Genomic Landscape of Germline RUNX1 variants in Adult Myeloid Neoplasia According to Functional Impact

Germline (GL) RUNX1 (RUNX1^GL) mutations result in autosomal-dominant familial platelet disorder (FPD) with heightened risk of myeloid neoplasia (MN). Platelet abnormalities and thrombocytopenia are characteristic of FPD, however the phenotype and age of MN onset have a marked high degree of variability, ranging from 2 to 72 yrs. Leukemic progression can be accompanied by biallelic RUNX1 hits or induced by additional somatic variants, and is potentially associated with distinct functional effects of RUNX1 variants including dominant negative (DN), haploinsufficient (HI), and hypomorphic (HM). We hypothesized that a large systematic study of RUNX1^GL including functional impact will elucidate the effects of different RUNX1 variants.

Irrespective of their derivation (GL or SOM), we categorized RUNX1 mutations (RUNX1^MT) according to their predicted functional effects as: i) HI (RUNT-truncating); ii) DN (C-terminal nonsense and frameshift, missense described as DN); iii) RUNT missense which functions as a hypomorphic category.¹ C-terminal was defined as downstream to the predicted location of nonsense-mediated decay.² Since our initial report we incorporated 1956 additional MN patients (pts) from our institution and public series.^1,3-4 We also devised a novel rational algorithm to curate pathogenic and likely pathogenic (P/LP) RUNX1^GL variants prioritizing exclusion of false positive. We then selected RUNX1^GL variants for VAF ≥40%, constant VAF in remission (blasts <5%), and description of GL in previous literature. Whenever possible, GL status (n=29) was assessed on CD3+ cells resulting in 3 entirely novel GL variants (c.843C>G; c.1229C>A; c.842dupA) with a clinical or family history suspicious for FPD and/or hematologic malignancy. The pathogenicity of GL P/LP variants were corroborated with the ClinGen MM/VCEP evidence repository.² As a result, we found 4 GL VUS selected based on suspicious clinical history, and included in our analysis 102 RUNX1^GL, including 4 novel variants (c.982dupA; c.548C>A; c.347delT; c.1413_1415delinsG) with stable VAF despite disease remission, consistent with GL derivation. The distribution of GL and SOM within functional impact categories (DN, HI, HM) overlapped.

When studying co-mutational landscape of GL mutants, PHF6^SOM (p=.018) hits were more frequent with RUNX1^GL, while U2AF1 (p=.042) and DNMT3A (p=.030) correlated with RUNX1^SOM. RUNX1^GL also attracted RUNX1^SOM mutations and thus, biallelic configuration was found in 13 (13%) of RUNX1^GL cases, while 2 simultaneous RUNX1^SOM, which occurred in 45 (14%) of RUNX1^SOM cases, could reflect either biallelic configuration or somatic mosaicism. We observed a trend for co-occurrence of +8 with RUNX1^SOM (p=.056), an association not found in RUNX1^GL.

The top SOM hits most overrepresented in RUNX1^GL vs wildtype (WT) were SRSF2, ASXL1, TET2, and PHF6, suggesting epistatic effects increasing fitness in RUNX1^GL background. When comparing hypomorphic GL cases to WT, the top hits most prevalent in HM RUNX1^GL were the same, indicating a potential selection of these mutations for somatic gene rescue (SGR). Mutations more prevalent in DN and HI RUNX1^GL cases compared to WT included SRSF2, ASXL1, TET2, and BCOR, pointing towards a linear model of somatic gene acquisition inducing leukemogenesis. No SOM hits showed a preference towards HM RUNX1^GL over DN and HI RUNX1^GL.

In general, RUNX1^MT exhibit a male predominance (65%) and this trend is also present in RUNX1^GL (65%). We hypothesize that an overrepresentation of X-linked mutations (BCOR/L1, KDM6A, PHF6, STAG2, ZRSR2) may be responsible for the observed sex disparity. Indeed, a greater proportion of male RUNX1 carriers harbor X-linked mutations (43%) vs females (34%), (p=.037). However, in RUNX1^GL, proportion of patients harboring X-linked mutations did not account for this discrepancy.

In summary, our results offer new insights on the genomic landscape of RUNX1 mutated MN. Specifically, different types of RUNX1^GL may select for certain SOM hits conferring clonal advantage. We illustrate the complexity of leukemogenesis as many hits may participate in SGR or linear progression.