Plasma proteomics identify fibroblast growth factor 5 as a biomarker for genetically proven Familial Hypercholesterolemia

Abstract

Abstract Background Subjects affected by genetically proven heterozygous Familial Hypercholesterolemia (FH) present lifelong exposure to high Low Density Lipoprotein Cholesterol ("LDL-C burden") causing premature ACVD. Nevertheless, the diagnosis of FH is still far from being optimal due to shortcomings of the clinical algorithms in detecting this form of severe ACVD since early decades of life (1) and because a significant proportion of FH present comparable LDL-C levels to that of subjects who are hypercholesterolemic but are not diagnosed as FH ("non-FH") (2). Purpose We previously demonstrated that targeted proteomics identify subjects at high Atherosclerotic Cardiovascular Diseases (ACVD) risk in the general populations beyond the known risk factors (3). Hence, we now tested whether targeted proteomics may differentiate FH subjects from the general population and from "non-FH". Methods We compared the plasma relative abundance of 264 proteins (Proximity Extension Assay; Olink) in 133 genetically proven FH with that of 586 normocholesterolemic subjects ("controls") and that of 55 non-FH (LDL-C between 190-220 mg/dL, not on statins). We trained and validated a Machine Learning (ML) boosting prediction model (4) in these cohorts to identify the protein markers representative of FH (Random Forest classifiers). Results Fibroblast Growth Factor 5 (FGF-5) resulted as the most powerful plasma protein in differentiating FH subjects from the controls, with an excellent performance (Area Under the Curve (AUC)=0.995 (0.985-1.000); sensitivity=100%; specificity=99.54%; P<0.0001). FGF-5 also emerged as the top protein in differentiating the FH subjects from non-FH (AUC= 1.000 (0.914-1.000; sensitivity=100%, specificity=100% and P<0.001). Finally, when the two groups were matched for LDL-C, FGF-5 still differentiated the FH from the non-FH subjects, (AUC=1.000 (0.914-1.000), sensitivity=100%, specificity=100%, p<0.001). Conclusions FGF-5 allows a clear-cut identification of FH versus controls and, of even more importance, from non-FH with the same LDL-C. The LDL-C burden might be responsible for our findings.