Proteomics of left ventricular structure in the Multi-Ethnic Study of Atherosclerosis

Abstract

Aims: Proteomic profiling offers an expansive approach to biomarker discovery and mechanistic hypothesis generation for LV remodelling, a critical component of heart failure (HF). We sought to identify plasma proteins cross-sectionally associated with left ventricular (LV) size and geometry in a diverse population-based cohort without known cardiovascular disease (CVD).

Methods and results: Among participants of the Multi-Ethnic Study of Atherosclerosis (MESA), we quantified plasma abundances of 1305 proteins using an aptamer-based platform at exam 1 (2000-2002) and exam 5 (2010-2011) and assessed LV structure by cardiac magnetic resonance (CMR) at the same time points. We used multivariable linear regression with robust variance to assess cross-sectional associations between plasma protein abundances and LV structural characteristics at exam 1, reproduced findings in later-life at exam 5, and explored relationships of associated proteins using annotated enrichment analysis. We studied 763 participants (mean age 60 ± 10 years at exam 1; 53% female; 19% Black race; 31% Hispanic ethnicity). Following adjustment for renal function and traditional CVD risk factors, plasma levels of 3 proteins were associated with LV mass index at both time points with the same directionality (FDR < 0.05): leptin (LEP), renin (REN), and cathepsin-D (CTSD); 20 with LV end-diastolic volume index: LEP, NT-proBNP, histone-lysine N-methyltransferase (EHMT2), chordin-like protein 1 (CHRDL1), tumour necrosis factor-inducible gene 6 protein (TNFAIP6), NT-3 growth factor receptor (NTRK3), c5a anaphylatoxin (C5), neurogenic locus notch homologue protein 3 (NOTCH3), ephrin-B2 (EFNB2), osteomodulin (OMD), contactin-4 (CNTN4), gelsolin (GSN), stromal cell-derived factor 1 (CXCL12), calcineurin subunit B type 1 (PPP3R1), insulin-like growth factor 1 receptor (IGF1R), bone sialoprotein 2 (IBSP), interleukin-11 (IL-11), follistatin-related protein 1 (FSTL1), periostin (POSTN), and biglycan (BGN); and 4 with LV mass-to-volume ratio: RGM domain family member B (RGMB), transforming growth factor beta receptor type 3 (TGFBR3), ephrin-A2 (EFNA2), and cell adhesion molecule 3 (CADM3). Functional annotation implicated regulation of the PI3K-Akt pathway, bone morphogenic protein signalling, and cGMP-mediated signalling.

Conclusions: We report proteomic profiling of LV size and geometry, which identified novel associations and reinforced previous findings on biomarker candidates for LV remodelling and HF. If validated, these proteins may help refine risk prediction and identify novel therapeutic targets for HF.

Keywords: Biomarkers; Cardiovascular magnetic resonance imaging; Left ventricular remodelling; Proteomics.

FIGURE 1

Plasma protein abundances associated with indexed left ventricular mass in the Multi‐Ethnic Study of Atherosclerosis. (A) Volcano plot of mean differences in left ventricular mass index (LVMi) per standard deviation (SD) increment in plasma protein levels versus −log₁₀ false discovery rate (FDR) estimated at exam 1 using linear regression with robust variance (n = 763). Blue indicates proteins associated at FDR < 0.10 adjusting for clinic center, age, sex, race/ethnicity, systolic blood pressure, blood pressure‐lowery therapy, total: HDL‐cholesterol ratio, lipid‐lowering therapy, diabetes, body mass index, physical activity, smoking, alcohol use, and estimated glomerular filtration rate. (B) Scatterplot of association point estimates, MESA exam 1 versus MESA exam 5. Proteins depicted (n = 4) were those associated with LVMi at exam 1 and depicted in red are those with reproduced associations at exam 5 (FDR < 0.05), estimated using the same modelling approach and adjustment variables noted in (A).

FIGURE 2

Plasma protein abundances associated with indexed left ventricular volume in the Multi‐Ethnic Study of Atherosclerosis. (A) Volcano plot of mean differences in left ventricular end‐diastolic volume index (LVEDVi) per standard deviation (SD) increment in plasma protein levels versus −log₁₀ false discovery rate (FDR) estimated at exam 1 using linear regression with robust variance (n = 763). Blue indicates proteins associated at FDR < 0.10 adjusting for clinic center, age, sex, race/ethnicity, systolic blood pressure, blood pressure‐lowery therapy, total: HDL‐cholesterol ratio, lipid‐lowering therapy, diabetes, body mass index, physical activity, smoking, alcohol use, and estimated glomerular filtration rate. (B) Scatterplot of association point estimates, MESA exam 1 versus MESA exam 5. Proteins depicted (n = 132) were those associated with LVEDVi at exam 1 with and depicted in red are those with reproduced associations at exam 5 (FDR < 0.05), estimated using the same modelling approach and adjustment variables noted in (A).

FIGURE 3

Plasma protein abundances associated with left ventricular geometry in the Multi‐Ethnic Study of Atherosclerosis. (A) Volcano plot of mean differences in left ventricular mass‐to‐volume ratio (LVM/V, g/mL) per standard deviation (SD) increment in plasma protein levels versus −log₁₀ false discovery rate (FDR) estimated at exam 1 using linear regression with robust variance (n = 763). Blue indicates proteins associated at FDR < 0.10 adjusting for clinic center, age, sex, race/ethnicity, systolic blood pressure, blood pressure‐lowery therapy, total: HDL‐cholesterol ratio, lipid‐lowering therapy, diabetes, body mass index, physical activity, smoking, alcohol use, and estimated glomerular filtration rate. (B) Scatterplot of association point estimates, MESA exam 1 versus MESA exam 5. Proteins depicted (n = 9) were those associated with LVM/V at exam 1 and depicted in red are those with reproduced associations at exam 5 (FDR < 0.05), estimated using the same modelling approach and adjustment variables noted in (A).