JMJD1C Exhibits Multiple Functions in Epigenetic Regulation during Spermatogenesis

Jmjd1C is one of the Jmjd1 family genes that encode putative demethylases against histone H3K9 and non-histone proteins and has been proven to play an indispensable role in mouse spermatogenesis. Here, we analyzed a newly-bred transgenic mouse strain carrying a Jmjd1C loss-of-function allele in which a β-geo cassette was integrated into the intron of the Jmjd1C locus. Jmjd1C gene-trap homozygous testes exhibited malformations in postmeiotic processes and a deficiency in the long-term maintenance of undifferentiated spermatogonia. Some groups of spermatids in the homozygous testis showed abnormal organization and incomplete elongation from the first wave of spermatogenesis onwards. Moreover, histone H4K16 acetylation, which is required for the onset of chromatin remodeling, appeared to be remarkably decreased. These effects may not have been a result of the drastic decrease in gene expression related to the events but instead may have been due to the lack of interaction between JMJD1C and its partner proteins, such as MDC1 and HSP90. Additionally, significant decreases in Oct4 expression and NANOG- and OCT4-expressing spermatogonia were found in the Jmjd1C homozygous mature testis, suggesting that JMJD1C may participate in the maintenance of spermatogonial stem cell self-renewal by up-regulating Oct4 expression. These results indicate that JMJD1C has multiple functions during spermatogenesis through interactions with different partners during the spermatogenic stages.


Introduction
Germ cells in mammals are the only cell lineage to transmit species-specific genetic and epigenetic information from one generation to the next. Therefore, the lineage leading to gamete production provides a simple but excellent model system to understand the molecular basis underlying every developmental event. Spermatogenesis features three major processes (maintenance of self-renewing stem cells, cyclic and well-organized differentiation through meiosis and maturation accompanying global chromatin remodeling). These processes are held under strict control via the systematic regulation of gene expression, epigenetic modification and genome DNA integrity. Among these processes, the methyltransferases and demethylases responsible for maintaining the dynamic balance associated with the methylation of lysine (K) residues of core histones have been demonstrated to play a central role in the epigenetic modification of transcription and chromatin structures during spermatogenesis [1,2]. For instance, G9a, which encodes the H3K9 methyltransferase, and Jmjd1A, which encodes the H3K9 demethylase, play essential roles in the regulation of the balanced control of spermatogenesisspecific gene expression by methylated H3K9 [3,4]. However, recent studies have revealed that many histone methylation and demethylation enzymes are not only specific for the K residues of core histones but also act on the K residues of non-histone proteins. Indeed, a methylated K residue is found in more than 400 non-histone proteins located both inside and outside of the nucleus; these proteins are involved in diverse biological events and diseases [5]. Moreover, the finding that G9a plays roles in the methylation of a histone acetyltransferase (HDAC1) and DNA methyltransferase (DMNT1) indicates that methyl modification consists of a comprehensive network that cooperates with other epigenetic regulation mechanisms [5].
Based on our microarray analysis that compared the expression levels of JmjC domain family genes between ES cells, the inner cell mass of blastocysts, epiblasts, and primordial germ cells (PGCs) from E8.5 embryos, we focused on the Jmjd1 family Jmjd1C gene (also known as Kdm3C and Trip8) because its expression level was much higher in ES cells and PGCs than cells of other stages (data not shown). Then, the Jmjd1C ortholog Jmjd1A (Kdm3A), which encodes a demethylase to remove mono/di-methyl modifications from histone H3K9, was revealed for the first time to play an essential role in spermatogenesis. A detailed analysis of the spermatogenic deficiency in Jmjd1A gene-trap mice demonstrated that JMJD1A participated in the transcriptional control of Transition protein 1 (Tnp1) and Protamine 1 (Prm1) through H3K9 demethylation of the upstream promoter regions, leading to postmeiotic chromatin condensation [6]. However, JMJD1A is known to have other functions in metabolic syndrome and sex determination [7,8]. During the progress of our work, a precedent study on Jmjd1C-deficient mice was reported that showed the male infertility phenotype of Jmjd1C null-knockout mice due to the progressive reduction of germ cells after 3 months of age, indicating that JMJD1C contributed to the long-term maintenance of male germ cells [9]. A similar phenotype involving the progressive loss of proliferating spermatogonia was reported in Promyelocytic leukaemia zinc finger (Plzf, also known as Zfp145 and Zbtb16)-knockout mice; the expression of this gene is restricted to undifferentiated spermatogonia [10]. The germ cell degeneration due to the Plzf defect starts at 2 weeks of age, which is apparently much earlier than the Jmjd1C defect; thus, it is unlikely that the progress of germ cells found in the mature testis of Jmjd1Cdeficient mouse is due only to the loss of Plzf expression. Furthermore, recent studies reported the contradictory finding that JMJD1C had no enzymatic histone H3K9 demethylation activity [9,11], which raised a possibility that Jmjd1C might play a non-enzymatic role, such as a serving as a scaffold for the construction of a regulatory apparatus, rather than contributing its enzymatic activity alone. In this study, we investigated the spermatogenic deficiency observed in Jmjd1C gene-trap mutant mice and revealed that Jmjd1C had multiple functions in chromatin remodeling during spermiogenesis and the maintenance of spermatogonial stem cells that were accompanied by multiple partner proteins.
The heterozygous mice (Jmjd1 +/gt ) were backcrossed with 129/svj and Jcl:ICR mice. Homozygous (Jmjd1C gt/gt ) mice were then generated by intercrossing Jmjd1C +/gt mice and identified by PCR based genotyping using the allele-specific primer pairs listed in S1 Table. 10, 94, 24, 8 and  38 mice were used for flow-cytometry, histochemical, immunoblotting, immunoprecipitation, experiments and RNA/DNA analyses, respectively, and total approximately 300 mice including ones for the maintenance of the strain were used for this study. Animal experiments was fully approved by the Keio University Institutional Animal Care and Use Committee (approval number: 12023-2) and all animal care and experimental procedures were performed in accordance with the Institutional Guidelines on Animal Experimentation at Keio University.

Testicular cell preparation and separation by flow cytometry
After treatment with collagenase type IV (Sigma), the testicular tubules were dissociated with a 0.05% Trypsin-EDTA solution (Sigma). The resulting single cell suspensions were used for cell fractionation, nuclear isolation and cell sorting. Subcellular fractionation was performed with the Subcellular Protein Fractionation Kit for Tissues (Thermo Scientific) according to the manufacturer's instruction. For flow-cytometry, the cells were stained with Hoechst 33342 and then analyzed with the FACS Aria III (BD Biosciences).

RNA preparation and QRT-PCR
Total RNA was prepared from mouse tissues with the Qiagen RNeasy kit or the Sepasol-RNAI SuperG kit (Nacalai Tesque, Inc.) and reverse transcribed using an oligo-dT primer and a Superscript III first-strand cDNA synthesis kit (Invitrogen) in accordance with the manufacturers' instructions. The real time (RT)-PCR was performed with Ex-Taq (Takara), and the quantitative real-time (QRT) PCR was analyzed using the SYBR Premix EX TaqII kit (Takara) and ViiA7 (Applied Biosystems). The QRT-PCR reactions were normalized against the β-actin transcript levels. The primer pairs used for the cDNA analysis are listed in S1 Table. Immunohistochemical staining, X-gal staining and immunoblotting analysis For immunohistochemistry, the testes and other tissues were fixed with either Bouin's fixative or 4% Paraformaldehyde (PFA) phosphate buffer solution and embedded in paraffin. After de-paraffinization, 5 μm-thick sections were stained with Hematoxylin or treated with Target Retrieval Solution (Dako, S1699) for unmasking, blocked with 3% bovine serum albumin (BSA) /0.1% Triton X-100, and then reacted with primary and secondary antibodies as previously described [12]. Fluorescent immunohistochemical images were obtained with an Olympus IX71 and the MetaMorph software. For X-gal staining, the tissues were fixed with 0.2% glutalaldehyde/2% formalin buffer and then colorized according to the X-gal staining protocol. For the immunoblotting analysis, testicular cell extracts (15 μg per lane) were separated by SDS-PAGE. The blotted membranes (Immobilon-XP) were immunoreacted with primary antibodies and visualized with the ECL Prime Western blotting Detection kit (GE Healthcare) containing secondary antibodies, HRP (horseradish peroxidase)-labeled anti-mouse IgG and anti-rabbit IgG, and the Quant LAS4000-mini (GE Healthcare) in accordance with the manufacturer's instruction. The antibodies used for immunofluorescence and immunoblotting detection are listed in S2 Table. lysates were prepared with RIPA lysis buffer (10 7 cells/ml). Then, 200 μl of each lysate was reacted with Protein G Sepharose beads pretreated with 2 μg of anti-MDC1 or normal rabbit IgG. IP proteins were extracted from the precipitated beads by the addition of 50 μl of 2x SDS sample buffer. A total of 8 μl of each IP sample and 2 μl of lysate for input per lane were applied for the SDS-PAGE.

Results
Spermatogenic deficiency observed in Jmjd1C gene-trap mice Two major Jmjd1C transcript variants have been demonstrated to be driven by alternative promoters. The long-form transcript encodes a 2530 amino acid protein that is 180 amino acids longer than the short-form protein (2350 amino acids) at the N-terminal end. The Jmjd1Cdeficient mouse line (Jmjd1C gt/gt ) used in this study was generated from a gene-trap ES cell clone carrying a mutation allele in which a SA(splicing acceptor)-β-geo cassette was inserted into the third intron of the long-form and the second intron of the short-form protein; the cassette was integrated approximately 1.0 kb upstream of the next exon (Fig 1). Theoretically, the β-geo insertion interrupts Jmjd1C transcription of both variants and the β-galactosidase translation products precisely reflect the Jmjd1C transcription activity in vivo, which provides us with a convenient and powerful tool to evaluate the Jmjd1C expression patterns in embryos and whole adult tissues. X-gal staining of Jmjd1C +/gt mice showed cell types with various Jmjd1C expression patterns. In mid-gestation embryos (E8.5 onward) undergoing organogenesis, major expression was detected in the neural fold (Fig 2a and 2b). The strong expression in neural cells continued throughout embryogenesis and extended to the brain, spinal cord and sympathetic ganglions during adulthood. Additionally, notable expression was observed in tissues such as the heart, blood vessels, whisker follicles, and stomach (Fig 2c-2j). Jmjd1C expression in the testis, where the X-gal staining intensity appeared to be much lower than neurons, was primarily detected in the inner cell layers of the seminiferous tubule in which meiotic and postmeiotic cells were located (Fig 3Aa). Meiotic and postmeiotic cell expression was confirmed by immunostaining with anti-β-gal and anti-JMJD1C antibodies (Fig 3Ab and 3Ac). Additionally, we found that JMJD1C localized to a part of the chromatin in the nuclei of Synaptonemal Complex Protein (SCP3)-positive spermatocytes ( Fig 3B). Conversely, X-gal staining of the adult ovary showed that Jmjd1C expression was primarily detected in the corpus luteum and was barely detected in the oocytes (Fig 2k).
Despite the wide range of Jmjd1C expression from early embryos to adult tissues, the homozygous mice (Jmjd1C gt/gt ) were normal in appearance and the abnormality was only found in the testes. Consistent with the previous report of null-knockout mice [9], the testes in Jmjd1C gt/gt mice 2-3 months of age appeared to be slightly smaller than the testes from the heterozygote (JmjdC +/gt ) and wild-type mice ( Fig 4A). Cross-sections of the adult testis and epididymis showed a remarkable loss of testicular germ cells expressing the VASA (also known as MVH  and DDX4) protein [12] and few spermatozoa in the Jmjd1C gt/gt mice ( Fig 4B). The previous report noted that the Jmjd1C deficiency appeared only during the long-term maintenance of spermatogonial stem cells [9]. However, in addition to the progressive loss of germ cells, we found another deficiency during spermatogenesis of the JmjdC gt/gt testis. Cross-sections of the Jmjd1C gt/gt adult testis (4-month-old) stained with spermatid-specific anti-Heat Shock Cognate protein 70t (HSC70t) [13] clearly showed the existence of disorganized stratification with abnormally shaped spermatids at the edge of the lumen ( Fig 4C). Additionally, double-staining with anti-HSC70t and spermatocyte-specificanti-SCP3 showed the presence of cell aggregates consisting of spermatocytes and spermatids, which were completely detached from the spermatogenic cell layers in many parts of the seminiferous tubules ( Fig 4D). These findings strongly suggest the importance of Jmjd1C expression for the completion of spermatogenesis in the meiotic and postmeiotic stages.
Examination of the Jmjd1C expression kinetics during the first wave of spermatogenesis showed that Jmjd1C short form expression was highest at birth and then gradually decreased with development, whereas expression of the long form reached its peak around the onset of meiosis, corresponding with the expression of Scp3 and Vasa (Fig 5). Most of the Jmjd1C transcripts appeared to be the long-form variant during testis development (S1A Fig), indicating Comparison of homozygote (gt/gt) testes with heterozygote (+/gt) testes. A) Heterozygote (+/gt) and homozygote (gt/gt) testes removed from 4-month-old littermate mice were compared. B) Histological sections of +/gt (Jmjd1C +/gt ) heterozygote (a, b) and gt/gt (Jmjd1C gt/gt ) homozygote (d, e) testes from 4-month-old littermate mice were stained with anti-VASA (MVH) (a, d), which recognized testicular germ cells at all spermatogenic stages, and DAPI for nuclear staining (b, e). Epididymis sections of the same +/gt (c) and gt/gt (f) littermate mice were counter-stained with Hematoxylin. Scale bars in (a) for (a, b, d, f) and in (c) for (c, f) are 100 μm. C) Testis sections from +/gt (a, b) and gt/gt (c, d) 2-month-old littermate mice were immunostained with anti-HSC70t, which specifically reacted with round and elongated spermatids. Section of a +/+ littermate testis stained with anti-HSC70t showed almost the same staining image with that of +/gt testis (data not shown). DAPI staining is shown in the same fields (b, d). Scale bar in (a) for (a-d), 100 μm. D) Section of a gt/gt testis double-stained with anti-HSC70t (a) and anti-SCP3 (b), which specifically recognized spermatids and spermatocytes, respectively. (c) is a merged image of (a+b), shown together with nuclearstaining with DAPI (d). Scale bar in (a) for (a-d), 100 μm. that the majority of Jmjd1C transcription in germ cells from the meiotic stage onwards was driven by the long-form promoter. The high level of Jmjd1C expression in the meiotic and postmeiotic cells was also confirmed at the protein level (S1B Fig). Here, although the anti-JMJD1C antibody used in this study was raised against an oligo-peptide located within exon 10 that was common between both the short and long forms (Fig 1), we barely detected a shortform protein band by immunoblotting.
When the difference in Jmjd1C expression was examined between genotypes, we found that our Jmjd1C gt/gt mouse was not completely null for Jmjd1C expression. The quantitative real time (QRT)-PCR analysis indicated that the Jmjd1C +/gt and Jmjd1C gt/gt testes expressed Jmjd1C transcripts at levels that were nearly 60% and 20% of the levels in the wild-type testis, respectively (Fig 6 and S2A Fig). Indeed, a small amount of full-length JMJD1C protein was detectable in the Jmjd1C gt/gt testis by immunoblotting (S2B Fig). Therefore, we think it is feasible that a small amount of wild-type transcript is produced when a β-geo cassette is integrated within the intron region of a target gene, possibly by alternative splicing of the read-through precursor transcript containing the inserted β-geo cassette. A similar phenomenon was reported in Jmjd1A-deficient mice [6] in which even an incomplete knockout led to the same phenotype as the null-type knockout. Therefore, Jmjd1C is likely to have a quantitative threshold for completing its function that is probably at least 20% greater than the wild-type level. In contrast, the QRT-PCR analysis did not reveal drastic changes in the expression levels of selected spermatogenesis marker genes, such as Scp3, Vasa, and Prm1 (Fig 6). These results suggest that JMJD1C is unlikely to have a direct effect on the transcriptional regulation of these germ cell-specific genes.
Based on their appearance, the Jmjd1C gt/gt testes from mice at two months of age were not different from the testes of their littermate controls. Nevertheless, cross-sections stained with antibodies against spermatid-specific proteins revealed a subtle contrast between the Jmjd1C gt/ gt and control mice (Fig 7). When sections of the wild-type and Jmjd1C +/gt testes from postnatal day 54 mice were stained with anti-HSC70t and ant-TNP1, the positively stained spermatids lining the lumen were localized in a well-ordered array. In contrast, in the Jmjd1C gt/gt testis of the littermate, the positively stained spermatids were located in a disarrayed manner and the elongation of the spermatids appeared incomplete (Fig 7A). Similar abnormalities in the spermatid arrangement were observed in the Jmjd1C gt/gt testis at postnatal day 30 at the final stage of first wave spermatogenesis. Newly generated sperm heads were radially arranged in the Jmjd1C +/gt testis but randomly localized in the Jmjd1C gt/gt testis (Fig 7B and S3 Fig). Thus, the morphogenetic disorganization found in Jmjd1C-deficient spermatogenesis may be the cause of the clustered detachment of spermatogenic cells at the later stage, as shown in Fig 4D.

Chromatin remodeling abnormality during spermiogenesis
Regarding the enzymatic activity of JMJD1C itself, several reports have provided negative evidence against demethylase activity for histone H3K9 [9,11]. Instead, demethylase activity for non-histone protein(s) has been recently reported. MDC1 (Mediator of DNA damage checkpoint protein 1), which is a core factor of the DNA damage response (DDR) complex, was first identified as a non-histone target protein of demethylation by JMJD1C [14]. The well-defined study demonstrated that JMJD1C-dependent demethylation at MDC1 K45 was not required for its function in the DDR. Mice lacking Mdc1 were demonstrated to undergo meiotic arrest at the mid-pachytene stage, leading to male sterility. Additionally, MDC1 plays an indispensable role in the formation and function of the XY body, which is a meiosis-specific structure formed around the paired X/Y chromosomes in a spermatocyte [15]. Staining of sections with antibodies against MDC1 and JMJD1C showed the co-localization of both proteins in the XY body (Fig 8), corresponding to their functional cooperation during the meiotic processes. However, the Jmjd1C gt/gt testis did not show the meiotic arrest found during MDC1 deficiency and did not exhibit abnormalities in MDC1 localization and XY body formation (S4A Fig), suggesting that JMJD1C was not essential for MDC1-mediated XY body formation. JMJD1C has not been demonstrated to promote an interaction between MDC1 and another core factor, Ring finger protein 8 (RNF8); these three factors bind to one another to form a core DDR complex [14]. Interestingly, mice lacking Rnf8 (encoding an ubiquitin ligase) also exhibited spermatogenic deficiency, which caused an imperfection at the postmeiotic stage similar to the Jmjd1C gt/gt phenotype described above [16]. The Rnf8 defect was revealed to appear during the histone-protamine transition process via histone H4K16 acetylation [16]. However, RNF8 primarily localizes in spermatocytes and showed no remarkable changes in the Jmjd1C gt/gt testis (S4B Fig).
Although the immunostaining of testis sections with anti-methylated lysine showed no differences between the Jmjd1C genotypes (S4C Fig), immunoblotting analysis with the same antibody revealed several non-histone protein bands with significantly increased methylation levels in the Jmjd1C gt/gt testis. A band representing an approximately 90 kDa protein appeared to be hypermethylated in the total extract from the Jmjd1C gt/gt testis; this band was also detected as a hypermethylated protein immunoprecipitated (IP) with anti-MDC1 from the Jmjd1C gt/gt testis ( Fig 9A). Judging from the size and the lysine-methylated modification, the hypermethylated protein should be HSP90. Indeed, we found that HSP90 was immunoprecipitated with anti-MDC1 ( Fig 9B) and co-localized with MDC1 in spermatogenic cells (Fig 8j-8l and S4A Fig).
Based on this line of analysis, we examined histone H4K16 acetylation (H4K16ac) in the Jmjd1C gt/gt testis. Immunostaining with anti-H4K16ac showed a remarkable decrease in the Jmjd1C gt/gt adult testis compared with the littermate controls (Fig 10). The significant decrease in H4K16ac in the Jmjd1C gt/gt testis was validated by the quantitative analysis of the protein bands detected with the anti-H4K16ac antibody, with the values of the histone H3, VASA or HSC70t bands corresponding to the number of total cells, germ cells or spermatids, respectively, that were utilized as the normalization standards (Fig 11 and S5 Fig). Furthermore, although immunostaining with anti-PRM1 clearly revealed expression in the spermatid nuclei of the control littermates, the staining signals were barely observed in the spermatid nuclei from the Jmjd1C gt/gt testis (Fig 10). Regarding histone H4K16 acetylation during spermiogenesis, Mof (Kat8), which encodes an acetyltransferase required for DDR, and Nat8f4 (1700019G17Rik), which encodes a testis-specific putative acetyltransferase, are responsible for H4 hyper-acetylation in a ubiquitination-dependent and -independent manner, respectively [17,18]. However, MOF expression showed no apparent difference in localization between the testes of Jmjd1C wild-type and homozygous littermate mice (S4D Fig). Additionally, the expression levels of selected genes related to the core histone removal (Rnf8, Rnf168, Mdc1, Nat8f4 and Mof), chromatin remodeling (Chd5, Brdt and Baz1a) and proteasome activation (PA200, also known as Psme4) pathways in spermatids showed no remarkable differences between the testes of the three different Jmjd1C genotypes, whereas a somewhat moderate reduction was detected in Rnf8 and Rnf168 (Fig 12) [15,16,[19][20][21][22][23][24]. Altogether, these results suggest that Jmjd1C has an indirect function in the regulation of histone H4K16 hyper-acetylation that leads to histone-protamine remodeling during sperm formation, although the action mechanism remains undefined.

Deficiency in the maintenance of spermatogonial stem cells (SSCs)
Previous reports demonstrated that Jmjd1C deficiency caused an abnormality in the long-term maintenance of spermatogonial stem cells (SSCs) leading to the degeneration of testicular tubules after 3 months of age [9]. Almost the same testis degradation time course was observed in the testis from our Jmjd1C gt/gt mouse (Fig 4). The developmental kinetics of Jmjd1C expression showed that the short-form variant of Jmjd1C was predominantly expressed in spermatogonia containing SSCs (Fig 5). The JMJD1C function related to stem cell potential has been well investigated in pluripotent ES cells. For instance, OCT4 (POU5F1) and primarily the short form of JMJD1C bind to one another's promoter regions to activate transcription in a positive feedback-like loop, and Nanog expression promoted by an OCT4 and JMJD1C compound plays a critical role in the maintenance of pluripotent stem cell-ness [17,25].
In this connection, an interesting finding was the existence of a rare subpopulation of undifferentiated spermatogonia expressing both NANOG and OCT4 that possibly corresponded to radiation-resistant SSCs in a certain quiescent state [26]. By immunohistochemical staining with anti-OCT4 and anti-NANOG, we detected OCT4-positive cells in spermatogonia residing in the outermost layer of the seminiferous tubule and extremely low numbers of NANOG-positive cells attached to the basement membrane of the Jmjd1C +/gt testis (Fig 13A and 13B). Moreover, Jmjd1C-positive cells recognized by X-gal staining of the Jmjd1C +/gt seminiferous tubules were detected not only in the meiotic and postmeiotic cell layers but also in a specific portion of the spermatogonial cell layers as scattered cells on the basement membrane ( Fig  13C). Additionally, extremely low numbers of strongly Oct4-GFP-positive cells were detected in a similar location in the Oct4-GFP/Vasa-RFP transgenic mouse testis (S6 Fig). In contrast, OCT4-positive cells were dramatically decreased, and NANOG-positive cells were barely detectable in the Jmjd1C gt/gt testis (Fig 13A and 13B). The reduction of OCT4 expression in the Jmjd1C gt/gt testis (less than half of the expression of the littermate controls) was validated by the immunoblotting analysis (Fig 13D).
Several candidate genes important for the maintenance of SSC potential, such as Bcl6b, Lhx1, Etv5 and Tex19.1, have been nominated based on findings that mouse SSCs can be  S5 Fig. B) Comparison of the detected protein bands between the +/+ and gt/gt extracts. C) Semiquantitative comparison of the acetylated H4K16 bands between the +/+ and gt/gt testes extracts. Values obtained from the +/+ and gt/gt testes were normalized with those of the histone H3, VASA and HSC70t protein bands, which roughly corresponded to the numbers of total cells, germ cells at stages from spermatogonium to late spermatid, and spermatids, respectively. Bars in the graph show relative ratios (%) of values of the H4K16ac band in gt/gt against those in +/+ after normalizing with values of standard protein bands as indicated below. Error bars indicate the SEM (n = 4-6).
doi:10.1371/journal.pone.0163466.g011 cultured in a chemically defined medium containing GDNF (which is required for SSC selfrenewal) and that withdrawal of GDNF from the medium and long-term culture in vitro leads to a loss of SSC potential [27,28]. To evaluate the effect of Jmjd1C deficiency on SSC potential, the expression of selected genes related to SSC self-renewal and proliferation was compared between the Jmjd1C gt/gt testis and the littermate controls. As shown in Fig 14, most SSC potential genes showed no significant differences in their expression levels. The most remarkable reduction in the Jmjd1C gt/gt testis was detected in Oct4 expression. Altogether with the finding that both in vivo and in vitro SSCs express Oct4, it is conceivable that Oct4 expression under the control of JMJD1C is closely related to the maintenance of SSC self-renewal and/or survival similar to the pluripotency of ES cells.

Multi-faceted functions of Jmjd1C during spermatogenesis
Jmjd1C is a member of the lysine demethylase (Kdm) 3 family, which consists of three members (Jmjd1A, B, and C). Among them, Jmjd1A is the first gene shown to play an essential role in the epigenetic regulation of spermatogenic gene expression [6]. The following studies have also revealed its multiple roles in metabolic homeostasis and sex determination through H3K9 demethylase activity and/or non-enzymatic activity [6][7][8]29], raising the possibility that the KDM3 family has a common function not only as an epigenetic regulator of transcription but also as a fine modulator of the post-translational modification of non-histone proteins. Indeed, Jmjd1C showed no demethylation activity towards dimethylated H3K9 when dimethyl-peptides were used as the substrates [11], suggesting that the spermatogenic abnormalities found in the Jmjd1C gt/gt testis were probably due to JMJD1C activities other than the demethylation of histone H3K9. Nevertheless, several findings showed a relationship between JMJD1C and H3K9 demethylation (i.e., the P450e17 gene promoter in Leydig cells [30]). Similarly, the Oct4 and miR302 promoter regions in ES cells have been validated to be under the regulation of JMJD1Cdependent hypomethylation of H3K9 modification to retain their active states of transcription [31,32]. The pluripotency-associated microRNA miR302 represses the neural differentiation of ES cells, whose transcription is enhanced by the OCT4 and JMJD1C compound [31]. Moreover, we found that JMJD1C was associated with a restricted portion of chromatin in pachytene spermatocytes (Fig 2). Altogether, it seems likely that a cell-type specific cofactor(s) may be required to fulfill the JMJD1C demethylase activity towards methylated H3K9 in a locus-specific manner or that locus-specific association of enzymatically inactive JMJD1C is required to maintain the hypomethylated H3K9 state upon chromatin remodeling by blocking repressive methyl modification of the target loci. Here, a precedent study suggested that Jmj-C family proteins had a non-enzymatic function that was required for normal development (i.e., the H3K27 demethylases JMJD3 and UTX played a crucial role in general chromatin remodeling in differentiated cells that was entirely independent of their demethylase potential) [33]. Therefore, we can generalize that Jmj-C family proteins also have a non-catalytic function through interactions between lineage-specific transcription factors and remodeling complex components to keep the promoter region active in the epigenetic microenvironment.
JMJD1C has been shown to have the ability to demethylate non-histone proteins. For instance, as a core component of the DNA damage response (DDR) complex, JMJD1C binds to and demethylates MDC1 at K45, thereby promoting the MDC1-RNF8/RNF168 interaction required for the recruitment of RAP80-BRAC1 to DNA double-strand break sites [14]. The function of the DDR complex in spermatogenesis was visualized as the co-localization of JMJD1C   Fig 14. QRT-PCR analyses of SSC related genes between the +/+, +/gt, and gt/gt testes. Singlestranded cDNA was prepared from the +/+, +/gt and gt/gt testes of adult (2-month-old) littermate mice. Expression levels of selected genes related to spermatogonial stem cell (SSC) characteristics were compared. The values with SEM (n = 4) were the relative gene expression values in the +/+ testis (set as 1.0) after standardization with the β-actin expression level in each testis. and MDC1 at the XY body found in the spermatocyte nuclei (Fig 8). Additionally, MDC1 is critical for XY body formation and meiotic progression [15]. However, Rnf8 and Rnf168, which encode E3 ubiquitin ligases, and Jmjd1C are dispensable for these processes [16,19]. Rnf8 functions upstream of Rnf168. Rnf168-deficient mice exhibit a defect in global nucleosome removal in the postmeiotic stages, whereas RNF8-dependent histone ubiquitination promotes H4K16 acetylation, which is an indirect trigger for the replacement of histones by protamines [16]. A reduced H4K16 acetylation level was also found in the Jmjd1C gt/gt testis (Figs 10 and 11), leading to the speculation that Jmjd1C defects were at least partially due to the insufficient interaction between DDR components and that the mutual interactions within the DDR machinery played an important role in general chromatin remodeling through modulating H4K16 acetylation.
Regarding other possible candidate non-histone protein targets of the JMJD1C demethylase, we found several proteins that appeared to be hypermethylated in the Jmjd1C gt/gt testis compared with the testes of control littermates. Among them, a 90 kDa protein appeared to be identical to a hypermethylated 90 kDa protein that co-immunoprecipitated with MDC1 from the Jmjd1C gt/gt testis lysate (Fig 9A). Our findings that the cytoplasmic protein chaperone HSP90α, which is a highly methylated protein, was immunoprecipitated with anti-MDC1 ( Fig  9B) and that both proteins were co-expressed in germ cells at the same spermatogenic stages (S4A Fig) strongly indicated that HSP90α was a possible target of the JMJD1C demethylase function. Moreover, previous studies showed that both HSP90α and JMJD1C co-immunoprecipitated with the histone methyltransferase WHISTLE, which raised a possibility that HSP90α had broad interacting activity with JmjC-family proteins, such as JMJD1A and KDM6A [29,30]. Together, these results suggest that a protein-protein interaction between HSP90α and JMJD1C (belonging to the DDR complex) plays a crucial role in the modulation of mutual functions, probably by altering the methylation status.
Intriguingly, Hsp90α-deficient mice also exhibit male infertility [34]. HSP90α is methylated by SMYD2 and is involved in the organization of cytoskeletal components, such as γ-tubulin and β-actin, leading to the disorganization and abnormal shaping of spermatids observed in the Hsp90α-deficient testis [34,35]. Moreover, a recent report revealed that the H3K9 demethylase activity of JMJD1A was dependent on the interaction with either HSP90α or HSP90β and that the Jmjd1A mutant showed cytoplasmic defects in cytoskeletal rearrangement during spermiogenesis in addition to the Jmjd1A defects that were reported previously [29]. Interestingly, because demethylation of the HSP90α K616 residue is not altered by the lack of Jmjd1A, another JmjC-containing protein may be responsible for the HSP90 demethylation. These results also support the possibility that HSP90α is a target of JMJD1C's demethylase activity and raise a question as to whether the JMJD1C-dependent defect in the HSP90 machinery leads to the disorganized rearrangement and malformation of spermatids found in the Jmjd1C gt/gt young testes. Altogether, it is quite comprehensible that JMJD1C serves as a scaffold to recruit various partners and promote mutual interactions by modulating their methylation statuses at each spermatogenic stages.

Progressive loss of spermatogonial stem cells (SSCs)
Jmjd1C gene-trap mice in this study were backcrossed on to a 129/svj or Jcl/ICR background, whereas the complete null knockout mice in the previous study were analyzed on a C57BL/6 background [9]. It has been reported that several lines of Jmjd1A knockout mice show a drastic change of phenotype on the different genetic background [6][7][8]36]. In contrast, although we can not deny a possibility that the difference between gene-trap and null-type knockout, and that of the genetic background causes a subtle phenotypic difference, the testis-restricted deficiency that asynchronous degradation of germ cell layers in the seminiferous tubules was common in both Jmjd1C-deficient mouse lines (Fig 4). In the null knockout mice, the progressive loss of undifferentiated spermatogonia was detected based on PLZF expression [9]. Plzf encodes a repressive transcription factor that is a well-known marker gene expressed in both SSCs and proliferating spermatogonia in the adult testis. Lack of Plzf function causes a progressive loss of spermatogonia without any differentiation defects, although the timing of degeneration appeared to be much earlier than the timing observed for the Jmjd1C defect [10]. Continuous proliferation of SSCs in vitro [referred to as germline stem cells (GSCs)] is maintained in a defined culture medium containing GDNF, LIF, EGF and bFGF. Stem cell potential is evaluated as a frequency of regenerated spermatogenesis based on transplantation into recipient testicular tubules in vivo. GSCs tend to lose their stem cell potential after long-term culture or withdrawal of GDNF. From this perspective, several genes, such as Lhx1, Etv5 and Bcl6b, have been identified to be important for stem cell potential [27]. However, we found that these genes showed no significant differences between the Jmjd1C gt/gt testis and the littermate control testis; instead, the most remarkable difference in the Jmjd1C gt/ gt testis was a reduction in Oct4 expression (Fig 14). GSCs express a set of iPS cell-inducing genes (Oct4, Sox2, Myc and Klf4), although the levels are somewhat lower than the levels in ES cells [37]. However, these cells do not express Nanog, which is in contrast to the ES and iPS cells [28]. Importantly, a few GSCs spontaneously dedifferentiate into pluripotent stem cells (referred to as multipotent GSCs) during the initiation of GSC cultures from postnatal testes [37].
A recent report that a small subpopulation of NANOG/OCT4-positive cells was found in the outermost cell layer attached to the basement membrane of seminiferous tubules in a specific spermatogenic stage was very intriguing [26] because it suggested that the conversion of cellular characteristics leading to NANOG expression might occur in both in vitro and in vivo self-renewing SSCs. NANOG-positive SSCs in the adult testis were also detected in this study but were undetectable in the Jmjd1C gt/gt testis, probably due to the remarkable decrease of OCT4-positive cells (Fig 13). JMJD1C has been reported to be essential for the maintenance of pluripotency via promoting the active expression of Oct4 and Nanog during ES cell proliferation. Additionally, up-regulation of Jmjd1C is required for the generation of completely reprogrammed iPS cells that express endogenous Oct4 [32]. Provided that the self-renewal of SSCs is based on maintaining their potentially pluripotent state, it is reasonable to speculate that JMJ1C plays a critical role in maintaining the self-renewal and stem cell potential of SSCs via essentially the same function described for pluripotent cells.
Currently, JMJD1C is thought to play an extensive role related to various stem cell potentials. As described above, OCT4 is regarded as a functional partner of JMJD1C in pluripotent stem cells and probably also SSCs. Similarly, JMJD1C is assumed to participate in the maintenance of proliferative stem cells through association with lineage-specific transcription factors. Indeed, several reports have noted that JMJD1C is not required for the survival and selfrenewal of leukemia stem cells. For instance, JMJD1C functions as a co-activator for Runx1--Runx1T1 in acute myeloid leukemia cell lines [38] and for HoxA9 in mixed lineage leukemia-AF9 and HoxA9-driven leukemia [39]. Therefore, it is conceivable that JMJD1C is recruited by partner transcription factors to their target genes and activates their expression by maintaining low H3K9 methylation levels, although whether JMJD1C acquires demethylase activity through association with a partner is not clear.
As shown in Fig 2, Jmjd1C expression was not restricted to the testis. Instead, testicular expression was at a relatively moderate level among adult tissues, and high level expression was primarily observed in the developing nervous system in embryos and neurons in the adult brain. Nevertheless, an apparent morphological or behavioral disorder was not found in the Jmjd1C gt/gt adult mice, which might be due to a functional redundancy with other Jmj family genes. However, recent progress in the genetic analysis of human diseases noted possible causal relationships between Jmjd1C mutations and mental disorders as well as relationships with various cancers [25,40]. For instance, Jmjd1C is a candidate gene for autism [41] because the breakpoint of chromosomal abnormality in the patient is located in the first intron of Jmjd1C. Moreover, Jmjd1C mutations have been identified in a subpopulation of Rett syndrome patients [42], suggesting that Jmjd1C could be related to higher order brain functions, such as mental faculties.
Further detailed investigations to clarify the partners and targets of JMJD1C functions in each cell lineage will provide important clues to understanding the action mechanism of the JMJD1C-dependent regulatory network and to develop novel therapeutic targets for cancers and mental disorders. Approximately 15 μg protein extracts from +/+ testes (4-month-old) were applied on SDS-PAGE (10-20%). The western-blotted membranes were separately reacted with antibodies against HSC70t, VASA, histone H3 (H3) and acetylated H4K16 (H4K16Ac) and then detected with HRP-linked anti-rabbit IgG and an ECL detection reagent. (TIF) S6 Fig. OCT4-expressing cells localized beneath the basal membrane of the seminiferous tubule. Testicular tubules were prepared from an adult (8-month-old) transgenic mouse carrying Oct4-GFP and Vasa-RFP [43]. Strong GFP-positive cells with Oct4 expression were observed just inside the basement membrane only in a restricted area of the seminiferous tubules. Merged GFP and phase contrast (a) or RFP and phase contrast (b) images in the same field are presented. Scale bar, 50 μm. (TIF) S1