Genetic and Molecular Dynamics of DDX41-variants: Insights on the Variant Effect Prediction

About 50% of germline DDX41 (DDX41^GL) occur with biallelic somatic hits (including the recurrent R525H) while the remainder seems to harbor only monoallelic (germline or somatic) hits. DDX41^GL are heterozygous and typically occur in a frameshift fashion (including the canonical D140fs) implying a loss of function and defining DDX41 a bona fide TSG. Lack of inactivating biallelic somatic hits and presence of the conserved R525H led to unanswered questions as to the impact of variants on the protein function.

We reviewed our tally of 158 DDX41 mutants obtained by screening 1,893 patients with BMF. Overall, 61% of the patients carried sole germline, 25% somatic + germline and 14% carried sole somatic hits. Among biallelic configurations, there were canonical D140fs/R525H (n=4), germline missense/R525H (n=16), D140fs/non R525H (n=1) and missense germline/somatic (n=10).

We then undertook an integrated genomic, enzymatic and molecular dynamics analysis of DDX41 variants to clarify their effect including maladaptive RNA unwinding properties of mutant DDX41.

Our original study1 postulated that R525H is possibly hypomorphic because R525 is in the ATP binding domain. This location makes R525 a preferential sweet spot forming hydrogen bonds with the α and β phosphates of bound ATP to potentially regulate the dynamical motion of DDX41 required for RNA unwinding. The histidine replacement may alter but retain the basal unwinding activity albeit through a different hydrogen bonding pattern. To clarify the impact of missense hits on RNA unwinding rate and the efficiency of remodeling activity, we applied state-of-the-art computer modeling. The AI based system, AlphaFold2, was deployed to predict 3D structure and generate DDX41/RNA models for molecular dynamics simulation to study the effect of R525H on DDX41/RNA structure.

To explore the unwinding impairment, we examined R525 and the close R522 in the ATP binding site. In the wild type DDX41, R525 forms strong salt bridge interactions with 2 oxygens (O2A, O2G) in ATP at shorter distances whereas R522 maintains weaker interactions with another oxygen (O3G) in ATP. In R525H, H525 loses the strong salt bridge interaction with ATP; however, R522 replaces H525 to form a strong salt bridge interaction with ATP based on a distance <4 Å deemed a strong salt bridge interaction. The simulations further clarified that the distances between R525 in DDX41 and O2G in ATP followed a bimodal distribution likely reflecting the on (<4Å) and off (peaked at 5 or 5.5 Å) switch between DEAD and HELICc. Loss of salt bridge interaction between H525 with ATP undermines the role of R525 to control the relative motion between two domains required for efficient progressive RNA unwinding and potentially it reduces/ slows down the unwinding rate further supported by kinetics experiments measuring the actual RNA unwinding.

We then studied differences in other missense variants. We first analyzed 15 somatic mutants at residues located distantly and proximally to R525H and predicted the impact of each of them on DDX41 function. In 2 missense, the amino acid changes altered the residues binding ATP with lesser impact than R525H, favorably (Q208E) and unfavorably (E345D). 3 variants (D446A, H463L, R471Q) abolished the RNA grab, 4 affected allosteric regulation (P321L, M378I, L548H, L552R), 2 affected conformational changes (A225D, G586R) while in 2 variants the impact could not be predicted (K482R, A488T). Of note is that the variant located in the closest triplet to R525H (G524R) seemed to exhibit different functional effects compared to R525H.

A similar in silico analysis was conducted by modeling 5 assumed non pathogenic DDX41^GL (M155I, Y151H, K184R, R164Q, I169V) and predicted that such spots might alter the binding with other protein partners possibly expanding the function of DDX41 which to date is restricted to pre mRNA and innate immunity, both failing to explain the smoldering phenotype of cases lacking the biallelic configuration.

Our study suggests that we can’t paint all mutations with the same brush and their functional impact is likely diverse. Novel insights in mutant DDX41 structure further clarify the impact of single variants which may exhibit functional effects distinct from those of the canonical configuration. In our presentation we will show alternate modes of biallelic somatic vs germline hits including loss of TSG function vs somatic gene rescue contingent on the lesions involved.