Manipulation of Nephron-Patterning Signals Enables Selective Induction of Podocytes from Human Pluripotent Stem Cells

Abstract

Visual Abstract

Background 

Previous research has elucidated the signals required to induce nephron progenitor cells (NPCs) from pluripotent stem cells (PSCs), enabling the generation of kidney organoids. However, selectively controlling differentiation of NPCs to podocytes has been a challenge.

Methods 

We investigated the effects of various growth factors in cultured mouse embryonic NPCs during three distinct steps of nephron patterning: from NPC to pretubular aggregate, from the latter to epithelial renal vesicle (RV), and from RV to podocyte. We then applied the findings to human PSC-derived NPCs to establish a method for selective induction of human podocytes.

Results 

Mouse NPC differentiation experiments revealed that phase-specific manipulation of Wnt and Tgf-β signaling is critical for podocyte differentiation. First, optimal timing and intensity of Wnt signaling were essential for mesenchymal-to-epithelial transition and podocyte differentiation. Then, inhibition of Tgf-β signaling supported domination of the RV proximal domain. Inhibition of Tgf-β signaling in the third phase enriched the podocyte fraction by suppressing development of other nephron lineages. The resultant protocol enabled successful induction of human podocytes from PSCs with >90% purity. The induced podocytes exhibited global gene expression signatures comparable to those of adult human podocytes, had podocyte morphologic features (including foot process–like and slit diaphragm–like structures), and showed functional responsiveness to drug-induced injury.

Conclusions 

Elucidation of signals that induce podocytes during the nephron-patterning process enabled us to establish a highly efficient method for selective induction of human podocytes from PSCs. These PSC-derived podocytes show molecular, morphologic, and functional characteristics of podocytes, and offer a new resource for disease modeling and nephrotoxicity testing.

Podocyte disorders manifest as nephrotic syndrome and/or glomerulosclerosis, which progress to renal failure. Thus, growing attention has been paid to podocyte research in recent years.^1,2 Owing to the poor availability of primary human podocytes, artificially immortalized podocyte cell lines^3,4 have made great contributions to many podocyte studies. However, these cells do not retain the original characteristics of podocytes, including abundant expression of slit diaphragm–associated genes and proteins.⁵ The lack of resources for podocytes with sufficient functional characteristics has been a bottleneck in this field.

We and others previously developed methods for induction of nephron progenitor cells (NPCs) from pluripotent stem cells (PSCs), enabling derivation of kidney organoids.^6–9 Molecular profiling of the sorted podocytes, comprising approximately 7.5% of the human kidney organoids, confirmed characteristic features that were shared with murine and human podocytes.¹⁰ Recent progress in the kidney organoid field has achieved higher-order organization.⁹ However, it remains a challenge to selectively induce podocytes by controlling the nephron-patterning process from NPCs. Although several groups have reported methods for induction of podocyte-like cells from human induced PSCs (hiPSCs),^11–13 the resultant cells expressed only a few selected marker genes at quite low levels and lacked typical slit diaphragm formation. We reasoned that this issue could be addressed by sufficient understanding of the podocyte specification process and signaling from NPCs.

The kidney develops by interactions of ureteric bud (UB) and metanephric mesenchyme (MM). The MM includes NPCs and stromal progenitors,¹⁴ the former of which express transcription factor Six2 and give rise to epithelial nephrons.¹⁵ Wnt signaling from the UB triggers condensation of a subset of NPCs below the UB tip to form the pretubular aggregate (PA), followed by the epithelial renal vesicle (RV).^16,17 These steps are designated mesenchymal-to-epithelial transition (MET). The RV shows proximodistal polarization, at least by gene expression levels.^18–20 Each part of the RV further elongates along the proximodistal axis and differentiates into committed nephron segments, including podocytes, parietal epithelial cells (PECs), proximal tubules (PTs), and distal tubules (DTs).

Previous genetic studies revealed the requirement and sufficiency of Wnt signaling for the MET process.^{16,17,21–23} Accordingly, in vitro experiments demonstrated the sufficiency of transient Wnt signaling for MET induction in the isolated MM.^24,25 Furthermore, a recent study showed both promotional and suppressive roles of Wnt signaling during the later phase (after RV formation) of distal and proximal nephron development, respectively.²⁶ However, the patterning mechanism for the proximodistal domain of the RV as well as the signals that specify the podocyte lineage during the later process of nephron patterning remain to be elucidated.

In this study, we investigated the podocyte lineage–specification factors by dissecting the nephron development process into three distinct steps: NPCs to PA, PA to RV, and RV to podocytes. For this purpose, we initially employed mouse embryonic NPCs, and then applied the findings in the mouse experiments to hiPSC-derived NPCs to establish a method for selective induction of human podocytes.

Methods

Animals

MafB-GFP knock-in (MafB-GFP) mice were described previously.²⁷ Six2-GFP-Cre transgenic (Six2-GFP) mice¹⁵ were kindly provided by Dr. Andrew P. McMahon (University of Southern California). All animal experiments were performed in accordance with institutional guidelines and approved by the Licensing Committee of Kumamoto University (A29–040).

Podocyte Induction from Mouse NPCs

Metanephroi were isolated from embryonic day (E) 15.5 MafB-GFP embryos and manually minced in PBS(−) using forceps. Minced tissues were dissociated by incubation in 0.25% trypsin-EDTA at 37°C for 8 minutes. After blocking with normal mouse serum (Thermo Fisher Scientific), dissociated cells were stained with anti-Robo2 and anti-Pdgfrb primary antibodies for 30 minutes on ice. After secondary antibody staining, NPCs were sorted as a Robo2^high/Pdgfrb⁻/MafB-GFP⁻ population using a FACS SORP Aria (BD Biosciences). NPCs were seeded at 100,000 cells/150 µl serum-free differentiation medium with various concentrations of CHIR99021 (CHIR) (Axon Medchem) and 10 µM Y27632 (Wako) in 96-well low-cell-binding U-bottom plates (Thermo Fisher Scientific). The plates were centrifuged (210 × g, 4 minutes) and cultured at 37°C. After initial CHIR induction, aggregated cells were transferred to a 3.0-µm pore transwell insert (Corning) in serum-free differentiation medium containing Fgf9 (10 µg/ml; R&D Systems) and cultured with the following factors: CHIR (0.5, 2 µM); IWR-1 (2 µM; Sigma-Aldrich); activin A (10 ng/ml; R&D Systems); SB431542 (SB) (5, 25, 100 µM; Wako); retinoic acid (RA) (10 µM; Sigma-Aldrich); BMS493 (10 µM; Tocris Bioscience); Jagged1-Fc (10 µg/ml; R&D Systems); ɤ-secretase inhibitor (2 µM; Merck). The culture medium was replaced at specified points or every 2 days. Induced tissues were harvested for analysis at day 6. The serum-free differentiation medium comprised DMEM/F12 supplemented with 1% insulin-transferrin-selenium and 1% penicillin-streptomycin (Thermo Fisher Scientific). Detailed antibody information is provided in Supplemental Table 1.

Cell Culture

The NPHS1-GFP knock-in hiPSC line was generated and maintained as described.¹⁰ The 201B7 hiPSC line²⁸ was maintained in the same way. The former line was used to investigate the induction efficiency, whereas the latter was used for immunohistochemical, TUNEL assay, qRT–PCR, RNA-seq, electron microscopy, and western blot analyses, as well as drug-induced injury tests. The RN7 hiPSC line was established from peripheral blood of a healthy volunteer using Sendai virus vectors as described.²⁹ The detailed characterization of this line will be described elsewhere. The experiments were performed in accordance with the institutional guidelines and approved by the licensing committee of Kumamoto University (Nos. 359 and 1453). A conditionally immortalized human podocyte cell line was purchased from Bristol University and cultured as described.⁴ Briefly, cells were cultured in RPMI 1640 supplemented with 10% FBS at 33°C and passaged every 4–5 days. For induction of the podocyte phenotype, cells were cultured at 37°C for 14 days. Cells with fewer than four passages since arrival were used for analyses.

Podocyte Induction from hiPSC-Derived NPCs

NPCs were induced from hiPSCs as described.⁹ NPCs were sorted as an ITGA8⁺/PDGFRa⁻ population using the FACS SORP Aria.³⁰ Sorted human NPCs were seeded at 50,000 or 100,000 cells/150 µl serum-free differentiation medium containing CHIR (3 µM) and Y27632 (10 µM) in 96-well low-cell-binding U-bottom plates. The plates were centrifuged (210 × g, 4 minutes) and cultured at 37°C. After 24 hours, aggregated cells were transferred to a 3.0-µm pore transwell insert in serum-free differentiation medium containing Fgf9 (10 µg/ml) and cultured with step 2 factors including IWR-1 (2 µM), SB (5 µM), and RA (10 µM). At 48 hours after culture initiation, differentiating cells on the transwell were transferred to fresh medium containing step 3 factors including IWR-1 (2 µM) and SB (5 µM). The culture medium was replaced at specified points or every 3 days. The induced podocytes were harvested for analysis on day 9 or 12.

Kidney Organoid Induction from hiPSC-Derived NPCs

Kidney organoids were differentiated by coculture of induced NPCs with mouse embryonic spinal cord taken from E12.5 embryos, and cultured at the air–fluid interface on polycarbonate filters (0.8 µm; Whatman) supplied with DMEM/F12 supplemented with 10% FBS and 1% penicillin-streptomycin.³¹ The culture medium was refreshed every 3 days. The induced organoids were harvested for analyses on day 9.

Collection of Human Adult Podocytes

Human adult kidney tissues were obtained from the normal parts of excised kidney specimens from three patients with renal cancer under approval from the Ethics Review Committee of the Faculty of Life Sciences, Kumamoto University (No. 1050) after receiving informed consent from each patient. The patients were aged 49, 69, and 80 years, and all three had normal renal function and normal urinalysis findings. The normal parts of the kidney specimens were immediately immersed in high-glucose DMEM (Sigma-Aldrich) supplemented with 10% FBS on ice. Glomeruli were isolated as described³² and sequentially digested with an enzyme mixture of collagenase XI (1 mg/ml) and dispase (2.4 U/ml) at 37°C for 20 minutes, followed by 0.25% trypsin-EDTA at 37°C for 8 minutes. After the glomerular dissociation, erythrocytes were hemolyzed with RBC lysis buffer (Thermo Fisher Scientific). After blocking, the dissociated cells were stained with anti-human NEPHRIN and anti-human PODOCALYXIN primary antibodies for 30 minutes on ice. After secondary antibody staining, human adult podocytes were sorted as a NEPHRIN⁺/PODOCALYXIN⁺ population using the FACS SORP Aria.

Drug Injury of Induced Podocytes

Induced podocytes in 96-well low-cell-binding U-bottom plates were treated with 30 µg/ml puromycin aminonucleoside (PAN) (Wako) in 200 µl DMEM/F12 supplemented with 5% FBS, 1% insulin-transferrin-selenium, and 1% penicillin-streptomycin. After 48 hours of treatment at 37°C, podocytes were harvested and analyzed.

Statistical Analyses

Data are presented as mean±SEM. t test was applied for statistical analyses of differences between two groups. Differences with values of P<0.05 were considered statistically significant.

Data Availability

The RNA-seq data have been deposited in the NCBI Bio Sample database under GEO accession number GSE116471.

A detailed description of methods is included as Supplemental Information.

Results

Optimal Duration and Strength of Wnt Signaling Are Essential for MET and Podocyte Differentiation

We initially developed a culture system to enable differentiation of epithelialized nephrons from purified mouse embryonic NPCs. For quantitative evaluation of podocyte induction efficiency, we used MafB-GFP mice, which specifically express strong GFP signals in differentiated podocytes.^27,33 On the basis of previous reports,^34,35 we identified the Robo2^high/Pdgfrb⁻/MafB-GFP⁻ fraction highly enriched with NPCs in the E15.5 kidney (Figure 1A). The purity of this fraction was confirmed using Six2-GFP mice (NPCs: 95.2%±0.37%) (Supplemental Figure 1) and the isolated cells expressed NPC-specific genes at comparable levels to those in Six2-GFP⁺ NPCs (Figure 1B). Thus, we isolated NPCs from MafB-GFP mice as the Robo2^high/Pdgfrb⁻/MafB-GFP⁻ fraction and used these cells in the following experiments.

Figure 1.: Optimal duration and strength of Wnt signaling are essential for MET and podocyte differentiation. (A) Flow cytometry analysis of E15.5 MafB-GFP embryonic kidneys with anti-Robo2 and anti-Pdgfrb antibodies (n=3). NPCs were purified as a Robo2^high/Pdgfrb⁻/MafB-GFP⁻ population. (B) Relative expression levels of NPC-associated genes to β-actin expression (n=3). Blue and gray bars, Six2-GFP⁻ and Six2-GFP⁺ cells sorted from E15.5 Six2-GFP embryonic kidneys, respectively; orange bars, Robo2^high/Pdgfrb⁻/MafB-GFP⁻ NPCs sorted from E15.5 MafB-GFP embryonic kidneys. Ab-sorted, sorted upon antibody-based staining. (C) Schematic representation of the experiments investigating the initial CHIR treatment of NPCs for MET induction. (D) Bright-field (left panels) and fluorescence (MafB-GFP, right panels) images of induced tissues at day 6 by modulating the concentration and duration of initial CHIR treatment (n=3). Scale bars, 200 µm. (E) Results of flow cytometry analyses at day 6 by modulating the concentration and duration of initial CHIR treatment (n=3). *P<0.05, **P<0.01, t test. (F) Immunostaining for Lhx1 (PA and RV marker) and Cdh1 (RV marker), and nuclear DAPI staining in differentiating cells at the indicated time points (n=3). Low-magnification (upper panels) and high-magnification (lower panels) images are presented. Epithelial vesicles are outlined by yellow dashed lines. Scale bars, 20 µm. (G) Temporal expression levels of stage-specific marker genes at the indicated time points (n=3). Relative expression levels of transcripts to β-actin expression are presented. (H) Schematic representation of the time course for the early stages of differentiation. basal, basal condition. Data are shown as means±SEM.

We next examined the optimal duration and concentration of Wnt signaling for the initial differentiation of NPCs (step 1). For this, NPCs were transiently treated with 1–10 µM CHIR, a chemical Wnt agonist, for 12–48 hours and analyzed at day 6 (Figure 1C). After the transient treatment, we continuously administered 0.5 µM CHIR to support tissue growth and defined this as the basal condition (Supplemental Figure 2). Macroscopic examination revealed that only treatment with 3 or 5 µM CHIR for 24 or 48 hours induced epithelial tissue formation (Supplemental Figure 3). Interestingly, MafB-GFP⁺ podocytes were robustly induced upon treatment with 3 µM CHIR for 24 hours, whereas few podocytes were induced with 5 µM CHIR or 48 hours of treatment (Figure 1, D and E). Therefore, we concluded that 3 µM CHIR treatment for the initial 24 hours was optimal for MET induction and subsequent podocyte differentiation. These results suggest that NPCs require optimal strength and duration of Wnt treatment for epithelialization, and that the initial Wnt signaling has a strong effect on the eventual podocyte formation.

Gene expression profiling and histologic analysis revealed that the induced cells started to show Lhx1⁺, Wnt4⁺, and Pax8⁺ aggregates at 24 hours, and subsequently formed an epithelial vesicle at 48 hours (Figure 1, F and G). These results indicated that NPCs underwent MET upon transient CHIR treatment, and that the differentiating cells at 24 and 48 hours corresponded to the PA and RV stages, respectively (Figure 1H).

Inhibition of Tgf-β Signaling in the PA-to-RV Differentiation Step Supports Domination of the RV Proximal Domain

Because the RV proximal domain is considered to contain the precursors of podocytes,¹⁹ we hypothesized that directed differentiation of NPCs into the proximal RV would enhance the podocyte induction efficiency. Therefore, we examined the effects of growth factors or inhibitors on the PA-to-RV differentiation step (24–48 hours, step 2; Figure 2A). Consistent with our earlier findings (Figure 1, D and E), a higher concentration of CHIR (2 µM) suppressed the ratio and number of podocytes compared with the basal condition (0.5 µM CHIR) (Figure 2, B–D). IWR-1, an inhibitor of Wnt/β-catenin signaling, inhibited epithelialization (Supplemental Figure 4B). In contrast, 25 µM SB, a Tgf-β/SMAD signaling inhibitor, significantly increased the ratio and final yield of podocytes (Figure 2, B–D, Supplemental Figure 4A). All other reagents investigated, including activin A, RA, BMS493 (a pan-RA receptor antagonist), Jagged-1 Fc, and ɤ-secretase inhibitor (DAPT; a Notch signal antagonist), failed to increase the podocyte induction efficiency (Supplemental Figure 4, B–D).

Figure 2.: Inhibition of Tgf-β signaling in the PA-to-RV differentiation step expands the RV proximal domain. (A) Schematic representation of the experiments examining the differentiation step from the PA to the RV. (B) Bright-field (upper panels) and fluorescence (MafB-GFP, lower panels) images of induced tissues at day 6 by modulating the differentiation factors in the PA-to-RV step (n=4). Scale bars, 200 µm. SB, 25 µM. (C) Results of flow cytometry analyses at day 6 by modulating the differentiation factors in the PA-to-RV step (n=4). **P<0.01, t test. (D) Estimated podocyte numbers in induced tissues at day 6 by modulating the differentiation factors in the PA-to-RV step (n=4). MafB-GFP⁺ cell numbers were obtained by multiplying the percentages of MafB-GFP⁺ cells and the total cell numbers in the induced tissues. *P<0.05, **P<0.01, t test. (E) Relative expression levels of proximal and distal RV marker genes to β-actin expression at 48 hours (n=3). (F) Immunostaining for Wt1 (proximal RV marker) and nuclear DAPI staining in cytocentrifuged differentiating cells at 48 hours (n=3). Scale bars, 25 µm. (G) Percentages of Wt1⁺ proximal RV cells at 48 hours (n=3). **P<0.01, t test. (H) Estimated numbers of Wt1⁺ and Wt1⁻ RV cells in differentiating tissues at 48 hours (n=3). Wt1⁺ cell numbers were obtained by multiplying the percentages of Wt1⁺ cells and the total cell numbers in the tissues. *P<0.05, **P<0.01; n.s., not significant; t test. Data are shown as means±SEM.

Examination of the gene expression profile at the RV stage (48 hours) revealed that SB treatment upregulated proximal domain markers such as Wt1, MafB, and FoxC2 (Figure 2E). In contrast, the expression levels of distal RV markers were slightly decreased (Figure 2E). A cellular-level analysis showed that both the ratio and number of Wt1⁺ proximal RV cells were increased in the SB-treated tissue, whereas 2 µM CHIR treatment decreased proximal RV cells (Figure 2, F–H).

These results suggest that inhibition of Tgf-β/SMAD signaling in the PA-to-RV differentiation step supports domination of the RV proximal domain, thereby increasing the final ratio and number of podocytes in the tissue.

Inhibition of Tgf-β Signaling after the RV Step Enriches the Podocyte Fraction by Suppressing the Development of Other Nephron Lineages

Next, we investigated the factors that control podocyte lineage specification after proximalized RV induction (step 3; Figure 3A). Continuous administration of SB from 48 hours to day 6 significantly increased MafB-GFP⁺ podocytes (Figure 3, B and C, Supplemental Figure 5A). Treatment with 2 µM CHIR completely abolished podocyte differentiation, whereas treatment with IWR-1 did not significantly increase the podocyte proportion (Supplemental Figure 5, B and C). These results suggest that strong Wnt signaling suppresses podocyte differentiation, consistent with a previous report.²⁶ Other tested factors, including activin A, RA agonist/antagonist, and Notch agonist/antagonist, did not significantly affect the podocyte proportion (Supplemental Figure 5, B and C). SB treatment at this step slightly increased the final number of podocytes and decreased the final number of DTs, although the difference was NS (Figure 3D, Supplemental Figure 5D). Meanwhile, the number of LTL⁺ PTs was significantly decreased (Figure 3D, Supplemental Figure 5D). A gene expression analysis revealed elevated expression of podocyte-related genes and decreased expression of PEC-, PT-, and DT-related genes (Supplemental Figure 5E). Importantly, podocytes developed by this high–induction efficiency method coexpressed a set of typical podocyte-specific proteins (Figure 3E). These results indicate that inhibition of Tgf-β/SMAD signaling during the nephron segment specification step suppresses the development of other nephron segments, resulting in enhanced podocyte induction efficiency (Figure 3F).

Figure 3.: Inhibition of Tgf-β signaling after the RV step enriches the podocyte fraction by suppressing the differentiation of other nephron lineages. (A) Schematic representation of the experiments examining the differentiation step after the RV step. (B) Bright-field (upper panels) and fluorescence (MafB-GFP, lower panels) images of induced tissues at day 6 by modulating the differentiation factors after the proximalized RV step (n=4). Scale bars, 200 µm. SB, 25 µM. (C) Results of flow cytometry analyses at day 6 by modulating the differentiation factors in the proximalized RV step (n=4). **P<0.01, t test. (D) Estimated cell numbers of podocytes (MafB-GFP⁺), PTs (LTL⁺), and DTs (Cdh1⁺) in induced tissues at day 6 by modulating the differentiation factors after the proximalized RV step (n=4). The cell numbers in each segment were obtained by multiplying the percentages of segment-specific marker-positive cells (see Supplemental Figure 5D) and the total cell numbers in the induced tissues. *P<0.05; n.s., not significant; t test. (E) Immunostaining for characteristic podocyte markers (Wt1, Nephrin, and Podocin) and nuclear DAPI staining in induced tissues at day 6 (n=3). Scale bars, 10 µm. (F) Model of step-by-step differentiation of mouse NPCs toward podocytes. Data are shown as means±SEM.

Selective Podocyte Induction from hiPSCs

Next, we pursued a high-efficiency human podocyte induction method from hiPSCs. For this purpose, we employed the NPHS1-GFP line¹⁰ and its parental 201B7 line,²⁸ and mainly utilized the NPHS1-GFP line for monitoring of podocyte induction efficiency. We induced human NPCs by a previously described protocol⁹ and sorted pure NPCs as an ITGA8⁺/PDGFRa⁻ fraction (Figure 4A, Supplemental Figure 6A).³⁰ First, we titrated the concentration and duration of CHIR treatment in the NPC-to-PA phase (step 1) and cultured the cells in the basal condition up to day 9 (steps 2 and 3), on the basis of the timing of human nephron differentiation.¹⁰ The final podocyte induction efficiency at day 9 was examined by the NPHS1-GFP⁺ cell ratio. We found that 3 µM CHIR treatment for 24 hours was most efficient for podocyte induction (Figure 4B, Supplemental Figure 6B). Histologic analysis showed that NPCs differentiated into the LHX1⁺/CDH1⁻ PA after 24 hours of CHIR induction and formed the LHX1⁺/CDH1⁺ RV after 48 hours (Supplemental Figure 6C). During the PA-to-RV differentiation step (step 2), similar to the findings in mice, SB treatment increased the ratio of podocytes (Figure 4C, Supplemental Figure 6D). Administration of IWR-1 and RA in step 2 also increased the proportion of podocytes in hiPSC-derived NPCs, and combination of all three factors synergistically increased the final podocyte ratio (Figure 4C). Analysis at the RV stage (48 hours) showed that the combination of IWR-1, SB, and RA strongly induced the expression of proximal RV markers (Figure 4D) and significantly increased the ratio of proximal RV cells in the tissue (Figure 4E). Although the difference was NS, the estimated number of WT1⁺ proximal RV cells was slightly increased, reflecting the decreased total cell number resulting from the dramatic decrease in WT1^– nonproximal RV cells (Figure 4F). Immunohistochemical staining revealed that the pHH3⁺ cell ratio in distal RV cells (JAG1⁺²⁰) was decreased (Figure 4G, Supplemental Figure 6E), suggesting suppression of cell proliferation. After RV formation (step 3), the combination of IWR-1 and SB showed maximal induction efficiency for podocytes (Figure 4H, Supplemental Figure 6F). In this step, the optimized condition mainly suppressed the differentiation of LTL⁺ PTs, whereas no significant change was seen in CDH1⁺ DTs (Figure 4I). When we compared the optimized condition (podocyte condition) with the control condition (culture in the basal condition in step 2 and step 3) side by side (Supplemental Figure 7A), the podocyte condition suppressed the proliferation of PT precursors (JAG1^+,20), but not DT precursors (POU3F3^+,20) (Supplemental Figure 7, B and C). TUNEL assays confirmed that the podocyte condition did not significantly increase cell apoptosis in all three lineages at day 6 (Supplemental Figure 7, D and E). We further evaluated the gene expression kinetics during differentiation into podocytes. Podocyte markers were dramatically increased during step 3 (from day 2 to day 9), whereas PT and DT segment markers were not upregulated (Supplemental Figure 7F), suggesting that our condition selectively allows podocyte differentiation. We also noted that PECs were absent in the induced podocytes, but detected in the control condition (Supplemental Figure 7, G and H). The finalized protocol allowed us to induce podocytes with >90% purity (Figure 4, H and J), and produced a two times higher yield of podocytes compared with the control condition (Supplemental Figure 7I) through RV proximalization and subsequent suppression of the PT lineage.

Figure 4.: Step-by-step optimization enables selective podocyte induction from hiPSC-derived NPCs. (A) Schematic representation of the experiments examining podocyte differentiation from hiPSC-derived NPCs. (B) Results of flow cytometry analyses at day 9 by modulating the concentration and duration of the initial CHIR treatment (step 1) (n=5). **P<0.01, t test. (C) Results of flow cytometry analyses at day 9 by modulating the differentiation factors in the PA-to-RV step (step 2) (n=3). SB, 5 µM. *P<0.05, **P<0.01, t test. (D) Relative expression levels of proximal and distal RV marker genes to β-actin expression at 48 hours (n=3). (E) Percentages of WT1⁺ proximal RV cells at 48 hours (n=3). **P<0.01, t test. (F) Estimated numbers of WT1⁺ and WT1⁻ RV cells in differentiating tissues at 48 hours (n=3). WT1⁺ cell numbers were obtained by multiplying the percentages of WT1⁺ cells and the total cell numbers in the tissues. **P<0.01; n.s., not significant; t test. (G) Percentages of pHH3⁺ proliferating cells within proximal and distal RV cells at 48 hours (n=3). (H) Results of flow cytometry analyses at day 9 by modulating the differentiation factors after the proximalized RV step (step 3) (n=4). SB, 5 µM. **P<0.01, t test. (I) Estimated cell numbers of podocytes (NPHS1-GFP⁺), PTs (LTL⁺), and DTs (CDH1⁺) in induced tissues at day 9 by modulating the differentiation factors after the proximalized RV step (n=3). The cell numbers in each segment were obtained by multiplying the percentages of segment-specific marker-positive cells and the total cell numbers in the induced tissues. **P<0.01; n.s., not significant; t test. (J) Bright-field (left panel) and fluorescence (NPHS1-GFP, right panel) images of induced podocytes at day 9 (n=4). Scale bars, 200 µm. Data are shown as means±SEM.

To further test the versatility of the protocol, we applied the same condition to the parental hiPSC line (201B7) and another hiPSC line derived from a different donor (RN7). We checked the podocyte induction selectivity by comparison with the conventional kidney organoid differentiation method⁶ that allows unbiased differentiation of all nephron segments (Figure 5A). The kidney organoids derived from each hiPSC line contained around 10%–30% podocytes, whereas the selective podocyte induction method achieved 65%–95% efficiency (Figure 5, B and C, Supplemental Figure 8, A–C). The resultant tissues were discoid with thicker central regions similar to conventional kidney organoids, and histologic analysis showed that podocytes were distributed throughout the tissues (Figure 5, D and E). Thus, by optimizing the combination of factors for each step of the differentiation procedure, we succeeded in developing an efficient podocyte induction method for multiple hiPSC lines.

Figure 5.: The established protocol induces human podocytes much more efficiently than the conventional kidney organoid system. (A) Outline of the protocol for selective podocyte induction from hiPSCs compared with the conventional kidney organoid induction. (B and C) Flow cytometry analyses of NEPHRIN and PODOCALYXIN expression (B) and percentages of NEPHRIN⁺/PODOCALYXIN⁺ cells (C) at day 9 in kidney organoids and selectively induced podocytes from hiPSCs (201B7) (n=3). **P<0.01, t test. (D) Horizontal (left panel) and coronal (right panel) histologic views of induced podocytes (201B7) at day 9 stained with hematoxylin and eosin (n=3). Scale bars, 200 µm. (E) Immunostaining for podocyte (WT1), proximal tubule (LTL), and distal tubule (CDH1) markers, and nuclear DAPI staining in resultant cells in each condition at day 9 (201B7) (n=3). Scale bars, 100 µm. (F) Results of qRT–PCR analyses for comparisons of podocyte-associated gene expression levels (n=3). Relative expression levels of transcripts to β-actin expression are presented. Cell line, conditionally immortalized human podocyte cell line cells; hiPSCs, undifferentiated hiPSCs; human adult podocytes, freshly sorted podocytes from human adult kidney; induced podocytes, selectively induced podocytes from hiPSCs by our protocol; NPCs, ITGA8⁺/PDGFRa⁻ NPCs induced from hiPSCs. Data are shown as means±SEM.

Induced Podocytes Exhibit Characteristic Gene Expression Profiles of Human Podocytes

The induced podocytes were able to survive and maintain their characteristics until day 12, but gradually regressed thereafter with continuous culture in step 3 medium (Supplemental Figure 8D). Thus, we evaluated the gene expression profile of the induced podocytes at day 12 compared with hiPSCs, immortalized human podocyte cell line cells,⁴ and human adult podocytes sorted as a NEPHRIN⁺/PODOCALYXIN⁺ population from sieved human adult kidney glomeruli (Supplemental Figure 9, A and B) as a genuine positive control. Quantitative RT–PCR analyses demonstrated that induced podocytes showed high expression levels of podocyte-specific genes comparable to those in human adult podocytes (Figure 5F). Of note, even in the well established podocyte cell line cells, the expression levels of NPHS1 and NPHS2 were quite low compared with those in human adult podocytes, whereas the expression of SYNPO was maintained to some extent, consistent with a previous report.⁵ In addition, we detected modest NPHS1 expression in undifferentiated hiPSCs, at about 1/1000 the level in human adult podocytes, as reported previously.^11,13 Thus, human podocytes in vivo, which were not employed in previous studies,^11–13 were indispensable as a positive control to precisely verify the quality of induced podocytes.

We also assessed the global transcriptional profile of our induced podocytes compared with hiPSCs, NPCs, immortalized human podocyte cell line cells, and sorted human adult podocytes by RNA-seq analysis. The results confirmed that many of the signature genes, including slit diaphragm–associated genes and core transcriptional factors, were highly expressed and comparable to the levels in human adult podocytes (Table 1). PLA2R1 and THSD7A, recently identified as endogenous antigens responsible for membranous nephropathy,^36,37 were moderately expressed. Meanwhile, some glomerular basement membrane–associated genes were missing in induced podocytes (Table 1). Clustering analyses with previously reported podocyte-specific gene entities^38,39 revealed significant similarity between the induced podocytes and human adult podocytes (Figure 6, Supplemental Figure 9C). These results indicate that our induced podocytes have a global signature gene expression profile resembling that in human podocytes.

Table 1. - Comparisons of podocyte-associated gene expression levels in the RNA-seq analysis

Gene Symbol	Expression Level
	hiPSCs		Cell Line		Induced Podocytes		Human Adult Podocytes
High-expression genes
PODXL	12.30	12.46	4.01	5.90	13.97	13.97	14.49	16.00	15.94
CLIC5	0.00	1.71	0.00	1.07	12.24	12.05	13.11	14.99	14.32
NPHS2	0.00	0.00	0.00	0.00	11.91	11.79	12.47	13.42	13.78
FAT1	10.67	10.55	11.83	11.44	12.20	12.05	12.06	13.55	13.23
NPHS1	3.05	3.20	0.00	0.07	11.54	11.29	10.32	12.36	12.35
ITGB1	10.43	10.13	13.57	13.22	11.54	11.52	12.49	12.49	12.29
SYNPO	3.35	3.85	6.71	7.88	10.28	10.03	9.34	11.66	11.57
MAFB	3.35	3.48	4.63	4.97	12.05	11.99	9.69	11.23	11.41
SULF1	4.40	4.57	3.11	2.07	10.22	9.96	10.64	11.52	11.32
DACH1	2.97	1.39	0.00	0.00	9.95	9.99	9.04	11.95	11.18
WT1	0.65	0.00	0.00	0.00	11.04	10.88	9.54	11.25	10.88
MAGI2	6.79	6.86	5.50	4.71	9.36	9.36	9.43	11.69	10.88
TCF21	0.00	0.00	5.53	4.77	9.79	9.59	9.55	11.42	10.86
FOXC1	0.00	0.00	6.79	6.67	9.35	9.26	9.63	11.17	10.73
LAMA5	3.89	9.11	9.56	10.20	8.72	8.68	9.57	9.76	9.91
COL4A5	7.54	7.54	6.65	6.39	8.83	8.68	8.23	8.95	9.44
BMP7	5.56	5.66	1.69	1.65	8.25	9.14	8.87	9.75	9.28
CD2AP	8.27	8.32	9.68	9.49	8.20	8.16	8.00	10.10	9.13
LMX1B	0.65	0.97	0.00	0.07	7.60	7.26	7.73	8.49	8.50
NPR1	2.23	0.97	0.00	0.00	7.47	7.24	7.08	8.92	8.49
ROBO2	4.87	4.72	1.11	0.07	8.84	9.48	9.30	9.28	7.42
FOXC2	0.00	0.00	2.52	1.70	5.20	7.01	6.58	5.44	7.06
Moderate-expression genes
PLA2R1	2.82	3.48	6.65	6.06	9.35	9.37	12.49	14.49	14.11
NPNT	0.65	2.20	4.28	3.65	10.16	10.08	11.92	13.29	13.24
THSD7A	4.51	4.68	3.43	0.00	8.62	9.09	10.38	12.34	12.39
VEGFA	5.74	6.68	7.78	7.70	9.44	9.49	10.96	12.49	12.26
ITGA3	4.91	5.27	11.22	11.56	8.52	8.61	10.80	11.18	11.77
LAMB2	7.11	7.27	8.88	9.39	7.75	7.68	10.94	10.75	11.61
Low-expression genes
COL4A3	0.00	0.97	2.69	0.00	0.00	0.00	10.67	12.26	11.58
COL4A4	4.93	4.48	5.19	6.28	1.90	0.00	6.86	11.54	11.09
DDN	0.00	0.00	0.00	2.33	2.90	2.33	8.47	10.35	10.88

Numbers represent normalized read counts.

Figure 6.: Induced podocytes have a global signature gene expression profile resembling that in human adult podocytes. Heatmap image and hierarchic clustering of RNA-seq analysis data using the podocyte gene list published by Karaiskos et al. ³⁸ Red, upregulation; blue, downregulation; cell line, conditionally immortalized human podocyte cell line cells (n=2); hiPSCs, undifferentiated hiPSCs (n=2); human adult podocytes, freshly sorted podocytes from human adult kidney (n=3); induced podocytes, selectively induced podocytes from hiPSCs by our protocol (n=2); NPCs, ITGA8⁺/PDGFRa⁻ NPCs induced from hiPSCs (n=2).

Induced Podocytes Display Morphologic and Functional Characteristics of Podocytes

Immunohistochemical analyses showed that the induced podocytes expressed WT1 in their nucleus and NEPHRIN on the basolateral cell membrane (Figure 7A). They also showed colocalization of NEPH1 and PODOCIN with NEPHRIN (Figure 7A) and apicobasal polarity with PODOCALYXIN and NEPHRIN expression at the apical and basolateral regions, respectively (Supplemental Figure 9D). Transmission electron microscopy showed the presence of protrusions at the basolateral domain of the podocytes (Figure 7B), and NEPHRIN was localized on the surface membrane of these protrusions (Figure 7C). Higher magnification observation revealed slit diaphragm–like structures recognized as filamentous bridges between adjacent podocytes and an actin lining structure recognized as an electron-dense material at the cytoplasmic insertion site (Figure 7D). These features resembled those of the slit diaphragm precursors we recently identified in iPSC-derived conventional kidney organoids in vitro.⁴⁰

Figure 7.: Induced podocytes can be used for measurement of slit diaphragm–associated proteins upon drug injury. (A) Immunostaining for characteristic podocyte markers (WT1, NEPHRIN, NEPH1, and PODOCIN) and nuclear DAPI staining in induced podocytes at day 12 (n=3). Scale bars, 20 µm. (B) Transmission electron microscopic images of induced podocytes at day 12 (n=2). The black arrowhead in the right panel indicates cellular protrusions extending from the bottom of adjacent induced podocytes. Scale bars, 2 µm (left panel), 500 nm (right panel). (C) Immunoelectron microscopic image of NEPHRIN expression (blue arrows) on the surface of podocyte protrusions (n=2). Scale bars, 200 nm (left panel), 100 nm (right panel). (D) Highly magnified transmission electron microscopic images of induced podocytes (n=2). The black arrowheads indicate filamentous bridges between adjacent induced podocytes. The black arrows in the right panel indicate an actin lining structure in a cellular protrusion of induced podocytes. Scale bars, 200 nm. (E) Western blots of podocyte-associated proteins in hiPSCs, immortalized human podocyte cell line cells, induced podocytes by our protocol, and hiPSC-derived kidney organoids (n=3). The numbers on the right side represent the mol wt of the proteins. (F) Western blots of induced podocytes at 48 hours after PAN treatment compared with nontreated control cells (n=3). (G) Immunostaining for NEPHRIN and WT1, and nuclear DAPI staining in induced podocytes at 48 hours after PAN treatment (n=3). Scale bars, 20 µm.

Western blotting analyses revealed that the induced podocytes abundantly expressed major slit diaphragm–associated proteins (Figure 7E). They also expressed PLA2R and WT1 at much higher levels than immortalized podocyte cell line cells. The induced podocytes also showed higher expression levels of podocyte-related proteins than hiPSC-derived conventional kidney organoids, indicating an advantage of the pure population system for studies on low-expression proteins over the heterogeneous organoid system.

To further test the biologic functionality, induced podocytes were treated for 48 hours with PAN, used as a model of podocyte injury and nephrotic syndrome.^41–43 Consistent with the phenotype of PAN-induced podocyte injury in vivo,^44,45 western blotting analyses showed reduced expression of slit diaphragm–associated proteins in PAN-treated podocytes (Figure 7F). Reductions in the phosphorylated and mature (upper band) forms of NEPHRIN were also observed. Immunohistochemical analyses revealed decreased intensity of cell surface–located NEPHRIN (Figure 7G), reminiscent of the decrease/translocation of NEPHRIN observed in an in vivo model of PAN-induced injury⁴⁶ and patients with nephrotic syndrome.⁴⁷ Thus, our hiPSC-derived podocytes, which show molecular, morphologic, and functional characteristics of podocytes, will serve as a valuable resource for disease modeling, nephrotoxicity testing.

Discussion

In this study, we have established a highly efficient podocyte induction method from hiPSCs by combining an NPC sorting method and elucidation of the podocyte specification signals.

In the kidney developmental biology field, the signals that dictate the differentiation and patterning of nephron segments from NPCs have been a major focus of interest. Although in vivo genetic studies identified Wnt signaling as the core component of the MET process,^16,17 these studies had limitations, especially with regard to time resolution and the roles of signal intensity and gradient. Meanwhile, in vitro studies have exploited their higher time resolution and more flexible manipulation of growth factor signaling.^24–26 However, most previous in vitro studies investigated the bulk embryonic kidney or MM tissue, thereby hampering elucidation of the cell-autonomous signal requirements for the nephron-patterning process. In this study, we combined the advantages of an in vitro directed culture system and purified NPCs to investigate the podocyte lineage–specification signals.

In the initial step (NPC-to-PA; step 1), optimal strength and duration of Wnt treatment were critical for MET induction in NPCs. Interestingly, among the conditions enabling MET, we found that slight differences in Wnt signal intensity and duration during this phase strongly affected the final endowment of podocytes. This may be partly consistent with a recent study, which led to a proposed model in which the length of NPC exposure to Wnt determines the future proximodistal fate of nephron epithelia, with NPCs exposed for a shorter time to Wnt differentiating into podocytes.⁴⁸

In the second step (PA-to-RV; step 2), we observed that inhibition of Tgf-β/SMAD signaling enriched the RV proximal domain. A previous report revealed Tgf-β1 expression in the UB tip,⁴⁹ which is adjacent to the distal side of the RV. This may indicate that Tgf-β/SMAD signaling is a key regulator with a role in the patterning process in vivo. It will be interesting to perform genetic loss-of-function studies to prove the endogenous role of Tgf-β/SMAD signaling. Inhibition of Wnt signaling and RA treatment were also effective for enrichment of the RV proximal domain in human NPCs. Because RA is known to promote proximal nephron differentiation in zebrafish,⁵⁰ our results suggest a possible role for RA signaling during the formation of nephron patterning in humans.

In the lineage specification step (RV-to-podocytes; step 3), inhibition of Tgf-β/SMAD signaling enhanced podocyte induction, mainly by suppressing proliferation of the PT lineage. Inhibition of Wnt signaling in this step also suppressed the PT lineage in humans, which may be consistent with a previous report that Wnt/β-catenin activity was lowest in the proximal end (future podocytes) and weakly present in the future PT segment of the S-shaped body.²⁶ We also noted an absence of PECs in our induced human podocytes, which is partly consistent with the previous report showing the requirement of Wnt signaling for PEC formation.⁵¹

Another signaling pathway that presumably plays a role in the nephron-patterning process is the Notch pathway,^52–56 although it remains controversial whether this signaling has a role, especially for proximal nephron fate specification.^53,56 Indeed, we did not observe significant changes in podocyte induction efficiency after treatment with a recombinant Notch agonist in our results. However, given that a chemical Notch signal agonist has not been well established, our results may not be strongly conclusive.

We observed some differences between murine embryonic NPCs and hiPSC-derived NPCs. For example, even in the optimized condition, murine embryonic NPCs could not differentiate into podocytes at >50%. Recent studies have shown that in vivo NPCs are heterogeneous,^48,57 whereas our hiPSC-derived NPCs were considered to be rather homogeneous because they do not receive prepatterning signals from UB or stromal progenitors. Alternatively, it is possible that the NPCs induced from hiPSCs in our protocol may be somewhat biased toward the podocyte lineage compared with those in vivo. Interestingly, a recent report revealed different developmental tendencies of nephron segments among kidney organoid induction protocols.⁵⁸ Although species differences could be responsible for such differences, developmental stage–matched gene profiling of mouse and human embryonic NPCs will be indispensable to address these issues.

We verified the quality of our induced podocytes by extensive gene expression profiling compared with freshly isolated human podocytes, morphologic analysis, and drug sensitivity. Although our podocytes showed proper localization of slit diaphragm–associated proteins on the cell membrane, the expression patterning of NEPHRIN was not sufficiently mature to exhibit a distribution along the basement membrane, as seen in adult podocytes. Electron microscopy showed that the induced podocytes lacked a typical interdigitated structure, although slit diaphragm–like filamentous bridges were detected between adjacent podocytes. RNA-seq analyses revealed low expression levels of COL4A3 and COL4A4 in the induced podocytes compared with human adult podocytes. This insufficient maturation as well as the long-term culture limitation of our condition may suggest a requirement for blood flow and interactions between podocytes and endothelial cells.¹⁰ Thus, it will be desirable to develop an in vitro podocyte maturation/long-term maintenance system in the future. When we tried an extended culture of the induced podocytes in a two-dimensional setting, the cells formed cellular processes with actin bundles and maintained WT1 expression for 5 days (Supplemental Figure 10, A and B). However, NEPHRIN expression at the membrane disappeared by day 2 (Supplemental Figure 10C). These results indicate that the three-dimensional culture system is advantageous for maintaining the podocyte characteristics.

In summary, our newly developed robust podocyte induction system will facilitate global accessibility to human podocytes, and serve as a valuable resource for disease modeling and nephrotoxicity testing.

Disclosures

None.