Genome-wide analysis of DNA methylation identi es the apoptosis-related gene UQCRH in renal cancer

Kosuke Miyakuni The University of Tokyo: Tokyo Daigaku Jun Nishida The University of Tokyo: Tokyo Daigaku Daizo Koinuma The University of Tokyo: Tokyo Daigaku Genta Nagae The University of Tokyo: Tokyo Daigaku Hiroyuki Aburatani The University of Tokyo: Tokyo Daigaku Kohei Miyazono The University of Tokyo: Tokyo Daigaku Shogo Ehata (  ehata-jun@umin.ac.jp ) The University of Tokyo: Tokyo Daigaku https://orcid.org/0000-0002-6740-9391

therapy directed toward vascular endothelial growth factor (VEGF) and mammalian target of rapamycin (mTOR) have developed over the past two decades. [8][9][10][11][12] However, their therapeutic effects remain limited. [13,14] Other mechanisms have been revealed to play important roles in the ccRCC progression and are thus expected to be potential targets for treatment. For instance, epigenetic modi cations are important not only for carcinogenesis but also for metastasis of ccRCC. [15] Particularly, somatic mutations of genes related to histone modi cation have been con rmed in ccRCC. Alterations in polybromo 1 (PBRM1), breast cancer early onset (BRCA) associated protein 1 (BAP1), SET domain containing 2 (SETD2), and enhancer zeste 2 polycomb repressive complex 2 subunit (EZH2) are closely associated with clinical outcomes in ccRCC patients. [15][16][17][18][19] These chromatin regulators affect a large number of gene transcriptions, which promote the heterogeneity and evolution of ccRCC cells. [20,21] We have also previously demonstrated that in ammation-related signalling is constitutively activated in advanced ccRCC through the formation of a super-enhancer. [22] In addition to histone modi cation, cytosine residue of DNA methylation is implicated in the ccRCC progression. [23] de novo DNA methylation is induced mainly at 5′-C-phosphate-G-3′ (CpG) dinucleotides by DNA methyltransferase (DNMT)3A and DNMT3B, and this methylation process is maintained by DNMT1. [24] The methylation of DNA cytosine bases leads to the inaccessibility of transcription factors to DNA regulatory elements, which in turn silences the transcription of tumor-suppressor genes. [25][26][27][28] Recent studies have revealed that DNA hypermethylation is frequently observed in ccRCC and correlates with poor prognosis. [16] However, unlike in other cancers, the methylated genes responsible for cancer progression are still unclear in ccRCC. Here, we identi ed the target(s) for DNMTs in ccRCC cells using genome-wide analysis, and their function in the regulation of cellular survival was evaluated. Highly malignant derivatives (OS5K-1, -2, and -3 cells) were established and maintained as previously described. [22] The protein kinase inhibitor staurosporine (Abcam, Cambridge, UK) and the DNMT inhibitor 5-aza-deoxycytidine (dC), (Sigma-Aldrich, St Louis, MO) were reconstituted in dimethyl sulfoxide (DMSO). The mTOR inhibitor everolimus (RAD001, Selleck Chemicals, Houston, TX) was used.

Lentiviral vector construction and production
Page 4/30 The lentiviral vector system (provided by H. Miyoshi, deceased, formerly Keio University) was used for speci c gene overexpression and knockdown as previously described. [29] For UQCRH overexpression, the complementary DNA encoding human UQCRH was inserted into the multiple cloning site of the empty vector pENTR201. Recombination between pENTR201 and the destination vector CSII-CMV-RfA was performed using Gateway cloning technology (Thermo Fisher Scienti c). The pCSII-EF-enhanced green uorescent protein (GFP) was produced as previously described. [29] For UQCRH knockdown, short hairpin RNAs (shRNAs) targeting UQCRH were inserted into the entry vector pENTR4-H1. The target sequences for shRNA are shown in Table 1 or as reported previously. [30] Recombination between pENTR4-H1 and the destination vector pCS-RfA was performed using Gateway cloning technology. The prepared plasmids, pCAG-HIVgp, and pCMV-VSV-G-RSV-Rev, were transfected into 293FT cells using Lipofectamine 2000 (Thermo Fisher Scienti c). The lentiviral vectors were collected from the culture supernatants and concentrated using the Lenti-X Concentrator (Clontech, Mountain View, CA).
Quantitative RT-PCR (qRT-PCR) analysis qRT-PCR analysis was performed as previously described. [30] Total RNA was extracted using the ISOGEN Reagent (Nippon Gene, Toyama, Japan) or an RNeasy Mini Kit (Qiagen, Hilden, Germany). cDNA was synthesized using the PrimeScript II 1st strand cDNA synthesis kit (Takara Bio, Shiga, Japan), and the cDNA products were mixed with FastStart Universal SYBR Green Master Mix with ROX (Roche Diagnostics, Basel, Switzerland) and analyzed on StepOnePlus Real-Time PCR system (Thermo Fisher Scienti c). The expression levels of human UQCRH mRNAs were normalized to that of human ACTB mRNA. The primer sequences are shown in Table 2.

Immunoblotting
Immunoblotting was performed as previously described. [ East Hills, NY) blocked with 5% skim milk containing tris-buffered saline with Tween20 (Sigma-Aldrich) (TBST). The membranes were incubated with primary antibodies and the appropriate secondary antibodies (Table 3) diluted in TBST, Can Get Signal 1 (Toyobo, Osaka, Japan), or Can Get Signal 2 (Toyobo). Chemiluminescence images were captured using an ImageQuant LAS 4000 device (Fuji Film, Tokyo, Japan).
Immunohistochemistry and TdT-mediated dUTP nick end labeling (TUNEL) assay For immunostaining, mouse tissues were xed with Mildform (Wako Pure Chemical, Tokyo, Japan), whereas for human renal tumor tissues, a pre-xed human tissue array was purchased (KD2082a, US Biomax Inc., Rockville, MD). After para nisation, the tissues were sectioned and depara nized using xylene and ethanol, followed by antigen retrieval using Universal HIER Antigen Retrieval Reagent (Abcam). Samples were subjected to hematoxylin and eosin staining or immunostaining. After blocking with Block ACE (Bio-Rad, Hercules, CA), the samples were incubated with primary antibodies and the appropriate secondary antibodies (Table 4) and stained using a Dako Liquid DAB+ Substrate Chromogen System (Agilent Technologies, CA) and Mayor's hematoxylin. Images were captured using an AX80 microscope (Olympus, Tokyo, Japan).
To detect apoptosis, In Situ Cell Death Detection Kit (TMR red; Roche Diagnostics) and DAPI Fuloromount-G (Southern Biotech, Birmingham, AL) were used as previously described. [32] Fluorescent images were captured using a BZ-X710 microscope (KEYENCE, Osaka, Japan).

Mouse renal orthotopic tumor models
All experiments were approved by the Animal Ethics Committee of the University of Tokyo. Mouse renal orthotopic tumor models were generated as previously described. [29] Brie y, BALB/c-nu/nu male mice (5weeks-old) were purchased from Sankyo Labo Service Corporation (Tokyo, Japan). ccRCC cells (1.0 × 10 5 ) expressing Luc2 and mCherry were inoculated into the subrenal capsule of mice. For in vivo bioluminescence imaging, D-luciferin potassium salt (200 mg/kg; Promega, Madison, WI) was diluted in PBS and injected into mice intra-peritoneally. For ex vivo bioluminescence imaging, the harvested kidneys and lungs were reacted with D-luciferin potassium solution for 10 min, and images were captured using Night OWL LB981 (Berthold Technologies, Bad Wildbad, Germany). Quantitative analysis was conducted using the IndiGO software (Berthold Technologies). Everolimus was reconstituted in saline solution (Otsuka, Tokyo, Japan) containing 5% Tween20 and 30% propylene glycol (Sigma-Aldrich) and administered to mice (2.5 mg/kg) thrice weekly.

Cell proliferation assay
Cell Counting Kit-8 (Dojindo Laboratories, Kumamoto, Japan) was used according to the manufacturer's protocol.

Colony formation assay
The colony formation assay was performed as previously described. [33] Flow cytometry analysis Flow cytometry analysis was performed as previously described. [22] Brie y, the cells were collected, washed with Annexin V binding buffer (Thermo Fisher Scienti c), and reacted with FITC-conjugated Annexin V (Thermo Fisher Scienti c) at room temperature for 10 min. Apoptotic cells were detected using a Gallios ow cytometer (Beckman Coulter, Brea, CA).

Bisul te-sequencing analysis
Genomic DNA extraction and bisul te conversion were performed as described previously. [22] Brie y, genomic DNA was puri ed using a Gentra Puregene Cell Kit (Qiagen). Bisul te conversion was performed using an EpiTect Bisul te Kit (Qiagen). Bisul ted DNA was ampli ed with Takara Epi-Taq HS ( Takara Bio) using a speci c primer for the human UQCRH CpG island shore. The primer sequences are listed in Table  5. After ligation with the pCR4-TOPO vector using the TOPO TA Cloning Kit (Thermo Fisher Scienti c), products were transformed into DH5a and sequenced.

Methylation array
Genomic DNA extraction was performed as described previously. [22] Five hundred ng of genomic DNAs were quanti ed by Qubit Fluorometer (Life Technologies, Carlsbad, CA) and bisul te-converted using an EZ DNA Methylation Kit (Zymo Research, Irvine, CA). Methylation array was conducted using the In nium Human MethylationEPIC BeadChip Kit (Illumina, San Diego, CA) according to the manufacturer's protocol.
The raw signal intensity for methylated and unmethylated DNA was measured using a BeadArray Scanner (Illumina). After color-bias correction, background subtraction of the signal intensities, and interarray normalization on Genome Studio (Illumina), the raw methylation value (β-value) for each CpG was de ned as M / (M + U + 100), where M and U were the intensities of methylated and unmethylated probes, respectively. CpG loci located-0-500 bp upstream of transcript start sites (TSS) were used for the analysis of methylation status in promoters.

Public database
Data for gene expression and DNA methylation were obtained from public databases, i.e., The Cancer Genome Atlas (TCGA) program, Gene Expression Omnibus (GSE131137, GSE53757, and GSE83820) of the National Center for Biotechnology Information (NCBI), and Cancer Cell Line Encyclopedia (CCLE) of the Broad Institute.

Statistical analysis
Graph generation and statistical analysis were performed using Excel (Microsoft, Redmond, WA), JMP Pro 14.2, R (v4.0.2), and Python 3. No method was used to analyse the sample sizes. Two-group comparison was analyzed using Student's t-test or Welch's t-test based on the results of the F-test. For multiple comparisons, one-way analysis of variance (ANOVA), Tukey's test, and Dunnett's test were used. For the Kaplan-Meier plot analysis, a log-rank test was used.

Increased expression of DNMT3B contributes to renal cancer progression
In our previous study, orthotopic transplantations were employed to establish highly malignant derivatives of human ccRCC cells (Fig. S1A). Parental OS-RC-2 (OSPa) cells were repeatedly exposed to the renal microenvironment. After ve serial orthotopic transplantations, three derivatives were obtained as OS5K-1, -2, and − 3 cells. Although the proliferative ability of OS5K cells did not increase in cell culture ( Figure S1B), they exhibited increased tumor formation and metastasis in 3D culture conditions or in vivo ( Fig. S1C-F).
To examine the role of DNA methylation during renal cancer progression, the expression levels of DNMTs were determined using our previous RNA-sequencing (RNA-seq) data (GSE131137) [22] and immunoblotting. The expression of DNMT3B was upregulated in OS5K cells (Fig. 1A, B). The activity of DNMTs in these cells was diminished following treatment with 5-aza-dC, a DNMT inhibitor. Although the viability of OSPa and OS5K cells was decreased by 5-aza-dC, 5-aza-dC was more potent in suppressing that of OS5K cells (Fig. 1C). When xenografted, 5-aza-dC pre-treatment attenuated the formation of primary tumor in OS5K cells, although lung metastasis was not signi cantly affected (Fig. 1D, E). Histological examination revealed a decrease in the number of cells with 5-methylated cytidine in the nuclei, which was accompanied by an increase in TUNEL-positive cells (Fig. 1F, G). These results suggest that DNA methylation is accelerated by DNMT3B in OS5K cells, which may account for their pro-survival phenotype.
Next, the involvement of DNA methylation in ccRCC progression was con rmed using clinical datasets. In ccRCC cases, increased expression of DNMT1 and DNMT3A was observed in a stage-dependent manner ( Fig. 2A). Moreover, poor prognosis of ccRCC patients correlated with the upregulation of DNMT3A and more signi cantly with DNMT3B (Fig. 2B). Overall, these data suggest that the expression of DNMT3B enhances DNA methylation during ccRCC progression, which may be important in tumor formation and related to poor patient outcomes.

Identi cation of targets for DNMT3B in renal cancer cells
To uncover the role of DNA methylation, genome-wide screening of DNA methylated sites in ccRCC cells was performed using a methylation array (Fig. 3A). We also re-analyzed the previous RNA-seq data (GSE131137) (Fig. 3B). We con rmed that the decreased expression of several genes was correlated with DNA methylation (Fig. 3C, D). To assess the features of methylated genes, the hypermethylated genes were extracted in OS5K cells and subjected to Gene Ontology (GO) analysis. GO analysis of the biological process revealed that the genes governing cellular survival were methylated in OS5K cells, including apoptosis and necrosis (Fig. 3E).
Next, among these methylated genes in OS5K cells, we identi ed those that are important for renal cancer progression. Ubiquinol cytochrome c reductase hinge protein (UQCRH), one of the components of the mitochondrial complex III, was extracted as a methylation target. Compared with normal proximal tubule HK-2 cells, UQCRH expression was decreased in OSPa cells and further reduced in OS5K cells (Fig. 4A, B). When all of the components of the electron transport chain in OS5K cells were examined, neither gene expression nor methylation status was altered, except for that of UQCRH (Fig, S2).
Clinical database analysis showed that decreased expression of UQCRH was con rmed in cancer cells derived from the kidney and ovary (Fig. S3A). Using the GEO data, we found that the loss of UQCRH cells was observed during the establishment of patient-derived xenograft (PDX) using ccRCC tissues (Fig.  S3B). To assess the clinical signi cance of the decreased expression of UQCRH, its expression in renal tumor tissues was analyzed using clinical samples. Immunohistochemical analysis revealed that the expression of UQCRH in ccRCC tissues was lower than that in normal adjacent tissues, irrespective of the tumor grade (Fig. 4C, D). Moreover, UQCRH expression in other histological types of renal tumor tissues, including papillary RCC, squamous cell carcinoma, and sarcomatoid carcinoma, was also lower than that in normal adjacent tissues.
Based on these observations, we focused on the role of UQCRH in ccRCC cells in subsequent experiments. Particularly, the methylation status of the CpG island of the UQCRH promoter locus was investigated. The methylation array revealed that all of the GpG loci in UQCRH were methylated in OS5K cells but not in OSPa cells (Fig. 5A). Similar results were yielded by bisul te-sequencing analysis (Fig. 5B).
Clinical database analysis revealed that the methylation of the promoter in UQCRH was frequently observed in cancer cells from the kidney and ovary (Fig. S4A). Notably, the expression of UQCRH was inversely correlated with that of DNMT3B in ccRCC (Fig. S4B).
To directly examine the regulation of UQCRH by DNA methylation, OS5K cells were treated with the DNA methyltransferase inhibitor 5-aza-dC. The expression of UQCRH was restored at both the mRNA and protein levels (Fig. 5C, D). Taken together, these data suggest that the DNMT3B-mediated decrease in UQCRH may contribute to renal cancer progression.
Renal cancer cells acquire apoptosis resistance through the decrease of UQCRH The role of UQCRH expression in apoptosis induction was further investigated. In OSPa cells, the translocation of cytochrome c from the mitochondria to the cytosol, an indicator of the initiation of the apoptotic process, was observed following staurosporine treatment (Fig. 6A, B). Subsequently, the cleavage of poly (ADP-ribose) polymerase (PARP), which indicates caspase activation (Fig. S5A), was observed. However, these apoptotic processes were suppressed in OS5K cells even after treatment with staurosporine.
Next, the involvement of UQCRH in apoptosis induction was examined using overexpression and knockdown experiments. The expression of UQCRH in OS5K-3 cells was recovered using lentiviral vectors (OS5K-UQCRH cells) (Fig. S5B, C). Cytochrome c was not released into the cytoplasm in the control OS5K-GFP cells after staurosporine treatment, whereas it was e ciently released in OS5K-UQCRH cells (Fig.  S5D). We also established UQCRH-silenced OSPa cells using shRNAs (OSPa-shUQCRH #1, #2 cells) (Fig. 6C, D). Immunoblot analysis revealed that the introduction of shUQCRHs inhibited the translocation of cytochrome c and the cleavage of PARP (Fig. 6E, Fig. S5E). Although 5-aza-dC treatment decreased the viability of the control OSPa-shNTC cells, this effect was partially attenuated in OSPa-shUQCRH cells (Fig. 6F).
Then, the tumorigenic potentials of OSPa-shNTC and OSP-shUQCRH cells were compared using a mouse renal orthotopic tumor model. When xenografted, OSPa-shUQCRH cells exhibited signi cantly faster primary tumor formation than OSPa-shNTC cells, while lung metastasis was not different in each cell type (Fig. 6G, Fig. S5F). The number of apoptotic cells was lower in tumor tissues derived from OSPa-shUQCRH cells than that from OSPa-shNTC cells (Fig. 6H). These results suggest that UQCRH is essential for the induction of apoptosis and tumor suppression during renal cancer progression.

DNMT inhibitor enhances sensitivity to mTOR inhibitor in ccRCC cells
Finally, the pharmacological effects of a DNMT inhibitor on drug-induced apoptosis were examined.
Treatment with the mTOR inhibitor everolimus successfully inhibited the activation of mTOR and the signaling of its downstream ribosomal protein S6 kinase β-1 (S6K1) in OSPa and OS5K cells equally (Fig.  S6). Nonetheless, everolimus signi cantly induced apoptosis in OSPa cells, and less potently that of OS5K cells (Fig. 7A). When OS5K cells were pre-treated with 5-aza-dC, the everolimus-induced apoptosis was enhanced (Fig. 7B).
The sensitizing effect of 5-aza-dC was further examined in vivo. Mice were orthotopically inoculated with OS5K cells, in which the reduced expression of UQCRH was recovered by treatment with 5-aza-dC in advance as shown in Fig. 5. Although 5-aza-dC pre-treatment did not enhance the effect of everolimus on lung metastasis (at least in the concentration we tested), it signi cantly augmented the therapeutic effect on primary tumor formation (Fig. 7C, D). These results suggest that DNMT inhibitor potently enhances the sensitivity of ccRCC cells to mTOR inhibitors through the recovered expression of UQCRH.

Discussion
Epigenetic alterations are widely recognized in various human cancer cells. [34][35][36][37][38][39] DNA hypermethylation of promoters is induced by the inactivation of DNA demethylases or the overexpression of DNA methyltransferases. [40,41] Additionally, mutations in genes encoding metabolism-related enzymes enhance DNA methylation status. Here, we found the up-regulation of DNMT3B in the highly malignant derivatives of ccRCC cells obtained from serial orthotopic inoculations (Fig. 1A, B). Although we have previously established highly metastatic derivatives of pancreatic cancer cells using a similar strategy, [42] the expression level of DNMTs was not increased in highly metastatic derivates of pancreatic cancer cells (GSE107960), suggesting that the interactions between cancer cells and the renal microenvironment may be crucial for the upregulation of DNMT3B. Several regulatory mechanisms are speculated to be involved in the increased expression of DNMT3 in cancer cells. For instance, interleukin (IL)-6 induced the expression of DNMT3B in a STAT3-dependent manner, resulting in the accelerated proliferation of ccRCC cells or oral squamous cell carcinoma cells. [43,44] Similarly, here, RNA-seq results revealed the enrichment of hallmarks of IL-6-janus kinase (JAK)-signal transducer and the activator of transcription (STAT)3 signaling in OS5Ks (unpublished data). The regulation of DNMTs by microRNAs (miRNAs) has also been described in various cancer cells. [45] Although miR4465 and miR-29c-3p may regulate the expression of DNMT3B, [46,47] this was not evident in OS5K cells (unpublished data).
In addition, recent studies have identi ed many targets for DNMT3B, including tumor-suppressor genes in cancer cells. [48] Colorectal cancer, one of the most common DNA methylated cancers, exhibits decreased expression of the cyclin-dependent kinase inhibitor 2A (CDKN2A) and Ras association (RalGDS/AF-6) domain family member 1 (RASSF1A). [49,50] Cadherin 1 (CDH1), CDKN2A, the runt-related transcription factor 3 (RUNX3), BRCA1, and RASSF1A are reported to be suppressed by DNMT3B-dependent DNA hypermethylation in gastric or breast cancers. [51,52] In the present study, we conducted genome-wide screening of methylation targets in ccRCC cells using RNA-seq analysis and methylation array. UQCRH was chosen as a candidate gene whose expression was epigenetically decreased in highly malignant derivatives in ccRCC cells. Decreased expression of UQCRH was also validated using a tissue array of clinical ccRCC cases (Fig. 4). These observations are in accordance with the results obtained using clinical ccRCC samples in other cohorts, [16] indicating that the downregulation of UQCRH is common in ccRCC. The expression of UQCRH is similarly decreased in several types of cancers. [53] Among them, we found that decreased expression of UQCRH in clear cell carcinoma of ovarian cancer (Fig. S3A). The histological phenotype of clear cell carcinoma is likely to be associated with the downregulation of UQCRH, irrespective of the origin of cancers.
UQCRH is highly conserved in many species and serves as a component of the mitochondrial complex III. It is closely associated with the function of cytochrome c and cytochrome c1. [54][55][56] Although UQCRH is also involved in electron transport and the maturation of cytochrome c1, [57][58][59] its detailed role in the maintenance of the proton gradient in the inner mitochondrial membrane remains unknown. The survival of blood cells was enhanced by the introduction of UQCRH. [60] During the preparation of this manuscript, Luo and colleagues reported a decrease in UQCRH and its functions in the Warburg effect in renal cell carcinoma. [61] Given these reports, we speculate that UQCRH contributes to the induction of apoptosis. In the absence of an apoptotic stimuli, cytochrome c is anchored to the inner mitochondrial membrane by binding to cardiolipin. This binding is attenuated when the inner mitochondrial membrane is oxidized by the accumulation of reactive oxygen species (ROS), resulting in the release of cytochrome c. [62][63][64] Since UQCRH regulates the production of ROS, [58] we hypothesized that UQCRH may be essential for the induction of apoptosis in cancer cells. Accordingly, we showed that the silencing of UQCRH attenuated the translocation of cytochrome c and the apoptosis in ccRCC cells treated with staurosporine ( Fig. 6 and  Fig, S5). These ndings suggest that UQCRH may serve as a potential tumor suppressor in ccRCC through the regulation of apoptosis.
DNMT inhibitors, namely 5-azacytidine and 5-aza-2'-dC, were initially considered for cancer treatment. However, due to their toxicity, only a low dose is recommended. A clinical trial of DNMT inhibitors in combination with other anti-cancer drugs, such as IL-2, interferon (IFN)-α, and bevacizumab, for renal cancer treatment has been conducted. [23] Here, our results demonstrated that the silencing of UQCRH enhances the drug-induced apoptosis in ccRCC cells (Fig. 6) and that the 5-aza-dC treatment sensitizes the anti-tumor effect of mTOR inhibitor both in vivo and in vitro (Fig. 7). Our results contribute to the further understanding of the molecular mechanisms involved in the drug resistance of ccRCC cells and provide an important insight into therapeutic options for ccRCC.

Conclusions
Here, we report the tumor progressive role of DNMT3B in ccRCC. Our results suggest that DNA methylation causes the decreased expression of the potential tumor-suppressor gene UQCRH, which is essential for the completion of the apoptotic process in ccRCC cells. In our preclinical study, DNA demethylation induced by 5-aza-dC enhanced the tumor-suppressive ability of everolimus. These ndings con rms the clinical potential of DNMT inhibitors for ccRCC treatment.

Availability of data and materials
Raw and processed RNA-seq data are available at Gene Expression Omnibus (GEO). The additional data that support the ndings of this study are available from the corresponding author upon reasonable request.