Improving stroke risk prediction and individualised treatment in carotid atherosclerosis

Background: Unstable carotid atherosclerosis causes stroke, but methods to identify patients and lesions at risk are lacking. Currently, this risk estimation is based on measurements of stenosis and neurological symptoms, which determines the therapy of either medical treatment with or without carotid endarterectomy. The efficacy of this therapy is low and higher accuracy of diagnosis and therapy is warranted. Imaging of carotid plaque morphology using software for visualisation of plaque components may improve assessment of plaque phenotype and stroke risk. These studies aimed firstly to investigate if, and if yes, how, the carotid plaque morphology with image analysis of CTA associated with on-going biology in the corresponding specimen. Secondly, if risk stratification in clinical risk scores can be linked to the aforementioned associations. Finally, if the on-going biological processes can be specifically predicted out of the CTA imaging analysis.

Methods: Plaque features were analysed in pre-operative CTA with dedicated software. In study I and II, the plaques were stratified according to quantified high and low of each feature, profiled with microarrays, followed by bioinformatic analyses. Immunohistochemistry was performed to evaluate the findings in plaques. In study III, patient phenotype, according to clinical stroke risk scores of CAR and ABCD2 stratified the cohorts of high vs low scores which were subsequently profiled with microarrays, followed by bioinformatic analyses and correlation analyses of plaque morphology in CTA. In study IV, the microarray transcriptomes were individually coupled to morphological data from the CTA analysis, developing models with machine intelligence to predict the gene expression from a CTA image. The models were then tested in unseen patients.

Results: In study I, stabilising markers and processes related to SMCs and ECM organisation were associated with highly calcified plaques, while inflammatory and lipid related processes were repressed. PRG4, a novel marker for atherosclerosis, was identified as the most up-regulated gene in highly calcified plaques. Study II showed that carotid lesions with large lipid rich necrotic core, intraplaque haemorrhage or plaque burden were characterized by molecular signatures coupled with inflammation and extracellular matrix degradation, typically linked with instability. Symptomatology associated with large lipid rich necrotic core and plaque burden. Cross-validated prediction model for symptoms, showed that plaque morphology by CTA alone was superior to stenosis degree. Study III revealed that a high clinical risk score in CAR and ABCD2, reflect a plaque phenotype linked to immune response and coagulation, where the novel ABCB5, was one of the most up-regulated genes. The high risk scores correlated with the plaque components matrix and calcification but no positive association with stenosis degree. Study IV resulted in 414 robustly predicted transcripts from the CTA image analysis, of which pathway analysis showed biological processes associated with typical pathophysiology of atherosclerosis and plaque instability. The model testing demonstrated a good correlation between predicted and observed transcript expression levels and pathway analysis revealed a unique dominant mechanism for each individual.

Conclusions: Biological processes in carotid plaques associated to vulnerability, can be linked to plaque morphology analysed with CTA image analysis. Patient phenotype classified with clinical risk scores associates to plaque phenotype and morphology in CTA. The biological processes in the atherosclerotic plaque can be predicted with plaque morphology CTA analysis in this small pilot study, providing a possibility to precision medicine after validation in larger scale studies.

List of scientific papers

I. Correlation of Computed Tomography with Carotid Plaque Transcriptomes Associates Calcification to Lesion-Stabilization. Eva Karlöf, Till Seime, Nuno Dias, Mariette Lengquist, Anna Witasp, Håkan Almqvist, Malin Kronqvist, Jesper R. Gådin, Jacob Odeberg, Lars Maegdefessel, Peter Stenvinkel, Ljubica Perisic Matic, Ulf Hedin. Atherosclerosis. 2019 Sep;288:175-185.
https://doi.org/10.1016/j.atherosclerosis.2019.05.005

II. Carotid Plaque Phenotyping by Correlating Plaque Morphology from Computed Tomography Angiography with Transcriptional Profiling. Eva Karlöf, Andrew J. Buckler, Moritz Lindquist Liljeqvist, Mariette Lengquist, Malin Kronqvist, Mawaddah A. Toonsi, Lars Maegdefessel, Ljubica Perisic Matic, Ulf Hedin. [Submitted]

III. Clinical risk scores for stroke associate with carotid plaque vulnerability. Katarina Wadén, Eva Karlöf, Maria Ioanna Kotopouli, Sampath Narayanan, Mariette Lengquist, Ulf Hedin, Joy Roy, Ljubica Perisic Matic. [Manuscript]

IV. Virtual Transcriptomics: Phenotyping of Atherosclerosis by Decoding Plaque Biology from Non-invasive Computed Tomography Angiography Imaging. Andrew J. Buckler, Eva Karlöf, Mariette Lengquist, T Christian Gasser, Lars Maegdefessel, Ljubica Perisic Matic, Ulf Hedin. Arterioscler Thromb Vasc Biol. 2021 May 5;41(5):1738-1750.
https://doi.org/10.1161/ATVBAHA.121.315969