Abstract
Patients with recurrent glioblastoma achieving response to bevacizumab combined with chemotherapy have clinical improvement and prolonged survival. High gene expression of angiotensinogen (AGT) is associated with a poor bevacizumab response. Because AGT expression is epigenetically regulated, we aimed to investigate whether AGT promoter methylation in tumor tissue predicts response to bevacizumab combination therapy in patients with recurrent glioblastoma. The study included 159 patients with recurrent glioblastoma, treated with bevacizumab combination treatment (training cohort, n = 77; validation cohort, n = 82). All patients could be evaluated for treatment response and biomarkers. DNA methylation of 4 CpG sites in the AGT promoter was measured using pyrosequencing. A model for nonresponse was established using logistic regression analysis. In the training cohort, lower methylation of each of the four CpG sites in the AGT promoter was significantly associated with nonresponse (all P < 0.05). Moreover, the mean methylation level of all four CpG sites was associated with an increased likelihood of not achieving response to bevacizumab combination therapy (twofold decrease: odds ratio = 3.01; 95% confidence interval: 1.41-6.44; P = 0.004). We developed a model for nonresponse in the training cohort, where a threshold of mean AGT promoter methylation levels was set to below 12%. The model could predict bevacizumab nonresponse with 96% specificity. Importantly, this predictor was also significantly associated with nonresponse in the validation cohort (P = 0.037). Taken together, our findings suggest that low AGT promoter methylation in tumor tissue predicts nonresponse to bevacizumab combination treatment in patients with recurrent glioblastoma. We have, thus, established and successfully validated a predictor for nonresponse that can be used to identify patients who will not benefit from bevacizumab combination therapy.