A scalable organoid model of human autosomal dominant polycystic kidney disease for disease mechanism and drug discovery

SUMMARY

Human pluripotent stem cell-derived organoids are models for human development and disease. We report a modified human kidney organoid system that generates thousands of similar organoids, each consisting of 1 to 2 nephron-like structures. Single-cell transcriptomic profiling and immunofluorescence validation highlighted patterned nephron-like structures, utilizing similar pathways, with distinct morphogenesis, to human nephrogenesis. To examine this platform for therapeutic screening, the polycystic kidney disease genes PKD1 and PKD2 were inactivated by gene-editing. PKD1 and PKD2 mutant models exhibited efficient and reproducible cyst formation. Cystic outgrowths could be propagated for months to centimeter-sized cysts. To shed new light on cystogenesis, 247 protein kinase inhibitors (PKI) were screened in a live imaging assay identifying compounds blocking cyst formation but not overall organoid growth. Scaling and further development of the organoid platform will enable a broader capability for kidney disease modeling and high-throughput drug screens.

INTRODUCTION

The human kidney performs the vital task of filtering blood through a network of around a million nephrons, maintaining the homeostasis of tissue fluid (Nyengaard and Bendtsen, 1992; Sasaki et al., 2019). Hormonal interactions also regulate hematopoiesis, blood pressure and bone composition. As a consequence, the loss of renal activity in chronic disease results in secondary complications including anemia, increased cardiovascular risk, bone disorders, and nutritional problems (Thomas et al., 2008). Chronic kidney disease affects up to 15% of the adult population in the United States (Johansen et al., 2021). Treatment options remain largely limited to dialysis and kidney transplantation, casting a burden on the healthcare system. Improving approaches to better understand and model human kidney disease is an important priority.

Multiple research groups have pioneered the generation of kidney organoid systems (Czerniecki et al., 2018; Freedman et al., 2015; Kumar et al., 2019; Lam et al., 2014; Low et al., 2019; Morizane et al., 2015; Taguchi and Nishinakamura, 2017; Taguchi et al., 2014; Takasato et al., 2014, 2015). In these model systems, human pluripotent stem cells (PSCs) differentiate into organized 3-D structures comprising multiple kidney cell types of the nephron lineage and potentially the ureteric epithelial collecting system lineage (Howden et al., 2019; Taguchi and Nishinakamura, 2017; Uchimura et al., 2020; Howden et al., 2021). Kidney organoids have been used to explore developmental processes (Czerniecki et al., 2018; Low et al., 2019; Przepiorski et al., 2018; Taguchi et al., 2014), examine physiological activities (van den Berg et al., 2018; Kumar et al., 2019; Low et al., 2019), model disease (Forbes et al., 2018; Freedman et al., 2015; Hale et al., 2018; Kim et al., 2017; Low et al., 2019) and identify disease modulators (Czerniecki et al., 2018). Grafted organoids have been reported to generate a renal filtrate (van den Berg et al., 2018; Low et al., 2019; Xinaris et al., 2012).

Despite the burden of kidney disease, only modest progress has been made in identifying effective drugs. Autosomal dominant polycystic kidney disease (ADPKD), with ~93% of the cases attributed to mutations in the PKD1 or PKD2 gene, is the most common monogenic cause of end-stage kidney disease and amongst the most common autosomal dominant gene mutations in the human population (Harris and Torres, 2018). Until recently, there has been no pharmacological intervention to help ADPKD disease patients. In 2018, the US Food and Drug Administration approved Tolvaptan, an AVPR2 inhibitory drug that slowed disease progression in a subset of ADPKD patients (Beaudoin et al., 2019; Higashihara et al., 2011; Hopp et al., 2015; Reif et al., 2011; Torres et al., 2016, 2017). AVPR2 is restricted to a subset of principal cell and principal cell-like epithelial cell types in the connecting region of the nephron and collecting system (Ransick et al., 2019; KidneyCellExplorer). Broader acting drugs are needed to target both cyst initiation and cystic growth throughout the nephron and collecting system.

Several groups have reported cystic growth in PKD1 and PKD2 mutant kidney organoid cultures where rare cystic outgrowths are augmented by forskolin, consistent with cAMP levels driving cystogenesis (Cruz et al., 2017; Czerniecki et al., 2018; Freedman et al., 2015; Kuraoka et al., 2020). Others have failed to observe cyst formation in vitro in a different organoid model system (Kumar et al., 2019). To date screens have been low-throughput, relying on manually displacing organoids from a retaining extracellular gel, and focused on identifying small molecules enhancing cyst formation (Czerniecki et al., 2018). Though cyst suppressing screens using mouse 3D cysts have been reported (Asawa et al., 2020; Booij et al., 2017; Booij et al., 2020; Chang et al., 2018), there are no reports of human organoid-based screens that could potentially identify therapeutic small molecule candidates for ADPKD.

Cost, scalability and reproducibility are further challenges to the wider adoption of kidney organoid models. Here, we developed, characterized and validated a simple and readily scalable, cost-effective kidney organoid platform for modeling ADPKD, identifying novel compounds blocking cyst initiation and inhibiting cyst expansion in both PKD1 and PKD2 mutant models.

Results

A scalable platform for generating human kidney organoids with nephron-like structures

To achieve a reproducible, large-scale production of 3-D kidney organoids, we utilized commercially available EZSPHERE 12-well plates, which are constructed using laser-based microfabrication to contain uniform microwells of 800-μm diameter and 400-μm depth (Sato et al., 2016). As each well contains 420 microwells, a single plate can generate over 4,000 organoids. At differentiation day 8 (dd8), cells were dissociated into single cells, and 600,000 cells were reseeded in each well of the EZSPHERE plate to produce about 400 3-D aggregates comprising approximately 1,500 cells per aggregate in each well (Figure 1A). We used a validated MAFB-P2A-eGFP H9 hESC line (Tran et al., 2019) to visualize the formation of podocyte-like cells in the organoids, and observed the emergence of eGFP+ cells at day 14 of differentiation (Figure 1B), a similar timing to published kidney organoid platforms which typically use many more cells (100,000 or greater) to seed organoid formation in 96-well plates - (Tran et al., 2019). At dd25, each organoid in the EZSPHERE platform had 1 or 2 eGFP+ clusters, reflecting the formation of 1 to 2 nephron-like structures per organoid (Figure 1B).

Figure 1: A Scalable Platform for Human Kidney Organoid Production (see also Figure S1).

(A) Schematic diagram of directed differentiation to generate kidney organoids.

(B) Brightfield (grey scale) and fluorescent (green) images of kidney organoids derived from MAFB-P2A-eGFP H9 hESC. Scale bars indicate 50 μm unless labeled differently.

(C to F) Immunofluorescent analyses of MAFB-P2A-eGFP organoids at various differentiation time points. Scale bars indicate 50 μm.

(G and H) qPCR analyses of MAFB, SLC3A1, SLC12A1 and GATA3 expression in individual organoids from 2 batches of differentiation (batch 1: no. 1 to 10; batch 2: no. 11 to 20) at dd16 (G) and dd28 (H).

A number of recent studies have provided insight into human nephrogenesis (Cao et al., 2020; Hochane et al., 2019; Kim et al., 2019; Lindström et al., 2018a, 2018b, 2018c, 2018d; Menon et al., 2018; Tran et al., 2019), identifying stage-specific regional markers (Figures S1A), that can be applied to the characterization of organoid development. Immuno-analysis (Figures 1C–F, and S1B–S1D) and transcriptional profiling using quantitative PCR (qPCR; Figure S1E) highlighted robust upregulation of nephrogenic signatures (including WT1, PAX2, MAFB, HNF4A, GATA3, SLC3A1, and SLC12A1) over the differentiation time course.

On dd13, we documented the presence of CDH1+ epithelial structures with WT1-high and WT1-low domains, reminiscent of polarization in the early renal vesicle in vivo (Figure 1C; S1A). This time point also marked the emergence of cells with signatures of proximal nephron segment precursors (MAFB+ or WT1+), medial precursors (MAFB-/SOX9-/JAG1+, or HNF4A+), and distal nephron precursors (SOX9+ or GATA3+) (Figures 1C–E; S1A–B). At dd14 and dd16, the segmentation of developing nephron-like structures was more apparent with serially ordered proximal, medial and distal precursor domains (Figures 1C–E; S1A). In dd25 organoids, we also detected the presence of WT1+ HNF4A+, HNF4A+POU3F3+ and GATA3+POU3F3+ nephron subdomains (Figure S1D). Though these segments have not been fate mapped in the mammalian kidney, HNF4A+ cells are predicted to give rise to proximal tubule segments, consistent with genetic studies in the mouse (Marable et al., 2018), and GATA3+ cells to distal nephron structures (Lindström et al., 2021). Expression of genes associated with segmental functions were also detected at dd25: MAFB+ podocyte-like cell clusters were adjacent to a CUBN+ proximal tubule-like segment, which connected with an SLC12A1+ ascending LoH-like segment (Figure 1F; S1C).

A more extensive immunofluorescence analysis comparing dd13, 14, 22, and 28 highlighted epithelial structures with the concerted emergence of distinct regional domains of the transcriptional regulators MAFB and SOX9 in PAX2+ nephron-like tubules progressing to PODXL+ podocytes connected to a LTL+ proximal tubule region, followed by an SLC12A1+ distal segment and terminal MECOM+ domain (Figures S1B and S1C). In summary, although the kidney organoids collectively lacked the stereotypical morphogenesis reported for mouse and human nephrogenesis (Lindström et al., 2021), polarization and patterning generated segmented early nephron-like structures with reasonable uniformity in the kidney organoids. Further, consistent with all published studies to date, we only observed weak expression of the distal convoluted tubule gene SLC12A3 (Figure S1E): SLC12A3 could not be detected by immunofluorescence. Further optimization is needed to improve distal patterning outcomes.

To assess the uniformity of development in individual organoids, quantitative PCR was performed on 30 individual organoids at dd16 and dd28, sampling three independent batches of differentiation (10 individual organoids/batch). Each organoid showed a strong transcriptional signature for nephron segmentation, including MAFB, HNF4A, GATA3, ANXA1, SLC3A1 and SLC12A1 (Figures 1G, 1H and S1F). Whereas MAFB levels where relatively stable to the last stage of culture, there is a trend in the downregulation of the expression of early nephron TF genes, HNF4A and GATA3, though mature nephron signature genes ANXA1, SLC3A1, and SLC12A1 were up-regulated in dd28 organoids (Figure S1F).

To demonstrate reproducibility of the cell culture model, organoids were generated from the KOLF2.1J human induced pluripotent stem cell (iPSCs) line (Skarnes et al., 2021) adjusting the CHIR99021 level to 7 μM during the initial 4 days of differentiation (detailed in Methods). Transcriptional profiling using qPCR highlighted the formation of nephron progenitor cells at dd8 (SIX1, SIX2, PAX2, and WT1), early patterned nephron structures at dd16 (PAX2, WT1, GATA3 and MAFB) and expression of mature regional cell markers at dd28 (MAFB, SLC3A1, SLC12A1 and SLC12A3) (Figures S1H and S1I). Immunofluorescent analyses validated the formation of early nephron-like segments at dd16 and more mature nephron fates at dd28 (Figure S1G)

Developmental trajectories of nephron-like cell types

To further examine the emergence of nephron-like cells in these organoids, we applied single-cell RNA-sequencing (scRNA-seq) technology to capture single-cell transcriptomic profiles at dd8, dd10, dd14, dd16 and dd28 (Figures 2A, S2A–D). After filtering low quality cells and clustering in Seurat version 3, cluster identities were assigned through the expression of well-established cell markers (Figures S2A and S2B, Supplemental Table 1). Nephrogenic signatures were identified in clusters 2, 4, 5, 6, 7, 11, 12, 19, 22. In addition, and in agreement with single cell analysis from several groups of organoids founded on aggregation of much larger numbers of NPCs (Combes et al., 2019; Kumar et al., 2019; Tran et al., 2019; Wu et al., 2018), we observed a variety of non-nephron cell types including: interstitial cells expressing PDGFRB (clusters 0 and 18), SOX17+ CDH5+ endothelial cells (cluster 24), NEUROD1 and NEUROG1 expressing neuron-like cells (clusters 8, 14 and 16), SOX10+ neural crest-like cells (cluster 1, 10 and 15), TNNI1+ ACTC1+ muscle-like cells (cluster 23), as well as developing neuron-like cells (clusters 13, 20 and 21 as identified by GO terms analysis) (Figures S2A and S2B, Supplemental Table 1).

Figure 2: Single-cell Transcriptomic Profiling of Kidney Organoids (see also Figure S2 and Supplemental Table 1).

(A) Schematic diagram describing scRNA-seq time points.

(B) UMAP reduction of nephrogenic cell clusters in kidney organoids colored by clusters.

(C) Dotplot of marker genes used for identification of nephrogenic cell clusters in organoids.

(D) UMAP reduction of nephrogenic cells in kidney organoids colored by their origins.

(E) Bar graph of cell count presenting contribution of different original identities to various clusters of the nephrogenic cells in kidney organoids.

(F) UMAP reduction of integrated dd28 nephrogenic cells from two independent batches of kidney organoid differentiation colored by clusters, and hierarchical clustering of cell identities.

(G) Dotplot of gene markers used for identification of integrated dd28 nephrogenic cells from two independent batches of kidney organoid differentiation.

(H) UMAP reduction of integrated dd28 nephrogenic cells from two independent batches of differentiation colored by their origins.

(I) Bar graph of cell count presenting contribution of different original identities to various clusters of the integrated dd28 nephrogenic cells from two independent batches of differentiation.

To interrogate the developmental trajectory of the nephrogenic lineage in our culture model, we subset cells annotated to nephron-related cell types, and reclustered these cells to refine the analysis and to remove proliferating cells and additional non-nephrogenic cell types. The final nephrogenic lineage extraction contained 8,654 cells from 4 timepoints: day 8, 14, 16 and 28 (Figures 2B–E). Re-clustering identified cells with strong transcriptional characteristics of induced nephron progenitors, early and late podocytes, medial and distal precursors, and proximal tubule cells (Figures 2B and 2C). When examining the origins contributing to the various nephrogenic identities, we noticed induced nephron progenitor-like clusters (clusters 4 and 7) were predominantly represented in the day 8 samples as expected, while day 10, 14 and 16 cells highlighted nephron segment precursors. Late podocyte-like cells and proximal tubule-like cells were detected in day 28 organoids and medial/distal precursor cells were also observed in organoids of this late timepoint (Figures 2D and 2E).

Consistent with a nephron segmental patterning, organoid nephrogenic lineage cells expressed transcription factor genes shown to be essential for mouse nephron segmentation and also known to be present in early nephron structures in the human kidney (Lindström et al., 2018c, 2018d) : EYA1, MEOX1 and SIX1 (Xu et al., 1999, 2003), SIX2 (Kobayashi et al., 2008; Self et al., 2006), SALL1 (Basta et al., 2014), HNF4A (Marable et al., 2018), WT1 (Berry et al., 2015; Guo et al., 2004; Hartwig et al., 2010; Kann et al., 2015), PAX2 and PAX8 (Grote et al., 2006; Naiman et al., 2017), MAFB (Moriguchi et al., 2006; Sadl et al., 2002), FOXC1 (Kume et al, 2000; Motojima et al., 2017), LHX1 (Cirio et al., 2011; Kobayashi et al., 2005), LEF1, which is an effector of the canonical Wnt pathway (Park et al., 2012), HNF1B (Heliot et al, 2013; Naylor et al., 2013), POU3F3 (Nakai et al., 2003), and SOX9 (important for ureteric branching and present in the distal nephron though its role in nephron patterning has not been explored) (Reginensi et al., 2011). Thus, it is likely core nephrogenic programs were activated in the in vitro system and underlie the specification and patterning of nephron-like cell types (Figure S2E).

To assess the reproducibility of organoid differentiation outcomes in our model system we used scRNA-seq to compare two independent batches at dd28: 14,193 cells from batch 1 and 7,388 cells from batch 2 (Figures S2F–I). Nephrogenic (clusters 0, 1, 3, 8, 9, 14, 16, 19, 25, and 27), interstitial (clusters 2 and 20) and non-kidney lineages (clusters 4, 5, 6, 7, 10, 12, 13, 15, 17, 18, 21, 22, 23, and 26) showed broad representation in both batches (Figures S2F–I) though batch 1 dominated the composition of clusters 17 & 18 (neural crest-like cells), 24 (muscle-like cells), 25 (podocytes) and 26 (developing neuron-like cells), suggesting some quantitative variability from batch-to-batch. Cells of the nephrogenic lineage (7055 cells) were re-clustered and clusters identified based on marker gene expression (Figures 2F–I). Both batches showed overlapping segment-specific representation in podocytes (clusters 0, 1, 4, 6, 7, 9 and 13), early podocyte or parietal epithelium (cluster 14), early parietal epithelium or proximal tubule (cluster 2), parietal epithelium (cluster 11), proximal tubule (cluster 8), and medial/distal precursors (clusters 3, 5, 10, 12 and 15), though variability was observed in the proportional contribution of each batch to individual clusters (Figures 2F–I).

To compare in vitro derived cells to cells of the human kidney, we subset 4,405 nephrogenic cells from week 17 human fetal kidney datasets (Tran et al., 2019) and merged the in vivo with the in vitro derived nephrogenic cells (Figure 3A). Cluster identification validated the presence of nephrogenic cell types (Figures 3A and 3B, Supplemental Table 1). Nephron progenitor, nephron segment precursor, late podocyte and proximal tubule identities were composed of both human fetal kidney and cells of in vitro origin, highlighting similarities between the in vivo and in vitro derived cells. Genes exhibiting correlated expression with MAFB, HNF4A, SLC12A1 and GATA3 in the organoids emphasized expected transcriptional signatures of podocyte (i.e. PODXL, PTPRO, NPHS2, TCF21, CLIC5, etc.), proximal tubule (i.e. CUBN, LRP2, HNF4G, SLC3A1, SLC34A1, etc.), loop of Henle (i.e. TFCP2L1, IRX1, DEFB1, ERBB4, MAL, etc.), and distal precursor (i.e. CALB1, MECOM, MAL, ALDH1A1, EMX2, etc.) cell identities (Supplemental Table 1).

Figure 3: Single-cell Transcriptomics-Driven Comparison of In Vitro Organoid and Fetal Nephrogenic Cells (see also Figure S3 and Supplemental Table 1).

(A) UMAP reduction of integrated nephrogenic cells from kidney organoids and human fetal kidney colored by clusters, and hierarchical clustering of cell identities.

(B) Dotplot of marker genes used for cell cluster identification of integrated nephrogenic cells from kidney organoids and the human fetal kidney.

(C) UMAP reduction of integrated nephrogenic cells from kidney organoids and human fetal kidney colored by their origins.

(D) Bar graph of cell count presenting contribution of different original identities to various clusters of integrated nephrogenic cells from kidney organoids and human fetal kidney.

(E) UMAP reduction of integrated nephrogenic cells from kidney organoids (of various research groups) and human fetal kidney colored by clusters, and hierarchical clustering of cell identities.

(F) Dotplot of marker genes used for cell cluster identification of integrated nephrogenic cells from kidney organoids of various research groups and the human fetal kidney.

(G) UMAP reduction of integrated nephrogenic cells from kidney organoids of various research groups and human fetal kidney colored by their origins.

(H) Bar graph of cell count presenting contribution of different original identities to various clusters of integrated nephrogenic cells from kidney organoids generated by various research groups and human fetal kidney.

Importantly, differences are observed in the transcriptional profiles of in vitro and in vivo derived cells (Figures 3A–D). Cluster 7 (in vitro NPC) and clusters 0 and 15 (in vitro podocytes) were predominantly composed of organoid cells. Furthermore, while core transcription factors driving nephrogenesis marked the segmentation of in vitro nephron-like structures (Figure S2E), incomplete maturation was documented in an unbiased comparison with the human fetal kidney (Supplemental Table 1, and Figures S3A–C). For example, in vitro derived podocytes exhibited lower expression of late-activated genes encoding structural proteins or extra cellular matrix components as previously reported in conventional kidney organoids (e.g., COL4A4, TNNT2, DCN, and CLXC12) (Tran et al., 2019; Yoshimura et al., 2019). Generally, solute carriers and transporters mediating key kidney functions were expressed at significantly lower levels in organoid-derived cell types (e.g., SLC9A3, SLC44A3, SLC5A2 and SLC6A19) (Figures S3A–C, and Supplemental Table 1).

To compare the nephrogenic cells in the organoids in our model with those generated by other groups, we performed an extensive comparative scRNA-seq analysis. We first extracted the nephrogenic lineage single-cell transcriptomes from the published datasets of the human fetal kidney (Tran et al., 2019) and kidney organoids (Combes et al., 2019; Czerniecki et al., 2018; Kumar et al., 2019; Subramanian et al., 2019; Tran et al., 2019; Wu et al., 2018), and then constructed a merged dataset consisted of in vivo nephrogenic cells and in vitro derived cells (from our group and others) using the integration function within Seurat (Figures 3E–F). Clusters showed signatures of NPCs, early developing nephron, early and late podocyte, parietal epithelium, proximal tubule, and medial/distal precursors with contributions from fetal kidney and organoid cells across the different groups (Figures 3G and 3H).

These data support the conclusion of similar differentiation outcomes amongst PSC-derived organoids profiled by several different groups, including the organoid data presented here, and clear homology between these in vitro generated cell types and in vivo counterparts in the developing fetal kidney. However, the merged analysis also highlighted in vitro and in vivo specific signatures. Cluster 13 was almost entirely composed of NPCs from the week 17 human kidney consistent with the absence of NPCs in late-stage kidney organoids (Figure 3G, H). Clusters 6 and 17 (early developing nephron) and 10 and 18 (podocyte) were non-overlapping with in vivo cells indicating an in vitro difference with in vivo cellular programs (Figure 3G, H).

We also compared cellular diversity of the interstitial cell component. Nephron and interstitial progenitors likely share a common origin (Mugford et al., 2008) and may play a role in organizing nephron structures (Das et al., 2013; England et al., 2020). Interstitial cells extracted from the week 17 human fetal kidney data (expressing MEIS1, PDGFRA and/or PDGFRB) (Tran et al., 2019) generated 14 clusters (Figure S3D. Together with MEIS1, PDGFRA and PDGFRB, differential expression of FAM162B, LUM, KISS1, REN, TYMS, MEF2C, MYH1, TOP2A, COL8A1, FOXD1, CADM1, C7, POSTN, TAGLN, SHISA3, DKK1, USP53, GATA3, CCL2, IRF1, TMX4, and GRID2 highlighted the cellular diversity and replicative state of interstitial cells in the developing human kidney (Figure S3E). Merging these data with the organoid scRNA-seq-derived interstitial subset (MEIS1, PDGFRA and/or PDGFRB) highlighted the absence in vitro of specific interstitial subtypes: COL8A1+/FOXD1+ (cluster 10), MEF2C+/MYH11+ (cluster 12), and TMX4+/GRID2+ (cluster 6) and a reduced representation of CCL2+/IRF1+ (cluster 0), USP53/GATA3+ (cluster 7), and TYMS+/MYH11+ (cluster 13) cell types (Figures S3F–I) indicating a less complex interstitial cell diversity in the organoid model.

Vascularization of kidney organoids

Published organoid models comprising many nephron-like structures have been shown to recruit vasculature and undergo extended development on grafting to the adult mouse kidney (Tran et al., 2019, van den Berg et al., 2020). To examine the ability of our kidney organoids with just 1–2 nephron-like structures to recruit the vasculature, we transplanted organoids derived from MAFB-P2A-eGFP hESCs under the kidney capsule of NOD/SCID mice (Figure 4A). Twenty-days post-transplant, MAFB+ in vitro derived podocytes, positive for a human nuclear antigen (HuNu+) were surrounded by HuNu- VEGFR2+ mouse endothelial cells (Figure 4A). Vascularized podocytes were PODXL+ and ANXA1+, indicative of mature podocyte-like signatures (Figure S4A) (Tran et al., 2019). Proximal tubule-like structures (HNF4A⁺ CUBN⁺ LRP2⁺ ACE2⁺) were identified in transplanted organoids together with an SLC12A1⁺ distal tubule-like segment (Figures S4B–D). Interestingly, we observed rare SLC12A3+ organoid derived cells suggesting additional maturation of distal convoluted tubule cells not observed in vitro (Figure S4E). Thus, the kidney organoids demonstrated vascular recruitment, segmented nephron signatures and the additional maturation of nephron cell types.

Figure 4: Vascularization of Human Kidney Organoids (see also Figure S4).

(A) Left: Schematic diagram describing the transplantation procedure to vascularize MAFB-P2A-eGFP organoids. Right: Immunofluorescent analyses of cryosectioned vascularized MAFB-P2A-eGFP organoids. Scale bars indicate 50 μm.

(B) Upper Panel: Schematic diagram describing the transplantation procedure to vascularize MAFB-P2A-eGFP organoids and injection of Dextran (3 and 70 kDa). Lower Panel: Immunofluorescent analyses of cryosectioned vascularized MAFB-P2A-eGFP organoids collected 10 minutes post dye injection. Scale bars indicate 50 μm.

To evaluate the functional properties of vascularized in vitro derived nephrons, we performed an in vivo dextran uptake assay injecting texas red conjugated (3 kDalton) and tetramethylrhodamine conjugated (70 kDalton) compounds into the retro-orbital venous sinus of the NOD/SCID mice two weeks after implanting organoids under the kidney capsule (Figure 4B). Kidneys were collected for analysis 10 minutes after the injection. As expected, mice injected with PBS showed no fluorescent signature in the kidney (Figure S4F). In the host mouse kidneys, 70 kDalton dextran was retained at the apical surface of proximal tubules, and in glomerular and peritubular capillaries (Figure 4B; Figure S4F). In contrast, the smaller 3 kDalton dextran showed only a weak signal in the glomerular and peritubular capillaries but a strong signal at the apical surface of proximal tubules consistent with rapid filtering of the smaller dextran molecule (Figure 4B; Figure S4F). Whereas the higher molecular weight compound could not be detected within implanted organoids, the 3 kDa dextran was evident as puncta predominantly associated with the apical region of LRP2+ proximal tubule cells (Figure 4B; Figure S4F, G) consistent with limited vascular permeability and filtering into organoid-derived nephrons.

Modeling polycystic kidney diseases using kidney organoids

To explore the utility of our scalable organoid platform to model disease and screen for phenotypic modifiers, we generated mutations in the two key genes associated with autosomal polycystic kidney disease (ADPKD): PKD1 and PKD2 (Bergmann et al., 2018; Harris and Torres, 2018). Both mutations were generated through CRISPR-Cas9 directed double strand-DNA cleavage and non-homologous end joining repair in the H9 hESC line (Figures 5A, S5A and S5B). Acknowledging the genetic complexity of the PKD1 gene, which has six known highly homologous pseudogenes (Bogdanova et al., 2001), we designed a guide RNA specific to the bone-fide PKD1 gene sequence (Figure S5A). Sanger sequencing identified clones carrying deletions predicted to generate a loss of function for both alleles of PKD1 and PKD2 in distinct targeting events (Figures S5A and S5B). To demonstrate the expected loss of function for each gene, we used validated PKD1 and PKD2 antibodies (Yu et al., 2007; MacKay et al., 2020). Wild-type PKD2 has a predicted molecular weight of 110 kDalton but frequently forms higher molecular aggregates in the process of Western blot analysis (Yu et al., 2009). PKD2 protein was detected in wild-type and PKD1 mutant organoids, but not in PKD2−/− mutant organoids with biallelic deletions (Figure S5C). PKD1 encodes a 462 kDalton-primary polypeptide (PC1) which can be observed on induction of PKD1 expression in HEK cells (arrowhead in Figure S5D). However, analysis of fetal kidney samples only detects a smaller product of 81 kDalton, consistent with published data (Lea et al., 2020), suggesting in vivo processing. In control samples and wild-type organoids, a higher molecular weight (140 kDalton PKD1 cleavage product was observed which was lost on biallelic mutation of the PKD2 gene (arrow in Figure S5E). Thus, both PKD mutant clones lost detectable levels of PKD protein encoded by the mutated allele.

Figure 5: Cyst Formation in PKD1−/− and PKD2−/− Human Kidney Organoids (see also Figure S5, Supplemental Table 2, and Supplemental Movie 1).

(A) Schematic diagram describing the CRISPR-Cas9-induced mutations on PKD1 or PKD2 alleles and resulting deletions and frame-shifts leading to premature termination of translation.

(B) Schematic diagram describing the timeline of the live imaging assay to track cyst formation.

(C and D) Brightfield images showing progressive cyst formation from differentiation day 14 to 20 in PKD1−/− or PKD2−/− mutant human kidney organoids alongside their wildtype controls. Red arrows: epithelial outpocketing observed at low frequency in wildtype organoids. Blue arrows: forming cysts in PKD mutant organoids.

(E) Boxplots comparing area increases of PKD1−/− or PKD2−/− mutants with their isogenic wildtype controls from differentiation day 14 to 20 (Wilcoxon Rank Sum test).

(F) Immunofluorescent analyses showing contribution of different nephron segment-like cells to cystic epithelial cells. Scale bars indicate 50 μm.

To determine whether PKD mutant hESC lines can replicate cystic outgrowth, PKD1−/− and PKD2−/− hESCs were differentiated alongside wild-type hESC lines. Seeding ~1,500 cells/micro-well, we examined cyst formation by high-resolution automated imaging of wild type, PKD1 and PKD2 mutant organoids cultured in methylcellulose supplemented medium, to prevent movement of the organoids, visualizing cyst growth over multiple days of culture (dd14 to dd20) (Figure 5B; Supplemental Movie 1). Quantitative PCR confirmed PKD mutant organoids underwent a similar differentiation trajectory to wild-type hESCs (Figures S5F–H). However, in contrast to wild-type organoids, cysts emerged (blue arrows in Figure 5C–D) and expanded from PKD1 and PKD2 mutant organoids (Figure 5E–F).

Cysts were recognizable as early as dd15 and morphologically distinct from low frequency epithelial protrusions observed in wild-type organoids that grow but do not balloon into cystic structures (red arrow in Figure 5C). Interestingly, cyst formation was dependent on the seeding density; cyst formation was delayed, and cyst frequency decreased, as the number of seeding cells was increased from 1,500 cells to 5,000 or 7,000 cells in the PKD2 mutant model (Figures S5I and S5J). Cysts expanded continuously on extended culture eventually detaching from the parent organoid generating free-growing cystic structures that reached more than 1 cm in diameter after 3 months of culture (Figure S5K). Antibody profiling showed cystic epithelia displayed a complex profile with contributions from proximal (JAG1+, HNF4A+ or SLC3A1+) and distal nephron segments (SOX9+ or POU3F3+) (Figures 5F and Figure S5L). Examination of tight junctions (TJP1/ZO-1), an apical polarity marker (PRKCI/aPKC) and the apically localized primary cilium (ARL13B) showed an expected luminal facing apical polarity of nephric epithelium within the organoid that was reversed in cystic outgrowths with the primary cilium projecting into the culture medium (Figure S5M). Polarity reversal has been reported in another in vitro model (Cruz et al., 2017), though not in cysts within the kidney of mouse Pkd mutants (Ma et al., 2013; Shao et al., 2020; Yu et al., 2008).

Phenotypic screening to identify protein kinase inhibitors inhibiting cyst initiation

With the advantage of mass production of organoids, we screened an annotated small molecule protein kinase inhibitor (PKI) library (see Methods) to identify potential inhibitors of cyst initiation (Figure 6A). PKIs can provide broad insight into signaling pathway activities which may uncover novel mechanisms or avenues for therapeutic approaches beyond the primary PKI hit. Cyst formation was initially calibrated using vehicle (DMSO-) treated PKD1−/− (n=933) and PKD2−/− (n=1241) mutant organoids. We performed automated bright-field imaging of each organoid daily from dd14–20. At dd20, 52.3% of PKD1−/− and 25.3% of PKD2−/− organoids formed cysts (Figure 6B). Positive Z-scores at dd20 for both the PKD1−/− and PKD2−/− lines (0.40 and 0.37 respectively) demonstrated that positive and negative outcomes could be determined with confidence (Zhang et al., 1999) (Figure S6A).

Figure 6: High-throughput Screening to Identify Compounds Inhibiting Cyst Initiation (see also Figure S6, Supplemental Table 2, and Supplemental Movies 2, 3, and 4).

(A) Schematic diagram describing the timeline of the phenotypic assay to identify compounds inhibiting cyst formation.

(B) Quantification of cyst formation rate of DMSO-treated PKD1−/− or PKD2−/− organoids from 96-wells of methylcellulose-embedded organoids for each line.

(C) Schematic diagram describing the screening process to identify protein kinase inhibitors impeding cyst formation.

(D) Compounds validated for cyst inhibition in both PKD1−/− and PKD2−/− kidney organoids. Celastrol, Carfilzomib and Rapamycin were included as literature-based positive controls, and Tolvaptan as a negative control based on the absence AVPR2 expression.

We therefore launched a series of screens to identify PKIs impeding cyst initiation (summarized in Figure 6C). The primary screen was performed with a library of 247 PKIs which was assembled by combining the EMD Protein Kinase Inhibitor 2, 3 and 4 collections (see Materials and Methods) using a single dose of 1μM in a primary screen of PKD2−/− organoids (Supplemental Table 2). Approximately 12–15 organoids were screened per well scoring cyst formation within 9 wells containing each compound, distributed across three separate plates. Cultures were tracked over a 6-day period of culture (dd14–20). Importantly, the compound set-up on the plates was blinded from the operator scoring outcomes to reduce operator bias. Screen outcomes were categorized into three groups: 1) “negative hits” defined as wells with continued cyst formation, 2) “positive hits” with cyst suppression but continued growth and development of the organoid 3) “non-specific hits” (NS hits) with no visible cyst formation but evidence of general growth retardation or cell death (Figure S6B and Supplemental Movies 2, 3 and 4). To increase the stringency of the screen, only compounds identified as “positive hits” in all 9 wells for each compound were considered true “hits” and selected for a secondary screen. Among the 247 initial screen compounds, 11 compounds were identified as NS hits and 9 as positive hits (Supplemental Table 2).

Hits and various controls (Supplemental Table 2) underwent a similarly structured secondary screen on both PKD1−/− and PKD2−/− organoids examining three different concentrations of each compound: 0.1, 1.0 and 10.0 μM. For controls, we included Tolvaptan, an inhibitor of AVPR2 and the first FDA approved pharmacological treatment for ADPKD. Importantly, as AVPR2 is activated in differentiated cells of the connecting segment and collecting duct (Beaudoin et al., 2019; Higashihara et al., 2011; Hopp et al., 2015; Reif et al., 2011; Torres et al., 2016, 2017; Ransick et al., 2019) that are generally absent from PSC-derived kidney organoids, Tolvaptan was not expected to inhibit cyst growth in our assay. The mammalian target of rapamycin (mTOR) pathway is activated in cystic epithelia in patients with ADPKD and rodent models of polycystic kidney disease. Rapamycin inhibits mTOR in rodent ADPKD models (Holditch et al., 2019; Shillingford et al., 2010; Tao et al., 2005). Furthermore, rapamycin treatment significantly improved renal function in vivo in animal models (Shillingford et al., 2010; Tao et al., 2005). Therefore, rapamycin was added as a potential inhibitor of cyst formation. Two compounds shown to inhibit cyst formation in ADPKD mouse models: Carfilzomib, a proteasome inhibitor and Celastrol, a pentacyclic triterpene (Booij et al., 2020; Chang et al., 2018; Fedeles et al., 2011) were added as further potential inhibitory controls. Additionally, we also examined the effects of forskolin and blebbistatin, compounds previously identified to enhance cystogenesis (Czerniecki et al., 2018; Low et al., 2019).

Celastrol, carfilzomib, and rapamycin, but not tolvaptan, scored as positive hits in the secondary screening assay inhibiting cyst formation in PKD1−/− and PKD2−/− organoids (Figure 6D). Carlfilzomib inhibited cyst formation at all concentrations evaluated in PKD2−/− organoids, but only at the highest concentration in PKD1−/− organoids. Neither forskolin or blebbistatin showed any enhancement of cyst outgrowth (Figures S6C–H) though surprisingly we observed a significant decrease in growth at the highest concentration, but only in the PKD2 mutant organoid cultures (10μM; Figure S6F). Of the compounds identified in the primary PKD2−/− screen, the protein kinase C pathway inhibitors UCN-01 and UCN-02 showed variable results in PKD1/- and PKD2−/− cyst inhibition, whereas QNZ and IKK inhibitor VII, potential NF-ĸB pathway modulators (Tobe et al., 2003; Waelchli et al., 2006), inhibited both PKD1−/− and PKD2−/− cyst formation (Figure 6D). Of note, QNZ prevented cyst formation in both PKD1 and PKD2 mutants at all doses.

Given the efficacy of QNZ, we examined the effective levels of QNZ for inhibiting cyst formation. Our data showed the concentration dependent inhibition of cyst formation after QNZ treatment with an IC₅₀ of 6.6 nM for PKD1 mutant and 4.5 nM for PKD2 mutant organoids and a complete inhibition of cyst formation at 20 nM (Figure 7A and 7B). Next, we examined if QNZ modified cyst growth after cyst formation had initiated, at early (a few days after cyst initiation) and late (free-growing cysts detached from the organoid) stages. QNZ was added to PKD1−/− and PKD2−/− organoids at concentrations ranging from 1 to 25 nM when early cysts were visible (dd16) and cyst outgrowth compared over 2 days of culture together with parallel assays of celastrol, tolvaptan, rapamycin and a DMSO non-compound control. As expected, DMSO and tolvaptan-treated organoids underwent cyst expansion, while 10 μM Celastrol reduced cyst areas (Figure 7C). Rapamycin treatment did not significantly reduce cyst growth in this short-term assay (Figure 7C). Consistent with the cyst initiation assay, QNZ treatment (5–25nM) significantly reduced cyst growth in both PKD1−/− and PKD2−/− organoids (Figure 7C). Additionally, QNZ administered to late cysts, (see Methods for information on late cyst assay; cysts diameter >= 1mm) led to a dramatic reduction in the size of PKD1−/− (0.1 and 1 μM QNZ) and PKD2−/− (0.01, 0.1 and 1 μM QNZ) cysts (Figures 7D and S7A).

Figure 7: Effects of QNZ on cyst initiation and progressive cyst growth (see also Figure S7 and Supplemental Table 1).

(A) Summary of effects of low QNZ doses on inhibiting the initiation of cyst outgrowth in PKD1−/− and PKD2−/− mutant kidney organoids compared to Celastrol, Tolvaptan and Rapamycin.

(B) Determination of IC₅₀ of QNZ in PKD1−/− or PKD2−/− organoids based on cyst formation rates in organoids (Data are represented as mean ± SEM).

(C) Effects of low QNZ doses on inhibition of growth in newly emerged cysts (monitored from dd16 – dd19) arising from PKD1−/− and PKD2−/− mutant kidney organoids: early cyst inhibition assay. Celastrol, Tolvaptan and Rapamycin serve as controls (Data are represented as mean ± SEM).

(D) Effects of low QNZ doses on inhibition of growth of cysts arsing from free-growing cysts detached from the parent organoid following extended culture: late cyst inhibition assay. QNZ-treated PKD2−/− late cysts were compared to DMSO controls from day 0 to day 9 of culture; QNZ administered on day 3 (Data are represented as mean ± SEM). Scale bars indicate 500 μm.

Working within the concentration range above, we examined the effects of QNZ on wildtype organoid development. MAFB-eGFP wildtype organoids were treated continuously with QNZ (10 nM to 1 μM) over the post aggregation period (dd8 to dd28). Though growth was reduced at the higher concentrations, 10 nM QNZ-treated organoids were comparable in size range to the DMSO control group (Figures S7B and S7C). eGFP+ podocytes first emerged as expected at dd14. Further, qPCR analysis of differentiation markers at the end of the culture period indicated DMSO and QNZ-treated MAFB-eGFP underwent comparable differentiation (Figure S7D). Transcriptional profiling of organoids at dd28 was consistent with qPCR assays but suggested QNZ treatment initiated metabolic changes and activated stress responses (Figure S7E and Supplemental Table 1). Transcriptional analysis of PKD2−/− organoids (Figure S7F and Supplemental Table 1) and late cysts (Figure S7G and Supplemental Table 1) indicated QNZ affected cell growth and proliferation in addition to inducing stress responses. No significant impact was observed on apoptosis measured by TUNEL when organoids were cultured with 20nM QNZ, a concentration sufficient to block all cyst formation (Figures S7H–J). Further, FACS analyses quantifying the percentage of live cells in PKD1 and PKD2 mutant organoids cultured in up to 1 μM QNZ showed no decrease in the live cell FACS index relative to DMSO controls, contrasting with a significant loss of cell viability and marked elevation in apoptosis in kidney cell types treated with 0.1μM of staurosporine as positive toxicity control (Figure S7K and S7L). Finally, no significant increase in apoptosis was observed in ZO1+ epithelial cells in late-cyst assays treated with 0.01 – 1 μM QNZ (Figure S7K and S7L). Collectively, these results highlight the potency and specificity of QNZ in inhibiting cyst formation and later growth of independent cystic structures, independent of a marked elevation in apoptosis.

DISCUSSION

Recognizing the need for a scalable system to generate kidney organoids for developmental studies and translational application, we developed a culture model that provides a good solution to generating tens of thousands of relatively homogenous human kidney organoids of a size and cellular complexity well suited for screening purposes. While a detailed developmental analysis of nephrogenesis in organoids in vitro highlights morphologically distinct structures to those mediating normal nephron formation in vivo, our analyses point to similarities in the patterning processes and cellular outcomes. As with all reports of kidney organoid systems, eliminating unwanted cell types and normalizing programs for developmentally and functionally interconnected cell types (vascular, interstitial and collecting system) will further improve kidney-specific modelling capability.

Utilizing the organoid model, we developed a robust platform to study cyst initiation and expansion in ADPKD. Organoids carrying homozygous loss of function in either PKD1 or PKD2 formed reproducible and robust cystic structures, with a more robust cyst forming capability in the PKD1 mutant organoids. The platform also provides a model for studying aspects of late-stage cystic expansion as cysts continue to grow to centimeter-sized structures, detached from the parental organoid body.

Previous studies have documented cyst formation promoted by small molecule simulators of cAMP production in alternative PKD1 and PKD2 PSC-generated organoid models (Czerniecki et al., 2018; Kuraoka, et al., 2020; Shimizu et al., 2020). In the absence of stimulation, cysts have been reported in the aforementioned organoid model, albeit at low frequency (~5%). Here, the requirement for manual extraction of cyst forming organoids from an adherent matrix encased culture further complicates high throughput screening (Czerniecki et al., 2018). Cyst formation is clearly assay dependent as others have not observed cyst formation in an alternative kidney organoid model (Kumar et al., 2019). In our organoid platform, both the timing and frequency of cyst formation depend on the initial seeding density of cells in the kidney organoid aggregate. The percentage of cyst-forming organoids is also markedly improved (25–50% to ~80%) by adding a preselection based on initial outgrowth at dd14. Coupling machine learning with an organoid-sized sorting system (Pulak, 2006) could generate a highly efficient, automated platform for future large-scale screening.

Previous kidney organoid screens identified compounds enhancing cystogenesis: forskolin and 8-Bromoadenosine 3’,5’-cyclic monophosphate (8-Br-cAMP) which elevate intracellular levels cAMP levels, and blebbistatin, which inhibits myosin ATPase activity (Czerniecki et al., 2018; Low et al., 2019). However, neither forskolin or blebbistatin enhanced cystogenesis over the time course investigated in our organoid model (Figures S6C–H). The model replicated cyst-inhibitory properties of rapamycin which has been widely reported to modify cystic growth in ADPKD although clinical trials of rapamycin in ADPKD patients showed minimal effects on the progression of kidney failure or improved renal function, and strong side-effects of the drug treatment (Serra et al., 2010; Walz et al., 2010). Celesterol and carlfilzomib inhibited cystogenesis in the human kidney organoid model supporting reports of cyst growth inhibition in transgenic mouse models (Booij et al., 2020; Chang et al., 2018; Fedeles et al., 2011; Holditch et al., 2019; Shillingford et al., 2010; Tao et al., 2005).

Screening protein kinase inhibitor libraries for compound suppressing cystogenesis in PKD1−/− and PKD2−/− kidney organoids shed new light on inhibitory compounds identifying UCN-01, UCN-02, IKK inhibitor VII and QNZ as novel inhibits of cyst formation in the organoid assay. UCN-01 and UCN-02 are derivatives of PKC inhibitor Staurosporine (Karaman et al., 2008; Rüegg and Gillian, 1989; Takahashi et al., 1989). Staurosporine appeared to generally inhibit cell growth even at lowest concentration evaluated (0.1 μM), agreeing with previous reports on its role in triggering apoptosis (Bertrand et al., 1994). In contrast, UCN-01 and UCN-02 dose response experiments point to a more marked effect on cyst growth than general organoid growth though a more rigorous measure of organoid growth would be required to draw this conclusion. UCN-01 and UCN-02 are stereoisomers, which may explain different dose-response on PKD1 and PKD2 mutant cysts. UCN-01, UCN-02 and staurosporine have promiscuous kinase inhibition (Tamaoki and Nakano, 1990), so non-PKC mechanisms of action cannot be ruled out.

IKK inhibitor VII is a selective competitive inhibitor of NF-κB signaling blocking activity of IκB kinases (IKKs) and transcriptional activation of the NF-κB pathway. IKK inhibitor VII targets both IKK-1 and IKK-2. However, annotated IKK-2-specific kinase inhibitors in the library did not prevent cyst formation in our primary screen suggesting a link to IKK-1 inhibition and potentially non-canonical NF-κB pathway activity in cystogenesis (Supplemental Table 2). A study of PKD in a rat model has highlighted the expression of noncanonical NF-κB pathway transcriptional components in cystic epithelial cells (Ta et al., 2016).

Interestingly, QNZ (quinazoline), the most effective inhibitory compound identified in our assay, is reported to inhibit NF-κB pathway activation (Scheurer et al., 2019; Tobe et al., 2003) though other studies suggest QNZ indirectly modulates NF-κB signaling through the control of Ca2+ entry into the cell (Choi et al., 2006). Further, phenotypic screens in Drosophila and induced pluripotent stem cell-derived Huntington’s disease model showed QNZ inhibition of neuronal store-operated Ca2+ entry pathway activity (Nekrasov et al., 2016; Wu et al., 2011). PKD2, a selective cation channel in complex with PKD1, has been linked to primary cilium transport of Ca2+, and potentially monovalent cations (DeCaen et al., 2013; Kleene and Kleene, 2017; Koulen et al., 2002; Liu et al., 2018; Su et al., 2018). Elevated Ca²⁺ levels have been reported in proximal tubule cell cultures from PKD1 mutant mice (Yanda et al., 2019). Our initial transcriptional profiling suggest QNZ induces stress responses in wild-type organoids, though this response does not inhibit differentiation of organoids at low cyst-inhibitory doses, and no elevation was observed in apoptosis or FACS assessed cell viability in either normal or PKD1 mutant organoids, or outgrowing cystic structures, at low QNZ doses that were completely refractory to cyst initiation and cyst growth. Potentially, enhanced stress in PKD deficient cysts may synergize with QNZ to inhibit cell division at low QNZ concentrations. The strong inhibitory activity of QNZ on cystogenesis at nanomolar concentrations in both PKD1−/− and PKD2−/− models supports further evaluation of the mechanistic action of this compound for potential therapeutic insight.

In conclusion, the reasonable uniformity in size, cell types and developmental progression in the kidney organoid platform we describe here are features well suited to high-throughput image-based screening. These organoids may also be better suited for renal implantation, and further evaluating functional integration will be an interesting avenue for future studies. In a proof-of-principle, we confirmed cyst inhibition in ADPKD organoid kidney models with small molecule inhibitors active in ADPKD animal models, and identified inhibitory compounds. The kidney organoid system we describe complements other approaches to dissect kidney development and study renal functions. Further, our platform adds to existing strategies to identify therapeutic leads for the treatment of ADPKD, and potentially other kidney diseases.

Limitations of the study

The in vitro ADPKD model investigates cystogenesis outside of normal kidney function which may modify cyst forming pathways. These findings were obtained using an in vitro model and need further studies to examine the effects of QNZ in vivo.

STAR METHODS

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Dr. Andrew P. McMahon (amcmahon@med.usc.edu).

Materials Availability

Cell lines generated in this study can be shared upon request following submission of a Material Transfer Agreement.

Data and Code Availability

Single-cell RNA-seq datasets collected in this study have been deposited at GEO, and accession number has been detailed in the Key Resources Table.

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
WT1	abcam	ab89901; RRID: AB_2043201
JAG1	R&D	AF599; RRID: AB_2128257
LAMB1	Santa Cruz	sc-33709; RRID: AB_627868
SOX9	abcam	ab185230; RRID: AB_2715497
HNF4A	R&D	MAB4605
CUBN	R&D	AF3700; RRID: AB_2086138
SLC12A1	Sigma	HPA018107; RRID: AB_1854504
LTL	Vector Laboratories	FL-1321; RRID: AB_2336559
SLC3A1	Sigma	HPA038360; RRID: AB_2675975
NPHS1	abcam	ab136927
POU3F3	ThermoFisher	PA5–64311; RRID: AB_2645790
MAFB	R&D	MAB3810; RRID: AB_2137675
PAX8	abcam	ab189249; RRID: AB_2801268
CDH1	Biosciences	610182; RRID: AB_397581
PAX2	R&D	AF3364; RRID: AB_10889828
GATA3	R&D	AF2605; RRID: AB_2108571
ACE2	R&D	AF933; RRID: AB_355722
PKD1	Kerafast (Yu et al., 2007)	EMD303
PKD2	Santa Cruz	sc-28331; RRID: AB_672377
DRAQ5	Novus Biologicals	NBP2-81125-50ul
DAPI	ThermoFisher	D1306
Chemicals, Peptides, and Recombinant Proteins
Bacillus licheniformis cold active protease	Creative Enzymes	NATE-0633
Collagenase type 2	Worthington	#LS00417
DNase I	Worthington	#LS002058
AutoMACS Running Buffer	Miltenyl Biotec	130-091-221
Methylcellulose powder	Sigma-Aldrich	M0512
CHIR99021	Sigma Aldrich	SML1046
ActivinA	R&D	338-AC-050
FGF9	R&D	279-F9
Click-iT Plus TUNEL Assay	Invitrogen	C10618
EMD-Protein Kinase Inhibitor 2	EMD Calbiochem®	539745
EMD-Protein Kinase Inhibitor 3	EMD Calbiochem®	539746
EMD-Protein Kinase Inhibitor 4	EMD Calbiochem®	539747
Critical Commercial Assays
10x Genomics Chromium Single Cell 3’ GEM, Library & Gel Bead Kit	10X Genomics	PN-1000075
HiSeq 3000/4000 SBS PE clustering kit	Illumina	PE-410–001
150 cycle flow cell	Illumina	FC-410–1002
Deposited Data
scRNA-seq of Miniature Kidney Organoids	This study	GSE164564
Bulk RNA-seq of PKD mutant kidney organoids and cysts	This study	GSE195717
scRNA-seq of Kidney Organoids	Czerniecki et al., 2018	GSE109718
scRNA-seq of Kidney Organoids	Subramanian et al., 2019	GSE136314
scRNA-seq of Kidney Organoids	Wu et al., 2018	GSE118184
scRNA-seq of Kidney Organoids	Combes et al., 2018	GSE114802
scRNA-seq of Kidney Organoids	Kumar et al., 2019	GSE117211
scRNA-seq of Kidney Organoids	Tran et al., 2019	GSE124472
scRNA-seq of Human Fetal Kidney (Week 17)	Tran et al., 2019	GSE124472
Experimental Models: Cell Lines
H9 hESC	WiCell	WA09
PKD2−/− H9 hESC	This paper
5T H9 hESC	This paper
5T PKD1−/− H9 hESC	This paper
MAFB-P2A-eGFP H9 hESC	Tran et. al., 2019
KOLF2.1J iPSC	Skarnes et al., 2021
Experimental Models: Organisms/Strains
Immunocompromised mice	The Jackson Laboratory	NOD.CB17-Prkdc<SCID>/J, RRID: MGI:5652139
Oligonucleotides
gRNA PKD1 (TGGCAACGGGCACTGCTACC)	This paper
gRNA PKD2 (CCCGGATGATGTCACAGCTCTTC)	This paper
Primer1_PKD1_Sanger_seq_genotype (TCCAGATGGGGCAGAGCCTG)	This paper
Primer2_PKD1_Sanger_seq_genotype (CCTCCTTCCTCCTGAGACTC)	This paper
Primer1_PKD2_Sanger_seq_genotype (CTGTGTTCCAGTGACCTACG)	This paper
Primer2_PKD2_Sanger_seq_genotype (AAGGCACAGGCAAAGTTCTCA)	This paper
Software and Algorithms
Seurat 3.0	Stuart et al., 2019	http://satijalab.org/seurat/
CellRanger 3.1	10x Genomics	https://support.10xgenomics.com

This paper uses referenced sources of code; no novel code was created for the analyzes in this study. Details of key parameters are provided in the Method Details section. If additional information is required for reanalysis of the data reported here, the lead contact will provide the information on request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Human Kidney Studies

Collection of de-identified human fetal tissue from elective terminations with informed consent was approved by the Institutional Review Boards of both Children’s Hospital of Los Angeles and the Keck School of Medicine of the University of Southern California. Guidelines from the American College of Obstetricians and Gynecologists were used to determine the gestational age using a combination of ultrasound and last menstrual measurements (O’Rahilly and Müller, 2010; O’Rahilly et al., 1987). Samples were transported on ice at 4°C in high glucose DMEM (Gibco, 11965–118) supplemented with 10% fetal bovine serum (Genesee Scientific, 25–550) and 25mM HEPES (Gibco, 15630080).

Human Embryonic Stem Cell Lines

H9 hESC line (female) was obtained from WiCell (WA09). CRISPR-Cas9 gene-editing (Jinek et al., 2012; Qi et al., 2013) was adopted to generate the PKD1−/− and PKD2−/− H9 hESC lines. For PKD2 mutation, a gRNA (5’-CCCGGATGATGTCACAGCTCTTC-3’) targeting coding sequences in exon 3 was delivered to H9 hESCs together with Cas9 protein (Thermofisher MPK5000S) by electroporation. Targeted hESCs were dissociated into single cells using Accutase (Gibco, A1110501) and seeded on a Geltrex-coated 96-well plate at 1 cell/well density, in mTeSR (StemCell Technologies, 85850) supplemented with 10 μM Y27632 (Tocris, 1254). Single cell-derived colonies were expanded and targeting events analyzed by qPCR and validated by Sanger sequencing (see below).

The PKD1−/− H9 line was generated using CRISPR-Cas9 technology on the background of a fluorescent reporter line generated on the H9 line background to visualize different components of the nephron (5-T H9 ESCs, shortened here to 5-T). PKD1 was targeted with a gRNA (5’-TGGCAACGGGCACTGCTACC-3’) homologous to sequences in coding exon 6 of the PKD1 gene. Clonal targeted colonies were identified similar to the PKD2−/− hESC line.

Kolf2.1J iPSC line was a gift from Dr. William Skarnes (Skarnes et al., 2021).

METHOD DETAILS

1). Kidney organoid cultures

a). hPSC Maintenance

Geltrex-coated plate preparation

DMEM/F12 (Life Technologies, 11320–033) was aliquoted into a 50-ml conical vial and 120μl of Geltrex added to make a 1% Geltrex mix. After thorough mixing, 2ml of 1% Geltrex was pipetted into each well of a 6-well plate (or 1ml/well for a 12-well plate). The Geltrex plates were incubated at 37°C/5% CO₂ overnight before use.

hPSC expansion and maintenance

This study utilized six hPSC lines (listed in Key Resources Table). (Tran et al., 2019)hPSCs were thawed in StemFit media (Ajinomoto, ASB01-R) supplemented with 100ng/ml of FGF2 (R&D, 273-F9) and 10μM Y27632 (Tocris, 1254) on 1% Geltrex-coated plates (ThermoFisher, A1413302). On reaching 70–80% confluency (1–2 days), cells were passaged in StemFit media + 100ng/ml of FGF2 + 10μM Y27632 into a 12-well plate at a seeding density of 6,000 cells/well. The medium was changed 48 hours later to StemFit media + 100ng/ml of FGF2 for cell expansion and replenished every 2 days. For freezing, when wells reached 70–80% confluency each well of hPSCs was mixed with 1ml of 10% DMSO/90% fetal bovine serum (FBS) and stored in insulated styrofoam boxes at −80°C overnight before transferring to liquid nitrogen storage.

b). Directed Differentiation to Generate Kidney Organoids

The differentiation protocol was developed based on published protocols (Morizane et al., 2015; Morizane and Bonventre, 2017) adapted in our laboratory. Each biological replicate was generated from a distinct frozen vial of hPSCs. After thawing and growth to 70% confluency, hPSCs were dissociated using Accutase (Gibco, A1110501) and seeded on 12-well plates and cultured as above. The differentiation procedure was initiated at 60% confluency. Briefly, culture medium was supplement for 4 days with 8μM CHIR99021 (Sigma Aldrich, SML1046) for differentiation of all hESC lines (or 7μM CHIR99021 for the kolf2.1 iPSC line), followed by 3 days with 10ng/ml ActivinA (R&D, 338-AC-050), and 1 day of 10ng/ml FGF9 incubation (R&D, 273-F9). At day 8, the cells were dissociated using TrypLE dissociation enzyme (Gibco, 12563011), and 600,000 cells seeded into each well of a 12-well EZSPHERE plate (Nacalai USA, TCI-4815–903SP-50P) in 3μM CHIR and 10ng/ml FGF9. Each well generated ~400 organoids and each organoid comprised ~1,500 cells. On differentiation day (dd) 10, the medium was switched to Advanced RPMI 1640 (Gibco, 12633020) + 1X Glutamax (Gibco, 35050079) + 1% Penicillin-Streptomycin (Invitrogen, 15070063), denoted basal differentiation medium, with 10ng/ml FGF9. From dd13 to dd28, cultures were maintained in basal differentiation medium. For the kolf2.1 iPSC differentiation, Essential 6 (Gibco, A1516401) was used as the basal differentiation medium.

c). Validation of PKD1 and PKD2 mutant alleles

Single cell-derived clones of the PKD1 and PKD2 targeted hESC lines were expanded and genomic DNA was extracted using the DNeasy Blood & Tissue Kit (Qiagen, 69504). CRISPR-Cas9 targeted regions were amplified using the Q5 High-Fidelity 2X Master Mix (PKD1: primers: 5’-TCCAGATGGGGCAGAGCCTG-3’ and 5’-CCTCCTTCCTCCTGAGACTC-3’ and PKD2: 5’-CTGTGTTCCAGTGACCTACG-3’, and 5’-AAGGCACAGGCAAAGTTCTCA-3’), and cloned into the pCR^™-Blunt II-TOPO^™ vector using the Zero Blunt^™ TOPO^™ PCR Cloning Kit (Invitrogen K280002). The inserted TOPO plasmids were expanded and Sanger sequencing performed to validate CRISPR mutations on each allele.

d). Embedding of Kidney Organoids for Observation and Phenotypic Drug Screening

Preparation of methycellulose plates

To prepare cultures for fixed organoid tracking over time, 15g of methylcellulose powder (Sigma-Aldrich, M0512) was autoclaved in a 500-ml Erlenmeyer flask. The autoclaved methylcellulose was dissolved in 60°C 450 ml of Advanced RPMI 1640 Medium (Gibco, 12633020). Fifty milliliters of Advanced RPMI 1640 Medium + 1X Glutamax (Gibco, 35050079) and 1% Penicillin-Streptomycin (Invitrogen, 15070063) was then added at room temperature to a final volume of 500ml and a stock concentration of 30 μg/ml methylcellulose. The final stock solution was cleared by centrifugation at 4000 × g for 2 hours. For organoid culture and imaging, 34 μl of the 30 μg/ml methylcellulose stock solution was added to 136 μl of basal differentiation media to each well of a 96-well plate (brand) to achieve the final concentration of 6 μg/ml methylcellulose optimal for organoid embedding.

Organoid embedding for growth assays

At dd13, kidney organoids were transferred from the EZSPHERE plates to a sterile 35-mm dish by gentle pipetting with wide-bore P1000 tips. Under a dissecting microscope, 10–12 organoids were picked up in 10μl of media and released into methylcellulose in a single well of the imaging plate.

e). Protein kinase inhibitor PKD screening of human kidney organoid cultures

Protein kinase inhibitor library and additional small molecule compounds:

Two hundred and fourty-seven annotated small molecule protein kinase inhibitory compounds were screened at 1 μM in the primary screen. Purchased compounds represented the following commercial libraries: EMD-Protein Kinase Inhibitor 2 (EMD Calbiochem^®, Catalog no. 539745), EMD-Protein Kinase Inhibitor 3 (EMD Calbiochem^®, Catalog no. 539746), EMD-Protein Kinase Inhibitor 4 (EMD Calbiochem^®, Catalog no. 539747).

Primary screen:

Protein kinase inhibitors were diluted in DMSO to generate 10 μM stocks. At dd14, 20 μl of each diluted compound or DMSO was added to a methycellulose well with embedded organoids (180μl of media) to achieve a final concentration of 1 μM. The plates were loaded onto an ImageXpress Micro System for live imaging throughout the screening period. The imaging was performed using the “Standard” algorithm, at 4X magnification, 2 camera binning, with laser-based and image-based focusing enabled, and well-to-well autofocus was set to “focus on plate bottom and well bottom”. To avoid observer bias, we performed blinded experiments in which the compound maps were not revealed to the researcher processing the ImageXpress output until all analyses were completed.

Scoring phenotypes:

Initially, we screened organoids generated from the PKD2−/− hESC line with the full compound library set. We categorized the outcomes of compound treatments into 3 groups: 1) “no-hit” wells were cyst formation and growth appeared normal at dd20 (cyst area ≥30% organoid size), 2) “hit” wells in which no cyst formation was observed but epithelial structures remained healthy, and 3) “non-specific hit” (NS hit) wells were no cyst formation occurred but there was clear evidence of cell death and general growth retardation in the culture.

Secondary screen:

Compounds scored as “hits” in the primary PKD2−/− focused screen were selected for validation in a secondary screen inhibiting cyst initiation in PKD1−/− and PKD2−/− mutant organoids testing a concentration range of each small molecule along with “no hit” and “NS-hit” controls. At dd14, 20 μl of each diluted compound stock or DMSO was added to a methycellulose well with embedded organoids (180 μl) to achieve a final concentration of 0.1, 1 or 10 μM (with the exception of UCN-01, which was examined at 0.2, 2 or 20 μM). The plates were imaged for 7 days using the ImageXpress Micro System as described above. Compounds that inhibit cyst formation in both PKD1−/− and PKD2−/− organoids were classified as final “hits”.

To determine the lowest concentration of QNZ capable of inhibiting cyst initiation, PKD1−/− and PKD2−/− mutant organoids were embedded as described previously, and were treated with QNZ at 0.25, 1, 2, 5 and 20 nM.

Effects of QNZ on early and late cyst growth:

To assay the effects of compounds on early cyst growth shortly after cyst initiation, dd15, PKD1−/− and PKD2−/− mutant organoids with recognizable tubular protrusions were embedded in methylcellulose wells. The organoids were treated with DMSO, Celastrol (0.1, 1 and 10 μM), Tolvaptan (0.1, 1 and 10 μM), Rapamycin (0.1, 1 and 10 μM), or QNZ (0.25, 1, 2, 5 and 20 nM) at dd16, and were imaged live as described above from dd16 to dd23). Cyst areas were quantified using ImageJ, and statistically significant size changes were determined using paired t-test (each time point compared to dd16).

To assay the effects of compounds on late cyst growth when cysts had grown extensively and detached from the parent organoid, cysts where expanded and passaged through dissociation and reformation in low adherence cell plates. PKD1−/− and PKD2−/− cysts reaching >= 1mm in diameter were embedded in methylcellulose (day 0) and imaged at day 0 and day 3 using the Zeiss AxioZoom.V16 microscope. Growing cysts were treated with DMSO or QNZ (0.01, 0.1 and 1 μM) for 6 days, and were imaged every 3 days. Cyst areas were quantified using ImageJ, and statistically significant size changes were determined using ANOVA test.

2). Characterization of the organoid system

a). Single-cell RNA-seq Data Collection and Analysis

Approximately 300 organoids were collected at dd8, dd10, dd14, dd16 and dd28 for scRNA-seq. The organoids were dissociated using 7.5 mg/ml Bacillus licheniformis cold active protease (Creative Enzymes, NATE-0633) mixed with 10 mg/ml collagenase type 2 (Worthington, #LS00417) and 125 U/ml DNase I (Worthington, #LS002058) in DPBS (150 μl) at 12°C for 20 min. The digestion mix was mixed twenty times with P-1000 wide-bore pipette tips. The dissociation reaction was terminated by mixing with 150 μl of 20% fetal bovine serum in DPBS. The cells were filtered through a pre-wetted 40-μm strainer (Falcon), and 1 ml of DPBS was used to wash cells off the strainer. The 1.3 ml of dissociated cell mix was combined with 3 ml of AutoMACS Running Buffer (Miltenyl Biotec, 130-091-221) and cells were briefly pelleted at 1250 rpm at 4°C. The cell pellet was then resuspended in 350 μl of AutoMACS Running Buffer, with 14 μM DAPI and 5 μM DRAQ5 added freshly. Single live cells (DAPI- and DRAQ5+) were selected by fluorescence-activated cell sorting (FACS) and single cell profiling with a 10x Genomics Chromium Single Cell 3’ GEM, Library & Gel Bead Kit (10X Genomics, PN-1000075). After recovery from the emulsion, cDNA was cleaned-up and amplified by PCR, examined on a 4200 Tape station (Agilent) for yield assessment, and then processed into barcoded library for Illumina sequencing. Paired-end sequencing on the Illumina HiSeq 4000 platform was performed using the HiSeq 3000/4000 SBS PE clustering kit (PE-410–001) and 150 cycle flow cell (FC-410–1002). From fastq files, quality control, alignment to reference genome (hg38) and generation of count tables of the five libraries were done using CellRanger 3.1 (10X Genomics).

The Seurat 3.0 package was used for scRNA-seq analyses (Stuart et al., 2019). The five datasets were merged using the merge function. To filter out low-quality cells, we kept cells that had between 500 and 5,500 features, fewer than 20,000 RNA counts, and less than 20% mitochondrial gene content. We integrated the datasets using the “Fast integration using reciprocal PCA” workflow presented by the Satija group (https://satijalab.org/seurat/articles/integration_rpca.html). To summarize, the merged dataset was split into a list of two Seurat objects based on their origins: group 1 consisted of day 8, 10, 13 and 16, and group 2 composed of day 28 cells (the day 28 dataset was sequenced at a different sequencing depth from other timepoints). We then normalized and identified variable features for each dataset independently using NormalizeData and FindVariableFeatures. Features that were repeatedly variable across the two Seurat objects were selected for integration run PCA on each dataset using SelectIntegrationFeatures. The FindIntegrationAnchors function was used to identify the anchors, and these anchors were used to integrate the two Seurat objects using IntegrateData with reduction specified as “rpca”. The default assay was switched to “integrated” before the integrated dataset was scaled and centered using. The RunPCA function was applied to calculate principle components (PCs), and 40 PCs were used to determine neighbor cells and cluster assignment (using the FindNeighbors and FindClusters functions). The UMAP reduction was calculated using RunUMAP to determine UMAP embedding. Differentially expressed genes of each cluster were found using the FindAllMarkers function. Clusters 2, 4, 5, 6, 7, 11, 12, 19 and 22 were subset for nephrogenic lineage examination. Using similar integration procedure, the nephrogenic subset was integrated based on their origins (day 28 versus other timepoints) before standard workflow was applied to identify clusters. Further filtering against non-nephrogenic clusters was performed based on differentially expressed gene list to achieve a “clean” in vitro nephrogenic subset.

The in vivo datasets of human week 17 fetal kidney from our 2019 study (Tran et al., 2019) (GSE124472) were used for comparison with the scRNA-seq profiles of the in vitro derived nephrogenic cell subset. After the nephrogenic cells were subset from the week 17 datasets, the in vitro and in vivo nephrogenic cells were merged using the merge function. The merged dataset was first split based on in vitro or in vivo origin of the cells. The split Seurat objects were then integrated using the Fast integration with reciprocal PCA as described above. The default assay was switched to “integrated” before the integrated dataset was analyzed using the standard workflow (ScaleData, RunPCA, RunUMAP, FindNeighbors, and FindClusters) as described above. Cell embedding was presented as a UMAP reduction output (Becht et al., 2019). “RNA” was used as the default assay (DefaultAssay function), and the integrated dataset was re-normalized using NormalizeData before assaying gene expression levels.

Similarly, the interstitial cell populations were extracted from the week 17 human fetal kidney dataset and merged with the organoid dataset. The in vitro and in vivo interstitial cells were then integrated using “rpca” reduction.

To compare single-cell transcriptomes of our organoids with kidney organoids generated by other groups, we retrieved published scRNA-seq datasets from various research groups (Combes et al., 2019; Czerniecki et al., 2018; Kumar et al., 2019; Subramanian et al., 2019; Tran et al., 2019; Wu et al., 2018)(listed in the Key Resources Table). From each dataset, we extracted cells that had between 500 and 5,500 features, fewer than 20,000 RNA counts, and less than 20% mitochondrial gene content, and performed the standard Seurat workflow to find clusters, and annotated the clusters using the differentially expressed genes (identified using the FindAllMarkers function). Nephrogenic cells were further subset based on the annotations. The subset was subjected to the standard workflow to identify cell clusters, and the procedure would be repeated if non-nephrogenic cell groups were detected to further refine the nephrogenic cell extraction. The “clean” nephrogenic subsets from all datasets were merged with the human fetal kidney nephrogenic lineage subset, and then split based on their origins before being integrated using the fast integration with reciprocal PCA workflow as detailed above. The “integrated” default assay was selected before ScaleData, RunPCA, RunUMAP, FindNeighbors, and FindClusters were used to cluster cells (presented in Figures 3E–H). The default assay was then switched to “RNA” before gene expression levels were assayed (presented in Figures 3F and S3A–C).

To compare the transcriptomic profiles of in vitro and in vivo nephron segment cells, the following clusters were extracted from the integrated in vivo/in vitro nephrogenic cell dataset: clusters 0 and 1 for podocyte, cluster 3 for proximal tubule, and clusters 9 and 13 for putative medial/distal nephron precursor. Each subset was merged, and clusters were identified. “RNA” was used as the default assay (DefaultAssay function), and the integrated datasets were re-normalized using NormalizeData before examination of gene expression levels. Differential gene test was performed using FindAllMarkers to look for genes highly expressed in the in vitro or the in vivo cells. To account for batch differences, these gene lists were compared with differentially expressed gene list from comparing the in vitro and in vivo interstitial cells, and genes that were present in all four lists (interstitium, podocyte, proximal tubule, and medial/distal nephron precursor) were considered “background differences”. Cell-type specific differences were presented in Supplemental Table 1. Expression patterns and levels of representative differentially expressed genes were examined in the all-group comparison using feature plots with UMAP reduction with three subsets presented separately based on their origins (figures S3A–C).

b). Histology

Organoids were fixed in 4% paraformaldehyde for 10 minutes at 4°C temperature and were washed three times in 1XPBS. Samples were then transferred to an embedding mold with 15% sucrose/7.5% gelatin in PBS and incubated in the gelatin solution at 37°C until the organoids sink. The organoids in gelatin solution was then frozen in a dry ice/ethanol slurry. Samples were stored at −80°C until cryosectioning and processing.

c). Immunohistochemistry and in situ hybridization

Frozen sections were warmed to room temperature for 10 minutes before the staining procedure. Citrate Buffer pH 6.0 (Sigma) was used for antigen retrieval in a pressure cooker. The slides were then washed with water and air dried for 5 min. 1.5% Seablock (ThermoFisher) in PBS + 0.25%TritonX block buffer was applied on the tissue for 1 hour at room temperature for blocking. The slides were then incubated with primary antibody mixture (diluted in block buffer) at 4°C overnight. Primary antibodies used in the study are listed as follow: WT1 (abcam, ab89901, 1:5000), JAG1 (R&D, AF599, 1:300), LAMB1 (Santa Cruz, sc-33709, 1:50), SOX9 (abcam, ab185230, 1:1000), HNF4A (R&D, MAB4605, 1:500), CUBN (R&D, AF3700, 1:500), SLC12A1 (Sigma, HPA018107, 1:500), LTL (Vector Laboratories, FL-1321, 1:300), SLC3A1 (Sigma, HPA038360, 1:500), NPHS1 (abcam, ab136927, 1:5000), POU3F3 (ThermoFisher, PA5–64311, 1:500), MAFB (R&D, MAB3810, 1:500), PAX8 (abcam, ab189249, 1:1000), CDH1(Biosciences, 610182, 1:300), PAX2 (R&D, AF3364, 1:500), GATA3 (R&D, AF2605, 1:300), ACE2 (R&D, AF933, 1:500). Secondary antibodies conjugated with AlexaFluor 488, 555, 594, and 647 (diluted to 1:1000 in block buffer) purchased from Molecular Probes. To stain the nuclei, slides were treated with 1 mg/ml Hoechst 33342 (Molecular Probes) in PBS for 5 min. ProLong Gold Antifade Reagent (Life technologies) was applied on the tissue for mounting, and images were acquired at 40X using the Leica SP8 confocal microscope.

d). RNA extraction, cDNA synthesis and quantitative polymerase chain reaction

About 200 organoids were collected for transcriptional analyses for each time point. The RNeasy Micro Kit (Qiagen, 74004) was used for RNA extraction following the manufacturer’s protocol. cDNA was synthesized from 200 μg of RNA for each sample using the SuperScript IV VILO Master Mix with ezDNase enzyme (Invitrogen, 11766050).

Quantitative polymerase chain reaction (qPCR) was performed using the Taqman Fast Advanced Master Mix (ThermoFisher, 444557) following the manufacturer’s instruction on the ViiA 7 Real-Time PCR System (ThermoFisher). The following probes from ThermoFisher were used for transcriptional analyses: WT1 (Hs01103751_m1), MAFB (Hs00534343_s1), PAX2 (Hs01057416_m1), HNF4A (Hs00230853_m1), GATA3 (Hs00231122_m1), SLC3A1 (Hs00942976_m1), SLC12A1 (Hs00165731_m1) and SLC12A3 (Hs01027568_m1).

e). Western Blot

To prepare protein lysate samples, organoids and control samples were suspended and homogenized in lysis buffer (RIPA buffer (Pierce, 89901) supplemented with 1 mM benzamidine hydrochloride (TCI America, TCB0013), 1X protease inhibitor cocktail (Cell signaling, 5871), 100 μM PMSF (Sigma-Aldrich, 11359061001), and protease inhibitor cocktail tablets (one tablet/10ml of buffer) (Sigma, 11836170001)), and left on ice for 30mins. Next, samples were centrifuged for 15 min at 16,000×g at 4°C, and supernatants were transferred to low protein binding tubes (Eppendorf). Total protein concentrations were measured using the BCA protein assay kit (Pierce, 87003–294), according to manufacturer’s instructions. Protein lysates were flash frozen and stored at −80°C.

Protein lysates were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE). 4–15% Mini-PROTEAN^™ TGX Stain-Free^™ Protein Gels (Bio-Rad, 4568086) were used. After electrophoresis, protein lysates were electroblotted to methanol activated Low-Fluorescence PVDF Transfer Membranes (Bio-Rad, IPFL20200). Membranes were then dried at 37°C for 5 minutes and then re-activated with methanol. Blots were stained with Li-Cor’s Revert-700 Total Protein Stain (Li-Cor, 926–11010) for normalization and imaged using a Li-Cor Odyssey Clx. In the subsequent incubation and washing steps, blots were placed on an orbital shaker. Blots were then de-stained following the manufacturer’ instruction and were incubated in block buffer (Li-Cor Intercept block, 13 927–60001) for 1 hour at room temperature. Blots were then transferred to a primary antibody mix (in block buffer supplemented with 0.1% Tween20) and were incubated overnight at 4°C. The following primary antibodies were used: PKD1 (Kerafast, Clone E8–8C3C10, catalog number EMD303)(Yu et al., 2007), PKD2 (Santa Cruz, sc-28331). On the following day, blots were washed four times in TBS-T (5 min each) at room temperature, and then incubated in secondary antibody (1:10,000) in block buffer with 0.1% Tween20 and 0.1% SDS for 1 hour at room temperature. The following secondary antibodies were used: IRDye^® 800CW Goat anti-Rat IgG (H + L) (Li-Cor, 926–32219) for PKD1; IRDye^® 680RD Donkey anti-Mouse IgG (H + L) (Li-Cor, 926–68072) for PKD2. Blots were then washed twice with TBS-T for 5 minutes each at room temperature, followed by two 5-minute TBS washes at room temperature. Blots were imaged using the Li-Cor Odyssey Clx system.

f). mRNA-Seq and data analysis

Samples were prepared according to library kit manufacturer’s protocol (Clontech SMARTer), indexed, pooled, and sequenced on an Illumina NovoSeq 6000. Basecalls and demultiplexing were performed with Illumina’s bcl2fastq software and a custom python demultiplexing program with a maximum of one mismatch in the indexing read. RNA-seq reads were then aligned to the Ensembl release 76 primary assembly with STAR version 2.5.1a (Dobin et al., 2013). Gene counts were derived from the number of uniquely aligned unambiguous reads by Subread:featureCount version 1.4.6-p5 (Liao et al., 2014). logCPM values were obtained with the ‘EdgeR’ package (Robinson et al., 2010) and used to calculate the pairwise correlation coefficients between samples. TPM values were calculated based on CPM and length of exon models of genes in the corresponding annotation. Differential expression analysis was performed with DESeq2 (Love et al., 2014), with thresholds described in supplementary tables. To generate the heatmaps in Fig. S7C–F, top 50 differentially expressed genes by adjusted p-values were extracted and ranked by fold change; for each gene, the log2-transformed TPM in each sample was normalized by the mean value across all samples in comparison before plotting with ‘heatmap.2’ function in the ‘gplots’ package in R. Gene Ontology (GO) analysis was performed with the ‘clusterProfiler’ package in R (Yu et al., 2012), with only top 3 terms by p-value shown. For a full list of TPM values of all samples, please refer to Supplemental Table 1.

g). TUNEL Assay

Cryo-sections were stained with the Click-iT Plus TUNEL Assay kit (Invitrogen) by following the manufacturer’s instruction. Briefly, the sections were incubated with the TdT reaction buffer for 10 min at 37°C. The slides were then rinsed with deionized water and incubated with 3%BSA and 0.1% Triton X-100 in PBS for 5 minutes followed by PBS washing. Next, the samples were incubated with the Click-iT Plus TUNEL reaction cocktail in a humidified chamber at 37 °C for 30 min protected from light. This was followed by PBS washes and incubation with 2% sea block for 1hrs. After that, we performed immunofluorescent staining as detailed above.

h). FACS analysis

Upon dissociation using 7.5 mg/ml Bacillus licheniformis cold active protease (Creative Enzymes, NATE-0633) mixed with 10 mg/ml collagenase type 2 (Worthington, #LS00417) and 125 U/ml DNase I (Worthington, #LS002058) in DPBS (150 μl) at 12°C for 20 min, organoid cells were pelleted. The pellet was resuspended in autoMACs running buffer containing DAPI (1:1000) and DRAQ5 (1:2000) then analyzed on an ARIA II FACS using gates to detect DRAQ5+ and DAPI- cells.

3). Vascularization of transplanted kidney organoids

a). Renal Capsule Transplant

All surgical procedures were carried out with appropriate oversight and compliance following guidelines after institutional review by USC’s Institutional Animal Care and Use Committee (IACUC). The procedure was adapted from (Yoshimura et al., 2017). Week 8–12 NOD.CB17-Prkdc<SCID>/J mice were anesthetized with Ketamine/Xylazine. The surgery site on the dorsal flank was shaved and swabbed with Proviodine/alcohol. An 8–10 mm incision was made in the flank and the fascia was incised before the kidney was externalized. The kidney capsule was kept moist with sterile saline during the procedure. A small incision was made in the outer membrane of the renal capsule at the caudal end, using a sharp 24g needle and the sub capsular space is flushed with 1ml of basal differentiation media using a blunted 24g needle 30g needle (B30–50, Strategic Applications, Inc.) attached to a 1 ml syringe. Two agarose rods (2mm long, 0.5mm diameter) were pushed into the sub capsular space in the shape of an open V using forceps. A 20g indwelling needle (SURFLO^® PTFE I.V. Catheter needle, VWR, TESR-OX2025CA) attached to a 1 ml syringe and draw up 3–4 dd13–14 organoids basal differentiation media into the needle. The indwelling needle was inserted under the renal capsule to place organoids between the agarose rods. The capsule incision was cauterized, and the kidney was replaced into the retroperitoneum. The muscle layer was sutured, and the skin was closed with wound clips.

b). In vivo dextran uptake assay

Two weeks after the implantation of kidney organoids, NOD/SCID mice were anesthetized with isoflurane (3% in induction chamber). 200 μl of fluorescently labelled dextran (3 and 70 kDa) diluted in PBS or sterile PBS (negative control) was administered by retro-orbital injections. 10 minutes after the injection, the mice were euthanized, and the kidneys were collected. The mixture of fluorescent dextran was consisted of 100 μl Tetramethylrhodamine-conjugated 70 kDa lysine fixable dextran (Invitrogen, D1818) (1 mg/animal) and 100 μl of 3000 Da MW Texas Red^™ conjugated dextran (Invitrogen, D3328) (1 mg/animal). We chose the 3 kDa dextran as the low molecular weight and the 70 kDa dextran as the high molecular weight dextran in the physiologically relevant ranges to investigate the glomerular filtration barrier based on previous studies by van den Berg et al. (2020). To label the proximal tubules on the cryosections of the dextran-injected kidneys, immunofluorescence was performed using LRP2 antibody (MyBioSource, MBS690201). All animal protocols were approved by the Institutional Animal Care and Use of Committee at the University of Southern California.

c). Image acquisition and analysis

Image acquisition of sections was performed using Leica SP8-X confocal fluorescence imaging system (Leica Microsystems, Germany) in 1024×1024 pixels using a 63X Leica oil immersion objective (NA 1.6). Image masking was performed and quantified with Imaris 9.7 software (Oxford Instruments, United Kingdom). The thresholding of the signal of 3000 Da dextran was based on the intensity histogram of Texas Red^™. Fluorescent punctae larger than 1.25 μm2 were scored. Masks were classified to their respective anatomical structures based on the presence of LRP2/MAFB markers.

QUANTIFICATION AND STATISTICAL ANALYSIS

Details of sample size were provided in the Method Details section for each experiment. Wilcoxon Rank Sum test was performed, and p-values were presented in Figure 5D. Kruskal-Wallis test was performed, and p-values were presented in Figures S6D–I. Paired t-test was performed, and p-values were summarized in Figure 7C. ANOVA test was performed for Figures 7D, S7B, S7I and S7L.

Supplementary Material

Supplementary Video 1a

Supplemental Movie 1: Movie showing progress of wildtype (a) PKD1−/− mutant organoids and (b) PKD2−/− mutant organoids from day 14 to day 20. Scale bars denoted 200 μm.

Supplementary Video 2a

Supplemental Movie 2: Movie showing progress of DMSO-treated PKD mutant organoids from day 14 to day 20 at low (2a) and high (2b) power. Scale bars denoted 100 μm. (Related to Figure 6).

Supplementary Video 3a

Supplemental Movie 3: Movie showing progress of growth inhibitor-treated PKD mutant organoids from day 14 to day 20 at low (3a) and high (3b) power. Scale bars denoted 100 μm. (Related to Figure 6).

Supplementary Video 4a

Supplemental Movie 4: Movie showing progress of cyst initiation inhibitor-treated PKD mutant organoids from day 14 to day 20 at low (4a) and high (4b) power. Scale bars denoted 100 μm. (Related to Figure 6).

Supplemental Table 1

Supplemental Table 1: Differentially expressed gene lists that were used to identify clusters and for comparisons of in vitro and in vivo cell types (Related to Figures 2, 3, S2, and S3). TPM values and GO terms from bulk RNA-seq analyses for various comparisons (related to Figures 7 and S7).

Supplemental Table 2

Supplemental Table 2: Annotated list of all PKI compounds used in the study (Related to Figure 6).

Supplemental Table 3

Supplemental Table 3: gRNA and primer sequences used to generate and validate PKD1−/− and PKD2−/− hESC lines (Related to Figures 5 and S5, and Method Details).

INCLUSION AND DIVERSITY STATEMENT

One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in science. While citing references scientifically relevant for this work, we also actively worked to promote gender balance in our reference list.