Urinary Proteomics Identifies Cathepsin D as a Biomarker of Rapid eGFR Decline in Type 1 Diabetes

Diabetic kidney disease (DKD), defined as a reduction in estimated glomerular filtration rate (eGFR) or development of albuminuria, affects as many as 30–40% of people with type 1 diabetes (T1D) and is a major cause of morbidity and mortality in this population (1). The rate of eGFR decline, even when eGFR is in the normal range, has been associated with future incidence of kidney failure (2,3).

Identification of urine markers of rapid eGFR decline in T1D using proteomics techniques may provide insight into the molecular mechanisms underlying this process and enable the development of prognostic tools and targeted therapies. Differential urine protein patterns may be linked to functional processes including inflammation, fibrosis, and aberrant glomerular and tubular activity and used to better understand the role of these disease pathways in DKD pathophysiology (4). Urine peptides have been associated with eGFR decline and incident microalbuminuria in diabetes, though most large cohort studies have been in populations with type 2 diabetes (T2D) (5,6).

The JDRF Biomarkers Consortium was established with the aim of identifying protein, metabolite, and lipid markers associated with rapid eGFR decline in T1D using novel, targeted multiomics assays (7). Here we report findings from the consortium’s urine proteomics analyses along with results of orthogonal confirmatory studies.

The JDRF Biomarkers Consortium performed a multinational case-control study nested within four longitudinal T1D cohorts, those of the Finnish Diabetic Nephropathy (FinnDiane) Study, the Steno Diabetes Center Copenhagen study (Steno), the Epidemiology of Diabetes Complications (EDC) study, and the Coronary Artery Calcification in Type 1 Diabetes study (CACTI) (7). Case subjects were categorized based on having an annual eGFR decline of ≥3 mL/min/1.73 m² and control subjects an annual eGFR decline of <1 mL/min/1.73 m² over a median 8 years (interquartile range 5–12) of follow-up. For all study cohorts the Chronic Kidney Disease Epidemiology Collaboration creatinine equation was used for glomerular filtration rate estimation (8). eGFR slope was derived with regression lines fitted to longitudinal eGFR values (FinnDiane Study, Steno) or calculation of the difference between eGFR values from the first and last study visits and dividing by the number of years between these (EDC, CACTI) (7,9).

A total of 1,430 participants across the four cohorts were included in the Biomarkers Consortium based on the following criteria: baseline eGFR ≥30 mL/min/1.73 m², follow-up of ≥2 years, three or more eGFR measurements, and baseline blood and urine sample availability. Because of the extended duration of each cohort, only participants examined between 1995 and 2011 were included to maintain consistency across cohorts. Proteomics measurements were performed on baseline urine samples from 1,270 participants (508 case and 762 control subjects) with sufficient urine available for measurement. Participants were randomly divided into a discovery set (n = 630) and validation set (n = 640).

Urine proteins were selected for measurement following a comprehensive literature review of DKD pathophysiology. Thirty-eight tryptic peptides from 20 proteins were determined to be most amenable for precise multiplex quantification with liquid chromatography–tandem mass spectrometry (LC-MS/MS) assay (Supplementary Table 1). We developed and validated a novel targeted LC-MS/MS assay for this purpose (10). Prior to LC-MS/MS quantification, urine proteins were precipitated from solution by mixing of 200 µL of each urine sample with 800 µL of a 50:50 mixture of methanol and acidified acetone (i.e., acetone supplemented with 10 mmol/L HCl) after spiking of each sample with α-amylase from Aspergillus oryzea. Proteins were cleaved into peptides by trypsin after denaturation, reduction, and alkylation (Supplementary Materials and Supplementary Table 2).

Peptides were separated with liquid chromatography prior to being analyzed on a hybrid quadrupole-Orbitrap tandem mass spectrometer. Selective measurement of the 38 target peptides was accomplished with parallel reaction monitoring. Stable isotope standards for each of the peptides using C¹³ or N¹⁵ were spiked into the samples immediately prior to analysis with LC-MS/MS. Endogenous peptides were quantified in terms of their peak area ratio (PAR), which is calculated as the ratio of the peak area of the endogenous peptide sample to that of its corresponding stable isotope internal standard. Samples were analyzed in three subsets. The discovery set was measured in subsets 1 and 2 with samples distributed randomly across batches. The validation set was measured in subset 3 (Supplementary Table 3).

For facilitation of comparison across peptides, peptide PAR values were log₂ transformed, centered to zero, and scaled, dividing by the SD. We calculated scaled protein PAR values by averaging scaled peptide PAR values corresponding to the same protein. Six peptide PAR with extremely high values (four within the discovery set and two within the validation set) believed to reflect laboratory error were removed (set to missing) because they were substantially different from another peptide in the same protein (residual >5 SD) and at least 1.5-fold different from the second-highest PAR. Peptide measurements with undetected quantities (n = 324) were set to half of the smallest detected value.

Peptide and protein associations with rapid eGFR decline were tested using logistic regression models. Three models were generated: 1) a minimally adjusted model including age, sex, and log₂(urine creatinine); 2) an intermediately adjusted model including these covariates plus baseline diabetes duration, antihypertensive medication use, SBP, and hemoglobin A_1c; and 3) a fully adjusted model including these covariates plus baseline urine albumin-to-creatinine ratio (UACR) (log₂ transformed) and eGFR.

A false discovery rate <0.05 was used to identify proteins of interest in the discovery set. These proteins and their corresponding peptides were then tested in the validation set with logistic regression models with additional adjustment for sample processing site. Discovery and validation set results corresponding to proteins of interest were meta-analyzed with use of inverse variance–weighted random-effects models, since laboratory processing differences in these two groups precluded individual-level logistic regression models. Subgroup associations were determined with stratified logistic regression analyses. Missing covariate and peptide data were accounted for with multiple imputation.

We tested associations between serum and urine cathepsin D concentrations and kidney histopathology in 56 subjects with either T1D or T2D and a primary biopsy diagnosis of diabetic nephropathy from the Boston Kidney Biopsy Cohort (BKBC) (Supplementary Table 4). The BKBC enrolled adults who underwent a for-cause native kidney biopsy between 2006 and 2016 across three Boston-area hospitals (11). Linear regression models were used to determine associations of urine and serum cathepsin D with histopathological lesion severity, with adjustment for age and sex.

We additionally examined cross-sectional associations between urine and serum cathepsin D concentrations in 96 adults with diabetes from the BKBC (n = 56), the Healthy Kidney Study (n = 19), and a pilot vitamin D trial in subjects with T2D and microalbuminuria (n = 21) (Supplementary Materials and Supplementary Table 4). Urine and serum cathepsin D measurements were conducted with a commercially available ELISA from Abcam (ab119586; Cambridge, MA [Supplementary Materials]).

Cathepsin D gene (CTSD) intrarenal profiles were evaluated in the microdissected kidney biopsy specimens from Pima Indians with T2D that underwent transcriptomic profiling (12). In short, the gene expression profile from Affymetrix platform underwent standard quality control processing, normalization, and log₂ transformation. Glomerular filtration rate was measured annually by the urinary clearance of iothalamate. Kidney morphometric measures evaluated included cortical interstitial fractional volume (VvInt) as previously described (13).

To define the RNA expression patterns of CTSD in kidney cells of patients with DKD, we compared CTSD expression in kidney cell populations using single-cell RNA sequencing (scRNAseq) in 10 adults with DKD and 18 living donors. Participants with DKD were enrolled in the Kidney Precision Medicine Project (KPMP) at recruitment sites located in Cleveland, Dallas, or Boston. Living donor biopsies were obtained before perfusion of living donor kidneys and placement in the recipient (14,15). For all kidney biopsies, tissue processing, single-cell isolation, and scRNAseq data generation were performed according to the protocol developed by the KPMP (16–18).

The F1 Akita mouse line (DBA/2 × C57BL/6-Ins2^Akita) was used to model diabetic kidney pathology, and DBA/2 × C57BL/6 mice were used as controls (19,20). Cathepsin D enzyme activity was measured in mouse urine and plasma samples by biochemical assay. Immunofluorescent histochemistry was used to localize cathepsin D in mouse kidney tissue samples (Supplementary Materials).

Given existing literature indicating that advanced glycation end products (AGEs) play an important role in lysosomal dysfunction in DKD pathology, we investigated cathepsin D responses to AGE exposure in mouse proximal tubular epithelial cells (MCT) in culture. MCT were cultured in growth media and exposed to 5 mmol/L glucose, 25 mmol/L glucose, 100 μg/mL control-BSA (ctr-BSA), 200 μg/mL ctr-BSA, 100 μg/mL AGE-BSA (AGE-BSA), or 200 μg/mL AGE-BSA for 24 h (Supplementary Materials). Cathepsin D and Lamp1 (lysosome-associated membrane glycoprotein 1) immunofluorescent staining was performed. Additionally, cathepsin D enzyme activity and interleukin-1 (IL-1), interleukin-6 (IL-6), and kidney injury molecule-1 (KIM-1) measurements were conducted in cell media using a biochemical assay and ELISA, respectively. MCT were then transfected with cathepsin D small interfering RNA (siRNA), which blocked cathepsin D translation, or control siRNA, which had no effect on cathepsin D (Supplementary Fig. 1). Cathepsin D enzyme activity and IL-1, IL-6, and KIM-1 measurements were then compared in siRNA-transfected cells exposed to either 100 μg/mL ctr-BSA or 100 μg/mL AGE-BSA.

The JDRF Biomarkers Consortium study was exempt from institutional review board approval, as it did not constitute human research. Mouse studies were conducted with approval from the Institutional Animal Care and Use Committees of UC San Diego and The University of Texas Health Science Center at San Antonio.

Mean age of JDRF cohort participants was 43 years at baseline; 51% were female, and 98% identified as White (Table 1). Mean baseline diabetes duration was 27.1 years and mean hemoglobin A_1c 8.5%. Most participants had no evidence of kidney disease at baseline, with an overall mean eGFR of 94 mL/min/1.73 m². Approximately 36% of participants had a UACR ≥30 mg/g. The percentage of missing data among covariate variables ranged from <1 to 4%, except for smoking history, which had 29% missingness.

In the discovery set, median annual eGFR slope over the follow up period was −4.3 and 0 mL/min/1.73 m² for case and control subjects, respectively. In the validation set, median annual eGFR slope was −4.4 and 0.1 mL/min/1.73 m² for case and control subjects. Compared with control subjects, case subjects were younger and had higher baseline eGFR and UACR (7). Demographic and clinical variables were well-balanced across discovery and validation sets.

In the discovery set, 8 and 4 of 38 urine peptides were significantly associated with rapid eGFR decline with false discovery rate <0.05 in the minimally and intermediately adjusted models, respectively (Supplementary Table 5). In these models, the greatest magnitudes of association with rapid eGFR decline were observed in two peptides corresponding to cathepsin D. In the fully adjusted model, only 1 cathepsin D peptide out of all 38 peptides remained significantly associated with rapid eGFR decline (odds ratio 1.44, 95% CI 1.17–1.78) (Fig. 1A). Of 20 proteins, 4 and 2 corresponding to measured peptides were associated with rapid eGFR decline in minimally and intermediately adjusted models, respectively (Supplementary Table 6). Of these, only cathepsin D remained associated with case status in the fully adjusted model (odds ratio 1.41, 95% CI 1.14–1.74) (Fig. 1B).

In the validation set, the concentrations of both urine cathepsin D peptides and cathepsin D protein were significantly associated with rapid eGFR decline in minimally and intermediately adjusted, but not fully adjusted, models (Supplementary Table 7). On sequential adjustment for covariates from the fully adjusted model, baseline UACR most strongly attenuated the association between cathepsin D protein and rapid eGFR decline in both discovery and validation sets (Supplementary Tables 8 and 9).

In meta-analysis of discovery and validation sets, both cathepsin D peptide concentrations and the combined protein concentration were significantly associated with rapid eGFR decline in all models (Fig. 2). In the fully adjusted model, 1 SD of log₂-transformed protein higher urine cathepsin D was associated with 29% greater odds of rapid eGFR decline (random-effects model 95% CI 1.07–1.55). No significant heterogeneity was observed by baseline albuminuria or eGFR subgroups in the validation set (Supplementary Tables 10 and 11).

In the BKBC cohort, significantly higher concentrations of urine cathepsin D were observed with increasing severity of interstitial fibrosis/tubular atrophy, with adjustment for age and sex (Table 2). Significantly higher concentrations of urine cathepsin D were also observed with increasing severity of tubulointerstitial inflammation in the fibrosed interstitium. No significant associations were observed between urine cathepsin D concentrations and glomerular or vascular compartment lesions. Serum cathepsin D concentrations were not associated with increasing severity of any histopathologic lesion (Supplementary Table 12). Serum and urine cathepsin D concentrations correlated only weakly in combined BKBC, Healthy Kidney Study, and pilot vitamin D trial cohorts (r = 0.19) (Supplementary Fig. 2 and Supplementary Table 13).

In the Pima Indian cohort (Supplementary Table 14), CTSD mRNA level from tubulointerstitium but not glomeruli of microdissected kidney biopsy samples significantly correlated with VvInt (r = 0.35, P = 0.018) (Fig. 3). Tubulointerstitial CTSD mRNA was also significantly correlated with age (r = 0.42, P = 0.003) and diabetes duration (r = 0.36, P = 0.013). No significant correlation was observed between glomerular CTSD mRNA and any of the clinical parameters.

In analyses of 10 KPMP participants with DKD and 18 living donors, 25 cell clusters that covered a wide spectrum of kidney cell types and tissue-resident immune cells were defined (Supplementary Table 15 and Supplementary Fig. 3). CTSD mRNA was ubiquitously expressed in these kidney cell types. Abundant CTSD mRNA was observed in tubular epithelial cells, including three subsets of proximal tubular cells as well as thin ascending loop of Henle, thick ascending loop of Henle, distal convoluted tubular, collecting duct intercalated, and principal cells. Expression of CTSD was detected with less abundance in glomerular cells, such as podocytes, endothelial cells, and mesangial cells (Supplementary Fig. 3). In comparisons of KPMP participants with DKD with living donors, CTSD mRNA was expressed in larger proportions of proximal tubular cells (Fig. 4A) and average expression levels were significantly higher (Fig. 4B).

Cathepsin D urine enzyme activity was significantly higher in Akita mice compared with controls at both 8 weeks (1,287 vs. 415 relative fluorescent units [RFU], respectively; P < 0.0001) and 16 weeks (1,339 vs. 530 RFU; P < 0.0001), while no difference was observed in plasma cathepsin D activity (Supplementary Fig. 4). Kidney immunofluorescent staining of cathepsin D demonstrated more abundant staining in the tubules and glomeruli of Akita mice compared with controls.

Immunofluorescent staining of cathepsin D and Lamp1 in MCT was more intense with exposure to 100 μg/mL and 200 μg/mL AGE-BSA compared with different concentrations of ctr-BSA or glucose alone (Fig. 5A). Greater cathepsin D with AGE-BSA appeared to be both localized to lysosomes and cytoplasmic, as evidenced by yellow and red immunofluorescence in merged panels, respectively.

Cathepsin D enzyme activity was higher in the media of MCT exposed to AGE-BSA compared with cells exposed to ctr-BSA (Fig. 5B). Exposure to AGE-BSA versus ctr-BSA additionally also resulted in greater release of IL-1, IL-6, and KIM-1 from MCT (Supplementary Fig. 5). Cells transfected with cathepsin D siRNA released significantly more IL-1 and KIM-1 in AGE-BSA than cells transfected with control siRNA (Figs. 5C–E). A similar, nonsignificant, trend was observed with IL-6.

Using a targeted proteomics approach, we found that urine cathepsin D was associated with rapid eGFR decline in our large, multinational T1D cohort. In subsequent studies, proximal tubular cathepsin D expression was higher in subjects with DKD compared with healthy living donors and both urine cathepsin D and tubulointerstitial cathepsin D expressions were associated with tubulointerstitial pathology among people with diabetes. Furthermore, in mouse proximal tubular epithelial cells, exposure to AGE-BSA increased cathepsin D activity and increased inflammation and cellular injury, which was exacerbated by blocking cathepsin D transcription. Together, these studies suggest that kidney cathepsin D production rises in response to activation of pathological molecular pathways in diabetes and that elevated urine concentrations may indicate early tubulointerstitial injury that leads to rapid eGFR decline in T1D.

Cathepsin D is a ubiquitous lysosomal enzyme involved in cellular homeostatic processes including collagen and intracellular protein degradation, apoptosis, and autophagy (21–23). Impaired lysosomal function and autophagy are linked to inflammation and oxidative stress and have been increasingly recognized as key drivers of tubulointerstitial and glomerular injury in diabetes (24,25). Increased urine cathepsin D has been described in humans and animals with diabetes (26–28). In addition, greater cathepsin D kidney tissue expression has been observed in conjunction with nonspecific tubular injury in human and animal studies of acute kidney injury and chronic kidney disease (CKD) (29–31). With our study we build on the work of published studies by demonstrating a temporal relationship between urinary cathepsin D excretion early in T1D and subsequent rapid eGFR loss, suggesting that lysosomal dysfunction may be important in early stages of human DKD, and characterizing the pathophysiology that underlies urinary cathepsin D excretion.

The results of our follow-up studies suggest that urinary cathepsin D largely reflects tubulointerstitial damage. Cathepsin D expression was apparent across all kidney cell types but most abundant in tubular cells. In focusing on proximal tubular cell types, cathepsin D expression was greater in KPMP participants with DKD compared with living kidney donors. In the Pima Indian cohort, tubulointerstitial cathepsin D expression was directly correlated with VvInt, a marker of tubulointerstitial injury that was previously linked to mitochondrial dysfunction, inflammation, and tubular metabolic pathways (12). In the BKBC, higher urine cathepsin D was associated with greater severity of interstitial fibrosis and tubular atrophy, which is associated with risk of kidney disease progression, and tubulointerstitial inflammation in the fibrosed interstitium (11). Our observed lack of correlation of plasma and urine cathepsin D further suggests that urinary cathepsin D is predominantly of kidney origin, consistent with the published observation that cathepsin D has distinct oligosaccharide patterns in human serum and urine (32).

Exposure to AGE-BSA increased cathepsin D in addition to markers of inflammation (IL-1, IL-6) and kidney injury (KIM-1) in mouse proximal tubular epithelial cells in culture. Elements of the hyperglycemic milieu, namely, AGE-modified albumin and lipid molecules, are known to interfere with lysosomal protein degradation and autophagic flux in various kidney compartments, particularly the proximal tubule (33–36). Proximal tubule cells in culture metabolize AGE-modified albumin differently from unmodified albumin, demonstrating impaired protein binding and intracellular degradation as well as abnormal lysosomal activity (37). Interaction of AGEs with specific AGE receptors on proximal tubule cells also triggers inflammatory and intracellular oxidative stress pathways, which contribute to lysosomal dysfunction, autophagy inactivation, and cell injury (36,38). Consequences of these intracellular metabolic derangements may include impaired tubular albumin reabsorption and fragmentation (39–41).

Increased cathepsin D may represent a compensatory, protective response to elements of the diabetic milieu. Following siRNA-mediated blockage of cathepsin D transcription, AGE-BSA induced even greater elevations of IL-1 and KIM-1 in mouse proximal tubular epithelial cells. Greater cathepsin D production and activity at sites of tubulointerstitial damage, as we observed in BKBC and Pima Indian cohorts, respectively, may thus reflect an effort to attenuate actively injurious pathways. Greater cathepsin D and Lamp1 immunofluorescent staining in mouse proximal tubular epithelial cells exposed to AGE-BSA suggests lysosomal protein upregulation and potentially also cytoplasmic release of cathepsin D, such as via lysosomal membrane permeabilization.

It is possible that cathepsin D production may be triggered by the presence of AGEs or increased tubular albumin (resulting from glomerular filtration barrier disruption in diabetes) in an effort to increase their degradation, as both of these products undergo metabolism by cathepsin D (41,42). Along these lines, higher urine cathepsin D correlated with albuminuria in our JDRF cohort, as well as in other rat and human studies (27,43). Increased cathepsin D may also result from AGE-triggered lysosomal membrane permeabilization (44). Lysosomal membrane permeabilization results in lysosomal dysfunction and impaired autophagy, which contribute to cellular injury. Notably, overexpression of cathepsin D in kidney tubular epithelial cells in culture has been shown to protect cells from injury by attenuating AGE-mediated lysosomal membrane permeabilization and loss of mitochondrial membrane potential (45). Similarly, cathepsin D knockout mouse models demonstrate autophagy impairment resulting in increased susceptibility to ischemia/reperfusion injury and proteinuria development (46,47). However, in other studies with use of mouse models of acute and chronic kidney injury, cathepsin D was believed to directly contribute to injury, with inhibition reducing tubular cell damage and tubulointerstitial collagen deposition (29,30). While it is evident that homeostatic control of cathepsin D activity is important for normal lysosomal function and autophagy, more research is needed to elucidate the role played by cathepsin D in specific pathologic scenarios.

Our findings add to growing evidence of lysosomal dysfunction, indicated by change in cathepsin D expression and activity, as an important contributor to DKD pathophysiology (46,48). Elevated urine cathepsin D in people with diabetes who develop a rapid decline in eGFR and in those with lower eGFR and greater albuminuria may reflect heightened lysosomal stress and demand, which contributes to tubular injury and CKD progression. Proteinuria may trigger or aggravate tubular injury, accounting for why albuminuria attenuated the relationship between cathepsin D and rapid eGFR in the JDRF cohort (49).

Strengths of our work include the use of well-characterized cohorts with large sample sizes and longitudinal data and measurement of urine proteins with targeted mass spectrometry. Orthogonal studies further exploring the kidney histological correlates and potential mechanisms of cathepsin D in DKD strengthen our findings. Limitations of our work include the primarily White study population, which precludes extrapolation of findings to other racial and ethnic groups. eGFR slopes were used to define case-control status in lieu of clinical end points, though recent evident suggests that eGFR slope may serve as a viable surrogate end point for CKD progression (3). Regression to the mean may have caused misclassification of case and control subjects. Prolonged storage and inconsistencies in biosample collections across measurement batches may have influenced peptide measurements. Our follow-up cathepsin D cohort study sample sizes were relatively small and may have been subject to confounding.

In conclusion, we identified urine cathepsin D, a lysosomal protease, as a potential indicator of tubulointerstitial dysfunction associated with rapid eGFR decline in T1D. This work adds to existing evidence of an important role of lysosomal dysfunction in DKD. Further investigation of cathepsin D and other elements of the lysosomal/autophagy pathway are warranted to better understand how these can be used to facilitate DKD diagnosis and develop targeted therapies to prevent and treat DKD.

Acknowledgments. The authors are grateful for the participation of the study volunteers and for the effort and dedication of all the staff who took part in subject recruitment.

Funding. JDRF Network grant 3-SRA-2016-104 (PI:KS) provided major support for this study. The FinnDiane Study was funded by the Folkhälsan Research Foundation, the Wilhelm and Else Stockmann Foundation, the Liv och Hälsa Society, the Novo Nordisk Foundation (NNF OC0013659), the Helsinki University Hospital Research Funds, and the Academy of Finland (299200, 275614, 316664). The EDC study was funded by National Institutes of Health (NIH) grant DK34818 and the Rossi Memorial Fund. CACTI was funded by NIH grants R01HL113029, R01HL079611, RC1DK086958, and R01DE026480. C.P.L. was funded by NIDDK grants T32DK007467 and R01DK088762 and Northwest Kidney Centers. R.G.N. was supported by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), and the imaging of kidney tissue specimens was also supported in part by the American Diabetes Association (Clinical Science Award 1-08-42). L.N. and J.Z. were partially supported by NIDDK grant 1R01DK110541-01A1. The development of the urine proteomics assay was partially supported by the University of Washington Nutrition Obesity Research Center (P30 DK035816) and Diabetes Research Center (P30 DK017047). This study was also supported, in part, by the George M. O’Brien Michigan Kidney Translational Core Center, funded by NIH/NIDDK grant 2P30-DK-081943. The KPMP is funded by the following grants from the NIDDK: U2C DK114886, UH3DK114861, UH3DK114866, UH3DK114870, UH3DK114908, UH3DK114915, UH3DK114926, UH3DK114907, UH3DK114920, UH3DK114923, UH3DK114933, and UH3DK114937. I.H.d.B. has obtained research funding from the NIH and the American Diabetes Association. S.E.R. reports grants through the Joslin Diabetes Center with NIDDK. M.K. reports grants outside the submitted work through the University of Michigan with NIH and JDRF.

Duality of Interest. I.H.d.B. has received equipment and supplies for research from Medtronic and Abbott and consults for Boehringer Ingelheim and Ironwood. P.-H.G. has received investigator research grants from Eli Lilly and Roche and lecture honoraria from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Elo Water, Genzyme, Medscape, Merck Sharp & Dohme (MSD), Novartis, Novo Nordisk, and Sanofi. P.-H.G. is an advisor for AbbVie, AstraZeneca, Boehringer Ingelheim, Cebix, Eli Lilly, Janssen, Medscape, MSD, Novartis, Novo Nordisk, and Sanofi. P.R. has received consultancy and/or speaking fees (to his institution) from AbbVie, Astellas, AstraZeneca, Bayer, Boehringer Ingelheim, Eli Lilly, Gilead Sciences, Novo Nordisk, Vifor Pharma, and Sanofi and research grants from AstraZeneca and Novo Nordisk. K.S. is on the advisory board for Boehringer Ingelheim and Janssen and has received research support from Merck and Boehringer Ingelheim. S.E.R. reports grants and contracts through the Joslin Diabetes Center with Bayer and AstraZeneca. M.K. reports grants and contracts outside the submitted work through the University of Michigan with Chan Zuckerberg Initiative, AstraZeneca, Novo Nordisk, Eli Lilly, Gilead, Goldfinch Bio, Janssen, Boehringer Ingelheim, Moderna, European Union Innovative Medicine Initiative, Certa, Chinook, amfAR, Angion, Renalytix, Travere Therapeutics, Regeneron Pharmaceuticals, and IONIS and consulting fees through the University of Michigan from Astellas Pharma, Poxel, Janssen, and UCB. Additionally, M.K., W.J., and V.N. have a patent, PCT/EP2014/073413 “Biomarkers and methods for progression prediction for chronic kidney disease,” licensed. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.P.L. drafted the manuscript. E.V. conducted statistical analyses for the primary cohort study. C.P.L., V.D., D.M., W.J., V.N., and S.A. additionally conducted secondary statistical and bioinformatics analyses. V.D. led mouse and cell culture studies. A.N.H., K.S., and I.H.d.B. serve as primary investigators for this proteomics project and obtained funding. C.F. and N.S. contributed to the study design. T.S.A. participated in data management and cleaning. P.R., P.-H.G., J.K.S.-B., T.C., and T.J.O. serve as primary investigators of the four studied cohorts and obtained funding. S.S.W. serves as primary investigator for the BKBC. R.G.N. serves as primary investigator for the Pima Indian cohort, and M.K. serves as the principal investigator for the Pima cohort transcriptome data analysis. J.F.O., R.D.T., S.E.R., and S.S.W. serve as heads of KPMP recruitment sites. A.N.H., K.S., and I.H.d.B. supervised the project. All authors contributed to interpretation of results and read, edited, and approved the final manuscript. C.P.L. and E.V. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the American Society of Nephrology's Kidney Week 2020, Denver, CO, 20–25 October 2020.