SETDB1-like MET-2 promotes transcriptional silencing and development independently of its H3K9me-associated catalytic activity

Transcriptionally silenced heterochromatin bearing methylation of histone H3 on lysine 9 (H3K9me) is critical for maintaining organismal viability and tissue integrity. Here we show that in addition to ensuring H3K9me, MET-2, the Caenorhabditis elegans homolog of the SETDB1 histone methyltransferase, has a noncatalytic function that contributes to gene repression. Subnuclear foci of MET-2 coincide with H3K9me deposition, yet these foci also form when MET-2 is catalytically deficient and H3K9me is compromised. Whereas met-2 deletion triggers a loss of silencing and increased histone acetylation, foci of catalytically deficient MET-2 maintain silencing of a subset of genes, blocking acetylation on H3K9 and H3K27. In normal development, this noncatalytic MET-2 activity helps to maintain fertility. Under heat stress MET-2 foci disperse, coinciding with increased acetylation and transcriptional derepression. Our study suggests that the noncatalytic, focus-forming function of this SETDB1-like protein and its intrinsically disordered cofactor LIN-65 is physiologically relevant.

T he proper segregation of active and inactive regions of the genome during differentiation is critical for the establishment and maintenance of tissue and genome integrity in eukaryotes. Heterochromatic regions of the genome are condensed and sequestered at the nuclear periphery or around nucleoli and are typically not transcribed 1,2 . Constitutive heterochromatin is associated with H3K9me, which is necessary for the repression of satellite repeats and transposable elements 3 . Tissue-specific genes can also be repressed by H3K9me (refs. [4][5][6]. Finally, H3K9me mediates chromatin anchoring at the nuclear periphery 7 and helps prevent interaction between euchromatic and heterochromatic compartments 8,9 . In mammals, the loss of H3K9me is associated with aging, cancer and the impaired maintenance of tissue integrity [10][11][12][13] . The enzymes that establish and maintain heterochromatin form subnuclear clusters or foci in many species [14][15][16][17][18][19] . In budding yeast, a function has been attributed to these stable perinuclear foci containing the Silent Information Regulator complex: the clustering promotes histone deacetylation and contributes to the stability and spread of repressed chromatin 19 . In mammals, the H3K9 histone methyltransferases (HMTs) SUV39H and SETDB1 remain bound to clusters of heterochromatin throughout the cell cycle 20,21 . In the case of SUV39H, this localized accumulation of the HMT requires an N-terminal chromodomain, which recognizes H3K9me2/me3 and thereby sequesters the enzyme at its chromatin targets 21 . To date, however, no function was assigned to the localized retention of H3K9 HMTs.
Recently, heterochromatic focus formation has been proposed to depend on disordered protein domains and multivalent interactions that promote local phase separation, as exemplified by the H3K9me reader HP1 protein 22,23 . In Caenorhabditis elegans, however, foci of MET-2, the SETDB1 homolog responsible for H3K9me deposition, and its disordered domain cofactor LIN-65, remain stable despite downregulation of the worm HP1 homologs 14 . Moreover, MET-2, unlike SUV39H, has no chromodomain. We therefore set out to examine what drives the focal clustering of heterochromatin proteins such as MET-2, and the relationship of such foci to histone H3K9me.
Although several C. elegans proteins have been implicated in methylation of H3K9 (refs. [24][25][26][27], only the double deletion of genes encoding two SET domain-containing proteins, MET-2 and SET-25, eliminates all detectable H3K9me, both in embryos 6,7,28 and at later stages of somatic development 6,7,29 . The loss of H3K9me compromises the transcriptional repression of tissue-specific genes and repetitive elements 6,30 , although no other common histone methylation marks were altered upon met-2 and set-25 ablation 7 . The SET domain of MET-2 has 50% identity to that of SETDB1 (refs. 18,28,31,32 ), which has been shown to methylate histone H3K9 both in vivo 33 and in vitro 34,35 . Moreover, MET-2 harbors a domain organization very similar to SETDB1, including a methyl CpG binding domain, Pre-SET domain and bifurcated SET domain at the C terminus (Fig. 1a). Nonetheless, the demonstration of HMT activity for recombinant MET-2 has been hindered because the purification of this large, partially unstructured protein with its cofactor, LIN-65, has proven difficult. We will therefore refer to MET-2 as a putative H3K9 HMT.
In cells, MET-2 and its two cofactors, the intrinsically disordered protein 18 ), and the poorly characterized but conserved ARLE-14, form foci that coincide with H3K9 methylation 14,18 . Not only the MET-2 partners, but also their regulatory interactions, are conserved in flies (Windei and CG14464 bind dSETDB1) 36,37 and mammals (ATF7IP and ARLE14EP bind SETDB1) [38][39][40] . Ablation of lin-65 (or atf7ip) leads to a loss of MET-2 (or SETDB1) in nuclei, respectively 14,18,41 , and in worms, lin-65 deletion compromises the formation of MET-2 foci and derepresses a large subset of MET-2-sensitive targets 14,18 . Consistent with the notion that the unstructured domains present in LIN-65 contribute to MET-2 focus formation, these foci are sensitive to interventions that disrupt phase-separated condensates, such as treatment with 1,6-hexanediol and heat 14 .

SETDB1-like MET-2 promotes transcriptional silencing and development independently of its H3K9me-associated catalytic activity
Here we examined the mechanisms of MET-2-mediated heterochromatin silencing and its impact on development and fertility. Unexpectedly, we discovered a function for MET-2 protein that operates alongside its role in H3K9 methylation and correlates with focus formation. Restoring foci with a catalytic-deficient MET-2 maintained low acetylation levels on heterochromatin and rescued somatic developmental rates, mitigating inappropriate transcription and infertility. This was achieved in an ARLE-14-dependent manner. We argue that MET-2 promotes gene repression through both catalytic and noncatalytic functions, the latter coinciding with heterochromatic MET-2 focus formation.
Loss of lin-65 impairs both MET-2 nuclear accumulation and concentration into subnuclear foci (Fig. 1c) 41 , and thus largely phenocopies met-2 deletion 14,18 . To distinguish roles of nuclear accumulation from focus formation, we introduced a nuclear localization signal (NLS) to the endogenously tagged met-2::mCherry locus by CRISPR (NLS::met-2). The NLS::MET-2::mCherry nucleoplasmic signal was stronger than the WT MET-2::mCherry that traffics between the cytosol and nucleus 14,18 , but in both cases foci were clearly distinguishable (Extended Data Fig. 1a). Whereas the ablation of lin-65 eliminated the nuclear enrichment of WT MET-2::mCherry, the NLS::MET-2 fusion protein remained nuclear but failed to form foci in the absence of LIN-65. Western blotting confirmed that H3K9me2 levels were not altered by the NLS tag on MET-2 as long as LIN-65 was present, and dropped similarly upon deletion of lin-65 (Extended Data Fig. 1b). We conclude that nuclear accumulation of MET-2 is not sufficient for H3K9 HMT activity and focus formation, and that LIN-65 promotes both.
In contrast to LIN-65, ARLE-14 is not required for general silencing of repetitive elements 14,18 , and its loss compromised neither the nuclear import of MET-2 nor the number of foci formed (Fig. 1c). Nevertheless, embryos carrying either arle-14Δ or two other independent arle-14 mutant alleles lost a majority of H3K9me2, as shown by quantitative fluorescence and western blots (Extended Data Fig. 1c,d). The impacts of lin-65Δ and of arle-14Δ on H3K9me2 levels in embryos were similar. Whereas we previously found that MET-2 stability is reduced in embryos lacking lin-65 (ref. 14 ), the loss of arle-14 did not reduce MET-2::mCherry levels (Extended Data Fig. 1e). These data argue that both ARLE-14 and LIN-65 promote MET-2 methylation activity, while only LIN-65 is essential for MET-2 focus formation.
Because the loss of H3K9me is associated with a failure to silence heterochromatin 5,6,33,44 , we examined gene expression in arle-14Δ using RNA sequencing (RNA-seq). Remarkably, only 14% of the genes derepressed by loss of met-2 (met-2Δ genes) were upregulated in arle-14Δ (Fig. 1d,e), while 59% of the met-2Δ-derepressed genes were upregulated in lin-65Δ. These data suggested that residual low levels of H3K9me2 in combination with the preserved accumulation of MET-2 into foci (promoted by LIN-65 in the arle-14 mutant) restricts gene derepression.
Because arle-14 and lin-65 mutants had different effects on transcription and focus formation, we performed epistasis analysis for the two genes, monitoring H3K9 methylation. We found that the lin-65Δ;arle-14Δ double mutant displayed a greater loss of H3K9me2 than either single mutant alone, achieving the same drop in methylation as in met-2Δ embryos (Fig. 1f). Consistent with the fact that ARLE-14 and LIN-65 interact with MET-2 independently of each other, we conclude that they function in parallel to promote MET-2's role in H3K9 methylation.
C. elegans animals lacking all H3K9me reproduced efficiently when cultivated at 20 °C or below, yet the loss of met-2 strongly accentuated the reduced fertility of worms at 25 °C, coincident with elevated levels of p53-dependent germline apoptosis (refs. 6,30 ). This loss of germline arises largely from the formation of R-loops at repeat elements, which are promiscuously transcribed upon loss of met-2 (refs. 6,30 ). To test whether ARLE-14 and LIN-65 differentially affect worm fertility, we quantified the number of progeny in arle-14Δ and lin-65Δ mutants. We observed an enhanced loss of viable progeny in both met-2Δ and lin-65Δ animals at 25 °C, while the arle-14Δ population was only modestly decreased in comparison with WT worms, despite having a near equal drop in H3K9me2 ( Fig. 1g and Extended Data Fig. 1f). The striking difference in the impact of lin-65Δ and arle-14Δ on fertility was substantiated independently at 20 °C (ref. 14 , and see Fig. 4a below). The residual fertility in arle-14Δ animals was not dependent on SET-25, which is required for all H3K9me3 in embryos 7,28,30 , as the arle-14Δ set-25Δ double mutant was no less fertile than arle-14Δ alone (Fig. 1g).
We note that in both lin-65Δ and arle-14Δ embryos substantially more H3K9me2 is retained than in met-2Δ animals ( Fig. 1f and Extended Data Fig. 1d), yet lin-65Δ phenotypes parallel those of met-2Δ more closely than those of arle-14Δ. This suggested that H3K9me2 and met-2Δ might not correlate strictly for all phenotypic read-outs, raising the possibility that MET-2 has functions beyond H3K9me deposition. These functions nonetheless require LIN-65 and its incumbent focus formation.

Catalytically deficient MET-2 attenuates defects of met-2Δ.
To test directly whether MET-2 has a function beyond ensuring H3K9me deposition, we used CRISPR-mediated mutagenesis to compromise the active site of MET-2's SET domain. The C1237A mutation met-2(gw1660) (hereafter met-2-CD) alters a highly conserved motif that is essential for the catalytic activity of SET domain proteins generally [45][46][47] (Fig. 2a), and specifically of SETDB1 (ref. 48 ). Consistently, we find that H3K9me2 and H3K9me3 levels dropped to the same level in met-2-CD as in met-2Δ ( Fig. 2b-d and Extended Data Fig. 2a,b). Moreover, H3K9me2 levels were as low in met-2Δ set-25Δ as in met-2-CD set-25Δ double mutants (Fig. 2c). Mass spectrometry suggested that H3K9me1 might not be completely abolished in the met-2-CD strain (Extended Data Fig. 2b), although this potential difference in H3K9me1 could not be detected by western blot (Extended Data Fig. 2a), nor was met-2-CD sufficient to preserve H3K9me1/2/3-dependent perinuclear anchoring of a well-characterized heterochromatic reporter (Extended Data Fig. 2c) 7,49 . Taken together, we conclude that met-2-CD is catalytically deficient for H3K9 methylation yet does not compromise MET-2 protein levels (Extended Data Fig. 2d) or MET-2's ability to bind ARLE-14 or LIN-65 (Extended Data Fig. 2e). This mutant made it possible to monitor H3K9me2-independent functions of MET-2.
We introduced the met-2-CD mutation (C1237A) into the met-2::mCherry locus to track catalytically deficient MET-2 in live embryos. We found that MET-2-CD forms foci, and the number of foci did not differ from WT MET-2 (Fig. 2e). Both WT and MET-2-CD foci were dependent on LIN-65, and thus we concluded that MET-2 foci can form independently of HMT catalytic activity. This allowed us to distinguish the physiological effects arising from a loss of H3K9me2 from those caused by the dispersion of MET-2 foci. Because fertility is sensitive to the loss of H3K9me, we next scored progeny in met-2-CD animals. Despite being H3K9me-deficient, the met-2-CD population retained higher levels of fertility than met-2Δ animals, and this was independent of set-25 (Fig. 2f). Previous work has shown that met-2Δ animals also display a stochastic delay in development 30 , which we quantified by scoring the number of L1 larvae that develop to the L4 larval stage by 42 h in a synchronized culture at 20 °C. In contrast to met-2Δ, met-2-CD animals had no notable developmental delay (Fig. 2g). Moreover, the rescue of development timing was reduced upon lin-65 ablation (Fig. 2g). In conclusion, a LIN-65-dependent function of MET-2-CD is important for normal development in both somatic and germline tissues, suggesting a role for MET-2 foci that is independent of MET-2's HMT activity.
MET-2-CD can repress heterochromatic genes. It was unclear whether MET-2-CD would be sufficient to preserve gene repression, despite the drop in H3K9me2 found in this mutant. We performed transcriptome analysis in met-2-CD, met-2Δ and WT early embryos. Importantly, most of the genes upregulated in met-2Δ were also upregulated in met-2-CD, highlighting the importance of H3K9me in maintaining repression (Fig. 3a, 'common' genes, N = 1225). However, there were 271 met-2Δ targets that were upregulated in met-2Δ but not in met-2-CD ('met-2Δ-specific' genes) and 179 genes that exhibited a mitigated response (derepression was more than 2-fold lower in met-2-CD than in met-2Δ, 'met-2Δ ≫ met-2-CD' , Fig. 3b). In other words, 28% of met-2 target genes retained complete or partial repression in met-2-CD embryos.
Heterochromatin is not only marked by H3K9me but is also characterized by a general hypoacetylation of histone tails. In contrast, gene activation is associated with increased histone tail acetylation. Intriguingly, mass spectrometry showed a global increase in H3K9 acetylation in met-2Δ embryos that was absent in met-2-CD embryos (Extended Data Fig. 2b). We confirmed this result by western blot and immunofluorescence, showing that both histone H3 and H4 acetylation (Extended Data Fig. 3a), including H3K9ac and H3K27ac (Fig. 3c,d), were increased in met-2Δ embryos, but not in met-2-CD embryos. These results suggested that the aberrant gain of acetylation in met-2Δ animals is not an inevitable consequence of H3K9me loss.
To see which loci gain histone acetylation in met-2 mutants, we performed Chromatin immunoprecipitation followed by genome-wide sequencing (ChIP-seq) in early embryos. We concentrated on H3K27ac due to its well-described role in transcriptional activation and enhancer function 50 . Relative to WT embryos, we defined 565 differentially enriched H3K27ac peaks in met-2Δ, while only 38 changes were detected in met-2-CD (Fig. 3e,f). All but one of the rare met-2-CD peaks were also observed in met-2Δ embryos (Fig. 3g). In addition, we examined the enrichment for pan-histone H3 acetylation (H3ac) at a subset of met-2Δ-specific genes by Chromatin Immunoprecipitation followed by quantitative real-time PCR (ChIP-qPCR). Compared with WT, these loci were enriched for H3ac in met-2Δ but not in met-2-CD embryos (Fig. 3h), whereas for common targets or a nonrepressed control, H3ac enrichment was very similar in met-2Δ and met-2-CD strains (Fig. 3h). The lack of acetylation in the met-2-CD embryos was not linked to residual H3K9me1, as the two mutants (met-2-CD and met-2Δ) had equally low levels of H3K9me1 at these loci (Extended Data Fig. 3b). Instead, the data suggested that MET-2 contributes to transcriptional repression and blocks the acetylation of specific loci, independent of its HMT activity.

ARLE-14 promotes MET-2 association with chromatin.
After determining that catalytic-dependent and -independent roles for MET-2 exist in gene repression, we sought to see if ARLE-14 plays a role in the noncatalytic MET-2 function. If ARLE-14 simply promotes MET-2 HMT activity, met-2-CD would be epistatic with arle-14. We examined fertility at both 20 °C and 25 °C in arle-14Δ met-2-CD double mutant animals. Remarkably, at 20 °C neither arle-14Δ nor met-2-CD single mutants had significantly fewer progeny than WT animals (Fig. 4a). However, the arle-14Δ met-2-CD population showed a synthetic loss of viable offspring at both temperatures, phenocopying met-2Δ mutants ( Fig. 4a and Extended Data Fig. 4a). This was surprising, because arle-14Δ did not affect foci of WT MET-2 ( Figs. 1c and 2e). Nonetheless, we found that arle-14 was required both for MET-2-CD::mCherry foci number and the overall nuclear abundance of the HMT (Fig. 4b-d), an effect that was enhanced at 25 °C (Extended Data Fig. 4b,c). These data argued that ARLE-14 acts redundantly with H3K9me to stabilize MET-2 in foci.
Although the function of worm ARLE-14 is unclear, recent evidence in human cells suggested that the human homolog, ARL14EP, may facilitate SETDB1 interaction with chromatin 43 . To see if ARLE-14 promotes MET-2 binding to chromatin, we performed ChIP-qPCR using antibodies against MET-2::mCherry in both WT and met-2-CD backgrounds, both in the presence and absence of ARLE-14. MET-2-CD showed similar enrichment as WT MET-2 at five of seven tested targets (Fig. 4e). Loss of arle-14 showed little effect on WT MET-2::mCherry enrichment at most MET-2 gene targets (with the exception of col-166), but arle-14 did lead to the loss of MET-2-CD at these same sites (Fig. 4e). Consistently, the loss of ARLE-14 in met-2-CD embryos led to the upregulation of met-2Δ-specific targets based on RNA-seq data, which was not observed in arle-14Δ alone (Fig. 4f,g). Taken together, these data suggested that ARLE-14 acts redundantly with H3K9me to stabilize both MET-2 foci and its association with MET-2-controlled genes. Thus, arle-14 is needed for the noncatalytic function of MET-2 on both physiological and molecular levels.  2Δ N = 2). ChIP-qPCR, Chromatin Immunoprecipitation followed by quantitative real-time PCR; *P < 0.01, **P < 0.001, ****P < 0.00001. Acute heat stress dissociates MET-2 foci. Having shown by genetics that MET-2 has a noncatalytic function that correlates with its concentration into foci, we wondered whether foci dynamics might be physiologically relevant in animals containing normal levels of H3K9me. We previously showed that MET-2 foci disperse upon transient exposure to 37 °C heat shock (HS) (Fig. 5a and Extended  Fig. 5a) 14 , coincident with the canonical heat shock response (HSR) 51 . Intriguingly, although the HSR is clearly associated with well-described signaling pathways that dampen transcription while specifically upregulating protein chaperone genes 52,53 , earlier studies in other species also linked acute heat stress to the transcriptional derepression of silent chromatin [54][55][56][57][58] . Thus, HS conditions provided a physiological system in which we could monitor a transcriptional response that coincides with the induced dispersion of MET-2 foci, without genetic alteration. We performed RNA-seq on early embryos exposed to either normal temperature (20 °C) or HS (1 h at 37 °C). We observed a broad upregulation of genes at 37 °C relative to embryos kept at 20 °C (2026 genes up, 314 genes down, fold change (FC) > 4, false discovery rate (FDR) < 0.01, Fig. 5b). Consistent with the dispersion of MET-2 foci, we found that most HS-upregulated genes, similar to met-2Δ-upregulated genes, had very low expression levels in embryos cultured at 20 °C (80% reads per kilobase million < 4 at 20 °C Fig. 5b). Strikingly, we observed a positive correlation (r = 0.65) between the transcriptional changes induced by 37 °C versus those induced by the loss of met-2 at 20 °C versus WT (20 °C) embryos ( Fig. 5b-d). In fact, 76% of the met-2Δ target genes were upregulated under HS conditions, constituting 56% of all HS-derepressed genes (Fig. 5c). The majority of the shared targets were genes that are normally expressed in postembryonic, differentiated tissues (Extended Data Fig. 5b). These data suggest that embryos exposed to HS not only lose MET-2 foci but also derepress many genes that are normally repressed by MET-2 (Fig. 5c,d). On the other hand, met-2Δ embryos respond to HS similarly to WT animals with respect to global gene expression, HSP gene induction and HSF-1 foci formation (Extended Data Fig. 5c-f).
Because met-2Δ mutants present enhanced phenotypes at 25 °C (refs. 6,30 ), we wondered whether the shift from 20 °C to 25 °C also induces a similar stress response as exposure to 37 °C. However, no HSF-1 foci formed in either WT or met-2Δ embryos at 25 °C, nor did the transcriptome of animals grown at 25 °C correlate with the pattern of genes induced by HS (Extended Data Fig. 5f,g). Thus, the loss of fertility at 25 °C is not a heat shock response.
Given that the loss of MET-2 foci correlates with a loss of H3K9me (Extended Data Fig. 1b), we examined how H3K9me levels change during 30 min at 37 °C. Indeed, embryos exposed to HS displayed a drop in H3K9me2 to 45% or 50% of untreated levels, as measured by western blot or immunofluorescence, respectively (Fig. 5e,f). Whereas the met-2 messenger RNA level upon HS was unchanged (log 2 FC = −0.094), we scored a ~24% reduction in MET-2 protein level (Extended Data Fig. 6c), which may influence chromatin states. We examined regions that bear H3K9me2 in embryos at 20 °C by ChIP-seq, to see if H3K9me2 is also reduced in a localized manner after exposure to 37 °C for 1 h. Indeed, we observed a general reduction in read counts for H3K9me2 to 63% of level scored at 20 °C (Extended Data Fig. 6a).
Whereas the transcriptional induction following HS was stronger than what we observed in met-2Δ, the loss of H3K9me2 was less. Given our earlier data that link the loss of MET-2 foci with increased acetylation levels, we checked to see if histone acetylation changes during HS. Western blots showed that HS-treated embryos have a substantial increase in pan-H3ac, including increases in H3K9ac and, more modestly, H3K27ac (Fig. 5g). Taken together, our results argue that HS-induced dispersion of MET-2 foci coincides with a partial destabilization of H3K9me2, a strong gain of H3ac and widespread transcriptional activation of loci silenced by MET-2 under normal conditions.
To compare the changes during HS to those incurred by met-2Δ, we examined changes in H3K27ac in WT embryos (HS versus 20 °C) by ChIP-seq, and compared these peaks with data from our met-2Δ and met-2-CD animals. HS (30 min at 37 °C) led to an enrichment of H3K27ac on 1646 peaks (Fig. 5h). Importantly, 49.7% of the H3K27ac peaks in met-2Δ were shared with those in HS embryos (Fig. 5i). Nonetheless, there were three times as many peaks after HS as in unstressed met-2Δ embryos (1646 in HS versus 565 in met-2), consistent with the idea that HS mobilizes a broad transcriptional response, as well as destabilizing MET-2 foci. Other regions of the genome of HS-treated embryos actually lost H3K27ac, reflecting the fact that a certain degree of transcriptional downregulation occurs during the canonical HSR 59 . The sites that gained acetylation in both HS and met-2Δ embryos were mainly promoters and repeats, while met-2-CD embryos showed almost no gain in acetylation over WT (Extended Data Fig. 6b). Intriguingly, the met-2 target genes that remain fully or partially repressed in the met-2-CD background (presumably because they remain in foci) showed particularly pronounced derepression during HS-induced loss of focus formation (Extended Data Fig. 6d,e). In other words, the dissolution of MET-2 foci upon HS is coupled with increased acetylation and gene derepression, beyond the program induced by HSF-1, reinforcing the correlation we found between MET-2-CD foci and transcriptional repression.

Discussion
The sequestration and self-association of heterochromatin into foci has been observed since the early days of microscopy 1 . While these foci coincided with HP1 and with H3K9me-specific HMTs, it was unclear whether the localized concentration of these proteins simply enhanced the propagation of H3K9 methylation, or served another role. Using a mutant of a key residue in the SET domain of the SETDB1-like enzyme MET-2 (met-2-CD), we have been able to distinguish a repressive role for MET-2 focus formation distinct from the catalytic function of MET-2's SET domain. H3K9me is clearly the key regulatory mark for the stable silencing of lineage-specific genes and repetitive elements across animal species, and the majority of loci repressed by heterochromatin in worms depend on the activity of MET-2's SET domain. Yet the careful analysis of the met-2-CD allele has revealed a second, noncatalytic role for MET-2 that requires its binding partners LIN-65 (H.s. ATF7IP) and ARLE-14 (H.s. ARL14EP), and which correlates with the accumulation of MET-2 and its ligands in subnuclear foci (Fig. 6).
LIN-65 is necessary for focus formation, while ARLE-14 may stabilize MET-2 interaction with chromatin within foci, acting redundantly with H3K9me with respect to MET-2 recruitment and focus formation (ref. 14 ) (Fig. 6). Foci of MET-2-CD appear to be sufficient to block local acetylation and mitigate inappropriate transcription of selected developmentally regulated genes, in an ARLE-14-dependent manner. We infer that MET-2 foci not only correlate with efficient H3K9me, but they create a semirepressed state that is refractory to histone acetylation. Consistent with the repressive function, MET-2-CD was able to largely preserve fertility and somatic development in the absence of MET-2's catalytic activity under normal growth conditions (20-25 °C). For this function, however, it requires ARLE-14.
The comparison of the met-2Δ and met-2-CD transcriptomes showed that even among the loci that were derepressed in both met-2Δ and met-2-CD embryos, the transcriptional levels of many genes were higher in met-2Δ. Genes that retained full or partial repression constitute 28% of met-2 target genes. This argues that H3K9me and MET-2's noncatalytic function silence genes in a partially additive manner, reinforcing each other to maintain heterochromatic repression.
The mechanism of focus-mediated repression appears to be largely structural: we found that MET-2-CD foci were able to block the increase in histone tail acetylation found in met-2Δ mutants. We note that similar to H3K9me, histone hypoacetylation is a hallmark of heterochromatin across eukaryotic species 60,61 . In Schizosaccharomyces pombe the histone deacetylases (HDACs) Clr3 and Sir2 are essential to restrict RNA Pol II access to heterochromatin 62,63 , and the treatment of murine cells with HDAC inhibitors resulted in a widespread loss of heterochromatic repression 64 and impaired the establishment of lineage-specific expression patterns in differentiating mouse embryos 65 . Finally, histone acetylation itself is known to promote chromatin decompaction, weaken the interaction between histones and DNA, and antagonize phase separation of chromatin arrays in vitro [66][67][68][69] . Therefore, MET-2-CD's ability to prevent hyperacetylation could explain how it maintains the semirepressed state that we observe by RNA-seq. MET-2-CD may block acetyltransferase access directly or recruit the factors, such as HDACs, that ensure that heterochromatin remains silent.
We cannot yet determine whether MET-2/LIN-65/ARLE-14 preserve hypoacetylation directly or indirectly. Although no HDACs were enriched in our MET-2 pulldown 14 (Fig. 1b), it was recently suggested that MET-2 can physically associate with HDA-1 (HDAC1) in a SUMOylation-dependent manner 70 . In addition to its physical interaction with LIN-65 and ARLE-14, and potentially HDAC1, MET-2 interacts genetically with the H3K4 demethylase SPR-5 (ref. 71 ), the DREAM complex 72 and SMARC-1 chromatin remodelers 73 . While we do not exclude it, the hypothesis that these factors serve as effectors of MET-2's noncatalytic function is made less likely by the fact that each of the others was shown to act either redundantly with MET-2 or through H3K9me. Alternatively, recent investigations into the function of subnuclear condensates suggest that these foci may physically impair the access of transcription activating complexes, such as RNA Pol II, Mediator, basal transcription factors or histone acetyltransferases 74 . Because foci of MET-2 and its cofactors share properties of condensates, they may also block recruitment of the transcription machinery and thereby inhibit the activation and acetylation of silenced genes, despite reduced levels of H3K9me.
We note that we were unsuccessful in expressing recombinant MET-2 or LIN-65, potentially due to their large size and highly disordered domains 75,76 . This precluded in vitro tests with recombinant protein for direct binding and/or HMT activity, which would complement the extensive genetic and biochemical evidence implicating MET-2 as the primary H3K9me1/me2 HMT in C. elegans embryos 7,15,28,30,32,71,[77][78][79][80][81][82] . Nevertheless, we were able to demonstrate H3K9me activity of MET-2 recovered in a pull-down assay. We note that MET-2-CD has the same protein stability, localization, dynamics and protein interactions as WT MET-2, and yet H3K9me2 levels drop strikingly. This supports a direct role of MET-2 catalytic activity in H3K9me deposition, and a direct role for MET-2-CD in a secondary repressive function associated with LIN-65 dependent focus formation.
We find that ARLE-14 becomes essential to stabilize the MET-2-CD interaction with chromatin and its accumulation in foci, when catalytic function is compromised (Fig. 4). Consistently, the ARLE-14 homolog ARL14EP has been shown in vitro to bind H3 peptides in a SETDB1-independent manner 43 . We therefore propose that ARLE-14 helps recruit MET-2 to chromatin, acting independently of the H3K9me mark. It is possible that ARLE-14 has other roles in gene activation that are independent of MET-2, or else that ARLE-14 also prevents MET-2 mistargeting, given that some genes actually acquire, rather than lose, repression upon loss of arle-14 (Fig. 4f).
Our study is not the first to suggest noncatalytic functions for a diverse group of HMTs. In mammalian cells, the H3K9 HMT G9a is able to promote repression by recruiting DNA methylation enzymes, independently of its catalytic activity 44,83 . Similarly, mammalian MLL3 and MLL4 do not require the enzymes' H3K4 methylation activity to maintain enhancer activation in mice 84 . We speculate that in the case of MET-2, the noncatalytic function may allow for multiple layers of repressive control that respond to different stimuli and react with different kinetics, an example of which is the cellular response to heat stress (HS, 30 min to 1 h at 37 °C). HS resulted in hyperacetylation and activation of silenced genes, coincident with MET-2 focus dispersion, even though the    Fig. 5h and Fig. 3e relative to WT (20 °C) (FDR < 0.01). WT 37 °C versus met-2Δ: P < 2.2 × 10 −16 by two-sided Fisher's exact test. *P < 0.01, **P < 0.001, ***P < 0.0001, ****P < 0.00001.
loss of genome-wide H3K9me was less severe than that triggered by arle-14Δ or met-2Δ strains. The dissolution of MET-2 foci by HS illustrates a rapid transcriptional response that promotes chromatin opening while retaining H3K9me, potentially allowing a quick post-HS recovery of the prestress transcriptome.
In summary, we have uncoupled the effect of MET-2 concentration into subnuclear foci from its lysine methyltransferase activity. While H3K9me deposition is the primary driver of gene silencing, both H3K9me and MET-2 focus formation contribute to the repression of transcription and may mutually reinforce the heterochromatic state. Whereas MET-2 foci do not require the HP1-like factors HPL-1 and HPL-2 (ref. 14 ), they do depend on LIN-65, a protein rich in low-complexity sequence and disordered domains 14,18 . They are also stabilized by ARLE-14 in a manner redundant with H3K9 methylation, and they attenuate histone tail acetylation. This indicates that the SETDB1-like complex containing MET-2, ARLE-14 and LIN-65 contributes to gene repression beyond the targeted methylation of histone H3K9, with physiological consequences for organismal growth and fertility.