CO082 | IN SILICO ANALYSIS THROUGH A NEW-GENERATED FACTOR VIII MODEL SUPPORTS TOLERANCE OF B DOMAIN TO MULTIPLE AMINO ACID SUBSTITUTIONS, LIMITING THE LOCAL IMPACT OF MISSENSE VARIANTS

A. Branchini¹, M.F. Testa¹, M. Pinotti¹, R. De Cristofaro^2|3 | ¹Department of Life Sciences and Biotechnology, University of Ferrara, Italy; ²Department of Translational Medicine and Surgery, Catholic University of the Sacred Heart, Roma, Italy; ³Center for Hemorrhagic and Thrombotic Diseases, Foundation University Hospital “A. Gemelli” IRCCS, Roma, Italy

Background and Aims. Missense variants represent a relevant cause of factor VIII (FVIII) deficiency in haemophilia A (HA). Intriguingly, a very few HA-associated missense variants, particularly those associated with severe phenotypes, are reported in the B domain that, among FVIII domains, remains the least characterized and structurally unresolved. Here, to provide insights into the impact of multiple amino acid substitutions in the B domain at a single-residue scale, we performed systematic in-silico analyses through a newly generated FVIII model (FVIII_IT) with the assembled B domain.
Methods. Generation of the FVIII_IT model by the I-TASSER tool and in silico analysis through DDmut and REVEL tools.
Results. The integrated inspection of HA databases indicates that HA-associated missense variants are mainly reported in FVIII A1-A2-A3-C1-C2 (hereafter referred to as A1-C2) domains, as indicated by the local (A1-A3, 77%; C1-C2, 20%) distribution in comparison with the B domain (3%). In particular, missense variants associated with severe phenotypes in the B domain are the least represented, as also highlighted by their relative frequency normalized per domain size (A1-C2, 0.36-0.49, mean 0.45; B, 0.02). To extensively and systematically dissect the impact of amino acid substitutions in the B domain, the newly generated FVIII_IT model was first validated for resolved A1-C2 domains towards reference structures by alanine scanning, then extended to the B domain, and in correlation with REVEL scores, which showed a minor impact in the B domain. Noticeably, a mutational scanning taking advantage of DDmut/REVEL at a single-residue scale on all possible single-nucleotide changes resulting in missense variants (>6,000 for the B domain) showed significant differences (p<0.0001) between B and A1-C2 domains, with a high percentage (~90%) of B-domain value localization in favorable score ranges corresponding to neutral variants. Single-nucleotide changes are also associated with the clinically relevant nonsense mutations causing premature termination codons (PTCs), whose suppression by the so-called translational readthrough (TRT) may result in the synthesis of full-length proteins bearing amino acid substitutions, with pathophysiological implications even for inhibitor susceptibility. Analysis of TRT-deriving missense variants from either possible (n=2439) or HA-associated (n=1094) PTCs revealed favorable values (p<0.0001) for B-domain variants. Noticeably, scores for B-domain missense variants remained significant even by analysis of those arising from possible (n=457) or HA (n=190) PTCs at which TRT-mediated re-insertion of the original amino acid is not predicted. This favorable output for the B domain included TRT-deriving missense variants at tryptophan residues, where missense changes are expected to be detrimental, as indicated by their high association with severe phenotypes in A1-C2 domains, whereas among the very few HA-associated B-domain missense variants, no Trp missense changes have been reported as causative of HA. Moreover, our extensive analysis was further validated in association with experimentally obtained data.
Conclusions. Altogether, these original insights, obtained by rational in silico analyses through the newly generated FVIII_IT model, strongly support tolerance of the B domain to multiple amino acid changes and support potential underestimation of HA-associated B-domain missense variants due to limited local impact of amino acid substitutions.