Genome-wide association analysis reveals regulation of at-risk loci by DNA methylation in prostate cancer

Abstract

Epigenetic changes are potentially important for the ontogeny and progression of tumors but are not usually studied because of the complexity of analyzing transcript regulation resulting from epigenetic alterations. Prostate cancer (PCa) is characterized by variable clinical manifestations and frequently unpredictable outcomes. We performed an expression quantitative trait loci (eQTL) analysis to identify the genomic regions that regulate gene expression in PCa and identified a relationship between DNA methylation and clinical information. Using multi-level information published in The Cancer Genome Atlas, we performed eQTL-based analyses on DNA methylation and gene expression. To better interpret these data, we correlated loci and clinical indexes to identify the important loci for both PCa development and progression. Our data demonstrated that although only a small proportion of genes are regulated via DNA methylation in PCa, these genes are enriched in important cancer-related groups. In addition, single nucleotide polymorphism analysis identified the locations of CpG sites and genes within at-risk loci, including the 19q13.2–q13.43 and 16q22.2–q23.1 loci. Further, an epigenetic association study of clinical indexes detected risk loci and pyrosequencing for site validation. Although DNA methylation-regulated genes across PCa samples are a small proportion, the associated genes play important roles in PCa carcinogenesis.

INTRODUCTION

Prostate cancer (PCa) is a common malignancy and leading cause of cancer death among men in the US where it is estimated that approximately 191 930 new PCa cases are expected to be diagnosed in 2020.1 Further, according to a Chinese statistical report, 60 300 PCa cases were confirmed in 72 population-based cancer registries in China from 2009 to 2011, with a mortality of 26 600, resulting in rankings of sixth and eleventh place, respectively, among all cancers affecting men, in China.2 As the Chinese population continues to age, both the incidence and mortality of PCa are expected to continue increasing over the next two decades. PCa cells are known to harbor a variety of genetic defects, including gene mutations and translocations, all of which provide the cells with new capabilities for dysregulated proliferation, immune system evasion, tissue invasion and destruction, inappropriate survival, and metastasis.3 Furthermore, there is abundant evidence that along with genetic changes,4 somatic epigenetic alterations also contribute to PCa carcinogenesis and metastasis.5,6,7,8 Epigenetic gene inactivation in cancer cells is largely based on transcriptional silencing mediated by the aberrant CpG methylation of CpG-rich promoter regions.6,9,10,11,12 DNA methylation is a widely recognized epigenetic marker associated with diagnosis and prognosis in many malignancies.6,13,14 Notably, abnormal DNA methylation has been reported to contribute to the occurrence and progression of PCa.15,16 Previous studies of DNA methylation and PCa risk have found that specific promoter sequences are hypermethylated at a higher frequency in PCa tumor tissues than those in nontumor tissues.4,6,10,17

Quantitative genetics has made a significant progress in revealing the genetic bases of complex traits, especially in developing sophisticated tools to identify the location of genes that impact complex traits. A region of the genome contributing to the variation in a quantitative trait, also known as quantitative trait loci (QTLs), has been used to study gene expression phenotypes (expression quantitative trait loci [eQTLs]) on a massive scale. DNA methylation quantitative trait loci (meQTLs) have been identified in pathological and physiological contexts. The genome-wide gene expression studies can provide information on genetic variation that affects gene expression levels.18 We can use linkage or association mapping to map cis- and trans-acting factors for many genes to explain the inheritance patterns. Previous reports showed that cis/trans-meQTLs could target different CpGs and clarified DNA methylation involvement in diseases and cancers.19,20,21,22,23 Thus, eQTL analysis is a straightforward and popular method for discovering regulatory genome sites24,25,26 and detecting underlying associations across the genome in PCa studies.27,28

The Cancer Genome Atlas (TCGA, http://cancergenome.nih.gov/) is a database that collects multiple types of “omics” from thousands of samples and provides public data access to researchers. Accordingly, modified eQTL methods involving calculations of correlations among gene expression, phenotype, DNA methylation, copy number variations, and single-nucleotide polymorphisms (SNPs) have proven to be a powerful tool.29,30,31 However, genome-wide correlations between gene expression, DNA methylation, and clinical phenotypes in PCa are not yet understood.

Therefore, we performed a meQTL analysis of PCa samples from the TCGA database. Although genes regulated via methylation comprise a minor proportion of the genome (0.8%), these genes are enriched in some gene ontology (GO) groups and in important canonical cancer-related pathways. In our study, we mainly identified meQTL pairs on chromosomes 16q and 19q, which have been reported as high-risk regions for PCa, using SNP analyses. We also identified DNA methylation regions and genes associated with clinical indexes at 11q13 and 16q13 and found several genes regulated by DNA methylation that are important for prognosis. According to our meQTL analyses, we selected some androgen receptor (AR) gene-related CpG sites and some sites that are altered during PCa genesis, and correlated them with Gleason score.

MATERIALS AND METHODS

Data preanalysis

A transcriptome of a prostate adenocarcinoma (PRAD) gene evaluation was downloaded with the TCGA dataset. The nonprimary PCa samples were discarded. Furthermore, the primary tumor samples not providing either gene expression (evaluated with RNA sequencing [RNA-seq]) or DNA methylation information also were excluded from later analysis. As result, a total of 419 samples were used for meQTL identification. Methylation data also were downloaded for the TCGA dataset and were correlated with expression data. Methylation levels at the evaluated sites were estimated using an Illumina Human Infilium 450k BeadChip (Illumina, San Diego, CA, USA). After sample normalization, samples were combined to a methylation matrix according to the gene ID.

meQTL analysis

We used R software MatrixeQTL package32 (version 2.15.1; R Project for Statistical Computing, Vienna, Austria) for eQTL analysis. Each methylated CpG level was regarded as a continuous variable rather than a discrete variable. Correlations between the methylation level of each CpG site and each gene were evaluated using MatrixeQTL. As DNA methylation mostly influences the expression of genes via promoter regions, we distinguished cis- and trans-regulation for further analysis. The gene and methylation CpG sites were extracted from Illumina Human Infilium 450k BeadChip annotation files and University of California, Santa Cruz (UCSC) reference gene list locations. For reference genes, the gene location was counted from the transcription start site to terminal site. We set a P value threshold of 1 × 10⁻⁵ for cis-regulated and 1 × 10⁻⁶ for trans-regulated eQTLs. Cis-regulated meQTLs were defined as interacting pairs with tested DNA methylation sites and genes <1 megabase (MB), whereas trans-regulated meQTLs were defined as pairs with genes >1 MB or located on other chromosomes.

Tissue specimens and bisulfite modification of DNA

A total of 70 PCa patients who underwent radical prostatectomy at Xinhua Hospital affiliated to Shanghai Jiao Tong University School of Medicine (Shanghai, China), between July 2014 and January 2017, were enrolled in this study. Informed consent was obtained from all subjects. All cases were histologically confirmed and had clinical stage II or III tumors, no clinical evidence of lymph node or distant metastasis, available pathology specimens, and complete clinical and serum prostate-specific antigen (PSA) data. Patients with missing any variable needed to accurately assign them to a risk group (PSA, tumor [T] stage, or Gleason score) were excluded. Patients with missing information on any available demographic variables were excluded. All samples were retrieved from the archive of the Institute of Pathology, Xinhua Hospital, and were anonymously analyzed in accordance with the guidelines of the Ethics Committee of Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine. We confirm that all experimental protocols were approved by the local ethics committee (approval No. XHEC-D-2016-005).

Patients were categorized as low-, moderate-, or high-risk PCa based on the 2015 National Comprehensive Cancer Network guidelines (Table 1). All 70 cases included tumor tissues and adjacent nontumor tissues. Prostate tissue samples from these cases were obtained at the time of radical prostatectomy, and tumor cell contents were determined to exceed 70.0% of all tissues. Tissue microdissection yielded tumor and adjacent nontumor tissues from which DNA was extracted using an FFPE DNA Kit (Omega Bio-Tek, Norcross, GA, USA) according to the manufacturer's instructions. Extracted DNA concentrations were measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, USA) and subsequently adjusted to approximately 40 ng ml⁻¹. Prepared DNA was subjected to bisulfite treatment using the EZ DNA Methylation-Gold kit (ZYMO, Orange, CA, USA); converted DNA was dissolved in Tris-EDTA (TE) buffer and stored at or below −20°C for later use.

Table 1.

Patient baseline characteristics

Variable	Low risk	Moderate risk		High risk
Patient (n)	10	34	14	12
Age (year), mean±s.d.	65.4±7.7	67.3±9.5	65.6±8.7	68.1±12.3
PSA (ng ml-1), mean±s.d.	8.6±1.2	15.4±2.4	17.2±1.7	32.7±10.5
Gleason score	≤ 6 (3+3)	7 (3+4)	7 (4+3)	≥ 8 (4+4, 4+5, 5+3, 5+4)
Clinical T-stage	≤ T2a	T2b	T2b	≥ T2c
Clinical N-stage	N0	N0	N0	N0
Clinical M-stage	M0	M0	M0	M0

Patients who met the study criteria with 14.3%, 68.6%, and 17.1% were classified as low, moderate, and high risk, respectively. PSA: prostate-specific antigen; T: tumor; N: node; M: metastasis; s.d.: standard deviation

Primer design and PCR

Specific primers were designed for 30 loci using the publicly available MethPrimer software package (http://www.urogene.org/methprimer/). A complete list of the primer pairs is available in Supplementary Table 1. Biotinylated reverse primers were substituted with 5′-tailed unlabeled reverse primers (aaccttcaacaccccaaccatata), allowing single (expansive) biotinylated primers to be used for subsequent pyrosequencing™. All primers and tag sequences were provided by Sangon (Shanghai, China).

Supplementary Table 1.

The primers of the polymerase chain reaction amplification and pyrosequencing

Sites	F1	F2
cg14338887, chr6:42928000,42929000	ATTTGTTATTGGTTAGGTGGGGT	TTGTTAGTAGTGTTTATGTTTTAAGTG
cg02213678, chr16:85112035,85113035	TTGGTTATGGTTAGTAGAAAAGTTA	TTGGTTATGGTTAGTAGAAAAGTTA
cg09207578, chr13:114781449,114782449	GTTTTAGGTTTGGTTGAGATGTAGATT	GTTTTAGGTTTGGTTGAGATGTAGATT
cg15492820, chr11:61322613,61323613	TTTTTAAAGTTTTATTATATTTTTGGGT	TTTTTAAAGTTTTATTATATTTTTGGGT
cg12918275, chr17:18150454,18151454	GTTATAGGGGTTAGGGTATGTTAGG	GTTATAGGGGTTAGGGTATGTTAGG
cg08170140, chr22:31090247,31091247	GATTTAGGAAGGATTTAGGGTTTAAATATGG	GATTTAGGAAGGATTTAGGGTTTAAATATGG
cg20664996, chr17:17359321,17360321	TATAATGAAGTGTGTATTAYGGGAGYGT	TTTTAAGTTGATTTTTGGTTGTAGTG
cg09093477, chr5:34838689,34839689	GAGATATAATATGTAGAGGTTAGTTTTGGGG	TTTAATTTGYGGGGAAGGGAAAGYGGAG
cg14927321, chr15:89491114,89492114	GAAGTAATTAGTATATTAATGTGGTATTTTGGG	TGGGGGGTGATATTTTTTGGGT
cg15854606, chr19:18654048,18655048	TTGGTTTYGTTTTTGTYGTGTTTATTG	TTGGTTTATGTAGAYGTTYGTTTYGGG
cg186688 36, chr2:113102684,113103684	GTTAGTTTTTAATTTTTGGGTTTAGGGATT	GTATTGTATTAATTTTATTTATATATAGGTTTGTAT
cg19283427, chr18:29522880,29523880	GAGGTGTATAATTGTAGGTYGGGTTTTT	AAGTTYGTTTTTAGTAGAGTAGAGTTGAGAG
cg06577658, chr19:46295770,46296770	TTYGTTTATAGGTGTTGTTAGAGTTTT	GATTTAGGAAGGATTTAGGGTTTAAATATG
cg23158051, chr2:65454192,65455192	TTAGTTTTTTYGTTATATTTTATTGTYGTAGTAG	TGTTTGYGTTTTTTGTAGTTYGTTAAGTAG
cg04744134, chr19:8407628,8408628	ATTTTTAATGTGGGAGGGGTTT	GGTAGAGTTTTTAGTGGGGTTGGAAA
cg18770350, chr1:236849466,236850466	TGGGTTYGTTTGTTAGTTAGTTYGTG	TGGGTTYGTTTGTTAGTTAGTTYGTG
cg23139584, chr8:56987006,56988006	TTGAGTTTAGTTGATATGAGTATTGGTAG	TTGAGTTTAGTTGATATGAGTATTGGTAG
cg05407490, chr7:23749167,23750167	TTAAATTTAYGTGAATAGATTAGGAAAGATTA	TTAAATTTAYGTGAATAGATTAGGAAAGATTA
cg23410069, chr22:20287122,20288122	TAGTTTGTTGGTGTTGYGTTTTTG	TTGTTGGYGYGTTAYGAGTAGGTG
cg05754508, chr12:65672031,65673031	TTTTATGTTTATATTTGGATTYGGTTGTT	AGGAATTTTTYGTTATGTTTTTYGTYGT
cg15850918, chr15:89900241,89901241	TTTAGGTTAGATTAGGGATTAGTTTT	AAGGGAGGATGAAGGATATAGAAG
cg25902616, chr11:47788693,47789693	AAGGGTGAGAGTTTAGTTAATAGTG	AAGGGTGAGAGTTTAGTTAATAGTG
cg23303108, chr8:23083078,23084078	AAGAGTTTTGAAAGTTTTTAAATTA	GGGGATAGGTTTTTTTAGYGGT
cg27069132, chrX: 135991043,135992043	GAAATTTTTAGGATAGTTTTTTTT	AGGYGGGGAAGTTTGGGGAT
cg09804858, chr6:42396621,42397621	GAGTATTGGTGTTTGGAGAAATTTT	TGGATGGATTTTGYGAGGTTTGG
cg14338887, chr6:42928000,42929000	aaccttcaacaccccaaccatataCAACCCAAACAACCATACCTTATAC	aaccttcaacaccccaaccatata	AGTGTTTATGTTTTAAGTGYGG
cg02213678, chr16:85112035,85113035	aaccttcaacaccccaaccatataATACAAATAAATCTTCTCCCCCTTC	aaccttcaacaccccaaccatata	AGTGGGTGYGTAATTAGAGTTGTTTTTAG
cg09207578, chr13:114781449,114782449	aaccttcaacaccccaaccatataCCAAAATTACCCCCTTACAACTAA	aaccttcaacaccccaaccatata	TAYGGTGAGGTYGGATTGTAGGGTTAT
cg15492820, chr11:61322613,61323613	aaccttcaacaccccaaccatataACCCAAAACCATACATCTAAAAACA	aaccttcaacaccccaaccatata	TTGGGYGGGGYGGYGYGATG
cg12918275, chr17:18150454,18151454	aaccttcaacaccccaaccatataCCCTCAAACTAAATCAAAAACTATAC	aaccttcaacaccccaaccatata	AGGGTATGTTAGGGTGYGAAGGAT
cg08170140, chr22:31090247,31091247	aaccttcaacaccccaaccatataCTTTCCCCATAAAACCRACCRACC	aaccttcaacaccccaaccatata	TATGGTTYGAGGTTGGTYGAGAG
cg20664996, chr17:17359321,17360321	aaccttcaacaccccaaccatataATTTAAAATAAATACCCATCCAAAC	aaccttcaacaccccaaccatata	TGTAGTGGTTTGGGTTTGTGTGG
cg09093477, chr5:34838689,34839689	aaccttcaacaccccaaccatataCAACCCAAATCCRACCAAAAAAACAATAC	aaccttcaacaccccaaccatata	AGGGTAAGGTAAGGGTTGTTGTAGG
cg14927321, chr15:89491114,89492114	aaccttcaacaccccaaccatataACACCACCCTATAAAATTACTCATTAAC	aaccttcaacaccccaaccatata	GTTTGGTAGTTTAATTTTTTTATTGTA
cg15854606, chr19:18654048,18655048	aaccttcaacaccccaaccatataTATTAAAATCCAACCCAAAATCCCC	aaccttcaacaccccaaccatata	TTAAAGGGATYGAAGTTTTTYGGTTG
cg186688 36, chr2:113102684,113103684	aaccttcaacaccccaaccatataTATTTCCTCCAACTTTCTAATTATTTAAAAA	aaccttcaacaccccaaccatata	TATAGGTTTGTATTTTTTTGTTAAAATT
cg19283427, chr18:29522880,29523880	aaccttcaacaccccaaccatataCCAAATCRATTACCCACAACAACCAAAAA	aaccttcaacaccccaaccatata	AGTAGAGTAGAGTTGAGAGTTATTTA
cg06577658, chr19:46295770,46296770	aaccttcaacaccccaaccatataCTTTCCCCATAAAACCRACCRACC	aaccttcaacaccccaaccatata	GGGTTTAAATATGGTTYGAGGTTGGTYG
cg23158051, chr2:65454192,65455192	aaccttcaacaccccaaccatataATAAACTCCAATAAAAAATAAAAAAAC	aaccttcaacaccccaaccatata	AGTTTTTGTGATATTTTAGGGAGAAAG
cg04744134, chr19:8407628,8408628	aaccttcaacaccccaaccatataCAATCACCACCCCTACATTCCAAACCC	aaccttcaacaccccaaccatata	GTGYGTTTTYGAATTTATTTATTTAGGGG
cg18770350, chr1:236849466,236850466	aaccttcaacaccccaaccatataTACTTCTCCCAAACTAAATCCAAAAAC	aaccttcaacaccccaaccatata	TTYGTGYGTTYGAGTTTTTYGYGTTT
cg23139584, chr8:56987006,56988006	aaccttcaacaccccaaccatataCCTTTCCAAAAATAAATATAAATACTTCTACTAAAC	aaccttcaacaccccaaccatata	GTTGATATGAGTATTGGTAGAAATGTTT
cg05407490, chr7:23749167,23750167	aaccttcaacaccccaaccatataCAAAAACCCCACACTACRCCTA	aaccttcaacaccccaaccatata	GTTTGAAGTTTTGYGAAAGGAATTGG
cg23410069, chr22:20287122,20288122	aaccttcaacaccccaaccatataCCTCCCTTCCTAACTTCCCTTAC	aaccttcaacaccccaaccatata	GGGTTTTTGTTYGTTTYGGGAG
cg05754508, chr12:65672031,65673031	aaccttcaacaccccaaccatataAACTACRCACTTCCCAAAAC	aaccttcaacaccccaaccatata	TTYGGAGTYGGGGAGGGAGGGAG
cg15850918, chr15:89900241,89901241	aaccttcaacaccccaaccatataCCTCTAAAATATCTTTCAAATTCTC	aaccttcaacaccccaaccatata	GGTTAGTTAGGTTTATTTAGTTTTTAGG
cg25902616, chr11:47788693,47789693	aaccttcaacaccccaaccatataAATAAAAAACAAATAAACAAAAAAAA	aaccttcaacaccccaaccatata	AGTGYGGTAGGGTTTGAATTAGAA
cg23303108, chr8:23083078,23084078	aaccttcaacaccccaaccatataAAAAAAAATAAACATCAATCAAC	aaccttcaacaccccaaccatata	TTTGTATTTAGYGTTGTAYGGG
cg27069132, chrX: 135991043,135992043	aaccttcaacaccccaaccatataACTAATAACACTATAACTCCCAAAAC	aaccttcaacaccccaaccatata	GTTYGGTTYGGYGGGGATTTTTA
cg09804858, chr6:42396621,42397621	aaccttcaacaccccaaccatataAACAAAAAATACTTCCTTCCATTTC	aaccttcaacaccccaaccatata	GAGGTTTGGGAAGTTTTATTTATTTTA

F: forward primer; R: reverse primer; SEQ: sequencing primer

PCR amplification of DNA was conducted using a nested-PCR protocol, using the primers shown in Supplementary Table 1. Two rounds of amplification reactions were performed. In the first round, a total reaction volume of 10 μl contains 1 mmol l⁻¹ first primers (1 μl), bisulfite converted DNA (1 μl), ×2 PCR master mix (5 μl), and ddH₂O (2 μl). In the second round, the total reaction volume was 30 μl: 4 μl of first-round PCR product as template, plus 10 mmol l⁻¹ second primers (0.5 μl), ×2 PCR Master Mix (15 μl), and ddH₂O (10 μl). PCR conditions were as follows: 94°C for 5 min, 30 cycles of denaturation at 94°C for 30 s, annealing at 55°C for 40 s, elongation at 72°C for 30 s, and an additional elongation step at 72°C for 7 min. Approximately 8 μl of each PCR product was separated by electrophoresis using a 3% agarose gel stained with GelRed (Invitrogen, Carlsbad, CA, USA) for 40 min at 120 V and visualized using a Gene Genius bio-imaging system (Syngene, Cambridge, UK).

Pyrosequencing™ methylation analysis

PCR products (20 μl) were added to a mix comprising Streptavidin Sepharose HP™ (3 μl; GE Healthcare, Dornstadt, Germany) and binding buffer (37 μl; Qiagen, Hilden, Germany). The contents were mixed at 2000g (Centrifuge 5810R, Eppendorf, Hamburg, Germany) for 10 min at room temperature. Using the Vacuum Prep Tool™ (Qiagen), according to the manufacturer's instructions, single-stranded PCR products were prepared. Sepharose beads with attached single-stranded templates were released into a PSQ 96 Plate Low™ (Qiagen) containing a mix of 40 μl annealing buffer (Qiagen) and the corresponding sequencing primer at 400 nmol l⁻¹ (Supplementary Table 1). Pyrosequencing™ reactions were performed in a PyroMark ID System (Qiagen), according to the manufacturer's instructions, using the PyroMark Gold 96 Reagent Kit (Qiagen). CpG site quantification was performed using Pyro Q-CpG™ methylation software (Qiagen).

Statistical analyses

The cis- and trans-meQTL sites were identified using R package “MatrixeQTL”. The P values were adjusted using Bonferroni method, and cis-meQTLs with false discovery rate (FDR) <1 × 10⁻⁵ and trans-meQTLs with FDR <1 × 10⁻⁶ were identified as meQTLs. Association significance between clinical indexes and gene expression/methylation levels was identified using analysis of variance (ANOVA) test (P < 0.05 being considered as significant). Gene expression comparison between recurrence and nonrecurrence group was performed using Student's t-test, and P < 0.05 was considered as statistically significant.

RESULTS

Methylated meQTLs in PCa

We subjected PCa samples including 419 samples from TCGA to a meQTL analysis using expression levels evaluated via RNA-seq and CpG site methylation levels in CpG island (CGI) regions. Among 20 321 genes, the methylation levels of 485 513 CpG loci were used for meQTL identification. We identified 5852 cis-regulation and 5 156 662 trans-regulation meQTL pairs in our dataset. Cis-regulation pairs included 1717 genes (8.4% of all genes tested) and 4895 corresponding CpG sites, with an FDR (Bonferroni) of <0.01. Among these pairs, 784 genes (45.7%) were regulated by multiple CpG sites, and 1661 CpG sites (33.9%) regulated multiple genes. We noticed that adenosine triphosphate (ATP)-binding cassette subfamily A member 17, pseudogene (ABCA17P), chromosome 11 open reading frame 85 (C11orf85), and cyclic adenosine monophosphate (cAMP) responsive element binding protein 3 like 3 (CREB3L3) were regulated according to the methylation levels of 135, 95, and 93 nearby CpG sites, respectively, which are the top three CpG sites-related genes, suggesting the importance of DNA methylation on these genes. We also found that significantly high numbers of CpG sites in meQTLs were located on 19q13.2–q13.43 (P = 1 × 10⁻¹¹) and 16q22.2–q23.1 (P = 1 × 10⁻¹¹), as shown in Figure 1 and Supplementary Table 2, consistent with previous reports.33,34 According to GATHER (http://gather.genome.duke.edu/), the genes with significant involvement are located at 19q13, 16p13, 11q13, and 17q25 (Supplementary Table 3). Through a GO and pathway analysis using the database for annotation, visualization, and integration discovery (DAVID),35 we discovered that although most GO groups appeared stochastic, several signaling pathways associated with PCa, including the Janus kinase/signal transducer and activator of tran-ions (JAK/STAT), vascular endothelial growth factor (VEGF), and cytokine–cytokine interactions, also are involved (Supplementary Table 4 and 5). Finally, the analysis of trans-regulation pairs identified 18 999 genes (93.5% of tested genes) and 239 808 CpG sites, with an FDR of <1 × 10⁻⁶. In summary, SNP analysis identified the locations of CpG sites and genes within at-risk loci, including the 19q13.2–q13.43 and 16q22.2–q23.1 loci.

Figure 1.

Circos plot showing the CpG density of eQTLs. The density of CpG sites associated with eQTLs (from inside to outside) across the genome. Each wedge represents a chromosome and different colors show different cytobands. In the inner track, from inner to outer, the fold line represents the values. Higher values mean that more meQTL-related CpG sites are over-present in these regions. eQTLs: expression quantitative trait loci; meQTL: methylation quantitative trait loci.

Supplementary Table 2.

The CpG sites were significantly enriched on some cytobands

Locus	Total genes with annotation	Genes involved	LOG10 P	Genes
19q13	110	110	11.19	A1BG APOC1 APOE ATP4A BCL2L12 CABP5 CAPN12 CEACAM4 CGB1 COX7A1 CPT1C CRX CYP2A13 CYP2B6 CYP2B7P1 CYP2F1 DBP DLL3 DMPK DPF1 EHD2 FLJ16165 FLJ23569 FLJ26850 FLJ32658 FLJ40125 FLJ40235 FLJ40321 FOSB FUT2 GIPR GPR32 GPR77 GRIN2D HSPB6 IGFL1 IL28B IL29 IL4I1 KCNA7 KCNJ14 KIR3DL3 KLC2L KLK12 KLK13 KLK14 KLK8 KLK9 LGALS14 LIG1 LMTK3 LOC112703 LOC126147 LOC147645 LOC199800 LOC342900 LOC388550 LOC440533 LRRC4B MAG MGC17986 MGC34799 MUC16 MYADM MYBPC2 NALP13 NALP4 NALP9 NANOS2 NAPSB NKG7 NKPD1 NPHS1 NUMBL NYD-SP11 PPP1R13L PPP1R15A PRKCG PRODH2 PSG1 PSG11 PSG3 PSG4 PSG7 R30953_1 RCN3 RPL18 RPS19 RSHL1 RUVBL2 SELV SIPA1L3 SLC6A16 SLC7A10 SPACA4 SPIB SYCN SYT3 TEX101 TIP39 TNNI3 TNNT1 UNQ467 VN1R4 ZFP36 ZNF30 ZNF331 ZNF342 ZNF473 ZNF541
16p13	47	47	11.19	C16orf35 C16orf5 C1QTNF8 CASKIN1 DNASE1L2 EME2 FBXL16 FLJ32252 FLJ34512 GNG13 HAGH HAGHL HBA1 HBAP2 HBQ1 HBZ IGFALS IL32 KIAA1924 KREMEN2 LOC146562 LOC283951 LOC342346 LOC400500 MGC2494 MMP25 MPN Magmas NME4 NOD3 NOXO1 PAQR4 PDIP PRM1 PRM2 PRM3 PRSS21 RAB11FIP3 RAB40C RGS11 RNU64 RPL3L SOCS1 SYNGR3 TNP2 TPSD1 WFIKKN1
11q13	46	46	11.19	ACY3 AIP B3GNT6 Bles03 CABP2 CABP4 CD5 CFL1 CST6 ESRRA FBXL11 FGF19 FGF3 FGF4 FIBP FLJ33790 FOLR1 FOSL1 GAL3ST3 GIF GPHA2 GPR152 HTATIP KIAA1394 LOC144097 LOC283129 LOC387778 MGC11102 MGC20410 MTL5 NXF OVOL1 SCGB1D1 SCGB1D2 SCYL1 SLC22A12 SLC29A2 SPTBN2 SSSCA1 SUV420H1 TNFRSF19L TSARG6 TSGA10IP TncRNA WNT11 YIF1
17q25	32	32	11.19	ANAPC11 CARD14 CBX8 DKFZP434P0316 EVPL EXOC7 FADS6 FASN FLJ31882 FLJ35767 FSCN2 GCGR GRIN2C HCNGP IREM2 LOC147111 LOC255275 LOC284001 LOC92659 MGC29814 OTOP3 PDE6G RAC3 RFNG RNF157 SECTM1 SPHK1 SRP68 USP36 UTS2R WBP2 ZC3HDC5
17q21	37	37	9.24	AOC2 C1QL1 CA10 CCR10 DLX3 DLX4 FLJ33318 FLJ40137 FLJ40342 G6PC GCN5L2 GIP GSDM1 HOXB1 HOXB2 HOXB7 HOXB8 IMP-1 IMP5 KRT20 LOC284067 LOC388389 LOC388394 MGC16309 MRPL10 NAGS PHOSPHO1 PLCD3 PNPO PPY PRAC PYY ProSAPiP2 SCAP1 TBX4 TREM4 ZNFN1A3
19p13	61	61	8.19	ANKRD24 ANKRD25 APC2 BRUNOL5 C19orf23 C19orf30 CAPS CASP14 CD37 CREB3L3 DF DNAJB1 EMR2 EPOR FLJ11535 FLJ25758 FLJ32416 FLJ35784 FLJ39369 FLJ40365 FLJ45910 FUT5 FUT6 GRIN3B HCN2 IER2 IL27RA ISYNA1 KIAA1086 LOC126536 LOC199675 LOC255809 LOC390874 LRG1 MADCAM1 MBD3L1 MGC15631 MGC17791 MGC23244 MGC24975 MGC39581 NANOS3 NRTN OR10H4 OR1I1 OR1M1 OR7A17 OR7C2 OR7D4 PCP2 PKN1 RAB3A RPL36 SAFB2 SHD SIRT6 THSD6 TIMM13 TMPRSS9 UNC13A VMD2L1
1q21	37	37	7.18	ANKRD34 ANXA9 APCS APOA2 AQP10 ATP8B2 C1orf45 CRP FCRH1 FCRH3 FLJ20519 FLJ37964 HAX1 IRTA1 IRTA2 KCNJ9 KCNN3 LCE1A LCE1D LCE2B LCE2C LCE3A LCE3E LCE4A LOC388698 LRRN6D MCSP MUC1 RHBG S100A2 S100A3 SH2D2A SPAP1 SPRR2A SPRR2B SV2A THHL1
12q13	33	33	5.95	AMHR2 AQP6 CACNB3 CYP27B1 FMNL3 GEFT HNRPA1 HOXC11 HUMCYT2A IGFBP6 IL23A INHBC ITGB7 K6IRS4 KRT6E KRTHB3 KRTHB5 KRTHB6 LETMD1 MGC17301 MLC1SA NXPH 4 OR10P1 OR6C1 OR6C65 OR6C68 OR6C74 OR9K2 SLC4A8 STAT2 TAC3 WNT10B ZNFN1A4
11p15	43	43	5.67	ASCL2 BRSK2 CALCA CNGA4 DKFZp761L1518 HBE1 HBG2 HCCA2 HSD-40 IFITM1 IFITM3 IGF2 IGF2AS KCNQ1DN KRTAP5-6 LDHA LOC387733 LOC439915 MRGX3 MRVI1 MUC2 MUCDHL NALP6 ODF3 OR51B4 OR51B5 OR51B6 OR51D1 OR51F2 OR51I2 OR52A1 OR52E6 OR52E8 OR52K1 OR52N1 OR52N2 OR56A4 OR56B1 SAA3P SYT9 TH USH1C ZDHHC13
20q13	29	29	5.14	BHLHB4 C20orf111 C20orf166 C20orf58 C20orf59 C20orf67 CBLN4 DNTTIP1 EDN3 FLJ30313 GATA5 GPR8 HRH3 KCNG1 KCNQ2 LOC198437 MC3R MYT1 NTSR1 PHACTR3 PPGB RIMS4 SGK2 SLC35C2 STMN3 SYCP2 UBE2C WFDC10A WFDC5

Supplementary Table 3.

Gene cytogenetics band enrichment loci of methylation quantitative trait loci-related genes according to a gene annotation tool to help explain relationships (http://gather.genome.duke.edu/)

Annotation	Total genes with annotation	Your genes (with annotation)	Your genes (no annotation)	Genome (with annotation)	Genome (no annotation)	Ln (Bayes factor)	Negative ln (P)	FE: negative ln (P)	FE: negative ln (FDR)
19q13	110	110	1494	857	27869	24.38	11.19	31.16	25.64
16p13	47	47	1557	307	28419	11.95	11.19	18.82	13.99
11q13	46	46	1558	320	28406	9.91	11.19	16.76	12.34
17q25	32	32	1572	184	28542	8.85	11.19	15.8	11.67
17q21	37	37	1567	276	28450	5.53	9.24	12.33	8.42
19p13	61	61	1543	598	28128	4.39	8.19	10.96	7.23
1q21	37	37	1567	306	28420	3.57	7.18	10.28	6.71
12q13	33	33	1571	277	28449	2.35	5.95	9.08	5.64
11p15	43	43	1561	410	28316	2.09	5.67	8.7	5.38
20q13	29	29	1575	240	28486	1.64	5.14	8.37	5.15

FE: Fisher’s exact test; FDR: false discovery rate

Supplementary Table 4.

Methylation quantitative trait loci related gene-enriched gene ontology terms according to database for annotation, visualization, and integration discovery

Term	Count	Percentage	P	List total	Pop hits	Pop total	Fold enrichment	Bonferroni	Benjamini	FDR
GO: 0007186~G-protein coupled receptor protein signaling pathway	179	10.49853372	9.80E-19	1121	1123	13528	1.923540154	3.27E-15	3.27E-15	1.78E-15
Signal peptide	399	23.40175953	4.18E-18	1588	3250	19113	1.477637473	1.58E-14	5.28E-15	7.72E-15
Topological domain: Extracellular	330	19.35483871	9.83E-14	1588	2719	19113	1.460774214	3.73E-10	9.32E-11	1.82E-10
Membrane	599	35.13196481	5.76E-06	1594	6256	19235	1.155404237	0.003759097	2.90E-04	0.008605043
Cytokine	34	1.994134897	1.32E-05	1594	181	19235	2.266753087	0.008565601	6.14E-04	0.019654069
GO: 0007267~cell–cell signaling	76	4.457478006	1.92E-04	1121	600	13528	1.528587571	0.472641555	0.044676586	0.349044181
GO: 0051302~regulation of cell division	13	0.762463343	3.35E-04	1121	47	13528	3.33790119	0.672253793	0.063512001	0.607711083
Growth factor	24	1.407624633	4.48E-04	1594	131	19235	2.210771308	0.254263894	0.010807233	0.668115409

GO: gene ontology; FDR: false discovery rate

Supplementary Table 5.

Methylation quantitative trait loci-related gene-enriched signaling pathways

Category	Term	Count	Percentage	P	List total	Pop hits	Pop total	Fold enrichment	Bonferroni	Benjamini	FDR
KEGG_PATHWAY	hsa04740:Olfactory transduction	71	4.164222874	1.22E-09	458	379	5085	2.079910359	2.14E-07	2.14E-07	1.49E-06
KEGG_PATHWAY	hsa04080:Neuroactive ligand-receptor interaction	44	2.580645161	3.13E-05	458	256	5085	1.908262828	0.005497044	0.00275231	0.038425378
KEGG_PATHWAY	hsa04060:Cytokine-cytokine receptor interaction	44	2.580645161	5.55E-05	458	262	5085	1.864562152	0.009719381	0.003250347	0.068074717
KEGG_PATHWAY	hsa05320:Autoimmune thyroid disease	12	0.703812317	0.004747675	458	51	5085	2.612381197	0.56724373	0.188925323	5.672613309
KEGG_PATHWAY	hsa00590:Arachidonic acid metabolism	12	0.703812317	0.009864494	458	56	5085	2.379132876	0.825314991	0.294574544	11.45423332
KEGG_PATHWAY	hsa04610:Complement and coagulation cascades	13	0.762463343	0.018625588	458	69	5085	2.091797987	0.963447888	0.423916532	20.60338795
KEGG_PATHWAY	hsa04672:Intestinal immune network for IgA production	10	0.586510264	0.028282991	458	49	5085	2.265840834	0.993587821	0.513911784	29.67691858
KEGG_PATHWAY	hsa04630:Jak-STAT signaling pathway	22	1.290322581	0.035933062	458	155	5085	1.575855754	0.998404616	0.552947878	36.17726313
KEGG_PATHWAY	hsa05310:Asthma	7	0.410557185	0.040786546	458	29	5085	2.67994278	0.999343725	0.557062045	40.01015413
KEGG_PATHWAY	hsa04950:Maturity onset diabetes of the young	6	0.351906158	0.06779727	458	25	5085	2.664628821	0.999995696	0.70934184	57.74731756
KEGG_PATHWAY	hsa04370:VEGF signaling pathway	12	0.703812317	0.07000343	458	75	5085	1.776419214	0.999997164	0.686886675	58.95815416
KEGG_PATHWAY	hsa04640:Hematopoietic cell lineage	13	0.762463343	0.081608456	458	86	5085	1.678303036	0.999999689	0.713092311	64.8191755
KEGG_PATHWAY	hsa04664:Fc epsilon RI signaling pathway	12	0.703812317	0.087535029	458	78	5085	1.708095398	0.9999999	0.710673264	67.50599408

FDR: false discovery rate; KEGG: Kyoto Encyclopedia of Genes and Genomes

Gene/CpG methylation and clinical information

To identify the mechanism by which these CpG sites affect carcinogenesis and development, we performed a correlation analysis and an ANOVA between gene expression levels and clinical indexes. The evaluated genes included 20 532 genes from the TCGA database that were evaluated via RNA-seq and correlated with clinical information, including primary and secondary Gleason score, node invasion stage, biochemical recurrence indicators, most recent PSA level, and clinical and pathological primary tumor stages. Each clinical index was associated with a number of genes, as shown in Table 2.

Table 2.

Number of genes identified significantly associated with clinical observations

Clinical variable	Gene (n)
Primary Gleason score	376
Secondary Gleason score	13
Gleason score	726
Nodes examined (entered N stage)	58
Biochemical recurrence indicator	14
PSA level	25
Clinical T stage	68
Pathological T stage	96

N: node; PSA: prostate-specific antigen; T: tumor

Genes associated with the same clinical information may affect each other. We further performed a similar analysis to compare methylation sites and clinical information and identified 57 719 CpG sites significantly correlated with the aforementioned clinical information (Table 3).

Table 3.

Number of CpG sites significantly associated with clinical observations

Clinical variable	CpG sites (n)
Primary Gleason score	4087
Secondary Gleason score	6518
Gleason score	11 321
Nodes examined (entered N stage)	2271
Biochemical recurrence indicator	7667
PSA level	5350
Clinical T stage	2583
Pathological T stage	11 615

N: node; PSA: prostate-specific antigen; T: tumor

Methylation sites influence clinical performance by regulating gene expression

A canonical function of DNA methylation is the blockade of transcription factors from binding gene regulatory elements, thus inhibiting gene expression. To reduce the FDR and determine causality, we sought CpG sites that were significantly correlated with gene expression via cis-regulation according to methylated QTLs. These genes were positively correlated further with clinical information (Supplementary Table 6), and the CpG sites also were positively correlated with the same clinical indexes. We identified 92 paired methylated QTLs that fulfilled the aforementioned criteria. Most methylated QTLs were associated with Gleason score; a DAVID-based GO analysis36 revealed that these genes were significantly enriched for DNA repair and cell cytoskeleton. Most of the sites associated with Gleason score were located in 11q13 and 16q13, and their correlated genes included essential meiotic structure-specific endonuclease subunit 2 (EME2), potassium channel tetramerization domain containing 13 (KCTD13), kelch like family member 17 (KLHL17), and WD repeat domain 90 (WDR90). Of these, EME2 is associated with genomic stability maintenance, whereas reports of the other genes in the context of all types of cancer are scarce.

Supplementary Table 6.

CpG sites and gene expression significantly associated with clinical information

SNPs	Gene	Statistic	P	FDR	Beta	CHR	MAPINFO	Strand	PSA_most_recent_results	P of methyl-clinical	Clinical Type 1	P gene_exp-clinical	Clinical Type 2
cg09906402	8519	−7.909108707	2.35E-14	1.99E-10	−0.002515419	11	1033236	R	2.77E-06	0.917349226	PSA_recent_methyl	0.917349226	PSA_recent
cg00027499	253980	4.429630563	1.21E-05	0.0170886	0.000321812	16	30759507	F	0.002535845	0.02366169	gleason_methyl	0.02366169	gleason
cg00086266	339451	4.122329189	4.53E-05	0.048383444	0.00021161	1	1242895	R	0.000152682	0.128466513	gleason_methyl	0.128466513	gleason
cg00112517	7153	4.263246691	2.49E-05	0.030359331	3.74E-05	17	37783011	R	0.00374	0.973106811	pathological_T_methyl	0.287392771	clinical_T.x
cg00112517	7153	4.263246691	2.49E-05	0.030359331	3.74E-05	17	37783011	R	1.15E-05	0.620386184	gleason_methyl	0.620386184	gleason
cg00112517	7153	4.263246691	2.49E-05	0.030359331	3.74E-05	17	37783011	R	2.46E-06	0.591742547	primary_gleason_methyl	0.591742547	primary_gleason
cg00139807	253980	5.567943716	4.62E-08	0.000144184	5.38E-05	16	29874869	F	0.002535845	0.002673337	gleason_methyl	0.002673337	gleason
cg00905951	5347	11.21066444	1.13E-25	2.40E-21	0.000148888	16	23652875	F	0.00174	0.189760603	pathological_T_methyl	0.130660172	clinical_T.x
cg00905951	5347	11.21066444	1.13E-25	2.40E-21	0.000148888	16	23652875	F	3.28E-06	0.211537011	gleason_methyl	0.211537011	gleason
cg00905951	5347	11.21066444	1.13E-25	2.40E-21	0.000148888	16	23652875	F	6.80E-08	0.441059277	primary_gleason_methyl	0.441059277	primary_gleason
cg00981594	23204	4.019791988	6.91E-05	0.066194509	0.000530386	16	18813271	R	0.000118541	0.004915781	gleason_methyl	0.004915781	gleason
cg00981594	23204	4.019791988	6.91E-05	0.066194509	0.000530386	16	18813271	R	0.002053963	0.010918013	primary_gleason_methyl	0.010918013	primary_gleason
cg01110955	57524	−4.866450486	1.61E-06	0.003217512	−1.34E-05	16	1706924	F	0.001185047	0.765035156	biochemical_recurrence_methyl	0.765035156	biochemical_recurrence
cg01110955	146330	−5.156566599	3.89E-07	0.000937772	−0.00011792	16	1706924	F	0.000607564	0.004459742	gleason_methyl	0.004459742	gleason
cg01110955	197342	−5.228939751	2.70E-07	0.000684276	−4.10E-05	16	1706924	F	0.000164007	0.004459742	gleason_methyl	0.004459742	gleason
cg01113246	57007	−3.988105148	7.86E-05	0.073139644	−2.10E-05	2	237033449	F	0.005941232	0.506264545	node_methyl	0.506264545	node.x
cg03054303	146330	4.233731796	2.83E-05	0.033499333	0.000419375	16	1662301	R	0.000607564	0.013219481	gleason_methyl	0.013219481	gleason
cg03054303	197335	4.141616607	4.18E-05	0.045480898	0.000450108	16	1662301	R	3.86E-07	0.013219481	gleason_methyl	0.013219481	gleason
cg03054303	197335	4.141616607	4.18E-05	0.045480898	0.000450108	16	1662301	R	0.00121	0.205222679	primary_gleason_methyl	0.205222679	primary_gleason
cg03228985	339451	−8.097854516	6.22E-15	5.63E-11	−8.10E-05	1	1083642	F	0.000152682	0.319957573	gleason_methyl	0.319957573	gleason
cg03588007	10541	−4.177780095	3.59E-05	0.040550739	−0.000156142	9	100849274	R	2.46E-05	0.049332958	gleason_methyl	0.049332958	gleason
cg03588007	10541	−4.177780095	3.59E-05	0.040550739	−0.000156142	9	100849274	R	0.000399	0.007561488	primary_gleason_methyl	0.007561488	primary_gleason
cg03588007	10541	−4.177780095	3.59E-05	0.040550739	−0.000156142	9	100849274	R	0.00478	0.461942007	pathological_T_methyl	0.461942007	pathological_T
cg03780338	2259	4.934626959	1.16E-06	0.002420551	1.39E-05	13	103244558	F	2.49E-06	0.958792934	gleason_methyl	0.958792934	gleason
cg03780338	2259	4.934626959	1.16E-06	0.002420551	1.39E-05	13	103244558	F	0.00709	0.550619929	primary_gleason_methyl	0.550619929	primary_gleason
cg04007841	283131	−4.945989784	1.10E-06	0.002305322	−0.001312476	11	65307006	F	0.001181542	0.300259221	node_methyl	0.300259221	node.x
cg04125223	9319	4.414024306	1.29E-05	0.018086909	1.57E-05	5	218200	R	1.22E-08	0.006144373	gleason_methyl	0.006144373	gleason
cg04125223	9319	4.414024306	1.29E-05	0.018086909	1.57E-05	5	218200	R	1.77E-06	0.099088812	primary_gleason_methyl	0.099088812	primary_gleason
cg04588774	57524	4.510523012	8.42E-06	0.012694387	7.04E-06	16	3207967	R	0.001185047	0.000893246	biochemical_recurrence_methyl	0.000893246	biochemical_recurrence
cg04738496	57524	5.196395311	3.19E-07	0.000788403	1.18E-05	16	2021690	R	0.001185047	0.067732981	biochemical_recurrence_methyl	0.067732981	biochemical_recurrence
cg04738496	197342	5.526480395	5.77E-08	0.00017534	3.60E-05	16	2021690	R	0.000164007	0.454694526	gleason_methyl	0.454694526	gleason
cg04977610	253980	5.192466243	3.25E-07	0.000802703	8.05E-05	16	30366665	R	0.002535845	0.814245385	gleason_methyl	0.814245385	gleason
cg05065926	286076	4.01583131	7.03E-05	0.067012266	1.08E-06	8	144329806	F	0.003055894	0.000258271	gleason_methyl	0.000258271	gleason
cg05177437	7058	−4.949962791	1.08E-06	0.002267946	−5.02E-05	6	170582289	F	5.26E-05	0.060200416	gleason_methyl	0.060200416	gleason
cg05317498	57524	−4.69690954	3.59E-06	0.006257313	−1.21E-05	16	1425256	F	0.001185047	0.592952261	biochemical_recurrence_methyl	0.592952261	biochemical_recurrence
cg05317498	146330	−5.773151395	1.52E-08	5.30E-05	−0.000122313	16	1425256	F	0.000607564	0.651898021	gleason_methyl	0.651898021	gleason
cg05317498	197335	−5.863174643	9.25E-09	3.39E-05	−0.00013601	16	1425256	F	3.86E-07	0.651898021	gleason_methyl	0.651898021	gleason
cg05317498	197342	−5.323550283	1.67E-07	0.000447261	−3.89E-05	16	1425256	F	0.000164007	0.651898021	gleason_methyl	0.651898021	gleason
cg05317498	197335	−5.863174643	9.25E-09	3.39E-05	−0.00013601	16	1425256	F	0.00121	0.380661447	primary_gleason_methyl	0.380661447	primary_gleason
cg05970437	399664	4.11905522	4.59E-05	0.048902771	7.31E-05	19	1864622	R	0.001667999	0.008596126	gleason_methyl	0.008596126	gleason
cg06758143	79000	4.814460112	2.07E-06	0.003967388	2.90E-06	1	26798177	R	0.00131	0.391977245	pathological_T_methyl	0.085384829	clinical_T.x
cg06758143	79000	4.814460112	2.07E-06	0.003967388	2.90E-06	1	26798177	R	4.95E-05	0.012484752	gleason_methyl	0.012484752	gleason
cg06758143	79000	4.814460112	2.07E-06	0.003967388	2.90E-06	1	26798177	R	5.34E-09	0.912885436	primary_gleason_methyl	0.912885436	primary_gleason
cg07008945	339451	3.952807907	9.07E-05	0.080922794	0.00028418	1	955720	F	0.000152682	0.205918508	gleason_methyl	0.205918508	gleason
cg08861556	197335	4.176276659	3.61E-05	0.040719792	0.000175177	16	850614	R	3.86E-07	0.007086157	gleason_methyl	0.007086157	gleason
cg08861556	9727	4.578985236	6.18E-06	0.009861319	0.000240103	16	850614	R	4.31E-07	0.007086157	gleason_methyl	0.007086157	gleason
cg08861556	8786	4.711802234	3.35E-06	0.005932623	0.000258129	16	850614	R	0.000382769	0.007086157	gleason_methyl	0.007086157	gleason
cg08861556	197335	4.176276659	3.61E-05	0.040719792	0.000175177	16	850614	R	0.00121	0.266894552	primary_gleason_methyl	0.266894552	primary_gleason
cg10286829	2491	4.705099824	3.46E-06	0.006074553	1.66E-06	X	100913838	F	5.02E-06	0.590714131	gleason_methyl	0.590714131	gleason
cg10286829	2491	4.705099824	3.46E-06	0.006074553	1.66E-06	X	100913838	F	0.000107	0.685867114	primary_gleason_methyl	0.685867114	primary_gleason
cg10498434	7153	4.56601381	6.55E-06	0.010356325	2.83E-05	17	37783488	F	0.00374	0.029055459	pathological_T_methyl	0.088721131	clinical_T.x
cg10498434	7153	4.56601381	6.55E-06	0.010356325	2.83E-05	17	37783488	F	1.15E-05	0.970155723	gleason_methyl	0.970155723	gleason
cg10498434	7153	4.56601381	6.55E-06	0.010356325	2.83E-05	17	37783488	F	2.46E-06	0.72449589	primary_gleason_methyl	0.72449589	primary_gleason
cg10705488	339451	6.488035843	2.47E-10	1.22E-06	0.000446796	1	1550648	R	0.000152682	0.039983988	gleason_methyl	0.039983988	gleason
cg11574970	4053	4.23776505	2.78E-05	0.03309595	9.23E-05	14	75469471	F	0.003138501	0.051219574	gleason_methyl	0.051219574	gleason
cg11847126	339451	−8.156421723	4.10E-15	3.77E-11	−4.28E-05	1	1248233	F	0.000152682	0.985911022	gleason_methyl	0.985911022	gleason
cg11855022	57047	4.473969431	9.92E-06	0.014580089	5.27E-07	3	145968700	F	0.0039	0.539977096	node_methyl	0.539977096	node.x
cg12431977	643866	5.689476523	2.40E-08	7.99E-05	2.52E-05	14	24769227	R	0.002053472	0.001252535	gleason_methyl	0.001252535	gleason
cg13444392	677770	−5.585073988	4.22E-08	0.000132899	−6.71E-06	4	1608283	R	7.19E-09	0.003853796	node_methyl	0.003853796	node.x
cg13508391	2305	4.351759854	1.70E-05	0.022508505	0.000516527	12	3068378	R	0.000484124	0.097979899	primary_gleason_methyl	0.097979899	primary_gleason
cg13917964	222161	4.47316485	9.95E-06	0.014614108	3.78E-05	7	30635712	F	4.45E-05	0.955004144	gleason_methyl	0.955004144	gleason
cg13917964	222161	4.47316485	9.95E-06	0.014614108	3.78E-05	7	30635712	F	0.00731	0.239622641	primary_gleason_methyl	0.239622641	primary_gleason
cg14820176	80198	4.480408126	9.64E-06	0.014229571	6.00E-05	11	66346739	FALSE	0.00373	0.621062934	secondary_gleason_methyl	0.621062934	secondary_gleason
cg14820176	1072	4.873586553	1.56E-06	0.003120113	0.002576989	11	66346739	F	0.005005142	0.700418497	gleason_methyl	0.700418497	gleason
cg14820176	80198	4.480408126	9.64E-06	0.014229571	6.00E-05	11	66346739	F	7.30E-10	0.700418497	gleason_methyl	0.700418497	gleason
cg14820176	80198	4.480408126	9.64E-06	0.014229571	6.00E-05	11	66346739	F	0.00108	0.990656757	primary_gleason_methyl	0.990656757	primary_gleason
cg15968307	784	4.81358966	2.08E-06	0.003981341	1.54E-05	12	48577213	R	1.51E-05	0.103009304	gleason_methyl	0.103009304	gleason
cg16638425	57524	−4.095229532	5.07E-05	0.052583939	−1.11E-05	16	2946805	F	0.001185047	0.394681318	biochemical_recurrence_methyl	0.394681318	biochemical_recurrence
cg16712664	253980	4.586041366	5.98E-06	0.009598823	0.000347931	16	30787355	F	0.002535845	0.815286106	gleason_methyl	0.815286106	gleason
cg17398515	25897	4.180309201	3.55E-05	0.040217881	1.34E-05	8	102092795	R	0.000304168	0.378109272	gleason_methyl	0.378109272	gleason
cg17696563	50636	−4.584466672	6.02E-06	0.009657957	−0.000113607	2	242756362	R	0.000184501	0.082855296	primary_gleason_methyl	0.082855296	primary_gleason
cg18123911	339451	10.89872597	1.66E-24	3.32E-20	0.000298968	1	1550773	F	0.000152682	0.670851582	gleason_methyl	0.670851582	gleason
cg18225536	643866	4.89908745	1.38E-06	0.002805114	0.000126544	14	24658182	F	0.002053472	0.493252557	gleason_methyl	0.493252557	gleason
cg19092396	339451	3.930642765	9.92E-05	0.086345169	7.16E-06	1	1143751	F	0.000152682	0.059898103	gleason_methyl	0.059898103	gleason
cg19165854	23251	4.18198955	3.52E-05	0.03999685	3.99E-06	15	79725126	R	0.004028996	0.17701659	biochemical_recurrence_methyl	0.17701659	biochemical_recurrence
cg19414302	339451	−4.48755798	9.33E-06	0.01387107	−2.41E-05	1	1234354	F	0.000152682	0.07787519	gleason_methyl	0.07787519	gleason
cg19901381	1609	4.267056743	2.45E-05	0.029992356	6.86E-06	4	1577971	F	0.002992571	0.147706171	gleason_methyl	0.147706171	gleason
cg21908287	57657	4.06324886	5.78E-05	0.0579969	1.14E-05	1	155225385	R	2.61E-06	0.867305735	gleason_methyl	0.867305735	gleason
cg21908287	57657	4.06324886	5.78E-05	0.0579969	1.14E-05	1	155225385	R	0.00255	0.954639475	primary_gleason_methyl	0.954639475	primary_gleason
cg22062432	2324	5.217476829	2.86E-07	0.000720528	1.36E-05	5	180634502	F	0.009901839	0.402124768	gleason_methyl	0.402124768	gleason
cg23058863	643866	5.449034648	8.68E-08	0.000250473	3.73E-05	14	24610866	F	0.002053472	0.005346928	gleason_methyl	0.005346928	gleason
cg24378253	339451	−8.425695272	5.87E-16	5.91E-12	−0.000298151	1	1017383	R	0.000152682	0.046752456	gleason_methyl	0.046752456	gleason

SNP: single nucleotide polymorphism; FDR: false discovery rate; CHR: chromosome; PSA: prostate-specific antigen

We identified that calcium/calmodulin-dependent serine protein kinase interacting protein 1 (CASKIN1) was enriched with CpGs and CASKIN1 itself was also significantly associated with genes related to biochemical cancer recurrence (P = 0.001; Figure 2). According to the methylated QTL algorithm, these CpG sites’ methylation levels may act as cis-regulators of CASKIN1, suggesting that the methylation of nearby CpG sites probably influences the expression of CASKIN1 and thus affects the recurrence in these cases.

Figure 2.

The violin plot of relative expression of CASKIN1 between biomedical recurrence and nonrecurrence subgroup evaluated by TCGA. The expression of CASKIN1 is significantly different between recurrence and nonrecurrence group (P = 0.001). CASKIN1: calcium/calmodulin-dependent serine protein kinase interacting protein 1; TCGA: The Cancer Genome Atlas.

Methylated QTLs in PCa-related genes

An analysis of these regulation pairs should include the regulatory status in both prostate and PCa cells. To extract useful information for later analysis, we selected meQTLs concerning genes associated with PCa genesis and metastasis.37,38,39,40,41,42 Ninety-two genes were used for meQTL selection. In our meQTL dataset, 38 cis-regulated meQTLs existed near these sites, and 1753 trans-regulated eQTLs were identified. Genes of interest among cis-eQTLs included alanyl aminopeptidase, membrane (ANPEP), activating transcription factor 3 (ATF3), B cell leukemia/lymphoma 2 (BCL2) associated X (BAX), and early growth response 1/3 (EGR1/3), suggesting that the expression levels of these PCa-related genes are regulated by the methylation levels of nearby CpG sites. For further analysis, these gene and methylation findings were validated in 70 pairs of clinical tumorous and nontumorous tissue samples. The patients from whom samples were collected were classified as low, moderate, and high risk according to Gleason score, mean PSA, and tumor/node/metastasis (TNM) clinical staging; a variety of tissues from different risk levels were subjected to methylation quantification via pyrosequencing™.

In our results, the false discovery rate was relatively high because we did not detect the expression levels of the selected meQTL-related genes. However, we still detected several differences in methylation sites between normal and tumorous samples (Figure 3). These sites included meQTL pairs involving FosB proto-oncogene, activator protein-1 (AP-1) transcription factor subunit (FOSB), ANPEP, and ras-related dexamethasone-induced 1 (RASD1). In tumorous samples, the methylation levels at these sites, particularly cg20664996 near RASD1, were low. We also noticed that in high- and moderate-risk samples, the methylation levels at this site reached 0. RASD1 has been reported as an apoptosis- and ras-related gene that prevents aberrant cell growth in several cell lines.43 In short, an epigenetic association study of clinical indexes detected risk loci and pyrosequencing for site validation. Although the number of DNA methylation-regulated genes across PCa samples is a small proportion, the associated genes play essential roles in PCa carcinogenesis.

Figure 3.

Four CpG loci that were significantly differently methylated according to the validated dataset. In each CpG locus, the left dots indicate the relative methylation level of the normal tissues and the right is the corresponding cancerous tissues, the P values between the normal tissues and cancerous tissues are shown from left to right.

DISCUSSION

DNA methylation is one among the most common and well-characterized epigenetic changes in PCa.5,6,7,8,9,10,11 The importance of large-scale methylome studies has increased in biomedical fields, and high-throughput sequencing technologies promise sensitive, quantitative, and high-resolution large-scale DNA methylation analyses. A genome-wide association study (GWAS) based on QTL analysis is an effective method with which cancer-related genes and risk loci throughout the genome can be studied;14,15,44 similarly, the usefulness of methylation-based GWAS has been reported.45 However, despite reports on the association of DNA methylation with carcinogenesis and development in the prostate,46,47,48,49 genome-wide analyses of the associations between gene expression, DNA methylation, and clinical information related to PCa continue to yield vague results. Here, we utilized TCGA to obtain DNA methylation data from more than 400 000 CpG sites, expression levels of over 20 000 genes, and clinical indexes for 419 PCa samples, to perform a methylated QTL analysis in which we identified 5852 cis-regulating pairs comprising 1712 genes. Compared with previous breast cancer studies, only 0.8% of genes associated with PCa are affected by DNA methylation, and these genes exhibit significant involvement in several pathways that previously have been reported to play important roles in PCa genesis, including the JAK/STAT and VEGF pathways.50,51 We also noticed that the associated genes and locations were significantly enriched at several genome locations, including 19q13.2–q13.43 and 16q22.2–q23.1. As these are cis-interactions within 1 MB, the regulated genes also are enriched at these sites. Previous reports of SNPs and PCa noted an association between SNPs at 19q13 and PCa risk.52 Another GWAS of a Chinese population demonstrated 19q13 as a novel risk locus for PCa,33 and a decade-long study detected chromosomal deletion and gene suppression at 16q22 in PCa.53 It was worth noting that this region was reported to show a loss of heterogeneity in breast cancer.54 In addition to somatic mutations, SNPs, and copy number variations, our data suggest that DNA methylation may also influence PCa genesis and probably plays important roles in this process by regulating the transcription of this gene nexus.

Although a previous study based on eQTL algorithms aimed to define the relationships between genomic, epigenomic, and transcriptomic changes,26,31 the link between the data obtained and clinical information remains unclear. Heyn et al.55 reported a TCGA QTL analysis using 13 tumor types from TCGA, including PCa, but their study lacks correlations to PCa-specific clinical variables such as Gleason score and PSA. Their data sets (cancer samples and adjacent normal tissues) did not include control subjects without the disease; future studies are needed to determine if GWAS risk alleles exhibit similar relationships in cancer-unrelated donors.

To further investigate the relationships of DNA methylation and gene expression with clinical information, we performed an association study between the three pairs (expression-methylation, methylation-clinical information, and clinical information-expression) and identified 92 genes regulated by CpG site methylation. Furthermore, we associated the expression levels of these genes with clinical information, particularly the Gleason score. GO analysis revealed that these 92 genes were significantly enriched with respect to DNA repair and cell cytoskeleton, suggesting that the genes probably direct the development and prognosis of PCa. Of these genes, we noted that EME2 previously was reported to maintain genomic stability,56 whereas the roles of the other genes were poorly elucidated. A locus analysis of these 92 genes demonstrated significant enrichment at 11q13 (P = 0.008, 3 genes) and 16q13 (P < 0.0001, 6 genes). The genes identified were enriched according to an in-house R script of the hypergeometric distribution, using the genes assayed for meQTLs as background. Enriched genomic loci with P < 0.01 were identified, and some genes associated with the mQTL were significantly enriched for DNA repair and cell cytoskeleton. Most of the sites associated with Gleason score were in 11q13 and 16q13, and their correlated genes included EME2,56 KCTD13,57 KLHL17,58 and WDR90.59 These special genes played critical roles in the biological and pathological characteristics of an organism.

Studies based on SNPs or somatic mutation profiling have shown that 11q13 and 16q13 loci are associated with aggressiveness, invasion, and poor prognosis of PCa. Our results show that epigenetic alterations in these regions are associated with genes considered necessary for the progression of PCa and suggest that methylation at 11q13 and 16q13 loci is essential to PCa progression. We were the first to prove that CASKIN1 was enriched with CpGs, and CASKIN1 itself was also significantly associated with genes related to biochemical cancer recurrence. Although the mechanism of how CASKIN1 affects PCa is unclear, its interacting protein, CASK has been associated with survival of patients with colorectal cancer.60

For validation, we collected 70 clinical tissue samples from 2014 to 2017. Although these patients are Asian, we used the 2020 National Comprehensive Cancer Network (NCCN) guidelines as a criterion to classify distinct groups. Some studies suggested racial differences in PCa. Non-Whites were associated with a significantly higher likelihood of presenting with high-risk PCa (22.9% of Hispanics, 23.8% of Blacks, and 23.3% of those of other races compared with 19.0% of Whites; all P < 0.001).61 Age-adjusted PCa mortality in black men was more than double that of white men (42.0/100 000 vs 18.7/100 000). Asian men had a lower incidence of PCa and death than white men. Chinese and Japanese men have a relatively low risk of PCa than their European and North American counterparts.62

The method we describe is based on detecting methylation and expression differences between samples of PCa. Therefore, it aims to identify correlations that occur within particular subsets of cases. For example, we found that some genes of interest among cis-eQTLs have played essential roles in PCa, including ANPEP,63,64 ATF3,65,66 BAX,67 and EGR1/3.68 These genes are involved in tumorigenesis, metastasis, and recurrence of PCa.

We selected 30 high-confidence CpG sites that have been reported to associate with PCa-related genes and evaluated methylation at related CpG sites (Supplementary Table 7). Despite a relatively high FDR, we determined differences in the methylation levels of several CpG sites between cancerous and noncancerous tissues (associated genes included ANPEP and FOSB). FDR is a tremendous concern for this study. To reduce FDR, we first enrolled the samples using the inclusion criteria described in the methods part. Second, we selected the P-value of meQTL pairs as 1 × 10⁻⁵ instead of 0.01/0.05. Last, the FDR values were calculated using Bonferroni, which is the most rigorous method available, and 0.01 was used for cutoff.

Supplementary Table 7.

The details of selected 30 high-confidence CpG sites via methylation quantitative trait loci reported to associate with prostate cancer-related genes

Gene	Symbol	Chromosome	SNPs	Position	FDR	P	References
2354	FOSB	chr16	cg02213678	85112535	3.03E-44	3.66E-49	1, 2
7356	SCGB1A1	chr11	cg02950427	61323284	7.99E-03	4.82E-06	3, 4
94025	MUC16	chr19	cg04744134	8408128	1.72E-14	1.36E-18	5, 6
3569	IL-6	chr7	cg05407490	23749667	4.35E-05	1.22E-08	7 8 9
11197	WIF1	chr12	cg05754508	65672531	5.10E-04	1.94E-07	10 11 12 13
2354	FOSB	chr19	cg06577658	46296270	9.10E-03	5.62E-06	1, 2
7273	TTN	chr22	cg08170140	31090747	1.62E-32	3.26E-37	14 15 16
94025	MUC16	chr19	cg08341673	8408202	1.28E-03	5.56E-07	5, 6
7356	SCGB1A1	chr11	cg08360511	61322922	9.72E-04	4.06E-07	3, 4
23600	AMACR	chr5	cg09093477	34839189	4.50E-02	4.11E-05	17 18 19
7356	SCGB1A1	chr13	cg09207578	114781949	5.43E-21	2.09E-25	3, 4
27232	GNMT	chr6	cg09804858	42397121	7.66E-04	3.08E-07	20 21 22 23
51655	RASD1	chr17	cg12918275	18150954	2.28E-04	7.79E-08	24, 25
27232	GNMT	chr6	cg14338887	42928500	1.69E-08	2.57E-12	20,21,2223
290	ANPEP	chr15	cg14927321	89491614	8.79E-03	5.40E-06	26 27 28
7356	SCGB1A1	chr11	cg15492820	61323113	6.78E-06	1.60E-09	3, 4
290	ANPEP	chr15	cg15850918	89900741	1.28E-02	8.47E-06	26 27 28
4852	NPY	chr19	cg15854606	18654548	8.51E-22	3.09E-26	29 30 31 32
3552	IL-1A	chr2	cg18668836	113103184	1.79E-03	8.22E-07	33, 34
6262	RYR2	chr1	cg18770350	236849966	4.31E-02	3.90E-05	35, 36
2354	FOSB	chr18	cg19283427	29523380	6.27E-38	9.96E-43	1, 2
51655	RASD1	chr17	cg20664996	17359821	2.22E-02	1.67E-05	24, 25
94025	MUC16	chr19	cg22886512	8408083	7.60E-09	1.10E-12	5, 6
5179	PENK	chr8	cg23139584	56987506	1.56E-03	6.98E-07	37 38 39
2354	FOSB	chr2	cg23158051	65454692	4.02E-39	6.05E-44	1, 2
1960	EGR3	chr8	cg23303108	23083578	1.26E-02	8.31E-06	8, 40
2354	FOSB	chr22	cg23410069	20287622	2.15E-39	3.20E-44	1, 2
1960	EGR3	chr8	cg24278165	23083551	1.83E-02	1.31E-05	8, 40
84419	C15orf48	chr11	cg25902616	47789193	3.59E-16	2.02E-20	41 42 43
7356	SCGB1A1	chrX	cg27069132	135991543	3.93E-21	1.50E-25	3, 4

PCa: prostate cancer; SNP: single nucleotide polymorphism; FDR: false discovery rate; FOSB: FosB proto-oncogene, activator protein-1 (AP-1) transcription factor subunit; SCGB1A1: secretoglobin family 1A member 1; MUC16: mucin 16; IL-6: interleukin 6; WIF1: Wnt inhibitory factor 1; TTN: titin; AMACR: alpha-methylacyl-coenzyme A racemase; GNMT: glycine N-methyltransferase; RASD1: ras-related dexamethasone induced 1; ANPEP: alanyl aminopeptidase, membrane; NPY: neuropeptide Y; IL-1A: interleukin 1A; RYR2: ryanodine receptor 2; PENK: proenkephalin; EGR3: early growth response 3; C15orf48: chromosome 15 open reading frame 48

We further identified a correlation between the Gleason score and methylation level at cg20664996, a CpG location associated with RASD1 expression, according to an eQTL analysis. RASD1 is a 30 kDa G-protein that belongs to the Ras superfamily of small GTPases. Several robustly upregulated bicalutamide-dependent genes were identified by micro-array in LNCaP-AR_W741L cells that were not significantly enhanced by dihydrotestosterone (DHT) in the LNCaP-LacZ line, including RASD1.69 Liu et al.70 showed that formononetin inhibited cell proliferation and induced apoptosis in DU-145 cells throughout the RASD1/mitogen-activated protein kinase (MAPK)/BAX pathway in PCa. The selected high-confidence CpG sites are reported to associate with some PCa-related genes, including FOSB,71 secretoglobin family 1A member 1 (SCGB1A1),72 mucin 16 (MUC16),73 alpha-methylacyl-coenzyme A racemase (AMACR),74 and glycine N-methyltransferase (GNMT).75 Emerging evidence has shown that these genes may be useful diagnostic and prognostic biomarkers for PCa.

CONCLUSIONS

We demonstrated that novel meQTL pairs were associated with PCa, and for the first time, we identified the locations of CpG sites and genes within at-risk loci, including the 19q13.2–q13.43 and 16q22.2–q23.1 loci. We further used pyrosequencing™ for validation and identified several genes that may impact the development and prognosis of PCa.

AUTHOR CONTRIBUTIONS

QL conceived and participated in its design, searched databases, and extracted and assessed studies. QL and GL carried out the statistical analysis and interpretation of data and drafted the main manuscript. QL and DTM prepared the tables and figures. QL, YTX, and RMW prepared the figures and supplementary tables. QL, JK, and JQ participated in the conceptualization and design of the manuscript, performed the selection of studies, and drafted the manuscript. All authors reviewed and approved the final manuscript.

COMPETING INTERESTS

All authors declare no competing interests.