Differences in gene expression between atherosclerotic plaques have revealed five novel subtypes with distinct underlying biology and clinical presentation.
The findings, presented in Nature Cardiovascular Research, have been welcomed in an accompanying editorial as heralding “a new era in understanding atherosclerotic plaques.”
The discovery challenges the established dogma of “stable” and “unstable” lesions, refining this through the application of artificial intelligence to molecular signatures.
Intriguingly, the newly defined subtypes do not always align with prevailing wisdom, with the most severe form containing a mixture of characteristics traditionally categorized as “stabilizing” and “destabilizing.”
“We clearly show that plaques from patients differ in expression levels of multiple important genes, from which many are known drug targets,” lead researcher Michal Mokry, an assistant professor at UMC Utrecht in The Netherlands, told Inside Precision Medicine.
“This likely means that patients benefit differently from drugs that aim to prevent stroke or myocardial infarction. The main clinical implication is the possibility for more personalized and targeted care and risk prevention in atherosclerotic disease.”
The researchers applied machine learning algorithms to the whole transcriptome—the entire range of messenger RNA molecules expressed by an organism—of 654 carotid plaques.
Through this, they discovered groups of lesions with similar molecular footprints, akin to a system used previously to identify tumor subtypes in cancers.
Combining transcriptomics and histological data revealed five plaque subtypes, 0 to 4, identified as fibro-collagenous, intermediate, lipomatous, fibro-inflammatory, and fibro-cellular.
These plaque subtypes had different cellular compositions and were associated with the severity of symptoms before carotid endarterectomy surgery.
Specifically, lipomatous and fibro-inflammatory plaques were linked with a higher percentage of more severe symptoms, such as transient ischemic attack and stroke, compared with intermediate and less severe fibro-collagenous and fibro-cellular plaques.
Plaque subtype 3 had the highest polygenic risk score and the highest expression of genes associated with cardiovascular disease.
Patients with the different plaque subtypes did not differ significantly in major adverse cardiac events within three years of surgery. However, adding transcriptomic information to histology data did significantly improve the performance of a prediction model.
The researchers maintain: “These results suggest that the plaque transcriptomes encompass substantial added value for the association with clinically relevant atherosclerotic disease.”
They further identified candidate biomarkers whose plasma levels differed between the different plaque phenotypes.
A separate analysis of coronary arteries from 162 patients found that the fibro-inflammatory plaque type was strongly associated with coronary ischemic events.
In an accompanying editorial, Alexander Bashore, Lucie Zhu and Muredach Reilly from Columbia University point to the potential importance of high-resolution technologies in this field, such as Merfish and HybRISS.
They note that these, together with new 10x Genomics platforms that integrate spatial transcriptomics with microscopy-based in situ technology, will make it possible to map the expression of hundreds of genes at single-cell and single-molecule levels.
“These exciting technologies, along with innovations in machine-learning analyses of lesion histology, will build on the work of Mokry et al. to create cell- and gene-specific spatial maps of stable lesions versus those of unstable lesions,” they predict.
“Such data provide great promise for novel therapeutic targeting of cell types and their regulatory genes to prevent plaque instability and clinical events.”