Maintenance of Cell Fates by Regulation of the Histone Variant H3.3 in Caenorhabditis elegans

Highlights TLK-1 maintains cell fates by repression of selector genes TLK-1 and downstream H3 chaperone CAF1 inhibit H3.3 deposition Loss of sin-3 suppresses the defect in cell-fate maintenance of tlk-1 mutants AcH4-binding protein BET-1 is necessary for sin-3 suppression Summary Cell-fate maintenance is important to preserve the variety of cell types that are essential for the formation and function of tissues. We previously showed that the acetylated histone H4-binding protein BET-1 maintains cell fate by recruiting the histone variant H2A.z. Here, we report that Caenorhabditis elegans tousled-like kinase TLK-1 and the histone H3 chaperone CAF1 maintain cell fate by preventing the incorporation of histone variant H3.3 into nucleosomes, thereby repressing ectopic expression of transcription factors that induce cell-fate specification. Genetic analyses suggested that TLK-1 and BET-1 act in parallel pathways. In tlk-1 mutants, the loss of SIN-3, which promotes histone acetylation, suppressed a defect in cell-fate maintenance in a manner dependent on MYST family histone acetyltransferase MYS-2 and BET-1. sin-3 mutation also suppressed abnormal H3.3 incorporation. Thus, we propose that the regulation and interaction of histone variants play crucial roles in cell-fate maintenance through the regulation of selector genes.


Introduction
Defects in cell-fate maintenance cause aberrant cell-fate transformation, which can induce tumor formation and tissue malfunction. Conversely, suppression of the mechanisms that maintain cell fate is necessary for efficient reprogramming such as the generation of induced pluripotent stem (iPS) cells (Takahashi and Yamanaka, 2015). Aberrant activation of genes that induce specific cell fates causes abnormal cell-fate transformation (Halder et al., 1995;Riddle et al., 2013).
Thus, the repression of the genes that specify cell fates is critical for maintaining individual cell fates.
Epigenetic marks including histone modifications play important roles in transcriptional repression during development. For example, methylation on lysine 27 of histone H3 (H3K27me) is required to silence developmentally regulated genes such as Hox genes (Ringrose and Paro, 2004). In contrast, the roles of histone variants in transcriptional repression are poorly understood. We previously showed that a histone H2A variant, H2A.z, is required to maintain cell fate in multiple cell lineages in Caenorhabditis elegans .
Subnuclear localization of H2A.z is regulated by an acetylated histone H4binding protein, BET-1, that is also required to maintain cell fate. BET-1 represses selector genes that encode DNA-binding transcription factors (TFs) such as LIM homeodomain protein MEC-3 and CEH-22/Nkx2.5, which induce specific cell fates Shibata et al., 2010).
The selector gene activates transcription of itself and of genes that are required for the specific function of each cell type (Hobert, 2008). Thus, although many studies suggest a role for H2A.z in transcriptional activation, H2A.z also preserves transcriptional repression in the maintenance of cell fate.
In addition to the H2A variant, another major histone variant is the histone H3 variant H3.3, which is often observed on actively transcribed loci (Wirbelauer et al., 2005). Canonical histone H3 and H3 variant H3.3 are deposited by chromatin assembly factor 1 (CAF1) and histone regulator A (HIRA), respectively (Tagami et al., 2004). In cultured cells, CAF1 depletion causes alternative deposition of H3.3 to fill the nucleosome gap at the replication site by HIRA (Ray-Gallet et al., 2011). CAF1 deficiency promotes artificial transdifferentiation, such as induction of iPS cells and the generation of neurons from fibroblasts and of macrophages from pre-B cells (Cheloufi et al., 2015). However, the roles of CAF1 and H3.3 in cell-fate maintenance during development are not known.
Several studies have suggested a relationship between H2A.z and H3.3.
Genome-wide analyses of H2A.z, RNA pol II, and transcription suggest that H2A.z is correlated with transcriptional repression at the poised state (Raisner et al., 2005). Biophysical evidence indicates that H2A.z promotes chromatin compaction (Chen et al., 2013). In contrast, H3.3 promotes gene activation, counteracting H2A.z-mediated transcriptional repression, and impairs H2A.zmediated chromatin compaction, thus reducing higher-order chromatin folding (Chen et al., 2013). Because transcriptional repression by H2A.z is necessary to maintain cell fate, it is important to clarify the role of H3 variants and their regulation to understand the maintenance of cell identity.
Tousled-like kinases (TLKs) are conserved protein kinases in multicellular organisms. They phosphorylate anti-silencing factor 1 (ASF1), which interacts with CAF1 (Klimovskaia et al., 2014). Arabidopsis TLK, Tousled, acts in the maintenance of transcriptional gene silencing and is required for leaf and flower development (Roe et al., 1993;Wang et al., 2007). In C. elegans early embryos, the ortholog TLK-1 is required for chromosome segregation and cytokinesis and promotes transcription (Han et al., 2005;Han et al., 2003;Yeh et al., 2010). Our genetic screening for mutants that are defective in cell-fate maintenance resulted in the isolation of tlk-1 mutants. Here, we analyzed the roles of TLK-1 and CAF1 in cell-fate maintenance and the regulation of H3.3. We also investigated the relationship between mechanisms that regulate H3.3 and H2A.z.

Isolation of tlk-1 mutants by screening for aberrant cell-fate transformation mutants
We previously showed that, in C. elegans, malfunction of the machinery that maintains cell fate induces the production of extra distal tip cells (DTCs) (Shibata et al., 2010). In wild-type animals, there are two DTCs that function as leader cells during gonad formation. To identify additional genes that are required for the maintenance of cell fate, we screened for mutants that have extra DTCs and isolated two mutants of tlk-1 ( Fig. 1A-E, Fig. S1A). tlk-1 encodes a serine/threonine kinase that is a member of the TLK family ( Fig. 1F-H). Although humans and mice have two TLK family proteins, TLK-1 is the sole family member in C. elegans. tlk-1(tk158) has a nonsense mutation at Q44stop, and tlk-1(tk170) has a missense mutation at T846I (Fig. 1H, Fig. S1B). The translational termination near the N terminus suggests that tk158 is a null allele. A DNA fragment containing the coding region and 3.5 kb of upstream sequence fully rescued the tlk-1(tk158) mutant phenotype (Fig. 1E, Fig. S1A). tlk-1::gfp expression was observed in the nuclei of all somatic cells including cells of the somatic gonad, neurons in the posterior lateral ganglia (PLG), and the hypodermis ( Fig. 1I-L). TLK-1 is also expressed in the nuclei of embryos (Han et al., 2003). The knockdown of tlk-1 by feeding RNAi resulted in embryonic lethality (data not shown). However, we observed the postembryonic extra-DTC phenotype in tlk-1 homozygous mutants from heterozygous mutant hermaphrodites because the embryonic lethality was rescued by the maternal effect.

TLK-1 regulates cell fate in multiple cell lineages
In the wild-type somatic gonad, the two DTCs express lag-2::gfp (Kostic et al., 2003). Extra DTCs were observed in half of the tk158 mutants (Fig. 1E). The maximum number of DTCs was five cells in tk158 mutants (Fig. S1A). In addition to expressing lag-2::gfp, DTCs were positioned at the tip of the gonad arms and showed a cup-like shape in wild-type animals ( Fig. 2A). tlk-1 mutants had extra DTCs at the tips of the extra gonad arms. The extra DTCs were also cup shaped ( Fig. 2B), suggesting differentiation into DTCs rather than simple ectopic expression of lag-2::gfp. We examined whether extra DTC formation depended on the NK-2 family homeodomain DNA-binding TF CEH-22, which induces DTCs . Because ceh-22 is required for the production of mother cells of DTCs (Lam et al., 2006), we performed partial knockdown of ceh-22 by feeding RNAi. ceh-22 RNAi in tlk-1 mutants partially suppressed the extra-DTC phenotype (Fig. 2C, Fig. S2A). Therefore, inappropriate expression of ceh-22 induced extra DTCs in tlk-1 mutants.

Cell-fate transformation in tlk-1 mutants
In wild-type animals, distal granddaughters of the Z1/Z4 cells that are born at the L1 stage differentiate into DTCs until the early L2 stage (Shibata et al., 2010). In tlk-1 mutants, we compared the extra-DTC phenotype at the L2 and adult stages and found a lower penetrance at the L2 stage (Fig. 3O, Fig. S3A). Analysis of the DTC numbers in the L2 and adult stages in the same animals revealed that they increased in 12 of 29 tlk-1 mutant animals (data not shown). These results indicated that extra DTCs are produced even after the production of normal DTCs.
To elucidate the cause of extra marker-positive cells, we observed the PLG using the PDE marker osm-6::gfp and the PVD marker dop-3::rfp at the L3 and adult stages in the same animals. The position of the cells is variable in each animal, but the relative position of cells is conserved in the same animal during development (Shibata et al., 2010). In wild-type animals, cells that expressed

Nuclear H3.3 levels are up-regulated in tlk-1 mutants
In cultured cells, CAF1 depletion causes alternative deposition of H3.3 that is observed in actively transcribed loci (Ray-Gallet et al., 2011;Wirbelauer et al., 2005). To examine the level of H3.3 deposition onto chromatin in chaf-1 or tlk-1 mutants, we observed the expression of his-72, which encodes H3.3 (Boeck et al., 2011;Ooi et al., 2006). Higher his-72::gfp expression was observed in the nuclei of the hypodermis of tlk-1 and chaf-1 mutants than in the wild type ( Fig.   5A, B, E). The expression level was higher in tlk-1 mutants than in chaf-1 mutants, which was consistent with higher penetrance of the extra-DTC phenotype in tlk-1 mutants. The granular pattern of his-72::gfp expression in the nucleus suggested that HIS-72::GFP was deposited in the chromatin. Higher his-72::gfp expression was also observed in the somatic gonad of tlk-1 mutants relative to the wild-type somatic gonad (data not shown). In contrast to tlk-1 mutants, mutants of bet-1, which functions through the deposition of H2A.z , showed only weak up-regulation of his-72::gfp expression (Fig. 5E). These results indicated that TLK-1 and CAF1 reduce his-72::gfp accumulation in chromatin.

Loss of SIN-3 suppresses extra-DTC phenotype of tlk-1 mutants
We performed RNAi screening for suppressors of tlk-1 mutants using the chromatin subset of the Ahringer's feeding RNAi library (Kamath et al., 2003). We found that sin-3 RNAi suppressed the extra-DTC phenotype of tlk-1 mutants (data not shown). A sin-3 deletion mutant also suppressed the tlk-1 phenotype (Fig. 6A,   Fig. S5A, B). Most sin-3 tlk-1 double mutants had two DTCs. The frequency of animals with one or no DTCs was similar between the double mutants and the sin-3 single mutants (Fig. S5A). Of note, sin-3 RNAi also suppressed the extra-DTC phenotype of chaf-1 mutants (Fig. 6B, Fig. S5C). Thus, sin-3 was antagonistic to tlk-1 rather than having a role in DTC differentiation.
We also examined the expression of his-72::gfp in the tlk-1 sin-3 background. Interestingly, the level of his-72::gfp expression also decreased in tlk-1 sin-3 relative to its expression in tlk-1 (Fig. 5C, E). There was a positive correlation between the his-72::gfp expression level and the extra-DTC phenotype. For example, half of tlk-1 mutants and 8% of tlk-1 sin-3 mutants showed the extra-DTC phenotype. The median his-72::gfp expression level in tlk-1 mutants was almost the same as the 80 th percentile for his-72::gfp expression in tlk-1 sin-3 mutants.

AcH4-binding protein BET-1 is necessary for suppression by sin-3 disruption
The SIN3 complex contains histone deacetylase, HDAC (Hassig et al., 1997;Laherty et al., 1997). However, HDAC (had-1, 2, 3, 4, and 6) RNAi did not suppress the extra-DTC phenotype in tlk-1 mutants (data not shown). If histone hyper-acetylation in tlk-1 sin-3 double mutants is responsible for suppression of the extra-DTC phenotype, disruption of histone acetyltransferase may induce extra DTCs. We found that RNAi of mys-2, which encodes MYST family histone acetyltransferase, induced the extra-DTC phenotype in the tlk-1 sin-3 background ( Fig. 6C, Fig. S5D). mys-2 RNAi did not enhance the extra-DTC phenotype in the tlk-1 background and did not induce extra DTCs in the wild-type or sin-3 background (Fig. 6D, Fig. S5E). Therefore, the effect of mys-2 RNAi was specific to the tlk-1 sin-3 background. These results suggested that the extra-DTC phenotype in tlk-1 mutants can be compensated for by hyper-acetylation.
Because the enhancement of the extra-DTC phenotype by the hira-1 mutation was similar to that of the bet-1 or chaf-1 mutations, we examined whether sin-3 could suppress the extra-DTC phenotype in tlk-1 hira-1 mutants. The results revealed that the sin-3 mutation showed weak suppression in tlk-1 hira-1 mutants (Fig. 6A, Fig. S5A, B). These results suggested that the functions of HIRA proteins may have diverged between C. elegans and mammals.
We also examined whether hira-1 is required for down-regulation of his-72::gfp by sin-3 depletion. sin-3 RNAi suppressed the level of his-72::gfp expression in tlk-1 hira-1 background. However, compared to tlk-1 or tlk-1 bet-1 background, the suppression was weak. This result indicated that the downregulation of his-72::gfp by sin-3 depletion is partially dependent on hira-1.

TLK-1 and CHAF-1 maintain cell fates
In this paper, we showed that TLK-1 and CHAF-1 maintain the cell fates of multiple cell types by inhibiting their fate transformation into gonadal DTCs, PDE and PVD interneurons, and AVM and PVM touch receptor neurons. TLK-1 represses, directly or indirectly, ceh-22 and mec-3, which encode the NK-2 family homeodomain and LIM homeodomain DNA-binding TFs, respectively. We speculate that TLK-1 regulates transcription of the selector genes that encode DNA-binding TFs (Garcia-Bellido, 1975;Hobert, 2008). Although it is not clear whether the mec-3 locus is a direct target of TLK-1, ectopic production of mec-3expressing cells in tlk-1 mutants and the loss of mec-4 expression in tlk-1 mec-3 double mutants suggest this possibility. The reduction of ectopic DTCs by ceh-22 RNAi in tlk-1 mutants also suggests that ceh-22 is a functional target.

TLK-1 and BET-1 act in parallel pathways
We found that loss of SIN-3 strongly suppressed the extra-DTC phenotype of tlk-1 and chaf-1 mutants. The SIN3 complex contains HDAC (Hassig et al., 1997;Laherty et al., 1997), suggesting that the acetylation of histones is elevated in sin-3 mutants. As expected, RNAi knockdown of one of the histone acetyl transferases, mys-2, reversed the suppression effect of sin-3 mutants.
Furthermore, the suppressor activity of sin-3 mutants depended on an acetylated histone H4 binding protein, BET-1, required for fate maintenance of cell types whose fates are also maintained by TLK-1 and CHAF-1 .
Because the cell fate-maintenance defect was enhanced in tlk-1 and bet-1 double mutants, it is likely that TLK-1 and BET-1 act in distinct pathways to regulate cell-fate maintenance.
We previously showed that the BET-1 pathway represses transcription of target genes by recruiting H2A.z to maintain cell fates . Loss of H2A.z is the major cause of defects in cell-fate maintenance in bet-1 mutants (Fig. S6B). In contrast, repression of the extra-DTC phenotype in tlk-1 sin-3 double mutants appears to be caused by an H3-dependent mechanism, because there is a positive correlation between the level of H3.3 and the strength of the extra-DTC phenotype in tlk-1 sin-3 double mutants, tlk-1 mutants, chaf-1 mutants, and tlk-1 hira-1 double mutants. Therefore, it is likely that the TLK-1 and CHAF-pathway positively regulates H2A.z deposition.

Model for cell-fate maintenance by histone variants
Based on our findings, we propose the following model. Localization of histone H3 on selector gene loci is critical for transcriptional repression (Fig. 7A). H3 deposition by TLK-1 and CAF1 inhibits deposition of H3.3. In tlk-1 mutants, reduction of CAF1 activity causes ectopic H3.3 deposition when the SIN-3-HDAC complex deacetylates histone H4 (Fig. 7B). Stochastic expression of selector genes by ectopic H3.3 deposition causes cell-fate transformation. In tlk-1 sin-3 double mutants, loss sin-3 activates two independent downstream machineries.
One is a HIRA-1-dependent mechanism which causes down-regulation of H3.3 ( Fig. 7C). Although HIRA is known as H3.3 chaperone in mammal, C. elegans HIRA-1 may promote H3 deposition. The other mechanism is activated by hyperacetylation of H4 that is caused by malfunction of the SIN-3/HDAC complex.
Hyper-acetylation of H4 is recognized by BET-1 that promotes H2A.z incorporation. BET-1-dependent mechanism does not affect the level of H3.3.
Thus, loss of sin-3 causes down-regulation of H3.3 and deposition of H2A.z in a manner dependent on HIRA-1 and BET-1, respectively.
Although it is thought that H2A.z localization is correlated with transcriptional activation, we previously showed that H2A.z represses transcription in cell-fate maintenance . These results, together with the current research, lead us to speculate that H3.3 localization is correlated with transcriptional activation, whereas H2A.z localization is correlated with transcriptional repression in cell-fate maintenance. In agreement, H2A.zmediated chromatin compaction is repressed by H3.3 in vitro (Chen et al., 2013).
These histone variants may co-regulate transcription through chromatin compaction in cell-fate maintenance. Chromatin compaction may affect the binding of TFs to cis elements, association of transcriptional machinery, or transcriptional elongation by RNA pol II.
Although HIRA is known as a chaperone of H3.3 in mammals (Tagami et al., 2004), a hira-1 deletion up-regulated the H3.3 level in C. elegans. One explanation is that C. elegans HIRA-1 function as a chaperone of H3. Alternatively, this may be explained by the activation of another H3.3 chaperone, for example, the counterpart of Daxx (Lewis et al., 2010). However, there is no Daxx homolog Daxx in C. elegans. C. elegans may have other H3.3 chaperones. It is reported that knockdown of the H2A.z chaperone EP400 hampers H3.3 deposition in a human cell line (Pradhan et al., 2016). However, a similar mechanism is inconsistent with down-regulation of HIS-72::GFP by activation of BET-1, which promotes H2A.z deposition. Thus, it is likely that the function of C. elegans HIRA-1 may have diverged from HIRA proteins of other organisms during evolution.
Recent studies revealed that artificial induction of trans-differentiation, including generation of iPS cells, is improved by repression of CAF1 (Cheloufi et al., 2015), which is known as a downstream factor of TLK (Klimovskaia et al., 2014). The current study is the first example showing the role of CAF1 and TLK-1 in cell-fate maintenance during normal development. The roles of CAF1, and probably TLK, in the maintenance of cell fate appear to be conserved in multicellular organisms. Our research also showed that there is a strong correlation between the H3.3 level and defects in cell-fate maintenance, suggesting that the regulation of H3 variants is important for cell-fate maintenance and trans-differentiation. We also unveiled the regulation of H3 variants through BET-1, which promotes H2A.z deposition. H3.3 and H2A.z are major histone variants that are conserved in yeast, C. elegans, and mammals.
Crosstalk between these histone variants may be a fundamental mechanism in the repression of trans-differentiation. The roles of H3.3, BET family proteins, and H2A.z remain unknown in artificial trans-differentiation. Thus, the functional conservation of TLK, CAF1, and histone variants between cell-fate maintenance in C. elegans and artificial trans-differentiation in mammals is a compelling issue for future studies.

Materials and Methods
Strains and culture N2 Bristol was used as the wild-type C. elegans strain (Brenner, 1974). Animals were cultured at 20°C. The bet-1 (Shibata et al., 2010), chaf-1, rba-1 (Nakano et al., 2011), and tlk-1 mutants are sterile and were maintained as heterozygotes over the hT2[qIs48] balancer. The phenotypes of homozygotes generated from the heterozygous hermaphrodites were analyzed.
Cloning of tlk-1 Single-nucleotide polymorphism (SNP) mapping indicated that both tk158 and tk170 are positioned on LG III. A complementation test revealed that tk158 and tk170 are allelic (data not shown). Because tk158 showed a more severe phenotype (Fig. 1E, Fig. S1A), we used tk158 for further analysis. tk158 was positioned at the center cluster of LGIII between 0.92 and 1.13 by SNP mapping (Fig. 1F). Because tk158 and tk170 are sterile, heterozygous animals that are balanced by hT2 were used for genome sequencing. Within the candidate region, tlk-1 is the sole gene that has a non-synonymous mutation in both tk158 and tk170 mutants. For the rescue experiment, a PCR fragment that contained tlk-1 and 3.5 kb of upstream sequence was amplified from the fosmid WRM0631bB10 using primers 5′-CTCTCTTTGCCACTTTATCGTTTGT-3′ and 5′-AAGTTTGCGCATGTAGTAAGTTTCA-3′. The transgenic marker was myo-3::mCherry.
Microscopy and statistical analysis Expression of lag-2::gfp, dat-1::gfp, osm-6::gfp, dop-3::rfp, and mec-3::gfp was detected by epifluorescence microscopy (AxiosImagerM2 and Axioplan2; Zeiss, Jena, Germany). Expression of his-72::gfp was detected by confocal microscopy (LSM510 and Pascal; Zeiss) in the L3 animals. We used V5.ppp to quantify his-72::gfp because V5.ppp is easy to identify in wild-type animals and mutants, has a large nucleus, and is positioned near the body surface. The average intensity in the V5.ppp nucleus was measured using ImageJ (National Institutes of Health, Bethesda, MD).
Background fluorescence was measured from the adjacent region of the nucleus of V5.ppp and subtracted from the average intensity.        Whiskers indicate the 10th and 90th percentiles. Boxplots represent the medians and the 25th-75th percentile. *0.05 < p < 0.01, ***p < 0.005.