Research reveals potential drug targets for heart disease and stroke, using a combination of genetic and observational data to separate causal from non-causal proteins.
Letter: Measured and genetically predicted protein levels and cardiovascular diseases in UK Biobank and China Kadoorie Biobank. Image Credit: Lightspring / Shutterstock
In a recent study published in the journal Nature Cardiovascular Research, researchers assessed the association between 2,919 distinct plasma proteins and cardiovascular disease (CVD) outcomes in the UK Biobank (UKB) and China Kadoorie Biobank (CKB), identifying potential causal proteins for novel drug targets.
Background
Many studies have measured plasma protein levels in individuals with CVD outcomes. However, due to the interrelated nature of proteins, observational studies often struggle to pinpoint those with true causal relevance.
One approach to address this is through Mendelian randomization (MR) and genetic analyses, investigating whether genetic loci regulating protein levels are also associated with CVD outcomes.
Previous research has found hundreds of proteins linked to myocardial infarction (MI) (a heart attack caused by blocked blood flow to the heart), ischemic stroke (IS) (a stroke caused by a blockage in blood vessels supplying the brain), and heart failure (HF) (a condition where the heart cannot pump blood efficiently), with a limited number showing evidence of causality.
Further research is needed to better understand these causal relationships and discover new therapeutic targets for CVD treatment.
About the study
The UKB is a large, multicenter, prospective cohort study conducted across the UK between 2006 and 2010, involving over 500,000 individuals aged 40-69. Data from 52,164 participants with valid proteomics data were included. The North West Multicentre Research Ethics Committee (UK) and the Swedish Ethical Review Authority granted ethical approval, with participants providing written informed consent. Baseline measurements, including plasma levels of glucose, cholesterol, and creatinine, were collected alongside blood pressure and estimated glomerular filtration rate (eGFR).
Participants’ ethnicity was categorized, and socioeconomic status was assessed using the Townsend social deprivation index. Smoking status was recorded as never, previous, or current smokers. Disease outcomes, including MI, IS, and HF, were classified using the International Classification of Diseases (ICD-10) codes. Plasma levels of 2,923 proteins were initially measured using the Olink EXPLORE assay, with 2,919 passing quality control.
The CKB study included 512,000 Chinese adults recruited between 2004 and 2008. Data were collected on smoking, medical history, and education. Plasma proteins were measured using the same analytical platform as in UKB. Key findings from UKB were successfully replicated in CKB, strengthening the study’s conclusions.
Study results
In the UKB, 126 proteins were found to be significantly associated with all three CVD outcomes. Of these, 118 were associated with at least one CVD outcome in the replication phase conducted in CKB. Among these, 87 proteins were linked to more than one CVD, while 31 proteins were related to all three CVD outcomes (MI, IS, and HF) at a significance level of P < 0.05.
Using Mendelian randomization (MR) analysis, genetically predicted levels of 33 proteins were linked to coronary heart disease (CHD), with the top proteins being Lipoprotein(a) (LPA), Cadherin EGF LAG Seven-Pass G-Type Receptor 2 (CELSR2), Apolipoprotein E (APOE), Feline Sarcoma Oncogene (FES), and Vesicle-Associated Membrane Protein 5 (VAMP5). Four proteins were linked to IS, while 18 proteins were associated with HF, with LPA, CELSR2, and Fibroblast Growth Factor 5 (FGF5) emerging as key players in CHD and HF. Sensitivity analyses using multiple cis-SNPs for MR analysis produced results consistent with the initial findings, enhancing the robustness of these associations.
Colocalization analysis provided further evidence, showing that 10 proteins had strong shared causal variants with CVD outcomes. Notably, CELSR2 and FGF5 were associated with both CHD and HF, while FURIN was linked to both CHD and stroke. These proteins, especially FURIN, were identified as promising targets for future drug development, supported by additional findings in subclinical CVD markers.
Interestingly, observational analyses demonstrated a weaker relationship between protein levels and CVD outcomes compared to MR results. Only 6% of proteins associated with CVD outcomes in the observational analyses were found to have causal relationships in the MR analysis, suggesting that many observational associations were non-causal. This discrepancy is likely due to biases such as residual confounding and reverse causality in observational studies.
Proteins like FGF5, Protein C Receptor (PROCR), and FURIN showed consistent evidence of causal involvement in CVD outcomes across both observational and genetic analyses, highlighting their potential as therapeutic targets. FGF5 was implicated in hypertension and carotid artery distensibility, while FURIN, a protease involved in protein activation, was associated with both MI and IS.
Conclusions
To summarize, the study identified numerous proteins associated with CVD outcomes, but genetic analyses revealed that only a small subset showed evidence of causality. Key proteins such as FGF5, PROCR, and FURIN emerged as potential targets for CVD prevention and treatment.
The findings highlight the importance of distinguishing between observational and causal associations in protein studies. Although many observational associations were non-causal, the use of Mendelian randomization and colocalization analyses provided a clearer understanding of which proteins may have direct roles in CVD pathogenesis, offering promising directions for future drug development.