An integrated organoid omics map extends modeling potential of kidney disease

Abstract

Kidney organoids are a promising model to study kidney disease, but their use is constrained by limited knowledge of their functional protein expression profile. Here, we define the organoid proteome and transcriptome trajectories over culture duration and upon exposure to TNFα, a cytokine stressor. Older organoids increase deposition of extracellular matrix but decrease expression of glomerular proteins. Single cell transcriptome integration reveals that most proteome changes localize to podocytes, tubular and stromal cells. TNFα treatment of organoids results in 322 differentially expressed proteins, including cytokines and complement components. Transcript expression of these 322 proteins is significantly higher in individuals with poorer clinical outcomes in proteinuric kidney disease. Key TNFα-associated protein (C3 and VCAM1) expression is increased in both human tubular and organoid kidney cell populations, highlighting the potential for organoids to advance biomarker development. By integrating kidney organoid omic layers, incorporating a disease-relevant cytokine stressor and comparing with human data, we provide crucial evidence for the functional relevance of the kidney organoid model to human kidney disease.

Fig. 1. Proteome of kidney organoids evolves with duration in culture.

a Volcano plot of proteomic differential expression analysis (log₂ fold change of label-free quantification intensity comparing D29 with D21). Proteins are represented by dots and the black line illustrates the significance cut-off (two sided t test, FDR < 0.05 and s0 = 0.1). Red colored proteins meet the significance threshold. Examples of strongly regulated proteins are labeled. Blue colored and labeled dots represent proteins highlighted in (b). b Immunofluorescence imaging of sectioned kidney organoids showing expression of (top panel) podocyte markers nephrin (NPHS1) and synaptopodin (SYNPO), and (bottom panel) cell structure and differentiation markers smooth muscle actin (ACTA2) and platelet-derived growth factor receptor alpha (PDGFRA), plus nuclear marker DAPI, n = 3, representative images shown, scale bar: 50 µm. c Heatmap (maximum distance) of normalized protein expression (mean subtracted label-free quantification values (average of three replicates)) during organoid differentiation. The zoomed-in region expands on the clusters that undergo marked changes during differentiation. Five clusters (of size >25 proteins) were distinguished and are indicated with colored rectangles. These five clusters are shown in (d). d GO enrichment analysis of unbiased clustering (clusters >25 proteins) of proteins during organoid differentiation (D21–D29) (Fisher’s exact test, FDR < 0.05). Source data are provided as a Source Data file.

Fig. 2. Integrated expression analysis of proteome-transcriptome trajectories over organoid culture duration.

a Scatterplot of RNA copy number (D21, D25, D29) and protein copy number with light blue color indicating high data point density and green for low density. Associated two-dimensional UniProt-keyword enrichment with basic functional terms, including ‘Differentiation’, ‘Glycolysis’, ‘Mitochondrion’, ‘TCA-cycle’ and ‘Protein-biosynthesis’ are highlighted in red. b Barplots of basic functional terms from 2D UniProt-keyword enrichment (from a). c Uniform Manifold Approximation and Projection (UMAP) representation of combined datasets from single cell transcriptomes of D25 untreated plus D24 TNFα-treated and vehicle control (VC) distinguished 14 cell type clusters. Visualization was carried out using Cell x Gene software; the contribution to the total cell number by each sample and cell type cluster is shown on the right. d The summed protein expression direction of corresponding transcript markers which were used to define the 14 cell clusters. Proteins were classified as overexpressed (FDR < 0.05 and log₂ fold-change >0), under-expressed (FDR < 0.05 and log₂ fold-change <0) or not differentially expressed (FDR ≥ 0.05) between D21 and D29. e Heatmap k-means clustering of bulk RNA and bulk protein of transcript cell-type markers for early glomerular epithelial 1 (EGE1), maturing podocyte (Podo), proximal tubular (PT) and stromal cells. The top 30 differentially expressed proteins D29 versus D21 (FDR < 0.01) were plotted. Source data are provided as a Source Data file.

Fig. 3. Organoid proteome organization shows similarities with human kidneys.

a Venn diagram of the organoid proteome compared with microdissected single glomeruli and tubules proteomes. Four compartments indicate proteins identified in (i) single glomeruli only, (ii) both single glomeruli and organoids, (iii) single tubules only, and (iv) both single tubules and organoids. b Over-representation plots corresponding to Venn diagram compartments in (a) for Gene Ontology (GO) terms related to biological process (BP), molecular function (MF), cellular component (CC), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (Fisher’s exact test, adjusted p-value < 0.05). The x-axis corresponds to the combined score from EnrichR analysis. c Venn diagram (top) with associated over-representation plot (bottom) of terms uniquely mapping to the organoid proteome but not to cultured podocyte proteome. d Protein copy number based on intensity based absolute quantification (IBAQ) of podocyte markers expressed in organoids (n = 3) compared to human tissue (n = 2), mouse tissue (n = 3) and cultured human podocytes (undifferentiated and differentiated, both n = 3). Human (H); Mouse (M). Data are represented as mean ± SEM. Source data are provided as a Source Data file. Source data are provided as a Source Data file.

Fig. 4. TNFα treatment significantly alters protein expression and secretion in kidney organoids.

a UMAP highlighting 6 kidney cell type clusters identified in Fig. 2c, representing 60% of cells from a total of 35,036 analyzed organoid cells. b Dot plot showing transcript expression of TNFα-responsive receptors TNFRSF1A and TNFRSF1B in kidney cell clusters in (a). Bar length and associated numbers indicate the number of cells in the clusters. c Immunofluorescence imaging of sectioned kidney organoids showing expression of TNFRSF1A (TNFRSF1A) in co-staining with either synaptopodin (SYNPO, top row), stromal cell marker Meis Homeobox 1/2 (MEIS1/2, middle row), or tubular epithelial cell marker N-cadherin (CDH2, bottom row). DAPI, nuclear marker; n = 3, representative images shown, scale bar: 50 µm. d Differential proteome expression analysis (log₂ fold change of 24 h and 48 h) in cell lysates of day 25 TNFα-treated organoids compared with VC (two-sided t test, FDR < 0.1 and s0 = 0.1). e All quantified cell cluster marker proteins categorized into over-expressed, under-expressed or not differentially expressed upon TNFα stimulation 48 h vs VC. f Differential expression analysis (log₂ fold change) of 24 h and 48 h day 25 TNFα-stimulated organoid supernatant compared with VC. g Expression of CXCL10 transcript and protein following treatment with recombinant TNFα for 24 h and 48 h. Data are represented as mean ± SEM of 5 independent experiments for qPCR and ELISA on culture media (top left and lower right graphs) and 3 independent experiments for ELISA on lysates (top right graph). Unpaired t test. VC, vehicle control. Source data are provided as a Source Data file.

Fig. 5. Proteome-based gene signature of TNFα activation in organoids identifies group of individuals with poorer clinical outcomes in proteinuric kidney disease.

a Schematic defining origins of Organoid and Tissue TNF gene signatures (left column) as well as Organoid TNF score (bottom middle column) used in this manuscript, with special attention to FSGS/MCD cohort with the NEPTUNE study (bottom left column) demonstrating poorer outcomes and higher TNF activity for individuals in cluster 3 as in Mariani et al. (2023); right column shows strategy for identifying potential biomarker candidates using Organoid TNF signature genes. b Z-scores showing summary expression of organoid TNF signature genes in human kidney diseases, generated from ERCB microarray data from microdissected human kidney biopsy tissue (boxplots for the 268 in ERCB of the total 322 genes assessed, Supplementary Data 10). Plots: median, boxes 25–75% percentile, whiskers represent the min. to max. values. Numbers of individual samples are indicated in parentheses. LD = Living Donor. Unpaired t test. c Venn diagram comparing gene sets of Organoid and Tissue TNF signatures (Fig. 4d, f, Supplementary Data 10). d Literature‐based network generated from the combination of Organoid and Tissue TNF signature genes (total 584). e, f box and whisker plots (white line, mean; box, 75%; whiskers, 90%) showing summary gene expression in diseased human kidney based on the Organoid TNF signature score in FSGS/MCD clusters (319 of the 322 genes detected) as in (a), and divided by proteins in organoid secretome (top, all 22 genes) and cell lysates (bottom, 299 of 302 genes); 2 proteins were expressed in both; ***p < 1.85 × 10E−11. Unequal variance two-tailed t test. g Dot plots of 10 TNF signature overlap gene expression from (c) in individuals with FSGS/MCD separated by TNFα activity status, generated from snRNA-seq data. h Dot plots of CXCL10 expression separated by diseased kidney TNFα activity status and cell type. Arrows highlight cell types of interest with higher expression in individuals with high TNFα status; cell type cluster names as in Supplementary Fig. 5C. a, c created using BioRender. Source data are provided as a Source Data file.

Fig. 6. Organoid omics data identify potential biomarkers of kidney disease.

a Expression of C3 and VCAM1 from Fig. 5c increased in D25 kidney organoids treated with TNFα for 24 h and 48 h, as measured in cell lysates by qRT-PCR (top) and in organoid culture supernatants by ELISA (bottom). Means of 3 separate experiments indicated by bold horizontal lines, with SEM error bars. Unpaired t test. b Dot plots and UMAPs of C3 and VCAM1 expression in kidney cell types from scRNA-seq analysis of TNFα-treated D24 kidney organoids (Fig. 4a). See also, Supplementary Fig. 6A, B. c Expression of C3 and VCAM1 in kidney organoids co-treated with TNF receptor inhibitor, R-7050, starting 1 h prior to treatment with TNFα for 24 h, as measured in cell lysates by qRT-PCR (top) or in organoid culture supernatants by ELISA (bottom). Results were reported as a percentage relative to TNFα alone. Data are represented as mean ± SEM of 3 independent experiments. Unpaired t test. d Mean C3 and VCAM1 expression in the same bulk transcriptional profile clusters from diseased human kidney tissues as in Fig. 5e. *p < 0.01. Unequal variance two-tailed t test. e Top: dot plots of C3 and VCAM1 expression by cell type in diseased human kidney generated from snRNA-seq analysis (Fig. 5g, h). Bottom: dot plots of the cell type cluster with the highest C3 and VCAM1 expression (DTL – distal thin limb) separated by TNFα status as in Fig. 5g, h. Cell cluster names as in Supplementary Fig. 5C. f Immunofluorescence imaging of sectioned human kidney biopsies of patients with FSGS showing expression of VCAM1 in the descending thin limb (DTL) compartment in patient 2 (eGFR = 39 mL/min/1.73 m²) relative to patient 1 (eGFR = 102 mL/min/1.73 m²) where no VCAM1 expression was observed. DTL segments were defined as AQP1+/LTL- tubules within the medullary region (Supplementary Fig. 8). DAPI, nuclear stain. Staining conditions were optimized in human nephrectomy tissue. 10 patient samples were stained once due to limited human biobank samples. Representative images of four patients are shown here and in Supplementary Fig. 8. Source data are provided as a Source Data file.

Fig. 7. Omics-expanded organoid modeling potential of kidney disease.

a Number of proteins associated with monogenic kidney disease expressed in kidney organoids. b Heatmap of protein expression of genes associated with nephrotic syndrome expressed during organoid differentiation, D21 to D29 (visualization using Kidney Disease Explorer https://kidneyapp.shinyapps.io/kidneyorganoids/). c Protein–protein interaction networks created with STRING database (V.11.0) and Cytoscape (V 3.8.2), demonstrating differences in extracellular matrix protein expression in aging (upper left) and TNFα-treated (upper right) organoids. As comparisons, corresponding networks of human FSGS tissue vs. control (lower left) as well as puromycin aminonucleoside (PAN)-treated rat kidney tissue vs. control (lower right) are illustrated. d Schematic of how organoids can contribute to our understanding of kidney disease through the integration of pre-clinical models with human data to drive better outcomes for individuals with kidney disease. d created using BioRender. Source data are provided as a Source Data file.