Genome-wide differentially methylated genes in prostate cancer tissues from African-American and Caucasian men

Abstract

Increasing evidence suggests that aberrant DNA methylation changes may contribute to prostate cancer (PCa) ethnic disparity. To comprehensively identify DNA methylation alterations in PCa disparity, we used the Illumina 450K methylation platform to interrogate the methylation status of 485,577 CpG sites focusing on gene-associated regions of the human genome. Genomic DNA from African-American (AA; 7 normal and 3 cancers) and Caucasian (Cau; 8 normal and 3 cancers) was used in the analysis. Hierarchical clustering analysis identified probe-sets unique to AA and Cau samples, as well as common to both. We selected 25 promoter-associated novel CpG sites most differentially methylated by race (fold change > 1.5-fold; adjusted P < 0.05) and compared the β-value of these sites provided by the Illumina, Inc. array with quantitative methylation obtained by pyrosequencing in 7 prostate cell lines. We found very good concordance of the methylation levels between β-value and pyrosequencing. Gene expression analysis using qRT-PCR in a subset of 8 genes after treatment with 5-aza-2′-deoxycytidine and/or trichostatin showed up-regulation of gene expression in PCa cells. Quantitative analysis of 4 genes, SNRPN, SHANK2, MST1R, and ABCG5, in matched normal and PCa tissues derived from AA and Cau PCa patients demonstrated differential promoter methylation and concomitant differences in mRNA expression in prostate tissues from AA vs. Cau. Regression analysis in normal and PCa tissues as a function of race showed significantly higher methylation prevalence for SNRPN (P = 0.012), MST1R (P = 0.038), and ABCG5 (P < 0.0002) for AA vs. Cau samples. We selected the ABCG5 and SNRPN genes and verified their biological functions by Western blot analysis and siRNA gene knockout effects on cell proliferation and invasion in 4 PCa cell lines (2 AA and 2 Cau patients-derived lines). Knockdown of either ABCG5 or SNRPN resulted in a significant decrease in both invasion and proliferation in Cau PCa cell lines but we did not observe these remarkable loss-of-function effects in AA PCa cell lines. Our study demonstrates how differential genome-wide DNA methylation levels influence gene expression and biological functions in AA and Cau PCa.

Introduction

Prostate cancer (PCa) is a common malignancy and a leading cause of cancer death among men in the United States.¹ The incidence of PCa is approximately 60% higher, and the mortality rate is 2 to 3 times greater in African-American (AA) men compared with Caucasian (Cau) men, with AA men experiencing some of the highest rates worldwide.² These findings have persisted for more than 20 years, before and after the prostate-specific antigen (PSA) testing era.³ However, the disparity (differences among racial or ethnic groups) in PCa is believed to be a complex combination of socioeconomic factors, environment, and genetics.⁴ PCa cells are known to carry a variety of genetic defects, including gene mutations, deletions, translocations, and amplifications, which endow the cells with new biological capabilities known as hallmarks of cancer.^5,6 To date, results from both family studies and genome-wide association studies (GWAS) indicate that many genetic loci contribute to disease risk with varying levels of penetrance.^7,8 In addition to genetic alterations, it has become apparent that PCa cells also carry epigenetic defects, including changes in cytosine methylation patterns and chromatin structure and organization, which are equivalent to genetic changes affecting and maintaining neoplastic and malignant phenotypes.⁹

For human PCa, abundant evidence has accumulated to suggest that somatic epigenetic alterations such as DNA methylation may appear earlier during cancer development than genetic changes, as well as more commonly and consistently than genetic changes.¹⁰ If tumor biology does differ between AA and Cau men with PCa, it is also possible that biologic processes, including methylation, would differ by race as well. DNA methylation changes may represent an integration of lifestyle and genetic predisposing factors to create a more aggressive disease milieu in AA patients. We have previously shown differential methylation patterns between AA and Cau men with PCa.¹¹ In a study of both normal human prostate tissue and PCa, we used pyrosequencing to quantitatively measure the methylation status of 6 genes; namely, GSTP1, AR, RARβ2, SPARC, TIMP3 and NKX2–5. Five out of the 6 genes (except GSTP1) showed significant differences (P < 0.05) in AA samples vs. Cau samples. Furthermore, regression analysis revealed significantly higher methylation for NKX2-5 (P = 0.008) and TIMP3 (P = 0.039) in the normal prostate tissue samples from AA vs. Cau patients. These observations are provocative and suggest that methylation patterns for PCa may differ by ethnic/racial groups. With the advent of DNA methylation microarrays, including the Illumina Infinium HumanMethylation450 BeadChip, which offers high-throughput quantitative assessment of methylation across the genome, we were able to interrogate the DNA methylation pattern levels in prostate tissues from AA and Cau men.

In this paper, we compared the genome-wide DNA methylation pattern in normal and PCa tissue samples from AA and Cau men (focusing on gene promoter associated regions) and correlated with gene expression in PCa samples from AA and Cau men. The use of prostate cell lines derived from both AA and Cau men allowed us to validate the methylation of top candidate genes by quantitative pyrosequencing technique and to investigate the biological functions of novel genes in AA and Cau cells.

Results

Genome-wide methylation analysis of prostate tissues from African-American and Caucasian men with prostate cancer

In order to find CpG sites that are differentially methylated in normal and PCa tissue samples in AA vs. Cau men, we used the Illumina Infinium platform to conduct a genome-wide DNA methylation analysis. For our analysis, we included 3 matched pairs of normal and PCa tissue samples from both AA and Cau men who had undergone radical prostatectomy. In addition, we included 4 and 5 normal organ donor prostate tissue samples from AA and Cau men, respectively. Unsupervised hierarchical clustering analysis of all samples from AA and Cau men (Fig. 1) shows heterogeneity in individual prostate tissue samples. Nevertheless, the normal prostate tissue samples from organ donors and radical prostatectomy samples clustered together, whereas the cancer tissues clustered together for both the AA and Cau patient samples. Because the normal prostate tissues from the organ donors and radical prostatectomies clustered together, we grouped these samples together and hereafter refer to them as normal prostate tissues.

Figure 1.

Hierarchical cluster analysis of methylation findings from microarray data. DNA methylation hierarchical clustering shows high within-sample group similarity and between-sample heterogeneity. These analyses were done using Partek genome studio software with 450K methylation probes showing greatest variability across all samples. CpG methylation differences were considered significant above a cut-off P-value < 10⁻³ and 0.2-fold change in the β-value. The scale is based on β-value (M-value, as determined by Partek Genome Studio) for methylation score. AA: African-American; Cau: Caucasian; NI: Normal; Ca: Cancer.

Comparisons of the genome-wide methylation levels at CpG sites expressed as the ratio of cancer to normal tissue samples for both AA and Cau samples were made. The results, presented as Manhattan Plot (Fig. 2A), showed significant differences in hypermethylated and hypomethylated genes in the cancer vs. normal prostate tissues for both AA and Cau samples. Overall, we observed higher frequencies of differential methylated genes in AA samples when compared to Cau samples. We identified 3,303 and 1,075 probe-sets for AA and Cau patient samples, respectively, for which the difference of cancer vs. normal tissue samples showed at least 1.5-fold change in methylation level. The result shown in the Venn diagram (Fig. 2B) indicates that there are 2,973 and 745 probe-sets unique to the AA and Cau groups, respectively, and 330 probe-sets that is shared between the 2 groups. We have previously reported that several of these CpG sites are distributed in genomic regions that are not associated with gene promoter loci.¹²

Figure 2.

(A) Manhattan plot for differentially methylated gene analysis in cancer vs. normal prostate tissues in both African-American and Caucasian men. The Y-axis shows the –log₁₀ P-values of probe sets (P < 0.0005, and the X-axis shows their chromosomal position. Horizontal red line represented an arbitrary P-value of 0.001. The P-values were obtained using rank-sum Man-Whitney's U-test to compare cancer vs. normal for each group. (B) Venn diagram showing the number of differentially methylated genes probe sets of cancer vs. normal for AA and Cau groups and the number of overlapping probe sets between the 2 groups.

We focused on CpG sites in gene-associated regions of the human genome, defined as the sum of CpG sites located within 200 bp downstream or 1,500 bp upstream of the transcription start site in order to assess the effects of promoter methylation on gene expression. The top 25 novel genes most differentially methylated by race are shown in Table S1.

Pyrosequencing analysis of novel methylated genes

To confirm the accuracy of the Illumina methylation array, pyrosequencing assays were performed on the top 25 novel differentially methylated genes in normal and PCa cell lines derived from both AA and Cau men. A heat map hierarchical cluster analysis for the methylation status in the prostate cell lines is shown in Figure 3. The results showed that 21 of the 25 genes identified by the genome-wide array to be hypermethylated were also hypermethylated in at least one PCa cell line with higher methylation frequency observed for the majority of genes in the PCa cell lines. However, the methylation frequency was mixed across the PCa cell lines derived from AA and Cau men. Conversely, 4 genes, namely MBD6, MED17, TERF2IP, and TRAM2, which showed low methylation by β-value, also showed low methylation in PCa cells. Thus, the pyrosequencing approach confirmed the genome-wide array results. Overall, we identified ABCG5, CSDE1, CECR1, MST1R, PRAM1, PROCR, SHANK2, SNRPN, SPTB, ST3GAL6, STOX1, TRIO, TECTA, and UBE3C as most frequently hypermethylated in PCa cell lines.

Figure 3.

Heat map hierarchical cluster analysis of gene validation of 25 novel selected promoter-associated DNA methylated genes in a panel of prostate cell lines: RWPE1 and pNT1A are primary immortalized prostatic epithelial cell lines derived from Caucasian (Cau) patients; Cau PCa cell lines are LNCaP, PC3, and DU145; African-American (AA) PCa cell lines are MDA-PCa-2b and E006AA. Percent methylated cytosines in the cell lines were obtained by pyrosequencing. Methylated DNA control (M-DNA) and unmethylated DNA control (UM-DNA) were purchased from Qiagen. Fold change calculation is based on a ratio of log transformed values (FDR adjusted P-value < 0.05; β-value > 0.2, except for 4 CpG loci included in our study). Unsupervised hierarchical clustering analysis of the most variable β-values was done with a false discovery rate (FDR) adjusted P-value < 0.05 as significant.

Gene expression and induction after 5-aza-deoxycitidine treatment in prostate cell lines

For the top 25 candidate genes we observed an inverse correlation between DNA methylation and gene expression array data (Affymetrix data files were stored at the NCBI Gene Expression Omnibus, Accession # GSE64331; Table S2). In other words, genes that were highly methylated in AA cancer samples vs. Cau cancer samples exhibited lower expression at the mRNA transcript level in AA vs. Cau cancer, and genes with lower methylation in AA cancer vs. Cau cancer tended to have higher mRNA expression in AA vs. Cau cancer. To verify that hypermethylation affects gene expression, we tested the hypothesis that pharmacological modulation of methylation can reactivate gene expression.¹³ The LNCaP and DU145 PCa cell lines were each treated with the demethylating agent 5′-aza-deoxycitidine (5′-aza-dC), the histone deacetylase inhibitor, trichostatin A (TSA), or the combination of the 2 drugs, and gene expression analyzed by quantitative RT-PCR (qRT-PCR) using untreated and treated cells to assess whether promoter CpG hypermethylation was closely associated with gene expression in the newly identified cancer-specific methylated genes (Fig. 4). Treatment with either 5′-aza-dC or TSA caused over 2-fold increase in mRNA transcript expression levels for all genes at least in one cell line. The combined 5′-aza-dC and TSA treatment synergistically increased the mRNA transcripts for all genes in at least one cell line tested. The results indicate that epigenetic mechanisms, including DNA methylation, are likely involved in the reduced expression of ABCG5, ACOT7, MST1R, SPTB, SHANK2, SNRPN, and WDR70 genes.

Figure 4.

Demethylation and gene expression. The androgen-dependent prostate cancer (PCa) cell line, LNCaP and the androgen independent PCa cell line DU145 from Caucasian patients were treated with 5′-aza-2′-dC (5 μmol⁻¹), TSA (250 noml⁻¹), or a combination of the 2 drugs. These cell lines had high methylation levels for the genes tested and all had low levels of expression at baseline. The fold change in gene expression level relative to mock-treated cells was determined by quantitative RT-PCR and expressed relative to GAPDH to correct for variation in the amounts of reverse-transcribed RNA. The data are representative of 3 independent experiments. The standard deviation of the mean is shown as error bar.

Validation of methylated genes in prostate tissue samples

To determine whether the genes that were hypermethylated in PCa cell lines were also hypermethylated in a cancer-specific manner, we analyzed the methylation status of a sub-set of genes in the same tissue samples used in the array analysis, and additional 26 DNA samples from matched normal and PCa tissue samples obtained from AA men and 30 DNA samples from matched normal and PCa tissue samples obtained from Cau men who underwent radical prostatectomy for PCa (age range, 52–75 years). Patient clinical data for the 2 groups, with the Cau samples showing a non-significant slightly higher Gleason score, have previously been reported.¹¹ We analyzed 4 novel genes identified to be hypermethylated from the Illumina array and in the PCa cell lines, namely, ABCG5, MST1R, SNRPN, and SHANK2, that also demonstrated large increase in mRNA transcript expression levels in response to 5′-aza-dC treatment, as shown in Figure 4. For each studied gene, the percentage of methylation at a specific promoter was compared between the matched normal and PCa tissue samples (Fig. 5A). We observed hypermethylation in both the matched normal and PCa cells for all genes studied, although the methylation level was significantly higher in the cancer samples for SNRPN (P < 0.0001 for AA; P = 0.0021 for Cau), SHANK2 (P = 0.044 for AA), MST1R (P = 0.013 for AA), and ABCG5 (P = 0.0085 for AA). Regression analysis to examine whether methylation frequency in the normal and PCa tissue samples differed by race showed significantly higher methylation prevalence for SNRPN (P = 0.012), MST1R (P = 0.038), and ABCG5 (P = 0.00019) in AA samples vs. Cau samples. The concordance in the β-values and pyrosequencing observed in the cell lines above is also observed in the prostate tissue samples. Thus, high methylation by pyrosequencing corresponded with high β-values in the normal and cancer samples used in the Illumina array data (Supplementary data 3).

Figure 5.

(A) Quantitative DNA methylation analysis in human prostate tissues is shown in the box plot. The percent DNA methylation levels of promoter CpGs were analyzed in bisulfite-modified genomic DNA extracted from matched pair normal and prostate cancer (PCa) tissues (26 samples from AA; 30 samples from Cau) from men. Y-axis, percentage of methylated cytosines in the samples as obtained from pyrosequencing; X-axis, Nl (Normal) and Ca (Cancer) tissues. The box plot describes the median, interquartile range and maximum/minimum methylation. The P values were obtained using Mann-Whitney t-test. Pearson correlation and simple and multiple regression methods were used to compare methylation changes in cancer by race and cancer x race interactions. *Significant (P < 0.05) difference comparing normal and tumor. (B) Gene expression in prostate tissue samples. The relative mRNA transcript expression levels for SHANK2, MST1R, and SNRPN were analyzed in 32 matched pairs of Nl (normal) and PCa (prostate cancer) tissue samples by RT-PCR and expressed relative to GAPDH to correct for variation in amount of reverse-transcribed RNA. *Significantly different compared to cells treated with nonsense siRNA using an ANOVA with Holm post hoc test (P < 0.05).

To investigate if methylation is associated with the silencing of SHANK2, SNRPN, and MST1R genes in PCa tissues, we performed expression analysis by qRT-PCR using Cau samples that were available. The results, presented in Figure 5b, showed a modest but significant (P < 0.05) reduction in gene expression in PCa when compared with the matched normal prostate tissues for SHANK2 (1.3-fold change), MST1R (2-fold change), and SNRPN (1.65-fold change) at the mRNA transcript level. Our data suggest an inverse association between DNA methylation and gene expression for all 3 genes investigated. This data further support an inverse association between methylation and the gene expression data reported in Table S2. We observed that in cancer samples that showed higher methylation, this was associated with low levels of gene expression, whereas the normal prostate samples had lower methylation and higher gene expression level to indicate that methylation leads to some loss of gene expression.

Functional validation of methylated genes in PCa cell lines

For the ABCG5 and SNRPN genes, which were found to be preferentially hypermethylated in AA compared to Cau PCa specimens, we assessed the functional consequence of knocking down these genes in Cau PCa cell lines PC3 and LNCaP, and in AA PCa cell lines E006AA and MDA-PCA2b. Prior to knockdown, baseline ABCG5 and SNRPN protein levels were evaluated by Western blot analysis in all 4 PCa cell lines. We found markedly good agreement between protein expression and DNA methylation patterns. Specifically, protein expression was inversely correlated with DNA methylation of ABCG5 and SNRPN across the 4 PCa cell lines (compare Fig. 6A and Fig. 3). Targeted knockdown of ABCG5 and SNRPN was accomplished using 2 different siRNA for each gene and confirmed by both qRT-PCR and Western blot analyses. Gene knockdowns ranged from 60–90%, as determined by qRT-PCR, in agreement with Western blot analysis (Fig. 6B). Compared to nonsense sequence siRNA control, knockdown of either ABCG5 or SNRPN resulted in a significant decrease in both invasion and proliferation in Cau cell lines PC3 and LNCaP (Fig. 6C and D). In contrast, knockdown of the same 2 genes in AA cell lines E006AA and MDA PCa 2b did not result in the same profound overall loss of oncogenic effects (Fig. 6C and D). Notably, proliferation was not affected by either ABCG5 or SNRPN-knockdown, while a loss of invasion was only seen with SNRPN knockdown in the AA cell lines.

Figure 6.

(A) Basal ABCG5 and SNRPN protein expression in Caucasian (Cau) prostate cancer (PCa) cell lines PC3 and LNCaP and African-American (AA) PCa cell lines E006AA and MDA PCa 2b. (B) Western blot validation of ABCG5 and SNRPN gene knockdowns in PCa cell lines. Cells were transfected with a nonsense sequence siRNA control (siNS) or one of 2 different gene-specific siRNAs (1 or 2). Knockdown of protein levels ranged from 60–90% compared to cells transfected with siNS, in agreement with qRT-PCR results (data not shown). Data shown are representative of 3–6 independent experiments. The amount of lysate loaded on the gel varied across the different knockdowns. If a protein was weakly expressed in 6A, we loaded more lysate in the knockdown experiments to get better signal to noise with the knockdowns. (C) and (D) Effects of gene knockdown on proliferation (C) and invasion (D) in Cau and AA PCa cell lines. siRNA-mediated knockdown of either ABCG5 or SNRPN led to a decrease in both invasion and proliferation by PC3 and LNCap cells. In contrast, loss of oncogenic function in AA cell lines was limited to a decrease in invasion in only the SNRPN knockdown. Data are represented as the mean ± SEM of 3–6 independent assays. *Significantly different compared to cells treated with siNS using an ANOVA with Holm post hoc test (P < 0.05).

Discussion

The biological contribution to prostate cancer disparity in AA vs. Cau is an area of great research interest; however, very few studies have looked at the biological consequences of epigenetic DNA methylation and prostate cancer disparity. In the present study we have carried out genome-wide DNA methylation analysis to more comprehensively interrogate DNA methylation alterations and gene expression in PCa disparity.

Hierarchical clustering analysis of genome-wide DNA methylation provided partial differentiation of normal and tumor samples from both AA and Cau men who had undergone radical prostatectomy. We focused on CpG-methylated sites in proximal promoters defined as the sum of CpG sites located within 200 bp downstream or 1,500 bp upstream of the transcription start site in order to assess the effects of promoter methylation on gene expression and found numerous differences. Fidelity of the Illumina Infinium array was confirmed using pyrosequencing for genes selected according to our criteria described above. Overall, we identified 25 novel promoter-associated CpG sites as top candidate genes to be differentially methylated in AA vs. Cau PCa cell lines and whose methylation pattern inversely correlated with gene expression not only in the PCa cell lines but also in PCa tissues at least for 3 genes that were investigated. The observation of inverse correlation of gene expression and DNA methylation changes in AA and Cau PCa samples suggests that DNA methylation could contribute to the differential aggressiveness of PCa in AA and Cau patients. Overall, we observed significantly higher methylation prevalence in the PCa tissue samples from AA vs. Cau patients. The Cau samples used in this study showed slightly higher Gleason score (but not significant) and similar pathologic staging when compared with the AA samples. Thus, the higher prevalence of methylation seen in AA cancer samples is not simply reflective of differences in disease aggressiveness or stage between the 2 groups. In support of our studies, a previous report that compared the global gene expression and DNA methylation profiles in the endothelial cells isolated from tumor vs. normal human prostate from AA and Cau PCa patients illustrated a wide spectrum of expression and methylation perturbation of endothelial cells between tumor and normal prostate tissues in both AA and Cau study subjects.¹⁴ While this study, however, was limited to endothelial cells, it suggests the existence of ethnic group-specific alterations in tumor samples from AA and Cau cancer patients.

Most of the genes identified through our array analysis have never been shown to be methylated in PCa or in other cancer types. Four of the genes, namely MST1R, SHANK2, SNRPN, and ABCG5, were selected for further analysis in human prostate tissues obtained from both AA and Cau men because these genes demonstrated markedly good agreement between the differential methylation pattern, as demonstrated by the Illumina array, pyrosequencing, and gene expression data.

These novel genes are suspected to have diverse roles in normal and PCa pathogenesis, which underscore the myriad ways whereby these epigenetic aberrations may contribute to PCa disease etiology and/or progression. For instance, the MST1R (RON) is one of 2 members of the MET receptor tyrosine kinase family and is primarily expressed on epithelial cells and macrophages¹⁵ and plays a role in promoting epithelial tumorigenesis and macrophage activation.^16,17 In human prostate cancers, MST1R is reported to be overexpressed in about 90% of cases. However, hypermethylation of MST1R gene has been observed. One report indicates that hypermethylation of MST1R proximal promoter is associated with silencing of the full-length MST1R gene transcript, whereas hypermethylation of the distal promoter is associated with transcription of the short form of MST1R which has a constitutively active tyrosine-kinase activity that drives cell proliferation,¹⁸ suggesting that hypermethylation of MST1R may play a role in the biological activity of different MST1R transcripts. Another novel gene that we investigated, SHANK2, plays a role as a scaffolding protein and participates in the organization of the post-synaptic density region synaptogenesis.¹⁹ One observation reported by Beri et al.²⁰ suggests that SHANK2 and its homologues SHANK1 and 3 possess several methylated CpG sites; however, in hippocampal neurons, only SHANK3 but not SHANK 1 or 2 expression was downregulated in response to DNA methylation, indicating that DNA methylation may play a role in tissue-specific expression of SHANK2.

The ABCG5 gene belongs to a large family of ATP-binding cassette (ABC) transporters that are involved in the regulated transport of hydrophobic compounds principally lipids (dietary) across cellular membranes.²¹ Specifically, ABCG5 functions to limit intestinal absorption and promote biliary excretion of sterols.²² One study that investigated epigenetic changes with dietary soy in cynomolgus monkeys observed that ABCG5 methylation was among those that changed between diets.²³ Another genome-wide DNA methylation profiling identified ABCG5 methylation in close association with histone H3 in the mouse liver.²⁴ One study indicated that the methylation of ABCG2 was correlated with the downregulation of the gene expression in brain tumor stem cells,²⁵ suggesting that DNA hypermethylation may be playing a role in the downregulation of ABC transporter genes expression. Yet, another report indicates enhanced expression of ABCG2 in the recurrent PCa tissues.²⁶ Thus, additional studies are necessary in order to fully understand the biological consequences of ABCG5 hypermethylation in prostate cancer. The SNRPN gene is a maternal imprinted gene identified on human chromosome 15 whose paternal absence is responsible for Prader-Willi/Angelman syndrome (PWS).^27,28 The SNRPN gene encodes for a small nuclear ribonucleoprotein polypeptide whose protein product plays a role in pre-mRNA processing, possibly via tissue specific alternative splicing events. Tumor specific methylation changes has been reported for SNRPN in invasive breast cancer,²⁹ indicating that abnormal methylation of imprinted genes, such as SNRPN, may play a role in cancer progression. Studies carried out by Ribarska et al.³⁰ provide further support that imprinted genes are frequently dysregulated in prostate cancer. The expression of some imprinted genes, such as H19, maybe due to altered DNA methylation, while a group of imprinted genes are coordinately dysregulated independently of DNA methylation changes in prostate cancer.³⁰

To address the issue whether the Illumina array methylation data as measured in β-values for individual gene loci can be directly translated into percent methylation levels, we compared the β-values and pyrosequencing for the 4 genes using matched normal and PCa tissues from AA and Cau patients. We observed very high concordance between the β-values and the pyrosequencing methylation data. For instance, where we observed very high β-values, we also observed very high methylation by pyrosequencing in both normal and PCa samples for both AA and Cau men. The high methylation frequency observed in normal prostate tissues using either methods may be partially explained by the increase in DNA methylation as a function of prostate cancer. The samples that were used in this study were from individuals between 52 and 75 y old (mean age 62.7 ± 7.2 and 62.88 ± 8.8 for the AA and Cau groups, respectively) and we have previously observed high methylation of several genes in benign prostate tissues for this age range.³¹ The sensitivity of MST1R, SHANK2, ABCG5, and SNRPN methylation for distinguishing between matched normal and PCa tissues was lower to that of other well-known methylated genes, including GSTP1, RARβ2, and RASSF1A, which we have tested using the pyrosequencing technique.³¹ Nonetheless, we did, however, see significant differences in the methylation patterns in AA when compared to Cau. Overall, we observed significantly higher methylation prevalence for 3 out of the 4 genes (namely ABCG5, SNRPN, and MST1R) in AA prostate tissue samples when compared to Cau prostate tissue samples. To study whether these DNA methylation events correspond to functional changes, we assessed the functional consequence of knocking down ABCG5 and SNRPN on cell proliferation and invasion in AA and Cau cell lines. While the gene knockdown demonstrated significant inhibition of cell proliferation and cell invasion in Cau PCa cell lines, we did not observe such remarkable effect on cell proliferation and invasion in AA PCa cell lines. These results indicate that ABCG5 and SNRPN may play a role in PCa aggressiveness in Cau patients, but maybe a less important one in AA PCa. These studies are preliminary and limited to the few AA and Cau PCa cell lines available for functional studies. In addition, we only investigated one methylation loci for each gene promoter. Because our candidate genes have multiple CpG loci, it is important to investigate multiple methylation loci for each gene in order to identify which putative CpG loci is playing a major role in regulating gene expression. Despite these limitations, the differential patterns in genome-wide methylation levels, gene expression, and functional differences observed in AA and Cau prostate specimen suggests that genome-wide DNA approaches may provide a useful resource for testable hypothesis in prostate cancer disparity.

In summary, we have found that genome-wide methylation patterns differ by ethnic/racial groups, which suggest distinct differences in the etiology of PCa in AA vs. Cau. To our knowledge, this present study is the first to apply a genome-wide approach to investigate the association between DNA methylation and PCa disparity. Although we were able to determine that racial differences in methylation were greatest in AA compared to Cau, disentangling the relationships between DNA methylation and race will require analysis of a much larger sample set to evaluate the independent effects of tumor characteristics on DNA methylation patterns. Our future test will use genetic ancestry informative markers as proxy in the AA samples to determine whether genetic differences are correlated with the higher methylation among AAs. Finally, large studies and functional experiments are necessary to elucidate the role of these DNA methylation differences in PCa disparity.

Materials and Methods

Prostate Cell lines

The immortalized normal prostate epithelial cell line RWPE1 and the human PCa cell lines, PC3, DU145, and LNCaP, were obtained from the American Type Culture Collection (ATCC, Manassas, VA). The immortalized normal prostate epithelial cell line pNT1A was obtained from the European Collection of Cell Culture (Salisbury, UK). The above cell lines were all derived from Caucasian patient samples. The androgen-dependent PCa cell line, E006AA, and androgen-independent cell line, MDA-PCa-2b, were derived from AA PCa patients and obtained from ATCC, respectively. All cell lines were maintained in RPMI supplemented with 10% fetal bovine serum (Invitrogen, Carlsbad, CA), unless stated otherwise.

Prostate tissues

High molecular weight genomic DNA was extracted from freshly frozen PCa tissues and matched benign tissue were from radical prostatectomy specimens, as described previously.³² Cancer tissues were at least 70% cell positive cancer and benign tissues were free of cancer or high-grade prostatic intra-epithelial neoplasia, as confirmed by histology of frozen sections before DNA extraction. All tissues were from self-identified AA or Cau men. This study was performed with institutional review board approval. Pathological characteristics of these cases were previously described.³²

Genome-wide methylation analysis

DNA (0.5 µg) was bisulfite converted using the EZ DNA Methylation Gold kit (Zymo Research, Irvine, CA). Whole-genomic DNA was then amplified, enzymatically fragmented, precipitated, re-suspended, and hybridized at 48°C for 16 h to an array containing 485,577 locus-specific oligonucleotide primers (Illumina). The hybridization procedure and methylation score has previously been described.¹² All other computations and statistical analyses were performed using Partek Genome Studio. The methylation score for each CpG site is represented as a Beta value (β-value). The β value is a continuous variable ranging between 0 and 1, representing the ratio of the intensity of the methylated-probe signal to the total locus signal intensity. A β-value of 0 corresponds to no methylation while a value of 1 corresponds to 100% methylation at the specific CpG site measured. P-values were calculated to identify failed probes as per Illumina's recommendations and no arrays exceeded our quality threshold of >5% failed probes. In addition, we removed CpG sites on the X and Y chromosome^33-35 and removed CpG sites from the analysis that contained a single nucleotide polymorphism (SNP) or a SNP within 10 base pairs of the methylation probe,³³ according to dpSNP132. We only focused on CpG sites within known CpG islands from the UCSC database (n = 200,419 CpG sites). The sites in the UCSC database were discovered using a modified method from Gardiner-Garden and Frommer.³⁶ Raw data were normalized using Illumina's control probe scaling procedure³³ and background subtracted.

The β-values were imported into Partek Genomics Suite (version 6.6; Partek Incorporated, St. Louis, MO, USA) and underwent a logit transformation (M-value).³⁷ The M-value is calculated as the log₂ ratio of the intensities of the methylation probes vs. unmethylated probe. The problem of heteroscedasticity in the high and low ranges of methylation (<0.2 and ≥0.8) is resolved with the transformation of β-value to M-value. In addition, we removed noise from our analysis by examining only CpG islands with a β-value ≥ 0.1. The data was analyzed using a T-test between normal and PCa samples. The M-value score was treated as a continuous variable that is the response for the ANOVA model. An FDR correction was implemented but none of the identified CpG islands were significant. Therefore, we validated the results by pyrosequencing.

These β-values were used to calculate a ratio of relative methylation between samples, with higher values corresponding to greater levels of methylation (hypermethylation) in tumor tissue relative to normal. Hypomethylation is defined as methylation in normal prostate samples but undermethylation in tumor samples. The CpG methylation differences were considered significant above a cut-off P-value < 10⁻³ and 0.2-fold change in the β-value, unless specifically indicated otherwise. Fold change calculation is based on a ratio of log transformed M-values (FDR adjusted P-value < 0.05; β-value > 0.2; except for 4 CpG loci included in our study). Unsupervised hierarchical clustering analysis of the most variable β-values was done with a false discovery rate (FDR) adjusted P-value < 0.05 as significant. All other computations and statistical analyses were performed using Partek Genome Studio.

Pyrosequencing

DNA was bisulfite-converted using Epitect Bisulfite kit (Qiagen, Valencia, CA). Pyrosequencing was done using the PSQ HS96 Gold SNP Reagents on a PSQ 96HS machine (Qiagen, Gaithersburg, MD, USA). Primers designed in-house or commercially available PyroMark CpG assays consisting of PCR and sequencing primers were used according to the manufacturer's protocol (pyrosequencing primers are provided in Table S4). Epitect ready-to-use completely methylated and bisulfite-converted human control DNA was used as positive control and unmethylated and bisulfite-converted human control DNA was used as negative control. Each bisulfite PCR and pyrosequencing reaction was done at least twice.

Gene expression

Total RNA extracted from cells and prostate tissues using TRIzoL Reagent (Invitrogen) was used in cDNA synthesis using Invitrogen SuperScript^TM first strand synthesis system for qRT-PCR and according to the manufacturer's protocol. TaqMan^TM assays were used to quantitatively measure mRNA expression. Real-time PCR was carried out in a Bio-Rad iCycler real-time thermal cycler (Hercules, CA), as described previously,³⁸ and incorporating optimized PCR reaction conditions for each gene. The threshold cycle (C_t) in the PCR cycle at which fluorescence exceeds background was then converted to copy number based on a cDNA standard curve. Each experiment was carried out in duplicate.

Treatment with 5-aza-2′-deoxycytidine (5-aza-dC) and/or Trichostatin A (TSA)

LNCaP, DU145, and HPC8L cells were seeded at 5 × 10⁵ cells/100-mm tissue culture dish. After 24 h of incubation, the culture media was changed to a media containing 5′-aza-dC (5 µM) for 96 h and/or TSA (250 nM) for an additional 24 h. Total RNA extracted from cells and tissues using TRIzol Reagent (Invitrogen) were first used in first strand DNA (cDNA) synthesis using Invitrogen Super-Script^TM first strand synthesis and then used in real-time quantitative PCR, as previously described.³⁹ Mock-treated cells were cultured similarly. Three independent experiments were carried out for each analysis.

siRNA knockdown

PCa cell lines were plated at approximately 20% confluence in 6-well plates 24 h prior to transfection with 50 nM siRNAs (Thermo Fisher Scientific, Lafayette, CO). Cells were transfected separately with 2 different siRNAs targeting ABCG5 and SNRPN using DharmaFECT 4 reagent (Thermo Fisher Scientific) in penicillin/streptomycin-free DMEM media. A nonsense sequence siRNA (Thermo Fisher Scientific) was used as a negative control. All siRNA sequences are provided in Supplementary Table 5. Cells were incubated with siRNAs for 24 h before harvesting for invasion, proliferation, and qRT-PCR assays. Successful knockdown of genes was confirmed by qRT-PCR (knockdowns ranged 60–90% compared to nonsense siRNA-treated cells) and Western blot analyses.

Functional assays

For Matrigel invasion assays, invading PCa cell lines were measured at 48 h using BD BioCoat Matrigel Invasion Chambers (BD, San Jose, CA), as previously described.⁴⁰ Cell proliferation assay was performed using BrdU cell proliferation assay kit (Calbiochem, MA), as described by the manufacturer. Briefly, 3 × 10⁴ PCa cells were seeded in 96-well culture dishes containing 20 µl of BrdU labeling reagent and 100 µl culture medium, and cells were incubated for 16 h at 37°C. PCa cells were then fixed and denatured for 30 min at room temperature, and incubated with anti-BrdU antibody for 1 h at room temperature. Cells were washed 3 times, and incubated with peroxidase goat anti-mouse IgG HRP antibody (Calbiochem, Billerica, MA) at room temperature. After 30 min incubation, cells were washed (3 times) and 100 µl of substrate solution was added to each well and cells were incubated in the dark at room temperature for 15 min. The reaction was terminated by adding 100 µl of stop solution to each well. The relative apoptosis indexes were determined by measuring absorbance at dual wavelengths of 450–540 nm.

Statistical analysis

The methylation index at each gene promoter and for each sample was calculated as the average value of mC/(mC + C) for all examined CpG sites in the gene and expressed as the percentage of methylation. Statistical significance was judged by the appropriate Mann-Whitney t test, Student's t test, Pearson correlation, and simple and multiple regression methods were used to compare the categorical variables of methylation changes in cancer by race and cancer x race interactions. Data analysis was performed using either Prism 4 software (GraphPad Software, Inc.) or SPSS for Windows (version 13.0, SPSS). For each gene, the frequency of methylated and unmethylated cases was analyzed with nonparametric tests, i.e., Kruskal-Wallis one-way ANOVA, followed by the Bonferroni-adjusted Mann-Whitney U test, where appropriate. Significance was set at P < 0.05.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplemental Material

Supplemental data for this article can be accessed on the publisher's website.

Funding

This work was supported by grant from Department of Defense Program Idea Award; PC101996 to Bernard Kwabi-Addo and PC121975 to Norman H Lee. This work is also supported by the use of facilities at Howard University Department of Biochemistry & Molecular Biology, Children's National Medical Center and George Washington University.