Rescuing the aberrant sex development of H3K9 demethylase Jmjd1a-deficient mice by modulating H3K9 methylation balance

Histone H3 lysine 9 (H3K9) methylation is a hallmark of heterochromatin. H3K9 demethylation is crucial in mouse sex determination; The H3K9 demethylase Jmjd1a deficiency leads to increased H3K9 methylation at the Sry locus in embryonic gonads, thereby compromising Sry expression and causing male-to-female sex reversal. We hypothesized that the H3K9 methylation level at the Sry locus is finely tuned by the balance in activities between the H3K9 demethylase Jmjd1a and an unidentified H3K9 methyltransferase to ensure correct Sry expression. Here we identified the GLP/G9a H3K9 methyltransferase complex as the enzyme catalyzing H3K9 methylation at the Sry locus. Based on this finding, we tried to rescue the sex-reversal phenotype of Jmjd1a-deficient mice by modulating GLP/G9a complex activity. A heterozygous GLP mutation rescued the sex-reversal phenotype of Jmjd1a-deficient mice by restoring Sry expression. The administration of a chemical inhibitor of GLP/G9a enzyme into Jmjd1a-deficient embryos also successfully rescued sex reversal. Our study not only reveals the molecular mechanism underlying the tuning of Sry expression but also provides proof on the principle of therapeutic strategies based on the pharmacological modulation of epigenetic balance.

Histone H3 lysine 9 (H3K9) methylation is a hallmark of heterochromatin. H3K9 demethylation is crucial in mouse sex determination; The H3K9 demethylase Jmjd1a deficiency leads to increased H3K9 methylation at the Sry locus in embryonic gonads, thereby compromising Sry expression and causing male-to-female sex reversal. We hypothesized that the H3K9 methylation level at the Sry locus is finely tuned by the balance in activities between the H3K9 demethylase Jmjd1a and an unidentified H3K9 methyltransferase to ensure correct Sry expression. Here we identified the GLP/G9a H3K9 methyltransferase complex as the enzyme catalyzing H3K9 methylation at the Sry locus. Based on this finding, we tried to rescue the sex-reversal phenotype of Jmjd1a-deficient mice by modulating GLP/G9a complex activity. A heterozygous GLP mutation rescued the sex-reversal phenotype of Jmjd1a-deficient mice by restoring Sry expression. The administration of a chemical inhibitor of GLP/ G9a enzyme into Jmjd1a-deficient embryos also successfully rescued sex reversal. Our study not only reveals the molecular mechanism underlying the tuning of Sry expression but also provides proof on the principle of therapeutic strategies based on the pharmacological modulation of epigenetic balance.

Author summary
In eukaryotes, DNA wraps an octamer of the core histones. Covalent modifications on the histones have diverse biological functions including transcriptional regulation. Histone H3 lysine 9 (H3K9) methylation is a hallmark of transcriptionally silenced chromatin. In mammals, the sex-determining gene Sry initiates testis differentiation in embryonic gonads. Sry expression in gonads is fine-tuned in both space and time. Here, we demonstrated that fine-tuning of Sry expression is achieved by the balance in activities between H3K9 demethylase and H3K9 methyltransferase. We found that the GLP/ G9a complex is the enzyme catalyzing H3K9 methylation of Sry. Based on this finding, Introduction Covalent modifications of histone tails are epigenetic marks that play roles in many nuclear processes. Among them, methylation of histone H3 lysine 9 (H3K9) is a hallmark of transcriptionally silenced heterochromatin. Various types of H3K9 methyltransferases ("writers") and demethylases ("erasers") have been discovered in mammals. Considering that these H3K9 methylation "writers" and "erasers" are expressed in a cell-type-specific and developmental-stage-specific manner, H3K9 methylation levels are regulated not statically but dynamically during development [1]. In this situation, a specific combination of H3K9 methylation "writer" and "eraser" may antagonistically tune the expression levels of their target genes. We previously demonstrated that H3K9 demethylation plays an indispensable role in mouse sex development [2]. XY mice lacking Jmjd1a (also called Kdm3a), an "eraser" for H3K9 methylation, showed male-to-female sex reversal. Jmjd1a demethylates H3K9 of the sex-determining gene Sry in sexually undifferentiated gonads at embryonic day 11.5 (E11.5), thereby activating Sry transcription. Jmjd1a deficiency induced a decrease, but not a delay of Sry expression in the developing gonads. We found a significant increase of dimethylated H3K9 (H3K9me2) at the Sry locus in embryonic gonads at E11.5 [2], suggesting the existence of an H3K9me2 "writer" that catalyzes H3K9 methylation at the Sry locus.
Several lines of evidence suggest that aberrant histone methylation levels are associated with diseases, including cancer, and intellectual disability [1]. Therefore, normalizing histone modification levels by manipulating the activity of the corresponding modifier is proposed as a potentially powerful therapeutic strategy [3]. Therefore, we speculated that manipulation of the activity of the H3K9me2 "writer" responsible for H3K9 methylation at the Sry locus normalizes Sry expression in the mice lacking the H3K9me2 "eraser" Jmjd1a.
In this study, we identified the H3K9 methyltransferase GLP/G9a complex as the enzyme responsible for H3K9 methylation at the Sry locus. Based on this finding, we aimed to rescue the aberrant sex development of Jmjd1a-deficient mice by modulating the activity of the GLP/ G9a complex. The GLP heterozygous mutation rescued not only H3K9 hypermethylation at the Sry locus but also the perturbed Sry expression in Jmjd1a-deficient embryos. Strikingly, the sex-reversal phenotype in Jmjd1a-deficient mice was completely rescued by the GLP heterozygous mutation. We also aimed to rescue the phenotype by artificially manipulating the activity of the GLP/G9a complex. The administration of the GLP/G9a complex inhibitor into Jmjd1adeficient embryos at a specific developmental time point rescued the aberrant sex differentiation of these mice. Our studies provide a novel strategy by which diseases attributed to the dysfunction of an epigenetic "eraser" can be rescued by blocking the activity of the corresponding epigenetic "writer."
Somatic cells of E11.5 embryonic gonads contain subpopulations with high and low expression levels of an orphan nuclear receptor, Nr5a1 (also called Sf-1/Ad4BP) [12]. Because a previous study demonstrated that the Nr5a1-high population contains Sry-expressing pre-Sertoli cells [13], we examined mRNA expression levels of GLP/G9a in this population ( Fig 1C). We had established a transgenic mouse line Nr5a1-hCD271-tg, in which the human cell surface marker CD271 (also called LNGFR) is expressed depending on the Nr5a1 promoter [2] [14]. We prepared a single cell suspension from the gonads/mesonephroi of E11.5 Nr5a1-hCD271tg embryos and then fractionated it according to the expression level of hCD271 by fluorescence-activated cell sorting (FACS) (S2 Fig). The hCD271-negative population contained mesonephric cells and germ cells [2]. As expected, quantitative RT-PCR (RT-qPCR) analysis demonstrated that the endogenous Nr5a1 expression levels were high and low in hCD271-high and -low populations, respectively (Fig 1C, left). Concordant with the previous study [13], Sry transcript was substantially enriched in the hCD271-high population (Fig 1C, right). Jmjd1a transcript was also significantly enriched in the hCD271-high population. GLP and G9a transcripts were detected in all populations at similar levels, suggesting the ubiquitous expression of GLP/G9a complex in the developing gonads (Fig 1C, right). To address whether GLP/G9a complex and Jmjd1a were co-expressed in Sry-expressing pre-Sertoli cells, we performed triple immunostaining analyses of E11.5 embryonic gonads with antibodies against GLP/G9a, Jmjd1a, and Sry. As shown in Fig 1D and 1E, GLP/G9a complex was expressed in Sry-expressing pre-Sertoli cells. Furthermore, Sry-and GLP/G9a complex-positive cells contained robust signals for Jmjd1a (Fig 1D and 1E). Cells containing both signals of GLP (or G9a) and Jmjd1a among the Sry-expressing pre-Sertoli cells ware calculated. As summarized in Fig 1D and 1E (right panels), almost all Sry-positive cells contained the signals of both GLP/G9a complex and Jmjd1a. We therefore concluded that GLP/G9a H3K9 complex and Jmjd1a are actually coexpressed in Sry-expressing pre-Sertoli cells.
GLP/G9a complex-mediated H3K9 methylation counteracts Jmjd1amediated H3K9 demethylation in gonadal somatic cells We previously demonstrated that GLP is a limiting factor that controls the stability of the GLP/G9a heterodimer complex. In addition, the heterodimer formation of GLP/G9a was shown to be essential for H3K9 methylation in vivo [10] [15]. Thus, we first examined whether a GLP mutation might rescue the increased H3K9me2 levels in Jmjd1a-deficient mice. A homozygous GLP mutation in mice leads to embryonic lethality around E9.5 [10]. We therefore generated mice heterozygous for the GLP mutation (GLP Δ ; lacking the coding sequences for the catalytic SET domain) combined with a Jmjd1a-null (Jmjd1a Δ/Δ ) background. Embryonic gonads/mesonephroi at E11.5 were stained with antibodies against H3K9me2 and Nr5a1. The H3K9 methylation levels of Nr5a1-positive gonadal somatic cells were compared by FACS analysis (Fig 2A). Jmjd1a deficiency resulted in an increased H3K9me2 level in Nr5a1-positive gonadal somatic cells, indicating the substantial contribution of Jmjd1a to H3K9 demethylation (Fig 2B and 2C). Notably, introduction of the GLP mutation into the Jmjd1a Δ/Δ Jmjd1a transcripts were substantially enriched in the hCD271-high population whereas GLP/G9a transcripts were detected in each population (right). mRNA expression levels in the hCD271-low population were defined as 1. Data are presented as mean ± SD. * P < 0.05, *** P < 0.001; n.s., not significant. (D, E) Triple immunofluorescence analyses for GLP (D) and G9a (E), counterstained with anti-Jmjd1a and anti-Sry in the center regions of E11.5 gonads. Enlarged boxes indicate co-expression of GLP (D) and G9a (E) with Jmjd1a in Sry-expressing pre-Sertoli cells (arrowheads). Asterisks represent germ cells. The population of the cells containing both signals of GLP (or G9a) and Jmjd1a among the Sry-expressing pre-Sertoli cells is presented at right. More than 200 cells per embryo (n = 3) were examined. Data are presented as mean ± SD. Scale bar, 50 μm.  Tuning of Sry expression by H3K9 methylation and demethylation background significantly reduced the H3K9me2 level in gonadal somatic cells (Fig 2B and 2C). Thus, we concluded that the GLP/G9a complex counteracts Jmjd1a-mediated H3K9 demethylation in the developing gonads in the sex-determining period.

GLP/G9a complex catalyzes H3K9 dimethylation at the Sry locus
To examine the possible counteracting role of the GLP/G9a complex on Jmjd1a-mediated H3K9 demethylation at single gene level, especially at the Sry locus, chromatin immunoprecipitation (ChIP) analyses were performed. We previously demonstrated that Jmjd1a is enriched at the linear promoter region of Sry in E11.5 gonadal somatic cells [2] (Fig 3A). We purified gonadal somatic cells from E11.5 XY Nr5a1-hCD271-tg embryos and then subjected these cells to ChIP-qPCR analyses. As shown in Fig 3B, we found that G9a is also accumulated in the linear promoter region of Sry. We used Npas4 as a positive control locus, that had been identified as one of the target loci of G9a [16]. We therefore concluded that H3K9 methyltransferase GLP/G9a complex and H3K9 demethylase Jmjd1a both target the Sry locus in embryonic gonadal somatic cells in the sex-determining period.
We next aimed to elucidate the impact of the GLP mutation on the H3K9me2 level of the Sry locus. Gonadal somatic cells were immunomagnetically purified from XY Jmjd1a Δ/Δ , GLP Δ/+ , and Nr5a1-CD271-tg embryos for ChIP analysis (the experimental scheme is shown in S3 Fig). Importantly, the numbers of purified cells were similar among XY Jmjd1a Δ/+ -, XY Jmjd1a Δ/Δ -, and XY Jmjd1a Δ/Δ ;GLP Δ/+ gonads, indicating that these mutations did not affect gonadal somatic cell numbers (S3 Fig). Purified gonadal somatic cells were then subjected to native ChIP-qPCR analyses. Consistent with our previous study, Jmjd1a deficiency resulted in an increase of H3K9me2 at the Sry locus in E11.5 gonadal somatic cells, compared with the level in control cells ( Fig 3C). Importantly, the increased level of H3K9me2 at the Sry locus was significantly rescued by the GLP mutation ( Fig 3C). We also demonstrated the inverse relationship between H3K9me2 and H3K4me2 at the Sry locus. The latter is an epigenetic mark for transcriptionally activated chromatin ( Fig 3C). Taking these findings together, the GLP/G9a complex is the bona fide enzyme responsible for H3K9 methylation at the Sry locus in E11.5 gonadal somatic cells.
Jmjd1a-and GLP/G9a complex-mediated expression tuning is confined to Sry within the Y chromosome genes We next elucidated whether the Jmjd1a mutation and/or Jmjd1a/GLP compound mutations may also induce transcriptional perturbation of Y chromosome genes other than Sry. Gonadal somatic cells were immunomagnetically purified from E11.5 embryos and then subjected to mRNA expression analysis. As shown in Fig 4A, we could not detect significant differences in the mRNA expression levels of Sry-neighboring genes, Uty, Ddx3y, Usp9y, and Zfy2, between control and mutant gonads. Accordingly, our previous microarray analysis showed that the expression levels of Y chromosome genes other than Sry were not affected by Jmjd1a deficiency [2]. Next, we evaluated the H3K9me2 levels of Uty, Ddx3y, Usp9y, and Zfy2 by ChIP-qPCR analysis using purified E11.5 gonadal somatic cells ( Fig 4B). Again, we could not detect significant differences in the H3K9me2 levels of these genes between control and mutant gonadal somatic cells. Taking these results together, Jmjd1a-and GLP/G9a complex-mediated expression tuning is highly confined to the Sry locus and is not extended to other genes on Y chromosome.
It is known that H3K4me3 and H3K9ac are enriched at the linear promoter region of Sry in E11.5 gonadal somatic cells [13]. To address whether Jmjd1a and/or Jmjd1a/GLP compound mutations may influence these modifications, we performed ChIP-qPCR analysis using Gonadal somatic cells were purified from E11.5 XY Nr5a1-hCD271-tg embryos, pooled, cross-linked, and then introduced into ChIP-qPCR analysis. We used Npas4, that had been identified as one of the target loci of G9a, as a positive control locus [16]. Data are presented as mean ± SD. * P < 0.05; *** P < 0.001. (C) ChIP-qPCR analyses for H3K9me2 (left) and H3K4me2 (right) at the Sry locus. Gonadal somatic cells of the indicated genotypes were purified according to the method described in S3 Fig, pooled for each genotype (2 to 4 embryos) and then subjected to native ChIP analysis. ChIP experiment was performed independently twice and gave similar results. Data are presented as mean ± SD. * P < 0.05; ** P < 0.01; n.s., not significant.   [17]. To address whether Jmjd1a deficiency may influence DNA methylation of Sry promoter, we fractionated E11.5 gonadal somatic cells carrying the Nr5a1-hCD271 transgene into hCD271-high and -low populations by FACS and introduced them into bisulfite sequence analysis. In control gonads, the DNA methylation level of Sry promoter was significantly low in the hCD271-high population compared to that of the hCD271-low population in E11.5 embryonic gonads (S4 Fig). These results are in accordance with a previous study [13]. We next compared DNA methylation levels of the Sry .5 XY gonads of the indicated genotypes were dissected and subjected to mRNA expression analysis. We examined the expression levels of Sryneighboring genes (Uty, Ddx3y, Usp9y, and Zfy2), which are located on the short arm of the Y chromosome. There was no significant difference in the expression levels of these genes between control and mutant gonads. mRNA expression levels in the adult testis (3 months) were defined as 1. All data are presented as mean ± SD. (B) ChIP-qPCR analyses for H3K9me2 at the Uty, Ddx3y, Usp9y, and Zfy2 loci. Primer locations are shown at the top. There was no significant difference of the H3K9me2 levels between control and mutant gonads. All data are presented as mean ± SD. promoter of the hCD271-high population between Jmjd1a Δ/+ and Jmjd1a Δ/Δ littermates. However, we could not find significant difference levels (S4 Fig). We therefore concluded that Jmjd1a mutation did not induce a significant alteration of DNA methylation at the Sry locus.
The GLP mutation rescues the reduced expression of Sry in Jmjd1adeficient embryos Sry activation is the first event in mammalian sex differentiation. In mice, sufficient and temporally accurate expression of Sry in sexually undifferentiated gonads at E11.5 is critical for triggering the testis-determining pathway [18]. To address whether the GLP mutation also rescues the perturbed expression of Sry in Jmjd1a-deficient gonads, we examined Sry expression by co-immunofluorescence analysis for Sry and Gata4, a marker of gonadal somatic cells. As shown in Fig 5A and 5B, the number of Sry-positive cells was reduced to approximately 25% in XY Jmjd1a Δ/Δ gonads at E11.5. The number of Sry-positive cells was significantly, although not completely, rescued in XY Jmjd1a Δ/Δ ;GLP Δ/+ littermates, indicating that the GLP/G9a complex and Jmjd1a antagonistically tune Sry expression in E11.5 gonadal somatic cells. We also performed expression analysis of Sox9, a downstream target of Sry, in E11.5 gonads ( Fig  5C and 5D). The number of Sox9-positive cells was also increased by the GLP mutation. Interestingly, the GLP mutation had a more profound effect on the increase of Sox9-positive cells compared with that of Sry-positive cells, presumably reflecting the activation of a non-cellautonomous pathway of Sox9 expression [19].
The GLP mutation rescues abnormal sex differentiation of XY Jmjd1adeficient embryos To examine embryonic gonadal cell differentiation just after sex determination, we investigated the expression of the testicular Sertoli cell marker Sox9 and the ovarian somatic cell marker Foxl2 in XY Jmjd1a Δ/Δand XY Jmjd1a Δ/Δ ;GLP Δ/+ embryonic gonads at E13.5 (Fig 6A). Control XY gonads had Sox9-positive cells and did not contain Foxl2-positive cells. Furthermore, multiple tubule-like structures were found in control XY gonads. On the other hand, Jmjd1a-deficient XY gonads were ovotestes containing not only Sox9-but also Foxl2-positive cells and had no tubule-like structures. As shown in Fig 6A and 6B, the GLP mutation restored the number of Sox9-positive cells and testicular tubule formation, both of which were perturbed by Jmjd1a deficiency, indicating that the GLP mutation successfully rescued gonadal sex differentiation of Jmjd1a-deficient embryos.
GLP and G9a form a heterodimer, which is essential for H3K9 methylation in vivo. We thus also performed epistatic analyses between G9a and Jmjd1a in mouse sex development. Notably, a G9a heterozygous mutation did not rescue the sex-reversal phenotype of XY Jmjd1a-deficient embryos (S5 Fig). Our previous studies indicated that GLP, but not G9a, is a limiting factor controlling the amount of GLP/G9a holoenzyme complex [10]. Consistent with this, we found that the GLP heterozygous mutation induced a significant reduction of G9a protein in embryonic gonads (S6 Fig). Thus, the decreased level of H3K9me2 associated with the GLP heterozygous mutation may be attributable to the reduction in the amount of the GLP/G9a complex.
We had previously established GLP-tg mice [20]. In this line, cDNA for Flag-tagged GLP was inserted in the Rosa26 locus and was designed to be expressed ubiquitously depending on an artificial CAG promoter [20]. To learn whether the overexpression of GLP affects sex determination, we compared the expression levels of Nr5a1, Sry, and Sox9 in gonads of GLP-tg XY embryos at E11.5. As shown in S7 Fig, although  The GLP mutation rescues XY sex reversal in Jmjd1a-deficient adult mice We finally verified the impact of the GLP mutation on the sex-reversal phenotype of Jmjd1adeficient mice in adults. Consistent with our previous study [2], XY mice lacking Jmjd1a alone were frequently sex-reversed; for example, analysis of the external genitalia revealed that of 15 XY Jmjd1a Δ/Δ animals, one carried male-type genitalia, two carried intersex-type genitalia with a micropenis and well-developed mammary glands, and another carried typical female-type Tuning of Sry expression by H3K9 methylation and demethylation genitalia (Fig 7A). In addition, analysis of the internal genitalia demonstrated that one exhibited bilateral testes, two exhibited a testis and an ovary, and another had two ovaries. In contrast, all XY Jmjd1a Δ/Δ ;GLP Δ/+ littermates exhibited male-type external genitalia with bilateral testes (Fig 7B). Taking these findings together, we conclude that the GLP mutation completely rescued the adult sex-reversal phenotype of Jmjd1a-deficient XY mice.

Embryonic administration of the GLP/G9a inhibitor UNC0642 reverses aberrant sex development of Jmjd1a-deficient mice
Our genetic experiments revealed that modulation of H3K9 methyltransferase activity of the GLP/G9a complex might be therapeutically effective for rescuing the aberrant sex Tuning of Sry expression by H3K9 methylation and demethylation development of Jmjd1a-deficient mice. We next aimed to rescue the phenotype in a different way using UNC0642, a chemical inhibitor of the GLP/G9a complex, which was recently developed by Liu et al. [21]. The experimental scheme of UNC0642 administration to Jmjd1a-deficient embryos is shown in Fig 8A. Briefly, 0.5 mg of UNC0642 was intraperitoneally injected once into pregnant females carrying E10.5 Jmjd1a-deficient embryos, and then the gonadal differentiation of the embryos was examined. As shown in Fig 8B and 8C, UNC0642 administration to Jmjd1a-deficient embryos resulted in a significant increase in the number of Sox9-positive male somatic cells at E13.5, while solvent only did not (compare with Fig 6B). This indicates that UNC0642 partially, but significantly, rescued the gonadal sex differentiation of Jmjd1a-deficient embryos after sex determination. We next investigated the impact of the embryonic administration of UNC0642 on the sex development of Jmjd1a-deficient mice by examining the external and internal genitalia of adult mice (Fig 8D). Although partially or completely sex-reversed mice were still found in the UNC0642-administered Jmjd1a-deficient mice, five out of 12 UNC0642-administered animals carried bilateral testes. In contrast, only one out of 11 animals in the solvent-injected control group exhibited bilateral testes (Fig 8D  and 8E). Taking these findings together, we conclude that administration of UNC0642 into E10.5 embryos successfully rescued the subsequent sex development of Jmjd1a-deficient mice.

Discussion
Here, we identified GLP/G9a H3K9 methyltransferase complex as an enzyme counteracting Jmjd1a-mediated H3K9 demethylation at the Sry locus in gonadal somatic cells. To our knowledge, this is the first study to identify the set of histone methyltransferase and demethylase that in combination account for stage-and cell-type-specific gene regulation in mammalian development.
Our data show that the molecular balance of the GLP/G9a complex and Jmjd1a is a critical factor for the tuning of Sry expression (Fig 9). We previously showed that G9a and GLP are expressed in almost all adult tissues in mice [9,10]. On the other hand, previous studies demonstrated that Jmjd1a is expressed in a tissue-and developmental stage-specific manner Tuning of Sry expression by H3K9 methylation and demethylation [22][23][24]. Considering that the expression of Sry is suppressed in almost all adult tissues in mice, this Sry silencing might be explained, at least in part, by the robust H3K9 methylation of the GLP/G9a complex and the absence of H3K9 demethylase in these tissues. We previously demonstrated the temporally specific expression of Jmjd1a in embryonic gonadal somatic Tuning of Sry expression by H3K9 methylation and demethylation

Fig 9. Fine-tuning of Sry expression is achieved by the balance in activities between H3K9 demethylase
Jmjd1a and H3K9 methyltransferase GLP/G9a complex. In wild-type embryonic gonads, Jmjd1a removes H3K9 methylation marks, which were deposited by GLP/G9a complex, from the Sry locus, thereby ensuring Sry activation. In Jmjd1a-deficient embryonic gonads, the absence of Jmjd1a results in the GLP/G9a complexmediated H3K9 hypermethylation at the Sry locus, thereby compromising Sry expression and causing male-tofemale sex reversal. Normalization of the H3K9 methylation balance of the Sry locus by a genetic or a pharmacological approach rescues the aberrant sex development of Jmjd1a-deficient mice by restoring Sry expression.
cells, which reaches a plateau around E11.5 [2]. In addition, we demonstrated the cell-typespecific expression of Jmjd1a in this study, as Jmjd1a transcript was substantially enriched in the Nr5a1-high population of E11.5 embryonic gonads (Fig 1C). High levels of Jmjd1a expression may overcome GLP/G9a complex-mediated H3K9 methylation, thereby inducing Sry expression in the pre-Sertoli cells.
Although the GLP mutation significantly rescued the perturbed Sry expression in Jmjd1adeficient embryonic gonads, the Sry-positive cell population in Jmjd1a Δ/Δ ;GLP Δ/+ gonads remained approximately half of that detected in control gonads (Fig 5B). Therefore, it is a surprising finding that the GLP mutation completely rescued the sex reversal of Jmjd1a-deficient XY mice in adults (Fig 7). The GLP mutation reduces global H3K9me2 levels of Jmjd1a-deficient gonadal somatic cells (Fig 2). Thus, it is also possible that GLP/G9a complex has a role in the sex differentiation pathway, independent of Sry regulation. In this regard, the GLP mutation may inhibit the ovarian development pathway in Jmjd1a-deficient gonads. Although we could not rule out this possibility, it seems likely that the restored Sry expression is a primary cause for the rescue of the sex reversal of Jmjd1a-deficient XY mice. Because there is a certain threshold level for Sry expression to induce the male pathway [18], it is conceivable that Sry expression substantially exceeds the threshold level as a result of the GLP mutation in the Jmjd1a-deficient background, thereby conferring profound effects on the subsequent male pathway.
Eset is known as an enzyme responsible for tri-methylation of H3K9 [6]. A previous study demonstrated that Eset is expressed in embryonic gonads [25]. Accordingly, we also confirmed the expression of Eset in E11.5 XY gonadal somatic cells by RT-qPCR analysis [2]. We introduced the Eset heterozygous mutation into the Jmjd1a-mutant background, and then the sex development of XY Jmjd1a Δ/Δ ; Eset Δ/+ embryos was examined ( S8 Fig). Consequently, we could not find any profound effect of the Eset mutation on the perturbed sex development of Jmjd1a-deficient mice (S8 Fig). Our previous study indicated that the H3K9me3 level of Sry was unchanged by Jmjd1a deficiency in E11.5 embryonic gonads. Altogether, it seems likely that Jmjd1a-mediated H3K9 demethylation does not counteract Eset-mediated H3K9 trimethylation, at least in the Sry locus.
Previous studies revealed the enrichment of H3K4me3/H3K9ac and that low levels of CpG methylation are characteristic in the Sry promoter region of gonadal somatic cells during the sex-determining period [17] [13]. In this study, we demonstrated that Jmjd1a deficiency and/ or the increased level of H3K9me2 did not affect H3K4me3/H3K9ac and CpG methylation levels of the Sry locus (S4 Fig). On the other hand, it seems likely that H3K9me2 and H3K4me2 marks are exclusively deposited mutually and these marks exert antagonistic functions for transcription, at least in the Sry locus. Jmjd1a deficiency and/or the increased levels of H3K9me2 resulted in the decrease of H3K4me2 of this locus, concomitantly with the suppression of Sry [2], also shown in Fig 3. The fact that Jmjd1a deficiency result in the loss of H3K4me2, but not H3K4me3, in the Sry locus warrants further discussion. It is one possible explanation that Jmjd1a or Jmjd1a-mediated H3K9 hypomethylation may prevent the accession of specific enzyme(s) responsible for H3K4me2 demethylation.
We have demonstrated that just a single administration of GLP/G9a inhibitor to E10.5 embryos significantly rescues the aberrant sex development of Jmjd1a-deficient mice. Mutation, silencing, or downregulation of histone methylation "erasers" was found in several types of cancer [26]. Our experiments suggest a new therapeutic strategy, in which diseases arising from the dysfunction of an epigenetic "eraser" can be rescued by blocking the activity of the counteracting epigenetic "writer."

Ethics statement
Animal experiments were performed under the animal ethical guidelines of Tokushima University and Kyoto University. The Ethics Committee of Tokushima University for Animal Research (Approval number: 14108) and the Animal Experimentation Committee of Kyoto University (Approval number: A12-6-2) approved this study.

Animals
Mouse lines of GLP Δ/+ , G9a Δ/+ , Jmjd1a Δ/+ , and Nr5a1-hCD271-tg were sequentially backcrossed to C57BL/6J, and then the F 5 or later generation was used. Since the sex reversal frequencies of Jmjd1a-deficient mice were dependent on the origin of the Y chromosome [2], we used mice carrying a Y chromosome of CBA origin in this study. However, we only used mice carrying a Y chromosome of B6 origin in the experiments shown in S5 Fig.

Histology and immunohistochemistry
Tissues were fixed in either Bouin's solution or 4% paraformaldehyde, embedded in paraffin, and cut into 4-μm sections. For histological analysis, sections were stained with hematoxylin/ eosin or hematoxylin/PAS. For immunohistochemistry, sections were deparaffinized and rehydrated, and then autocleaved at 105˚C for 5 min in 10 mM citric acid buffer (pH 6.0). To quench endogenous peroxidase, the sections were treated with 0.3% (v/v) hydrogen peroxide. After blocking with TBS containing 2% skim milk and 0.1% Triton-X100 at room temperature for 1 h, sections were incubated with primary antibodies overnight at 4˚C. For fluorescence staining, the sections were incubated with Alexa-conjugated secondary antibodies (Life Technologies) at room temperature for 1 h and counterstained with DAPI. The sections were mounted in Vectashield (Vector) and observed with a confocal laser scanning microscope (LSM700, Carl Zeiss).

Flow cytometry and cell sorting
Isolated gonads and mesonephroi from E11.5 embryos were digested with Accutase (Nacalai) to obtain a single cell suspension. For flow cytometric analysis, cells were fixed with 2% paraformaldehyde (PFA) in PBS for 10 min, permeabilized with ice-cold ethanol for 20 min, and blocked with 0.5% skim milk in PBS for 1 h. They were then stained with primary antibodies overnight at 4˚C and subsequently incubated with Alexa-conjugated secondary antibodies (Life Technologies) for 1 h at room temperature. Data were collected using FACSCanto 2 (BD Bioscience) and analyzed with FlowJo software (TreeStar). For FACS sorting, cells were stained with FITC-labeled anti-LNGFR and sorted based on fluorescence intensity using FACS Aria 2 (BD Bioscience) as shown in S2 Fig.

Bisulfite sequencing analysis
Genomic DNA was isolated using the All DNA/RNA Micro kit (QIAGEN). Genomic DNA was treated with sodium bisulfite using the MethylEasy Xceed Rapid DNA Bisulfite Modification Kit (Human Genetic Signatures) following the manufacturer's instructions. The bisulfitetreated DNA was PCR-amplified using the primer pair 5 0 -TTTATATTGGGTTATAGAGTTA GAATAGAT-3 0 and 5 0 -CCAAAATATACTTATAACAAAAATTTTAAT-3 0 . PCR products were subcloned into the pGEM-T Easy vector (Promega) and sequenced.