Algorithmic assessment of missense mutation severity in the Von-Hippel Lindau protein

Abstract

Von Hippel-Lindau disease (VHL) is an autosomal dominant rare disease that causes the formation of angiogenic tumors. When functional, pVHL acts as an E3 ubiquitin ligase that negatively regulates hypoxia inducible factor (HIF). Genetic mutations that perturb the structure of pVHL result in dysregulation of HIF, causing a wide array of tumor pathologies including retinal angioma, pheochromocytoma, central nervous system hemangioblastoma, and clear cell renal carcinoma. These VHL-related cancers occur throughout the lifetime of the patient, requiring frequent intervention procedures, such as surgery, to remove the tumors. Although VHL is classified as a rare disease (1 in 39,000 to 1 in 91,000 affected) there is a large heterogeneity in genetic mutations listed for observed pathologies. Understanding how these specific mutations correlate with the myriad of observed pathologies for VHL could provide clinicians insight into the potential severity and onset of disease. Using a select set of 285 ClinVar mutations in VHL, we developed a multiparametric scoring algorithm to evaluate the overall clinical severity of missense mutations in pVHL. The mutations were assessed according to eight weighted parameters as a comprehensive evaluation of protein misfolding and malfunction. Higher mutation scores were strongly associated with pathogenicity. Our approach establishes a novel in silico method by which VHL-specific mutations can be assessed for their severity and effect on the biophysical functions of the VHL protein.

Fig 1. Score distributions for the VHL missense mutations used in the multiparametric approach.

A. A fitted Gaussian distribution (red) of scores for all 1379 possible missense mutations from a SNP in VHL B. A fitted Gaussian distribution (red) of scores for the 285 ClinVar missense mutations used in this study. C. Relationship between the All Mutation data set and the ClinVar data set. D. Mutation algorithm scores plotted according to their ClinVar pathogenicity. Each dot is a mutation. All error bars represent the standard deviation. A * represents a P < .05 according to a Kolmogorov Smirnov test. All statistics done in Graph Pad Prizm.

Fig 2. Association of missense mutation algorithm score to its spatial distribution on pVHL.

A. Algorithm scores for mutations according to secondary structure. B. pVHL domain C. or pVHL binding interfaces. Significance was determined using an ANOVA or Kruskal-Wallis test and followed up with Tukey HSD or Dunn’s MCT as appropriate. Error bars represent the standard deviation. * represents a significant difference with a p < .05. D. Algorithm Score for mutations according to their depth within the structure of VHL. Each dot is a mutation. Error bars represent the standard deviation. * represents a significant difference with a p < .05 as determined by Student’s t-test. All statistics were done using GraphPad Prizm.

Fig 3

VHL missense mutations algorithm scores associated with onset of the VHL related cancers: A. pheochromocytoma (PCC) B. central nervous system hemangioblastoma (CHB) C. retinal angioma (RA) and D. clear cell renal carcinoma (ccRCC). Each dot is the average age of onset for a missense mutation. Error bars represent the standard deviation. P-values were determined using Student’s t-test.