Infrequent promoter methylation of the MGMT gene in liver metastases from uveal melanoma

Abstract

Uveal melanoma is associated with a high mortality rate once metastases occur, with over >90% of metastatic patients dying within less than 1 year from metastases to the liver. The intraarterial hepatic (iah) administration of the alkylating agent fotemustine holds some promise with response rates of 36% and median survival of 15 months. Here, we investigated whether the DNA-repair-protein MGMT may be involved in the variability of response to fotemustine and temozolomide in uveal melanoma. Epigenetic inactivation of MGMT has been demonstrated to be a predictive marker for benefit from alkylating agent therapy in glioblastoma. We found a methylated MGMT promoter in 6% of liver metastases from 34 uveal melanoma patients. The mean MGMT activity measured in liver metastases with negligible liver tissue content was significantly lower than in liver tissue (146 versus 523 fmol/mg protein, p = 0.002). Expression of the MGMT protein was detectable in 50% of 88 metastases by immunohistochemistry on a tissue microarray. Expression was heterogeneous, and in accordance with MGMT activity data, usually lower than in the surrounding liver. Differential MGMT activity/expression between metastasis and liver tissue and more efficient depletion of MGMT with higher doses of alkylating agent therapy using iah delivery may provide the pharmacologic window for the higher response rate. However, these results do not support MGMT methylation status or protein expression as predictive markers for treatment outcome to iah chemotherapy with alkylating agents. © 2008 Wiley-Liss, Inc.

Uveal melanoma is a rare disease with an estimated incidence of 0.6 per 100,000 persons/year. At diagnosis, tumors are often at an advanced stage, with 5-year survival rate of almost 50%.1–3 In contrast to its cutaneous counterpart, melanoma of the eye primarily metastasizes hematogenously, with a particular tropism to the liver, thus displaying distinct biological and clinical features.4, 5 Once liver metastases are diagnosed, patients usually die within less than 1 year, and standard systemic chemotherapy regimens are often ineffective with response rates below 10%. Locoregional intraarterial hepatic (iah) chemotherapy has demonstrated some promise. Recently, we have reported the updated results of our previous phase II trial in a series of over 100 patients demonstrating that the 3rd generation chloroethylnitrosurea fotemustine leads to significant disease control when administered directly into the hepatic artery.6, 7 Response rate was 36%, and even more importantly the median overall survival was 15 months and 29% of the patients were alive at 2 years.

An important role for resistance to alkylating agent therapy has been attributed to the DNA repair enzyme O⁶-methylguanine-DNA methyltransferase (MGMT) (reviewed by Gerson).8 MGMT removes the alkyl groups from the O⁶-position of guanine, an important site of DNA alkylation, thereby blunting the therapeutic effect of DNA-alkylating agents such as fotemustine or temozolomide. Left unrepaired, these chemotherapy-induced lesions trigger cytotoxicity and apoptosis.9 Epigenetic silencing of the gene encoding MGMT has recently been established as a predictive factor for benefit from alkylating agent chemotherapy in patients with glioblastoma.10–12 There is evidence that epigenetic silencing of MGMT by promoter methylation is involved in the development and progression of skin melanoma.13 Although the results and the sample size in the studies reported are generally small and heterogeneous, there seems to be an increase in frequency of MGMT-promoter methylation during tumor progression from 10% in primary melanomas to 30% in metastases.13–17

No data are yet available on MGMT in uveal melanoma in the literature. Herein, we investigated the frequency of the epigenetic silencing of MGMT gene in liver metastases from uveal melanoma patients. We present a comprehensive investigation on the potential role of MGMT in treatment resistance by determining gene promoter methylation, enzyme activity and protein expression in these liver metastases.

Material and methods

Patients and tumor samples

Frozen liver metastasis samples from 34 uveal melanoma patients were available for molecular analysis. Standard hematoxylin–eosin stainings were performed prior to molecular analysis and all samples were reviewed by a pathologist estimating the percentage of normal liver tissue in the sample (A.M.). All patients had undergone resection of liver metastases between 1990 and 2003 and 31 patients were treated with intrahepatic fotemustine,7 including 14 patients who received 2 low doses of temozolomide preceding fotemustine administration aiming at depleting MGMT (Voelter et al., manuscript on combination treatment submitted). Our study has been approved by the local ethics committee.

DNA extraction and methylation-specific PCR

The methylation status of the MGMT gene promoter was determined by methylation-specific PCR (MSP), according to methods reported previously.11 In brief, genomic DNA was isolated from frozen sections of the metastasis (DNeasy Tissue kit, Qiagen 69504) and chemically transformed with sodium-bisulfite. The treated DNA was amplified in a 2-step approach18: The specific primers for the modified methylated (M) and the unmethylated MGMT-promoter (U), respectively, were described previously.10 The results were confirmed in an independent experiment starting at the bisulfite treatment step of genomic DNA.

MGMT activity

Frozen tissues of 19 liver metastases and 6 liver samples were available for the analysis of MGMT activity. The method is based on a radioactive assay that measures the transfer of tritium-labeled methyl groups from the O⁶-position of guanine to the MGMT protein as detailed elsewhere.19

Tissue microarray and immunohistochemistry

The tissue microarrays (TMAs) were constructed using paraffin-embedded, formalin-fixed, nonirradiated primary and metastatic uveal melanomas resected between 1960 and 2005, and collected at the Jules Gonin Eye Hospital and the Institute of Pathology, Centre Hospitalier Universitaire Vaudois in Lausanne. Immunohistochemical determination of MGMT (dilution, 1:50; clone MT3.1; NeoMarkers) was performed using a standard high temperature epitope retrieval method (citrate buffer pH 6.0; pressure cooker, 15 min). The staining was evaluated independently by 2 pathologists (A.M., P.W.) using the following semiquantitative criteria: 0, negative; low index, <10% positive tumor cells; moderate, 10–50%; and high index, >50%.

Statistics

Statistical analyses were carried out in R, a free software environment available at http://www.r-project.org/. The linear mixed effect model fitted to MGMT activity included liver content as fixed factor and patient as random factor (triplicate measurements were available for most patients). The percentage of normal liver content in the samples was treated as a categorical variable: low <5%, high 10–45%, and liver tissue >95%. MGMT activity in the three groups was not normally distributed and was therefore logarithmically transformed before performing any statistical analysis. The linear mixed effect model was fitted using maximum likelihood estimation available within R version 2.5.1, library nlme, version 3.1-83.

Results

Determination of the MGMT promoter status revealed that epigenetic inactivation of MGMT in liver metastases from uveal melanoma is infrequent. A methylated MGMT-promoter was only detected in 2 out of 34 metastases (6%), as illustrated in Figure 1.

Figure 1. Methylation status of the MGMT-promoter in liver metastases. The MGMT status was determined by MSP, depicted here for representative tumor samples. U, Unmethylated MGMT promoter; M, methylated MGMT promoter. The specimen # 2060 contains a methylated MGMT promoter. The peripheral blood lymphocyte (PBL) DNA was used as control for unmethylated and the cell line SW48 served as positive control for methylated MGMT. Water was used as a negative control for the PCR. The size of the PCR products is indicated on the right, L indicates the 100-bp marker.

Given the low frequency of epigenetic inactivation of MGMT, we further investigated whether differences in enzymatic activity of MGMT in these uveal melanoma metastases may account for the variability of response to therapy. Since liver is known for high MGMT activity,20 we estimated the percentage of normal liver content by histology for each sample. A linear mixed-effect model was fitted to estimate the activity in the metastasis in comparison to liver tissue, taking into account the different degree of liver content they comprised. Metastases estimated to have less than 5% normal liver tissue (n = 11) displayed significantly lower MGMT activity than liver tissue (n = 6, p = 0.002), and were also significantly different from samples with high liver content (n = 6, p = 0.018) as visualized in Figure 2. The average activity in the low liver content group (n = 11) was 146 fmol/mg (95% confidence interval [CI]: 96–223), as compared to 523 fmol/mg (95% CI: 252–1,086) in liver tissue (n = 6). The average activity in the high liver content group (n = 8) was 340 fmol/mg (95% CI: 174–663).

Figure 2. MGMT activity in liver metastases as compared to liver tissue. The average of the log2 of replicate measurements for each sample was considered. In the linear mixed model used to evaluate the impact of liver contamination in the metastasis (Met) on overall MGMT activity, both high contamination (grey, >10%) and liver tissue (black, >95%) emerged as significant covariates (p-values 0.018 and 0.002, respectively). The average activity in log2 scale for the low contamination group (white, <5%) is 7.19 (95% CI: 6.58–7.80; corresponding to 146 fmol/mg, 95% CI: 96–223; measurements n = 32), while the average activity in log2 scale for the liver (dark grey, >95% liver tissue) is 9.03 (95% CI: 7.98–10.08, corresponding to 523 fmol/mg, 95% CI: 252–1,086; measurements n = 16). The average activity in the high contamination group (bright grey, >10%) is 8.41 (95% CI: 7.45–9.37, corresponding to 340 fmol/mg, 95% CI: 174–663; measurements n = 23).

To extend the investigations to a larger panel of uveal melanoma metastases the MGMT expression was evaluated on a TMA by immunohistochemistry within a parallel study of the Pathology Institute of Lausanne that takes advantage of a uncommonly important collection of paraffin-embedded tissue samples of this rare disease (Fig. 3). There were 88 interpretable cases, including 77 liver metastases: MGMT was below the limit of detection in 50% of metastases, while in the remainder a high, moderate and low index was observed in 28% (n = 25), 15% (n = 13) and 7% of the cases (n = 6), respectively. For most positive metastases, MGMT staining of the surrounding liver tissue was more intense than in the tumor, in concordance with the observation of the significantly lower MGMT activity measured in metastases as compared to liver tissue. Of note, in patients, for whom more than one metastasis was evaluable, staining patterns could be different, e.g. with negative and positive metastases within the same patient. MGMT expression in primary uveal melanoma samples was detected at a similar frequency as determined on a TMA with 71 interpretable cases. Again, MGMT was not detectable in 48% of the samples (n = 35), while a high, moderate, and low index was found in 32% (n = 23), 15% (n = 11) and 3% of cases (n = 2), respectively.

Figure 3. MGMT expression in liver metastases. Immunohistochemical determination of MGMT expression was performed on tissue microarrays comprising uveal melanoma metastases. High nuclear MGMT expression in all tumor cells is displayed by the tumor in (a). In (b) a MGMT negative tumor is shown infiltrating the liver with endogenous, high MGMT expression. (c) Magnified examples are shown for a strongly positive metastasis; (d) a moderately positive metastasis; (e) a negative metastasis; (f) a negative metastasis with enclosed, highly positive, liver cells; and (g) a negative metastasis with infiltrating lymphocytes exerting high expression of MGMT.

The attempt to correlate the molecular findings with survival was precluded for the MGMT methylation status due to the low frequency of samples with a methylated MGMT promoter (2 of 34), and for the measurements of MGMT activity in the metastasis by the confounding activity of the intermingled liver tissue in almost half of the cases (8 of 20). The association of MGMT protein expression in the metastasis as determined by IHC was not significantly associated with survival (scores 0, 1 vs. 2, 3). The latter is not surprising given the fact that distinct metastasis of a patient may exhibit markedly different expression levels as revealed by the TMA.

Discussion

Here, we aimed at determining the role of the MGMT methylation status and its potential relationship with sensitivity of uveal melanoma metastases to alkylating agent chemotherapy. Epigenetic inactivation of MGMT by promoter methylation was infrequent (6%), suggesting other regulatory mechanisms to be relevant for MGMT expression in this tumor type. This low rate of MGMT methylation contrasts the clinical response rate of ∼40% to iah fotemustine.7 Intriguingly, it corresponds to the low response rate of less than 10% upon systemic delivery of alkylating agents.21 The higher clinical response rate of iah fotemustine versus systemic therapy may suggest that the increased dose intensity obtained by locoregional delivery may sufficiently deplete MGMT in tumors of a subgroup of patients to overcome MGMT-mediated resistance. A relationship between chemotherapy dose and MGMT depletion has been previously demonstrated for temozolomide in various clinical studies.22, 23 Currently, dose-dense regimens with the aim to yield maximal MGMT depletion and thus potentially enhancing antitumor efficacy are investigated within clinical trials in melanoma (EORTC trial 18032) and glioma patients (RTOG-0525/EORTC 26052-22053 trial/NCT 00304031). Furthermore, this interpretation is supported by the finding that both protein expression and MGMT activity were significantly lower in metastases than in surrounding liver tissue, providing a pharmacologic window: increased cytotoxicity on malignant tissue with lower sensitivity of normal tissue and thus limited toxicity. In addition, our experience suggests enhanced clinical efficacy of iah fotemustine when preceded by the alkylating agent temozolomide (Voelter et al., manuscript submitted), suggesting an important role for MGMT-mediated resistance. The highest levels of MGMT activity have been described for normal liver tissue.20 Here, we show that liver metastases from uveal melanoma exhibited an average of 30% of the activity measured in the liver, which is higher than the activity reported from lung and brain tumors with 10–20% of liver activity, respectively.24, 25

Altogether, our data raise the question whether determination of MGMT expression in uveal melanoma metastasis would be helpful for predicting response to alkylating agent therapy. It should be noted, however, that MGMT measurements reported here are baseline values before chemotherapy. This is not taking into account the potential induction of MGMT expression by alkylating agents through activation of DNA-repair pathways and stress response. Furthermore, our data suggest that intra-individual MGMT protein expression between metastases may vary substantially from a negative to a high MGMT index. Such intra-individual heterogeneity has also been reported for metastases from skin melanoma.16 However, we cannot exclude that these differences are based on artifacts due to sampling errors and the small diameter of the tissue cores (2 mm in diameter) on the TMA.

In conclusion, the low frequency of MGMT methylation, the heterogeneity of MGMT expression, together with poorly known aspects such as inducibility of the enzyme during alkylating chemotherapy or enzyme exhaustion, seem to preclude the precise determination of the role of MGMT in treatment resistance to alkylating agent therapy. Our data do not support the use of either, the methylation status of the MGMT promoter or MGMT expression or activity as predictive markers for outcome in uveal melanoma. In glioblastoma, no correlation between a unmethylated MGMT promoter and MGMT protein expression has also been found.26, 27 The ongoing randomized phase III EORTC study of iah versus intravenous fotemustine represents an ideal opportunity to prospectively investigate resistance factors, including MGMT and other cellular repair mechanisms on response and outcome.

Acknowledgements

The authors thank Roger Stupp for his constructive review of the manuscript.