Glycosyltransferases in Cancer: Prognostic Biomarkers of Survival in Patient Cohorts and Impact on Malignancy in Experimental Models

SIMPLE SUMMARY: Cancer-associated glycosylation changes are widely used as biomarkers and strongly impact malignancy. However, the clinical significance of the deranged expression of glycosyltransferases observed in specimens is not always consistent with their role in experimental systems. We analyzed the overall survival curves of patients expressing high or low mRNA levels of 114 glycosyltransferases from the 21 cohorts of The Cancer Genome Atlas (TCGA). We identified 17 glycosyltransferases associated with poor prognosis and 4 associated with good prognosis in a large number of cohorts. In addition, we identified several glycosyltransferases with a very high prognostic value in only one or a few cohorts. Comparisons with published experimental works reveal partial consistency with TCGA clinical data. These data pave the way for the use of glycosyltransferases as prognostic markers and potential therapeutic targets and place experimental studies in an appropriate clinical context. ABSTRACT: Background: Glycosylation changes are a main feature of cancer. Some carbohydrate epitopes and expression levels of glycosyltransferases have been used or proposed as prognostic markers, while many experimental works have investigated the role of glycosyltransferases in malignancy. Using the transcriptomic data of the 21 TCGA cohorts, we correlated the expression level of 114 glycosyltransferases with the overall survival of patients. Methods: Using the Oncolnc website, we determined the Kaplan–Meier survival curves for the patients falling in the 15% upper or lower percentile of mRNA expression of each glycosyltransferase. Results: Seventeen glycosyltransferases involved in initial steps of N- or O-glycosylation and of glycolipid biosynthesis, in chain extension and sialylation were unequivocally associated with bad prognosis in a majority of cohorts. Four glycosyltransferases were associated with good prognosis. Other glycosyltransferases displayed an extremely high predictive value in only one or a few cohorts. The top were GALNT3, ALG6 and B3GNT7, which displayed a p < 1 × 10(−9) in the low-grade glioma (LGG) cohort. Comparison with published experimental data points to ALG3, GALNT2, B4GALNT1, POFUT1, B4GALT5, B3GNT5 and ST3GAL2 as the most consistently malignancy-associated enzymes. Conclusions: We identified several cancer-associated glycosyltransferases as potential prognostic markers and therapeutic targets.