Previous findings from several laboratories including our laboratory have demonstrated that Nrf2 plays an essential role in the development of various cancers 
. Nrf2 regulates the expression of antioxidant and phase II detoxifying enzymes including NQO1, HO-1 and GST 
. Therefore, the control of transcriptional activation of Nrf2 and Nrf2-target genes would appear to be an important homeostatic mechanism that protect cellular injuries or damages resulted from oxidative stress 
. Nrf2 deficiency could lead to defect in the cellular defense system against oxidative stress, potentially resulting in cancer initiation, promotion and progression 
. The repressed expression of antioxidant and detoxifying enzymes such as GSTP1 in prostate cancer has extensively been studied 
. However the role of Nrf2 in prostate cancer have not received enough attention until recently 
Frolich et al. reported that the down-regulation of Nrf2 appears to be responsible for the reduced GST expression, elevated oxidative stress and DNA damage in prostate tumorigenesis in TRAMP mice 
. We have recently found that the expression of Nrf2 as well as Nrf2-target genes is gradually down-regulated during the progression of prostate cancer in TRAMP mice 
. Previous analysis of the online human prostate gene expression data sets demonstrated that the expression of Nrf2 and GST 
as well as NQO1 was gradually decrease during human prostate carcinogenesis (Figure S4
). Furthermore, it has been reported that several GST genes are down-regulated in primary but not in metastatic TRAMP tumors 
. In current study, we found that the expression of Nrf2 and the induction of NQO1 was compromised in tumorigenic TRAMP C1 cells but not in non-tumorigenic TRAMP C3 cells (). This suppression of Nrf2 expression and Nrf2-target gene NQO1 in both of these TRAMP cell lines would exclude the possibility that Nrf2 expression would be affected by the SV40 transgene, since these TRAMP cell lines do not express the SV40 transgene 
DNA methylation has been implicated in the silencing of the GSTP1 gene in human prostate cancer, and similarly DNA-methylation silencing of several other genes are also implicated in TRAMP prostate tumor 
. However, interestingly MassARRAY Quantitative DNA Methylation Analyses (MAQMA) analysis of the 5′ region of several GST genes displayed no significant differences between normal prostatic epithelial cells and prostate tumor from the TRAMP mice 
. Recently, several reports show that in both TRAMP and Rb−/− prostate tumors, an Rb/E2F-dependent increase of DNMT1 expression and methylation activity 
. Hypermethylation Nrf2 promoter was ruled out using MSP and that 5-aza treatment had no effect on Nrf2 expression (with data not shown) 
. We analyzed the 5′-flanking region of Nrf2 gene and identified a CpG island that extends to position -1175 (). Using bisulfite sequencing, which would be more specific in identifying CpG methylation and would reveal more details about DNA methylation than MSP, we found that the first 5 CpGs in the CpG island are hypermethylated in TRAMP prostate tumors and in the tumorigenic TRAMP C1 cells but not in normal prostate tissues and non-tumorigenic TRAMP C3 cells (). Remarkably, these 5 CpGs are located adjacent to the previously reported Nrf2 promoter 
. Thus, the methylation status of these specific CpGs appears to be correlated with the tumorigenicity as well as Nrf2 expression and NQO1 induction ( and ). Notably, similar pattern of specific methylation of the distal CpG island has also been observed in the Keap1 gene in lung cancer cells 
. It is important to note that some of our current findings appear to be somewhat contradictory with the results reported previously 
, however previous findings of extensive down-regulation of Nrf2 and GST during prostate tumor progression in TRAMP mice 
, are consistent with our current results.
To determine the functional role of methylation of these 5 CpGs in the suppression of Nrf2 expession, luciferase reporters of the Nrf2 promoter with or without these 5 CpGs were constructed (). The Nrf2 promoter possesses very GC-rich non-canonical promoter which contains neither TATA box nor a CCAAT box 
, however, this Nrf2-promoter potently activated the transcription of luciferase reporter gene (). Interestingly, the addition of sequences from −1065 to −1367 appears to be repressive to the transcriptional activity of the Nrf2 promoter (). Such repressive sequence could function by recruiting specific repressing factors 
, but the exact mechanism accounting for this repressive function of this sequence even in the absence of methylation would require further investigation. Nevertheless, when the reporters were methylated in vitro
by CpG methyltransferase, the luciferase reporter activity of the Nrf2 promoter with the additional sequence containing the 5 CpGs (pGL-1367) was reduced by about 84%. In contrast, methylation of the reporter without the additional sequence resulted in about 23% reduction (). This is probably due to the heavy methylation of the whole construct including the luciferase gene (but further study would be needed to prove this). Altogether, these results suggest that the extra 5 CpGs (−1065 to −1367) could play a critical role in methylation-dependent suppression of Nrf2 promoter activity.
The role of CpG methylation in suppressing Nrf2 expression and activation was tested by treatment with 5-aza in TRAMP cells. 5-Aza has previously been shown to be able to prevent early disease progression, delay androgen-independent disease and improve survival of TRAMP mice 
. In addition, stage and phenotype-specific CpG island methylation and DNA methyltransferase expression have been well documented during prostate cancer progression in TRAMP mice 
. These published findings suggest the relevance of using this TRAMP system to interrogate the possible role of epigenetic alterations in prostate carcinogenesis. 5-aza treatment of TRAMP C1 cells modestly increased the mRNA level of Nrf2, while combined treatments of 5-aza and TSA induced a more prominent increase of Nrf2 (). This result is consistent with the report by Mavis et al., in which combined 5-aza and TSA treatments significantly enhanced the expression of GST genes 
. The protein level of Nrf2 in TRAMP C1 cells remained unaffected by either 5-aza or 5-aza/TSA treatments, however, addition of a potent Nrf2-activator tBHQ would enhance the accumulation of Nrf2 protein (). As expected, the induction of NQO1 mRNA and protein levels displayed a similar trend with that of the Nrf2 protein. It is highly likely that since Nrf2 signaling is primarily regulated via post-translational mechanisms, without challenges with Nrf2-activators such as tBHQ, Nrf2 protein would be rapidly turned over by proteosome-dependent degradation 
. The precise reason as to why tBHQ is needed for NQO1 induction would require further study.
To further delineate the molecular mechanism by which the specific CpG-methylation suppresses Nrf2 expression, ChIP assays were performed. The result reveals that the binding of MBD2 and H3K9m3 to the specific CpGs was substantially higher in TRAMP C1 cells than in TRAMP C3 cells correlating with the fact that the CpGs were minimally methylated in TRAMP C3 cells (). In contrast, AcH3 displayed an opposite binding pattern to the same CpGs sequence in TRAMP C1 cells as compared to TRAMP C3 cells, and similarly the binding of Pol II to the transcription start site showed similar pattern as AcH3 (). These methylation-dependent associations of corepressors could be modulated by 5-aza and TSA treatments and our results () correlated very well with the transcription level of Nrf2 (mRNA level) in these two cell lines (). MBD2 has been reported to mediate epigenetic silencing of 14-3-3σ in TRAMP C1 cells and human LNCaP prostate cells 
, and has been shown to be involved in the transcriptional repression of GSTP1 in MCF-7 breast cancer cells 
. MBD2 and other MBD proteins bind to methylated CpGs and recruit corepressor complexes which contain HDACs, chromatin remodeling proteins as well as other proteins leading to the repression of the expression of hypermethylated genes 
. Interestingly, the protein level of MBD2 was much higher in TRAMP C1 cells than in TRAMP C3 cells (), suggesting a possible common MBD2-mediated epigenetic suppression of Nrf2 in TRAMP C1 cells. In contrast, the protein level of AcH3 was apparently lower in TRAMP C1 cells than in TRAMP C3 cells but was increased tremendously by TSA treatment (). Since the expression of MBD2 would be dependent on HDAC activities, our above results would potentially explain the requirement of HDAC inhibitors in order to effectively modulate corepressors binding and to maximally restore the reexpression of Nrf2 as well as induction of NQO1 in TRAMP C1 cells. In addition, when this suppressive sequence containing the extra 5 CpGs was further analyzed using the Transcriptional Elements Search System (TESS, http://www.cbil.upenn.edu/tess
) based on the TRANSFAC V6.0 database, several transcription factor binding sites were identified, including the binding sites of E2F1-p107 and NF-E2 (data not shown). The exact binding proteins and their functions leading to the suppression of Nrf2 expression would require further investigation.
In summary, our present results clearly demonstrate that the expression of Nrf2 is suppressed by promoter CpG methylation in TRAMP prostate tumors. The existence and possible biological consequences of such epigenetic mechanism in the regulation of Nrf2 expression in human prostate cancer is currently under investigation in our laboratory. To the best of our knowledge, this is the first report revealing the epigenetic regulation of Nrf2 in prostate tumorigenesis. These findings would certainly open the door for further study on the role of Nrf2 as a plausible target for cancer chemoprevention and a possible diagnostic marker for detection of human prostate cancer.