A multimodal atlas of COVID-19 severity identifies hallmarks of dysregulated immunity

The alpha-variant wave of the COVID-19 pandemic provided a unique opportunity to study, at single-cell resolution, how near-universal exposure to the same pathogen can lead to either effective or dysfunctional immune responses in humans. Although single-cell RNA-sequencing studies have characterized immune cellular features of COVID-19, they have not shown how tocilizumab treatment changes these features at single-cell resolution, or which features might persist into convalescence. In this study, we analyzed 2.5 million circulating immune cells from 428 patients across time points (840 PBMC samples), encompassing three contemporaneous SARS-CoV-2 cohorts: acutely infected patients across five WHO disease severity levels and three time points, patients from the first randomized control trial to study the efficacy of tocilizumab in the management of COVID-19, and convalescent patients three months after infection. We used linear modeling to integrate multiple data types - including single-cell RNA-seq, CITE-seq, TCR and BCR sequencing, viral load measurements, viral neutralization assays, detection of 75 autoantibodies, HLA genotype data, and serum proteomics covering 1,463 targets - to derive the most comprehensive view to-date of the biological features of COVID-19 disease severity. Our findings show that myeloid-derived suppressor cells (MDSCs) act as a key immunologic pivot point in severe COVID-19. Myeloid dysfunction, which is marked by impaired antigen presentation, drives a non-productive adaptive immune response, as reflected by reduced expression of B and T cell gene programs involved in antigen recognition, immune synapse formation, and cytotoxicity. Severe disease is also linked to autoantibodies targeting type I interferons, influenced by specific HLA-DQB1 allelic variants, and strongly correlated with serum IL-6 levels. Tocilizumab treatment eliminates CLU-expressing MDSCs and ISG-positive myeloid subsets, restores antigen presentation, and reactivates productive adaptive immunity. These changes align with improved clinical outcomes and better clinical laboratory measures, including reduced CRP. While many immunologic abnormalities in acute severe COVID-19 resolve during convalescence 3-months post-infection, we observed persistently high ICOS expression in regulatory T cells, potentially linking acute infection to chronic post-COVID syndromes. Overall, we define distinct innate and adaptive host immune responses associated with acute, IL-6-responsive, and convalescent SARS-CoV-2 infection. Our multimodal and high-dimensional dataset with curated clinical metadata provides a foundational and clinically relevant resource for modeling host immune response biology in health and disease.