This prospective cohort study was conducted at Hospital Sao Paulo, the University Hospital of the Federal University of Sao Paulo, located in Sao Paulo, Brazil (15). The study was submitted to and approved by the National Research Ethics Committee (Comitê Nacional de Ética em Pesquisa – CONEP), Process number 3.978.709. All volunteers provided written informed consent prior to enrolment. The study included adult patients (≥18 years old) diagnosed with COVID-19 based on positive results from quantitative reverse transcription polymerase chain reaction (RT-qPCR) testing of nasopharyngeal swabs. These patients were admitted to the hospital wards between May 10 and September 26, 2020. This period coincided with the first wave of the COVID-19 pandemic in Brazil (February 25, 2020, to November 5, 2020) (16), when the predominant variants observed were B.1.1.28 (20–30%) and B.1.1.33 (10–35%) (16). Patients included in the cohort had moderate to severe illness, as defined by guidelines from the National Institutes of Health and the World Health Organization (17, 18). Exclusion criteria encompassed patients referred to outpatient clinics or admitted directly to the ICU. Among the initial 68 COVID-19 patients in the cohort, 17 KTRs were selected for this study [for a comprehensive overview of the entire cohort, please refer to Peçanha-Pietrobom et al. (15)]. Additionally, a control group comprised 10 KTR volunteers without COVID-19 symptoms (KTR controls), matched to the patient group for time since transplantation, immunosuppressive therapy, sex, and age. None of the participants in this study were vaccinated due to the unavailability of vaccines at the time of enrolment.

AKI was prospectively defined according to the Kidney Disease Improving Global Outcomes (KDIGO) criteria (19, 20). Baseline graft function was assessed using creatinine levels recorded three months prior to COVID-19 diagnosis, and AKI was identified based on creatinine levels at hospital admission.

Blood samples were collected from both patients and KTR controls into ethylenediaminetetraacetic acid (EDTA)-treated tubes (BD Biosciences, San Diego, CA, USA). Plasma and peripheral blood mononuclear cells (PBMCs) were separated using a Ficoll gradient method (Ficoll-Paque PLUS, GE Healthcare Biosciences, Uppsala, Sweden). Plasma samples were stored at –80°C, and PBMCs were preserved in liquid nitrogen for subsequent use.

For detailed information on the proteomics experiments, data processing protocols, plasma biomarker assays, and the corresponding analysis methods, please refer to the supplementary methods in Supplementary Material 1 .

Statistical analyses were performed using R (version 4.3.0). Normality of data was assessed using the Shapiro–Wilk test and Quantile–Quantile (Q-Q) plots. Non-normally distributed data were analyzed using the Mann–Whitney U test or the Kruskal–Wallis test, while normally distributed data were analyzed using Welch’s t-test or analysis of variance (ANOVA). For post hoc testing of non-normally distributed data, Dunn’s test of multiple comparisons using rank sums was employed; for normally distributed continuous data, Tukey’s post hoc test was conducted. Categorical variables were compared using the chi-square test or Fisher’s exact test, as appropriate. Biomarker data were log-transformed to normalize their distribution. To visualize the overall differences among plasma biomarkers, principal component analysis (PCA) was conducted following previously established method (21, 22). Correlation analyses were performed using Spearman’s rank correlation coefficient (Rho). Differences between groups were quantified using standardized mean differences (SMD) (23).

Differential protein abundance analysis was performed using the R/Bioconductor package limma. The model was adjusted for additional covariates: age, sex, time after transplantation (months), maintenance immunosuppressive therapy (class), and donor status (deceased or living). Differences in PBMC proteome profiling between the two groups were detected using the empirical Bayes moderated t-statistics, and Benjamini-Hochberg corrections were applied to all p-values to calculate the false discovery rates (FDR). Protein set enrichment analysis (PSEA) was performed using the fgsea package, with annotation information from the Reactome pathway database and hallmark gene sets from MSigDB. Additionally, single-sample PSEA (ssPSEA) was performed using the hacksig package and the Reactome pathway database. We focused on Reactome-defined immune-related pathways implicated in the host response to infection: innate immune system, adaptive immune system, cytokine signaling in the immune system, hemostasis, programmed cell death, and metabolism.

Seventeen kidney transplant recipients (KTRs) with COVID-19, a subset of a hospital-admitted cohort with diverse clinical outcomes (15) were analyzed in this study ( Table 1 ). According to the KDIGO criteria for staging AKI, we observed that six patients (35%) did not present with AKI (stage 0), while five (29%) were classified as stage 1, three (17%) as stage 2, and three (17%) as stage 3. The mean age was 54.0 years, and most were male (n = 10, 59%). The time between COVID-19 symptom onset and hospital admission was 7.5 ± 4.6 days. Primary symptoms included dyspnea (n = 12, 71%), cough (n = 12, 71%), fever (n = 7, 41%), diarrhea (n = 7, 41%), and nausea/vomiting (n = 7, 41%). Arterial hypertension was the most prevalent comorbidity (n = 13, 76%). Additionally, 53% of patients required ICU admission, 41% underwent mechanical ventilation, 29% required renal replacement therapy, and 41% died. Among the survivors, six participated in a clinical follow-up at an average of 29.3 ± 5.0 days after hospital discharge (designated as Convalescent Sample 30 [CS30]). No significant differences were observed in demographic data, maintenance immunosuppressive therapy class, or donor status when comparing KTRs with COVID-19 and KTR controls ( Table 1 ).

A TMT-based quantitative proteomic approach combined with liquid chromatography-tandem mass spectrometry (LC-MS/MS) was employed to quantify differentially abundant proteins (DAPs) among KTRs. The cohort was categorized based on the timing of blood sampling: patients sampled during their hospital stay (D0-wards) and those sampled approximately 30 days after discharge (CS30). These groups were compared with KTRs non-COVID-19 (KTR controls). Across seven TMT batches, a total of 2,330 proteins were identified. Focusing on proteins quantified in at least 50% of samples reduced this number to 1,403 proteins, with an average of 1,242 proteins per sample ( Supplementary Material 1 : Supplementary Figure S1 ). Among these 1,403 proteins, 363 were differentially abundant between D0-wards and KTR controls ( Supplementary Material 1 : Supplementary Figure S2A and Supplementary Material 2 ), while 111 showed differential abundance between CS30 and KTR controls ( Supplementary Material 1 : Supplementary Figure S2B and Supplementary Material 3 ).

Reactome pathway enrichment analysis revealed that D0-wards patients exhibited alterations in 15 pathways compared with KTR controls, characterized by positive normalized enrichment scores (NES) in pathways associated with interferon response, innate immune system, and neutrophil degranulation. Conversely, pathways related to glycolysis and glucose metabolism exhibited negative NES. Similar NES patterns were observed in hallmark gene set enrichment analysis, with positive NES in pathways related to interferon response and inflammatory processes, and negative NES in metabolism-related pathways ( Figure 1A ). Notably, only metabolism-related pathways remained altered in CS30 patients ( Figure 1A ).

Plasma biomarker levels indicative of cytokine/inflammation and endothelial cell activation were significantly elevated in D0-wards patients ( Figure 1B , Supplementary Material 1 : Supplementary Figure S3 ). Additionally, two proteins, NGAL and myeloperoxidase (MPO), were identified in both plasma and PBMCs (proteomics), demonstrating a positive moderate correlation ( Figure 1C ). Furthermore, the absolute number of lymphocytes, specifically T lymphocytes (CD3⁺CD45⁺), T cytotoxic (CD3⁺CD8⁺), and T helper (CD3⁺CD4⁺) were lower in D0-wards compared to KTR controls ( Figure 1D ). In contrast to admission samples, all plasma biomarkers ( Supplementary Material 1 : Supplementary Table S1 ) and absolute lymphocyte counts were comparable between CS30 and KTR controls ( Supplementary Material 1 : Supplementary Table S2 ), suggesting a potential return to a homeostatic state approximately 30 days after hospital discharge.

We compared KTR COVID-19 patients who presented with AKI (n = 6) to those who did not (n = 11) during ward admission (Table 2). Patients with AKI presented with higher Sequential Organ Failure Assessment (SOFA) scores (SMD: 0.90), increased C-Reactive Protein (CRP) levels (SMD: 0.60), and elevated creatinine levels (SMD: 1.40) upon ward admission. The AKI group experienced worse outcomes, including a higher rate of ICU transfers (SMD: 0.64), increased need for mechanical ventilation (SMD: 0.86), and greater hospital mortality (SMD: 0.86). Interestingly, creatinine showed a strong positive correlation with CRP (Rho = 0.50, p-values < 0.05) and SOFA respiratory subscores (Rho = 0.68, p-values = 0.002) ( Supplementary Material 1 : Supplementary Figure S4 ).

We compared the PBMC proteomic alterations in AKI and non-AKI COVID-19 patients using ssPSEA for Reactome-defined immune-related pathways. Our analysis revealed alterations in seven innate immune-related pathways, three adaptive immune system pathways, and five cytokine signaling pathways in the immune system. Additionally, no alterations were observed in hemostasis, while four metabolic pathways and one programmed cell death pathway showed alterations ( Figure 2A ). Indeed, the innate immune-related pathways exhibited the most differences when comparing patients with AKI and non-AKI; these pathways also demonstrated a pattern of positive correlation with creatinine, SOFA, and CRP levels ( Figure 2B ).

Next, we explored the differences between KTR COVID-19 patients with AKI and non-AKI based on 15 host response biomarkers reflective of two key pathophysiological domains (cytokine/inflammatory response and endothelial cell activation/procoagulant responses) in plasma obtained on admission to the wards. First, we generated domain-specific PCA plots to compare the groups ( Figure 3 ). There were significant differences in plasma biomarkers grouped in the cytokine/inflammatory response domain between patients with AKI and non-AKI. We also identified a substantial overlap in the endothelial cell activation/procoagulant responses between the AKI and non-AKI groups ( Figure 3A ). The complete contribution of each biomarker to a principal component (PC) score is depicted in Supplementary Material 1 : Supplementary Table S3 .

When comparing the magnitude of individual biomarker differences expressed as Hedges’ g, we observed moderate to large increases in cytokine/inflammatory markers among AKI patients, while endothelial cell activation/procoagulant markers showed small or negligible differences ( Figure 3B ). Subsequently, we evaluated the absolute number of total lymphocytes, which showed statistical differences when comparing the KTR control and COVID-19 AKI groups ( Figure 3C ). We also observed differences between the control and COVID-19 AKI groups for T helper cells. For T lymphocytes and T cytotoxic cells, differences were found in both comparisons between the KTR control vs. COVID-19 AKI groups, as well as between the KTR control vs. non-AKI groups. No differences were found in the comparison between the groups for the helper/cytotoxic ratio, B lymphocytes, and NK cells. Interestingly, for total lymphocytes, T lymphocytes, T cytotoxic cells, and T helper cells, the AKI group consistently showed lower absolute values than the non-AKI group, but with no statistical difference in the pairwise comparison.

Previous studies suggest an increase in immature cells, such as low-density neutrophils (LDNs) or granulocytic/polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs), in COVID-19 patients, which correlate with T cell suppression (24, 25). LDNs commonly co-purify with PBMCs in various infections, enriching pathways like “neutrophil degranulation” in omics studies (25–30). Therefore, we tested the correlation between the absolute values of T lymphocytes and the “neutrophil degranulation” ssPSEA scores, which showed a moderate negative correlation between T lymphocytes (Rho = –0.43, p-values = 0.02), T helper cells (Rho = –0.45, p = 0.03), and the “neutrophil degranulation” score ( Supplementary Material 1 : Supplementary Figure S5 ). A weak negative and not statistically significant correlation was found between T cytotoxic cells and the “neutrophil degranulation” score (Rho = –0.35, p-values = 0.08).