Urinary proteomics identifies distinct immunological profiles of sepsis associated AKI sub-phenotypes

Abstract

Background: Patients with sepsis-induced AKI can be classified into two distinct sub-phenotypes (AKI-SP1, AKI-SP2) that differ in clinical outcomes and response to treatment. The biologic mechanisms underlying these sub-phenotypes remains unknown. Our objective was to understand the underlying biology that differentiates AKI sub-phenotypes and associations with kidney outcomes.

Methods: We prospectively enrolled 173 ICU patients with sepsis from a suspected respiratory infection (87 without AKI and 86 with AKI on enrollment). Among the AKI patients, 66 were classified as AKI-SP1 and 20 as AKI-SP2 using a three-plasma biomarker classifier. Aptamer-based proteomics assessed 5,212 proteins in urine collected on ICU admission. We compared urinary protein abundances between AKI sub-phenotypes, conducted pathway analyses, tested associations with risk of RRT and blood bacteremia, and predicted AKI-SP2 class membership using LASSO.

Measurement and main results: In total, 117 urine proteins were higher in AKI-SP2, 195 were higher in AKI-SP1 (FDR < 0.05). Urinary proteins involved in inflammation and chemoattractant of neutrophils and monocytes (CXCL1 and REG3A) and oxidative stress (SOD2) were abundant in AKI-SP2, while proteins involved in collagen deposition (GP6), podocyte derived (SPOCK2), proliferation of mesenchymal cells (IL11RA), anti-inflammatory (IL10RB and TREM2) were abundant in AKI-SP1. Pathways related to immune response, complement activation and chemokine signaling were upregulated in AKI-SP2 and pathways of cell adhesion were upregulated in AKI-SP1. Overlap was present between urinary proteins that differentiated AKI sub-phenotypes and proteins that differentiated risk of RRT during hospitalization. Variable correlation was found between top aptamers and ELISA based protein assays. A LASSO derived urinary proteomic model to classify AKI-SP2 had a mean AUC of 0.86 (95% CI: 0.69-0.99).

Conclusion: Our findings suggest AKI-SP1 is characterized by a reparative, regenerative phenotype and AKI-SP2 is characterized as an immune and inflammatory phenotype associated with blood bacteremia. We identified shared biology between AKI sub-phenotypes and eventual risk of RRT highlighting potential therapeutic targets. Urine proteomics may be used to non-invasively classify SP2 participants.

Keywords: Acute kidney injury; Intensive care unit; Sepsis; Urinary proteomics.

Fig. 1

Comparison of urinary proteomic profile between patients with AKI-SP1 and AKI-SP2. Volcano plot showing 117 urinary proteins with higher abundance in AKI-SP2 and 195 urinary proteins with higher urinary abundance in AKI-SP1 adjusted for age, sex, COVID-19 diagnosis, and body mass index

Fig. 2

Pathway analysis of urinary proteins between AKI-SP1 and AKI-SP2. 17 pathways were upregulated in AKI-SP2, while 3 pathways were downregulated in AKI-SP2 using WebGestalt based on a FDR ≤ 0.1. To the left of the figure includes each pathway ID and to the right of the figure includes the total number of proteins within the pathway

Fig. 3

Association of urinary proteins with risk of RRT during hospitalization. Among 86 patients with AKI-SP1 or AKI-SP2, 13 (15%) developed RRT. A greater abundance of urinary proteins to the right of 1 (n = 206) demonstrate a higher odds of receiving dialysis and a greater abundance of urinary proteins to the left of 1 (n = 179) demonstrate a higher odds for not receiving dialysis. Urinary proteins that are also significantly associated with AKI-SP1 are colored purple and urinary proteins significantly associated with AKI-SP2 are colored gold to demonstrate the overlap in biological processes leading to AKI sub-phenotypes and also to risk of dialysis