PAX2 is dispensable for in vitro nephron formation from human induced pluripotent stem cells

The kidney is formed by reciprocal interactions between the nephron progenitor and the ureteric bud, the former of which gives rise to the epithelia of nephrons consisting of glomeruli and renal tubules. The transcription factor PAX2 is essential for this mesenchymal-to-epithelial transition of nephron progenitors, as well as ureteric bud lineage development, in mice. PAX2 mutations in humans cause renal coloboma syndrome. We previously reported the induction of nephron progenitors and three-dimensional nephron structures from human induced pluripotent stem (iPS) cells. Here we generate iPS cells lacking PAX2, and address the role of PAX2 in our in vitro induction protocol. While PAX2-null human nephron progenitors were properly formed, they unexpectedly became epithelialised to form glomeruli and renal tubules. However, the mutant glomerular parietal epithelial cells failed to transit to the squamous morphology, retaining the shape and markers of columnar epithelia. Therefore, PAX2 is dispensable for mesenchymal-to-epithelial transition of nephron progenitors, but is required for morphological development of glomerular parietal epithelial cells, during nephron formation from human iPS cells in vitro.

renal vesicles, S-shaped bodies, and distal renal tubules. Pax2 deletion in mice leads to kidney agenesis, as well as defects in the genital tract, optic nerve, retina, inner ear, and midbrain 9,10 . When Pax2-deficient metanephric mesenchyme is isolated and co-cultured with spinal cord, no tubulogenesis or nephron formation occurs 11 , indicating an indispensable role of mouse Pax2 in mesenchymal-to-epithelial transition of nephron progenitors. Pax2 is also expressed in the ureteric bud, another precursor population of the kidney, and is important for its development. During the initial phase of kidney development, the nephric duct (Wolffian duct) is formed and elongates caudally, followed by sprouting out of the ureteric bud. Pax2 deficiency causes impaired epithelial integrity of the nephric duct, thereby affecting proper ureteric budding 12 , which is likely to contribute to the kidney agenesis observed in the absence of Pax2.
In humans, PAX2 frameshift mutations cause renal coloboma syndrome, an autosomal dominant disorder manifesting as kidney hypoplasia and various optic nerve defects, accompanied by auditory and central nervous abnormalities in some cases 13,14 . Heterozygous Pax2 1Neu mice, which harbour a frameshift mutation found in some human patients, exhibit similar phenotypes 15 , and genetically-engineered Pax2 heterozygous mice show kidney hypoplasia 9 . These data indicate that the renal coloboma phenotype is caused by PAX2 haploinsufficiency. This is likely due, in large part, to a role of PAX2 in the ureteric bud, because the kidney size in heterozygous Pax2 1Neu mice is restored by apoptosis suppression in the ureteric bud 16 . However, the precise role and expression patterns of PAX2 in human kidney development are not fully understood, despite the accumulated findings in mice.
We and others previously reported the induction of nephron progenitors from human induced pluripotent stem (iPS) cells [17][18][19][20] . In our protocol, the induced progenitor aggregates form three-dimensional glomeruli and renal tubules ex vivo upon spinal cord recombination 17,18 , thus mimicking the situation in murine progenitors. Thus, our protocol should serve as a useful tool to analyze the role of PAX2 in differentiation of nephron progenitors in humans. For this purpose, we disrupted the PAX2 gene in human iPS cells by homologous recombination using transcription activator-like effector nucleases (TALENs) 21 , and unexpectedly found a dispensable role of PAX2 under our nephron differentiation conditions.

Results
Generation of PAX2-deficient human iPS cells. First, we confirmed PAX2 expression during in vitro nephrogenesis from human iPS cells. PAX2 was detected in the renal vesicles, S-shaped bodies, and distal renal tubules (Fig. 1A), similar to the expression pattern in mouse embryonic kidneys. To examine the role of PAX2 in human kidney development, we inserted a gene encoding green fluorescent protein (TurboGFP) into the PAX2 locus of human iPS cells by homologous recombination (Fig. 1B). For this, we constructed a pair of plasmids that expressed TALENs targeting sequences in close proximity to the PAX2 start codon, and introduced these plasmids, along with a targeting vector expressing the homology arms, into human iPS cells. We successfully obtained heterozygous and homozygous GFP knock-in clones (−/−), determined by PCR and Southern blotting (Fig. 1C,D). Subsequently, we induced the correctly targeted clones toward nephron progenitors, and confirmed the reduction and absence of PAX2 protein in the heterozygous and homozygous clones, respectively (Fig. 1E).
Next, we co-cultured the induced nephron progenitors with murine spinal cord, as a potent inducer of nephrogenesis, based on our previous reports 17,18 . At day 9 of co-culture, round-shaped glomeruli and renal tubules with a clear lumen were detected in all three genotypes histologically ( Fig. 2A). While PAX2, but not GFP, was detected in wild-type renal tubules, heterozygous clones exhibited weak but faithful GFP expression relative to the expression of endogenous PAX2 (Fig. 2B). In the homozygous mutant clones, a clear GFP signal, but no PAX2 signal, was detected in the renal tubules, indicating successful deletion of PAX2 with compensatory expression of GFP. The GFP signal at the periphery of the glomeruli may reflect PAX2 expression in the glomerular parietal epithelial cells. The successful nephron formation in the absence of human PAX2 was unexpected, because no tubulogenesis occurred when Pax2-deficient mouse embryonic metanephric mesenchyme (containing nephron progenitors) was co-cultured with spinal cord 11 .
During the course of the gene targeting, we obtained one cell line that consisted of both wild-type and homozygous mutant cells, probably through picking up two adjacent colonies. The kidney tissues induced from this mixed colony consisted of either PAX2 + or GFP + epithelia (Fig. 2C). These findings indicate that cells lacking PAX2 can differentiate into nephron epithelia even in a competitive environment with wild-type cells, again supporting a dispensable role of human PAX2 in our in vitro differentiation protocol.

Nephron progenitors are properly formed from PAX2-null iPS cells. Nephron formation in vitro
can be divided into two steps: nephron progenitor induction from iPS cells and subsequent nephron formation from the progenitors. To analyse these two steps precisely, we developed a method to isolate the nephron progenitors, thereby enabling examination of both the induction efficiency of nephron progenitors and the nephron formation from the purified progenitors from each clone. We previously reported that nephron progenitors in mouse embryos can be sorted by the presence of integrin α8 (ITGA8) and absence of PDGF receptor α (PDGFRA) 17 . Meanwhile, O'Brien et al. 22 showed that human embryonic nephron progenitors are also ITGA8-positive. Therefore, we analysed the nephron progenitors induced from one of our wild-type clones (+/+) by FACS analysis, and found that 15.4% of the cells constituted the ITGA8 + /PDGFRA − population (Fig. 3A). This population expressed higher levels of multiple nephron progenitor markers than the ITGA8 − /PDGFRA − population, including OSR1, WT1, SIX1, PAX2, and SALL1 (Fig. 3B). GDNF was also enriched, although the difference was not statistically significant. Importantly, when co-cultured with murine spinal cord, the ITGA8 + /PDGFRA − population exhibited robust tubulogenesis, in contrast to the ITGA8 − /PDGFRA − population (Fig. 3C). Thus, nephron progenitors were enriched in the ITGA8 + /PDGFRA − fraction. However, the observation that SIX2 was not enriched in the ITGA8 + /PDGFRA − fraction indicated that this population may not consist solely of nephron progenitors. We then examined another wild-type clone (+/+) and two PAX2-null clones (−/−), and observed variable percentages of the ITGA8 + /PDGFRA − fraction depending on the experiments (  variations, which were not apparent when using the parental iPS cells 17 , may partly result from the extensive passages required to establish the genetically manipulated clones. Nevertheless, the sorted fractions expressed similar levels of nephron progenitor markers, with the exception of PAX2 (Fig. 3D). Although GDNF was reported to be a  Sorted PAX2-null human nephron progenitors undergo mesenchymal-to-epithelial transition to form glomeruli and renal tubules. The isolation of nephron progenitor fractions enabled precise comparisons of their competence. The same numbers of ITGA8 + /PDGFRA − nephron progenitors from wild-type (+/+) and PAX2-null (−/−) clones were co-cultured with spinal cord, and their development into nephron structures was examined histologically. At 3 days of co-culture, renal vesicles, marked with Lim Homeobox 1 (LHX1), were formed both in the wild-type and mutant clones (Fig. 4A). At 6 days of co-culture, CADHERIN6 + proximal renal tubules and E-CADHERIN + distal tubules were observed from both genotypes (Fig. 4B). These data indicate that mesenchymal-to-epithelial transition, as well as establishment of proximo-distal polarity, was not impaired in the absence of human PAX2. In addition, NCAM + S-shaped bodies, of which the proximal regions are precursors of glomeruli, were observed in both the control and mutant clones. At 9 days of co-culture, proximal (Lotus tetragonolobus lectin: LTL + ) and distal (E-CADHERIN + ) renal tubules were formed (Fig. 4C). In addition, the glomeruli showed WT1 + podocytes with basally-distributed NEPHRIN expression (Fig. 4C). Therefore, nephron formation was comparable between the wild-type and PAX2-mutant clones, even when started from the same numbers of nephron progenitors, clearly indicating the dispensable role of PAX2 in mesenchymal-to-epithelial transition of nephron progenitors in the differentiation process from human iPS cells.
Human PAX2 is required for morphological development of glomerular parietal epithelial cells. Despite the apparently normal glomerulogenesis in the absence of PAX2, we noticed that the mutant glomerular parietal epithelial cells (epithelia of Bowman's capsule) were not flattened and remained in the columnar shape (Fig. 5A). The thickness of the epithelial layers was significantly larger in the mutant clones compared with the wild-type clones (Fig. 5B). In addition, CADHERIN6, which showed restricted expression to columnar proximal renal tubules in the wild-type clone, was detected in the basolateral domains of the mutant glomerular parietal epithelial cells (Fig. 5C). Furthermore, atypical protein kinase C (aPKC), which is normally expressed in the apical domains of columnar epithelia, was detected in the mutant glomerular parietal cells (Fig. 5D). Therefore, PAX2-deficient glomerular parietal epithelial cells failed to undergo columnar-to-squamous transition, and retained columnar marker expression, suggesting a requirement for human PAX2 in the morphological development of this cell lineage. mesenchymal-to-epithelial transition under PAX2 deficiency may be explained through compensation by other redundant genes. In mice, Pax8, which belongs to the same Pax paralogue group as Pax2, functions redundantly with Pax2 in kidney development, whereas Pax5 is not expressed in the kidney. While Pax8 deletion alone exhibits no apparent phenotypes, Pax2/Pax8 deficiency causes defects in nephric duct formation, which occurs much earlier than metanephros formation 23,24 . During metanephros formation, Pax8 also functions redundantly with Pax2, because Pax2/Pax8 double-heterozygous mice exhibit more severe kidney hypoplasia than Pax2 heterozygous mice 23,25 . As determined by in situ hybridisation analyses, mouse Pax8 is expressed in renal vesicles and S-shaped bodies, but not in the metanephric mesenchyme (nephron progenitors) where Pax2 exists 25 . We found that one of the commercially available anti-PAX8 antibodies showed consistent results with the reported patterns obtained by in situ hybridisation 25 . By using this anti-PAX8 antibody, we confirmed that PAX2, but not PAX8, was expressed in the murine metanephric mesenchyme and ureteric buds, while both PAX2 and PAX8 were expressed in the renal vesicles ( Fig. 6A and Supplementary Fig. S1). However, PAX8 expression was confined to the distal part of the renal vesicles, while PAX2 was expressed in all regions. As development proceeded, PAX2 was expressed in the middle and distal parts of the S-shaped bodies ( Fig. 6A and Supplementary Fig. S1). In addition, although weak, PAX2 was more abundantly expressed in the precursors of glomerular parietal epithelial cells than in those of podocytes. In contrast, PAX8 was expressed in both lineages, and weakly in the middle part of the S-shaped bodies. These expression patterns were almost conserved in human embryonic kidneys (Fig. 6B). PAX2 was detected in the metanephric mesenchyme (nephron progenitors) and the entire regions of the renal vesicles, as well as in the ureteric bud tips. Weak PAX2 expression was also observed in the precursors of glomerular parietal cells. PAX8 was expressed in the distal region of renal vesicles, middle part of S-shaped bodies, and precursors of glomerular podocytes and parietal cells.
We further examined the expression pattern of PAX8 during nephrogenesis from the wild-type and PAX2-null human iPS cells. Our qPCR analysis detected PAX8 expression in the ITGA8 + /PDGFRA − progenitor fraction In the latter case, GFP staining was used to identify the cells that should have expressed PAX2. Note that PAX8 is expressed in most of the GFP + epithelia, while PAX2 + /PAX8 − vesicles were detected in the wild-type epithelia (asterisk). (D) Expression of PAX2 and PAX8 in S-shaped bodies induced from the wild-type and PAX2-null iPS cells. White arrowheads: renal vesicles; arrows: precursors of glomerular parietal epithelia; black arrowheads: precursors of podocytes. np: nephron progenitors; ub: ureteric bud. Scale bars: 50 µm. (Fig. 3B). However, its expression level was not high, given the absence of PAX8 immunostaining in the nephron progenitor region (Fig. 6B). Moreover, PAX8 expression in PAX2-null progenitors was unaltered (Fig. 3D). At the renal vesicle stage, PAX2 was expressed in most of the wild-type epithelia, while PAX8 was expressed in a subset of PAX2 + epithelia (Fig. 6C). There were PAX2-single-positive renal vesicles, and even in PAX2 + /PAX8 + vesicles, PAX8 high cells were clustered at one end, consistent with the findings in vivo. In the PAX2 mutant clones, we stained GFP to detect the cells with PAX2 promoter activity, and found that it marked most of the epithelia. Interestingly, PAX8 was detected in most of the PAX2 + epithelia (Fig. 6C), suggesting compensatory expression of PAX8 in the absence of PAX2. However, at the S-shaped body stage, PAX2, and more abundantly PAX8, were expressed in the precursors of glomerular podocytes and parietal cells in the wild-type epithelia (Fig. 6D), and there was no compensatory PAX8 expression in the mutant epithelia. Thus, the compensatory expression of PAX8 upon PAX2 deletion in the renal vesicles may explain the successful mesenchymal-to-epithelial transition to form nephrons, whereas failure of such compensation in the glomerular parietal cells could lead to the morphogenetic abnormalities of this cell lineage.

Discussion
We have examined the role of PAX2 in human nephron formation, by utilizing an induction protocol for nephron progenitors from iPS cells. By generating homozygous PAX2-null iPS cells, we have demonstrated that PAX2 is dispensable for the formation of nascent nephrons. These findings are significantly different from those obtained in Pax2-deficient mice. In mice, Pax2 is expressed in both lineages derived from nephron progenitors and the ureteric bud. Pax2/Pax8 double-heterozygous mice, as well as Pax2/Wt1 double-heterozygous mice, show more severe renal hypoplasia than Pax2 heterozygous mice 23,25,26 . Because Pax8 and Wt1 are exclusively expressed in the metanephric mesenchyme (including nephron progenitors), Pax2 is likely to have a role in this cell lineage, in addition to the ureteric bud. Indeed, mouse Pax2-null metanephric mesenchyme co-cultured with spinal cord fails to undergo mesenchymal-to-epithelial transition, thereby leading to a lack of nephron formation 9 . In contrast, human iPS cell-derived PAX2-null nephron progenitors in the same setting successfully epithelialise to form nephrons. This may be due to the compensatory expression of other genes at the mesenchymal-to-epithelial transition stage, including PAX8, because the PAX8 expression domain is expanded in the PAX2-null renal vesicles. Meanwhile, PAX8 is not up-regulated in the PAX2-null nephron progenitors, the stage before epithelial transition occurs. Thus, deletion of both PAX2 and PAX8 is required to prove the compensatory role of PAX8. Nonetheless, it is clear that mesenchymal-to-epithelial transition is less dependent on PAX2 in nephron formation from human iPS cells.
PAX2 is also expressed in precursors of the glomerular parietal epithelia in both humans and mice 27 . However, owing to the severe defects in nephron formation in Pax2-deficient mice, the importance of this gene in parietal cells has not been addressed. Based on the unaffected nephron formation from PAX2-deficient human nephron progenitors, we have shown that PAX2 is indeed required for the proper morphology of glomerular parietal cells. Our data revealed sustained expression of characteristic markers of columnar epithelia in PAX2-null parietal cells. In particular, the existence of CADHERIN6 suggests that the mutant cells retained the character of adjacent proximal renal tubules. PAX2 may inhibit CADHERIN6 expression directly or indirectly, thereby regulating the morphological changes of the glomerular parietal epithelia.
The discrepant data between human iPS cells and genetically engineered mice should be carefully interpreted, before concluding that there are species-specific differences in the requirement for PAX2. First, our protocol for nephron formation in vitro may not completely recapitulate nephron formation in vivo. As previously demonstrated, nephron-forming competence is retained in progenitors with a broader spectrum of gene expression patterns than expected 28,29 . Thus, it is possible that nephron formation based on the present protocol is less dependent on PAX2. Application of the protocol to Pax2-null mouse embryonic stem cells would be helpful to address the species-specific differences in the role of PAX2, at least in the in vitro setting. However, careful extrapolation is needed for its role in humans in vivo. Second, the spinal cord is a potent inducer of nephrogenesis from nephron progenitors, but does not extensively support their propagation, thereby leading to rapid depletion of the progenitors. Thus, the present study may not address the role of PAX2 in the propagation and maintenance of nephron progenitors. Recently, we and others reported the in vitro propagation of nephron progenitors induced from human iPS cells [29][30][31] . If a complete propagation method can be established, it would be worthwhile applying this method to PAX2-null progenitors.
PAX2 heterozygous iPS cells showed unaltered nephron formation, while accumulating evidence indicates that the renal coloboma phenotype is caused by PAX2 haploinsufficiency 9, 13, 15 . However, the kidney hypoplasia in heterozygous Pax2 1Neu mice could largely result from defects in the ureteric bud, because suppression of apoptosis in the ureteric bud restores this phenotype 16 . This notion may be consistent with the unaltered nephrogenesis from heterozygous human iPS cells, because our protocol selectively induces nephron progenitors, but not the ureteric bud. The present study does not address the role of human PAX2 in the ureteric bud, another important progenitor that constitutes the kidney. The Pax2-deficient murine Wolffian duct, as the precursor of the ureteric bud, exhibits impaired epithelial integrity, comprising loss of polarity and reduced intercellular adhesion 12 . Although a few groups have reported protocols for induction of ureteric buds from human iPS cells 32, 33 , the branching capacity and nephron-inducing potential were not sufficient. Thus, more robust protocols for the formation of competent Wolffian ducts/ureteric buds are under development in many laboratories, including ours, and these protocols will be useful for examining the role of PAX2 in this important lineage.
Regarding the technical aspects of the iPS cell-based differentiation, we established a method to sort human nephron progenitors by utilising ITGA8 expression on the cell surface, which enabled us to precisely compare the competence of nephron progenitors from different iPS clones. Clonal and experimental variations in the differentiation efficiencies of any organs are the bottleneck of iPS cell technology. Even when induction protocols are modified for individual clones, induction efficiencies may still vary depending on the experiments. While several factors, such as gene manipulation, passage numbers, and colony conditions, may cause these variations, the gene expression and differentiation competence of the sorted progenitors were consistent. Our data suggest that differentiation potentials can be precisely testable by starting with purified progenitors derived from individual clones. In this way, the subtle differences caused by gene deletion should be detected over clonal and experimental variations.
Taken together, we have established PAX2-deficient human iPS cells and revealed the dispensable role of PAX2 in nephron formation in vitro. Further improvement of the induction methods from human iPS cells will accelerate our understanding of kidney development in humans. Furthermore, human PAX2 mutations manifest many extra-renal symptoms, including optic nerve colobomas, auditory abnormalities, and in some cases, central nerve malformation 13,14 . Thus, our PAX2-deficient iPS cells will be useful tools to address the role of PAX2 in various tissues in humans.

Methods
Generation of PAX2-TALEN plasmids and the targeting vector. PAX2-TALENs were designed to bind to the following sequences and cleave close to the start codon of PAX2: 5′-TCCTCTGCCTCCCCATGG-3′ for the left TALEN and 5′-CAGACCCCTTCTCCGCGA-3′ for the right TALEN (underlined sequence corresponds to the start codon). A Platinum Gate TALEN Kit (Addgene; #1000000043) was used to construct the TALEN expression vectors as described previously 34 . The PAX2-TALEN plasmids were transfected into HEK293 cells using Lipofectamine 2000 (Thermo Fisher Scientific) to evaluate their activity. The target region was amplified with the following primers: 5′-CACCGTCCCTCCCTTTTCT-3′ and 5′-CCAAGATGGGACCTGAGCG-3′ (primers A and A′, respectively). When the resulting fragment was denatured and annealed, a clear band shift was observed, indicating the formation of a mismatched duplex resulting from deletions or insertions 35 .
For the targeting vector, 5′ and 3′ homology arms (0.71 kb and 0.82 kb, respectively) of PAX2 were amplified from the genomic DNA of iPS cells with the following primers: 5′-TGAATTCTTAGAGAGACACAC ACCGGG-3′ and 5′-TGCTAGCGGGGAGGCAGAGGAGCGGGA-3′ for the 5′ homology arm, and 5′-AACCTAGATCGGATCTGCACCGTGAGTACCGGCGC-3′ and 5′-ATTACGCCAAGCTTGCTG GCTCTCTCCCTGACTTC-3′ for the 3′ homology arm. Upon sequence verification, the 5′ homology arm was cloned in the EcoRI-NheI site of the modified HR120PA-1 vector (Systems Biosciences), such that the start codon of PAX2 was replaced with that of TurboGFP, followed by red fluorescent protein (RFP) and puromycin-resistance gene (Puro) cassettes flanked by loxP and insulator sequences. The 3′ homology arm was cloned in the BamHI-SphI site using an In-fusion HD Cloning Kit (Takara Bio).

Generation of PAX2-deficient iPS cells.
The human iPS cell line (201B7) was maintained on mouse embryonic fibroblasts as described 36 . The cells were pre-treated with Y27632 (10 µM) at 1 h prior to electroporation, and dissociated into single cells with dissociation solution (Reprocell; RCHETP002) followed by Accutase (Millipore). The targeting vector (10 µg), as well as the pair of TALEN plasmids (5 µg each), were electroporated into the dissociated human iPS cells using a SuperElectroporator NEPA21 (Nepagene) under the following conditions: two poring pulses (125 V, 2.5 ms) followed by five transfer pulses (20 V, 50 ms). The cells were then plated onto puromycin-resistant DR4 feeders, and puromycin (0.25 µg/ml) was added at 2 days after electroporation 37 . After 2 days of incubation, the puromycin concentration was increased to 0.5 µg/ml. The primers used for PCR screening of 5′ recombination were 5′-CCCATTTCTCCCTCCCCTG-3′ and 5′-CCACCAGCTCGAACTCCA-3′ (primers B and B′, respectively; product size: 1029 bp), and those for 3′ recombination were 5′-GGCTGTCCCTGATATCAAACA-3′ and 5′-GGCTGCGTGATCCTCAATG-3′ (primers C and C′, respectively; product size: 1041 bp). The PCR amplifications were performed as described 18 . The digoxygenin-labelled probes for Southern blot analysis were amplified using a PCR Dig Probe Synthesis Kit (Roche) and the following primers: 5′-AGCTCGATTCTGAACCAAGC-3′ and 5′-GGGAGCCCGGGATTAAAACT-3′ for probe A, and 5′-GGAAAGCCTCGGTCCTTTTC-3′ and 5′-CTCTAGCCCCACTTCTCACC-3′ for probe B. We obtained two homozygous GFP knock-in clones (−/−) among 91 clones, as determined by PCR and Southern blotting. Although we obtained 35 candidate heterozygous clones, most of the clones had mutations at the TALEN target sites in the non-GFP-containing allele, probably arising from the high activity of TALEN plasmids, resulting in the acquisition of only one heterozygous clone (+/−). Likewise, only two wild-type clones were confirmed to be devoid of mutations at their TALEN target sites (+/+). All of the clones showed nephron formation, and GFP faithfully mimicked PAX2 expression without deletion of the puromycin-resistance cassette, as shown in Fig. 2. All experiments were approved by the Committee on Living Modified Organisms of Kumamoto University (#A27-064), and performed in accordance with the institutional guidelines of Kumamoto University.
For staining of PAX8, sections were treated for antigen retrieval and subsequently incubated with an anti-PAX8 antibody (Abcam; ab53490) followed by an anti-mouse secondary antibody conjugated with peroxidase polymers from an ImmPress Reagent Kit (Vector Laboratories; MP-7402). The sections were then incubated with Alexa Fluor 594-tyramide (Thermo Fisher Scientific) for 15 min at room temperature. The stained expression patterns were consistent with the reported results obtained by in situ hybridisation 25 . In contrast, staining with another anti-PAX8 antibody (Proteintech; 10336-1) showed almost identical patterns to those of PAX2, indicating the cross-reactivity of this antibody with PAX family proteins. Sections of human embryonic kidney were purchased from USbiomax, and stained for PAX2 and PAX8 using the ImmPress Reagent Kit and Alexa Fluor-tyramide, as described above.
Quantitative RT-PCR analysis. RNA was isolated using an RNeasy Plus Micro Kit (Qiagen), and then reverse-transcribed with random primers and a Superscript VILO cDNA Synthesis Kit (Life Technologies). Quantitative PCR was carried out using a Dice Real Time System Thermal Cycler (Takara Bio) and Thunderbird SYBR qPCR Mix (Toyobo). The primer sequences are listed in Supplementary Table S2. For the data shown in Fig. 3B, fractions from three independent experiments were analysed. For the data shown in Fig. 3D, three spheres were combined to represent one sample, and three samples were analysed for each clone. The data were analysed by Student's t-test.