp88, which was appended to an existing non-coding locus, but contains genetic material from the trio of GD15542 gene copies located on the X chromosome (Fig 5A), much like amplifications of Dox family genes trigger hpRNA birth [15, 16].

Functional repression by hpRNAs is most overt for the youngest siRNA loci

With an expanded view of hpRNAs and targets in hand, we addressed the larger consequences of RNAi loss on the testis transcriptomes of both species. Interestingly, although cytological consequences of RNAi loss on spermatogenesis is substantial in D. melanogaster [19] and severe in D. simulans [14], their respective transcriptome responses were relatively restricted (Fig 2A and 2C). For example, in D. melanogaster, only 245/16773 annotations detected >10RPM were 2-fold upregulated in dcr-2 mutants compared to dcr-2/+ heterozygous controls (FDR <1%) (S2 Table). However, all hpRNAs (except hpRNA1, upregulated only 1.5-fold) were amongst the top 20 upregulated transcripts in dcr-2 mutants (S2 Table), confirming their efficient metabolism by Dcr-2. hpRNA targets also responded directionally to RNAi loss, in that none were downregulated. However, many hpRNA targets were not detected in adult testis. Of the 8 conserved hpRNA loci between D. melanogaster and D. simulans, the pncr009 cluster hpRNAs (hp-pncr009, hp-CR32205, and hp-CR32207) bear homology to 10 target genes belonging to the 825-Oak family [19] (S3 Fig). In D. melanogaster dcr-2 mutant testis, all 825-Oak family genes were <10RPM threshold, or indeed not detected (S7 Fig). The D. melanogaster orthologs of 825-Oak loci are restricted to pupal gonads (http://flybase.org/), even though their corresponding hpRNA-siRNAs are detectable in adult testis. Of the remaining D. melanogaster hpRNAs targets, only two (ATP-synthase ß and mus308, targeted by hpRNA1 and hp-CG4068 respectively), were elevated in dcr-2 mutants (Fig 6A). The targets of D. melanogaster-D. simulans conserved hpRNAs exhibited similar expression profiles in D. simulans dcr-2 mutants (Figs 6B and S7). Thus, there is functional repression of targets of the older, conserved hpRNAs in both species, but the effects are generally modest. Nevertheless, the regulatory effects on siRNA targets are still greater than with most miRNA targets [32].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Preferential suppression of targets of young hpRNAs vs. older hpRNAs.

Differential expression analysis of the targets of conserved and novel hpRNAs in D. melanogaster and D. simulans testis RNA-seq data. (A, B) Targets of hpRNAs conserved between D. melanogaster and D. simulans. Fold change values comparing wildtype and dcr-2 mutant shown in yellow. Black and grey dotted lines show 2-fold and 1.5 fold changes in dcr-2 mutant compared to wild type. Fold change values were estimated using two replicates each for the mutant and wildtype samples using DEseq package in R with an FDR < 1%. (C) Targets of D. simulans-specific hpRNAs (red) and their fold changes in dcr-2 mutant compared to wild type. Note the functional repression of targets of young hpRNAs is much greater than for older hpRNAs. (D) The expression changes of additional testis-expressed genes with similar domains as the hpRNA targets (in grey) are plotted for comparison, which emphasizes specificity of upregulation of direct hpRNA targets (in red).

https://doi.org/10.1371/journal.pgen.1010787.g006

In striking contrast, the targets of the youngest siRNA loci, i.e. of D. simulans-specific hpRNAs, showed substantial greater directional change upon dcr-2 loss (Fig 6C). In D. simulans, 1021/15119 loci expressed >10RPM exhibit at least two-fold upregulation in dcr-2 mutant compared to w[XD1] (FDR <1%) (S3 Table). Of these, 22 novel and 7 conserved hpRNAs were in the top 200 derepressed genes. Moreover, 14/20 targets of novel hpRNAs in D. simulans were among top 200 upregulated genes (S3 Table). Of note, more genes were deregulated in dcr-2 mutants in D. simulans compared to D. melanogaster (1021 vs. 245), consistent with the more severe cytological defects in dcr-2 mutant testis of D. simulans (Lin et al. 2018) [14]compared to D. melanogaster {Wen et al and Lai Molecular Cell 2015}. To further assess whether the upregulation of de novo hpRNA targets is specific, we compared de novo hpRNA targets to related paralogs in D. simulans that are also expressed in testis. For example, we detected directional change in expression of Dox family members upon loss of Dcr-2, but not their related HMG-box paralogs that are also expressed in testis (Fig 6D). Similar specificity was observed for other families of de novo targets (zinc-finger factors, Tudor proteins, kinases, and BTBD factors, Fig 6D). However, we wish to note that the mRNA for the tudor protein Tapas is not selectively upregulated, although hp-ODox-1 generates a population of siRNAs antisense to Tapas. In accord with the striking derepression of the targets of de novo hpRNAs in RNA-seq data, they are targeted by far more antisense siRNAs than are targets of hpRNAs that are conserved between D. simulans and D. melanogaster (S4 Table).

Overall, the analysis of D. simulans was particularly informative, since we could compare the properties of "young" and "older" hpRNA-target interactions. This was perhaps unexpected, since we had earlier used the latter cohort to derive clear evidence of adaptive co-evolution between hpRNAs and their targets [19]. Due to insufficient orthologs, we lack statistical foundation to assess co-evolution with simulans-specific hpRNAs. Nevertheless, the picture is clear that younger hpRNAs mediate quantitatively greater target suppression than older hpRNAs. In particular, these data flip the rationale for small RNA mediated regulation relative to the miRNA pathway, for which recently-evolved loci might generally be neutral and only the oldest miRNAs appear to have biologically significant effects [33].

Biased X-linkage of de novo D. simulans hpRNA targets, along with some hpRNAs, reflects likely roles in sex chromosome conflict

D. simulans hpRNAs that are conserved in D. melanogaster preserve their documented targeting of a cohort both young and ancient genes [19]. Overall, the targets of these conserved hpRNAs are not biased in their genomic location, as they are distributed across all the chromosomes (with multiple targets on each of the larger chromosomes, S8 Fig). Thus, there is no overt bias to the age and location of "old" hpRNA targets, beyond the fact that many hpRNAs and targets in the hp-pncr009/825-Oak network are clustered within a small genomic interval, and continue to expand actively (Figs 7 and S3).

On the other hand, a distinct pattern emerges with our collection of "young" D. simulans hpRNAs. These very young hpRNAs (blue loci, Fig 7A) are distributed across the major chromosomes, in a pattern relatively similar to the older hpRNAs (orange loci, Fig 7A). On the other hand, the targets of these recently-emerged D. simulans hpRNAs, exhibit statistically significant bias for X-chromosome localization (Fisher’s exact test, p = 0.0416; blue loci, Fig 7B). In addition to Dox and MDox, 14/20 (70%) other targets of D. simulans-emerged hpRNAs are found on the X, whereas only 2/11 (18%) targets of hpRNAs shared with D. melanogaster are found on the X. These genes are all testis-specific paralogs of older gene families, and many encode proteins with roles putatively related to meiosis or chromosome segregation. As documented, these include four paralogous loci Dox, MDox, PDox1 and PDox2, which define a rapidly evolving set of young protamine-like gene copies with known or inferred meiotic drive activities [9, 11, 14–16]. Moreover, we reported the existence of additional X-linked tHMG loci, for which several protein-coding copies within the tHMG-cluster and the Xhet region are targeted by hpRNAs (Figs 4, S5, and S6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Biased location of de novo hpRNA targets on the D. simulans X chromosome.

(A) Genomic location of D. simulans hpRNAs, separated into ones that are conserved in D. melanogaster (in orange) and ones that are D. simulans-specific (in blue). Both classes of hpRNAs are distributed across all the major chromosome arms. (B) Genomic location of D. simulans hpRNA targets, separated into ones whose hpRNA is conserved in D. melanogaster (in orange) and ones whose cognate hpRNA is D. simulans-specific (in blue). The target of hp-ODox-2 (BS2/Jockey) was not included in this visualization, as the BS2/Jockey transposable element has numerous copies across all major chromosomes. The concentration of targets of D. simulans-specific hpRNAs on the X suggests that they may comprise novel selfish genes involved in sex ratio meiotic drive.

https://doi.org/10.1371/journal.pgen.1010787.g007

Hun Hunaphu (Hun) is another intriguing X-linked hpRNA target. Hun is a newly-emerged X-linked derivative of ballchen, encoding a histone kinase that is essential for germline stem cell renewal (in both sexes) and meiotic chromosomal architecture [34–37]. However, like Dox family genes [15, 16], Hun is a chimeric, young gene that has lost parts of Ballchen and gained new coding sequence [38]. Of note, Hun orthologs exhibit a large excess of nonsynonymous substitutions compared to Ballchen [38]. Ballchen is moderately upregulated in dcr-2 mutant testis but Hun is much more strongly derepressed (Fig 6). In light of their strong suppression by endo-RNAi, we take this as a strong suggestion for a selfish function of Hun that necessitates silencing by a hpRNA. Overall, we infer the biased location of target genes of de novo hpRNAs on the X chromosome reflects its disproportionate involvement in active genomic conflicts.

The hpRNA targeting landscape is diametrically opposite to miRNA regulation

We conceive two general classes of hpRNA targets. While hpRNAs are all evolutionary young, some of their targets are relatively old genes; e.g. targeting of ATP synthase-ß by hpRNA-1. We imagine there are adaptive reasons for why mild suppression of such loci imparts beneficial regulatory consequences. This may be due directly to the acquisition of elevated transcriptional properties of the targets, and/or by emergence of duplicated loci with preferred testis expression, which appears to be a relatively common process [39, 40]. In any case, the role for endo-RNAi here is to modulate target activity, since full suppression of these genes is clearly deleterious. Still, we may speculate further that as several well-conserved targets of hpRNAs encode protein activities that have been linked to speciation, such as heterochromatin, DNA damage, and energy homeostasis [19], these conserved genes may harbor selfish activities in certain species that warrants adaptive suppression by RNAi. Since the miRNA pathway is not typically involved in adaptive targeting, and instead relies upon capture of targets bearing invariant miRNA seed matches, the RNAi pathway may be more flexible to suppress such genes. A counterpoint to this is the fact that specific testis-expressed, evolutionarily young, miRNA clusters diverge even within their seed regions [21]. This suggests a possible atypical role for specialized miRNA loci in "hpRNA"-like evolutionary targeting dynamics.

Recurrent targeting of HMG-box factors in distinct hpRNA/RNAi networks

Our annotation of hpRNAs in two species indicates that, as a rule, these endogenous siRNA loci comprise very short-lived genes. Thus, they can at best only mediate modestly conserved regulation. If this is the case, can we learn any general principles from such fast-evolving regulatory networks? In fact, when considering this study alongside recent literature, we find several recurrent themes that provide a framework for understanding how endogenous RNAi is harnessed in biology.

We recently found that de novo hpRNAs in the simulans clade are required to silence a newly-emerged, amplifying, and selfish set of X-linked protamine derivatives, namely the Dox family [15, 16]. Protamines are central factors that condense the sperm genome, and therefore seem ripe for co-option by selfish factors to disrupt paternal inheritance. Indeed, following removal of histones, multiple sperm nuclear basic proteins (SNBPs) play roles in packaging sperm chromatin, and most of these contain HMG-box domains. We find that beyond the X-linked Dox family, there is a separate amplification of X-linked tHMG box genes in D. simulans that are concomitantly associated with silencing by cognate hpRNAs. Thus, we infer there is recurrent innovation of selfish SNBP activities by X chromosomes, consistent with the notion of sex chromosome meiotic drive that requires silencing by endogenous RNAi. Protamines are also functionally relevant to activity of Segregation Distorter, an autosomal meiotic drive system in D. melanogaster [41, 42]. Moreover, the Malik group recently reported high turnover of testis HMG-box loci across the Drosophilid genus, supporting the notion that their rapid evolution is due to recurrent intragenomic conflict between sex chromosomes {Chang and Malik eLIFE 2023}. We predict that hpRNAs may be employed to silence other protamine-based meiotic drive phenomenon in other species.

Within the Dox superfamily system itself, we document the innovation of novel hpRNAs that bear chimeric domain structures characteristic of coding Dox genes, but that lack the HMG-box. Instead, they incorporate sequences from the piwi-interacting RNA (piRNA) factors Tapas and Krimper [43, 44], also bear TE fragments. Homology to piRNA factors Tapas and Krimper were recognized in a prior study in distinct duplicate copies of the ancestral “ODox” gene, termed X:17.1 and X:17.2 [16]. However, our analyses revealed that the duplicate copies of the ancestral ODox gene bear inverted repeats and now generate hpRNA-siRNAs (two independent loci, which we termed hp-ODox1 and hp-ODox2) in contemporary D. simulans (Fig 3). Overall, the genetic conflict that led to identification of Dox and Nmy [9, 11] is actually part of a far more extensive and rapidly evolving network of putative meiotic drivers and suppressors, and may in fact integrate activities of the siRNA and piRNA pathways. Relevant to this, the third recognized sex ratio meiotic drive system in D. simulans ("Paris"), is driven by HP1D2 [13], a derivative of the core piRNA factor Rhino [45, 46]. Thus, there are recurrent linkages of piRNA factors to sex ratio meiotic drive, notwithstanding that TEs are themselves selfish genetic elements.

Of course, TEs are intrinsically selfish elements that are targeted by host genomic defenses, most famously by piRNAs. However, recent studies provide analogous conceptual involvement for co-option of the piRNA pathway by drive systems. As mentioned, the Paris SR system utilizes a de novo copy of the HP1-like factor Rhino [13], a central nuclear piRNA factor that defines piRNA cluster transcription [45, 46]. As another example, telomeric TART elements were found to have captured a fragment of Nxf2 [47], a piRNA-specific copy of the mRNA export machinery that gained activity in co-transcriptional silencing [48–51]. We surmise that the capture of some piRNA factors by hpRNA loci may in fact reflect their selfish activities of such defense factors.

The adaptive deployment of hpRNAs in the testis in D. simulans highlights that some of the most important biologically overt manifestations of endo-RNAi cannot be studied in the major model system D. melanogaster. Looking to other Drosophilids, the recognition of rampant duplications of the RNAi effector AGO2 in various obscura clade species, resulting primarily in testis-restricted paralogs [52–55], provides a further hint into active genomic conflicts that may be playing out in these species. Indeed, intragenomic conflicts that mediate aberrant sex ratio and/or sterility, specifically in male fathers, have been documented in the obscura clade [56, 57]. The genetic factors in these conflicts remain to be documented fully, but it is intriguing to hypothesize whether de novo hpRNAs might be involved in any of these scenarios. Overall, the Drosophila RNAi/hpRNA pathway provides a policing system that helps to surveil and silence gene expression in the testis against selfish meiotic loci. We speculate that such cycles of drive and repression are poised to underlie speciation [5, 58, 59].

Drosophila strains

Stocks bearing D. melanogaster dcr-2 alleles wIR; dcr-2[L811fsx]/CyO and wIR; dcr-2[R416X]/CyO were obtained from Richard Carthew (Northwestern) [60]. D. simulans w[XD1] wild-type strain was obtained from BestGene, Inc. and used as the control strain for mutant comparisons. dcr-2 loss of function mutant with DsRed allele replacing the endogenous locus was made in the w[XD1] strain background as reported in [14]. Similar to dcr-2 DsRed mutant allele, we also generated dcr-2 mutant allele replacing the endogenous locus with a mini-white⁺ marker for efficient selection of dcr-2 trans-heterozygous mutants by crossing dcr-2[DsRed] and dcr-2[white⁺] parental files. Oligo sequences for dcr-2 targeting are listed in S5 Table. All flies were reared on standard cornmeal molasses food. As D. simulans lack balancer chromosomes, and as dcr-2 homozygous mutants are also male sterile [14], we maintained dcr-2[DsRed] and dcr-2[white⁺] alleles by visual selection of markers every few generations to maintain the alleles.

Testis dissection and RNA preparation

For D. melanogaster testis dissections dcr-2 mutant testis were collected from wIR; dcr-2[L811fsx]/dcr-2[R416X] trans-heterozygotes and wIR; dcr-2[R416X]/+ heterozygous flies were used as controls. For D. simulans, we collected testis from dcr-2[DsRed]/[white⁺] trans-heterozygous mutants, and used the parental strain w[XD1] as control. Briefly, testis from 3 days old flies were extracted in TRIzol (Invitrogen) in batches of 10 flies at a time and the testis samples were flash frozen in liquid nitrogen. RNA was extracted from 25–50 testis per genotype.

Small RNA and RNA-seq library preparation

RNA extraction was performed as described in [14], and the quality of RNA samples were assessed with the Agilent Bioanalyzer. RNA samples with RIN >6.5 were used for library preparation using the Illumina TruSeq Total RNA library Prep Kit LT. Briefly, for RNA-seq libraries we used 650 ng of total RNA, and we used the Manufacturer’s protocol except for reducing the number of PCR cycles from 15 as recommended to 8, to minimize artifacts that may arise from PCR amplification. We prepared stranded RNA-seq libraries for D. simulans and unstranded libraries for D. melanogaster as RNA samples were extracted and processed in different time points. Samples were pooled using barcoded adapters provided by the manufacturer and the paired-end sequencing was performed at New York Genome Center using PE75 in the Illumina HiSeq2500 sequencer.

We prepared small RNA libraries using ~20 μg total RNA, as previously described [14]. To the total RNA pool, we added a set of 52 RNA spike-ins, spanning a range of concentrations (QIAseq miRNA Library Spike-In kit #800100). A list of spike-in sequences used for small RNA library preparation is provided in S5 Table. Briefly, small RNAs of size 18- to 29-nt-long small RNAs were purified by preparative PAGE. Next, the 3′ linker (containing four random nucleotides) was ligated overnight using T4 RNA ligase 2, truncated K227Q (NEB), after which the products were recovered by a second PAGE purification. 5′ RNA linkers with four terminal random nucleotides were then ligated to the small RNAs using T4 RNA ligase (NEB) followed by another round of PAGE purification. The cloned small RNAs were then reverse transcribed, PCR amplified and sequenced using P50 single-end sequencing on the Illumina HiSeq 2500 sequencer.

5’-seq and 3’-seq library preparation

To map 5’ ends, we used the parallel analysis of RNA 5′ ends from low-input RNA (nanoPARE) strategy [24]. For D. simulans libraries, testis was extracted from <1-week males and total RNA was extracted using TRIzol. cDNA was prepared using Smart-seq2 [61] and tagmented using the Illumina Nextera DNA library preparation kit, purified using the Zymo 5x DNA Clean and Concentrator kit (Zymo Research), and eluted with resuspension buffer. For 5’-end enrichment PCR, the purified reaction was split and amplified either Tn5.1/TSO or Tn5.2/TSO enrichment oligonucleotide primer sets. PCR reaction products with Tn5.1/TSO enrichment oligonucleotide and Tn5.2/TSO enrichment oligonucleotide primer sets were pooled and purified using AMPureXP DNA beads. Final libraries were checked for quality on an Agilent DNA HS Bioanalyzer chip. Libraries with size ranges between 150 and 800 bp were diluted and sequenced to 10–15 million single-end 50-bp reads per sample using a custom sequencing primer (TSO_Seq) and a custom P5/P7 index primer mix on an Illumina HiSeq 2500 instrument.

To annotate 3’ transcript termini, we used the QuantSeq 3’ mRNA-seq library preparation REV kit for Illumina (Lexogen) with a starting material of 50 ng total RNA from D. melanogaster and D. simulans control and dcr-2 mutant samples, according to manufacturer’s instructions. cDNA libraries were sequenced on Illumina HiSeq-1000 sequencer with single-end SE 50 mode.

Genomic analyses of RNA-seq data in D. melanogaster and D. simulans

De novo annotation of testis transcriptome

In addition to previously annotated transcripts/genes from the FlyBase annotation, we performed de novo annotation of our transcriptome data to identify additional, novel testis-expressed transcripts in D. melanogaster and D. simulans. The novel annotated transcripts were then supplemented with known annotations to make a combined set of 17285 transcripts in D. melanogaster and 15119 transcripts in D. simulans. We employed two independent, genome assembly guided transcript prediction algorithms, Cufflinks [71] and StringTie [67]. For both methods, de novo transcripts were predicted for each RNA-seq dataset, and a merged transcript model was generated encompassing the transcriptome from WT and mutant datasets. hpRNAs were predicted using the scheme shown in S2 Fig, and visualized using the Integrated Genomics Viewer (IGV) [72]. The termini of primary hpRNA transcripts were refined using the 5’-seq and 3’-seq data.

Homology searches for hpRNA orthologs

For each novel hpRNA identified in D. simulans, we first determined its genomic location based on flanking genes, and searched for homologous sequences (with synteny) in other closely related simulans clade species (D. mauritiana and D. sechellia), as well as D. melanogaster and D. yakuba as a close outgroup species. No matches to newly-identified D. simulans hpRNAs were recorded in D. melanogaster and D. yakuba, whereas most of these had hits in D. mauritiana and/or D. sechellia. In the latter cases, we assessed them further for presence at the syntenic location of the D. simulans hpRNA. This proved to be the case for all loci with homologs, in which case we infer their birth within the simulans clade ancestor. In some cases no hits were identified in other species, suggesting its potential birth in the lineage leading to D. simulans.