Presence of 4 or More Driver Mutations Predicts Poor Response to Hypomethylating Agent (HMA) Therapy and Poor Overall Survival in MDS

Background: Recent advance in massively parallel sequencing technology has enabled identification of a number of novel somatic driver mutations in myelodysplastic syndromes (MDS). While these efforts have clearly advanced our understanding of MDS pathogenesis, clinical implication of driver mutations in MDS is less studied.

Methods: We performed whole exome sequencing on bone marrow aspiration samples obtained from 114 consecutive patients with untreated MDS. Exome capture was performed using Agilent's SureSelect V4 and sequencing was conducted using Illumina's HighSeq 2000 platform. Sequencing achieved median 124x coverage for the targeted exons. Mutect and Pindel algorithm were used to call single nucleotide variants (SNV) and small indels against virtual common normal reference. Annotation of high-confidence driver mutations followed the previous publication by Pappaemmanuil et al. (Blood 2013). Clonal heterogeneity of driver mutations was assessed in patients who have 2 or more driver mutations by Pearson's goodness of fit test.

Results: Among the 114 patients with MDS, total 221 high-confidence driver mutations were detected in 39 genes by sequencing. Eighty eight percent (100/114) of the patients were found to have at least one driver mutation. The number of driver mutation ranged from 1 to 5 per case. Commonly detected driver mutations include TET2 (25%), SRSF2 (22%), ASXL1 (20%), RUNX1 (19%), TP53 (12%), SF3B1 (9%), and U2AF1 (9%). As a rare driver mutation, we confirmed ETNK1 p. N244S mutation in 2 MDS patients (2%). This mutation was recently described as a recurrent somatic mutation in atypical CML (Gambacorti-Passerini et al. Blood 2015). Clonal heterogeneity of driver mutations was evaluable in 65 patients (57%). Nineteen patients were found to have clonal heterogeneity in driver mutations (29%).

Among the 114 patients, 61 patients (54%) were treated with HMA therapy. Complete response (CR), partial response (PR), and hematological improvement (HI) was observed in 22 (36%) patients, 4 (7%) patients, and 5 (8%) patients, respectively. Presence of TET2 mutation did not predict response to HMA therapy in this series (P = 0.57) even when we restricted to TET2 mutations with variant allele frequency (VAF) >10%. There was a trend toward poor response to HMA therapy in ASXL1 mutated patients (P = 0.074). None of the other driver mutations were predictive of response to HMA therapy as a sole. However, patients who were found to carry 4 or more driver mutations had significantly poor response to HMA therapy (CR rate 0%) compared to patients with less than 4 driver mutations (P = 0.035). Presence of clonal heterogeneity in driver mutations was not predictive of response to HMA therapy (P = 0.43)

In regards to survival outcome, presence of SF3B1 mutation predicted favorable overall survival (OS, P = 0.02) while TP53, and DNMT3A mutations were associated with worse OS (P < 0.001 and P = 0.02, respectively). Presence of clonal heterogeneity in driver mutations was not prognostic for OS (P = 0.71). Patients who were found to have 4 or more driver mutations were associated with significantly worse OS (P = 0.014). None of the patients with 4 or more driver mutations had SF3B1 mutation. Multivariate Cox proportional hazard regression analysis considering dichotomized variables relevant to IPSS-R classification (absolute neutrophil count < 0.8 x 10³ / µ l, hemoglobin < 8 g/dL, platelet count < 50 x 10³ / µ l, and bone marrow blast > 10%, and poor or very poor risk cytogenetics), SF3B1 mutation, DNMT3A mutation, TP53 mutation, and the number of driver mutations (≥ 4) revealed that the presence of 4 or more driver mutations (HR = 2.72 95% CI: 1.34-5.53, P = 0.06), platelet count < 50 x 10³ / µ l (HR = 4.73, 95% CI: 2.53-8.85, P < 0.001), and TP53 mutation (HR = 3.34, 95% CI: 1.65-6.75, P = 0.001) significantly predicted worse OS.

Conclusion: With the modern sequencing technology, approximately 90% of MDS patients were found to have at least one known myeloid driver mutation. Presence of 4 or more driver mutations in MDS patients predicted poor response to HMA therapy. Multivariate model incorporating mutation profile showed that the presence of 4 or more driver mutations and TP53 mutation status were significantly prognostic in MDS independent of IPSS-R variables. Screening for driver mutations in MDS has clinical impact and mutation profiles should be incorporated into the existing prognostic model.