The histone methyltransferase SETD2 is required for expression of acrosin-binding protein 1 and protamines and essential for spermiogenesis in mice

Spermatogenesis is precisely controlled by complex gene expression programs and involves epigenetic reprogramming, including histone modification and DNA methylation. SET domain–containing 2 (SETD2) is the predominant histone methyltransferase catalyzing the trimethylation of histone H3 lysine 36 (H3K36me3) and plays key roles in embryonic stem cell differentiation and somatic cell development. However, its role in male germ cell development remains elusive. Here, we demonstrate an essential role of Setd2 for spermiogenesis, the final stage of spermatogenesis. Using RNA-seq, we found that, in postnatal mouse testes, Setd2 mRNA levels dramatically increase in 14-day-old mice. Using a germ cell–specific Setd2 knockout mouse model, we also found that targeted Setd2 knockout in germ cells causes aberrant spermiogenesis with acrosomal malformation before step 8 of the round-spermatid stage, resulting in complete infertility. Furthermore, we noted that the Setd2 deficiency results in complete loss of H3K36me3 and significantly decreases expression of thousands of genes, including those encoding acrosin-binding protein 1 (Acrbp1) and protamines, required for spermatogenesis. Our findings thus reveal a previously unappreciated role of the SETD2-dependent H3K36me3 modification in spermiogenesis and provide clues to the molecular mechanisms in epigenetic disorders underlying male infertility.

is a well-characterized feature of transcriptionally active genes and usually enriched in the gene body regions. H3K36me3 is also implicated in pre-mRNA splicing, DNA mismatch repair, and recruitment of Dnmt3b. SET domain-containing 2 (SETD2; also known as HYPB and KMT3A) is the major methyltransferase responsible for H3K36me3 in mammals (29 -34). Setd2 ablation in mice leads to global loss of H3K36me3 and embryonic vascular remodeling defects, indicating a critical role of SETD2-mediated H3K36me3 during development (32, 34 -36). Although the functions of Setd2 and its associated H3K36me3 have been studied in diverse biological processes, the requirement for Setd2 and H3K36me3 during spermatogenesis remains elusive.
In this report, we show the spatiotemporal distribution of SETD2 and its associated H3K36me3 in mouse testes. We demonstrate that loss of H3K36me3 by germ cell-specific disruption of Setd2 leads to impaired spermiogenesis due to significant down-regulation of its direct target genes, which include many critical master regulators for spermiogenesis. This study thus reveals a crucial role of Setd2 and its associated H3K36me3 in spermiogenesis and uncovers Acrbp1 and protamine genes as direct targets of SETD2-dependent H3K36me3 during this process.

Spatial and temporal patterns of Setd2 and H3K36me3 distribution during spermatogenesis
To explore the role of Setd2 and the associated H3K36me3 during spermatogenesis, we first examined the expression pattern of Sedt2 in mouse testes. In postnatal mouse testes, the Setd2 mRNA level dramatically increased at 14 days old, a stage that corresponds to the appearance of pachytene spermatocytes in the seminiferous epithelium (Fig. S1A), and remained at a high level afterward, suggesting that Setd2 is abundantly expressed in pachytene spermatocytes and spermatids. We then performed immunofluorescence staining to further investigate the subcellular localization of the SETD2 protein in the testis. Consistent with the mRNA expression, we found that SETD2 protein is mainly detected in the nucleus of pachytene spermatocytes and round spermatids, which were confirmed by the counterstaining of peanut agglutinin (PNA), a marker for the acrosome, and ␥H2AX, a marker for the XY body, respectively ( Fig. 1A and Fig. S1B) (37). Moreover, SETD2 is also present in Sertoli cells as confirmed by counterstaining of GATA4, a marker for Sertoli cells (and round spermatids) (38) (Fig.  S1C). Collectively, Setd2 is predominantly expressed in pachytene spermatocytes and postmeiotic germ cells in the adult mouse testes.
As SETD2 is considered the main histone H3K36me3 methyltransferase in mammals, we next analyzed the distribution of H3K36me3 in mouse testis. Interestingly, we found that H3K36me3 distribution did not show a pattern identical to that of Setd2 in the testes. As shown in Fig. 1B, H3K36me3 exhibited strong staining in the nucleus of all stages of spermatocytes and the nucleus of round spermatids, whereas weak H3K36me3 signal was detected in spermatogonia and elongated spermatids. Taken together, these observations indicated that SETD2 might be responsible for cellular H3K36me3 in late-stage spermatocytes (pachytene/diplotene spermatocytes) and spermatids. However, whether it is predominantly responsible for H3K36me3 in the early-stage spermatocytes, such as preleptotene spermatocytes, remains to be determined in the future.

Setd2 is essential for murine spermiogenesis
To determine the role of Setd2 during mammalian spermatogenesis, we used a conditional knockout approach that generated a Setd2 floxed line (Setd2 fl/fl ) in which exons 6 and 7 of the Setd2 allele are flanked by loxP sites (Fig. S2A). To generate germline-specific deletion, mice bearing loxP-flanked alleles of Setd2 were crossed with Stra8-GFPCre knockin mice (Setd2 fl/del , Stra8-GFPCre) (Fig. S2A). The Stra8-GFPCre knockin mice express CRE from A1 spermatogonia onward in which both alleles of Setd2 could be excised at A1 spermatogonia (Setd2 fl/del , Stra8-GFPCre) (10). Thus, we designate the two different genotypes of mice as control (Setd2 fl/del ) and Setd2 cKO (Setd2 fl/del , Stra8-GFPCre). Immunostaining verified that SETD2 protein was absent in the mutant germ cells (Fig. S2B). Notably, in Setd2 cKO mice, H3K36me3 signals were also completely absent in pachytene spermatocytes and round spermatids, whereas signal remained in preleptotene spermatocytes. These results further confirmed SETD2 as the major H3K36me3 methyltransferase in late-stage spermatocytes and round spermatids but not in early-stage spermatocytes (Fig. S2C). We also found that the levels of H3K9me2, H3K9me3, H3K27me3, H3K79me3, and H3K27Ac were not affected in the mutants, indicating that Setd2 loss did not cause global changes of other histone modifications (Fig. S2D). 4

SETD2 is crucial for spermatogenesis
Although the Setd2 cKO males copulate normally, they were completely sterile. Testes from adult Setd2 cKO mice were significantly smaller than those of the littermate controls (Fig. 2, A and B; age, 8 weeks; control, 0.3867 Ϯ 0.0746 g; Setd2 cKO mice, 0.2425 Ϯ 0.0416 g; p Ͻ 0.001, n ϭ 8). Histological examinations of adult germ cell mutant testes show that the seminiferous tubules are devoid of mature spermatozoa and contain exclusively round-headed spermatids. Moreover, round spermatids in Setd2 cKO mice did not undergo cellular elongation, indicating a developmental arrest before step 8 spermatids (Fig. 2C). Consistent with this, no sperm was found in the epididymis of germ cell mutants (Fig. 2C). These results suggest that Setd2 plays an essential role in spermatogenesis, and its loss leads to developmental arrests at round-spermatid stage.
Close examination of the mutant seminiferous tubules showed a high frequency of degenerated germ cells in the spermatocyte and spermatid layers. TUNEL staining further identified a dramatic increase of apoptotic spermatocytes and spermatids in the mutant testes (Fig. 2, D and E). To distinguish whether the defects in spermatocytes resulted from an intrinsic requirement of Sedt2 function or secondary effects of impaired spermiogenesis, we examined the first occurrence of the apoptotic germ cells in Setd2 cKO mouse testes. At 2 weeks old, when the pachytene spermatocytes start to occur in the testis, we found that there was no significant difference between Setd2 cKO and control mouse testes (Fig. S3, A and C). In 4-week-old mice, when the first population of spermatids reaches the elongated spermatid in the control testes, elevated apoptosis in germ cells was observed in Setd2 cKO mouse testes (Fig. S3, B and C). Similar apoptotic spermatocyte defects are also observed in Crem, Trf2, Miwi, Tdrd5, and Mrg15 mutants, although these mutants show impaired spermatid differentiation (39 -45). Taken together, these findings suggest that, in Setd2 cKO mouse testes, spermatocyte defects resulted from the spermiogenic arrest.
To verify the requirement for Setd2 in spermiogenesis but not in the early stages of spermatogenesis, we conditionally deleted Setd2 allele in germ cells as early as embryonic day 15 using the Vasa-Cre transgenic line (Setd2 fl/del , Vasa-Cre). Inactivation of the Setd2 in gonocytes resulted in a similar phenotype, and the loss of H3K36me3 was only observed in pachytene spermatocytes and spermatids as shown in Setd2 fl/del , Stra8-GFPCre mice (Fig. S4, A-C). Thus, these results further demonstrated an essential role for Setd2 and its associated H3K36me3 in spermiogenesis.

Loss of Setd2 results in malformation of acrosome
To elucidate the nature of the defects in spermiogenesis of Setd2 cKO mice, we performed immunostaining using FITCconjugated PNA, capable of binding the outer acrosomal membrane, and anti-Golgi 58K antibody, a marker for the proacro-

SETD2 is crucial for spermatogenesis
some in spermatids (46). Compared with a single acrosomal vesicle in the Golgi phase at steps 2 and 3 of spermiogenesis in controls (Fig. 3A), we found several acromsomal vesicle-like structures in Golgi-phase spermatids of Setd2 cKO mice (Fig.  3B), indicating that disruption of Setd2 could lead to acrosomal malformation as early as in the Golgi phase during spermiogenesis. In the cap phase at steps 4 -7 of spermiogenesis, the acrosome expanded to form a caplike structure covering the nucleus of the round spermatids (Fig. 3A). However, in Setd2 cKO mouse spermatids, the caplike acrosome failed to form, resulting in a discontinuous acrosomal structure in the anterior region of the nucleus (Fig. 3B).
To characterize the defective acrosome biogenesis in spermiogenesis of Setd2 cKO mice in further detail, we performed a transmission EM (TEM) study (Fig. 4). Consistent with immunofluorescence analyses, the first acrosomal defects were detected in the Golgi-phase spermatids in which several acrosomal structures appeared in Setd2 cKO mouse round spermatids, whereas a single uniformed acrosomal structure was present in control spermatids (Fig. 4, A and B). Compared with control spermatids, acrosomal cap structures were discontinuous and abnormally fragmented in Setd2 cKO mouse spermatids at the cap phase (Fig. 4, C and D). It is of note that the fragmented acrosome persisted in all subsequent stages of spermiogenesis in Setd2 cKO mutants. Taken together, these observations reveal that Setd2 plays crucial roles in acrosome biogenesis and is required for spermiogenesis.

Setd2 inactivation led to complete loss of H3K36me3 in pachytene spermatocytes and round spermatids
To better gain mechanistic insights into the spermiogenesis defects caused by Setd2 loss, we set out to determine the genome-wide distribution of H3K36me3 by chromatin immunoprecipitation (ChIP)-seq in the pachytene/diplotene sper-matocytes (PDs) and round spermatids (RSs) from the control and mutant mice. Our H3K36me3 ChIP-seq analyses identified 28,270 and 26,109 peaks, which cover 10,027 and 10,164 genes in PD and RS cells, respectively (Fig. 5A). In both PD and RS cells, we found that H3K36me3 is mainly enriched in the gene bodies, especially the exonic, intronic, and 3Ј-UTR regions, in agreement with the canonical understanding of this histone mark (Fig. 5B). Consistent with our hypothesis of SETD2 being the major H3K36me3 methyltransferase in the late stages of spermatogenesis, we found that the mutant cells showed a drastic loss of H3K36me3 (Fig. 5C, normalized by spike-in approach; see "Experimental procedures"). The ChIP-seq results were also confirmed by ChIP-quantitative PCR (ChIP-qPCR) of a few select targets (Fig. 5D).

Setd2 loss led to misregulation of critical regulators in spermiogenesis
To further investigate the expression changes caused by Setd2 loss, we profiled the transcriptomes of the PD and RS from control and Setd2 cKO mouse testes, respectively. We found that Setd2 loss led to a total of 2,48 and 1,546 transcripts and of 1,088 and 1,145 transcripts significantly (false discovery rate Ͻ0.05, Ͼ2-fold difference) down-and up-regulated in PD and RS cells, respectively ( Fig. 6A and Fig. S5A). Further analyses found that the down-regulated genes are generally marked with higher levels of H3K36me3 in contrast to the up-regulated genes, suggesting that the down-regulated expression is likely the direct effect of Setd2 loss (Fig. 6B). Consistent with the higher H3K36me3 level, the down-regulated genes were also expressed at higher levels than the up-regulated genes in the

SETD2 is crucial for spermatogenesis
wide-type cells (Fig. 6C). Gene ontology (GO) analysis of the top-ranked differential genes indicated enrichment in genes associated with system development and motility for up-regulated genes and cell-cell signaling and cytoskeleton for downregulated genes in RS, respectively (Fig. S5B), suggesting that the developmental defects caused by Setd2 loss might be due to the compromised functions of these pathways in RS.
It is of note that the expressions of many critical regulators for spermiogenesis, such as Acrbp1, Pick1, Odf1, and Spem1, were also significantly impaired upon Setd2 loss (Fig. 6D and Fig. S5C), which was also confirmed by RT-qPCR (Fig. 6E) (47)(48)(49)(50). Among these genes, both Acrbp1 and Pick1 are known to be required for acrosome development, and knockout of either Acrbp1 or Pick1 led to fragmentation of the acrosome, which is similar to that seen in Setd2 mutants. Moreover, loss of Spem1 resulted in sperm deformation. Thus, the malformation of the acrosome caused by Setd2 loss could be due to the misregulation of these genes, which are likely SETD2 direct targets.

Impaired activation of transition nuclear protein and protamine genes in Setd2 cKO mutants
Among the deregulated genes identified in Setd2 cKO mice, we also noticed aberrant expressions of many replication-dependent histone genes (Fig. 6F). It is well-known that, during late stages of sperm differentiation, the majority of the core histones are subsequently replaced by transition nuclear proteins (TNPs) and protamines (PRMs), leading to the formation of highly condensed chromatin in elongated spermatids (51). Either Tnp-or Prm-deficient mice are infertile due to disruption of chromatin condensation in elongating spermatids (52,53). Such processes are also accompanied with a coordinated and gradual transcriptional silencing of the canonical replication-dependent histone genes and activation of the germline-specific histone and Tnp genes. Consistent with the spermatid developmental defects of the Setd2 cKO mice, we identified significant down-regulation of Tnps, Prm1/2/3, and several testis-specific histone variant genes, such as H1fnt, H2afb1, and H2afv, accompanied with upregulation of many replication-dependent histone genes (Fig. 6F) in both isolated pachytene spermatocyte and purified round spermatids. These observations were further validated by RT-qPCR (Fig. 6E). These findings indicate that the transition from histoneto-protamine replacement could also be compromised upon Setd2 loss, thereby contributing to the defective spermiogenesis.

Discussion
SETD2 is the major methyltransferase of histone H3K36me3, a mark associated with active transcription. Here, we show that SETD2 is the primary H3K36me3 methyltransferase in PD and

SETD2 is crucial for spermatogenesis
RS and is essential for spermiogenesis. Loss of Setd2 led to complete loss of H3K36me3 in PD and RS cells and stalled development at round-spermatid stage, thereby resulting in sterility. Mechanistically, we found that the activation of several critical master regulators for spermiogenesis is impaired in Setd2 cKO germ cells accompanied with compromised expression of motility-related genes. Interestingly, we also found that the expressions of many histone genes and Tnp as well as protamine genes were significantly affected, indicating an impaired histone-to-protamine transition during spermiogenesis. Taken together, we have revealed an essential role of Setd2 during spermiogenesis but not early stages of spermatogenesis.
SETD2-mediated H3K36me3 is a transcriptionally coupled histone methylation mark, enriched over gene bodies undergoing active transcription (31). Although the enzymology and molecular function of SETD2 have been well-documented, the biological consequences of Sedt2 inactivation in animals and at the organism level have not been well-explored. To understand its biological significance during spermatogenesis, we generated a Setd2 cKO model in this study. Consistent with previously reports, we found that H3K36me3 patterns in PD and RS are also enriched in gene bodies in PD and RS cells. Setd2 loss leads to expression changes of hundreds to thousands of genes in PD and RS cells ( Fig. 6A; false discovery rate Ͻ0.05, Ͼ2-fold difference). However, because H3K36me3 is a mark associated with active transcription, we were surprised to find fewer down-regulated genes than up-regulated genes in both PD and RS cells from Setd2 cKO mice. Further analyses found that the down-regulated genes are generally marked by higher levels of H3K36me3 in the WT germ cells (Fig. 6, B and C) and are expressed at higher levels. These observations not only suggest that the down-regulated expression is likely the direct effect of Setd2 loss but also indicate that Setd2 and H3K36me3 are required for transcriptional activities of more highly expressed genes, which may play important roles in spermiogenesis. In contrast, the up-regulated genes are generally expressed at low levels, and the elevated expression may be due to the compromised distribution of transcriptional machineries upon global loss of H3K36me3. Indeed, we observed significant impaired expression of critical spermiogenesis regulators, including Acrbp1, Pick1, Odf1, Spem1, Tnp, and Prm genes, and misexpression of histone genes, which normally undergo transcriptional silencing at these stages. For example, loss of either Acrbp1 or Pick1 resulted in acrosomal fragmentation similar to

SETD2 is crucial for spermatogenesis
that observed in Setd2 mutants. This suggests that malformation of the acrosome in Setd2 mutants results from decreased expression of Acrbp1 and Pick1. Moreover, down-regulation of Tnp and Prm genes could disrupt histone-to-protamine transition, thereby causing the spermatid chromatin condensation defects. However, we cannot directly observe the spermatid chromatin condensation phenotype because arrest of spermiogenesis in Setd2 mutants occurred long before spermatid chromatin condensation. Thus, dysregulation of multiple pathways might together contribute to the spermatid developmental defects in Setd2 cKO mice.
Notably, at global levels, we found that SETD2 and H3K36me3 patterns displayed dissimilarities during spermatogenesis. The highest level of SETD2 was observed in pachytene spermatocytes and postmeiotic germ cells in the adult mouse testes, whereas strong H3K36me3 levels are maintained throughout all stages of spermatocytes and round spermatids with spermatogonia and elongated spermatids showing weaker H3K36me3. Importantly, we found that the H3K36me3 level was largely unchanged in preleptotene spermatocytes of the Setd2 cKO mice when compared with control littermates. Such findings suggested that H3K36me3 might be subjected to the regulation of other methyltransferases in the early stages of spermatogenesis. In support of this, Prdm9 was recently reported to mediate noncanonical H3K36me3 at hot spots of recombination sites during meiosis (54). As the epigenome undergoes rapid programming during spermatogenesis, whether there are other unknown methyltransferases responsible for H3K36me3, especially in the early stages of spermatogenesis, will need further investigation.

Mice
The Setd2 conditional knockout mouse line was produced by Shanghai Biomodel Organism Co. as shown in Fig. S2A. The germline-specific deletion of Setd2 was generated by crossing with Stra8-GFPCre mice or Vasa-Cre mice as described previously (10). All mice were maintained on the C57BL/6J background. All animal experiments were conducted in accordance with the guidelines of the Animal Care and Use Committee at Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences.

Histology, immunofluorescence, and TUNEL staining
For testis histology, Bouin's fixed sections were deparaffinized, rehydrated, and stained with hematoxylin and eosin (H&E). For immunofluorescence analysis, sections were retrieved in 10 mM sodium citrate buffer (pH 6.0) by boiling in a microwave and further washed in PBS with 0.1% Triton X-100. The sections were blocked with blocking buffer (10% donkey serum and 0.1% Triton X-100 in PBS) and then incubated with the primary antibodies overnight at 4°C followed by Alexa Fluor 488 -and Alexa Fluor 594 -conjugated donkey secondary antibodies (Jackson ImmunoResearch Laboratories) for 1 h at room temperature. The fluorescent sections were mounted with Prolong Gold Antifade medium containing DAPI (Molecular Probes) and then analyzed by fluorescence microscopy (Olympus) or confocal microscopy (Olympus). Apoptotic cells were analyzed using the In Situ Cell Death Detection kit, Fluorescein (Roche Applied Science) according to the manufacturer's instructions. The following primary antibodies were used in this study: rabbit anti-SETD2 (LS-B12260 and LS-C332416, LSBio), rabbit anti-H3K36me3 (4909S, Cell Signaling Technology), mouse anti-␥H2AX (05-636, Millipore), goat anti-GATA4 (SC-1237, Santa Cruz Biotechnology), and FITC-conjugated PNA (L7381, Sigma).

Isolation of pachytene/diplotene spermatocytes and round spermatids
Spermatocytes and round spermatids were enriched using Hoechst 33342/propidium iodide-based fluorescence-activated cell sorting (FACS) as described previously with minor modifications (55). In brief, testes from adults were collected in PBS and placed on ice. After removal of the tunica albuginea and interstitial testicular cells, the seminiferous tubules were digested with PBS containing 120 units/ml collagenase type I (Worthington), further digested with 0.25% trypsin (Gibco) and 5 mg/ml DNase I (Gibco), and then terminated by adding fetal bovine serum (FBS). The dissociated testicular cell suspension was collected and resuspended at a concentration of 1 ϫ 10 6 cells/ml in DMEM (Hyclone) with 0.1 mg/ml DNase I. The testicular cell suspension was then stained with Hoechst 33342 at a final concentration of 6 g/ml for 30 min and stained with propidium iodide prior to sorting immediately. Sorting of the spermatocytes and round spermatids was performed using flow cytometric analysis (BD Biosciences). The collected spermatocytes contained ϳ85 and ϳ10% of pachytene and diplotene spermatocytes, respectively, based on nuclear spreading analysis. About 90% of the sorted spermatids were round spermatids as determined by PNA staining.

Transmission electron microscopy
Fresh testes were fixed in 2.5% (v/v) glutaraldehyde in 0.1 M phosphate buffer, pH 7.4, for 2 h at 4°C; washed with 0.1 M phosphate buffer; postfixed in 2% OsO 4 for 1.5 h; dehydrated in a graded ethanol series before transfer to acetone; and embedded in Poly/Bed 812. Ultrathin sections were taken with a Leica EM UC7 ultramicrotome (Leica, Inc.), doubly stained with uranyl acetate and Reynold's lead citrate, and then imaged on a FEI Tecnai G2 Spirit TEM (FEI Co.) at 120-kV accelerating voltage.

ChIP and ChIP-seq
ChIP assay was carried out as described previously (56). The sorted pachytene spermatocytes and round spermatids were cross-linked with 1% formaldehyde for 10 min and then stopped by adding 125 mM glycine. Chromatin samples were SETD2 is crucial for spermatogenesis lysed with lysis buffer (20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS) and sonicated using a Qsonica R2. A total of 2 g of H3K36me3 antibody (4909S, Cell Signaling Technology) was incubated with each chromatin sample overnight at 4°C together with 0.5 g of spike-in antibody (61686, Active Motif) and 20 ng of spike-in chromatin (53083, Active Motif) for normalization purposes. The protein-DNA complexes were immobilized on 15 l of protein A/G beads (SA032005, Smart Lifesciences Inc., Changzhou, China) and then washed four times with lysis buffer, twice with low-salt buffer (10 mM Tris-HCl, 250 mM LiCl, 1 mM EDTA, 0.5% Nonidet P-40, 0.5% sodium deoxycholate), and once with 10 mM Tris-HCl, pH 8.0. Decross-linking was carried out in elution buffer (50 mM Tris-HCl, pH 8.0, 10 mM EDTA, 1% SDS) at 65°C for 5 h. Proteinase K and RNase A digestion was performed at 55°C for 1 h. DNA samples were then purified using a PCR extraction kit (28006, Qiagen) and analyzed by real-time PCR or prepared for deep sequencing according to the manufacturer's guidelines (KK8503, Kapa Biosystems).

RNA-seq and qRT-PCR
RNA samples were prepared using an RNeasy minikit (Qiagen). mRNA pools were enriched from total RNA by poly(A) beads. Barcoded RNA-seq libraries were constructed using an NEBNext Ultra Directional RNA Library Prep kit for Illumina according to the manufacturer's instructions (New England Biolabs). Nonstrand-specific pair end 150-bp sequencing was by X-Ten (Illumina, run by WuXi App Tec). qRT-PCR was performed using a SuperScript III Platinum SYBR Green One- Step qRT-PCR kit (Invitrogen).

ChIP-seq peak calling and distribution analysis
ChIP-seq reads were mapped with Bowtie (version 0.12.7) (parameters, Ϫm 1). Duplicate reads were removed by Samtools. The mapped unique and monoclonal reads were extended to 350 bp based on the average sonicated chromatin DNA length. H3K36me3 peaks were called using MACS1.4 with default settings. For H3K36me3 distribution analysis, the ChIP-seq signals over gene bodies of the up-and down-regulated genes in Setd2 mutant PD and RS cells were analyzed by a homemade script called Signalplot based on the normalized ChIP-seq tag density; all the gene bodies were aligned as the same length.

Heat map analysis
The RNA-seq heat map were generated based on the expression -fold changes in WT and Setd2 mutant cells using Pheatmap, which is an R package. The H3K36me3 heat map over gene bodies was generated using Deeptools2.0.

RNA-seq-related bioinformatics analysis
Low-quality bases and adapter-containing reads were trimmed from raw data by Trim-galore with default parameters. Then the remaining trimmed sequence were mapped against the reference genome (mm9) with Bowtie/TopHat version 2, which allowed mapping across splicing junctions by read segmentation. All programs were performed with default settings unless otherwise specified. Unique mapped reads were subsequently assembled into transcripts guided by reference RefSeq annotation with Cufflinks. The expression level of each gene was quantified with normalized FPKM with FPKM_count in RSeQC software. Differentially expressed genes were identified by asking for a log 2 (-fold change) Ն1 or ՅϪ1 with FPKM Ͼ0.3 in at least one sample.

GO analysis
GO analysis for differentially expressed genes was performed on the database for annotation, visualization and integrated discovery (DAVID) website.