ABSTRACT
Repression of germline-promoting genes in somatic cells is critical for somatic development and function. To study how germline genes are repressed in somatic tissues, we analyzed key histone modifications in three Caenorhabditis elegans synMuv B mutants, lin-15B, lin-35, and lin-37, all of which display ectopic expression of germline genes in the soma. LIN-35 and LIN-37 are members of the conserved DREAM complex. LIN-15B has been proposed to work with the DREAM complex but has not been shown biochemically to be a complex member. We found that in wild-type worms synMuv B target genes and germline genes are enriched for the repressive histone modification dimethylation of histone H3 on lysine 9 (H3K9me2) at their promoters. Genes with H3K9me2 promoter localization are distributed across the autosomes and not biased toward autosomal arms like broad H3K9me2 domains. Both synMuv B targets and germline genes display dramatic reduction of H3K9me2 promoter localization in lin-15B mutants, but much weaker reduction in lin-35 and lin-37 mutants. This is the first major difference reported between lin-15B and DREAM complex mutants, which likely represents a difference in molecular function for these synMuv B proteins. In support of the pivotal role of H3K9me2 in regulation of germline genes through LIN-15B, global loss of H3K9me2 but not H3K9me3 results in phenotypes similar to synMuv B mutants, high temperature larval arrest and ectopic expression of germline genes in the soma. We propose that LIN-15B-driven enrichment of H3K9me2 on promoters of germline genes contributes to repression of those genes in somatic tissues.
INTRODUCTION
Repression in somatic cells of genes that promote germline development and function is a vital cell fate regulatory mechanism, which when disrupted leads to developmental problems and is a hallmark of aggressive cancer (Janic et al. 2010; Petrella et al. 2011; Whitehurst 2014; Al-Amin et al. 2016). Repression of germline genes in the soma poses a unique challenge for cells. First, like other genes expressed in specific tissues, germline genes can be found clustered along chromosomes; however, within a given cluster genes with ubiquitous, germline-specific, germline-enriched, and non-germline expression are interspersed (Spellman and Rubin 2002; Roy et al. 2002; Reinke and Cutter 2009).
Therefore, somatic cells require a mechanism to repress germline-specific genes without disrupting expression of important flanking genes. Second, because embryos start life as the fusion of two germline cells, an egg and a sperm, they inherit an epigenetic state that has been driving germline gene expression (Furuhashi et al. 2010; Rechtsteiner et al. 2010; Zenk et al. 2017; Tabuchi et al. 2018; Kreher et al. 2018). This chromatin state must be reset during development to turn off germline gene expression in differentiating somatic cells (Morgan et al. 2005; Fraser and Lin 2016). There has been no investigation to date of the unique patterns of chromatin modifications or regulatory protein binding that lead to repression of germline-specific genes across somatic tissues.
synMuv (for synthetic Multivulva) B proteins are a diverse class of transcriptional repressors that are involved in a number of different fate decisions in C. elegans (Unhavaithaya et al. 2002; Wang et al. 2005; Fay and Yochem 2007). A subset of synMuv B genes show a distinct set of mutant phenotypes, which include ectopic expression of germline genes in somatic cells and larval arrest at high temperature (called HTA for high temperature arrest) (Wang et al. 2005; Petrella et al. 2011; Wu et al. 2012). Of this subset, a large proportion encode proteins that exist in two proposed complexes: the HP1-containing heterochromatin complex (HPL-2, LIN-13, LIN-61 and the DREAM complex (EFL-1, DPL-2, LIN-35, LIN-9, LIN-37, LIN-52, LIN-53, LIN-54) (Coustham et al. 2006; Harrison et al. 2006; Wu et al. 2012). Several additional synMuv B mutants, including lin-15B and met-2, also display ectopic germline gene expression in the soma, but have not been shown biochemically to encode members of the HP1 or DREAM complex (Wu et al. 2012; Petrella et al. 2011). lin-15B mutants, like mutants in genes encoding DREAM complex members, also display an HTA phenotype, show changes in regulation of somatic RNAi, and cause transgene silencing in the soma (Wang et al. 2005; Petrella et al. 2011; Wu et al. 2012).
While mutations in genes encoding the heterochromatin complex, the DREAM complex, LIN-15B, and MET-2 all lead to ectopic expression of germline genes in the soma, the precise way these different complexes/proteins function in parallel or together to repress germline genes in the somatic tissues of wild-type animals is not understood.
Several lines of evidence point to synMuv B complexes repressing gene expression by altering chromatin. First, synMuv B mutant phenotypes, including HTA and ectopic germline gene expression, are strongly suppressed by loss of chromatin factors (Unhavaithaya et al. 2002; Wang et al. 2005; Cui et al. 2006; Petrella et al. 2011; Wu et al. 2012). Second, the DREAM complex has been shown to promote enrichment of the H2A histone variant HTZ-1 in the body of a subset of genes that the DREAM complex represses in L3 larvae (Latorre et al. 2015). Finally, HPL-2 is a homolog of heterochromatin protein 1 (HP1) (Couteau et al. 2002). HPL-2, in a complex with LIN-13 and LIN-61, localizes to areas of histone H3 methylated at lysine 9 (H3K9me) and helps create repressive heterochromatin (Wu et al. 2012; Garrigues et al. 2015). Together these data indicate that changes to chromatin may underlie the ectopic expression of germline genes in synMuv B mutants.
One of the best studied aspects of chromatin regulation is covalent modifications on histone tails. Specific histone modifications are often associated with repressive or active chromatin compartments and can be a read-out of the expression state of a gene. Histone H3 lysine 4 methylation (H3K4me) and H3 lysine 36 methylation (H3K36me) are generally associated with areas of previous or active gene expression (Ho et al. 2014; Evans et al. 2016). In contrast, histone H3 lysine 9 methylation (H3K9me) and histone H3 lysine 27 methylation (H3K27me) are associated with areas of low/no expression of coding genes and repression of repetitive elements (Ahringer and Gasser 2018). Of particular interest for the regulation of germline gene expression in somatic cells is histone H3K9 methylation. In C. elegans, mono- and dimethylation of H3K9 (H3K9me1 and H3K9me2, respectively), are primarily catalyzed by MET-2. met-2 mutants lose 80-90% of H3K9me1 and H3K9me2 in embryos (Towbin et al. 2012). met-2 is a synMuv B gene and mutants have been previously shown to ectopically express germline genes in somatic cells (Wu et al. 2012).
Trimethlyation of H3K9 (H3K9me3) is catalyzed by a separate histone methyltransferase, SET-25 (Towbin et al. 2012). set-25 is not a synMuv B gene and its potential role in regulating germline gene expression in the soma has not been tested. Several studies have analyzed the roles in C. elegans of H3K9me2 and H3K9me3 in regulating the interaction of heterochromatin with the nuclear periphery and repression of repetitive elements (Meister et al. 2010; Towbin et al. 2012; Guo et al. 2015; Zeller et al. 2016). Both of these functions primarily rely on the high enrichment of H3K9 methylation on the heterochromatic arms of the autosomes (Ikegami et al. 2010; Liu et al. 2011; Garrigues et al. 2015; Evans et al. 2016). However, little work has been done to look at how H3K9 methylation localizes to or regulates protein-coding genes in the euchromatic central regions of autosomes, especially germline genes. To fill this gap, we sought to determine the changes in the levels and distributions of active and repressive histone modifications in the soma of synMuv B mutants and whether such changes may underlie ectopic expression of germline genes.
In this study we used chromatin immunoprecipitation with genome-wide high-throughput sequencing (ChIP-seq) to investigate histone modifications in wild type and three synMuv B mutants, lin-15B, lin-35, and lin-37. We found that in wild-type L1 larvae, which are composed of 550 somatic cells and 2 germ cells and are therefore primarily somatic, H3K9me2 is enriched on the promoters of a subset of genes that display germline-specific expression. The genes that have H3K9me2 in their promoters in wild type are generally up-regulated in synMuv B mutants, suggesting that H3K9me2 plays a role in their repression. In support of this, the localization of H3K9me2 to gene promoters is largely lost in lin-15B mutants and is diminished but not lost in lin-35 and lin-37 mutants. Loss of H3K9me2 on promoters in mutants is associated with an increase in H3K4me3 on promoters and H3K36me3 in gene bodies, modifications associated with gene expression, suggesting that these genes go from a repressed to expressed state. Finally, we found that global loss of H3K9me2 but not H3K9me3 results in both the HTA and the ectopic germline gene expression phenotypes seen in lin-15B mutants. We propose that LIN-15B and DREAM repress a subset of germline genes in somatic tissues by promoting enrichment of H3K9me2 on those genes’ promoters.
MATERIALS AND METHODS
C. elegans strains and culture conditions
C. elegans were cultured using standard conditions (Brenner 1974) at 20°C unless otherwise noted. N2 (Bristol) was used as wild type. Mutant strains were as follows:
MT10430 lin-35(n745) I
SS1110 hpl-2(tm1489) III
MT5470 lin-37(n758) III
MT13293 met-2(n4256) III
MT17463 set-25(n5032) III
LNP0036 met-2(n4256) set-25(n5032) III
MT2495 lin-15B(n744) X
ChlP-seq from L1s
Worms were grown from synchronized L1s in standard S-basal medium with shaking at 230 rpm and fed HB101 bacteria until gravid. Embryos were harvested using standard bleaching methods, and L1s were synchronized in S-basal media with shaking for 14-18 hours in the absence of food. For 26°C samples, worms were grown to the L4 stage at 20°C, when they were up-shifted to 26°C until gravid and L1s were harvested as described above. Extracts were made as described in (Kolasinska-Zwierz et al. 2009) with the following modifications. Cross-linked chromatin was sonicated using a Diagenode Bioruptor at high setting for 30 pulses, each lasting 30 sec followed by a 1 min pause. ChIP was performed as described by (Kolasinska-Zwierz et al. 2009) with the modification of using 0.5 mg of protein and 1 μg antibody or by using an IP-Star Compact Automated System (Diagenode) as described in (Tabuchi et al. 2018). Sequencing libraries were prepared in two ways. Some libraries were prepared with the NEBNext Ultra DNA library Prep Kit (NEB) following the manufacturers’ instructions. 1 ng of starting DNA was used, adapters were diluted 1:40, and AMPure beads were used for size selection before amplification to enrich for fragments corresponding to a 200 bp insert size. The other libraries were prepared using Illumina Truseq adapters and primers with the following custom mixtures. ChIP or input DNA fragments were end repaired with the following: 5 μl T4 DNA ligase buffer with 10 mM ATP, 2 μl dNTP mix, 1.2 μl T4 DNA polymerase (3U/μl), 0.8 μl 1:5 Klenow DNA polymerase (diluted with 1X T4 DNA ligase buffer for final Klenow concentration 1U/μl), 1 μl T4 PNK (10U/μl). This 50 μl reaction was incubated at 20°C for 30 minutes and purified with a QIAquick PCR spin column (elution volume 36 μl). ‘A’ bases were then added to the 3’ end of the DNA fragments with the following: 5 μl Klenow buffer/NEB 2, 10 μl dATP (1 mM), 1 μl Klenow 3’ to 5’ exo-(5U/μl).
This mixture was incubated at 37°C for 30 min, and the DNA was purified with a QIAquick MinElute column (11 μl of DNA was eluted into a siliconized tube). Illumina TruSeq adapters were ligated to DNA fragments with the following: 15 μl 2x Rapid Ligation buffer, 1 μl adapters (diluted 1:40), 1.5 μl Quick T4 DNA Ligase. This 30 μl reaction was incubated at 23°C for 30 min. The mixture was then cleaned up 2X with AMPure beads (using 95% volume beads) and DNA was eluted in 22 μl. The Adapter-Modified DNA fragments were amplified by PCR with the following mixture: 6 μl 5X Phusion Buffer HF, 2 μl Primer cocktail (from TrueSeq kit), 0.5 μl 25 mM dNTP mix, 0.5 Phusion polymerase (2U/μl) using the following PCR program: 98°C 30 min, 98°C 10 min and 60°C 30 min and 72°C 30 min repeated 16 cycles, 72°C 5 min. The amplified DNA was concentrated and loaded onto a 2% agarose gel, and DNA between 300-400 bp was recovered from the gel. The multiplexed libraries were sequenced on an Illumina HiSeq4000 or HiSeq2500 at the Vincent J. Coates Genomics Sequencing Laboratory at University of California, Berkeley.
ChIP-chip from embryos
Late-stage embryos were obtained and chromatin extracts prepared as described in (Latorre et al. 2015). Chromatin immunoprecipitation and subsequent LM-PCR, microarray hybridization, and scanning were performed as in (Garrigues et al. 2015).
Antibodies used for ChIP
Mouse monoclonal antibodies for H3K9me2 (Wako MABI0307, #3023069), H3K36me3 (Wako MABI0333, #300–95289), H3K27me3 (Wako MABI0323, #309–95259), and H3K4me3 (Wako MABI0304, #305–34819) as used in (Liu et al. 2011; Egelhofer et al. 2011). Rabbit polyclonal LIN-15B antibody (SDQ2330, Novus #38610002) was used at a concentration of 2.5 μg per mg of chromatin extract.
Analysis of ChlP-seq data
Raw sequence reads from the Illumina HiSeq (50 bp single-end reads) were mapped to the C. elegans genome (Wormbase version WS220) using Bowtie with default settings (Langmead et al. 2009). MACS2 (Zhang et al. 2008) was used to call peaks and create bedgraph files for sequenced and mapped H3K4me3 ChIP samples and corresponding Input DNA samples with the following parameters: callpeak -t H3K4me3.mapped.reads.sampleX -c Input.mapped.reads.sampleX -g ce --bdg -- keep-dup=auto --qvalue=0.01 --nomodel --extsize=250 —call-summits
MACS2 was used to call peaks and create bedgraph files for sequenced and mapped H3K9me2 ChIP samples and corresponding Input DNA samples with slightly different parameters to account for the broader domains of H3K9me2: callpeak -t H3K9me2.mapped.reads.sampleX -c Input.mapped.reads.sampleX -g ce --bdg -- keep-dup=auto --broad --broad-cutoff=0.01 --nomodel --extsize=250
A peak was considered to be associated with a gene’s promoter if it overlapped at least 100bp with the region 750bp upstream from the gene’s TSS to 250bp downstream from the TSS. A peak was considered to be associated with the body of a gene if it overlapped at least 250bp with the region from 250bp downstream from the TSS to the TES. A gene’s promoter or gene body was considered bound by H3K4me3 or H3K9me2 in one of the conditions if for all replicates of that condition a peak was associated with the gene’s promoter or body, respectively. The distribution of genes with peaks in promoters or gene bodies along an autosome are shown in Fig. 3A in 200kb windows.
Bedgraph files for genome browser displays were scaled to 5 million total reads for all H3K4me3 ChIP samples, 10 million reads for all H3K36me3 samples, 15 million reads for all H3K9me2 samples, and 20 million reads for all H3K27me3 samples. The different scaling factors roughly correspond to the different genome-wide coverages of the different ChIP factors, e.g. H3K4me3 being found mostly on promoters of expressed genes, H3K36me3 mostly on gene bodies of expressed genes, and H3K9me2 mostly on chromosomal arms. Further data analysis below was based on these scaled read coverages. Scaled bedgraph files were converted to bigwig using the bedGraphToBigWig UCSC Genome Browser tool (Kent et al. 2010) and displayed on the UCSC Genome Browser.
Analysis of LIN-15B ChIP-chip data
NimbleGen 2.1M probe tiling arrays (DESIGN_ID = 8258), with 50 bp probes, designed against WS170 (ce4) were used. Two independent ChIPs were performed. Amplified samples were labeled and hybridized by the Roche NimbleGen Service Laboratory. ChIP samples were labeled with Cy5 and their input reference with Cy3. For each probe, the intensity from the sample channel was divided by the reference channel and log2 transformed. The enrichment scores for each replicate were calculated by standardizing the log ratios to mean zero and standard deviation one (z-score) and the average z-score across replicates was calculated and displayed in the UCSC Genome Browser (Fig S3). Peak calling was performed with the MA2C algorithm (Song et al. 2007) using Nimblegen array design files 080922_modEncode_CE_chip_HX1.pos and 080922_modEncode_CE_chip_HX1.ndf and MA2C parameters METHOD = Robust, C = 2, pvalue = 1e-5, BANDWIDTH = 300, MIN_PROBES = 5, MAX_GAP = 250. The resulting peak calls were associated with gene promoters and bodies as described in the previous section.
Correlation heatmap of samples
The scaled bedgraph files were used to calculate for each sample the average read coverage in 1kb windows across all autosomes and the X chromosome. The resulting read coverage data were log transformed and normalized for each ChIP sample by dividing by the standard deviation across all 1kb windows and subtracting the 25th percentile across all 1kb windows. For each 1kb window and condition, the resulting data were averaged across replicates. The data were used to calculate the Pearson Correlation coefficient r between all conditions once for autosomes and once for the X chromosome. The distance d = 1 – r was calculated, and hierarchical clustering was used with the complete linkage method to cluster the conditions. The results are displayed in a heatmap where the cell coloring indicates r between two conditions (Fig. S1). The analysis was performed in R version 3.5.1 (R Core Team 2018).
Metagene plots
Metagene plots for the various ChIP targets and conditions (e.g. Fig. 2C, 4A, and S7) were generated by aligning genes of length greater than 1.25 kb at their TSS and TES using WormBase WS220 gene annotations. Regions 1 kb upstream to 1 kb downstream from the TSS and TES were divided into 150bp windows stepped every 50bp. The mean read coverage within each of these 150bp windows was calculated and normalized for each ChIP data set by dividing by the standard deviation across all 150bp windows and subtracting the 25th percentile across all 150bp windows. For each 150bp window the normalized data were averaged across replicates. A metagene profile for a set of genes was generated by averaging and plotting for each 150bp window the data across the genes in the set. Light vertical lines indicate 95% confidence intervals for the mean of each 150bp window. The analysis was performed in R version 3.5.1.
Boxplots and scatterplots
To display promoter ChIP signal in boxplots (Fig. S8) and scatter plots (Fig. 4B and S9), the mean read coverage for each protein-coding gene was calculated over the region 250bp up and down from the TSS. For boxplots (Fig. S8) resulting mean read coverage data were log transformed and normalized for each ChIP data set by dividing by the standard deviation across all genes and subtracting the 25th percentile across all genes. For each promoter the resulting signals were averaged across replicates and plotted in Fig S8. In scatterplots (Fig. 4B and S9) the wild-type log2 normalized read coverage was subtracted from the mutant log2 normalized read coverage for each promoter, resulting in a log2 fold change of mutant over wild-type promoter signal.
Gene set definitions
Ubiquitous genes, originally defined and discussed in (Rechtsteiner et al. 2010), are genes that were found to be expressed in germline, muscle, neural, and gut tissues (Wang et al. 2009; Meissner et al. 2009). Germline-enriched genes are as defined in (Reinke et al. 2004). Germline-specific genes are genes whose transcripts were found to be expressed exclusively in the adult germline and maternally loaded into embryos; these genes were defined using multiple datasets as described in (Rechtsteiner et al. 2010). Soma-specific genes are genes expressed in at least 1 of 3 somatic tissues (muscle, gut, and/or neuron) with at least 8 SAGE tags (Meissner et al. 2009) but not enriched (Reinke et al. 2004) or detectably expressed (Wang et al. 2009) in the adult germline. Silent genes are 415 serpentine receptor genes that are expressed in a few mature neurons and are not detectably expressed in L1 larvae, originally defined in (Kolasinska-Zwierz et al. 2009). HTA germline genes, as defined in (Petrella et al. 2011), are genes that were significantly up-regulated in lin-35(n745) mutants versus wild type and also significantly down-regulated in lin-35(n745) mes-4(RNAi) versus lin-35(n745), and that have germline-enriched expression (Reinke et al. 2004).
HTA larval arrest assays
L4 larvae were placed at 26°C for ~18 hours and then moved to new plates and allowed to lay embryos for 8 hours. Progeny were scored for L1 larval arrest (Petrella et al., 2011).
Immunohistochemistry
L1 larvae were obtained by hatching embryos in the absence of food in M9 buffer and fixed using methanol and acetone (Strome and Wood 1983). Anti-PGL-1 primary antibody (Kawasaki et al. 1998) was diluted 1:30,000 and larvae were stained for ~18 hours at 4°C (Petrella et al. 2011). Alexa Fluor 488 (Invitrogen) secondary antibody was used at a 1:500 dilution for 2 hours at room temperature. Slides were mounted in Gelutol and imaged using a Nikon Inverted Microscope Eclipse Ti-E confocal microscope at 60X.
RESULTS
lin-15B mutants lose a large proportion of H3K9me2 promoter peaks; lin-35 and lin-37 mutants lose fewer
To better understand how synMuv B proteins regulate germline gene expression in somatic cells, we sought to identify changes in histone modification patterns in mutants compared to wild type. We profiled the distributions of two histone modifications associated with active chromatin (H3K4me3 and H3K36me3) and two histone modifications associated with repressive chromatin (H3K27me3 and H3K9me2) using chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq). Experiments were done on L1 animals that experienced embryogenesis at 20°C or 26°C for four genotypes: wild type and three synMuv B mutants, lin-15B(n744), lin-35(n745), and lin-37(n758). Because L1 stage worms have 550 somatic cells and only 2 germline cells, extracts from L1s contain genomic material primarily from somatic tissue. Analysis of H3K4me3 and H3K36me3 patterns showed increased enrichment of these marks in mutants compared to wild type on classes of genes that are up-regulated in synMuv B mutants (discussed below). As the level of enrichment of these marks generally correlates well with expression level, this change was expected. We saw no changes in the pattern of the repressive modification H3K27me3 between mutants and wild type. However, we observed significant changes in the pattern of the repressive modification H3K9me2 between synMuv B mutants and wild type, especially on germline-expressed genes. We analyzed the changes to H3K9me2 patterns in detail to investigate whether this particular histone modification is important for repression of germline gene expression by synMuv B complexes.
Analysis of H3K9me2 showed there are not genome-wide changes in the distribution of H3K9me2 enrichment on autosomes and the X chromosome between mutants and wild type (Fig. S1). However, a subset of H3K9me2 peaks were observed to be lost or reduced in synMuv B mutants (Fig. 1A and B). To investigate the pattern of this loss/reduction, we performed peak calling for H3K9me2 and designated two types of peaks depending on the location of H3K9me2 relative to coding gene bodies. “Gene body peaks” are those peaks where H3K9me2 overlapped with at least a portion of the coding region of the gene that is more than 250bp downstream of the transcription start site (TSS) (Fig. 1A). The distribution of genes with gene body peaks mirrors what has been previously described for the general pattern of H3K9me2 and H3K9me3 enrichment in the C. elegans genome (Fig. 3A; Liu et al. 2011; Evans et al. 2006). “Promoter peaks” are those peaks where H3K9me2 overlapped with a region 750 bp upstream to 250 bp downstream of the TSS, but not further than 250bp downstream of the TSS (Fig. 1B). In wild-type extracts, H3K9me2 gene body peaks were generally broader than promoter peaks (Fig. 1A and B), and genes with body peaks (2991 at 20°C/2871 at 26°C) were about three times more abundant than genes with promoter peaks (984 at 20°C/981 at 26°C) (Fig. 1C and D).
Our analysis showed that loss of synMuv B proteins had a smaller effect on H3K9me2 in gene bodies than in promoters. In lin-15B mutants grown at 20°C ~12% fewer genes had a gene body peak compared to wild type; there was no reduction in the number of genes with H3K9me2 gene body peaks at 26°C (Fig. 1C). In contrast, in lin-15B mutants there were significantly fewer genes with H3K9me2 promoter peaks at both 20°C (~42% fewer) and 26°C (~25% fewer) when compared to wild type (Fig. 1D). The H3K9me2 promoter peaks found in lin-15B are for the most part a subset of the H3K9me2 promoter peaks in wild type (Fig. S2). Unlike the significant loss of genes with H3K9me2 promoter peaks seen in lin-15B mutants, fewer genes with H3K9me2 promoter peaks were lost in lin-35 and lin-37 mutants.
In both lin-35 and lin-37 mutants, there was no decrease in the number of genes with H3K9me2 gene body peaks (Fig. 1C). There was a small but significant decrease in the number of genes with H3K9me2 promoter peaks compared to wild type at 20°C, but no significant change in the number of genes with promoter peaks at 26°C (Fig. 1D). This is the first description of a molecular difference in phenotypes seen between mutants in DREAM complex members and lin-15B mutants and may represent a difference in their molecular function at target loci.
Genes with an H3K9me2 promoter peak are enriched for DREAM and LIN-15B target genes in wild type but not in lin-15B mutants
If localization of H3K9me2 to promoters is driven by synMuv B binding, we predict that genes with H3K9me2 promoter peaks are bound by synMuv B proteins in wild-type animals. There is a high co-occurrence of DREAM complex and LIN-15B binding, with 70% of DREAM bound loci also bound by LIN-15B (Fig. S3). We compared genes with an H3K9me2 promoter peak with genes that we defined as DREAM complex targets and/or LIN-15B targets. We defined 170 DREAM complex targets as those genes bound by the DREAM complex in their promoter by ChIP-seq in late embryos (Goetsch et al. 2017) that are also significantly up-regulated in lin-35 mutant L1s at 26°C (Petrella et al. 2011). We defined 115 LIN-15B targets as those genes bound by LIN-15B in their promoter by ChIP-chip in late embryos (this paper) that are also significantly up-regulated in lin-15B mutant L1s at 26°C (Petrella et al. 2011). Genes with an H3K9me2 promoter peak are enriched for DREAM complex and LIN-15B target genes in wild type, lin-35, and lin-37 mutants but not in lin-15B mutants (Fig. 2A). Thus, genes that have H3K9me2 promoter localization are correlated with DREAM complex and LIN-15B binding, and this correlation is disrupted when LIN-15B is absent.
Genes with an H3K9me2 promoter peak lose enrichment for germline genes in lin-15B mutants
One of the major phenotypes of many synMuv B mutants, including lin-15B mutants, is the ectopic expression in somatic cells of genes whose expression is normally restricted to the germline (Wang et al. 2005; Petrella et al. 2011; Wu et al. 2012). We investigated if genes that have an H3K9me2 promoter peak are enriched for genes that are specifically expressed in the germline. We analyzed four categories of expression: genes that are broadly expressed in all tissues (2576: ubiquitous, ubiq), genes that are repressed in most tissues (415: silent), genes that are expressed specifically in somatic tissues (1181: soma), and genes that are expressed specifically in the germline (169: gl-spec). Genes with H3K9me2 promoter peaks in wild type are enriched for genes with germline-specific expression, but not for genes with ubiquitous, silent, or somatic expression (Fig. 2B and S4). These enrichments are mirrored when plotting H3K9me2 ChIP-seq signal around the transcription start site (TSS) averaged over the genes in each expression category (Fig. 2C). If H3K9me2 at germline gene promoters is correlated with synMuv B repression of germline gene expression in the soma, then we would predict that germline genes would lose H3K9me2 promoter peaks in synMuv B mutants. Indeed, in lin-15B mutants, there were many fewer germline-specific genes with an H3K9me2 promoter peak, and there was a large decrease in the signal of H3K9me2 at the TSS of germline-specific genes (Fig. 2B and 2C). lin-35 and lin-37 mutants resembled wild type in showing genes with an H3K9me2 promoter peak enriched for germline-specific genes (Fig. 2B and 2C).
We also examined germline genes whose misregulation is correlated with the high temperature larval arrest (HTA) phenotype (Petrella et al. 2011). HTA-germline targets are defined as genes normally expressed in the germline that are up-regulated in arrested lin-35 mutant L1s at 26°C and whose expression returns to near wild-type levels in HTA-suppressed lin-35; mes-4(RNAi) double mutant L1s at 26°C (48: HTA-gl) (Petrella et al. 2011). Similar to what was seen with germline-specific genes, genes with an H3K9me2 promoter peak were enriched for HTA-germline genes in wild type, lin-35, and lin-37 mutants, but this enrichment was reduced in lin-15B mutants (Fig. 2B). These data together reveal a striking loss of H3K9me2 over the promoters of germline-specific and HTA-germline genes in lin-15B mutants, but not lin-35 or lin-37 mutants.
H3K9me2 promoter peaks are distributed along the length of autosomes
Previous work on H3K9me2 in C. elegans focused on its distribution in broad domains on autosomal arms and the role of H3K9me2 in repressing repetitive sequences (Ikegami et al. 2010; Liu et al. 2011; Guo et al. 2015; Zeller et al. 2016). Little investigation has been done into what role the more narrowly focused H3K9me2 found in promoters may be serving in gene regulation. In C. elegans, genes with expression that is higher in the germline than other tissues (germline-enriched genes) or with expression exclusive to the germline (germline-specific genes), show a biased localization to the centers of autosomes compared to the localization of all coding genes (Fig. S5). Therefore, if H3K9me2 promoter peaks are associated with regulation of germline gene expression, we would predict that H3K9me2 promoter peaks would also be found in the center regions of chromosomes and not be biased to the arms. We compared the distributions along autosomes of genes with H3K9me2 in their gene body versus in their promoter. In wild type, genes with H3K9me2 in their gene body demonstrated the previously reported pattern of H3K9me2 enrichment on autosomal arms compared to centers (Fig. 3A). For genes with an H3K9me2 gene body peak, all mutants showed the same autosomal arm bias as seen in wild type (Fig. 3A, 3B, and S6). In contrast, genes with an H3K9me2 promoter peak in wild type were more evenly distributed across autosomes, with weak or no depletion from autosomal centers (Fig. 3A and B). Notably, lin-15B mutants showed strong depletion of genes with H3K9me2 in their promoter on all autosomal centers (Fig. 3A and B), suggesting that LIN-15B is needed for H3K9me2 localization on gene promoters in autosomal centers where germline genes are enriched. lin-35 and lin-37 mutants showed distributions of genes with H3K9me2 in their promoter similar to wild type (Fig. S6). H3K9me2 promoter peaks in chromosome centers in wild type represent a pattern not previously described for H3K9me2 in C. elegans and place H3K9me2 promoter peaks in mainly euchromatic regions where they may affect coding gene expression. Additionally, the loss of H3K9me2 from promoter peaks in autosomal centers in lin-15B mutants suggests that LIN-15B plays a specific role in directing H3K9me2 to areas of the genome where there are fewer repeats and more coding genes, especially germline genes.
Loss of H3K9me2 in mutants is associated with increased H3K4me3 on germline genes
Trimethylation of histone H3 on lysine 4 (H3K4me3) and lysine 36 (H3K36me3) are correlated with active gene expression (Liu et al. 2011; Ho et al. 2014; Evans et al. 2016). Thus, we expected to see increases in H3K4me3 and H3K36me3 on germline genes in synMuv B mutants. Indeed, synMuv B mutants displayed increases in H3K4me3 and H3K36me3 on germline-specific and HTA-germline but not on other categories of genes (Fig. 4A, S7, and S8). The increased levels of both H3K4me3 and H3K36me3 on germline-specific and HTA-germline genes in mutants is consistent with these genes being expressed at higher levels, most likely in a larger population of cells (i.e. somatic cells in addition to the 2 primordial germ cells) in these mutants.
Because loss of H3K9me2 may not be sufficient to lead to gene expression changes, we investigated if there is a correlation between loss of the repressive H3K9me2 chromatin modification and gain of the active H3K4me3 chromatin modification in gene promoters. To compare those marks in promoter regions, we calculated the log2 fold change of the signal of each modification in lin-15B mutant/wild type within 250bp upstream and downstream of the transcription start site (TSS). A higher histone modification signal in lin-15B mutants than wild type would result in a positive log2 fold change; a lower histone modification signal in lin-15B mutants than wild type would result in a negative log2 fold change. In lin-15B mutants, 25% of all genes (122 of 448) that had reduced H3K9me2 promoter signal also had increased H3K4me3 promoter signal, using a 1.5-fold cut-off (Fig. 4B). Strikingly, 40% of germline-specific (6 of 15) and 75% of HTA-germline genes (12 of 16) that had reduced H3K9me2 promoter signal also had increased H3K4me3 promoter signal (Fig. 4B). Thus, in lin-15B mutants loss of H3K9me2 enrichment on the promoter of germline genes is more likely to result in increased H3K4me3 promoter enrichment than on genes in other categories. Interestingly, although fewer genes showed a decrease in promoter H3K9me2 enrichment in lin-35 and lin-37 mutants, the percentages of germline-specific and HTA-germline genes that showed reduced promoter H3K9me2 and increased promoter H3K4me3 were similar to the percentages in lin-15B mutants (Fig. S9). Thus, the germline genes that lose H3K9me2 promoter enrichment in any of the three mutants have a correlated increased enrichment of H3K4me3 and up-regulation in synMuv B mutants.
Global loss of H3K9me2 leads to phenotypes similar to lin-15B and DREAM complex mutants
To investigate if loss of H3K9me2 promoter localization plays an important role in lin-15B mutant phenotypes, we analyzed mutants for the histone methyltransferases (HMTs) responsible for H3K9 methylation. Loss of these HMTs leads to a global loss of all H3K9 methylation, which may result in a phenocopy of lin-15B mutants. H3K9 methylation in C. elegans embryos is catalyzed by two HMTs, MET-2 and SET-25, which primarily catalyze H3K9me1/2 and H3K9me3, respectively (Towbin et al. 2012). If loss of H3K9 methylation is associated with ectopic germline gene expression and the HTA phenotype, we would expect that met-2 and set-25 mutants would show these phenotypes. set-25 single mutants, which lose H3K9me3, showed neither an HTA phenotype nor an ectopic germline gene expression phenotype, as assessed by staining for the germline-specific protein PGL-1 (Fig. 5A and B).
Therefore, H3K9me3 does not appear to be important for repression of germline genes in the soma. In contrast, met-2 single mutants, which lose 80-90% of H3K9me2 and ~70% of H3K9me3 (Towbin et al. 2012), displayed ~80% larval arrest around the L3 stage at 26°C, but no larval arrest at 24°C (Fig. 5A). Thus, met-2 mutants show an HTA phenotype similar to but weaker than lin-15B mutants (Fig. 5A; Petrella et al. 2011). We also observed ectopic expression of PGL-1 in met-2 mutants at 26°C similar in level and distribution to lin-15B mutants, with a combination of diffuse and punctate cytoplasmic PGL-1 in intestinal cells (Fig. 5B). Unlike the previously published pattern of ectopic PGL-1 expression in met-2 (Wu et al. 2012), we did not see perinuclear punctae in intestinal cells (Fig. 5B). Perinuclear PGL-1 in intestinal cells has been described for hpl-2 mutants, the worm homolog of HP1, which is known to bind H3K9me2 (Petrella et al. 2011; Wu et al. 2012). Thus, met-2 single mutants have phenotypes similar to, but in the case of HTA weaker, lin-15B mutants. To test if the remaining 10-20% of H3K9me2 catalyzed by SET-25 in met-2 mutants (Towbin et al. 2012) partially represses germline gene expression in somatic cells, we analyzed met-2 set-25 double mutants, which have been shown to completely lack H3K9 methylation during embryonic stages (Towbin et al. 2012; Garrigues et al. 2015). Consistent with residual SET-25-mediated H3K9me2 serving a role in repression of germline genes in somatic cells, met-2 set-25 double mutants showed significantly enhanced larval arrest at 26°C when compared to met-2 single mutants (Fig. 5A). Similar to lin-15B, lin-35, and lin-37 mutants, met-2 set-25 double mutants did not show increased larval arrest at 24°C. The level and distribution of ectopic PGL-1 in met-2 set-25 double mutants at 26°C were similar to those seen in met-2 single mutants (Fig. 5B). Altogether, our results show that a global loss of H3K9me2 phenocopies both the HTA and ectopic germline gene expression seen in synMuv B mutants.
Discussion
Repression of germline gene expression in the soma is vital, as loss of germline gene repression is a hallmark of various disease states including cancer. Investigating the changes to chromatin that occur when germline genes are misexpressed in the somatic cells of mutants is a first step in understanding the mechanisms that repress germline genes to protect somatic fate and development. Here we investigated the changes to histone modifications that occur in a subset of synMuv B mutants that misexpress germline genes in the soma. We defined a new localization for the repressive histone modification H3K9me2 in wild-type extracts, at the promoter of coding genes, which unlike the previously described broad domains of H3K9me2 are not enriched on autosomal arms (Liu et al. 2011; Garrigues et al. 2015; Evans et al. 2016; Ahringer and Gasser 2018). Promoter enrichment of H3K9me2 in autosomal centers provides a new regulatory role for H3K9me2 in addition to its well described regulation of repetitive elements on autosomal arms. We also found that in wild-type somatic cells genes with an H3K9me2 promoter peak are enriched for genes expressed specifically in the germline and genes that are synMuv B targets. The localization of H3K9me2 to germline genes and synMuv B targets is disrupted strongly in lin-15B mutants and weakly in DREAM complex mutants. We additionally showed that loss of H3K9me2 but not H3K9me3 phenocopies synMuv B mutants. Our data implicate H3K9me2 promoter enrichment as an important aspect of repression of germline gene expression in somatic cells.
There is strong evidence that a memory of gene expression/repression and associated chromatin modifications are transmitted from the parental germline to the developing embryo (Furuhashi et al. 2010; Rechtsteiner et al. 2010; Zenk et al. 2017; Tabuchi et al. 2018). For example, genes that were expressed in the germline continue to be marked with MES-4-generated H3K36me3 in embryos, even in the absence of ongoing transcription in embryos (Furuhashi et al. 2010; Rechtsteiner et al. 2010; Kreher et al. 2018). It is thought that H3K36me3 marks these genes for re-expression in the germline during post-embryonic development. How then are germline genes repressed properly in somatic tissues when those tissues inherit marks of active expression that potentially set germline genes up for reexpression? Our data, along with other recent work, strongly implicate deposition of H3K9me2 at the proper time in development in creating proper expression patterns in differentiating somatic cells. Recent experiments have shown that in C. elegans H3K9me2 and H3K9me3 levels are very low in the nuclei of early stage embryos and only start to accumulate when cells are transitioning from early embryogenesis to mid-embryogenesis (Mutlu et al. 2018). This is in part driven by the nuclear import of an active MET-2 complex that catalyzes conversion of H3K9me1 to H3K9me2. The timing of MET-2 import just precedes the stage in embryogenesis when zygotic transcription is up-regulated and when tissue-specific expression patterns emerge (Spencer et al. 2011; Levin et al. 2012; Robertson and Lin 2015; Mutlu et al. 2018). Our data suggest that loss of H3K9me2, either through loss of the MET-2 and SET-25 HMTs that catalyze the mark or through loss of proper localization of H3K9me2 to germline genes in lin-15B mutants, leads to misexpression of germline-specific genes in somatic cells. We hypothesize that specific localization of H3K9me2 to germline gene promoters facilitated by LIN-15B is an important aspect of resetting the chromatin landscape of germline genes at this stage of development to prevent their expression in somatic lineages.
A striking aspect of our findings is the difference in changes to promoter-enriched H3K9me2 between lin-15B mutants and DREAM complex mutants. It was previously proposed, based on phenotype analysis, that LIN-15B is a member of the DREAM complex (Wu et al. 2012). Our data indicate that, although LIN-15B binds to and represses many of the same genes as the DREAM complex, its molecular function at those genes is probably distinct. The proposed DNA-binding domain of LIN-15B may allow it to be independently recruited to similar targets as the DREAM complex, where the two may function together to repress genes. This scenario has implications for regulation of gene expression in the germ line as well as in the soma. Recent work from the Seydoux lab has further implicated the differential presence of LIN-15B in the germline versus soma in this regulation (Lee et al. 2017). Maternally provided LIN-15B is normally removed from the primordial germ cells (PGCs), while DREAM components are not (Lee et al. 2017). Our work suggests that loss of LIN-15B from the PGCs may protect essential germline genes from being H3K9 methylated and repressed in those cells. How the different synMuv B complexes work together to fully repress germline-specific genes in somatic cells is still an open question. The establishment of H3K9me2 may be an initiating step in germline gene repression or may be one aspect of a series of redundant steps necessary to repress germline genes. Analysis of the order and dependency of DREAM, LIN-15B, and MET-2 binding to germline genes is necessary to address these questions.
The work presented here focuses on a subset of germline-promoting genes that are regulated through the LIN-15B/H3K9me2/DREAM complex pathway. Although this pathway may only regulate a small subset of genes in this way, the repercussions to development are clear, in that the organisms cannot thrive in the face of challenges (e.g. high temperature) when these fate changes occur. Recent work in Drosophila underscores the importance of H3K9 methylation in repression of a small subset of coding genes to maintain proper cell fate. In the Drosophila ovary, loss of H3K9me3 leads to up-regulation of testis-specific transcripts that change the fate of ovarian germ cells, leading to sterility (Smolko et al. 2018). As in C. elegans, prior investigations of H3K9 methylation loss in Drosophila had focused primarily on up-regulation of repetitive elements (Rangan et al. 2011; Wang et al. 2011; Guo et al. 2015; Zeller et al. 2016). However, it is clear that H3K9me2/3 loss leading to up-regulation of small sets of coding genes in a tissue-specific manner can have profound effects on cellular fate and function. As more studies investigate the roles of H3K9me2/3 in repression of coding genes, it seems likely that new pathways will be uncovered that are necessary to create different patterns of H3K9me2/3 in different tissues for maintenance of proper cell fate.
The expression of germline-specific genes in somatic tissues leads to a variety of adverse consequences in diverse animal species. These include L1 starvation and reduced apoptosis during development in C. elegans synMuv B mutants, tumor formation in Drosophila l(3)mbt mutants, and poor outcomes in human tumors that express germline genes (Janic et al. 2010; Petrella et al. 2011; Whitehurst 2014; Al-Amin et al. 2016). Thus, there is a need across species to repress germline gene expression in the soma to facilitate proper development and somatic function. Our data suggest that repression of germline genes during development in somatic tissues through H3K9me2 may be a conserved mechanism. As in C. elegans embryonic somatic cells, mammalian ES cells also repress expression of germline genes (Blaschke et al. 2013). Mouse ES cells have been shown to lose repression of germline genes when treated with vitamin C (Blaschke et al. 2013; Ebata et al. 2017). Interestingly, ectopic expression of germline genes upon exposure of ES cells to vitamin C is dependent on loss of H3K9me2 and DNA methylation. Loss of H3K9me2 at germline genes in response to vitamin C appears to be through the activity of H3K9 demethylases (Ebata et al. 2017). In contrast, in C. elegans synMuv B mutants, germline genes likely lose H3K9me2 due to failure in the initial HMT-catalyzed deposition of H3K9me2 during development. The conservation of H3K9me2 on germline genes and its role in repressing these genes in developing somatic lineages may represent an ancient regulatory role for H3K9me2. Since in both C. elegans and Drosophila repression of germline genes in the soma is through complexes known to interact with chromatin (Janic et al. 2010; Petrella et al. 2011; Wu et al. 2012), it will be interesting to see if ectopic expression of germline genes in human somatic tumors is due to loss of these conserved complexes. Finally, not all germline genes, but only a specific subset, are ectopically expressed in these models. Why only certain germline genes are vulnerable to misexpression, if those genes are the same across species, and which cellular processes are disrupted as a result of germline gene misexpression singularly or as a group, are open questions. Further investigation into these questions could have broad implications for understanding conserved basic chromatin mechanisms and therapeutic targets for cancer treatment.
ACKNOLEDGMENTS
Many thanks to Anita Manogaran for comments and discussion of the manuscript.
Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). This work was supported by a NIH grants R00GM98436 and R15GM122005 to L.N.P and NIH grant R01GM34059 to S.S.