Association of TNFRSF10D DNA-Methylation with the Survival of Melanoma Patients

In this retrospective pilot study, the DNA-methylation status of genes that have been demonstrated to be involved in melanoma carcinogenesis was analyzed in order to identify novel biomarkers for the risk assessment of melanoma patients. We analyzed DNA extracted from punch-biopsies from 68 formalin-fixed paraffin-embedded (FFPE) melanoma specimens. Using MethyLight PCR, we examined 20 genes in specimens from a training set comprising 36 melanoma patients. Selected candidate genes were validated in a test set using FFPE tissue samples from 32 melanoma patients. First, we identified the TNFRSF10D DNA-methylation status (TNFRSF10D methylated vs. unmethylated) as a prognostic marker for overall (p = 0.001) and for relapse-free survival (p = 0.008) in the training set. This finding was confirmed in the independent test set (n = 32; overall survival p = 0.041; relapse-free survival p = 0.012). In a multivariate Cox-regression analysis including all patients, the TNFRSF10D DNA-methylation status remained as the most significant prognostic parameter for overall and relapse-free survival (relative-risk (RR) of death, 4.6 (95% CI: 2.0–11.0; p < 0.001), RR of relapse, 7.2 (95% CI: 2.8–18.3; p < 0.001)). In this study, we demonstrate that TNFRSF10D DNA-methylation analysis of a small tissue-punch from archival FFPE melanoma tissue is a promising approach to provide prognostic information in patients with melanoma.


Introduction
The incidence of melanoma is increasing in white populations worldwide [1]. In advanced disease, median survival is very poor. Melanomas account for 90% of the deaths associated with cutaneous neoplasms [1]. The five-year risk of death of patients with advanced (Stage IV) disease is about 80% [2]. Thus, the recognition of patients at risk for progression is crucial. The most relevant prognostic factors for primary melanoma without metastases are vertical tumor thickness (Breslow's depth) and the presence or absence of histological ulceration; to a lesser extent, also mitotic activity and invasion level (Clark's level). However, clinical experience demonstrates that some patients with thin neoplasms face recurrence, metastases and death after surgical excision, while others with thick melanomas do not. New prognostic markers, such as metallothionines or genetic subtypes defined by gene expression profiling, have been established; however, additional reliable markers to select patients for early therapy are still lacking [3,4]. Besides genetic alterations, changes in the status of DNA-methylation, a type of epigenetic alteration, are among the most common molecular alterations in human neoplasia, including melanoma [5][6][7][8][9]. Irreversible silencing of certain genes by DNA-methylation might enable cells to acquire new capabilities that may drive tumorigenesis. Here, we analyzed the DNA-methylation status of the 5' regions of 20 different genes that are involved in melanoma carcinogenesis and that have previously been shown to be aberrantly methylated in melanoma [7,[9][10][11][12][13][14][15][16][17] or other human cancers [18].
In the present pilot study, we aimed to explore whether differences in the DNA-methylation pattern from formalin-fixed paraffin-embedded (FFPE) tissues may be useful as a prognostic marker for the further outcome of non-metastatic melanoma patients.

Test Set for the Validation of Relevant Genes
For validation of the results obtained with the training set, we analyzed the TNFRSF10D DNA-methylation status in an independent test set consisting of FFPE tissues of 32 melanoma patients. Furthermore, in this analysis, TNFRSF10D DNA-methylation was confirmatively shown to be highly significantly associated with a poor overall and relapse-free survival (p = 0.041 and p = 0.012, respectively; Table 1B).

Overall Prognostic Significance Merging the Training and the Test Set
The comprehensive univariate analysis of all 68 patients together identified age (<60 vs. ≥60), the Clark level, mitotic rate and TNFRSF10D DNA-methylation status as prognostic parameters for poor overall survival (p < 0.001, <0.001, 0.005, <0.001; Table 1C), as well as for poor relapse-free survival (p = 0.013, 0.011, 0.012, <0.001; Table 1C). The Kaplan-Meier survival analysis for TNFRSF10D DNA-methylation is depicted in Figure 1. Finally, we analyzed all patients, including the variables, sex, age, Clark level, presence of ulceration, tumor thickness, interferon alpha therapy and TNFRSF10D DNA-methylation status, using a multivariate Cox regression model. The TNFRSF10D DNA-methylation status remained the most significant prognostic parameter for overall and relapse-free survival. Patients with methylated TNFRSF10D showed a 4.6-fold higher risk of death (95% CI: 2.0-11.0; p < 0.001) and a 7.2-fold higher risk of relapse (95% CI: 2.8-18.3; p < 0.001) than patients with unmethylated TNFRSF10D (Table 2). After exclusion of a subgroup of 16 patients, who received an adjuvant interferon alpha therapy, TNFRSF10D DNA-methylation still remained the highest significant prognostic parameter in the multivariate analysis for overall (p = 0.004) and relapse-free survival (p = 0.001).

Discussion
The prognostic biomarkers currently used in melanoma do not adequately predict disease recurrence and overall survival in a significant subset of patients. Therefore, novel biomarkers are highly required to overcome these problems.
In this pilot study, we identified TNFRSF10D DNA methylation status in paraffin-embedded melanoma tissues to be an independent prognostic biomarker for relapse-free survival and overall mortality in non-metastatic melanoma patients. Surprisingly, ulceration and tumor thickness were significantly associated with survival only in the training set, whereas the invasion level (Clark's level) was a significant, prognostic feature consistently in all analyses performed in this work. Due to the small sample size of this pilot study, a subsequent validation study in a larger, independent patient cohort must be performed. In our study, the prognostic value of TNFRSF10D concerns mainly the group of patients with methylated TNFRSF10D in the tumors. Recently, TNFRSF10D DNA-methylation has also been shown to be an independent prognostic and predictive marker in the serum of patients with MYCN nonamplified neuroblastoma [19]. Interestingly, recently published data indicated that TNFRSF10D is epigenetically silenced in human melanoma [9,16], as well as in cancers of breast, lung, mesothelioma, prostate, bladder, cervix, ovary, brain and in hematopoietic malignancies [20]. Bonazzi et al. found that 72% of the analyzed melanoma cell lines had no TNFRSF10D mRNA expression and that the transcript level of TNFRSF10D was correlated inversely with promoter methylation [16]. They identified that 66% of the analyzed cell lines and 30% of the analyzed fresh frozen melanoma samples showed a high degree of methylation. In our study, we identified TNFRSF10D DNA-methylation in only 16% of all analyzed specimens. This discrepancy probably reflects the different levels of TNFRSF10D DNA-methylation in melanoma cell lines, metastatic tumor tissues and primary tissues. Bonazzi et al. observed a five-fold average increase in TNFRSF10D mRNA expression in five melanoma cell lines after treatment with the DNA-demethylating agent, 5-aza-2'-deoxycytidine [16]. TNFRSF10D belongs to the tumor necrosis factor (TNF) receptor superfamily. This receptor contains an extracellular tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) binding domain, a transmembrane domain and a truncated cytoplasmic death domain. The second known TRAIL decoy receptor, TNFRSF10C, lacks this intracellular death domain completely. Therefore, both receptors appear unable to induce apoptosis. Considering that TNFRSF10D, as well as TNFRSF10C have been presumed to function as oncogenes, because of their postulated anti-apoptotic effect [20], our data seem to be controversial at first sight. However, recently, Venza et al. have shown that an ectopic overexpression of TNFRSF10C and/or TNFRSF10D in melanoma cell lines led to a significant reduced growth rate and a clear increased apoptotic response [21]. In the context of our data, it seems that the methylation and silencing of TNFRSF10D may represent a special feature of more aggressive cancer cells.

The Patient Study Cohort and Study Design
We retrospectively analyzed prospectively collected melanoma specimens (FFPE tissues) from melanoma patients treated at the Department of Dermatology and Venereology, Innsbruck Medical University, Innsbruck, Austria. Samples have been collected during primary surgery. Sixty-eight patients (32 women and 36 men) diagnosed between 1983 and 2004 with primary, invasive, non-metastasized melanoma were included in this study (T1-T4, tumor-node-metastasis (TNM) classification American Joint Committee of Cancer 2001). Patients had no metastases at the time of diagnosis or surgery, respectively. Thirty-four melanomas were located on the trunk, 34 on the limbs. The tumor thickness was 0.5-2 mm (T1/T2), 2.01-4 mm (T3) and >4 mm (T4) in 25, 23 and 18 patients, respectively; tumor thickness was unknown in two patients. Twenty-three patients showed ulcerated melanomas. All patients underwent surgery with 1-2 cm excision margins, according to standard guidelines, and 16 patients received adjuvant interferon alpha. The median age was 53.4 years (21.9-90.7 years). Thirty and 32 patients relapsed/died, respectively, due to the consequences of the melanoma after a median follow up period of 2.0 (interquartile (IQ)-range 6.2) and 3.4 (IQ-range 5.5) years, respectively. The patient study cohort was a priori randomly split into a training and a test set, consisting of 36 and 32 patients, respectively. The study was approved by the local medical research ethics committee (Reference Number UN3856, 26 January 2010) and conducted in accordance with the Declaration of Helsinki. Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) were adhered to where applicable [22,23].

DNA Extraction and Bisulfite Conversion from Formalin-Fixed Paraffin-Embedded (FFPE) Tissues
DNA was isolated from punches gained from FFPE melanoma specimens using the DNeasy Tissue Kit (Qiagen, Hilden, Germany) in order to assure that mainly melanoma tissue was collected.
Sodium bisulfite-modification of genomic DNA (700 ng) was performed using the EZ DNA Methylation-Gold Kit (Zymo Research, Orange, CA, USA), according to the manufacturer's instructions.
To control for the amount of input bisulfite-modified DNA, this value was normalized to the extent of amplification of a COL2A1 DNA sequence lacking CpG dinucleotides. The specificity of the reactions for methylated DNA was confirmed separately using SssI (New England Biolabs, Ipswich, MA, USA)-treated human white blood cell DNA (heavily methylated). The SssI treated DNA was additionally used for the standard curve preparation, which is required for the quantification. The percentage of fully methylated molecules at a specific locus was calculated by dividing the GENE:COL2A1 ratio of a sample by the GENE:COL2A1 ratio of SssI-treated white blood cell DNA and multiplying by 100 (the percentage of fully-methylated reference, PMR). PMR values have been calculated for all analyzed genes. If less than 50% of the samples were methylated for a specific gene (a gene was deemed methylated if the PMR value was >0), we categorized the samples in "unmethylated" and "methylated" and performed the analyses with these dichotomized variables. Only in 3 genes were more than 50% of the samples methylated (ESR1, SFN and MAGE1). For these genes we used the PMR values for the following statistical analysis.

Mitotic Rate
The mitotic rate per square millimeter of tumor tissue was evaluated by counting mitotic figures on hematoxylin and eosin (H&E)-stained tissue sections.

Statistical Analysis
Disease-free and overall survival were calculated from the date of diagnosis to the date of relapse, death or last follow-up. Disease-free and overall survival curves were calculated with the Kaplan-Meier method. Univariate analysis of overall survival according to clinicopathological factors (age, Clark-level, tumor thickness, presence of ulcerations, etc.) or DNA-methylation status was performed using a two-sided log rank test. A multivariate Cox proportional hazards model was applied to estimate the prognostic effect of TNFRSF10D DNA-methylation. A p-value <0.05 was considered statistically significant. SPSS 19.0 was used for all statistical analyses (SPSS Inc., Chicago, IL, USA).

Conclusions
Our data demonstrate that DNA-methylation analysis of the TNFRSF10D promoter from a small tissue punch from archival paraffin-embedded melanoma tissue is able to provide independent prognostic information in order to identify patients with a higher risk for aggressive progress.
Further studies are needed to elucidate how TNFRSF10D promoter hypermethylation is associated with poor prognosis and aggressive proliferation in melanoma. Additionally, further research needs to be conducted to assess if TNFRSF10D hypermethylation in serum samples from melanoma patients could be an indicator of poor prognosis in melanoma, as has been shown in neuroblastoma patients [19].