Kidney organoids as a platform to study plasminogen-induced injury and uPA-based therapies Authors: Marina de Cos, Samuel Mon-Wei Yu, Jacob Wright, Nanditha Anandakrishnan, Kristin Meliambro, Jenny S Wong, Evren Azeloglu, Kirk N Campbell Published: 2026 ABSTRACT AbstractUrinary plasminogen has been implicated in the progression of proteinuric kidney diseases, yet its direct role in mediating injury along the nephron remains incompletely understood. Experimental models commonly used to study plasminogen-mediated damage are limited by incomplete phenotypic fidelity or interspecies differences, hindering translational relevance. Here, we used kidney organoids generated from H9 human embryonic stem cells to investigate the effects of plasminogen exposure and to evaluate urokinase-type plasminogen activator (uPA) inhibition with amiloride as a therapeutic strategy. Kidney organoids were exposed to plasminogen in the presence or absence of amiloride, and uPA activity was quantified using a fluorescence-based assay. Compartment-specific injury was assessed by confocal microscopy, automated image analysis, and AI-assisted nuclear segmentation, while targeted transcriptomic changes were evaluated using bulk quantitative PCR. Plasminogen exposure induced glomerular injury, characterized by reduced podocalyxin, nephrin, podocin, and WT1 expression, decreased glomerular cell density, and increased apoptosis. Proximal tubules exhibited reduced LTL and megalin expression, with increased KIM-1, whereas distal tubules were relatively preserved. Amiloride effectively inhibited uPA activity and partially restored podocyte and proximal tubular marker expression, but did not prevent apoptosis, indicating incomplete protection against plasminogen-induced injury. These findings support preferential plasminogen-induced injury along the glomerular–proximal tubular axis and highlight the utility of human kidney organoids for modeling this process, while underscoring the need for further validation and broader molecular profiling to strengthen their application in mechanistic studies and therapeutic discovery.Keywords: plasminogen, kidney organoids, podocyte injury, glomerular disease, proteinuria, urokinase-type plasminogen activator, amiloride Most kidney diseases are associated with the leakage of protein into the urine, where some of these proteins may actively worsen kidney damage. Plasminogen has been linked to faster progression of proteinuric kidney disease, but its effects on human kidney tissue have been difficult to study in models that fully reflect human biology. Using mini-kidneys grown from human stem cells, we show that plasminogen injures key filtering and reabsorbing cell types and that blocking its activation with an available pharmacologic agent is partially protective. This work establishes a human disease-modeling platform for studying toxic urinary proteins and for screening therapies. Introduction Proteinuria, defined as the pathological excretion of plasma proteins in the urine, is a central driver of kidney disease progression irrespective of the underlying etiology. Once disruption of the glomerular filtration barrier permits the urinary loss of medium-to-large plasma proteins, these filtered proteins actively induce and propagate injury along the nephron.1 Among these proteins, plasminogen, a circulating zymogen best known for its role in fibrinolysis, has emerged as a potential mediator of kidney injury in the proteinuric milieu, given its abundance and known roles in inflammation, tissue remodeling, and cell migration.2,3 In our previous work, we demonstrated that plasminogen directly injures human immortalized podocytes, inducing loss of focal adhesions and impaired cell spreading, phenotypes that resemble some of the features of foot process effacement in vivo.4,5 Notably, these effects were partially rescued by amiloride, an off-target inhibitor of urokinase-type plasminogen activator (uPA)6 which also reduced proteinuria in a puromycin aminonucleoside (PAN) rat model of nephropathy.4,7,8 Importantly, we recently showed that urinary plasminogen levels independently predicted renal disease progression in a large cohort of patients with glomerular disease, and others have reported its presence across diverse etiologies of kidney disease, underscoring the clinical relevance of this hypothesis.9–11 However, experimental systems used to study plasminogen-mediated podocyte injury have important limitations. Immortalized human podocytes, while highly tractable, incompletely recapitulate the differentiated podocyte phenotype and exhibit time- and culture-dependent loss of key slit-diaphragm and cytoskeletal proteins, limiting their physiological relevance.12 Animal models are further constrained by interspecies differences in the plasminogen–uPA axis, complicating direct translation to human disease.13 In addition, in vivo models inherently couple plasminogenuria to global proteinuria, making it difficult to isolate the specific contribution of filtered plasminogen from the broader toxic effects of protein overload on the nephron. Stem cell-derived kidney organoids, despite their inherent limitations, offer a powerful platform to model renal disease and study podocyte injury in a human-specific context. These 3-dimensional structures recapitulate key aspects of nephron architecture and enable mechanistic interrogation of injury responses to defined stimuli.14 A growing number of studies have used kidney organoids to model podocyte injury.15 Notably, organoids can simultaneously capture tubular responses to podocyte injury, enabling investigation of damage propagation along the glomerular–tubular axis. In this study, we sought to explore the utility of kidney organoids as a platform to model plasminogen-mediated injury and to evaluate uPA inhibition as a targeted therapeutic strategy. Methods Kidney organoid generation and treatment Kidney organoids were generated from H9 human embryonic stem cells (WiCell) and differentiated according to the protocol described by Morizane et al.16 After 21 days of differentiation, organoids were treated for 24 hour with 0.25 µM human plasminogen (Enzyme Research Laboratories, HPg2001), with or without 10 µM amiloride (MilliporeSigma, CAS 17440-83-4). Control organoids received equivalent volumes of vehicle (DMSO). Because treatments were performed under serum-free conditions, plasminogen was not present in the culture medium unless added exogenously. Day 21 was selected based on prior publications16 and internal maturation benchmarks, and only batches demonstrating robust nephron differentiation were included for treatment and downstream analyses. Details of quality control performed throughout differentiation are provided in the Supporting Information. Kidney organoid uPA activity assay The kidney organoid uPA activity assay was adapted from the cell-surface protocol described by Buckley et al.8 and is illustrated in Figure 1A. Briefly, organoids were transferred individually into separate wells of black-wall, clear-bottom 96-well plates, incubated with 25 nM high–molecular-weight active human uPA (molecular innovations, SKU: HTC-UPA), washed with assay buffer, and then exposed to serial dilutions of amiloride or vehicle (DMSO). Both control and amiloride-treated conditions received identical concentrations of exogenous uPA. The reaction was initiated by addition of 1 mM urokinase fluorescent substrate III (Calbiochem, Cat. #672159) and monitored using a SpectraMax M3 plate reader (acquisition parameters shown in Supporting Information.). Reaction linearity over the acquisition period was confirmed (Figure 1B), and changes in fluorescence over 10-minute intervals were used to calculate IC50 values. To account for variability in baseline uPA activity and degree of differentiation, activity levels were normalized within each organoid to its own baseline. Assays were performed using 3 independent biological replicates, each analyzed in technical triplicate, and data were log10-transformed and background-corrected prior to analysis. Inhibition of uPA activity by amiloride in kidney organoids. (A) Schematic representation of the uPA activity assay (created with BioRender). The experiment was performed using 3 independent biological replicates. Each biological replicate included 3 technical replicates (1 organoid/well) per condition across 7 amiloride concentrations or dose-matched DMSO controls, in addition to a non-DMSO-treated control. (B) Blank-corrected fluorescence intensity from non-DMSO control organoids was monitored over 60 minutes to confirm reaction linearity and validate the assay for quantitative analysis (see Figure S1.I). Each line represents fluorescence kinetics from 1 well over time (n = 3 technical replicates). (C) Results from a representative uPA assay showing Δ fluorescence per 10 minutes as a function of amiloride concentration compared with DMSO control, including technical triplicates. Each line represents one of the 3 technical replicates. (D) Compiled uPA activity data showing normalized uPA activity (uPA/min) as a function of amiloride concentration (red) versus DMSO (blue). Data were fitted using a nonlinear variable-slope model [log(inhibitor) vs. normalized response]. Points represent the mean of 3 biological replicates and error bars indicate the SD. DMSO, dimethyl sulfoxide; uPA, urokinase-type plasminogen activator. Confocal microscopy and image analysis Kidney organoids were cryosectioned at 6 μm thickness, immunostained, and imaged as confocal z-stacks using 20× or 40× objectives on a Nikon AX/AX R confocal microscope with electronic scan zoom applied during acquisition. Prior to export, images were preprocessed by generating maximum intensity projections (MIPs) and applying denoising using Nikon NIS-Elements NX. Segmentation of kidney organoid compartments into glomerular and tubular regions was performed using CellProfiler v4.2.6 (see Supporting Information for details). Segmentation was fully automated and unbiased, and overlay images were reviewed by a blinded investigator to confirm accuracy. For each parameter analyzed, at least 3 independent biological replicates were included, each comprising 8-12 organoids per condition. Data were expressed as the mean per organoid after normalization to the mean of the corresponding batch control, allowing integration across differentiation batches. Statistical analyses were performed at the organoid level following batch normalization to minimize inter-differentiation variability. AI-based nuclear segmentation Due to poor segmentation of individual nuclei using traditional thresholding methods, Cellpose v4.0.7 in combination with the Cellpose-SAM model—a deep-learning algorithm for robust and generalizable cell and nuclear segmentation—was employed.17,18 MIP nuclear images were normalized for background intensity and used as input for the nuclei model in Cellpose, with a target object diameter of 50 pixels. The segmentation workflow generated 16-bit labeled mask images, with each nucleus assigned a unique integer identifier corresponding to an individual segmented object. Following segmentation, the Cellpose-generated nuclear masks were imported into CellProfiler v4.2.6 for downstream analysis, where nuclei were mapped to their corresponding tubular and glomerular compartments, as defined by the CellProfiler compartment segmentation pipeline described above (see Supporting Information for details). RNA extraction and qPCR 3D organoids were washed with cold PBS and homogenized in TRIzol. RNA was extracted using the RNeasy Mini Kit (Qiagen, Cat #74104). Six biological replicates were performed, with each replicate consisting of pooled RNA from 8 to 12 organoids per treatment condition within the same experiment and differentiation batch, following confirmation of adequate nephron differentiation. Technical qPCR replicates were averaged before analysis. Target-gene expression was normalized to GAPDH, and relative expression was calculated using the 2−ΔΔCt method, with the corresponding batch-matched control used as the calibrator. Data are displayed as fold change relative to control. The list of primers used is shown in the Supporting Information. Statistical analysis Analyses were conducted using GraphPad Prism 10.4.1. Data are presented as mean ± SD. uPA assay data were analyzed with nonlinear regression (log[inhibitor] vs normalized response, variable slope). Group comparisons were performed using 1-way analysis of variance with Tukey post hoc correction. P < .05 was considered significant. Ethics approval This study did not involve human participants, identifiable human material, or animals, and therefore did not require Institutional Review Board approval or informed consent. The research used the commercially available, established human embryonic stem cell line H9 (WA09; WiCell Research Institute), which is registered on the NIH Human Embryonic Stem Cell Registry (NIHhESC-10-0062). No new embryonic stem cell lines were derived. Results uPA activity assay The results of the uPA fluorogenic assay used to validate uPA inhibition by amiloride in our experimental model are shown in Figure 1. After confirming reaction linearity over the acquisition period (Figure 1B), amiloride significantly reduced uPA activity compared with DMSO (Figure 1C), with an IC50 of 3.91 µM (95% confidence interval [CI], 3.59-4.23 µM) (Figure 1D), supporting the utility of the assay and effective inhibition of uPA by amiloride. Confocal microscopy and high-content analysis Exposure to plasminogen was associated with evidence of glomerular injury within podocalyxin-positive (PDXL+) regions. Specifically, plasminogen reduced podocalyxin signal intensity and decreased cellular density (cells/µm²) within PDXL+ areas compared with controls. In parallel, plasminogen also affected tubular compartments: proximal tubules (PT), identified by LTL staining, showed a reduction in LTL expression, whereas PT cell density was not significantly altered. In contrast, distal tubules (DT), labeled by E-cadherin, did not exhibit detectable changes in either marker expression or cell density following plasminogen treatment, suggesting compartment-specific susceptibility to injury (Figure 2). Renal compartment–specific effects of plasminogen ± Amiloride Assessed by Immunofluorescence. (A) Representative confocal images stained for podocalyxin (glomeruli), lotus tetragonolobus lectin (LTL; proximal tubules), and E-cadherin (distal tubules). Scale bar, 75 μm. (B) Mean organoid fluorescence intensity of podocalyxin, LTL, and E-cadherin. For each organoid, fluorescence intensity values were averaged across all segmented structures of the corresponding compartment. Each point represents 1 organoid, normalized to the mean control value of the corresponding differentiation batch. Bars represent mean ± SD. (C) Mean cell density, expressed as nuclei counts per μm², in glomerular, proximal tubular, and distal tubular regions. For each organoid, nuclei density values were averaged across all segmented structures of the corresponding compartment. Each point represents 1 organoid, normalized to the mean control value of the corresponding differentiation batch. Bars represent mean ± SD. (D) Cleaved caspase-3 expression in organoids exposed to plasminogen ± amiloride. Representative confocal images are shown (merged and single-channel cleaved caspase-3). Scale bar, 75 μm. (E) Mean cleaved caspase-3 intensity in glomerular (podocalyxin+) and proximal tubular (LTL+) regions. For each organoid, cleaved caspase-3 fluorescence intensity values were averaged across all segmented structures of the corresponding compartment. Each point represents 1 organoid, normalized to the mean control value of the corresponding differentiation batch. Bars represent mean ± SD. AML, amiloride; CCAS3, cleaved caspase-3; CTRL, control; E-CAD+, E-cadherin, LTL, lotus tetragonolobus lectin; PLS, plasminogen. To assess whether plasminogen exposure induced cell death, cleaved caspase-3 was quantified as a marker of apoptosis (Figure 2D). Mean cleaved caspase-3 levels were significantly increased in both PDXL+ (glomerular) and LTL+ (PT) regions relative to controls, indicating activation of apoptotic pathways in these compartments (Figure 2E). Notably, co-treatment with amiloride did not attenuate the increase in cleaved caspase-3 in either PDXL+ or LTL+ regions. Consistent with the absence of structural changes in DT, cleaved caspase-3 levels in E-cadherin–positive areas were not significantly different from controls after plasminogen exposure (Figure S2). Given the significant changes within glomerular and PT regions, we next performed additional marker-based analyses to further characterize podocyte and PT injury (Figure 3). Quantification of nephrin, podocin, and WT1 demonstrated that plasminogen significantly reduced all 3 markers compared with controls, supporting a podocyte injury phenotype. Co-treatment with amiloride partially restored podocin and WT1 expression, but not nephrin, indicating partial preservation of podocyte marker expression under uPA inhibition (Figure 3A and B). Additional analysis of LTL+ tubular regions showed increased KIM-1 and reduced megalin expression following plasminogen exposure, supporting PT injury beyond changes in LTL staining. Amiloride produced only minimal improvement in megalin expression and did not attenuate KIM-1 induction (Figure 3C and D). Effects of plasminogen ± amiloride on additional podocyte markers and apoptosis. (A) Representative confocal images of glomeruli stained for nephrin, podocin, and WT1. Scale bar, 35 μm. (B) Mean organoid fluorescence intensity of nephrin, podocin, and WT1. For each organoid, fluorescence intensity values were averaged across all segmented glomerular structures. Each point represents 1 organoid, normalized to the mean control value of the corresponding differentiation batch. Bars represent mean ± SD. (C) Representative confocal images of LTL+ proximal tubular regions stained for KIM-1 and megalin. Scale bar, 75 μm. (D) Mean organoid fluorescence intensity of KIM-1 and megalin in LTL+ tubular regions. For each organoid, fluorescence intensity values were averaged across all segmented LTL+ structures. Each point represents 1 organoid, normalized to the mean control value of the corresponding differentiation batch. Bars represent mean ± SD. AML, amiloride; CTRL, control; KIM-1, kidney injury molecule-1; LTL, lotus tetragonolobus lectin; PLS, plasminogen; WT1, Wilms tumor 1. Transcriptomic changes assessed by qPCR At the transcript level, plasminogen exposure led to a coordinated reduction in the expression of key podocyte-associated genes, including NPHS1, NPHS2, PDXL, and WT1, together with decreased expression of the PT-associated genes HNF4, LRP2, and CUBN. HAVCR1 expression showed a non-significant increase after plasminogen exposure. In contrast, expression of SYNPO and DT markers (CDH1, CDH16, and SCNN1A) were not significantly altered, indicating a selective transcriptional response in glomerular and PT compartments. Consistent with the imaging-based analyses, co-treatment with amiloride partially or fully restored the expression of the plasminogen-sensitive genes, suggesting partial reversal of plasminogen-induced transcriptional changes, although this rescue was highly variable across biological replicates (Figure 4). Bulk qPCR analysis of organoids treated with plasminogen ± amiloride. Gene expression was normalized to GAPDH and is presented as fold change relative to the corresponding batch-matched control, calculated using the 2−ΔΔCt method. Each point represents 1 independent biological replicate (n = 6), consisting of pooled RNA from 8 to 12 organoids per treatment condition within the same differentiation experiment. Bars represent mean ± SD. Statistical testing was performed on the corresponding − ΔΔCt values using a 1-way repeated-measures analysis of variance followed by Tukey’s multiple-comparisons test. Bars represent mean ± SD. AML, amiloride; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; PLS, plasminogen. Discussion These results demonstrate that plasminogen exposure induces coordinated podocyte and PT injury in human kidney organoids. Plasminogen reduced the expression of key structural podocyte proteins, decreased podocyte number, and increased apoptosis, phenotypes that resemble podocyte dysfunction observed in patients with proteinuric glomerular diseases. PT injury was supported by reduced LTL and megalin staining, downregulation of HNF4, LRP2, and CUBN, and increased KIM-1 staining, paralleled by a trend toward increased HAVCR1 by qPCR. In contrast, DT markers remained unchanged, indicating relative sparing of distal segments. This pattern suggests preferential involvement of the glomerular–PT axis in this model, while distal nephron effects of plasminogen described in nephrotic syndrome may primarily reflect ENaC activation rather than direct distal tubular injury.8,19–21 Our findings extend previous clinical observations linking urinary plasminogen to kidney disease progression by providing mechanistic evidence in a human-derived experimental system. Although kidney organoids do not recapitulate proteinuria physiologically, given the absence of vascularization, a functional filtration barrier, and compartmentalized delivery of filtered proteins, their uniform exposure to plasminogen enables direct interrogation of how this proteinuric factor may affect human renal epithelial cells that are not normally exposed to it. Future incorporation of microfluidic flow and organ-on-chip technologies may further improve the physiological relevance of these systems by enabling spatially and temporally controlled delivery of proteinuric factors.14,22,23 Together, these advances support kidney organoid–based platforms as relevant models for studying plasminogen-mediated injury and evaluating targeted therapeutic strategies. The partial rescue observed after uPA inhibition is consistent with previous findings by us and others.4,7 Although amiloride partially restored podocyte and, to a lesser extent, PT marker expression by qPCR and immunofluorescence, it failed to prevent plasminogen-induced apoptosis, indicating that injury-related changes persisted despite uPA inhibition. Moreover, amiloride-mediated rescue was variable across biological replicates, particularly by qPCR, whereas plasminogen-induced injury was more consistent. Together, these findings suggest that plasminogen-driven toxicity may involve a complex injury response only partially modulated by uPA inhibition. However, the current data are not sufficient to definitively define distinct uPA-dependent and uPA-independent pathways, and broader transcriptomic approaches will be needed to better characterize plasminogen-mediated injury and guide the development of more selective interventions. Kidney organoids may be particularly valuable for this purpose, especially when used in conjunction with complementary in vivo systems. Several limitations should be acknowledged. First, all experiments were performed using organoids from a single human embryonic stem cell line (H9) and the findings should be interpreted as H9-specific until validation in additional cell lines. Kidney organoids also lack fully mature glomerular architecture, a functional filtration barrier, and immune components, all of which influence how proteinuric injury is initiated and propagated in vivo.24 In addition, the immature state, inter-organoid variability, and relatively proximalized nephron phenotype inherent to organoid differentiation protocols remain important challenges, warranting caution when interpreting segment-specific injury patterns, such as the apparent sparing of DT injury observed in this study. To minimize bias in our study, multiple organoids were included per condition and predefined quality-control criteria were applied before downstream analysis, as described in the Supplementary. Future approaches such as genetically encoded reporter lines and longitudinal live imaging may enable more precise tracking of injury dynamics over time. An additional limitation is that we did not perform a detailed analysis of the effect of amiloride treatment alone on podocyte/ PT marker expression, cell density, or cleaved caspase-3 at baseline/without plasminogen treatment. Finally, our analyses were limited to bulk qPCR and immunostaining, and more advanced transcriptomic approaches will be important to better resolve segment-specific responses, cellular heterogeneity, and the molecular mechanisms underlying plasminogen-mediated injury. Integration of these organoid-based data with human tissue and clinical datasets will be essential to validate the translational relevance of this model.25 Despite these constraints, our study provides proof-of-concept evidence that H9-derived human kidney organoids can recapitulate key features of plasminogen-mediated injury and offers a new framework for studying mechanisms of proteinuric kidney disease progression. This platform may facilitate the identification and testing of targeted therapies aimed at interrupting plasminogen-driven injury pathways in proteinuric kidney disorders. Supplementary Material Contributor Information Marina de Cos, Dr. Barbara T. Murphy Division of Nephrology, Icahn School of Medicine at Mount Sinai, New York, NY, United States. Samuel Mon-Wei Yu, Dr. Barbara T. Murphy Division of Nephrology, Icahn School of Medicine at Mount Sinai, New York, NY, United States. Jacob Wright, Dr. Barbara T. Murphy Division of Nephrology, Icahn School of Medicine at Mount Sinai, New York, NY, United States. Nanditha Anandakrishnan, Dr. Barbara T. Murphy Division of Nephrology, Icahn School of Medicine at Mount Sinai, New York, NY, United States. Kristin Meliambro, Dr. Barbara T. Murphy Division of Nephrology, Icahn School of Medicine at Mount Sinai, New York, NY, United States. Jenny S Wong, Division of Renal-Electrolyte and Hypertension, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, United States. Evren Azeloglu, Dr. Barbara T. Murphy Division of Nephrology, Icahn School of Medicine at Mount Sinai, New York, NY, United States. Kirk N Campbell, Division of Renal-Electrolyte and Hypertension, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, United States. Author contributions Nanditha Anandakrishnan (Data curation [equal], Formal analysis [equal], Methodology [equal], Validation [equal], Writing—review & editing [equal]) Conception and design: Marina de Cos, Jenny S. Wong, Evren Azeloglu, Nanditha Anandakrishnan, Kristin Meliambro, and Kirk N. Campbell. Financial support: Evren Azeloglu and Kirk N. Campbell. Collection and/or assembly of data: All authors. Data analysis and interpretation: All authors. Manuscript writing: Marina de Cos, Samuel Mon-Wei Yu, Jacob Wright, Evren Azeloglu, Nanditha Anandakrishnan, and Kirk N. Campbell. Final approval of manuscript: All authors. Supplementary material Supplementary material is available at Stem Cells Translational Medicine online. Funding This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health under award number R01DK126477 to K.N.C. Conflict of interest The authors declare no potential conflicts of interest. Data availability The data underlying this article are available in the article and in its online supplementary material. References Associated Data Supplementary Materials Data Availability Statement The data underlying this article are available in the article and in its online supplementary material.