Immune Aging is an Independent Risk Factor for Cardiovascular Disease

Figures

• 

  Extended Data Figure 1:. PBHS COX modeling

• 

  Figure 1.. IMM-AGE is a quantifiable metric of cardiovascular morbidity and mortality risk. (A) A schematic outlining the development of a molecular-clinical translational framework from a high-resolution molecular multi-omic longitudinal immune-aging study to multiple validated clinical fine-grained clinical associations using large cross-sectional studies. (B) Validation of the ratio-based IMM-AGE signature across blood fractions and transcriptomic platforms. IMM-AGE gene expression ratios signature was validated across whole blood and PBMC samples profiled by both Affymetrix and RNA-seq, with strong correlations to mass cytometry (CyTOF)-derived IMM-AGE in all four settings (R = 0.753–0.824; all P < 0.001). (C) Independent prognostic value of IMM-AGE in the Framingham Heart Study. Forest plot of hazard ratios from a multivariable Cox model in 2,131 participants. IMM-AGE remained independently associated with mortality following adjustment for conventional cardiovascular risk factors (HR = 1.23, p = 7×10–3). Abbreviations: CVD - History of cardiovascular disease; BP Class [High]- sysBP≥130 OR diaBP≥85. (D) Kaplan–Meier curves for overall survival in the Framingham cohort stratified by…

• 

  Figure 2.. IMM-AGE captures residual cardiovascular risk in the UK Biobank cohort. (A) Decomposition of IMM-AGE variance across clinical domains in 43,608 UK Biobank participants. Demographic, inflammatory, metabolic, and disease-burden factors jointly explain 43.1% of total variance (R2 = 0.431), with age (10.7%), BMI (10.4%), CRP (7.8%), and eGFR (5.1%) as the largest contributors, while 56.9% remains unexplained. (B) IMM-AGE is elevated in individuals who subsequently develop cardiovascular events. Left, baseline IMM-AGE levels in participants who developed incident cardiovascular events compared with those who remained event-free, showing significantly higher values among future cases (two-sided test, p < 2.2 × 10−16; Cohen’s d = 0.645). Right, multivariable-adjusted IMM-AGE remained significantly elevated in individuals who subsequently developed cardiovascular events, indicating that the association was not explained by conventional clinical risk factors (P < 2.2×10−16; Cohen’s d = 0.231). (C) IMM-AGE is an independent risk factor for CV morbidity. Multivariable Cox proportional hazards model showing that higher baseline IMM-AGE is independently associated with incident cardiovascular…

• 

  Figure 3.. A scalable framework for monitoring immune aging in standard clinical settings. (A) Identification of optimized marker panels for clinical translation. Schematic illustrating the computational framework used to select low-dimensional cell combinations that preserve the IMM-AGE signal. We identified optimized configurations that maintain the original ordering on the IMM-AGE trajectory and predict mortality in the Framingham Heart Study. (B) Selection of high-fidelity marker panels based on trajectory and mortality prediction. Scatter plot ranking all simulated eight-marker panels by their ability to recapture the original sample trajectory (Kendall’s τ x-axis) and their prognostic performance in mortality prediction within the Framingham dataset (C-index, y-axis). Selected combinations (colored red), and original CyTOF combination (blue). Dashed lines highlight the selected cutoff for each axis. (C) Preservation of the IMM-AGE signal in the discovery cohort. Correlation analysis between the original full-panel IMM-AGE scores (x-axis) and the optimized reduced-panel estimates (y-axis) in the SELA discovery cohort. Pearson R = 0.96, p < 2.2×10−16. (D) Selected IMM-AGE cell…

• 

  Figure 4.. IMM-AGE is an independent predictor of post-TAVR ventricular adaptation and mortality. (A) Schematic of the prospective study design involving 96 patients undergoing transcatheter aortic valve replacement (TAVR). Assessments were conducted at baseline (pre-procedure), 1 month, and 1 year post-procedure, incorporating clinical history, echocardiography, and IMM-AGE scoring via optimized flow cytometry. (B) Association between 1-month cardiac phenotype and 1-year mortality. Sankey diagram illustrating the progression from baseline echocardiography to 1-month echocardiographic clusters to 1-year survival outcomes. (C) Association of baseline and 1-month echocardiographic phenotypes with 1-year mortality. Dot plot showing the significance of the association between echocardiographic phenotype and 1-year mortality at each time point. Only the 1-month phenotype was significant (permutation P = 0.044). (D) Baseline IMM-AGE scores stratified by 1-month cardiac response. Boxplot comparing pre-procedural IMM-AGE scores (y-axis) between patients who subsequently developed adaptive vs. maladaptive 1-month phenotypes (x-axis). Higher baseline IMM-AGE was significantly associated with…

• 

  Figure 5.. IMM-AGE independently predicts major adverse cardiovascular events among clinically matched individuals. (A) Retrospective study design and multidimensional phenotyping of the PBHS cohort. Schematic illustrating the longitudinal workflow for 2,234 participants in the Project Baseline Health Study (BHS). Participants were followed for a median of 1,463 days to record incident major adverse cardiovascular events (MACE; n = 122). (B) Distribution of IMM-AGE across categories of cardiovascular burden and clinical risk in the Baseline Health Study. IMM-AGE is shown across coronary artery calcium (CAC) categories, 10-year PREVENT risk groups, echocardiographic E/e′ groups and ASE-defined diastolic dysfunction grades. P values for pairwise t test comparisons are indicated in the plot. Boxes show the interquartile range, center lines indicate medians and points represent individual participants. (C) Network-based similarity matching of cardiovascular event cases and controls. Schematic of the high-dimensional matching strategy used to define clinically similar case-control groups based on 13 clinical and laboratory covariates. Each case was matched to its ten nearest controls, and…