All patients with PCR-based SARS-CoV-2 infection confirmed by nasopharyngeal swabs were admitted to the Department of Anesthesiology, Emergency and Intensive Care Medicine, University Medical Center Göttingen, Germany and those requiring ICU supportive care were included. The institutional ethical board of the University Medical Center Göttingen, Germany approved the study (reference number 25/4/19Ü), and informed written consent was obtained. All urinary samples were collected at the University Medical Center Göttingen, Germany between March and June 2020. Urinary samples from severe septic shock patients collected in 2015 served as control and have been previously described and recruited at the Department of Intensive Care Medicine, University Medical Center Hamburg-Eppendorf, Germany (12). The study was approved by the ethical committee Hamburg (Aerztekammer Hamburg, reference number PV4550). A detailed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) flow chart of patient disposition is shown in Supplementary Figure 1.

Comorbidities were assessed using reported information by hospitals or private practices. Arterial hypertension, diabetes, hyperlipoproteinemia, or obesity was considered as metabolic diseases. Liver diseases, malignancies, and autoimmune diseases were considered as other diseases. Group allocation into AKI was defined according to KDIGO by laboratory findings, oliguria, or the need for dialysis. The simplified acute physiology score (SAPS) II and sequential organ failure assessment (SOFA) score were calculated according to published guidelines (13, 14). Requirement of intensive care treatment was defined at admission and calculated by the time between admission to the intensive care unit (ICU) or intermediate care unit (IMC) and relocation to the non-ICU/non-IMC medical ward; all patients required critical care treatment >24 h.

Investigators were blinded to clinical data collection and urinary enzyme-linked immunosorbent assay (ELISA) measurements. According to the manufacturer's protocols, the urinary levels of SARS-CoV-2 N (KIT40588, Sino Biological), human ACE2 (NBP2-78734, Novus Biologicals), and human TMPRSS2 (NBP2-89170, Novus Biologicals) were analyzed. In brief, 100 μL of native patient urine was analyzed by sandwich ELISA using precoated plates with antibodies specific to SARS-CoV-2 N, human ACE2, and human TMPRSS2. After adding a biotinylated detection antibody, avidin-horseradish peroxidase conjugate was added successively to each microplate well and incubated. After washing, the substrate solution was added to each well. The enzyme–substrate reaction was terminated by addition of a stop solution, and the optical density was measured spectrophotometrically at a wavelength of 450 nm. Measurements were done in triplicates for each urinary sample and compared to the standard curve. Negative test results were declared as not detectable. Photomicrographs of urinary SARS-CoV-2 N ELISA measurements in 20 cases with confirmed SARS-CoV-2 infection requiring ICU supportive care are shown in Supplementary Figure 2.

Tissues were dissolved in TRIzol (Zymo) and shredded using QiaShredder (Qiagen). RNA was extracted according to the manufacturer's protocol. RNA (0.5–1 μg) was added to the final PCR reaction. SARS-CoV-2 genome equivalents were detected by quantitative RT–PCR targeting the SARS-CoV-2 E gene as previously reported using the following primers: E_Sarbeco_F, ACAGGTACGTTAATAGTTAATAGCGT; E_Sarbeco_P1, FAM-ACACTAGCCATCCTTACTGCGCTTCG-BBQ; E_Sarbeco_R, ATATTGCAGCAGTACGCACACA (15, 16). The quantitative RT-PCR experiment and data processing were carried out using the LightCycler 480 Real-Time PCR System (Roche) and LightCycler 480 Software (version 1.5, Roche Molecular Systems). Absolute quantification was performed using SARS-CoV-2-specific in vitro–transcribed RNA standards, as previously described (15, 16).

At the time of the study design, data on markers for viral infection (SARS-CoV-2 N) and shedded cellular membrane proteins (ACE2, TMPRSS2) in urinary samples of AKI in COVID-19 patients were lacking. Hence, a convenience sample of 20 consecutive patients was chosen to provide a timely report. Variables were tested for normal distribution using the Shapiro–Wilk test. Non-normally distributed continuous variables are expressed as median and interquartile range (IQR); categorical variables are presented as frequency and percentage. Statistical comparisons were not formally powered or prespecified. For group comparisons, the Mann–Whitney U-test was used to determine differences in medians. Non-parametric between-group comparisons were performed with Pearson's chi-square test. Correlation between parameters was calculated by Spearman's rank correlation and is shown by heatmap reflecting the mean values of Spearman's ρ; asterisks indicate p < 0.05. We retained those covariates found to be significantly associated with AKI in a multivariable regression model (limiting the model to four covariates to avoid model overfit). To establish a cutoff for each parameter, the ability of prognostic factors to discriminate groups was evaluated by receiver operator curves (ROC) and the area under the curve (AUC), as well as sensitivity and specificity. Sensitivity and specificity were based on selection of the cutoff point on the ROC that maximized Youden's index (sensitivity+specificity-1); comparison of survival curves was performed with log rank (Mantel–Cox) testing. For multiple logistic regression, comparison of models was performed with the likelihood ratio test. We reported the hazard ratio (HR) with 95% confidence interval (CI) for each covariate of interest to assess the association between AKI and subsequent mortality. Data analyses were performed with GraphPad Prism (version 8.4.0 for MacOS, GraphPad Software, San Diego, California, USA).

Among the 20 cases with confirmed SARS-CoV-2 infection requiring ICU supportive care, 10/20 (50%) had AKI of whom 2 presented already at ICU admission and 8 developed AKI within 8 days after admission (Table 1). Interestingly, risk of AKI in severe COVID-19 was not associated with disease severity including respiratory failure at ICU admission (Table 1), implicating specific pathomechanisms that contribute to AKI in COVID-19. The urinary levels of SARS-CoV-2 N, ACE2, and TMPRSS2 were measured at ICU admission (day 1), at day 3, and at day 8 during further clinical course. The AKI stage in COVID-19 was associated with elevated urinary levels of SARS-CoV-2 N measured at ICU admission (Figure 1A, Table 2). A similar association was observed for day 3 and 8 measurements, but the effect was less pronounced (Table 2). The specificity of urinary SARS-CoV-2 N measurements was validated by analyzing an organ failure matched historical ICU cohort with no SARS-CoV-2 infection and AKI in 12/22 (54.5%) of the cases, in which urinary SARS-CoV-2 N was not detectable (Supplementary Table 1) (12). ROC analysis revealed that a cutoff urinary SARS-CoV-2 N level higher than 512.2 pg/mL at ICU admission identified patients with AKI (AUC 0.81, p = 0.0211, Figure 1B). Survival analysis for cumulative incidence of AKI confirmed that urinary SARS-CoV-2 N levels at ICU admission identified patients at risk for AKI (HR 5.9, 95% CI 1.4–26, p = 0.0095, Figure 1C). In contrast, urinary ACE2 or TMPRSS2 levels did not correlate with urinary SARS-CoV-2 N measurements or AKI at any time point (Table 2), implicating that urinary clearance of SARS-CoV-2 N more reflects systemic inflammation due to viral spread rather than renal cell death by direct viral infection of the kidneys. This is further supported by post-mortem analysis from deceased patients with AKI and high urinary SARS-CoV-2 N, where SARS-CoV-2 RNA was not detectable in corresponding kidneys (Supplementary Table 2). These findings suggested that urinary SARS-CoV-2 N occurred independently of shedded cellular membrane proteins essential for viral entry, associated with risk of AKI in severe COVID-19.

Plasma albumin levels at the time of ICU admission have previously been established as a risk marker for AKI and COVID-19 severity (17–19). Therefore, we next compared clinical and routine laboratory parameters assessed at ICU admission for association with AKI. In line with previous findings, hypoalbuminemia at the time of ICU admission also identified patients at risk for AKI in our cohort (Figure 2A, Table 3). Interestingly, the levels of urinary SARS-CoV-2 N and plasma albumin did not correlate directly (Figure 2A), suggesting that combining urinary SARS-CoV-2 N and plasma albumin could further improve risk stratification for AKI in COVID-19. Furthermore, hypoalbuminemia did not correlate with proteinuria including albuminuria (Table 3), indicating that observed hypoalbuminemia reflects systemic inflammation rather than urinary loss. Confirmed by ROC analysis, combining urinary SARS-CoV-2 N higher than 512.2 pg/mL and hypoalbuminemia below 2.05 g/dL (two-variable model, AUC 0.94, p = 0.0009, Figure 3A) outperformed urinary SARS-CoV-2 N alone (AUC 0.81, p = 0.0211, comparison of models: p = 0.0016) or plasma albumin alone (AUC 0.78, p = 0.03, comparison of models: p = 0.0061). Thus, combined urinary SARS-CoV-2 N and plasma albumin levels assessed at ICU admission identified patients at risk for AKI (two-variable model, HR 11.4, 95% CI 2.7-48, p = 0.0016, Figure 3B). In summary, combining urinary SARS-CoV-2 N and plasma albumin levels at ICU admission allowed for a more robust identification of patients at risk for AKI in COVID-19 as compared to analysis of the single markers alone.

AKI has previously been considered as an independent risk factor for premature death in patients with COVID-19 (1–5). Therefore, we next compared the association between AKI, disease course, outcome, and urinary SARS-CoV-2 N and plasma albumin levels assessed at ICU admission. AKI was associated with a prolonged ICU supportive care before patients could be relocated to a non-ICU medical ward (HR 2.8, 95% CI 0.97–8.1, p = 0.0269, Figure 4A). Urinary SARS-CoV-2 N and plasma albumin measured at ICU admission were equally associated with the length of ICU supportive care (two-variable model, HR 3.3, 95% CI 1.1–9.9, p = 0.0273, Figure 4B). In addition, combining urinary SARS-CoV-2 N and plasma albumin levels at ICU admission identified patients at risk for premature death in COVID-19 (two-variable model, HR 7.6, 95% CI 1.3–44, p = 0.0240, Figure 4C). In summary, AKI was associated with disease severity reflected by prolonged ICU length of stay. Furthermore, urinary SARS-CoV-2 N and plasma albumin levels at ICU admission equally allowed for a robust identification of patients at risk for prolonged ICU length of stay and premature death in COVID-19.