Azacitidine (AZA) for higher risk MDS patients is a standard therapy with limited durability. To monitor mutation dynamics during AZA therapy we utilized massive parallel sequencing of 54 genes previously associated with MDS/AML pathogenesis. Serial sampling before and during AZA therapy of 38 patients (reaching median overall survival 24 months (Mo) with 60% clinical responses) identified 116 somatic pathogenic variants with allele frequency (VAF) exceeding 5%. High accuracy of data was achieved via duplicate libraries from myeloid cells and T-cell controls. We observed that nearly half of the variants were stable while other variants were highly dynamic. Patients with marked decrease of allelic burden upon AZA therapy achieved clinical responses. In contrast, early-progressing patients on AZA displayed minimal changes of the mutation pattern. We modeled the VAF dynamics on AZA and utilized a joint model for the overall survival and response duration. While the presence of certain variants associated with clinical outcomes, such as the mutations of CDKN2A were adverse predictors while KDM6A mutations yield lower risk of dying, the data also indicate that allelic burden volatility represents additional important prognostic variable. In addition, preceding 5q- syndrome represents strong positive predictor of longer overall survival and response duration in high risk MDS patients treated with AZA. In conclusion, variants dynamics detected via serial sampling represents another parameter to consider when evaluating AZA efficacy and predicting outcome.
Keywords: AML; MDS; NGS; azacitidine; somatic mutation.