Multi-omics integration and machine learning identify an ACKR1+ endothelial cell-based lactate-immune signature associated with carotid plaque vulnerability and progression.

BACKGROUND

Early identification of vulnerable carotid plaques is crucial for stroke prevention, but the upstream metabolic processes linked to endothelial transition and immune recruitment remain poorly defined. We characterised lactate metabolism in ACKR1+ endothelial cells (ECs) in human carotid plaques and derived a machine-learning signature for discriminating plaque vulnerability.

METHODS

Internal and external (GSE253903) single-cell datasets were integrated to identify endothelial subpopulations, and bulk transcriptomic (GSE163154) and in-house proteomic data were used to assess metabolism-immune correlations. Least absolute shrinkage and selection operator (LASSO) regression selected core features for a logistic model, which was locked and applied to three independent cohorts: GSE43292 (Stary stage ≥ IV vs. I-II), GSE28829 (early vs. advanced lesions) and GSE100927 (normal vs. atherosclerotic arteries). Key molecules were validated in situ by immunohistochemistry (IHC) and spatial transcriptomics in independent plaques (five stable, five unstable).

RESULTS

An ACKR1+ EC subpopulation linked to neovascularisation showed hyperactive glycolysis and enhanced lactate metabolism. Computational inference linked lactate accumulation to HIF-1α stability, CXCL12 release and CD8+ T-cell infiltration via the CXCL12-CXCR4 axis, and the monocarboxylate transporters SLC16A3 and SLC16A4 emerged as quantitative readouts of this 'lactate-immune' state. In unstable plaques, IHC confirmed higher atypical chemokine receptor 1 (ACKR1; geometric mean ratio [GMR] = 6.82), CD8+ T-cell density (GMR = 4.63), SLC16A3 (GMR = 4.74) and HIF-1α (GMR = 4.02; all q < .05), with lactylation elevated in the lipid core. A four-feature model (ACKR1, SLC16A3, SLC16A4, CD8+ T-cell cytotoxicity score) performed well in training and retained discrimination in the progression-stratified cohorts, while in GSE100927 it showed high disease-identification performance across three arterial beds.

CONCLUSIONS

Lactate metabolic reprogramming under hypoxia is associated with ACKR1+ EC-related CD8+ T-cell recruitment and plaque instability. This chain is an association-based hypothesis requiring functional confirmation. The signature discriminates robustly at both the vulnerability-stratification and disease-identification levels, providing a quantifiable framework for molecular classification of atherosclerotic plaques.