Caenorhabditis elegans germ granules are present in distinct configurations and assemble in a hierarchical manner

Thousands of transcripts undergo RNA surveillance to ensure proper gene expression in Caenorhabditis elegans. Transcript monitoring and silencing is performed by conserved regulatory small RNAs comprising three distinct branches: microRNAs (miRNAs), piwi-interacting RNAs (piRNAs) and small interfering RNAs (siRNAs). Each small RNA branch differs in biogenesis, protein interactors, transcript targets, and primary mechanisms of action, but the fundamental means of RNA surveillance are conserved: small RNAs, ranging in size from 18 to 30 nucleotides, are bound by Argonaute proteins and together target fully or partially complementary RNAs to direct RNA silencing through transcriptional or post-transcriptional mechanisms (reviewed by Ketting et al., 2021).

Small RNA-directed silencing is efficient. In C. elegans, exogenously introduced double-stranded RNAs are routed through the siRNA pathway and can silence transcripts to biochemically undetectable levels (Fire et al., 1998). A reduction in transcript levels can be detected by 4 h and silencing outcomes can be phenotypically observed by 24 h (Kamath et al., 2000; Ouyang and Seydoux, 2022). Silencing signals directed at germline-expressed genes can also be inherited through multiple generations, mediated by siRNA amplification, maternal and paternal small RNA inheritance, and small RNA-directed accumulation of repressive chromatin marks on targeted loci (Alcazar et al., 2008; Ashe et al., 2012; Burkhart et al., 2011; Grishok et al., 2000; Shirayama et al., 2012). RNA silencing also functions to ensure proper gene expression and protect against deleterious transcripts. In brief, (1) miRNAs regulate genes involved in cell patterning, development and lifespan (reviewed by Ambros and Ruvkun, 2018), (2) piRNAs engage most germline genes and can target transposons, including the Tc3 transposon (Bagijn et al., 2012; Batista et al., 2008; Das et al., 2008; Lee et al., 2012; Shen et al., 2018), and (3) endogenous siRNAs target additional transposons, gene duplications, pseudogenes, germline genes, and repetitive elements (Fischer et al., 2011; Gu et al., 2009). ‘Primary’ small RNAs, including piRNAs and a subset of siRNAs, feed into a downstream amplification step that produces ‘secondary’ siRNAs. The amplification of primary small RNAs to secondary siRNAs creates a robust, heritable, efficient silencing signal (Aoki et al., 2007; Grishok et al., 2000; Pak and Fire, 2007; Vasale et al., 2010). Together, small RNA pathways accomplish widespread regulatory roles and are required for proper development and fertility.

In the germline, the subcellular organization of small RNA pathways occurs within perinuclear germ granules. At least four distinct compartments, collectively referred to here as nuage, facilitate efficient RNA surveillance, coordinate small RNA amplification, and ensure small RNA inheritance (Manage et al., 2020; Phillips et al., 2012; Pitt et al., 2000; Sheth et al., 2010; Wan et al., 2018). P granules, the first discovered C. elegans nuage compartment, contain both nascent transcripts and key small RNA pathway components, including the Argonautes PRG-1, ALG-3, ALG-5, CSR-1 and WAGO-1 (Batista et al., 2008; Brown et al., 2017; Claycomb et al., 2009; Conine et al., 2010; Gu et al., 2009; Sheth et al., 2010). P granules also contain other small RNA-associated factors, including the Dicer-related helicase DRH-3 and the RNA-dependent RNA polymerase (RdRP) EGO-1, which act together to produce a subset of secondary siRNAs (Claycomb et al., 2009; Gu et al., 2009). Also localized to P granules are piRNA-induced silencing-defective (PID-1) and the RNase PARN-1, which are required for the synthesis and processing of piRNAs (Cordeiro Rodrigues et al., 2019; de Albuquerque et al., 2014; Tang et al., 2016). Further, most nuclear pores (75%) associate with P granules, indicating that the majority of nascent transcripts enter P granules upon nuclear export (Pitt et al., 2000). The colocalization of key small RNA components and a direct association with newly exported transcripts place P granules as a central compartment for RNA surveillance.

Adjacent to P granules are Mutator foci, which are considered hubs of small RNA amplification. Mutator foci are nucleated by the intrinsically disordered protein MUT-16, which recruits key secondary siRNA synthesis proteins, including the RdRP RRF-1 and all Mutator complex proteins (Phillips et al., 2012; Uebel et al., 2018). Recent work proposes that a Mutator complex protein, MUT-2 (also known as RDE-3), a nucleotidyltransferase, marks transcripts for secondary siRNA synthesis by the addition of 3′ poly-UG (pUG) repeats (Shukla et al., 2020). Loss of Mutator foci results in loss of many secondary siRNAs and an inability to respond to exogenous RNA interference (RNAi) (Ketting et al., 1999; Phillips et al., 2012; Zhang et al., 2011). Therefore, Mutator foci represent a compartment of nuage in which siRNA amplification is organized.

A third nuage compartment, Z granules, are found between P granules and Mutator foci (Wan et al., 2018). Z granules are named after the first identified component, a conserved RNA helicase, ZNFX-1, which interacts with a number of small RNA proteins including the RdRP EGO-1, and the Argonaute proteins CSR-1, WAGO-1 and PRG-1 (Ishidate et al., 2018). Notably, Z granules also contain WAGO-4, a secondary Argonaute required for transmission of secondary siRNAs to progeny; znfx-1 mutants respond normally to RNAi, but are unable to pass the silencing signal on to progeny (Wan et al., 2018; Xu et al., 2018). Additionally, ZNFX-1, in coordination with other factors such as PID-2 (also known as ZSP-1) and LOTR-1, appears to balance the production of some secondary siRNAs across transcripts (Ishidate et al., 2018; Marnik et al., 2022; Placentino et al., 2021). Z granules, therefore, organize the transgenerational inheritance of small RNAs and balance the distribution of secondary siRNA synthesis across targeted transcripts.

The most recently discovered nuage compartment involved in siRNA pathways are SIMR foci, of which only three associated proteins are currently known: SIMR-1, an extended Tudor domain protein; HPO-40, a SIMR-1 paralog; and RSD-2 (Manage et al., 2020). SIMR-1 was first identified in a MUT-16 immunoprecipitation, but was found to create spatially separate foci adjacent to Mutator foci. The name SIMR-1 is derived from siRNA-defective and mortal germline, as simr-1 mutants lose some Mutator-dependent siRNAs and become sterile over multiple generations at elevated temperature (Manage et al., 2020). Interestingly, simr-1 mutants de-silence a piRNA sensor and lose small RNAs specifically mapping to piRNA targets (Manage et al., 2020). This mutant phenotype, along with a conserved role of Tudor domain proteins in both piRNA pathways and in protein–protein interactions (Pek et al., 2012), leads to the hypothesis that SIMR foci act in part to direct piRNA target genes to Mutator foci for downstream siRNA amplification (Manage et al., 2020). Furthermore, RSD-2 acts in the exogenous siRNA pathway to ensure efficient RNAi, suggesting that SIMR foci are a compartment of nuage that more generally mediates the transition between primary and secondary small RNA pathways (Han et al., 2008; Manage et al., 2020; Sakaguchi et al., 2014; Zhang et al., 2012).

All four of these compartments are not bound by lipid membranes. Specifically, P granules, Z granules and Mutator foci have been shown to form via biomolecular phase separation, a process by which proteins and RNA de-mix from the surrounding bulk phase to form distinct, liquid-like condensates held together by transient, weak interactions and/or multivalent interactions. P granules exhibit controlled dissolution and condensation in embryos, rapid intramolecular rearrangement, and the ability to drip and fuse off nuclei when a shearing force is applied (Brangwynne et al., 2009). P granules also dissolve during heat stress and in 1,6-hexanediol, an aliphatic alcohol that disrupts the weak hydrophobic interactions that promote phase separation (Putnam et al., 2019; Updike et al., 2011). Z granules are colocalized with P granules until later in development, when they de-mix to form separate, adjacent foci, which also exhibit rapid molecular rearrangement (Wan et al., 2018). Z granule viscosity is regulated by PID-2 (Placentino et al., 2021; Wan et al., 2021). Finally, Mutator foci are concentration dependent, dissolve during heat stress and 1,6-hexanediol treatment, and are able undergo rapid molecular exchange with the surrounding bulk phase (Uebel et al., 2018). Thus, at least four distinct perinuclear nuage compartments, multiple of which are phase separated, are involved in RNA surveillance.

The initial characterization of these compartments demonstrates individual contributions to small RNA pathway organization, but the physical interaction and spatial configuration of all four compartments has only been briefly explored. It is also unclear how biomolecule exchange of RNAs, proteins or small RNAs occurs between compartment boundaries. The current incomplete understanding of the C. elegans nuage assemblage limits formation of a comprehensive model to describe how multiple phase-separated compartments organize small RNA pathways and facilitate RNA silencing. Here, we demonstrate that individual components of the Mutator foci and P granules can maintain separation in an ectopic environment and that interaction between germ granule compartments is dynamic and able to be re-established after perturbation. We use 3D-structured illumination microscopy (SIM) to visualize the spatial organization of Mutator foci, P granules, Z granules and SIMR foci. We discover a previously uncharacterized toroidal morphology of P granules, which we term ‘P granule pockets’, that interacts with all currently known compartments of nuage. Moreover, we quantify the relationships between germ granule compartments to reveal discrete populations of nuage that appear to assemble in a hierarchical manner. Finally, our quantification of RNA interaction with germ granule compartments demonstrates that nuage populations associated with Mutator foci are more likely to interact with RNAi-targeted RNAs. Ultimately, our data constructs a more cohesive model of how multiple phase-separated condensates organize small RNA pathways.

In the endogenous germline environment, Mutator foci and P granules exist as separate, yet adjacent, phase-separated perinuclear condensates (Phillips et al., 2012). We were interested in determining how these two compartments with liquid-like properties exist adjacently at the nuclear periphery without co-mixing. We first visualized the juxtaposition between Mutator foci and P granules in the germline at widefield resolution using endogenously expressed MUT-16::mCherry and PGL-1::GFP, respectively (Fig. 1A). To determine whether individual proteins within P granules or Mutator foci are able to form separate, unmixed compartments, we overexpressed MUT-16 and PGL-1 in C. elegans muscle tissue using the myo-3 promoter (Fig. 1B). MUT-16 is required for Mutator foci formation and has been previously shown to form ectopic foci in the cytoplasm of muscle cells when overexpressed, and PGL-1 is a major constituent of P granules that forms ectopic foci when overexpressed in intestinal tissue (Uebel et al., 2018; Updike et al., 2011). In previous work, we determined that myo-3-driven overexpression of either mCherry or GFP alone did not form condensates; thus, overexpression of MUT-16 and PGL-1 is driving ectopic condensate formation (Uebel et al., 2018). Similar to P granule and Mutator foci interaction at the germline nuclear periphery, the PGL-1::GFP and MUT-16::mCherry condensates maintained their separate and adjacent relationship in the cytoplasm of the ectopic muscle environment (Fig. 1B). To ensure that the ectopic interaction was not tag specific, we also overexpressed PGL-1::mKate2 and MUT-16::GFP in muscle cells and obtained the same result (Fig. S1A). Thus, the separate and adjacent relationship between PGL-1 and MUT-16 does not require germline cellular environment or association with nuclear pores, but instead may rely on the intermolecular interactions or phase-separation properties of key proteins in each condensate.

To test whether the molecular interactions contributing to phase separation differ significantly between Mutator foci and P granules, we exposed mut-16::mCherry; pgl-1::gfp gonads to 1,6-hexanediol, an aliphatic alcohol proposed to disrupt weak, hydrophobic interactions and dissolve phase-separated condensates (Kroschwald et al., 2015). Note that, in previous work, we have demonstrated that MUT-16 is required for recruitment of all known Mutator proteins to Mutator foci (Phillips et al., 2012; Uebel et al., 2018); thus, loss of MUT-16::mCherry foci in this experiment is indicative of loss of the entire Mutator compartment. In contrast, PGL-1 can be lost from P granules without dissolution of the entire P granule compartment (Kawasaki et al., 1998); therefore, PGL-1::GFP is only reporting on the presence or absence of PGL-1 in P granules. In a dissection in control buffer, PGL-1 foci and MUT-16 foci were clearly visible (Fig. 1Ci). Both PGL-1 and MUT-16 foci dissolved in 10% hexanediol and 5% hexanediol, as previously observed (Fig. 1Cii,Ciii) (Uebel et al., 2018; Updike et al., 2011). We further diluted the concentration of hexanediol and observed that PGL-1 foci were present in both 2.5% and 1.25% hexanediol, yet MUT-16 foci remained absent (Fig. 1Civ,Cv). MUT-16 foci were present only in very dilute 0.625% hexanediol (Fig. 1Cvi). Repeating the experiment with GFP-tagged Mutator foci yielded similar results (Fig. S1B). These data indicate that MUT-16 foci are more sensitive to perturbations of hydrophobic interactions caused by 1,6-hexanediol than are PGL-1 foci. It is possible that this differential reliance on hydrophobicity for focus formation may be one factor preventing co-mixing between adjacent condensates.

Many phase-separated condensates, including Mutator foci, dissolve with the addition of heat (Nott et al., 2015; Uebel et al., 2018). We previously discovered that Mutator foci dissolved by heat stress quickly reassemble as punctate foci during room temperature recovery (Uebel et al., 2018). To determine whether Mutator foci re-establish adjacency to P granules after heat-stress dissolution, we subjected mut-16::mCherry, pgl-1::gfp animals to heat shock for 2 h at 32°C. Upon immediate imaging after removal from heat, we discovered that MUT-16 no longer formed punctate Mutator foci, but instead faintly colocalized with P granules in enlarged foci in 80% of samples (Fig. 1Dii). In previous MUT-16::GFP heat-stress experiments, we observed a similarly enlarged Mutator focus phenotype, but did not have P granule markers to compare localization (Uebel et al., 2018). At permissive temperatures, P granules create a size-exclusion barrier that excludes molecules ≥70 kD (Updike et al., 2011). It is possible that the addition of heat alters the properties of the P granule size exclusion barrier and allows for colocalization of the 149 kD MUT-16::GFP protein. The 2-h heat shock did not completely dissolve PGL-1 in the late pachytene region, as observed previously (Fritsch et al., 2021; Watkins and Schisa, 2021). After 30- and 60-min room temperature recovery post heat shock, MUT-16 was less visibly colocalized with PGL-1, and MUT-16 cytoplasmic signal remained high (Fig. 1Diii-Div). By 90- and 120-min recovery post heat shock, MUT-16 formed foci that were punctate and adjacent to P granules, appearing strikingly similar to the non-heat-shocked control (Fig. 1Di,Dv,Dvi). To expand this analysis to Z granules and SIMR foci, we applied heat stress to ZNFX-1 and SIMR-1, respectively (Fig. S1C). We found that ZNFX-1 behaves similarly to PGL-1, i.e. foci failed to fully dissolve, and that SIMR-1 foci are disrupted and reappear in a similar manner to MUT-16 foci. Interestingly, SIMR-1 foci may recover faster than MUT-16, as 30% of gonads had SIMR foci at 60 min recovery compared to no gonads with MUT-16 foci at 60 min recovery. The ability to quickly re-establish granule boundaries and interaction after disruption suggests the adjacent relationship between granules is not necessarily established at an earlier developmental time point and continually maintained, but rather that Mutator foci and SIMR foci have the ability to disassemble and reassemble next to P granules and Z granules.

To facilitate RNAi, exchange of biomolecules must occur within C. elegans nuage, but the physical interface between the different phase-separated compartments is not well understood. We sought to gain a high-resolution view of the interface between Mutator foci and P granules using 3D-SIM, which has nearly twofold higher resolution than widefield microscopy (Gustafsson et al., 2008). In the transition zone, 3D-SIM shows that Mutator foci and P granules interact adjacently, as previously observed in widefield microscopy (Fig. 2A). Unexpectedly, however, imaging the mid- and late-pachytene regions revealed a previously uncharacterized P granule morphology. In these pachytene regions, a subset of P granules appeared to surround Mutator foci in a toroidal ‘donut-like’ morphology, which we further refer to as P granule pockets (Fig. 2B,C, arrows and insets). We also observed a gap between Mutator foci and the encircling P granule, suggesting the two condensates do not directly interact in P granule pockets (Fig. 2C, insets), consistent with previous ideas of compartment interaction (Manage et al., 2020; Wan et al., 2018). After identifying P granule pockets at high resolution, we began to recognize P granule pockets in widefield microscopy. P granule pockets are visible in widefield live imaging of undissected gonads, although the gap between P granules and Mutator foci is not resolved (Fig. S2A). As P granule pockets are present in live samples, the unique morphology does not appear to be a result of gonad fixation. To test whether fluorescent protein tags influence the formation of P granule pockets, we used anti-PGL-1 to visualize untagged P granules and used anti-HA to visualize mut-16::SNAP::HA, which contains a SNAP-tag approximately 10 kD smaller than an mCherry tag. Subsequent 3D-SIM imaging revealed that antibody-visualized untagged P granules still form P granule pockets, further validating the authenticity of the P granule structures (Fig. S2B). Of note, Pitt et al. (2000) briefly describe arch-shaped P granules, which may be related to P granule pockets. Altogether, our high-resolution imaging reveals a previously uncharacterized toroidal P granule morphology, in which some P granules encircle Mutator foci in the mid- and late-pachytene regions.

From our microscopy, we observed that not all P granules are associated with a Mutator focus. We hypothesized that nuage compartments interact in distinct populations and sought to determine the stoichiometry between P granules, Z granules, Mutator foci and SIMR foci to define nuage populations. To avoid possible artifacts from non-specific antibody binding, we used native fluorescence from endogenously tagged MUT-16, ZNFX-1, PGL-1 and SIMR-1 at widefield resolution. Quantification of fluorescent signal revealed that each nucleus (n=30) in the late pachytene is associated with on average 22.4 (±3.6) P granules, 18.4 (±2.0) Z granules, 12.4 (±2.3) SIMR foci and 10.8 (±2.3) Mutator foci (Fig. 3A). P granules were roughly twice as abundant as Mutator foci (2.1-fold), and this ratio was consistent between quantifications using different fluorescent tags (Fig. S3A). Z granules were 1.7-fold more abundant than Mutator foci. SIMR foci were more similar in quantity to Mutator foci, yet still had statistically significant higher foci counts per nucleus than Mutator foci (P=0.007; Fig. 3A). Our quantifications reveal that P granules are the most abundant nuage compartment in the germline, closely followed by Z granules, with Mutator foci as the scarcest nuage compartment.

Although our quantification demonstrates the abundance of each nuage compartment in a given nucleus, it is limited in terms of directly addressing the relationships between the different compartments. To accurately assess compartment associations, we simultaneously visualized P granules, Z granules and Mutator foci and manually evaluated each compartment for proximity to the other nuage compartments. We found three main populations of P granules (n=183) (Fig. 3B-D). The first population consisted of P granules not associated with any other visible compartment (P; 22%) (Fig. 3B,C). These solitary P granules were generally smaller than other P granules (Fig. 3D, asterisks; Fig. S3B). The second population contained P granules associated only with Z granules (PZ; 35%) (Fig. 3B,C). This population ranged more broadly in size, but was generally composed of medium and large P granules (Fig. 3D, arrowheads; Fig. S3B). The third and most prevalent population, constituting 43% of assessed P granules, associated with both Z granules and Mutator foci (PZM) (Fig. 3B,C). The majority of these P granules were large and included P granule pocket conformations (Fig. 3D, arrows; Fig. S3B). All assessed Z granules and Mutator foci were adjacent to P granules, indicating there are no solitary Z granules or Mutator foci (Fig. 3C). To incorporate all known nuage compartments into our analysis, we next assessed the proximity of SIMR foci (S) to P granules and Mutator foci, and found that the majority of SIMR foci were adjacent to both compartments (PSM; 87%) (Fig. 3B, Fig. S3C). In agreement with our nuage compartment quantification, some SIMR foci were not associated with Mutator foci (PS; 12%). Only one SIMR focus did not appear to be associated with any other granule (Fig. 3B,C); however, Manage et al. (2020) found that 100% of SIMR foci are adjacent to P granules and all SIMR foci are closely associated with Z granules (100%). Lastly, we found that all Mutator foci are associated with SIMR foci (100%) (Fig. 3C, Fig. S3C). From these relationship ratios, we extrapolate that P granules associated with Mutator foci are also associated with all other known germ granules and that PZSM constitutes 43% of all P granule populations. Previous work has indicated that disruption of P granules results in disruption of both Z granules and Mutator foci, whereas loss of Mutator foci has no effect on P granules or SIMR foci (Manage et al., 2020; Phillips et al., 2012; Singh et al., 2021). These data, together with our findings that nuage compartments are found in distinct populations with reproducible stoichiometry, suggest a hierarchy of nucleation in which P granules constitute a base granule for the subsequent nucleation of Z granules, SIMR foci and Mutator foci.

Because not all P granules form P granule pockets, we quantified P granule pockets per nucleus visible at widefield resolution and discovered that nuclei in the late pachytene (n=18) have on average 3.8 (±1.2) P granule pockets (Fig. S3D). Each P granule pocket (n=39) was associated with a Mutator focus. Given that we demonstrate that 100% of Mutator foci are also associated with both SIMR foci and Z granules (Fig. 3C), we extrapolate that Z granules, SIMR foci and Mutator foci are all present within P granule pockets. To assess the physical interaction between P granule pockets and Z granules or SIMR foci, we used SIM to visualize RFP::ZNFX-1, SIMR-1::GFP and P granules. We were often able to detect a gap between SIMR-1 and P granules (Fig. 4A, insets), indicating their localization within P granule pockets is similar to Mutator foci. In contrast, we found that Z granules generally occupy the entirety of P granule pocket interiors (Fig. 4A), and can also appear to encompass Mutator foci in pocket-like formations (Fig. 4B,C, insets). This observation suggests that P granule pockets directly interact with Z granules, and that Z granules bridge the gap between P granule pockets and Mutator foci and SIMR foci, similar to previous descriptions of stacked PZM interaction, but with an exterior-to-interior organization (Manage et al., 2020; Wan et al., 2018).

To conceptualize the trajectory of RNA through nuage, we sought to determine which compartments interact with nuclear pores and, therefore, could interact with newly exported RNA. Pitt et al. (2000) report that 75% of nuclear pores are adjacent to P granules and that 96% of P granules are adjacent to nuclear pores, with the remaining 4% being very small or separated from the nucleus. We immunostained nuclear pore complexes, Z granules and Mutator foci and discovered that nuclear pore complexes create similar patterns as P granules, and, as such, some nuclear pores are arranged in a similar morphology to P granule pockets, with rafts of pores surrounding the other germ granule compartments (Fig. 4C). We observed that Z granules occupy the inner space of the nuclear pore pocket and appear to have minimal overlap with nuclear pores (Fig. 4C, insets). Mutator foci occupied the innermost position, and, similar to their position within P granule pockets, a distinct gap was seen between Mutator foci and the surrounding nuclear pore complexes (Fig. 4C, insets). Next, to determine whether P granules lie closer to nuclear pores than Z granules, we quantified the distance between nuclear pores and P granules, and between nuclear pores and Z granules (Fig. 4Di-Diii). We found that P granules are significantly closer to nuclear pores than Z granules. We therefore infer that Z granules and Mutator foci do not significantly interact with nuclear pores and thus are not directly involved in the capture of newly exported RNAs. This conclusion is consistent with distance quantifications between compartments described previously (Manage et al., 2020). The consistent exterior-to-interior organization of P granule pockets, wherein P granules appear to be the predominant nuage compartment interacting with newly exported RNAs, suggests that RNA follows a distinct trajectory through the nuage compartments: P granules, to Z granules, to SIMR and Mutator foci.

We next sought to address whether nuclear pore association differs among nuage populations. By visualizing nuclear pore complexes, P granules and Z granules, we found that all perinuclear P granules, regardless of size or compartment interactions, appear to be associated with nuclear pores, although some nuclear pores could be seen that were not associated with P granules (Fig. 4D). This finding is consistent with observations by Pitt et al. (2000) who reported that 25% of nuclear pores are not associated with P granules. These data indicate that all germ granule configurations, including solitary P granules, are in contact with nuclear pores and able to capture nascent RNAs.

To investigate whether RNAs targeted by RNAi preferentially associate with germ granules of a particular composition, we performed RNAi against the germline-expressed gene mex-6 and visualized association of mex-6 RNA with P or PM granules by single-molecule fluorescence in situ hybridization (smFISH). We observed that the amount of cytoplasmic mex-6 RNA was substantially reduced in the germline of mex-6 RNAi-treated animals, compared with animals on control RNAi for 6 h (Fig. 5A,B). In contrast, a control RNA, oma-1, which is not targeted by RNAi, was visible throughout the cytoplasm of both control and mex-6 RNAi animals (Fig. 5A,B). While there was no significant enrichment of mex-6 RNA in germ granules of gonads treated with control RNAi (Fig. 5C), the mex-6 RNA remaining after mex-6 RNAi treatment appeared to be enriched in the germ granules of oocyte nuclei (Fig. 5D), as previously observed (Ouyang and Seydoux, 2022), indicating that RNA localization with nuage is RNAi-targeting dependent. Moreover, we observed that the mex-6 RNA signal often localizes between the Mutator foci and P granules (Fig. 5D). Although the resolution of our experiments was not sufficient to conclude definitively with which granule the RNA associates, these data are consistent with the association of RNA with Z granules, which localize between P granules and Mutator foci (Wan et al., 2018). In further support, Ouyang and Seydoux (2022) have shown that ZNFX-1 is essential for accumulation of RNAi-targeted RNAs in germ granules.

We next sought to determine the frequency with which mex-6 RNA associates with P or PM granules. Note that when imaging only P granules and Mutator foci, we can score ‘P’ (inclusive of P, PZ and PZS granules) or ‘PM’ (PZSM granules). We found that, of the granules scored as P (P, PZ or PZS), 54% were associated with mex-6 RNA and 46% were not associated with RNA (Fig. 5E). In contrast, the granules scored as ‘PM’ (PZSM granules) were associated with mex-6 RNA 98% of the time. These data indicate that germ granule configurations that include all currently known subcompartments associate with RNAi-targeted RNAs at a higher frequency than germ granule configurations lacking Mutator foci (P, PZ or PZS). Thus, although RNAi-targeted RNAs were not associated exclusively with one population of nuage, they also do not associate indiscriminately with the different nuage populations.

In this work, we have set out to establish a comprehensive model of how the various compartments of nuage assemble and associate with each other, nuclear pores, and RNAs. First, we demonstrate that individual components of Mutator foci and P granules can form separate but adjacent condensates in an ectopic environment, and find that nuage compartments disrupted via heat stress are able to re-establish adjacent boundaries upon return to the permissive temperature. We next examine the spatial organization and stoichiometry of germ granules, where we discover a toroidal granule morphology and identify populations of germ granules with distinct composition, suggesting that granules assemble in a hierarchical manner beginning with P granules at the nuclear pore, followed by Z granules, SIMR foci and then Mutator foci. Lastly, we demonstrate that RNAi-targeted RNAs preferentially associate with nuage populations of more complex composition, suggesting that different granule configurations may have distinct functions. Together, our data provide a more nuanced understanding of how phase-separated condensates organize small RNA pathways.

Previous to our work, C. elegans nuage had not been characterized using super-resolved microscopy techniques. Our work adds three key details to the physical organization of C. elegans nuage. First, nuclear pores do not appear to significantly colocalize with Z granules, SIMR foci or Mutator foci (Fig. 4). As previous literature demonstrates a direct interaction between P granules and nuclear pores (Pitt et al., 2000; Sheth et al., 2010), we conclude that P granules are the first nuage compartment to directly capture newly exported RNA (Fig. 6). Second, a subset of P granules exhibit a toroidal morphology that encompasses Z granules, SIMR foci and Mutator foci in a consistent exterior-to-interior organization. This organization is distinct from previous findings (Manage et al., 2020; Wan et al., 2018) in that the P granule surrounds a Z granule, which in turn encompasses both a SIMR focus and Mutator focus (Fig. 6). Third, P granule size correlates with more-complex arrangements of nuage compartments and the formation of P granule pockets (Fig. S3). Our model of the spatial configuration of nuage provides a foundation for understanding the trajectory of an RNA as it enters small RNA pathway compartments. In future studies, it will be necessary to determine whether P granule pocket morphology and organization actively promote RNA surveillance, or if it is simply a biophysical outcome of granule size and interaction.

We find that the separation of P granule and Mutator compartments does not rely on features of the germline environment, such as association with the nuclear periphery or germline mRNAs. Rather, separation of the condensate-promoting proteins PGL-1 and MUT-16 can be maintained in an exogenous environment, as demonstrated by our ectopic overexpression experiments (Fig. 1). Moreover, we find that PGL-1 and MUT-16 proteins have drastically different responses to perturbation of weak, hydrophobic interactions, suggesting that compartment separation may also be conferred by dissimilar compositions of the multivalent interactions that promote phase separation, such as differences in amino acid contents of intrinsically disordered regions. Finally, we show that, following disruption, Mutator foci and SIMR foci are able to re-form and restore adjacency to P granules. This has exciting implications for the dynamic nature of nuage compartments: Mutator foci and SIMR foci may disassemble or reassemble adjacent to P granules, possibly in reaction to changing RNA and/or small RNA populations. Our data on granule recovery after disruption are also consistent with our proposed hierarchy of granule assembly (Fig. 6). These findings begin to illuminate mechanisms that may maintain and promote multiple phase-separated structures as distinct compartments within nuage.

Our work reveals discrete populations of perinuclear nuage (P, PZ, PZS, PZSM) that appear to assemble in a hierarchical manner, in that the association of Z granules with P granules is required for the nucleation of SIMR foci and then Mutator foci (Fig. 3). These results are supported by prior work demonstrating that disruption of essential P granule components leads to disruption of Z granules and Mutator foci (Ishidate et al., 2018; Singh et al., 2021). Thus far, we cannot assign distinct functional roles for the different granule configurations in RNA surveillance, but our evidence that RNAi-targeted RNAs associate at a higher frequency with germ granules containing Mutator foci compared with germ granules lacking Mutator foci is encouraging (Fig. 5). However, it is also possible that germ granule proteins and associated RNAs can freely exchange between germ granules via the cytoplasm. Thus, while germ granules may provide a hotspot for screening RNAs as they are exported from the nucleus, RNA surveillance need not be limited to granules with particular biochemical activities, such as siRNA synthesis, or, for that matter, to granules at all.

One caveat to this work is the possibility that germ granule compartments may be present below the level of detection of confocal microscopy or 3D-SIM. For example, PZS granules may contain Mutator complex proteins at visually undetectable, but functionally relevant, levels. Perhaps the compartments that we can visualize within a given germ granule are indicative of which molecular processes are occurring in that granule rather than indicating a restriction on the molecular processes that can occur there. Ishidate et al. (2018) have proposed that subdomains of nuage could be formed by synchronous molecular events occurring on such a massive scale that they become visible as distinct phase-separated domains within nuage. From this vantage point, we can speculate that the hierarchical assembly of germ granule compartments may be determined by the order of the molecular events required for RNA silencing. As we identify new RNA-silencing factors associated with nuage and re-image previously characterized proteins at higher resolution, we may find more and more domains within nuage that define distinct compartments with unique molecular functions.

Together, our work uncovers new details of the organization and compartmentalization of C. elegans nuage and advances the understanding of how RNA surveillance is organized through multiple phase-separated compartments.

Ectopic expression plasmids for myo3p::pgl-1::mKate::HA, myo-3p::mut-16::mCherry::HA and myo-3p::pgl-1::gfp::FLAG were created by PCR amplification (Table S1) and assembled into a SpeI-digested pCFJ151 (Addgene, #19330) vector using isothermal assembly (Gibson et al., 2009). DNA templates for PCR products were as follows: myo-3p from pCFJ104 (Addgene, #19328), pgl-1::GFP::FLAG from DUP75 (Andralojc et al., 2017), mut-16::mCherry::HA from USC896 (Uebel et al., 2018), pgl-1 gene from N2 genomic DNA, and let-858 3′ UTR and mKate::HA::let-858 3′UTR from pDD287 (Addgene, #70685). Correct sequences of constructed plasmids were confirmed with Sanger sequencing. myo-3p::mut-16::gfp was constructed previously (Uebel et al., 2018). We generated extra-chromosomal arrays as follows: 10 ng/µl myo-3p::mut-16::gfp, 10 ng/µl myo-3p::pgl-1::mKate2 and 70 ng/µl pBlueScript was injected into HT1593 unc-119(ed3) III and Unc-rescued animals were selected to create USC1242 (cmpEx95). 10 ng/µl myo3p::mut-16::mCherry::2xHA, 10 ng/µl myo3p::pgl-1::GFP::3xFLAG and 70 ng/µl pBlueScript was injected into HT1593 unc-119(ed3) III and Unc-rescued animals were selected to create USC1392 (cmpEx97).

For mut-16(cmp249[SNAP::HA::mut-16]), znfx-1(cmp294[GFP::3xFLAG::znfx-1]) and mut-16(cmp259[2xMYC::mTagBFP2::mut-16]), PCR repair templates for CRISPR genome editing were designed with primers (see Table S1) and injected as previously described (Paix et al., 2017). Injection mix was prepared as follows: 2.5 µg/µl Cas9 protein (IDT, 1081059), 100 ng/µl tracrRNA (IDT, 1072534), 14 ng/µl dpy-10 crRNA (IDT Alt-R system), 42 ng/µl gene-specific crRNA (IDT) were incubated at 37°C for 10 min. Following incubation, 10 pmol of PCR repair template was amplified from pSNAP-tag plasmid (Addgene, #101135), GFP::3xFLAG containing genomic DNA (strain USC717) or mTagBFP2 plasmid (Addgene #75029), and 110 ng/µl of dpy-10 ssODN repair template were added. The injection mix was centrifuged at maximum speed (21,100 g for 2 min), transferred to a fresh tube, and microinjected into N2 animals for cmp249[SNAP::HA::mut-16] and cmp294[GFP::3xFLAG::znfx-1] or into USC1229 [simr-1(cmp112[simr-1::GFP::3xFLAG]) I; znfx-1(gg634[HA::tagRFP::znfx-1])] for cmp259[2xMYC::mTagBFP2::mut-16]. F1 animals with Rol or Dpy phenotypes were singled to individual plates and screened for insertion of the tag by PCR. Transgenic animals were confirmed by Sanger sequencing.

Strains were maintained at 15°C, 20°C or room temperature (∼21°C) except during heat-stress experiments. A complete strain list is available in Table S1. All other new strains created for this study were made by mating previously generated strains. All C. elegans strains used in this manuscript will be made available upon request.

For fixed widefield, confocal and 3D-SIM fluorescent imaging, adult animals were dissected in egg buffer containing 0.1% Tween-20 and fixed for 3 min in a 1% v/v formaldehyde solution. Animals were permeabilized by freeze-cracking and further fixed in ice-cold 100% methanol for 1 min. Slide preparation was then performed as described by Phillips et al. (2009).

Antibody staining of fixed germlines was performed with 1:500 rat anti-HA 3F10 (Roche, 11867423001), 1:500 rabbit anti-GFP (Thermo Fisher Scientific, A-11122), 1:50 mouse IgM anti-PGL-1 (Developmental Studies Hybridoma Bank, K76), 1:2000 mouse IgG anti-FLAG M2 (Sigma-Aldrich, F-1804) and 1:5000 mouse IgG anti-Nup107 (Covance, mAb414). Fluorescent secondary antibodies used at 1:1000 were: goat anti-rabbit Alexa Fluor 488 (Thermo Fisher Scientific, A-11008), goat anti-mouse IgG Alexa Fluor 488 (Thermo Fisher Scientific, A-11029) and goat anti-rat Alexa Fluor 555 (Thermo Fisher Scientific, A-21434). Secondary antibodies used at 1:500 were: goat anti-mouse IgM Alexa Fluor 647 (Thermo Fisher Scientific, A-21238) and goat anti-mouse IgG Alexa Fluor 647 (Thermo Fisher Scientific, A21236).

Young adult animals were standardized for age by selection of the L4 larval stage on the day preceding imaging. For live imaging, undissected animals were mounted on glass slides in <1% sodium azide in M9 buffer solution to prevent movement. For dissected gonads, animals were quickly dissected in M9 with a #11 Feather Blade scalpel. Coverslip edges were sealed with clear nail polish to prevent buffer evaporation. All live imaging and some immunostaining were performed on a GE Healthcare DeltaVision Elite microscope using a 60× NA 1.42 oil-immersion objective. Unless otherwise stated, five z-stacks (0.2 µm z-step) were compiled as maximum intensity projections to create each image. Adobe Photoshop was used to pseudocolor and adjust the brightness/contrast of each image for clarity.

Confocal imaging was performed on a Leica Stellaris 5 confocal microscope using a 63× NA 1.40 oil-immersion objective and the LIGHTNING deconvolution package. z-stack spacing was system optimized and maximum intensity projections of the entire nucleus or oocyte were generated using Fiji. Adobe Photoshop was used to pseudocolor and adjust the brightness/contrast of each image for clarity.

3D-SIM was performed at the USC Core Center of Excellence in Nano Imaging using a GE Healthcare DeltaVision OMX V4 microscope with an Olympus 60× NA 1.42 oil-immersion objective. Three angles and five phases were collected for each z-stack (0.125 z-step). Laser wavelengths of 405 nm, 488 nm, 568 nm and/or 642 nm were used for excitation. Image processing was performed with Softworx software for structured illumination image reconstruction (Weiner filter=0.005) and channel alignment. Final images were created by compiling maximum intensity projections of the reconstructions in Softworx. Channels were pseudocolored and brightness/contrast was adjusted for clarity in Adobe Photoshop.

For 1,6-hexanediol experiments, young adult animals were dissected in either M9 buffer as a control, or a solution of 10%, 5%, 2.5%, 1.25% or 0.625% 1,6-hexanediol in M9 buffer. Animals were imaged within 5 min after dissection. At least four gonads were assessed for each dilution. For concentrations with variable germ granule morphology (2.5% and 1.25% 1,6-hexanediol), additional gonads were imaged (ten and nine gonads, respectively).

For heat-shock experiments, NGM plates of young adult animals were wrapped in Parafilm and placed in an incubator at 32°C for 2 h. For room-temperature recovery, plates were allowed to recover on the benchtop at room temperature (21°C) for the specified time. Images of undissected gonads were collected and imaged within ±10 min from each indicated time point. Ten gonads were assessed for each time point.

Forty-eight smFISH probes were designed for each mex-6 and oma-1 mRNAs using the Stellaris Probe Designer, and labeled with Quasar 570 and Quasar 670 dyes, respectively. smFISH was performed on dissected germlines as described previously (Ouyang and Seydoux, 2022). Briefly, worms were dissected in 1× PBS and fixed using 3.7% (v/v) EM-grade paraformaldehyde in 1× PBS for 15 min. Post-fixation, excess liquid was removed and animals were permeabilized by freeze-cracking, rinsed twice in 1× PBS with 0.1% Tween-20 (PBST) and stored in 70% ice-cold ethanol overnight. Samples were washed with Stellaris Wash Buffer A solution [2 ml Stellaris RNA FISH Wash Buffer A (Biosearch Technologies SMF-WA1-60), 7 ml nuclease-free water, and 1 ml deionized formamide] followed by 100 μl of freshly prepared Hybridization Buffer [90 μl Stellaris RNA FISH Hybridization Buffer (Biosearch Technologies, SMF-HB1-10), 10 μl deionized formamide, and 1 μl of each 12.5 μM RNA FISH probe] and incubated at 37°C overnight. The samples were briefly rinsed again in 500 μl of Stellaris Wash Buffer A mixture, then incubated in 500 μl of Stellaris Wash Buffer A mixture for 30 min at 37°C, washed with 500 μl Stellaris RNA FISH Wash Buffer B (Biosearch Technologies, SMF-WB1-20), and incubated in 500 μl Stellaris RNA FISH Wash Buffer B for 5 min at room temperature. Lastly, the samples were mounted in Vectashield Mounting Medium and sealed under a coverslip with nail polish.

Quantification of P granules, Z granules and Mutator foci for Fig. 3 was performed on strain USC1409. SIMR foci quantification was performed on strain USC1401. Animals were dissected in 1× PBS and fixed for 20 min by adding 1:1 EM-grade paraformaldehyde for a 4% final solution. Animals were permeabilized by freeze-cracking, rinsed twice in 1x PBST and stored in 70% ice-cold ethanol overnight. Gonads were mounted in NPG glycerol and sealed with nail polish. The endogenous fluorescence of each granule was used for quantification to avoid any off-target or background artifacts that could arise from immunostaining. Widefield fluorescence images were taken using a GE DeltaVision Elite imaging system and deconvolved with Softworx software. Ten complete nuclei from the late-pachytene region of three gonads were selected at random for quantification for a total n=30. TIFs for individual channels of each nucleus were converted to 8-bit files, thresholded using Fiji 3D Object Counter, and manually inspected to ensure appropriate granule coverage through all z-stacks. Granule counts were obtained using Fiji 3D Object Counter and plotted using ggPlot.

Granule adjacency quantification was performed on three randomly selected late pachytene nuclei from each of the three previously assessed germlines in both USC1401 for SIMR populations and USC1409 for P granule populations (n=9 nuclei per genotype). To aid in adjacency visualization, fluorescent signal from each channel was converted into a 3D object using the Fiji 3D-Viewer plugin. Fluorescent thresholding was the same as for granule quantification, the re-sampling factor was set to 1, and the display was set to ‘surface’. All three channels were layered into one 3D-viewer image for adjacency assessment and was cross-checked with the original TIF files. For USC1409, P granules (n=183), Z granules (n=142) and Mutator foci (n=78) were assessed. Raw quantification data showed that 142 Z granules were adjacent to P granules, and 78 Mutator foci were adjacent to both P granules and Z granules. There were no solitary Z granules or Mutator foci. Of all P granules assessed, six P granules were associated with two Z granules (PZZ, 3%); three P granules had two Z granules and one Mutator foci (PZZM, 2%); and one P granule had two Z granules and two Mutator foci (PZMZM, 1%). These outliers were grouped as PZ and PZM, respectively. For USC1401, SIMR foci (n=117) and Mutator foci (n=102) were assessed: 116 SIMR foci were adjacent to P granules, and 102 Mutator foci were adjacent to both SIMR foci and P granules. P granules were not individually counted for USC1401. P granules for both USC1401 and USC1409 were manually assessed from the TIF files for P granule pocket morphology and qualitative size.

Quantification for Fig. S3A was performed on immunostained USC1266 whole-mount germlines. Widefield images were taken using the GE DeltaVision Elite imaging system and deconvolved with Softworx software. Granules were manually scored from z-stacks containing ten late pachytene nuclei from five germlines (n=50 nuclei). Granule counts were plotted using ggPlot.

To measure the distance between nuclear pores and P granules or Z granules in strain USC1532, 40 perinuclear Z granules were selected randomly (without referencing P granule and nuclear pore channels). Line profiles were generated for each granule in a radial orientation from the center of the nucleus and line profile plots were generated using the RGB Profile Tool macro in Fiji. The peak fluorescence intensity along the line was identified for each fluorescent channel and the corresponding pixel distance along the line was used to calculate pixel distance between fluorescent peaks.

RNA adjacency quantification was performed on USC1408 using the Fiji 3D Objects counter plugin. 16-bit TIFs of maximum intensity projections for individual channels of each oocyte were thresholded in Fiji 3D Object Counter and manually inspected to ensure appropriate granule coverage. The size filter minimum was set to 10 and map display was set to ‘objects’. Object maps for each channel (P granules, Mutator foci, and RNA) were merged and manually inspected for object colocalization/overlap. Three oocytes from three different gonads were quantified (n=9 total oocytes), providing a total of 597 P granules for assessment of adjacency to Mutator foci and/or RNA. P granules were scored as solitary P granules (P=208), P granules with RNA (PR=247), P granules with Mutator foci (PM=3), or P granules with Mutator foci and RNA (PMR=139).

We thank Scott Kennedy and Heng-Chi Lee for sharing strains. 3D-SIM images presented in this article were acquired at the Core Center of Excellence in Nano Imaging at the University of Southern California with the help of Kevin Keomanee-Dizon.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202284.reviewer-comments.pdf.