Serial Sequencing in Myelodysplastic Syndromes Reveals Dynamic Changes in Clonal Architecture and Allows for a New Prognostic Assessment of Mutations Detected in Cross-Sectional Testing

MDS and related disorders, including MDS/MPN and sAML that evolved from these conditions constitute disease continuum characterized by a wide spectrum of molecular lesions which often overlap. Here, we defined general mutational spectrum and clonal architecture in a large cohort (n=718) of MDS studied by whole exome sequencing (WES) and target deep sequencing. Within this cohort 97 cases were studied at multiple time points to clarify the clinical impact of clonal dynamics on phenotype commitment or outcomes. All samples were obtained after informed consent, according to protocols approved by the respective ethics boards of the participating institutions.

When mean and maximum variant allele frequency (VAF) for whole mutations were at one time-point evaluated in disease phenotypes, significantly higher averaged values suggested their larger clones in sAML and CMML compared to MDS. Clustering analysis of multiple mutational events by Pyclone software discriminated the cases with multiple mutational clones (positive heterogeneity) and those with a single expansion of MDS clone (no heterogeneity detected). Over 80% of low-risk MDS and all the sAML harbored multiple clusters of mutations. These results suggest that intra-tumor heterogeneity of MDS is most likely due to various sizes of clonal and subclonal mutations, likely impacting clinical behavior.

To delineate clonal dynamics in MDS, we assessed mutational burden and their temporal changes in serially collected samples (n=97). Among these, Pyclone analysis was applied to exome sequencing at two time points (n=11 pairs). All cases showed various mutational clusters with individual expansions and declines, including initially present, newly acquired or disappearing during clinical course. Initial subclones were identified at disease presentation in 55% of cases, of which in 86% the subclones expanded to occupy whole MDS population with clonal sweep. New subclones acquired during clinical course were identified in 91%, in which 60% cases harbored clonal sweep. Disappearing clones were observed in 55% of cases. Next, we applied clustering analysis on clonal size of driver mutations evaluated at multiple time points (n=97 cases) to categorize the most frequently mutated genes into 3 subtypes. Mutational burden of PTPN11 most frequently increased and were associated with leukemic evolution (an example of type I gene). Similarly, CBL, NRAS, STAG2, RUNX1, and IDH1 were categorized into the type I genes, demonstrating increased clonal size resulting in the evolutions into high-risk phenotypes. Although JAK2 mutations were related to the stable clinical course when the mutational burden decreased, cases with highly expanded JAK2 mutations resulted in leukemic evolution (occasional evolution or expansions; type II gene). DNMT3A, SRSF2, TP53, U2AF1, and ASXL1 mutations were also categorized into such type II consequences with occasional progression. The last category (type III) included clonal/founder genes EZH2, TET2, SF3B1 and PRPF8, demonstrating random shifts of clonal size and lack of association with leukemic evolution.

The proposed hierarchical categorization correlates with clinical parameters. Cases with the increasing burden of type I gene mutations showed most significant increases in myeloblasts. Overall survival measured from second sampling time points in the cases with increasing type I mutations was significantly shorter in the whole cohort (HR=2.05, 95%CI; 1.14-3.79, P=0.016) and in the cases solely with IPSS INT-1 (HR=2.37, 95%CI; 1.01-5.97, P=0.048). Subcohorts classified according to the presence or absence of increasing type I mutations did not differ with regard to the IPSS categories. In contrast, increased mutational burden of type II and III genes did not correlated with any of the clinical parameters examined, even though some gene mutations including TP53, EZH2, and U2AF1 represented poor prognostic factors at disease presentation.

In conclusion, this work demonstrates that detailed understanding of clonal dynamics allows for new insights into clinical significance of somatic mutations, made possible only by serial sample sequencing at multiple time points. Increasing clonal burden of extracted genes associated with predictive prognostic impact should be prospectively validated in more uniform and larger cohort of MDS.