Genetics
Proteomic screen of 48,859 UK Biobank participants finds ITIH3 and CBLN4 as a non-heritable Lp(a) axis linked to atherosclerotic events (medRxiv preprint 2025)
Original title: Proteomic analysis identifies lipoprotein(a)-associated proteins linked to incident atherosclerotic cardiovascular disease events
This study screened 1,459 circulating proteins in 48,859 UK Biobank participants (mean age 57, 93.9% European, 53.8% male, median follow-up 8.9 years) for association with Lp(a), replicating findings in 9,416 ARIC study participants on a separate proteomic platform. After FDR correction, 164 proteins were significantly associated with Lp(a), of which 10 replicated in ARIC; most were independent of an LPA genetic risk score, marking a non-heritable Lp(a) axis. Five replicated proteins also predicted incident atherosclerotic disease: ITIH3 was positively associated with coronary disease (HR 1.13), peripheral artery disease (HR 1.42), major limb events (HR 1.65), carotid stenosis (HR 1.45) and ischaemic stroke (HR 1.33), while CBLN4 was inversely associated with coronary disease (HR 0.88), peripheral artery disease (HR 0.78) and stroke (HR 0.72). This is a preprint, not yet peer-reviewed, but its proteomic map points to specific downstream mediators of Lp(a)-driven risk beyond genetics.
Original abstract
Background: The pathways linking lipoprotein(a) (Lp[a]) to atherosclerotic cardiovascular disease (ASCVD) are unclear. This study aimed to discover Lp(a)-associated plasma proteins and estimate their associations with incident ASCVD.
Methods: We analyzed 48,859 UK Biobank participants with measured Lp(a) and proteomic profiles, with replication in 9,416 individuals in the Atherosclerosis Risk in Communities (ARIC) study cohort utilizing a separate proteomic platform. Linear models assessed associations between Lp(a) and protein concentrations adjusted for age, sex, cigarette smoking, diabetes diagnosis, body mass index, systolic blood pressure, hypertension, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol, triglycerides, estimated glomerular filtration rate, statin prescription, and the first 10 components of genetic ancestry. Multiple testing correction was performed using the Benjamini-Hochberg FDR method (P < 0.05). We examined how the protein effect sizes from the primary analysis using the outcome of Lp(a) aligned with those for the outcomes of an LPA genetic risk score (GRS) and LDL-C. Cox proportional hazards models quantified hazard ratios (HRs) for protein associations with incident ASCVD.
Results: Participants were a mean age of 57 years (SD 8.22), 93.9% European, and 53.8% male, with median follow-up of 8.9 years (IQR 8.3-9.7). Of 1,459 circulating proteins, 164 were significantly associated with Lp(a) after FDR correction, with enrichment for lipid degradation, metabolism, and insulin secretion. In the ARIC study, 10 proteins were replicated with consistent effect estimates. Of these replicated proteins, there were no significant associations observed with an LPA GRS. Only REG4 and VWC2 showed concordant associations with LDL-C (P < 0.001), consistent with their association with Lp(a). Five proteins exhibited concordant associations with Lp(a) and incident ASCVD (ITIH3, DLL1, REG4, VWC2, CBLN4). ITIH3 was positively associated with coronary artery disease (HR 1.13, 95% CI 1.04-1.23), peripheral artery disease (HR 1.42, 95% CI 1.19-1.69), major adverse limb events (HR 1.65, 95% CI 1.14-2.40), carotid stenosis (HR 1.45, 95% CI 1.13-1.85), and ischemic stroke (HR 1.33, 95% CI 1.13-1.55). CBLN4 uniquely showed inverse associations with Lp(a) and disease: higher levels were linked to lower risk of CAD (HR 0.88, 95% CI 0.80-0.96), PAD (HR 0.78, 95% CI 0.64-0.96), and ischemic stroke (HR 0.72, 95% CI 0.60-0.85).
Conclusion: Using high-throughput proteomics, we discovered and replicated 10 proteins associated with circulating Lp(a), several of which were independent of genetically-predicted Lp(a). While Lp(a) is highly heritable, these atherogenic proteins represent a non-heritable Lp(a) axis.
Summary written by lp-a.org from the published abstract; figures as published. Page updated 17 August 2026. Methods.