Molecular Predictors of Outcome in Patients with MDS and AML Following MDS after Allogeneic Hematopoietic Stem Cell Transplantation

Introduction: The landscape of molecular aberrations in patients with myelodysplastic syndromes (MDS) has been well characterized and has identified ASXL1, BCOR, CUX1, IDH1, IDH2, SRSF2, RUNX1, U2AF1, TP53 and others as negative prognostic markers for overall survival (OS). We comprehensively investigated the prognostic impact of genetic aberrations in the context of allogeneic hematopoietic stem cell transplantation (alloHSCT) in a large cohort of patients with high risk MDS or secondary AML following MDS (sAML).

Patients and Methods: 308 Patients with a diagnosis of MDS (47.4%) or sAML (52.6%) who received an alloHSCT at four German university medical centers and for whom genomic DNA was available at a time with active disease before transplantation were evaluated for the presence of mutations in 54 genes by Illumina high-throughput sequencing.

Results: At least one mutation was identified in 82% of our patients with a median number of 2 mutations per patient. The most frequently mutated genes were ASXL1 (24.4%), DNMT3A, (23.1%), RUNX1 (16.9%), TET2 (16.9%), STAG2 (12%), TP53 (11.7%), and SRSF2 (11%) in agreement with previous reports. Mutation frequencies were similar between MDS and sAML patients for all mutated genes except SRSF2, TET2 and WT1, which were more frequently mutated in sAML. We grouped the mutated genes into functional classes and found that patients most frequently had mutations in modifiers of DNA methylation (42.2%), followed by chromatin modifiers (41.5%), splicing genes (31.1%), transcription factors (30.8%), signal transducers (28.8%) and tumor suppressors (14.6%). Mean variant allele frequencies were highest in modifiers of DNA methylation (33.2%), while signal transducers had the lowest allele frequency (20.4%).

We next assessed the prognostic impact of gene classes and individual genes on outcome of patients after alloHSCT. Median follow up from transplantation was 4.15 years. Median patient age at time of HSCT was 58 years (range 19-75). 76 patients (25%) werein complete remission and 232 patients (75%) had active disease before transplantation. Low, intermediate, and high risk cytogenetics according to IPSS were found in 116 (38%), 59 (19%), and 115 (37%) patients, respectively. Matched and mismatched related donor HSCT was performed in 71 and 4 patients, respectively (23.1 and 1.3%), and matched and mismatched unrelated donor HSCT in 171 and 62 patients, respectively (55.5 and 20.1%).

The six functional gene classes had no prognostic impact on survival, cumulative incidence of relapse and non-relapse mortality. We therefore evaluated the prognostic impact of individual gene mutations, of aberrations of chromosomes 3, 5, 7, 8, 17, 20 or a complex karyotype, and of transplant characteristics on OS. Parameters with significant impact on OS in univariate analysis were included in a multivariate cox proportional hazards model. Significant predictors of OS in multivariate analysis were mutations in PTPN11 (HR 3.1, present in 2.3% of pts.), IDH2 (HR 2.6, present in 4.2%), PHF6 (HR 2.2, present in 4.9%), NRAS (HR 1.8, present in 7.5%), presence of a complex karyotype (HR 1.8, present in 16.6%), transplantation from haploidentical donor (HR 5.5), RAEB/sAML not in complete remission before transplantation compared to untreated RA/RARS or RAEB/sAML and treated RAEB/sAML in remission (HR 2.0), GvHD prophylaxis other than calcineurin inhibitor with methotrexate or mycophenolate mofetil (HR 1.7) and female sex of the donor (HR 1.7). TP53 mutations lost their unfavorable prognostic impact when complex karyotype was added to the multivariate model. Competing risk analysis for cumulative incidence of relapse and non-relapse mortality showed that IDH2 and NRAS mutations and a complex karyotype were significantly associated with higher risk for relapse while PTPN11 and PHF6 mutations predicted for a higher incidence of non-relapse mortality. Importantly, a negative prognostic impact of ASXL1, BCOR, CUX1, IDH1, SRSF2, RUNX1 and U2AF1 seen previously in MDS patients not undergoing alloHSCT was not found in the transplant setting, suggesting that alloHSCT may overcome the unfavorable effect of these mutations.

Conclusion: By extensive genetic characterisation of 308 MDS or sAML patients undergoing alloHSCT we identified mutations in IDH2, NRAS and complex karyotype as predictors of relapse and reduced OS and provide a matrix to refine risk prediction for allogeneic HSCT.