Chlon, Timothy M.

Abstract Germline variants in DDX41 are the most frequent genetic predisposition to adult hematologic malignancies. The most common variants are truncating, implicating loss of function in the pathogenesis. However, non-truncating variants account for 30-40% of cases, and their impact on essential DDX41 functions remains unknown. We utilized a genetic complementation assay to assess the functionality of 10 recurrent germline non-truncating variants of DDX41. All variants restored viability to Ddx41 -deficient hematopoietic progenitor cells at exogenous expression levels. In contrast, the hotspot mutant p.R525H, which is somatically acquired at disease onset in >50% of patients, failed to restore viability. CRISPR-based modeling in cell lines and mice revealed heterogeneity: some variants were non-functional at endogenous expression levels whereas others maintained complete functionality, supporting normal cell proliferation and even lifelong hematopoiesis in a homozygous setting. Notably, co-expression of p.R525H with some variants caused impaired hematopoietic progenitor cell viability, indicating a dominant-negative effect of p.R525H. In contrast, other variants, all classified as variants of unknown significance, were unaffected by the presence of p.R525H. A screen of 100 disease-associated variants confirmed that many non-truncating germline variants are susceptible to p.R525H-mediated dominant-negative effects, whereas wild-type DDX41 is not. These findings indicate that DDX41 variant curation is complicated by variable effects on functionality and variant-specific interactions with somatically-acquired DDX41 mutations. The dominant-negative effect of p.R525H provides a mechanistic basis for the conclusion of recent patient cohort analyses that co-occurrence with a somatic hotspot mutation is a reliable indicator of DDX41-driven disease in carriers of non-truncating variants.

Abstract Introduction: DDX41 (DEAD-box helicase 41) mutations (mt) represent the most prevalent germline predisposition syndrome (GPS) in adult patients (pts) with myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML). This molecular subtype is typically associated with late-onset disease, runs an indolent clinical course, with a relative absence of high-risk co-mutations. While several retrospective studies have reported favorable responses to venetoclax (VEN)-based therapies, anecdotal observations suggest variable outcomes. We aimed to evaluate the clinical characteristics and treatment outcomes of patients with DDX41-mt MDS/AML treated at the Mayo Clinic. Methods: We conducted a retrospective study of pts diagnosed with DDX41-mt germline predisposition syndrome with concurrent myeloid neoplasms at our institution. Germline testing was conducted under an IRB approved protocol (Mayo Clinic IRB# 16-004173; clinical-trials gov. Identifier: NCT02958462). Of 198 pts with DDX41-mt myeloid neoplasms, 77 had high-risk disease—43 with MDS with increased blasts (MDS-IB1, n = 11; IB2, n = 32) and 34 with AML (WHO 2022). We assessed how clinical, and disease features influenced outcomes in this subgroup. Among these pts, 69 (90%) pts had likely pathogenic/pathogenic variant (DDX41path) variants and 8 (10%) pts had variant of undetermined significance (DDX41VUS) according to variant classification criteria from the American College of Medical Genetics/ the Association of Medical Genetics/ the Association for Molecular Pathology (ACMG/AMP) guidelines. Results: The median age was 68 years (yrs) (range [R],34-92), 70% were male and the median bone marrow blast % was 15 (R,5-60). The most common DDX41 germlinevariants (> 5% cases) included p.D140fs (26%), p.M1I (22%), and p.P258L (6.5%). 74% had DDX41 truncating variants. Thirty-one (40%) acquired a somatic DDX41-mt (R525H [81%]) and 43 (56%) had other somatic mutations. Among these, the most frequent were ASXL1 (n= 9, 21%), TP53 (n= 8, 10%), splicing factor (n= 7, 9%), and DNMT3A (n= 5, 6%). Twenty-six, (34%), 26 (34%), 15 (19%) and 9 (12%) pts were treated with hypomethylating agents (HMA), HMA+VEN, intensive chemotherapy (IC) and active surveillance/supportive care, respectively. Overall, the composite complete remission rate (cCR) in high-risk MDS and AML was 56%. The cCR rates (41%, 59% and 68.7%, p= 0.10) and median duration of response (DOR; 43.7, 44.9 and 28.9 months [mo], p= 0.46) were not significantly different with HMA, HMA+VEN, and IC induction. Thirty-one (40%) pts received allogeneic stem cell transplantation (allo-HCT); 4 (36%), 15 (47%) and 12 (35%) in MDS-IB1, MDS-IB2 and AML groups, respectively (p= 0.56). With a median follow up of 34.7 mo (R, 0.7-136.7), the median overall survival (OS) from the time of diagnosis was 56.9 mo (95% CI, 50.7-63.0) and was not significantly (NS) different between MDS-IB1 (not reached, 70% at 3 yrs), MDS-IB2 (56.9 mo, 86% at 3 yrs) and AML (55.7 mo, 60% at 3 yrs) groups, p= 0.16). The median OS was NS different between DDX41path and DDX41VUS (58.6 vs 55.7 mo, p= 0.50), respectively. We conducted a multivariable analysis for OS using variables that showed significance or trend towards significance in univariable analysis. Concurrent splicing factor mutations (HR; 5.29, 95% CI: 1.29-21.6, p= 0.02) remained negatively prognostic, while there was a trend towards better survival outcomes with allo-HCT (HR; 0.48, 95% CI: 0.21-1.07, p= 0.07). Bone marrow blast ≥ 20% (HR; 1.87, p= 0.12) and TP53-mt (HR; 2.14, p= 0.12) did not retain significance. Conclusion: In our cohort of patients with germline DDX41-mt high-risk MDS and AML, VEN-based regimens were not superior to HMA alone or IC, in terms of response rates and outcomes. On multivariable analysis, the presence of concurrent splicing factor mutations independently and negatively impacted survival with a trend towards better outcomes with allo-HCT. These findings highlight the heterogeneity within DDX41-mt myeloid neoplasms and underscore the need for future studies in larger cohorts to better define predictors of response and survival.

Abstract Germline mutations in the RNA-helicase gene DDX41 cause predisposition to myeloid malignancies (MDS/AML) with a penetrance of 50% by age 90. DDX41 germline mutations are heterozygous and predominantly truncating variants, indicative of loss-of-function. More than half of DDX41-mutated carriers that develop MDS/AML will acquire a second-hit mutation in the other allele of DDX41, frequently causing the amino acid substitution p.R525H. Single-cell analysis from a biallelic DDX41-mutated MDS patient showed that the p.R525H mutation was highly enriched in dendritic cells (DCs) but less frequent in other cell types. We hypothesized that p.R525H causes the hematopoietic stem and progenitor cells (HSPCs) to differentiate preferentially into DC. To precisely model the germline of DDX41-mut patients, we generated human induced pluripotent stem cells (iPSC) with a frameshift mutation in exon 6 of one allele of DDX41, modeling the most common germline mutation. We attempted to modify the other allele of DDX41 to express the p.R525H mutation but could not derive viable cells. To circumvent potential lethality of the acquired mutation, we inserted a Doxycycline (DOX)-inducible wild-type DDX41 expression cassette in the AAVS1 safe harbor locus. Then we modified the endogenous allele to cause the p.R525H mutation in trans to the existing frameshift mutation. In this manner, we derived iPSCs that would model the most common configuration of DDX41 mutations (DDX41R525H/-) upon the withdrawal of DOX from the culture. In undifferentiated iPSC cultures, we observed profound defects in cell proliferation and survival upon DOX withdrawal. To determine the effect of the p.R525H mutation on hematopoietic cells, we differentiated the iPSCs into CD34+ HSPCs in the presence of DOX, and then withdrew DOX and studied the effect of the DDX41R525H/- genotype on HSPC function. In colony formation assays, DDX41R525H/- cells produced fewer and smaller myeloid and erythroid colonies after DOX withdrawal. In liquid culture, DDX41R525H/- cells showed cell cycle arrest by EdU incorporation assay and protein synthesis deficit by HPG assay. Thus, we conclude that DDX41 is required for proliferation and survival in iPSC and hematopoietic progenitor cells and that p.R525H lacks the essential function. Next, we tracked HSPC differentiation in liquid culture containing Flt3 ligand, stem cell factor and thrombopoietin with or without GM-CSF. Identification of DC subsets by flow cytometry was based on the literature (Sontag S, et al. Stem Cells 2017;35(4)). We observed reduced overall differentiation in DDX41R525H/- cultures, evidenced by maintenance of CD34 positivity. However, among the CD45+CD34- cells, DDX41R525H/- cells were 3-4 fold more likely to be identified as plasmacytoid DC (pDC) compared to control conditions (+DOX or DDX41+/- cells). Notably, in DDX41R525H/- CD45+CD34- cells, overall CD123 and CD304 expression were profoundly upregulated and CD11c was downregulated, indicative of increased pDC character. We used a similar DOX-inducible approach to generate Molm13 human AML cell line with DDX41R525H/- and homozygous knockout (DDX41-/-) genotype and found that these cells had profound survival defects. Both genotypes exhibited DC-like morphological change with extended multiple projections upon DOX withdrawal. In flow cytometry analysis, both genotypes exhibited upregulated Clec9a (conventional DC1 (cDC1) marker) and Clec10a and CD1c (cDC2 marker) upon DOX withdrawal. DDX41R525H/- cells specifically exhibited pDC features shown by upregulated CD123 by flow cytometry and increased IRF7 by western blotting, while DDX41-/- had decreased IRF7. Furthermore, we found that DDX41R525H/- cells specifically showed increased IRF9 and phosphorylated STAT1 and STAT2, which are downstream of interferon receptor activation, indicating that type-I interferon produced by DDX41R525H/- pDC activates interferon signaling in autocrine/paracrine manner. These results indicate that p.R525H promotes differentiation into pDC, which is not observed in DDX41-/- cells. In conclusion, DDX41 p.R525H mutation has bimodal functions in HSPC: loss-of-function for proliferation/survival and gain-of-function for skewed differentiation into pDC. Dysregulated DCs may contribute to the pathogenesis of DDX41-mutated MDS/AML, particularly in light of recent descriptions of AML with pDC differentiation and the broader role of DC dysregulation in MDS/AML.

Abstract Heterozygous germline variants of DDX41 occur in 2-4% of adult Myelodysplastic Syndrome (MDS) and Acute Myeloid Leukemia (AML) cases, making them the most prevalent cause of inherited predisposition to these diseases. Homozygous deleterious variants are never observed in the germline, and loss of both DDX41 alleles causes apoptosis of proliferative cells, consistent with essentiality. Most disease-associated germline variants cause loss of full-length protein expression (truncating). However, 30-40% are missense (non-truncating) and their role in disease risk compared to truncating variants remains in question. Previous reports have shown that both truncating and non-truncating Pathogenic/Likely Pathogenic (P/LP) variants (defined by ACMG scoring) are associated with DDX41-specific disease features, including co-mutation of the contralateral allele of DDX41. In contrast, variants of unknown significance (VUS) are less associated with these features. Acquired co-mutations of DDX41, of which p.R525H accounts for 50-70%, are restricted to DDX41-mutated patients and are always missense. Our previous studies indicated that p.R525H lacks the essential DDX41 function and cannot support hematopoietic progenitor proliferation. However, the frequency with which this mutation occurs and the absence of acquired truncating mutations suggest that p.R525H has a specific effect on DDX41 function that is distinct from complete loss and promotes pathogenesis specifically in the context of existing P/LP DDX41 mutation. To understand how germline non-truncating variants predispose to disease, we characterized the effect of six recurrent non-truncating P/LP variants (p.G173R, p.R219H, p.P258L, p.Y259C, p.K331del, and p.I396T) and three VUS (p.K9E, p.Y33H, and p.M155H) on DDX41 function. We utilized a genetic complementation approach in murine lineage-negative (Linneg) bone marrow cells with inducible deletion of DDX41. We found that all non-truncating P/LP and VUS forms of DDX41 could fully rescue the colony forming potential of Ddx41-deficient cells to a similar level as wild-type protein, whereas the p.R525H variant completely failed to rescue. To determine the effect of the non-truncating variants on DDX41 function at endogenous expression levels, we utilized CRISPR to modify the DDX41 alleles in MOLM13 human AML cells. We produced viable, proliferating cell lines with homozygous mutations for all 3 VUS and 4/6 P/LP variants. For p.P258L and p.K331del, like p.R525H, we could not produce homozygous mutant cell lines despite screening hundreds of clones and finding several with heterozygosity. To further study the P/LP variants that were functional in the cell line, we generated C57Bl/6 mouse lines harboring p.G173R, p.Y259C, and p.I396T mutations in the endogenous Ddx41 alleles. Heterozygous mutant mice were phenotypically normal in all three strains. However, we could not obtain homozygous mutant mice for p.G173R or p.Y259C, indicating these cannot support embryonic/fetal development. In contrast, p.I396T-homozygous mutant mice were born at Mendelian ratios and were phenotypically normal until at least 15 months of age. Since p.R525H mutation is common in patients with non-truncating P/LP variants, we sought to model the combined effect of these mutations. We utilized the Linneg cells with inducible knockout of Ddx41, co-expressed each non-truncating variant with p.R525H by viral transduction, and monitored the colony formation potential. Remarkably, cultures expressing P/LP variants or M155I combined with p.R525H had dramatically reduced colony formation, while wild-type DDX41 and the other 2 VUS supported normal colony formation when co-expressed with p.R525H. Thus, we conclude p.R525H has a dominant-negative effect on certain non-truncating DDX41 variants. This is consistent with the emerging idea that the presence of an acquired hotspot DDX41 co-mutation is sufficient to classify a non-truncating DDX41 variant as pathogenic.Our results demonstrate that many DDX41 missense variants retain protein function when overexpressed. However, at endogenous expression levels, some P/LP variants are unable to support cell proliferation, indicative of hypomorphic effects on protein function. In patients, acquisition of the p.R525H mutation in the contralateral allele disrupts the functionality of P/LP variants through dominant-negative effects, leading to hematopoietic progenitor defects that are hallmark of the DDX41 pathogenesis.

SummaryGermline variants in DDX41 (DDX41MT‐germline predisposition syndrome [GPS]) are associated with predisposition to haematological malignancies (HM), including lymphoid and myeloid neoplasms (MN). We retrospectively analysed the clinical and molecular features of 195 patients diagnosed and treated at Mayo Clinic with DDX41MT‐GPS. Patients with germline DDX41 pathogenic variants (42.3%) and variants of unknown significance (VUS, 57.6%) were included. The median age was 68.6 years (16.2–93.4). Ninety‐two per cent were Caucasian, 64.1% were male and 30.8% had a family history of HM. There were 92 distinct germline variants among our cohort, and the most common was p.Met1? (15.9%), followed by p.Asp140Glyfs*2 (9.2%). Clinical diagnoses included asymptomatic carriers (10.2%), clonal cytopenia of undetermined significance (CCUS, 6.1%), myeloproliferative neoplasms (6.7%), myelodysplastic syndrome (40.5%), acute myeloid leukaemia (20.5%), lymphoid neoplasms (9.2%), plasma cell dyscrasias (6.1%) and solid tumours (22.5%). Patients with MN were older (median age 70 vs. 63.5 years) and more likely to be male (M:F ratio 2.3 vs. 1.0) and most patients (78.8%) with MN had a normal karyotype. The most common somatic variants involved DDX41 (34.4%), followed by TET2 (11.2%), DNMT3A (9.6%) and ASXL1 (9.2%). In summary, we have comprehensively described the spectrum of clinical phenotypes within the Mayo Clinic DDX41MT‐GPS cohort.

DDX41 MT Germline Predisposition Syndrome (GPS) is the most common adult GPS for myeloid neoplasms (MN), including MDS and AML. Although there have been reports of non-myeloid neoplasms, particularly lymphoid malignancies, in DDX41MT-GPS, these occurrences are infrequent and poorly characterized. We carried out this study in a large cohort of DDX41MT-GPS patients to address this deficiency. Following IRB approval, we retrospectively analyzed our database of 174 patients with DDX41MT-GPS to determine the rates of non-myeloid malignancies. Our previous data revealed that certain DDX41MT genotypes classified as variants of uncertain significance (VUS) exhibit a similar phenotypic spectrum to those classified as pathogenic (Badar et al Hematologica). Consequently, we included all these variants in our analysis. The median age of the cohort was 68.6 yrs (range, 13-4-93.4); 66% (n=114) male, 93% (n=162) Caucasian, with a median follow-up of 1.4 yrs (range, 0.03-5.6) from DDX41MT-GPS detection. The cohort included 34 (19.5%) asymptomatic carriers, 17 (9.8%) with clonal hematopoiesis (CH) or clonal cytopenias (CCUS), and 123 (70.7%) with MN. Eighteen (10%) patients had a lymphoid neoplasm (78% male, 100% Caucasian), with a median age at lymphoid neoplasm diagnosis of 67.7 yrs (range, 12.2-85.9). The most common lymphoid neoplasm was chronic lymphocytic leukemia (CLL) seen in 2.3% (n=), followed by follicular lymphoma (FL) and acute lymphoblastic leukemia (ALL) present in 1.7% each (n=3; 1 T-ALL and 2 B-ALL: 1 Ph+ and1 Ph-like). Additionally, there was 1 patient each with monoclonal B cell lymphocytosis (MBL), Waldenstroms macroglobulinemia , peripheral T cell lymphoma , large granular lymphocytic leukemia, and marginal zone lymphoma. Germline genotypes were classified as pathogenic in 50% (n=9, D140Gfs*2 x2, K331del, M1I x3, M316Dfs*31, P258L, Y340N) and VUS in 50% (n=9, E426K, E69D, G313D, K184R, P206A, P70S, Q246E, R164W, c.572-3T>C. Among these 18, 9 (50%) patients had co-existing myeloid disorders (CCUS x3, MDS x5, AML x1), and 5 (27.8%) patients were later diagnosed with therapy-related myeloid neoplasms (TMN) at a median of 5.9 yrs (range,1.6-38.7) after lymphoid neoplasm diagnosis and treatment. Seven (4%) patients had plasma cell dyscrasias (PCD; 3 with monoclonal gammopathy of undetermined significance(MGUS) and 4 with multiple myeloma (MM)); 57% (n=4) were male, with a median age of 61.3 yrs (range, 41.4-83.8) at PCD diagnosis. Subtypes of MGUS included IgG Kappa (n=2) and IgG Lambda (n=1), subtypes of MM included IgG Kappa (n=2), IgA Kappa (n=1), and Kappa light chain only (n=1). DDX41MT-GPS genotypes were pathogenic in 28.6% (n=2, G67Vfs*13, G313S) and VUS in 71.4% (n=5, A555T, R164W, Q246E x2, M307T). All patients with MM (n=4) received an autologous stem cell transplant, with 3 (75%) being subsequently diagnosed with a t-MN; median 3.6yrs after PCD diagnosis (range, 3.2-12.3). Seventeen percent (n=31) had visceral malignancies; 2.8% (n=5) having >1malignancy, 61% (n=19) male, 93.5% (n=29) Caucasian and 6.5% (n=2) Hispanic. Prostate cancer was most common (n=8, 4.6%), followed by breast (n=7, 4.0%), colon (n=5, 2.9%), squamous cell carcinoma (n=4, 2.3%, 3 of the head/neck and 1 of the leg), bladder (n=4, 2.3%), testicular (n=3, 1.7%), gynecologic (n=3, 1.7%), lung (n=2, 1.2%), thyroid (n=1, 0.6%) and pancreatic neuroendocrine tumor (PNET, n=1, 0.6%), respectively. The median age at first solid tumor diagnosis was 62.4 yrs (range, 10.9-81.9). DDX41MT-GPS was pathogenic in 64.5% (n=20, D140Gfs*2 x3, E113Kfs*14, G67Vfs*13, I396T, L283Cfs*21, M1I x4, M316Dfs*31 x2, p.M1?, P258L, Q368*, Q41* x2, R159*, R311*) and VUS in 35.5% (n=11, D30Y, H203Y, I215L, I337V, M155I, p.?, S289L, T13I, V318G, A555T). Of the 31 patients with visceral malignancies , 51.6% (n=16) were diagnosed with t-MN at a median of 4.9 yrs (range, 1.1-34.3) after malignancy diagnosis. While DDX41MT-GPS are clearly associated with an increased risk of myeloid neoplasms, we demonstrate potential associations with other neoplasms, especially lymphoproliferative disorders (10%). We for the first-time report associations with PCD (4%) and visceral malignancies (17%) and also highlight a worrisome trend for t-MN in DDX41MT-GPS treated with genotoxic agents. Given limitations of a retrospective analysis, this study underscores the need for a comprehensive natural history study in patients with DDX41MT-GPS .

Background: While myeloproliferative neoplasms (MPN), including essential thrombocythemia (ET), polycythemia vera (PV), and primary myelofibrosis (MF), are usually sporadic, adult-onset neoplasms, an increasing number of germline risk alleles have been described (e.g. EGLN1, ETV6, SH2B3, and ATM) in these disorders. DDX41MT Germline Predisposition Syndrome (GPS) is the most common myeloid GPS in adults and while there is extensive data on MDS/AML, little is known regarding MPN occurring in this context. We carried out this study to define the prevalence and characteristics of DDX41MT-GPS associated MPN. Methods: After IRB approval, we retrospectively assessed the Mayo Clinic cohort of 174 patients with DDX41MT-GPS, which included both pathogenic and clinically relevant germline variants of undetermined significance (VUS), based on our previous analysis (Badar et al Haematologica 2023). Standard statistical measures were applied. Results: We identified 12 (6.9%) patients with MPN (ET-5, PV-2, PMF-4, MPN-NOS-1). The median age at MPN diagnosis was 61.2 years (range, 8.-70.4); 58.3% (n=7) were male, and 83.3% (n=10) were Caucasian. A family history of hematologic malignancy was present in 16.7% (n=2, AML and T-LGL), while 50% (n=6) had a family history of visceral malignancies and 41.7% (n=5) had no family history of malignancies. Among ET patients (n=5) with DDX41MT-GPS, JAK2V617F was the MPN driver in 4 (80%) while type 1-CALR (L367fs*46) was a driver in 1 (20%). This included a 9-year-old child diagnosed with JAK2V617FMT ET. The DDX41 variant was classified as pathogenic in 40% (n=2; E113Kfs*14, D140Gfs*2) and VUS in 60% (n=3; V544L, V591I, P38L). Bone marrow (BM) aspirates in the ET patients were hypercellular in 40% (n=2), normocellular in 40% (n=2) and hypocellular in 20% (n=1), and none had increased blasts, or reticulin fibrosis (RF). In addition to JAK2V617F, somatic mutations seen included SF3B1 (K666T, VAF 16%) in 1 patient and IDH2 (R140Q, VAF 6%) and DNMT3A (R882H, VAF 7%) in 1 patient, respectively. With a median follow-up from diagnosis of 3.9yrs (range 2.9-8.9), there were no thrombotic events and no disease progressions was documented. There were 2 patients with JAK2V167FMT PV in whom the DDX41MT were clinically significant VUS (c.138+5G>T and M155I). Both patients had hypercellular BM, with no increase in blasts, no RF, normal karyotypes and no additional somatic mutations. With a median follow up of 7.5yrs (range, 2.1-12.9), 1 patient developed a deep vein thrombosis (DVT) and there were no fibrotic or leukemic transformations. There were 4 patients with PMF; 3 (75%) with JAK2V617F and 1(25%) with a type 1-CALR driver mutation (L367fs*46). DIPPS-plus scores ranged from 1 to 4 (median 2.5). The DDX41 variants were pathogenic in 50% (n=2, E256K and P78Qfs*3) and VUS in 50% (n=2, P510S and M155I). The median grade of RF at PMF diagnosis was grade 2 (range, 2-3). BM were hypercellular in 75% (n=3) and hypocellular in 25% (n=1), and 1 patient had 3% peripheral blasts (and no BM blasts). Somatic co-mutations included TET2 in 1 patient with a JAK2V617F, and CBL, EZH2 and SRSF2 in 1 patient with a CALR driver-mutation, respectively. PMF patients were treated with ruxolitinib (n=3), stem cell transplant (n=1), and pIFNα (n=1). With a median follow up of 3.9yrs (range,1.1-21.9), all patients were alive, and none had leukemia transformation. There was 1 patient with a JAK2V617F MPN-NOS with trisomy 9, where the DDX41MT-GPS was a VUS (I215L). There were no somatic co-mutations and at last follow up (3.1 years), there was no evidence for progression. Conclusions: While associations between late onset MDS/AML and DDX41MT-GPS has been well established, we systematically describe the MPN landscape seen in this context. MPN was diagnosed in 7% of our cohort, with JAK2V617F being the most common driver mutation. Most importantly, unlike in AML/MDS, where somatic DDX41 mutations are frequently associated with neoplastic transformation (30-50%; R525H being most common), none of our MPN patients had a somatic DDX41 mutation. Our work underscores the importance of germline mutation testing in patients with MPN.

Germline variants in DDX41 are the most common cause of inherited predisposition to Myelodysplastic Syndromes (MDS) and Acute Myeloid Leukemia (AML). DDX41 codes for an RNA helicase with many described functions, and thus the precise role of these variants in disease is uncertain. Approximately 2/3 of disease-associated DDX41 variants cause truncation of the DDX41 protein due to frameshift, loss of a translation start, or splice alterations, while the other 1/3 of DDX41 variants are missense. The effect of these missense variants on DDX41 function is largely unknown and thus they are typically categorized as variants of unknown significance (VUSs). Importantly, 30-50% of MDS/AML patients with either truncating or missense DDX41 variants acquire a somatic mutation in the other allele of DDX41, and this mutation has strong preference for the amino acid substitution R525H. Our previous work demonstrated that the R525H mutation causes loss of DDX41 function. To better understand how germline variants affect DDX41 function, we designed a genetic complementation strategy in immortalized hematopoietic progenitor cells with inducible deletion of DDX41. We screened six of the most common missense VUSs by expressing them from a lentivirus in DDX41-proficient cells and then inducing deletion of endogenous DDX41 to determine the effect of each VUS on proliferation and survival of the cells. Empty vector-transduced DDX41-deleted cells undergo cell-cycle arrest and apoptosis within 3-4 days. In contrast, viral expression of wild-type DDX41 rescues the survival and proliferation of the cells, while expression of the R525H mutant fails to rescue the cells. We screened the Y33H, M155H, G173R, R219H, P258L, and I396T variants of DDX41 in this assay, and all six of the mutants successfully rescued the survival and proliferation of DDX41-deleted cells long-term. This indicates that the mutant proteins coded for by these variants are functional. To test the effect of commonly observed truncating variants in the same assay, we expressed M1I, Q52fs, and D140Gfs variants of DDX41. Surprisingly, we saw complete rescue of cell proliferation and survival by the M1I and Q52fs variants. In contrast, the D140Gfs variant failed to rescue the cells. These data indicate that frameshift at D140 or later causes loss of DDX41 function. It is likely that translation from alternative methionine residues located at M127 or M132 of DDX41 accounts for the rescue of the cells by truncating variants affecting sites upstream of M127. In support of this theory, we targeted DDX41 for CRISPR-mediated knockout in MOLM13 cells and found that guide RNAs targeting exons 1-4, which are upstream of M127, produced homozygous frameshift mutations in proliferating cells from single-cell clones. However, guides targeting exon 6, which is downstream of M127 and M132, did not produce any surviving cells with homozygous frameshift mutations, despite screening 96 clones. This finding indicates that short DDX41 isoforms expressed from alternative start sites are functional and can support cell proliferation when full-length protein is lost. Collectively, these data raise important questions about the precise role of DDX41 variants in hematologic malignancy. Data from recently published patient cohorts indicates that patients with missense DDX41 variants have similar patient characteristics to those with truncating variants and thus these aberrations are thought to cause disease by similar mechanisms. While these cell line models may not fully recapitulate the pathogenic mechanism by which DDX41 variants cause MDS/AML predisposition, the results of this study demonstrate that the germline variants are not complete loss-of-function alleles in all cases and more complex mechanisms may be involved.

Introduction: Germline DEAD-box RNA helicase-1 mutant (DDX41) predisposition syndrome ( DDX41-MT-GPS) is one of the most frequent hereditary predisposition syndromes in adults with myeloid neoplasms (MN-2-5%). Somatic mosaicism involving the unmutated DDX41 allele (e.g., R525H) as a potential mechanism for myeloid evolution has been reported in 39-50% of patients (pts). In addition, given the later age of onset, somatic mosaic states composed of age-related clonal hematopoiesis (ARCH) (e.g., DNMT3A, ASXL1, TET2, and TP53), with reported frequencies less than expected in MDS and AML, have been documented. We conducted this study to define the somatic mosaic landscape in DDX41-MT-GPS . Methods: After approval from the Mayo Clinic IRB the prospective GPS database was queried for pts with DDX41-MT GPS. Germline confirmation was conducted on skin fibroblast-, or hair follicle-derived DNA. Somatic CH testing was conducted using a targeted NGS panel as previously described. Given the high frequency of variants of undetermined significance (VUS), comparisons were made between pathogenic ( DDX41path) and DDX41VUS. Results: We identified 106 pts with DDX41-MT-GPS, of which 83 (78%) met criteria for MN, 16 (15%) for CH/CCUS (clonal cytopenias of undetermined significance), while 7 (6%) were unaffected carriers. Somatic mosaicism in DDX41-GPS patients without MN Sixteen DDX41-MT pts without MN met criteria for CH/CCUS ( DDX41path [n=4] and DDX41VUS [n=12]; median age 69 years (yrs) (range [R], 19-78) ( DDX41path [69 yrs] and DDX41VUS [67 yrs], p=0.85), with a male predilection (75%). Four (25%) pts had somatic DDX41 MT (R525H [n=3] and A376T [n=1]) with median variant allele frequency (VAF) 10.5% (R, 5-27); 2 (50%) pts with germline DDX41path had R525H (VAF; 5% and 13%) and 1 (8%) pt each with germline DDX41VUS had R525H (VAF 5%) and A376T (VAF 27%) somatic variants, respectively. Abnormal cytogenetics (CG) were observed in 2/13 evaluable (15%) pts ([+8 and del20q]) in the germline DDX41path group only (p= 0.05). ARCH co-mutations were observed in 11 (69%) pts, not significantly (NS) different between DDX41path (75%) and DDX41VUS (75%), p= 0.66. Individual co-mutations: DNMT3A (25% vs 16%, p= > 0.99), ASXL1 (25% vs 8%, p=0.71), TET2 (25% vs 8%, p= >0.99), and JAK2 (0% vs 8%, p= >0.99) were NS different between the two groups. Somatic mosaicism in DDX41-GPS patients with MN Eight-three pts met criteria for MN ( DDX41path [n=43] and DDX41VUS [n=40]; median age 67 yrs (range [R], 27-92), with a male predilection (58%). Fifty-four (65%) pts had MDS ( DDX41path [55%] and DDX41VUS [45%], p=0.55) and 29 (35%) had AML ( DDX41path [41%] and DDX41VUS [59%], p=0.65). Overall, 15 (18%) pts had somatic DDX41 MT; 12 (80%) pts with DDX41path and 3 (20%) pts with DDX41VUS (p= 0.003). The median VAF of somatic DDX41 MT was 12% (R, 5-37), NS different between DDX41path (10.5%) and DDX41VUS (12%) (p=0.08). The most common somatic DDX41 MT was R525H (66% [n=10]), significantly associated with DDX41path (9/10 [90%]), p=0.01. ARCH was observed in 17 (20%) pts, significantly higher in DDX41path (76%) compared to DDX41VUS (24%), p= 0.03. Abnormal CG were observed in 10/73 evaluable (14%) pts (-Y [n=2], del 20 [n=3], del 5q [n=2], del 7q [n=1], t (3;8) [n=1] and complex CG [n=1]), higher in DDX41path (70%) compared to DDX41VUS (30%) with p value >0.99. We compared the spectrum of CH between DDX41 MT CH/CCUS and MN pts ( Table 1) and found no difference in the proportion of pts with somatic DDX41 MT (p= 0.52) and ARCH/myeloid MT (p= 0.56). The individual occurrences of DNMT3A (p= 0.49), ASXL1 (p= 0.71), TET2 (p=0.59), JAK2 (p>0.99) and TP53 (p= 0.56) mutations were NS different between the two groups ( Figure 1). Conclusion: We define the somatic mosaic landscape in DDX41-MT GPS and demonstrate that somatic mutations involving the other allele, especially R525H, are seen in ~20% of pts (>80% of R525H present in MN). The frequency and composition of ARCH and myeloid driver mutations was lower than expected for the age of presentation, in comparison to de novo and secondary MDS/AML and occurred more frequently in DDX41path compared to DDX41VUS. Forty-two % (n=41), including 37% with MN, did not have somatic mosaicism, underscoring the fact that mechanisms of leukemia progression remain to be defined in subsets of affected pts.