Circulating Cardiovascular Proteomic Associations With Genetics and Disease

Abstract

Background: The analysis of the circulating proteome can identify translational modifiers and biomarkers of disease expressivity and severity at a given time point. Here, we explore the relationships between protein measures implicated in cardiovascular disease and whether they mediate causal relationships between cardiovascular risk factors and disease development.

Methods: To understand the relationships between circulating biomarkers and genetic variants, medications, anthropometric traits, lifestyle factors, imaging-derived measures, and diagnoses of cardiovascular disease, we undertook in-depth analyses of measures of 9 plasma proteins with a priori roles in genetic and structural cardiovascular disease or treatment pathways (ACE2 [angiotensin-converting enzyme 2], ACTA2 [actin alpha 2], ACTN4 [actinin alpha 4], BAG3 [BAG cochaperone 3], BNP [B-type natriuretic peptide], CDKN1A [cyclin-dependent kinase inhibitor 1A], NOTCH1 [neurogenic locus notch homolog protein 1], NT-proBNP [N-terminal pro-B-type natriuretic peptide], and TNNI3 [troponin I]) from the Pharma Proteomics Project of the UK Biobank cohort (over 45 000 participants sampled at recruitment).

Results: We identified significant variability in circulating proteins with age, sex, ancestry, alcohol intake, smoking, and medication intake. Phenome-wide association studies highlighted the range of cardiovascular clinical features with relationships to protein levels. Genome-wide genetic association studies identified variants near GCKR, APOE, and SERPINA1, that modified multiple circulating protein levels (BAG3, CDKN1A, and NOTCH1). NT-proBNP and BNP levels associated with variants in BAG3. ACE2 levels were increased with a diagnosis of hypertension or diabetes, particularly in females, and were influenced by variants in genes associated with diabetes (HNF1A and HNF4A). Two-sample Mendelian randomization identified ACE2 as protective for systolic blood pressure and type-2 diabetes.

Conclusions: From a panel of circulating proteins, the results from this observational study provide evidence that ACE2 is causally protective for hypertension and diabetes. This suggests that ACE2 treatment may provide additional protection from these cardiovascular diseases. This study provides an improved understanding of the circulating pathways depicting cardiovascular disease dynamics.

Keywords: angiotensins; diabetes; genetic variation; heart failure; hypertension; proteomics.

Figure 1.

Study flow chart and protein summary. A summary of the main analysis steps and data available for the analysis of 9 plasma proteins and the genetic and outcome associations. HES indicates hospital episode statistics; MRI, magnetic resonance imaging; WES, whole exome sequencing; and SR, self-reported.

Figure 2.

The relationships between the proteins and ancestry. The plots depict the significant differences in protein levels across self-reported ancestries. Participants of self-reported African or Caribbean ancestry had increased average ACE2 (angiotensin-converting enzyme 2) and decreased ACTA2 (actin alpha 2), ACTN4 (actinin alpha 4), BAG3 (BAG cochaperone 3), NT-proBNP (N-terminal pro-B-type natriuretic peptide), and BNP (B-type natriuretic peptide), compared with British ancestry. Participants with Chinese ancestry had decreased average ACTA2, BAG3, CDKN1A (cyclin-dependent kinase inhibitor 1A), and NT-proBNP, and participants with Indian ancestry had increased average BAG3, CDKN1A, and NOTCH1 (neurogenic locus notch homolog protein 1), compared with British ancestry. The significance of differences in means as derived by the Student t test is denoted as stars compared with British ancestry. The y axis units are Olink arbitrary units in log₂ scale. The sample sizes were as follows (African, n=584; British, n=40228; Caribbean, n=434; Chinese, n=131; Indian, n=495; Irish, n=1200; other, n=2736; and Pakistani, n=143). TNNI3 indicates troponin I.

Figure 3.

The relationships between the proteins and medication intake. The plots depict the proteins measured at recruitment that were significantly increased with medication intake reported only at recruitment (current), or the imaging visit on average 8 years later (future), or reported during both visits (current [long-term]). The significance of differences in means as derived by the Student t test is denoted by stars compared with no report of the medication (never). The y axis units are standardized residuals after adjustment for covariates. The data only includes those with proteomics who attended the imaging visit (n=5324). ACE2 indicates angiotensin-converting enzyme 2; BNP, B-type natriuretic peptide; NOTCH1, neurogenic locus notch homolog protein 1; and NT-proBNP, N-terminal pro-B-type natriuretic peptide.

Figure 4.

Phenome-wide association study results of the plasma protein levels with selected circulatory disorders. Phenotypes as phecodes are described on the y axis and the protein traits on the x axis. Each point denotes a significant phenome-wide association study (PheWAS) association with a Bonferroni correction for the number of analyzed phecodes. The shape and color denote the direction of effect and the odds ratio. Only the most significant associations with selected, nonredundant phenotypes of the circulatory disorder category are presented for clarity. See Table S3 for the full PheWAS results. As no negative direction of association was identified, points do not have the shape of an inverted triangle. ACE2 indicates angiotensin-converting enzyme 2; ACTA2, actin alpha 2; BAG3, BAG cochaperone 3; BNP, B-type natriuretic peptide; CDKN1A, cyclin-dependent kinase inhibitor 1A; NOTCH1, neurogenic locus notch homolog protein 1; NT-proBNP, N-terminal pro-B-type natriuretic peptide; and TNNI3, troponin I.

Figure 5.

Significant genome-wide association study results. The Manhattan plots present the genome-wide association study (GWAS) significant SNPs for the 9 protein levels. The prioritized gene is noted for the significant loci identified. A subset of gene labels for ACE2 (angiotensin-converting enzyme 2), CDKN1A (cyclin-dependent kinase inhibitor 1A), and NT-proBNP (N-terminal pro-B-type natriuretic peptide) has been selected to allow for presentation. The y axis is cut at a minimum of 5×10⁻⁸. Please see Table S6 for the full GWAS results. ACTA2 indicates actin alpha 2; BAG3, BAG cochaperone 3; BNP, B-type natriuretic peptide; CDKN1A, cyclin-dependent kinase inhibitor 1A; NOTCH1, neurogenic locus notch homolog protein 1; NT-proBNP, N-terminal pro-B-type natriuretic peptide; and TNNI3, troponin I

Figure 6.

Increasing levels of NT-proBNP (N-terminal pro-B-type natriuretic peptide) and ACE2 (angiotensin-converting enzyme 2) are observed with cardiovascular disease diagnoses. Increasing levels of NT-proBNP (N-terminal pro-B-type natriuretic peptide) are observed in participants diagnosed with incident heart failure (HF), atrial fibrillation (Afib), and myocardial infarction (MI). Increasing levels of ACE2 (angiotensin-converting enzyme 2) are observed in participants diagnosed with incident hypertension and type-2 diabetes. A, The figure shows the sequential increase in mean NT-proBNP with overt diseases and other modifiers influencing the protein’s levels. This includes a HF diagnosis alongside sex, age at recruitment, a diagnosis of cardiomyopathy (CM) or Afib, and carriers of pathogenic CM-associated variants. The Student t test was used to compare with the no HF diagnosis reference group. The groups contained the following sample sizes, respectively: no HF diagnosis, 44 050; female groups, 103, 286, 12, 33, 234, 2; male groups, 202, 496, 7, 42, 470, 14. NT-proBNP units are Olink’s arbitrary unit on log₂ scale. The forest plots (B–F) of Cox proportional hazards regression models assessed death or diagnosis from recruitment, with those diagnosed before recruitment excluded. Sex (increasing risk is male), European ancestry (increasing risk is European), and age at recruitment (incremental risk per year lived) were added to this multivariable analysis for comparison. Forest plots are presented for deciles of NT-proBNP levels with incident (B) HF, (C) Afib, (D) MI, from recruitment, and ACE2 levels by decile with incident (E) hypertension and (F) type-2 diabetes, from recruitment. Plp indicates P/LP variant carrier.

Figure 7.

Evidence of a causal relationship between systolic blood pressure, type-2 diabetes, and ACE2 (angiotensin-converting enzyme 2). Increased circulating ACE2 can decrease blood pressure through the creation of vasodilators and is causally associated with type-2 diabetes. A genetic predisposition for decreased systolic blood pressure is associated with increased ACE2. A, Mendelian randomization (MR) genetic determination model of systolic blood pressure (mm Hg decrease), genetic instruments as exposures for the ACE2 outcome. Two-sample MR was undertaken with ACE2 (using the GWAS [genome-wide association study] results) and decreased systolic blood pressure (from GWAS summary statistics of published data). B, MR genetic determination model of type-2 diabetes genetic instruments as exposures for the ACE2 outcome. Two-sample MR was undertaken with ACE2 (using the GWAS results) and type-2 diabetes (from GWAS summary statistics of published data). C, MR genetic determination model of the ACE2 genetic instrument as an exposure for type-2 diabetes. See Table S8 and Figures S3 and S4 for further details.