Comparison of Bioinformatics Prediction, Molecular Modeling, and Functional Analyses of FOXC1 Mutations in Patients with Axenfeld-Rieger Syndrome

Mutations in the forkhead box C1 gene (FOXC1) cause Axenfeld-Rieger syndrome (ARS). Here, we investigated the effect of four ARS missense variants on FOXC1 structure and function, and examined the predictive value of four in silico programs for all 31 FOXC1 missense variants identified to date. Molecular modeling of the FOXC1 forkhead domain predicts that c.402G> A (p.C135Y) alters FOXC1's structure. In contrast, c.378A> G (p.H128R) and c.481A> G (p.M161V) are not predicted to change FOXC1's structure. Functional analysis indicates that p.H128R reduced DNA binding, transactivation, nuclear localization, and has a longer protein half-life than normal. p.C135Y significantly disrupts FOXC1's DNA binding, transactivation, and nuclear localization. p.M161V reduces transactivation capacity without affecting other FOXC1 functions. C.1103C> A (p.T368N) is indistinguishable from wild-type FOXC1 in all tests, consistent with being a rare benign variant. Comparison of these four variants, plus 18 previously characterized FOXC1 missense variants, with predictions from four commonly used in silico bioinformatics programs indicated that sorting intolerant from tolerant (SIFT), polymorphism phenotyping (PolyPhen-2), and MutPred can sensitively identify as pathogenic only FOXC1 mutations with significant functional defects. This information was used to predict, as disease-causing, nine additional FOXC1 missense variations. Importantly, our results indicate SIFT, PolyPhen-2, and MutPred can reliably be used to predict missense variant pathogenicity for forkhead transcription factors.

Axenfeld-Rieger syndrome; FOXC1; bioinformatics programs; functional analysis; missense variants; mutation prediction.