How ruxolitinib impacts on driver and additive mutations in patients with myelofibrosis

Background: Introduction of ruxolitinib in the clinical practice has changed the outcome of patients with myelofibrosis (MF), offering longer survivals and improvement of the quality of life. Nevertheless, about 50% of patients loss clinical response, and some authors ascribed this phenomenon to driver and non-driver mutations: Patel et al. (Blood 2015) reported that having more than 3 mutations well correlated with shorter time to discontinuation and shorter overall survival (OS), but other investigators in the COMFORT-II trial reported that non-driver mutations did not correlate with response or survival (Guglielmelli, Blood 2014). Aims: in order to further investigate if ruxolitinib could play any role in changing the mutational landscape in MF patients, we assessed the 3 driver and 8 non-driver mutations in 36 MF patients at diagnosis; 19 were tested also after 12 months of ruxolitinib, and compared with 4 cases receiving hydroxyurea. Methods: JAK2, CALR, and MPL mutations were screened by qualitative/quantitative PCR. For the non-driver mutations, we designed a PCR plate with pre-spotted primers able to amplify ASXL1, EZH2, DNMT3A, IDH1, IDH2, SRSF2, TET2, TP53, for total 38 hot-spot sites (Custom qBiomarker Somatic Mutation PCR Array® - Qiagen, Italy). These genes were chosen because already included in the high molecular risk subgroup (ASXL1, EZH2, SRSF2, IDH1/2) or for their prognostic negative role in myeloid hematological neoplasias (DNMT3A, TP53). Results: JAK2 was mutated in 70% of cases, CALR in 20%, whereas 10% were triple-negative. The median OS was significantly longer for primary MF (160 months) vs post-ET (80 months) or post-PV MF (35 months)(p=0.03), and for CALR- vs JAK2-mutated patients. At the last follow-up, 4 patients (11%) progressed to AML, and 12 (33%) died. The non-driver mutations were found at diagnosis in 33% of cases receiving ruxolitinib and in one/4 patients treated with hydroxyurea. Considering both driver and non-driver mutations, 24 cases (67%) were mutated, with 16 carrying one mutation, and 10 two mutations. The most frequently detected mutations belonged to the methylation pathway (DNMT3A, IDH, TET2=75%), followed by TP53 (17%), SRSF2 (8%), ASXL1 (8%), and EZH2 (8%). During treatment, JAK2 VAF remained stable, whereas non-driver mutations changed in 13 cases: 9 acquired a new mutation, while 4 lost the previously detected mutations. Acquisition of DNMT3A mutation was found in 5 patients, of IDH2 in one, and of TP53 in another one. None of the CALR-mutated cases carried non-driver mutations. In the 4 cases treated with hydroxyurea, during treatment one acquired TP53 and another one DNMT3A mutation. On the other hand, 4 cases lost mutations previously present at diagnosis (TP53, IDH2, ASXL1, DNMT3A) in the group of ruxolitinib, but nobody in the group of hydroxyurea. Presence/absence of non-driver mutations, their number (>1), the molecular subgroup (methylation, splicing, chromatin) did not significantly condition OS. Summary/Conclusion: In this work, even if on a small series of patients, we showed that during ruxolitinib about the half of cases develops non-driver mutations, a percentage overlapping to that observed in cases receiving hydroxyurea. Interestingly, ruxolitinib allowed disappearance of mutations in one third of cases. Acquisition of new mutations or the type of non-driver mutations did not correlate with higher rate of death or ruxolitinib failure.