Unsupervised clustering of DDX41 mutants informs on the somatic landscape of single germline hits

Abstract

Abstract Germline DDX41 mutations can occur in monoallelic form or in the context of somatic hits, predominantly R525H. While germline DDX41 can be easily identified, the sensitivity for detection of somatic hits in this gene is limited. Cell sorting experiments have shown that in seemingly monoallelic germline DDX41 cases, somatic hits might remain undetected as they are limited to early precursor cells not contributing to blood production. Kroger B et al. Evaluating Stem Cells to Predict Post-Transplant Relapse in the Myelodysplastic Syndromes. Blood (2022); 140 (Supplement 1): 9833-98351 Clinically, germline DDX41 defects indicate that DDX41 is a bone marrow failure (BMF) gene and leukemias constitute, potentially, a molecularly separated entity of myeloid neoplasia (MN). However, current prognostic schemes are inaccurate or unable to assess clinical impact of various DDX41 configurations. Similarly, molecular classifications (e.g., our molecular nosology or taxonomy) did not account for the diversity of molecular hits in DDX41 gene. Indeed, we reported a dedicated prognostic scheme to capture relevant prognostic information in classical DDX41 constellation. Gurnari C et al. Blood (2024) 144 (Supplement 1):1332.2 However, the mutational landscape of single germline DDX41 remains to be defined. Here, we determine the molecular and clinical context of DDX41 configurations focusing on those with single germline hits. For this purpose, we screened a total of 11,500 cases with available clinical grade NGS which included MN (n=3,000) and BMF patients (n=266) among others. We found a total of 154 DDX41 mutant patients: single somatic (14%), single germline (61%) or biallelic hits (25%) in patients diagnosed with AML (36%), MDS (28%), MDS/MPN (14%) and other clinical presentations (22%; with history of cytopenias or aplastic anemia). To address the heterogeneity of mutational configurations of DDX41 we took advantage of unbiased/unsupervised clustering. We used a binary latent factor model and identified 8 molecular DDX41 associated clusters (MC1-8). Based on the size we identified 1 major (MC1), 2 medium (MC3, 5) and 5 minor (MC2, 4, 6, 7, 8) clusters. We categorized the clusters according to the abundance of the three configurations. MC1 together with MC5 (40%) were the only two clusters with the highest fraction of cases with germline/somatic hits. MC1 was predominantly composed of single germline (71%) and was enriched with 44% (67/154) of cases diagnosed with AML (43%) and MDS (25%). BMF (25% of MC1) accounted for 50% of the entire BMF cohort. Only 7% of the cases were diagnosed with MDS/MPN or MPN only. This cluster was characterized by patients with a low mutational burden (≤2% of the targeted panel) with the most common somatic mutations (although rare) being SUZ12 and ZRSR2. MC3 contained 66% of germline and was characterized by TET2 (63%) with a relatively large fraction of cases having SF3B1 (36%) and TP53 (31%) mutations. MC4 was mostly characterized by mutations in the cohesin complex(STAG2, SMC3) suggesting that disruptions in chromosome segregation are a key pathogenic driver for this subgroup of patients. MC5 was the only cluster with co-occurrent DNMT3A (50%) and JAK2 (45%) mutations and more MPN diagnoses. Among the minor clusters, nearly all patients in MC6 harbored a TP53mutation, MC7 showed the highest proportion of splicing factor mutants (SF3B1, U2AF1) and MC8 showed an unusually high proportion of DNMT3A mutants with 90% of the cases harboring mutations in this gene. Minor clusters were also the most heterogeneous with MC2 having almost all patients carrying co-occurrent EZH2, BCORL1, CUX1, TET2 and ASXL1 mutations. We then analyzed the somatic contributors in the context of single germline DDX41 and we explained 60% of the patients by having at least one somatic lesion including 2 patients carrying somatic mutations in other RNA helicases (DDX10, DHX16). The most common somatic hits in this cohort occurred in TET2 (25%), DNMT3A (22%), ASXL1 (18%) and JAK2 (12%) while common cytogenetic abnormalities included complex karyotypes. Notably, 30 cases had no detectable somatic mutations and were all present in MC1 and almost equally distributed in terms of disease types (AML, 35%; MDS, 28%; BMF, 25%, MDS/MPN, 12%) suggesting other disease driven pathways. In summary, molecular assignment suggests possible molecular interactions and somatic contributors to MN development in the pathogenesis of DDX41 mutant hematologic diseases including the ones with history of cytopenias and may offer a clinically applicable guide for management of DDX41 mutations carriers.