DNMTs and SETDB1 function as co-repressors in MAX-mediated repression of germ cell–related genes in mouse embryonic stem cells

In embryonic stem cells (ESCs), the expression of development-related genes, including germ cell–related genes, is globally repressed. The transcription factor MAX represses germ cell–related gene expression in ESCs via PCGF6-polycomb repressive complex 1 (PRC1), which consists of several epigenetic factors. However, we predicted that MAX represses germ cell–related gene expression through several additional mechanisms because PCGF6-PRC1 regulates the expression of only a subset of genes repressed by MAX. Here, we report that MAX associated with DNA methyltransferases (DNMTs) and the histone methyltransferase SETDB1 cooperatively control germ cell–related gene expression in ESCs. Both DNA methylation and histone H3 lysine 9 tri-methylation of the promoter regions of several germ cell–related genes were not affected by knockout of the PRC1 components, indicating that the MAX-DNMT and MAX-SETDB1 pathways are independent of the PCGF6-PRC1 pathway. Our findings provide insights into our understanding of MAX-based repressive mechanisms of germ cell–related genes in ESCs.

Introduction (G9a -/-) ESCs [17] were cultured as described previously. All ESCs were cultured in conventional ES medium with serum and leukemia inhibitory factor (LIF). Max-null and Dnmts-TKO ESCs were cultured without feeder cells. VV3 ESCs were cultured on STO feeder cells inactivated with mitomycin C, whereas Ring1a/b-DKO and G9a-KO ESCs were cultured on inactivated mouse embryonic fibroblasts. In Max-null ESCs, both alleles of the Max gene are disrupted and Max cDNA was introduced into the ROSA26 locus under the control of a tetracycline-off system [4]. In Ring1a/b-DKO ESCs, both alleles of the Ring1a gene are disrupted and the Ring1b gene are floxed, and Ring1b could be conditionally deleted by 4-hydroxy tamoxifen (OHT) treatment [16]. For KO of Max and Ring1b, cells were treated for 3 days with doxycycline (Dox) (1 μg/ml) or 4-hydroxytamoxifen (4OHT) (800 nM), respectively.

Conventional bisulfite sequencing
Bisulfite sequencing analysis by Sanger sequencing was carried out as described previously [18]. Max-KD VV3 ESCs with a Vasa-Venus reporter or control ESCs (ESCs transfected with AllStars negative control siRNA) were cultured for approximately 72 h and sorted using an S3e cell sorter (Bio-Rad). Max-KD VV3 ESCs were purified based on Vasa::Venus-positivity, and control ESCs were purified based on Vasa::Venus-negativity. Genomic DNA was extracted from both cell types using a Qiagen DNeasy blood & tissue kit or Qiagen All-prep DNA/RNA micro kit and converted with sodium bisulfite using an EZ DNA methylationdirect kit (Zymo Research) according to the manufacturer's instructions. The targeted regions were amplified from bisulfite-converted DNAs using BIOTAQ HS DNA Polymerase (Bioline). The sequences of the PCR primers used for this assay are shown in S1 Table. The PCR products were cloned into respective pGEM-T easy vectors (Promega) and sequenced using a Big-Dye Terminator v1.1 cycle sequencing kit (Applied Biosystems).

Targeted methylome sequencing (TMS)
DNA preparation for TMS. DNA was isolated from sorted Max-KD VV3 ESCs and control ESCs using a DNeasy blood & tissue kit (Qiagen); 1 μg of DNA was dissolved in 130 μl of 10 mM Tris-HCl (pH 8.0) and sheared using an S220 focused ultrasonicator (Covaris) to yield 500-bp fragments. An AMPure XP system (Agilent Technologies) was used to purify the fragmented DNA as follows. Sheared DNA (130 μl) was mixed with 1.8 volumes (234 μl) of AM-Pure XP reagent and allowed to stand for 15 min at room temperature. The beads were collected using a magnetic stand, the supernatant was removed, and pelleted beads were rinsed with 70% ethanol and dried by incubation at 37˚C for 5 min. DNA was then eluted from the beads using 20 μl of RNase-free water. The eluted DNA was dried under vacuum and then dissolved in 7 μl of RNase-free water.
Target enrichment for TMS. A SureSelect Mouse Methyl-Seq kit (Agilent Technologies) was used for target enrichment by liquid-phase hybridization capture [19]. The probe set used in this study is designed by Agilent Technologies to comprehensively detect promoters, enhancers, and gene bodies. Genomic DNA (7 μl) fragmented and purified as described above was supplemented with 3 μl of formamide (biochemistry grade; Wako) and overlaid with 80 μl of mineral oil (Sigma-Aldrich). The DNA was then completely denatured by incubating at 99˚C for 10 min; the sample was then cooled to and maintained at 65˚C for at least 5 min before adding the following reagents. Hybridization buffer and capture probe mix were prepared according to the manufacturer's protocol, and they were each overlaid with 80 μl of mineral oil and incubated at 65˚C for 10 min. The two solutions were then combined and mixed thoroughly by pipetting. The combined solution was transferred to a tube containing the denatured input DNA (maintained at 65˚C as described above), and the solution was thoroughly mixed by pipetting. The sample was incubated at 65˚C for 24 h to allow for probe/target hybridization. A 50-μl volume of well-suspended DynaBeads MyOne streptavidin T1 solution (Life Technologies) was placed in a 1.5-ml tube, and the beads were washed twice with 200 μl of binding buffer. The hybridization reaction mixture, supplemented with 200 μl of binding buffer, was then added to the pelleted beads and thoroughly mixed. After incubation at room temperature for 30 min with agitation, the beads were collected using a magnetic stand and washed with 500 μl of wash buffer 1, subjected to three rounds of washing and resuspension in pre-warmed buffer 2, then incubated at 65˚C for 10 min. After removing the washing solution, the enriched DNA was eluted by incubating the beads in 20 μl of elution buffer at room temperature for 20 min. The eluate was immediately subjected to bisulfite treatment.
Bisulfite treatment for TMS. An EZ DNA methylation-gold kit (Zymo Research) was used for bisulfite treatment of target-enriched DNA according to the manufacturer's instructions. Enriched DNA solution (20 μl) was mixed with 130 μl of freshly prepared CT conversion reagent, and the mixture was incubated at 64˚C for 2.5 h. The 10-min incubation step at 98˚C was omitted because the target-enriched DNA was already denatured. After purification and desulfonation, bisulfite-treated DNA was eluted with 20 μl of M-elution buffer.
TMS library construction and illumina sequencing. We used bisulfite-treated DNA for library preparation according to the PBAT protocol [20] (also available from http://crest-ihec. jp/english/epigenome/index.html), except for use of the primers described below. The primer used for first-strand synthesis was 5'-biotin ACA CTC TTT CCC TAC ACG ACG CTC TTC CGA TCT WWW WNN NN-3' (W 1/4 A or T). The indexed primer used for secondstrand synthesis was 5'-CAA GCA GAA GAC GGC ATA CGA GAT XXX XXX GTA AAA CGA CGG CCA GCA GGA AAC AGC TAT GAC WWW WNN NN-3', where XXX XXX represents the index sequence of each primer. The constructed TMS libraries were sequenced as previously described [20][21][22][23] using an HiSeq2500 system (illumina).
TMS alignment and statistical analysis. TMS reads were aligned to the mouse genome (mm10; Genome Reference Consortium Mouse Build 38) using the Bismark tool (v.0.10.0; http://www.bioinformatics.babraham.ac.uk/projects/bismark/), with the following specific options: q n 2 -l 93 -pbat. MOABS module [24] was applied to detect differentially methylated regions (DMRs) from TMS reads of Max-KD VV3 ESCs and control ESCs (2 biological replicates). The MOABS pipeline calls DMR candidates by 3 different methods (M1, M2, and M3). In our case, M2 method identified the largest numbers of DMRs (no false positives). Since M2-called DMRs include almost M1-and M3-called DMRs, we used the M2-called DMRs for following analyses. In MOABS M2 method (according to the credible methylation difference metric), DMRs were defined as minC (minimum coverage of targeted regions) � 10, maxDist (maximum distance of DMRs from genes) 300 base pairs, cMethDif (credible Methylation Difference Cutoff) > 0.2. cMethDif > 0.2 means that only the regions where the DNA methylation levels in Max-KD ESCs show differences larger than 20% compared with those in control ESCs identify as DMRs. Motifs in DMRs were identified using the motif call tool, findMotifs. pl, from HOMER (http://homer.salk.edu/homer/motif/).

RNA preparation and real-time PCR
Total RNA isolated from cells was purified using an RNeasy Plus Mini kit (Qiagen) or RNeasy Micro kit (Qiagen) according to the manufacturer's instructions. RNAs were reverse-transcribed using SuperScript III (Invitrogen) and random primers (Promega). Gene expression was quantified using SYBR Green master mix (Applied Biosystems) with the primers shown in S1 Table. PCR signals were detected using CFX Connect (Bio-Rad). Arbp was used as an internal control.

Microarray analysis of L3mbtl2-KD ESCs
Microarray analyses were carried out as described previously [2]. VV3 ESCs were transfected with non-silencing negative control siRNA (AllStars) or siRNA against the L3mbtl2 gene. Vasa-positive cells were purified using fluorescence-activated cell sorting (FACS). Total RNA (100 ng) was isolated and purified using an RNeasy micro kit (Qiagen). The quality and quantity of purified total RNA were verified by Agilent 2100 Bioanalyzer (Agilent) and NanoDrop ND-1000 (Thermo Fischer Scientific), respectively. DNA microarray analysis was carried out according to manufacturer's instruction. In brief, cyanine3-labelled cRNA was obtained from 100 ng of purified total RNA using a Low Input Quick Amp Labeling kit (Agilent). The cRNA was purified, fragmented and then hybridized to an Agilent Whole Mouse Genome Oligo DNA Microarray kit, Ver 2.0 (Agilent) containing over 44,000 probes for mouse genes. Following hybridization at 65˚C for 17h, the arrays were washed and fluorescence signals were scanned using an Agilent DNA microarray scanner. Agilent Feature Extraction software was used to reduce the array images to the intensity of each probe (TXT files). Each cell type was analyzed in four biological replicates. All the microarray data have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (http:// www.ncbi.nlm.nih.gov/geo/, GEO Series accession number GSE102610).

Microarray analysis of G9a-KO ESCs
Total RNAs of G9a-KO ESCs and control ESCs (TT2) were isolated and purified using an RNeasy mini kit (Qiagen). The quality and quantity of purified total RNA were verified by agarose gel electrophoresis and spectrophotometry, respectively. DNA microarray analysis was carried out according to manufacturer's instruction. In brief, biotinylated cRNA was obtained from 200 ng of purified total RNA using a GeneChip 3' IVT Express Kit (Affymetrix). The cRNA was purified, fragmented and then hybridized to an Affymetrix Mouse Genome 430 2.0 Array containing over 45,000 probe sets for mouse genes. Following hybridization at 45˚C for 16h, the arrays were washed and labeled with phycoerythrin. Fluorescence signals were scanned using the Affymetrix GeneChip System. Affymetrix GeneChip Command Console software was used to reduce the array images to the intensity of each probe (CEL files). All the microarray data are MIAME compliant and have been deposited in a MIAME compliant database, the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/, GEO Series accession number GSE102423), as detailed on the FGED Society website (http://fged.org).

Transcriptome data analysis
GeneSpring (version 12.6, Tomy Digital Biology) was used for the identification of differentiallyexpressed genes (DEGs), statistical analysis, gene ontology (GO) analysis and description of Venn diagrams for microarray data. Seventy-five percentile shift was used to obtain normalized intensities for every feature on the array. We calculated the ratio of intensity in KO or KD samples to the intensity in the respective control samples as expression change. DEGs were determined as the genes in which the expression change of at least one probe is more than the expected fold change. Multiple testing corrections were performed using the Benjamini-Hochberg false-discovery rate correction. For the microarray data of Max-KD ESCs (GSE45181) [2] and L3mbtl2-KD ESCs obtained in our previous and this studies, respectively, gene expression profiles of Vasa-positive cells isolated from VV3 ESCs transfected with siRNAs for Max or L3mbtl2 were compared to those of VV3 ESCs transfected with non-silencing negative control siRNA (AllStars). For the microarray data of G9a-KO ESCs and Dnmts-TKO ESCs (GSE20177) [25] obtained in this study and by another group, respectively, gene expression profiles of wildtype (WT) and KO ESCs were compared. For the microarray data of Ring1a/b-DKO ESCs (GSE10573) [16], and Setdb1-KO ESCs (GSE28593) [26] obtained by other groups, gene expression profiles of conditional KO ESCs with OHT were compared with those of control ESCs without OHT. For RNA-seq analysis, RNA-seq datasets for Pcgf6-KO ESCs (GSE84480) [13], Dnmt1-cKO E13.5 PGCs (GSE74938) [27] and Setdb1-cKO E13.5 PGCs (GSE60377) [28] published by other groups were downloaded from GEO. The RNA-seq reads were aligned to the mouse reference genome (UCSC mm9 and RefSeq) using TopHat (ver. 2.0.8) [29]. Cufflinks (ver. 2.0.10) was used to estimate gene expression levels on the basis of fragments per kilobase of exon model per million mapped fragments [30]. For Pcgf6-KO ESCs, gene expression profiles of conditional KO ESCs with OHT were compared with those of control ESCs without OHT. For Dnmt1-cKO and Setdb1-cKO E13.5 PGCs, gene expression profiles of WT and KO PGCs were compared. The details of each sample can be accessed via these GEO accession numbers.

Immunoprecipitation and Western blotting
Anti-MAX antibody (Santa Cruz, sc-197x) (2 μg) were bound to Protein G Dynabeads (Invitrogen 10007D) in PBS for 2 h at 4˚C. After washing with Cross-linking buffer (50 mM phosphate buffer [pH 8.0], 20 mM triethanolamine [pH 8.0]), the precipitate was incubated in Cross-linking buffer with 5 mM Dimethyl pimelimidate dihydrochloride (SIGMA) for 30 min at room temperature. After washing with PBS, the precipitate was incubated in blocking buffer (50 mM phosphate buffer [pH 8.0], 10 mM triethanolamine [pH 8.0]) for 1 h at 4˚C. After washing with PBS and 0.1 M glycine (pH 3.5), the precipitate was suspended with PBS and used for immunoprecipitation. VV3 ESCs were harvested and suspended with Buffer A (10 mM Hepes-NaOH [pH 7.9], 10 mM KCl, 1.5 mM MgCl 2 , 1× cOmplete protease inhibitor cocktail [Roche]). The cell suspension was homogenized using a 120 Vac Overhead Stirrer (Wheaton), and the nuclear fraction was extracted by centrifugation. Collected nuclei were washed with Buffer A and re-suspended with Buffer B (20 mM Hepes-NaOH [pH 7.9], 350 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.1% NP-40, 10% glycerol, 0.5 mM DTT, 1× protease inhibitor). The nuclear suspension was centrifuged, and the supernatant was incubated with antibody-bound beads overnight at 4˚C. After three washes with Buffer B, the precipitate was eluted with 20 μl of 0.1 M glycine (pH 3.5). The elution was performed twice for the Western blotting using anti-DNMT1 and anti-DNMT3B antibodies (S4A Fig). The eluted proteins were used for Western blotting as described previously [2]. The antibodies used for these assays are listed in S2 Table. Anti-DNMT3L antibody were kindly provided from Dr. Keisuke Nimura [31].

Chromatin immunoprecipitation (ChIP) by cross-linking
For ChIP of MAX, RING1B, and SETDB1, ChIP-qPCR experiments were carried out as described previously [32], with some modifications. In brief, 1-5 μg of antibodies were bound to Dynabeads Protein G (Invitrogen) overnight at 4˚C. Cells were fixed using the ethylene glycol bis (succinimidyl succinate)/formaldehyde dual cross-linking method, as described previously [33]. The cross-linked cells were washed, collected in pellets by centrifugation, and flash frozen with liquid nitrogen. The cells were then lysed in SDS lysis buffer (1% SDS, 50 mM Tris-HCl [pH 8.0], 10 mM EDTA), and genomic DNA was sheared by ultrasonic fragmentation using a Bioruptor UCD-300 over 12 medium cycles (Cosmo Bio). After centrifugation, the cleared lysates were incubated with antibody-bound Dynabeads overnight at 4˚C. The beads were then washed, and chromatin was eluted using ChIP direct elution buffer (10 mM Tris-HCl [pH 8.0], 300 mM NaCl, 5 mM EDTA, 0.5% SDS). The eluted chromatin was subjected to reverse crosslinking with 10% SDS for 8 h at 65˚C. DNA was purified using a Qiagen PCR purification kit and analyzed via real-time PCR using Power SYBR Green PCR master mix (Applied Biosystems) and primers that spanned the TSSs of the genes of interest. The primer sequences and antibodies used for these assays are listed in S1 and S2 Tables, respectively.

ChIP using micrococcal nuclease (MNase)
For ChIP of H3K9me2 and H3K9me3, cells were fixed with 1.0% formaldehyde for 10 min at room temperature, after which glycine was added to the medium to a final concentration of 125 mM. The cells were then incubated in 1% NP-40, 50 mM Hepes/NaOH (pH 7.5), 10 mM KCl, 15 mM MgCl 2 , and 1× cOmplete protease inhibitor cocktail (Roche) for 15 min at 4˚C. After centrifugation, the cells were stored at -80˚C until analyzed. The thawed lysates were subjected to MNase treatment with MNase mixture (100U MNase [New England BioLabs], 15 mM Hepes/NaOH pH7.5, 60 mM KCl, 15 mM NaCl, 0.32 mM sucrose, 3 mM CaCl 2 , 1× cOmplete protease inhibitor cocktail [Roche]) for 20 min at 37˚C in order to obtain oligo-and mononucleosomes. Subsequent immunoprecipitation and detection processes were performed using the ChIP cross-linking method described above.

Immunoprecipitation of each fraction
Subsequent immunoprecipitation of each fraction was performed using the immunoprecipitation method described above with some modifications. The nuclear fractions were incubated with anti-MAX antibody-bound Protein G Dynabeads (Invitrogen) for 12 h in the presence of Benzonase nuclease (Novagen) at 4˚C. After washing three times with washing buffer (0.15 M NaCl, 0.1% NP-40, 50 mM HEPES [pH 7.4]), the antibody-Protein G beads were suspended in SDS-PAGE sample buffer. After boiling for 5min, the samples were resolved by SDS-PAGE and probed by western blotting with indicated antibodies.

Contribution of PCGF6-PRC1 to MAX-mediated gene repression
To identify additional epigenetic factors associated with MAX, but not included in PCGF6-PRC1, for repression of germ cell-related genes, we used Max-null ESCs for RT-qPCR and ChIP-qPCR analyses because we can easily obtain a large number of Max-depleted cells compared with Max-KD ESCs, and do not need feeder cells and transfection of siRNAs for the Max-null ESCs. In agreement with our previous results [3,4], Dox-induced Max KO in ESCs led to the reduced expression of MAX protein and significant upregulation of Ddx4 (also known as Mvh, mouse Vasa homologue), Dazl, Stra8, and Sycp3 (genes defined as the late PGC markers in this manuscript) expression (S1A and S1B Fig). We also confirmed that MAX was enriched in the transcription start sites (TSSs) of these genes and that enrichment declined dramatically following Dox treatment, whereas only slight enrichment was observed in the TSS of the hemoglobin-ß gene (Hbb-b1), which is not a target of MAX, with or without Dox (S1C Fig).
We then confirmed involvement of MAX in PCGF6-PRC1-dependent repression of germ cell-related genes, and evaluated the MAX dependency of the localization of RING1B (a catalytic subunit of PRC1) to germ cell-related genes. RING1B was enriched in the TSSs of the late PGC markers compared with the TSS of Hbb-b1, and Max KO decreased RING1B enrichment ( Fig 1A). We also confirmed that H3K9me2 These results were in agreement with our previous report [2,13] and indicate that MAX is required for the recruitment of PCGF6-PRC1 to its target genes.
To further assess contributions of PCGF6-PRC1 to the regulation of germ cell-related genes under the control of MAX (S3 Table), we compared expression change of these genes in Max-KD ESCs [2] and L3mbtl2-KD ESCs. L3mbtl2 disruption has been reported to abolish PCGF-PRC1-mediated gene repression [12]. We found limited up-regulation of germ cellrelated genes in L3mbtl2-KD ESCs compared to Max-KD ESCs (Fig 1B and 1C). Limited upregulation of the late PGC markers in Ring1a/b-DKO ESCs compared with Max-null ESCs was also confirmed by qRT-PCR (Figs 1D and S1B), as reported in the previous study [13]. The results suggest that MAX fully repress the expression of germ cell-related genes by PCGF6-PRC1-dependent and independent mechanisms. We also performed Gene Ontology (GO) analysis of up-regulated genes in Max-KD ESCs and several ESCs with functional deficiency of the PCGF6-PRC1 complex, including Ring1a/b-DKO ESCs [16], G9a-KO ESCs [17], L3mbtl2-KD ESCs, and Pcgf6-KO ESCs [13]. We found germ cell-related genes and/or germcell related GO terms in genes up-regulated only in Max-KD ESCs as well as in genes commonly upregulated in Max-KD ESCs and in those with KD or KO of other factors containing PCGF6-PRC1 (S2C-S2F Fig). The results suggest that PCGF6-PRC1 repress only a subset of germ cell-related genes under the control of MAX.
PCGF6-PRC1 also contains the histone deacetylases HDAC1 and HDAC2 (HDAC1/2) [9], and we confirmed interaction between MAX and HDAC1 using a co-immunoprecipitation assay (S3A and S3B Fig). However, the late PGC markers were not up-regulated, and only a few germ cell-related genes were up-regulated in Hdac1/2-DKD ESCs (S3C-S3E Fig), suggesting that contribution of HDAC1/2 on repression of germ cell-related genes in PCGF6-PRC1 is limited. Collectively, these results suggest that MAX-containing PCGF6-PRC1 partially represses subset of germ cell-related genes, but additional MAX-interacting proteins could play a role in further repression of these genes.

MAX-mediated repression of germ cell-related genes through DNA methylation
Since DNMT3B and DNMT1 are involved in repression of germ cell-related genes in PGCs [27,34,35] and we recently identified DNMT3A and DNMT3L as MAX-interacting proteins in ESCs by mass spectrometry (data not shown), we first focused on DNMTs as additional candidate co-repressors associated with MAX. We confirmed co-immunoprecipitation of DNMT1, DNMT3A, DNMT3B, and DNMT3L with MAX in ESCs (Figs 2A and S4A). To elucidate the significance of functional interactions between MAX and DNMTs on a genome-wide level, we performed targeted methylome sequencing (TMS) [19,36] for semi-comprehensive DNA methylome analysis of Max-KD ESCs. For consistency with the transcriptome data in the previous study [2], we performed methylome analysis using VV3 ESCs with or without Max-KD in this study. A differentially methylated region (DMR) calling identified 17 genes which had hypomethylated DMRs in close proximity (±300 bp) compared with control ESCs (Fig 2B and  S4 Table). Notably, germ cell-related GO terms were enriched in these 17 genes (Fig 2C), and motif analyses revealed that the DMRs in Max-KD ESCs frequently contain E-box sequences (CACGTG), a binding motif of MAX (Fig 2D). Bisulfite sequence analysis by conventional Sanger sequencing of the late PGC markers in Max-KD ESCs confirmed that the levels of TSS DNA methylation were clearly lower than in control ESCs (Fig 2E and 2F). These results demonstrate that MAX associates with DNMTs and contributes to the maintenance of DNA methylation and/or de novo methylation of germ cell-related genes.
To verify the contribution of DNA methylation to the repression of germ cell-related genes, existing microarray data for Max-KD ESCs and Dnmts-TKO ESCs [15,25] were re-analyzed. We found that 266 genes, including Dazl and Stra8, were commonly up-regulated by  Table). These results indicate that MAX represses some germ cell-related genes through DNA methylation.

MAX-mediated repression of germ cell-related genes through H3K9me3
Since Sycp3 was not significantly up-regulated in any ES cell lines examined except Max-KD or Max-null ESCs (Figs 1C, 1D and S4C) [2], Sycp3 may be repressed by a mechanism that does not dependent on PCGF6-PRC1 nor DNMTs, but dependent on MAX. Thus, we hypothesized that additional factors associate with MAX to cooperatively repress germ cell-related genes, including Sycp3. Using previously reported ChIP-seq data for ESCs, we found that SETDB1 (a histone methyltransferase that catalyzes H3K9me3) localized in the TSSs of the late PGC markers, including Sycp3 (S5A Fig) [37]. Data in a previous study also showed that some germ cell-related genes were targets of SETDB1 and H3K9me3 in ESCs [38].
To test the possible cooperative role of MAX and SETDB1 in repressing germ cell-related genes, we first examined the interaction between MAX and SETDB1 using a co-immunoprecipitation assay (Figs 3A and S5B). We then examined MAX dependency of SETDB1 localization in the late PGC markers. The levels of H3K9me3 and SETDB1 in the TSSs of the late PGC markers (except Stra8) were decreased in Max-null ESCs compared with control ESCs, whereas levels of H3K9me3 and SETDB1 in the TSS of Hbb-b1 did not significantly differ from controls (Fig 3B and 3C).
To verify whether the decline in H3K9me3 levels causes up-regulation of germ cell-related genes, previously reported microarray data for Setdb1-KO ESCs [26] were re-analyzed. We found that 238 genes, including Vasa, Dazl, and Sycp3, were commonly up-regulated by Max-KD and Setdb1-KO, and germ cell-related GO terms were enriched in these genes (S5C Fig). In addition, 1,274 genes, including Stra8, were up-regulated by Max-KD alone, and germ cellrelated GO terms were also enriched in these genes. By contrast, germ cell-related GO terms were not enriched in genes up-regulated by Setdb1-KO alone. The results suggest that MAX cooperatively represses a subset of germ cell-related genes with SETDB1. Furthermore, Setdb1-KD ESCs showed higher expression of all of the late PGC markers including Sycp3 than control ESCs as in Max-null ESCs (S1B Fig), and Dazl was particularly up-regulated compared with other genes (Fig 3D and 3E). The results together suggest that MAX-dependent recruitment of SETDB1 catalyzes H3K9me3 in a subset of germ cell-related genes, resulting in their repression.

Relationship between DNA methylation, H3K9me3, and PCGF6-PRC1 in repression of germ cell-related genes
Although DNMTs and SETDB1 have not been identified as components of PCGF6-PRC1 [12,13,39], we investigated whether or not DNA methylation and H3K9me3 are regulated by PCGF6-PRC1. To this end, we determined the levels of DNA methylation and H3K9me3 in the TSSs of the late PGC markers in Ring1a/b-DKO ESCs. Bisulfite sequence analyses revealed that Ring1a/b-DKO did not decrease the level of DNA methylation in the Stra8 and Sycp3 TSSs (Fig 4A) or the level of H3K9me3 in the TSSs of the late PGC markers (Fig 4B). These data suggest that the regulation of DNA methylation and H3K9me3 on the late PGC markers are independent of PCGF6-PRC1.
We also investigated to what extent DNMTs and SETDB1 contribute to the repression of germ cell-related genes through MAX-mediated pathways. Max-KD in Dnmts-TKO ESCs ( Fig  4C and 4D), as well as Max-KO in Setdb1-KD ESCs (Fig 4E and 4F), remarkably enhanced the expression of the late PGC markers. These results suggest that DNMTs and SETDB1 are partially contribute for the repression of germ cell-related genes mediated by MAX, and complete repression of these genes are achieved through multiple epigenetic function based on MAX.
We further investigated the relationship between DNA methylation and H3K9me3. Setdb1 KD in Dnmts-TKO ESCs caused additional up-regulation of the late PGC markers, especially Dazl and Stra8, compared with Dnmts-TKO ESCs exposed to control siRNA (Fig 4G and 4H). The expression of Sycp3 was clearly decreased in Dnmts-TKO ESCs compared with control ESCs, indicating that DNA methylation resulted in transcriptional activation for Sycp3 as is the case with a previous report [40]. These results suggest that DNMTs and SETDB1 function through distinct pathways for the repression of germ cell-related genes. Taken

Fractionation of MAX-interacting complexes
The abovementioned results suggest that MAX functions in the repression of germ cell-related genes by forming multiple complexes with various epigenetic factors in addition to PCGF6-PRC1. To further clarify this possibility, we sequentially fractionated ESC nuclear extracts using heparin sepharose and Q sepharose columns. Each fraction collected was subjected to immunoprecipitation using an anti-MAX antibody with subsequent Western blotting (Fig 5A). We observed coimmunoprecipitation of MAX, DNMT3A and DNMT3L with or without RING1B in some fractions (Figs 5B-5E and S6A-S6D), suggesting the existence of both PCGF6-PRC1-associated andunassociated MAX-DNMT complexes. In D-III fraction, the MAX signal was strongest but the signals of RING1B, DNMT3A and DNMT3L were comparable to other fractions. It suggests that D-III fraction contains multiple MAX-mediated complexes including PCGF6-PRC1. These results suggest that MAX, DNMT3A, and/or DNMT3L could regulate DNA methylation independent from PCGF6-PRC1 in ESCs.

Discussion
In this study, we showed that MAX represses germ cell-related genes in ESCs via interaction with the epigenetic regulators DNMTs and SETDB1 in addition to PCGF6-PRC1. The target genes of  MAX/L3MBTL2-containing PCGF6-PRC1, MAX-DNMT1, and MAX-SETDB1 in ESCs may partially overlap, which suggests that those complexes redundantly repress some of their target genes, but many genes upregulated by Max-KD were affected either by L3mbtl2-KD, G9a-KO, Dnmts-TKO, or Setdb1-KO alone (S7A Fig). As PRCs are required for the establishment of H3K27me3 and subsequent silencing of developmental genes in bivalent domains involving both a repressive modification (H3K27me3) and a permissive modification (H3K4me3) [41], we examined the overlap between bivalent genes and genes regulated by MAX (S7B Fig). A majority of bivalent genes in ESCs [42,43] did not overlap with genes up-regulated by Max-KD. We further revealed the presence of complexes composed of MAX and DNMTs without RING1B in ESCs (Fig 5B and 5C). Collectively, these data suggest that MAX is involved in multiple regulatory mechanisms that differ from that involving PRC1, according to the target genes. Furthermore, MAX-DNMTs and MAX-SETDB1 may function, at least to some degree, as part of distinct pathways to repress their target genes, because Dnmts-TKO and Setdb1-KD additively repressed the expression of the late germ cell markers (Fig 4G and 4H). It was also reported that Dnmts-TKO ESCs exhibit minimal changes in genome-wide H3K9me3 occupancy compared with wild-type ESCs, supporting the hypothesis that DNA methylation and H3K9me3 act non-redundantly in ESCs [44]. An important future task would be to identify MAX-containing complexes other than PCGF6-PRC1 that repress germ cell-related genes.
We summarized the quantitative data for RT-qPCR (fold changes), ChIP-qPCR (fold changes of localization), and bisulfite sequence (Δ %CpGme) compared with each control condition obtained in this study (S5 Table).  (Fig 4H and S5 Table). Since these fold changes were higher than those in Max-null ESCs (Dazl: 32 fold, Stra8: 13 fold) (S1B Fig  and S5 Table), DNMTs and /or SETDB1 may have additional roles for the repression of Dazl and Stra8 other than MAX-containing complexes.
Sycp3 showed decrease of H3K9me3 (0.205 fold) and SETDB1 localization (0.407 fold) by Max-KO and its expression was up-regulated by Setdb1-KD (2.48 fold). Though Sycp3 also showed considerable decrease of DNA methylation (-21.7%) by Max-KD (Fig 2E and 2F), Dnmts-TKO rather down-regulated Sycp3 expression (0.26 fold, Fig 4H). The results suggest that DNMTs and DNA methylation do not result in transcriptional repression on Sycp3, and SETDB1 and H3K9me3 make a major contribution on its repression. In addition, Sycp3 was not up-regulated by Ring1a/b-DKO (0.77 fold) and by GLP-KO (0.87 fold), and therefore PCGF6-PRC1 may not play a role on repression of Sycp3. Although these four genes are all categorized as the late PGC markers, the regulatory mechanisms and responsible modifications seem quite different one another.
Although Dnmts-TKO ESCs exhibited robust growth and maintained their undifferentiated characteristics [15], DNMTs reportedly play several important roles in ESCs. DNMT1 and DNMT3A/3B suppress long terminal repeats and long interspersed elements of retrotransposons, respectively, possibly through interaction with UHRF1 [45]. Furthermore, DNA methylation in imprinting control regions (ICRs) by DNMT3A/3B is stably maintained via the interaction between G9a and GLP [46]. In this study, we demonstrated interactions between DNMTs and MAX, and preferential enrichment of DMRs at the promoters of meiotic genes in ESCs after Max-KD (Fig 2C), suggesting additional roles for DNMTs in the repression of meiotic programming in association with MAX as discussed below.
SETDB1 also plays a variety of roles in ESCs. Several studies have reported that SETDB1 and ATF7IP form a complex and that both are required for proviral silencing in mouse ESCs, especially for class I and II endogenous retroviruses (ERVs) [44,47,48]. We examined whether MAX is also involved in the regulation of ERVs silenced by SETDB1 and ATF7IP. The expression of MLV and IAP Ez (class I and II ERVs, respectively) and a class III ERV MaLR was not affected by Max-KD, whereas the expression of another class III ERV, MERVL, was significantly higher in Max-KD ESCs than in control ESCs (S8 Fig). These results suggest that MAX does not function in concert with the ATF7IP-SETDB1 complex with regard to proviral silencing. It has been reported that a transcription regulator, KAP1/TRIM28, is enriched in MaLR and MERVL at a similar level, but only MERVL is upregulated in Kap1-KO ESCs [49]. It suggests that KAP1/TRIM28 represses MERVL, but not MaLR in class III ERV. Since we have found the interaction between MAX and KAP1 in ESCs by the immunoprecipitation of FLAG-tagged MAX protein (unpublished data), MAX may exert silencing particular ERVs including MERVL with KAP1, but not with SETDB1.
As described above, MAX is associated with several epigenetic mechanisms that lead to repression of germ cell-related genes in ESCs, but whether the mechanisms in ESCs are also functional in PGCs is unclear. DNMT1 preserves DNA methylation of ICRs and meiotic gene promoters in PGCs, and conditional deletion of Dnmt1 in PGCs was shown to cause a decrease in the number of germ cells and their precocious differentiation, including up-regulation of spermatogenesis-and meiosis-related genes in male and female germ cells, respectively, consequently leading to hypogonadism and infertility [27]. Setdb1 KO E13.5 PGCs exhibit derepression of many ERVs, a reduced number of male PGCs, and postnatal hypogonadism [28]. We re-analyzed existing data to estimate whether germ cell-related genes repressed by MAX in ESCs are also repressed by DNMTs or SETDB1 in PGCs. Eighty-five genes, defined as germ cell-related according to the GO term "reproduction", were found to be up-regulated by Max-KD in ESCs (S3 Table). Among those, up-regulated genes in Dnmt1-cKO or Setdb1-cKO PGCs were extracted as genes with > 1.3-fold change compared with control PGCs. 52 and 55 genes were found to be up-regulated by Dnmt1-cKO and Setdb1-cKO in female E13.5 PGCs, respectively, whereas 32 and 41 genes were found to be up-regulated by Dnmt1-cKO and Setdb1-cKO in male E13.5 PGCs (S9 Fig). Considering that decreases in MAX levels induce meiosis-like cytologic changes in cultured germline stem cells [3], MAX, DNMTs, and SETDB1 could repress meiosis-related genes in PGCs in a similar manner as in ESCs. Future research should focus on determining whether MAX, DNMTs, and/or SETDB1 function in concert to control initiation of meiosis in PGCs.  [37] were re-analyzed using Integrative Genomics Viewer (IGV). Neighboring regions of TSSs of the late PGC markers are shown. (B) Un-cropped data of Western blotting corresponding to Fig 3A. Immunoprecipitated samples by anti-MAX antibody or control IgG were subjected to Western blotting by using anti-SETDB1 antibody. Red indicates the data shown in Fig 3A. (C) Venn diagram of genes up-regulated in Max-KD ESCs (GSE45181) [2] and Setdb1-KO ESCs (GSE28593) [26] (Max-KD ESCs; n = 3, > 2-fold change, one-way ANOVA P < 0.05, Setdb1-KO ESCs; n = 3, > 1.3-fold change, one-way ANOVA P < 0.05). GO analyses of genes representing each category were performed. GO terms with the lowest corrected P value (top 7) are shown.  Table) were selected and expression change of these genes in E13.5 Setdb1 cKO PGCs (GSE60377) [28] (n = 2) (A) or Dnmt1 cKO PGCs (GSE74938) [27] (n = 3) (B) compared with control PGCs were represented as heat maps. (TIF) S1  Table. List of germ cell-related genes up-regulated in Max-KD ESCs compared with control ESCs. Eighty-five of germ cell-related genes with gene ontology term "reproduction" were extracted as up-regulated genes in Max-KD ESCs compared with control ESCs (GSE45181) [2] (n = 3, > 2-fold change, one-way analysis of variance [ANOVA] P < 0.05). (TIF) S4 Table. List of genes with differentially methylated region (DMR). (TIF) S5 Table. Summaries of qPCR, ChIP and bisulfite sequence in this study. Red and orange indicate > 5 fold and > 2 fold up-regulated genes in RT-qPCR, respectively. (TIF)