CFP1 coordinates histone H3 lysine-4 trimethylation and meiotic cell cycle progression in mouse oocytes

Trimethylation of histone H3 on lysine-4 (H3K4me3) is associated with gene-regulatory elements, but its transcription-independent function in cell division is unclear. CxxC-finger protein-1 (CFP1) is a major mediator of H3K4 trimethylation in mouse oocytes. Here we report that oocyte-specific knockout of Cxxc1, inhibition of CFP1 function, or abrogation of H3K4 methylation in oocytes each causes a delay of meiotic resumption as well as metaphase I arrest owing to defective spindle assembly and chromosome misalignment. These phenomena are partially attributed to insufficient phosphorylation of histone H3 at threonine-3. CDK1 triggers cell division–coupled degradation and inhibitory phosphorylation of CFP1. Preventing CFP1 degradation and phosphorylation causes CFP1 accumulation on chromosomes and impairs meiotic maturation and preimplantation embryo development. Therefore, CFP1-mediated H3K4 trimethylation provides 3a permission signal for the G2–M transition. Dual inhibition of CFP1 removes the SETD1–CFP1 complex from chromatin and ensures appropriate chromosome configuration changes during meiosis and mitosis.

H istone posttranslational modifications (PTMs)-and specific combinations they create-mediate a wide range of nuclear events. Methylation of histone H3 on lysine-4 (H3K4me) is evolutionarily conserved from yeast to humans and is widely associated with gene-regulatory elements 1 . Particularly, the trimethylated species (H3K4me3), which is found primarily in active promoters, is implicated in transcriptional regulation 2 . In contrast to lysine methylation, phosphorylation of serine and threonine residues in histone H3 fluctuates substantially throughout the cell cycle, rising sharply when cells enter mitosis 3 . Histone H3 phosphorylation at threonine-3 (hereafter referred to as H3T3ph) is carried out by the kinase haspin early in prophase 4 . The close proximity of residues T3 and K4 in histone H3 potentially enables functional and regulatory interplay between the instances of phosphorylation and methylation 5,6 . Nonetheless, the possible direct function of H3K4me3 in cell division has been elusive because deficiency in this histone modification in cultured cells affects transcription of a broad spectrum of genes and causes cell cycle arrest in the G1 or S phase 7,8 . In this respect, a fully grown mammalian oocyte is an ideal model for studies on transcription-independent functioning of histone modifications because transcription is neither active nor required for the two sequential meiotic divisions 9,10 .
An oocyte-specific knockout of one of the genes coding for histone H3K4 methyltransferases or demethylases leads to impaired meiotic cell cycle progression, including delayed germinal vesicle breakdown (GVBD), distorted spindle assembly, error-prone chromosome separation, and compromised polarbody emission rates [11][12][13] . It has been proposed that overexpression or underexpression of certain key cell cycle regulators is responsible for these meiosis-related phenotypes 14,15 . Nevertheless, the evidence has been indirect and weak. The transcriptomes of fully grown GV oocytes are only moderately affected by epigenetic factors because their genome is transcriptionally quiescent 16 . In addition, epigenetic regulation of transcription is genome-wide. It is unlikely that genes of certain cell cycle regulators are specifically targeted in growing oocytes. In contrast, we hypothesized that multiple methylation of histone H3 at lysine-4 directly drives oocyte meiotic progression by regulating chromosome behaviors.
There are six major histone methyltransferases that catalyze H3K4 methylation: SETD1A, SETD1B, MLL1, MLL2, MLL3, and MLL4, which are subdivided into three distinct groups according to their sequence similarity and complex formation: SETD1A/B, MLL1/2, and MLL3/4 17,18 . In most cell types, SETD1-based complexes are the predominant H3K4 methyltransferases 19,20 . The SETD1 complex is driven to trimethylate H3K4 via its essential subunit CxxC-finger protein 1 (CFP1, encoded by the Cxxc1 gene in mammals), which engages in multivalent chromatin binding to recognize both nonmethylated DNA and existing H3K4me3 21,22 . Zygotic knockouts of Setd1a, Setd1b, or Cxxc1 in mice each results in early embryonic mortality due to a failure in gastrulation 23,24 . These pleiotropic phenomena illustrate fundamental roles for the SETD1-CFP1 complex in mammalian development but, at the same time, prevent detailed studies of its functions in adult cell types, including oocytes.
Recently, we generated an oocyte-specific Cxxc1 knockout mouse strain using transgene Zp3-Cre, which is selectively expressed in growing oocytes, and demonstrated that SETD1-CFP1 is the major complex that mediates H3K4me3 accumulation in mouse oocytes 16 . Physiologically, maternal CFP1 and H3K4me3 are essential for the establishment of zygotic developmental competence. Meanwhile, spindle assembly was partially impaired in some CFP1-deleted oocytes that underwent meiotic maturation. Nevertheless, the oocyte transcriptome was only mildly affected in the CFP1-null oocytes; there was no sign of oocyte death and premature ovarian insufficiency in these mice before 6 months of age 16 . A conditional knockout of Setd1b in oocytes also caused oocyte maturation defects and female infertility, but the mechanism was investigated insufficiently 25 . We thus speculate that the SETD1-CFP1 complex and H3K4me3 may be involved in the regulation of meiotic cell division through a previously unidentified transcription-independent mechanism.
Given the fundamental role that the SETD1-CFP1 complex plays in depositing maternal H3K4me3 and sustaining female fertility, we investigate the transcription-independent function of CFP1 and its regulation during oocyte meiotic maturation. In contrast to our previous study in which Cxxc1 was selectively knocked out in growing oocytes by Zp3-Cre, Gdf9-Cre was employed in the current investigation to knock Cxxc1 out in oocytes as early as the primordial follicle stage, thereby giving the oocytes a longer time window to remove the existing CFP1 mRNA and protein after the gene knockout. We demonstrate that CFP1-mediated H3K4me3 accumulation in fully grown oocytes is a priming signal for meiotic cell cycle progression, particularly for T3 phosphorylation of histone H3. Moreover, we find that activity of the SETD1-CFP1 complex is dually inhibited by CDK1-triggered CFP1 phosphorylation and degradation at the onset of G2-M transition. Thus the downregulation of H3K4me3 and removal of CFP1 from chromatin are crucial for proper chromatin structure that is required for successful spindle assembly, accurate chromosome separation, and proper meiotic cell cycle progression.

Results
Dynamic histone H3K4 trimethylation and CFP1 accumulation in cell cycles. In wild-type (WT) oocytes, chromatin-bound H3K4me3 levels increased during transition of chromatin configurations from the non-surrounded nucleolus (NSN) to surrounded nucleolus (SN) type, as determined by immunofluorescence analysis of GV chromatin spreads (Fig. 1a,  b). During meiotic maturation, H3K4me3 levels gradually decreased after GVBD and reached the lowest point in anaphase I (Fig. 1c, d). In contrast, H3K4me1 and H3K4me2 levels increased during meiotic maturation (Fig. 1c, d). The meiotic maturationcoupled fluctuation of H3K4me3 levels suggested that this type of histone modification may perform a direct function in regulating meiotic cell cycle progression of oocytes.
It is noteworthy that the CFP1 protein level underwent a dramatic decrease during the G2-M transition and remained undetectable in maturing oocytes (Fig. 1c). Western blotting of oocyte samples collected at closer time points showed that CFP1 protein level quickly decreased within 1-2 h after oocytes were released from the GV arrest (Fig. 1e) and CFP1 remained undetectable in metaphase II (Fig. 1f). The CFP1 protein reaccumulated in zygotes after fertilization throughout the time period in which the fertilized eggs exited meiosis and entered interphase (Fig. 1f, g). Upon entering the first mitosis, CFP1 was degraded again (Fig. 1f, g). Furthermore, immunofluorescence results in 2-8-cell embryos revealed that CFP1 was degraded in each blastomere when it entered the second, third, and fourth mitosis (Fig. 1g).
In HeLa cells, CFP1 was detected in the nucleus of interphase cells but not in the cells that underwent mitotic divisions (Fig. 2a). In HeLa cells in which the cell cycle was arrested in metaphase by the microtubule disruptor nocodazole, the CFP1 protein was undetectable (Fig. 2b, c). When these cells were released from nocodazole treatment, CFP1 reappeared, accompanying the mitotic exit (Fig. 2b, c). Taken together, these results indicated that the CFP1 protein is transiently degraded during the meiotic and mitotic division. The observation of cell division-coupled CFP1 degradation inspired us to study the function of CFP1 and regulation of protein stability in oocytes and preimplantation embryos.
CFP1 is crucial for spindle assembly and chromosome in murine oocyte. CFP1 is expressed in the nucleus of oocytes as early as the primordial follicle stage (Fig. 3a). To study the function of CFP1 in oocyte maturation, we specifically knocked out Cxxc1 in mouse oocytes as early as the primordial follicle stage using the Gdf9-Cre. Efficient CFP1 protein depletion in oocytes was confirmed by immunohistochemistry (Fig. 3a). Cxxc1 fl/fl ;Gdf9-Cre female mice (hereafter referred to as Cxxc1 oo-/ -) are completely infertile. Nonetheless, the CFP1-null oocytes were able to develop to the fully grown GV stage ( Fig. 3a and   a b  Fig. 1a) and could be ovulated by superovulation treatment (Supplementary Fig. 1b). We collected these oocytes from oviducts and analyzed their potential defects in oocyte meiotic maturation, at 16 h after human chorionic gonadotropin (hCG) injection. Immunofluorescence and confocal microscopic results showed that the majority of ovulated CFP1-null oocytes contained distorted spindles and chromosomes were not properly aligned at the equatorial plate (Fig. 3b). To examine the meiotic maturation process more closely, we isolated fully grown GV oocytes from pregnant mare serum gonadotropin (PMSG)primed Cxxc1 oo-/mice and cultured them in vitro. These oocytes resumed meiosis more slowly than WT oocytes did and manifested reduced GVBD and polar body 1 (PB1) emission rates (Fig. 3c). Similar to those that underwent meiotic maturation in vivo, the in vitro-matured CFP1-null oocytes failed to assemble bipolar spindles, and chromosomes were not aligned at the equatorial plates at both the MI and MII stages (Fig. 3d). We observed that 30% of CFP1-deleted oocytes released PB1s and developed to metaphase II (MII), but the spindles were distorted (Fig. 3d, e). In addition, their chromosomes were mostly aneuploid (Fig. 3f), as counted in chromosome spreads (Fig. 3g).
In Cxxc1 fl/fl ;Gdf9-Cre mice, CFP1 was deleted in oocytes as early as at the primordial follicle stage (Fig. 3a). Therefore, the defects of meiotic maturation in CFP1-null oocytes might have been caused by some indirect and nonspecific factors accumulated during oocyte growth. To test whether CFP1 is directly involved in oocyte meiotic maturation, we expressed a dominantnegative CFP1 mutant (CFP1 C379A ) in fully grown GV oocytes by mRNA microinjection. This mutant protein can bind to DNA but fails to recruit SETD1, and therefore has a dominant-negative effect 26 . The expression of hemagglutinin (HA)-tagged CFP1 C379A at a level comparable with that of the endogenous CFP1 was detected by western blotting (Fig. 3h). In addition, western blotting and immunofluorescence results also indicated that the H3K4me3 level was significantly lower in CFP1-null and CFP1 C379A -overexpressing oocytes (Fig. 3h, i). Notably, in the in vitro maturation experiment, the CFP1 C379A -overexpressing oocytes had a low GVBD rate, similar to that of oocytes isolated from Cxxc1 oo-/mice (Fig. 3c, upper panel). In addition, the PB1 emission rate in CFP1 C379A -overexpressing oocytes was significantly lower than that of the control oocytes (Fig. 3c, lower panel). Furthermore, overexpression of CFP1 C379A impaired spindle formation at both stages MI and MII (Fig. 3d, e). Overall, overexpression of a dominant-negative CFP1 mutant in fully grown oocytes mimicked the phenotype of the oocyte-specific Cxxc1 knockout. Considering the fact that transcriptional activity   is absent in fully grown oocytes, our results indicated that CFP1 has a transcription-independent role in the regulation of oocyte meiotic maturation.
To obtain more direct evidence, we ectopically expressed the H3.3 K4R mutant protein in GV oocytes by mRNA microinjection. According to other studies, H3.3 is the major H3 variant that undergoes K4 trimethylation in mouse oocytes 27,28 . This H3.3 K4R mutant deficient for methylation was expressed and efficiently incorporated into chromatin ( Supplementary Fig. 2a). As predicted, the chromatin-bound H3K4me3 amount was low in oocytes expressing Flag-H3.3 K4R but not in those expressing exogenous Flag-H3.3 WT (Supplementary Fig. 2a). Furthermore, the H3.3 K4R -expressing oocytes had meiosis defects resembling to those of CFP1-deleted oocytes: GVBD and PB1 emission rates CFP1 deletion in oocytes caused meiotic cell cycle arrest in metaphase I. Employing live cell imaging microscopy, we next compared the dynamic spindle assembly and meiotic cell cycle progression between WT and CFP1-deleted oocytes. GV-arrested oocytes were microinjected with mRNA encoding green fluorescent protein (GFP)-tagged α-tubulin (to label spindles) and were cultured in a medium containing Hoechst-33342 (to label DNA). Oocytes that underwent GVBD within 4 h after being released from milrinone were subjected to live cell imaging. Results in Fig. 4a reveal that WT oocytes finished meiosis I and developed to MII at the end of cultivation. In contrast, bipolar spindle formation, equatorial chromosome alignment, and PB1 emission failed in Cxxc1 oo−/− oocytes.
To assess the kinetics of metaphase-to-anaphase transition, we microinjected mRNA encoding mCherry-securin into WT and Cxxc1 oo−/− oocytes and monitored the dynamics of mCherry fluorescence. As shown in Fig. 4a and quantified in Fig. 4b, securin was degraded in WT oocytes approximately 10-12 h after meiotic resumption but remained stable in CFP1-deleted oocytes up to 18 h after meiotic resumption, suggesting that the CFP1deleted oocytes failed to enter anaphase I.
At the end of in vitro maturation culture, oocytes were collected for preparing chromosome spreads. Immunofluorescence results in Fig. 4c show that all WT oocytes under study (n = 26) contained 20 pairs of coherent sister chromatids, and the cohesion core subunit SMC3 was detected only at the kinetochore (labeled with CREST antibody). On the other hand, most Cxxc1 oo−/− oocytes (21 out of 32) contained unseparated homologous chromosomes with remarkable SMC3 distribution on chromosome arms. Thus, in CFP1-null oocytes, the anaphase-promoting complex, which removes securin, and the separase, which in turn removes cohesin, were not activated, and the meiotic cell cycle was arrested at pre-MI.
CFP1-dependent H3K4me3 accumulation is a prerequisite for H3T3 phosphorylation. Phosphorylation of histone H3 at threonine-3 during the G2-M transition is required for oocyte meiotic maturation 3 . T3 of histone H3 is promptly phosphorylated during the G2-M transition. High H3T3ph levels were maintained throughout meiosis I (Fig. 1c, d). Blocking this K4adjacent H3 modification causes similar defects of oocyte maturation as we observed in Cxxc1 oo−/− oocytes 3 . Therefore, we hypothesized that CFP1-dependent H3K4 trimethylation is a priming signal for H3T3 phosphorylation. Consistent with this hypothesis, the H3T3ph level was significantly compromised in Cxxc1-null oocytes at 6 h after in vitro culture (Fig. 5a). Similarly, overexpression of the dominant-negative CFP1 (CFP1 C379A ) also blocked H3T3 phosphorylation on chromosomes (Fig. 5b). In contrast, phosphorylation of histone H3 at Ser-10 (H3S10ph), another cell division-related histone H3 modification, was not affected in Cxxc1 oo−/− oocytes ( Supplementary Fig. 2b, c). This result indicated that the role of CFP1 in regulating H3T3ph is specific.
More directly, when the oocytes overexpressing H3.3 K4R were allowed to undergo meiotic maturation, total and chromosomal H3T3ph levels also decreased ( Fig. 5c-e). The H3T3ph western blot and immunofluorescence results in Fig. 5b, c, e represent only endogenous protein because the anti-H3T3ph antibody does not recognize any K4-mutation-bearing H3 forms, due to the close proximity of these mutations to the T3ph epitope. A coimmunofluorescence assay of H3T3ph and H3K4me3 in MI oocytic chromosome spreads (using a rabbit-and a mousederived antibody, respectively) indicated that these two types of histone H3 modifications did not overlap strongly (Fig. 5f). Rather, H3T3ph preferentially accumulated at the interaction interface between homologous chromosomes, whereas H3K4me3 tended to be located at the surface of paired chromosomes ( Fig. 5f). Collectively, these results strongly indicated that methylated H3K4 regulates T3 phosphorylation in an intermolecular manner.
Histone H3T3 phosphorylation partially reverses the defects in CFP1-null oocytes. The evidence so far indicated that CFP1mediated H3K4 trimethylation is upstream of H3T3 phosphorylation in the control over meiotic progression. Therefore, we determined whether a histone H3.3 mutant that mimics T3 phosphorylation (H3.3 T3E ) can partially reverse the meiosis defects in CFP1-null oocytes. This histone mutant was expressed in GV oocytes of WT and Cxxc1 oo−/− mice by mRNA microinjection. After being released from the GV arrest, the H3.3 T3E expressing CFP1-null oocytes has a GVBD rate close to that of normal oocytes (Fig. 5g). Although the PB1 emission rate was not rescued by H3.3 T3E , the spindles were assembled with better morphology than in CFP1-null oocytes expressing H3.3 T3E (Fig. 5g, h). These results suggested that an important function of CFP1-mediated H3K4 trimethylation is to facilitate H3T3 phosphorylation and that these hierarchical histone modifications are crucial for meiotic cell cycle progression.
CDK1 triggers CFP1 phosphorylation and degradation. The results so far indicated that CFP1-dependent H3K4me3 accumulation in fully grown GV oocytes is a priming signal for proper meiotic cell cycle progression. In contrast, we also observed prompt CFP1 degradation during the G2-M transition in both oocytes and early blastomeres. Therefore, we further investigated the physiological importance and biochemical mechanism of CFP1 removal during meiotic and mitotic divisions. To easily monitor CFP1 degradation in oocytes, we constructed a plasmid expressing a CFP1-GFP-HA fusion protein. After in vitro transcription (IVT) and microinjection of the mRNA into GV oocytes, the CFP1-GFP-HA fusion protein was abundantly expressed. Just as endogenous CFP1, this fusion protein was degraded within 3 h of meiotic resumption (Fig. 6a). Furthermore, its degradation was prevented by addition of the 26S proteasome inhibitor MG132 into the culture medium, suggesting that CFP1 was degraded via the canonical ubiquitin-proteasome pathway (Fig. 6b). After a series of domain-mapping analyses, we found that deleting the C-terminal 65 amino acids (aa) stabilized CFP1 in maturing oocytes (Fig. 6c, d). In HeLa cells transfected with a CFP1-expressing plasmid, deletion of the C-terminal 65 aa significantly inhibited the polyubiquitination of CFP1 (Fig. 6e) and prolonged half-life of the CFP1 protein when protein synthesis was inhibited by cycloheximide treatment (Fig. 6f). Moreover, we noticed an upshift of the CFP1 band on a gel (after sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE)), specifically in MI oocytes, when CFP1 degradation was blocked by MG132 treatment (Fig. 6b) (Fig. 6d). This band upshift was abrogated when the protein sample was pretreated with protein phosphatase (Fig. 6g), indicating that the upshift was caused by CFP1 phosphorylation. Based on this observation, we attempted to identify the protein kinase that mediates meiotic maturation-coupled CFP1 phosphorylation and degradation. In a co-immunoprecipitation (co-IP) experiment, CFP1 ΔC interacted with CDK1, the major kinase that triggers the G2-M transition in the cell cycle (Fig. 6h). We generated constitutively active CDK1, in which the inhibitory phosphorylation sites (Thr-14 and Tyr-15) were mutated to alanine and phenylalanine, respectively. This CDK1 mutant (CDK1 14A,15F ) bound to CFP1 ΔC as well (Fig. 6h). When we coexpressed CDK1 14A;15F and CFP1 ΔC in GV oocytes by mRNA microinjection, CDK1 14A;15F triggered premature degradation of endogenous CFP1 and phosphorylation of CFP1 ΔC (Fig. 6i). On the other hand, specific inhibition of haspin and aurora A, two kinases that also mediate the G2-M transition but downstream to CDK1, did not prevent the GVBD-associated CFP1 phosphorylation ( Supplementary Fig. 3a, b). Taken together, these results strongly indicated that CDK1 directly interacts with CFP1 and phosphorylates it at the onset of the G2-M transition. The CFP1 protein contains multiple consensus CDK1 phosphorylation sites (SP and TP) that are conserved among vertebrates (Fig. 6c). By a series of point mutagenesis analyses, we found that mutation of two SP sites (Ser-138 and Ser-143) close to, but not within, the CXXC finger domain of CFP1 completely blocked CFP1 phosphorylation in maturing oocytes (Fig. 6j). Nevertheless, the full-length CFP1 protein that bears these two point mutations (CFP1 2SA ) was degraded normally after meiotic resumption ( Supplementary Fig. 3c). This result suggests that, although CDK1 triggers both phosphorylation and degradation of CFP1 protein, phosphorylation of CFP1 by CDK1 is not a prerequisite for its degradation during cell division. Therefore, CDK1 may trigger CFP1 degradation through some indirect mechanisms rather than CFP1 phosphorylation. The biochemical and cellular function of CFP1 phosphorylation was further addressed below.
Preventing CFP1 degradation and phosphorylation impairs oocyte maturation. To assess the cellular importance of CFP1 degradation, we microinjected mRNAs encoding WT and nondegradable CFP1 (CFP1 ΔC and CFP1 ΔC;2SA ) into GV oocytes and examined the process of their meiotic maturation. In addition, to test whether the potential function of the stabilized CFP1 depends on its DNA-binding ability, we generated a CFP1 mutation in which a key cysteine residue required for DNA binding (C173) 29 was mutated to alanine in the CFP1 ΔC background (CFP1 ΔC; C173A ). In GV oocytes, both CFP1 WT and CFP1 ΔC were relatively ubiquitously distributed in the nucleus except for the nucleolus (Fig. 7a). CFP1 ΔC;C173A was also located in the nucleus but formed multiple foci that did not overlap with DNA (Fig. 7a). When the microinjected oocytes developed to the MI stage, CFP1 WT was degraded and became undetectable (Fig. 7b). In contrast, CFP1 ΔC and CFP1 ΔC;2SA were stably expressed and enriched on chromosomes, particularly in the kinetochores (Fig. 7b). CFP1 ΔC;C173A was also stabilized but existed as large cytoplasmic particles without co-localization with chromosomes (Fig. 7b).
Ectopic expression of CFP1 WT in GV oocytes did not affect meiotic maturation (Fig. 8a, b). This finding is conceivable because CFP1 was quickly degraded during the G2-M transition. On the other hand, both GVBD and PB1 emission rates decreased in oocytes overexpressing stabilized CFP1 ΔC (Fig. 8b). Confocal microscopy showed that spindle assembly was severely affected in these oocytes (Fig. 8a, middle panels, and Fig. 8b). Moreover, these phenotypes are remarkably strengthened in oocytes overexpressing CFP1 ΔC;2SA (Fig. 8a, bottom panels, and 8b). Although CFP1 ΔC;C173A was not degraded, it did not cause any abnormality during meiotic maturation (Fig. 8a, b), indicating that only the DNA-binding CFP1 is detrimental to meiotic progression of the oocyte.
We further examined the chromosome spreads made from MI oocytes. While the ectopically expressed CFP1 WT was degraded and not detected on chromosomes, CFP1 ΔC was present on the chromosome arms but, unexpectedly, was more concentrated in centromeres (Fig. 8c). Furthermore, the chromosomal H3K4me3 amount was high in oocytes expressing the degradation-resistant CFP1 ΔC (Fig. 8d, e). CFP1 ΔC;2SA overexpression not only caused a more significant increase in the H3K4me3 level than CFP1 ΔC did (Fig. 8d, e) but also impaired the condensation and organization of chromosomes: poorly condensed chromosomes formed a clump in oocytes and single chromosomes were difficult to distinguish (Fig. 8a bottom panels; Figs. 8d, f; 7b, bottom panels). In contrast, CFP1 ΔC;C173A was neither present on chromosomes (Fig. 8c) nor did it affect the H3K4me3 level in oocytes (Fig. 8d, e).
Phosphorylation of CFP1 decreases its chromatin-binding ability. Western blotting analysis of whole-oocyte lysates confirmed that the H3K4me3 level increased at the MI stage in CFP1 ΔC -expressing but not in CFP1 ΔC;C173A -expressing oocytes (Fig. 9a, b). Although CFP1 ΔC was not degraded, it did not completely prevent the decrease of total H3K4me3 levels in MI oocytes (Fig. 9a-c). On the other hand, overexpression of phosphorylation site-mutated CFP1 ΔC (CFP1 ΔC;2SA ) was more effective than CFP1 ΔC at maintaining a high H3K4me3 level in MI oocytes (Fig. 9b, c). Moreover, we generated phosphorylationmimicking mutations in CFP1 ΔC (Ser-138 and Ser-143 were mutated to aspartic acid, CFP1 ΔC;2SD ). When CFP1 ΔC;2SD was overexpressed in GV oocytes, it had a dominant-negative effect by decreasing the H3K4me3 level (Fig. 9d). Taken together, these results indicated that phosphorylation of CFP1 at Ser-138 and Ser-143 inhibits its ability to mediate H3K4 trimethylation, and this phenomenon is independent of CFP1 degradation.
We investigated the mechanism by which CFP1 phosphorylation inhibits H3K4me3 accumulation. In co-IP experiments, mutations of CFP1 phosphorylation sites did not affect CFP1 binding to SETD1 (Supplementary Fig. 3d) but reduced the interaction between CFP1 and histone H3 (Fig. 9e). We ectopically expressed CFP1 WT or CFP1 2SD in GV oocytes at comparable levels (Fig. 9f). CFP1 WT showed an uneven distribution pattern in the nucleus and was more concentrated in the chromatin area (Fig. 9g). On the other hand, CFP1 2SD was evenly distributed in the entire nucleus excluding the nucleolus. In addition, H3K4me3 levels slightly increased in oocytes expressing CFP1 WT but remarkably decreased in oocytes Centromeres and DNA were labeled by CREST immunofluorescence (green) and DAPI staining (blue), respectively. Scale bar, 5 μm. c Western blot of the indicated proteins in WT oocytes ectopically express Flag-tagged histone H3.3 WT and H3.3K4 K4R . d Quantification of H3T3ph signals in (e). Error bars, S.E. M. ***P < 0.001 by two-tailed Student's t tests. n.s., non-significant. e Immunofluorescence of Flag-tagged histone H3.3 (H3.3 WT and H3.3K4 K4R ) and H3T3ph on chromosome spreads prepared at MI stage. Fully grown WT oocytes at the GV stage were microinjected with mRNAs encoding histone H3.3 WT or H3.3 K4R and were released to resume meiotic maturation at 12 h after microinjection. Scale bar, 5 μm. f Co-immunofluorescence of H3K4me3 and H3T3ph on MI chromosome spreads prepared from WT oocytes. Scale bar, 5 μm. g Rates of GVBD, PB1 emission, and normal spindle assembly in oocytes cultured in vitro. Fully grown WT oocytes at the GV stage were microinjected with mRNAs encoding histone H3.3 WT or H3.3 T3E and were released to resume meiotic maturation at 12 h after microinjection. Error bars, S.E.M. ***P < 0.001, **P < 0.01, and *P < 0.05 by two-tailed Student's t tests. n.s., nonsignificant. h Immunofluorescent staining of α-tubulin showing spindle assembly in WT oocytes microinjected with mRNAs encoding Flag-tagged histone H3.3 WT (Fig. 9g, h). This observation is in agreement with the western blot results in Fig. 9d. Furthermore, we prepared chromatin spreads from these oocytes. With this procedure, free proteins and proteins loosely bound to chromatin in the nucleus were removed. Only proteins tightly bound to chromatin remained. Immunofluorescence results showed that CFP1 WT was detected in GV chromatin spreads, but the signal of CFP1 2SD was much weaker in comparison to CFP1 WT (Fig. 9l, j).
The identified CFP1 phosphorylation sites are adjacent to the CXXC domain, which mediates DNA binding (Fig. 6c). Therefore, the potential influence of CFP1 phosphorylation on DNA-binding ability was evaluated by chromatin immunoprecipitation (ChIP) assay in HeLa cells transfected with plasmid expressing HA-CFP1 (WT, 2SD, and the non-DNA binding mutant C173A as a negative control). The binding of CFP1 with its reported target gene promoter regions 22 Fig. 8 CFP1 degradation is required for appropriate histone H3 modifications and normal spindle assembly. a Immunofluorescent staining of α-tubulin showing spindle assembly in oocytes microinjected with mRNAs encoding WT or mutated CFP1. Oocyte subcortical actin microfilaments and DNA were labeled by phalloidin (red) and DAPI (blue), respectively. Scale bar, 20 μm. Fully grown GV oocytes from WT mice were used for mRNA microinjection throughout the experiments presented in this figure. b GVBD and PB1 emission rates in oocytes microinjected with mRNAs encoding CFP1 (WT and mutants), as well as the percentages of oocytes containing a normal spindle at MI stage. GV oocytes were microinjected with mRNAs encoding CFP1 (WT and mutants) and were allowed to resume meiotic maturation for 8 h. Error bars, S.E.M. ***P < 0.001 and **P < 0.01 by two-tailed Student's t tests. n.s., non-significant. c, d Immunofluorescence of CFP1 (c) and H3K4me3 (d) on oocyte chromosome spreads prepared at MI stage. Centromeres and DNA were labeled by CREST immunofluorescence (green) and DAPI staining (blue), respectively. Scale bar, 5 μm. e Quantification of H3K4me3 fluorescent signals in (d). Error bars, S.E.M. ***P < 0.001, **P < 0.01, and *P < 0.05 by two-tailed Student's t tests. n.s., non-significant. f Chromosome spreads made from MI oocytes overexpressing WT and mutated CFP1. Centromeres, chromosome arms, and DNA were labeled by CREST (green), Topoisomerase II (TOP2, red), and DAPI (blue) staining, respectively. Scale bar, 5 μm was evaluated. For the four genes detected, CFP1 WT and CFP1 2SD had comparable binding ability with the genomic DNA, whereas the same genomic DNA fragments were not co-precipitated with CFP1 C173A (Supplementary Fig. 3e). These results confirmed that phosphorylation inhibited CFP1 by impairing its histone-binding, but not DNA-binding, ability.
CFP1 inhibition are crucial for embryo development. We found that CFP1 is degraded not only in dividing mouse oocytes but also in blastomeres and somatic cells in metaphase of the cell cycle. To evaluate the physiological importance of CFP1 inhibition for mitosis, we overexpressed WT or mutated CFP1 in mouse zygotes by mRNA microinjection and monitored the preimplantation developmental process for the next 72 h. While overexpression of CFP1 WT in the zygote did not significantly affect embryo development to the 8-cell stage, the constitutively expressing CFP1 (CFP1 ΔC;2SA ) blocked embryo development beyond a 2-cell stage (Fig. 10a, b; arrows and hollow arrows indicated embryos arrested at 1-and 2-cell stages, respectively). In comparison, overexpression of CFP1 ΔC or CFP1 2SA caused intermediate phenotypes: although CFP1 ΔC decreased embryonic developmental rates to a lesser extent than CFP1 ΔC;2SA did, CFP1 2SA impaired the development from the 1-to 2-cell stage as significantly as CFP1 ΔC;2SA did but was less effective in the next two cell cycles (Fig. 10a, b). This is conceivable because the CFP1 2SA protein was degraded during the first mitotic division. Overall, these results indicated that dual inhibition of CFP1 by phosphorylation and degradation is not only important in oocytes but also crucial for successful preimplantation embryo development.

Discussion
Although H3K4me3 is an evolutionarily conserved and common histone modification, its transcription-independent role in cell division has not been investigated before this study. Several research groups have demonstrated that the amount of H3K4me3 remarkably increases during oocyte maturation, and H3K4me3 is abundantly deposited at numerous broad regions on chromatin of mature oocytes [31][32][33] , but the functional significance of these phenomena remain unclear. It is well recognized that oocytes with an SN chromatin configuration have a better developmental potential than those with an NSN configuration 34 . This notion is in agreement with our finding that the H3K4me3 level in SN oocytes is higher than that in NSN oocytes. Therefore, H3K4me3 is a reliable marker of oocytes with high maturation rates and a post-fertilization developmental potential. Our results suggest that reinforcing H3K4me3 accumulation in human oocytes may be a promising approach to optimizing oocyte quality during human-assisted reproduction.
It is worth noticing that the adjacent threonine residue (T3) of H3K4 is modified by phosphorylation during cell division 35 . Mitotic and meiotic cell cycle-coupled phosphorylation of H3T3 by haspin kinase performs a pivotal function in spindle assembly and chromosome alignment 4,36,37 . The H3T3ph mark recruits survivin (a subunit of the chromosomal passenger complex) to chromatin and is essential for proper chromosome segregation and for modulation of aurora B kinase activity 38 . The haspin inhibitor 5-ITu was reported to impair spindle assembly and chromosome separation in mouse oocytes 3 . In our experiment, CFP1 deletion in mouse oocytes led to phenotypes similar to those in oocytes with inhibited haspin, including compromised GVBD and PB1 emission rates, defective spindle assembly, and failure of homologous chromosome separation. Evidently, meiotic maturation-coupled H3T3 phosphorylation was significantly decreased in Cxxc1 oo−/− oocytes. According to these observations, we hypothesized that CFP1-mediated H3K4 trimethylation is a priming signal for H3T3 phosphorylation and facilitates meiotic resumption (Fig. 10c). This previously unrecognized scheme may ensure proper spindle assembly and chromosome organization during cell division.
During further testing of our hypothesis, ectopic expression of a H3.3 mutant (H3.3 T3E ; which mimics H3T3 phosphorylation) partially rescued GVBD and spindle assembly, but not PB1 emission rates, in Cxxc1 oo−/− oocytes. The rescue effect was not very efficient possibly because it is difficult to control the expression level of H3.3 T3E in oocytes, and abnormally low and high H3T3ph levels are all detrimental to meiotic spindle assembly 4 . Besides, H3.3 T3E may not effectively mimic the precise chromosomal localization and function of H3T3ph.
Interplays between H3T3 phosphorylation and H3K4 methylation have been investigated before by biochemical studies, but contradictory results were obtained. Protein structural analyses reveals that the bulkiness of K4 methylation compromises H3 tail interactions with the narrow substrate-binding groove of haspin 5 . Correspondingly, in vitro kinase assay results indicated that H3K4 methylation impaired H3T3 phosphorylation by haspin 5 . This finding is consistent with the observation that H3T3ph was detected adjacent to, but not overlap with H3K4me3, on chromosome spreads of oocytes (Supplemental Fig. S3D of this study). In return, in vitro and in vivo results derived from nuclear magnetic resonance chemical shift perturbation analysis, fluorescence binding assays, pull-downs, peptide microarrays, and cell fluorescence microscopic studies revealed a strong anticorrelation between histone H3T3 phosphorylation and the ability of PHD finger-containing proteins (including CFP1) to recognize H3K4me3 or to retain in chromatin during mitosis 6 . Contradictorily, co-occurrence of H3T3ph/K4me, H3K9me/ S10ph, and other phospho/methyl modifications has also been experimentally detected, and combinatorial PTM-specific antibody studies reveal that K4me3 and T3ph can be present on the same histone H3 tail 39,40 .
Our results revealed that H3K4 methylation affects H3T3 phosphorylation in an intermolecular manner because the ectopically expressed H3K4 mutant (H3.3 K4R ) decreased T3 phosphorylation of endogenous histone H3. Technically, we could not distinguish the intermolecular effects of monomethylation, dimethylation, and trimethylation of K4 on T3 phosphorylation in this study because all these modifications were abrogated in the H3.3 K4R mutant. Nonetheless, only trimethylation, but not the other modifications of H3K4, was significantly repressed in Cxxc1 oo−/− oocytes. Therefore, we concluded that H3K4me3, rather than H3K4me1 or H3K4me2, triggers H3T3 phosphorylation. In our previous study, a decreased H3K4me3 level in CFP1-null oocytes caused tightening of the chromatin structure and reduced the rate of histone H3 exchange 16 . As a result, the T3 sites in histone H3 on the chromatin of Cxxc1 oo−/− oocytes may not be accessible to haspin. Although our results indicate that defective T3 phosphorylation is an important reason for meiotic maturation failure of Cxxc1 oo−/− oocytes, it is not the only abnormality that matters here. Insufficient accumulation of maternal mRNAs and aberrant distribution of cytoplasmic organelles caused by CFP1 deletion may also indirectly influence meiotic cell cycle progression of oocytes 16 .
In this study, we observed that CFP1 is phosphorylated and degraded during the G2-M transition in both meiosis and mitosis. Chromatin undergoes extensive reorganization throughout the cell cycle, transitioning from a relatively relaxed state in interphase to a highly condensed state in mitosis and returning back to the interphase state after cell division is completed 27 . The degradation of CFP1 is a robust way to remove the SETD1-CFP1 complex from chromatin and therefore facilitates chromosome condensation. Otherwise, chromosome-bound CFP1 proteins, particularly those in the centromeric region, may become an obstacle for chromosome organization and microtubule attachment during spindle assembly. The finding of complete degradation of CFP1 during each cell division reveals a surprising fact that all SETD1-CFP1 H3K4 methyltransferase complexes need to be de novo assembled in the cell after each division, and the distribution pattern of this complex on chromatin needs to be re-established in every cell cycle. This knowledge is conceptually important for understanding the genomewide regulation of H3K4 methylation dynamics in general. For instance, this concept implies that quickly proliferating cells tend to have a more dynamic H3K4me3 distribution pattern, whereas quiescent non-proliferating cells have relatively stable SETD1-CFP1 and H3K4me3 localization on chromatin.
Although our data indicate that CDK1 is involved in CFP1 degradation and phosphorylation, which ubiquitin E3 ligase directly binds to CFP1 and mediates its polyubiquitination is currently unclear. It is conceivable that this ubiquitin E3 ligase is activated by CDK1 during the G2-M transition and may directly interact with the C terminus of CFP1. Nor do we know the signal that recruits nondegradable CFP1 to the centromere region.
CDK1 is a multifunctional kinase and is essential for oocyte meiotic resumption 41,42 . Therefore, we could not test whether In GV oocytes, SETD1-CFP1 complex mediates accumulation of H3K4me3, which is a permissive signal for haspin-mediated H3T3 phosphorylation upon meiotic resumption. During meiotic resumption, CDK1 triggers CFP1 phosphorylation and degradation. Phosphorylation weakened the interaction between CFP1 and histone H3, and ubiquitin-proteasome pathway-mediated degradation removes CFP1 from the condensing chromosomes. This dual inhibition mechanism maintains an appropriate H3K4me3 level that is essential for meiotic cell cycle progression CDK1 inhibitors, such as roscovitine, prevent CFP1 phosphorylation and degradation, because the inhibitors would block oocyte GVBD as well. We cannot completely rule out the possibility that CDK1 activation leads to CFP1 degradation and phosphorylation indirectly as consequences of the G2-M transition. Nonetheless, our experiments showed that (1) CDK1 directly interacted with CFP1; (2) mutations at two putative CDK1 phosphorylation sites (S138 and S143) abrogated meiotic division-coupled CFP1 phosphorylation; and (3) inhibition of other G2-M transition-related kinases, such as haspin and aurora A 43 , did not prevent CFP1 phosphorylation. Together, these results strongly suggest that CDK1 directly phosphorylates CFP1. Although CFP1 phosphorylation is not a prerequisite for CFP1 degradation, the dual inhibition of CFP1 is physiologically important. Stabilization of CFP1 in a maturing oocyte by Cterminal deletion did not completely prevent the reduction in H3K4me3 levels. Only when both CFP1 phosphorylation and degradation were abrogated, was H3K4me3 maintained at high levels in maturing oocytes, and strong defects in oocyte meiotic maturation were observed. The CFP1 phosphorylation sites are conserved among vertebrates and are close to the CxxC finger domain. Mutations that mimic phosphorylation (CFP1 2SD ) did not impair the SETD1-CFP1 interaction but weakened the chromatin binding of CFP1. In agreement with this observation, CFP1 2SD has a dominant-negative effect on H3K4 trimethylation in GV oocytes. Therefore, phosphorylation of CFP1 provides a mechanism for SETD1-CFP1 regulation and may be involved in fine-tuning of genomic histone modification processes. CFP1 is also degraded in mitosis, and constitutively active CFP1 blocks preimplantation murine embryo development. These facts suggest that the dual inhibition of the SETD1-CFP1 complex is a widespread mechanism in cell cycle regulation that is not restricted to oocytes. Furthermore, unidentified extracellular signals, ubiquitin E3 ligases, or kinases other than CDK1 may regulate activity of the SETD1-CFP1 complex via this dual inhibition mechanism in interphase cells in a time-and locationspecific manner.

Methods
Animals. Three-to-four-week-old C57BL/6 mice maintained under specific pathogen-free conditions were used for all experiments. Cxxc1 fl/fl ;Gdf9-Cre mice were produced by crossing mice bearing the previously reported Cxxc1 fl allele with Gdf9-Cre transgenic mice 8,44 . Animal care and experimental procedures were in accordance with the Animal Research Committee guidelines of Zhejiang University.
Oocyte culture. Female mice were primed with PMSG (5 IU) and full-grown oocytes were collected 44 h later by ovarian puncture. Healthy looking fully grown oocytes were collected from antral follicles and cultured in M16 medium (M7292; Sigma-Aldrich) covered with mineral oil at 37°C in a 5% CO 2 atmosphere 45 .
Plasmid construction and IVT. cDNAs encoding human histone H 3.3 and CFP1 were subcloned into a Flag-tagged and a HA-tagged expression plasmid, respectively, and used for IVT. IVT was performed using the T7 and SP6 mMESSAGE mMACHINE Kit (Invitrogen, AM1344 and AM1340) according to the manufacturer's instructions 46 .
Immunofluorescence. Oocytes were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) at room temperature for 30 min. They were then permeabilized with 0.3% Triton X-100 in PBS for 15 min. Antibody staining was performed using standard protocols described previously. Antibodies used in the experiments are described in Supplementary Table 1. Imaging was performed on a Zeiss LSM710 confocal microscope. Semiquantitative analysis of the fluorescence signals was conducted using the NIH Image program Image-J, as previously described 27 .
Chromosome spreading. Zona pellucida-free oocytes were fixed in a solution containing 1% paraformaldehyde, 0.15% Triton X-100, and 3 mmol/L dithiothreitol (Sigma-Aldrich) on glass slides for 30 min and air dried. Immunofluorescent staining was performed as in oocytes described above.
Microinjection of mRNAs and small interfering RNAs. All microinjections were performed using an Eppendorf transferman NK2 micromanipulator. Fully grown GV oocytes were harvested in M2 medium with 2 μM milrinone to inhibit spontaneous GVBD. Approximately 5 pl of 200 μg/ml mRNAs was microinjected into the ooplasm.
Live cell imaging. For live imaging, mRNAs encoding for GFP-tubulin and mCherry-securin were microinjected into WT and CFP1-deleted oocytes and released from milrinone after 8 h. Oocytes were cultured in M16 medium containing Hoechst-33342 (to label DNA) and subjected to live cell imaging at 5 h after meiotic resumption. Live oocytes were acquired on a DV ELITE High Resolution Invented Living Cell Work station. Image acquisition was performed using Zeiss LSM-780 confocal microscope (Zeiss) equipped with PC-Apochromat 20×/0.8 NA objective lenses at 6 min intervals for 13 h.
Immunoblot. Oocytes were lysed in protein-loading buffer and heated at 95°C for 5 min. SDS-PAGE and immunoblots were performed following the standard procedures using a Mini-PROTEAN Tetra Cell System (Bio-Rad, Hercules, CA). The antibodies used are listed in Supplementary Table 1. Uncropped scans of the most important blots are supplied as Supplementary Figures 4, 5.
Cell culture. HeLa cells were from the American Type Culture Collection and were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone) and 1% penicillin-streptomycin solution (Gibco) at 37°C in a humidified 5% CO 2 incubator. Cells were in healthy condition but were not tested for mycoplasma contamination.
Immunoprecipitation assay. The immunoprecipitation (ChIP) assay in HeLa cells was performed using the SimpleChIP Enzymatic Chromatin IP Kit (CST, 9003) according to the manufacturer's protocol. Approximately 4 × 10 6 cells were used for each immunoprecipitation. Immunoprecipitation of immunoglobulin G (CST#2729; 1 µg, as a negative control) or HA (CST#3724; 1 µg per sample) was performed at 48 h after transfection. In all, 10% total DNA was used for input evaluation. PCR products of immunoprecipitated and input samples were analyzed on a 2% agarose gel. Sequences of primers for PCR amplification are provided in Supplementary Table 2.
Statistical analysis. Two-tailed Student's t test was used to calculate P values. Statistically significant values for <0.05, <0.01, and <0.001 are indicated by single, double, and triple asterisk, respectively.
Data availability. The authors declare that all data supporting the findings of this study are available within the article and its supplementary information files or from the corresponding author upon reasonable request.