A Nuclear RNA Degradation Pathway Helps Silence Polycomb/H3K27me3-Marked Loci in Caenorhabditis elegans

LSM PROTEINS REDUCE mRNA FROM HETEROCHROMATIC REPORTERS SELECTIVELY

The initial screen for reporter derepression monitored GFP expression in embryos (Fig. 1B). The elimination of LSM proteins by lsm-2, lsm-5, or lsm-6 RNAi, on the other hand, was detectable throughout worm development—that is, in larvae and in adult worms (L4 larvae shown in Fig. 1C; see also Mattout et al. 2020). This was also true for other RNAi hits from the initial screen—namely, sams-3 RNAi and mes-4 RNAi (Fig. 1C). Elevated GFP expression from the heterochromatic array after RNAi treatment was scored at all stages of worm development and in nearly every somatic cell type, except germline cells (see below). Given that Piwi-interacting RNAs (piRNAs) are known to contribute to silencing uniquely in germline cells, we hypothesize that the effect of lsm gene down-regulation may be masked by redundant silencing pathways, like piRNAs, that act specifically in the gonad. Alternatively, the LSM-mediated silencing pathway may not be functional in germ cells.

To obtain a quantitative value for reporter expression, we monitored the increase in GFP fluorescence intensity using the COPAS BioSort, which measures fluorescence on a worm-by-worm basis for hundreds to thousands of worms, using flow cytometry. Using the worm sorter, we compared the effects of lsm-6, lsm-2/gut-2, and lsm-5 RNAi with the knockdown of further shared subunits of the two LSM complexes: LSM-3, LSM-4, and LSM-7 (Fig. 1D). The RNAi clone we used against lsm-3 had several predicted targets; therefore, the results were inconclusive (data not shown). However, in case of both lsm-4 and lsm-7 RNAi, a significant (P < 10⁻¹³) increase in average GFP fluorescence of twofold to fourfold was scored (see Mattout et al. 2020). The variable degrees of GFP derepression following RNAi against these LSM subunits likely stems from differences in RNAi efficiency.

Given that the LSM complex has been implicated in RNA processing (Kufel et al. 2004; Beggs 2005; Tharun 2009; Perea-Resa et al. 2012; Golisz et al. 2013; Cornes et al. 2015), it was possible that somatic cell derepression was specific to the reporter construct used in the screen. We ruled this out by monitoring the effects of lsm-2, lsm-5, or lsm-6 RNAi on the expression of four different heterochromatic reporters each integrated in the C. elegans genome as a large array (200–300 copies) at a single locus (Mattout et al. 2020). Each reporter carried a different promoter, gene, and 3ʹ UTR, yet all showed reporter up-regulation following lsm-2, lsm-5, or lsm-6 RNAi (Mattout et al. 2020). In contrast, two euchromatic reporters, which were single-copy transgenes integrated into a nonheterochromatic region of the genome, showed no change in expression under the same conditions (data shown for GW1108, eft-3p::gfp cb-unc-119(+) chr II in Fig. 1E). This was true for euchromatic reporters with both high and low levels of expression, indicating that LSM complex subunits are selectively involved in the repression of loci that have heterochromatic, but not euchromatic features. Derepression levels were comparable or greater than the loss of our positive control, MES-4, a histone H3K36 methyltransferase recovered in the original screen (Towbin et al. 2012), and persisted during somatic cell differentiation and in postmitotic cells.

We confirmed that the up-regulation of GFP in these worms indeed reflects changes on the mRNA level, rather than GFP protein synthesis or turnover. This was achieved by quantifying the reporter mRNA by quantitative polymerase chain reaction using reverse transcription (RT-qPCR) following RNAi against various LSM subunits, notably lsm-6 or lsm-7 down-regulation (Mattout et al. 2020).

LSM2-8 CONTRIBUTES TO HETEROCHROMATIC REPORTER SILENCING

Studies in yeast and plants have established that there are two LSM complexes (LSM1-7 and LSM2-8) (Fig. 2A) that share six subunits, whereas the seventh subunit in each complex, LSM-1 or LSM-8, participates only in one (Beggs 2005; Tharun 2009). The two complexes also interact with distinct cofactors. RNAi against lsm-1, one of the nonshared subunits, did not derepress the heterochromatic reporter nor did the depletion of the 5ʹ → 3ʹ exoribonucleases, XRN-1, which is reported to act together with LSM1-7 in the cytoplasm to mediate cytoplasmic RNA decay (Figs. 1D and 2A). On the other hand, RNAi against xrn-2, a nuclear RNase that is thought to degrade RNA together with LSM2-8 in budding yeast (Tharun 2009), did show significant heterochromatic reporter derepression (Fig. 1D). Null-allele worms lacking lsm-1 or the decapping factor dcap-2 also failed to show array derepression (Mattout et al. 2020). This argued against a role for LSM1-7 in heterochromatic reporter silencing and suggested instead that LSM2-8 and an evolutionarily conserved nuclear 5′ → 3′ exoribonuclease called XRN-2 (HsXRN2, SpDhp1, or ScRat1) are involved (Miki et al. 2016).

View larger version:

• In this window
• In a new window

Figure 2.

LSM-8 is required for heterochromatic reporter silencing, implicating the LSM2-8 complex. (A) The two main LSM complexes and functions (Beggs 2005; Tharun 2009). (B) Schematic view of the lsm-8 deletion/gene replacement created by CRISPR–Cas9. (C) Differential interference contrast (DIC) images of L1/L2 larvae (upper panel) and young adults (lower panel) together with dual-channel fluorescence for lsm-8^+/− (yellow pharynx in merge) and lsm-8^−/− (red pharynx only) worms. Red and green channels are merged to distinguish heterozygous from homozygous worms. The inset in the upper panel shows the protruding vulva of an lsm-8^−/− adult (marked with white triangle) and the accumulating vesicles (black triangles). The normal vulva is boxed in the lower panel. Bars = 100 μm. (D) Red and green fluorescence channels are merged and coupled with DIC images of one lsm-8^−/− worm and one lsm-8^+/− heterozygote bearing the pkIs1582 reporter (GW1120). Note that this strain has no nT1 balancer with the green fluorescent pharynx marker. GFP fluorescence in nuclei throughout the lsm-8^−/− worm is visible. Bar, 100 μm. (E) Merged DIC images with fluorescent images as in D. lsm-8^−/− mutant L4 larvae bearing pkIs1582 reporter is shown with fully developed gonad (encircled in red). The lsm-8^−/− mutant is 100% sterile. Somatic nuclei and somatic cells of the gonad show GFP expression from the heterochromatic array. Bar, 50 μm. (This image also appeared in Mattout et al. 2020. Reproduced with permission.) (F) Quantification of the length, width, and gonad nuclei count from the DAPI staining of L3 WT and lsm-8^−/− larvae. n = 8.

Because LSM-8, the only subunit unique for the nuclear LSM2-8 complex, was unresponsive to RNAi in our hands, it was essential to create a genomic deletion allele of the lsm-8 gene. A full lsm-8 deletion was generated by CRISPR–Cas9, and simultaneously we inserted a red fluorescent marker gene with pharynx-specific expression. We could track the null allele by fluorescence microscopy, but found that the homozygous lsm-8^−/− worms were 100% sterile. To propagate the lsm-8 deletion, the mutation was balanced with a green fluorescence–marked balancer nT1 [qIs51], which is also expressed in the pharynx. With dual fluorescence, we could distinguish offspring that are heterozygous for lsm-8^−/+ (i.e., that have red and green fluorescence in the pharynx) from those that are homozygous (lsm8^−/−; red pharynx only) (Fig. 2C). GFP fluorescence from the heterochromatic array was then monitored in red pharynx–marked homozygous, lsm-8^−/− animals with or without (Fig. 1D) the nT1 balancer. Indeed, deficiency for LSM-8 strongly derepressed the heterochromatic array, yielding GFP fluorescences at levels similar to those observed after lsm-2, lsm-5, lsm-6, or lsm-7 RNAi (Fig. 2D). Although mutant germline and gonad development up to L3 and L4 larval stages was similar to that in wild-type (N2) worms (Fig. 2E,F), there was no oocyte maturation in the lsm-8^−/− mutant. Nonetheless, we could monitor GFP derepression in all somatic tissues of larvae and adults, including the somatic cells of the gonad, but not in the germline itself (Fig. 2E).

Although the lsm-8^−/− animals can develop to adulthood, the adults have protruding vulva (see inset, Fig. 2C, white arrowhead), show empty cavities or vacuoles in differentiated tissues (black triangle), and die prematurely (Mattout et al. 2020). Other lsm mutants, such as lsm-2 and lsm-5, are phenotypically similar to lsm-8 mutants, with clear protruding vulva, the presence of vacuoles, and 100% sterility. Adult populations of the lsm-8 homozygous phenotype started dying after 6 days, whereas a similar loss of viability in wild-type worms would not occur before 10–12 days. Worms mutant for lsm-1 did not present these phenotypes (Cornes et al. 2015), confirming the distinct roles played by the two LSM complexes.

LSM2-8 IS REQUIRED TO MAINTAIN SILENT ENDOGENOUS HETEROCHROMATIN

Although LSM2-8 was clearly implicated in heterochromatic array silencing, it was important to establish whether or not it also controls the expression of endogenous transcripts. The data sets for total RNA-seq of WT and homozygous lsm8^−/− sorted L3 and L1 larvae are presented in Mattout et al. (2020) and data are deposited with the NCBI Gene Expression Omnibus (GSE92851). The lsm-8 data sets were compared with RNA-seq data from similarly sorted met-2^−/− set-25^−/− L3 larvae, which lack detectable H3K9 methylation (Towbin et al. 2012), and from L3 worms of the triple mutant, met-2^−/− set-25^−/−; lsm8^−/−, to determine if the LSM2-8 silencing pathways would be epistatic with the classic H3K9 methylation–mediated repression. In each case, the mutant and wild-type worms were sorted by fluorescence (nongreen pharynx indicating homozygosity) and by size, to generate a uniform population of L3 stage larvae. Mutant and wild-type L1 populations were also examined following sorting for L1 size larvae (Mattout et al. 2020). For the L3 populations, shifts in developmental timing arising during the sorting process were compensated by comparing matched replicates by making use of a characteristic, temporal fluctuation in gene expression characterized by the Großhans laboratory (Hendriks et al. 2014). With developmentally matched mutants and control larvae, we analyzed the RNA-seq data using edgeR, to determine the genes that changed significantly in each of three genetic backgrounds and WT strains (Mattout et al. 2020). Two or three analytical criteria were used, with the most stringent being the combined criteria of a false discovery rate (FDR) < 0.05 and a fold change (Fc) in average RNA level >4 (Mattout et al. 2020). In Figure 3A, we highlight genes with average altered expression rates of Fc > 4 only.

View larger version:

• In this window
• In a new window

Figure 3.

LSM2-8 silences hundreds of genes and 95% of these carry the Polycomb mark H3K27me3. (A) Deletion of lsm-8 (lsm-8^−/−) derepresses significantly more than 200 genes (Fc > 4). Relative gene expression profiles are shown as scatter plots, with fold change (Fc > 4) for two RNA-seq replicas of L3 sorted worms of the indicated genotype versus WT. Each dot corresponds to a gene. The original RNA-seq data are available at NCBI Gene Expression Omnibus number GSE92851. The same data set was analyzed in Mattout et al. (2020), but with (FDR < 0.05 and Fc > 4). (B) Scatter plot comparing the relative gene expression between the lsm-8 (x-axis) and the met-2 set-25 double mutant (y-axis). Common up-regulated genes are shaded yellow; 36% of genes up-regulated in the lsm-8 mutant (FDR < 0.05 and Fc > 4) are also up-regulated (FDR < 0.05 and Fc > 4) in the met-2 set-25 mutant. lsm-8^–/–-specific up-regulated genes are shaded in pink; met-2^−/− set-25^−/–-specific are in blue. (This panel is reproduced from Figure 3C in Mattout et al. 2020 with permission.) (C) LSM2-8-mediated silencing of the heterochromatic reporter is independent of H3K9 methylation. Quantitation of derepression of GFP expressed from the gwIs4 heterochromatic reporter in L1 progeny in WT and met-2 set-25 mutant genotypes, after control or lsm-7 RNAi. Quantitation of GFP derepression expressed from the gwIs4 heterochromatic reporter in L1 progeny in WT and met-2 set-25 mutant genotypes, respectively, from strains GW76 and GW637, after control or lsm-7 RNAi. P-values indicated; two-tailed unpaired t-test. Quantification and statistical analyses were based on n = 1460, 2399, 1593, and 1189 worms for conditions indicated from left to right, pooled from three independent experiments. N = 339, 1004, 426, 673 same after lsm-7 RNAi in the hpl-2^−/− mutant background. (D,E) Scatter plots compare the average gene expression changes in lsm-8^−/− worms (x-axis in log₂, RNA-seq L3 stage) versus enrichment for the indicated euchromatic or heterochromatic histone modification (y-axis in log₂, ModEncode data of WT L3 stage). Up-regulated genes (FDR > 0.05 and Fc > 4) in the lsm-8^−/− mutant are in red to the right of the black line, and genes enriched for the histone mark are above the red line (enriched over input). % indicates genes in upper right zone. (The panels in E are reproduced from Fig. 4B of Mattout et al. 2020, with permission.) (F) Scatter plot of absolute gene expression (log₂) of lsm-8^−/− versus WT. Red dots as in E and each dot corresponds to a gene. Values <6 (log₂) are considered as very low expression: 6 (log₂) corresponds to about 60 RNA-seq reads per gene, whereas 10 (log₂) corresponds to about 1000 reads per gene. Most lsm-8^−/− derepressed genes have low steady-state level expression in WT worms but moderate expression levels in the mutant (pink shading). (This panel is reproduced from Fig. 4C of Mattout et al. 2020, with permission.)

We found that the deletion of lsm-8 led to the strong and reproducible up-regulation of 122 genes (FDR < 0.05 and Fc > 4), with very few genes being down-regulated (Mattout et al. 2020). With less stringent cutoff values, even higher numbers of loci were selectively derepressed in lsm-8^−/− (Fig. 3A): Only FDR < 0.05 identified 147 genes up-regulated; only Fc > 4 identified 204 genes; and only Fc > 2 identified 1332 genes. Interestingly, a significant number of transcripts sensitive to lsm-8^−/− were involved in the innate immune response and the regulation of cell shape, but they did not reflect a single developmental stage. Genes derepressed in the mutant in L1 larvae overlapped only partially (5%–10% based on the thresholds used) with those derepressed in the L3 stage, and among those overlapping, many are involved in the innate immune response (Mattout et al. 2020). We confirmed the lsm 8^−/− larvae results by performing RNA-seq experiments on the lsm-4 mutant (Fig. 4A). The patterns of genes derepressed were overlapping but not identical, in lsm-4 and lsm-8 mutants (Fig. 4B). This is to be expected, given that lsm-4 also eliminates the activity of the LSM1-7 complex as well as the LSM2-8. Indeed, the distribution of LSM-4-GFP is diffuse in both cytoplasm and nucleus, and in the nucleus it appears to be excluded from foci that bind the H3K9me3 histone methyltransferase (HMT), SET-25 (labeled with mCherry, Fig. 4C).

View larger version:

• In this window
• In a new window

Figure 4.

lsm-8 and lsm-4 deletions derepress overlapping gene sets in L1 and L3 larvae. (A) Deletion of lsm-4 (lsm-4^−/−) leads to L1 larval arrest and derepresses more than 100 genes significantly. Relative gene expression profiles are shown as scatter plots of fold change (Fc) in log₂ of two RNA-seq replicas of L1 stage worms, lsm-4^−/− versus WT. Each dot corresponds to a gene. The class of genes that were most significantly affected control body morphogenesis (P = 0.0072). (B) Scatter plot comparing the relative gene expression between the lsm-8 (x-axis) and the lsm-4 mutant (y-axis). Common up-regulated genes are shaded yellow; 53% of genes up-regulated in lsm-8 (FDR < 0.05 and Fc > 4) are also up-regulated to some extent (50% increase) in lsm-4. Genes up-regulated in lsm-8^−/− only are shaded pink, and those in lsm-4^−/− only in blue. (C) Z-projection of live confocal images of embryonic progeny issued from the cross between the strain (GW1004) expressing LSM-4-Flag-GFP (GFP, green channel) and the strain (GW694) SET-25, the H3K9me3 HMT, fused to mCherry. The merged image shows no significant overlap on the brightest foci of SET-25. The boxed area is enlarged to show nuclear foci in bottom panels. Bars, 10 µm.

In yeast and plants, RNA-binding complexes feed back to establish H3K9 methylation on centromeric and other heterochromatin (Buhler et al. 2007; Vasiljeva et al. 2008; Keller et al. 2012; Zofall et al. 2012; Yamanaka et al. 2013; Egan et al. 2014; Chalamcharla et al. 2015; Tucker et al. 2016). It was therefore of interest to compare RNA-seq changes provoked by the loss of LSM2-8 with loss of canonical heterochromatin marks, H3K9me2 or me3. We found that 36% of the genes derepressed because of the loss of H3K9 methylation (met-2^−/− set-25^−/−) overlapped with those depressed in the lsm-8^−/− mutant at the L3 stage (Fig. 3B; Mattout et al. 2020). To test the additivity of the LSM2-8 and H3K9me-mediated repression pathways, we compared the RNA-seq data of the single mutants with the triple mutant (lsm 8^−/−; met-2^−/− set-25^−/−) (Fig. 3A) and found that there were clear subsets of up-regulated genes that were unique to either the lsm-8 or the met-2 set-25 mutant, and that genes showing mild up-regulation upon loss of either lsm-8 or met-2 set-25 (Fc ∼ 2) were more strongly expressed when combined (FDR < 0.05 and Fc > 4) (Mattout et al. 2020). Thus, the LSM2-8 pathway of silencing for endogenous loci is distinct from that mediated by H3K9me2/3, even though some genes are targeted by both pathways. Consistently using the heterochromatic reporter pkIS1582, we could show that lsm-7 RNAi derepresses in an additive fashion with the loss of either H3K9me3 (met-2 set-25) or the loss of the H3K9me3 reader, HPL-2 (Fig. 3C).

LSM2-8 TARGETS POLYCOMB-MARKED GENES

To ask what the lsm-8-sensitive genes have in common, we compared our RNA-seq data to the normalized ChIP-seq data of common histone modifications generated by ModEncode (Fig. 3D–E). In worms, as in most organisms, H3K4me2, H3K4me3, and H3K27ac are associated with active genes (Liu et al. 2011; Wenzel et al. 2011; Ho et al. 2014), whereas H3K9me2/3 and H3K27me3 are repressive histone marks that colocalize with heterochromatin (Wenzel et al. 2011). The genes that were up-regulated in the lsm-8^−/− mutant were depleted for active marks (H3K4me3) (Fig. 4E) in WT L3 larvae, and showed no particular enrichment for H3K9me2 nor for H3K9me3 (Mattout et al. 2020). Strikingly, however, 95% of genes that were derepressed by loss of LSM-8 were enriched for the repressive Polycomb mark, H3K27me3 (Fig. 3E). This was true not only for strongly up-regulated genes, but also for genes that were mildly up-regulated upon loss of LSM-8 (less than fourfold) and for windows of 500 bp across the genome. Thus, the overriding feature of genes or domains silenced by LSM2-8 was their enrichment for H3K27me3.

Consistent with the fact that H3K27me3 is the hallmark of Polycomb-repressed genes (Liu et al. 2011; Margueron and Reinberg 2011; Grossniklaus and Paro 2014; Conway et al. 2015), the absolute steady state level of the lsm-8-sensitive mRNAs was very low or almost undetectable in WT larvae (Fig. 3F). Although H3K27me3 is deposited in C. elegans by a PRC2-like complex comprising MES-2/E(z)/Ezh2, MES-3, and MES-6/Esc (Ketel et al. 2005; Yuzyuk et al. 2009; Gaydos et al. 2014) and mediates cell type–specific repression of developmentally regulated genes, in worms H3K27me3 and H3K9me3 ChIP-seq signals colocalize at many sites across the C. elegans genome (Ho et al. 2014). Interestingly, based on the ModEncode data sets from WT L3 larvae, 83% of the genes that are reproducibly enriched for H3K9me3 are also enriched for H3K27me3, and 41% of the genes enriched for H3K27me3 also carry H3K9me3. Among the genes up-regulated by lsm-8 mutation, we find that 40%–43% also carry H3K9me3 (Mattout et al. 2020). Because this rate is the same as the genome-wide coincidence of paired K9/K27 methylation, it appears that nucleosomes or domains bearing both repressive marks are not significantly enriched among lsm-8 targets. Most likely, H3K9me2 or me3 marks are irrelevant, for LSM2-8 recruitment or silencing.

LSM2-8 SILENCES THROUGH RNA DEGRADATION

Given that the LSM2-8 complex is known to stabilize and bind U6 snRNA in yeast and plants (Beggs 2005; Perea-Resa et al. 2012; Zhou et al. 2014), and that the LSM2-8 complex coprecipitates both with a factor involved in U6 snRNA stability (Ruegger et al. 2015) and with U6 snRNA itself in C. elegans (Mattout et al. 2020), we analyzed the RNA-seq data for splicing defects. We compared the exon–exon junction reads in the lsm-8 RNA-seq data with those from WT worms, to see if LSM2-8-mediated silencing coincides with defects in splicing, but we found irregular splicing events at only 18 exon–exon junctions out of 134,836 splice junctions (<0.02%) (Mattout et al. 2020). Thus, impaired splicing is most likely not the cause of lsm-8^−/−-triggered derepression.

If lsm 8^−/− worms reduced the expression of PRC2 or PRC1 homologs, then it would follow that histone H3K27me3 levels would be reduced, and Polycomb-target genes would be derepressed. By checking the relevant mRNA levels, we could rule this out as a mechanism (Mattout et al. 2020). Nonetheless, to explore the relationship of LSM2-8-mediated silencing to H3K27me3, we treated the heterochromatic reporter–bearing lsm-8^−/− worms with RNAi against either mes-2 (EZH2) or xrn-2. As controls, we combined the lsm-8 mutant with RNAi against lsm-7 for epistasis and with mes-4 (H3K36me HMT) or set-25 (H3K9me HMT), which both show additivity with lsm-8 (Fig. 5A). Although xrn-2 RNAi is epistatic with the lsm-8 background, the depletion of mes-2 increased reporter expression, albeit less than that observed for set-25 or mes-4 RNAi (Fig. 5A). This argues that Polycomb-mediated silencing is not exclusively dependent on LSM-8. Again, this can be explained by the fact that H3K27me3 is known to repress on the level of transcription. The strong epistasis observed with XRN-2, on the other hand, argued that LSM2-8 may silence through RNA turnover or degradation (Fig 5A). Comparison of RNA-seq data from worms treated with xrn-2 RNAi at L4 stage (Miki et al. 2016) and lsm-8 mutant worms at L3 confirms significant overlap of up-regulated endogenous genes (Fig. 5B). Indeed, 71% of the genes up-regulated by lsm-8^−/− were also up-regulated by xrn-2 RNAi (Fig. 5B, yellow) and 95% of those genes are enriched for H3K27me3. Again, this argues that LSM-8 and XRN-2 function on the same pathway. Nonetheless, a subset (<1/3) of LSM2-8 target genes (pink) were not affected by the down-regulation of XRN-2. It is unclear if those reflect experimental differences or a subset of genes silenced by LSM2-8 through another mechanism.

View larger version:

• In this window
• In a new window

Figure 5.

XRN-2 and other RNA factors show epistasis with lsm-8 deficiency for heterochromatic array derepression. (A) RNAi experiments were performed in the lsm-8^−/− (GW1119) strain. The GFP derepression of the heterochromatic reporter pkIs1582 was quantified by semiautomated fluorescence microscopy on homozygous lsm-8^−/− progeny under indicated RNAi conditions. P-values (Students t-test) are indicated: (*) P < 0.005. This indicates additivity between the lsm-8 derepression and derepression by the RNAi used. Bars = S.E.M., N = 2, n = 55, 45, 22, 11, 10, 23, 25, 25, 85. (These data are reproduced from Fig. 6F of Mattout et al. 2020, with permission.) (B) Scatter plot comparing relative gene expression changes of lsm-8^−/− L3 larvae and xrn-2 RNAi treated L4 (Miki et al. 2016). Common up-regulated genes are shaded yellow; 71% of genes up-regulated in the lsm-8 mutant (Fc > 4) are also up-regulated to some extent (50% increase) in xrn-2 depleted worms. lsm-8^–/–-specific up-regulated genes are shaded pink. (This panel is reproduced from Fig. 6A of Mattout et al. 2020, with permission.) (C) WT and lsm-8^−/− worms were sorted, refed with OP50 in liquid culture for 1 h at room temperature, and treated with 50 μg/mL final concentration of α-amanitin, which inhibits Pol II and Pol III transcription. RNA was isolated at time 0, 4.5, and 6 h, as indicated for each independent experiment. RNA levels of three transcripts affected by LSM-8 and two control transcripts (expression not affected by LSM-8) were determined by RT-qPCR and normalized to 18S rRNA levels, which are insensitive to α-amanitin. The value at 0 h is defined as 100%. Error bars represent standard deviation of the mean from at least three independent experiments. (Data are reproduced from Extended Data Fig. 8A in Mattout et al. 2020, with permission.)

LSM2-8 ACTS TOGETHER WITH RNA DEGRADATION ENZYMES

To find other genes implicated in this potential nuclear RNA decay function, we tested candidate cofactors for heterochromatic reporter derepression and examined their genetic relationship with lsm-8^−/−. Heterochromatic array derepression at the L1 stage was monitored by GFP expression, following a targeted RNAi screen against genes that are co-regulated with LSM2-8 subunits and/or functionally implicated in RNA processing—namely, RNA Pol II subunits (rpb-12, rpb-7) and the type II poly(A) binding protein (pabp-2). Like xrn-2 RNAi, RNAi against rpb-12, rpb-7, or pabp-2 on their own derepress the heterochromatic reporter (Fig. 5A), whereas the depletion of six additional genes that are implicated in aspects of RNA metabolism does not (F49C12.11, T13F2.2, cgh-1—the human ortholog of DDX6, B0495.8, C50D2.8, and pab-2—the human ortholog of PABPC1) (Mattout et al. 2020). Important, nonetheless, is the epistasis between derepression by xrn-2 RNAi, rpb-12, rpb-7, and pabp-2 and the lsm-8 mutant. In conclusion, PABP-2, RPB-12, and RPB-7 may act with XRN-2 on the LSM2-8-mediated silencing pathway.

LSM2-8 CONTRIBUTES TO THE DEGRADATION OF mRNAs FROM POLYCOMB-MARKED GENES

The cooperation between the RNA-binding LSM2-8 complex and XRN-2 suggests that LSM2-8 may silence genes by triggering mRNA degradation. It would follow that lsm-8 deletion might increase the RNA stability of specific LSM2-8-sensitive transcripts in vivo because of a failure to degrade these RNAs by XRN-2. This was confirmed for the mRNAs of three endogenous lsm-8-sensitive genes—far-3, grl-23, and ZK970.2—by monitoring mRNA stability after the addition of α-amanitin, an inhibitor of RNA Pol II and Pol III elongation. The rate of mRNA decay was monitored at the L3 stage of sorted WT and lsm-8^−/− worms over 6 h using RT-qPCR. RNA levels were normalized to that of the 18S rRNA. Deletion of lsm-8 delayed the turnover of far-3, grl-23, and ZK970.2 mRNAs, whereas the mRNA and pre-mRNA of two control genes (eft-3 and F08G2.8) showed no change in turnover rates (Fig. 5C; Mattout et al. 2020). Consistently, transcript levels from these lsm-8-sensitive genes, far-3 and ZK970.2, were also strongly up-regulated in xrn-2 RNAi treated worms as well, with a fold change in log₂ of 4.9 and 4.0, respectively (Mattout et al. 2020). The decay of the three lsm-8-sensitive mRNAs was also attenuated in the Polycomb-deficient strain (mes-2) (Mattout et al. 2020). Thus, our data suggest that MES-2-mediated repression of LSM-8 regulated genes acts at least in part through mRNA degradation (Mattout et al. 2020).

LSM2-8 ACTION FEEDS BACK TO MAINTAIN H3K27me3 LEVELS

Although mes-2 and lsm-8 depletion appeared to be additive with respect to derepression, one cannot conclude that the LSM2-8 complex acts independently of MES-2 and H3K27me3, especially considering the striking correlation of LSM-8-sensitivity with H3K27me3 (Fig. 3E). We initially tested for global changes in repressive epigenetic marks, H3K9me3 and H3K27me3, by immunostaining of embryos following lsm-7 RNAi. However, we found no global drop in staining (Fig. 6A). Nonetheless, when we performed quantitative chromatin immunoprecipitation (ChIP) for H3K27me3 on genes that are sensitive to lsm-8 ablation (i.e., genes silenced by the LSM2-8 complex in WT worms), we observed a significant decrease (>50%) in H3K27me3 levels (Fig. 7B). This was not detected for lsm-8-insensitive H3K27me3-marked genes. This suggests that the LSM2-8 complex selectively feeds back to maintain H3K27me3 levels, either directly or indirectly, at a subset of H3K27me3-silenced loci. The fact that it is not detectable by immunostaining argues that H3K27me3 levels may not be affected globally, although many Polycomb targets do show significant derepression.

View larger version:

• In this window
• In a new window

Figure 6.

LSM2-8 helps maintain full H3K27me3 levels on silenced loci. (A) Z-projection of confocal images of wild-type (GW566) embryos treated with control or lsm-7 RNAi (GFP, live imaging) or stained with antibodies against H3K9me3 and H3K27me3, as indicated. Nuclei are enlarged in the insets. Bar, 5 µm. (B) H3K27me3 ChIP-qPCR on target genes in WT and lsm-8^−/− worms. Three categories of genes were assessed. Five genes were monitored that are up-regulated in lsm-8^−/− worms and enriched for H3K27me3 mark in WT (lsm-8 target genes), three genes were monitored that are enriched for the H3K27me3 mark in WT but do not change in expression in the lsm-8 mutant (nontarget genes), and the control category is represented by ska-1, a gene not enriched for H3K27me3 in WT and with no expression change (control). N = 3, n = 3, bars (mean ± S.E.M.). (Data appear in Fig. 7A of Mattout et al. 2020 and are reproduced with permission.)

View larger version:

• In this window
• In a new window

Figure 7.

Model for the LSM2-8 pathway of silencing. We showed that the LSM-8-mediated silencing pathway makes use of XRN-2 ribonuclease and involves other factors, such as PABP-2 (H.s: PABPN1). We hypothesize that RNA arising from H3K27me3 genomic regions that are controlled by the LSM2-8 complex acquire a specific feature during transcription (e.g., a specific structure, RNA modification, 3ʹ UTR, poly(A/U) tail, or specific RNA-binding protein(s)), which allows recognition and processing by LSM2-8. LSM2-8-mediated silencing also feeds back to regulate H3K27me3 levels on LSM-8-regulated genes, although it is unclear if the interaction with PRC2 complex is direct (dotted arrow). LSM-2-8-mediated silencing of H3K27me3-bound regions defines a new mechanism for selective post-/cotranscriptional silencing for facultative heterochromatin through RNA decay.

CONCLUSION

We have shown that in nematodes a conserved nuclear RNA-binding complex, LSM2-8, contributes to full repression of heterochromatic reporters and genomic regions that carry the H3K27me3 epigenetic mark. A related, cytoplasmic complex, LSM1-7, appears to have no specific role in heterochromatic silencing. Unlike heterochromatin-linked RNA degradation pathways in plants and fission yeast, LSM-2-8-mediated repression is independent of and additive with H3K9me3 and H3K9me HMTs. Our data suggest that the LSM2-8 complex acts primarily on RNA stability, selectively mediating the decay of transcripts from H3K27me3-tagged genes through the 5ʹ–3ʹ exoribonuclease XRN-2. The pathway does not, however, involve XRN-1 nor the decapping enzyme DCAP-2. The LSM2-8-mediated silencing occurs at Polycomb H3K27me3-marked genes, with only ∼40% of the LSM2-8 targets also carrying H3K9me3. This represents the stochastic frequency of H3K27me3 and H3K9me3 coincidence on genes in L3 stage larvae, suggesting that H3K27me3 and not H3K9me3 is the recognition signal for mRNA degradation. As sketched in our model in Figure 7, we propose that the LSM2-8 complex mediates gene silencing primarily through the exonuclease XRN-2, which would degrade the transcripts from a subset of H3K27me3-marked genes either co- or posttranscriptionally. This represents a novel pathway for Polycomb-mediated gene silencing in animals, which appears to act in addition to transcriptional repression.

The fact that only a subset of transcripts arising from H3K27me3-marked genes is found to be significantly derepressed in lsm-8 mutant worms suggests that there may be additional criteria that provide specificity for LSM2-8 control. Using a less stringent cutoff for derepression in the lsm-8 RNA-seq data set (Mattout et al. 2020), we find many more genes up-regulated, and >93% still bear H3K27me3. It is important to note that the derepression of tissue-specific transcripts (i.e., those that are expressed in only a few cells of larvae or adult worms) will not be detectable by our RNA-seq methods that monitor RNA in whole worm extracts. This was confirmed to be the case for one HOX gene called egl-5, which is known to be regulated by MES-2 (EZH2) (Mattout et al. 2020). By analyzing the cell type–specific derepression of egl-5 using fluorescence microscopy, we were able to show similar levels of derepression of this gene in male tail cells, in both lsm-8 and mes-2 mutants (Mattout et al. 2020), although its derepression level by RNA-seq was only ∼1.5-fold.

The molecular basis of LSM2-8 specificity for H3K27me3 provides a framework for future studies exploring how the epigenetic information can be transferred from the chromatin to an RNA-binding complex. We propose that the RNAs arising from H3K27me3-marked genomic regions are recognized and bound selectively by the LSM2-8 complex, and that this likely requires a feature that is acquired during transcription from that locus—for example, a specific structure, RNA modification, cap, a poly-A/U tail, or specific RNA-binding protein(s) recognized by LSM2-8. It is important to note that we scored a drop in H3K27me3 specifically on LSM2-8-sensitive genes in lsm-8^−/− animals, even though the overall staining level of H3K27me3 appeared unchanged (Fig. 6). This suggests that RNA degradation helps maintain H3K27me3 levels. In other words, transcriptional repression by this epigenetic mark may be transferred from the RNA-binding complex LSM2-8 to chromatin.

An RNA decay-mediated pathway involving the cofactors PABP-2, RPB-12, and RBP-7 appears to function independently of DACP-2 and LSM-1 and additively with HPL-2 and the H3K9me3 HMT, SET-25. The epistasis with type II poly(A) binding protein (PABP-2) (Hurschler et al. 2011) (HsPABPN1 and SpPab2) is particularly interesting, because it appears to regulate 3ʹ UTR and poly(A) tail length in the nucleus (Kuhn et al. 2009). Moreover, it binds nascent RNAs early during the elongation step (Lemieux and Bachand 2009; Beaulieu et al. 2012). Given that the LSM2-8 complex is known to bind to the 3ʹ oligo(U) tail of the U6 snRNA (Zhou et al. 2014), as well as the 3ʹ poly(A+) tail of nuclear RNAs (Kufel et al. 2004), PABP-2 may regulate its specificity by modulating the 3ʹ end of mRNAs at H3K27me3-marked domains. The participation of RNA Pol II subunits RPB-7 and -12 in LSM2-8-mediated gene silencing is also suggestive of an involvement of RNA Pol II in silencing, just as the Schizosaccharomyces pombe RBP-7 homolog has been implicated in centromeric repeat RNAi-directed silencing (Djupedal et al. 2005). Interestingly, fission yeast lacks H3K27me3; thus, the link to chromatin status may have evolved at a later stage. In S. cerevisiae, the same subunit plays a different, but direct role in Pat1/Lsm1-7-mediated mRNA decay in the cytoplasm (Lotan et al. 2007; Haimovich et al. 2013). One speculative scenario, based on functions described in other organisms, is that rpb-7 and rpb-12 RNA Pol II subunits might be loaded on LSM2-8-regulated transcripts to signal to that these should be degraded by XRN-2. An unknown decapping enzyme(s) may also be involved, which would allow RNA degradation by XRN-2 in the nucleus.

We conclude that facultative heterochromatin in metazoans can be silenced through a mechanism of selective transcript degradation and not only by transcriptional repression. LSM2-8 may mediate gene silencing by linking a specific epigenetic state to transcript degradation, adding an additional layer of control over differentiation and development.