Rsp-derived small RNAs are underrepresented in the small RNA transcriptome of SD-Mad testes

Because there is evidence that satellite-derived small RNAs (smRNAs) are important for sperm development [36,37], we hypothesized that SD may interact with these smRNAs as part of its mechanism. We explored the small RNA transcriptome of testes heterozygous for combinations of two different driving SD haplotypes (SD-Mad and SD-5) (Fig 1A) and wild type chromosomes with different Rsp alleles (Fig 1A). Because our previous work found a correlation between Rsp copy number and RNA abundance [34], all of our comparisons are relative to Rsp copy number controls (R16/Gla or R16/Iso1 where R16 contains a Rsp deletion allele and Iso1 and Gla contain sensitive and super-sensitive Rsp alleles, respectively; Fig 1A) [26]. The SD haplotypes we use differ by their inversion types (Fig 1A) but both exhibit strong drive, measured as k, or the proportion of SD offspring from a cross with SD heterozygous males (Table A in S1 Appendix).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Analysis of small RNAs in SD-Mad heterozygotes show a lower abundance of small RNAs corresponding to Rsp.

A) A schematic of the 2nd chromosomes used in this study. Iso1 and Gla are sensitive backgrounds with varying copy number of Rsp. Iso1 contains ~1100 copies of Rsp making it sensitive and Gla contains >3000 making it super sensitive. R16 is a copy number control for SD and does not drive. SD-Mad and SD-5 are driving chromosomes containing different inversions (highlighted in colored boxes). They contain the main drive locus, Sd-RanGAP and show perfect or near perfect drive against sensitive chromosomes. SD-Mad^Rev is a haplotype control for SD-Mad. It is identical to SD-Mad except for a deletion of Sd-RanGAP. B) A comparison of the DESeq2 normalized counts of Rsp in each comparison shows that Rsp smRNAs are less abundant in SD-Mad heterozygotes relative to R16 heterozygotes (Log2FC = -0.95, P_adj = 5.1e − 9). Rsp smRNAs are apparently unaffected in SD-Mad^Rev heterozygotes indicating that the reduction of smRNAs in SD-Mad heterozygotes is correlated with drive. Rsp RNAs are also unaffected in the SD-5 background indicating that SD-5 does not perturb Rsp smRNA abundance. C) A heatmap showing the differential abundance of smRNAs for major piRNA source loci in the genome. 1.688, flamenco, and 38C2 are differentially expressed in this background.

https://doi.org/10.1371/journal.pgen.1012235.g001

We found significantly fewer Rsp-derived smRNAs (piRNAs and endo-siRNAs) in some driving testes. Rsp-derived smRNAs are significantly less abundant in the testes of SD-Mad/Iso1 heterozygotes relative to the non-driving copy number control (R16/Iso1) (Fig 1B; Log2FC = -0.95, P_adj = 5.1e^-9, DESeq2). To confirm that the low abundance of Rsp smRNAs correlates with drive rather than non-drive related factors in the genetic background, we created a revertant SD-Mad background (SD-Mad^Revertant or SD-Mad^Rev, Fig 1A). We used CRISPR/Cas9 to delete the Sd-RanGAP duplication from SD-Mad, leaving one full length copy of RanGAP and the rest of the chromosome with putative drive modifiers intact. This chromosome is therefore identical to SD-Mad outside of Sd-RanGAP but no longer exhibits strong drive (k* = 0.54). We expect that any differences in smRNA abundance between SD-Mad/Iso1 and SD-Mad^Rev/Iso1 are therefore due to drive or Sd-RanGAP itself. Indeed, we found that Rsp smRNAs are more abundant in SD-Mad^Rev/Iso1 relative to SD-Mad/Iso1 (Log2FC = 0.66, P_adj = 0.004, DESeq2) and at similar levels as the copy number control, R16/Iso1 (Fig 1; Log2FC = -0.21, P_adj = 0.19, DESeq2). This indicates that the difference in Rsp smRNA abundance in SD-Mad/Iso1 is likely due to the drive phenotype. We see a similar dearth of Rsp-derived smRNA in driving testes in an independent background with a super-sensitive Rsp allele (SD-Mad/Gla) compared to its non-driving copy number control (R16/Gla) (S1 Fig; Log2FC = -0.68, P_adj = 2.6e^-14, DESeq2). This suggests that the effects on Rsp smRNA abundance are related to the driving SD-Mad chromosome rather than the target chromosome.

To determine if the effect on Rsp smRNAs is specific to SD-Mad, we explored the small RNA transcriptome in driving testes with another SD haplotype (SD-5) that differs in its chromosomal inversions [29]. Intriguingly, Rsp smRNA abundance is similar between SD-5 and R16 in an Iso1 background (Fig 1; Log2FC = -0.07, P_adj = 0.79, DESeq2) and higher than expected levels in the Gla background (S1 Fig; Log2FC = 0.33, P_adj = 2.3e^-11, DESeq2). Therefore, the disruption of Rsp RNAs is not a universal feature of SD chromosomes.

Outside of Rsp, we see shifts in many repeat-derived smRNAs in driving testes of both SD haplotypes: for example, 30% (12/40) and 27.5% (11/40) of repeat annotations (satellite DNAs, piRNA clusters, and TEs) were differentially expressed (P_adj ≤ 0.01) in SD-Mad and SD-5, respectively (S2 Fig; with similar—56.4% (22/39) and 50% (20/40)—effects in the Gla background; S1 File). There are only a few notable patterns that arise from these comparisons (Table B in S1 Appendix). We found that another complex satellite DNA on the X chromosomeunrelated to Rsp, 1.688, is significantly less abundant in all SD heterozygote testes in the Iso1 background comparisons (Fig 1C, Table B in S1 Appendix; Log2FC^SD-Mad = -1.1, P_adj^SD-Mad = 4.8e^-14; Log2FC^SD-MadRev = -0.64, P_adj^SD-MadRev = 1.16e^-12; Log2FC^SD-5 = -0.71, P_adj^SD-5 = 0.02, DESeq2) and in SD-Mad/Gla testes (S1 Fig; Log2FC^SD-Mad = -0.28, P_adj^SD-Mad = 0.001). SmRNAs derived from a piRNA cluster linked to Sd-RanGAP on chromosome 2L, 38C2, and the X-linked somatic piRNA cluster, flamenco, is differentially expressed for all genotypes in both backgrounds except for flamenco in SDMad/Gla (Figs 1C, S1). 38C2 and flamenco are not known to be involved in SD-mediated drive.

Our results demonstrate that SD-Mad and SD-5 haplotypes, despite sharing the same main drive locus, have different effects on smRNAs derived from repetitive loci such as complex satellites (including Rsp), transposable elements, and piRNA clusters. Since these SD chromosomes differ in their inversions, we propose other linked loci might affect the drive mechanisms and phenotypes.

The dearth of Rsp smRNAs in SD-Mad heterozygotes could be due to a disruption in transcription of the locus or subsequent processing steps. Many factors can influence piRNA production. For example, the piRNA pathway can amplify piRNAs independently of transcription, such as the ping pong cycle, [41]. Notably, Rsp piRNAs do not have a strong ping pong signature in testes [34,42]. To distinguish between a disruption in transcription or some downstream process, we examined total RNA.

Broad differences in the long RNA transcriptome of different driving backgrounds

We performed total RNAseq to determine how the expression landscape of coding and non-coding sequences differs in testes with drive. We found that 24.2% (3359/13825) of the total RNA transcriptome is differentially expressed (P_adj ≤ 0.01) in the SD-Mad background (SD-Mad/Gla vs R16/Gla; S3B Fig). Driving testes with SD-5 had similarly divergent transcriptomes compared to copy number controls (24.3% (3352/13747) DEG in SD-5/Gla vs R16/Gla; S3E Fig). Repetitive loci were more frequently differentially expressed (P_adj ≤ 0.01) in SD-Mad heterozygotes compared to SD-5 heterozygotes (41% (16/39) vs 23% (8/40), S4 Fig, Table C in S1 Appendix). Despite the effect on Rsp smRNA abundance in SD-Mad heterozygotes, the abundance of Rsp long non-coding RNAs (lncRNAs) does not show any consistent pattern between the backgrounds (S5A-S5B Fig, Table C in S1 Appendix). This could be due to low counts for Rsp. On the other hand, we did find some differences in repetitive elements related to rDNA (R1, R2, and IGS) and Tc1-Mariner family TEs (all backgrounds; S6 Fig). Interestingly, there was no correlation between the expression of TEs and the expression of piRNA clusters that contain fragments of these TEs in the total RNA, nor was there any correlation between the small RNAs from piRNA clusters and the total RNAs for those TEs. PiRNA clusters are usually defined in one isolate of Iso1: rapid turnover of TEs and piRNA sources could explain why we do not see a correlation between piRNA cluster expression and TE expression in our backgrounds.

The reduction in Rsp smRNAs without any corresponding change in the lncRNA in SD-Mad-mediated drive is intriguing, as piRNA precursors should be among these RNAs. Because previous studies based on immunofluorescence in testes showed that Sd-RanGAP is mislocalized in driving spermatocytes [23], we asked if the difference in smRNAs could be due to a mislocalization of the lncRNA. We performed RNA FISH for Rsp and the 1.688 satellite as a fluorescence control in testes of SD-Mad heterozygotes. The most abundant variant of 1.688 is the 359-bp satellite on the X chromosome, which is controlled for in our comparisons. Curiously, we do see some apparent differences in the pattern of 1.688 and, to a smaller extent, Rsp signals between driving and non-driving genotypes. However, the signals are nuclear (S7 Fig), which suggests that mislocalization of the lncRNA, at least at a gross scale, may not explain the reduction in Rsp smRNAs.

We performed complementary analyses to identify patterns in gene expression. Genes make up 99.6 – 99.8% of the differentially expressed sequences in driving testes (S1 File). However, many of these differences likely reflect divergence among the second chromosomes used in each condition rather than the drive phenotype itself. To address this, we focused on pathway-level changes and coordinated gene expression rather than individual genes.

We conducted gene set enrichment analyses using PANGEA [43], and compared all differentially expressed genes (P ≤ 0.01) to Reactome gene sets [44] (S2 File). These analyses revealed a consistent enrichment in terms related to metabolism, the immune system, and the Rho GTPase cycle across driving testes (S8 Fig). In addition, genes involved in sperm individualization (e.g., asp, didum, bug22, chic, ctp, jar, nsr) [45] were differentially expressed in both SD-Mad/Gla and SD-5/Gla (Fig 2A, S3 File), which is consistent with the hallmark chromatin defect [30–32].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Analysis of total RNA from SD heterozygotes shows a broad dysregulation of genes involved in individualization and the cell cycle.

A) A heatmap showing the direction and significance of differential expression for genes whose perturbation has been shown to result in defects at the individualization stage ([45]). Boxes outlined in red are genes that are differentially expressed in both SD haplotypes and in the same direction. B) WGCNA analysis of the total RNA identified modules of genes that are upregulated or downregulated in SD heterozygotes relative to the R16/Gla control. These modules were enriched for genes in Reactome gene sets. Module B contained genes that are downregulated and which are in gene sets related to metabolism, the immune system, and the Rho GTPase cycle. Module H contained genes that are upregulated and which are in gene sets related to the cell cycle and post translational modification.

https://doi.org/10.1371/journal.pgen.1012235.g002

To further identify coordinated expression patterns, we performed weighted gene co-expression network analysis (WGCNA) [46]. WGCNA identified modules of co-expressed genes that distinguish driving genotypes from R16/Gla (Fig 2B, Table D in S1 Appendix, S2 File). Gene set enrichment analysis of these modules using PANGEA [43] revealed that the drive phenotype is negatively correlated with several terms related to metabolism, the Rho GTPase cycle, and the immune system (Module B, Fig 2B).

Together, both pathway-level and network-based analyses implicate disruptions in cytoskeletal organization and/or sperm individualization. Whether these effects are directly related to drive mechanisms or downstream consequences remains unclear, but they are consistent with known phenotypes of spermiogenic failure.

WGCNA also identified two modules that were upregulated in driving genotypes. Our gene set enrichment analysis with PANGEA showed that one module (Module A) was not associated with any enriched pathways, whereas the other (Module H) was enriched for genes involved in the cell cycle and post-translational protein modification (Fig 2B).

Rsp is expressed well before individualization [34], during the period when male germ cells are actively dividing. If Rsp RNAs indeed contribute to the SD mechanism, the enrichment of cell cycle-related genes in Module H is consistent with a mechanism that originates earlier in spermatogenesis, around or before meiosis. Although the phenotype manifests after meiosis, temperature shift experiments suggest that the critical window for SD activity occurs before or during meiosis [47]. Together, these results further support a model in which SD acts early in spermatogenesis, with defects only becoming apparent later, during individualization.

To identify genes that might interact to cause drive, we compared the gene expression of SD-Mad/Iso1 to SD-Mad^Rev/Iso1. These genotypes only differ by the presence of the main drive locus, Sd-RanGAP. We performed both totRNA and 3’ Digital Gene Expression (DGE) RNA sequencing and examined the overlap in differential expression between the totRNA and DGE sequencing. There are 69 differentially expressed genes where the DGE comparison is significant (P^DGE ≤ 0.01), and the sign of the Log2FC of the totRNA matches that of the DGE. Among this set of differentially expressed genes, 57 show at least a 50% difference in gene expression (absolute Log2FC value of at least 0.58 in DGE). These genes are not enriched in any Reactome gene sets. The top 20 most differentially expressed genes consists of 9 lncRNAs (3 anti-sense RNAs) and 11 protein coding genes: 8 of which are uncharacterized. The 3 characterized genes are Artemis (Arts), Gr61a, and Tono (S9 Fig, S1 File).

Loss of Kipferl results in the upregulation of satellite DNAs and weaker SD-Mad drive

Our RNA seq analysis suggests that the reduction in Rsp smRNAs in SD-Mad/Iso1 may be involved in the mechanism of drive. To test this hypothesis, we sought to determine if overexpressing Rsp is sufficient to rescue wild-type spermatids from SD. If drive is a consequence of reducing Rsp smRNAs in an SD-Mad background, then overexpressing Rsp RNAs in SD-Mad/Iso1 testes may reduce drive strength. To overexpress Rsp, we took advantage of a recently-discovered component of the piRNA pathway called Kipferl (kipf) [48]. In ovaries, kipf recruits piRNA transcriptional machinery to so-called kipf-dependent piRNA clusters. In its absence, the piRNA machinery becomes enriched at complex satellite DNAs, including Rsp and an unrelated repeat called 1.688, leading to their upregulation [48] (Fig 3A-3B). While previous studies did not detect evidence for kipf expression in testes [49,50] (S10 Fig), we found that kipf transcripts are indeed present in testes, albeit at low levels (S10 Fig). We also found that, like in ovaries, satellite DNAs are overexpressed in kipf knock out (KO) testes, as demonstrated by reverse transcription quantitative polymerase chain reaction (RT-qPCR) (Fig 3B; testis fold change: 359-bp, a component of the 1.688 complex satellite = 2.85, P = 0.064; Rsp = 1.94, P = 0.038, t-test).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Satellite DNA overexpression via kipf-KO is sufficient to rescue drive in SD-Mad heterozygotes.

A) A schematic of how we expect the localization of the RDC complex, the piRNA transcriptional machinery, to change upon knockout of kipf in testes. Typically, kipf recruits the RDC complex to specific piRNA source loci throughout the genome. Data from ovary suggests that knocking out kipf leads to an enrichment of the RDC complex at satellite DNAs. B) qRT-PCR in kipf KO homozygotes (kipf/kipf) and heterozygotes (kipf/TM3) shows that complex satellites, and specifically Rsp, is upregulated in both ovaries and testes when kipf is knocked out. C) Drive strength assays show that kipf-KO individuals exhibit much lower drive strength when compared to the heterozygote in an SD-Mad background. We see a mild but significant reduction in drive in the SD-5 genotype despite not seeing a Rsp smRNA phenotype in the smRNA seq.

https://doi.org/10.1371/journal.pgen.1012235.g003

To determine if a loss of kipf (and thus overexpression of Rsp) is sufficient to rescue wild-type sperm from SD-Mad, we used a knockout generated by deleting the genomic locus [48]. We examined the effects of kipf-KO on the drive strength of SD-Mad and SD-5 in an Iso1 background. We estimate drive strength as k: the proportion of the progeny of SD heterozygotes that contain the SD chromosome. A k value of 1 represents perfect drive and a k value of 0.5 represents fair Mendelian segregation. We found that knocking out kipf was sufficient to reduce the strength of drive in SD-Mad heterozygotes from ~0.975 to ~0.815 (P < 0.00001, arcsine transformation then t-test (asin-t), Fig 3C, S4 File). We observed a similar reduction in drive strength (from k = 0.98 to k = 0.88; P < 0.00001, asin-t, S11A Fig) when we knock down kipf (kipf-KD) compared to a white shRNA control using RNA interference [48]. We also tested kipf-KD in a semi-sensitive Rsp deletion line (Iso1ΔC8, hereafter C8) derived from the Iso1 chromosome but with ~50% of the Rsp locus deleted using CRISPR/Cas9 [51]. In a SD-Mad/C8 background, kipf-KD reduced drive from k = ~0.85 to k = ~0.72 (P < 0.00001, asin-t, S11B Fig). Kipf-KO had a mild but statistically significant effect on the drive of SD-5 heterozygotes (from k = ~1 to k = ~0.94, P < 0.00001, asin-t, Fig 3C).

The reduction in drive strength in the absence of kipf is consistent with the hypothesis that Rsp transcription, transcripts, or their processed products (i.e. piRNAs) are involved in the mechanism of SD-Mad and are somehow important for the development of Rsp-bearing sperm. However, the loss of kipf also overexpresses other satellite DNAs, like 1.688 (Fig 3B), and likely dysregulates kipf-dependent piRNA clusters [48].

Overexpression of Rsp piRNAs and weaker drive in SD-Mad heterozygotes

To specifically overexpress the Rsp repeat, we inserted 3 copies of the Rsp monomer into the abundantly expressed piRNA cluster, 38C1, in an Iso1 2^nd chromosome (Iso1^Rsp>38C1) (Fig 4A). This insertion results in the overexpression of Rsp in both SD backgrounds (SD-Mad background: Log2FC = 0.45, P_adj = 1.6e^-10, DESeq2; SD-5 Log2FC = 0.47, P_adj = 0.008, DESeq2) (Figs 4B-4C, S12, Table E in S1 Appendix, S1 File). We detect a mild effect on the expression of 38C1 (Fig 4B; SD-Mad: Log2FC = -0.34, P_adj = 2.3e^-11, SD-5: Log2FC = -0.37, P_adj = 4.3e^-5, DESeq2) but the Rsp insertion generally does not influence transcription at other piRNA clusters (Fig 4B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Overexpression of Rsp via 38C1 insertion is sufficient to reduce drive strength of SD-Mad heterozygotes but has no effect on SD-5 heterozygotes.

A) Schematic of the overexpression construct. Three copies of Rsp were inserted roughly 1/3 of the way into the piRNA cluster 38C1 of Iso1 using CRISPR/Cas9-mediated homology directed repair. The construct included a floxed 3xP3-DsRed reporter which was removed using Cre after the insertion was validated. AttP sites were also included to facilitate additional insertions of Rsp. B) Rsp is significantly overexpressed in the RspOE strain for both SD heterozygotes relative to the same genetic background with the unmodified 2nd chromosome. 38C1 on the other hand was significantly downregulated. Other piRNA clusters were largely unaffected. C) A comparison of the DESeq2 normalized counts from SD/Iso1 and SD/RspOE smRNA. Rsp is significantly overexpressed in both conditions. D) Drive strength assay for SD-Mad in an RspOE background relative to Iso1. Drive is significantly and substantially reduced the SD-Mad/RspOE genotype. E) Drive in SD-5/RspOE was unaffected. These observations demonstrate that SD-Mad and SD-5 appear to interact with Rsp differently.

https://doi.org/10.1371/journal.pgen.1012235.g004

We tested the transgenic flies with the full Rsp overexpression construct including the 3xP3-DsRed reporter (Iso1^{Rsp-DsRed>38C1} and did not see a change in drive strength, Figs 4A,4D, S13, S4 File). We suspect that the insertion of a strong euchromatic promoter into an otherwise heterochromatic locus may affect the RNA species produced from the locus, so we removed the 3xP3-DsRed reporter via recombination using Cre (Figs 4A, S13, S4 File).

When we tested the drive sensitivity of Iso1^Rsp>38C1 in an SD-Mad heterozygote, we found that drive strength is significantly lower relative to Iso1 (Fig 4D, S4 File). This provides further evidence that SD-Mad somehow perturbs Rsp RNAs in the presence of Sd-RanGAP.

In contrast, Rsp overexpression had no effect on drive strength in SD-5 heterozygotes in an otherwise identical background (Fig 4E, S4 File), consistent with the smRNA results suggesting that Rsp RNAs are not perturbed in SD-5. Taken together, these results suggest that SD-Mad and SD-5 have diverged in the factors and features involved in their drive.

smRNAs derived from the target satellite DNA affect drive

Our data indicate that SD-Mad, but not SD-5, disrupts the abundance of Rsp-derived smRNAs and that this phenotype correlates with drive. Our observations add SD to the growing list of meiotic drive systems that involve smRNAs [11]. We observed an effect on drive strength—and the increased survival of wild-type Rsp-bearing sperm—even with a modest overexpression of Rsp RNA (~2 fold with kipf KO or KD and ~1.5-fold with Iso1^Rsp>38C1). The ability of the Iso1^Rsp>38C1 transgene to rescue wild-type sperm may depend on transcription of the source RNAs and genetic background: we did not see an effect on drive without removing 3xP3-DsRed from the 38C1 insertion (Figs 4D-4E, S13). The difference in drive suppression phenotypes between Iso1^Rsp>38C1 and Iso1^{Rsp-DsRed>38C1} may suggest that although transcription is taking place at the Rsp > 38C1 locus in Iso1^{Rsp-DsRed>38C1} (Fig 4B), the transcripts may not be shunted into any functional RNA pathways. Nonetheless, the kipf experiments (Fig 3) overexpress Rsp from the endogenous locus rather than a transgene in a piRNA cluster, albeit not specifically, and therefore bolsters our conclusions that Rsp RNAs are involved in SD-Mad drive. Future studies that focus on how different species of RNA might protect sperm and influence drive phenotypes may provide important clues about drive mechanisms and the functions of these RNAs.

We know relatively little about the rules that govern transcription at satellite DNAs. Previously, we found that the Rsp satellite is regulated by non-canonical heterochromatin-dependent transcription machinery, similar to dual-stranded piRNA clusters [34,35]. There may be specific transcription factors important for licensing transcription at Rsp, similar to the role of HP2 in regulating the transcription of AAGAG [52]. Identifying potential proteins that interact with Rsp may fill in an important piece of the puzzle linking satellite transcripts to drive phenotypes. Satellite transcripts and their interacting proteins play similar roles in spermatogenesis, provide important clues about why satellites like Rsp are targets of drive.

Several drive systems involve smRNAs as suppressors of drive [12–14,53]. Rsp smRNAs on the other hand appear to, either directly or indirectly, be targets of drive. The protective effect of increased Rsp RNAs in SD-Mad-mediated drive suggests that these smRNAs may play a role in chromatin regulation but these Rsp-derived smRNAs are not required for spermatogenesis in the absence of the Rsp genomic locus. Spermatogenesis normally proceeds without issue for chromosomes with Rsp deletions [34]. Interestingly, despite being a target of SD drive, large Rsp loci are abundant in natural populations [33] and there is some evidence that there may even be a fitness advantage associated with large Rsp alleles [54], which could explain its persistence.

We propose a model where satellite DNAs require epigenetic regulation by smRNAs in a dose-dependent manner during spermatogenesis and SD exploits this feature to bias its transmission. The epigenetic profile of male germ cells is dynamic. Histones experience multiple post-translational modifications before ultimately being replaced by protamines [55,56]. We hypothesize that the smRNAs direct H3K9 trimethylation at satellite DNAs. Without proper coordination, we propose that the satellite cannot be correctly repackaged with the rest of the pericentromere during meiosis or the histone to protamine transition, triggering a checkpoint on sperm chromatin quality that leads to an individualization failure. In SD-Mad drive, disrupting Rsp smRNA would affect the wild type locus with the large block of repeats, but SD-Mad chromosomes would be unaffected because they lack the repeat.

Alternatively, if a lack of smRNA regulation makes Rsp more accessible, it may become a sink for transcription factors or other proteins. For example, the pioneering factor GAF binds to the simple satellite, AAGAG, during embryogenesis and is required for its transcriptional silencing [57]. Either of these defects could be the cause for the eventual elimination of the affected spermatids. Without a better understanding of the dynamic epigenetics at the Rsp locus during spermatogenesis, it is difficult to infer why small RNAs are important for spermatids that contain large Rsp loci. Identifying proteins that interact with Rsp DNA and RNA will be important for understanding its regulation during spermatogenesis.

It is intriguing that while SD-Mad involves Rsp-derived RNAs, we do not detect any effects of SD-5 on Rsp-derived RNAs, and we see no effect on the strength of drive upon overexpressing Rsp RNA in an SD-5 background using Iso1^Rsp>38C1. This suggests that SD-Mad and SD-5 drive via different specific mechanisms. It is interesting that although SD-5 heterozygotes did not affect Rsp RNA, the kipf-KO experiments revealed a small but significant effect on drive strength. This suggests that other piRNA clusters, transposons, or satellite DNAs can affect drive. This also suggests that, while SD-5 and SD-Mad have different effects on smRNAs, SD-5 associated drive mechanisms may operate downstream of Rsp piRNAs.

Mechanistic variation in drive haplotypes

We suggest that the mechanistic variation between SD-Mad and SD-5 arises from molecular differences in the haplotypes. The most conspicuous differences between these and other SD haplotypes are in their chromosomal inversions (Fig 2A). These inversions may link different genetic variants, including enhancers of drive, to the main drive locus, creating different co-adapted gene complexes. While there is abundant genetic evidence for linked modifiers that enhance SD (i.e., Enhancer of SD (E(SD)) [26], Modifier of SD (M(SD)) [27], and Stabilizer of SD (St(SD)) [28]) on some well-studied SD haplotypes, none have been molecularly identified. The suppression of recombination under chromosomal inversions and around the centromere, preventing us from mapping putative modifiers precisely. SD-Mad may contain a modifier that influences Rsp smRNA abundance that is absent in SD-5. Further experimentation is required to molecularly characterize the differences between these chromosomes and whether the effects on drive arise from a single locus or epistatic interactions between loci.

All SD chromosomes share the main drive locus, Sd-RanGAP, which is a duplication of an essential component of the Ran cycle [25]. We suspect that there are myriad ways to induce sperm dysfunction involving the Ran cycle, and that while different SD haplotypes share the Sd-RanGAP duplication, they may recruit different modifiers. This system-level property of SD might affect its population dynamics. SD drive prompts the evolution of drive suppressors [19]. SD chromosomes that acquire new modifiers that help escape suppression may gain a transmission advantage over SD haplotypes targeted by suppressors [10,33]. These dynamics could explain the rapid turnover of SD haplotypes in populations across the globe [58–61].

Drive-correlated transcriptional changes

Our differential expression analysis did not suggest specific modifier candidates. However, we identified potentially interesting genes among those most differentially expressed between SD-Mad testes with and without the main drive locus that may highlight processes involved in drive mechanisms.

First, Tono, a BTB zinc finger-containing transcription factor is upregulated (Log2FC^DGE = 1.7) in all SD-Mad comparisons. Tono plays a role in regulating transcription in muscle cells in response to mechanical pressure [62] but also shows enrichment in male germ cells [63]. The putative DNA-binding capacity and ability to form nuclear condensates [62] makes this an interesting candidate gene for interacting with the Rsp satellite. Second, the importin-4 ortholog, Artemis (Arts), which facilitates Ran-mediated import of H3 and H4 is overexpressed in SD-Mad (Log2FC^DGE = 2.5). Interestingly, Arts expression is antagonistic to male fertility [64]. Also of note, Apollo, a paralog of Arts which supports male fertility [64] is downregulated (Log2FC^DGE = -0.6) though it is not in the top-most differentially expressed genes.

Many of the differentially expressed genes in the transcriptomes of driving testes may be more related to the drive phenotype than the mechanism. First, we found that multiple repetitive loci are differentially expressed with little evidence that they influence drive. While this could be due to the different 2^nd chromosomes in the comparison, it could also be due to incidental effects from the drive mechanism on the Rsp locus. Second, we found that genes related to individualization and the cell cycle are broadly differentially expressed (Figs 2, S8). Third, we detect a consistent differential expression in genes involving cytoskeleton dynamics. SD-targeted sperm experience a delay in protamine loading and can have nuclear morphology defects [30] leading to SD-targeted sperm being eliminated around the individualization stage. Elongation and individualization heavily rely on actin and microtubule dynamics [65] and appears to be a common point of sperm elimination (reviewed in [45]). As SD-mediated drive eliminates wild type sperm at the individualization stage, the enrichment of genes involved in cytoskeleton-related processes in our analyses could be related to the drive phenotype rather than its proximal cause. We suspect that the timing of the proximal cause of SD-mediated drive may align with early spermatogenetic processes; perhaps where cell cycle-related genes are active and appear to be broadly differentially expressed (Fig 2B, Module H). This earlier timing is consistent with temperature shift experiments that place the critical period for SD at or before meiosis [47].

Regardless of the precise mechanism, different SD chromosomes may converge on similar phenotypes if the dysfunctional sperm targeted by SD trigger a sperm quality checkpoint that eliminates spermatids [30,66]. Many have proposed the existence of such a checkpoint. Events that take place during prophase/metaphase may be linked with late-stage sperm elimination [67,68]. Lastly, a recent study identified a component of this putative checkpoint [66]. The enrichment of genes related to cytoskeleton pathways found in our gene set analysis could be associated with the common elimination mechanism of this putative checkpoint. Alternatively, the SD-induced defect may not trigger this putative checkpoint and instead, the spermiogenic failure at individualization is simply the end point of a cascade of perturbations that begins prior to meiosis.

In this paper, we make significant progress toward understanding molecular mechanisms of SD in connecting Rsp RNA to SD. Future studies of SD-related phenotypes and effects on satellite DNA transcription and chromatin, may have general implications for understanding how spermatogenesis is vulnerable to drive and other cheaters.

Drive is widespread across species and has significant consequences for the fundamental process of gametogenesis and its evolution (reviewed in [3,69,70]). Revealing drive mechanisms could help us understand fundamental aspects of spermatogenesis, with implications for our understanding of male infertility, and the development of efficacious and safe synthetic drivers to control vectors of human disease, address agricultural challenges, and species conservation.

Fly stocks

We maintained all fly stocks and crosses at 25°C on cornmeal medium. Iso1 [71], R16 [26], CyO/Gla, SD-Mad and SD-5 [72]. Iso1 chromosomes are typically sensitive to drive (Rsp^s allele with ~1100 copies) and Gla chromosomes are typically super-sensitive to drive (Rsp^ss allele with >2000 copies), although the Gla chromosome we use here has lower drive strength in the SD-Mad background (Table A in S1 Appendix). R16 is a complete deletion of Rsp and the surrounding sequences [26].

The stocks for the kipf knockdown experiments were provided as a generous gift from Julius Brennecke [48] and are now available from the Vienna Drosophila Stock Center: Kipf-sh VDRC# 314048, w-sh VDRC# 313772.

The Sd-RanGAP deletion line SD-Mad^Rev is a transgenic line that we made using CRISPR. We generated a gRNA-expressing construct by inserting the gRNA sequence (AGGAGGATTTGGAATAGTC) into pCFD3 vector following the protocol in [73]. This gRNA sequence matches the sequence near the breakpoint between the Sd-RanGAP and wild type RanGAP gene. We also generated a repair construct with the homology arms matching the sequence around the CRISPR cut site and a tag. To perform CRISPR on the SD-Mad chromosome, we generated a fly stock with the X chromosome from the Bloomington Drosophila Stock Center #51323 (y[1] M{vas-Cas9}ZH-2A w[1118]), and the 2nd chromosome from SD-Mad. The injection of the pCFD3 gRNA construct into this Cas9;SD-Mad fly was done by GenetiVision. The transgenic flies were screened, and the deletion of Sd-RanGAP was verified by PCR.

Rsp overexpression transgenic fly

We used the CRISPR/Cas9 system to insert Rsp sequences into the 38C1 piRNA cluster to generate the RspOE_DsRed line. We generated a gRNA expressing construct by inserting the gRNA sequence (GATTCGATCTCAAGACCCAC) into a pCFD3 vector following the protocol in [73]. The designed gRNA targets position 20,154,890 on contig 2L_1 in the heterochromatin-enriched genome assembly [74]. To generate the repair construct, we synthesized two gBlock fragments (Integrated DNA Technologies, IDT) based on the homology arms around the CRISPR cut site. We inserted the gBlock with the left homology arm into the pDsRedattp (Addgene #51019) vector with SphI and NdeI (New England biolabs, NEB), and then inserted the gBlock with the right homology arm with BglII and XhoI (NEB). We synthesized a Rsp trimer sequence by PCR using primers (forward: GGAAAATCACCCATTTTGATCGC, reverse: CCGAATTCAAGTACCAGAC) described before in [34], and inserted it into the repair construct with AvrII and SacII (NEB). All insertions were verified with Sanger sequencing. To accomplish CRISPR on the Iso1 chromosome 2, we generated a fly stock with the X chromosome from the Bloomington Drosophila Stock Center #51323 (y[1] M{vas-Cas9}ZH-2A w[1118]), and the 2nd chromosome from Iso1. The injection of the pCFD3 gRNA construct and the repair construct into this Cas9;Iso1 fly was done by GenetiVision. The transgenic flies were screened using the DsRed visible marker and we verified the DsRed insertion by PCR. To remove the DsRed marker, we crossed the RspOE_DsRed line to a Cre-expressing line y[1] w[67c23]; sna[Sco]/CyO, P{w[+mC]=Crew}DH1 (BDSC #1092) following the cross scheme described in section 3.1.3 in [75]. The removal of the DsRed sequence was verified by PCR.

gBlock sequence with the left homology arm: GCACATgcatgctagcTTAATTATTAAGTACAAAAGGAAAAATTTATCTAGTTATTAATTACTTAATGTGTAATAATATATTACTATCTTACTATCCTCCGGGTAGGGAGATCGCTGGGGTGTACCCTCTTCAGCCTTCGAAGGGAGATGGGATTAACTAGGATATTTGCCAGTCGGTTCGGGTGGACCATAAGCCTGCCGTCTGCAGAAGCAACAATCACTTCAACAACGAGAAGACAGTAATCGACAGCACAGATTTCTTGATGTAAAATCGCTTCTGGAGCAACAACAACAACAATTTCAGCAATGGCAACAGCAATTTCGCTTGTGGCTTCGACAGGAAAATCAACAGCAAAACAAACATGTCAACCAGCGCTTAAAAAAGCTTGAAATATCGTATTTGAAATCGCTAATAATATAAAACAATGAGCTGGAGCTAAACACCAGTCTCATGACAAAGTTTCGGCCTTGCAATGATCCATCTTAAGGTTCTTATATGGAGTGTCAATGGCATTTCATGCAAAGCCAGAGAAATTGAGCGCTTCGCAACTGACCTTGGTGACATGGAGTTCAGTGCCATTTACTGCCCTCCAATGAACAGATTAGAAGAAAGACATTTCACTAATCTACTCCGTGCTTGCAGGCAAAGGTACTTGGTAGATAGTGACTGGAATGCGCGACACTGGCTGTGGGGAGATAGATACAACTCACCCAGGGGGCGAGAACTAGCTAAAGCCATTTCTAGCTGTGGGGCTAATATTCTTGCAGACGTGCGATTCGATCTCAAGACCCACgcggccgcgagctcCCCAGGTCAGAAGCGGTTTTCGGGAGTAGTGCCCCAACTGGGGTAACCTTTGAGTTCTCTCAGTTGGGGGCGTAGGGCCGCCGACATGACACAAGGGGTTGTGACCGGGGTGGACACGTACGCGGGTGCTTACGACCGTCAGTCGCGCGAGCGCGAcctaggTCCccgcggAGGCATTcatatgGAATTCC

gBlock sequence for the right homology arm: GGAagatctggcgcgccTCGCGCTCGCGCGACTGACGGTCGTAAGCACCCGCGTACGTGTCCACCCCGGTCACAACCCCTTGTGTCATGTCGGCGGCCCTACGCCCCCAACTGAGAGAACTCAAAGGTTACCCCAGTTGGGGCACTACTCCCGAAAACCGCTTCTGACCTGGGaggcctgtcgacggtaccATGACTTCTTTGGAACCGGGCGGCAAATCAGCTAAAAAACTTGCTAAGACTTAGAAACGAATTCTTTGAGCAAAAGTTGACGTCGCTGGACTAGACTAGTGGGCATAAGTATTTGGACAAGAAATATTTTCAATATATTTGTTTGTAACTCATTTTTTACTTAAAGCAAAATTGGTTAATTATTTTGTTGATTTAACAAAAATTACTTTCTACAAATTAGCAGTGGCATAAGTACTTGGACAAATTATTTAAATGTTATAAATTGAAAAAAAAAGTCAACTACAAAATTATGACAAATTAATAAAAAATATACACTCCCTCAGCGTCTATCACCTTTTGCAGACGTTTTGGTACAGACTGTACCAAGCTGTGGATATAATAATCTATAATCTATAATCTATTGAAATATCCTTCCATAAGGTTTGAATTTCCAAAATCGTTTATTCGCGACTTTTAGATAAGTTTGCACTCCATTTTCTTTTTAAATATGACCAGAGGTTTTCTATTATGTTGAGGTCTGGACTTTGAGGTGGCCAATCCAATGTGTTTACGCCAACATCTTTCAGAAACTTCGTAATCAGTTTGGCCTTATGACAGGGAGCATTATCCTGCTGGAGTATGAAGGACTCTCCAATCAATTTATCACCAGAGGGGAATGCATGATTATTAAGGACATCTAAATATTTTCCTTGATTCATGGTTCCATCAACAGGAACCAAATCTCCCAAGCCAGTGAAAGGATATTTTGCTGTTCACAAAGCTTCTTATAAGCAAATATTGGTTTTTAGACgctcgaggggcgccAAA

Rsp PCR product sequence: CCGAATTCAAGTACCAGACAAACAGAAGATACCTTCTACAGATTATTTAAACCTGGTACACAAAAACAATAAATTGACTAAGTTATGTCATTTTAAGCGGTCAAAATTGGTGATTTTTCGATTTCAAGTACCAGGCGAACAGAAGACACCTTCTAGAGATTCTGTTCAACTGGTAAGCAAAAACAGTAAATTGCCTAAGTTTTACATTTTAAGCGGTCAAAATGGGTGATTTTCCGATTTCAAGTACCAGACAAACAGAAGATACCTTCTACAGATTATTTAAACCTGGTACACAAAAACAATAAATTTACTAAGTTATGTCATTTTAAGCGATCAAAATGG GTGATTTTCC

Testis RNA extraction for (q)RT-PCR

10-15 pairs of testes per replicate from 3-5 day old flies were dissected in cold PBS and immediately placed in 300ul of 1X RNA Protect from the Monarch Total RNA Miniprep kit (New England Biolabs Cat # T2010S). The tissues were stored in a -80C freezer until all replicates were dissected. When RNA was to be isolated, we added proteinase K and buffer to the sample and incubated the testes at 55C for 20–30 minutes. Genomic DNA was removed using a gDNA removal column and one volume of 95% EtOH was added to the eluent (the fraction containing the RNA). The RNA was run through an RNA capture column and was subsequently treated with on-column DNAse I for 45–60 mins at RT (extended DNase treatment allows more time to degrade repetitive DNA). The RNA was then primed with RNA Prime buffer and washed twice. After the second wash, the tubes were spun for an additional 2 minutes to dry the membrane. RNA was then eluted in 40–50 ul of RNase free water.

cDNA generation for RT-PCR

200 ng of total RNA mixed with oligo(dT), satellite DNA-specific primer, and nucleotides was brought up to 13 ul with water and incubated at 65C for 5 minutes and allowed to cool on ice for at least 1 min. After cooling the RNA, we added 4 ul 5X reaction buffer, 1 ul DTT, 1 ul RNaseOUT Recombinant Ribonuclease Inhibitor (ThermoFisher Cat # 10777019), and 1 ul SuperScript III reverse transcriptase (ThermoFisher Cat # 18080093) (water was added for no RT controls) up to 20 ul total reaction volume. The reaction was then incubated at 25C for 5 min, 55C for 1 hr, and 75C for 15 min. If the sample was going to be used for RT-PCR the cDNA was frozen at -80C until use. If the cDNA was to be used for qRT-PCR it was diluted 3X by adding 40 ul RNase-free water and then stored at -80C until use.

RT-PCR

1 ul of cDNA was added to 5 ul Q5 High-Fidelity DNA Polymerase rxn buffer, 2.5 ul of dNTPs, 2.5 ul of 6 uM primers, 13.5 ul of water, and 0.5 ul of Q5 High-Fidelity DNA Polymerase (New England Biolabs Cat # M0491S) for a total reaction volume of 25 ul. The samples were then cycled between 98C, 60C, and 72C 38 times and the amplicons were run on a DNA gel. Primers for kipf: FWD: ACCAGCGAAACGGGATTATTA REV: CAAGGCGTAGACACTACTGAAC and RP49: FWD: GCTTCTGGTTTCCGGCAAGGTATGT REV: ACGTTTACAAATGTGTATTCCGACCACGTT.

qRT-PCR

4 ul of cDNA was added to 7.5 ul of PerfeCTa SYBR Green FastMix (Quantabio Cat # 95072-012), 1.5 ul 6 mM primers (Rsp For: GGAAAATCACCCATTTTGATCGC Rev: CCGAATTCAAGTACCAGAC, 359 bp. For: TACGATCTCAGCGAGGTATGA Rev: TTCCAAATTTCGGCCATCAAAT), and 2 ul water for a reaction volume of 15 ul. The reactions were then run on a BioRad C1000 thermal cycler (BioRad Cat # 1851196) equipped with a BioRad CFX96 attachment (BioRad Cat # 1845097). The program consists of a 98C melting temp, 60C annealing temp, and a 10 second extension at 72C.

qRT-PCR data were analyzed using the ΔΔCt method [76] and represented using fold change (FC) relative to a housekeeping gene (rbp4. For: CAATCCCAACTACAATCCCTATCT Rev: CTTGGGTGCCATCGGAATTA or rps3. For: AGTTGTACGCCGAGAAGGTG Rev: TGTAGCGGAGCACACCATAG).

RNA FISH

To image Rsp and 1.688 we used a custom Stellaris probe set for the former and a previously described probe labeled with Cy5, (5’-Cy5TTTTCCAAATTTCGGTCATCAAATAATCAT-3’) [77] for the latter.

Testes were dissected in cold PBS and fixed using 4% FA in PBST (0.1% Tween or Triton, either produced good results) for 30 minutes on a nutator at RT. After the fixative was removed, the tissues were washed with 1 ml PBS twice, 5 minutes each on a nutator. The tissues were prehybridized for 5 minutes using 2X SSC with 10% formamide. The probe was diluted in Stellaris hybridization buffer mix (final concentration of probes: 0.1 uM for oligo probes and 0.125 uM for stellaris probes) to the samples and placed them at 37C on a nutator overnight.

After overnight incubation we washed the tissues with 1 ml of 2X SSC with 10% formamide for 30 min twice at 37C. The tissues were then mounted using SlowFade Diamond Antifade mountant (ThermoFisher, cat # S36963). The DAPI was allowed to soak into the tissue overnight at 4C before imaging within 1 week.

K value assays

To assay the strength of segregation distortion k values were estimated from heterozygous SD/Rsp^s males (where Rsp^s are chromosomes with sensitive Rsp alleles) as described previously [78]. At least 10 replicate crosses of single 3–5-day old males with two 3–5-day old Iso-1 female virgins were prepared. Each cross was transferred to a fresh food vial after 3 days, and the parents discarded after 4 more days. We counted the progeny from each vial through day 18 after the parents were first introduced to the vial. Only replicates with more than 50 total progeny were included in our analysis. The transmission rate of the SD allele depends on not only its distortion strength, but also its viability relative to the other allele. Thus, to correct for viability effects, the transmission rate of the SD chromosome was measured in SD/Rsp^s females. The SD/Rsp^s females have a Mendelian segregation ratio with k = 0.5, and any distortion from this equal segregation can be attributed to the difference in the relative viability of the two alleles. 3 replicate crosses of five 3–5-day old females with three 3–5-day old Iso1 males were prepared, and then transferred to a fresh food vial after 4 days and kept for 4 more days. The progeny from each vial were then counted through day 18 after the parents were first introduced to the vial. The relative viability factor w = n^SD/n^Iso1 for each SD/Rsp^s genotype, where n^SD and n^Iso1 are the numbers of SD and Iso1 bearing progeny produced by the SD/Rsp^s female, respectively. This viability factor was then used to calculate a corrected strength of segregation distortion k* = N^SD/ (w * N^Gla + N^SD), where N^SD and N^Iso1 are the numbers of SD and Iso1 bearing progeny produced by the SD/Rsp^s male, respectively. The viability corrected k value (k*) was reported where we were able to estimate viability (Table A in S1 Appendix; Figs 4, S11, S12), otherwise the reported value is k (Fig 3).

Imaging

A Leica SP5 laser scanning confocal microscope was used to acquire images of our slides.

Small RNA Seq

Six- to eight-day-old testes were dissected in RNase free PBS buffer. Total RNA was extracted using mirVana miRNA Isolation Kit (Ambion) with procedures for isolating RNA fractions enriched for small RNAs (<200nt), then treated with RNase free DNase I (Promega) at 37°C for 1 hour. Library preparation and sequencing were performed by Genomics Research Center at University of Rochester. Briefly, 2S rRNA was depleted [79], small RNA library was prepared with TruSeq Small RNA Library Prep Kit (Illumina) and sequenced by Illumina platform HiSeq2500 Single-end 50 bp.

Total RNA-seq

Six- to eight-day-old testes were dissected in RNase free PBS buffer. Total RNA was extracted using mirVana miRNA Isolation Kit (Ambion) with procedures for isolating RNA fractions enriched for long RNAs (>200nt), then treated with RNase free DNase I (Promega) at 37°C for 1 hour. Library preparation and sequencing were performed by Genomics Research Center at University of Rochester. Briefly, rRNA was removed and total RNA library was prepared with TruSeq Stranded Total RNA Library Prep Human/Mouse/Rat (Illumina) and sequenced by Illumina platform HiSeq2500 Paired end 125 bp. Care was taken when extracting all RNA samples that all chromosomes outside of the 2^nd were controlled for to control for copy number of other repetitive elements.

Total RNA Seq analysis

Fastq files were trimmed using trimgalore with a minimum read length of 20 bp. FastQC was used to assess the quality of the remaining reads. Total RNA reads were mapped to a heterochromatin enriched genome using both Bowtie2 (for repeats) [80] and STAR (for genes) [81] and expression was calculated using HTSeq (for genes) [82] or a custom proportional counting script (for repeats; https://github.com/LarracuenteLab/SD-RNAseq). To calculate differentially expressed genes we used DESeq2 [83]. We additionally used PANGEA [43] to find enriched gene sets.

Weighted Gene Correlation Network Analyses (WGCNA) [46] was carried out using the WGCNA package in R using the 50^th percentile of variable genes and a minimum R^2 value of 0.8. We used the spearman method for calculating correlations then found the difference between the non-driving genotype (R16/Gla) and the driving genotypes (SD-Mad/Gla, SD-5/Gla). We calculated significance using a t-test with Tukey’s multiple hypothesis correction. We then performed PANGEA gene set analysis on the modules correlating to the drive and SD haplotype. Gene sets were labeled significant if their Benjamin & Hochberg false discovery rate <0.05.

Small RNA Seq analysis

SmRNA seq fastq files were trimmed for quality and length with a minimum length of 18 and a maximum length of 32 using cutadapt. Remaining reads were then mapped to a library of rRNAs and tRNAs and unmapped reads were saved to a separate file. The unmapped reads were then mapped to a heterochromatin enriched genome using Bowtie [84]. Counts for miRNAs and complex satellite repeats were calculated using a custom python script that assigned reads to features according to the proportion of the read mapping to that feature (https://github.com/LarracuenteLab/SD-RNAseq). Differential expression was then calculated using DESeq2 [83].