Linking genotypes database with locus-specific database and genotype-phenotype correlation in phenylketonuria

Abstract

The wide range of metabolic phenotypes in phenylketonuria is due to a large number of variants causing variable impairment in phenylalanine hydroxylase function. A total of 834 phenylalanine hydroxylase gene variants from the locus-specific database PAHvdb and genotypes of 4181 phenylketonuria patients from the BIOPKU database were characterized using FoldX, SIFT Blink, Polyphen-2 and SNPs3D algorithms. Obtained data was correlated with residual enzyme activity, patients' phenotype and tetrahydrobiopterin responsiveness. A descriptive analysis of both databases was compiled and an interactive viewer in PAHvdb database was implemented for structure visualization of missense variants. We found a quantitative relationship between phenylalanine hydroxylase protein stability and enzyme activity (r(s) = 0.479), between protein stability and allelic phenotype (r(s) = -0.458), as well as between enzyme activity and allelic phenotype (r(s) = 0.799). Enzyme stability algorithms (FoldX and SNPs3D), allelic phenotype and enzyme activity were most powerful to predict patients' phenotype and tetrahydrobiopterin response. Phenotype prediction was most accurate in deleterious genotypes (≈ 100%), followed by homozygous (92.9%), hemizygous (94.8%), and compound heterozygous genotypes (77.9%), while tetrahydrobiopterin response was correctly predicted in 71.0% of all cases. To our knowledge this is the largest study using algorithms for the prediction of patients' phenotype and tetrahydrobiopterin responsiveness in phenylketonuria patients, using data from the locus-specific and genotypes database.

Figure 1

Distribution of mutations tabulated in the PAHvdb according to gene region, affected protein domains, amino acid boundaries of the exons and BH₄ cofactor binding regions (CBR). Red bars, exons; blue bars, introns.

Figure 2

Relationship between different PKU phenotypes and (a) mean FoldX value (r_s (70)=−0.446, P<0.001); (b) mean SNPs3D value (r_s (89)=0.479, P<0.001); (c) mean PolyPhen-2 value (no relationship); and (d) mean SIFT values (no relationship). Error bars represent the 95% confidence interval.

Figure 3

Percentage of BH₄ responders, slow responders and non-responders within PKU phenotype groups.

Figure 4

(a) Percentage of BH₄ responders, slow responders and non-responders in PKU genotypes with different domain combinations; (b) percentage of MHP, mild PKU and classic PKU patients in genotypes with different domain combinations.