Mutation Profiles Identify Distinct Clusters of Lower Risk Myelodysplastic Syndromes with Unique Clinical and Biological Features and Clinical Endpoints

Background. The severity of hematopoietic impairment and the kinetics of disease progression in lower risk myelodysplastic syndromes (LR-MDS) are extremely variable. Genomic profiling has the potential to inform the clinical management of these disorders, including improved classification, risk assessment and therapeutic choice. In the present study, based on a comprehensive mutation analysis in a large and clinically well-characterized cohort of LR-MDS patients, either recruited into the European MDS Registry or referred to European excellence centers involved in the MDS-RIGHT project, we adopted unsupervised hierarchical clustering analyses to identify relevant genetically defined disease subtypes within early stage MDS.

Methods. The dataset comprised 856 cases identified as LR-MDS based on IPSS risk low or intermediate-1. Median age was 73 years (range 36-98); IPSS-R risk was very low in 30.1% of patients, low in 50.4%, intermediate in 19.5%. We investigated possible sub-structure amongst patients according to their mutational profiles, and correlated this sub-structure with relevant endpoints. For this analysis, unsupervised clustering was used, based on a mixture model of multivariate Bernoulli distributions. The optimal number of clusters was chosen using the Bayes Information Criterion (BIC), with secondary structure identified with the Akaike Information Criterion (AIC).

Results. This analysis identified three distinct clusters within LR-MDS. Cluster 1 comprised exclusively patients with SF3B1 mutation, either isolated or associated with other mutations (SF3B1-mutant cluster) (37% of patients). Cluster 2 was characterized by excess mutations associated with higher risk disease (high-risk (HR) cluster) (27% of patients), including a significantly higher prevalence of ASXL1, IDH1/IDH2, SRSF2, RUNX1, CBL and EZH2 mutations (P<.001). This cluster also showed a significantly higher number of mutations per patient compared to other groups (P<.001), suggesting a subtending clonal progression resulting in the accumulation of sub-clonal mutations. Finally, cluster 3 was characterized by mutation profiles as observed in Clonal Hematopoiesis of Indeterminate Potential (CHIP) (CHIP-like cluster) (36% of patients), mainly including isolated DNMT3A, TET2 or ASXL1 mutations, pointing toward the contribution of extra-clonal factors to disease expressivity. In addition, this cluster showed enrichment in TP53 mutations, as recently reported in community-dwelling elderly individuals with unexplained anemia (Blood 2020;135:1161-70).

The three recognized clusters showed distinct clinical features and outcome measures. Patients within HR cluster were significantly older (P=.008) and showed significant enrichment in WHO categories with multi-lineage dysplasia and excess blasts (P<.001) and IPSS-R intermediate risk scores (P<.001), as well as significantly lower platelet count (P=.001). Conversely, patients within the CHIP-like cluster showed significantly higher hemoglobin values compared with the other two clusters (P=.001). As expected, the SF3B1-mutant cluster was significantly enriched for MDS with ring sideroblasts (MDS-RS) and showed significantly lower hemoglobin values (P=.001) and increased values of serum ferritin and transferrin saturation compared to other clusters (P=.001 and P=.002, respectively).

HR-cluster showed significantly lower overall survival (OS) compared to CHIP-like and SF3B1-mutant clusters (median 2.6 vs 6.8 or 6.4 years; P<.001), and higher risk of progression into higher-risk MDS or acute myeloid leukemia (AML) (median 4.2 vs 12.7 years or not reached; P<.001). No significant difference in either OS or risk of disease progression was noticed between SF3B1-mutant and CHIP-like clusters. However, a significantly shorter time-to-treatment with erythropoiesis stimulating agents was noticed in the SF3B1-mutant cluster (P=.007), suggesting a more rapid erythropoietic impairment that did not translate into a worse outcome.

Conclusion. Mutation profiling identifies meaningful clusters of lower risk MDS with distinct molecular pathways, clinical features and endpoints. These results represent a robust basis to inform genetic ontogeny-based classification and individual risk assessment, as well as to inspire biology-driven clinical trials in lower risk MDS.