Granulocyte maturation and splicing defects in DDX41-mutated leukemia

Abstract Deleterious germ line DEAD-box helicase 41 (DDX41) variants cause an inherited predisposition to adult-onset acute myeloid leukemia (AML) accounting for ~5% of cases. DDX41 encodes a DEAD-box type RNA helicase potentially involved in RNA processing, ribosomal biogenesis, metabolism, cell cycle progression, and innate immunity. To date, how DDX41 loss-of-function causes leukemia remains unclear. To evaluate the role of DDX41 in AML, we performed total RNA-Sequencing (RNA-seq, N=9) and Tandem-Mass-Tag Mass Spectrometry (TMT-MS, N=11) on peripheral blood and bone marrow from patients (pts) with deleterious germ line DDX41 variants comparing these to age-matched healthy individuals (N=8 and N=7, respectively). Five (45%) pts had secondary somatic DDX41 mutations, and eight (73%) had somatic mutations in genes recurrently mutated in AML such as TET2, DNMT3A, FLT3, and BCOR. First, we questioned if differential gene expression via RNA-seq is recapitulated in TMT-MS data and if shared pathways could be identified. We identified 24,336 unique genes using RNA-seq and 4787 unique proteins using TMT-MS. 4729 were common in both datasets and 209 (4.4%) were mutually differentially expressed (p<0.01, false discovery rate (FDR)<0.01). Using over-representation analysis of KEGG gene sets, we identified nucleocytoplasmic transport (hsa03013) and spliceosome (hsa03040) pathways as shared over-expression signatures. Under-expression signatures included adherens junction (hsa04520), leukocyte-endothelial migration (hsa04670), and Fc gamma R-mediated phagocytosis (hsa04666), pathways suggesting compromised granulocyte differentiation and innate immune function. To study this, we engineered pro-myeloblast HL-60 cells with CRISPR/CAS9 fusing a HaloTag to the C-terminus of endogenous DDX41. These cells were differentiated with ATRA (5 µM) or DMSO (1.25%) and DDX41-Halo was selectively degraded with PROTAC3 (300 nM) which targets Halo-proteins to the E3 ligase. Using flow cytometry, we found that DDX41-Halo degradation induced partial differentiation along the granulocyte lineage identified by acquisition of CD15. However, cells were unable to progress to mature granulocytes and did not express CD66b to the same extent as those differentiated with ATRA (4% versus 26%, p<0.05). Furthermore, degradation of DDX41 induced an apoptotic phenotype notably pronounced upon differentiation (2% control, 16% PROTAC3, 10% ATRA, 41% PROTAC3 and ATRA; p<0.05). Second, we sought to characterize splicing defects associated with DDX41-mutated AML. Using our RNA-seq data, we identified 133 confident splicing defects in DDX41-mutated AML samples compared to healthy controls (deltaPSI>10%, FDR<0.05). Retained introns (RI, N=103) were the predominant event identified and shared a common signature: short intron length (median 87 nucleotides) and enrichment of cytosine residues upstream of the 3' intron-exon junction. Notably, we found RIs predicted to undergo nonsense-mediated decay within TP53 in 7 (78%) germ line DDX41 AML samples. To better understand the mechanisms driving splicing dysfunction, we overexpressed DDX41 p.E345A, a helicase dead mutant in HEK293 cells. Comparing DDX41 p.E345A cells to wild-type DDX41, we identified 2378 splicing defects (deltaPSI>10%, FDR<0.05), of which 78% were RIs with the same features as pts. To identify the direct RNA targets of DDX41, we used Photoactivatable-Ribonucleoside-Enhanced Crosslinking and Immunoprecipitation and observed that most of DDX41 binding sites are located at the 3' splice sites of the introns with high retention in the DDX41 p.E345A cells. We next performed co-immunoprecipitation coupled to MS and found direct interactions between DDX41 and the catalytic spliceosome complex. Altogether, our results support that DDX41 is required for the removal of short introns, potentially by facilitating the accessibility of the 3' splice site to the splicing machinery. Overall, these data raise the question of whether transcript levels are sufficient to predict downstream biological mechanisms in diseases with splicing anomalies. Our findings suggest that ineffective granulocyte maturation and splicing defects are characteristics of DDX41-mutated AML. Our tissue culture model in HL60 cells suggests that depletion to DDX41 may induce apoptosis predominantly upon differentiation and our mechanistic studies highlight a direct role of DDX41 in RNA splicing and identify a signature for DDX41-associated splicing events.