Proteomic Analysis of COVID-19 Plasma Reveals Dysregulated TREM-1, I-17, and Tumor Microenvironment Pathways Associated with Disease Severity

Abstract

Rationale: To utilize high-dimensional proteomic data to identify dysregulated pathways that are associated with COVID-19 disease severity and suggest potential therapeutic targets. Methods: We enrolled 161 COVID-19 inpatients admitted at two tertiary care hospitals. Plasma samples collected within 48 hours of admission were analyzed with the Olink Proximity Extension Assay;713 unique proteins were assayed. The WHO COVID-19 ordinal severity scale at enrollment was dichotomized into moderate (levels 3-4) and severe (levels 5-7). Normalized protein expression (NPX) values were generated in relation to a common pooled control plasma on each plate. The association between NPX values and disease severity on admission was estimated with logistic regression (LR) after adjustment for age, sex, race, and select comorbidities. Ingenuity Pathway Analysis (IPA) was employed after application of the Benjamini-Hochberg procedure with a false discovery rate of 5% to all proteins for which the NPX difference was +/-0.8 between groups. Predictive models of disease severity on hospital day 7 using all proteins as potential features were fit using elastic net LR (ENLR) and gradient boosting (GBM). Performance was estimated on a held-out test set (40% of the data) with area under the receiveroperator characteristic curve (AUROC). Results: Of 161 subjects, 85 (53%) were classified as having severe COVID-19. A total of 552 proteins were differentially expressed (Figure 1), and 31 of these proteins met criteria for inclusion in pathway analysis. IPA identified the triggering receptor expressed on myeloid cells 1 (TREM-1) signaling pathway (4 members, p=3.8E-3), the tumor microenvironment (TME) pathway (5 members, p=4.1E-3), and the interleukin 17 (IL-17) signaling pathway (4 members, p=1.8E-2). Interleukin 1 receptor-like 1, a member of the TREM-1 pathway, was the protein most associated with disease severity (OR=3.18, p=1.82E-08). Tumor necrosis factor ligand superfamily member 11 (TNFSF11), a member of the IL-17 signaling pathway was the only factor whose enrichment was associated with less severe disease (OR=0.39, p=2.3E-05). ENLR and GBM predicted disease severity on day 7 with AUROC values of 0.908 (0.828, 0.968) and 0.882 (0.788, 0.957), respectively. Conclusion: We identified pathways differentially expressed between patients with severe and nonsevere COVID-19 associated with immune function and angiogenesis. Several agents currently being investigated to treat severe COVID-19 act on these dysregulated pathways, and future investigations could test whether these proteins act as enrichment markers or response indicators. Integrating protein expression with cellular immune phenotype may help explain COVID-19 pathophysiology.