During the first hours of embryogenesis, formation of higher-order heterochromatin coincides with the loss of developmental potential. Here, we examine the relationship between these two events, and we probe the processes that contribute to the timing of their onset. Mutations that disrupt histone H3 lysine 9 (H3K9) methyltransferases reveal that the methyltransferase MET-2 helps terminate developmental plasticity, through mono- and di-methylation of H3K9 (me1/me2), and promotes heterochromatin formation, through H3K9me3. Although loss of H3K9me3 perturbs formation of higher-order heterochromatin, embryos are still able to terminate plasticity, indicating that the two processes can be uncoupled. Methylated H3K9 appears gradually in developing C. elegans embryos and depends on nuclear localization of MET-2. We find that the timing of H3K9me2 and nuclear MET-2 is sensitive to rapid cell cycles, but not to zygotic genome activation or cell counting. These data reveal distinct roles for different H3K9 methylation states in the generation of heterochromatin and loss of developmental plasticity by MET-2, and identify the cell cycle as a crucial parameter of MET-2 regulation.
Upon fertilization, embryos initiate developmental events to control cell division, pluripotency, differentiation, reorganization of the nucleus, onset of transcription and repositioning of cells into germ layers that ultimately generate organs and tissues. These processes are coordinated in time and space. In C. elegans, for example, the reorganization of cells during gastrulation coincides with loss of developmental plasticity, the formation of heterochromatin and a surge of transcription. Although the timing of these events has been defined (Mango, 2009; Mutlu et al., 2018; Robertson et al., 2004; Seydoux and Fire, 1994; Sulston et al., 1983; Yuzyuk et al., 2009), little is known about the processes that regulate the timing or synchronization of different events (Detwiler et al., 2001; Mutlu et al., 2018; Yuzyuk et al., 2009). To begin to address this issue, we have examined how early embryonic events are synchronized and regulated during early embryogenesis.
C. elegans embryos, like those of other animals, are born pluripotent and lack higher-order heterochromatin, which packages the region of the genome that is less transcriptionally active, carries repetitive DNA and becomes visible as electron-dense regions by electron microscopy (Mango, 2009; Mutlu et al., 2018). C. elegans embryos develop within 14 h, with initiation of zygotic transcription at the four-cell stage and gastrulation at the 28-cell stage (Schauer and Wood, 1990; Seydoux and Fire, 1994; Sulston et al., 1983). Prior to gastrulation, cells are developmentally plastic, and their normal pattern of development can be reprogrammed by ectopic expression of selector genes; however, this flexibility is lost during gastrulation (Djabrayan et al., 2012; Fukushige and Krause, 2005; Gilleard and McGhee, 2001; Horner et al., 1998; Mango, 2009; Priess and Thomson, 1987; Sulston et al., 1983; Wood, 1991; Zhu et al., 1998). As embryos transition towards gastrulation, they lose developmental potential and they also generate higher-order heterochromatin (Mango, 2009; Mutlu et al., 2018; Yuzyuk et al., 2009).
In cultured mammalian cells, the enzymes that promote heterochromatin formation also inhibit reprogramming into embryonic stem cells, suggesting a possible link between these processes (Gaspar-Maia et al., 2009; Kang, 2014; Soufi et al., 2012; Zaret and Mango, 2016). For example, in mammals, the H3K9 methyltransferase SETDB1 both promotes heterochromatin formation and dampens reprogramming efficiency (Becker et al., 2015). However, in mammals, multiple enzymes contribute to a given modification, which complicates studying the function of a single histone modification. We use C. elegans, which enabled us to dissect the roles of different H3K9 methylated states. C. elegans MET-2 is homologous to SETDB1 (Andersen and Horvitz, 2007; Poulin et al., 2005) and is responsible for virtually all embryonic H3K9me1/me2, but has only partial effects on H3K9me3; SET-25 is necessary for virtually all H3K9me3, but not for H3K9me1/me2 (Towbin et al., 2012). The separation of enzymatic activities provides an opportunity for determining which modifications govern which processes. Studying an intact animal also enables us to establish the temporal order of events in relation to developmental milestones.
The formation of heterochromatin during C. elegans embryogenesis depends on MET-2 (Mutlu et al., 2018). Early embryos have almost no H3K9me2 and low levels of H3K9me1 and H3K9me3. Transmission electron microscopy reveals that their nuclei lack the electron-dense regions that represent heterochromatin. As embryos develop, MET-2 becomes enriched in the nucleus, promotes an increase in all three methylated forms of histone H3 and generates heterochromatin (Mutlu et al., 2018). It is unknown whether MET-2 also influences the ability of cells to alter their fates.
Our previous studies have shown that H3K9me2 is regulated by nuclear import of MET-2 and its binding partners LIN-65 and ARLE-14, which resemble human ATF7IP and ARF-like 14 effector protein, respectively (Mutlu et al., 2018). Accumulation within nuclei occurs slowly, culminating at the onset of gastrulation, when H3K9me2 becomes readily visible within nuclei. A central issue is what processes dictate the timing of MET-2 localization.
Here, we examine two crucial issues for early embryo development. First, we test whether MET-2 or SET-25 is required to terminate plasticity, and whether developmental plasticity and heterochromatin are linked. Second, we investigate how MET-2 and H3K9 methylation are regulated in early embryos to establish the appropriate timing of events during gastrulation.
MET-2 promotes loss of plasticity
The loss of plasticity in C. elegans involves the Polycomb component mes-2/E(z) and the Notch ortholog glp-1 (Djabrayan et al., 2012; Yuzyuk et al., 2009). However, neither factor is essential for terminating plasticity, suggesting additional regulators must exist. We took advantage of the cell fate challenge assay (CFC) (Horner et al., 1998; Mango, 2009) to test the importance of met-2 for developmental plasticity (Fig. 1). Embryos were induced to alter their cell identity and acquire muscle fate by ectopic expression of hlh-1/MyoD under control of the heat-shock promoter (HS::hlh-1; Fukushige and Krause, 2005). We focused on the 80- to 100-cell stage (mid-gastrulation), when wild-type embryos have an intermediate response to the CFC assay, thereby providing a sensitive assay that can detect increases or decreases (Fig. 1A,B). In addition, H3K9me2 is readily visible at this developmental stage in wild-type embryos but absent from met-2 mutants (Mutlu et al., 2018). The foregut marker PHA-4 was used to identify cells that retained their endogenous identity and resisted exogenous HLH-1, and muscle paramyosin was used to track conversion to muscle fate (Horner et al., 1998; Yuzyuk et al., 2009). We concentrated on endogenous fate markers because exogenous reporters can sometimes lead to spurious expression or regulation (Jiao et al., 2018; Mango, 2007).
When challenged with HS::hlh-1, 35% of 100-cell stage wild-type embryos converted to muscle fate completely, consistent with previous studies (Fukushige and Krause, 2005; Mango, 2009; Yuzyuk et al., 2009). In met-2 mutants, 54% of embryos responded to HS::hlh-1 with a complete cell-fate transformation (Fig. 1C, 100-cell stage, n=4, >100 embryos each, two-tailed Student's t-test P=0.008). This result suggests that, normally, met-2 promotes loss of plasticity during early gastrulation. By the ∼200-cell stage, none of the wild-type or met-2 mutant embryos remained fully plastic (Fig. 1C). Moreover, both wild-type and met-2 mutant embryos had a similar number of cells that resisted changing fate, indicating that met-2 mutants eventually terminate plasticity (Fig. S1A). The data reveal that met-2 restricts developmental plasticity during gastrulation but is not absolutely required. The partial requirement for met-2 may reflect that other regulators also control developmental plasticity in the embryo (Djabrayan et al., 2012; Joshi et al., 2010; Yuzyuk et al., 2009).
H3K9me1/me2 exist in the embryo and, in addition, are used to produce H3K9me3 by the methyltransferase SET-25 (Towbin et al., 2012). We wondered whether either the plasticity or the heterochromatin phenotype of met-2 could be explained by its role in generating H3K9me3. First, we examined developmental plasticity with the CFC assay. Mutations in set-25 lead to the opposite result of met-2 mutations: only 20% of set-25 embryos were developmentally plastic, about half of wild type under the conditions used (Fig. 1C, n=3 experiments, >60 embryos each, two-tailed Student's t-test P=0.011). This result demonstrated that H3K9me3 was not required to terminate plasticity during gastrulation. Rather, developmental plasticity was anti-correlated with H3K9me2, in which we observed lower plasticity/higher H3K9me1/me2 in set-25 (Fig. S2), or higher plasticity/lower H3K9me1/me2 in met-2 embryos compared with wild type.
To test the importance of H3K9me2, we examined met-2; set-25 double mutants at the 100-cell stage, which lack mono, di and tri H3K9me (Garrigues et al., 2015; Towbin et al., 2012). The double mutant had prolonged plasticity, ∼50% higher than the wild type, which is similar to the met-2 single mutant (Fig. 1C, n=2, 30 embryos each, two-tailed Student's t-test P=0.045). By the 200-cell stage, none of the met-2; set-25 mutant embryos remained fully plastic, similar to single met-2 or set-25 mutants (data not shown). These data show that the reduced plasticity of set-25 mutants requires wild-type met-2 activity. A simple hypothesis to explain these results is that H3K9me1 and/or H3K9me2 is required to terminate developmental plasticity but H3K9me3 is not.
As a control, we examined hlh-1 mRNA expression before and after heat shock, and observed no difference in induction between wild-type and mutant embryos (Fig. 1D). We also tested whether wild-type and mutants progressed through embryonic development at the same pace under our assay conditions (Fig. S1B,C). We found that the time from the two-cell stage to hatching was not significantly different between wild-type and mutant embryos (Fig. S1B). To compare embryonic stages before the hs::hlh-1 heat shock, we staged embryos by identifying intestinal nuclei under DIC imaging and found that they were all at the 8E (endodermal) stage, which corresponds to ∼100 cells (Yuzyuk et al., 2009) (Fig. S1C, n>10 embryos for each genotype, ∼3 h after the two-cell stage) or at the 200-cell stage (4.5 h). Consistent with our results, the duration of mitosis remains unchanged between wild-type and met-2; set-25 mutant embryos (Zeller et al., 2016).
To gain insight into genes regulated by H3K9me1 and H2K9me2, we examined modENCODE data for genes bearing either of these marks. Among genes with peaks of H3K9me1, there was enrichment for GO terms ‘intracellular signaling cascade’ and regulation of ‘developmental process’ and ‘growth’ (Fig. 1E). For H3K9me2, there was enrichment for ‘cell-fate determination’ and ‘embryonic pattern specification’, which included genes expressed in the pre-gastrula embryo (e.g. par-1, par-4, mbk-2 and others). These GO terms were not enriched for H3K9me3, possibly explaining the distinct phenotypes of met-2 versus set-25.
Our results do not distinguish between H3K9me1 and H3K9me2, but we note that H3K9me1 does not correlate well with plasticity temporally (Mutlu et al., 2018). In addition, H3K9me1 is distributed throughout the genome and is not restricted to silent regions (Gerstein et al., 2010; Liu et al., 2011). Nevertheless, as MET-2 controls both H3K9me1 and H3K9me2, it is possible that both modifications are important. For example, H3K9me1 could act as a substrate to produce H3K9me2 or H3K9me3 (Loyola et al., 2009).
The H3K9me3 methyltransferase SET-25 is required for heterochromatin formation
Next, we examined the role of MET-2 and SET-25 for higher-order heterochromatin, as visualized by transmission electron microscopy (TEM). Wild-type nuclei from young embryos were predominantly electron translucent whereas nuclei from older embryos contained many electron-dense regions (EDRs) (Fig. 2A), as observed previously (Mutlu et al., 2018). EDRs represent heterochromatin regions based on morphology and on the enrichment for silencing marks such as H3K9me3 (Rübe et al., 2011). In met-2 mutants, EDRs are paler than normal and delayed in their appearance (Mutlu et al., 2018; Fig. 2B). In set-25 mutants, we did not detect any EDRs through the 200-cell stage, a phenotype that was stronger than met-2 mutants (Fig. 2A). As a control for TEM fixation and imaging, we examined the cytosol of wild-type and set-25 mutants, using a threshold value to define electron-dense regions. Both genotypes had darkly stained cytosolic regions (Fig. S3). These data show that disruption of EDR formation tracks with H3K9me3: intermediate EDR formation in met-2 mutants correlates with intermediate H3K9me3 levels and absent EDRs in set-25 mutants with absent H3K9me3. These results demonstrate that neither H3K9me3 nor the formation of large-scale, higher-order heterochromatin is required to terminate developmental plasticity. The data suggest that synchronization between termination of plasticity and generation of heterochromatin relies on distinct readouts from MET-2: formation of heterochromatin from H3K9me3 and termination of plasticity by H3K9me1/me2 (Fig. 2C).
H3K9 de-methylation is not a timer for establishing H3K9me2 domains at gastrulation
The above data suggest that MET-2 helps regulate two crucial events in the early embryo: loss of developmental plasticity by H3K9me1/me2 and generation of higher-order heterochromatin by H3K9me3. Both of these processes occur during gastrulation as H3K9me accumulates, raising the issue of what mechanisms regulate H3K9me temporally (Fig. 3A). As MET-2 acts upstream of SET-25 (Towbin et al., 2012) and H3K9me2 increased most dramatically from fertilization to gastrulation (Mutlu et al., 2018), we focused on the timing of H3K9me2 onset.
Initially, we wondered whether an H3K9me2 demethylase might act in early embryos and account for the low levels of H3K9me2 observed at this time. To test this idea, we examined a mutant for the H3K9me2 demethylase jmjd-1.2 (Kleine-Kohlbrecher et al., 2010). In jmjd-1.2(tm3713) mutants that do not produce JMJD-1.2 (Myers et al., 2018), H3K9me2 was extremely low in early embryos, as in wild type, and was established normally at gastrulation (Fig. 3B-D). We also checked jmjd-2 mutants, either alone or in combination with jmjd-1.2 RNAi. jmjd-1.2 and jmjd-2 are both highly expressed in the early embryo (Levin et al., 2012). JMJD-2 has been shown to demethylate H3K9me3, and the tm2966 mutation, which deletes the catalytic domain, is predicted to be enzymatically inactive (Greer et al., 2014; Whetstine et al., 2006). In both cases, H3K9me2 levels were very low in early embryos and increased with dynamics similar to wild type, or even a little slower (Fig. 3E). These data suggested that histone de-methylation is not part the H3K9me2 timer and led us to focus on MET-2.
RNA polymerase II transcription is not rate limiting for H3K9 di-methylation
The onset of gastrulation and surge in H3K9me2 is accompanied by a big wave in zygotic transcription (Baugh, 2003; Edgar et al., 1994; Hsu et al., 2015; Levin et al., 2012; Storfer-Glazer and Wood, 1994; Yuzyuk et al., 2009). We hypothesized that met-2 or its co-factors could be activated in the embryo as zygotic transcription begins (Fig. 4A, model 1). In addition, transcription elongation could be rate limiting for recruiting MET-2 to specific loci through interactions with the RNAi machinery (Guang et al., 2010, Fig. 4A, model 2). We tested these models in three ways. First, we blocked zygotic transcription and examined whether the onset of H3K9me2 was disturbed. Second, we tested whether a zygotic copy of met-2 could rescue H3K9me2 in the absence of maternal MET-2. For these experiments, we were able to compare H3K9me2 levels across different genotypes by including wild-type embryos marked with a fluorescent tag as an on-slide control. Control embryos and test embryos were processed together on the same slide and imaged with identical settings (Fig. 4B). Third, we examine lin-41 mutants, which activate transcription prematurely (Tocchini et al., 2014), to determine whether transcription was sufficient for H3K9me2.
To block transcription, we inactivated the canonical TFIID subunit, taf-6.2, with a temperature-sensitive mutation (ax514) (Bowman et al., 2011), and a core subunit of RNA polymerase II, ama-1, by RNAi. To assess the strength of the block on transcription, we stained embryos using the H5 polymerase II (Pol II) antibody against the phosphorylated Ser2 on the CTD domain, which is a hallmark of transcriptional elongation (Bregman et al., 1995; Seydoux and Dunn, 1997). Under our conditions, H5 signal was undetectable after blocking zygotic transcription, whereas non-specific staining of P-granules by the H5 antibody served as a positive control (Fig. 4C,D). Despite the lack of detectable H5, H3K9me2 levels were identical to wild-type embryos (Fig. 4C,E), suggesting that neither zygotic genes nor elongating polymerase II was rate-limiting for H3K9me2 onset.
The transcriptional block suggested that MET-2 and its regulators must be contributed maternally and independently of zygotic transcription. To test this idea further, we examined RNA expression of met-2 and its binding partners lin-65 and arle-14 (Mutlu et al., 2018), which are required for H3K9me2. met-2, lin-65 and arle-14 transcripts were abundant in early embryos, before the initiation of zygotic transcription (Levin et al., 2012; Seydoux and Fire, 1994; Tintori et al., 2016), revealing that these RNAs were deposited by the mother. To test whether zygotic MET-2 could rescue H3K9me2 in embryos, we mated met-2 mutant mothers with wild-type males and analyzed the progeny, which carried a single zygotic copy of the wild-type met-2 locus. Zygotic met-2 could not restore H3K9me2 to mutant embryos (Fig. 4F). Similarly, neither zygotic lin-65 nor zygotic arle-14 could rescue H3K9me2 from maternally mutant mothers (Fig. 4G). Moreover, when we mated wild-type mothers with met-2::gfp males, the resulting met-2::gfp progeny never expressed GFP (Fig. 4H,I). These results demonstrate that maternally deposited MET-2 and its binding partners are responsible for establishing H3K9me2 in embryos. We note that maternal MET-2 is bound by the DNA repair protein SMRC-1 in the germ line (Yang et al., 2019), which itself is involved in the onset of heterochromatin formation in zebrafish (Laue et al., 2019). However, inactivation of smrc-1 failed to alter the accumulation of nuclear MET-2 or the onset of H3K9me2 in embryos (Fig. S4).
Third, we examined lin-41 mutants for precocious H3K9me2. In animals lacking lin-41, transcription is often activated prematurely in germ-line oocytes as they initiate somatic development prematurely (Tocchini et al., 2014). We found that H3K9me2 did not correlate with precocious transcription, monitored with the H5 antibody (Fig. 4J): some oocytes had H5 immunoreactivity but no H3K9me2, suggesting elongating Pol II was not sufficient to initiate H3K9me2. Other oocytes had precocious H3K9me2 but lacked H5, indicating that elongating Pol II was not necessary. Together, these data indicate that the onset of Pol II transcription cannot account for the timing of H3K9me2 deposition. We note that this conclusion shows that worms differ from zebrafish, where transcription is required to initiate heterochromatin formation (Laue et al., 2019).
Length of interphase, but not cell counting mechanisms, dictates timing of H3K9me2
The transcription block ruled out some expected mechanisms for H3K9me2 onset. To gain a broader perspective on regulation of H3K9me onset, we considered alternative models. Embryonic timing has been studied extensively for the onset of zygotic genome activation (ZGA), and processes that dictate ZGA could, in theory, apply to H3K9me. One model postulates that cell counting mechanisms such as the amount of DNA in the embryo could be a timer (Dekens et al., 2003; Newport and Kirschner, 1982a). A second model focuses on the changes in the nuclear-to-cytoplasmic ratio as embryonic cells divide, which could be crucial for diluting or concentrating a maternal repressor/activator (Newport and Kirschner, 1982b; Pritchard and Schubiger, 1996). A third model predicts that fertilization initiates a process that builds gradually as a direct readout of time passed (Ferree et al., 2016; Howe and Newport, 1996; Kimelman et al., 1987; Treen et al., 2018; Yuan et al., 2016). We tested whether any of these models could explain timing of H3K9me2 onset.
To determine whether cell counting is important, we uncoupled the number of cells produced over a given unit of time. We inactivated div-1, a subunit of DNA polymerase-alpha primase complex, by using a temperature-sensitive mutation (or148) to extend the duration of S phase and slow down cell divisions (Encalada et al., 2000). One hour after the two-cell stage, wild-type embryos had 25-30 cells, whereas div-1 mutant embryos had 5-15 cells (Fig. 5A). div-1 embryos exhibited precocious H3K9me2 based on cell number, but wild-type levels based on time post-fertilization (Fig. 5B,C). The amount of H3K9me2 per nucleus was similar in 25-30 cell wild-type embryos and 5-15 cell div-1 embryos. These delayed div-1 cells had a similar volume to wild-type embryos with the equivalent number of cells (Fig. 5D) and comparable amounts of DNA to wild-type embryos that contain the same number of cells (Fig. 5E). This result rules out some counting models, specifically the nuclear-to-cytoplasmic ratio, the number of cells or the amount of DNA. In short, it is possible to acquire high levels of H3K9me2 without reaching a specific cell number or nuclear-to-cytoplasmic ratio, or undergoing a specific number of cell divisions. Instead, this result suggests that the timing of H3K9me2 is dictated by absolute time post-fertilization (or ∼1 h after the two-cell stage). We note that cells spend most of this time in interphase as the early embryonic cell divisions are rapid, suggesting time in interphase may be the key variable.
One concern was that the replicative stress or DNA damage in early div-1 embryos caused higher levels of H3K9 methylation to engage DNA repair pathways (Ayrapetov et al., 2014). We hypothesized that if DNA damage in div-1 mutants caused the increase in H3K9me2, one might expect restoring the faster cell cycle in div-1 mutants to lead to even more DNA damage and H3K9me2 (Fig. 5F). Alternatively, if the amount of time from fertilization was the crucial parameter, then restoring the faster cell cycle to div-1 mutants would rescue normal timing of H3K9me2 accumulation. To restore faster cell cycle progression to div-1 mutant embryos, we inactivated the ATR-related gene atl-1, which leads to faster cell cycles but potentially more DNA damage (Brauchle et al., 2003). atl-1(RNAi); div-1 double mutants partially suppressed the increase in H3K9me2 (Fig. 5G-I), suggesting precocious H3K9me2 in div-1 mutants was not due to DNA damage. Instead, timing of H3K9me2 depends on the amount of time from fertilization.
Our previous work revealed that MET-2 is initially partitioned between the nucleus and cytoplasm, but becomes concentrated within nuclei as embryos mature (Fig. 6A; Mutlu et al., 2018). Nuclear MET-2 is rate limiting for H3K9me2 levels and gradually accumulates in nuclei with its binding partners LIN-65 and ARLE-14 (Mutlu et al., 2018). We wondered whether precocious H3K9me2 in div-1 mutants might reflect precocious accumulation of the MET-2 complex in nuclei. Indeed, MET-2 and its binding partners each accumulated in nuclei earlier compared with wild-type embryos (Fig. 6B-G).
We used temperature as an independent means to modulate embryonic timing. Wild-type embryos were raised at 15°C and 24°C or 25°C. The cell cycle at 15°C is approximately twice as long as that at 25°C, suggesting that MET-2 might accumulate at earlier stages (cell numbers) for embryos grown at cooler temperatures compared with warmer. We observed that, at 15°C, MET-2::GFP was already detectable in nuclei by the two- to four-cell stage (Fig. S5A). At 25°C, there was not much enrichment for nuclear MET-2::GFP at the two- to four-cell stage, but it started to accumulate in the nucleus at the eight-cell stage. We also analyzed embryos that contained a FLAG tag at the endogenous met-2 locus and compared 15°C and 24°C (Fig. S5B,C). This configuration showed a similar but less pronounced trend for earlier accumulation of nuclear MET-2 at lower temperatures. At 15°C, there was more 3xFLAG::MET-2 in nuclei compared with 24°C in experiment 1, whereas the effect was subtler in experiment 2 (Fig. S5B,C). These data reveal that MET-2 and its partners are sensitive to the speed of the cell cycle and time from fertilization. We note that many cellular activities are sensitive to temperature, such as enzymes, and this likely explains the variability of the results. Recent results indicate interphase timing is important for heterochromatin onset in Drosophila, suggesting the early phenomenon is conserved (Seller et al., 2019). In summary, the amount of time spent in interphase dictates the timing of MET-2 localization within nuclei during early embryogenesis.
This study has made two contributions towards understanding the timing and function of H3K9me during embryogenesis. First, the methyltransferase MET-2 promotes loss of developmental plasticity through H3K9me1/me2, and formation of higher-order heterochromatin through H3K9me3 and SET-25. Surprisingly, developmental plasticity is still lost in the absence of H3K9me3 and higher-order heterochromatin. Second, MET-2 activity and H3K9me2 are sensitive to the rapid cell cycles in early embryos and the gradual accumulation of MET-2 and H3K9me2 can be explained by the cumulative amount of time early embryonic cells spend in interphase, after fertilization. Neither zygotic transcription nor the nuclear to cytoplasmic ratio was important for the onset of H3K9me2.
Distinct roles for di versus tri-methylated H3K9 in cell fate potential and formation of higher-order heterochromatin
The presence of partially redundant H3K9 methyltransferases in mammals (including SETDB1, SUV39h1/2, G9a and PRDM2/3/16) and lethal phenotypes associated with loss of some of the enzymes have made it hard to dissect the roles of di- versus tri-methylation of H3K9. By sequence, MET-2 is most similar to vertebrate SETDB1 (Andersen and Horvitz, 2007; Mutlu et al., 2018; Poulin et al., 2005). Both enzymes catalyze mono- and di-methylation on H3K9 (Basavapathruni et al., 2016; Loyola et al., 2009; Towbin et al., 2012; Wang et al., 2003). One speculative but intriguing idea is that the catalytic domain may restrict MET-2 activity to mono and di, and perhaps to SETDB1 also. SET domain methyltransferases contain a ‘switch position’ in their catalytic site that determines the degree of methylation, with bulkier residues able to accommodate mono- and di- but not tri-methylation (Jih et al., 2017). MET-2 and SETDB1 both have a bulky tryptophan residue in the switch position, suggesting that these enzymes may favor mono- and di-methylation in the absence of regulatory partners. In mammals, SETDB1 exists in a complex with additional methyltransferases, which generate H3K9me3; it is unclear whether SETDB1 can generate H3K9me3 itself (Basavapathruni et al., 2016; Fritsch et al., 2010). In worms, the differential effects of the two H3K9 methyltransferases met-2 and set-25 on mono-, di- and tri-methylated H3K9 provided a useful tool for distinguishing the roles of these histone marks (Mutlu et al., 2018; Zheng et al., 2013).
We found that developmental potential was inversely correlated with H3K9me2 (Figs 1 and 2C): met-2, wild-type and set-25 embryos had low, average and high levels of H3K9me2, respectively. met-2 mutants were able to alter their developmental fate robustly, wild-type embryos less so and set-25 mutants least of all. On the other hand, H3K9me3 did not correlate with plasticity, as met-2, wild-type and set-25 embryos had low, high and no H3K9me3, respectively. As predicted by the H3K9me2 hypothesis, met-2 and set-25 double mutants lacked H3K9me2 and extended plasticity like met-2 single mutants (and unlike set-25). We note that the effect of met-2 on plasticity may be direct, e.g. by repressing specific genes required for plasticity, or it may be indirect.
Ultimately, met-2 mutants lost developmental plasticity, revealing that while H3K9me2 is important for the timely loss of plasticity, it is not essential. Additional regulators, such as Polycomb and Notch contribute to reduced embryonic potential, which may explain the partial effects of each of these mutants alone (Djabrayan et al., 2012; Joshi et al., 2010; Yuzyuk et al., 2009).
EDRs were reduced in met-2 mutants and absent in set-25 embryos. Similarly, H3K9me3 was reduced in met-2 and absent in set-25 mutants, making H3K9me3 an excellent candidate for mediating higher-order heterochromatin. An intriguing idea, based on these data, is that emergence of EDRs depends on H3K9me3 quantitatively rather than responding to an all-or-nothing threshold. This idea fits well with observations in Drosophila and mammals, where high concentrations of HP1 form liquid droplets that coalesce into heterochromatin (Larson et al., 2017; Strom et al., 2017). HP1 binds the H3K9me3 (Bannister et al., 2001; Garrigues et al., 2015; Jacobs et al., 2001; Lachner et al., 2001), suggesting a route whereby variable levels of H3K9me3 could produce variable EDRs.
Higher-order heterochromatin and loss of developmental plasticity
In theory, an absence of heterochromatin could provide a permissive environment for transcribing diverse genes, an important feature of pluripotent embryos. However, neither loss of H3K9me3 and EDRs in set-25 mutants nor the absence of all H3K9me in met-2; set-25 double mutants enabled HLH-1 to activate muscle genes ubiquitously. Therefore, additional mechanisms must exist to curtail the ability of cells to respond to a heterologous regulator. H3K27me3 is present in met-2; set-25 mutants (Towbin et al., 2012) and could compensate for the loss of H3K9 methylation (Yuzyuk et al., 2009). Alternatively, the deposition of nucleosomes over transcription factor binding sites or within transcribed regions could interfere with plasticity (Gaspar-Maia et al., 2011). In vertebrates, H3K9me-rich domains can interfere with transcription factor binding to block reprogramming of induced pluripotent stem cells, likely H3K9me3 domains (Soufi et al., 2012). This result suggests that ‘reverse development’, as seen in reprogramming, may be qualitatively different from ‘forward development’, when plasticity is lost during embryogenesis. It is possible that H3K9me3 initiates processes that eventually block reprogramming in differentiated cells, but these processes may not yet be complete in gastrula-stage embryos. Alternatively, the vertebrate studies did not distinguish between different degrees of methylation of H3K9, and it is possible that H3K9me2 plays a role in reprogramming.
Classical models of time-keeping and H3K9 methylation
Classical models of embryonic time-keeping include: (1) a clock that depends on the time from fertilization (Howe and Newport, 1996; Treen et al., 2018); (2) the increasing nuclear-to-cytoplasmic ratio (Newport and Kirschner, 1982b; Pritchard and Schubiger, 1996); and (3) cell counting via increasing DNA content in the embryo, which titrates a maternal repressor (Dekens et al., 2003; Newport and Kirschner, 1982a). These models have mostly been studied in the context of zygotic genome activation and not histone modifications. One study analyzed local changes in chromatin organization during fly embryogenesis and found that promoter accessibility is controlled by the nuclear-to-cytoplasmic ratio (Blythe and Wieschaus, 2016). However, large-scale changes in chromatin organization during development, such as heterochromatin formation, had not been studied through the lens of these models.
In this study, we tested these models with regard to H3K9 methylation, which emerges during gastrulation (Mutlu et al., 2018). Our analysis of div-1 mutants ruled out cell counting as a potential timer and suggested that H3K9me2 is sensitive to cell cycle timing. MET-2 moves gradually into nuclei in the pre-gastrula embryo, but is released into the cytosol during mitosis (Mutlu et al., 2018). Early embryonic cells divide rapidly, with a 40-min cell cycle that could interfere with the accumulation of MET-2 in nuclei. div-1 mutants specifically extend S-phase compared with wild-type embryos (Encalada et al., 2000), suggesting that the cumulative time that early embryonic cells spend in S-phase may be crucial for MET-2 accumulation in nuclei and deposition of H3K9me by MET-2.
Spatial regulation of proteins provides a rapid means with which to restrict their activity and is a common theme in temporal regulation of developmental processes. For example, SETDB1, Prmt1 and SIRT1 each transition from the cytosol to the nucleus, or vice versa, to alter their activity upon differentiation (Ancelin et al., 2006; Hisahara et al., 2008; Tachibana et al., 2015). Similarly, the onset of zygotic transcription in C. elegans is dictated by the OMA proteins, which sequester a crucial co-factor for TFIID in the cytoplasm (Guven-Ozkan et al., 2008). Our study provides a new example, where the embryonic clock for H3K9me under normal conditions depends on the spatial regulation of MET-2 and its binding partners.
MATERIALS AND METHODS
Strains were maintained at 20°C according to (Brenner, 1974) apart from EU548 div-1(or148ts) III and KW1975 taf-6.2(ax514); unc-17(e113) IV:
N2 (wild-type Bristol)
EU548 div-1(or148ts) III (Encalada et al., 2000)
KW1975 taf-6.2(ax514); unc-17(e113) IV
SM2440 jmjd-1.2 (tm3713) IV (Kleine-Kohlbrecher et al., 2010)
EL597 omIs 1 [Cb-unc-119 (+) met-2::gfp II]
SM2575 lin-65::3xflag I
SM2333 pxSi01 (zen-4::gfp, unc-119+) II; unc-119(ed3) III
KM167 HS::hlh-1, (Fukushige and Krause, 2005)
SM1623 HS::hlh-1; met-2 (ok2307) III
JAC500 his-72(csb43[his-72::mCherry]) III (Norris et al., 2015)
DG3784 lin-41 (tn1487) (Spike et al., 2014)
SM2635 N2 (sibling of jmjd-2 tm2966)
SM2636 FX02966 (Jmjd-2 (tm2966)
WM182 csr-1(tm892) IV/nT1 [unc-?(n754) let-?] (IV;V)
EL663 3xMYC::smrc-1; 3xFLAG::met-2 III
Antibody staining of embryos
Antibody staining was performed as described previously (Burger et al., 2013; Mutlu et al., 2018). Antibodies against the following proteins were used for immunostaining embryos fixed for 5 min in 2% paraformaldehyde (PFA) and incubated for 3 min in 100% methanol (for all except PHA-4, which was fixed for 10 min in 2% PFA and incubated for 3 min in 100% methanol): H3K9me2 (1:200; Abcam, ab1220; MABI0307, Kimura 6D11); Histone H3 (1:500; Abcam, ab1791); pan-histone (1:500; Chemicon/Millipore, MAB052); FLAG M2 (1:100; Sigma-Aldrich, F1804); MET-2 (1:500; raised against the first 17 amino acids of MET-2 and affinity purified, a gift from Eleanor Maine, Syracuse University, NY, USA); ARLE-14 (1:500; generated by our lab as described by Mutlu et al., 2018); H5 polymerase II (1:100; Covance, MMS129-R); paramyosin (1:50; Developmental Studies Hybridoma Bank, 5-23); PHA-4 N-terminus (1:1000) (Kaltenbach et al., 2005); and MYC (1:200; Cell Signaling Technology, 71D10). H5 Pol II staining followed a completely different protocol described elsewhere (Kaltenbach et al., 2000).
Antibody staining and imaging of germline cells
Staining of dissected gonads was performed as previously described (Burger et al., 2013) with the exception of using 3.7% formaldehyde instead of paraformaldehyde and Tween-20 instead of Triton X-100. Imaging was carried out on a Zeiss LSM880 confocal microscope with Airyscan, and images were processed in an identical manner using Fiji and Adobe Illustrator CC 2018.
Transmission electron microscopy
Transmission electron microscopy was carried out as described previously (Mutlu et al., 2018). In brief, adult animals were fixed with high-pressure freezing and substitution with 2% osmium tetroxide and 0.1% uranyl acetate in acetone. Samples were rinsed in 4% distilled H2O in acetone, followed by propylene oxide and finally embedding in plastic resin. We sectioned adult worms with Pioloform-coated slot grids on a diamond knife, counterstained using uranyl acetate followed by lead citrate, and examined using an electron microscope. Individual nuclei were chosen from embryos encased within adult mothers, and many cells from many embryos were examined. To control for fixation, we compared the cytosol of wild-type versus mutant animals.
Quantitation of histone modifications and nuclear proteins
Analysis was carried out as described by Mutlu et al. (2018). Briefly, embryos were imaged with a Zeiss LSM700 or a LSM880 confocal microscope and analyzed by Volocity Software. Signal intensities of marks were calculated for each nucleus and average values for nuclei at designated embryonic stages plotted.
KW1975 taf-6.2(ax514) was maintained at 15°C. For experiments, KW1975 L4s were fed bacteria containing an ama-1 RNAi vector. In parallel, SM2233 L4s were fed with bacteria containing empty vector. Both strains were grown at 15°C for 50-60 h. Adult worms were dissected at 26°C and one- to four-cell embryos were transferred onto the same poly-L-lysine slide. Embryos were aged for 1 h at 26°C in a humidity chamber and stained for H3K9me2. For H5 polymerase II staining, wild-type JAC500 worms fed with empty vector were used as an on-slide control instead of SM2233.
Generating RNAi plates
RNAi clones from the Ahringer library were used unless stated otherwise. First, identity of clones was confirmed by sequencing. To pour plates, bacteria were grown in 5 ml LB with 5 μl carbenicillin (100 mg/ml) for 6-8 h at 37°C and pelleted at 2500 g for 10 min. The bacterial pellet was resuspended in 400 μl 0.5 M IPTG, 30 μl 100 mg/ml carb and 70 μl nuclease-free water. NGS plates (5 ml) were seeded with 100 μl of resuspended bacterial solution and kept at room temperature for 2 days before use.
The smcr-1 clone was a generous gift from Eleanor Maine. For each experiment, three adult EL663 worms were placed on RNAi plates at 20°C and the embryos of their offspring used in experiments.
div-1(or148ts) was maintained at 15°C and shifted to 26°C for experiments. Two-cell stage wild-type zen-4::gfp (SM2233) and div-1 (EU548) embryos were picked and aged for 1 h on the same poly-L-lysine slide in a humidity chamber at 26°C. H3K9me2 staining after embryonic temperature shifts compared 25- to 30-cell wild-type embryos with 10- to 15-cell div-1 embryos.
Temperature shifts to 26°C that started with L4 animals instead of embryos produced identical results in terms of H3K9me2 levels at given embryonic stages. In L4-shift experiments, mixed stage SM2233 and EU548 embryos were dissected from gravid adults and stained for H3K9me2 on the same slide. Cells that contained the same number of cells were compared using L4 shifts.
For atl-1 rescue experiments, N2 L4s were fed with atl-1 RNAi bacteria. div-1(or148ts) L4s were fed with either empty vector or atl-1 RNAi bacteria overnight at 26°C. SM2233 L4s were fed with bacteria containing an empty vector at 26°C and included as an on-slide control on all experiments. Embryos were dissected from gravid adults and stained for H3K9me2.
A cell fate challenge assay was conducted similarly to that described by Kiefer et al. (2007). In brief, two-cell embryos were collected from wild-type, set-25 or met-2 mothers carrying an integrated HS::hlh-1 array (Fukushige and Krause, 2005). Embryos were incubated at 20°C for ∼3 h until they reached the 100-cell stage, determined by DIC imaging and cell counts (for wild type and mutants). Heat shock was administered at 33°C for 30 min on poly-L-lysine-coated slides in a humidity chamber and embryos were incubated at 15-20°C for 20 h. Terminally differentiated embryos were stained for paramyosin (muscle; Fukushige and Krause, 2005) and PHA-4 (foregut; Horner et al., 1998). Embryos were imaged using the Zeiss LSM 700 confocal microscope. RNA expression analysis for hlh-1 was carried out as described previously (Mutlu et al., 2018). For 200-cell stage embryos, two-cell embryos were aged for ∼4.5 h at 20°C.
Time from two-cell embryos to hatching
Adults of the indicated genotype were dissected in M9/double-distilled H2O at a 1:1 ratio and two-cell stage embryos were transferred to an unseeded NGM plate. Embryos were processed in batches of two to six and the time noted when they were placed on the NGM plate. After 10-15 min, progression into the four-cell stage was confirmed using a dissecting scope. After 13 h at 20°C the plates were checked every 5 min and the time noted down when embryos hatched.
H3K9me1 (GSE49744), H3K9me2 (GSE49736) and H3K9me3 (GSE49732) methylated regions were defined by a MACS2 broad peak call (Zhang et al., 2008) and the centers of peaks were assigned to genes to curate a list. DAVID (david.ncifcrf.gov/) was used for GO term analysis. modENCODE ‘early embryos’ were analyzed, which we consider late stage (the 300- to 500-cell stage; Fig. S1D).
We thank B. Yang, M. Sullenberger and E. Maine for smrc-1 reagents, the Harvard Center for Biological Imaging, and K. Nguyen and B. Raja for help on TEM procedures. Some strains were provided by the Caenorhabditis Genetics Center funded by NIH P40OD010440.
Conceptualization: B.M., S.E.M.; Methodology: B.M., S.G., S.K.-R., S.E.M.; Validation: B.M.; Formal analysis: B.M., H.-M.C., S.G., D.H.H., S.K.-R.; Investigation: B.M.; Writing - original draft: B.M., S.E.M.; Writing - review & editing: S.E.M.; Visualization: B.M., H.-M.C.; Supervision: S.E.M.; Project administration: D.H.H., S.E.M.; Funding acquisition: D.H.H., S.E.M.
S.E.M. acknowledges support from the National Institutes of Health (R37GM056264), the John D. and Catherine T. MacArthur Foundation, Harvard University and The Biozentrum of the University of Basel. B.M. is supported by an American Association of University Women International Fellowship. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.