Multiomic analyses uncover immunological signatures in acute and chronic coronary syndromes

The study addresses the limited understanding of systemic immunological signatures underlying acute and chronic coronary syndromes (ACS and CCS). Coronary artery disease manifests as CCS and its acute complications as ACS, notably ST-elevation myocardial infarction (STEMI) arising from plaque rupture in pre-existing CCS. Local and systemic immune responses drive atherosclerosis, thrombosis and post-infarction remodeling, yet the precise immune states in humans and their clinical implications are unclear. Single-cell omics can define immune signatures at high resolution and have characterized immune cells in atherosclerotic plaques and myocardial infarction tissue. However, translation to clinical stratification in coronary syndromes and integrative multi-omic analysis is lacking. The authors hypothesized that integrating blood-derived multiomic data with an unsupervised framework, MOFA, would uncover hidden axes of immune variation that map to disease state and clinical outcomes, thereby informing mechanisms and biomarker development.

Prior studies established single-cell profiling as a powerful approach to dissect immune states in cardiovascular disease, including single-cell analyses of human atherosclerotic plaques and spatial multi-omic maps of myocardial infarction tissue. Single-cell genomics has shown diagnostic value in oncology but has not been widely applied to coronary syndromes. Multi-omics factor analysis (MOFA) was previously introduced as a framework for unsupervised integration across omics layers to identify shared factors representing major axes of variation. Immune-system pathway knowledge bases (REACTOME, KEGG) and intercellular communication modeling (NicheNet) provide resources to interpret ligand-receptor-target relationships. Collectively, these works motivate applying integrative, factor-based multi-omics to uncover clinically relevant immune programs in ACS and CCS.

Study design and cohorts: A prospective multi-omics study was performed in a Munich cohort (n=62), including ACS (STEMI; n=28 with four longitudinal timepoints: TP1M peri-intervention, TP2M ~14 h, TP3M ~60 h, TP4M pre-discharge ~5–8 days), CCS (n=16), and non-CCS controls (n=18). Key comparisons focused on sterile ACS (acute symptom onset, immediate recanalization, no infection) versus CCS. An independent validation cohort from Groningen (n=55; ACS measured at three timepoints TP1G-TP3G and controls TP0G) was used to replicate signatures, focusing on 10x v2 chemistry samples.

Data modalities: Multi-omic measurements included (1) clinical blood tests (CK, CK-MB, troponin T, CRP, leukocytes, neutrophils); (2) single-cell RNA-seq of PBMCs with unsupervised clustering and pseudo-bulk aggregation by cell-type clusters; (3) plasma cytokine/chemokine multiplex (71-plex, 65 analytes retained); (4) plasma proteomics (DIA on Orbitrap Exploris 480; 490 proteins); and (5) neutrophil bulk transcriptomics via prime-seq (892 genes after QC). Flow cytometry provided orthogonal immune-phenotyping for baseline characterization and in vitro assays.

Preprocessing and integration: scRNA-seq data were QC-filtered, integrated with Scanorama, clustered (Leiden) and annotated; pseudo-bulk means were computed per patient-timepoint per cluster after normalization and gene filtering. Cytokines were log-transformed with pseudo-count and filtered; plasma proteomics normalized and imputed; neutrophil RNA-seq normalized, filtered for variability; clinical markers log-transformed. Feature-wise quantile normalization to standard normal quantiles was applied across 13,382 features (views: clinical, scRNA-seq cluster-specific profiles, cytokines, proteomics, neutrophils).

MOFA modeling: MOFA2 was trained with 20 factors (maxiter 50,000). Variance decomposition and factor values were examined to identify factors aligning with disease states and time courses. A second MOFA excluding clinical variables assessed dependence on clinical data. Pathway enrichment (REACTOME, KEGG immune and signal transduction) on feature weights (top features, positive/negative separately and jointly) identified enriched pathways per factor.

Cell-cell and plasma-cell communication: Using NicheNet ligand-receptor-target resources, ligands and target genes among features were identified. Spearman correlations between ligand-target pairs were computed with constraints on receptor detectability in target cell types. Analyses included contemporaneous and lagged ligand-target associations across ACS timepoints (mapping Ligand_TP1→Target_TP2, Ligand_TP2→Target_TP3, Ligand_TP3→Target_TP4).

In vitro functional assays: Monocytes from healthy donors were incubated with patient plasma (CCS or ACS at multiple timepoints) with/without IL-6 pathway perturbation (gp130/IL6ST inhibitor SC144 or anti-IL-6 antibody) and analyzed by flow cytometry for surface phenotype (e.g., CD93, CD88, CD14, CD86, CD89, CD11a, CD16, CCR2, HLA-DR, SLAN, CD36). Functional assays measured efferocytosis of apoptotic Jurkat cells, survival, ROS production (DCFDA), and CCL2-driven chemotaxis. PBMCs incubated with patient plasma were profiled for T-cell activation/maturation markers (CD69, PD1, TIM3, CCR7, CD45RO) over time. Additional in vitro scRNA-seq examined medication effects (heparin, aspirin, prasugrel) on Factor 2.

Prediction analyses: Factor 4 (IAR) values at admission (TP1M) were benchmarked versus GRACE score and clinical markers (CK, troponin, CRP) using ROC AUC for good vs poor in-hospital EF trajectory. A lasso logistic regression trained on top IAR features at TP1M in Munich was tested on Groningen TP1G (longer-term EF outcomes). Separate logistic regression models assessed a small NK-feature panel (CD53, GZMB, TXNIP). Statistical tests included ANOVA/Kruskal-Wallis with multiple-comparison corrections, mixed-effects models, and appropriate paired/unpaired tests as detailed in Methods.

• Data overview: 117 individuals total (Munich n=62; Groningen n=55) across 838 modality samples and multiple timepoints. Unsupervised scRNA-seq identified immune cell subsets and compositional shifts during ACS, including increases in CD14high monocyte clusters (clusters 4 and 6) and decreases in activated CD4+ T cell clusters (clusters 0 and 2) and NK cells.

• Factor 2: Integrative ACS Ischemia (IAI). MOFA Factor 2 explained substantial inter-patient variance across views and most variance in the clinical view, tracking myocardial ischemia onset and resolution and cleanly discriminating sterile ACS from CCS. • Replication: Applying Munich-derived feature weights to Groningen scRNA-seq reproduced the same temporal IAI pattern. • Independence from clinical variables: Re-running MOFA without clinical markers preserved the IAI pattern and factor structure. • Dominant contributors: High-weight features from CD4+ and CD8+ T cells (clusters 0,1,2) and monocytes (cluster 4). Notable genes included EIF3E and HINT1 in T cells and HMGB1 across T cells, tracking IAI over time. • Monocyte/T cell states: ACS plasma induced a proinflammatory monocyte phenotype in vitro (↑CD93, CD88, CD14, CD86, CD89, CD11a; ↓CD16 early; ↓CCR2 later) and increased efferocytosis without affecting survival or ROS; ACS plasma also drove early T-cell maturation/activation and later inhibition. • Pathway enrichment: Strong enrichment of interleukin signaling pathways (IL-6, IL-10, IL-12, IL-27). IL-6 pathway weights were prominent in T cells (clusters 2, 5) and monocytes (cluster 4), with plasma IL-6 high. • Functional validation: Inhibiting IL-6 signaling in monocytes reduced effector functions (ROS, survival, efferocytosis, chemotaxis) and cytokine secretion; IL-6 blockade partially reversed ACS plasma-induced monocyte activation while showing no effect with CCS plasma. • Ligand-target associations: IL-6 levels correlated positively with monocyte PIM1 and negatively with monocyte CD74, and associated with increased VCAN (cluster 7 monocytes). HMGB1 expression in T cell clusters correlated with many high-ranking IAI targets, including negative associations with T-cell UBC and monocyte PSME2, suggesting altered proteasome/ubiquitination activity. TGFBI in T cells correlated with monocyte ODC1, consistent with efferocytosis-promoting, anti-inflammatory macrophage programs that were suppressed early and recovered later in ACS. Lagged analyses confirmed IL-6 leading increases in CRP and acute-phase proteins (ORM1, SAA2, HP, SERPINA3).

• Factor 4: Integrative ACS Repair (IAR). IAR captured a repair-associated immune trajectory and stratified short-term outcomes. • Outcome association: Patients with poor in-hospital EF trajectory had significantly lower IAR already at admission (TP1M). IAR outperformed GRACE and single clinical markers (CK, troponin, CRP) at admission for classifying good vs poor outcome (higher ROC AUC; exact figure text cites superior association). • Replication and prediction: A lasso model trained on top IAR features at TP1M achieved ROC AUC 0.83 in the Groningen cohort (TP1G). A compact NK-feature model (CD53, GZMB, TXNIP) yielded ROC AUC 0.79 (Munich) and 0.91 (Groningen), indicating potential targeted biomarker panels. • Defining features: IAR had high negative weights for NK cytotoxicity/activation genes (TXNIP, PRF1, LITAF, GZMB, FYN, CST7, CD53) and for cytokines TRAIL/TNFSF10, CXCL9, CXCL10 (temporal patterns opposite to IAR), and CCL24. Plasma SERPINs (SERPINA1/2/3), anti-trypsin protease inhibitors protecting from granzyme-mediated cytotoxicity, showed positive weights, associating with a protective/repair state. General inflammatory markers (CRP, SAA1/2, C9) also tracked the repair signature within the study’s inclusion criteria.

• Factor 1: Integrative CCS (IC). IC was high in CCS, intermediate in coronary sclerosis, and low/negative in healthy coronaries, delineating chronic disease burden. • Dominant contributors: Many CD4+ and CD8+ T-cell features with positive weights (CD3E, ICAM3, TRAC, PRDX2, CORO1A, JUNB, CD37); negative weights included FOSB in activated T cells and NK cells. • Ligand-target network: ICAM3 among top ligands (positive), with CALM1 and CALM2 (negative) in T and other immune clusters. CALM1/ICAM3 correlated with PTMA (prothymosin alpha) in CD8+ T cells, a disease-protective factor. Monocyte NAMPT (negatively associating with IC) correlated inversely with T-cell JUNB and related genes, suggesting a dysregulated CD16high monocyte–NAMPT–T cell axis in CCS.

• Medication sensitivity: In vitro exposure of PBMCs to heparin, aspirin, and prasugrel did not reproduce IAI shifts; pre-medication in patients did not meaningfully confound IAI factor values.

Overall, MOFA uncovered multicellular immune axes: IAI (ischemia response), IAR (repair/outcome), and IC (chronic disease signature), linking plasma cytokines (notably IL-6), T-cell programs, monocyte phenotypes, NK cell cytotoxic activity, and protease inhibitors to ACS/CCS states and outcomes.

The findings support the hypothesis that multi-omic factor analysis of blood reveals coordinated, multicellular immune programs that map onto clinical states and outcomes in coronary syndromes. IAI captures the temporal systemic immune response to myocardial ischemia, integrating clinical biomarkers with transcriptional programs of T cells and monocytes and plasma cytokines. Mechanistic inference and in vitro validation highlight IL-6–driven monocyte activation and T-cell activation followed by quiescence, consistent with dynamic immune phases post-MI. IAR defines a repair-associated axis negatively linked to NK cytotoxic signatures and positively associated with anti-protease SERPINs, correlating with favorable EF trajectories; its performance at admission and replication in an external cohort underscores translational potential for early risk stratification. IC delineates chronic coronary disease characterized by T-cell dysregulation and a putative monocyte–T cell signaling imbalance. Together, these factors provide a compressed representation of complex immune landscapes with diagnostic and prognostic relevance and suggest therapeutic avenues (e.g., IL-6 pathway modulation, targeting NK cytotoxicity) to improve myocardial healing.

This proof-of-concept multi-omics study introduces MOFA-derived immune signatures for coronary syndromes: IAI (ischemia), IAR (repair/outcome) and IC (chronic). These factors integrate cell-intrinsic gene programs, plasma cytokines, proteomics, and clinical data to explain inter-patient variance, discriminate ACS from CCS, and predict short-term myocardial recovery. Mechanistic analyses implicate IL-6–dominated plasma signaling and T-cell–monocyte communication in shaping ACS immune states, while reduced NK cytotoxicity and elevated SERPINs associate with repair and improved outcomes. The approach demonstrates the potential of multi-omic liquid biopsies and factor-based integration for biomarker discovery, patient stratification, and identification of immunomodulatory targets. Future work should validate these signatures in larger, diverse cohorts, extend to non-STEMI and other ACS subtypes, evaluate targeted biomarker panels in pragmatic clinical workflows, and test therapeutic modulation of identified axes (e.g., IL-6, NK cytotoxic pathways) on cardiac recovery.

The study is observational with a limited sample size, constraining generalizability and precluding definitive clinical performance claims; large-scale validation trials are needed. Multi-omics profiling is cost-intensive, though small targeted panels of top-ranking features may be clinically feasible. The ACS cohorts focus on STEMI, limiting applicability to other ACS phenotypes. Guideline-recommended treatments could influence immune signatures, though analyses suggest minimal confounding. External validation used scRNA-seq only and did not include CCS vs non-CCS stratification for the IC factor.