CALON-FH: AlphaFold3-derived structural severity score enables domain-specific ASCVD penetrance prediction and mutation-guided treatment selection in familial hypercholesterolaemia [Abstract]

Genedy, N. 2026. CALON-FH: AlphaFold3-derived structural severity score enables domain-specific ASCVD penetrance prediction and mutation-guided treatment selection in familial hypercholesterolaemia [Abstract]. Atherosclerosis Plus 65 (Supple) , p. 2. 10.1016/j.athplu.2026.100591

PDF - Published Version Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (359kB)

Abstract

Background: Current familial hypercholesterolaemia (FH) management applies uniform treatment algorithms regardless of mutation location, despite the LDL receptor (LDLR) comprising functionally distinct domains with different mechanisms of dysfunction. The ligand-binding repeats (LA1–LA7) capture LDL particles, the EGF-A domain mediates PCSK9 binding, the beta-propeller enables receptor recycling at endosomal pH, and the cytoplasmic NPXY motif signals clathrin-mediated endocytosis. Disruption of each mechanism has distinct therapeutic implications, yet no framework exists to translate mutation location into treatment selection. We developed the Structural Severity Score (SSS) to classify LDLR mutations by mechanism of receptor dysfunction and predict domain-specific ASCVD penetrance and treatment response. Methods: We constructed the SSS by integrating 14 independent data sources across all 860 LDLR residue positions: AlphaFold3 structure prediction (pLDDT confidence), FoldX 5.0 thermodynamic stability (ΔΔG) for 359 clinical variants and saturation mutagenesis, Rosetta energy decomposition, AlphaFold3 mutant LDLR–PCSK9 complex modelling, interface distance to the LDL binding surface derived from the Reimund et al. (Nature 2025) cryo-EM structure, and Islam et al. (2024) cell-surface LDLR functional assay (234 variants, 200 positions). ClinVar pathogenicity classifications were incorporated. Clinical validation used three independent cohorts: the Wales FH Registry (n=7,253; 2,405 genetically positive; 323 unique LDLR mutations; 1,515 ASCVD events), the DRAGON3 clinical cohort (n=1,362; 424 with identified mutations), and UK Biobank whole-exome sequencing (n=426,732; 1,623 FH carriers). Kaplan–Meier survival analysis with time-to-first-ASCVD methodology was used for domain-level penetrance estimation. ASCVD prediction used logistic regression adjusted for age, sex, LDL-C, statin use, diabetes, smoking, hypertension, and tendon xanthomata. An interactive clinical atlas with 3D AlphaFold structure visualisation was deployed at https://nadergenedy.github.io/calon-fh-atlas/ Results: Domain-level SSS correlated with ASCVD penetrance across 19 LDLR domains (Spearman ρ=0.644, P=0.003), with a 5.7-fold range: ligand-binding R4 (21.4%, n=42) versus ligand-binding R3 (3.8%, n=26). SSS was significantly higher in patients with tendon xanthomata (0.343 vs 0.297, P=2.7×10-4). In 1,095 patients with longitudinal LDL-C data, treatment response varied significantly by domain: ligand-binding R4 achieved 25.6% LDL-C reduction (n=21), EGF-A 21.9% (n=31), versus cytoplasmic domain 4.1% (n=19) — consistent with the mechanistic prediction that NPXY internalisation defects are refractory to conventional lipid-lowering therapy. Twenty-eight patients on PCSK9 inhibitors spanning 14 LDLR domains and 23 unique variants achieved 52% mean LDL-C reduction, demonstrating efficacy across all mutation types through wild-type allele rescue in heterozygous FH. VUS carriers (n=232) demonstrated ASCVD rates comparable to pathogenic carriers (18.1% vs 15.6% vs 3.5% in mutation-negative). FoldX thermodynamic predictions were independently validated against the Islam et al. functional assay: variants classified as decreased binding/uptake had mean ΔΔG=3.37 kcal/mol versus 0.06 kcal/mol for undetermined variants (56-fold difference). A fully adjusted ASCVD prediction model incorporating SSS with conventional risk factors and xanthomata achieved AUC=0.85 (n=350, 39 events). Conclusions: CALON-FH demonstrates that LDLR domain architecture determines both ASCVD penetrance and treatment response in heterozygous FH. The 5.7-fold variation in domain-specific ASCVD risk and the 6.2-fold variation in treatment response (25.6% vs 4.1%) support a paradigm shift from uniform genotype-aware management to structure-informed, mutation-guided precision treatment. PCSK9 inhibitor efficacy across all 14 tested domains provides mechanistic reassurance for early use through wild-type allele protection, while the identification of cytoplasmic NPXY mutations as treatment-refractory supports early referral for LDLR-independent therapies including evinacumab.

Item Type:	Short Communication
Date Type:	Publication
Status:	Published
Schools:	Schools > Medicine
ISSN:	2667-0895
Date of First Compliant Deposit:	15 September 2026
Last Modified:	15 Sep 2026 09:30
URI:	https://orca.cardiff.ac.uk/id/eprint/189600

Actions (repository staff only)

Edit Item