Epigenome-wide DNA methylation profiling of periprostatic adipose tissue in prostate cancer patients with excess adiposity-a pilot study

RESEARCH Open Access Epigenome-wide DNA methylation profiling of periprostatic adipose tissue in prostate cancer patients with excess adiposity—a pilot study Yan Cheng 1,2 , Cátia Monteiro 3,4 , Andreia Matos 5,6 , Jiaying You 1 , Avelino Fraga 6,7 , Carina Pereira 3,8 , Victoria Catalán 9,10 , Amaia Rodríguez 9,10 , Javier Gómez-Ambrosi 9,10 , Gema Frühbeck 9,10,11 , Ricardo Ribeiro 3,5,6,12,13*† and Pingzhao Hu 1*† Abstract Background: Periprostatic adipose tissue (PPAT) has been recognized to associate with prostate cancer (PCa) aggressiveness and progression. Here, we sought to investigate whether excess adiposity modulates the methylome of PPAT in PCa patients. DNA methylation profiling was performed in PPAT from obese/overweight (OB/OW, BMI > 25 kg m −2 ) and normal weight (NW, BMI < 25 kg m −2 ) PCa patients. Significant differences in methylated CpGs between OB/OW and NW groups were inferred by statistical modeling. Results: Five thousand five hundred twenty-six differentially methylated CpGs were identified between OB/OW and NW PCa patients with 90.2% hypermethylated. Four hundred eighty-three of these CpGs were found to be located at both promoters and CpG islands, whereas the representing 412 genes were found to be involved in pluripotency of stem cells, fatty acid metabolism, and many other biological processes; 14 of these genes, particularly FADS1,MOGAT1, and PCYT2, with promoter hypermethylation presented with significantly decreased gene expression in matched samples. Additionally, 38 genes were correlated with antigen processing and presentation of endogenous antigen via MHC class I, which might result in fatty acid accumulation in PPAT and tumor immune evasion. Conclusions: Results showed that the whole epigenome methylation profiles of PPAT were significantly different in OB/OW compared to normal weight PCa patients. The epigenetic variation associated with excess adiposity likely resulted in altered lipid metabolism and immune dysregulation, contributing towards unfavorable PCa microenvironment, thus warranting further validation studies in larger samples. Keywords: DNA methylation, Periprostatic adipose tissue, Obesity, Prostate cancer, Microenvironment Background Prostate cancer (PCa) is one of the most frequent malignancies in men and the second leading cause of cancerrelated death in the North America and most western European countries [1,2]. Epidemiological studies support obesity or excess adiposity as an important environmental risk factor for PCa, being primarily associated with advanced disease and death [3]. Periprostatic adipose tissue (PPAT), a white fat depot surrounding the prostate capsular-like structure, has been recognized to have the potential to exert pro-tumoral endocrine and paracrine influences on prostate cancer cell’s biological phenotypes [4]. There is now evidence that obesity and overweight result in excess fat deposit at PPAT [5], altered fatty acid profile [6], migration of tumor cells [7], secretion of a variety of adipokines, such as interleukin-1 beta (IL-1b), osteopontin, leptin, tumor necrosis factor alpha (TNF-a), * Correspondence: [email protected];[email protected] † Equal contributors 3 Molecular Oncology Group, Portuguese Institute of Oncology, Porto, Portugal 1 Department of Biochemistry and Medical Genetics & Department of Electrical and Computer Engineering, University of Manitoba, Winnipeg, Canada Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Cheng et al. Clinical Epigenetics (2018) 10:54 https://doi.org/10.1186/s13148-018-0490-3 and decreased adiponectin, thus contributing to a tumor microenvironment that ultimately facilitates PCa aggressiveness [7,8]. DNA methylation is a well-known epigenetic mechanism resulting from the interaction between environmental factors and the genome [9]. DNA methylation with variation of CpG sites is associated with tissue-specific gene modulation and involved in phenotype transmission and in the development of diseases [10]. Excess adiposity, as a consequence of environmental factors such as excessive food consumption or inactive lifestyle, has been identified as a regulator of epigenetic modification in adipose tissue. Recent findings from experimental studies suggested that modification of DNA methylation pattern in adipose tissue and adipocytes was related with development of cancer, type 2 diabetes, and cardiovascular diseases through influencing metabolism and inflammation [11–13]. Additionally, several studies reported altered DNA methylation in PCa cells as compared with adjacent benign tissue, and some significantly methylated CpG sites and genes were found to be responsible for the occurrence and progression of PCa [14–16]. Nevertheless, the epigenome-wide DNA methylation profile of PPAT from excess adiposity PCa patients is currently unknown despite its potential mechanistic involvement in obesity association with PCa. The aim of this study was to perform a epigeneticwide association study (EWAS) in order to evaluate DNA methylation profile of PPAT obtained from obese/ overweight (OB/OW) in comparison with normal weight (NW) PCa patients and identify differentially methylated sites. We also explored the consequential potential biological functions that account for the effect of PPAT from OB/OW subjects in PCa molecular mechanisms. Methods Study samples This study included ten prostate cancer patients from the Portuguese Institute of Oncology, Porto Centre. Inclusion criteria and conditions of this study have been previously reported, including the procedures for PPAT collection, handling, and storage [4]. Briefly, PPAT was collected and immediately processed in the operating room and transported to the laboratory within 2 h in appropriate culture media and temperature conditions, in order to minimize pre-analytical errors. Patients’ signed informed consent and research procedures were approved by the institute’s ethics committee. The clinical and pathological characteristics of participants are presented in Table 1. The ten subjects were selected from a larger group of patients undergoing prostate surgery (n= 51) [4,17] that fitted the strict inclusion and exclusion criteria, in order to control for variables that might influence adipose tissue gene expression or methylation (e.g., anti-diabetic or antidyslipidemia drugs, stage of disease and PSA, concomitant diseases such as diabetes, other neoplasia or metabolic syndrome). Subjects were matched for age at diagnosis, PSA value, Gleason grade, and stage of disease, which differed in body mass index (BMI). BMI was calculated by dividing weight in kilograms by the squared height in meters and categorized using the WHO (World Health Organization) criteria: normal weight, BMI < 25 kg m −2 , overweight, 25 ≤BMI < 30 kg m −2 , and obese, BMI ≥ 30 kg m −2 . Obese and overweight were combined into one excess adiposity group (n=5, BMI≥25 kg m −2 ) versus normal weight group (n= 5, BMI < 25 kg m −2 ). Therefore, the two groups were selected to differ only by BMI, in order to reflect our objective of assessing whether excess adiposity (BMI) influences PPAT methylation profile. Epigenome-wide DNA methylation analysis DNA was isolated from PPAT using Puregene hisalt extraction method (Qiagen/Gentra). Briefly, the tissue was minced with scalpels in a sterile petri dish on ice and then transferred to Puregene Cell Kit for overnight Proteinase K digest at 55 °C. A second Proteinase K digest was done the next morning for 5 h. DNA from the digested tissue was purified using Puregene extraction protocol (Qiagen/Gentra). Purified DNA was washed 2× with 70% ethanol and DNA pellet air dried and rehydrated in TE (10 mM Tris-Cl, 1 mM EDTA pH 7.5). Epigenome-wide DNA methylation was analyzed using the Infinium Human Methylation450 (HM450) BeadChip (Illumina, San Diego, CA, USA) in the Center for Applied Genomics (Toronto). This array Table 1 Clinicopathological characteristics of PCa patients by BMI category Character NW (n= 5) OB/OW (n=5) Pvalue Age (years) 65.2 ± 3.8 63.2 ± 2.5 0.67 a BMI (kg/m 2 ) 23.0 ± 0.3 29.0 ± 0.9 0.0003 a Gleason score < 7 2 (40%) 1 (20%) ≥7 3 (60%) 4 (80%) 1.00 b Stage OCPCa 2 (40%) 2 (40%) EPCa 3 (60%) 3 (60%) 1.00 b Smoking status Yes 1 5 No 4 0 0.05 b PSA (ng/ml) 10.7 ± 2.7 12.1 ± 3.23 0.74 a Data are presented as mean ± SD or number (%). Significant difference between OB/OW and NW was evaluated using a ttest and b Fisher’s exact test OB/OW obese/overweight, NW normal weight, BMI body mass index; PSA, prostate specific antigen; PCa, prostate cancer; OCPCa, organ-confined prostate cancer; EPCa, extra-prostatic PCa Cheng et al. Clinical Epigenetics (2018) 10:54 Page 2 of 15 contains 485,577 probes, which cover 21,231 (99%) RefSeq genes. Briefly, DNA was bisulfite-converted using the EZ DNA methylation kit (Zymo Research, Orange, CA, USA) and then used on the Infinium Assay® followed by the Infinium HD Assay Methylation Protocol (Illumina). The imaging data on the BeadChips was captured by Illumina iScan system. Data filtering and normalization Raw methylation level for each probe was represented by methylation βvalue, which was calculated based on β= intensity of the methylated allele/(intensity of the unmethylated allele + intensity of the methylated allele + 100). Mvalues were the logit transformation of βvalues based on M=log 2 (β/(1 −β)), which makes the data more homoscedastic and appropriate for further bioinformatic and statistical analysis. Methylation values were normalized using the functional normalization algorithm implemented in Minfi R package [18]. Quality control was performed by excluding CpG probes, which are found by Chen et al. to be cross-reactive with areas of the genome not at the site of interest [19], as well as control probes and probes on sex chromosomes. We analyzed a total of 438,458 CpG sites from the PPAT of 5 OB/OW PCa patients and 5 NW PCa patients. Differential methylation analysis A statistical linear modeling approach was applied to the detected differentially methylated CpG sites (DMCs) associated with obesity in PPAT using the Bioconductor “limma”package [20]. Hyperor hypomethylation was determined when methylation levels of CpGs increased or decreased between the OB/OW PCa group and the NW PCa group based on mean different β> 0 or < 0. False discovery rate (FDR)-corrected Pvalues were determined according to the method of Benjamin and Hochberg’s (BH method) multiple testing procedure [21]. Differentially methylated regions (DMRs) were identified using the “Bumphunter”method implemented in the “chAMP”Rpackage with the parameters (B= 1000, useWeights = TRUE, minProbes = 10, pickCutoff = TRUE, and other settings with default values) [22]. The proportions of significant hyperor hypomethylated CpGs were calculated and visualized according to their relation to the nearest genes or to the CpG islands, separately. Gene promoter region was defined as 1500 base pairs (bp) and 200 bp upstream of the transcription start site (TSS) (TSS1500 and TSS200) [23]. Identified genes were selected when more than two significantly hypermethylated CpGs were simultaneously located in the promoter region. Functions, pathway, and network enrichment analysis Gene ontology (GO) and KEGG pathway enrichment analyses were performed to explore the biological functions of significantly methylated genes using the online bioinformatic tool Enrichr [24]. Protein-protein interaction (PPI) analysis of all DMC-related genes was performed using NetworkAnalyst according to STRING database [25]. Association analysis between DNA methylation and gene expression We have previously performed gene expression experiment of the PPAT of the 5 OB/OW PCa patients and the 5 NW PCa patients using the HG-U133 Plus 2.0 Affymetrix GeneChip Array (Affymetrix, Santa Clara, CA, USA) [4]. Differential gene expression (DGE) analysis between the OB/OW PCa patients and the NW PCa patients was re-performed using the Bioconductor “limma”package as previously described [4]. Spearman’s rank correlation analysis was performed between the methylation profiles of the hypermethylated CpGs and the gene expression profiles of the genes in PPAT. Results Clinical characteristics Clinical characteristics of PCa patients in this study were stratified according to obesity classification groups and are presented in Table 1. Mean age, PSA level, Gleason sum score, and cancer stage in subjects with PCa were similar (Pvalue > 0.05) between OB/OW and NW groups. As expected, the mean BMI of the OB/OW group was significantly higher than that of the NW subjects (Pvalue < 0.01). All the patients in the OB/OW group are ex-smokers or active smokers, while only one patient in the NW group is a smoker (Pvalue = 0.05). Epigenome-wide DNA methylation profiling of PPAT To study the impact of obesity status on DNA methylation profiles and to identify differentially methylated CpG sites in PPAT from OB/OW and NW prostate cancer patients, we conducted epigenome-wide DNA methylation analyses. A flowchart of the data analysis is depicted in Additional file 1: Figure S1. After quality control and filtering, the Infinium array generated methylation data for 438,458 CpG sites, from which 5526 were differentially methylated after FDR control in the PPAT of OB/OW PCa patients compared to NW (adjusted Pvalue < 0.25; Additional file 2: Table S1 and Table 2). The unsupervised hierarchical clustering of DMCs showed differential DNA methylation patterns in PPAT between OB/OW and NW samples (Additional file 3: Figure S2). The majority of DMCs were hypermethylated (n= 4985, 90.2%), with 9.8% hypomethylated CpG sites (n= 541) in OB/OW versus NW prostate cancer patients (Fig. 1a, b, c). Cheng et al. Clinical Epigenetics (2018) 10:54 Page 3 of 15 Chromosomal distribution of the DMCs To further explore the methylation profile, we investigated the chromosome distribution of DMCs. Results showed that hypermethylated CpG sites were located at chromosomes 1, 6, 11, and 17 (proportion > 6%, Fig. 1d) and hypomethylated CpG sites were located at chromosomes 1, 2, 6, 7, and 11 (proportion > 6%, Fig. 1e). Methylation variations of hypermethylated DMCs and hypomethylated DMCs were found mainly distributed on chromosomes 1, 6, and 11, suggesting that the DNA methylation alterations in these chromosomes were correlated with the body weight changes in prostate patients. Furthermore, we compared the distribution of the DMCs (hyperand hypomethylated, separately) with the distribution of all evaluated CpG sites based on their relation to nearest gene regions (Fig. 2f, Additional file 4: Table S2) or their relation to CpG islands (Fig. 2g, Additional file 5: Table S3) using χ 2 test. The results showed that hypermethylated CpGs are mainly located at TSS1500 (transcription start sites 1500), IGR (intergenic Table 2 Differentially methylated CpG sites in PPAT between obese/overweight PCa patients and normal weight controls Probe ID Chromosome and coordinate (GRCh37) Nearest gene Relation to gene region Relation to CpG island DNAm β difference (%) Pvalue Adjusted Pvalue (< 0.25) Hypermethylated CpG sites cg09476130 chr1:159870086 CCDC19 TSS200 Island 12.1 1.87E−03 0.213 cg21293934 chr18:14748230 ANKRD30B TSS200 Island 11.2 1.83E−03 0.212 cg16925210 chr2:216946718 PECR TSS200 Island 11.2 2.44E−03 0.226 cg11625005 chr5:1295737 TERT TSS1500 Island 11.1 1.38E−03 0.196 cg07039560 chr5:140683681 SLC25A2 TSS200 Island 10.5 2.24E−03 0.222 cg00329447 chr8:145028170 PLEC1 TSS200 Island 10.1 3.41E−03 0.244 cg24463471 chr1:25257978 RUNX3 TSS1500 Island 9.9 3.58E−04 0.155 cg26149485 chr19:2428350 TIMM13 TSS1500 Island 9.7 3.36E−04 0.154 cg05156901 chr22:51016646 CPT1B TSS200 Island 9.3 3.05E−03 0.238 cg18689454 chr21:45705694 AIRE TSS200 Island 9.3 7.33E−04 0.174 cg01454592 chr3:49236800 CCDC36 TSS200 Island 9.3 2.89E−03 0.236 cg24041556 chr19:10736059 SLC44A2 TSS200 Island 9.1 2.15E−05 0.110 cg22257574 chr9:135754383 C9orf98 TSS200 Island 9.0 1.97E−05 0.110 cg23005885 chr15:90543450 ZNF710 TSS1500 Island 8.9 6.20E−04 0.169 cg05726756 chr17:46608288 HOXB1 TSS200 Island 8.6 1.67E−03 0.206 cg12782180 chr7:127880932 LEP TSS1500 Island 8.5 9.78E−04 0.184 cg04675542 chr5:150284416 ZNF300 TSS200 Island 8.4 3.42E−03 0.244 cg10134527 chr6:33283015 TAPBP TSS1500 Island 8.4 3.19E−03 0.241 cg23387569 chr12:58120011 LOC100130776 TSS200 Island 8.4 3.39E−04 0.155 cg17205324 chr14:23835595 EFS TSS1500 Island 8.3 1.46E−04 0.133 cg24402300 chr19:55591437 EPS8L1 TSS1500 Island 8.2 2.02E−03 0.216 cg18081258 chr14:21494161 NDRG2 TSS1500 Island 8.2 1.05E−03 0.187 cg00730561 chr10:102279703 SEC31B TSS200 Island 8.1 4.82E−04 0.162 cg17791651 chr1:38513489 POU3F1 TSS1500 Island 8.0 1.31E−04 0.133 Hypomethylated CpG sites cg03462171 chr16:1664488 CRAMP1L TSS200 Island −8.2 2.30E−03 0.223 cg11648730 chr5:92907151 FLJ42709 TSS1500 Island −6.4 1.72E−03 0.207 cg04558166 chr1:210001279 C1orf107 TSS200 Island −4.0 1.75E−03 0.209 cg25472897 chr8:145560555 SCRT1 TSS1500 Island −3.1 1.53E−03 0.201 cg17612948 chr5:110427863 WDR36 TSS200 Island −3.0 1.70E−03 0.207 cg21665057 chr3:196295764 WDR53 TSS1500 Island −2.4 2.42E−03 0.225 cg12683173 chr7:69063404 AUTS2 TSS1500 Island −1.4 3.30E−03 0.242 cg04872557 chr1:76190008 ACADM TSS200 Island −1.0 1.96E−03 0.215 Cheng et al. Clinical Epigenetics (2018) 10:54 Page 4 of 15 region), N-shore, and S-shore, and hypomethylated CpGs are mostly located at the gene body and open sea. Functional enrichment analysis of significantly obesity-associated DMCs To investigate the potential biological relevance of the significant DMCs, we further filtered 483 DMCs (distributed within 413 genes) from a total of 5526 DMCs according to their locations at both the gene promoter and CpG island (Additional file 6: Table S4). Four hundred seventyfive of the 483 DMCs (representing 404 genes) were hypermethylated. Functional enrichment analysis of the hypermethylated genes showed that these genes were enriched for biological processes, such as pattern specification process, neuron differentiation, neuron fate specification, and negative regulation of phosphate metabolic process (adjusted Pvalue < 0.05, Additional file 7: Table S5), as well as molecular functions, such as Fig. 1 Epigenetic profiles of differentially methylated CpGs of PPAT between OB/OW and NW groups. Manhattan plots show epigenetic profiles of all increased methylated CpGs (a) and all decreased methylated CpGs (b). The X-axis shows chromosomes, and the Y-axisisa−log10 (Pvalue). The black line represents the threshold of adjusted Pvalue = 0.25. CpGs above the black line are significantly hyperor hypomethylated. The volcano plot of DNA methylation (c) shows a significant difference in PPAT between the OB/OW and NW groups. Four thousand nine hundred eighty-five hypermethylated CpGs are labeled in red, and 541 hypomethylated CpGs are labeled in green (adjusted Pvalue > 0.25). The proportions of hyperand hypomethylated CpGs on each chromosome are shown in (d)and(e). The black line indicates if the proportions of hyper-andhypomethylatedCpGs on a chromosome are higher than 6%. The distribution of significant DMCs (hyperor hypomethylated CpGs) and globe DNA methylation CpGs in locations related to the nearest gene regions and CpG islands are shown in fand g. Hypermethylated CpGs are mainly located at TSS1500 (transcription start sites 1500), IGR (intergenic region), N-shore (the 2 kb regions upstream of the CpG islandboundaries),andS-shore(the2kbregions downstream of the CpG island boundaries), and hypomethylated CpGs are mostly located at the gene body and open sea. The difference of the proportion of CpGs among the three CpG groups was calculated based on the χ 2 test (*P< 0.05, **P< 0.01, ***P< 0.001, ns not significant). CpG islands were defined as DNA sequences (500 base windows; excluding most repetitive Alu-elements) with a GC base composition greater than 50% and a CpG observed/expected ratio of more than 0.6. The 2 kb regions immediately upstream (N_Shore) and downstream (S_Shore) of the CpG island boundaries were defined as “CpG island shores,”and the 2 kb regions upstream (N_Shelf) and downstream (S_Shelf) of the CpG island shores were referred as “CpG island shelves.”Open seas were the regions more than 4 kb from CpG islands Cheng et al. Clinical Epigenetics (2018) 10:54 Page 5 of 15 neuropeptide receptor activity and sequence-specific DNA-binding RNA polymerase II transcription factor activity (adjusted Pvalue < 0.1, Additional file 8: Table S6). KEGG pathway enrichment analysis showed that hypermethylated genes were involved in signaling pathways regulating pluripotency of stem cells, fatty acid metabolism, basal cell carcinoma, non-alcoholic fatty liver disease (NAFLD), and AMPK signaling pathway (Pvalue < 0.05, Additional file 9: Table S7). We mapped the 404 hypermethylated genes to the STRING database and generated a protein-protein interaction (PPI) network by the NetworkAnalyst. The largest subnetwork was identified to include 247 nodes (genes) and 403 edges (Fig. 2a). In the network, the size of the nodes was based on their degree values and the color of nodes was based on their Pvalues. This network contained 118 seed genes from the DMCs, and the enrichment pathway analysis showed that the genes of the Fig. 2 Protein-protein interaction analysis. aA subnetwork composing of 247 nodes and 403 edges was generated using methylated genes. Blue dots represent the genes involved in prostate cancer; red and pink dots represent the seeds (methylated genes) according to the different Pvalues; the gray dots represent the proteins which were closely interacted with the seeds, and the circle size represents the node degree. bThe pathway enrichment analysis shows the subnetwork is mainly enriched in cancer pathways (P<0.0001) Cheng et al. Clinical Epigenetics (2018) 10:54 Page 6 of 15 subnetwork were mostly involved in the pathways of prostate cancer and other cancers (Fig. 2b,Additionalfile10: Table S8, adjusted Pvalue < 0.05). Particularly, the gene UBC (ubiquitin C) was found to be a hub connecting with many other nodes in the network, suggesting that the gene may play important biological roles in the PPAT of obese PCa patients. Selected genes with multiple methylated CpG sites In order to explore repression of genes by DNA methylation modifications, we selected genes which had multiple hypermethylated CpG sites (the number of methylated CpG sites ≥2, in at least one of the sites with ameandifferenceofβ> 3% and an adjusted Pvalue < 0.25) (Additional file 1: Figure S1 and Additional file 11:TableS9). A total of 38 genes with 100 differentially methylated CpG sites were selected, which included TAPBP,RUNX3,CPT1B, CPT1C,MOGAT3,WNT2,andAIRE (Additional file 11: Table S9). Notably, the promoter region of TAPBP (TAPbinding protein) had eight hypermethylated CpG sites in the promoter (Fig. 3a), which were significantly more methylated in the OB/OW than those in the NW groups (Fig. 3b), with ameandifferenceofβvalue greater than 5% (Additional file 10: Table S8). Spearman’s rank correlation showed strong association (r 2 =0.73–0.97) of the eight hypermethylated CpGs in the TAPBP promoter with their methylation levels (Fig. 3c). Pathway analysis of these genes revealed enrichment for fatty acid metabolism, PPAR signaling pathway, glucagon signaling pathway, AMPK signaling pathway, glycerolipid metabolism, basal cell carcinoma, antigen processing and presentation, ECM receptor interaction, and insulin resistance (adjusted Pvalue < 0.25) (Additional file 12: Table S10). Differential methylated regions analysis Ten DMRs were identified (P< 0.01) in obesity PPAT samples compared to normal weight controls (Table 3). The size of the DMRs varied from 161 to 1287 bp. Noteworthy, four out of the ten DMRs were discovered on chromosome 6. Eight regions were located in genes, and two were in the intergenic region. Four regions were in the gene promoter of FAM104A,C17orf80,HOXA4A, and TAPBP. Association analysis between DNA methylation and mRNA expression Increased DNA methylation of promoter in CpG islands was obviously linked to gene transcriptional silencing [26]. Therefore, we related hypermethylated CpG sites in PPAT with genes showed decreased gene expression level from our previously generated mRNA expression data [4]. DNA methylation of 16 CpG sites, corresponding to 14 genes, was associated with significantly decreased transcripts in OB/OW group (Pvalue < 0.05) (Table 4). The Spearman’s rank correlation analysis showed that eight of the 14 genes have significantly negative association (Pvalue < 0.05) between the methylation profiles and the gene expression profilesofthesegenes(Table4). The repression genes were mainly involved in metabolic pathways (Additional file 13: Table S11, adjusted Pvalue < 0.25), such as MOGAT1 (glycerolipid metabolism), FADS1 (fatty acid metabolism and biosynthesis of unsaturated fatty acids), and PCYT2 (glycerophospholipid metabolism). The mRNA expression level of FADS1 was significantly decreased in the PPAT of obese with prostate cancers in our previous study using qRT-PCR [4]. Besides these, GO enrichment analysis showed that these genes are functionally related to receptor binding (neuropeptide receptor binding, dopamine receptor binding, and insulin receptor binding) and enzyme activity (acid phosphatase activity, metallocarboxypeptidase activity, and acylglycerol O-acyltransferase activity) (Additional file 14: Table S12, adjusted Pvalue < 0.25). Discussion This pilot study revealed significant differences of DNA methylation profiles between the PPATs from OB/OW versus NW PCa patients. Variations in global DNA methylation demonstrated that excess adiposity played an important role in DNA methylation level of PPAT tissues in prostate cancer patients, which provide an opportunity to explore the effect of obesity on PPAT epigenetic modification and subsequently on prostate cancer. These findings reported for the first time in PPAT depot are in concordance with previous works reporting that excess adiposity and BMI activate DNA methylation in adipose tissue [27–29]. Thus, considering the present understanding of the potential causal relationship between excess adiposity and cancer [30], diabetes [11], and cardiovascular disease [31], our results provide methylated candidate genes, which might foster research on the potential biological mechanisms underlying epigenetic regulation of PPAT by excess adiposity and prostate cancer. Given that DNA methylation of CpGs located at promoters and islands are associated with gene transcription silencing, we performed a strict filtering of DMCs and explored the biological functions of all promoter hypermethylated genes, aiming to find the critical methylated CpGs in the PPAT between the obese and normal weight PCa patients. Bioinformatic analysis showed that the enriched pathways were mostly involved in metabolic disorders, particularly fatty acid degradation and glycerolipid and choline metabolism. These pathways are known to mediate the pro-tumoral effect of white adipose tissue in tumors, thus contributing to tumorigenesis and metastasis [32,33], particularly in prostate cancer [5]. Findings from other oncological models highlight excess adiposity-associated impact in methylation Cheng et al. Clinical Epigenetics (2018) 10:54 Page 7 of 15 Fig. 3 Visualization and analysis of hypermethylated CpG sites in TAPBP promoter.aThe chromosome positions of hypermethylated CpG sites show that DMCs are located at chr6 (p21.32), which is in the region of TSS1500 (pink square) of TAPBP and at the location of CpG island 31. bMethylation levels of eight CpG sites in PPAT from OB/OW and NW PCa patients have shown a significant difference. cCorrelation analysis shows strong correlation (Spearman correlation coefficient r 2 0.73~ 0.97) between the eight hypermethylated CpG sites based on the mean βvalue difference of individual probes Cheng et al. Clinical Epigenetics (2018) 10:54 Page 8 of 15 markers known to associate with potential effect in the cancer microenvironment (e.g., aromatase, prostaglandin E 2 receptor in breast cancer) [34,35]. Obesity has also been shown to associate with methylation of cancerrelated genes (E-cadherin, p16, and RAR-β(2)) directly in malignant breast cells [36,37]. Pathway enrichment analysis showed a strong association between promoter hypermethylation of CPT1B, CPT1C,ACADM, and FADS1, with fatty acid metabolism. CPT1B (carnitine palmitoyltransferase 1B) and CPT1C (carnitine palmitoyltransferase 1C) genes encode rate-limiting enzymes in fatty acid degradation and play critical roles in long-chain fatty acid (LCFA) β-oxidation by controlling transportation of long-chain fatty acylCoAs from the cytoplasm across the outer mitochondria membrane [38]. Maple et al. reported that increased methylation of specific CpGs in the CPT1B promoter was correlated with decreased CPT1B transcripts in the skeletal muscle after lipid oversupply in severe obesity, which resulted in obese individual’s incapacity to increase fat oxidation, contributing to metabolic inflexibility [39]. Although the biochemical function of CPT1C has been verified to be necessary for the regulation of energy homeostasis in CPT1C knockout mouse brain [40], the study of CPT1C methylation was absent. CPT1B and CPT1C were previously reported to be highly expressed in the muscle, brain, and many other normal tissues including adipocytes [41]. Taken together, these findings suggest that methylation of specific CpG sites in the CPT1B and CPT1C promoters likely result in gene Table 3 Differentially methylated regions (DMR) in PPAT of obese/overweight PCa patients compared to normal weight controls DMR Chr Start–end (bp) Size (bp) Pvalue FDR Located gene DMCs* Relation to CpG island 1 6 30,038,791–30,039,801 1010 5.11E−05 2.07E−02 RNF39 37(0) Island 2 6 29,648,161–29,649,084 923 1.54E−03 2.08E−01 ZFP57 # 22(0) Open sea 3 17 71,228,123–71,228,832 709 3.10E−03 2.86E−01 FAM104A 4(3) Island 4 17 71,228,123–71,228,832 709 3.10E−03 2.86E−01 C17orf80 10(8) Island 5 12 42,720,006–42,720,167 161 5.56E−03 2.86E−01 PPHLN1 4(0) Island 6 6 31,650,735–31,651,158 423 5.59E−03 2.86E−01 MIR4646 # 16(0) Island 7 7 27,169,674–27,170,961 1287 5.66E−03 2.86E−01 HOXA4 17(11) Island 8 6 33,282,736–33,283,145 409 5.87E−03 2.86E−01 TAPBP 18(18) Island 9 20 57,463,763–57,464,129 366 6.74E−03 2.86E−01 GNAS 15(0) Island 10 16 86,546,938–86,547,322 384 7.03E−03 2.86E−01 FOXF1 4(0) Shore #The DMR is located at the intergenic region Chr chromosome *The number in the bracket is the quantities of DMCs located at the promoter (TSS200 and TSS1500) regions Table 4 Genes hypermethylated in promoters with significantly decreased gene expression Gene symbol DNA methylation Gene expression Correlation analysis Probe ID DNAm β diff. (%) Adjusted Pvalue (< 0.25) Probe ID FC Pvalue (< 0.05) Spearman’s rank correlation coefficient Pvalue (< 0.05) UCN cg20442078 5.6 0.17 8051061 −1.12 3.61E−02 −8.42E−01 2.23E−03 CCHCR1 cg00160818 1.9 0.17 8124868 −1.14 1.73E−02 −7.45E−01 9.21E−03 CRB3 cg14782015 4.3 0.20 8025041 −1.13 1.84E−02 −7.21E−01 1.21E−02 AGBL4 cg21834207 3.2 0.13 7915971 −1.17 1.29E−02 −6.73E−01 1.97E−02 INSL3 cg10174482 4.2 0.13 8035345 −1.13 4.94E−02 −6.36E−01 2.72E−02 ANKRD30B cg21293934 11.2 0.21 8069499 −1.17 2.24E−02 −6.08E−01 3.11E−02 FADS1 cg16213375 3.6 0.16 7948612 −1.8 9.55E−04 −5.88E−01 4.01E−02 PAPL cg18481683 2.3 0.24 8028570 −1.19 1.45E−02 −5.52E−01 5.21E−02 MOGAT1 cg12678667 4 0.15 8048725 −1.28 3.87E−02 −4.67E−01 8.91E−02 PPP1R1B cg09762778 5 0.12 8006865 −1.27 4.74E−02 −4.67E−01 8.91E−02 PRUNE2 cg00390775 4.4 0.15 8161884 −1.31 1.91E−02 −3.82E−01 1.39E−01 CIDEA cg18309817 1.8 0.18 8020211 −1.32 3.21E−02 −2.97E−01 2.03E−01 PCYT2 cg19583655 6.2 0.21 8019280 −1.26 6.38E−04 −1.88E−01 3.04E−01 SCUBE1 cg07697597 1.7 0.23 8076586 −1.23 9.03E−03 −4.24E−02 4.59E−01 FC fold change, DNAm βdiff. DNAm βdifference Cheng et al. Clinical Epigenetics (2018) 10:54 Page 9 of 15