Functional epigenomics approach to identify methylated candidate tumour suppressor genes in renal cell carcinoma

Promoter region hypermethylation and transcriptional silencing is a frequent cause of tumour suppressor gene (TSG) inactivation in many human cancers. Previously, to identify candidate epigenetically inactivated TSGs in renal cell carcinoma (RCC), we monitored changes in gene expression in four RCC cell lines after treatment with the demethylating agent 5-azacytidine. This enabled us to identify HAI-2/SPINT2 as a novel epigenetically inactivated candidate RCC TSG. To identify further candidate TSGs, we undertook bioinformatic and molecular genetic evaluation of a further 60 genes differentially expressed after demethylation. In addition to HAI-2/SPINT2, four genes (PLAU, CDH1, IGFB3 and MT1G) had previously been shown to undergo promoter methylation in RCC. After bioinformatic prioritisation, expression and/or methylation analysis of RCC cell lines±primary tumours was performed for 34 genes. KRT19 and CXCL16 were methylated in RCC cell lines and primary RCC; however, 22 genes were differentially expressed after demethylation but did not show primary tumour-specific methylation (methylated in normal tissue (n=1); methylated only in RCC cell lines (n=9) and not methylated in RCC cell lines (n=12)). Re-expression of CXCL16 reduced growth of an RCC cell line in vitro. In a summary, a functional epigenomic analysis of four RCC cell lines using microarrays representing 11 000 human genes yielded both known and novel candidate TSGs epigenetically inactivated in RCC, suggesting that this is valid strategy for the identification of novel TSGs and biomarkers.

(B75%) of the tumours are classified as clear cell (conventional) and the next most frequent subtype is papillary RCC (B15% of all cases) (Kovacs et al, 1997). The most frequent genetic event in the evolution of clear cell RCC is inactivation of the VHL tumour suppressor gene (TSG) (Latif et al, 1993;Foster et al, 1994;Herman et al, 1994;Clifford et al, 1998), but epigenetic inactivation of TSGs by methylation of CpG dinucleotides in the promoter region has also been implicated in the pathogenesis of RCC (Morrissey et al, 2001;Dallol et al, 2002;Morris et al, 2003). Tumour suppressor gene promoter methylation has been studied most extensively in colorectal cancer, but TSGs that are frequently methylated in human cancers (e.g. p16 ink4a , DAPK, NORE1A, MGMT, SDHB, RARB2 and CDH13) are infrequently methylated in RCC (Morris et al, 2003). This observation prompted us to speculate that, compared to other tumour types, a different repertoire of TSGs might undergo epigenetic inactivation in RCC. Therefore, to identify candidate novel epigenetically inactivated RCC TSGs, we performed a gene expression profiling of four RCC cell lines treated with the demethylated agent 5-azacytidine (5-AZA) (Morris et al, 2005). Analysis of genes differentially expressed before and after demethylation led us to identify HAI-2/SPINT2 as a novel epigenetically inactivated RCC TSG (Morris et al, 2005). In this study, we report the further analysis of our functional epigenomic screen to identify genes relevant in development of RCC.

MATERIALS AND METHODS
Cell Lines, 5-AZA-dC treatment and microarray analysis Full details of the gene expression microarray experiments on four RCC cell lines (KTCL26, SKRC39, SKRC45 and SKRC47) have been described previously (Morris et al, 2005). All RCC cell lines (KTCL26, RCC4, UMRC2, UMRC3, SKRC18, SKRC39, SKRC45, SKRC47, SKRC54, 786-0 and Caki-1) analysed in this study were routinely maintained in DMEM (Invitrogen, San Diego, CA, USA) supplemented with 10% FCS at 371C, 5% CO 2 . The demethylating agent 5-AZA-dC (Sigma, Gillingham, UK) was freshly prepared in ddH 2 O and filter sterilised. Cell lines were plated in 75-cm 2 flasks in DMEM supplemented with 10% FCS at differing densities, depending upon their replication factor, to ensure that both control and 5-AZA-dC-treated lines reached approximately 75% confluency at the point of RNA extraction. Twenty-four hours later, cells were treated with 5 mM 5-AZA-dC. The medium was changed 24 h after treatment and then changed again after 72 h. RNA was prepared 5 days after treatment using RNABee (AMS Biotechnology, Oxon, UK).

Patients and samples
DNA from a total of 127 primary RCCs and 6 non-cancer-related kidneys was analysed. Local ethics committees approved the collection of samples and informed consent was obtained from each patient.

RT -PCR conditions
PCR cycling conditions consisted of an initial denaturing step of 951C for 5 min, followed by 30 cycles of 951C denaturing at 45 s, primer annealing at 55 -601C (gene-dependent) and product extension at 721C for 45 s. Semiquantitative analysis of expression was performed using LabWorks software (Ultraviolet products, Upland, CA, USA). GAPDH primers were 5 0 TGAAGGTCGGAGT CAACGGATTTGGT3 0 and 5 0 CATGTGGGCCATGAGGTCCACC AC3 0 (other RT -PCR primers and conditions upon request).

Bisulphite modification and methylation analysis
Bisulphite DNA sequencing was performed as described previously (Morris et al, 2003). Briefly, 0.5 -1.0 mg of genomic DNA was denatured in 0.3 M NaOH for 15 min at 371C, and then unmethylated cytosine residues were sulphonated by incubation in 3.12 M sodium bisulphite (pH 5.0; Sigma)/5 mM hydroquinone (Sigma) in a thermocycler (Thermo Fisher Scientific, Waltham, MA, USA) for 20 cycles of 30 s at 991C and 15 min at 501C. The sulphonated DNA was recovered using the Wizard DNA cleanup system (Promega, Southampton, UK) in accordance with the manufacturer's instructions. The conversion reaction was completed by desulphonating in 0.3 M NaOH for 10 min at room temperature. The DNA was ethanol-precipitated and resuspended in water.

Promoter methylation analysis
CpG islands were identified on the human genome browser and putative promoter regions were predicted by Promoter Inspector software (Genomatix, www.genomatix.de). Primer details are shown in Supplementary Table 1.

Plasmid constructs and colony formation assay
The CXCL16 expression construct was made by cloning the fulllength human CXCL16 coding region from the SKRC 18 kidney cell line, into the EcoR1 -BamHII sites of pCDNA3.1 vector (Invitrogen). Plasmid constructs were verified by sequencing. Six micrograms of empty vector or 6.8 mg (equal molar amounts) of expression vector was transfected, by calcium phosphate method, into 5 Â 10 5 SKRC39 cells. Forty-eight hours after transfection, cells were seeded in a serial dilution and maintained in DMEM and 10% foetal bovine serum supplemented with 1 mg ml À1 G418 (Gibco, Invitrogen, Paisley, UK). Surviving colonies were stained with 0.4% crystal violet (Sigma) in 50% methanol, 21 days after initial seeding, and counted. Each transfection was carried out in triplicate. Additionally, replicate experiments were carried out to obtain further clones for expression analysis.

Identification and evaluation of differentially expressed genes after demethylation of RCC cell lines
A total of 56 genes (of 11 000 transcripts analysed), each with a 5 0 CpG island, that demonstrated 45-fold increased expression in one cell line or 42-fold in multiple cell lines were identified as candidate epigenetically inactivated RCC TSGs. In addition to SPINT2, four further genes (CDH1, PLAU, IGFB3 and MT1G) had previously been reported to undergo promoter region hypermethylation in RCC tumours (Yoshiura et al, 1995;Morris et al, 2003;Ibanez de Caceres et al, 2006). As we wished to identify novel genes, these were not analysed further. After bioinformatic evaluation (e.g. renal tissue specific expression patterns (array express (www.ebi.ac.uk/arrayexpress/), human genome browser (www.genome.ucsc.edu)), analysis of CpG islands for the presence of predicted promoter regions (www.genomatix.de) and consideration of likely role in tumorigenesis), 34 genes were selected for analysis of expression and/or CpG island methylation status in RCC cell lines±primary RCC tumours.
After analysis of expression pre-and post-treatment with 5-AZA in up to 11 RCC cell lines, nine genes (ATF5, SLC1A4, ID3, STC2, DUSP6, SEMA3C, CD44, HMGA1 and IRF7) were excluded from further investigation as they were not commonly silenced in the pretreatment cell lines (data not shown).

Promoter region methylation analysis of RCC cell lines and tumours
All genes analysed had a 5 0 CpG island (as identified by the human genome browser (http://genome.ucsc.edu/)) and promoter regions were predicted by Genomatix promoter inspector software (http://www.genomatix.de/). The methylation status of 5 0 CpG island promoter regions were analysed by direct sequencing of bisulphate-modified DNA of RCC cell lines and primary tumours for 25 genes. Methylation of 5 0 CpG dinucleotides was rare (ICAM1, IGSF4, EHM2, CLDN1, MUC1, SEMA5A) or absent (CTGF, PHD3, RRM2, PMAIP1, GPR39, MYL2, BAP1, SLC25A21, FBLN1, H2B, ECE1, FZD8) for 18 genes. For ICAM1, IGSF4, EHM2, CLDN1 and MUC1, no methylation was detected in 20 sporadic RCC tumours. These findings suggest that the alterations in expression after demethylation observed for these 18 genes were caused either by (a) methylation of other in cis regulatory regions or (b) that expression of these genes might be regulated by genes that were epigenetically inactivated.
Seven genes demonstrated promoter region hypermethylation concordant with gene expression in RCC cell lines (e.g. KTN19; Figures 1A -C). Four of the seven genes (SST, PTGS1, ISG15, THY1) were frequently methylated in RCC cell lines (Table 1) but were not methylated in primary RCC (n ¼ 20) and although ENG was frequently methylated in RCC cell lines and primary tumours (Table 1), methylation was also detected in adjacent normal tissues and in normal renal tissue from patients without cancer (n ¼ 6).
Two of the seven genes, CXCL16 and KTN19, demonstrated frequent promoter region hypermethylation in RCC cell lines (40 and 62%, respectively) and in primary RCCs (42% (n ¼ 62) and 38% (n ¼ 66), respectively, but not in normal renal tissue from patients without cancer (n ¼ 6). For primary RCC with KTN19 promoter methylation, methylation was rarely (14%, 3/22) detected in the adjacent normal tissue. In contrast, for tumours with CXCL16 methylation (but not for unmethylated tumours), methylation was also detected in the adjacent normal tissue.
We have expanded our previous analysis of HAI-2/SPINT2 promoter methylation in RCC (38%) to incorporate all tumours analysed for KTN19 and CXCL16 promoter methylation. However, no significant correlation was detected between methylation at SPINT2, KTN19 and CXCL16 (P40.1) (data not shown).

RE-expression of CXCL16 reduces the colony forming ability of kidney-derived cell lines
The effect of re-expression of CXCL16 on cell growth was assessed by in vitro colony formation assays. Following transfection of a wild-type CXCL16 expression plasmid (or empty vector) into SKRC39 (an RCC cell line, which is heavily methylated at the CXCL16 promoter), there was a significantly reduced number of G418 resistant colonies (mean 40.1%, t ¼ 4.16 P ¼ 0.014) compared to SKRC39 cells transfected with an empty vector control in three independent experiments (Figure 2).

DISCUSSION
Functional epigenomic screens have proven to be a successful strategy for identifying epigenetically inactivated TSGs in a number of different tumour types (Yamashita et al, 2002;Sato et al, 2003;Lodygin et al, 2005). However, to our knowledge, only one other functional epigenomic study of RCC has been reported. Thus, Ibanez de Caceres et al (2006) performed gene expression microarrays in four RCC cell lines (all different to those analysed in our study) after treatment with a demethylating agent (5-AZA-2 deoxycytidine) and a histone deacetylation inhibitor (trichostatin A) and found that between 111 and 170 genes demonstrated a X3fold upregulation of expression after treatment in each cell line. Then they proceeded to analyse 12 genes that were upregulated X3-fold in at least three of the four cell lines and were expressed in renal tubular cells (BIRC3, NP, GADD45A, NFKB1A, CYCS, TGM2, IGFBP1, COL1A1, CTGF, IGFBP1, GDF15 and PLAU) and found that three genes (IGFBP1, IGFBP3 and COL1A1) demonstrated tumour-specific methylation and seven genes did not show promoter methylation. Despite differences in experimental details and microarrays employed, their findings are similar to those from our more extensive analysis. Thus, it is clear that although upregulation of gene expression after demethylation treatment will often not indicate promoter region hypermethylation, and even if methylation is present, it may not be tumour specific, this experimental approach does provide a strategy for identifying novel TSGs. Keratin 19, an intermediate filament protein responsible for the structural integrity of epithelial cells as a gene, frequently demonstrated promoter methylation in sporadic RCC but not in normal kidney tissue from RCC patients and non-cancer patients. Previously, keratin 19 was reported to be downregulated in head and neck cancers. Furthermore, although re-expression of keratin 19 did not affect the growth rate of transfected cell lines, in vitro invasiveness after treatment with HGF was reduced after reexpression (Crowe et al, 1999).
We found that the transmembrane chemokine CXCL16 (Matloubian et al, 2000;Shimaoka et al, 2000;Wilbanks et al, 2001;Shimaoka et al, 2003) was also frequently methylated in RCC cell lines and in primary tumours but not in normal kidney from non-cancer controls. Although CXCL16 methylation was detected in adjacent normal renal tissue from RCC patients with tumour methylation, this might indicate a premalignant field defect (as described in bronchial epithelium); however, contamination by tumour cells cannot be excluded completely. Interestingly, re-expression of CXCL16 significantly reduced the formation of kidney cell line colonies.
A significant rationale for identifying RCC-associated hypermethylated TSGs is their potential role as biomarkers to identify high-risk individuals and presymptomatic tumours by analysis of urine samples (Battagli et al, 2003;Hoque et al, 2004). In this context, it is valuable to identify both genes that are methylated early in tumorigenesis (but which may not show tumour-specific methylation) and also genes that are specific for tumours.
The functional epigenomic approach we pursued, in addition to identifying genes associated with tumour-specific methylation, also identified significant numbers of genes that (a) appeared to demonstrate cell line-specific methylation (SST, PTGS1, ISG15, THY1) or (b) were upregulated by demethylation but had no apparent promoter region methylation. The former group, although apparently not frequently methylated in primary tumours, might be considered as candidate genes implicated in progression (assuming that acquisition of methylation in cell culture might have a growth advantage). In this context, it is interesting to note that PTGS1 methylation is frequent in prostate cancer (Bastian et al, 2005), THY1 has been implicated as a candidate TSG in nasopharyngeal cancer (Lung et al, 2005) and SST promoter methylation has been described in B90% of colorectal cancers (Mori et al, 2006).
Genes that were upregulated by demethylation but not methylated included genes that have been described previously as TSGs in other cancer types (e.g. BAP1 (Jensen and Rauscher, 1999), IGSF4 , RRM2 (Gautam and Bepler (2006) Figure 2 Re-expression of CXCL16 in RCC cells results in growth suppression. Equal (molar) amounts of empty vector (EV) and pCDNA3.1-CXCL16 (CXCL16) were transfected into SKRC39 cells. Following antibiotic selection, surviving colonies were stained. Each experiment was performed in triplicate and the means of these used to produce the bar chart. The mean number of colonies counted in the EV plates was taken as 100%. There was a statistically significant reduction of colonies in each of the CXCL16 transfectants (P ¼ 0.0141).

2007) and ICAM1
; (Georgolios et al, 2006)). These observations would be consistent with the hypothesis that expression of these genes is reactivated secondary to changes in promoter methylation at upstream regulators. Hence, further analysis of this class of genes may provide insights into key pathways that are dysregulated in RCC and identify candidate upstream regulators for epigenetic analysis.