Background: The DNA repair enzyme O(6)-methylguanine-DNA methyltransferase (MGMT) confers therapeutic resistance to DNA alkylating agents, including temozolomide. It is largely believed that MGMT promoter methylation is associated with down regulation of MGMT transcription and corresponding protein expression, thereby predisposing tumor cells to the toxic effect of temozolomide. Here we rigorously examined this underlying assumption.
Methods: We examined the correlation between MGMT promoter methylation, transcription, and protein expression using The Cancer Genome Atlas (TCGA) glioblastoma database as well as an independent collection of glioblastoma specimens.
Results: In both analyses, we found that MGMT promoter methylation status correlates well with low MGMT mRNA levels (p = 0.04). On the other hand, glioblastomas with unmethylated MGMT promoters exhibited a wide range of MGMT mRNA expression. Intriguingly, the MGMT mRNA levels correlated poorly with MGMT protein levels by Western blotting (R(2) = 0.04, p = 0.34) or by ImmunoHistoChemical (IHC) stain quantitation (R(2) = 0.02, p = 0.50). To exclude the possibility that the poor correlation was due to substandard specimens, we determined the mRNA and protein levels of Colony Stimulating Factor 1 (CSF1), a gene previously shown to exhibit excellent mRNA/protein correlation. In contrast to MGMT, the mRNA level of CSF1 correlated well with the protein level (R(2) = 0.47, p = 0.001). Importantly, long-term passaged glioblastoma cell lines with comparable MGMT transcript levels differed in MGMT protein levels, suggesting mechanisms of post-transcriptional regulation. Accordingly, the correlation between MGMT promoter methylation and MGMT protein expression was poor (p = 0.27). In silico analysis predicted potential binding sites for several miRNA within the 3'UTR of MGMT, suggesting a mechanism for the post-transcriptional of MGMT.
Conclusion: Our results suggest mechanisms such as miRNA mediated regulation for post-transcriptional regulation of MGMT. Identification of these mechanisms should enhance the value of MGMT based prognostic or predictive biomarker strategies.