Generation of kidney tubular organoids from human pluripotent stem cells

Recent advances in stem cell research have resulted in methods to generate kidney organoids from human pluripotent stem cells (hPSCs), which contain cells of multiple lineages including nephron epithelial cells. Methods to purify specific types of cells from differentiated hPSCs, however, have not been established well. For bioengineering, cell transplantation, and disease modeling, it would be useful to establish those methods to obtain pure populations of specific types of kidney cells. Here, we report a simple two-step differentiation protocol to generate kidney tubular organoids from hPSCs with direct purification of KSP (kidney specific protein)-positive cells using anti-KSP antibody. We first differentiated hPSCs into mesoderm cells using a glycogen synthase kinase-3β inhibitor for 3 days, then cultured cells in renal epithelial growth medium to induce KSP+ cells. We purified KSP+ cells using flow cytometry with anti-KSP antibody, which exhibited characteristics of all segments of kidney tubular cells and cultured KSP+ cells in 3D Matrigel, which formed tubular organoids in vitro. The formation of tubular organoids by KSP+ cells induced the acquisition of functional kidney tubules. KSP+ cells also allowed for the generation of chimeric kidney cultures in which human cells self-assembled into 3D tubular structures in combination with mouse embryonic kidney cells.

Scientific RepoRts | 6:38353 | DOI: 10.1038/srep38353 To obtain pure population of kidney tubular cells, we focused on one of cadherins, Kidney-specific protein (KSP; cadherin 16) which is exclusively expressed in kidney lineage cells including ureteric buds, developing nephrons, mesonephric tubules, Bowman's capsules, proximal tubules, loops of Henle, and distal and collecting tubules 26 . We previously generated a monoclonal antibody against the extracellular domain of KSP, which enabled to purify KSP+ cells by flow cytometry 13 . Building on these prior reports, the aim of this study was to establish a rapid and simple differentiation protocol of human embryonic stem cells (hESCs) into kidney tubular cells with purification of KSP+ cells and to generate tubular organoids in 3D culture systems. We employed a feeder-free monolayer culture system to differentiate hESCs into KSP+ cells, and established a simple two-step differentiation protocol using a glycogen synthase kinase-3β (GSK-3β ) inhibitor (BIO; 6-bromoindirubin-3′-oxime) in order to obtain sufficient number of KSP+ cells for flow sorting. We also showed that these KSP+ cells derived from hESCs exhibited characteristics of all segments of kidney tubular cell and acquired more functional characteristic when they formed tubular organoids in 3D Matrigel. In addition, we demonstrated that the KSP+ cells form tubular structures in chimeric culture with dissociated mouse metanephric kidneys, supporting our hypothesis that purification of KSP+ cells from differentiated hPSCs provides pure population of kidney tubular cells.

Differentiation of hESCs into KSP+ cells.
We developed a simple two-step protocol to differentiate KhES-1 hESCs into KSP+ cells within 10 days. Given the facts that embryoid body (EB) formation might be a time-consuming differentiation process and also hinder the effective delivery of differentiation-inducing signals to all cells 27 , we employed a two-dimensional feeder-free monolayer culture system. In vivo, the kidneys develop from the intermediate mesoderm (IM) which arises from the primitive streak 28 . Hence, KhES-1 cells were first differentiated into the cells of the primitive streak with a GSK-3β inhibitor (BIO, a WNT activator) (BIO(+ ); BIOtreated cultures) for 3 days, or without BIO (mock; BIO-untreated cultures) as a negative control (Fig. 1a) 14,29 . Accordingly, the expression of the primitive streak marker BRACHURY 30 measured by real time quantitative reverse transcription polymerase chain reaction (qRT-PCR) was transiently increased, whereas the expression of the pluripotency marker OCT3/4 was down-regulated with BIO treatment (Fig. 1b). Subsequently, those cells were stochastically differentiated with a commercially available renal epithelial growth medium (REGM) for 7 days, which consists of 0.5% fetal bovine serum (FBS), recombinant human epidermal growth factor, insulin, hydrocortisone, epinephrine, triiodothyronine and transferrin 15,31,32 . We found increased expression of the IM markers, PAX2 and OSR1 from day 4 to 6 of differentiation by qRT-PCR (Fig. 1b). The expression of WT1 was upregulated from day 8 to 10, which is consistent with a previous report showing that WT-1 is known to be required for kidney development 33 and restricted to podocytes at later stage 34 (Fig. 1b). The expression of KSP, which is exclusively expressed in kidney lineage cells and occurs during normal development at later stages of kidney tubular differentiation 26 , was significantly greater on day 10 (Fig. 1b). To determine whether additional growth factors to REGM further enhance KSP expression, we tested activin A (10 ng/ml), hepatocyte growth factor (HGF; 10 ng/ml) or insulin-like growth factor 1 (IGF-1; 10 ng/ml) from day 3 to day 10 of differentiation, since our previous study showed that HGF and IGF-1 increased KSP expression in mouse ESCs differentiated with DMEM supplemented with activin A 10 ng/ml 13 . Any of activing A, HGF or IGF-1, however, did not further increase KSP expression compared to REGM alone on day 10 of differentiation ( Fig. 1c), suggesting that REGM itself was optimized for differentiation of hESCs into kidney tubular cells. We also evaluated other lineage markers by qRT-PCR in hESCs on day 10 of differentiation; PAX6 (an endoderm/ectoderm marker) or NKX2-5 (a cardiac mesoderm marker) (Fig. 1d). Those two genes were not upregulated by our differentiation protocol with BIO and REGM while nephron progenitor markers including OSR1, WT1, GDNF, CITED1, HOXD11 ( Fig. 1d and Fig. S1) and kidney tubular markers including KSP, GTT, AQP1, CD13, ATP1B, MEGALIN, SGLT1, UROMODULIN, AQP 2 were significantly upregulated (Fig. 1e), suggesting that hESCs were differentiated into kidney lineage cells with the two-step differentiation protocol.
Collectively, these qRT-PCR results indicated that the first specification into the primitive streak cells with BIO for 3 days subsequently induced the intermediate mesoderm cells and kidney lineage cells which represent a heterogeneous population containing cells with characteristics of intermediate mesoderm, metanephric mesenchyme, developing nephron, and glomerular podocytes (WT1) and tubular cells (KSP) with stochastic differentiation within 10 days of the differentiation beginning.

Purification of KSP+ cells from differentiated hESCs.
To obtain pure population of kidney tubular cells from differentiated hESCs with our two-step protocol, we used anti-KSP antibody for flow sorting, which was designed to recognize the extracellular domain of KSP 13 . To evaluate the specificity and cross-reactivity of anti-KSP antibody to human kidney tubular cells, we performed a flow cytometric analysis and immunocytochemistry using a human embryonic kidney (HEK) 293 cell line as a positive control and human aortic smooth muscle cells (HASMCs) as a negative control as indicated by qRT-PCR and western blotting (Fig. 2a,b). Western blotting with anti-KSP antibody detected several bands (Fig. 2b, left). However, a band of 75-100 kDa was not reacted with the secondary antibody alone, suggesting that this band is specific for KSP (Fig. 2b, right). The flow cytometric analysis detected KSP+ cells in HEK293 cells but not in human aortic smooth muscle cells (HASMCs) (Fig. 2c). Immunocytochemistry showed a positive staining on the cell surface of HEK293, which is consistent with protein localization of KSP in kidneys in vivo (Fig. 2d) 18,35 . In addition, we performed immunohistochemistry of human kidney samples using the anti-KSP antibody, anti-AQP1 antibody, and anti-AQP2 antibody ( Fig. 2e-g). KSP+ cells were co-localized with AQP1+ cells which represent proximal tubules (Fig. 2f), AQP2+ cells which represent collecting ducts (Fig. 2g), which is in accordance with the results previously obtained in mouse neonatal kidney tissues and human kidney 13,36 . These data demonstrated that our original anti-KSP antibody detected human KSP as well as mouse KSP 13 To obtain KSP+ cells from stochastically differentiated hESCs, we evaluated the protein expression of KSP in differentiated hESCs using the anti-KSP antibody. Immunohistochemistry showed that KSP was expressed on the surface of cells in a small population of differentiated hESCs along the periphery of cell clusters on day 10 of the differentiation with BIO and REGM (Fig. 3b,c), whereas KSP staining was not detected in hESCs in BIO-untreated (mock) cells (Fig. 3a), which was consistent with the results obtained by qRT-PCR (Fig. 1b). Western blotting also detected positive bands in samples treated with BIO, but not in samples treated with mock in 2 independent experiments (Fig. 3d).
To exclude the possibility of nonspecific labelling of the anti-KSP antibody, we performed flow cytometry using anti-TRA1-60 (tumour rejection antigen 1-60) antibody in combination with the anti-KSP antibody in hESCs on day 3 and 10 of the differentiation. TRA1-60 is one of pluripotent markers of hESCs and hiPSCs 37,38 ; therefore, co-expression of TRA1-60 and KSP were very unlikely. On day 3 of the differentiation, KSP+ cells were not detected as predicted by qRT-PCR (Fig. 1b) while TRA1-60+ cells were detected in both mock (85%) and BIO (50%) treated cells (Fig. 3e). On day 10 of the differentiation, KSP+ cells were increased to 4.08 ± 0.514% of differentiated cells with BIO treatment (Fig. 3f), and, KSP and TRA1-60 double positive cells were not detected (< 1%). In addition, mock-treated cells were negative for KSP while 25% of cells were positive for TRA1-60

Characterization of KSP+ cells. We confirmed the presence of sufficient KSP+ cells in hESCs by flow
cytometry on day 10 of the differentiation with our two-step protocol (Fig. 3f). To characterize KSP+ cells purified from differentiated hESCs with our two-step differentiation protocol (Fig. 4a), we purified KSP+ cells from differentiated hESCs with flow cytometry using the KSP-antibody on day 10 of the differentiation and evaluated gene expression of kidney tubular markers by qRT-PCR. Gene expression in KSP+ cells were compared to that in whole differentiated hESCs on day 10 of the differentiation (BIO+ , unsorted, Fig. 4b), which were not subjected to flow sorting, in order to evaluate how efficiently flow sorting with the KSP-antibody improved purity of kidney tubular cells. KSP+ cells purified by flow sorting showed significantly higher expression of KSP than unsorted hESCs by qRT-PCR (Fig. 4b), demonstrating that flow sorting with the KSP-antibody significantly improved purity of KSP+ tubular cells. qRT-PCR also showed that MEGALIN, AQP1, UROMODULIN and AQP2 were higher in KSP+ cells than in unsorted hESCs (Fig. 4b).
In order to further characterize purified KSP+ cells, we compared gene expressions of kidney tubular markers in KSP+ cells to those in primary renal proximal tubular epithelial cells (RPTECs). KSP+ cells showed a comparable expression of AQP1 as RPTECs, whereas the expressions of UROMODULIN, SLC12A3 and AQP2 were higher in KSP+ cells than RPTECs (Fig. 4c). These data suggest that KSP+ cells have characteristics of proximal tubular, loops of Henle, distal tubular and collecting duct cells [39][40][41][42] . We then evaluated the functional characteristics of purified KSP+ cells by mRNA expression and activity of γ -glutamyl transferase (GGT), which is known to be expressed in the proximal tubule 15,43 . The modest increase of GGT in purified KSP+ cells compared with unsorted hESCs on day 10 of the differentiation (1.45 ± 0.25-fold) is most likely due to the fact that the basal level of GGT mRNA level in unsorted hESCs are already high (Fig. 4b), since GGT is expressed in various tissues 43 . In order to analyse the GGT activity, which reflects the functional characteristic of proximal tubular cells 15 , we measured p-nitroaniline which is released by the action of GGT 15 . KSP+ cells as well as RPTECs demonstrated GGT activity (Fig. 4e); however, the GGT activity in the RPTECs was about six times as high as that in KSP+ cells, consistent with the relative gene expression level of GGT in KSP+ cells and RPTECs (Fig. 4d). These results suggest that KSP+ cells are a heterogeneous population containing not only proximal tubules but also other segment cells, though another possible explanation is that KSP+ cells are functionally less mature than RPTECs. In addition, as our previous study showed that tubular formation in Matrigel with Wnt4 further promoted differentiation of KSP+ cells derived from mESCs into more functional proximal tubules and collecting duct cells 13 , we hypothesized that same approaches would further facilitate functional maturation of KSP+ cells derived from hESCs. First, we investigated the capacity of KSP+ cells for tubular organoids formation in vitro by transferring onto Matrigel (Fig. 4f). KSP+ cells formed small and sparse tubular structures in Matrigel alone (Fig. 4g). Cross-section slides of those tubular structures showed cilia-like structures at the luminal surface (Fig. 4h). Consistent with this, staining of the ciliary marker, acetylated TUBULIN was observed 18 , indicating that kidney tubular organoids generated from purified KSP+ cells have cellular polarity ( Fig. 4i). Furthermore, we tested a co-culture with NIH3T3-Wnt4 cells that constitutively express Wnt4, which is essential for tubulogenesis 13,44 . After 24 h of culture, the KSP+ cells formed abundant tubular organoids that co-expressed MEGALIN, AQP1 or AQP2 (Fig. 4j-m), demonstrating that KSP+ cells have characteristics of kidney tubular cells. In contrast, KSP negative cells just formed petal-like structures that were negative for KSP and MEGALIN (Fig. S2). Accordingly, the gene expression of KSP, AQP1, and AQP2 was increased upon Wnt4 stimulation after 48 h of co-culture (Fig. 4n), suggesting the induction and maturation of proximal tubules and collecting ducts. These findings indicate that KSP+ cells possess the functional characteristics of kidney tubular cells and formed tubular organodis in 3D Matrigel.

Formation of tubular structures in chimeric culture with dissociated mouse metanephric kidneys.
To confirm our results that KSP+ cells have characteristics of kidney tubular cells, we examined functional integration of KSP+ cells into 3D renal structures during the re-aggregation of embryonic mouse metanephric cells, as described previously 14,17,27,45 . In this assay, dissociated cells from wild-type E11.5 mouse embryonic kidneys, which can recapitulate nephrogenesis 27,45 were re-aggregated with KSP+ cells purified from differentiated hESCs on day 10 of the differentiation (Fig. 5a). Since a previous study demonstrated that undifferentiated hPSCs disrupted the formation of 3D structures by murine cells 27 , we excluded undifferentiated hESCs with anti-TRA1-60 antibody. After 7 days of chimeric culture in REGM, we performed immunohistochemistry for human mitochondria in order to identify hESC-derived cells (Fig. 5b,c). Human mitochondria positive cells formed 3D tubular structures positive for the proximal tubule marker, lotus tetragonolobus lectin (LTL) (Fig. 5b,c) in combination with mouse embryonic kidney cells. These data demonstrate that KSP+ cells are committed to kidney tubular cells.

Discussion
We demonstrated a rapid and simple differentiation protocol of hESCs into kidney tubular cells, direct purification of kidney tubular like cells with the anti-KSP antibody and generation of tubular organoids in 3D culture systems. We have previously shown that mESC-derived KSP+ cells exhibit characteristics of kidney tubular phenotypes 12,13 and have suggested that KSP can serve as a kidney lineage marker during mESC differentiation. Here, we successfully generated the KSP+ cell population from hESCs with stochastic differentiation after induction of primitive streak in approximately half the time compared to that required for the differentiation of mouse ES cells 12,13 . In addition, KSP+ cells derived from hESCs exhibited the functional characteristics of kidney tubular cells, formation of tubular organoids in 3D Matrigel, and the generation of chimeric kidney cultures.
In the present study, the differentiation of hESCs was achieved by a monolayer culture method that did not require feeders, high serum concentration, or EB formation. Previous studies have reported that the signalling induced by extracellular matrix components plays an important role in directing early kidney differentiation, suggesting that this process might be promoted in a monolayer culture 46 . On the other hand, EB formation and feeder cell contamination might hinder the initiation of such signalling 14,27 , and certain growth factors present in serum have been shown to antagonize hESC differentiation 47 .
The inhibition of GSK-3β , which is a serine/threonine kinase, leads to the activation of canonical Wnt/β -catenin signalling in ESCs, resulting in mesodermal differentiation 14,45,[48][49][50][51][52] . Therefore, during the first step of our protocol, KhES-1 cells were first differentiated into the cells of primitive streak cells with a GSK-3β inhibitor (BIO). As shown in this study, treatment of BIO induced BRACHYURY expression (Fig. 1b), indicating the induction of the primitive streak cells. Notably, previous studies suggested that the Brachyury-positive primitive streak population is capable of differentiating into the Osr1-positive IM cells, the origin of kidney lineage 16,28 . The induction of primitive streak cells from hPSCs is triggered by the activation of Wnt/β -catenin signalling, and subsequent differentiation into IM cells requires downregulation of Wnt activity 16,49 . Therefore, in the second step of our protocol, we removed the Wnt/β -catenin stimulation and provided culture conditions suitable for the efficient generation of IM cells and kidney tubular cells 15,31,32 . In our step-wise procedure, the treatment of KhES-1 cells with BIO for 3 d induced Brachyury expression resulting in a decrease of TRA1-60+ cells (Figs 1b  and 3e), and subsequent culture in REGM induced the differentiation into KSP+ cells (Figs 1b and 3f). In contrast, BIO-untreated cultures did not induce KSP+ /TRA1-60-negative cells on day 10 of differentiation (Fig. 3f). These results demonstrated that application of the GSK-3β inhibitor in conjunction with culture in REGM can efficiently promote the differentiation of hESCs into KSP+ kidney tubular cells.
Under these differentiation conditions, the gene expression of not only KSP but also WT1, which is known to be expressed in intermediate mesoderm, metanephric mesenchyme, developing nephron, and glomerular podocytes at the late stage of kidney differentiation 34 , was upregulated on day 8~10 of differentiation. Given that the cell population after differentiation was heterogeneous, the KSP+ /TRA1-60-negative cells were purified from differentiated hESCs using flow cytometry (Fig. 3f) to obtain pure population of kidney tubular cells.
Next, we evaluated the functional characteristics of purified KSP+ cells. GGT is expressed in the proximal tubule 15,43 and produces p-nitroaniline. Thus, GGT activity reflects the functional characteristic of kidney tubular cells 15 and KSP+ cells demonstrated GGT activity (Fig. 4e). KSP+ cells also formed tubular organoids that contained MEGALIN+ , AQP1+ or AQP2+ tubular cells in 3D Matrigel (Fig. 4j-m). By forming tubular organoids, KSP+ cells acquired more functional characteristic with higher expression of functional tubular proteins including AQP1 and AQP2 (Fig. 4n), which represent proximal tubules and collecting ducts respectively. Previously, Narayanan et al. and Kandasamy et al. showed the induction of proximal tubule-like cells with the high purity, by differentiation of hPSCs with REGM supplemented with bone morphogenetic protein (BMP) 2 and 7 and formed tubule-like structures in a two-dimensional culture 15,53 . In contrast to their studies, here we were able to purify KSP+ cells that exhibited characteristics of proximal tubular, loops of Henle, distal tubular and collecting duct cells, which is consistent with previous studies showing KSP expression in all segment of tubules from proximal tubule to collecting ducts 26 . KSP+ cells also generated tubular organoids that contained MEGALIN+ , AQP1+ or AQP2+ tubular cells from KSP+ cells in 3D Matrigel.
In addition, KSP+ /TRA1-60-negative cells formed tubular structures in chimeric kidney cultures with dissociated mouse metanephric kidneys in organ culture experiments, indicating that KSP+ cells purified from differentiated hESCs provides a population of definitive kidney tubular lineage cells.

Conclusions
In this study, we have demonstrated a method to purify human kidney tubular cells from hESCs with the anti-KSP antibody after differentiation with a simple two-step differentiation protocol. This approach is useful for generating cells of definitive kidney tubular lineage cells that might be usable as a source for generation of bioengineered kidneys, modeling kidney diseases, drug screening for new therapeutic approaches for kidney diseases, nephrotoxicity assays, and ultimately kidney regenerative medicine using hPSCs.
Flow cytometry. Cells were dissociated with trypsin/EDTA (Invitrogen) and treated with 0.2% collagenase type IV (Wako) for 15 min at 37 °C. The cells were mechanically dissociated by pipetting, then filtered through Scientific RepoRts | 6:38353 | DOI: 10.1038/srep38353 a 70-μ m cell strainer (Falcon, Corning, NY, USA) and maintained in DMEM containing 10% FBS for 1 h before treatment with 10 μ g/ml Fc blocker (#422301; Bio Legend, San Diego, CA, USA) for 20 min on ice. The cells were incubated with 0.08 mg/ml biotinylated anti-KSP antibody (originally produced using hybridomas) 13 for 30 min, 0.01 mg/ml streptavidin-conjugated Alexa Fluor 647 (Invitrogen) for 20 min on ice, and finally with a fluorescein isothiocyanate (FITC)-conjugated monoclonal anti-TRA1-60 antibody (#560173; 1:200; BD Biosciences). Just prior to analysis, the cells were transferred to a 5 ml polypropylene tube (Falcon, Corning) to which 1 μ g/ml propidium iodide (Sigma-Aldrich) was added. Cells were counted and sorted using a flow cytometer (MoFlo XDP; Beckman Coulter, Brea, CA, USA) according to the manufacturer's instructions. The KSP expression threshold was set relative to the signal intensity of negative control samples incubated without antibody or with the secondary antibody only.
Animal experiments. All mice experiments were approved by the Institutional Animal Center at Keio University School of Medicine in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Statistical analysis. The results are presented as the means ± S.E.M. Comparisons between groups were performed using a two-tailed Student's t test. P < 0.05 was considered significant.