In Analogy to AML, MDS Can be Sub-Classified By Ancestral Mutations

Somatic mutations constitute key pathogenetic elements in MDS. Unbiased whole exome sequencing (WES) and deep NGS led to discovery of new somatic mutations and also to the recognition of i) tremendous diversity of mutations and their combinations; ii) individual intra-tumor heterogeneity and clonal hierarchy. Chromosomal lesions further increase the complexity of molecular defects.

While in MDS molecular defects are acquired in order, observations made in AML highlight the importance of ancestral events; e.g., t(8;21), inv16 or t(15;17) and other lesions that are used as the basis for nosological sub-classification. Thus, it is the identity of individual ancestral events or their classes rather than the spectrum of secondary events or the distribution of mutations, that will allow for molecular, functionally-relevant and diagnostically useful classification within MDS. This would explain why only a few somatic mutations have been found to be prognostically important, as their position in the clonal hierarchy has not been accounted for. With this in mind, we applied WES (N=206) and targeted deep NGS (N=836) and studied 100 samples serially with analyses focused on ancestral events.

Globally, through WES we identified and validated 2386 mutational events in 1458 genes. Of these, 112 genes were mutated at significant frequencies (q<0.05); groups of affected genes involved in splicing, transcription, DNA methylation, histone modification, and others were distinguished. On average, 9 somatic events per MDS case, 10.7 in secondary AML, and 12.5 in MDS/MPN were found. Resequencing in combination with SNP-array karyotyping provided information on variant allelic frequency (VAF) adjusted for corresponding zygosity of mutations; 99% of cases displayed clear intra-tumor heterogeneity due to multiple clones defined by hierarchically acquired somatic mutational patterns.

Using cross-sectional analyses, the highest mean VAF could be interpreted as consistent with the ancestral nature of the mutations, as seen for instance in a proportion of TET2 and SF3B1 mutant cases. In contrast, the lowest mean VAF indicated secondary events, as occur in NPM1 and RAS pathway mutations. Similar conclusions were made based on cross-sectional analyses showing a similar distribution of ancestral but not secondary events in MDS and sAML. All gene mutations were categorized into those that are predominantly ancestral and those that are facultatively secondary. The most frequent founder mutations were identified (TET2, DNMT3A, SF3B1, ASXL1, TP53, U2AF1, RUNX1, SRSF2) and used to sub-classify approximately 80% of patients, with the remainder containing more infrequent ancestral mutations. While in a combined fashion (as both founder and secondary events) many of these mutations were not predictive of prognosis, they gained relevance when only cases affected by ancestral mutations were used for prognostication. Thus some of the mutations, when present as secondary events may not be predictive.

Founding mutations may determine subsequent clinical and molecular features. While other frequently affected genes, SF3B1 or ASXL1, are not associated with a significant increase in the number of concomitant mutations, cases with TET2 mutations showed significantly more frequent mutations per case than those with wild-type TET2 (14.6 vs. 9.1; p=0.001). Moreover, ancestral TET2 mutations were associated with concomitant mutations due to high C-to-T transitions, possibly because reduced 5-hydroxymethylcytosine might create the specific mutator milieu.

Most important is the association not of any type, but of ancestral mutations with certain pathomorphologic features and outcomes. Founding TET2 mutations are associated with MPN/MDS while secondary TET2 mutations are present in MDS. Ancestral DNMT3A mutations determine a rapid progression to AML, whereas subclonal DNMT3A mutations are also found in high-risk MDS. RAS pathway mutations are ancestral in CMML and also secondarily positive in the late stage of MDS (sAML). Specific ancestral events may determine subsequent mutational events, and while both types of mutation may affect the clinical phenotype, the initial events are less diverse and more subtype-specific. In conclusion, WES clarified the distinct landscape and ordering of the somatic mutational spectrum in MDS.