The prolyl 3-hydroxylases P3H2 and P3H3 are novel targets for epigenetic silencing in breast cancer

Expression of P3H2 (Leprel1) and P3H3 (Leprel2) but not P3H1 (Leprecan) is down-regulated in breast cancer by aberrant CpG methylation in the 5′ regulatory sequences of each gene. Methylation of P3H2 appears specific to breast cancer as no methylation was detected in a range of cell lines from other epithelial cancers or from primary brain tumours or malignant melanoma. Methylation in P3H2, but not P3H3, was strongly associated with oestrogen-receptor-positive breast cancers, whereas methylation in P3H3 was associated with higher tumour grade and Nottingham Prognostic Index. Ectopic expression of P3H2 and P3H3 in cell lines with silencing of the endogenous gene results in suppression of colony growth. This is the first demonstration of epigenetic inactivation of prolyl hydroxylases in human cancer, implying that this gene family represents a novel class of tumour suppressors. The restriction of silencing in P3H2 to breast carcinomas, and its association with oestrogen-receptor-positive cases, suggests that P3H2 may be a breast-cancer-specific tumour suppressor.

Epigenetics describes heritable changes in gene expression that occur in the absence of changes in DNA sequence (Herman and Baylin, 2003). The best characterised epigenetic alteration in cancer is hypermethylation of CpG rich regions, usually found in the promoter region of a gene (Jones and Baylin, 2002). Hypermethylation, along with other epigenetic events often associated with gene silencing, is crucial in the development of cancer (Baylin, 2005). The detection of methylation-associated gene inactivation is today widely used to identify candidate tumour suppressor genes (Baylin and Ohm, 2006).
The prolyl 4-hydroxylases (P4H) have been extensively studied and are known to reside in either the endoplasmic reticulum (ER) or cytoplasm, where their function is to hydroxylate proline residues in the X-Pro-Gly sequence in collagens (Kivirikko et al, 1989) or to hydroxylate the 564 proline residue in the a-subunit of the hypoxia-inducible factor (HIF; Bruick and McKnight, 2001;Epstein et al, 2001;Ivan et al, 2001). The c-P4H enzymes have a key function in the biosynthesis of collagen allowing appropriate folding of the procollagen chains to form a triple helical structure (Myllyharju, 2003). Furthermore, a decrease in oxygen tension has been found to result in an up-regulation of P4H genes, P4HA1 and P4HA2, as they have been found to be transcriptionally activated by HIF. Although the prolyl hydroxylase reaction does require O 2 , it is thought that the over-production of P4HA1 and P4HA2 in hypoxic conditions compensates for this (Hofbauer et al, 2003;Fähling et al, 2006).
In comparison to the P4H proteins, the function of the P3H proteins is less well defined. However, it is known that P3Hmodified residues are more abundant in basement membrane collagens. Prolyl 3-hydroxylation typically occurs in the Gly-3Hyp-4Hyp sequence (Gryder et al, 1975;Vranka et al, 2004;Myllyharju, 2005). P3H1 belongs to a family comprising two further genes, all three proteins sharing conserved catalytic residues of the 2-oxoglutarate and iron-dependent dioxygenases with the c-P4Hs and LHs (Vranka et al, 2004). P3H2 was initially identified as a protein mainly localised to the endoplasmic reticulum and Golgi (Jarnum et al, 2004), but more recently has been demonstrated in tissues rich in basement membranes, and participates in the hydroxylation of collagen IV (Tiainen et al, 2008). It has previously been hypothesised that prolyl 3-hydroxylation occurs after prolyl 4-hydroxylation, thus once the triple helix is formed, the 3-hydroxyproline results in destabilisation (Jenkins et al, 2003;Mizuno et al, 2004). There are no published reports on the function of P3H3.
This study examined the epigenetic regulation of P3H1, P3H2 and P3H3 expression in breast cancer cell lines and in a panel of breast carcinomas. We show that loss of P3H2 and P3H3 expression results from epigenetic silencing, and is associated with aberrant hypermethylation in the CpG islands around exon 1 of both P3H2 and P3H3.

Cell culture
The following breast carcinoma cell lines were used in this study and routinely maintained in Dulbecco's modified Eagle's media (Invitrogen, Paisley, UK), supplemented with L-glutamine (5 mM) and 10% heat-inactivated fetal bovine serum (Invitrogen) in 5%
For western blotting, cells were lysed with RIPA lysis buffer. Protein lysate (40 mg) was resolved on 8% SDS -PAGE gel and proteins were transferred onto nitrocellulose membrane that were incubated for 1 h with primary antibodies.
Rabbit antibody against P3H2 was described previously (Jarnum et al, 2004) and was used at a dilution of 1 : 1000. Polyclonal rabbit antibodies against P3H3 were raised against the peptide CHQRVQDKTGRAPRVREEL (Biogenes, Berlin, Germany) and used at a dilution of 1 : 1000. The secondary antibody was affinity-purified HRP-conjugated goat anti-rabbit and used at a dilution of 1 : 2000 (Dako, Cambridgeshire, UK). Anti-PCNA (1 : 10 000) was used as a loading control.

Bisulphite modification
Genomic DNA (gDNA) was extracted from cell pellets using the DNeasy kit (Qiagen Ltd.). Genomic DNA (500 ng) was used for bisulphite modification with the Zymo EZ DNA Methylation kit (Genetix, Hampshire, UK), and eluted in 200 ml dH 2 O. Included in each bisulphite modification were unmethylated human DNA, and CpGenome Universal Methylated DNA (Chemicon International, Temecula, CA, USA), which were used as negative and positive controls, respectively.
Methylation-specific PCR was performed in a 20 ml volume using Thermo-Start PCR Master Mix (Abgene). The standard thermal cycling conditions were an initial 'hotstart' of 8 cycles followed by a further 30 cycles, with a final extension. PCR products were resolved on a 2% agarose gel stained with ethidium bromide (Promega), and visualised under UV.

Clinical tissue
Genomic DNA was extracted from 184 primary, previously untreated breast cancers, using the M48 Qiagen DNA extraction robot, following Tayside Tissue Bank Local Research Ethics Committee approval. Cancers were subject to histopathological review before use for DNA extraction to ensure adequate representation of neoplastic cells. Expression of the oestrogen receptor, progesterone receptor and HER2 (using antibody CB11, supplemented by FISH for 2-positive cancers to confirm amplification), was measured as part of routine clinical care. Clinical and pathological data included tumour grade, tumour type, pathology node status, and relapse-free and overall survival.

Colony formation assay
Plasmids for ectopic expression of P3H2 and P3H3 were as follows: pEGFP-N1-P3H2 was as previously described (Jarnum et al, 2004). A full-length P3H3 cDNA in pCMV6-XL6 vector was purchased from Origene (Rockville, MD, USA) and the insert was subcloned into pcDNA3.1 as a Not1 fragment. Correct orientation was determined by sequencing multiple plasmid clones. To assess the effect on cell proliferation of ectopic expression of P3H2 and P3H3, cell lines were transfected with 4 mg of the above expression clones or empty vector alone, using Lipofectamine 2000 (Invitrogen). Media were changed 24 h after transfection, and transfected cells were selected in G418 (800 mg ml À1 ). After 16-day growth, surviving colonies were fixed in 4% paraformaldehyde, washed with phosphate-buffered saline, and dH 2 O, dried, then stained with liquid crystal violet (Sigma-Aldrich, Dorset, UK) and counted. Experiments were carried out in triplicate.

Statistical analyses
Assessment between two categorical variables was carried out using w 2 -or Fisher's exact test. Analysis of the cumulative survival was carried out by the Kaplan -Meier method and differences between the groups were tested with the log-rank test. All reported P-values were two sided and considered statistically significant if Po0.05. Tests were performed using GraphPad Prism version 5.0 software (GraphPad Software Inc, San Diego, CA, USA).

RESULTS
Transcriptional down-regulation of P3H2 and P3H3 in breast cancer cell lines Using RT -PCR, we analysed the expression of the P3H1, P3H2 and P3H3 genes in breast carcinoma cell lines ( Figure 1A). All three genes were expressed in HMEC. P3H1 was expressed in all 13 carcinoma cell lines in our panel, but there was no detectable expression of P3H2 mRNA in MDA MB 361, MDA MB 453, MCF7 and T47D cell lines, with only low levels of expression in BT474 and SKBR3. In the case of P3H3, expression was undetectable in the MDA MB 231, MDA MB 361, MDA MB 468, MCF7, BT474 and SKBR3 cell lines ( Figure 1A). Next, we analysed protein levels of P3H2 and P3H3. We used a previously described antibody to P3H2 and generated a new polyclonal antibody to P3H3 and performed western analysis of the breast carcinoma cell line panel. In general, protein levels for both P3H2 and P3H3 paralleled mRNA expression ( Figure 1A and 1B). Interestingly, however, P3H3 protein was barely detectable in T47D cells despite readily detectable expression of P3H3 mRNA, and the level of P3H3 protein was also reduced in MDA MB 453 relative to MDA MB 436, GI101 and Cal51 despite comparable expression of P3H3 mRNA. This may reflect other regulatory mechanisms operating at the level of mRNA translation or protein stability.

Aberrant methylation of P3H2 and P3H3 in breast cancer cell lines
We identified CpG islands in the 5 0 sequences of P3H1, P3H2 and P3H3 genes (http://genome.ucsc.edu). To address whether promoter methylation was the cause of loss of gene expression, we performed MSP analysis of the CpG islands of P3H1, P3H2 and P3H3 in each of the breast cancer cell lines. The CpG island of P3H1 was uniformly unmethylated in each cell line consistent with expression analysis ( Figure 1C). In the case of P3H2 methylation, initial analysis used primers located in the centre of the CpG island and these detected methylation in the MDA MB 453 and T47D cell lines, both of which lacked detectable expression of P3H2. However, analysis with this primer set did not detect methylation in some cell lines that lack expression of P3H2. We therefore designed a second primer set located further 3 0 in the CpG island. To characterise methylation in greater detail across the P3H2 and P3H3 CpG islands, we mapped each island, using bisulphite sequencing, in a panel of cell lines previously analysed by MSP (Figures 2 and 3). These studies closely paralleled the MSP analysis. For example, in the P3H2 CpG island, the region of the CpG island sampled by MSP primer pair 1 was methylated only in the MDA MB 453 and T47D cell lines. In contrast, the region of the CpG island sampled by MSP primer pair 2 contained methylation in all cell lines lacking expression of P3H2. Consistent with methylation-dependent transcriptional silencing, some of the cell lines expressing P3H2 mRNA show an extremely low frequency of CpG methylation (Figure 2). In the case of P3H3, methylation was observed across the entire CpG island consistent with MSP analysis ( Figure 3). As with P3H2, there was an extremely low level of methylation in some of the cell lines that express P3H3 mRNA ( Figure 3). The entire CpG islands of both P3H2 and P3H3 were unmethylated in normal mammary epithelium.

Methylation of P3H2 is specific for breast carcinomas
The observation of methylation-dependent transcriptional silencing in P3H2 and P3H3 prompted us to examine expression and methylation of the P3H genes in cell lines from other common solid tumour types. In ovarian, head and neck, vulval, melanoma, glioblastoma and renal carcinoma cell lines, P3H3 was clearly methylated and, as in breast cancer, this correlated with down-regulation of the mRNA ( Figure 1D; data not shown). In contrast, we found no evidence for methylation of P3H2 in analysis of cell lines from multiple other tumour types, including ovarian and renal adenocarcinomas, squamous carcinomas of the vulva and head and neck, malignant melanoma and glioma ( Figure 1D; data not shown). These results imply that whereas P3H3 is widely methylated in human cancers, methylation in P3H2 is restricted to breast cancer, at least within the tumour types we have analysed in the present study.
P3H2 and P3H3 genes are methylated in primary breast carcinomas Next we tested whether the CpG islands of P3H2 and P3H3 are methylated in a series of 184 primary breast carcinomas. From studies in cell lines described above, MSP analysis with primer set 2 of P3H2 was most strongly associated with down-regulated expression of the mRNA. To further confirm the utility of this primer set for methylation detection, we performed preliminary bisulphite sequencing on eight cancers designated as either methylated or unmethylated by MSP. These initial studies fully confirmed that primer pair 2 accurately assessed methylation status and so this primer pair was used for all subsequent analyses    of the full set of cases ( Figure 4). The frequency of methylation for P3H2 and P3H3 found in the 184 analysed breast samples was 42 and 26% respectively. Several associations were observed between methylation status and clinicopathological parameters (Table 1). First, methylation in the P3H2 CpG island was positively associated with a positive oestrogen receptor status (P ¼ 0.0053), whereas methylation in the P3H3 CpG island showed no such association (P ¼ 0.71) ( Table 1). The observation prompted us to determine whether there was a similar association in breast cancer cell lines. Other than MDA MB 453, all cell lines methylated in the P3H2 CpG island (MDA MB 361, MDA MB 453, MCF7, T47D, BT474 and SKBR3 express the oestrogen receptor). Methylation of the P3H3 CpG island was positively associated with increasing tumour grade (P ¼ 0.02) and with higher Nottingham Prognostic Index (P ¼ 0.02). However, we did not find evidence that methylation in either gene was associated with clinical outcome (Table 1).

Ectopic expression of P3H2 and P3H3 suppresses colony-forming ability
Genes found to be hypermethylated in their promoter region are often considered to be candidate tumour suppressor genes. One property that such genes may possess is the ability to suppress proliferation when ectopically expressed in cells lacking endogenous expression. To determine whether this was the case for P3H2 and P3H3, an expression plasmid for P3H2 was introduced into MCF7 and T47D cell lines and an expression plasmid for P3H3 was introduced into MCF7 and MDA MB 231, and we assessed the efficiency of colony formation after 16 days in G418 selection. In each case, expression of the transfected cDNA efficiently suppressed colony growth ( Figure 5). Using RT -PCR and western blotting we confirmed that the transfected sequences were expressed ( Figure 5). Ability to suppress proliferation demonstrated in these assays is consistent with a potential tumour suppressor function for P3H2 and P3H3.

DISCUSSION
The 2-oxoglutarate dioxygenases are a family of proteins required for modifications of collagen that are essential for its synthesis, folding and assembly. The collagen P3H are members of the 2-oxoglutarate dioxygenase family, which catalyse the posttranslational formation of 3-hydroxyproline in Gly-3Hyp-4Hyp  sequences in collagens, especially type IV and V collagens. The possible involvement of the P3H genes in human tumourigenesis was explored because ectopic expression of P3H1 was reported to cause growth arrest in fibroblasts (Kaul et al, 2000). There are three P3H proteins encoded in the human genome, P3H1, P3H2 and P3H3. Here, we show that both P3H2 and P3H3, but not P3H1, are frequent targets for epigenetic inactivation in human breast cancer. To the best of our knowledge, this is the first to report of an epigenetic inactivation of any prolyl hydroxylase gene in human neoplasia.
Breast carcinoma cell lines were screened by RT -PCR and western blotting for expression of the three P3H genes. Strikingly, whereas P3H1 was present in all cell lines, expression of both P3H2 and P3H3 was undetectable in several lines at both mRNA and protein levels. Bisulphite sequencing and MSP of the CpG islands located in the 5 0 sequences of each gene revealed a clear correlation between down-regulated expression and aberrant methylation for both genes. This implies that methylation-dependent transcriptional silencing is the mechanistic basis for the loss of mRNA expression, as is the case for a number of tumour suppressor genes in breast cancer, including p16 INK4a , Rassf1a and E-cadherin among others. Previous expression profiling studies of human breast cancer show that P3H2 mRNA (Radvanyi et al, 2005;Richardson et al, 2006) andP3H3 mRNA (van't Veer et al, 2002) are down-regulated in breast cancer, consistent with our results.
Taken together, our results suggest that P3H2 and P3H3 are candidate tumour suppressors in breast cancer, raising the question of which function(s) are selected against during tumourigenesis. The collagen prolyl hydroxylases are localised to the endoplasmic reticulum and their activity is required for proper collagen synthesis and assembly. Studies of inherited disorders of collagen biosynthesis suggest that loss of function in P3H proteins results in dysfunctional collagen; mutations in P3H1 are associated with oesteogenesis imperfecta type VIII (Cabral et al, 2007) and loss of function mutations in both P3H1-and P3H-related protein CRTAP have been described in oesteogenesis imperfecta types II and III (Baldridge et al, 2008). Evidence implicating collagen abnormalities in human tumours is afforded by studies showing methylation-dependent silencing of collagen-encoding genes in various tumour types (Sengupta et al, 2003;Ikeda et al, 2006). Type IV collagen, a major substrate for the P3H proteins, is an important component of the basement membrane, and impaired expression of type IV collagen has been reported to be an early event in acquisition of an invasive phenotype in some epithelial cancers (Ikeda et al, 2006). In this respect, it will clearly be of interest to determine whether loss of expression of P3H2 and/or P3H3 affects the properties of the basement membrane in breast cancer cells and thereby influences cancer-associated phenotypes such as invasiveness and metastasis. In addition to the possible effects of loss of P3H2 and P3H3 on collagen, our data demonstrate that both P3H2 and P3H3 have a direct anti-proliferative effect in breast cancer, suggesting additional tumour suppressor properties for each gene. Specifically, ectopic expression of both P3H2 and P3H3 in cells lacking endogenous expression due to epigenetic silencing resulted in a decrease in colony formation. Inhibition of colony formation in such assays has been demonstrated previously with known tumour suppressors such as p53 (Crook et al, 1994). Clearly, understanding the mechanism(s) by which the P3H genes negatively regulate proliferation will require additional studies.
A striking feature of the data is the restriction of methylation in P3H2 to breast cancer, with no detectable methylation in carcinoma cell lines in the other tumour types examined. Such tight specificity of methylation in one gene for a single tumour type is unusual and raises the potential to use detection of methylated DNA either in tissue or body fluids as a cancer biomarker. Verification that P3H2 is methylated only in breast cancer would make it an attractive candidate gene with potential utility in diagnosis and screening in breast cancer. The specificity of down-regulation of P3H2 mRNA in breast cancer may reflect the association with oestrogen-receptor-positive primary breast cancers, an association also noted in breast carcinoma cell lines. Selective methylation of the P3H2 CpG island in oestrogenreceptor-positive breast cancers is consistent with multiple arraybased expression profiling studies (Miller et al, 2005;Minn et al, 2005;Wang et al, 2005;Hess et al, 2006) and in array analysis comparing oestrogen-receptor-positive and -negative breast cancer cell lines (Neve et al, 2006). However, it remains to be determined whether P3H2 is an oestrogen-inducible gene and what the mechanistic basis is for the selective methylation of P3H2 in oestrogen-receptor-positive cases. One possibility is that P3H2 is an oestrogen-inducible negative regulator of proliferation. This hypothesis is supported by the demonstration that ectopic expression of P3H2 in cell lines lacking endogenous expression suppresses colony survival and growth. A second interesting association was that methylation in P3H3 was associated with higher histopathological grade, consistent with a number of expression profiling studies Ivshina et al, 2004;Zhao et al, 2004;Farmer et al, 2005;Miller et al, 2005;Ginestier et al, 2006;Hess et al, 2006) and with higher Nottingham Prognostic Index. From the relatively small number of cases of primary breast cancer analysed (n ¼ 184), we did not observe a significant association between P3H2 or P3H3 methylation and clinical outcome. However, mRNA analysis implies that downregulation of P3H2 is associated with less favourable prognosis in some breast cancer series Pawitan et al, 2006;Desmedt et al, 2007) and recurrence after tamoxifen (Ma et al, 2004). It will clearly be of interest to determine in large study populations whether analysis of P3H2 methylation has prognostic utility in breast cancer. This is the first demonstration of epigenetic inactivation of prolyl hydroxylases in human cancer. The prolyl 3-hydroxylases P3H2 and P3H3 are, therefore, novel candidate tumour suppressor genes in breast cancer.