Heart disease is the leading cause of death in the United States, costing the US over $250 billion in the span of a single year. The age of those affected is dropping. In 2022, 1 in 5 of those who died of heart disease was under the age of 65. This article covers the use of proteogenomic approaches in cardiovascular research. This is the fourth article in our Proteogenomics Blog Series — for our previous article, visit Proteogenomics in Neuroscience Research.
Identifying Causal Factors of CVD
With the recent buzz about proteogenomics in light of the pending UK Biobank study on 600,000 individuals, it might seem that combining genetics and circulating biomarkers is a new idea. However, this is a tried-and-true method in cardiovascular disease research.
In 2017, researchers at the Karolinska Institute used Olink Protemics to identify 79 SNP-trait associations, with 41 cis-effects and 38 trans-effects in a European population of over 3,000 subjects with cardiovascular disease (CVD) risk factors. The majority of the factors identified were cis-acting, and 27 proteins explained more than 2% of the variability in protein levels.
In a more recent study combining GWAS and Olink Proteomics, researchers have identified a causal factor in CVD risk: COL6A3 C-terminus (aka endotrophin). Scientists employed a rigorous multi-step design to identify the causal effect of BMI levels on nearly 5,000 plasma proteins. The analysis included replication in a separate cohort and wet bench testing via epigenomics and single-cell sequencing.
Using SomaScan aptamer technology, researchers were able to specifically assay the level of the N- and C-termini of COL6A3. It was clear that elevated levels of this cis-pQTL’s C-terminus/endotrophin impart a significantly increased risk of CVD in those with high BMIs. These exciting results have produced multiple potential new therapeutic targets, the most noteworthy being endotrophin.
Global Proteogenomic Studies Identify Novel Risk Factors
Cardiovascular disease is not a uniquely American problem. Globally, it takes nearly 18 million lives each year. The aforementioned studies are all based on European populations, which can miss targets that impact over 85% of the world’s population.
China has joined the US with CVD as its top killer. In 2019, CVD accounted for ~45% of all deaths in China. The Chinese population has a high percentage of smoking and an obesity crisis, both of which are strong contributors to poor cardiovascular health. Similar in format to the UK Biobank, the China Kadoorie Biobank (CKB) has over 500,000 participants. It includes GWAS, clinical metadata, and combined SomaScan and Olink assays for plasma proteins on nearly 4,000 subjects.
Scientists from China and the UK collaborated to conduct a multi-aim study to compare the two proteomics platforms, uncover cis-pQTLs utilizing the GWAS data, and identify the performance of proteins as predictors of both CVD and ischemic heart disease. 859 proteins shared associations with BMI and followed the same directionality on both Olink and SomaScan. For example, the volcano plots at the top of Figure 1 clearly show GHR, FURIN, and WFDC (a) when elevated are associated with high BMI and increased risk of IHD (d).
Figure 1. Comparison of number of proteins significantly associated with BMI and risk of incident IHD and their effect sizes between Olink and SomaScan platforms. Analyses were conducted for 1694 matched proteins among 3976 participants, including 1951 IHD cases and 2025 subcohort participants. a Associations between protein levels and BMI, with coloured dots indicating significant associations. b Comparison of effect sizes (beta coefficients) for associations with BMI, with darker dots indicating significant associations and error bars indicating 95% confidence intervals. c Observational correlations between Olink and SomaScan, with shaded areas indicating shared associations for BMI. d Associations between protein levels and IHD. e Comparison of effect sizes (beta coefficients) for associations with IHD, with darker dots indicating significant associations and error bars indicating 95% confidence intervals. f Observational correlations between Olink and SomaScan, with shaded areas indicating shared associations for IHD. Linear regression was used to test associations between protein levels and BMI, adjusted for age, age2, sex, ambient temperature (at sample collection) and its square, time since last meal and its square, plate ID, and case-subcohort ascertainment. Cox regression was used to test associations between protein levels and IHD, stratified by sex and region and adjusted for age, age2, fasting time and its square, ambient temperature and its square, plate ID, education, smoking, alcohol consumption, physical activity, systolic blood pressure, type 2 diabetes, ApoB/ApoA, and BMI. P values were corrected for multiple testing with a false discovery rate = 0.05 (two-sided) within each platform to define significant associations. BMI body mass index, IHD ischaemic heart disease.
This study was able to identify pQTLs in East Asian populations that have very low minor allele frequencies in European populations. For example, the cis-pQTL sentinel variant (rs76863441) for PLA2G7 identified in both platforms has a MAF of 0.06 in East Asians but <0.001 in other populations. Alleles with very low prevalence in European populations were also captured by both platforms for PCSK9 and ALDH2.
It is worth noting that combining the overlapping proteins from both platforms did not improve the risk model’s performance. As the plex of these technologies increase, it may be possible to run a single platform and address multiple aims, including phenotypic association and risk assessment.
Identifying AMI Biomarkers with Proteomics
Our next case study’s goal is to identify protein biomarkers for acute myocardial infarction (AMI) using Olink Proteomics. The initial cohorts included 20 AMI patients and 10 healthy controls. For validation (replication) of the results, samples from an additional 125 AMI subjects and 120 healthy controls were collected and tested by ELISA.
32 circulating proteins showed significant differential expression between AMI and healthy controls. Of those, five (PCOLCE, FCN2, REG1A, DEFA1, and CRTAC1) demonstrated strong associations with clinical indicators and were significantly upregulated in AMI vs. controls using ELISA. Two (FCN2, DEFA1) were subsequently shown to have causal relationships with AMI risk using Mendelian Randomization. That said, all five proteins have potential as diagnostic indicators and may also lead to novel therapeutic targets for AMI.
Uncovering Fracture Risk with Mendelian Randomization
In our final case study fresh off the presses from Lancet, researchers utilize Mendelian Randomization (MR) to investigate the association of 274 cardio-metabolic-related proteins to rates of fracture in >12,000 subjects. 24 proteins demonstrated an association with fracture risk and were subjected to Mendelian Randomization analysis. Causality of fracture risk was supported by MR for SOST, CCDC80, NT-proBNP, and BNP.
The association for PTX3 and SPP1 with low bone density was positive in that higher levels increase fall risk and decrease grip strength. Eight of the other proteins showed an effect on body fat mass. Further research is needed to investigate these potential therapeutic and prognostic targets.
Cardiovascular disease is a major global killer. The application of proteogenomics is showing great promise in identifying diagnostic, therapeutic, and prognostic protein targets.
