Abstract 4362727: Computational Modeling Predicts Benefits of Pro-efferocytic Therapy in Human Atherosclerosis

Efferocytosis is impaired in atherosclerosis and recognized as a hallmark of plaque vulnerability. Defective efferocytosis can be reactivated by ‘macrophage checkpoint inhibitors’. Only the key ‘don’t-eat-me’ molecule CD47, and its myeloid receptor, SIRPα, have entered prospective clinical trials. These anti-efferocytic molecules activate the phosphatase SHP-1. To target SIRPα, we developed a macrophage-specific nanotherapy loaded with a SHP-1 inhibitor. We demonstrated that this pro-efferocytic treatment reduces plaque inflammation both in mice and large animals. However, its efficacy in humans remains uncertain. Το address this question, we developed an in silico random forest classifier method to predict the effects on the transcriptomic profile of human atheromas. We leveraged transcriptomic information from 654 human carotid endarterectomy samples to model the clinical impact of SHP-1 inhibition (SHP-1i, Figure A ). We found that SHP-1 protein expression was upregulated in both lipo-necrotic and fibro-inflammatory lesions, as well as positively associated with histologically-assessed macrophage content and higher vulnerability index scores ( Figure B ). We mapped the gene expression changes that occur in human macrophages following SHP-1i onto genes linked to high plaque vulnerability scores. SHP-1i led to the downregulation of several genes associated with MACE. Pathway analyses indicated induced gene clusters that drive phagocytic signaling, leukocyte migration, and antigen presentation ( Figure C ). We then performed computational modeling to estimate how SHP-1i might influence the landscape of plaque subtypes. First, we trained a random forest classification model and then simulated the effects of SHP-1i by altering the gene expression profiles of these plaques based on the expression changes observed from in vitro studies. Finally, we used the trained model to reclassify the plaque subtypes. Notably, SHP-1i is predicted to manifest with a shift from a lipo-necrotic to a fibro-collagenous composition, indicating a reduction of lesion vulnerability ( Figure D ). To validate the method, we performed a neutral control and positive controls using two well-established treatments. Our method correctly predicted the impact of each drug on lesion vulnerability ( Figure E ).