HDAC1 SUMOylation promotes Argonaute-directed transcriptional silencing in C. elegans

Eukaryotic cells use guided search to coordinately control dispersed genetic elements. Argonaute proteins and their small RNA cofactors engage nascent RNAs and chromatin-associated proteins to direct transcriptional silencing. The small ubiquitin-like modifier (SUMO) has been shown to promote the formation and maintenance of silent chromatin (called heterochromatin) in yeast, plants, and animals. Here, we show that Argonaute-directed transcriptional silencing in Caenorhabditis elegans requires SUMOylation of the type 1 histone deacetylase HDA-1. Our findings suggest how SUMOylation promotes the association of HDAC1 with chromatin remodeling factors and with a nuclear Argonaute to initiate de novo heterochromatin silencing.


Introduction
Argonautes are an ancient family of proteins that utilize short nucleic acid guides (usually composed of 20-30 nts of RNA) to find and regulate cognate RNAs (reviewed in Meister, 2013). Argonautedependent small RNA pathways are linked to chromatin-mediated gene regulation in diverse eukaryotes, including plants, protozoans, fungi, and animals (reviewed in Martienssen and Moazed, 2015). Connections between Argonautes and chromatin are best understood from studies in fission yeast Schizosaccharomyces pombe, where the Argonaute, Ago1, a novel protein, Tas3, and a heterochromatin protein 1 (HP1) homolog, Chp1, comprise an RNA-induced transcriptional silencing (RITS) complex that maintains and expands heterochromatin (Verdel et al., 2004). The Chp1 protein binds H3K9me3 through its conserved chromodomain (Partridge et al., 2000;Partridge et al., 2002) and is thought to anchor Ago1 within heterochromatin, where it is poised to engage nascent RNA transcripts (Holoch and Moazed, 2015). Low-level transcription of heterochromatin is thought to create a platform for propagating small-RNA amplification and heterochromatin maintenance (reviewed in Holoch and Moazed, 2015).
In yeast, a protein complex termed SHREC (Snf2/Hdac-containing Repressor complex) has been linked to both the establishment and maintenance of heterochromatin and transcriptional silencing (Job et al., 2016;Motamedi et al., 2008;Sugiyama et al., 2007). SHREC contains a homolog of type 1 histone deacetylase (HDAC), a homolog of Mi-2 and CHD3 ATP-dependent chromatin remodelers, and a Krü ppel-type C2H2 zinc finger protein. SHREC therefore resembles the nucleosome remodeling and deacetylase (NuRD) complex in animals (Denslow and Wade, 2007;Torchy et al., 2015). NuRD complexes play a key role in converting chromatin from an active to a silent state, and can be recruited to targets through sequence-independent interactions (e.g., modified chromatin or methylated DNA) or through sequence-specific interactions via the Krü ppel-type C2H2 zinc finger protein and other DNA-binding factors (Ecco et al., 2017;Lupo et al., 2013). In animals, NuRD complexes function broadly in developmental gene regulation and transposon silencing (Ecco et al., 2017;Feschotte and Gilbert, 2012;Ho and Crabtree, 2010).
The post-translational modification of heterochromatin factors by the small ubiquitin-like protein SUMO has been implicated at several steps in the establishment and maintenance of transcriptional gene silencing and has been linked to silencing mediated by the Piwi Argonaute (reviewed in Ninova et al., 2019). The addition of SUMO (i.e., SUMOylation) requires a highly conserved E2 SUMO-conjugating enzyme, UBC9, which interacts with substrate-specific co-factor (E3) enzymes to covalently attach SUMO to lysines in target proteins (Johnson, 2004). Whereas ubiquitylation is primarily associated with the protein turnover (reviewed in Glickman and Ciechanover, 2002), SUMOylation is primarily associated with changes in protein interactions, especially with proteins that contain SUMO-interacting motifs (SIMs) (reviewed in Kerscher, 2007). In mammals, for example, SUMOylation of the KAP1 corepressor is required to recruit the NuRD complex and the SETDB histone methyltransferase via SIM domains in CHD3 and SETDB and to silence KRAB targets (Ivanov et al., 2007).
Here, we identify a connection between the SUMO pathway and transcriptional silencing initiated by the Piwi Argonaute pathway in the Caenorhabditis elegans germline. We show that SUMOylation of C-terminal lysines on the type 1 HDAC, HDA-1, is required for Piwi-mediated transcriptional silencing. SUMOylation of HDA-1 promotes its association with conserved components of the C. elegans NuRD complex, the nuclear Argonaute HRDE-1/WAGO-9, the histone demethylase SPR-5, and the SetDB-related histone methyltransferase MET-2. Our findings suggest how SUMOylation of HDAC1 promotes the recruitment and assembly of an Argonaute-guided chromatin remodeling complex that orchestrates de novo transcriptional gene silencing in the C. elegans germline.
Null alleles of many of these genes cause embryonic arrest, which precludes an analysis of silencing in the adult germline. To further explore the role of SUMO and HDAC factors in piRNA silencing, we therefore tested whether partial or conditional loss-of-function alleles activate the piRNA sensor. Auxin-inducible degron alleles of hda-1, let-418, mep-1, and mrg-1 and a truncation allele of sin-3 all desilenced the piRNA sensor in 100% of worms examined ( Figure 1C). We found that a 3xflag fusion to the endogenous smo-1 gene desilenced the piRNA sensor in~40% of adults analyzed ( Figure 1C), suggesting that this tagged smo-1 allele behaves like a partial loss of function. A strain expressing a temperature-sensitive UBC-9(G56R) protein desilenced the sensor in 90% of the animals at the semi-permissive temperature of 23˚C ( Figure 1C) (Kim et al., 2021). By contrast, Figure 1. SUMOylation and chromatin remodeling factors promote piRNA-mediated silencing. (A) Schematic of the piRNA sensor screen. The piRNA sensor strain contains a gfp::csr-1 transgene that is silenced by the piRNA pathway in the presence of an active oma-1::gfp transgene (Seth et al., 2018). OMA-1::GFP localizes to the cytoplasm of oocytes. Inactivation of the piRNA pathway (by RNAi, mutation, or auxin-inducible protein depletion) desilences the transgene, resulting in GFP::CSR-1 expression in perinuclear P-granules throughout the germline, as shown in (B). (B) Differential interference contrast and epifluorescence images of dissected gonads in wild-type (wt), prg-1(tm872), and wago-9/hrde-1(ne4769) worms. PRG-1 is required to initiate silencing, while WAGO-9 is required to maintain silencing. The percentage of desilenced worms and number of worms scored are shown. (C) Analysis of SUMO and chromatin remodeling factors required for piRNA-mediated silencing. Genes identified in the RNAi-based screen of chromatin factors are listed with their human homologs and with the percentage of worms that express GFP::CSR-1 among the total number of worms analyzed (n) when function is reduced by RNAi (blue column) or by either mutation or degron-dependent protein depletion (peach column).
presumptive null alleles of gei-17 completely failed to desilence the piRNA sensor strain (Kim et al., 2021). Together, these findings suggest that histone deacetylase complexes and components of the SUMO pathway promote piRNA-mediated silencing.

SUMOylation of HDA-1 promotes piRNA surveillance
In a parallel study, we found that C. elegans HDA-1 is SUMOylated in the adult germline (Kim et al., 2021). Moreover, we showed that SUMOylation of HDA-1, formation of an adult NuRD complex, and piRNA-mediated silencing depend redundantly on PIE-1, a CCCH zinc finger protein with SUMO E3-like function, and on GEI-17, a homolog of PIAS1/Su(var)2-10 SUMO E3 ligase, suggesting that SUMOylation of HDA-1 might promote piRNA surveillance (Kim et al., 2021). Human HDAC1 is SUMOylated near its C-terminus on consensus SUMO-acceptor lysines 444 and 476 ( Figure 2A; David et al., 2002), but HDA-1 does not contain consensus SUMO acceptor sites, and poor conservation between the C-termini of human HDAC1 and worm HDA-1 did not suggest the identity of potential SUMO-acceptor sites. However, GPS-SUMO prediction software identified lysines 444 and 459 of HDA-1 as possible non-consensus SUMO-acceptor sites ( Figure 2A; Zhao et al., 2014). To investigate whether one or both of these lysine residues is required for HDA-1 SUMOylation, we used CRISPR genome editing to mutate the endogenous hda-1 gene in a strain that expresses SUMO fused to 10 N-terminal histidines, his10::smo-1 (Kim et al., 2021). We then used nickel-nitrilotriacetic acid (Ni-NTA) affinity chromatography under stringent denaturing conditions (Tatham et al., 2009) to capture SUMOylated proteins from worm lysates. The Ni-column eluates were then analyzed by western blotting for HDA-1. SUMOylated HDA-1 was recovered by Ni-NTA affinity chromatography from wild-type lysate and from lysates of each singlesite lysine-to-arginine mutant ( Figure 2B, lanes 7-9), but was absent when both K444 and K459 were mutated together, HDA-1(KKRR) ( Figure 2B, lane 10). As a control, the protein MRG-1, which is highly SUMOylated in wild-type worms (Kim et al., 2021), was readily detected in worms expressing HDA-1(KKRR) ( Figure 2B). When we introduced the piRNA sensor into the HDA-1 SUMO-acceptor site mutants, we found that whereas the piRNA sensor remained silent in the single-site mutants (n = 30) (data not shown), it was expressed in 100% of HDA-1(KKRR) worms ( Figure 2C).
Since the SUMOylation sites in HDA-1 are very close to the C-terminus, we wondered if appending SUMO via translation fusion to the HDA-1(KKRR) mutant protein might restore HDA-1 function in the piRNA sensor assay. Using CRISPR, we inserted a modified smo-1 open reading frame just before the stop codon of the hda-1 [KKRR] gene at the endogenous hda-1 locus. The modified SMO-1 fusion cannot be transferred to other proteins because it lacks the GG amino acids required for conjugation (Dorval and Fraser, 2006;see Materials and methods). Surprisingly, the resulting hda-1 [KKRR]::smo-1 strain was homozygous viable, healthy, and expressed an HDA-1::SMO-1 fusion protein at levels similar to those observed for wild-type HDA-1 ( Figure 2D). Strikingly, appending SMO-1 to the C-terminus of HDA-1(KKRR) completely rescued the silencing defect of HDA-1(KKRR) ( Figure 2C). Moreover, HDA-1(KKRR)::SMO-1 rescued the piRNA-mediated silencing defects of smo-1 and ubc-9 mutants ( Figure 2C), suggesting that the SUMO pathway promotes piRNA-mediated silencing via C-terminal SUMOylation of HDA-1.

HDA-1 SUMOylation is not required for maintenance of piRNA-initiated silencing
Once silencing is established by the upstream components of the Piwi pathway, it can be maintained indefinitely by the co-transcriptional arm of the pathway without the continued need for prg-1 activity (Shirayama et al., 2012). For example, the initial silencing of the gfp::cdk-1 transgene requires PRG-1 activity, but maintenance of silencing does not (Shirayama et al., 2012). To ask if HDA-1 SUMOylation is required for the maintenance of piRNA-initiated silencing, we crossed an already silenced gfp::cdk-1 transgene into an hda-1[KKRR] mutant strain. We found that the gfp-1::cdk-1 reporter remained silent even after five generations in HDA-1(KKRR) worms (n = 26), supporting the idea that C-terminal SUMOylation of HDA-1 is not required to maintain piRNA-induced transcriptional silencing.  HDA-1 SUMOylation mutants cause temperature-dependent reductions in fertility Factors that promote genome integrity and epigenetic inheritance are required for germline immortality: their loss causes a mortal germline phenotype, whereby fertility declines in each generation and this decline is often exacerbated at elevated temperatures (Ahmed and Hodgkin, 2000).
In MEP-1 immunoprecipitates from hda-1[KKRR]::smo-1 lysates, we detected multiple HDA-1 bands, including a prominent band slightly larger than the expected size of HDA-1::SMO-1 (white stars in Figure 3D, E). Western blots with ubiquitin-specific antibody suggested that this prominent band is a mono-ubiquitinated form of the fusion protein ( Figure 3D, E). In both input and IP samples, we also observed an isoform similar in size to endogenous HDA-1 that was not detected by SUMO-specific antibodies ( Figure 3D, E), suggesting that the HDA-1::SMO-1 fusion protein may be cleaved near the C-terminus of HDA-1, removing the SUMO peptide.

HDA-1 SUMOylation promotes histone deacetylation in vivo
The findings above suggest that SUMOylation promotes the assembly and function of HDA-1 complexes. Our proteomic studies also revealed that SUMOylation promotes interactions between HDA-1 and other histone-modifying enzymes required for heterochromatin formation, including the demethylase SPR-5/LSD1, which removes activating H3K4me2/3 marks, and the methyltransferase MET-2/SetDB1, which installs silencing H3K9me2/3 marks (Greer and Shi, 2012). Consistent with the idea that SUMOylation of HDA-1 promotes silencing via SPR-5 and MET-2, immunostaining revealed greatly reduced levels of H3K9me2 and increased levels of the H3K4me3 and throughout the germline in HDA-1(KKRR) worms as compared to wild-type ( Figure 4A, B, Figure 4-figure supplement 1). Immunostaining also revealed higher levels of acetylated H3K9 (H3K9Ac) in germlines of HDA-1(KKRR) and mep-1-depleted worms than in wild-type ( Figure 4C). Moreover, we found that HDA-1, LET-418, and MEP-1 (including MEP-1 expressed from the germline-specific wago-1 promoter) bind heterochromatic regions of the genome, depleted of the activating H3K9Ac mark and enriched for the silencing marks H3K9me2/3 ( Figure 4D). Thus, SUMOylation of HDA-1 appears to drive formation or maintenance of germline heterochromatin.
Many of the factors that maintain heritable epigenetic silencing triggered by piRNAs-including WAGO-9/HRDE-1-are also required for multigenerational germline silencing triggered by exogenous dsRNA, that is, inherited RNAi (Ashe et al., 2012;Buckley et al., 2012;Shirayama et al., 2012). We therefore asked if HDA-1 SUMOylation is also required for inherited RNAi. Worms expressing a bright germline OMA-1::GFP were fed bacteria that express GFP dsRNA (i.e., RNAi by feeding) for one generation (P0). The subsequent F1 and F2 generations were removed from the RNAi food and cultured on a normal Escherichia coli diet (no gfp dsRNA). In wild-type worms, oma-1::gfp was silenced in the P0 generation (in the presence of gfp dsRNA) and remained silent in the F1 and F2 generations in the complete absence of gfp dsRNA ( Figure 6C). By contrast, as previously shown (Ashe et al., 2012;Buckley et al., 2012;Spracklin et al., 2017), oma-1::gfp is efficiently silenced in P0 wago-9/hrde-1 worms in the presence of gfp dsRNA, but oma-1::gfp is reactivated in the F1 and F2 generation after the removal from gfp dsRNA. hda-1[KKRR] animals were similarly defective for heritable RNAi: oma-1::gfp was efficiently silenced in the P0 worms in the presence of gfp dsRNA, but expression was fully restored in the F1 and F2 generations in the absence of gfp dsRNA ( Figure 6C). Finally, the hda-1[KKRR]::smo-1 translational fusion fully rescued the heritable RNAi defect of hda-1[KKRR] ( Figure 6C). Taken together, these findings suggest that HDA-1 SUMOylation promotes heritable transcriptional silencing in both the piRNA and RNAi pathways.

HDAC1 SUMOylation promotes Argonaute-directed transcriptional silencing
Argonaute small RNA pathways collaborate with chromatin factors to co-regulate gene expression, transposon activity, and chromosome dynamics (reviewed in Almeida et al., 2019). Here, we have identified components of the NuRD complex as well as the SUMO and its E2 protein ligase, UBC-9, as chromatin-remodeling and -modifying complexes required for the initiation of Piwi-mediated gene silencing. Using mutagenesis of candidate SUMO-acceptor sites and affinity enrichment of SUMO, we identified C-terminal lysines required for the SUMOylation of the HDAC1 homolog, HDA-1. Mutating these lysines to arginine in HDA-1(KKRR) abolished the initiation but not the maintenance of piRNA-mediated silencing and also abolished the initiation of transgenerational silencing in response to dsRNA. Appending SUMO by translational fusion to HDA-1(KKRR) completely rescued Argonaute-mediated silencing and even restored piRNA silencing in animals with mutations in the SUMO conjugating machinery. Taken together, these molecular genetic studies strongly implicate HDAC1 SUMOylation in promoting Argonaute-directed transcriptional silencing (Figure 7). SUMO: a potent genetic modifier with an elusive biochemical signature Given the strong genetic evidence that the modification of HDA-1 by SUMO promotes piRNA silencing in the adult germline, we were surprised that the conjugated isoform of HDA-1 was undetectable in our IP assays from adult animals. Only when Ni-Affinity purification was used under strict denaturing conditions was it possible to recover the HDA-1-SUMO isoform. A likely explanation for these findings is that the covalent attachment of SUMO to its target proteins is rapidly reversed in lysates from C. elegans adults (Kim et al., 2021).
The dynamic nature of SUMO conjugation is well known from studies on other organisms (Mukhopadhyay and Dasso, 2007), and the SUMO protease SENP1 has been identified as an important regulator of human HDAC1 SUMOylation (Cheng et al., 2004, see more discussion below). In the future, it will be interesting to explore whether any of the three C. elegans homologs of SENP1 regulate HDA-1 SUMOylation in the worm germline. The presence of such factors, which appear to be very active in worm lysates, likely explains why so little HDA-1-SUMO was recovered in our IP studies. Indeed, it seems likely that the labile nature of the SUMO interaction could explain the discordance between its robust genetic effect and its surprisingly weak physical association with its targets in co-IP assays. Consistent with this idea, an HDA-1::SMO-1 translational fusion protein with an abnormal linkage via the N-terminus of SUMO was stable in protein lysates, strongly rescued the silencing defects of the presumptive SUMO-acceptor mutant protein HDA-1(KKRR), and dramatically enhanced the detection of protein-protein interactions between HDA-1 and components of an adult-stage NuRD complex.
We were surprised that the HDA-1::SMO-1 fusion protein, which was constructed at the endogenous and essential hda-1 locus, was so well-tolerated. Conceivably, the cleavage and ubiquitination of HDA-1::SMO-1 that we detected in our IP assays provide alternative mechanisms to counter the effects of this translational fusion. The mono-ubiquitinated form of the HDA-1::SMO-1 protein exhibited reduced staining with a SUMO-specific antibody, raising the possibility that the SUMO moiety in this protein is itself modified by ubiquitin, a modification that has been reported for SUMO isoforms in humans (Danielsen et al., 2011;Tatham et al., 2008). Taken together, our findings suggest that HDA-1 SUMOylation is not only important genetically but is highly dynamic and may be regulated at multiple levels.

HDAC1 SUMOylation plays diverse roles in gene regulation
Studies in mice and in human cell culture have investigated SUMOylation of C-terminal HDAC1 lysines (Cheng et al., 2004;Citro et al., 2013;David et al., 2002;Joung et al., 2018;Tao et al., 2017). In at least three cases, mutating these lysines increased the levels of histone acetylation and transcriptional activation (David et al., 2002;Cheng et al., 2004;Joung et al., 2018). As mentioned above, the SUMO protease SENP1 was identified as a factor that promotes the removal of C-terminal SUMO moieties from human HDAC1 in a prostate cancer model, causing upregulation of the androgen receptor (Cheng et al., 2004). In a subsequent study, these authors identified SENP1 as upregulated in human prostate cancers and found that overexpression of SENP1 was sufficient to drive prostate neoplasia in a mouse model (Cheng et al., 2006). One very interesting study showed that Ab insult led to upregulation of the SUMO E3 factor PIAS and resulted in SUMOylation of C-terminal lysines in HDAC1 in the rat hippocampus (Tao et al., 2017). Remarkably, this study showed that a lentivirus-driven HDAC1::SUMO translational fusion protein could rescue the learning and memory deficits of an APP/Presenilin 1 murine model of Alzheimer's disease (Tao et al., 2017), suggesting that HDAC1 SUMOylation may be neuroprotective in response to Ab accumulation. Taken together, our worm studies and these studies in mammalian systems point to the likely importance of HDAC1 SUMOylation in the regulation of gene expression in diverse animals.

SUMO promotes the assembly of a germline MEP-1-NuRD complex
Precisely how HDA-1 SUMOylation promotes its downstream functions in piRNA silencing will require further work. One possibility is that SUMO acts as a bridge between HDA-1 and a nuclear Argonaute WAGO-9, while simultaneously promoting NuRD complex assembly (Figure 7 model). This model is supported by the dramatically enhanced association between MEP-1 and the HDA-1:: SMO-1 translational fusion and by the ability of HDA-1::SMO-1 to rescue the piRNA silencing defects of upstream SUMO pathway mutants.
Paradoxically, HDA-1 is not SUMOylated in the embryo and yet interacts robustly with MEP-1 (Kim et al., 2021). We do not know why the association of HDA-1 with MEP-1 requires SUMO in the adult germline but not the embryo. Perhaps unknown co-factors promote their SUMO-independent interaction in the embryo. Alternatively, adult germline-specific factors may inhibit or modify HDA-1 or other NuRD complex components to prevent their direct association. SUMOylation of HDA-1 in the adult may therefore enable it to associate with MEP-1 via a mechanism bridged by SUMO, a SUMO glue, or SUMO-SIM network mode of interaction (Matunis et al., 2006;Psakhye andJentsch, 2012 Pelisch et al., 2017). The MEP-1 protein has two consensus SIMs, which could be important for this association between MEP-1 and HDA-1. However, our preliminary studies on these motifs were confounded by the finding that their simultaneous mutation resulted in a completely sterile inactive mep-1 allele, similar phenotypically to a null allele. Thus, although the SIM motifs in MEP-1 may be important for its interaction with HDA-1-SUMO, they may also be required for other functions or for interactions with other essential co-factors.

Parallels in the role of SUMOylation in Piwi silencing in insects, mice, and worms
Histone deacetylation is a necessary step in de novo transcriptional silencing. Yet, precisely how nuclear Argonautes orchestrate both deacetylation and the subsequent installation of silencing marks on target chromatin is not known. The nuclear Argonaute WAGO-9/HRDE-1 initiates transcriptional silencing downstream of both the piRNA-and the dsRNA-induced silencing pathways (Ashe et al., 2012;Bagijn et al., 2012;Buckley et al., 2012;Shirayama et al., 2012), and we have shown here that HDA-1 interacts with WAGO-9 in a manner that partially depends on HDA-1 SUMOylation. In fruit fly, the nuclear Argonaute Piwi was recently shown to interact with the SUMO E3 PIAS homolog Su(var)2-10 and to promote the recruitment of the histone methyltransferase SetDB1/EGGless (Egg) and its cofactor MCAF1/Windei (Wde) (Ninova et al., 2020). A worm paralog of EGG, MET-2, was identified in our SUMO-dependent HDA-1 IP complexes. Although the worm PIAS/Su(var)2-10 homolog, GEI-17, scored negative in our piRNA sensor screen, in a parallel study we have identified a synthetic piRNA silencing defect between gei-17 and mutations in pie-1, which encodes a tandem zinc finger protein with properties consistent with SUMO E3 activity (Kim et al., 2021). Thus, an attractive possibility is that upon binding to nascent target transcripts, nuclear Argonautes recruit SUMO E3 factors to promote the SUMOylation and recruitment of both histone deacetylase and histone methyltransferase complexes.
Interestingly, in Drosophila Kc167 cells, a MEP-1:Mi-2 complex (called MEC) reportedly forms without HDAC1 (Kunert et al., 2009). Kc167 cells are derived from ovarian somatic cells and express components of the Piwi pathway (Vrettos et al., 2017), raising the possibility that HDAC1 SUMOylation might promote its association with the MEC complex in these cells. Interestingly, a recent paper showed that Drosophila MEP-1 and the HDA-1 homolog RPD3 interact with PIWI and with Su(var)2-10 in fly ovarian somatic cells, where they function together to promote transposon silencing (Mugat et al., 2020). Another recent study found that the mouse Piwi homolog MIWI engages NuRD complex components and DNA methyltransferases to establish de novo silencing of transposons in the mouse testes, suggesting additional parallels to the findings from worms and flies (Zoch et al., 2020). It will be interesting in the future to learn if mammalian and insect Piwi Argonautes target HDAC1 SUMOylation to promote de novo Piwi silencing.
The mep-1 gene was previously implicated in regulating the transition from spermatogenesis to oogenesis in hermaphrodites (Belfiore et al., 2002), and our findings suggest that HDA-1 SUMOylation also promotes this function. We were surprised, however, to find that spermatogenesis targets are also regulated by the WAGO-pathway factor RDE-3 and by the Piwi Argonaute PRG-1. Similar findings were reported recently by Reed et al., 2020. These observations raise the interesting possibility that C. elegans germline, Argonaute systems function with SUMO and HDAC1 in promoting the switch from spermatogenesis to oogenesis in hermaphrodites.

C. elegans strains and genetics
All the strains in this study were derived from Bristol N2 and cultured on nematode growth media (NGM) plates with OP50, and genetic analyses were performed essentially as described (Brenner, 1974). The strains used in this study are listed in Supplementary file 3.

RNAi screen
RNAi screen was performed against all 337 genes in the chromatin subset of the C. elegans RNAi collection (Ahringer). RNAi of smo-1 and ubc-9 were added to the screen as chromatin regulators. Synchronous L1 worms of the reporter strain were plated on the isopropyl-b-Dthiogalactoside (IPTG) plates with the corresponding RNAi food. Bacteria with empty L4440 vector served as negative control. The desilencing phenotype was scored when the worms on the control plates grew to young adult stage at 20˚C. In the first round of the screen, 78 RNAi clones scored positive. In the second round, we required that the sensor be desilenced in more than 5% of worms, resulting in 29 positive clones, including smo-1 and ubc-9 (Supplementary file 1).

Generation of strains expressing SUMO-conjugated HDA-1
To prevent transfer of SUMO from the HDA-1::SMO-1 fusion to other SUMO targets, we mutated the tandem C-terminal glycines of SMO-1/SUMO to alanines (GG to AA) (Dorval and Fraser, 2006). The modified smo-1 open reading frame was fused directly before the stop codon of hda-1 or hda-1 (ne4747 [KKRR]) gene at the endogenous hda-1 locus by CRISPR genome editing as described above.

Auxin treatment
For the auxin-inducible degron system (Zhang et al., 2015), tir-1::mRuby was expressed in the germline under the control of the sun-1 promoter and eft-3 3 0 UTR. The degron-tagged L1 larvae were plated on NGM plates with 100 mM auxin indole-3-acetic acid (IAA; Alfa Aesar, A10556) unless otherwise stated and kept in dark. Worms were collected at young adult stage for further analysis.

Immunoprecipitation-mass spectrometry (IP-MS)
Synchronous adult worms were used for IP experiments unless otherwise indicated. Worm lysates were prepared by combining 1 ml frozen worm pellet mixed with 1 ml IP lysis buffer (Kim et al., 2021) and 2 ml glass beads, and then vortexing on FastPrep (MP Biomedicals) at a speed of 6 m/s for 20 s for four times. Lysates were clarified twice by centrifugation for 30 min each at 12,000 rpm, 4˚C. GBP beads were prepared by conjugating GBP nanobody to N-hydroxysuccinimide (NHS)activated Sepharose beads (Rothbauer et al., 2008). GBP beads were incubated with lysates for 1 hr at 4˚C on a rotating shaker, and GFP-tagged proteins and complexes were eluted with 1% sodium dodecyl sulfate (SDS), 50 mM Tris, pH 8.0 at 95˚C for 10 min. Proteins were precipitated with trichloroacetic acid (TCA) and digested with trypsin. The resulting peptides were analyzed on a Q Exactive mass spectrometer (Thermo Fisher Scientific) coupled with an Easy-nLC1000 liquid chromatography system (Thermo Fisher Scientific). Peptides were loaded on a pre-column (75 mm ID, 6 cm long, packed with ODS-AQ 120 Å À10 mm beads from YMC Co., Ltd.) and separated on an analytical column (75 mm ID, 13 cm long, packed with Luna C18 1.8 mm 100 Å resin from Welch Materials) using a 78 min acetonitrile gradient from 0% to 30% (v/v) at a flow rate of 200 nl/min. The top 15-most intense precursor ions from each full scan (resolution 70,000) were isolated for HCD MS2 (resolution 17,500; NCE 27), with a dynamic exclusion time of 30 s. We excluded the precursors with unassigned charge states or charge states of 1+, 7+, or >7+ . Database searching was done by pFind 3.1 (http://pfind.ict.ac.cn/) against the C. elegans protein database (UniProt WS235). The filtering criteria were: 1% false discovery rate (FDR) at both the peptide level and the protein level; precursor mass tolerance, 20 ppm; fragment mass tolerance, 20 ppm; the peptide length, 6-100 amino acids. Spectral counts of HDA-1 in IP samples were used for normalization. Proteins either absent in N2 or with more than twofold of the spectra counts in the hda-1::gfp IP compared to those in N2 are shown in Supplementary file 2.

Silver staining
One-third of each sample from each IP-MS experiment was fractionated on a 4-12% SDS-PAGE gel. Gels were silver stained using ProteoSilver Plus Silver Stain Kit (Sigma-Aldrich). Visible bands were cut for trypsin digestion and MS identification, and the most abundant protein in each band is labeled in Figure 3A.

RNAi inheritance
L1 oma-1::gfp larvae were placed on IPTG plates (NGM plate with 1 mM IPTG and 100 mg/ml ampicillin) seeded with E. coli strain HT115 transformed with either control vector L4440 or gfp RNAi plasmid. The worms were scored for OMA-1::GFP signal in the oocyte at gravid adult stage and transferred to regular NGM plates. OMA-1::GFP was monitored at gravid adult stage for every generation until most worms recovered the expression of oma-1::gfp.

Germline mortal assay
For each strain, 18 worms at L4 stage were singled and grown at 25˚C. For each generation, the mothers were transferred to new plates every two days and their brood sizes were scored. One worm from each plate for every generation was randomly picked to continue scoring the brood size until plates became totally sterile.

Gonad mRNA-seq and analysis
For every mutant, about 100 gonads were dissected from the young adult worms (first day as adults). We carefully cut near the À1 oocyte to make sure every gonad is similar and transfer gonads into a 1.5 ml microcentrifuge tube. RNA was extracted from dissected gonads using Tri-reagent with a yield about 0.5 mg. Ribosomal RNA was depleted by RNase H digestion after being annealed with home-made anti-rRNA oligos for C. elegans. DNA was removed by DNase treatment. RNA-seq libraries were constructed using a KAPA RNA HyperPrep kit, and paired-end sequencing was performed on a Nextseq 500 Sequencer with the illumina NextSeq 500/550 high-output kit v2.5 (150 cycles).
Salmon was used to map the mRNA-seq reads with the worm database WS268, and its output files were imported to DESeq2 in R. The differentially expressed genes were filtered by fold change more than 2 and adjusted p-value <0.05. The scatter plots were generated by the plot function in R.

CHIP-seq
CHIP was carried out using a previous described protocol (Askjaer et al., 2014). Young adult worms were washed three times with M9 and once with PBS and then cross-linked with 1.1% formaldehyde in phosphate-buffered saline (PBS) with protease inhibitors for 10 min before being quenched with 125 mM glycine. Exchanging to CHIP Cell lysis buffer (20 mM Tris-Cl, 85 mM KCl, 0.5% NP40, pH 8.0), DNA was fragmented by sonication (Bioruptor, high intensity, 30 s on and 30 s off for 45 cycles). Samples were incubated with antibodies overnight at 4˚C, and then with magnetic beads, which were precleaned with 5% bovine serum albumin (BSA) in PBS + 0.02% Tween-20 (PBST) for 4 hr. After a series of washes in buffers of different stringency, DNA was eluted with CHIP elution buffer (1% SDS, 250 mM NaCl, 10 mM Tris pH 8.0, 1 mM EDTA) at 65˚C for 15 min twice. RNA and proteins were removed by RNase A and Proteinase K treatments. Samples were reverse cross-linked by incubating at 65˚C, and the DNA was purified using a Zymo DNA clean kit (cat # D5205). Libraries were prepared with NEBNext Ultra II DNA Library Prep Kit. Libraries were multiplexed and sequenced on HiSeq X or NovaSeq 6000 with paired-end 150 bp sequencing. With a pair-wised IP sample and its input, a ChIP-Seq Pipeline in the DolphinNext platform (built by the Bioinformatics core at UMass Medical School) was used to analyze the CHIP-seq data. The data processing and analysis pipeline includes adapter removal (cutadapt), reads mapping (Bowtie2-align-s v2.2.3), duplicates removal (Picard-tools v1.131), peak calling (MACS2 v2.1.1.20160309), peak location, and quantification (Bedtools v2.25.0, Quinlan and Hall, 2010). The output bed files were used to generate figures with IGV (v2.7.2) for Figure 4D. Background subtraction was applied with the intersect function of Bedtools for

Small RNA cloning and data analysis
The small RNA cloning was conducted as the previous report . Worms were synchronized and collected at young adult stage. Small RNAs were enriched using a mirVana miRNA isolation kit (Invitrogen) from Trizol purified total RNA. Homemade PIR-1 was used to remove the d or triphosphate at the 5 0 to generate 5 0 monophosphorylated small RNA. Products were then ligated to 3 0 adaptor (/5rApp/TGGAATTCTCGGGTGCCAAGG/3ddC/) by T4 RNA ligase 2(NEB) and 5 0 adaptor (rGrUrUrCrArGrArGrUrUrCrUrArCrArGrUrCrCrGrArCrGrArUrCrNrNrNrCrGrArNrNrNrUrA rCrNrNrN, N is for a random nucleotide) by T4 ligase 1(NEB) sequentially, followed by reverse transcription with RT primer (CCTTGGCACCCGAGAATTCCA) and SuperScript III (Invitrogen). PCR amplification was done with Q5 and primers with indexes (forward: AATGATACGGCGACCACC-GAGATCTACACGTTCAGAGTTCTACAGTCCGA, reverse: CAAGCAGAAGACGGCATACGAGAT [6bases index] GTGACTGGAGTTCCTTGGCACCCGAGAATTCCA). PCR productions around 150 bp were separated by 12% SDS-PAGE, extracted with TE buffer (10 mM Tris-HCl, 0.1 mM EDTA, pH 8.0), and purified with isopropanol precipitation. Libraries were equally mixed and sequenced on a NextSeq 550 sequencer using the Illumina NextSeq 500/550 high-output kit v2.5 (75 cycles) with 75 bp single-end sequencing. Adapters were trimmed by cutadapt and reads were mapped to a worm database (WS268) using Bowtie2. DESeq2 was used to normalized reads between samples. The following datasets were generated: