Changing Patterns, and Clinical Correlates, of Mutations in Patients with Myelofibrosis over Treatment with Ruxolitinib or Hydroxyurea

Background. About half of patients (pts) with primary (PMF) and post-polycythemia vera/essential thrombocythemia MF (PPV/PET-MF) lose clinical response over prolonged treatment (3y+) with ruxolitinib. Both driver (JAK2, MPL, CALR) and non-driver mutations lack predictive power for the short-term (1y) response (Guglielmelli P, Blood 2015) although pts with >3 mutations are less likely to respond to ruxolitinib (Patel K, Blood 2016). Despite some reductions of Variant Allele Frequency (VAF) of driver mutations, molecular remissions induced by ruxolitinib are exceptional.

Aims. To describe modifications of VAF of driver and non-driver mutations and changing patterns of mutation profile, and to correlate with response, in pts with MF receiving ruxolitinib or hydroxyurea.

Methods. Sixty-three pts (38 PMF, 25 PPV/PET-MF) in active follow-up (FU) at CRIMM, Florence, were included; 17pts (15 PMF, 2 PPV/PET-MF) received hydroxyurea (HU) and 46 (23 PMF, 23 PPV/PET-MF) ruxolitinib. All pts were annotated for driver mutations and an additional panel of 24 "myeloid" genes by NGS. Paired samples of blood granulocytes were analyzed before starting HU/ruxolitinib therapy (baseline) and at the last FU available while on therapy. For the analysis, we recorded any modification of VAF, either decreasing or increasing, of at least 20% compared to baseline. Mutations were defined "acquired" in case of de novo detection (>2% VAF) and "loss" in case of disappearance of pre-exiting mutation at baseline.

Results. The median follow-up from start of therapy was 2.9y for HU-pts and 3.4y Ruxo-pts. At last FU, among Ruxo-pts, 11 (23.9%%) had discontinued for failure, 1 progressed to AML and 16 (34.8%) died; in the HU group, all pts were still on treatment. JAK2 V617F mutation was found in 49 pts (77.8%; 22% HU, 76% Ruxo), CALR mut in 10 (15.9%; 40% HU, 60% ruxo), MPL mut in 1pt in each group; 3 pts were triple negative (2 HU, 1 Ruxo). The most frequent non-driver mutations were ASXL1 (33%), TET2 (20%), NFE2 (15%), ZRSR2 (12%), SRSF2 (8%).

HU-pts: During treatment period, JAK2 V617F VAF increased in 2 pts (18%; +43% and +60% vs baseline) and decreased in 1 pt (-63%). No VAF changes were observed in CALR mut pts while MPL VAF increased by 47%. As regards non-driver mutation, VAF increased in 4 pts (31%; median +112%, range +22% to +168%) and remained stable in 13; acquisition of new mutation occurred in 5 pts (29%; 1 each for TET2, TP53 and NRAS and 2 CBL); no pts displayed mutation loss.

Ruxo-pts: During treatment period, JAK2 V617F VAF increased in 6 pts (15%; median +27%, range +20% to +50%) and decreased in 9 (23%; median -38%, -21 to -62%). CALR VAF increased of 32% in 1 pt and was unchanged in 5 pts. In the MPL mut pt VAF decreased from 37% to 14%. As regards non-driver mutation, VAF increased in 26% of cases (median +103%) and decreased in 9 pts (23%; median -31%); in 3 pts, loss of 4 mutations was observed. Acquisition of new mutation was observed in 8 pts (17%; 1 each for EZH2, SF3B1, ASXL1, PTPN11, 2 NRAS and 3 KRAS).

Response of symptoms or splenomegaly by the IWG-MRT criteria (Tefferi et al, Blood 2013) was achieved in 86% and 57% of pts after a median of 2.9mo and 13.1mo; of these, 13 (34%) and 12 (48%) pts lost clinical response after a median of 3.3y for symptoms and 2.8y for spleen volume reduction. ASXL1 mut at baseline was associated with a significantly lower likelihood of IWG-MRT-defined spleen response over prolonged treatment (32% vs 63% in the absence of ASXL1 mut; P=.040) . The probability of maintaining a spleen response at 2y in ASXL1 mut pts was 33% compared with 87% of unmutated pts. No mutation individually predicted for symptoms response or therapy-discontinuation. Acquisition of at least one mutation was associated with higher rate of discontinuation (36% vs 11% in those pts without clonal evolution, P=.048). Acquisition of 1 and >2 mutations correlated with higher rate of death (63% vs 29% and 45% vs 7%, respectively, compared to pts without clonal evolution).

Conclusions. VAF changes of driver and non-driver mutations, as well as loss/acquisition of new non-driver mutations, occurred at comparable rate and with similar patterns in HU- and ruxolitinib-treated pts. Acquisition of new mutations while on ruxolitinib correlated with higher rate of discontinuation and death.