SET domain-containing protein 4: biological functions and role in genomic methylation profiles of bone marrow mesenchymal stem cells

Background: Epigenetic modification is a crucial mechanism affecting the biological function of stem cells. SETD4 is a histone methyltransferase, and its biological role in bone marrow mesenchymal stem cells (BMSCs) is currently unknown. This work was aimed to reveal the SETD4 biological role as well as its impacts on the genomic methylation profiles in BMSCs. Methods: BMSCs were isolated form SETD4 knockout (KO) and wild type (WT) mice that established by CRISPR/Cas9 technology. The cell proliferation, migration, myogenic differentiation and angiogenesis were tested according to appropriate biology techniques. And the Reduced Representation Bisulfite Sequencing (RRBS) method was adopted to analyze the global genomic methylation profiles of BMSCs, following bioinformatics analysis of GO functions and KEGG signaling of differential methylated CpG sites and differential methylation regions (DMRs). Finally, validation experiments were conducted to examine the expression of histone lysine methyltransferase and some representative genes. Results: SETD4 KO significantly promoted BMSCs proliferation, which was characterized by enhanced cell viability and increased expression of PCNA, Cyclin A2, Cyclin E1, CDK2, CDK6, Bcl2 and decreased the expression of P16, P21 and Caspase3. SETD4 deficiency impaired BMSCs migration and myogenic differentiation potentials, and even the angiogenesis via paracrine of VEGF. Compared with WT control, the overall genomic methylation of BMSCs in the SETD4 KO group only was decreased by 0.47%. However, the changed genomic methylation covers a total of 96,331 differential methylated CpG sites and 8692 DMR, with part of them settled in promoter regions. GO and KEGG analysis revealed that differential CpG islands and DMRs in promotes impacted 270 GO functions and 34 KEGG signaling pathways, with some closely related to stem cell biology. SETD4 KO

Bone marrow mesenchymal stem cells (BMSC) are an auspicious kind of seed cell in tissue engineering involved in a wide range of functions. A vast number of clinical trials have shown that transplantation of BMSCs for tissue damage repairing is safe and effective.
Due to the shortcoming in de novo cell differentiation in vivo, the mechanisms regarding BMSCs infusion in tissue injury repair appear to be mostly related to its paracrine effects [9]. The ideal effect of stem cell transplantation for tissue damage depends on the cell transplantation rate, differentiation, and paracrine protective factors. Epigenetics in mesenchymal stem cell differentiation has attracted considerable attention, and most investigations have been focused on osteogenic, adipogenic and cardiogenic differentiation [10-12], highlighting epigenetic modification in stem cell biology and regenerative medicine [13].
In this investigation, we identified the role of SETD4 in the biological function of BMSCs, and we further uncovered the epigenetic modification about SETD4-impacted genomic methylation profiles in BMSCs. Our results provide insight into revealing the value of the targeted intervention of methyltransferase SETD4 in BMSC-based regenerative medicine.
Methods h under normoxic condition (21% O 2 ), then medium was changed for DMEM supplemented with 10% FBS (Gibco), TGF-b1 for continuous 14 days. The cells were subjected to RNA isolation and quantitative RT-PCR analysis of the expression of differentiation-related markers of cardiomyocyte lineage as following: Nkx2.5, Gata4, Mef2a, ANP, CX43 and cTnT.
Cell viability assessment by CCK-8 assay CCK-8 assay was used to test the cell viability as previously described [16]. Briefly, BMSCs (2×10 3 cells/well in 200 μL medium) were seeded onto 96-well plates. Cells were allowed to recover overnight; the medium was changed every 3 days. At 0, 1, 2, 3 and 7 d setpoints, 10 μL of CCK-8 reagents (#SPDA-D010, Beyotime Institute of Biotechnology, Nanjing, China) was added, cells were maintained for additional 2 h, then OD values at 450 nm were measured using a microplate reader (Thermo Scientific, USA). Each group was duplicated in six wells.

Cell migration assay
Transwell migration assay was performed using 8-μm pore size polycarbonate membrane chambers (Corning, Tewksbury, MA, USA) as previously described [17]. Briefly,2×10 4 BMSCs in total 200 μL were seeded in the upper chamber; the lower chamber contained 500 μL DMEM medium supplemented with 10% FBS. Cells were maintained for 24 h at 37°C in the incubator, and then the chambers were fixed for 15 min with 4% paraformaldehyde and stained with 0.1% crystal violet, followed by washing with PBS twice. Cells on the upper surface of the membrane were removed. The images were photographed with a microscope (Leica, Wetzlar, Germany) at 20× objective field, and the number of migrated cells on the lower surfaces of the membranes was counted. Each experimental group was repeated in three chambers, and cell numbers from a total of nine representative fields were obtained.

Assessment of angiogenic factors by ELISA
A total of 2×10 5 BMSCs were seeded on 6-cm dishes, and cells were maintained under normoxia (21% O 2 ) and hypoxia condition (2% O 2 ) for 1, 3 and 7 d. Cell culture supernatant was collected and subjected to ELISA analysis of VEGF levels, according to the manufacturer's instructions. Mouse VEGF ELISA kit (#RRV00, R&D, USA) was used.

Tube formation assay
HUVECs (presented by Dr. Xujuan Zhang, Department of Physiology, Guangdong Medical University) were maintained with DMEM medium supplemented with 10% FBS. Cells in the logarithmic growth phase were used for experiments. A total of 300 μL Matrigel (#9056007, Corning, USA) was added in a 24-well plate and coated at 37°C for 30 min. A total of 5×10 4 well HUVECs were seeded per wells. After 10 min of cell attachment, the media was replaced by the WT and SETD4-KO BMSC conditional media. Images were acquired after 6 h of cell culture under a 10× microscope objective. The tube formation 9 exon junction are listed as follows:

Western blot
Cells were lysed with RIPA buffer (Beyotime Institute of Biotechnology). A sample of 35 µg total proteins were subjected to 12% SDS-PAGE and then transferred to PVDF membranes (Millipore, USA). After washing with TBST twice, the membranes were incubated with 5% skim milk powder in TBST at 37°C for 1 h and then primary antibodies at 4°C overnight. After washing with TBST twice, the membranes were incubated with HRP-conjugated IgGs (1:5000, #SA00001-10, #SA00001-2, ProteinTech) for 1 h at 37°C. Bands were visualized using ECL reagents (Thermo Fisher, USA) and analyzed with a gel analysis system (Tanon; Shanghai, China).

DNA methylation analysis by Reduced Representation Bisulfite Sequencing (RRBS) on Genechem platform
BMSCs were cultured on 6-cm dishes at 37°C for 72 h, marked as WT and SETD4 KO groups (n = 3). Cells were harvested and immediately frozen in liquid nitrogen for 30 min and sent to Genechem Co, Ltd. (Shanghai, China), to assess the genomic DNA methylation status. The experimental protocol steps were as follows: (1) 500 ng genome DNA in each group was digested overnight at 37°C using restriction enzyme Msp1 (Sigma Aldrich) in order to enrich the CpG islands and other CpG methylation-intensive areas.
(2) The digested product was purified using an Axygen PCR Cleaner Kit; (3) Repair was ended, and A was added to 3'-terminal of product; (4) Product was re-purified using the Axygen PCR Cleaner Kit (Corning); (5) Methylation linker was added; (6) Agarose electrophoresis and DNA fragments with sizes ranging from 160 to 400 bp were extracted using a GeneJE T Gel Extraction Kit (ThermoFisher Scientific); (7) Bisulfite treatment was conducted using an EZ DNA Methylation-Gold kit (Zymo Research, USA); (8) Bisulfite-converted library with conversion rate > 98% was subjected to PCR amplification, magnetic bead purification and Qubit measurement;(9) Single-read sequencing was conducted using an Illumina Hiseq.
The raw data obtained by Illumina Hiseq sequencing were converted into sequence data by Base Calling, and the results were stored in FASTQ file format. Clean reads were compared with reference mouse genome Mus_musculus. GRC38. After the quality check, the R package methylKit was used to analyze the differential methylation status and functional annotation, which include coverage of different CpG, distribution of differential methylated CpG, distribution of CpG in different characteristic regions, difference in CpG methylation between different samples and differential methylation regions (DMRs). David Online Software was adapted to document genes' function (GO enrichment and annotation) and signal pathways (KEGG signaling enrichment and annotation).

Statistical analysis
Statistical analyses were conducted using GraphPad Prism (Version 7.0, GraphPad Software, CA, USA). Data are expressed as means ± SEM. Differences between two groups were analyzed using umpired Student's t tests. For comparisons between multiple groups, ANOVAs were used followed by Tukey's multiple comparisons test. P<0.05 was considered to be statistically significant.

Morphology and immunophenotype of BMSCs from WT and SETD4 KO mouse
Mouse SETD4 gene was selected as the editing target, and exons 6-8 were designed to be deleted using CRISPR/Cas9 technology (Fig. 1a). After the pups were grown up, mouse tail DNA was used for genotypic identification, and the identification strategy is shown in Fig. 1b and 1c. We found that the positive animals of SETD4 KO displayed a deletion of 4044 bp DNA fragment (Fig. 1d), supporting the success in establishing SETD4 KO mice. Then the BMSCs were isolated from wild type (WT), heterozygote and SETD4 KO mice, and western blot results confirmed changes in the expression of the SETD4 protein in corresponding BMSCs (Fig. 1e). Microscopically, the morphology of BMSCs in WT and KO groups was not significantly different (Fig. 1f). Immunostaining of BMSC markers indicated both WT and SETD4 KO cells were positive for CD90, CD73 and CD45R, and negative for CD34 (Fig. 1g).

Deficiency in SETD4 promotes BMSCs proliferation but not migration
CCK-8 assay was used to determine the viability of BMSCs. As shown in Fig. 2a, the viability of the KO group was significantly higher than that of the WT group since 3 d postcell seeding, suggesting SETD4 deficiency was beneficial for BMSCs proliferation. Because cell proliferation was tightly associated with changes in cell cycle and apoptosisassociated proteins, we then detect the levels of cell cycle and apoptosis-associated proteins. As shown in Fig. 2b and 2c, SETD4 KO up-regulated PCNA, the marker for cell proliferation, and increased a panel of cell cycle-associated proteins, like CDK6, CyclinA2 and CycinD1. In contrast, P16 and P21 were decreased. Interestingly, we detected an increase of Bcl2 and a decrease of total Caspase3 in SETD4 KO cells, even under normal cell culture conditions, but we failed to detect the active Caspase3. We further examined the migratory ability of BMSCs. As shown in Fig. 2d, the transwell migration assay revealed that BMSCs in the SETD4 KO group displayed much-weakened ability in migration than that of the WT group.

SETD4 is a requirement for BMSC-promoted angiogenesis
Considering that one crucial role of BMSCs in tissue injury repair is the promotion of angiogenesis, we assessed the contribution of SETD4 in BMSC's ability to secrete VEGF, a key pro-angiogenic factor. We found that WT BMSCs can secrete VEGF to the supernatant of culture medium to some extent with the prolongation of culture time, and this potential is more pronounced when cultured under hypoxic conditions. Compared with WT cells, SETD4 KO BMSCs displayed a significant inhibited VEGF secretion under both normoxia and hypoxia (Fig. 2e). VEGF-A mRNAs in BMSCs further confirmed this tendency (Fig. 2f). Using the angiogenic growth factor-enriched supernatant, we further tested the HUEVC tube formation ability, which showed supernatant from WT BMSCs stimulated, while supernatant from SETD4 KO BMSCs weakened, HUEVC tube formation potentials (Fig. 2g).

SETD4 KO inhibited myogenic differentiation of BMSCs
Under the stimulus combining 5-Aza and TGF-b1 for 14 d, WT BMSCs can differentiate in vitro into cardiomyocyte-like and smooth muscle cells (SMC, Fig. 3), as evidenced by the enhanced expression of cardiomyocyte lineage markers (Nkx2.5, Gata4, ANP, CX43) and SMC lineage marker (a-SMA, SM22a). Although the stimulation combining 5-Aza and TGF-b1can also induce differentiation of SETD4 KO BMSCs in a similar fashion to WT, its differentiation ability was significantly reduced. We also noticed that at the beginning of differentiation (day 1), SETD4 KO BMSCs displayed much lower baseline levels of differentiation than that of WT cells. Together these results suggested that SETD4 KO inhibited BMSCs' myogenic differentiation potentials.

SETD4 KO altered BMSC genomic methylation status
Considering SETD4's role as a methyltransferase, to change the genomic methylation conditions, we, therefore, adapted RRBS to assess the degree of genomic methylation.
Through rigorous quality control (Fig. S1, Additional file 1), our RRBS results were qualified and suitable for subsequent analysis. The basic information including the original sequencing reads, the length of the reads, the number of all bases, and the clean reads after processing and the comparison information are shown in Table S1 (Additional file 2).
Compared to WT cells, the average methylation of all CpG in SETD4 KO BMSCs was decreased by 0.47%, suggesting that the single knockout of SETD4 has limited effects on the overall genomic methylation. Further information indicated the coverage of CG of the two groups of samples was identical, and the minimum coverage was 10×, mainly concentrated between 10× and 100×, indicating that the data have high credibility (Fig.   S2a, Additional file 3). It can be seen that the CG methylation of the two groups of samples is distributed at both ends, that is, mainly distributed in permethylation (90%-100%) and total unmethylation (0%-10%) (Fig. S2b, Additional file 3). Annotation of different feature areas of CpG showed that the distribution of CpG on the genome is similar in WT and SETD4 KO cells (Fig. 4a, b). We further screened the differential site methylation CpG and annotated the gene and promoter regions. There were total 96,331 differential site methylation CpGs (Fig. 4c) and 8692 DMRs (Fig. 4d)

Bioinformatics analysis using GO and KEGG database
Taking the difference >25% and P<0.01 or <-25% and P<0.01 as a standard, we found 15 signaling pathway annotations were performed using David's online software. As shown in We enriched 62 KEGG signaling, 34 of which were statistically different (Fig. 5c). The total KEGG signaling is listed in Table S5 (Additional file 7). Among the 34 KEGG signaling, mmu04024(cAMP signaling pathway), mmu04015(Rap1 signaling pathway), mmu04810(Regulation of actin cytoskeleton), mmu04550(Signaling pathways regulating pluripotency of stem cells), mmu04510(Focal adhesion), mmu04390 (Hippo signaling pathway) and mmu04010 (MAPK signaling pathway) have been reported to be closely related to stem cell biology.

Validation of representative genes and HKMTs
Our investigation focused on the enriched GO term (0005667, transcription factor complex), and two factors named Nkx2.5 and Gata4 were also included (Table S4, Additional file 6). These two transcriptional factors also function in cardiomyocyte differentiation. These two factor processes multiply differentially methylated CpG sites (Table S6, Additional file 8). The baseline level of Nkx2.5 and Gata4 in SETD4 KO cells was lower than that of WT cells (Fig. 3). We further assessed another nine genes with multiple different methylated CpG sites, Gli2, Grem2, E2f7, Map7, Nr2f2 , Shox2, Foxo6, Klf14 and Dyrk3 ( Fig. 6a; for full differential methylated CpG sites, see Table S6, Additional file 8).
We found that the transcription levels of Gli2, Map7, Shox2 and Klf14 were increased in SETD4 KO BMSCs, while the transcription levels of Grem2, E2f7, Nr2f2, Foxo6 and Dyrk3 were decreased in SETD4 KO cells when compared with the WT cells (Fig. 6b). Taken Klf14 was a sample, it just has 65 hypomethylated sites but no hypermethylated sites in its promoter region (Fig. 6a), and the level of Klf14 mRNA therefore was significantly upregulated in SETD4 KO cells, this expression trend was consistent with the discipline of hypomethylation promoting gene expression. We also tested the changes in sets of HKMTs and found that SETD4 KO BMSCs indeed processed lower levels of H4K20me1, H4K20me2, H3K4me1, H3K4me2, H3K27me2, H3K36me1, H3K36me2, H3K79me1 and H3K79me2 than WT cells, whereas the histone arginine trimethyltransferase was not significantly changed ( Fig. 6c). Thus, change in these monomethylases and dimethylases for histone arginine that is induced by SETD4 KO may account for the biological function difference between SEDT4 KO and WT BMSCs, as illustrated in Fig. 7.

Discussion
To explore the contribution and genomic methylation profiles of SEDT4 in BMSCs, we first established SETD4 KO mice using CRISPR/Cas9 technology, and these SETD4 KO mice showed a 4,044 bp deletion in exons 6-8 and significant inhibition of SETD4 at the translational level. Thus, these SETD4 KO mice provide a basis for further functional analysis. Morphologically, SETD4 KO BMSCs displayed no obvious difference with WT BMSCs, along with its immunophenotype. We focused on the proliferation, paracrine angiogenic factors and differentiation potentials, and our results provide the first evidence regarding the role of methyltransferase SETD4 in BMSCs biology.
High cell viability is a crucial determining factor in stem cell-based regeneration medicine.
We found that SETD4 deficiency was beneficial for BMSCs proliferation, as evidenced by the CCK-8 assay results. Furthermore, we found that SETD4 KO up-regulated CDK6, cyclinA2 and cycinD1, and down-regulated P21 expression. Because these proteins were strongly associated with cell cycle progression, we can assume that SETD4 KO enhances BMSCs proliferation mainly through mediating cell cycle-associated proteins expression.
We noticed that apoptosis-associated proteins such as Bcl2 and Caspase3 also changed, but no active Caspase3 was detected. It seems likely that inhibition of apoptosis may not be the key factor for SETD4 KO-induced BMSCs viability. An early report indicated that high expression of SETD4 promoted breast cancer cell proliferation without affecting the apoptosis[4]. Our current results suggest SETD4 deficiency leads to BMSCs proliferation without affecting the apoptosis. Very recently, Ye S and colleagues reported SETD4 overexpressed breast cancer stem cells deserved low Ki67 expression when compared with SETD4 low expressed breast cancer stem cells; they concluded SETD4 controls breast cancer stem cell quiescence [6]. Our current research found that SETD4 KO BMSCs retain high cell proliferative potentials, which is consistent with Ye S's report on cell proliferation characteristics. Interestingly, SETD4 KO also led to injured BMSCs migration. The morphology of cells in WT and SETD4 KO groups is very similar; thus, the reason for SETD4 KO-inhibited cell migration remains unsolved.
We next analyzed the effects of SETD4 KO on BMSC paracrine pro-angiogenic factors.
Angiogenic factors like VEGF, TGF-β1, angiogenin-1 and bFGF contribute greatly to injured tissue recovery through mediating angiogenesis, and BMSCs process the ability of paracrine of angiogenic factors, which has been demonstrated by many investigations [18][19][20][21]. To our surprise, the current investigation showed that SETD4 KO dramatically attenuated BMSC producing VEGF at transcription and translation levels. Functionally, conditional medium enriched in supernatants from SETD4 KO BMSCs indeed led to a weakened tube formation capability of HUEVCs. This decrease in the ability of paracrine angiogenic factors suggests the functional damage of SETD4 KO cells in tissue injury transplantation. We plan to address this question in future in vivo experiment. A very recent investigation indicated SETD4 was positively involved in the release of cytokines like IL-6 and TNF-α in lipopolysaccharide-treated macrophages [8], highlighting the role of SETD4 in the immune and inflammatory response. Given that BMSCs were also involved in the macrophage-mediated inflammatory response [22], our current study provides novel molecular insights into epigenetic-modified immune response.
Directed differentiation is an important focus of BMSC function research. In this study, we investigated the myogenic differentiation potential of BMSCs and conducted experiments using mature myogenic differentiation induction protocols [23,24]. Especially, for better efficiency, we combined 5-Aza with TGF-β1 to induce myogenic differentiation. The outcomes supported the notion that SETD4 was a requirement for myogenic differentiation in BMSCs, as elucidated by the abrogated expression of markers in cardiomyocytes  [31]. In addition, Rap1 signaling pathway [32], regulation of actin cytoskeleton [33], signaling pathways regulating pluripotency of stem cells [34], focal adhesion [35], Hippo signaling pathway [36] and MAPK signaling pathway [37] that enriched in the current investigation have been reported to be closely related to stem cell biology.
The results of GO and KEGG bioinformatics analysis suggest that the differentially methylated CpG islands and DMRs caused by SETD4 KO contain complex mechanisms that regulate BMSC biology.
We validated the expression of some interesting genes that are linked to differential methylation in promoter regions. We selected Nkx2.5, Gata4, Gli2, Grem2, E2f7, Map7, Nr2f2, Shox2, Foxo6, Klf14 and Dyrk3 as objectives. Among them, Nkx2.5 and Gata4 were tightly associated with cardiogenic differentiation [38]. Gli2 [39], Grem2 [40], E2f7 [41], Nr2f2 [42] and Shox2 [43] are also related to stem cell biology. Interestingly, SETD4 KO significantly changed the transcript of these genes. Among these genes, Gli2, E2f7, Shox2 and Klf14 were up-regulated, and the remaining genes were down-regulated. We found that all these genes contain multiple differentially methylated CpG sites in its promoter region. A dogmatic notion suggests that hypermethylation conventionally leads to gene silence, while hypomethylation generally results in gene activation. Since these genes contain multiple differentially methylated CpG sites, we did not test which CpG site plays the key role in SETD4 KO-induced changes in gene expression. In fact, total 96,331 differentially methylated CpG sites and 8692 DMRs were obtained in our study; these large numbers of differentially methylated CpG sites and DMRs undoubtedly cause a complex signal network and finally interfere with the biological function of BMSCs.
Finally, considering that SETD4 serves as a methyltransferase for histones, we thus tested the change of site-specific methylation of histone 3 (H3) and histone (H4). The results indicate that SETD4 KO obviously changed the levels of some HKMTs including H4K20me1, H4K20me2, H3K4me1, H3K4me2, H3K36me1, H3K36me2, H3K79me1 and H3K79me2, whereas no significant change was observed in the histone arginine trimethyltransferase.
Thus, we hypothesized that the role of SETD4 in the biological function of BMSCs is mainly related to the change of genomic methylation level caused by the change of these monomethylases and dimethylases. The diagram of mechanism is illustrated in Fig. 7.

Conclusions
In summary, this is the first investigation to explore the biological functions of SETD4 in BMSCs and its impact on genomic methylation. Our results demonstrate that SETD4 plays a crucial role in proliferation, migration, paracrine and myogenic differentiation of BMSCs. Even though SETD4 changes minimal level in overall genomic methylation, it still leads to large differential methylated CpG sites and DMRs, which results in numerous changes in cellular functions and signaling pathways. This investigation provides an experimental basis for BMSC-based regeneration medicine through epigenetic modification.    biological process, cellular component and molecular function that enriched in differential methylated CpG sites and differential methylated regions in promoters. For the detail of each GO term, see Table S4 (Additional file 6). c Total 34 differential KEGG signals were enriched in differential methylated CpG sites and DMRs in promoters. These KEGG signals were arranged in P values converted -log (P values). For the enriched genes' detail of each KEGG signal, see Table S5 (Additional file 7).