Membrane-associated cytoplasmic granules carrying the Argonaute protein WAGO-3 enable paternal epigenetic inheritance in Caenorhabditis elegans

Epigenetic inheritance describes the transmission of gene regulatory information across generations without altering DNA sequences, enabling offspring to adapt to environmental conditions. Small RNAs have been implicated in this, through both the oocyte and the sperm. However, as much of the cellular content is extruded during spermatogenesis, it is unclear whether cytoplasmic small RNAs can contribute to epigenetic inheritance through sperm. Here we identify a sperm-specific germ granule, termed the paternal epigenetic inheritance (PEI) granule, that mediates paternal epigenetic inheritance by retaining the cytoplasmic Argonaute protein WAGO-3 during spermatogenesis in Caenorhabditis elegans. We identify the PEI granule proteins PEI-1 and PEI-2, which have distinct functions in this process: granule formation, Argonaute selectivity and subcellular localization. We show that PEI granule segregation is coupled to the transport of sperm-specific secretory vesicles through PEI-2 in an S-palmitoylation-dependent manner. PEI-like proteins are found in humans, suggesting that the identified mechanism may be conserved. Schreier et al. report a previously undescribed cytoplasmic condensate, termed the paternal epigenetic inheritance (PEI) granule, that contains the Argonaute protein WAGO-3 and 22G small RNAs and mediates paternal epigenetic inheritance in C. elegans.


WAGO-3 is guided into sperm by PEI granules.
Confocal microscopy analysis of GFP::3×FLAG::WAGO-3 (hereafter, WAGO-3) revealed expression throughout the germline at all stages, with localization to P granules in mitotic, meiotic and primordial germ cells (Extended Data Fig. 4a,b). Notably, we found strong WAGO-3 signals within the sperm-containing spermatheca (Extended Data Fig. 4a). The presence of WAGO-3 in sperm was confirmed by analysing isolated, male-derived germ cells at different stages of spermatogenesis, which also revealed a punctate subcellular localization ( Fig. 1a and Extended Data Fig. 4c). We next performed IP analysis of WAGO-3 from late fourth larval stage (L4) hermaphrodites, a stage during which spermatogenesis is ongoing, followed by label-free quantitative mass spectrometry (IP-MS/MS). Besides known P granule components such as DEPS-1, PRG-1 and WAGO-1 (ref. 38 ; Fig. 1b), we identified F27C8.5 (PEI-1). We confirmed the interaction (Extended Data Fig. 4d), and found that PEI-1::mTagRFP-T (hereafter, PEI-1) was exclusively expressed during the later stages of spermatogenesis, both in L4 hermaphrodites (Extended Data Fig. 4e) and in males (Extended Data Fig. 4f).
WAGO-3 still co-localized with PGL-1 in pei-1 mutants, but was absent from spermatozoa and instead was found in the residual body ( Fig. 1g-i). Thereby, WAGO-3 followed the same fate as the Argonaute proteins WAGO-1, ALG-3 and CSR-1 in wild-type animals (Extended Data Fig. 4h-j,l-n).
We conclude that PEI-1 defines a spermatogenesis-specific germ granule-the PEI granule-that recruits WAGO-3 and enables its segregation into mature sperm.

Paternal epigenetic inheritance requires WAGO-3 and PEI-1.
The Mis phenotype 8,9 (described above) enables us to examine the relevance and mechanisms of epigenetic inheritance. The precise set-up that we used in the experiment is shown in Fig. 2a. Note that the mut-7 and mut-16 mutants can be used interchangeably in both sexes, as they both result in Mutator system dysfunction 8,9,14 . This set-up generates embryos that can make Mutator 22G-RNAs. Depending on the specific cross, the mother, the father or neither parent can make Mutator 22G-RNAs. All strains carry a prg-1 mutation to remove the partially redundant activity of inherited piRNAs in this system 8,9 . Using this set-up, we found that maternal or paternal 22G-RNAs were sufficient to prevent the Mis phenotype (Fig. 2b,c (top three bars)), enabling us to dissect male-and The blue and green data points represent above and below the threshold, respectively. WAGO-3 and PEI-1 are highlighted by red data points. c, Confocal micrograph showing spermatogenesis of late-L4 stage hermaphrodites expressing the indicated proteins. PGL-1::mTagRFP-T was used as a P granule marker. Germ cell development progresses from left to right. The areas indicated by dashed yellow boxes (i and ii) are magnified on the right. ROI1: PGL-1::mTagRFP-T and GFP::3xFLAG::WAGO-3 co-localize; ROI2: GFP::3xFLAG::WAGO-3 leaves P granules; ROI3: PGL-1::mTagRFP-T signal is not detectable anymore. rb, residual body; sc, spermatocyte; st, spermatid. Scale bars, 10 μm (proximal gonad) and 4 μm (magnified images). d, Co-localization analysis between GFP::3xFLAG::WAGO-3 and PGL-1::mTagRFP-T based on the image shown in c. Signals from ROI1-ROI3 are plotted in orange, blue and green, respectively. The x and y axes indicate fluorescence intensity. PC, Pearson's correlation coefficient. e, Confocal micrograph as in c for PEI-1::mTagRFP-T instead of PGL-1::mTagRFP-T. ROI1: no PEI-1::mTagRFP-T expression. ROI2: PEI-1::mTagRFP-T is expressed. The areas indicated by dashed yellow boxes (i and ii) are magnified on the right. Scale bars, 10 μm (proximal gonad) and 4 μm (magnified images). f, Co-localization analysis as in d for PEI-1::mTagRFP-T instead of PGL-1::mTagRFP-T. g, Confocal maximum intensity projections of male-derived budding spermatids expressing GFP::3xFLAG::WAGO-3 in the presence and absence of PEI-1. Scale bars, 4 μm. h, Confocal micrograph showing spermatogenesis of late-L4 stage hermaphrodites expressing GFP::3xFLAG::WAGO-3 and PGL-1::mTagRFP-T in the pei-1(ok1050) mutant background. Germ cell development progresses from left to right. The areas indicated by dashed yellow boxes (i and ii) are magnified on the right. ROI1: PGL-1::mTagRFP-T is expressed; ROI2: no PGL-1::mTagRFP-T expression is detectable. Scale bars, 10 μm (proximal gonad) and 4 μm (magnified images). i, Co-localization analysis as in d in the pei-1(ok1050) mutant background based on the image shown in h. The images in c, e, g and h represent three biologically independent experiments. Source data are available online.  female-specific contributions to the Mis phenotype. We next examined the roles of WAGO-3 and PEI-1 in this process, and found that both PEI-1 and WAGO-3 are specifically required in the male (Fig. 2b), but not in the female (Fig. 2c). We conclude that PEI-1 and WAGO-3 have critical roles in paternal epigenetic inheritance.

PEI-1 recruits WAGO-3 to PEI granules through its IDR.
The N-terminal region of PEI-1 is predicted to adopt a BTB fold followed by a BACK domain 40,41 , whereas the C-terminal part of PEI-1 is predicted to be an intrinsically disordered region (IDR) (Fig. 3a). Following these predictions, we edited the endogenous PEI-1::mTagRFP-T locus to generate five different PEI-1 variants (Fig. 3b), and analysed their effects on PEI-1 and WAGO-3 expression in primary spermatocytes ( Fig. 3c-k) and budding spermatids ( Fig. 4a-n). As a control, free GFP was expressed from the wago-3 locus (Figs. 3b,l,m and 4h). This revealed that the BTB and BACK domains primarily affected granule number and intensity (Figs. 3j and 4i,j,n), but had a very small effect on co-localization between PEI-1 and WAGO-3 (Figs. 3k and 4k). By contrast, deletion of the PEI-1 IDR resulted in WAGO-3 that mostly localized to the residual body (Fig. 4e,f,l,m). WAGO-3 signal was diffuse, as quantified by a loss of high-intensity pixels (Fig. 4i,j). The PEI-1 signal itself was weaker, but remained in foci that segregated into the spermatids (Fig. 4e,f). Deletion of both the BACK and IDR domains did not further affect WAGO-3, but did result in a diffuse PEI-1 signal that accumulated in the residual body, together with WAGO-3 (Fig. 4g,k).
Residual PEI-1 signal was always detected in spermatozoa within the spermatheca, even when both the BACK and IDR domains were deleted (Extended Data Fig. 5b-g). We observed the same for free GFP (Extended Data Fig. 5h). However, WAGO-3 is undetectable in mature sperm when PEI-1 misses its IDR (Extended Data Fig. 5f,g), indicating that WAGO-3 cannot be stably maintained in sperm without PEI-1 interaction.
We conclude that the IDR of PEI-1 is essential to recruit and stabilize WAGO-3, whereas the BTB, BACK and IDR domains have important roles in forming and stabilizing PEI-1 foci during spermatogenesis.
Characteristics of PEI granules. WAGO-3 was found to be highly sensitive to 1,6-hexanediol, a compound that is often used to probe condensates 42 , as no foci remained in the presence of only 1.25% (Extended Data Fig. 6a,b). By contrast, PEI-1 foci were more resistant, especially in budding spermatids in which even 5% 1,6-hexanediol treatment did not cause complete disassembly of PEI granules (Extended Data Fig. 6b). This resistance to 1,6-hexanediol possibly derives from additional interactions between the folded BTB and BACK domains, which have been shown to drive oligomerization 41,43 .
We also compared WAGO-3 mobility between PEI granules and P granules by measuring fluorescence recovery after photobleaching (FRAP) (Extended Data Fig. 6c,d). Proteins localizing to the liquid phase of P granules have been reported to exhibit high recovery rates 44,45 . Consistently, we found that WAGO-3 showed relatively

Fig. 2 | WAGO-3 and
PEi-1 are required for paternal epigenetic inheritance. a, Schematic of the crosses that were used to examine the specific effects of maternal and paternal epigenetic inheritance on the Mis phenotype. The mut-7(pk204) and mut-16(pk710) alleles both cause global depletion of Mutator 22G-RNAs, and can be used interchangeably. Small RNA colour code: grey, absent without parental influence; green, present; without parental influence; orange, present with maternal influence; blue, present with paternal influence. b,c, The percentage of fertile F 1 animals generated by crosses between males and hermaphrodites of the indicated genotypes. Fertility implies the presence of paternal (b) or maternal (c) epigenetic inheritance. Sterility implies no epigenetic inheritance. Statistical significance was tested using a Pearson's χ 2 test with yates continuity correction. NA, not available. . For c-i, GFP::3xFLAG::WAGO-3 appears in green and PEI-1::mTagRFP-T variants appear in magenta. The images represent two biologically independent experiments. The image in h was acquired with a higher gain compared with the image in g to visualize the remaining PEI-1::mTagRFP-T signal within the spermatocyte. l, Free GFP was expressed from the wago-3 locus. j,k, Quantification of the GFP::3xFLAG::WAGO-3 foci number (j), and co-localization of GFP::3xFLAG::WAGO-3 and PEI-1::mTagRFP-T (k) in isolated, male-derived primary spermatocytes expressing the indicated PEI-1::mTagRFP-T variants. FL, full length. n = 10 cells pooled from two independent experiments for each condition. Statistically significant differences were determined using one-way analysis of variance (ANOVA) (P ≤ 0.001) followed by Tukey's honest significant difference post hoc test (P ≤ 0.05). Different letters represent significant differences. The exact P values are provided as source data. The box plots show the median (centre line), 25th or 75th percentiles (box edges), and the whiskers indicate the median ± 1.5 × interquartile range. Note that the full-length data in j and k are the same as those shown in Extended Data Fig. 4c (primary spermatocyte) and Extended Data Fig. 10f,g, respectively. For c-i and l, scale bars, 4 μm. m, Confocal micrograph of an L4-stage hermaphrodite expressing free GFP from the endogenous wago-3 locus. The image represents two biologically independent experiments. Scale bar, 20 μm. Source data are available online.
rapid FRAP in P granules (t 1/2 = 4.9 s). In PEI granules of budding spermatids, WAGO-3 exhibited much slower exchange dynamics (t 1/2 = 42.2 s). Moreover, we found that the mobile fraction of WAGO-3 was reduced in PEI granules compared with in P granules. The prevalence of certain amino acids has been shown to modulate the material properties of condensates 46 . In particular, glycine residues maintain liquidity, whereas serine and glutamine residues promote hardening. We analysed the amino acid composition of the PEI-1 IDR, and compared it with the IDRs of PGL-1 and PGL-3, which are both known to localize to the liquid phase of P granules [47][48][49] , and MEG-3 and MEG-4, which are both reported to form gel-like assemblies 45 (Extended Data Fig. 7a-h). This revealed that the PGL-1 and PGL-3 IDRs were strongly enriched for glycine, due to a glycine-rich C-terminal domain (Extended Data Fig. 7b,e,f), whereas such enrichment was absent from PEI-1, MEG-3 and MEG-4 (Extended Data Fig. 7c,g,h). Introduction of this PGL-1-derived glycine-rich stretch into PEI-1::mTagRFP-T rendered the glycine enrichment of the PEI-1 IDR similar to that of PGL-1 (Extended Data Fig. 7d), and enhanced WAGO-3 recovery (Extended Data Fig. 7i-k), indicating that the PEI-1 IDR composition affects WAGO-3 mobility. Note that the mTagRFP-T tag reduced WAGO-3 recovery compared with untagged PEI-1 (compare Extended Data Fig. 6c with Extended Data Fig. 7k), indicating that the tags that we introduced affected PEI granule properties.
Finally, fusion and fission of individual foci is a strong indication for liquid-like properties of condensates. We therefore performed live imaging of PEI granules by monitoring WAGO-3. However, we found that they were rather static; we did not observe any major movements, and we therefore observed no fission or fusion events in a period of 1 h (Extended Data Fig. 6e and Supplementary Video 1).
PEI granules associate with membranous organelles. In our deletion analysis of PEI-1, we generated deletions that removed all of PEI-1, or all but a few amino acids at the N and C termini (Fig. 5a). Although the full deletion produced a diffuse signal that segregated significantly into the residual body, the remaining PEI-1 peptides guided the mTagRFP-T signal to discrete structures that were maintained in spermatids (Fig. 5b,c and Extended Data Fig. 8a-c), and that were distinct from PEI granules (Extended Data Fig. 8d,e). Besides the nucleus, only two organelles are sorted into spermatids-mitochondria and FB-MOs 26 . The latter are sperm-specific membranous organelles that help to sort major sperm protein (MSP) and other proteins into spermatids. They consist of a membranous part (MO) and a fibrous body (FB) made of MSP. The PEI-1-marked structures did not overlap with mitochondria (Extended Data Fig. 8f,g), and their numbers approximately match that of FB-MOs ( Fig. 5d), suggesting that the large PEI-1 deletion possibly marks FB-MOs. Furthermore, the sorting of PEI granules depends on the myosin VI motor protein SPE-15 (Fig. 5e), a protein that is known to drive FB-MO, but also mitochondria localization in sperm 50 .
To resolve PEI granule localization at a high resolution, we used correlative light and electron microscopy (CLEM) (Fig. 6a- Supplementary Information). In early primary spermatocytes, when the MOs are just starting to form from the Golgi, and no FBs are associated yet 26 , PEI granules were found close to and overlapping with MOs (Fig. 6a). This situation remained in the later stages of spermatogenesis: when the FB showed its typical fibrous structure and was enwrapped by the MO (Fig. 6b,c); when the MO started to retract from the FB (Fig. 6d); and when FBs were fully released from the MOs (Fig. 6e). From a total of 10 precise CLEM overlays, we found that 18 out of 18 and 17 out of 17 foci in spermatocytes and spermatids, respectively, were positioned immediately next to, or overlapping with, an MO. In both stages, only three of these foci were found to also contact a mitochondrion. We conclude that PEI granules associate with MOs.
In pei-2 mutants, FB-MOs, as visualized by the ΔH15-Q558 PEI-1 deletion (Fig. 5a,b), segregated normally into spermatids (Extended Data Fig. 10c,d). Moreover, PEI granules still formed (Extended Data Fig. 10e) and still recruited WAGO-3 (Fig. 7h,i and Extended Data Fig. 10e-g). However, in budding spermatids, these PEI granules did not properly segregate and were often lost in the residual bodies ( Fig. 7h-l), similar to spe-15 mutants (Fig. 5e). Mutants lacking SPE-15 are sterile, making it impossible to test the relevance of PEI-granule segregation in paternal epigenetic inheritance using the Mis phenotype. However, pei-2 mutants are fertile, enabling us to reveal that effective PEI-granule segregation is also required for epigenetic inheritance by sperm (Fig. 7m).

Segregation of PEI granules requires S-palmitoylation.
S-palmitoylation can guide proteins to membranes and typically occurs on Golgi-related membranes 51 . The palmitoyltransferase The image in f was acquired with higher gain compared with the image in e to visualize the remaining PEI-1::mTagRFP-T signal within the budding spermatids. h, Free GFP was expressed from the wago-3 locus. i,j, The fluorescence intensity (x axis) versus the mean of relative pixel count (y axis) of GFP::3xFLAG::WAGO-3 signal in budding spermatids (i) and residual bodies (j) expressing the indicated PEI-1::mTagRFP-T variants. n = 10 cells pooled from two independent experiments for each condition. The relative pixel count is the number of pixels with a given intensity within a selected region, divided by the total number of pixels in that region. In each plot, the curve derived from full-length PEI-1::mTagRFP-T is also shown in blue. The width of the curves reflects the s.d. of the mean. k-n, Quantification of the co-localization of GFP::3xFLAG::WAGO-3 and PEI-1::mTagRFP-T in budding spermatids + residual bodies (k), the fraction of GFP::3xFLAG::WAGO-3 foci in residual bodies (l), total GFP::3xFLAG::WAGO-3 signal in residual bodies (m) and GFP::3xFLAG::WAGO-3 foci number in budding spermatids + residual bodies (n) of male-derived cells expressing indicated PEI-1::mTagRFP-T variants. n = 10 cells pooled from two independent experiments for each condition. Statistically significant differences were determined using one-way ANOVA (P ≤ 0.001) followed by Tukey's honest significant difference post hoc test (P ≤ 0.05). Different letters represent significant differences. The exact P values are provided as source data. The box plots show the median (centre line), 25th or 75th percentiles (box edges), and the whiskers indicate the median ± 1.5 × interquartile range. Note that the full-length data in k, l, m and n are the same as the wild-type data shown in Fig. 7i, Fig. 7k, Fig. 7l,j and Extended Data Fig. 4c (budding spermatid), respectively. Representative images from two biologically independent experiments are shown in a-h. Source data are available online.
Fraction of GFP fluorescence in residual body SPE-10 localizes to MOs, and is required for their stable interaction with FBs 52 . We found that PEI granules were severely defective in spe-10 mutants. Large and irregularly shaped PEI granules formed along the cell periphery in spe-10 spermatocytes (Fig. 8a). Similar to the much smaller wild-type PEI granules, these structures were static and did not show signs of fusion or fission (Extended Data Movie 2). At the later stages, large PEI-1 aggregates were detected in the residual body (Fig. 8a). WAGO-3 and PEI-1 still co-localized in the absence of SPE-10 ( Fig. 8a), indicating that S-palmitoylation affects the subcellular localization of PEI granules, but not their recruitment of WAGO-3. Western blotting showed a clear doublet band for PEI-2 ( Fig. 8b), compatible with palmitoylation 53,54 . Interestingly, the top band of PEI-2 disappeared in spe-10 mutants (Fig. 8b), suggesting PEI-2 is a substrate of the SPE-10 enzyme. Our western blot analysis also revealed evidence for PEI-1 modification, although this appears more as a smear than as a discrete band (Fig. 8c). PEI-1 modification did not depend on SPE-10 ( Fig. 8c), suggesting that PEI-1 may carry a different kind of modification, or that another palmitoyltransferase may act on PEI-1. Strikingly, PEI-1 and PEI-2 affected each other's modification status-although PEI-2 was required for PEI-1 modification (Fig. 8c), PEI-1 inhibited PEI-2 modification (Fig. 8b). In a heterologous expression system, PEI-1 and PEI-2 also appeared as doublets, and showed decreased stability after palmitoylation inhibition (Extended Data Fig. 10h,i), similar to what has been reported for the palmitoylated protein PD-L1 (ref. 55 ). We conclude that PEI-2 is a potential direct substrate of SPE-10, and that PEI-1 can also be modified, but only in the presence of PEI-2. Budding spermatids + residual bodies

Discussion
We identified a previously undescribed, sperm-specific compartment-the PEI granule-and define its role in sperm-borne cytoplasmic inheritance of a specific Argonaute protein, WAGO-3 (Fig. 8d). PEI granules are made by PEI-1 and PEI-2 proteins, which contain a BTB-BACK domain, followed by an IDR. Interestingly, BTB and BACK domains can mediate homo-and heteromeric oligomerization 40,43 , providing multivalency, a property that is known to drive phase separation 24 . As has been found in other condensates 56 , our data are consistent with the idea that BTB-BACK domain interactions help to stabilize IDR-IDR interactions, together driving PEI granule formation. As such, we consider PEI-1 and PEI-2 as scaffold proteins 24 of PEI granules. Note that the material state of the PEI granules has not been clarified. To examine whether PEI granules have a liquid character, are more gel-like or represent some other form of complex, experiments with purified proteins will be required.
The PEI-1 IDR also recruits WAGO-3, which we propose is a client 24 of PEI granules. Even though WAGO-1 and CSR-1 have been proposed to be also present in sperm 27,57 , we found that N-terminally tagged versions of these proteins expressed from their endogenous loci accumulated in the residual bodies. We speculate that the PEI-1 IDR may create a condensate that is selective for some feature of the WAGO-3 protein, or possibly non-permissive for characteristics of depleted proteins, such as WAGO-1 and CSR-1. Such features could be sequence intrinsic, but could also relate to post-translational modifications. Further experiments will be required to test these ideas.
PEI granules interact with MOs through S-palmitoylation and PEI-2 is probably palmitoylated by SPE-10, providing an example of how acylation of germ granule components may promote their membrane-affinity. S-palmitoylation is reversible 51 , raising the possibility that PEI granules as a whole may be released from the MOs after fertilization by depalmitoylation. However, other processes that may affect PEI granules and their cargo in the zygote are the effect of dilution, which could directly affect PEI granule stability, and potential post-translational modification of PEI granules by maternal factors. We note that environmental cues, such as temperature, could also affect PEI granule behaviour and therefore paternal epigenetic inheritance. When domain organization is considered, PEI-1-related proteins can be easily identified within nematodes (Fig. 8e) and in humans (such as BTBD7; Fig. 8f). Interestingly BTBD7 carries a predicted myristylation site close to its N terminus, suggesting that it might be membrane-associated like PEI-2. The other human proteins shown in Fig. 8f are also known to be expressed in the testis and, for two of these, functions during spermatogenesis have been described previously 58,59 . BTBD18 forms nuclear foci 58 , and GMCL1 interacts with IDRs found in primate-specific GAGE proteins, and affects their localization 60 . Thus, the mechanism that we revealed may be broadly conserved.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41556-021-00827-2.   and c was performed once. d, We provide data for the top half of the model: WAGO-3 starts in P granules, gradually moves to PEI granules that associate with FB-MOs through S-palmitoylation, which ensures spermatid localization through SPE-15 dependent transport. The bottom half is hypothetical, and is depicted in reduced opacity. We speculate that PEI granules release their content into the oocyte, helping to establish/maintain silencing of specific targets (Extended Data Fig. 1). The schematic is not to scale. e, Phylogenetic analysis showing PEI-1 conservation within the Caenorhabditis genus. The phylogenetic tree was generated using EggNOG (v.4.5.1). PEI-1 was defined as a query and compared with all eukaryote entries. aa, amino acids. f, Protein length and domain composition of six human BTB-domain-containing proteins that resemble the PEI-1 protein composition.

C. elegans culture and strains.
Unless otherwise stated, all worm strains were cultured according to standard laboratory conditions at 20 °C on nematode growth medium (NGM) plates seeded with Escherichia coli OP50 (ref. 61 ). Animals for IP-MS/MS experiments were grown on egg plates (diameter, 90 mm) 62 for one generation, synchronized by bleaching, and then grown on standard NGM plates (diameter, 90 mm) for one generation before collection. Egg plates were generated by thoroughly mixing egg yolk with 50 ml LB medium/egg. After incubation at 65 °C for 2-3 h, the mixture was allowed to cool to room temperature before adding 10 ml OP50 culture per egg. About 10 ml was put on top of standard NGM plates (diameter, 90 mm) and incubated at room temperature. The next day, excess liquid was decanted and egg plates were incubated at room temperature for another 2 d. All of the strains are in the N2 Bristol background. A list of the strains used in this study is provided in Supplementary Table 1.
Mortal germline assay. All of the mutant strains were out-crossed four times with wild-type animals before starting the experiment, to clean the genetic background from potential mutations that occurred during culturing. For each strain, 90 L2 or L3 animals were distributed to 15 NGM plates (diameter, 90 mm), resulting in six larvae per plate. Animals were grown at 25 °C. Worms were picked onto fresh plates every second generation. The experiment was stopped after 17 generations.
Mutator-induced sterility crosses. All strains were confirmed and out-crossed twice before setting up crosses. Note that out-crossing ensured comparable results as an enhanced Mis phenotype was observed when using non-out-crossed animals. The transgenic allele otIs45[unc-119p::gfp] V was always present in paternal strains and served as mating control to avoid picking self-fertilized offspring. Only L2-stage F 1 animals were picked onto individual plates to avoid any biased selection. After 3 d, male or dead F 1 animals were excluded from the analysis. The fertility of F 1 animals was determined by the presence of F 2 animals after another 2-4 d.
PCR products served as linear, double-stranded DNA donor templates for the insertion of mTagRfp-t and gfp sequences. The mTagRfp-t coding sequence, including three introns and flanking homology regions, was amplified from pDD286, which was a gift from B. Goldstein (Addgene plasmid, 70684). The gfp coding sequence, including three introns and flanking homology regions, was amplified from pDD282. All PCR products were purified using the QIAquick PCR Purification Kit (28106, Qiagen), eluted in sterile water and confirmed by agarose gel electrophoresis. For all epitope tag insertions, co-conversions and precise deletions, we ordered 4 nmol Ultramer DNA oligodeoxynucleotides from Integrated DNA Technologies, which serves as linear, single-stranded DNA (ssODN) donor templates. All Ultramer DNA oligodeoxynucleotides were resuspended in sterile water. All linear DNA donor templates contained ~35 bp homology regions 71,72 . A list of all of the DNA donor templates is provided in Supplementary Table 3.
Immunoprecipitation experiments. Unless otherwise stated, synchronized animals were cultured at 20 °C until late-L4 stage, collected with M9 buffer and frozen on dry ice in sterile water and 200 µl aliquots. Aliquots were thawed on ice, mixed with same volume of 2× lysis buffer (50 mM Tris HCl pH 7.5, 300 mM NaCl, 3 mM MgCl 2 , 2 mM dithiothreitol (DTT), 0.2% Triton X-100, cOmplete Mini EDTA-free Protease Inhibitor Cocktail (11836170001, Roche)) and sonicated using a Bioruptor Plus device (B01020001, Diagenode) (4 °C, 10 cycles, 30 s on and 30 s off). After centrifugation for 10 min at 4 °C and 21,000g, supernatants were carefully transferred into a fresh tube without taking any material from the pellet or lipid phase. Pellet fractions were washed three times in 1× lysis buffer and resuspended in 1× Novex NuPAGE LDS sample buffer (NP0007, Invitrogen) supplemented with 100 mM DTT. Total protein concentrations of soluble worm extracts were determined using the Pierce BCA Protein-Assay (23225, Thermo Fisher Scientific) and an Infinite M200 Pro plate reader (Tecan). Extracts were diluted with 1× lysis buffer to reach 550 µl and a total protein concentration of 3 µg µl −1 . For each sample, 50 µl of this extract was added to 50 µl 1× Novex NuPAGE LDS sample buffer supplemented with 100 mM DTT and served as input control sample. For each IP experiment, 30 µl Novex DYNAL Dynabeads Protein G (10004D, Invitrogen) were washed three times with 500 µl 1× wash buffer (25 mM Tris HCl pH 7.5, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM DTT, cOmplete Mini EDTA-free Protease Inhibitor Cocktail), combined with the remaining 500 µl extract and incubated with rotation for 1 h at 4 °C. In the meantime, 8 µg antibody (monoclonal anti-Flag M2, F3165, Sigma-Aldrich; anti-Myc (9B11) mouse monoclonal antibodies, 2276, Cell Signaling Technology; monoclonal anti-HA (12CA5) mouse antibody, in-house production) was conjugated to another 30 µl Novex DYNAL Dynabeads Protein G according to the manufacturer's instructions. Extracts were separated from non-conjugated Dynabeads, combined with antibody-conjugated Dynabeads and incubated with rotation for 2 h at 4 °C. Following three washes with 500 µl 1× wash buffer, antibody-conjugated Dynabeads were resuspended in 25 µl 1.2× Novex NuPAGE LDS sample buffer supplemented with 120 mM DTT.
For RNA immunoprecipitation experiments, IPs were performed as described above with the following modifications: (1) adult animals were collected; (2) soluble worm extracts were diluted to 650 µl and a total protein concentration of 7 µg µl −1 , of which 150 µl served as input sample for later RNA extraction; and (3) antibody-conjugated Dynabeads were resuspended in 50 µl nuclease-free water.
For RIP experiments on males, synchronized wago-3(xf119) I;him-5(1490) V animals were cultured at 20 °C until adulthood. Adults were collected in M9 buffer and filtered through a 35 µm mesh using CellMicroSieves (35 µm pore size; N35R, BioDesign) 75 . Animals at the bottom of the mesh (>98% males) were collected and frozen on dry ice in sterile water and 200 µl aliquots.
MS and proteome comparison. IP was performed in quadruplicate. After resuspending the precipitates in Novex NuPAGE LDS sample buffer (NP0007, Invitrogen), samples were incubated at 70 °C for 10 min and separated on a Novex NuPAGE 4-12% Bis-Tris Mini Protein Gel (NP0321, Invitrogen) in 1× Novex NuPAGE MOPS SDS Running Buffer (NP0001, Invitrogen) at 180 V for 10 min. After separation the samples were processed by in-gel digest as previously described 76,77 . After protein digest, the peptides were desalted using a C18 StageTip 78 . For measurement, the digested peptides were separated on a 25 cm reverse-phase capillary (inner diameter, 75 µm) packed with Reprosil C18 material (Dr. Maisch). Elution was carried out along a 2 h gradient of 2-40% of a mixture of 80% acetonitrile/0.5% formic acid using the EASY-nLC 1000 system (LC120, Thermo Fisher Scientific). A Q Exactive Plus mass spectrometer (Thermo Fisher Scientific) operated with a Top10 data-dependent MS/MS acquisition method per full scan was used for measurement 79 . Processing of the obtained results was performed with the MaxQuant software (v.1.5.2.8) against the Wormbase protein database (version WS263) for quantification 80 . The processed data were visualized with R and R-Studio using in-house scripts.
RNA extraction, library preparation and sequencing. RNA of input and GFP::3×Flag::WAGO-3 IP samples was extracted using TRIzol LS Reagenz (10296010, Invitrogen) according to the manufacturer's instructions, and resuspended in nuclease-free water. RNA quality and quantity was assessed using the Bioanalyzer RNA 6000 Nano Kit (5067-1511, Agilent Technologies) and the Qubit RNA BR Assay Kit (Q10210, Invitrogen), respectively. RNA 5′ pyrophosphohydrolase (RppH) (M0356S, New England BioLabs) treatment was performed with a starting amount of 690 ng RNA. After purification, samples were quantified using the Qubit RNA HS Assay Kit (Q32852, Invitrogen). Next-generation sequencing library preparation was performed using the NEXTFLEX Small RNA-Seq Kit v3 (Bioo Scientific) following step A to step G of the manufacturer's standard protocol (v.16.06). Libraries were prepared with a starting amount ranging between 426 ng and 896 ng and amplified in 16 PCR cycles. Amplified libraries were purified by running an 8% TBE gel and size-selected for 15-50 bp. Libraries were profiled in a High Sensitivity DNA Chip on a 2100 Bioanalyzer Instrument (Agilent Technologies) and quantified using the Qubit dsDNA HS Assay Kit (Q32851, Invitrogen), in a Qubit 2.0 Fluorometer (Invitrogen). All of the samples were pooled at an equimolar ratio and sequenced on one NextSeq 500/550 flow cell, single read for 1 × 84 cycles plus seven cycles for the index read.
Read processing and mapping. Raw sequenced reads from high-quality libraries, as assessed by FastQC, were processed using Cutadapt 81 for adapter removal (-a TGGAATTCTCGGGTGCCAAGG -O 5 -m 26 -M 48) and low-quality reads were filtered out using the FASTX-Toolkit (fastq_quality_filter, -q 20 -p 100 -Q 33). Unique molecule identifiers were used to remove PCR duplicates using a custom script and were subsequently removed using seqtk (trimfq-l 4 -b 4). Finally, reads shorter than 15 nucleotides were removed using seqtk (seq -L 15).
Small RNA classification and target identification. All mapped reads were divided into sense and antisense reads using BEDTools intersect 83 , and reads of differing lengths and 5′ nucleotides were counted using a custom Python script.
Small RNA classes were then extracted from the mapped reads with the different classes defined as: 21U-RNAs, 21-nucleotide-long mapped reads that map sense to annotated piRNA loci; 22G-RNAs, mapped reads of lengths 20-23 nucleotides with a G at the 5′ position that map antisense to protein-coding genes/ ncRNAs/pseudogenes; 26G-RNAs, mapped reads 26 nucleotides long that map antisense to annotated protein-coding genes/ncRNAs/pseudogenes; miRNAs are 20-24 nucleotide reads that map sense to annotated miRNA loci; finally all mapped reads longer than 26 nucleotides were classed in a separate group. ncRNAs were defined to include not only annotated ncRNAs but also RNAs annotated as lincRNAs, snRNAs, snoRNAs, tRNAs and rRNAs. Read filtering was performed using a Python script based on pysam (v.0.15) 84 in combination with BEDTools intersect to extract miRNA and piRNA information 83 .
All mapped reads from sequences of 20-23 nucleotides in length, regardless of 5′-nucleotide and mapping direction, were counted using htseq-count (v.0.11.1) 85 (-s no -m intersection-nonempty). Reads per kb of transcript per million mapped reads (RPKM) values were calculated for 22G-RNAs mapping to protein-coding genes/ncRNA/pseduogenes relative to all mapped reads. Targets were defined as genes that were positive in at least two out of three replicates, with positive meaning that the 22G-RNA RPKM was (1) above 4 in the IP; (2) at least twice as high in the IP relative to the input; and (3) non-zero in the input.
For 22G-RNAs mapping to transposons, RPKM values were calculated relative to the predicted number and length of insertions in the custom annotation file and positives were defined using only requirements 2 and 3 above with no minimal RPKM requirement in the IP.
Reads mapping to intronic, exonic or untranslated regions were counted using a custom Python script with reads mapping at exon-intron junctions counted as 0.5 intronic and 0.5 exonic regardless of the spanned region.
Microscopy. For L4 larvae, adults and males, 20-30 animals were washed in a drop of 100 µl M9 buffer and subsequently transferred to a drop of 50 µl M9 buffer supplemented with 40 mM sodium azide on a coverslip. After 15-30 min, excess buffer was removed and a glass slide containing a freshly made agarose pad (2% (w/v) in water) was placed on top of the coverslip. For imaging embryos, adult hermaphrodites were washed and dissected in M9 buffer before mounting. To image sperm, L4 males were singled from hermaphrodites, grown over night, washed and dissected in SMG buffer (50 mM HEPES pH 7.5, 50 mM NaCl, 25 mM KCl, 5 mM CaCl 2 , 1 mM MgSO 4 , 10 mM glucose) by cutting near the vas deferens. Animals and sperm were immediately imaged using a TCS SP5 Leica confocal microscope equipped with the HCX PL APO ×63 water-immersion objective (NA, 1.2) or the HCX PL APO CS ×40 oil-immersion objective (NA, 1.3). Fluorescence emission was detected by either photomultiplier tubes or hybrid detectors. Depending on the experiment, SMG buffer was supplemented with 1:2,000 Hoechst 33342 (H1399, Invitrogen), 200 nM MitoTracker Green FM (M7514, Invitrogen) or 1,6-hexanediol (240117, Sigma-Aldrich), and sperm were imaged after 30 min incubation. To score the expression of a germline-specific mCherry::H2B transgene, we used a Leica DM6000 B research microscope equipped with a HC PL Fluotar ×20 dry objective (NA, 0.5). Images were processed using ImageJ v.1.53i. The following figures were deconvolved using the Huygens Remote Manager v3.6: Figs. 1c,e,h and 7b and Extended Data Figs. 4a,b,e-j, 5a and 8c.
Time series of spermatocytes expressing GFP::3×Flag::WAGO-3 were acquired using a fluorescence spinning-disk confocal microscope from Visitron Systems (VisiSope 5Elements) based on a Nikon Ti-2E stand and a spinning disk from Yokogawa (CSU-W, 50 µm pinhole) controlled by the VisiView software. The microscope was equipped with a ×60 plan apochromatic water-immersion objective (CFI Plan Apo VC; NA,1.2), a twofold magnification lens in front of the sCMOS camera (BSI, Photometrics), and a stage-top incubation chamber for live imaging (20 °C, ambient CO 2 ). The sample was excited by an argon laser at λ ex = 488 nm (200 mW, power set to 20%) and the emission was detected in a range of λ em = 500-550 nm (ET525/50m, Chroma).

Image quantification.
Co-localization analyses of confocal micrographs of C. elegans gonads were performed using ImageJ v.1.53i and the JACoP plugin. Foci quantification and co-localization analyses of confocal z stacks of isolated, male-derived germ cells were performed using IMARIS 9.7.2. The distribution of fluorescence intensities between budding spermatids and residual bodies, as well as quantification of pixel counts per fluorescence intensity (grey value) were determined using ImageJ v.1.53i. The relative pixel count is the number of pixels with a given fluorescence intensity within a selected region of interest (either spermatid or residual body), divided by the total number of pixels in that region. Every other data distribution is represented as a box plot, with the whiskers defining the median ± 1.5 × interquartile range, a rectangle marking the first and third quartile, and the centre line showing the median.
FRAP. FRAP measurements were performed on a TCS SP5 Leica confocal microscope equipped with a FRAP-booster and a HCX PL APO ×63 water-immersion objective (NA, 1.2). An entire granule was bleached in a fixed region of interest (diameter, 0.9 µm), while two additional control ROIs of same size were used to detect fluorescence emission of an unbleached granule and background signal, respectively. Five prebleach frames were recorded (5 × 0.374 s per frame), followed by two bleach frames (2 × 0.374 s per frame), and 3 sets of post-bleach frames (10 × 0.5 s per frame, 10 × 5 s per frame, 15 × 10 s per frame). Data analysis, including full-scale normalization and curve fitting using a double term exponential equation, was performed using EasyFRAP-web 89 .
MO counting. The LysoSensor Blue DND-192 stained MOs 90 within living spermatids were viewed using an Olympus BX60 with a ×100/1.35 NA oil-immersion objective lens. Epi-fluorescence of stained MOs was imaged using a DAPI filter pack and captured using a SensiCam digital camera (Cooke) controlled by SlideBook software (Intelligent Imaging Innovations). SlideBook software was used to collect z-axis stacks of 11-30, and 12-bit images were captured every 0.44-0.88 μm. The majority of images within the z stacks was collected approximately every 0.65 μm. A nearest-neighbour deconvolution algorithm within the SlideBook software was applied to the images. Images were then converted to z-axis projections, again using the SlideBook software. The diameters of individual spermatids were measured within SlideBook using the software's ruler function over images captured with a DIC filter. The manipulated images were exported from SlideBook as 16-bit tif images. The 16-bit tif images exported from SlideBook were reopened in ImageJ (NIH). The cell counter plugin within ImageJ was used to assist in counting MOs. Images were compiled using PhotoShop CS3 (Adobe).
For the IP experiments, 4.0 × 10 6 BmN4 cells were seeded in a 10 cm dish and transfected with 10 µg of plasmid DNA. Cells were lysed in IP-150 lysis buffer (supplemented with 0.5% Triton X-100) and IP experiments were performed using ChromoTek RFP-TRAP beads. After IP, beads were washed five times in IP-150 lysis buffer (supplemented with 0.5% Triton X-100).
CLEM analysis. For CLEM analyses, C. elegans males (wago-3(xf119) I; pei-1(xf193) IV; him-5(e1490) V) were selected using a stereomicroscope. Using a platinum wire, around 50 worms were transferred into the 100-µm-deep cavity of an A-type carrier (Wohlwend, imbibed with 1-hexadecene) filled with thick E. coli paste (serving as a cryo-protectant; thick OP50 E.coli suspension in M9 worm buffer, 20% BSA, 150 mM NaCl), closed with the flat side of a B-type carrier and subsequently high-pressure frozen (HPM010, AbraFluid). All samples were further processed by freeze-substitution in a temperature-controlling device (EM-AFS2, Leica Microsystems). Freeze-substitution was carried out at −90 °C for 72 h with 0.1% (w/v) uranyl acetate in glass distilled acetone (EMS). The temperature was then raised to −45 °C (3.5 °C h −1 ), and samples were further incubated for 5 h. After rinsing in acetone, the samples were infiltrated with increasing concentrations (10%, 25%, 50% and 75%; 6 h each) of Lowicryl HM20 resin (EMS) in acetone, while the temperature was further raised to −25 °C. Lowicryl (100%) was exchanged three times in 10 h steps, and samples were ultraviolet-light polymerized at −25 °C for 48 h, after which the temperature was raised to 20 °C (5 °C h −1 ), and ultraviolet-light polymerization continued for 6 h. Longitudinal sections (thickness, 300 nm) were cut using an ultra-microtome (UC7, Leica) and a diamond knife (ultra semi, DiATOME). Targeting of areas containing spermatocytes and spermatids was performed using toluidine blue staining. Sections of interest were picked up onto carbon-coated formvar-slot grids. Ultramicrotomy and acquisition of the in-resin retained fluorescence within the sections were best performed on the same day to avoid bleaching of the fluorescence. The fluorescence microscopy imaging of the sections (stained also with HOECHST) was performed as previously described 91 using a wide-field fluorescence microscope (Olympus IX81) equipped with an Olympus PlanApo ×100/1.40 NA oil-immersion objective. After post-staining, tilt series of the area of interest (1° increments, −60° to 60°) were acquired using a FEI TECNAI F30 TEM operated at 300 kV and a fast Gatan OneView 4K camera. Tomograms were reconstructed at a final voxel size of 1.56 nm using gold fiducials and weighted-back projection algorithms of the software package IMOD 92 . Correlation between light and electron micrographs was carried out using the plugin ec-CLEM 93 of the software platform Icy 94 . The coordinates of pairs of corresponding features in the two imaging modalities (50 nm Tetraspecks, HOECHST-stained condensed chromosomes, auto-fluorescent uranyl-acetate-stained E. coli) were used to calculate a linear transformation, which enabled mapping of the coordinates of the fluorescent spots of interest to overlay them onto the electron micrograph. Electron tomograms were displayed and analysed using the IMOD software package 92 .
Statistics and reproducibility. Statistical analyses were performed using R-based packages. The log-rank test was used for the mortal germline assay. For multiple group comparison, either one-way ANOVA followed by Tukey honest significant difference post hoc test, Pearson's χ 2 test with Yates continuity correction or Fisher's exact test was used. P < 0.05 was considered to be significant. Pearson's correlation analyses were performed to determine the relationship between two different factors. No statistical method was used to predetermine sample size. No data were excluded from the analyses. Fluorescence microscopy, CLEM, time-lapse microscopy, 1,6-hexanediol assays as well as C. elegans experiments comprising mCherry::H2B(RNAe) reactivation and Mutator-induced sterility were performed twice. Mortal germline assay (Extended Data Fig. 1a) and western blots (Fig.  8a,b and Extended Data Fig. 10a,b,h,i) were performed once unless specified in the legends. Immunoprecipitation experiments associated with MS and small RNA sequencing were performed in biological quadruplicates and triplicates, respectively. Quantification and statistical analyses of microscopy images were derived from the number of cells analysed across two independent experiments as indicated in the figure legends, with the presented data being derived from one representative independent experiment.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The accession number for the small-RNA-seq data generated in this study is PRJNA629991. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 95 partner repository under dataset identifier PXD019099. The BigDataViewer supporting the CLEM analyses is available at Mendeley Data (V1; https://doi.org/10.17632/dgb8d7h2hz.1; https:// data.mendeley.com/datasets/dgb8d7h2hz/1). All other data supporting the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper. Fig. 1 | WAGO-3 is required for germline immortality and transgenerational maintenance of RNAe. a, Mortal germline assay representing loss of fertility of strains with indicated genotype at 25 °C. Statistical significance was tested with a log-rank-test (n = 15 populations per strain assayed in a single experiment). b, Diagram displaying mCherry::H2B(RNAe) reactivation in prg-1(n4357), prg-1(n4357);mut-7(xf125) and prg-1(n4357);wago-3(pk1673) mutant generations. F2-F5: second-fifth homozygous generation. For each generation, reactivation in 10 populations of 50 animals each was scored. Each plotted point represents the fraction of 50 animals that express the mCherry::H2B transgene. Since no prg-1(n4357) single mutant animal was found to reactivate mCherry::H2B expression, the value of this group is deterministically zero due to lack of variability/statistical noise. Thus, any positive number of animals that expresses the mCherry::H2B transgene in either the prg-1(n4357);mut-7(xf125) or prg-1(n4357);wago-3(pk1673) group causes a significant difference from the prg-1(n4357) group. c, Micrographs of three example animals with the mCherry::H2B transgene in either RNAe (prg-1(n4357)) or activated (prg-1(n4357);wago-3(pk1673) and prg-1(n4357);mut-7(xf125)) status. Top panel shows schematic representation of an adult hermaphroditic gonad. Activity status of the transgene was homogeneous in F2 homozygous prg-1(n4357) and prg-1(n4357);mut-7(xf125) mutants. Images represent two biologically independent experiments. Scale bars: 30 μm. Source data are provided. Fig. 4 | PEi granules specifically recruit WAGO-3. a,b, Confocal maximum intensity projections of an adult hermaphrodite (a) and gastrula-staged embryo (b) expressing indicated proteins. Zooms show perinuclear co-localization of GFP::3xFLAG::WAGO-3 and PGL-1::mTagRFP-T in meiotic (a) and primordial (b) germ cells. Z2/Z3 are the primordial germ cells of C. elegans. Scale bars: 20 μm (a, adult), 20 μm (b, embryo), 10 μm (a, zoom), 4 μm (b, zoom). c, Quantification of GFP::3xFLAG::WAGO-3 foci number within indicated, male-derived germ cells (n = 10 cells pooled from two independent experiments, for each condition). Statistically significant differences were determined by one-way ANOVA (p ≤ 0.001) followed by Tukey's honestly significant difference post hoc test (p ≤ 0.05). Different letters represent significant differences. Note that the values for primary spermatocytes and budding spermatids (c) are the same as those shown in Fig. 3j (FL) and Extended Data Fig. 10f (wild-type), and Fig. 4n (FL) and Fig. 7j (wild-type), respectively. Secondary spermatocyte and budding spermatid stages are separated into 'c' and 's'. c: coupled; due to incomplete cytokinesis, leaving the two sister cells coupled and both cells were analyzed; s: separate, each of the coupled cells in 'c' was analyzed individually. Boxplot centre and box edges indicate median and 25th or 75th percentiles, respectively, while whiskers indicate the median ± 1.5 x interquartile range. d, Co-immunoprecipitation experiments using whole-worm extracts of late-L4 stage hermaphrodites analyzed by Western blotting. Sample processing control was run on a different gel. Data represent two biologically independent experiments. e-j, Confocal micrographs of an adult male (f) and late-L4 stage hermaphrodites (e,g-j) expressing indicated proteins. sc -spermatocyte, st -spermatid, rb -residual body. Scale bars: 10 μm (e-j). a,b,e-j, Images represent two biologically independent experiments. k-n, Co-localization analyses between PEI-1::mTagRFP-T and DEPS-1::GFP (k), GFP::3xFLAG::WAGO-1 (l), GFP::3xFLAG::CSR-1 (m) and GFP::3xFLAG::ALG-3 (n) based on the images shown in g-j, respectively (n = 10 worms for each condition). X and Y axes indicate fluorescence intensity. PC: Pearson's correlation coefficient. Exact P values (c), unprocessed original scans of blots and the source data for all graphical representations are provided. Fig. 7 | the amino acid composition of the PEi-1 iDR affects exchange dynamics of WAGO-3. a,b, Occurrence of glycine and serine residues in PEI-1 (a) and PGL-1 (b) was counted in amino acid 50-mers, starting at position one, shifting 5 residues at a time, and displayed as stacked columns. Indicated residue positions in the diagrams are the mid-point of the 50-mer. Y-axes display number of relevant residues in amino acid 50mers. X-axes indicate the position along the respective proteins. The various domains are indicated by vertical, dashed lines. NtDD and CDD indicate the N-terminal and C-terminal dimerization domains of PGL-1, respectively. The exact positions of glycine and serine residues for each protein are indicated above the stacked bar diagrams. IDR -intrinsically disordered region. c-h, Amino acid composition profiles of the intrinsically disordered region of the indicated proteins. Bars representing serine and glycine residues are highlighted in blue and orange, respectively. Panel d reflects a fusion between the PEI-1 IDR and the very C-terminal end of PGL-1(A711-F771). The profiles were generated using Composition profiler. Sequences were analyzed against the SwissProt database using 10,000 bootstrap iterations. Statistical significance was tested using the two sample t test (***: p ≤ 0.001, **: p ≤ 0.01, *: p ≤ 0.05, ns: p > 0.05). The exact P values are provided as source data. i-j, Time lapse images showing fluorescence recovery after photobleaching (FRAP) of GFP::3xFLAG::WAGO-3 localizing to PEI granules via PEI-1::mTagRFP-T (i) or via PEI-1::PGL-1(A711-F771)::mTagRFP-T (j) in isolated, male-derived budding spermatids. Residual bodies are marked by a dashed circle. Images represent two biologically independent experiments. Scale bars: 4 μm. k, FRAP recovery curves of GFP::3xFLAG::WAGO-3 localizing to PEI granules, containing indicated PEI-1 proteins, in male-derived budding spermatids. Normalized data is presented as mean + /− SD and was fitted to a double exponential curve (n = 5 granules pooled from one independent experiment). Source data are provided. Fig. 10 | PEi-2 interacts with PEi-1. a, Co-immunoprecipitation experiments using whole-worm extracts of late-L4 stage hermaphrodites analyzed by Western blotting. Labels above the blots indicate the presence (+) or absence (-) of the respective tags. Asterisks indicate non-specific signals. b, Pull-down experiments on extracts of transfected BmN4 cells expressing the indicated PEI-1 and PEI-2 variants. Full-length (FL) PEI-1-mCherry was pulled down, followed by detection of the various PEI-2 fragments. Expression of free 3xFLAG-eGFP served as negative control. c, Fraction of total mTagRFP-T signal within the residual body of male-derived budding spermatids expressing PEI-1_ΔH15-Q558::mTagRFP-T in wild-type or pei-2(xf270) mutant background (n = 10 cells pooled from two independent experiments, for each condition). Note that the wild-type data is the same as shown in Fig. 5c (ΔH15-Q558). d, Confocal maximum intensity projection of isolated, male-derived budding spermatids expressing PEI-1_ΔH15-Q558::mTagRFP-T in pei-2(xf270) mutant background. The residual body is marked by a dashed circle. Scale bar: 4 μm. e, Confocal maximum intensity projection of an isolated, male-derived primary spermatocyte expressing GFP::3xFLAG::WAGO-3 and PEI-1::mTagRFP-T in pei-2(xf270) mutant background. Scale bar: 4μm. d,e, Images represent two biologically independent experiments. f-g, Co-localization analysis between GFP::3xFLAG::WAGO-3 and PEI-1::mTagRFP-T (g), and quantification of GFP::3xFLAG::WAGO-3 foci number (f) in wild-type and pei-2(xf270) mutant, male-derived primary spermatocytes (n = 10 cells pooled from two independent experiments, for each condition). c,f,g, Statistically significant differences were determined by one-way ANOVA (p ≤ 0.001). Boxplot centre and box edges indicate median and 25th or 75th percentiles, respectively, while whiskers indicate the median ± 1.5 x interquartile range. Note that the wild-type data in f and g are the same as those shown in Fig. 3j (FL) and Extended Data Fig. 4c (primary spermatocyte), and Fig. 3k (FL), respectively. h-i, Transfected BmN4 cells were treated with the palmitoylation inhibitor 2-BP at indicated concentrations, followed by Western Blot detection of PEI-1-mCherry (h) and PEI-2-HA-eGFP (i). α-tubulin served as loading control. a,b,h,i, The experiment has been performed once. Unprocessed original scans of blots are provided in source data.