Frequently Methylated Tumor Suppressor Genes in Head and Neck Squamous Cell Carcinoma

Introduction

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide. According to the National Cancer Institute's Surveillance Epidemiology and End Results cancer statistics review, 34,360 individuals were predicted to be diagnosed with HNSCC in 2007. Treatment of HNSCC is complex, using radiation therapy, surgery, and chemoradiation. Patients with advanced HNSCC are limited to a complete response of 50% and often require long-term rehabilitation. However, early HNSCC detection increases survival to 80%. Therefore, early detection markers may potentially serve as a predictive tool for diagnosis and recurrence of HNSCC ( 1).

DNA methylation within the promoter region of a gene can result in chromatin compaction and inhibition or down-regulation of gene transcription. Aberrant promoter methylation is often responsible for gene silencing in a variety of malignancies, and several genes have been shown to be epigenetically down-regulated in HNSCC ( 2). Because methylation is an early event, identification of methylation markers could provide great promise for early detection and treatment in HNSCC ( 1).

Our laboratory uses a technique called Restriction Landmark Genomic Scanning (RLGS) for genome-wide promoter methylation analysis. Previous RLGS analysis on several HNSCC patient samples revealed more methylation in metastatic HNSCC samples compared with primary tumors, and the methylated loci in metastatic versus primary tumors within the same patient were found to be dissimilar ( 3). Also, RLGS analysis has been used to identify amplification of potential oncogenes in HNSCC ( 4). The study described here used these RLGS profiles to identify potential novel tumor suppressor genes that are frequently methylated in HNSCC. This analysis provided 5 genes (IRX1, FUSSEL18, EBF3, SLC5A8, and SEPT9) that were epigenetically silenced in >10% of the patient samples analyzed. Interestingly, these genes have previously been found to exhibit tumor suppressor activity in cancer ( 5– 9). Also, they have been shown to potentially be involved in transforming growth factor (TGF-β) signaling ( 5, 10– 13), which is often disrupted in HNSCC ( 14– 16).

Materials and Methods

Patient samples and cell lines. Frozen tumor tissues and adjacent normal mucosa tissue (normals) from HNSCC patients were attained from The Ohio State University Medical Center via the Cooperative Human Tissue Network. Surgery was performed on all patients at The Ohio State University Medical Center. Control samples were collected from morphologically normal tissue located at least 3 cm from the tumor margin. Histopathologic evaluation was performed on all samples for verification. A second set of HNSCC and normal mucosa was obtained from the Tissuebank of the National Center of Tumor Diseases, Heidelberg. All sample collections were done according to the NIH guidelines and under a protocol approved by The Ohio State University's or the Medical Faculty Heidelberg's Institutional Review Boards. The five different established human HNSCC cell lines used in the study (SCC8, SCC11B, SCC17AS, SCC22B, and SCC25) were kindly provided by Dr. Thomas Carey (University of Michigan, Ann Arbor, MI). Cell lines were maintained in DMEM with 10% fetal bovine serum and 1% streptomycin/penicillin antibiotics.

MassARRAY methylation analysis. Quantitative DNA methylation analysis was performed by MassARRAY technique. Briefly, 1 μg of HNSCC patient or cell line DNA was bisulfite treated, in vitro transcribed, cleaved by RNase A, and subjected to MALDI-TOF mass spectrometry analysis to determine methylation patterns, as previously described ( 17). To normalize for any bias, standards (0%, 5%, 10%, 20%, 40%, 60%, 80%, and 100% SssI methylated) were included in the assay to generate an “observed to expected plot.” The equation of the best-fitting line was then used to normalize the values for the HNSCC samples. Subsequent to normalization, tumor samples with more methylation than adjacent normal samples were considered methylated. The methylation results are displayed as a heat map using the Mutliple Experimental Viewer software ( 18).

RNA isolation and cDNA synthesis. RNA was isolated according to the manufacturer's protocol. DNase treatment was performed on 1 to 2 μg of RNA by adding 2 U of DNaseI (Invitrogen), 1 μL DNase buffer, and 0.4 μL RNase Out (Invitrogen) for 15 min at room temperature. One microliter of EDTA was then added to the mix for 10 min at 65°C, followed by an incubation on ice for 5 min. Complementary DNA synthesis was performed by the following reaction: 2 μL random hexamers and 1 μL deoxynucleotide triphosphates (10 mmol/L) for 5 min at 65°C, then 2 min at 4°C; 2 μL of 10× buffer, 4 μL MgCl₂, 2 μL DTT, and 1 μL RNase Out was added for 2 min at 25°C; 100 U of SuperScript II (Invitrogen) for 50 min at 42°C; 15 min at 70°C, then transferred to 4°C.

5 aza-2′-deoxycytidine treatment. HNSCC cell lines were incubated for 96 h with 1 μmol/L 5-aza-2′-deoxycytidine (5-aza-dC; Sigma) with medium changed every day. Treated cells were harvested for analysis 1 to 4 d after the procedure. Cells were suspended in Trizol for RNA isolation.

Real-time PCR. Quantitative mRNA expression was measured using SYBR Green I (Bio-Rad) in an I-Cycler (Bio-Rad). Expression of glycosylphosphatidylinositol was used as the internal control gene. I-Cycler conditions and real-time PCR (RT-PCR) primers can be provided upon request. For the RT-PCRs on the HNSCC patient samples, each tumor was normalized to the expression of its adjacent normal tissue, which was set to a value of 1. For the 5-aza-dC–treated HNSCC cell lines, each treated cell line was normalized to its untreated counterpart, which was set to a value of 1.

Plasmids. The SLC5A8 overexpression vector was provided by Dr. Joe Costello (University of California, San Francisco, CA). The 1.71-kb EBF3 cDNA was cloned into the EcoRI site of pBABE, and the 1.61-kb IRX1 cDNA was cloned into the BamHI and SalI sites of pBABE.

Transfections. Stable transfections were performed as previously described ( 19). Briefly, 120 h postinfection, the target cells were collected, RNA isolated, and cDNA synthesized for RT-PCR confirmation of overexpression.

Growth curve. Five thousand cells were plated in triplicates in a 6-well plate for a 5-d growth curve analysis. For each count, 500 μL of trypsin was added to each well for 3 min, after which the cells were suspended in 2 mL of PBS, and 500 μL was counted using a Coulter counter.

Colony formation assay. One thousand cells were plated in triplicates in 60-cm plates. After 10 d, the medium were removed and the cells were washed with PBS and fixed with 2:1 methanol:acetic acid for 15 min. Staining was performed using 0.1% crystal violet PBS for 30 min shaking at room temperature. Stain was removed, cells were washed with water, and colonies were counted.

Statistical analysis. The statistical significance of the results was calculated by unpaired Student's t test, and P < 0.05 was considered to be statistically significant.

Results

RLGS analysis in HNSCC patient samples. Global methylation analysis via RLGS has previously been performed on 42 HNSCC patient samples ( 3, 4). However, these critiques only focused on general changes in methylation between primary and metastatic tumors and demethylation of potential oncogenes. Therefore, these profiles were reevaluated to identify the specific genes (potential tumor suppressors) that are frequently methylated in HNSCC. This analysis yielded 23 different loci among the profiles that had decreased RLGS fragment intensity in >10% of the profiles [examples shown in Fig. 1A and complete list in Table 1 ; the 10% cutoff provides a significant difference (P < 0.05)]. Eleven of these fragments have been previously cloned, eight of which were found to be associated with a CpG island and a known gene: ALX3, HMX2, SLC5A8, SEPT9, FUSSEL18, EBF3, IRX1, and GBX2 ( Table 1). BLAT analysis revealed that 7 of 8 of these genes had CpG islands upstream of their transcription start sites, and SEPT9 contained several within the transcribed region. This analysis revealed that 69% (29 of 42) of the HNSCC patients had at least one of these 8 genes methylated. Of these, 31% (9 of 29) had shared methylation between 2 genes, and 14% (4 of 29) had shared methylation between 4 or 6 genes ( Table 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Methylation and expression of candidate genes in HNSCC. A, examples of decreased intensity fragments identified by RLGS analysis. Three loci are shown as examples (HMX2, SEPT9, and SLC5A8). The profiles on the right are generated from adjacent normal tissue, whereas the profiles on the left come from HNSCC tumor tissue. Arrows, fragment of interest. Decreased intensity is assessed by comparing with the profile of its adjacent normal. B, MassARRAY broad methylation analysis. The heat maps display the different genes tested for methylation. Each row represents a different sample, each column represents a CpG site or group of CpGs. The top section represents normals (N), the middle section consists of tumor samples (T), and the bottom section is HNSCC cell lines (C). Increasing color intensity represents increasing methylation. C, relative candidate mRNA expression in HNSCC tumors. The five candidates from the methylation summary were subjected to quantitative RT-PCR analysis. Each bar represents the expression of HNSCC tumor or adjacent normal. The tumor expression is normalized to its matched adjacent normal, which is set to a value of 1 (dashed line). Samples in which the expression level in the tumor is significantly (P < 0.05) reduced are marked (*).

Confirmation of DNA methylation by high-throughput MassARRAY DNA methylation analysis. The MassARRAY technique was used to attain a broad and quantitative DNA methylation analysis of the gene promoters. Primers were designed near or upstream of the transcription start site and/or within the associated CpG island. The genes found to be methylated by a preliminary methylation screen were selected for the MassARRAY (i.e., SEPT9, FUSSEL18, SLC5A8, EBF3, and IRX1). The MassARRAY analysis revealed 56% (20 of 36) of the tumor samples with SEPT9 methylation, 49% (17 of 35) with methylation for FUSSEL18, 32% (13 of 41) SLC5A8 methylation, 50% (28 of 56) EBF3 methylation in tumors, and 45% (21 of 47) showed IRX1 methylation ( Fig. 1B). These data were further validated in a separate set of 20 HNSCC samples and 18 samples of normal mucosa (Supplementary Fig. S1). Of the tumor samples found to be methylated by the MassARRAY analysis, >50% were methylated in at least 2 of the candidate genes. Also, 15% of the samples contained shared methylation between 3 of the candidate genes (most commonly between IRX1, SLC5A8, and EBF3; Table 2 ).

Messenger RNA expression of the methylated candidates is reduced in HNSCC tumor samples. To investigate whether methylation may provide decreased expression, quantitative RT-PCR analysis was performed on the top five genes with the highest methylation incidence (SEPT9, FUSSEL18, SLC5A8, EBF3, and IRX1; Fig. 1B). By normalizing expression to the adjacent normals, IRX1 showed decreased expression in 80% (12 of 15) of the tumors (P < 0.0353), SLC5A8 showed decreased expression in 77% (10 of 13) of tumor samples (P < 0.0400), EBF3 showed 46% (6 of 13) down-regulation (P < 0.0360), FUSSEL18 showed reduced expression in 44% (4 of 9) of tumors (P < 0.0160), and SEPT9 revealed decreased expression in 25% (3 of 12) of the tumor samples (P < 0.0430; Fig. 1C). In addition, we tested the correlation of DNA methylation and gene expression for EBF3, FUSSEL18, IRX1, and SLC5A8 (Supplementary Table S1). Promoter methylation events resulted in gene down-regulation in 25 tumor samples, whereas no change in gene expression was detected in five samples that showed a methylated promoter.

5-aza-2′-dC treatment restores expression. Several HNSCC cell lines were treated with 5-aza-dC, and mRNA expression was compared between treated and untreated cell lines to analyze the capacity for demethylation to restore gene expression. RT-PCR was performed on the 4 genes that had originally shown decreased expression in >25% of the HNSCC tumor samples tested ( Fig. 1C). This analysis revealed a significant restoration of mRNA expression in 4 of 5 cell lines for SLC5A8, 3 of 5 cell lines for FUSSEL18, and 2 of 5 cell lines for IRX1 and EBF3 ( Fig. 2A ; *, P < 0.050).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Demethylation increases candidate mRNA expression and candidate gene overexpression provides tumor suppressor activity in HNSCC cell lines. A, relative fold increase in candidate mRNA expression in HNSCC cell lines after demethylation treatment. cDNA was synthesized from HNSCC cell lines with and without 1 μmol/L 5-aza-dC for 96 h. Quantitative RT-PCR was performed on the 4 candidates that showed >25% of the samples with decreased mRNA expression in HNSCC tumors compared with normals. Each bar represents the expression of the 5-aza-dC–treated HNSCC cell lines nomalized to the untreated HNSCC cell line, which is set to a value of 1. *, P < 0.050. B, confirmed mRNA overexpression of EBF3, IRX1, and SLC5A8 in an HNSCC cell line SCC22B. cDNA was synthesized from HNSCC cell line SCC22B with and without the overexpression plasmids (EBF3-pBABE, IRX1-pBABE, and SLC5A8-pBABE). Quantitative RT-PCR was performed to validate overexpression. −, wild-type (WT) SCC22B; +, SCC22B infected with that overexpression vector. C, decreased colony formation with candidate gene overexpression (top). Colony formation assay was performed with wild-type SCC22B and SCC22B cells overexpressing EBF3, IRX1, and SLC5A8. One thousand cells were plated, and 10 d later, the medium was removed and colonies were stained. The graph represents the number of colonies. *, P < 0.007. Bottom, decreased cell numbers with candidate gene overexpression. Growth curve analysis was performed with wild-type SCC22B and SCC22B cells overexpressing EBF3, IRX1, and SLC5A8. Five thousand cells were plated, and each day, the plates were trypsinized and cells counted. *, P < 0.013. D, decreased Survivin expression in the presence of SLC5A8 overexpression. Quantitative RT-PCR was performed to check Survivin expression. *, P < 0.015.

SLC5A8, IRX1, and EBF3 tumor suppressor activity. To investigate the potential tumor suppressor activity of the differentially methylated genes, the three most frequently down-regulated candidates (EBF3, IRX1, and SLC5A8) were overexpressed in an HNSCC cell line. After validating overexpression ( Fig. 2B), colony formation assays and growth curves were performed. The colony formation assay revealed a significant decrease in colonies in the presence of EBF3, IRX1, and SLC5A8 overexpression (*, P < 0.007; Fig. 2C, top). Also, growth curve analysis revealed a significant reduction in cell number compared with the wild-type SCC22B cells (*, P < 0.013; Fig. 2C, bottom). We also found that increased SLC5A8 expression correlated with decreased SURVIVIN expression [an SLC5A8 target involved in apoptosis resistance ( Fig. 2D; Supplementary Fig. S2)]. Therefore, SLC5A8, EBF3, and IRX1 overexpression seems to provide tumor suppressor activity in HNSCC.

Discussion

HNSCC is an aggressive cancer and exhibits poor survival outcome for patients in advanced stages ( 20). Early detection markers would provide great promise for timely diagnosis and treatment, but very few markers are currently known. Furthermore, no methylation markers for HNSCC have been clinically developed.

This study revealed a group of candidate tumor suppressor genes that are frequently methylated and down-regulated in HNSCC, and demethylation was able to restore gene expression in HNSCC cell lines. SLC5A8, IRX1, and EBF3 have previously been shown to act as tumor suppressors in several cancer types ( 5– 7), and overexpression of these candidates in HNSCC also revealed tumor suppressor activity. Other frequently methylated candidates elucidated in this study (FUSSEL18 and SEPT9) have also shown tumor suppressor function in other malignancies ( 8, 9). However, fairly little is known about most of the candidate genes selected in this study, with the exception of SLC5A8.

The observation of comethylation of certain candidate genes at distant loci is a significant finding. This not only reveals the possibility for using comethylation for improved diagnosis but also suggests that the candidates described in this study may be involved in a similar pathway that is disrupted and epigenetically silenced in HNSCC. TGF-β signaling has been found to be disrupted in HNSCC progression; therefore, this pathway has been targeted for therapy ( 14). Aberrant expression of hypoxia-induced proteins has also been shown to have relevance for HNSCC progression and prognosis prediction ( 15). Furthermore, evidence supports the synergistic cooperation between the hypoxia and TGF-β signaling pathways ( 16). Interestingly, the candidates elucidated in our study may be involved in these pathways (Supplementary Fig. S2). FUSSEL18 has been shown to bind Smad 2 and 3 and inhibit TGF-β signaling ( 11). Also, SEPT9 interacts with HIF-1α to prevent its ubiquitination and degradation, which can allow TGF-β feedback activation ( 10). IRX-1 is a downstream target of Smad 2–mediated transcriptional activation ( 12), and SLC5A8 down-regulates SURVIVIN (target of HIF-1α) to induce apoptosis ( 5). EBF3 has been shown to activate p21 (which is a target of HIF-1α and Smad3), and an EBF–associated zinc-finger protein has been found to bind with Smad proteins ( 13). Therefore, epigenetic down-regulation of these candidate genes may be critical to disrupt their roles in TGF-β signaling, allowing uncontrolled proliferation and apoptosis resistance for HNSCC progression.

The research discussed in this study may lend the opportunity for improved HNSCC screening and detection. However, more comprehensive studies with large patient sets that incorporate stage and pathology should be performed to confirm whether these markers have prognostic implications. Methylation of the genes described in this study may occur at an early stage, providing great promise for earlier HNSCC detection. Therefore, the methylated candidates revealed in this study deserve further attention to discern their potential worth for HNSCC diagnosis and prognosis prediction.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.