Impaired removal of H3K4 methylation affects cell fate determination and gene transcription

Post-translational modifications of histone proteins play a role in many DNA-based biological processes (Bannister and Kouzarides, 2011; Strahl and Allis, 2000) and their regulation is essential during development (Greer and Shi, 2012; Kouzarides, 2007). Methylation of lysine residues, which can be mono-, di- and tri-methylated, is coupled to different outcomes in terms of transcription (Martin and Zhang, 2005). For example, tri-methylation of histone 3 lysine 27 or lysine 9 (H3K27me3 and H3K9me3) is often associated with transcription silencing, whereas tri-methylation of lysine 4 (H3K4me3) is generally considered a mark for active transcription (Ruthenburg et al., 2007). H3K4me3 is enriched at transcription start sites (TSSs) of active genes (Barski et al., 2007; Schübeler et al., 2004), where it acts as a docking site for the general transcription factor TFIID and for chromatin remodeling complexes (Vermeulen et al., 2007), thus facilitating transcription (Clapier and Cairns, 2009). Mono-methylated H3K4 (H3K4me1) marks enhancer regions and, when together with acetylated H3K27 (H3K27ac), is predictive for active enhancers (Heintzman and Ren, 2007; Hon et al., 2009; Ong and Corces, 2011). Thus, various methylation states of H3K4 can control transcription by defining different cis-regulatory regions. The level of H3K4 methylation is regulated by lysine methyltransferases (KMTs) and lysine demethylases (KDMs) that add or remove methyl marks, respectively. Members of the KDM1 and KDM5 families of histone demethylases specifically remove methyl groups from H3K4. Although the KDM1 (also known as LSD) family is required for H3K4me1/2 demethylation to H3K4me0 (Rudolph et al., 2013; Shi et al., 2004), the KDM5 (also known as JARID1) family is responsible for the demethylation of H3K4me2/3 to H3K4me1 (Christensen et al., 2007; Iwase et al., 2007; Lee et al., 2007; Sims and Reinberg, 2006; Tahiliani et al., 2007; Yamane et al., 2007).

The KDM5 family is composed of four genes in mammals; KDM5A-D, sharing similar domain structure comprising a JmjN, a Arid/Bright and a C5HC2 zinc-finger domain, two or three plant homeodomains (PHD) and the catalytic JmjC domain, which are present in almost all known histone demethylases (Pedersen and Helin, 2010). The catalytic activity of KDM5 family members as demethylases specific for H3K4me3/2 has been extensively tested in vitro (Christensen et al., 2007; Iwase et al., 2007). Functionally, several studies indicate that the KDM5 family members are recruited at TSSs where they modulate the level of H3K4me3, acting therefore as transcriptional repressors (Kooistra and Helin, 2012; Lopez-Bigas et al., 2008; Schmitz et al., 2011; Yamane et al., 2007). More recently, KDM5B and KDM5C have been found to be enriched at enhancer regions, which are characterized by the presence of high H3K27ac and H3K4me1, low H3K4me3 and the acetyltransferase P300. This genomic localization suggests the intriguing possibility that KDM5 members might also regulate enhancer activity by controlling the H3K4me3/H3K4me1 ratio at these regulatory sites (Kidder et al., 2014; Outchkourov et al., 2013). However, this hypothesis is still missing in vivo validation.

The biological functions of the KDM5 members have been studied by analyzing the phenotypes of knockout animals (Albert et al., 2013; Blair et al., 2011; Catchpole et al., 2011; Dey et al., 2008; Iwase et al., 2016; Schmitz et al., 2011) and the results support a role of the KDM5 family members in cell fate determination and differentiation. However, our understanding of the molecular mechanisms behind the observed phenotypes is limited in part by the functional redundancy of the KDM5 family members. The KDM5 family is conserved during evolution. Drosophila melanogaster possesses one ortholog of the KDM5 family, LID (Little Imaginal Discs), a trithorax group protein required for viability (Gildea et al., 2000; Secombe et al., 2007). Similarly, a unique member of the KDM5 class, RBR-2, is encoded in Caenorhabditis elegans. Interestingly, animals mutant for rbr-2 are viable and show mild phenotypes related to lifespan regulation (Alvares et al., 2014; Greer et al., 2010, 2011), germline maintenance (Alvares et al., 2014), axon guidance (Mariani et al., 2016) and vulva formation (Christensen et al., 2007). Therefore, C. elegans offers a unique opportunity to study the functional roles of H3K4me3/2 demethylase in the absence of gene redundancy.

During C. elegans vulva development, the coordinate action of three evolutionarily conserved signal transduction pathways (Ras, Notch and Wingless pathways) controls the differentiation of the six vulva precursor cells (VPCs), named P3.p to P8.p, which form the egg-laying organ during post-embryonic development. Each VPC has the potential to adopt a primary, secondary or tertiary fate (Sternberg and Horvitz, 1986; Sulston and Horvitz, 1977), depending on the signal received. Cells adopting a primary (P6.p) or secondary (P5 and P7.p) fate give rise to dividing vulva cells, whereas cells adopting a tertiary fate (P3.p, P4.p and P8.p) do not divide and fuse with the epidermis. Thus, vulva formation represents a powerful system to reveal the role of RBR-2 in cell fate acquisition and differentiation and to explore the relevance and the site of action of its catalytic activity.

Here, we show that the enzymatic activity of RBR-2 is required cell-autonomously in VPCs for the acquisition of the secondary cell fate, at least in part, by promoting the expression of lin-11 transcription factor. RBR-2 seems to exert this function by regulating the epigenetic signature of the lin-11 enhancer region. Furthermore, genome-wide analyses indicate that RBR-2 reduces H3K4me3 levels both at transcription start sites and in upstream regions and acts both as a transcriptional repressor and activator.

The rbr-2 gene encodes a protein of 1477 amino acids containing several evolutionarily conserved domains (Fig. 1A). Three deletion alleles of rbr-2 are available (Fig. 1A; Supplementary Materials and Methods), all lacking the catalytic JmjC domain. Consistently, an increase in global H3K4me3 levels at fourth larval stage (L4) is observed by western blot analysis (Fig. 1B) (Christensen et al., 2007; Greer et al., 2010; Mariani et al., 2016). For our analyses we used the out-of-frame tm3141 allele in which a large portion of the RBR-2 protein is deleted.

At L4 stage, the vulva of a wild­-type animal is symmetrically shaped with a so-called ‘Christmas tree’-like structure (Fig. 1C). In contrast, about one third of the rbr-2(tm3141) animals exhibit an asymmetrically shaped vulva and a small fraction (4%) had an additional invagination (Fig. 1C,D). Similar defects were observed in animals with mutations in the other rbr-2 alleles (Fig. S1A,B). After knockdown of rbr-2 by RNA interference (RNAi), the vulva showed, albeit at lower penetrance, morphological defects similar to the one observed in the tm3141 mutant (Fig. 1C,D). As re-expression of rbr-2 under the control of its own promoter rescued the vulval phenotype of the tm3141 allele (Fig. 1C,D; Table 1), we concluded that rbr-2 is required for proper vulva formation.

To understand whether the abnormal vulva morphology found in rbr-2 mutants is caused by altered cell fates, we examined the role of RBR-2 in specification of the vulva precursor cells (VPCs). VPC fate decisions were assessed in rbr-2(tm3141) and wild-type animals by analyzing the pattern of expression of previously described reporter constructs egl-17p::GFP (for the primary cell fate), lip-1p::GFP, lst-5p::YFP and lin-11p::GFP (for the secondary cell fate), and dpy-7p::GFP (for the tertiary cell fate). In wild-type L3 animals, lip-1p::GFP, lst-5p::YFP and lin-11p::GFP reporters are specifically expressed in P5.p and P7.p progeny cells, upon acquisition of the secondary cell fate (Berset et al., 2001; Gupta and Sternberg, 2002; Li and Greenwald, 2010). In rbr-2(tm3141) animals, expression of all three secondary cell fate transgenes was missing in a subset of P5.p or P7.p cells at the L3 stage (Fig. 2A-C,E). Furthermore, the expression of egl-17, which in wild-type at the L4 stage specifically marks the descendants of the secondary cell fate lineage, vulC and vulD, (Berset et al., 2005; Burdine et al., 1998), is partially lost in rbr-2(tm3141) animals, correlating with abnormal vulva morphology observed at this stage (Fig. 2D,E). In contrast, markers for the primary (egl-17p::GFP at L3 stage; Burdine et al., 1998) and the tertiary cell fate (dpy-7p::GFP; Myers and Greenwald, 2005) were normally expressed in rbr-2(tm3141) animals (Fig. S1C,D).

Aberrant acquisition of cell fate can interfere with VPC cell division (Sternberg and Horvitz, 1986; Sundaram and Greenwald, 1993) and toroid ring formation (Sharma-Kishore et al., 1999). Indeed, in rbr-2(tm3141) animals the descendants of the secondary cell lineage exhibit an abnormal pattern of cell division and an abnormal number of cells (Fig. S1E,F), and vulva toroids, analyzed using the ajm-1::GFP marker (Köppen et al., 2001), appearing disorganized and missing (Fig. 2F). Thus, RBR-2 is required for cell fate determination during vulva development and, when lost, the secondary cell fate induction is compromised, leading to abnormal cell division behavior in descendants of the P5.p and P7.p lineages and to aberrant morphogenesis of the vulva.

Consistent with its role in vulva development, a translational RBR-2::GFP reporter driven by the endogenous rbr-2 promoter showed RBR-2 expression in VPCs and in the anchor cell (Fig. S2). To determine the cellular focus of RBR-2 action, we re-expressed RBR-2 in an rbr-2(tm3141) background in different tissues and tested the transgenic lines for the rescue of the aberrant expression of lip-1p::GFP in P5.p and P7.p (Table 1). We used the rbr-2 endogenous promoter (rbr-2p) and well-characterized tissue-specific promoters expressed in tissues relevant for vulva formation, such as hypodermis (dpy-7p), VPCs (lin-31p) and anchor cell (cdh-3p), to drive the expression of RBR-2 (Bulow et al., 2004; Gilleard et al., 1997; Myers and Greenwald, 2005; Pettitt et al., 1996; Tan et al., 1998). Expression of RBR-2 under its endogenous promoter and in VPCs rescues the aberrant lip-1p::GFP expression in P5.p and/or P7.p in rbr-2(tm3141) worms, whereas expression of RBR-2 in the hypodermis or anchor cell did not (Table 1). These results are consistent with a cell-autonomous role of RBR-2 in the VPCs to determine the secondary cell fate.

During the L3 stage, VPCs are patterned through the interaction of two main signaling pathways acting in the vulva. The LIN-3 (a member of the EGF family) inductive signal, derived from the anchor cell, induces the closest VPC, P6.p, to adopt the primary vulval fate and to produce ligands for LIN-12/Notch. These ligands constitute a lateral signal activating the Notch pathway in the neighboring VPCs, P5.p and P7.p, thus promoting the secondary vulval fate. The findings that RBR-2 acts in VPCs to support the secondary cell fate specification and that its loss leads to downregulation of LIN-12/Notch targets suggest that LIN-12/Notch signaling, the main determinant of the secondary cell fate, might be compromised in rbr-2(tm3141) animals. To test this possibility, we assessed the sensitivity of this pathway to the lack of rbr-2, using previously described lin-12 mutant alleles (Ambros, 1999; Greenwald et al., 1983; Levitan and Greenwald, 1995; Sundaram and Greenwald, 1993). If RBR-2 promotes LIN-12/Notch signaling, we would expect that loss of rbr-2 would reduce the effect of the gain-of-function lin-12(n137n460) allele and enhance that of the lin-12(n676n930) hypomorphic allele. In agreement with our hypothesis, loss of rbr-2 slightly reduced the defects observed in lin-12(n137n460) and increased defects in lin-12(n676n930) animals (Table S1A and B).

The LIN-12/Notch pathway is thought to act in a dual manner in the vulva precursor cells, by promoting the secondary cell fate and by inhibiting the primary fate in P5.p and P7.p (Kenyon, 1995; Sternberg, 2004). Whereas promotion of the secondary cell fate includes the upregulation of LIN-12/Notch target genes, primary cell fate inhibition is achieved by the inhibition of the EGF/Ras pathway (Berset et al., 2001; Yoo et al., 2004). To test whether loss of rbr-2 results in increased sensitivity to EGF/Ras signaling, we used sensitized genetic backgrounds where the EGF/Ras pathway is compromised. Consistent with previous findings from our lab (Christensen et al., 2007), reduction of rbr-2 by RNAi in lin-15AB(n765ts) animals, a mutation that causes increased EGF levels in a temperature-dependent manner (Cui et al., 2006), resulted in enhanced multivulva (Muv) phenotype at permissive temperature (Table S2). Similar results were observed when rbr-2 was downregulated in let-60(n1046gf) animals, a gain-of-function mutation of let-60 (encoding a Ras family protein) (Table S2). However, lack of rbr-2 did not lead to a Muv phenotype in combination with any of the synMuv class A, B or C genes, indicating that rbr-2 is not a synthetic multivulva gene (Table S3) (Christensen et al., 2007).

Overall these analyses indicate that loss of rbr-2 affects the secondary cell fate by reducing LIN-12/Notch signaling and compromises the primary cell fate inhibition in sensitized genetic backgrounds.

To investigate if the catalytic activity of RBR-2 is required for vulva development, we performed rescue experiments using a rbr-2p::rbr-2(DD)::mCherry construct (where DD stands for demethylase dead) in which two conserved amino acids in the catalytic core of the JmjC domain, H514 and E516, were mutated to two alanine residues (Fig. S3A). RBR-2(DD) shows a similar expression level and pattern as the wild-type transgene (Fig. S3B), but it fails to restore normal H3K4me3 levels (Fig. S3C). We expressed RBR-2(DD) in rbr-2(tm3141) worms carrying the lip-1p::GFP marker and tested its ability to restore the correct GFP expression in P5.p and P7.p. Whereas the re-introduction of wild-type RBR-2 in rbr-2(tm3141) animals resulted in the rescue of the aberrant lip-1p::GFP marker expression, the expression of RBR-2(DD) did not (Table 1). This experiment demonstrates that RBR-2 enzymatic activity is strictly required for the secondary cell fate acquisition and led us to further explore the relevance of H3K4me3 regulation in cell fate commitment.

Although RBR-2 has been shown to regulate H3K4me3 at global level (Christensen et al., 2007; Greer et al., 2010; Wang et al., 2011) and at TSSs of specific target genes (Lopez-Bigas et al., 2008; Mariani et al., 2016), the impact of loss of RBR-2 on genome-wide H3K4me3 levels and distribution remains unexplored. To this end, we performed chromatin immunoprecipitation for H3K4me3 followed by deep sequencing (ChIP-seq) in rbr-2(tm3141) and wild-type worms. Consistent with a previous study (Liu et al., 2011), in wild-type animals H3K4me3 seems distributed along the chromosomes, with a slight depletion at the distal regions of the chromosomes, compared with the central regions (Fig. S4). The chromosomal distribution of H3K4me3 seems unchanged in rbr-2(tm3141) animals (Fig. S4). Heat-map analysis (Fig. 3A) of the H3K4me3 peaks clustered with respect to the TSS (±5 kb) reveals that in wild-type animals H3K4me3 is mainly located at TSSs (Fig. 3A, cluster 5), as previously reported (Gu and Fire, 2010; Liu et al., 2011). We also observed H3K4me3 peaks in region upstream of the TSS (clusters 6 and 3) and, to a minor extent, in the gene bodies (cluster 4). The general localization of the H3K4me3 peaks is maintained in rbr-2(tm3141) animals, however, in agreement with increased global levels of H3K4me3 observed by western blot, a higher number of peaks is found in rbr-2(tm3141) worms with respect to wild type [6780 in rbr-2(tm3141) compared with 5122 in wild type]. Accordingly, the heat-map analysis (Fig. 3A) shows that in rbr-2(tm3141) animals the number of genes with H3K4me3 in close proximity to the TSS (cluster 5) is increased. Similarly, in rbr-2 animals, we found a larger number of genes with H3K4me3 located upstream of the TSS (clusters 6 and 3) and in the gene bodies (cluster 4). Examples of aberrant H3K4me3 profiles in rbr-2 are shown in Fig. 3B.

To gain information about the biological processes that might be altered in rbr-2(tm3141) animals, we performed gene ontology (GO) analysis of the genes with aberrant H3K4me3 in rbr-2(tm3141) worms identified by differential analysis for ChIP-sequencing using macs2diff (Zhang et al., 2008) (Fig. 3C). In agreement with the role of rbr-2 in vulva development (Christensen et al., 2007) and lifespan (Greer et al., 2010, 2011), genes implicated in genitalia development and aging are significantly enriched. GO analysis of clusters 2-6 is presented in Fig. S5. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis reveals that several conserved pathways, including the Notch and EGF/MAPK pathways, important for vulva formation, are overrepresented among the genes with deregulated H3K4me3 in rbr-2(tm3141) worms (Fig. 3D).

Overall, these experiments indicate that RBR-2 is required to reduce the H3K4me3 levels in different regions of the genome and suggest, consistently with the phenotypes we described in rbr-2(tm3141) animals, a potential role of RBR-2 in regulating the H3K4me3 level in genes implicated in vulva development.

H3K4me3 level has been correlated to gene expression (Ruthenburg et al., 2007) and, accordingly, loss or downregulation of KDM5 members has been associated with aberrant transcription, albeit the effect is often limited to minor gene expression changes (Albert et al., 2013; Alvares et al., 2014; Christensen et al., 2007; Dey et al., 2008; Iwase et al., 2007, 2016; Schmitz et al., 2011; Tahiliani et al., 2007). As the contribution of RBR-2 to transcriptional control at a global level has not been assessed, we analyzed the transcriptome of rbr-2(tm3141) worms by RNA sequencing (RNA-seq). We found almost 1500 genes differentially regulated in mutant animals (P<0.005), of which 814 genes are upregulated and 650 genes are downregulated in rbr-2(tm3141) animals, compared with wild type. The median changes of gene expression range from 3.2- and 3.6-fold for up- and downregulated genes, respectively (Fig. 4A), indicative of a robust transcriptional deregulation occurring in mutant animals. GO analysis of genes upregulated in rbr-2 indicates that, consistently with the function of RBR-2 in aging (Greer et al., 2010, 2011), genes regulating lifespan are among the most enriched classes (Fig. 4B, top panel). Selected genes with aberrantly increased expression were validated by qPCR (Fig. 4C). In agreement with the notion that high level of H3K4me3 at TSSs is generally associated with active transcription, we found increased H3K4me3 levels at TSSs of genes upregulated in rbr-2 compared with wild type by ChIP-qPCR (Fig. 4D). Furthermore, by ChIP-qPCR using GFP antibody in a transgenic line expressing integrated RBR-2::GFP, we detected an efficient recruitment of RBR-2::GFP specifically at the TSSs of upregulated genes (Fig. 4E). Overall, these results are consistent with a role of RBR-2 as a transcriptional repressor by removing H3K4me3 at TSSs, and are in agreement with previous studies (Martin and Zhang, 2005).

More unexpectedly, several hundreds of genes are downregulated in rbr-2(tm3141) animals (Fig. 4A). Consistent with the role of RBR-2 in vulva formation and with our ChIP-seq results, GO analysis of genes downregulated in the rbr-2(tm3141) allele identified the ‘vulva development’ class as enriched (Fig. 4B, lower panel). Among others, lin-11, osm-11 and lin-31, known regulators of vulva development (Freyd et al., 1990; Komatsu et al., 2008; Miller et al., 1993) were validated as significantly reduced by qPCR (Fig. 4F). The overlap between genes with aberrant levels of H3K4me3 and altered expression is shown in the Venn diagram in Fig. S6. By using the hypergeometric test, we found that clusters 3 and 6 [H3K4me3 increased at 5′ regions in rbr-2(tm3141) versus wild-type animals] are significantly enriched in downregulated genes (P=0.00485) but not in upregulated genes (P=0.99999) and clusters 4 and 5 [H3K4me3 increased at TSSs and gene bodies in rbr-2(tm3141) versus wild-type animals] are significantly enriched in both upregulated (P=0.0002) and downregulated genes (P=2×10^–8). This analysis suggests that although increased levels of H3K4me3 at 5′ of the gene might be predictive of reduced expression, increased level of the mark at the TSS, in agreement with other studies, is not always prognostic for upregulated transcription.

These results, together with the fact that the catalytic activity of RBR-2 is required to positively regulate the expression of many secondary cell fate markers in VPCs (Fig. 2), suggest that RBR-2 can act as a transcriptional activator.

As ChIP-seq analysis of the rbr-2(tm3141) allele identified increased levels of H3K4me3 in regions upstream the TSSs (cluster 3 and 6), where distal enhancer elements are generally found (Bulger and Groudine, 2010), we postulated that RBR-2 could positively regulate the rate of transcription by controlling enhancer signatures and activities, as also recently proposed for RBR-2 mammalian homologues (Kidder et al., 2014; Outchkourov et al., 2013). Active enhancers are characterized by the presence of high levels of H3K4me1 and H3K27ac, low level of H3K4me3 and by the binding of CBP-1 (CBP/p300 homologue), responsible for H3K27 acetylation (Calo and Wysocka, 2013; Heintzman and Ren, 2007; Rada-Iglesias et al., 2011). In support of our hypothesis, we found that loss of rbr-2 results in a significant reduction of global level of H3K4me1 and of H3K27ac (Fig. 5A). To validate this hypothesis in vivo, we took advantage of the role of RBR-2 in vulva development and we analyzed the histone modifications in wild-type and rbr-2(tm3141) animals at the lin-11 locus, encoding a transcription factor previously implicated in vulva formation and regulated, at least in part, by LIN-12/Notch signaling (Gupta and Sternberg, 2002; Gupta et al., 2003; Newman et al., 1999). The selection of this gene was based on the evidence that its expression is reduced in rbr-2(tm3141) worms (as observed by RNA-seq, qPCR and marker analysis) and on the fact that an enhancer element (called here lin-11 enh) of 625 bp, specifically required for lin-11 expression in VPCs, was previously identified (Gupta and Sternberg, 2002; Gupta et al., 2003; Marri and Gupta, 2009; Newman et al., 1999). In wild-type animals, RBR-2 (Fig. 5B) and H3K4me1 and H3K27ac (Fig. 5C) are enriched at lin-11 enh when compared with control regions, confirming at the molecular level the regulatory role of this enhancer element. Strikingly, in rbr-2(tm3141) animals, H3K4me1 and H3K27ac in the lin-11 enh region seem reduced to background levels, as compared with wild type, with concomitant increased H3K4me3 (Fig. 5C). As expression of EGL-17 in vulC and vulD is reduced in a subset of rbr-2(tm3141) animals (Fig. 2D,E), we analyzed the egl-17 enhancer that drives the expression of egl-17 in vulC and vulD cells (Cui and Han, 2003) and obtained similar results for H3K4 methylation (Fig. S7A,B). Of note, when two genes, unc-54 and myo-3, unrelated to vulva development and with well-described enhancer regions (Jantsch-Plunger and Fire, 1994; Okkema et al., 1993), were analyzed (Fig. S7C-E), we did not observe any enrichment of RBR-2::GFP at their enhancer regions, nor changes in H3K4me1 and H3K27ac levels in rbr-2 animals, compared with wild type. In agreement, the expression levels of UNC-54 and MYO-3 seem to be unchanged in rbr-2 animals.

To functionally test whether the LIN-11 level is relevant for the proper secondary cell fate acquisition in our experimental setup, we analyzed lip-1p::GFP expression in wild-type animals after reducing the expression of lin-11 by RNA interference (lin-11 RNAi). Downregulation of lin-11 resulted in reduced lip-1p::GFP levels and a comparable effect was measured when lin-11 was reduced in the rbr-2(tm3141) allele (Table S4), indicating that lin-11 and rbr-2 act in the same pathway regulating lip-1 expression. Reduced expression of lip-1p::GFP was also observed in lin-11(n566) hypomorph mutant animals, further proving that lin-11 is required for proper lip-1p::GFP expression (Table S4).

Overall, these experiments indicate that RBR-2 can directly control the activity of lin-11 vulva enhancer and suggest that it can act as a transcriptional activator by regulating the epigenetic signature of enhancer elements.

The evidence that in rbr-2(tm3141) animals the level of H3K27ac is decreased suggests that the role of RBR-2 in regulating enhancer activity should be exerted in cooperation with a lysine acetyltransferase. In mammals, the establishment of H3K27ac at enhancer regions is mediated by the closely related members of the lysine acetyltransferase class 3 (KAT3); CBP (also known as CREBBP and KAT3A) and p300 (also known as EP300 and KAT3B) (Heintzman and Ren, 2007; Holmqvist and Mannervik, 2013; Jin et al., 2011; Visel et al., 2009) and, accordingly, their genomic distribution is commonly used, together with H3K4me1, for genome-wide enhancer mapping (Calo and Wysocka, 2013; Ghisletti et al., 2010; Kim et al., 2010; Rada-Iglesias et al., 2011). In C. elegans, a member of the KAT3 family, CBP-1, is required for embryonic development (Shi and Mello, 1998; Victor et al., 2002) and for antagonizing Ras activity during vulva formation (Eastburn and Han, 2005). We hypothesized that if cbp-1 is required, similar to rbr-2, for the regulation of enhancer signatures during vulva development, its removal should result in a rbr-2-like phenotype. As cbp-1 loss leads to embryonic lethality, we used RNA interference to explore the role of cbp-1 in vulva cell fate determination. Reduced expression of cbp-1 resulted in a decreased global level of H3K27ac (Fig. S8A), confirming the CBP-1 acetyltransferase activity towards H3K27 and indicating the effective downregulation of cbp-1 expression. Strikingly, RNA interfered animals showed aberrant secondary cell fate acquisition, as judged by the pattern of expression of lip-1p::GFP (Fig. 6A,B), thus reproducing rbr-2(tm3141) defects. Reduction of cbp-1 in a rbr-2 background resulted in a similar penetrance of the phenotype (Fig. 6A,B), suggesting a cooperation of rbr-2 and cbp-1 in secondary cell fate acquisition. Importantly, reducing cbp-1 levels by RNAi did not affect rbr-2 mRNA levels or its recruitment at lin-11 enh (Fig. S8B,C). Lack of specific antibodies or transgenic lines prevented us from exploring the recruitment of CBP-1 at the enhancer region of lin-11, however, after cbp-1 interference, we observed reduced levels of lin-11 expression (Fig. 6C), suggesting that CBP-1 might be the acetyltransferase cooperating with RBR-2 in the establishment of the epigenetic signature of lin-11 enhancer required for its correct expression.

In this paper we investigated the biological consequences of loss of RBR-2, the unique H3K4me2/3 demethylase in C. elegans. We found that RBR-2 acts cell-autonomously to regulate cell fate specification of vulva precursor cells and its loss results in a compromised expression pattern of secondary cell fate markers and abnormal vulva morphology. The reduced expression of secondary cell fate markers and the genetic interactions with lin-12 mutants suggest that rbr-2 acts as a modulator of the LIN-12/Notch pathway. The regulatory function of RBR-2 on LIN-12/Notch signaling occurs specifically in the VPCs, as loss of rbr-2 does not result in aberrant anchor cell/ventral uterine fate decision (Table S5), which is also under the control of LIN-12 expression in the anchor cell (Greenwald et al., 1983). This result is in line with previous studies on the Drosophila KDM5 homolog, LID (Di Stefano et al., 2011; Liefke et al., 2010; Moshkin et al., 2009); however, differently to LID, RBR-2 seems to have a positive role in the transcriptional regulation of Notch target genes. Loss of RBR-2 also lead to a multivulva phenotype when EGF/Ras signaling was increased. This evidence suggests that, in the absence of RBR-2, VPCs cannot fully commit to the secondary cell fate and are more sensitive to the inductive signaling initiated by EGF/LIN-3. However, RBR-2, differently from other chromatin factors (Fay and Yochem, 2007), does not act in the synthetic multi-vulva pathways. This is in agreement with the fact that RBR-2 acts in VPCs and that its expression is not required in hypodermal cells for VPC fate determination, where the action of many synMuv genes has been identified to suppress the expression of lin-3 (Cui et al., 2006).

Rescue experiments provide unequivocal evidence that the function of RBR-2 in cell fate determination depends on its catalytic activity, thus establishing a straightforward link between the regulation of H3K4me3 levels and cell fate commitment. By genome-wide analyses, we found that RBR-2 controls the level of H3K4me3 in many genes, both at TSSs and in regions upstream of the TSSs. The defective modulation of H3K4me3 in rbr-2 mutant animals is accompanied by aberrant transcription of several hundreds of genes, as estimated by the transcriptome analysis. In agreement with the role of the KDM5 family members acting as transcriptional repressors, we identified direct RBR-2 target genes with increased levels of H3K4me3 at TSSs and upregulated in rbr-2 mutant animals. We also found genes downregulated in rbr-2 mutants, including genes regulating vulva development. This latter result, together with the identification of increased H3K4me3 level in regions upstream of the TSSs, led us to investigate if RBR-2 could act also as a positive regulator of transcription by promoting enhancer activity. Recent publications hypothesized that mammalian homologs of RBR-2, KDM5B and KDM5C, might support transcription by removing spurious H3K4me3/2 and maintaining the level of H3K4me1 at enhancers (Kidder et al., 2014; Outchkourov et al., 2013). We could validate this hypothesis in vivo, by testing the molecular signature of the lin-11 locus, for which the enhancer element required for lin-11 expression specifically in VPCs is well characterized (Gupta and Sternberg, 2002; Gupta et al., 2003). Our results indicate that RBR-2 is recruited to the lin-11 regulatory element to maintain the characteristic enhancer signature; low levels of H3K4me3 and high levels of H3K4me1 and H3K27ac. Similar results were obtained for egl-17, another gene implicated in vulva formation and regulated by RBR-2. Owing to the lack of functionally defined enhancers for other downregulated genes, we could not extend our analysis to other candidate targets. However, western blot analysis on global H3K4me1 and H3K27ac level in rbr-2 mutant suggests that the RBR-2 role as a regulator of enhancer activity might be exerted on a large scale. Overall, this evidence strongly suggests that RBR-2 can have both positive and negative effects on transcription. It should be noted that despite the considerable effect on transcription, the vulva phenotype associated with rbr-2 loss is not very penetrant. This emphasizes that fact that cell fate determination of VPCs is robust and can accommodate variations in the developmental paths without changes in the final output, as previously suggested (Félix and Barkoulas, 2012).

The reduced level of H3K27ac in rbr-2 mutant animals suggests a correlated involvement of a histone acetyltransferase activity exerted, at least in part, by CBP-1. Interestingly, whereas qualitatively the defects observed in cbp-1 mutant animals are similar to those observed in animals lacking rbr-2, their penetrance is higher. This might suggest that lack of cbp-1 has a stronger effect on transcription activation, or it could reflect the ability of CBP-1 to regulate the transcription of many genes involved in cell fate acquisition, other than lin-11. This might also be true for RBR-2, as several genes implicated in vulva development seem reduced in the rbr-2 genetic background, suggesting that the rbr-2 vulval phenotype might be the final outcome of deregulation of multiple genes. In agreement with this possibility, re-expression of lin-11::GFP in rbr-2 mutants failed to rescue the lip-1p::GFP expression to wild-type level (Table S6). This result suggests that lin-11 expression is required but not sufficient for ensuring proper lip-1 transcription in a rbr-2 genetic background, and indicates that RBR-2 might also control the expression of other genes required in VPCs for secondary cell fate specification.

Based on the broad pattern of expression of RBR-2 and on the evidence that it can regulate the expression of hundreds of genes, it is expected that RBR-2 controls several biological processes. Indeed, RBR-2 has been previously implicated in the regulation of lifespan, germline functions and actin remodeling. In striking agreement with these reported functions, our global ChIP analysis identified, besides vulva genes, genes implicated in aging, cell cycle/meiosis and cytoskeleton dynamics to be deregulated in rbr-2 mutants, thus providing a list of potentially useful RBR-2 targets to study mechanisms related to aberrant H3K4 methylation.

The identification of RBR-2 target genes in specific cell types and of factors required for RBR-2 recruitment at different genomic regions will be the object of future investigations.

C. elegans worms were grown under standard laboratory conditions (Brenner, 1974) at 20°C unless otherwise indicated. Strains, genotype and source of C. elegans used are provided in Table S7.

Nomenclature for describing VPC fates follows that of Sternberg and Horvitz (1986) in which fates are described in terms of the axes of the last nuclear divisions of the lineage; each letter refers to the axis of a single division. N, no division. L, lateral division. T, transverse division. jcIs1[ajm-1::GFP+unc-29(+)+rol-6(su1006)] was used to delineate the cell borders.

Three mutant alleles of rbr-2 are available. The allele rbr-2(tm3141) carries a deletion of 365  bp and an insertion of 8 bp, leading to a premature stop codon. The rbr-2(ok2544) and rbr-2(tm1231) alleles contain in-frame deletions of 1311 bp and 648 bp, respectively, and they might be not null mutants. The alleles tm1231 and tm3141 were identified at the National BioResource Project (NBRP), Japan, and the allele ok2544 was provided by the C. elegans Gene Knockout Consortium (CGK) at OMRF. All alleles were backcrossed at least three times with wild-type animals before phenotypic analyses.

Equal amounts of whole worm lysates were run on SDS-PAGE. The following antibodies were used: H3K4me3 (Abcam, ab8580; 1:500), H3 (Abcam, ab1791; 1:30,000), H3K4me1 (Abcam, ab8895; 1:500) and H3K27ac (Abcam, ab4729; 1:1000). Western blots were quantified using ImageJ (National Institutes of Health).

For the RBR-2::GFP and RBR-2::mCherry constructs, a 6081 bp fragment of rbr-2 including the entire coding region and a 2540 bp promoter region was PCR-amplified from wild-type genomic DNA and inserted in the multiple cloning site of pDONR pCR8 and pDONR P4-P1R vectors, respectively (Gateway cloning system, Life Technologies). The mCherry followed by unc-54 3′UTR was cloned into pDONR P2-RP3 and the tissue-specific promoters into pDONR P4-P1R. The pDONR P2-RP3 vector containing the GFP sequence followed by unc-54 3′UTR was a generous gift from Erik Jorgensen (Howard Hughes Medical Institute, University of Utah, UT, USA). Final constructs were cloned into the pDEST R4-R3 destination vector. For the RBR-2(DD) construct, the RBR-2(WT) construct was mutated using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). Specifically, the DNA sequence was mutated so that the histidine at position 514 (H514) and the glutamic acid at position 516 (E516) were changed to alanines. All plasmids were verified by sequencing. To generate an integrated line, the vectors pDONR P4-P1R [rbr-2p], pDONR pCR8 [rbr-2] and pDONR P2-RP3 [GFP::unc-54 3'UTR] were recombined with the destination vector pCG150 (Addgene), which includes the unc-119 rescue fragment into the vector backbone pDEST R4-R3. Primer sequences are listed in Table S9.

Transgenic lines were obtained through microinjection (Mello et al., 1991). Constructs were injected as extra-chromosomal arrays at 20 ng/μl with ttx-3p::RFP (100 ng/μl) as co-injection marker. To obtain an integrated line, 20 ng/μl of the vector pCG150 [rbr-2p::rbr-2::GFP::unc-54 3'UTR] [unc-119] was injected in [unc-119(ed3)] genetic background, integrated by UV irradiation and backcrossed three times with wild-type animals before analyses.

RNA interference (RNAi) was performed by feeding and carried out as described previously (Timmons et al., 2001). For rbr-2, two different non-overlapping fragments were used as described previously with similar results (Christensen et al., 2007). The lin-11, cbp-1 and synMuv gene RNAi construct were obtained from the C. elegans RNAi feeding library (J. Ahringer's laboratory, Wellcome Trust/Cancer Research UK Gurdon Institute, University of Cambridge, Cambridge, UK). Empty L4440 vector was used as negative control. F2 progeny were scored for vulva defects at L4 stage or lip-1p::GFP expression at Pnp.xx stage (four-cell stage), with the exception of cbp-1 RNAi experiments, for which we plated starved L1 and we analyzed P0 animals. Three individual plates were scored for each variable.

Total RNA was isolated from synchronized worms using TRIzol reagent (Invitrogen) and RNAeasy Minikit (Qiagen). cDNA was synthesized using TaqMan Reverse Transcription kit (Applied Biosystems). qPCR was performed using SYBR Green 2× PCR Master mix (Applied Biosystems) in a Lightcycler 480 Instrument (Roche Life Science). The measures were normalized to rpl-26 mRNA levels unless otherwise indicated. All reactions were performed in duplicates, in at least two independent experiments. All primer sequences are available upon request. The control region used in the ChIP-qPCR experiments is an unrelated intergenic region on chromosome IV. The lin-11 vulva precursor cell-specific enhancer region is located 3169 bp upstream of the lin-11 translational start site (Gupta et al., 2003; Newman et al., 1999).

Vulva cell fate markers and morphology were scored in L3/L4 or at adult stage. Fluorescence microscope and DIC pictures were acquired using an Axiovert 135, Carl Zeiss, Inc. All pictures were exported in preparation for printing using Photoshop (Adobe). Vulva toroid formation was analyzed using the Deltavision microscope (Köppen et al., 2001). Images were deconvolved and merged using softWorRx (Applied Precision).

Chromatin immunoprecipitation (ChIP) was performed with a protocol modified from Kolasinska-Zwierz et al. (2009) using H3K4me3 (Abcam, ab8580), GFP (Abcam, ab290), H3 (Abcam, ab1791), H3K4me1 (Abcam, ab8895), H3K27ac (Abcam, ab4729) and rabbit IgG (Sigma-Aldrich, I8140) antibodies all at 1 μm/ml. At least two independent experiments were performed.

The DNA was sequenced by the Danish National High-Throughput DNA Sequencing Centre. Reads were checked for quality using FastQC and preprocessed to remove adaptors and E. coli sequences and mapped to the C. elegans genome (WS220) using Bowtie2 (Langmead and Salzberg, 2012; Langmead et al., 2009). The number of reads processed and percentage aligned were: wildtype, 40 million, 70.3%; rbr-2, 49 million, 75.2%; wild-type IgG control, 42 million, 50.1%; rbr-2 IgG control, 67 million, 63.3%. Peak calling was performed using macs2diff (Zhang et al., 2008) to pull out peaks differential to either rbr-2 mutant or wild type. UCSC tracks and profile heat maps were prepared using the DeepTools (Ramirez et al., 2014) package. Background subtraction was performed using the bamCompare tool. SES (Diaz et al., 2012) was used to normalize for read depth differences.

RNA was isolated from synchronized worms collected from two independent cultures using a Trizol/chloroform extraction followed by RNeasy Mini preparation with on column DNase I digestion (Qiagen). RNA amplification and sequencing were performed by the Beijing Genomics Institute (BGI). A list of genes up- and downregulated in the rbr-2(tm3141) animals is shown in Table S10.

Bar code and adaptor-cleaned sequences were checked for quality using FastQC and mapped to the C. elegans genome (WS220) with TopHat 2.0.9 (Trapnell et al., 2012) using parameters as described previously (Peltonen et al., 2013). Reads successfully mapped was ∼50% using a criteria of two mismatches. The number of reads processed and percentage aligned were: wild-type replicate 1, 14.6 million, 49.7%; wild-type replicate 2, 3.7 million, 45.0%; rbr-2 replicate 1, 5.1 million, 45.8%; rbr-2 replicate 2, 2.7 million, 57.2%. Mapped reads were analyzed for transcript assembly and differential expression using Cufflinks 2.1.1 with a filter of twofold difference and FDR correction (P<0.05). Gene set enrichment analysis (GSEA) was performed using DAVID 6.7 (Huang da et al., 2009).

Statistical analyses were performed in GraphPad Prism 6 using Fisher's exact test or Student's t-test, when applicable. Values are expressed as mean±s.d. Differences with a P-value <0.05 were considered significant.

We are grateful to Alexandra Avram, Andreea Talos and Line Kerimbäck Kikarsen for technical assistance, to Jens Vilstrup Johansen for data submission and to Kristian Helin for critical reading of the manuscript. Some strains were provided by the Caenorhabditis Genetics Center (CGC), funded by National Institutes of Health Office of Research Infrastructure Program (P40 OD010440). We thank the National BioResource project for C. elegans for providing strains.