Prognostic Impact of ASXL1 Mutations in MDS and CMML

Background Molecular profiling through next-generation sequencing (NGS) in myeloid diseases has greatly improved prognostic discrimination. In myelodysplastic syndromes (MDS), mutations of TP53, EZH2, ETV6, RUNX1 and ASXL1 were shown in a multivariate model to predict for inferior survival although ASXL1 mutations were least predictive (HR 1.38, P=.049; Bejar et al., 2011). In contrast, mutation analysis of two large chronic myelomonocytic leukemia (CMML) cohorts both identified ASXL1 mutation as the only mutation to be independently predictive of inferior survival in multivariable analysis when including only nonsense and frameshift mutations (Itzykson et al., 2013; Patnaik et al., 2014). Therefore, our goal was to investigate the relative impact of ASXL1 mutations in MDS and CMML.

Patients and Methods NGS profiled MDS and CMML cases were retrospectively identified from the Moffitt Cancer Center MDS database. We genotyped ASXL1 and up to 20 additional genes in our cohort. The lower limit of detection was set at 5% mutant allele reads and the minimum depth of coverage was 500X. Clinical variables and outcomes of MDS patients were characterized at the time of sample procurement. Fisher's exact and t-tests were used for comparative analyses. Kaplan-Meier estimates were used to estimate overall survival and analyzed from the date of mutation identification. Multivariate Cox regression models were created to adjust for clinical characteristics.

Results From May 2013 to October 2014, 248 patients with NGS for somatic mutations were identified in our database. The baseline characteristics of the MDS and CMML cohorts stratified by ASXL1 mutation status are shown in Table 1. In this cohort, 89% of patients (n=220) had at least one pathogenic mutation. ASXL1 frameshift or nonsense mutations occurred in 22% of patients (n=54) with increased frequency in the CMML cohort (50% versus 18%, P=0.0002). An additional 7% of patients (n=17) had a missense mutation of ASXL1 and were predominantly restricted to the MDS subgroup (16/17). The most common type of ASXL1 mutation were frameshift mutations in the MDS and CMML cohorts, 51% and 69% respectively, with the c.1934dupG; p.G646Wfs*12 variant observed in 33% of CMML patients (n=5) and 21% of MDS patients (n=8). ASXL1 mutant patients had similar survival to wild-type patients with median OS of 15.3 versus 13.7 months, respectively (P = 0.80). Exclusion of missense mutations had no effect on the prognostic impact of ASXL1 mutation (P=0.78). In the MDS cohort, there was again no difference in OS (median OS 14.3 versus 13.3 months; P=0.47) and type of ASXL1 mutation had no effect (P = 0.85). In the CMML cohort (n=30), median OS was 8.8 months in ASXL1 mutant patients and not reached in wild-type patients (HR 3.0, 95% CI 0.87 to 10.37, P=0.08). In comparison to wild-type patients, class of ASXL1 mutation was predictive of survival in the CMML cohort (P=0.008). CMML patients with frameshift mutations of ASXL1 had shorter OS in comparison to nonsense mutations although this did not reach significance (P=0.21). In multivariable analysis incorporating age and IPSS, ASXL1 mutation status was predictive for inferior survival among CMML patients (HR 2.5, 95% CI 1.08 to 5.85; P=0.03).

Conclusion In MDS patients, neither ASXL1 mutation status nor exclusion of missense mutations was predictive for inferior survival, while ASXL1 mutant CMML patients had inferior OS, particularly with frameshift mutations. Together, these results highlight biological differences between MDS and CMML and support collaborative efforts in mutational profiling of these patients to improve prognostication.