Epigenetic Screen of Human DNA Repair Genes Identifies Aberrant

ORIGINAL ARTICLE Epigenetic screen of human DNA repair genes identiﬁes aberrant promoter methylation of NEIL1 in head and neck squamous cell carcinoma

J Chaisaingmongkol1, O Popanda1,RWarta2,3, G Dyckhoff2, E Herpel4, L Geiselhart1, R Claus1, F Lasitschka5, B Campos3, CC Oakes1, JL Bermejo6, C Herold-Mende2,3, C Plass1 and P Schmezer1

Aberrant promoter methylation of different DNA repair genes has a critical role in the development and progression of various cancer types, including head and neck squamous cell carcinomas (HNSCCs). A systematic analysis of known human repair genes for promoter methylation is however missing. We generated quantitative promoter methylation profiles in single CpG units of 160 human DNA repair genes in a set of DNAs isolated from fresh frozen HNSCC and normal tissues using MassARRAY technology. Ninety-eight percent of these genes contained CpG islands (CGIs) in their promoter region; thus, DNA methylation is a potential regulatory mechanism. Methylation data were obtained for 145 genes, from which 15 genes exhibited more than a 20% difference in methylation levels between tumor and normal tissues, manifested either as hypermethylation or as hypomethylation. Analyses of promoter methylation with mRNA expression identified the DNA glycosylase NEIL1 (nei endonuclease VIII-like 1) as the most prominent candidate gene. NEIL1 promoter hypermethylation was confirmed in additional fresh frozen HNSCC samples, normal mucosa, HNSCC cell lines and primary human skin keratinocytes. The investigation of laser-microdissected tissues further substantiated increased methylation levels in tumor versus matched non-tumor cells. Immunohistological analysis revealed significantly less NEIL1 protein expression in tumor tissues. 5-Aza-20-deoxycytidine treatment and DNMT1 knockdown resulted in the re-expression of NEIL1 in HNSCC cell lines, which initially carried hypermethylated promoter regions. In conclusion, our results suggest that DNA methylation contributes to the downregulation of NEIL1 expression and might thus have a role in modulating the response to therapies of HNSCC.

Oncogene (2012) 31, 5108--5116; doi:10.1038/onc.2011.660; published online 30 January 2012 Keywords: HNSCC; DNA methylation; NEIL1 expression; DNA glycosylase

INTRODUCTION Aberrant DNA methylation of DNA repair genes MGMT, MLH1 or DNA repair has an important role in genetic instability, is involved MSH2 has been shown to have a critical role in the development 12 --14 in tumor initiation and progression and can modulate cancer and progression of HNSCC. A systematic screen of repair therapy outcome.1 Changes in DNA repair function were found to genes for promoter methylation is however missing. We therefore be induced by mutations and epigenetic mechanisms. It is now performed quantitative DNA methylation analyses in the promoter 15,16 recognized that epigenetic changes such as aberrant DNA region of all known DNA repair and DNA repair-related genes. methylation could be included as one of the ‘hits’ in Knudson’s We identified novel genes showing aberrant methylation in two-hit hypothesis. Epigenetic modifications, especially DNA HNSCC. The DNA glycosylase NEIL1 (nei endonuclease VIII-like 1) methylation, leading to gene silencing have been shown to be a was among those genes showing significant hypermethylation relevant factor in the disease progression of many cancers in tumor tissues. Our results suggest that DNA methylation including head and neck squamous cell carcinoma (HNSCC). contributes to the regulation of NEIL1 expression and might have HNSCC of the oral cavity, oropharynx, hypopharynx or larynx is a role in modulating the response to therapies of HNSCC. the sixth most common cancer by incidence worldwide, with B600 000 expected cases in 2011.2 Despite the treatment options of surgery, radiation and chemoradiation therapy, o50% of all RESULTS patients survive for 5 years, and this situation has not much Screening of promoter methylation in HNSCC revealed aberrantly improved within the last 20 years.3 The role of genetic alterations methylated target genes in the development of HNSCC has been extensively studied; Using evidence from the UCSC CpG Islands Track (NCBI build however, the role of many affected genes is still unknown.4,5 36/hg18 human assembly) we found that 158 out of 160 Several epigenetically silenced genes have been described in investigated genes contained CpG island (CGI) in the 50 regulatory HNSCC including CDKN2A, LHX6, TCF21, CEBPA, SFRP genes, and region offering the potential for epigenetic gene regulation members of the Fanconi anemia/BRCA pathway.6--11 as a regulatory mechanism for DNA repair genes. Two genes

1Division of Epigenomics and Cancer Risk Factors, German Cancer Research Center (DKFZ), Heidelberg, Germany; 2Department of Otorhinolaryngology, Head and Neck Surgery, University of Heidelberg, Heidelberg, Germany; 3Division of Neurosurgical Research, Department of Neurosurgery, University of Heidelberg, Heidelberg, Germany; 4NCT Tissue Bank, Institute of Pathology, University Hospital Heidelberg, Heidelberg, Germany; 5Institute of Pathology, University Hospital Heidelberg, Heidelberg, Germany and 6Institute of Medical Biometry and Informatics, University Hospital Heidelberg, Heidelberg, Germany. Correspondence: Dr P Schmezer, Division of Epigenomics and Cancer Risk Factors (C010), German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, 69120 Heidelberg, Germany. E-mail: [email protected] Received 18 September 2011; revised 9 December 2011; accepted 21 December 2011; published online 30 January 2012 The DNA repair methylome in head and neck cancer J Chaisaingmongkol et al 5109 (POLN and RPA4) contained no CGI within 10 kb upstream or differences lower than 10% between normal and tumor tissues. downstream of the transcription start site. The screen of promoter Thirty-six genes showed methylation differences 410% DNA methylation of 158 genes was performed using a set of 20 (Figure 1a). Out of these, 15 genes showed a higher than 20% HNSCC tissues from patients and 5 normal head and neck mucosa difference in their DNA methylation level and were classified as samples from non-cancer patients. For 13 genes, methylation data aberrantly methylated. Ten of them were located on autosomes could not be obtained due to problematic sequence composition and five genes on the X chromosome. of the region of interest. At the end, quantitative promoter For validation, these 15 genes were analyzed for promoter methylation data in single CpG resolution were obtained for 145 methylation in 47 tumor and 31 normal tissue samples (Table 1). DNA repair genes (5353 CpGs in total). Because of the low number Four genes showed an increased methylation (SALL3, NEIL1, FANCB of samples used in the screen, any statistical analysis would not be and MGMT) and three genes showed a decreased methylation in conclusive. Therefore, we decided to use a non-statistical tumor samples versus controls (APEX2, TREX2, MSH4; probability approach based on the range of methylation values in each values from Wilcoxon rank-sum tests under 0.05). The methylation group. A gene was defined as aberrantly hypermethylated when differences of X-linked DNA repair genes in males and females the maximum methylation value of tumors was 420% higher were separately calculated and displayed (FANCB, APEX2 and than the maximum of normal tissues, and as aberrantly TREX2). There was no indication that methylation levels of hypomethylated when the minimum value of tumors was autosomal genes were affected by gender. Methylation levels of 420% lower than the minimum value of normal tissues. The long interspersed nucleotide elements (LINE-1) were analyzed to majority of screened genes (109 genes) showed methylation obtain an estimate for global DNA methylation changes.17,18

Figure 1. (a) High-throughput methylation analysis of DNA repair genes. Graphical display of differentially methylated genes of 20 HNSCC samples. Each row of the heat map represents a tumor sample and each column represents a methylation difference between tumor and normal tissue. Methylation differences were estimated using the range of methylation values of each group (tumor versus normal), as described in the text. Red color indicates hypermethylation. Green color indicates hypomethylation in tumor compared with normal tissues. Gray indicates unavailable data. (b) mRNA expression of aberrantly methylated genes in HNSCC. Box-and-whisker plot of mRNA expression in a subset of HNSCC (T; n ¼ 27) compared with normal mucosa (N; n ¼ 18) as measured by qRT-PCR. Normalized expression levels were determined using HPRT1, GAPDH and ACTB as reference genes. The differences of mRNA expression between tumor and normal in NEIL1, APEX2 and MSH4 were statistically signiﬁcant (Wilcoxon rank-sum test; *Po0.01; **Po0.001). Whisker indicates 10--90% percentile. Correlation between methylation and mRNA expression was estimated by Spearman’s rank correlation (DPo0.01).

& 2012 Macmillan Publishers Limited Oncogene (2012) 5108 --5116 The DNA repair methylome in head and neck cancer J Chaisaingmongkol et al 5110 Table 1. DNA repair and repair-related genes with aberrant promoter methylation in HNSCC

Gene Chromosomal Accession Methylation Tissue Number Methylation P-valuea Frequency of sample name location number status in type of samples level in percent with aberrant cancer tissue (mean±s.d.) methylationb

SALL3 18q23 NM_171999 Hypermeth. Tumor n ¼ 43 26.3±21.1 o0.0001 35/43 (81%) Normal n ¼ 24 3.1±1.7 NEIL1 15q24.2 NM_024608 Hypermeth. Tumor n ¼ 45 30.9±22.1 0.0027 28/45 (62%) Normal n ¼ 23 11.4±8.9 FANCB Xp22.31 NM_152633 Hypermeth. Tumor (#) n ¼ 37 8.2±8.2 0.0009 17/37 (46%) Normal (#) n ¼ 12 3.8±2.1 Tumor (~) n ¼ 5 46.8±20.8 0.4185 --- Normal (~) n ¼ 11 42.7±14.5 MGMT 10q26.3 NM_002412 Hypermeth. Tumor n ¼ 44 8.0±18.4 0.0313 11/44 (25%) Normal n ¼ 23 0.04±0.21 APEX2 Xp11.21 NM_014481 Hypometh. Tumor (#) n ¼ 41 3.4±4.2 0.3226 --- Normal (#) n ¼ 15 5.0±8.1 Tumor (~) n ¼ 5 20.5±4.1 0.0012 5/5 (100%) Normal (~) n ¼ 15 33.7±7.0 TREX2 Xq28 NM_007205 Hypometh. Tumor (#) n ¼ 42 65.7±19.0 o0.0001 33/42 (79%) Normal (#) n ¼ 14 83.9±2.4 Tumor (~) n ¼ 5 69.7±11.8 0.0021 4/5 (80%) Normal (~) n ¼ 16 82.4±3.5 MSH4 1p31.1 NM_002440 Hypometh. Tumor n ¼ 47 71.1±23.0 0.0081 28/47 (60%) Normal n ¼ 25 88.0±4.8 LINE-1c ------Hypometh. Tumor n ¼ 26 58.9±2.1 o0.0001 22/26 (85%) Normal n ¼ 20 62.3±1.6 Abbreviations: HNSCC, head and neck squamous cell carcinoma; LINE-1, long interspersed nucleotide elements. aStatistically signiﬁcant differences (bold) between tumor and normal tissues according to the Wilcoxon rank-sum test. bFrequency of samples with aberrant methylation is a number of tumor samples which showed higher (Hypermeth.) or lower (Hypometh.) methylation level than mean methylation level±s.d. of normal samples. cLINE-1 methylation was used as an indicator for global methylation.

Tumor samples showed significantly lower methylation levels of primary tumors (n ¼ 34) and in primary human skin keratinocytes LINE-1 elements compared with normal mucosa. (n ¼ 4). The methylation level was elevated in HNSCC versus non-cancerous tissue (Po0.0001) with 71% of samples showing Correlation of mRNA expression with aberrant DNA methylation increased methylation (Figure 2b). To exclude the possibility that contamination with other tissue types might affect the methyla- Expression analyses of the seven selected genes were performed tion pattern of tumor samples, six HNSCC samples were laser in a subset of samples using quantitative real-time PCR (qRT-PCR). microdissected to enable a separate analysis of highly pure tumor Data for FANCB were not available due to technical limitations. and surrounding non-tumor tissue. In four out of six pairs, the Expression of mRNA of genes located on X chromosome (APEX2 tumor fraction showed at least a 20% higher methylation level and TREX2) was not affected by gender as would have been than the matched non-tumor tissue. In addition, high levels of expected due to X inactivation (Wilcoxon rank-sum test; P40.05; promoter methylation were observed in most of the tested HNSCC data not shown). Therefore, expression data of males and females cell lines (90% frequency) but not in primary human skin are shown together. NEIL1, APEX2 and MSH4 showed a decreased keratinocytes. Overall, NEIL1 promoter hypermethylation in HNSCC mRNA expression in tumors (Figure 1b). NEIL1, a bifunctional DNA tumors was consistent throughout the different samples sets used glycosylase involved in base excision repair of oxidatively in this study. damaged DNA,19 showed the most significant difference in the frequency of hypermethylated samples (62% of tumor samples showed hypermethylation in HNSCC, Po0.0001). This hyper- NEIL1 downregulation in HNSCC methylation corresponded to a decrease in mRNA expression Since we checked only mRNA expression of NEIL1_1 isoform in the (Figure 1b; P 0.001; Wilcoxon rank-sum test). o screening step, another primer pair was designed in order to distinguish both isoforms (Figure 2a). Both NEIL1_1 and NEIL1_2 Frequent DNA methylation of NEIL1 promoter in HNSCC were expressed in normal tissue samples, which showed a low Two isoforms of NEIL1 are annotated in the UCSC genome methylation level. This expression, however, was significantly browser. The first isoform contains nine coding exons and codes decreased in tumor tissue (Supplementary Figure S3). Estimates of for a protein with 390 aa (indicated here as NEIL1_1). The second Spearman’s rank correlation revealed an inverse relationship isoform contains 10 coding exons and codes for a protein with 478 between promoter DNA methylation and NEIL1_1 mRNA expres- aa (NEIL1_2). The PCR amplicon of NEIL1, which was used for the sion (rs ¼À0.63; Po0.0001; Figure 3a). This correlation was also methylation screening, was located between transcription start observed for the NEIL1_2 isoform (Supplementary Figure S4). site and the CGI (Figure 2a). A more detailed analysis revealed that NEIL1_1 expression in HNSCC cell lines was lower than in primary the informative CpG units of NEIL1 cover a region of 187 bp human skin keratinocytes, which also showed a lower NEIL1 located adjacent to the CGI, whereas other regions remained methylation (Figure 3b). unmethylated (Supplementary Figure S2). We analyzed NEIL1 protein expression in normal and tumor For further validation, NEIL1 promoter hypermethylation was tissues by immunohistochemistry using a monoclonal antibody confirmed in additional fresh frozen tissue samples (total number: recognizing the N-terminus of NEIL1. Representative staining 135 tumor and 38 normal tissues), in HNSCC cell lines derived from preparations are shown in Figure 3c. Statistical results are

Oncogene (2012) 5108 --5116 & 2012 Macmillan Publishers Limited The DNA repair methylome in head and neck cancer J Chaisaingmongkol et al 5111 tion of the promoter region, and not due to an unspecific effect of 5-Aza-dC, for example, by DNA damage. Interestingly, only NEIL1_1 could be induced by demethylation. To further confirm NEIL1 re-expression by demethylation, a second demethylation approach was carried out by siRNA knockdown of DNMT1. HNSCC cell lines were treated with DNMT1 siRNA for 72 h, and the knockdown efficiency was proven by qRT-PCR. DNMT1 mRNA expression was reduced by 90% in both cell lines (Figure 4e). In HNO223, which showed initially high promoter methylation, the knockdown induced a decrease of NEIL1 methylation by 18% (Figure 4d) and a concurrent increase in NEIL1_1 mRNA expression by a factor of 9.5 (Figure 4e). A smaller increase in NEIL1_2 could also be observed (5.2-fold; Figure 4e). An B2-fold induction of NEIL1_1 and NEIL1_2 expression could be observed in HNO97. Taken together, our data suggest that expression of NEIL1 might be regulated by promoter methylation affecting mainly the first isoform. The relevance of promoter methylation for NEIL1 expression was further studied using CpGL reporter assay in 293T cells. An NEIL1 promoter construct covering the studied 50 region of NEIL1 was treated with SssI or remained untreated. Luciferase expression of the methylated NEIL1 promoter construct was significantly reduced compared with the unmethylated construct (Po0.01, T-test; Figure 5), further strengthening our hypothesis that NEIL1 promoter methylation can reduce gene expression. Figure 2. Methylation of NEIL1 promoter in HNSCC. (a) Schematic representation of NEIL1 gene showing the location of CGI (open box) The impact of NEIL1 promoter methylation on survival end points and PCR amplicons used for methylation analysis (gray boxes). We additionally investigated the association between NEIL1 Arrows indicate transcription start sites (TSSs). Arrowheads indicate location of intron-spanning primers for qRT--PCR. Black boxes methylation and overall survival, cancer-specific survival and indicate exons. (b) Quantitative methylation analysis of the recurrence-free survival in a subset of HNSCC tumors with informative CpG units. Methylation differences were assessed by available clinical data (n ¼ 96). Patient characteristics are summar- Wilcoxon matched pairs test for matched microdissected samples ized in Supplementary Table S3. Log-rank tests identified survival (MD, n ¼ 6, P ¼ 0.0625) and Wilcoxon rank-sum test for tissue differences according to tumor localization (Supplementary Table samples (T, tumor, n ¼ 135; N, normal, n ¼ 38) and cell lines S4). Raw estimates suggested a lower recurrence risk with (H, HNSCC cell lines, n ¼ 34; K, keratinocytes, n ¼ 4; *Po0.01; increased methylation of CpG unit 1 (P ¼ 0.04), but this association **Po0.001). Bar indicates median. disappeared after adjustment by tumor localization (Supplemen- tary Table S5). No association between NEIL1 methylation and survival was found in location-stratified analyses. Methylation differences among tumor locations did not reach statistical summarized in Table 2. We observed less protein expression in significance. tumor (n ¼ 8) than in normal tissue (n ¼ 5). None out of eight tumor samples and three out of five normal tissue samples showed moderate-to-strong NEIL1 expression (Fisher’s exact test, DISCUSSION Po0.05). In this study, we performed a systematic screen of all known human DNA repair and repair-related genes and identified seven Promoter demethylation leads to upregulation of NEIL1 mRNA candidate genes that are frequently hypermethylated or hypo- expression methylated in HNSCC. The results include MGMT hypermethyla- To confirm the functional significance of DNA methylation for tion, which has been previously described in several studies, thus mRNA expression, we investigated whether demethylation of adding plausibility to our screening results.12,13 Information about highly methylated NEIL1 promoter regions leads to re-expression epigenetic regulation of SALL3 and FANCB in cancer is rather of NEIL1. A pair of HNSCC cell lines showing high and low NEIL1 rare.20,21 MSH4 was recently shown to be hypermethylated in methylation was selected. HNO97 showed low NEIL1 methylation HNSCC.22 No data on aberrant promoter methylation have been (2%) and a considerable amount of NEIL1 mRNA expression. published so far for NEIL1, APEX2 and TREX2. Differential HNO223 showed a very high methylation level (77%) and low methylation of MLH1 and MSH2 was not observed, maybe due mRNA expression (Figure 4a). Both cell lines were treated with to our detection method or the relatively small number of samples 5-aza-20-deoxycytidine (5-Aza-dC) for 72 h. A decrease of LINE-1 in our screening. The screening results suggest a propensity for methylation was used as an indicator for the effect of 5-Aza-dC aberrant methylation in DNA repair genes, which are located on treatment on global methylation changes (Figure 4b). NEIL1 the X chromosome. There is currently no conceivable explanation methylation levels decreased with increasing concentrations of for this observation and additional research is needed to attain 5-Aza-dC, especially in HNO223 that initially showed high further clues. promoter methylation (from 78 to 45%; Figure 4b). NEIL1 mRNA Three out of seven differentially methylated genes are expression in HNO223 was induced up to 15-fold (Figure 4c). Very hypomethylated in HNSCC (APEX2, TREX2 and MSH4). As the minor changes of methylation and mRNA induction after 5-Aza-dC promoter regions of these genes are highly methylated in normal treatment were observed in HNO97, the cell line with initially low mucosa, our results suggest that a loss of methylation in cancer NEIL1 methylation (Figure 4c), although its LINE-1 methylation might activate their expression as reported for specific genes.23,24 dropped by the same extent as in HNO223 (Figure 4b). These This hypothesis was however not confirmed by our expression results indicate a specific upregulation of NEIL1 due to demethyla- results. The correlation between promoter hypomethylation in

& 2012 Macmillan Publishers Limited Oncogene (2012) 5108 --5116 The DNA repair methylome in head and neck cancer J Chaisaingmongkol et al 5112

Figure 3. NEIL1 methylation and expression in HNSCC. (a) Scatter plot of methylation levels against mRNA expression levels of NEIL1_1 in tissue samples. Correlation between both parameters was quantiﬁed by Spearman’s rank correlation (rs). Expression values were normalized by using HPRT1, GAPDH and ACTB. Relative mRNA expression was calculated by setting the expression value of the calibrator sample as 1. T, HNSCC; N, normal tissue. (b) NEIL1_1 mRNA expression in HNSCCs (T; n ¼ 27), normal tissues (N; n ¼ 17), HNSCC cell lines (H; n ¼ 34) and human skin keratinocytes (K; n ¼ 4). Normalization and relative mRNA expression were calculated as described above. NEIL1_1 mRNA expression decreased in tumor compared with normal tissues and HNSCC cell lines showed lower NEIL1_1 mRNA expression than keratinocytes (Wilcoxon rank-sum test; *Po0.05; **Po0.001). Whisker indicates 10--90% percentile. (c) Immunohistochemical analysis of NEIL1 protein expression in representative tissue samples. All slides were counterstained with hemalaun for visualization of nuclei (blue). Arrows indicate NEIL1 expression (brown) in normal musoca (Nmuc) and normal tonsil (Nton); T, HNSCC samples. Scale bar, 100 mm.

Among the hypermethylated genes, NEIL1 showed the stron- Table 2. Summary of NEIL1 protein expression analysis performed in a gest correlation between methylation in HNSCC and decreased tissue samples gene expression. The NEIL1 gene product is involved in base excision repair of oxidatively damaged DNA.19 Its function as a Sample ID Tissue type Staining Methylation b c bifunctional DNA glycosylase has been studied in relative intensity level in percent detail.26,27 Cell lines lacking NEIL1 were more sensitive to 28,29 HNO356a Normal (tonsil) +++ NA g-irradiation, interstrand crosslinks and bulky adducts. NEIL1 HNO184 Normal (mucosa) ++ NA knockout mice showed increased levels of DNA damage, deletions 30 HNO107 Normal (mucosa) ++ 7% and metabolic syndrome. NEIL1 and NEIL1/NTH1 double knock- HNO163 Normal (mucosa) + 1% out mice developed pulmonary and hepatocellular tumors.31 HNO80 Normal (mucosa) À NA Genetic alterations of the NEIL1 gene in human gastric cancer lead HNO421a Tumor + 6% to impaired DNA glycosylase activity or a truncated protein.32 HNO341 Tumor + 32% Furthermore, NEIL1 has been shown to interact with other DNA HNO21 Tumor + 49% repair genes which have been linked to cancer.31,33,34 The level of HNO343 Tumor + 57% NEIL1 is greatly diminished in cells depleted for FANC proteins and HNO393 Tumor À 9% HNO87 Tumor À 30% Fanconi anemia cell lines. Expression of the NEIL1 protein partially HNO348 Tumor 35% reverses the hypersensitivity of Fanconi anemia cells to interstrand À 34 HNO479z Tumor À 54% crosslinks. However, very few data on a possible epigenetic a regulation of this gene are so far available. We show here for the Defined as score: negative (À); weak (+); moderate (++); and strong b first time that NEIL1 is consistently hypermethylated in HNSCC staining (+++). For Fisher’s exact test, negative and weak staining were tissues and HNSCC cell lines. It remains unclear why a smaller defined as low expression; moderate and strong staining were defined as subgroup of samples does not show increased methylation levels high expression. cMedian of methylation levels in normal tissues was 8% (range: 0 --46%; n ¼ 38); in tumor tissues was 27% (1 --80%; n ¼ 135). when compared with controls. This observation is unlikely to be caused by deletions on chromosome 15 as they have not been reported so far to occur within the chromosomal region of NEIL1 in cancer and gene re-expression seems less straightforward, HNSCC.4,5 No such chromosomal aberrations have been found in and this has been similarly reported for other genes.25 18 out of 20 (90%) of HNSCC cell lines which we used in Additional regulatory mechanisms such as microRNAs or histone our experiments.35 However, it is known that tumors have modifications might also result in reduced expression in tumors. heterogeneous patterns of somatic genetic and epigenetic The decreased expression of MSH4 and APEX2 in HNSCC did not alterations. Previously, methylation patterns in HNSCC have been correlate with methylation. associated with etiologic exposures, such as smoking and

Oncogene (2012) 5108 --5116 & 2012 Macmillan Publishers Limited The DNA repair methylome in head and neck cancer J Chaisaingmongkol et al 5113

Figure 4. Activation of NEIL1 expression after promoter demethylation. (a) NEIL1 mRNA expression (left) and DNA methylation (right) in HNSCC cell lines. Expression levels were normalized with reference genes and relative expression was calculated using a calibrator (set as 1). Gray bar, NEIL1_1 isoform; open bar, NEIL1_2 isoform; black bar, methylation value. Error bars indicate s.e.m. of three independent experiments. Methylation level (b) and mRNA expression (c)ofNEIL1 in HNSCC cell lines HNO223 and HNO97 after treatment with the demethylating agent 5-Aza-dC (0.5, 1 and 2 mM) for 72 h. LINE-1 methylation indicates global demethylation. Methylation level (d) and mRNA expression (e)ofNEIL1 in HNSCC cell lines after treatment with DNMT1-siRNA (20 nM) for 72 h. Fold change in expression levels was determined by normalizing expression values of treated against untreated samples. Error bars indicate s.e.m. of two independent experiments. alcohol.36 This makes NEIL1 methylation attractive for the use as a thymine-DNA glycosylase (TDG), might be implicated in the ten- biomarker. eleven translocation (TET)-mediated and base excision repair We found a significant correlation between increased methyla- (BER)-coupled demethylation pathways which involve oxidized tion of the upstream promoter region and a reduction in mRNA products of 5-mC such as 5-hydroxymethylcytosine or 5-carbox- expression of NEIL1 in our tumor samples. This observation was ylcytosine.41 An impaired demethylating activity could then experimentally confirmed using a collection of HNSCC cell lines. contribute to increased methylation in tumors. Since NEIL1 has been shown to be activated by acute oxidative Epigenetic silencing of DNA repair genes is thought to increase stress37 and downregulated probably by the long-term exposure genetic instability and thus promote carcinogenesis.42 Once the tumor to reactive oxygen species in HCV-infected liver cells,38 it is has developed, decreased DNA repair gene expression can however possible that methylation of NEIL1 promoter region controls the be beneficial to the patients by increasing the sensitivity of the tumor gene expression under these conditions. In addition, major risk cells to DNA-damaging therapies, as shown for MGMT hypermethyla- factors of HNSCC such as the chronic consumption of tobacco tion in glioblastoma.43 In our study, patients with high NEIL1 smoke and alcohol can induce a prolonged exposure to reactive methylation showed a significantly higher recurrence-free survival oxygen species,39,40 which might cause epigenetic changes, such butthiswasnotapparentafterstratificationbytumorlocalization.The as methylation, affecting NEIL1 expression either directly or statistical power of our study cohort might however be too small for a indirectly. Furthermore, it was also shown that HNSCC risk factors stratified analysis. As 490% of our patients received a DNA damage- (drinking, smoking and HPV) are associated with epigenetic inducing radiotherapy and/or chemotherapy post surgery (Supple- inactivation of genes, such as the SFRP genes.11 mentary Table S3), NEIL1 deficiency might nevertheless have the Future experiments will be necessary to further elucidate the suggested positive effect in patient survival due to an increased role of epigenetic regulation of NEIL1 expression in HNSCC and to sensitivity of the tumor cells to DNA-damaging chemotherapy or verify whether it is a cause or consequence of tumorigenesis. DNA radiotherapy. For possible clinical implications, comprehensive studies methylation changes might occur early in carcinogenesis and with more homogenous cohorts and larger patient numbers are could therefore be used as an early indicator or biomarker in warranted. In any case, our novel data on aberrant promoter clinical studies. With respect to methylation changes, one might methylation of NEIL1 will stimulate such studies and thus open the speculate that the glycosylase activity of NEIL1, in addition to possibility to improve HNSCC diagnosis and treatment.

& 2012 Macmillan Publishers Limited Oncogene (2012) 5108 --5116 The DNA repair methylome in head and neck cancer J Chaisaingmongkol et al 5114 Fisher Scientific, Lafayette, CO, USA) were prepared and added per well. Cells treated with only DharmaFECT Reagent served as a negative control. Knockdown efficiency was confirmed by qRT-PCR.

DNA and RNA isolation DNA from frozen tissue samples was isolated using DNeasy Blood and Tissue Kit (Qiagen). Total RNA from tissue samples was isolated using Trizol (Invitrogen, Karlsruhe, Germany). DNA and RNA from cell lines were isolated using AllPrep DNA/RNA Mini Kit (Qiagen).

Bisulfite conversion, primer design strategy and methylation analysis Genomic DNA was converted by bisulfite treatment using EZ DNA methylation kit (Zymo Research, Orange, CA, USA). After bisulfite conversion and PCR amplification of the DNA region of interest, MassARRAY technology (Sequenom, San Diego, CA, USA) was used to quantify the methylation status of a specific promoter region in set of 20 Figure 5. DNA methylation reduced NEIL1 promoter activity. 46 Luciferase CpG-free pCpGL vector constructs contained NEIL1 HNSCCs and 5 normal tissues samples as described. The list of human 15,16 50 region from methylation study were M.SssI methylated or left DNA repair genes is based on a compilation by Wood et al., which we unmethylated and transfected into 293T cells. The reporter activity supplemented with 10 additional genes (Supplementary Table S1). PCR was studied 48 h after transfection. Relative luciferase activity was primers were designed to cover a region located near the transcription determined by normalizing firefly luciferase against Renilla luciferase start site and associated with the CGI (Supplementary Figure S1). To activity. Empty vectors were used to normalize the background normalize for PCR induced biases, a 6-point methylation standard was noise. Error bars indicate s.e.m. of three independent experiments. included in the assay to generate a standard curve and used for data Luciferase expression of methylated version of NEIL1 promoter correction. The resulting percent methylation values are displayed as an construct was reduced comparing with unmethylated construct average of all CpG units within an amplicon. Informative CpG units are (T-test; **Po0.01). defined as the CpG units depicting the strongest methylation differences between normal and tumor tissues. A smaller PCR amplicon of NEIL1 (187 bp) located within the informative region was used to characterize MATERIALS AND METHODS methylation levels in laser-microdissected tissues. Display of methylation Patient samples and cell lines results as heat maps and unsupervised clustering was performed using the Frozen tissue specimens from HNSCC tumor patients and clinically normal Multiple Experimental Viewer software (http://www.tm4.org/mev.html). head and neck mucosa samples from non-cancer patients who underwent tonsillectomy were obtained from the Department of Otorhinolaryngology, Reverse transcription and qRT--PCR Head and Neck Surgery, University of Heidelberg via the tissue bank of the The total RNA (500 ng--1 mg) was treated with DNaseI (Fermentas, St Leon- NCT Tissue Bank (Heidelberg, Germany). The study was approved by the Rot, Germany) followed by cDNA synthesis using SuperScript III reverse institutional ethics committee (207/2005 and 206/2005). The number of transcriptase (Invitrogen) and random hexamers. Primers and PCR tissue samples used for each experiment is indicated in Supplementary conditions are shown in Supplementary Table S2. Quantitative mRNA Table S6. expression analysis was performed using the Absolute QPCR SYBR Green Clinical information and follow-up data were analyzed by reviewing the Mix (ABgene, Epsom, UK) on a LightCycler 480 (Roche Diagnostics, medical records, radiographic images, by telephone or written correspon- Mannheim, Germany) as described.47 Normalized expression ratios were dence and by review of death certificate. Mean follow-up time was 6.9 determined for each sample using HPRT1, GAPDH and ACTB as reference years (±3.9 years). Patients were followed from date of first diagnosis to genes. Stratagene QPCR Human Reference Total RNA (Agilent Technolo- the end of study (30 June 2010), whereas dead patients were censored. All gies, Santa Clara, CA, USA) was used as a calibrator for calibrator- patients were chemotherapy and radiotherapy naive at the time of surgery. normalized relative quantification. Histological evaluation of all tissue samples was performed by a single pathologist. A subset of frozen samples was cut into 18 mm thick sections Immunohistochemistry using a cryostat (Leica CM1850; Leica Microsystems, Wetzlar, Germany) and Immunohistochemical staining was performed using a mouse monoclonal was processed by laser microdissection and pressure catapulting using a antibody recognizing the N-terminus of NEIL1 (sc-271167; Santa Cruz Microbeam LMPC System (Carl Zeiss MicroImaging, Jena, Germany) as Biotechnology, Santa Cruz, CA, USA) at a concentration of 10 mg/ml. described.44 DNA was isolated using the Puregene Tissue Core Kit B Specificity of primary antibodies was ensured using an irrelevant control (Qiagen, Hilden, Germany) according to manufacturer’s instructions. antibody. Incubation with primary and secondary antibodies and detection HNSCC cell lines45 were cultured routinely in Dulbecco’s modified with Vectastain ELITE ABC Kit (Vector Laboratories, Burlingame, CA, USA) Eagle’s medium (DMEM) supplemented with 10% fetal calf serum (PAA was carried out as described.45 An amplification step was performed with Laboratorie, Co¨lbe, Germany) and 1% penicillin/streptomycin (Sigma, St the Tyramide Signal Amplification kit (NEL700; Perkin-Elmer, Rodgau, Louis, MO, USA) in an atmosphere containing 5% CO . 293T cell line was 2 Germany) according to manufacturer’s instructions. cultured in the same medium without antibiotics.

Luciferase reporter assay Treatment of cell lines with 5-Aza-dC and siRNA NEIL1 promoter construct covering the studied 50 region was cloned HNSCC cell lines HNO97 and HNO223 were seeded in 6-well plates directly into the CpG-free pCpGL vector and SssI methylated in vitro or left (1.5 Â 105 cells/well) containing 2 ml culture medium. Cells were treated unmethylated.48 The 293T cells (3000 cells/well) were transiently with 5-Aza-dC dissolved in phosphate-buffered saline (5-Aza-dC; Sigma) at transfected in 384-multiwell plates using 0.2 ml TransIT-LT1 Transfection 0.5, 1.0 and 2.0 mM for 72 h by adding freshly prepared 5-Aza-dC in culture Reagent (Mirus Bio, Madison, WI, USA) and 40 ng of reporter plasmid and medium and changed every 24 h. For transient transfection with siRNA, 10 ng of Renilla vector (Promega, Mannheim, Germany) according to cells were seeded as described. In all, 20 nM DNMT1 Dharmacon manufacturer0s instructions. After 48 h, cells were harvested, and cell ON-TARGETplus SMARTpool and 2.4 ml DharmaFECT Reagent (Thermo lysates assayed for ﬁreﬂy and Renilla luciferase activity using the dual

Oncogene (2012) 5108 --5116 & 2012 Macmillan Publishers Limited The DNA repair methylome in head and neck cancer J Chaisaingmongkol et al 5115 luciferase reporter assay system (Promega) on a SPECTRAmax 250 14 Sengupta S, Chakrabarti S, Roy A, Panda CK, Roychoudhury S. Inactivation of spectrophotometer (GMI Inc., Cold Spring, NY, USA). Firefly luciferase was human mutL homolog 1 and mutS homolog 2 genes in head and neck squamous normalized against Renilla luciferase activity. Empty vectors were used to cell carcinoma tumors and leukoplakia samples by promoter hypermethylation normalize the background noise. and its relation with microsatellite instability phenotype. Cancer 2007; 109: 703 --712. 15 Wood RD, Mitchell M, Sgouros J, Lindahl T. Human DNA repair genes. Science Statistical analysis 2001; 291: 1284 --1289. Percent methylation values were compared among groups using the non- 16 Wood RD, Mitchell M, Lindahl T. Human DNA repair genes, 2005. Mutat Res 2005; parametric Wilcoxon rank-sum test. The Wilcoxon matched pairs test was 577: 275 --283. used to compare matched tumor and normal microdissected tissue 17 Yang AS, Estećio MR, Doshi K, Kondo Y, Tajara EH, Issa JP. A simple method for estimating global DNA methylation using bisulfite PCR of repetitive DNA samples. Correlations between methylation and mRNA expression were 0 elements. Nucleic Acids Res 2004; 32: e38. assessed by the non-parametric Spearman s rank correlation coefficient 18 Nelson HH, Marsit CJ, Kelsey KT. Global methylation in exposure biology and (rs). Fisher’s exact test was chosen to examine the association between translational medical science. Environ Health Perspect 2011; 119: 1528 --1533. tissue type and protein overexpression. Survival differences between 19 Bandaru V, Sunkara S, Wallace SS, Bond JP. A novel human DNA glycosylase that groups of patients were explored by log-rank tests and subsequent removes oxidative DNA damage and is homologous to Escherichia coli multivariate Cox regression models. GraphPad Prism (GraphPad Software endonuclease VIII. DNA Repair (Amst) 2002; 1: 517 --529. Inc, La Jolla, CA, USA) was applied for the analysis. 20 Smith IM, Mithani SK, Mydlarz WK, Chang SS, Califano JA. Inactivation of the tumor suppressor genes causing the hereditary syndromes predisposing to head and neck cancer via promoter hypermethylation in sporadic head and neck CONFLICT OF INTEREST cancers. ORL J Otorhinolaryngol Relat Spec 2010; 72: 44 --50. 21 Yu J, Zhu T, Wang Z, Zhang H, Qian Z, Xu H et al. A novel set of DNA methylation The authors declare no conflict of interest. markers in urine sediments for sensitive/specific detection of bladder cancer. Clin Cancer Res 2007; 13: 7296 --7304. 22 Poage GM, Houseman EA, Christensen BC, Butler RA, Avissar-Whiting M, McClean ACKNOWLEDGEMENTS MD et al. Global hypomethylation identifies loci targeted for hypermethylation in head and neck cancer. Clin Cancer Res 2011; 17: 3579 --3589. We thank R Gliniorz, O Muecke, P Waas and O Zelezny (Division of Epigenomics and 23 Sato N, Fukushima N, Matsubayashi H, Goggins M. Identification of maspin and Cancer Risk Factors, German Cancer Research Center) and J Scheuerer (Institute of Pathology) for their excellent technical assistance, all other members of the Plass S100P as novel hypomethylation targets in pancreatic cancer using global gene laboratory for their help and thoughtful discussion, M Zucknick (Division of expression profiling. Oncogene 2004; 23: 1531 --1538. Biostatistics, German Cancer Research Center) for statistical support and P Boukamp 24 Wilson AS, Power BE, Molloy PL. DNA hypomethylation and human diseases. (Division of Genetics of Skin Carcinogenesis, German Cancer Research Center) for Biochim Biophys Acta 2007; 1775: 138 --162. providing primary human skin keratinocytes. This project was in part supported by a 25 Dokun OY, Florl AR, Seifert HH, Wolff I, Schulz WA. Relationship of SNCG, S100A4, scholarship of The Royal Thai Government to JC. This work was supported in part by S100A9 and LCN2 gene expression and DNA methylation in bladder cancer. Int J NIDCR, grant DE13123 (CP). FL is funded by the Deutsche Forschungsgemeinschaft, Cancer 2008; 123: 2798 --2807. grant SFB938 Z2. 26 Doublie S, Bandaru V, Bond JP, Wallace SS. The crystal structure of human endonuclease VIII-like 1 (NEIL1) reveals a zincless finger motif required for glycosylase activity. Proc Natl Acad Sci USA 2004; 101: 10284 --10289. 27 Dou H, Mitra S, Hazra TK. Repair of oxidized bases in DNA bubble structures by REFERENCES human DNA glycosylases NEIL1 and NEIL2. J Biol Chem 2003; 278: 49679 --49684. 1 Hoeijmakers JHJ. Genome maintenance mechanisms for preventing cancer. 28 Rosenquist TA, Zaika E, Fernandes AS, Zharkov DO, Miller H, Grollman AP. The Nature 2001; 411: 366 --374. novel DNA glycosylase, NEIL1, protects mammalian cells from radiation-mediated 2 Leemans CR, Braakhuis BJM, Brakenhoff RH. The molecular biology of head and cell death. DNA Repair 2003; 2: 581 --591. neck cancer. Nat Rev Cancer 2011; 11: 9 --22. 29 Couve´-Privat S, Mac G, Rosselli F, Saparbaev MK. Psoralen-induced DNA adducts 3 Seiwert TY, Salama JK, Vokes EE. The chemoradiation paradigm in head and neck are substrates for the base excision repair pathway in human cells. Nucleic Acids cancer. Nat Clin Prac Oncol 2007; 4: 156 --171. Res 2007; 35: 5672 --5682. 4 Schmezer P, Plass C. [Epigenetische Aspekte bei Karzinomen der Kopf-Hals- 30 Vartanian V, Lowell B, Minko IG, Wood TG, Ceci JD, George S et al. The metabolic Region]. HNO 2008; 56: 594 --602. syndrome resulting from a knockout of the NEIL1 DNA glycosylase. Proc Natl Acad 5 Ha PK, Chang SS, Glazer CA, Califano JA, Sidransky D. Molecular techniques and Sci USA 2006; 103: 1864 --1869. genetic alterations in head and neck cancer. Oral Oncol 2004; 45: 335 --339. 31 Chan MK, Ocampo-Hafalla MT, Vartanian V, Jaruga P, Kirkali G, Koenig KL et al. 6 Kraunz KS, Hsiung D, McClean MD, Liu M, Osanyingbemi J, Nelson HH et al. Targeted deletion of the genes encoding NTH1 and NEIL1 DNA N-glycosylases Dietary folate is associated with p16INK4A methylation in head and neck reveals the existence of novel carcinogenic oxidative damage to DNA. DNA Repair squamous cell carcinoma. Int J Cancer 2006; 119: 1553 --1557. (Amst) 2009; 8: 786 --794. 7 Estecio MRH, Youssef EM, Rahal P, Fukuyama EE, Gois-Filho JF, Maniglia JV et al. 32 Shinmura K, Tao H, Goto M, Igarashi H, Taniguchi T, Maekawa M et al. Inactivating LHX6 is a sensitive methylation marker in head and neck carcinomas. Oncogene mutations of the human base excision repair gene NEIL1 in gastric cancer. 2006; 25: 5018 --5026. Carcinogenesis 2004; 25: 2311 --2317. 8 Smith LT, Lin M, Brena RM, Lang JC, Schuller DE, Otterson GA et al. Epigenetic 33 Guan X, Bai H, Shi G, Theriot CA, Hazra TK, Mitra S et al. The human checkpoint regulation of the tumor suppressor gene TCF21 on 6q23-q24 in lung and head sensor Rad9-Rad1-Hus1 interacts with and stimulates NEIL1 glycosylase. Nucleic and neck cancer. Proc Natl Acad Sci USA 2006; 103: 982 --987. Acids Res 2007; 35: 2463 --2472. 9 Bennett KL, Hackanson B, Smith LT, Morrison CD, Lang JC, Schuller DE et al. 34 Mace-Aime G, Couve S, Khassenov B, Rosselli F, Saparbaev MK. The Fanconi Tumor suppressor activity of CCAAT/enhancer binding protein a is epigenetically anemia pathway promotes DNA glycosylase-dependent excision of interstrand down-regulated in head and neck squamous cell carcinoma. Cancer Res 2007; DNA crosslinks. Environ Mol Mutagen 2010; 51: 508 --519. 67: 4657 --4664. 35 Freier K, Hofele C, Knoepfle K, Gross M, Devens F, Dyckhoff G et al. Cytogenetic 10 Marsit CJ, Liu M, Nelson HH, Posner M, Suzuki M, Kelsey KT. Inactivation of the characterization of head and neck squamous cell carcinoma cell lines as model Fanconi anemia/BRCA pathway in lung and oral cancers: implications for systems for the functional analyses of tumor-associated genes. J Oral Pathol Med treatment and survival. Oncogene 2003; 23: 1000 --1004. 2010; 39: 382 --389. 11 Marsit CJ, McClean MD, Furniss CS, Kelsey KT. Epigenetic inactivation of the SFRP 36 Marsit CJ, Christensen BC, Houseman EA, Karagas MR, Wrensch MR, Yeh RF et al. genes is associated with drinking, smoking and HPV in head and neck squamous Epigenetic profiling reveals etiologically distinct patterns of DNA methylation in cell carcinoma. Int J Cancer 2006; 119: 1761 --1766. head and neck squamous cell carcinoma. Carcinogenesis 2009; 30: 416 --422. 12 Dikshit RP, Gillio-Tos A, Brennan P, De Marco L, Fiano V, Martinez-Pen˜uela JM et al. 37 Das A, Hazra TK, Boldogh I, Mitra S, Bhakat KK. Induction of the human oxidized Hypermethylation, risk factors, clinical characteristics, and survival in 235 patients base-specific DNA glycosylase NEIL1 by reactive oxygen species. J Biol Chem 2005; with laryngeal and hypopharyngeal cancers. Cancer 2007; 110: 1745 --1751. 280: 35272 --35280. 13 Martone T, Gillio-Tos A, De Marco L, Fiano V, Maule M, Cavalot A et al. Association 38 Pal S, Polyak SJ, Bano N, Qiu WC, Carithers RL, Shuhart M et al. Hepatitis C virus between hypermethylated tumor and paired surgical margins in head and neck induces oxidative stress, DNA damage and modulates the DNA repair enzyme squamous cell carcinomas. Clin Cancer Res 2007; 13: 5089 --5094. NEIL1. J Gastroenterol Hepatol 2010; 25: 627 --634.

& 2012 Macmillan Publishers Limited Oncogene (2012) 5108 --5116 The DNA repair methylome in head and neck cancer J Chaisaingmongkol et al 5116 39 DeMarini DM. Genotoxicity of tobacco smoke and tobacco smoke condensate: a 44 Funke B, Autschbach F, Kim S, Lasitschka F, Strauch U, Rogler G et al. review. Mutat Res 2004; 567: 447 --474. Functional characterisation of decoy receptor 3 in Crohn0s disease. Gut 2009; 40 Seitz HK, Stickel F. Molecular mechanisms of alcohol-mediated carcinogenesis. 58: 483 --491. Nat Rev Cancer 2007; 7: 599 --612. 45 Ninck S, Reisser C, Dyckhoff G, Helmke B, Bauer H, Herold-Mende C. Expression 41 He YF, Li BZ, Li Z, Liu P, Wang Y, Tang Q et al. Tet-mediated formation of proﬁles of angiogenic growth factors in squamous cell carcinomas of the head 5-carboxylcytosine and its excision by TDG in mammalian DNA. Science 2011; and neck. Int J Cancer 2003; 106: 34 --44. 333: 1303 --1307. 46 Ehrich M, Nelson MR, Stanssens P, Zabeau M, Liloglou T, Xinarianos G et al. 42 Esteller M, Risques RA, Toyota M, Capella G, Moreno V, Peinado MA et al. Promoter Quantitative high-throughput analysis of DNA methylation patterns by hypermethylation of the DNA repair gene O(6)-methylguanine-DNA methyl- base-speciﬁc cleavage and mass spectrometry. Proc Natl Acad Sci USA 2005; transferase is associated with the presence of G:C to A:T transition mutations in 102: 15785 --15790. p53 in human colorectal tumorigenesis. Cancer Res 2001; 61: 4689 --4692. 47 Wiebalk K, Schmezer P, Kropp S, Chang-Claude J, Celebi O, Debus J et al. In vitro 43 Hegi ME, Liu L, Herman JG, Stupp R, Wick W, Weller M et al. Correlation of O6- radiation-induced expression of XPC mRNA as a possible biomarker for developing methylguanine methyltransferase (MGMT) promoter methylation with clinical adverse reactions during radiotherapy. Int J Cancer 2007; 121: 2340 --2345. outcomes in glioblastoma and clinical strategies to modulate MGMT activity. J Clin 48 Klug M, Rehli M. Functional analysis of promoter CpG methylation using a CpG- Oncol 2008; 26: 4189 --4199. free luciferase reporter vector. Epigenetics 2006; 1: 127 --130.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene (2012) 5108 --5116 & 2012 Macmillan Publishers Limited