CSR-1 and P granules suppress sperm-specific transcription in the C. elegans germline

Germ cells are pluripotent and immortal. These properties are maintained through extrinsic signals from the germline stem cell niche, as well as through intrinsic signals originating from both the nucleus and cytoplasm. In the cytoplasm, non-membrane-bound organelles called germ granules are found in germ cells across phyla. In some species, their presence can induce soma to develop as germline, suggesting that germ granules function to promote cellular pluripotency (Illmensee and Mahowald, 1974; Tada et al., 2012).

In C. elegans, germ granules are called P granules (Strome and Wood, 1982). In the adult germline, they act to maintain totipotency by repressing somatic cell fate via an unknown mechanism. Most P-granule-associated proteins identified to date contain RNA-binding domains or are known regulators of translation, suggesting that P granules act post-transcriptionally to maintain germline pluripotency [reviewed by Sengupta and Boag (2012)]. When P granules are impaired in the adult germline, germ cells lose pluripotency, and a fraction initiate somatic reprogramming (Updike et al., 2014). The mechanisms behind the ability of P granules to maintain pluripotency by repressing somatic fate are currently not understood. One model is that P granules create a microenvironment in which nascent transcripts can be surveyed as they exit the nucleus, and those not licensed for germline expression can be selectively silenced (Kasper et al., 2013; Sheth et al., 2010; Updike et al., 2011). These transcripts include foreign and viral mRNAs that could compromise genome stability, as well as transcripts encoding somatic differentiation factors that might be stochastically expressed in the germline. If this model is correct, silenced expression of these transcripts could be accomplished either at the level of translation or at the level of transcript accumulation. If by translation, it implies that somatic mRNAs are already present in the germline, either at low levels or through stochastic expression in a subset of germlines, and that they need not be elevated to express somatic proteins when P granules are impaired. If by transcript accumulation, somatic mRNAs will be elevated in P-granule-depleted germlines, either through increased transcription of somatic mRNAs or through increased somatic mRNA stabilization. Microarray expression profiles in mutants lacking core P-granule components (i.e. DEPS-1 and GLH-1) have shown surprisingly few changes when compared with wild-type worms, supporting a model for translational regulation as opposed to the accumulation and/or degradation of mRNA (Spike et al., 2008a,,b). However, these expression profiles have been limited to single mutants in gene families, and in fertile worms that do not completely disrupt P-granule assembly. To fully understand how P granules regulate mRNA, transcript accumulation needs to be analyzed in dissected germlines in the presence and absence of P granules.

In C. elegans, germline expression is primarily regulated post-transcriptionally through 3′ UTR binding proteins. An exception is sperm-specific transcription, which instead depends on promoter sequences (Kulkarni et al., 2012; Merritt et al., 2008; Shim, 1999). In the fourth larval stage (L4), ∼40 germ cells in each of the gonad arms of the hermaphrodite undergo spermatogenesis and develop into 160 spermatids [reviewed by Chu and Shakes (2013)]. Mature sperm are then stored in spermathecae and used to fertilize the remaining germ cells that develop into oocytes. During the process of spermatogenesis, P granules are disassembled [reviewed by Updike and Strome (2010), figure 2]. First, the P-granule nucleating factors PGL-1 and PGL-3 are cleared from P-granules so that they are not present in secondary spermatocytes (Amiri et al., 2001; Kawasaki et al., 1998,, 2004). At the same time, two endo-siRNA-binding Argonaute proteins, ALG-3 and ALG-4, become expressed and overlap with PGL-1/3 expression in primary spermatocytes, but persist in secondary spermatocytes after PGLs are cleared (Conine et al., 2013). ALG-3 and ALG-4, along with other constitutive P-granule components, including the mRNA cap-binding translation factor IFE-1 and the Vasa homologs GLH-1 and GLH-4, remain localized to P granules in secondary spermatocytes until remaining P-granule components disassemble and are deposited in residual bodies during spermatid budding (Amiri et al., 2001; Conine et al., 2010; Gruidl et al., 1996). This small time window, when PGL-1/3 begins to be cleared from P granules in primary spermatocytes until spermatid budding, coincides with the onset of transcription and subsequent translation of sperm-specific mRNAs. IFE-1, which facilitates this sperm-specific translation (Henderson et al., 2009; Kawasaki et al., 2011), physically interacts with PGL-1 (Amiri et al., 2001). This interaction is predicted to repress IFE-1 activity until PGL-1 is cleared from P granules at the onset of spermatogenesis (Amiri et al., 2001). PGL-1 clearing might also be required to initiate alg-3 and alg-4 transcription. Once expressed, ALG-3 and ALG-4 bind to endo-siRNAs (26Gs) that complement sperm-specific transcripts, but it is unclear whether these two Argonaute proteins degrade or protect their mRNA substrates (Conine et al., 2010,, 2013; Han et al., 2009; Pavelec et al., 2009). Once sperm-specific genes are transcribed and translated, their proteins are packaged into transcriptionally and translationally silent spermatids.

A core component of P granules called CSR-1 is essential for proper spermatogenesis and germline expression. CSR-1 is an endogenous siRNA-binding Argonaute protein that specifically binds tri-phosphorylated 22-basepair endo-siRNAs (22Gs) that are complementary to germline-expressed transcripts (Claycomb et al., 2009). In adult hermaphrodites, CSR-1 22Gs do not target sperm-specific transcripts, but in males they do (Conine et al., 2013). A number of recent publications have shown that CSR-1 retains an epigenetic memory of germline expression from one generation to the next, and might function to license mRNAs for germline expression (Cecere et al., 2014; Conine et al., 2013; Gu et al., 2009; Lee et al., 2012; Seth et al., 2013; Shirayama et al., 2012; Wedeles et al., 2013). This ability probably ensures that somatic and foreign transcripts (those not licensed for germline expression) are either degraded or silenced within P granules [reviewed by Kasper et al. (2013)]. RNAi depletion of csr-1 (and its co-factors ekl-1, drh-3 and ego-1) causes a very specific, large P-granule phenotype and cytoplasmic pooling of RNA (Claycomb et al., 2009; Updike and Strome, 2009; Vought et al., 2005). Together, these findings suggest that P-granule and CSR-1 functions are tightly integrated in maintaining germline integrity.

Here, we examine the expression profile of C. elegans germlines that have been depleted of P granules, and compare this with the expression profile of csr-1-depleted germlines. Our findings suggest that translational regulation, and not mRNA accumulation/degradation, is the mechanism used to silence aberrant somatic transcripts and preserve germline pluripotency. Strikingly, we also find that P granules and CSR-1 both function to suppress sperm-specific transcription and masculinization in C. elegans hermaphrodites by regulating the germline sex determination pathway.

We previously demonstrated that impairing P granules causes the loss of germ-cell pluripotency. This is accompanied by a small fraction of P-granule-depleted germlines that express muscle-specific and pan-neuronal markers in germ cells (Updike et al., 2014). One possible mechanism to explain these germ cell reprogramming events is that P granules function to repress the accumulation of somatic transcripts that become stochastically transcribed in the germline. To test this, we performed both control (empty vector RNAi) and P-granule RNAi by simultaneously targeting four core P-granule components (PGL-1, PGL-3, GLH-1 and GLH-4) as previously described (Updike et al., 2014). We then prepared total RNA from C. elegans germlines dissected from young adult hermaphrodites (animals with a vulva, one day following the fourth larval stage, see supplementary material Fig. S1). Four biological replicates for each condition, consisting of ∼500 dissected germlines each, were obtained and prepared for mRNA-seq. Using the pipeline described in the Materials and Methods, we identified transcripts in the germline corresponding to 14,390 out of 20,259 annotated coding genes (WBcel215), each represented anywhere from a normalized average of 1 to 435,500 sequences (Fig. 1A; supplementary material Table S1, column D). We compared our data with germline- and soma-enriched datasets (Reinke et al., 2004), and found that 95% of previously defined germline-enriched genes were represented by our set of 14,390, with an average of 3002 sequences per gene. We also found that 88% of previously defined soma-enriched genes were represented in our set of 14,390, with an average of 680 sequences per gene. Next, we compared our list of germline-expressed genes with compiled germline- and soma-specific datasets (Rechtsteiner et al., 2010), and found all previously defined germline-specific genes present in our dataset, with an average of 5424 sequences per gene. Ninety-four percent of soma-specific genes were also represented, with an average of 861 sequences per gene. As dissected germlines contain somatic gonad cells, they are a potential source for these soma-specific transcripts; however, single-molecule fluorescent in situ hybridization (smFISH) of two ‘soma-specific’ transcripts, unc-16 and hbl-1, exhibited foci on the periphery of pachytene nuclei (supplementary material Fig. S2). Together, these results show that many traditionally defined soma-enriched and soma-specific genes might still be transcribed in the germline, suggesting a need for germline components that actively inhibit their translation to prevent germ-to-soma reprogramming.

Following P-granule RNAi, we observed 1149 genes with a P-value <0.05 that corresponds to at least a twofold change in transcript accumulation over the control (1043 up and 106 down), and we refer to this as ‘set 1’ (Fig. 1A; supplementary material Table S1, column I). To reduce false positives, we used an adjusted P-value <0.05, corresponding to at least a 3.6-fold change, to obtain a more stringent list of 102 genes, which we refer to as ‘set 2’ (supplementary material Table S1, column J). P-granule RNAi decreased transcripts from the target genes (pgl-1, pgl-3, glh-1 and glh-4) by a fold change of 9.4, 6.3, 1.5 and 4.7, respectively, and resulted in over 80% sterility in each of the replicates. To test the hypothesis that P-granule depletion causes increased accumulation of somatic transcripts, we compared our P-granule RNAi dataset with both germline- and soma-enriched datasets defined previously (Reinke et al., 2004) (Fig. 1B). Germline-enriched genes exhibited a slight but significant increase in transcript levels, whereas soma-enriched genes showed no change. We also did not observe increased transcripts from a previously defined set of neuronal genes (Watson et al., 2008), despite our previous observation that continuing P-granule RNAi for one day longer causes ∼10% of sterile germlines to express pan-neuronal markers (Updike et al., 2014). In fact, of the 5867 genes not expressed, only 45 (0.77%) showed a significant increase in transcripts after P-granule RNAi (3.9% of set 1), which is 7.4-fold lower than expected by chance (Fig. 1A, left, P=4E-103). Likewise, of the 5867 most abundantly expressed genes, only 47 showed a significant change in transcript levels (Fig. 1A, right, P=3E-101). The remaining genes from set 1 are transcribed and accumulate at very low levels in dissected germlines (represented by an average of only 1 to 213 sequences in our mRNA-seq data; Fig. 1A, center). These ‘low germline expression’ genes show a slight increase in transcript accumulation when P granules are impaired. However, we conclude that P granules have very little effect on the mRNA accumulation of abundantly expressed germline genes or on transcripts that do not accumulate in the germline. This suggests that, if P granules repress the expression of somatic transcripts, as we hypothesize, such regulation must occur at the level of translation and not mRNA accumulation/degradation.

Gene ontology analysis of upregulated set 1 genes following P-granule RNAi shows an enrichment in terms associated with spermatogenesis. We therefore compared upregulated set 1 genes with available gamete-enriched datasets (Chu et al., 2006; Ortiz et al., 2014; Reinke et al., 2004). We found that oocyte-enriched genes are slightly downregulated, whereas sperm-enriched genes are highly upregulated following P-granule RNAi (Fig. 1B). In fact, 81% (848/1043) of upregulated P-granule RNAi set 1 genes are in at least one of the sperm-enriched datasets, and this increases to 87% (76/87) in more stringent set 2 genes (supplementary material Table S1, column X). Subtracting sperm- and oocyte-specific genes from the germline-enriched dataset demonstrated that the increased transcript accumulation of germline-enriched genes following P-granule RNAi can be attributed entirely to the presence of sperm-enriched genes (Fig. 1B; supplementary material Table S1, column R). Genes upregulated by P-granule RNAi encode several major sperm proteins (MSPs), the sperm-specific PP1 phosphatases GSP-3 and GSP-4, other sperm-specific chromatin factors, including HTAS-1, SPCH-1, SPCH-2, SPCH-3, SPE-11, SMZ-1 and SMZ-2, and the sperm-specific small RNA-binding Argonautes ALG-3 and ALG-4 (supplementary material Table S1). mRNAs from these latter two genes increase 2.2- and 2.4-fold after P-granule RNAi. Interestingly, genes regulated by P granules do not include those encoding SPE-44 and ELT-1, which have been shown to promote sperm-specific transcription (Kulkarni et al., 2012; Shim, 1999), or IFE-1, which promotes sperm-specific translation (Kawasaki et al., 2011).

The core P-granule component CSR-1 has been ascribed with the ability to distinguish between transcripts that should be expressed in the germline and those that should not (Seth et al., 2013; Wedeles et al., 2014). In addition, CSR-1 is required for proper spermatogenesis in C. elegans males (Conine et al., 2013). In light of this, we also wanted to examine germline expression profile following the loss of csr-1. Microarray and Gro-seq expression profiles have previously been performed in csr-1 whole worms (Cecere et al., 2014; Claycomb et al., 2009). To examine germline-specific expression, we performed csr-1 RNAi and dissected germlines from young adult hermaphrodites. This was also carried out with four biological replicates consisting of ∼500 dissected germlines each. Correlation between germline mRNA-seq following csr-1 RNAi and previously performed microarray tiling profiles in a csr-1 mutant (Claycomb et al., 2009) was strong between transcripts that increased more than twofold (supplementary material Fig. S3A,B). Downregulated genes on the tiling arrays showed no change in transcript accumulation with csr-1(RNAi) mRNA-seq. One notable difference, because we sequenced mRNA and not total RNA, is that we were unable to detect a decrease in histone mRNA levels following csr-1 RNAi as previously reported (Avgousti et al., 2012). Previous findings (Claycomb et al., 2009) in hermaphrodites, suggesting that the loss of CSR-1 primarily affects the accumulation of non-CSR-1 target mRNAs (i.e. mRNAs not targeted by 22G siRNAs bound to CSR-1), were supported by our results (supplementary material Fig. S3B). We observed 2327 significantly regulated genes in germlines following csr-1 RNAi, with a P-value <0.05, corresponding to a fold change >1.9 (Fig. 2A; supplementary material Table S1, column O). An adjusted P-value <0.05 was calculated to obtain a more stringent list of 949 genes, corresponding to a fold change >2.6 (supplementary material Table S1, column P). Of the significantly regulated genes, only 4% of set 1 and 5% of set 2 were previously shown to complement CSR-1-bound 22G siRNAs (Claycomb et al., 2009), a percentage fivefold (set 1) and 18-fold (set 2) lower than expected by chance. Both glh-1 and glh-4 fall into this group of CSR-1 targets that are upregulated (2.6- and 2.3-fold, respectively) with csr-1 RNAi. This result was similar to what we previously reported by qRT-PCR (Updike and Strome, 2009), suggesting that CSR-1 negatively modulates transcript levels of core P-granule components. Increased expression of glh-1 and glh-4, although modest, might contribute to the large P-granule phenotype that is observed in germlines when csr-1 is depleted.

Strikingly, both csr-1 RNAi and P-granule RNAi expression profiles showed a high degree of correlation (Fig. 3;,supplementary material Table S1). We compared csr-1(RNAi) upregulated genes with available gamete-enriched datasets (Chu et al., 2006; Ortiz et al., 2014; Reinke et al., 2004). As with P-granule RNAi, we found that oocyte-enriched genes are slightly downregulated, whereas sperm-enriched genes are highly upregulated following csr-1 RNAi (Fig. 2B). Of upregulated csr-1 RNAi set 1 genes, 59% (1044/1760) are sperm-enriched, increasing to 68% (533/781) in set 2 genes (supplementary material Table S1, column X). Increased transcript accumulation of germline-enriched genes following csr-1 RNAi can also be attributed entirely to the presence of sperm-enriched genes (Fig. 2B; supplementary material Table S1, column R). Again, these upregulated genes encode several MSPs, the sperm-specific PP1 phosphatases GSP-3 and GSP-4, other sperm-specific chromatin factors, including HTAS-1, SPCH-1, SPCH-2, SPCH-3, SPE-11, SMZ-1 and SMZ-2, and genes encoding ALG-3 and ALG-4, which are upregulated 2.9- and 3.1-fold, respectively (supplementary material Table S1). SPE-44 and IFE-1 are also not differentially expressed. These results suggest that the functions of P granules and CSR-1 are tightly integrated.

Argonaute proteins, such as CSR-1, ALG-3 and ALG-4, are directed to their mRNA substrates through sequence-specific small RNAs. CSR-1-bound 22G siRNAs selectively target germline-expressed transcripts, whereas ALG-3/4 26G siRNAs selectively target sperm-specific transcripts (Conine et al., 2010). CSR-1 has recently been shown to act in the ALG-3/4 pathway to promote male spermatogenesis (Conine et al., 2013). In the hermaphrodite germline, however, our results showed that CSR-1 negatively regulates the expression of sperm-specific genes. This might reflect distinct gender roles for CSR-1, depending on whether ALG-3 and ALG-4 are present. In an alg-4; alg-3 double mutant, ALG-3/4 mRNA targets exhibit very little change in accumulation (Conine et al., 2013). By contrast, 45% (634/1403) of ALG-3/4 targets are significantly regulated, all positively, following P-granule RNAi (set 1), and the average expression of all ALG-3/4 targets is significantly increased (Fig. 1B). Likewise, 51% (721/1405) of ALG-3/4 targets are significantly regulated, 719/721 positively, following csr-1 RNAi (set 1), and the average expression of all ALG-3/4 targets is also significantly increased (Fig. 2B). One possible model is that, instead of regulating targets with bound 22G siRNAs, CSR-1 functions in conjunction with P granules to repress alg-3/4 transcription and ALG-3/4 target gene expression in the germline of adult hermaphrodites. CSR-1 was previously found to bind 22G siRNAs corresponding to sperm-specific genes in males; however, these sperm-mRNA-targeting 22Gs are not present in adult hermaphrodites (Conine et al., 2013). Interestingly enough, we observed that male-specific CSR-1 mRNA targets are significantly upregulated following either csr-1 or P-granule RNAi in the germlines of adult hermaphrodites (Fig. 1B and Fig. 2B). Our findings suggest that the ability of CSR-1 to repress sperm-specific mRNA accumulation in adult hermaphrodites might be indirect and/or independent of the subset of CSR-1-bound 22Gs that directly bind sperm-specific transcripts in males, and further testing is warranted to determine whether that is the case.

In order to visualize, and thereby verify, the accumulation of sperm-specific mRNA following either P-granule or csr-1 RNAi, we performed FISH on msp-3 transcripts in whole-mounted worms. Under control RNAi conditions, msp-3 becomes expressed in the L4 stage in the most proximal germ cells that undergo spermatogenesis. In young adult and 3-day-old adult worms, msp-3 transcripts are only detected in a handful of residual bodies that remain after the completion of spermatogenesis (Fig. 4A; Table 1). After depleting P granules or csr-1 with RNAi, the domain of msp-3 expression expands past the bend in the gonad in young adults, and in 3-day-old adults can even reach the distal end of the germline. We found that 16% of worms show a distal expansion of msp-3 in young adults following P-granule RNAi, whereas this expansion is observed in 35% of young adults following csr-1 RNAi (Fig. 4A; Table 1). Similar distal expansions were observed for the sperm-enriched transcripts ssq-1 and rmd-3 (supplementary material Fig. S4). As was previously described, P-granule RNAi causes premature cellularization of germ cells (Updike et al., 2014). These msp-3⁺ cells do not express LIN-41::GFP, a marker of maturing oocytes, and in csr-1- and P-granule-depleted worms, the domain of LIN-41::GFP expression is retracted or absent depending on the extent of msp-3 expression (Fig. 4B). This suggests that proximal germ cells that would otherwise develop as oocytes now initiate spermatogenesis when csr-1 or P granules are impaired. However, this germline masculinization phenotype is incomplete, as these cells in young adults and 3-day-old adults do not stain with an antibody (SP56) that recognizes epitopes in mature spermatids (data not shown). The expansion of msp-3 provides convincing evidence that a primary role of P granules, possibly through CSR-1, might be to limit the extent of germ cells that initiate spermatogenesis in the developing hermaphrodite by promoting the switch from spermatogenesis to oogenesis. The presence of a vulva (supplementary material Figs S5 and S6, arrowheads), the position of the distal end of the germline, and the expansion of msp-3 in 3-day-old adults suggests that this phenotype is not due to delayed germline development. This suggests that P-granule components actively suppress germline masculinization to promote proper germ cell and oocyte development.

Next, we wanted to determine whether the expansion of msp-3 expression is caused by compromised P granules or by the loss of specific P-granule components. This is always a difficult question to resolve, as P-granule integrity can be compromised by impairing a number of individual P-granule components. Single mutations in all four of our core target genes (pgl-1, pgl-3, glh-1 and glh-4) caused P-granule defects, but not the complete removal of P granules. We found that simultaneous RNAi depletion of the redundantly functioning paralogs pgl-1 and pgl-3 (Kawasaki et al., 2004) caused msp-3 expansion in 20% of worms (supplementary material Fig. S5;Table 1). This expansion was also observed in 18% of worms only fed pgl-1 RNAi, and in 30% only fed glh-1 RNAi. We also observed an expanded msp-3 domain in 18% of glh-4(gk224) glh-1(ok439) double mutants, in 31% of pgl-1(bn101) single mutants and in 20% of pgl-3(bn104) single mutants. Proximal germlines in pgl-1(bn101); pgl-3(bn104) did not show an expansion of the msp-3 domain, but these double mutants contained very severe germline proliferation defects that might have masked this phenotype. Our results suggest that the distal expansion of sperm-specific mRNAs can be caused by compromising P-granule integrity in several genetic backgrounds. The percentage of csr-1 animals showing msp-3 expansion did not increase in a pgl-1(bn101) background, possibly due to pgl-1 mutants being partially RNAi defective (Robert et al., 2005), but increased to 65% in pgl-3(bn104) mutants (supplementary material Fig. S6;Table 1). One interpretation is that the effects of CSR-1 and PGL-3 on msp-3 expansion are additive, and that csr-1 RNAi further compromises P granules to suppress the accumulation or transcription of sperm-specific mRNAs. A second interpretation is that csr-1 is epistatic to P granules, and that partial CSR-1 activity remains when P granules are disrupted. Because csr-1 mutants are RNAi defective with severe germline defects, attempts to distinguish these two interpretations were inconclusive.

Sperm-specific mRNA expression is regulated by the germline sex-determination pathway [reviewed by Zanetti and Puoti (2013)]. This pathway converges to increase the expression of FOG-3 during the L3-L4 stages of development in the C. elegans hermaphrodite. Expression of FOG-3 initiates sperm fate specification through unknown mechanisms. Almost half of the genes known to act upstream of FOG-3 in the germline sex determination pathway co-localize with P granules (Ariz et al., 2009; Draper et al., 1996; Eckmann et al., 2002; Jones et al., 1996; Rybarska et al., 2009; Voronina et al., 2012), and include proteins that promote (GLD-1 and GLD-3) and repress (MEX-3, PUF-8, GLS-1 and FBF-2) FOG-3 expression. We found that fog-3 transcription is upregulated following P-granule RNAi (1.5-fold) and csr-1 RNAi (9.5-fold, supplementary material Table S1), suggesting that the increase in sperm-specific transcription is dependent upon FOG-3. Other P-granule components known to promote proper spermatogenesis include the mRNA cap-binding protein IFE-1 (Amiri et al., 2001; Henderson et al., 2009), and the Argonautes ALG-3 and ALG-4 (Conine et al., 2010,, 2013). To determine whether the expansion of msp-3 transcripts is dependent upon these P-granule components, P-granule, pgl-1;pgl-3 and csr-1 RNAi was performed in ife-1(bn127) mutants, alg-4(ok1041); alg-3(tm1155) double mutants, and fog-3(q470) loss-of-function heterozygous mutants (supplementary material Fig. S6;Table 1). ife-1(bn127) single mutants and alg-4(ok1041); alg-3(tm1155) double mutants did not suppress msp-3 expansion upon csr-1, P-granule and pgl-1; pgl-3 depletion. Similarly, msp-3 expansion was not suppressed in fog-3(q470/+) heterozygotes. However, msp-3 transcripts are absent in fog-3(q470) homozygotes. This shows that CSR-1 and P granules either act indirectly to limit the extent to which fog-3 is expressed, or that spermatogenesis must first initiate in order to observe the distal expansion of sperm-gene expression. The large increase in fog-3 expression that is observed after csr-1 RNAi supports the former possibility, but cannot exclude the latter.

To determine how fog-3 is regulated, P-granule and csr-1 germline mRNA-seq data were examined for expression changes in genes known to act upstream in the germline sex determination pathway. We found that transcription of her-1 is increased 1.8- and 6.6-fold following P-granule and csr-1 RNAi, respectively. Hermaphrodites do not normally express her-1 mRNA. Instead, during the short window of spermatogenesis, hermaphrodites promote FOG-3 activity through other germline sex determination components. Therefore, one possibility is that P granules and CSR-1 negatively regulate fog-3 expression by suppressing her-1 transcript accumulation in the hermaphrodite germline.

We initially set out to test whether P granules maintain germline pluripotency by antagonizing the transcription or accumulation of somatic mRNAs. Instead, we discovered that most transcripts previously thought to be soma specific might already be expressed at low levels in the germline, suggesting that P granules inhibit translation and not the accumulation of these somatic mRNAs, and follow-up studies will need to determine whether this is the case. We also found that P granules and CSR-1 suppress the accumulation of sperm-specific transcripts in the germline of adult hermaphrodites. This does not come as a complete surprise, as the role of both germ granules and endogenous siRNA pathways in spermatogenesis is well documented in a number of species [reviewed by Yadav and Kotaja (2014)]. In C. elegans, the RNA-directed RNA polymerase EGO-1 functions to produce the 22G siRNAs bound by CSR-1, and ego-1 mutants prolong spermatogenesis and produce small ‘intersexual’ gametes before making recognizable oocytes (Smardon et al., 2000). Because these intersexual gametes are not produced in ego-1 fog-3 germlines, it was proposed that EGO-1 in some way promotes the sperm-oocyte switch. Sperm genes are also overexpressed in eri-1 and rrf-3 mutants (Asikainen et al., 2007), both of which encode Dicer complex components that produce 26G endo-siRNAs complementing spermatogenesis genes (Gent et al., 2009; Han et al., 2009; Pavelec et al., 2009). It has been proposed that spermatogenesis defects observed when impairing ERI-1, RRF-3 or the sperm-specific 26G-binding proteins ALG-3 and ALG-4 can be attributed to either under- or overexpression of sperm mRNAs (Conine et al., 2010; Gent et al., 2009; Han et al., 2009; Pavelec et al., 2009). This apparent discrepancy might have a simple resolution if an endogenous siRNA pathway represses sperm-specific transcription in adult hermaphrodites that have already undergone spermatogenesis, but promotes sperm-specific transcription in males and during the short window of spermatogenesis in hermaphrodites.

Our model is that both P granules and CSR-1 function to regulate the switch between sperm-specific activation and repression. Spermatogenesis occurs during the L4 stage of hermaphrodite development, but sperm transcription must be suppressed in the first wave of germ cells that develop into oocytes (Fig. 5A). A clear boundary must then be maintained between male and female germ cells until spermatogenesis is complete. Defects observed in proximal germ cells following csr-1 or P-granule RNAi were suppressed by inhibiting spermatogenesis, suggesting that germ cell crowding and premature cellularization in these worms are the result of partial masculinization. Our data suggest that P-granule components such as PGL-1 and PGL-3 act to define the boundary between spermatogenesis and oogenesis, and function with CSR-1 to suppress sperm-specific transcription. CSR-1 function might require perinuclear localization to interact with nascent mRNAs exiting P granules, and, in turn, P-granule integrity might depend on the ability of CSR-1 to escort highly expressed target mRNAs through and away from P granules. One model is that CSR-1-bound 22G RNAs simply correspond to germline transcripts that are most abundantly expressed. In complying with this model, when spermatogenesis is initiated and ALG-3 and ALG-4 are expressed, 26G RNAs that correspond to sperm genes are processed into CSR-1-bound 22G RNAs, which are then used to chaperone newly abundant sperm transcripts through P granules. In adult hermaphrodite germlines, the accumulation of CSR-1-targeted mRNAs does not change following csr-1 RNAi, supporting a more passive model in which CSR-1 primarily functions to chaperone its target mRNAs instead of regulating their accumulation.

If this passive model for CSR-1 and its bound 22Gs is correct, how do P granules and CSR-1 first promote the switch from spermatogenesis to oogenesis in the hermaphrodite? Our results suggest that they do so by regulating the germline sex determination pathway (Fig. 5B). Transcripts encoding FOG-3 were significantly upregulated when CSR-1 or P granules were impaired, and feminized fog-3 and fem-3 loss-of-function animals fully suppressed sperm-specific transcription. Similar results were observed previously with the loss of eri-1. Whereas eri-1 mutants exhibit increased sperm transcription in the germline, this is suppressed in fem-3 worms (Pavelec et al., 2009). Whereas our results suggest that CSR-1 and P granules repress her-1 in the germline, there are additional possibilities that could explain how CSR-1 and P granules regulate the germline sex determination pathway. For example, IFET-1 is an eIF4E-transporter in C. elegans that recruits a number of translational repressors to P granules, and impairing its function is sufficient to promote masculinization in the hermaphrodite germline (Sengupta et al., 2013). IFET-1 activity might require intact P granules. Another study shows that the germline sex determination pathway component, FBF-2, requires PGL-1 and P-granule recruitment to bind to and silence the translation of its mRNA targets in the more distal germline (Voronina et al., 2012). Both fem-3 and fog-3 are known FBF-2 targets [reviewed by Zanetti and Puoti (2013)]. A final study shows that the germline sex determination component FBF-1 forms a complex with EFT-3/eEF1A and CSR-1 in C. elegans lysates (Friend et al., 2012). This study also shows that CSR-1 represses the accumulation FBF targets gld-1 and cye-1, whereas the CSR-1 cofactors DRH-3, EGO-1 and EKL-1 do not. Therefore, the role of CSR-1 in regulating the germline sex determination pathway might be independent from the sequence of bound 22G siRNAs, instead acting through a conserved PUF-Ago-eEF1A complex that attenuates translational elongation of its mRNA targets. Although it is clear that P granules and CSR-1 regulate the germline sex determination pathway, the specifics of this regulation still need to be resolved.

C. elegans strains were maintained as per standard protocols (Brenner, 1974). The following strains were used and can be obtained through the CGC: N2(Bristol), SS579 pgl-1(bn101) IV, SS608 pgl-3(bn104) V, KB4 glh-4(gk255) glh-1(ok439) I/hT2[bli-4(e937) let-?(q782) qIs48] I;III, SS615 pgl-1(bn101) unc-24(e138) IV; pgl-3(bn104) dpy-11(e224) V, SS712 ife-1(bn127) III, WM300 alg-4(ok1041) III; alg-3(tm1155) IV, JK1403 fog-3(q470)/unc-13(e1091) lin-11(n566) I, CB3844 fem-3(e2006) IV and TH206 [pgl-1p::PGL-1::TY1::EGFP::3xFLAG(92C12)::pgl-1 3′UTR + Cbr-unc-119(+)] I. DG3913 lin-41(tn1541[GFP::tev::s::LIN-41]) I was kindly provided by David Greenstein (Spike et al., 2014).

P-granule RNAi (simultaneously targeting pgl-1, pgl-3, glh-1 and glh-4) has been described previously (Updike et al., 2014). The pgl-1;pgl-3 RNAi construct (pDU51) contains 320- and 349-base pair fragments from pgl-1(1374-1666) and pgl-3(265-613), respectively, cloned into L4440 and transformed into HT115 bacteria. csr-1 and pgl-1 RNAi was obtained from the Ahringer Library (Kamath et al., 2003). The glh-1 RNAi construct was described in Spike et al. (2008b). Control RNAi contained the L4440 plasmid in HT115 bacteria. RNAi was performed at 24°C for mRNA-seq experiments. RNAi feeding for FISH experiments was started on L1 P0 animals at 20°C; their F1 larval progeny were raised at 24°C and maintained through the F2 generation before being imaged as young F2 adults. CB3844 animals were fed RNAi at 15°C (P0, F1 generations) and their F2 larval progeny were shifted to 24°C.

RNAi feeding (empty vector, P-granule or csr-1) was started in L1 staged worms at 24°C; and germlines were dissected from their F1 progeny. Anti-PGL-1 antibody staining was performed in parallel to verify RNAi knockdown. Total RNA was collected from young adult hermaphrodite germlines (those with a vulva, one day past the L4 stage). Mouth-pipetting with a pulled capillary needle was used to isolate the gonad at the dorsal-to-ventral bend (supplementary material Fig. S1). Germlines were dissected in egg buffer (25 mM HEPES, 120 mM NaCl, 2 mM MgCl₂, 2 mM CaCl₂, 50 mM KCl, 0.5% Tween-20 and 1 mM levamisole) and immediately dispensed into Trizol (Life Technologies) on ice, and a total of 400-500 dissected germlines were collected for each replicate. RNA isolation proceeded as previously described (Gaydos et al., 2012), resulting in total RNA ranging from 1.5 to 4 μg per replicate sample, which was sufficient to sequence without linear amplification.

Four biological replicates of each condition (empty vector control, P-granule RNAi and CSR-1 RNAi germlines) were collected for total RNA. Using a TruSeq RNAv2 kit (Illumina), mRNA was polyA-selected, fragmented and used to create adapter-ligated cDNA for sequencing. One hundred base pair sequences were obtained on an Illumina HiSeq2500 and processed as single-end reads using the following pipeline: TopHat2 (v.2.0.8b) was run as described in Kim et al. (2013), using the WBcel215 assembly and a provided transcript annotation file (Ensembl release 70) retrieved from Illumina iGenomes (http://support.illumina.com/sequencing/sequencing_software/igenome.html) (Kersey et al., 2014). HTSeq (v.0.5.4p3) was then used to generate read counts per gene per sample and those counts were normalized across samples with DESeq (v.1.1.2.1) (Anders and Huber, 2010; Anders et al., 2015). Default parameters were used in most instances, with the following exceptions: -G, -M arguments were used in the TopHat2 protocol; -m union was specified in the htseq-count command; method=“blind” was used for dispersion estimation in DESeq. Fold change between the average of the four control replicates and the average of the four test replicates was used to calculate the significance using a negative binomial distribution. An adjusted P-value was calculated using the Benjamini–Hochberg method for multiple testing correction (Anders and Huber, 2010). Gene ontology was evaluated via Gorilla (Eden et al., 2009). The UCSC Genome Browser (Kent et al., 2002) was used to visualize aligned reads. P-granule RNAi dataset: Set 1 transcripts were defined as those with a P-value <0.05 and an expression change of at least >twofold over control. Set 2 transcripts had an adjusted P-value <0.05 and an expression change >3.6-fold over control. csr-1 RNAi dataset: Set 1 transcripts were defined as those with a P-value <0.05 and an expression change of >1.9-fold over control. Set 2 transcripts were defined as those with an adjusted P-value <0.05 and an expression change of >2.6-fold over control. mRNA-seq data have been deposited at the GEO database under the accession number GSE67954.

SP56 (Ward et al., 1986) staining was performed on dissected germlines as previously described (Updike et al., 2014).

Stellaris FISH probes (Biosearch Technologies) were designed for msp-3, ssq-1, rmd-3, unc-16 and hbl-1 with CAL Fluor Red 610. Whole-worm FISH for msp-3, ssq-1 and rmd-3 was performed as described by Ji and van Oudenaarden (2012) with the following modifications: worms were incubated in fix for 30 min, not 45, and kept rotating in ethanol at 4°C for 48 h, not 24. After the ethanol incubation, worms were resuspended in 1 ml wash buffer for 8 min. Worms were hybridized using ∼2.5 μl probe (at 25 μM stock oligo concentration) per worm plate washed, and incubated at 37°C for 4 h. The animals were dropped on GCP (0.2% w/v gelatin, 0.02% w/v chrome alum, 0.05% w/v poly-L-lysine)-coated slides, placed under a coverslip with 2× SSC and imaged. unc-16 and hbl-1 FISH was performed on dissected germlines as previously described (Updike et al., 2014). Images were acquired and deconvolved using Leica AF6000 acquisition software on an inverted microscope (Leica DMI6000B) with a cooled CCD camera (Leica DFC365FX) and a Leica 40× air objective with a 0.5 μm z-stack. Expanded msp-3, ssq-1 and rmd-3 expression was identified as expression expanding distally from the proximal germline past the bend region of the gonad.

We would like to thank Ben King at the MDI Biological Laboratory Comparative Functional Genomics Core for assistance with data analysis, the Genome Technologies division at the Jackson Laboratory which performed the mRNA-seq experiments, David Greenstein and Caroline Spike (University of Minnesota, MN, USA), and Susan Strome (University of California Santa Cruz, CA, USA) for reagents and discussions, and the C. elegans Genetics Center NIH-ORIP [P40 OD010440] for strains.