Hypermethylation trigger of the glutathione-S-transferase gene (GSTP1) in prostate cancer cells

Article metrics


Understanding what triggers hypermethylation of tumour suppressor genes in cancer cells is critical if we are to discern the role of methylation in the oncogenic process. CpG sites in CpG island promoters, that span most tumour suppressor genes, remain unmethylated in the normal cell, despite the fact that CpG sites are the prime target for de novo methylation by the DNA methyltransferases. The CpG island-associated with the GSTP1 gene is an intriguing example of a CpG rich region which is susceptible to hypermethylation in the majority of prostate tumours and yet is unmethylated in the normal prostate cell. In this study we evaluate a number of factors purported to be involved in hypermethylation to test their role in triggering hypermethylation of GSTP1 in prostate cancer DU145 and LNCaP cells. We find that hypermethylation is not associated with (1) elevated expression of the DNA methyltranferases, or (2) removal of Sp1 transcription factor binding sites in the CpG island or (3) removal of CpG island boundary elements or (4) prior gene silencing. Instead our results support a model that requires a combination of prior gene silencing and random ‘seeds’ of methylation to trigger hypermethylation of the GSTP1 gene in the prostate cancer cell. We propose that the GSTP1 gene is initially silenced in the prostate cancer and random sites of methylation accumulate that result in subsequent hypermethylation and chromatin remodelling.


Methylation at the 5′-position of cytosine in CpG dinucleotides is a common modification of DNA in vertebrate genomes. However not all CpG sites are susceptible to methylation resulting in compartmentalization of the ‘methyl’ genome. Repeated sequence DNA and the CpG-depleted regions of the genome are mostly hypermethylated, whereas the CpG rich clusters termed CpG islands, span promoter regions of house-keeping genes and are typically unmethylated. The biological significance of DNA methylation in vertebrates is still debated, but it is thought to play a major role in gene silencing of repeated sequences (Turker and Bestor, 1997) and possibly plays a role in modulating tissue-specific gene expression (Warnecke and Clark, 1999). What is clear is that DNA methylation is critical for normal development, as embryos die at implantation in knockout mice that are lacking the DNA methyltransferase gene Dnmt1 (Li et al., 1992). Three additional DNA methyltransferases have now been described; Dnmt2 displays no activity in vitro, but Dnmt 3a and 3b have been shown to exhibit de novo methylation activities (Hsieh, 1999) and are also critical for embryo development (Okano et al., 1999).

In human cancers the DNA methylation pattern of the genome is altered, with the repeated region of the genome often demethylated and the CpG island-associated genes often hypermethylated. In particular the CpG islands spanning the promoter regions of tumour-suppressor genes are methylated early in cancer progression (Baylin et al., 2001). Multiple tumour suppressor or tumour-associated genes are methylated in any one cancer cell but the sub-set of genes that are susceptible to methylation are cancer-subtype specific (Melki et al., 1999). Methylation of the CpG island region of genes is associated with their inactivation and therefore abnormal DNA methylation has been described as a contributory factor equal in importance to gene mutation and gene deletion in mediating gene silencing in cancer (Costello and Plass, 2001; Jones and Laird, 1999). What triggers the DNA methylation changes in the cancer cell and in particular what triggers the hypermethylation of normally unmethylated gene promoter regions is a critical but unanswered question in cancer biology.

One means to address the mechanism that is responsible for aberrant hypermethylation in cancer is to determine what protects CpG islands normally from methylation. Interestingly the CpG dinucleotide is the prime target for methylation by the DNA methyltransferase enzymes, however CpG sites in CpG islands are never methylated during development when the rest of the genome is undergoing major de novo methylation changes (Warnecke and Clark, 1998). A number of possibilities have been proposed. Firstly, that the DNA methyltransferase enzymes DNMT1, 3a and 3b are moderately elevated in cancer cells (Robertson et al., 1999) and this may be sufficient to elicit the changes observed in the cancer cell genomic methylation pattern. Over expression of DNMT1 can result in aberrant promoter CpG island methylation in vitro (Vertino et al., 1996), however a bi-allelic knockout of the same gene in a cancer cell did not abolish maintenance of the CpG island methylation (Rhee et al., 2000), suggesting other DNA methyltransferases may be responsible. Secondly, protein factors may protect CpG islands from methylation by binding interference. Mutation of Sp1 sites in the adenine phosphoribosyl transferase (APRT) gene led to the hypermethylation of the associated CpG island in transfection assays and transgenic mouse studies (Brandeis et al., 1994; Macleod et al., 1994). Thirdly, RNA transcripts from CpG islands may themselves protect the islands either directly or indirectly in combination with 5MeC-DNA glycosylase (Jost et al., 1997). Finally, replication timing or chromatin structure may inhibit CpG island methylation (Kass et al., 1997).

The CpG island promoter region spanning the glutathione S-transferase (GSTP1) gene becomes methylated in the majority of prostate tumours. The gene is expressed and unmethylated in all normal tissues and is not commonly methylated in other cancers, with exception of liver and breast cancer (Esteller et al., 1998; Jhaveri and Morrow, 1998; Tchou et al., 2000). Intriguingly no mutations or deletions have been reported for GSTP1 in prostate cancer, however the gene is inactivated and both alleles are commonly methylated (Millar et al., 1999). Therefore, the GSTP1 CpG island becomes an ideal candidate to study what triggers its hypermethylation in prostate cancer cells. We have previously described the methylation profile across the CpG island of the GSTP1 gene in normal and prostate cancer cells and found that the gene is extensively methylated and inactivated in prostate cancer cells, including the prostate cancer cell line LNCaP (Millar et al., 1999, 2000). In contrast, the cancer cell line DU145 behaves more like a normal prostate cell with GSTP1 gene essentially unmethylated and expressed. A detailed methylation analysis in normal cells revealed a marked boundary of methylation at the 5′ flank of the of the island that separated the methylated and unmethylated regions of the gene and this boundary correlated with an ATAAA19–24 repeated sequence. In prostate cancer cells this marked methylation boundary is disrupted or bypassed as the entire CpG island, on both alleles is methylated.

In this study we compare the methylation susceptibility of various constructs of the GSTP1 gene stably transfected into DU145 cells, that harbour an unmethylated endogenous gene and LNCaP cells, that harbour a methylated endogenous gene, in order to address the mechanism that may be responsible for triggering such hypermethylation (Figure 1). Using this system we evaluate the role of the DNA methyltransferase enzymes, promoter factors and putative boundary elements in eliciting hypermethylation of the GSTP1 CpG island in prostate cancer cells.

Figure 1

GSTPI vector contructs. (A) Map of the GSTPI gene showing the position of the 7 exons, and the ATAAA repeat element. Below is an expanded map of the GSTP1 shuttle vector map containing the 2.3 kb XbaI/SnaBI cloned into pBluescriptSK+. The restriction enzyme sites in the GSTP1 and plasmid sequence and co-ordinates are shown. (B) CpG map of the region of the GSTP1 gene from CpG −56 to CpG +10 is shown. The scale is marked relative to co-ordinates from GenBank Accession Number M24485). The modifications to the various GSTP1 constructs are indicated in each of the CpG maps. *, marks the location of HpaII sites within the GSTP1 CpG island map


GSTP1 methylation in LNCaP and DU145 cell lines

Hypermethylation of the GSTP1 CpG island is an early and common event in prostate cancer (Brooks et al., 1998; Cookson et al., 1997; Esteller, 2000; Lee et al., 1994; Millar et al., 1999; Moskaluk et al., 1997). We previously reported that the methylation pattern across the GSTP1 CpG island in the prostate cancer cell line LNCaP is extensive and the gene is silent (Millar et al., 1999, 2000). Essentially all the CpG sites are methylated extending past the ATAAA repeat motif through the promoter region and into the body of the gene (Figure 2). In contrast, the GSTP1 gene that is actively expressed in the prostate cancer cell line DU145, shows a methylation profile more typical of normal prostate cells where the ATAAA repeat sequence separates the upstream methylated CpG sites from the unmethylated CpG island. Previously, by direct bisulphite PCR sequencing, we showed that the promoter region of the GSTP1 gene is completely unmethylated (Millar et al., 1999). However by cloning the PCR product and sequencing individual clones we find that about one-third of the molecules in the DU145 cells show a small amount of methylation around the start of transcription (Figure 2). This pattern of low level methylation around the transcription start site is similar to the methylation pattern in normal liver in which expression of GSTP1 is reduced (Millar et al., 2000).

Figure 2

Methylation Map of the GSTP1 gene in LNCaP and DU145. The relative position of the 7 exons of the GSTP1 gene and the position of the (ATAAA)19–24 repeat element are shown. The CpG density plot of the GSTP1 gene is shown with the base number below the line and the number of the CpG site relative to the start of transcription indicated above. The transcription start site is shown by an arrow. The region of the CpG island that was analysed by bisulphite sequencing is shown with the individual CpG sites numbered and the corresponding base number indicated below (GenBank accession number M24485). The methylation profile of the endogenous GSTP1 gene from individual clones sequenced from bisulphite treated and amplified LNCaP and DU145 DNA is shown. Closed circles indicate a methylated CpG and open circles indicate an unmethylated CpG. The arrow indicates the start of transcription; X the lack of transcription

DNA methyltransferase expression in LNCaP versus DU145

To determine if the difference in the methylation state of the endogenous GSTP1 gene in LNCaP and DU145 is related simply to the level of the DNA methyltransferases between the two cell lines we determined the expression of the DNA methyltransferase enzymes DNMT1, 2, 3a and 3b by quantitative real-time RT–PCR (Taqman). Figure 3 shows the relative expression profile of the different DNA methyltransferases. DNMT1 is the most highly expressed at the RNA level in both cell lines and is a least threefold higher than DNMT3a expression and at least 30-fold higher than DNMT3b and DNMT2. Interestingly, both DNMT2 and 3b are expressed approximately threefold higher in LNCaP than DU145 cells whereas the expression of DNMT1 and DNMT3a is similar in both cell lines. These results suggest that the difference in the methylation state of GSTP1 CpG island between LNCaP and Du145 cells is not merely due to a substantial increase in the level of expression of the predominant DNA methyltransferases (DNMT1 and 3a). Any possible biological significance of a threefold increase of DNMT2 and 3b in LNCaP versus DU145 is not clear, however it has been proposed that DNMT3b may be a key player in de novo methylation and different forms of the same enzyme may have altered target site specificity (Robertson et al., 1999).

Figure 3

DNA methyltransferase expression levels in LNCAP and DU145 cells. Fluorescence quantitative RT–PCR was used to estimate the relative expression of DNMT1, DNMT2, DNMT3a, DNMT3b. The scales on the graph represent the mean expression levels as a ratio between GSTP1 and beta-actin expression

Is GSTP1 a target for hypermethylation?

To address what might trigger abnormal hypermethylation of the GSTP1 CpG island in prostate cancer cells, such as LNCaP, we initially asked whether the GSTP1 gene sequence itself is a target for abnormal hypermethylation in LNCaP cells. That is, have the LNCaP prostate cancer cells lost the ability to specifically protect the GSTP1 CpG island from methylation whereas the DU145 cells essentially maintain the protective mechanism? We cloned a 2.3 kb XbaI/SnaBI fragment containing the GSTP1 gene including 1.2 kb upstream from the start of transcription and exons 1–4 and inserted a polyadenylation signal 3′ to the GSTP1 fragment to ensure transcript stability (GSTP1 shuttle) (Figure 1a). The GSTP1 shuttle construct was co-transfected with pC1-neo into LNCaP and DU145 cells and G418 resistant colonies selected. Approximately 20 resistant colonies were pooled respectively from each cell line and passaged. DNA was isolated after a number of serial passages and analysed for methylation of the transfected GSTP1 gene. Figure 4 shows that in both cell lines the CpG island, spanning the promoter and start of transcription from the transfected GSTP1 shuttle, remained essentially unmethylated even after 22 doublings. It is interesting to note that random single CpG sites became methylated in a few molecules in both cell types. The GSTP1 island is not susceptible to de novo methylation in LNCaP or DU145 cells unlike the LNCaP endogenous gene counterpart. Therefore the GSTP1 CpG island sequence does not appear to be specifically targeted for methylation in the LNCaP prostate cancer cells.

Figure 4

Methylation profile of the GSTP1 shuttle transfected into DU145 and LNCAP. DNA was isolated after 22 doublings from LNCaP and DU145 cells that were stably transfected with the GSTP1 shuttle vector. After bisulphite-treatment and PCR amplification the CpG sites −41 to +10 were analysed for methylation by cloning and sequencing a number of clones from at least 2–3 PCR reactions, respectively. The GSTP1 sequence of transfected shuttle DNA was distinguished from endogenous GSTP1 sequence by the addition of an extra CpG at NarI (1061). Closed circles indicate a methylated CpG and open circles indicate an unmethylated CpG

Does transcription protect from hypermethylation?

It is possible that active transcription protects CpG island sequences from hypermethylation, either through an open chromatin structure, binding of transcription factors or possibly due to interaction of the RNA transcript itself. We therefore tested whether the GSTP1 shuttle construct used for the transfection experiments expressed GSTP1 in the two cell lines. Using quantitative real-time RT–PCR we find that in contrast to LNCaP cells where the endogenous GSTP1 is silent the GSTP1 transfected copy is actively transcribed (Figure 5). However the level of transcription in the LNCaP cells was 10-fold less than the level of transcription of the endogenous gene in DU145 cells, possibly indicating that LNCaP cells have a reduced capacity to support GSTP1 expression. The transfected copy of GSTP1 is also transcribed in DU145 cells as seen by the twofold increase of GSTP1 expression in the transfected DU145 cells versus the endogenous GSTP1 expression. The GSTP1 shuttle vector is transcriptionally active in both cell lines and therefore it is possible that the process of active transcription is protecting the transfected GSTP1 CpG island from de novo methylation.

Figure 5

Quantitative RT–PCR analysis of GSTP1 expression. Expression levels from the endogenous GSTP1 gene in DU145 and LNCaP cells and from the various GSTP1 constructs were estimated using fluoresence quantitative RT–PCR. The scale represents the mean expression levels as a ratio between GSTP1 and beta-actin expression

It has been proposed that Sp1 sites and the binding of Sp1 transcription factor may also protect CpG islands from methylation (Brandeis et al., 1994; Macleod et al., 1994). Therefore to test the effect of Sp1 sites and active expression on the methylation potential of the GSTP1 CpG island we transfected two modified GSTP1 constructs into LNCaP cells. We only used LNCaP cells for these transfection experiments because in contrast to DU145 cells, LNCaP cells contain the ability to abnormally de novo methylate the endogenous GSTP1 gene. Therefore we argued that LNCaP cells must have or have had the capacity to initiate and maintain hypermethylation of the transfected gene. The two modified constructs were (1) a Sp1 mutation construct where the two Sp1 sites spanning CpG site 6 and CpG site 8 in the promoter region of the GSTP1 gene were mutated; these sites have previously been shown to be essential for expression (Moffat et al., 1996b); (2) a promoter deletion construct where CpG sites −2 to −8 spanning the Sp1 sites were deleted; this included the minimal promoter region including an essential AP-1 recognition sequence (Moffat et al., 1996b). We co-transfected the two modified constructs into LNCaP cells with pC1-neo and selected G418 resistant colonies. Approximately 20 colonies were pooled and serially passaged. Mutation of the two SP1 sites and the deletion of the minimal promoter both resulted in silencing of GSTP1 expression, as shown in Figure 5. Therefore, to determine if GSTP1 inactivation was sufficient to trigger de novo methylation of the CpG island we isolated DNA from the stably transfected cell lines after a number of serial passages (p2–p35 corresponding to 22–70 doublings) and analysed the methylation state across the transfected CpG island using bisulphite sequencing. Figure 6a,b shows that neither mutation of the Sp1 binding sites nor promoter deletion was sufficient to trigger hypermethylation of the GSTP1 CpG island in the LNCaP cells even after 35 passages (70 doublings). As noted for the GSTP1 shuttle vector, that was actively expressed, we did observe a few single sites of apparently random CpG methylation in some molecules of the modified constructs. However only a limited number of molecules showed evidence of accumulation or spreading of methylation with increasing passage number. Therefore interference of Sp1 or AP1 binding or inactive transcription do not appear to be involved in triggering GSTP1 CpG island hypermethylation in LNCaP cells.

Figure 6

Methylation profile of the Sp1 mutation, promoter deletion and ATAAA deletion constructs transfected into LNCAP. (A) The promoter deletion construct and (B) The Sp1 mutation construct were stably transfected into LNCAP cells and DNA harvested after 22, 45 and 70 doublings (22d, 45d and 70d). (C) The ATAAA deletion construct was stably transfected into LNCaP cells and DNA was isolated after 22 doublings (22d). The methylation status of the CpG island from CpG site −41 to +10 was analysed by bisulphite sequencing a number of clones from 2–3 PCR reactions. Clones from the promoter deletion construct indicate that CpG sites −2 to −8 are deleted. Clones from the Sp1 mutation construct show grey shaded circles at the mutated Sp1 CpG sites that are not available for methylation

Does the ATAAA repeat domain protect from hypermethylation?

Our previous studies of the GSTP1 CpG island identified a boundary region that separates the methylated and unmethylated regions of the gene at the 5′-flank of the CpG island and this boundary correlated with an (ATAAA)19–24 repeated sequence (Millar et al., 1999). In DNA from normal tissue, including prostate tissue, methylation occurred 5′ to the repeat but not 3′ to the repeat. However, in most prostate tumour tissue and in LNCAP cells there is extensive CpG methylation 3′ to the repeat spanning the entire island (Figure 2). It is therefore possible that the ATAAA repeat may act as a physical barrier to methylation or may be a binding site for proteins which block methylation in the normal cell but may be lacking in the prostate cancer cell permitting hypermethylation. To determine the role of this repeat structure in the methylation potential of the GSTP1 DNA we prepared a GSTP1 construct where the entire repeat domain of the GSTP1 gene was deleted (Figure 1b). This ATAAA deletion construct was cotransfected with pC1-neo into LNCAP cells and stable colonies were selected by G418 resistance. Approximately 20 colonies were pooled and harvested after 22 doublings. Removal of the ATAAA repeat unit did not adversely affect transcription of the GSTP1 gene (Figure 5) indicating that this sequence does not harbour any critical transcription factors or enhancer elements. The ATAAA deleted transfected GSTP1 gene was analysed for methylation across the CpG island by genomic bisulphite clonal sequencing. Analysis of the individual clones revealed that like the other constructs there was little de novo methylation of the transfected GSTP1 sequence with only a few sites of methylation in some molecules (Figure 6c). Therefore the ATAAA repeat does not appear to contain binding sites for factors that protect the CpG island from hypermethylation, as removal of the repeat did not alter the methylation pattern observed in the transfected DNA.

Is hypermethylation triggered by ‘seeding’?

It was intriguing that all the GSTP1 constructs that we transfected into LNCaP cells were resistant to extensive de novo methylation, even though the endogenous GSTP1 gene was maintained in a hypermethylated state. It was also interesting that the de novo methylation we did observe in all the constructs both non-expressing (Sp1 mutation and promoter deletion) and expressing (shuttle and ATAAA deletion) showed a low level, random methylation which appeared to accumulate with multiple passaging in some molecules. We therefore decided to ask whether random methylation may provide a ‘seed’ to trigger subsequent hypermethylation in cancer cells. We used HpaII methylase to ‘pre-seed’ the GSTP1 shuttle (GSTP1 expression active) and the promoter deletion construct (GSTP1 expression inactive) and co-transfected the methylated constructs with pC1-neo into LNCaP and DU145 cells. After G418 selection we pooled approximately 20 colonies and serially passaged the cells and analysed the transfected GSTP1 DNA for methylation. Before transfection the HpaII methylated plasmid DNA was sequenced to confirm that all the HpaII sites were indeed methylated (Figure 7). However after stable transfection and 22 doublings the GSTP1 shuttle DNA was found to be substantially demethylated in both LNCaP cells (Figure 8a) and DU145 cells (data not shown). Demethylation of the HpaII sites was even more pronounced after 33 doublings. As observed in the other GSTP1 constructs a low level of de novo methylation was observed at random sites in a few molecules. In contrast, the promoter deletion construct had become extensively methylated at other CpG sites within the CpG island in both LNCaP cells (Figure 8b) and DU145 cells (Figure 8c). After a further 10 passages (33 doublings), the methylation pattern in the promoter deletion construct showed evidence of spreading, whereas the HpaII methylated shuttle construct had become more demethylated (Figure 8a,b). Interestingly, the methylation did not appear to be necessarily spreading from the ‘seeded’ HpaII sites but was localized to specific regions between the HpaII sites. Distinct peaks of de novo methylation were observed in the promoter deleted construct between the methylated HpaII sites. This wave pattern is visualized in Figure 9 which depicts only those CpG sites that were de novo methylated. In particular, the region 3′ to the start of transcription (CpG 1–10) appeared to be preferentially methylated, similar to the methylation region of the GSTP1 in DU145 DNA (Figure 2) and in the liver GSTP1 DNA (Millar et al., 2000).

Figure 7

Methylation profile of HpaII methylated GSTP1 pre-transfection. The GSTP1 plasmid DNA constructs were methylated with HpaII methylase prior to transfection. A sample of the HpaII methylated plasmid was treated with bisulphite and the GSTP1 core promoter region amplified. The efficiency of methylation was tested by sequencing six clones derived for the PCR amplification. (A) A sample sequence profile from the HpaII methylated GSTP1 shuttle vector and (B) the HpaII methylated promoter deletion construct is shown. The sequence is in reverse orientation. The CpG sites are numbered relative to the start of transcription. The Dcm+ bacterial methylated EcoRII site is underlined. *Indicates the HpaII CpG sites

Figure 8

Methylation profile of HpaII methylated GSTP1 after transfection. GSTP1 shuttle and promoter deletion constructs were methylated using HpaII methylase enzyme. The methylated constructs were transfected into LNCAP cells (A,B) or DU145 cells (C) and DNA isolated after 22 and 33 doublings (22d and 33d). The methylation status of the transfected DNA was analysed by bisulphite sequencing a number of clones from 2–3 PCR reactions. *Indicates the HpaII methylated CpG sites prior to transfection

Figure 9

De Novo methylation profile. The degree of de novo methylation was scored from the HpaII methylated promoter deletion construct after 33 doublings (Figure 8). The methylation profile is drawn to scale relative to the position in the sequence across the GSTP1 CpG island, from CpG −28 to +10. Only sites subjected to de novo methylation in the LNCaP cells are shown. Methylation at the HpaII sites (*) is not shown. Position number in the GSTPI sequence (GenBank accession number M24485) is indicated below the CpG sites

To determine the effect of the methylation pattern on the transcription of the constructs we assayed GSTP1 expression levels from the methylated constructs. Expression from the GSTP1 shuttle after 22 doublings was reduced threefold by the HpaII methylation, whereas the HpaII methylated promoter deletion construct failed to express as the essential promoter sequences had been deleted (Figure 10). Therefore active expression of GSTP1 appears to promote demethylation of the CpG island whereas inactive expression appears to promote extensive de novo methylation.

Figure 10

Quantitative RT–PCR analysis of HpaII methylated GSTPI constructs. GSTPI expression levels in LNCaP cells from the transfected GSTPI shuttle, the HpaII methylated GSTPI shuttle and the HpaII promoter deletion GSTPI constructs were estimated using fluoresence quantitative RT–PCR. The scale represents the mean expression levels as a ratio between GSTPI and beta-actin expression


In this study we address the process of abnormal hypermethylation of CpG islands that commonly occurs in cancer cells. We chose to study the mechanism responsible for triggering methylation of the CpG island spanning the GSTP1 gene, as both copies of this gene become methylated in nearly all prostate cancer cells and yet the island remains essentially unmethylated in normal cells (Millar et al., 1999). To identify possible methylation determinants, we compared the potential for aberrant methylation of GSTP1 in two prostate cancer cell lines; DU145 cells where the methylation profile of the endogenous GSTP1 gene is essentially unmethylated and the gene is expressed, and LNCaP cells where the endogenous gene is extensively methylated and silent. Even though we found that there was no intrinsic difference between either cell line in their capacity to express or methylate an introduced GSTP1 construct, this system did allow us to evaluate a number of possible mechanisms that have been proposed to be responsible for triggering hypermethylation of CpG islands in cancer.

We found that: (1) There is no significant difference in the levels of the predominant DNA methyltransferase DNMT1 and 3a between LNCaP and DU145 cells. Interestingly expression of DNMT2 and 3b was elevated threefold in LNCaP versus DU145. However, the moderate increase in expression we found in LNCaP cells does not appear to modulate hypermethylation of the transfected constructs. Similarly, overexpression of DNMT3b by up to 7.5-fold has been reported in some bladder, colon and kidney tumours (Robertson et al., 1999) and a 3.6–4.0-fold increase for colorectal tumours (Eads et al., 1999), but this over expression also did not appear to correlate with CpG island hypermethylation. Therefore at this stage we conclude that DNMT2 and 3b elevated expression alone is not responsible for the endogenous GSTP1 methylation. However it should be noted that the exact biological role of the methyltransferase enzymes is unclear and therefore it is possible that the enzymes may display an altered target specificity in the cancer cell. (2) The GSTP1 CpG island sequence itself is not a target for abnormal methylation in the prostate cancer cell lines because the transfected sequence remained unmethylated. Similar findings were also found for the VHL gene that was not methylated when transfected into a renal colorectal cell line that methylated its endogenous VHL gene (Kuzmin et al., 1999); suggesting that it is cis-specific local features and not trans-acting factors that are pivotal in maintaining and perpetuating the hypermethylation of the VHL CpG island. (3) Sp1 binding or binding of other protein factors in the essential promoter region are not responsible for protecting the GSTP1 CpG island from de novo methylation, because both specific mutation or removal of the essential promoter region does not trigger hypermethylation of the transfected DNA. This is in contrast to the finding that the APRT CpG island becomes methylated in CHO cells after inactivation of Sp1 sites (Brandeis et al., 1994; Macleod et al., 1994). However our results may not be so surprising since in Sp1−/− embryos CpG islands remain unmethylated (Marin et al., 1997). (4) The ATAAA repeat methylation boundary sequence does not harbour CpG island methylation protective factors because the GSTP1 gene did not become methylated when the boundary sequence was removed. (5) Active transcription alone is not responsible for CpG island methylation protection, because GSTP1 constructs that did not transcribe due to mutation or deletion of the essential promoter region failed to become hypermethylated, even after multiple passaging.

Intriguingly, hypermethylation of the GSTP1 CpG island was only triggered after a two step event, namely ‘seeding’ methylation in combination with inactivation of transcription. When the CpG island was ‘seeded’ with a limited number of methylated sites and active transcription was present, rapid demethylation was observed. In contrast, ‘seeding’ methylation in combination with inactive transcription resulted in extensive de novo methylation. Therefore active transcription can play an important role in maintaining the island in an unmethylated state possibly by promoting active or passive demethylation. However if the gene is inactivated and contains enough methylation ‘seeds’, de novo methylation of the island can be triggered. It is interesting to note that de novo methylation of the transfected GSTP1 gene did not appear to be as a result of spreading from the ‘seeded’ CpG sites but showed preferred sites or regional preference, thus giving rise to distinct peaks of methylation. We and others have previously have observed such a wave pattern of de novo methylation in the p16 gene in human mammary epithelial cells that had escaped senescence (Huschtscha et al., 1998) (Wong et al., 1997). Methylation in this case was also associated with p16 inactivation and lifespan extension. It is possible that the wave pattern reflects the nucleosome positioning or chromatin structure of the DNA. The relationship of the de novo methylation profile and chromatin structure will require further investigation.

Interestingly, both prostate cancer cell lines have the capacity to de novo methylate and support GSTP1 expression even though expression was reduced in the LNCaP cells. So therefore what could lead to the methylation of the endogenous GSTP1 in LNCaP and most prostate tumours and not in DU145 cells? We propose that prior gene silencing is an integral part to aberrant methylation in cancer cells. Our data suggests that GSTP1 gene silencing is not necessarily due to a lack of specific transcription factors in the pre-cancer prostate cell because the LNCaP cells remain transcriptionally competent. However a reduced expression capacity in the cancer cell, as seen in LNCaP cells, may be sufficient to override the balance between demethylation and de novo methylation. We propose that a more likely possibility is that GSTP1 gene silencing could be transient in the pre-cancer call, that is GSTP1 fails to successfully undergo transcription in a low proportion of normal prostate cells. In fact it has already been demonstrated that in normal prostate tissues expression of GSTP1 is heterogenous and expression in prostatic intraepithelial neoplasia (PIN) is diminished (Lee et al., 1994). Therefore the GSTP1 gene at any one time can be silent in an individual normal prostate cell leaving it susceptible to de novo methylation. However if the gene is actively transcribed, as in most of the normal prostate cells and in DU145, seeding methylation may occur randomly, but extensive de novo methylation is prevented because of active transcription.

We propose a trigger model to explain aberrant hypermethylation in cancer cells (Figure 11). This model requires a combination of two events namely, inactivation of the gene and random ‘seeds’ of methylation within the CpG island to act as a trigger for de novo methylation of the surrounding CpG sites. According to our model de novo methylation is normally occurring in CpG island regions at a very low and random level in the differentiated cell, a level similar to the low level methylation we observe in the transfected DNA constructs. In fact we have previously noted a low level of random methylation in some individual molecules of various CpG island associated genes in a variety of normal cells (Melki et al., 1999, 2000). However we propose that CpG methylation does not accumulate in the CpG islands because active transcription promotes demethylation of these sites. The processes of de novo methylation, maintenance methylation and demethylation are competing and therefore the amount of sporadic methylation is in constant flux. Methylation at a few CpG sites within the GSTP1 CpG island promoter region does not interfere with transcription. In fact, expression of a gene appears to result in demethylation of any randomly methylated sites within the island and therefore a critical density of CpG methylation, essential to promote extensive de novo methylation, cannot be obtained. However if the gene is inactivated, either transiently or via an independent oncogenic process, the low level CpG methylation within the CpG island may serve to trigger or promote hypermethylation of the rest of the island. Once hypermethylated, the gene is locked into an inactivated chromatin state by the attraction of methylation binding proteins (MBP) and histone de-acetylation (Bird and Wolffe, 1999).

Figure 11

Seeding and silencing: trigger hypermethylation model in cancer

Materials and methods

Cells and culture conditions

The prostate cancer cell lines used were LNCaP and DU145 (Lee et al., 1994). LNCaP cells were cultured in T-medium (Thalmann et al., 1996) with 10% heat inactivated foetal calf serum. DU145 cells were cultured in RPMI 1640 medium (Gibco BRL) supplemented with 10% heat inactivated foetal calf serum. The LNCaP cells and DU145 cells were split 1 : 3 and 1 : 6 respectively every 4–5 days.


The vector constructs (5–10 μg) were co-transfected with pC1-neo plasmid DNA (Promega) (10 : 1 ratio) using the cationic liposome reagent DOTAP and DOSPER (Boehringer Mannheim) according to the manufacturer's instructions. The day before transfection the cells were subcultured to a T75 cm2 flask to yield approximately 50% confluency at the time of transfection. LNCaP cells were incubated with the DNA/DOTAP mixture for 16 h and DU145 cells with a 1 : 1 mixture of DOTAP/DOSPER and DNA for 6 h. One day after transfection the cells were split 1 in 3 and cultured with 900 μg/ml GeneticinR (G418) (Gibco BRL) for LNCaP cells and 400 μg/ml G418 for DU145 cells. After 15–25 days the colonies (10–20) were pooled and grown to confluence. The cells were harvested for genomic DNA and RNA isolation. Copy number was estimated by Southern blot hybridization and found to be less than 1 copy per genome.

DNA extraction and genomic bisulphite sequencing

DNA was extracted using the Puregene extraction kit (Gentra Systems). To enrich for the transfected GSTP1 gene, we digested the genomic DNA prior to bisulphite treatment with either Tth111I that digests the endogenous GSTP1 gene but not the ATAAA gene construct or MaeIII that digests the endogenous gene but not the SPI mutation or promoter deletion construct. The bisulphite reaction was carried out on 2 μg of restricted DNA for 16 h at 55°C under conditions as previously described (Clark and Frommer, 1995; Clark et al., 1994). After neutralization, the bisulphite treated DNA was ethanol precipitated, dried, resuspended in 50 μl of H2O and stored at −20°C. Approximately, 2 μl of DNA was used for each of the nested PCR amplifications in a 50 μl final volume, 200 μM dNTPs, 300 ng primers, 1–2 mM MgCl2, 1.25 units AmpliTaq DNA polymerase (Perkin Elmer), and reaction buffer supplied with the enzyme. The GSTP1 primers used for the amplification of bisulphite-treated DNA from CpG −56 to –30 were for first round GST1 and GST4, and for second round GST 25 and GST3, as described in Millar et al. (2000). The GSTP1 primers used for the amplification of bisulphite-treated DNA from CpG −28 to +10 were for first round GST9 and GST10, and for second round GST 11 and GST12, as described in Millar et al. (1999). The primers used for PCR amplification from CpG sites −41 to +10 were for first round; GST55 (688–713) 5′-TTAGTTTGGGTTATAGCGTGAGATTA and GST10 and for second round; GST56 (859–882) 5′-TTTGTTTTGTGAAGGGGGTGTGTAAGTTT and GST12. GenBank Accession number M24485.

Cloning and sequence analysis

At least three PCR reactions were independently performed for each set of primers and target DNA to ensure a representative methylation profile. PCR fragments were gel-purified or directly purified using the Wizard® PCR preps DNA purification system and cloned into the pGEM®-T Easy Vector (Promega) using the Rapid Ligation Buffer system (Promega). Approximately 12 individual clones were sequenced from the pooled PCR reactions using the Dye terminator cycle sequencing kit with AmpliTaq® DNA polymerase, FS (PE Biosystems) and the automated 373A DNA Sequencer (Applied Biosystems).

Plasmid constructions

GSTP1 shuttle vector 

(Figure 1a) contains the GSTP1 gene sequence from XbaI (16) to SnaBI (2306) adjacent to a poly A signal. It was constructed in a number of steps. A 3.1 kb XbaI fragment, isolated from cosGSTrP6 cosmid (Cowell et al., 1988), containing the entire GSTP1 CpG island extending from 1.2 kb upstream of the transcription start site to intron 5 was cloned into pBluescriptSK+. The bovine growth hormone 3′ polyadenylation signal was cloned as a KpnI/EcoRI fragment, isolated from pCI-GFP, into the plasmid KpnI/EcoRI sites 3′ to the GSTP1 fragment, to provide for stability and detection of the mRNA produced from the introduced gene. For subsequent cloning steps it was necessary to remove the NarI sites at base positions 1061 and 2506. The NarI site (1061) was removed by subcloning the 1.1 kb NotI fragment from the GSTP1 shuttle vector into pBluescriptSK+, linearising with NarI, end-filling and blunt-end ligating. The 1.1 kb NotI fragment was then recloned into the GSTP1 shuttle vector. The modification of the NarI site resulted in an extra CpG dinucleotide at CpG −23, that was used as an informative marker to distinguish the endogenous GSTP1 gene from transfected GSTP1. The NarI site (2506) was removed by deletion of the SnaBI/SpeI fragment spanning the SnaBI site (2306) in the GSTP1 sequence and SpeI in the plasmid multiple cloning.

Promoter deletion plasmid

(Figure 1b) was constructed by deletion of the NarI(1150)/StuI(1216) fragment encompassing the minimal promoter region (Moffat et al., 1996a,b), followed by end-filling and blunt-end religation of the GSTP1 shuttle vector.

Sp1 mutation vector

(Figure 1b) was constructed by deletion of the NarI(1150)/StuI(1216) fragment from the GSTP1 shuttle vector and ligation of annealed oligos with NarI/StuI ends incorporating nucleotide changes such that the two Sp1 sites and MaeIII site (1555) were mutated but otherwise containing an identical sequence to the GSTP1 gene. The two oligos were: 5′ CGCCGTGATTCAGCACTGG A G TA GAGCGG A G TA GGACGACCCTTATAAGGCTCGGAGGCCGCGAGG and 5′ CCTCGCGGGGTCCGAGCCTTATAAGGGTCGTCC TA C T CCGCTC TA C T CCAGTGCTGAATCACGGCG with the italics indicating the mutated bases in the Sp1 sites and bold indicating the base change to mutate the MaeIII site. The Sp1 sites are underlined.

ATAAA deletion vector

(Figure 1b) was constructed by deletion of the Tth111I (709)/BssHII (927) fragment from the GSTP1 shuttle vector and ligation of an annealed and extended PCR fragment. TS1 primer 5′ ACGAGGCAATTTCCTTTCCTCTAAGCGGCCTCCACCCCTCTCCCCTGCCCTGT and BS1 primer 5′ CTTCGCTCCCGGAGCTTGCACACCCGCTTCACAGGGCAGGGGAGAGGGGTGGA was annealed and further extended by PCR amplification using TS1 primer and BS2 primer 5′-CCGGGACATCGCGGGGGGAAATTCCCTAAGACCGCTTCGCTCCCGGAGCTTGCAC. The extended PCR fragment was end-filled and ligated into the Tth111I/BssHII cut GSTPI shuttle vector. The extended PCR fragment spanned the GSTP1 sequence from 709 to 927 but all the ATAAAA repeats were deleted leaving the surrounding sequences intact, except for the Tth111I site that was mutated after blunted ligation.

HpaII methylation

GSTP1 shuttle and promoter deletion plasmid DNA was methylated using HpaII methylase (New England Biolabs) according to the manufacturer's protocol. Complete methylation was confirmed by resistance to restriction by HpaII enzyme and by bisulphite sequencing.

RNA extraction and quantitative real-time RT–PCR

RNA was extracted using Trizol reagent (Life Technologies, Inc.) according to the manufacturer's protocol. cDNA was reverse transcribed from 1–2 μg of total RNA in a 20 μl reaction using SuperScriptTM II RNase H Reverse Transcriptase (Life Technologies) as per the manufacturer's instructions. The reaction was primed with 200 ng of random hexamers (Boehringer). The reverse transcription reaction volume was adjusted to 100 μl with sterile H2O before addition to the RT–PCR.

The quantitation of mRNA was performed using a fluorogenic real time detection method using the ABI Prism 7700 Sequence Detection System. Beta-actin was also amplified to control for RNA amount and integrity. All reactions were amplified using a 96-well optical plate and caps with a 25 μl final volume. GSTP1 RT–PCR products were amplified in TaqMan® Universal PCR Master Mix (P/N 4304437) and 300 mM each primer, 200 mM probe and 5 μl of diluted cDNA. Beta-actin was amplified using a Human Beta-actin pre-developed Taqman® Assay Reagents Endogenous Control (P/N 4310881E) (Applied Biosystems, Foster City, CA, USA) in TaqMan® Universal PCR Master Mix (P/N 4304437) with 5 μl of diluted cDNA. Cycling conditions were 50°C for 2 min, 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and at 60°C for 1 min. The primers and probes used for the quantitative RT–PCR assays are summarised in Table 1. All reactions were performed in triplicate and the standard deviation was calculated using the Comparative method (ABI PRISM 7700 Sequence Detection system User Bulletin #2, 1997 P/N 4303859). The cycle number that the measured fluorescence crosses a threshold is directly proportional to the amount of starting material. The use of beta-actin as a reference gene yields the relative gene expression of each sample. The mean expression levels are represented as the ratio between GSTP1 and beta-actin expression. The amplification efficiency of all the DNA methyltranferase genes was compared to that of the Beta-actin pre-developed Taqman® endogenous control by measuring ΔCT in a template dilution series as outlined in ABI PRISM 7700 Sequence Detection system User Bulletin #2, 1997 (P/N 4303859). When regression analysis was performed on the plot of log input amount verses ΔCT all amplifications had a slope of <0.14 and therefore amplified with efficiencies similar to the Beta-actin control. It is therefore possible to make direct comparisons between the expression levels by factoring in the specific reaction efficiency for each gene.

Table 1 PCR primers and probes used for quantitative RT-PCR


  1. Baylin SB, Esteller M, Rountree MR, Bachman KE, Schuebel K, Herman JG . 2001 Hum. Mol. Genet. 10: 687–692

  2. Bird AP, Wolffe AP . 1999 Cell 99: 451–454

  3. Brandeis M, Frank D, Keshet I, Siegfried Z, Mendelsohn M, Nemes A, Temper V, Razin A, Cedar H . 1994 Nature 371: 435–438

  4. Brooks JD, Weinstein M, Lin XH, Sun YH, Pin SS, Bova GS, Epstein JI, Isaacs WB, Nelson WG . 1998 Cancer Epidemiol. Biomarkers Prev. 7: 531–536

  5. Clark SJ, Frommer M . 1995 DNA and Nucleoprotein Structure, In Vivo Saluzand HP and Wiebauer K (eds) Springer-Verlag, Austin, Texas: RG Landes Company pp 123–132

  6. Clark SJ, Harrison J, Paul CL, Frommer M . 1994 Nucleic Acids Res. 22: 2990–2997

  7. Cookson MS, Reuter VE, Linkov I, Fair WR . 1997 J. Urology 157: 673–676

  8. Costello JF, Plass C . 2001 J. Med. Genet. 38: 285–303

  9. Cowell IG, Dixon KH, Pemble SE, Ketterer B, Taylor JB . 1988 Biochem J. 255: 79–83

  10. Eads CA, Danenberg KD, Kawakami K, Saltz LB, Danenberg PV, Laird PW . 1999 Cancer Res. 59: 5860–

  11. Esteller M . 2000 Eur. J. Cancer 36: 2294–2300

  12. Esteller M, Corn PG, Urena JM, Gabrielson E, Baylin SB, Herman JG . 1998 Cancer Res. 58: 4515–4518

  13. Hsieh CL . 1999 Mol. Cell Biol. 19: 8211–8218

  14. Huschtscha LI, Noble JR, Neumann AA, Moy EL, Barry P, Melki JR, Clark SJ, Reddel RR . 1998 Cancer Res. 58: 3508–3512

  15. Jhaveri MS, Morrow CS . 1998 Gene 210: 1–7

  16. Jones PA, Laird PW . 1999 Nat. Genet. 21: 163–167

  17. Jost J-P, Frémont M, Siegmann M, Hofsteenge J . 1997 Nucleic Acids Res. 25: 4545–4550

  18. Kass SU, Pruss D, Wolffe AP . 1997 Trends Genet. 13: 444–449

  19. Kuzmin I, Geil L, Ge HY, Bengtsson U, Duh FM, Stanbridge EJ, Lerman MI . 1999 Oncogene 18: 5672–5679

  20. Lee W-H, Morton RA, Epstein JI, Brooks JD, Campbell PA, Bova GS, Hsieh W-S, Isaacs WB, Nelson WG . 1994 Proc. Natl. Acad. Sci. USA 91: 11733–11737

  21. Li E, Bestor TH, Jaenisch R . 1992 Cell 69: 915–926

  22. Macleod D, Charlton J, Mullins J, Bird AP . 1994 Genes Dev. 8: 2282–2292

  23. Marin M, Karis A, Visser P, Grosveld F, Philipsen S . 1997 Cell 89: 619–628

  24. Melki JM, Vincent PC, Clark SJ . 1999 Cancer Res. 59: 3730–3740

  25. Melki JR, Vincent PC, Brown RD, Clark SJ . 2000 Blood 95: 3208–3213

  26. Millar DS, Ow KK, Paul CL, Russell PJ, Molloy PL, Clark SJ . 1999 Oncogene 18: 1313–1324

  27. Millar DS, Paul CL, Molloy PL, Clark SJ . 2000 J. Biol. Chem. 275: 24893–24899

  28. Moffat GJ, McLaren AW, Wolf CR . 1996a J. Biol. Chem. 271: 20740–20747

  29. Moffat GJ, McLaren AW, Wolf CR . 1996b J. Biol. Chem. 271: 1054–1060

  30. Moskaluk CA, Duray PH, Cowan KH, Linehan M, Merino MJ . 1997 Cancer 79: 1595–1599

  31. Okano M, Bell DW, Haber DA, Li E . 1999 Cell 99: 247–257

  32. Rhee I, Jair KW, Yen RW, Lengauer C, Herman JG, Kinzler KW, Vogelstein B, Baylin SB, Schuebel KE . 2000 Nature 404: 1003–1007

  33. Robertson KD, Uzvolgyi E, Liang G, Talmadge C, Sumegi J, Gonzales FA, Jones PA . 1999 Nucleic Acids Res. 27: 2291–2298

  34. Tchou JC, Lin XH, Freije D, Isaacs WB, Brooks JD, Rashid A, DeMarzo AM, Kanai Y, Hirohashi S, Nelson WG . 2000 Int. J. Oncol. 16: 663–676

  35. Thalmann GN, Sikes RA, Chang SM, Johnston DA, von Eschenbach AC, Chung LW . 1996 J. Natl. Cancer Inst. 88: 794–801

  36. Turker MS, Bestor TH . 1997 Mutat Res. Rev. 386: 119–130

  37. Vertino PM, Yen RW, Gao J, Baylin SB . 1996 Mol. Cell. Biol. 16: 4555–4565

  38. Warnecke P, Clark S . 1998 PhD thesis DNA Methylation in Early Development University of Sydney, Sydney

  39. Warnecke PM, Clark SJ . 1999 Mol. Cell. Biol. 19: 164–172

  40. Wong DJ, Barrett MT, Stoger R, Emond MJ, Reid BJ . 1997 Cancer Res. 57: 2619–2622

Download references


We are grateful to Dr P Molloy from CSIRO Molecular Science, Sydney for helpful discussions and for the gift of pCI-GFP. We thank Dr P Board, from JCSMR Canberra, for the gift of Cosmid cosGSTpP6. J Song and C Stirzaker contributed equally to the manuscript and are supported by NH&MRC grant (9937987) and JR Melki is funded by an Anthony Rothe Memorial Trust grant.

Author information

Correspondence to Susan J Clark.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Song, J., Stirzaker, C., Harrison, J. et al. Hypermethylation trigger of the glutathione-S-transferase gene (GSTP1) in prostate cancer cells. Oncogene 21, 1048–1061 (2002) doi:10.1038/sj.onc.1205153

Download citation


  • DNA methylation
  • CpG islands
  • bisulphite genomic sequencing
  • GSTP1
  • prostate cancer

Further reading