The Impact of Clonal Architecture of IDH1 and IDH2 Mutant Cases on the Biology of Myeloid Malignancies

The rearrangements of core binding factors (CBF) are classical founder events in primary AML (pAML). In non-CBF AML, many new somatic mutations have been identified, some of which may also serve as initial ancestral events. In myelodysplastic syndromes (MDS) and secondary AML (sAML), the diversity of ancestral lesions may be even greater, with potential overlap with other myeloid neoplasm in initiating molecular events. E.g., endomorphic IDH1/IDH2 mutations have been discovered in pAML, but also can be present in sAML and MDS. Analysis of the clonal architecture of a limited number of AML cases suggested that IDH1/2 mutations are ancestral. Theoretically, any ancestral lesion could also occur at later stages as a subclone. It is not clear if IDH1/2 mutations play the same role in pAML vs. sAML and MDS. This is a critical distinction as drugs targeting founder events may be able to be curative, while targeting secondary events may reverse clonal evolution, but is unlikely to eliminate the founder clone. Moreover, initiating molecular lesions are more likely to be subcategory-defining than the secondary events.

We hypothesized that IDH1/2 mutations may occur at different stages in the clonal hierarchy in patients with various myeloid phenotypes and may affect both the nature of genetic landscape and clinical features should IDH1/2 mutations be ancestral. Conversely, should IDH1/2 be found to be subclonal it will be essential to identify the founder clone. To address these issues, we have investigated a cohort of 2033 patients with secondary AML (N=125), pAML (N=294), MDS (N=1404), MDS/MPN or MPN (N=201) for the presence of IDH1/2mutations and other clonal events using various forms of deep NGS.

IDH1 mutations were found in 3%, 1.3%, 7.8% and 8% of MDS, MDS/MPN, pAML and sAML, whereas those of IDH2 were found in 3.4%, 3.3%, 6.5% and 9.1%, respectively. While both IDH1/2 mutations occurred most frequently in pAML (P=.0001), IDH2 was commonly associated with higher-risk MDS (P=.013). IDH1 mutations coincided with younger age (P=.01) while IDH2 mutants were likely to have a normal karyotype (P=.01). The average copy number-adjusted clonal size was 58% for IDH2 and 56% for IDH1 (P=.78). DNMT3A and NPM1 mutations most frequently co-occurred with IDH1/2. RUNX1, ASXL1, SRSF2, and STAG2 were significantly associated with IDH2, whereas PHF6 mutations with IDH1.

To reconstruct the clonal architecture in IDH1/2 cases, variant allelic frequencies were compared, along with cross-sectional comparisons of mutational frequencies and serial samples. Cross-sectional analysis of the distribution of IDH1/2 mutations in patients with lower-risk and higher-risk MDS and sAML showed increasing proportions of both IDH1 and IDH2 mutant cases with more advanced stages of the disease, suggesting that in some cases IDH mutations evolved during progression. This finding was substantiated when we analyzed serial samples. In 5/6 cases, IDH1 mutations constituted ancestral events with PHF6, NPM1, JAK2, NRAS and SMC3 serving as the most common subsequent subclonal events. The sixth case had U2AF1 as the ancestral event. IDH2 was an ancestral event in 3/4 serial cases with STAG2, ASXL1, NRAS, TET2 and NPM1 as common subclonal events. When clonal hierarchy of concurrently mutated genes was analyzed using the diversity score calculation of VAFs, IDH1 mutation was identified as a dominant clone in 45% of cases; when it was subclonal, DNMT3A, SRSF2 and SF3B1 were the most commonly detected dominant mutations. In contrast, in pAML cases, IDH1/2 constituted dominant mutations is 80%, with NPM1, FLT3 mutations being the most common secondary events to IDH1, and RUNX1 being the most common secondary event to IDH2. When IDH1 mutations were analyzed, they constituted ancestral events in 59% of cases (P=.9, vs. IDH1). When IDH was found to be subclonal, the ancestral events were heterogeneous. OS was analyzed in 786 patients within this cohort. IDH1 mutational status had no effect on survival of pAML or MDS patients, whereas IDH2 mutant MDS patients had an inferior OS compared to WT patients (24 vs. 40 months, P=.05) in multivariate analysis.

In sum, our results suggest that in pAML, IDH1/2 mutations are primary events, whereas in sAML, MDS and MPN/MDS, IDH1/2 mutations may also be subclonal. As a primary event, IDH1 or IDH2 may influence clinical features and be used to categorize a proportion of patients in particular with regard to use of targeted therapies.