Longitudinal proteomic characterization of AKI in hospitalized COVID-19 patients

2021 
Background: Acute kidney injury (AKI) is a known complication of COVID-19 associated with increased in-hospital mortality. Methods: We longitudinally measured serum levels of 4,497 proteins (SomaScan) in 437 COVID-19 patients at multiple timepoints along their hospital course and identified associations with AKI. Using single cell transcriptomic data from healthy human kidney specimens, we identified cell-specific kidney intracellular markers and quantified their leakage in sera from AKI patients. We also investigated whether serum proteomics improves AKI prediction. Results: We identified 408 upregulated and 107 downregulated proteins in COVIDAKI (144 cases, 293 controls, FDR<0.05, Fig 1A). Downregulated proteins included coagulation cascade inhibitors (protein C, heparin cofactor 2) and platelet dysregulation markers (Fig 1B), including platelet factor 4 (PF-4). Given the role of PF-4 in heparin induced thrombocytopenia (HIT), we then retrospectively analyzed 4,035 COVID-19 hospitalizations and found a significant association of HIT suspicion with COVID-AKI (aOR = 12.6, p <0.001). Intracellular AKI associated proteins were enriched for markers of the Loop of Henle, descending vasa recta endothelium, and NK cells (Fig 1C), which all have low ACE2 (Fig 1D) and TMPRSS2 expression (SARS-CoV2 receptor and activator respectively), suggesting bystander damage within the kidney, not direct viral invasion likely drives COVID-AKI. Finally, a random survival forest model incorporating protein levels had lower prediction error for incident AKI than one using only clinical variables (Fig 1E). Conclusions: The COVID-AKI serum proteome is characterized by dysregulated platelets with clinical evidence of HIT, improves prediction of incident AKI in a machine learning model and suggests inflammation mediated renal cell death, rather than direct viral invasion via the renal ACE2 receptor.
    • Correction
    • Cite
    • Save
    • Machine Reading By IdeaReader
    0
    References
    0
    Citations
    NaN
    KQI
    []