Introduction

The tubules play a crucial role in optimal functioning of the kidneys, and their impairment is responsible for substantial morbidity and mortality from conditions such as acute kidney injury (AKI) and progressive chronic kidney disease (CKD) culminating in kidney failure1,2. Tubular lesions on histopathology have been described across virtually all forms of chronic and acute kidney diseases. The kidney tubules comprise > 60% of the kidneys’ cellular mass and have among the highest density of mitochondria and metabolic workloads of any cell type in the body3,4. The high energy requirements of the tubules result from ATP-consuming reabsorption of over 99% of filtered sodium, glucose, and amino acids from the glomerular ultrafiltrate3. Several studies suggest that kidney function decline is more closely correlated with tubulointerstitial damage than glomerular injury5,6,7.

Acute tubular injury (ATI) describes a combination of pathologic findings, including tubular dilatation and epithelial flattening, tubular cell sloughing, and loss of nuclei, that reflect the morphologic responses of the tubules to a diverse range of insults8,9,10. ATI stands as a hallmark in the diagnosis of AKI but also manifests in the context of CKD10. Episodes of ATI and subsequent maladaptive repair can lead to the development and progression of CKD2,11. Therapies focusing on preventing or treating the consequences of ATI are lacking, but could play key roles in preventing morbidity and mortality from various kidney diseases.

Several studies have illustrated how large-scale proteomics approaches can identify important protein markers to prognosticate the risk of adverse clinical outcomes in patients with kidney disease12,13,14,15,16,17,18. Few studies have used this approach to determine non-invasive correlates of kidney histopathology, which could potentially lead to the identification of new therapeutic targets19.

In the present study, we used an unbiased proteomics assay to measure 6592 plasma proteins in the Boston Kidney Biopsy Cohort (BKBC), a cohort study of individuals with biopsy-confirmed kidney disease and adjudicated semi-quantitative assessment of ATI. We first assessed associations of each protein biomarker with ATI severity and determined enriched pathways that may have biological relevance for ATI pathogenesis. We then investigated the expression of biomarker proteins and corresponding biomarker-genes in regional proteomics and transcriptomics as well as single-cell RNA sequencing (scRNA-seq) from human kidney. Lastly, we explored associations between our findings and the development of AKI in three additional cohort studies with available proteomic profiling: the Kidney Precision Medicine Project (KPMP), the Atherosclerosis Risk in Communities (ARIC) study, and a cohort of critically ill patients from an intensive care unit (COVID-19 Host Response and Clinical Outcomes (CHROME)) study.

Results

As outlined in Fig. 1, we analyzed data from (i) individuals with biopsy-confirmed ATI, (ii) individuals biopsied for AKI and healthy controls, (iii) individuals from the general population at risk for developing AKI, and (iv) critically ill individuals at risk of severe AKI.

Fig. 1: Outline of studies, study participants, and primary outcomes.
figure 1

ATI: acute tubular injury, AKI: acute kidney injury. This figure was created with BioRender.com and released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license.

Full size image

Identification and characterization of plasma biomarkers of ATI

Baseline characteristics of the BKBC study cohort are shown in Table 1. The mean age was 54 ± 16 years and 48% were women. The mean eGFR was 51 ± 33 ml/min/1.73m2 and the median proteinuria (IQR) was 1.4 (0.3, 3.9) g/g creatinine. The most common primary clinicopathologic diagnoses were glomerulopathies (40%), diabetic nephropathy (16%), advanced chronic changes (15%), vascular disease (12%), and tubulointerstitial disease (11%). Figure 2a shows the breakdown of ATI severity by clinicopathologic diagnosis: 53% of BKBC participants had no ATI, 30% had mild ATI, 13% had moderate ATI, and 4% had severe ATI.

Table 1 Baseline characteristics of BKBC participants
Full size table
Fig. 2: Acute tubular injury (ATI) scores and identification of ATI biomarkers.
figure 2

a Distribution of semiquantitative severity scores for ATI by clinicopathologic diagnostic category in the Boston Kidney Biopsy Cohort (BKBC), Study 1. b Circulating plasma proteins associated with ATI severity in native kidney biopsy specimens. Beta coefficients are derived from multivariable linear regression models, adjusted for age, sex, race, and eGFR. The horizontal dotted line shows the Bonferroni-adjusted significance threshold. P-values are two-sided. ATI: acute tubular injury, BKBC: Boston Kidney Biopsy Cohort, eGFR: estimated glomerular filtration rate, GN: glomerulonephritis, GP: glomerulopathy.

Full size image

Associations between plasma biomarkers and ATI severity are shown in Fig. 2b and Supplemental Data 1. After multivariable adjustment and correction for multiple testing, 156 unique proteins (170 aptamers; some proteins are measured by 2 or more aptamers) were associated with ATI. Among those, higher levels of 126 proteins and lower levels of 30 proteins were independently associated with ATI severity. Higher levels of the following plasma biomarkers were associated with greater ATI severity (top 5 findings in the order of level of significance): osteopontin (SPP1), macrophage mannose receptor 1 (MRC-1), tenascin C (TNC), netrin-4 (NTN4), and marginal zone B- and B1-cell-specific protein (MZB1). The top five plasma proteins negatively associated with ATI severity were plasma serine protease inhibitor (SERPINA5), cholinesterase (BCHE), neuropeptide S (NPS), kallistatin (SERPINA4) and alpha-2-HS-glycoprotein (AHSG). Additional proteins associated with ATI with biological relevance in kidney disease include kidney injury molecule-1 (KIM-1; HACVR1), WAP four-disulfide core domain protein 2 (HE4; WFDC2), growth/differentiation factor-15 (GDF-15), serum amyloid A-1 protein (SAA1), and serum amyloid A-2 protein (SAA2).

Of the 156 ATI biomarkers identified in the BKBC, 78 were analyzed within the KPMP regional proteomics dataset. Comparison with glomerular expression revealed elevated protein expression of several ATI biomarkers in the tubulointerstitium (Fig. 3a). When comparing regional protein expression between individuals with AKI and healthy controls, we observed protein expression of eight ATI biomarkers. All except one (apolipoprotein A-1; APOA1) showed protein expression in direction concordant with greater ATI severity as observed in Study 1 (BKBC). We observed protein expression of our top ATI biomarker findings, osteopontin (SPP1), macrophage mannose receptor 1 (MRC-1), and tenascin C (TNC), in the tubulointersitium of kidneys with AKI (Fig. 3b). Utilizing regional transcriptomics data, 42 of the 156 ATI biomarkers were found to be expressed in kidney tubules, 29 of which showed higher tubular compared to glomerular expression. We observed the highest tubular expression for WFDC2, GDF15, and SPP1 when compared to glomerular expression (Fig. 3c). Detailed expression of these three markers in different kidney cell types is shown in Supplemental Fig. 1. We observed prominent expression of WFDC2, GDF15, and SPP1 in cell clusters of the thick ascending limb. SPP1 was also strongly expressed in distal convoluted tubule cell clusters, while GDF15 and WFDC2 expression was prominent in clusters of connecting tubule cells.

Fig. 3: Expression of acute tubular injury (ATI) biomarkers in regional proteomics and transcriptomics data and pathway analysis.
figure 3

a ATI biomarkers with elevated protein expression comparing tubulointerstitial to glomerular levels in KPMP regional proteomics data (AKI, CKD, HRT combined). b Protein expression of ATI biomarkers in the tubulointerstitium of individuals with AKI compared to healthy controls using Wilcoxon Rank Sum Test and Benjamini–Hochberg FDR to account for multiple comparisons. Shown are all proteins with a two-sided p < 0.05. c Gene expression of ATI biomarkers in KPMP regional transcriptomics data (AKI, CKD, HRT combined) using ANOVA. Two-sided p-values are adjusted for multiple comparisons using Benjamini–Hochberg FDR. d Pathway analysis of biomarkers associated with ATI severity. The top-ranked pathways are listed in black (one-sided p < 0.05) and gray (one-sided p < 0.25) following Benjamini–Hochberg correction. ATI: acute tubular injury, AKI: acute kidney injury, CKD: chronic kidney disease, HRT: healthy reference tissue, NES: normalized enrichment score, R-HSA: Reactome-Homo sapiens, GO: Gene Ontology, COPII: Coat Protein II.

Full size image

Of the 156 unique proteins linked to ATI, several play roles in the regulation of the immune system and intracellular pathways. We identified seven key pathways (all p < 0.05) with relevance for ATI pathogenesis. Of these, the top-ranked pathways were: ‘Intracellular Organelle Lumen’, ‘Endoplasmic Reticulum Lumen’, and ‘Immune System’ (Fig. 3d). Additional pathways that had immunoregulatory capacities included ‘Interactions Between Lymphoid and Non-Lymphoid Cells’ and ‘Neutrophil Degranulation.’ Proteins involved in these pathways included several cytokines, interleukins, members of the TNF receptor superfamily, and regulators of the complement cascade (Supplemental Data 2).

Difference of ATI biomarker levels in individuals with AKI and healthy controls

In the KPMP study, we analyzed plasma concentrations of the 156 ATI biomarkers comparing individuals with AKI to healthy subjects. After Bonferroni correction, 93 unique ATI biomarkers had significantly different plasma levels in AKI vs. healthy plasma (Fig. 4a). Of those, 77 ATI biomarkers were elevated and 16 were lower in individuals with AKI. The direction of these changes in all 93 biomarkers was consistent with the observations made in the BKBC (Fig. 4d). We observed the largest differences comparing AKI to healthy plasma levels for kidney injury molecule-1 (HACVR1), WAP four-disulfide core domain protein 2 (WFDC2), and growth/differentiation factor-15 (GDF-15).

Fig. 4: ATI biomarkers and acute kidney injury (AKI).
figure 4

a Comparison of ATI biomarker levels between healthy participants and those with AKI in the KPMP study using Student’s t-test. b Plasma ATI biomarkers associated with incident AKI in ARIC. Hazard ratios are derived from Cox proportional hazards models, adjusted for age, sex, Black race, hypertension, diabetes, systolic blood pressure, current smoking, eGFR, and log(UACR). p-values were calculated using the Wald test. c Plasma ATI biomarkers associated with severe AKI 7 days after ICU admission in a cohort of critically ill patients (CHROME). Odds ratios are derived from multivariable logistic regression models, adjusted for age, sex, and COVID-19 status. p-values were calculated using the Wald test. The horizontal dotted lines (a–c) show the Bonferroni-adjusted significance thresholds. p-values (a–c) are two-sided. d The heatmap illustrates the associations of 156 ATI biomarkers with specific outcomes across studies. Each biomarker is represented by a color: red signifies a statistically significant positive association, blue denotes a statistically significant negative association, and gray indicates that the biomarker was not available in the respective cohort. The biomarkers are organized based on their frequency of associations across studies and are further ordered according to the magnitude of their association with ATI severity in Study 1 (BKBC). ARIC: Atherosclerosis Risk in Communities study, ATI: acute tubular injury, BKBC: Boston Kidney Biopsy Cohort, CHROME: COVID-19 Host Response and Clinical Outcomes study, eGFR: estimated glomerular filtration rate, KPMP: Kidney Precision Medicine Project, UACR: urine albumin to creatinine ratio.

Full size image

Associations of ATI biomarkers with incident AKI

The baseline characteristics of ARIC participants are shown in Supplemental Data 3. Over a median of 6.8 years, 1084 ARIC participants developed AKI. Of the 156 unique proteins identified as ATI biomarkers in Study 1 (BKBC), 122 proteins were measured in ARIC. Of those, 35 were significantly associated with higher risks and 10 with lower risks of incident AKI after multivariable adjustment (Fig. 4b and Supplemental Data 4). The direction of these associations was found to be consistent (Fig. 4d): the same 35 proteins associated with elevated risks of AKI in ARIC were also linked to more severe ATI in the BKBC. Among those were osteopontin (SPP1), tenascin C (TNC), as well as WAP four-disulfide core domain protein 2 (HE4; WFDC2) and growth/differentiation factor-15 (GDF-15). Similarly, the 10 proteins associated with lower risks of AKI in ARIC corresponded to less severe ATI in the BKBC (Fig. 4d).

Associations of ATI biomarkers with severe AKI

Among the 268 critically ill patients enrolled into the CHROME study, 34 developed severe AKI on day 7 after ICU admission (Supplemental Data 5). Of the previously identified ATI biomarkers, 38 unique markers were significantly associated with severe AKI after adjustment for age, sex, and COVID-19 status. Of those, 36 were associated with higher risks and 2 with lower risks of severe AKI (Fig. 4c, d, and Supplemental Data 6) and were in a direction concordant with that observed in Study 1 (BKBC). Thirteen ATI biomarkers associated with severe AKI were also associated with higher risks of incident AKI in the general population (Study 3, ARIC) and had higher plasma levels in patients with AKI compared to healthy controls (Study 2, KPMP). These included osteopontin (SPP1), tenascin C (TNC), and mannose-binding lectin 2 (LMAN2), among others (Fig. 4d).

Discussion

This study provides an assessment of the plasma proteome in a cohort of individuals with biopsy-confirmed ATI followed by further study of the proteomic results in external cohorts. We identified biomarkers of ATI severity, assessed their tubular expression, and conducted pathway enrichment analyses that highlighted immune regulation and organelle stress responses as central mechanisms in ATI pathogenesis. Additionally, we expanded our investigation to examine associations between our findings and the development or prognosis of AKI in three distinct cohorts with available proteomic data.

Previous proteomics research has mainly centered on analyzing the plasma proteome to assess risks of adverse clinical outcomes in individuals with kidney disease18,20,21,22. Fewer studies have utilized proteomics to interrogate associations between circulating proteins and kidney histopathologic changes including ATI19,23,24. Kidney tubular injury can lead to a significant decrease in kidney function. While it is commonly linked to the clinical syndrome of AKI, it may also influence the onset and progression of CKD25. Some of the individual proteins we investigated have previously been identified to be associated with ATI. Among those, plasma KIM-1 stands as the most extensively studied marker12,26. In prior studies using both ELISA and proximity extension assay (Olink platform)27, we observed consistent associations of plasma KIM-1 with the degree of ATI19,28. Many other findings from the present study unveil new insights that may pave the foundation for future investigation.

The biomarker most strongly associated with more severe ATI was osteopontin (OPN; SPP1). OPN is a 44 kD glycoprotein predominantly secreted in bone and epithelial tissues. In the kidney, expression increases during injury29,30. Recent mechanistic studies have identified circulating OPN derived from kidney tubule cells as a key mediator of AKI-induced acute lung injury31. Another study showed that plasma OPN levels in individuals recovering from AKI were significantly higher when compared to individuals with irreversible loss of kidney function after AKI32. In the setting of CKD, higher serum OPN levels were associated with worse kidney function and greater risk of kidney failure and death33. Our results provide further evidence for a strong association of OPN with tubular injury and suggest its potential as a specific tool to assess for the degree of ATI. Similarly, we found a consistent relationship between higher tenascin C (TNC; TNC) levels, more severe ATI, and greater risk of AKI across cohorts. TNC, an extracellular matrix protein, is upregulated in the tubulointerstitium of CKD patients and has been shown to be involved in the transition from AKI to CKD by impairing tubular integrity through αvβ6 integrin signaling34.

We also observed strong associations between ATI severity and human epididymis protein-4 (HE4; WFDC2) and growth differentiation factor-15 (GDF-15; GDF-15). Both markers had higher levels in AKI, were associated with incident AKI, and were also among the top markers expressed in kidney tubules. HE4 is a serine protease inhibitor that has recently been shown to promote kidney fibrosis by inhibiting the degradation of type I collagen35. Studies also found a positive correlation between higher HE4 levels and more severe kidney fibrosis36,37. A prior investigation in diabetic kidney disease demonstrated that higher HE4 levels associated with increased risks of future kidney function decline38. Our study adds that HE4 may play an important role in the development of ATI. Additionally, the upregulation of this marker in both CKD (as shown in prior studies) and AKI underscores the interconnectedness between the two conditions and may provide additional evidence for the importance of kidney tubular injury in CKD progression12,13,26,39.

GDF-15 is a member of the TGF-β superfamily and has recently been described as a marker of oxidative stress in mitochondrial diseases40,41. In the kidney, GDF-15 exhibits a range of functions that are variably antagonistic or complementary, contingent upon the cellular state and surrounding microenvironment42,43,44,45. While some studies have found a potentially nephroprotective role for GDF-15 through the regulation of tubular Klotho expression43 and downregulation of inflammatory activities45,46,47,48, elevated levels of circulating and urinary GDF-15 have been linked to increased risks of incident CKD, CKD progression, and death in individuals with kidney disease44,45, as well as higher risk of cardio-renal outcomes in individuals with type 2 diabetes49. After ischemia-reperfusion injury, GDF-15-deficient mice had more severe ATI whereas recombinant GDF-15 attenuated kidney injury50. It is possible that in patients with kidney disease, GDF-15 levels may not have a pathogenic role but rather reflect a compensatory response to acute or chronic injury that is present in the setting of oxidative stress. To our knowledge, this is the first study to investigate the association between biopsy-confirmed ATI and GDF-15 in a large cohort of individuals with a diverse spectrum of kidney diseases. The observed association between higher GDF-15-levels and more severe ATI may yield further confirmation for the central role of mitochondrial dysfunction in acute and chronic kidney disease where GDF-15 may act as an important mediator of mitochondrial stress responses41.

Another ATI marker associated with AKI across all cohort studies was LMAN2, which has been previously shown by Mendelian Randomization (MR) analyses to be causally associated with kidney function. LMAN2, a protein also known as vesicular integral-membrane protein (VIP36), is implicated in transport processes and metabolic activities occurring on the apical surface of kidney epithelial cells51. Prior studies have demonstrated that LMAN2 correlates with albuminuria and is associated with kidney function decline51,52. A number of biomarkers identified in this study, especially those with previously demonstrated significant MR findings, merit additional evaluation for their therapeutic potential in a range of kidney diseases.

We found lower levels of seven plasma proteins to be associated with more severe ATI that were also significantly lower in individuals with AKI and conferred a potentially protective effect against AKI. Among those was plasma kallikrein (KLKB1), a serine protease inhibitor implicated in coagulation and blood pressure regulation53. Concordant with our observations, another investigation in type 1 diabetes revealed that plasma kallikrein activity decreased in more advanced stages of diabetic nephropathy, reaching its lowest levels in individuals undergoing dialysis54. Additionally, a recent genome-wide association study in the German CKD Study on OPN revealed a significant association at a locus within KLKB1, suggesting a potential link between OPN and the kallikrein-kinin system55. We also found that levels of complement C1q tumor necrosis factor-related protein 3 (C1QTNF3; C1QTNF3), a member of the C1q/TNF-related protein family, were lower in individuals with AKI. This is congruent with a smaller investigation focused on diabetic patients, which reported lower serum levels of C1QTNF3 in those with type 2 diabetes when compared to healthy controls56. These findings are further supported by in vitro experiments demonstrating that C1QTNF3 mitigates TGF-β1-mediated kidney fibrosis57 and ameliorates lipid accumulation and necroinflammation induced by high glucose concentrations in renal tubular cells58.

While the functional significance and potential prognostic value of these proteins remain to be fully investigated, many of them are involved in common biological pathways relevant to tubular injury such as regulation of immune and cellular stress responses. The leading pathways in our study related to cellular organelle stress including mitochondrial or endoplasmic reticulum damage. These pathways are critical in AKI onset and can also drive inflammation and fibrosis, key factors in the transition from AKI to CKD when stress is sustained59. In fact, several markers found in this study are markers involved in the pathogenesis of both AKI and CKD. Traditionally, ATI parameters have not been systematically assessed in biopsies obtained for CKD evaluation. It is possible, however, that the extent of ATI could serve as a valuable prognostic indicator for assessing the likelihood of transitioning from acute to chronic kidney injury, underscoring the potential benefit of incorporating routine ATI assessment in CKD biopsies to enhance prognostic accuracy.

Significant strengths of our study include the large number of protein biomarkers included in the analyses. Detailed adjudicated histopathologic scores by two kidney pathologists allowed us to test associations between markers and the degree of ATI. We were able to replicate our findings in three independent study cohorts and evaluated the expression of biomarker proteins and genes in regional proteomics, transcriptomics, and single-cell analyses, confirming many of our findings. Our study has several limitations that warrant consideration as well. Given the intrinsic limitations of cross-sectional studies, which prevent the establishment of causality, additional experimental and clinical studies are needed to establish the causal role of the ATI biomarkers identified in this study. While our study cohort underwent comprehensive phenotypic characterization, enabling extensive multivariable adjustment, we cannot exclude the potential influence of unmeasured confounding variables on our results. Our study delineates several proteins that warrant further investigation in animal models to elucidate their tissue-specific function within the kidney.

In conclusion, our proteomics study identified over 150 proteins in the plasma of individuals with histologically confirmed tubular injury, with subsequent interrogation of findings in separate cohort studies with plasma proteomics, regional proteomics, and single-cell transcriptomics.

Methods

Ethical compliance

We have complied with all ethical regulations related to this study. All studies included received approval from their respective institutional review boards (IRB) at each participating center which included the Mass General Brigham IRB (BKBC study), the University of North Carolina at Chapel Hill IRB (ARIC Study), the Johns Hopkins University IRB (ARIC Study), the University of Minnesota IRB (ARIC Study), the University of Mississippi Medical Center IRB (ARIC Study), and the University of Washington IRB (KPMP and CHROME studies). Informed consent was obtained from participants of the BKBC, ARIC, and KPMP studies. In the CHROME study, participants were enrolled under an IRB-approved waiver of informed consent. All studies were conducted in accordance with the principles of the Declaration of Helsinki.

Study populations

The BKBC is a prospective, observational cohort study of patients who underwent native kidney biopsy at three tertiary care hospitals in Boston, Massachusetts, including Brigham and Women’s Hospital, Massachusetts General Hospital, and Beth Israel Deaconess Medical Center. The study includes adults ≥18 years of age who underwent a clinically indicated kidney biopsy between September 2006 and October 2018. Exclusion criteria were the inability to provide written consent, severe anemia, pregnancy, and enrollment in competing studies. Details of the study design have been previously described60. Patients provided blood samples on the day of kidney biopsy. For this study, we evaluated 434 participants with available plasma samples. The KPMP is a multicenter prospective cohort study of people with CKD or AKI who undergo a protocol kidney biopsy at study entry as part of the KPMP consortium (https://KPMP.org)61. For plasma proteomics analyses, we evaluated protein measurements of 44 participants (26 with AKI and 18 who provided healthy reference tissue). The ARIC study is a prospective cohort study of individuals recruited from four US communities62. Participants were enrolled between 1987 and 1989, with subsequent visits in 1990–1992 (visit 2), 1993–1995 (visit 3), 1996–1998 (visit 4), 2011–2013 (visit 5), 2016–2017 (visit 6), and 2018–2019 (visit 7), visit 8 (2020), visit 9 (2021–2022), and visit 10 (2023). In this study, we included 4,610 participants from visit 5 who had available plasma proteomic profiling and non-missing covariates at baseline. The CHROME study is a prospective cohort study of critically ill patients admitted to three hospitals affiliated with the University of Washington in Seattle, WA between March 2020 and May 2021. Details of the study design have been previously described63. In brief, patients were eligible if admitted to a medical intensive care unit (ICU) with signs or symptoms of acute respiratory illness. Exclusion criteria included being under 18 years of age, incarcerated, pregnant, or undergoing chronic maintenance hemodialysis. In this study, we included 268 individuals with available SOMAScan plasma proteomic profiling. All studies included in this manuscript received approval from their respective institutional review boards at each participating center.

Sample collection and proteomics assays

Proteomic profiling was performed on blood samples from the baseline visit of the BKBC, KPMP, and CHROME cohort, as well as visit 5 of the ARIC study. After collection, blood samples were aliquoted and stored at −80 °C. The SOMAscan assay was utilized for proteomic measurements64. This assay employs SOMAmers (selective single-stranded deoxyoligonucleotides) for protein binding and quantifies proteins based on fluorescence intensity, indicative of relative protein concentrations. The ARIC and CHROME plasma samples were analyzed using the SOMAscan 5k platform (approximately 5000 proteins), while the BKBC and KPMP samples were assessed using the SOMAscan 7k platform (approximately 7000 proteins). In the BKBC, 6592 aptamers passed quality control metrics and were included in subsequent analyses; the mean coefficient of variation (CV) on 8 blind duplicate pairs was 4.7%. In ARIC, KPMP, and the CHROME cohort, we evaluated only those proteins that were significantly associated with ATI in the BKBC (156 unique proteins). For ARIC visit 5, the mean Bland Altman coefficient of variation was 6.6% from 26 samples in blind triplicate. In KPMP, the mean CVs on 2 sets of 4 blind duplicates was 4.9%. For all studies, protein aptamers were log2-transformed and winsorized at mean±5×SD and adaptive normalization was performed by maximum likelihood as previously described64.

Histopathologic outcome

In the BKBC, kidney biopsy specimens were adjudicated under light microscopy by two experienced kidney pathologists who provided semiquantitative scores of ATI scored from 0 to 3 reflecting none, mild, moderate, and severe lesion severity. Methods to evaluate and score histopathologic lesion severity were previously described in detail60. The weighted kappa statistic (95% CI) from 26 randomly selected biopsies for repeat review months after the initial scoring for ATI was 0.67 (0.45–0.89)60. All participants’ charts were reviewed alongside histopathologic evaluations to provide the final primary clinicopathologic diagnosis.

Outcome of acute kidney injury

In KPMP, patients eligible for percutaneous kidney biopsy for AKI must have elevated serum creatinine that is either sustained or accompanied by evidence of parenchymal injury. Detailed inclusion criteria for AKI biopsies were previously described in detail61. In ARIC, individuals included in this study were free of AKI at baseline and followed prospectively for incident AKI. Incident cases of AKI were identified by International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes 584.5 to 584.9 and Tenth Revision, Clinical Modification (ICD-10-CM) codes N17.0 to N17.9 or a 50% increase from outpatient serum creatinine during hospitalization. ICD codes were retrieved from hospital discharge billing and death certificates65. In the CHROME Cohort, the primary outcome was severe AKI within 7 days, defined by either initiation of kidney replacement therapy (KRT) or a doubling of serum creatinine from ICU admission63. In the ARIC and CHROME studies, blood samples were collected prior to the diagnosis of AKI or severe AKI, respectively.

Covariates

In the BKBC, detailed patient information was collected at the biopsy visit, including demographics, medical history, medication lists, and pertinent laboratory data and stored using REDCap electronic data capture tools hosted at Partners Health Care. We obtained serum creatinine (SCr) from the electronic medical record (EMR) on the day of biopsy. In participants for whom this was unavailable, we measured SCr in available blood samples collected on the day of the biopsy. We obtained spot urine protein-to-creatinine ratio (UPCR) or urine albumin-to-creatinine ratio (UACR) from the date of kidney biopsy to 3 months before biopsy from the EMR. If both were available, the UACR was used. If a participant did not have any of these values, we measured urine albumin-to-creatinine ratio from urine collected on the day of the kidney biopsy. SCr and urine creatinine were measured using a Jaffe-based method and urine albumin was measured by an immunoturbidometric method. The creatinine-based Chronic Kidney Disease Epidemiology Collaboration 2021 equation was used to calculate the eGFR66. In ARIC, covariates included age, sex, self-reported race, systolic blood pressure (SBP), UACR, smoking status, diabetes (fasting glucose of ≥126 mg/dL or non-fasting glucose level of ≥200 mg/dL, self-reported history of diabetes diagnosed by a physician, or use of medications for diabetes), hypertension (SBP ≥ 140 mm Hg and DBP ≥ 90 mm Hg, or use of medication for high BP), and eGFR (calculated using the CKD-EPI 2021 equation, which takes both serum creatinine and cystatin C into account)66. Serum creatinine was measured using a Roche enzymatic method (visit 5) and serum cystatin C was measured using the Roche Cobas 6000 chemistry analyzer. The UACR was calculated using urine albumin and creatinine (measured using an immunoturbidometric method on the ProSpec nephelometric analyzer and the Roche enzymatic method, respectively). In CHROME, covariates including age, sex, and COVID-19 status were extracted from the EMR.

Regional proteomics

We investigated the expression of ATI biomarker proteins in regional tissue proteomics from KPMP kidney biopsy samples (https://atlas.kpmp.org/explorer/regionalpro; access date: April 20, 2024). Detailed protocols and dataset information are available at https://www.kpmp.org/help-docs/technologies. In brief, kidney tissue was laser-microdissected to isolate glomerular and tubulointerstitial compartments. Following protein extraction from targeted tissue sections, proteins were analyzed using high-performance liquid chromatography coupled with mass spectrometry (HPLC-MS) for comprehensive identification and quantification. The study included tissue samples from 31 individuals (12 with AKI, 14 with CKD, and 5 healthy controls). We analyzed and compared the expression profiles of ATI biomarkers between the tubulointerstitial and glomerular compartments across all participants, and separately, evaluated differences in tubulointerstitial protein expression between healthy controls and those with AKI.

Kidney gene expression analysis

To map biomarker proteins to gene expression data, we investigated the expression of ATI biomarker-corresponding genes using regional transcriptomics and scRNA-seq data from KPMP (https://atlas.kpmp.org/explorer; access date: August 1, 2023). Tissue samples used for regional transcriptomics were drawn from living donor biopsy participants and used to compare expression (Fold Change) of our ATI biomarker genes of interest in the tubulointerstitium (n = 36; 9 healthy reference, 22 CKD, 5 AKI). Tissue samples analyzed using scRNA-seq were drawn from 47 participants (12 with AKI, 15 with CKD, and 20 healthy controls) and used to test for differential gene expression in kidney cell types. Details on these datasets have been described previously67,68.

Pathway analysis

To obtain basic functional information on biomarker proteins that were significantly associated with ATI severity and to investigate potentially relevant biological pathways, we applied Pathway Enrichment Analysis using gene sets obtained from publicly available databases including Gene Ontology, KEGG, and Reactome69,70. We ranked proteins based on their strength of association with ATI. We then calculated normalized enrichment scores to identify pathways with significant overrepresentation of ATI biomarkers. The Benjamini–Hochberg approach was employed to account for multiple testings and used to rank pathways based on the obtained p-value. Analyses were performed using the R package fgsea.

Statistical analysis

We summarized descriptive statistics as count with percentages for categorical variables and mean ± standard deviation or median with interquartile range for continuous variables. For skewed data distributions, we performed logarithmic transformation as appropriate. In the BKBC, multivariable linear regression models were used to assess associations of each plasma biomarker protein with ATI severity. In these models, the ATI severity score was used as the independent variable and each log2-transformed biomarker as the dependent variable. The adjusted model included the covariates age, race, sex, and eGFR. A prespecified α level of 7.58×10−6 set by Bonferroni correction (0.05/6592 proteins) was used to determine statistical significance. In ARIC, we used Cox proportional hazards models to test associations between the ATI biomarkers identified in the BKBC and the outcome of incident AKI. Models were adjusted for age, sex, self-reported race, SBP, smoking status, diabetes, hypertension UACR, and eGFR. In KPMP, we compared plasma levels of the ATI biomarkers in individuals with AKI and healthy controls using Analysis of Variance (ANOVA). In the CHROME cohort, we explored associations between the ATI biomarkers and severe AKI within 7 days of ICU admission using logistic regression models adjusted for age, sex, and COVID-19 status. In KPMP, we used a Bonferroni-corrected significance threshold of p < 3.21 × 10−4 (0.05/156 proteins). In ARIC and CHROME, we adjusted this threshold to p < 4.1 × 10−4 (0.05/122 proteins) and p < 3.79 × 10−4 (0.05/132 proteins), respectively, reflecting that only measurements of 122 (ARIC) and 132 (CHROME) of the 156 ATI biomarkers were available. In other analyses, we considered a two-sided p-value < 0.05 statistically significant. Statistical analyses were performed using R Version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria) and STATA 18.0 (STATACorp, College Station, TX).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.