Stromal Fat4 acts non-autonomously with Dchs1/2 to restrict the nephron progenitor pool

Determining how organ size is regulated, and how progenitor cells self-renew, is crucial for understanding normal development. The mammalian kidney derives from three interdependent lineages: the ureteric bud (UB), which forms the collecting ducts; the nephron progenitors, also known as condensing or cap mesenchyme (CM), which both self-renew and undergo a mesenchymal-to-epithelial transition (MET) to give rise to the nephron; and the loosely associated mesenchymal-like cells known as the stroma.

Stromal cells regulate both UB renal branching and the balance between self-renewal and nephron differentiation of the CM (Batourina et al., 2001; Das et al., 2013; Hatini et al., 1996; Hum et al., 2014; Levinson et al., 2005; Mendelsohn et al., 1999, 1994; Paroly et al., 2013). Loss of the stromal transcription factor gene Foxd1 results in reduced branching and an expanded CM (Hatini et al., 1996). Similarly, loss of the stromal transcription factor genes Pbx1 (Schnabel et al., 2003), Pod1 (Tcf21 ­– Mouse Genome Informatics) (Quaggin et al., 1999) or the atypical cadherin gene Fat4 (Das et al., 2013; Saburi et al., 2008) also results in an excessive CM. How these stromally expressed genes regulate self-renewal and/or differentiation of the CM remains to be determined.

The Drosophila tumour suppressor Fat (Ft) is a large atypical cadherin that regulates cell adhesion (Fanto et al., 2003; Matakatsu and Blair, 2006), planar cell polarity (PCP) (Fanto et al., 2003; Rawls et al., 2002; Yang et al., 2002), growth (Bennett and Harvey, 2006; Bryant et al., 1988; Silva et al., 2006; Willecke et al., 2006) and metabolism (Sing et al., 2014). The vertebrate family of FAT cadherins is composed of four cadherins (FAT1-4). FAT4 is considered to be the orthologue of Ft based on homology of the intracellular domain and the ability of the FAT4 cytoplasmic domain to rescue PCP defects of Drosophila ft mutants (Pan et al., 2013). Fat4^−/− mice fail to establish proper PCP in the kidney tubules, resulting in the formation of renal cysts (Saburi et al., 2008). Fat1 is required for renal slit junction formation (Ciani et al., 2003), and synergises with Fat4 in cyst repression (Saburi et al., 2012). Fat2 is the only FAT family cadherin not expressed in the developing kidney (Barlow et al., 2010; Rock et al., 2005). Fat3 is expressed in the developing kidney (Rock et al., 2005), and acts synergistically with Fat4 to regulate cyst development (Saburi et al., 2012).

Dachsous (Ds) is a large atypical cadherin that binds Drosophila Ft, and regulates growth, adhesion and PCP [reviewed by Sharma and McNeill (2013)]. Dachsous 1 (DCHS1) and Dachsous 2 (DCHS2) are the two mammalian orthologues of Drosophila Ds. Mao et al. (2011) showed that Dchs1 mutants have similar phenotypes to Fat4 mutants, including cystic kidneys. Dchs2 is also expressed in the kidney, but no genetic study has yet addressed the function of Dchs2.

Fat4 mutants have an expanded CM population surrounding the UB tips (Das et al., 2013; Mao et al., 2011; Saburi et al., 2008). Drosophila Ft regulates the Hippo kinase pathway, a conserved growth-controlling pathway that regulates the transcriptional co-activator protein Yorkie via inhibitory phosphorylations. The mammalian homologues of Yorkie are called YAP and TAZ. Recently, Das et al. (2013) reported that YAP levels were increased and phosphorylated-YAP (pYAP) was decreased in the CM of Fat4 mutants. They proposed that FAT4 regulates the progenitor pool by increasing nuclear YAP in the CM non-autonomously, possibly by binding to FAT3 or DCHS1 in the CM.

Here, we use tissue-specific analyses with Cre recombinase in mice to demonstrate that Fat4 regulates the CM non-autonomously by acting in the stroma. We show that CM formation is not affected by the core PCP pathway, and that Dchs1 and Dchs2 act in a partially redundant fashion to control mesenchymal progenitor cell number. We further show that Fat1 and Fat3 are not required to regulate nephron progenitors, and that loss of Yap in the CM does not reduce the excess progenitors found in Fat4 mutants. Electron microscopy (EM) analysis reveals that Fat4 mutants have defective cellular organisation in both the CM and the UB-derived epithelial tubules. mRNA expression analysis of Fat4 mutants reveals alterations in Notch and FGF pathway components. Taken together, our data indicate that FAT4 in the stroma binds to DCHS1 and DCHS2 in the CM to regulate nephron progenitors, independently of YAP in the CM, and that loss of Fat4 disrupts multiple signalling pathways.

Loss of Fat4 results in an expanded CM that is visible in histological sections (Fig. 1A,B). To understand better how loss of Fat4 leads to excess CM, we carefully characterised Fat4^−/− kidney development, using antibodies against SIX2 to label the CM and cadherin 1 (ECAD) to label the UB. This analysis showed that loss of Fat4 results in an expanded CM population that is obvious from embryonic day (E) 12.5 to birth (Fig. 1G-L; supplementary material Fig. S1A-F). Quantification of the CM progenitor pool surrounding each UB confirms an almost twofold increase in the number of nephron progenitors (e.g. at E13.5 there was an average of 95 cells/UB in wild type versus 181 cells/UB in Fat4^−/− mutants; Fig. 1M). Staining with other CM markers, such as CITED1, PAX2 and SALL1, confirm expansion of the CM, and the presence of the stromal markers PBX1, FOXD1 and tenascin C (TEN; TNC ­– Mouse Genome Informatics) (Fig. 1G-L; Fig. 4E-J) demonstrate an apparently normal-sized stroma. Quantification of glomeruli number showed reduced nephrogenesis at postnatal day (P) 0 in Fat4 mutants, suggesting a delay in epithelialisation (supplementary material Fig. S1M-O).

To obtain a better insight into the cellular mechanism underlying the increased CM observed in Fat4 mutants, we used ultrathin sections and EM to examine Fat4 mutants. In controls at E12.5, the normal tight organisation of mesenchymal cells surrounding an inducing UB tip was easily observed (Fig. 1C). EM analysis of Fat4^−/− kidneys at E12.5 showed a larger CM surrounding the epithelial tubules (Fig. 1D). In addition, both the epithelial tubules and the CM of Fat4 nulls were distorted and showed a lack of tight organisation (Fig. 1C-F). This lack of tissue organisation could reduce the efficacy of signalling from the UB tip to the CM.

It has been speculated that Fat4 might act through other FAT cadherins, such as FAT3, to regulate the CM progenitor pool (Das et al., 2013). Fat1, Fat3 and Fat4 are the only FAT cadherins expressed during kidney development (Rock et al., 2005). We therefore examined the CM of Fat1^−/− and Fat3^−/− single mutants by staining embryonic kidneys with antibodies to SIX2 and ECAD (Fig. 2A-D). This revealed a normal CM in both Fat3^−/− and Fat1^−/− single mutants. We also found that Fat1^−/−;Fat3^−/− double mutants had a normal-sized CM (Fig. 2I). Thus, Fat4 has a unique role in regulating the CM, and does not act through Fat1 or Fat3 to control the nephron progenitor pool.

Drosophila dachsous (ds) has two mammalian orthologues (Dchs1 and Dchs2). Previous work has shown that loss of Dchs1 results in renal cysts, albeit to a lesser extent than that observed in Fat4 mutants (Mao et al., 2011). We examined Dchs1^−/− mutants and found that loss of Dchs1 phenocopies the expanded Fat4 CM (Fig. 3A,B). Staining with several CM markers confirmed excess progenitors in Dchs1 mutants (Fig. 3E-J). Quantification of the CM population surrounding each UB confirmed expansion of the CM in Dchs1^−/− kidneys (average 115 cells/UB in wild type versus average 171 cells/UB in Dchs1^−/−) (Fig. 3K). Dchs1^−/− mutants had a slightly smaller CM population compared with Fat4^−/− mutants (average 171 cells/UB versus 181 cells/UB); however, this was not statistically significant. Immunofluorescence (IF) and in situ hybridisation (ISH) analyses of Dchs1 revealed highest expression in the stroma and mesenchyme, with weak expression in the UB (supplementary material Fig. S2B,C,G,H). Interestingly, DCHS1 expression is increased in Fat4 mutants (supplementary material Fig. S2I,J; Mao et al., 2011).

We wondered whether the Dchs1 paralogue Dchs2 might also play a role in the regulation of nephron progenitors. Dchs2 is expressed at low levels throughout the developing kidney (supplementary material Fig. S2). We generated Dchs2-null mice, and determined that they were viable and fertile (F.H., unpublished). Analysis of Dchs2 mutants showed a normal-sized CM (Fig. 3C,K). To determine whether there is a synergistic relationship between Dchs1 and Dchs2, we generated and examined Dchs1^−/−;Dchs2^−/− double mutants and stained for SIX2 and ECAD (Fig. 3D). Importantly, quantification revealed that Dchs1^−/−;Dchs2^−/− mutants have an expanded CM population that is larger than that of Dchs1 single mutants (Fig. 3D,K). Thus, Dchs1 and Dchs2 are partially redundant in the regulation of the progenitor pool.

Drosophila Ft and Ds regulate PCP in the wing and eye [reviewed by Sharma and McNeill (2013)]. Atrophin (Grunge – FlyBase) is a nuclear co-repressor that functions with Ft to regulate PCP in the Drosophila eye (Fanto and McNeill, 2004). Two mammalian orthologues of Atrophin exist in mammals: atrophin 1 (Atn1) and atrophin 2 (Rere). Atn1 mutants are viable and fertile, with a normal CM (data not shown). Rere^−/− mutants die at ∼E9.5 (Zoltewicz et al., 2004). We find that Rere^+/− single mutants had a normal CM (supplementary material Fig. S3B,C) and Fat4^−/−;Rere^+/− double mutants have an expanded CM that is similar to that of Fat4 single mutants (Fig. 2J). To assess more directly RERE function in the CM, we generated Rere^flox/flox conditional null mutants (D.A.S., unpublished). Removal of Rere with Pax2-Cre did not affect the size of the CM (Fig. 2K,L). Thus, alterations in Atn1 and Rere do not disrupt CM self-renewal and/or differentiation.

Vangl2 is an essential component of the core PCP pathway (Simons and Mlodzik, 2008). Loss of Vangl2 results in a reduced number of glomeruli and reduced branching (Babayeva et al., 2011; Yates et al., 2010). Staining with a SIX2 antibody revealed that Vangl2^−/− single mutants have a normal CM, and that Fat4^−/−;Vangl2^+/− double mutants have a CM population similar in size to that of Fat4 single mutants (Fig. 2G,H).

Drosophila Four-jointed (Fj), is a Golgi-localised kinase that has roles in both PCP and Hippo growth regulation [reviewed by Sharma and McNeill (2013)]. fj has a single mammalian orthologue, Fjx1, which has increased expression in both Drosophila fat and mouse Fat4^−/− mutants (supplementary material Fig. S1P,Q; Saburi et al., 2008). Staining of Fjx1^−/− kidneys with a SIX2 antibody showed a normal CM, suggesting that Fjx1 is not essential for regulating the size of the progenitor pool (Fig. 2E). To determine whether increased Fjx1 expression in Fat4 mutants is responsible for the enhanced CM of Fat4 mutants, we examined Fat4^−/−;Fjx1^−/− double mutants. Fat4^−/−;Fjx1^−/− double mutants phenocopy Fat4 single mutants (Fig. 2F), indicating that increased FJX1 is not responsible for the increased CM size of Fat4 mutants. Taken together, our results indicate that Atn1, Rere, Vangl2 and Fjx1 are not required for nephron progenitor regulation.

To understand better how Fat4 and Dchs1 regulate progenitors, we examined their expression patterns. Using both ISH and a Fat4^EGFP/+ knock-in reporter line (Wu et al., 2008), we found that Fat4 has highest expression in the stroma, moderate expression in the CM and weak expression in the UB (Fig. 4A-D). The stromal compartment of Fat4 mutants still expresses the stromal markers FOXD1 and PBX1, along with increased tenascin C expression in the stroma (Fig. 4E-H).

To determine where Fat4 acts to control progenitor pool size, Fat4 was removed from different lineages. Removal of Fat4 from the UB (Hoxb7-Cre;Fat4^flox/−) or the CM (Six2-Cre;Fat4^flox/−) lineages did not result in an expanded CM population (Fig. 5D,E). Thus, Fat4 does not act in the epithelium or in the CM to regulate MET. However, by contrast, we found that removal of Fat4 only from the stroma (Foxd1^Cre/+;Fat4^flox/−) resulted in an expanded CM population, similar in size to that of Fat4 mutants (Fig. 5F). Therefore, Fat4 in the stroma non-autonomously regulates the size of the CM.

The stromal compartment can be subdivided into three layers: (1) the capsular stroma (single layer of cells surrounding the developing kidney proper); (2) the cortical stroma (peripherally localised cells that surround the UB and the CM); and (3) the medullary stroma (the inner kidney) (Cullen-McEwen et al., 2005; Li et al., 2014). Foxd1^Cre/+ excises from both cortical and medullary stromal compartments (Humphreys et al., 2010; Kobayashi et al., 2014). Pax3 is expressed throughout the kidney, with mosaic expression in both the UB and the CM, and is strongly expressed in the medullary and cortical stroma (Engleka et al., 2005). We found that Pax3^Cre/+;Fat4^flox/− kidneys also exhibited large CM condensates similar to those observed in Fat4 nulls and Foxd1^Cre/+;Fat4^flox/− conditional mutants (Fig. 5C). To determine whether Fat4 is acting in the cortical or medullary stroma, we examined Rarb2-Cre;Fat4^flox/− kidneys because Rarb2-Cre excises from the CM and medullary stroma but not from the cortical stroma (Di Giovanni et al., 2011; Kobayashi et al., 2005). Interestingly, Rarb2-Cre;Fat4^flox/− kidneys (supplementary material Fig. S3D,E) have a normal-sized CM. Taken together, our results indicate that Fat4 acts non-autonomously in the cortical stroma to regulate the renal progenitor pool.

Previous studies stated that loss of Fat4 in E15.5 embryos result in increased Yap expression in the CM and decreased pYAP, and proposed that increased nuclear YAP in the CM regulates nephron progenitor pool size (Das et al., 2013). To explore this potential link, we stained for YAP and phosphorylated-YAP (pYAP) in Fat4 null and Foxd1^Cre/+;Fat4^flox/− mutants. We were unable to detect any differences in YAP or pYAP in Fat4 null or Foxd1^Cre/+;Fat4^flox/− conditional mutants by IF at E13.5, E14.5 and P0 (supplementary material Fig. S4; data not shown). Western blots of Fat4 mutants and controls also did not reveal consistent differences in pYAP expression levels (supplementary material Fig. S4).

It is possible that subtle changes in YAP levels or phosphorylation were not detectable with our approaches. We reasoned that if loss of Fat4 indeed functions by increasing nuclear YAP, then removal of Yap in the CM of a Fat4 mutant should rescue the expanded CM phenotype. We examined Fat4^−/−;Yap^+/− double mutants, and found that they phenocopy Fat4 single mutants in CM size (Fig. 6A-C), suggesting that changes in YAP levels in the CM are not the crucial mediators of excess CM in Fat4 mutants.

Yap^−/− mutants die at ∼E9.5, so to test directly whether increased YAP mediates the increased progenitors in Fat4 null mutants, we used conditional alleles to remove Yap solely from the CM in Fat4 mutants (i.e. a Fat4^−/−;Six2-Cre;Yap^flox/flox mutant) (Fig. 6D-I). We found that Yap is efficiently excised from the CM (supplementary material Fig. S4). Significantly, these mutants phenocopied Fat4 single mutants (i.e. there was no rescue of the expanded CM phenotype), demonstrating that the increased CM in Fat4 mutants is not due to excess nuclear YAP in the CM.

To obtain an unbiased global view of changes in gene expression in Fat4 mutants, we isolated kidneys from Fat4^−/− mutants and Fat4^+/+ controls at E13.5, and conducted RNA sequencing of the transcriptome. Consistent with the increased CM of Fat4 mutants seen in histological and marker analysis, Fat4^−/− kidneys have increased expression of progenitor genes, including Six2, Sall1, Gas1, Meox2 and Eya1 (Fig. 7A). There is a moderate reduction of some stromal genes (Pod1, Pbx1; Fig. 7A). No change in expression was detected in the YAP target gene Ctgf or the canonical Wnt pathway target Axin2 (Fig. 7C). Interestingly, global analysis of gene expression showed significant changes in several components of the Notch and FGF pathways (Fig. 7B,C). These data confirm that loss of Fat4 leads to increased expression of progenitor markers, and show that multiple pathways known to regulate the CM (Boyle et al., 2011; Di Giovanni et al., 2015) are disrupted in Fat4 mutants.

Our staining and transcriptome analyses indicate that loss of Fat4 results in increased expression of Six2 and expansion of SIX2⁺ cells. Previous studies have shown that Six2 is a crucial regulator of the progenitor pool (Kobayashi et al., 2008; Self et al., 2006), suggesting that the increase in Six2 expression might be important for the increased CM of Fat4 mutants. We therefore examined Six2^−/−;Fat4^−/− mutants. Strikingly, Six2^−/−;Fat4^−/− mutants have a depleted progenitor pool, with precocious epithelialisation similar to that observed in Six2^−/− mutants (Fig. 7D). These data indicate that Six2 is epistatic to Fat4, and suggest that Six2 acts downstream of, or in parallel to, Fat4 to regulate the progenitor pool.

We have shown here that Fat4 functions non-autonomously to regulate the size of the nephron progenitor pool. We further show that Fat4 is the only FAT cadherin the loss of which results in an expanded CM and find similar phenotypes in Dchs1^−/− mutants. Importantly, we also find there is a strongly expanded CM in Dchs1^−/−;Dchs2^−/− double mutants. These data suggest that FAT4 in the stroma binds to DCHS1 and DCHS2 to non-autonomously restrict the number of CM progenitors. How and where does this signalling occur? In Drosophila, when Ft and Ds bind, there is evidence for signalling occurring in both directions (reviewed by Sharma and McNeill, 2013), with some of this signalling mediated by the Ft and Ds intracellular domains. As there is clear conservation not only between the cytoplasmic domains of Fat and FAT4, but also between Ds and DCHS1 and DCHS2, this raises the possibility of bidirectional signalling during regulation of the nephron progenitor pool. A signal could be generated via Dachsous cytoplasmic signalling in the nephron progenitors or downstream of FAT4 in the stromal cells, leading to restriction of the progenitor pool.

Dchs1 regulates CM size cell-autonomously (Mao et al., 2015), and, as we have shown here, Fat4 is needed in the stromal population to regulate CM size. Therefore, the interface between the CM and the stroma is a likely place for a FAT4-DCHS1/2 signal to occur. The CM is normally two to three cell layers thick. How the FAT4-DCHS1/2 signal passes through this cell layer is not clear. Cell motility could result in different cells being involved in a FAT4-DCHS1/2 contact over time, or alternatively a diffusible signal could be generated as a result of a FAT4-DCHS1/2 interaction. Further detailed molecular and cellular studies will be needed to ascertain how FAT4 and DCHS1 signal at this interface.

Drosophila Ft and Ds binding regulates PCP. To determine whether mammalian PCP affects nephron progenitors, we examined Fjx1, Atn1, Rere and Vangl2 mutants. Loss of these genes, or loss of one of these genes in a Fat4-null background did not affect the renal progenitor pool. Together, these data suggest that Fat4 regulation of progenitors is not via the core PCP pathway or via the atrophin-PCP pathway.

Fat4 is expressed predominantly in the stroma, with moderate expression in the mesenchyme and weak expression in the UB. Removal of Fat4 from the stromal population using Foxd1^Cre/+ resulted in excess nephron progenitors, demonstrating that Fat4 acts non-autonomously to control the CM. Foxd1 is expressed in both cortical and medullary interstitial stromal cells (Kobayashi et al., 2014). When Fat4 was excised only from the medullary stroma using Rarb2-Cre, the CM appeared to be normal. These results suggest that Fat4 acts from the cortical stromal layer to regulate nephron progenitors.

Recently, a stromaless kidney (produced by expression of diphtheria toxin in the Foxd1 lineage) was shown to phenocopy Fat4 nulls (Das et al., 2013; Hum et al., 2014). Owing to complete removal of the stromal compartment, the authors were unable to determine if a single molecular factor was responsible for the increased CM observed. We note that the increase of CM pool size in the stromaless kidney is greater than that observed in the Fat4 mutant (Das et al., 2013), which may be due to other factors, or due to secondary proliferation subsequent to death/injury caused by the diphtheria toxin. We note that dying cells are known to release signals that promote proliferation of adjacent cells (Chera et al., 2009; Huang et al., 2011; Smith-Bolton et al., 2009), and that Yki/YAP levels increase after tissue damage, and are essential for repair and regeneration. Our genetic data clearly show that Fat4 is a crucial stromal regulator of the CM. As we see only subtle decreases in Foxd1, Pbx1 and Pod1, the effect of Fat4 is unlikely to be mediated by regulation of the expression of these genes.

Our studies indicate that Fat4 acts non-autonomously in the stroma, and suggests a model in which FAT4 signals to DCHS1 in the CM. We find that YAP and pYAP levels do not change in Fat4 mutants during renal development (E11.5-P0), suggesting that YAP is not essential for the excess CM seen in Fat4 mutants. We cannot account for the difference in staining between our analysis and that of Das and co-workers (Das et al., 2013); however, we note that they only reported changes in YAP expression at E15.5, days after the excessive CM becomes obvious. We note that YAP levels are slightly lower in the CM of wild-type animals, and as the CM is expanded in Fat4 mutants, this could give an impression of lower YAP in Fat4 nulls.

Our genetic analyses clearly demonstrate that Fat4 does not control nephron progenitors solely through regulation of YAP in the CM, as removal of Yap from the SIX2⁺ renal population does not rescue the excess CM observed in Fat4 nulls. It is possible that the YAP paralogue TAZ could play a role; however, Fat4^−/−;Taz^−/− kidneys have an expanded CM (Mao et al., 2015), indicating that TAZ is also not essential for the increased CM of Fat4 mutants. As we have yet to examine Fat4^−/−;Six2-Cre;Yap^flox/flox;Taz^flox/flox mutants, we cannot exclude the possibility that TAZ could compensate for the loss of YAP. However, we see no increase in TAZ levels in Fat4 mutants (data not shown), and loss of both YAP and TAZ in the CM does not cause an early block in CM renewal (Reginensi et al., 2013). Taken together, these data argue against a crucial role for either YAP or TAZ in the CM in Fat4-dependent nephron progenitor expansion. By contrast, we find that Six2 is needed to maintain the progenitor pool (Kobayashi et al., 2008; Self et al., 2006). Our data indicate that Six2 is needed for the expanded CM of Fat4^−/− mutants, as Six2^−/−;Fat4^−/− mutants show loss of the characteristic expanded CM of Fat4 mutants, and instead display the precocious epithelialisation of Six2 mutants.

FAT4-DCHS1/2 signalling could be affecting progenitor maintenance, proliferation or the ability of the progenitors to commit and convert to an epithelial renal vesicle. Wnt9b induces the commitment of a subset of progenitors to an epithelial fate (Carroll et al., 2005). While canonical WNT signalling promotes epithelialisation, stabilisation of β-catenin blocks transition of induced CM to renal vesicles, and progression of the nephrogenic programme (Park et al., 2007). Thus, excess canonical signalling could also potentially account for the excess CM. However, we found no changes in the TCF/Lef1-lacZ WNT reporter in Fat4 mutants (Saburi et al., 2008), nor in expression of the WNT target gene Axin2, suggesting that Fat4 acts in a parallel pathway to Wnt9b to regulate progenitor gene expression and CM size. Terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) analysis shows a minor increase in cell death in Fat4 mutants, and analysis of 5-ethynyl-2′-deoxyuridine (EdU) incorporation indicates that there is no increase in proliferation of the CM in Fat4 mutants (supplementary material Fig. S5). Although decreased branching could lead to reduced division of CM at tips, we see no increase in the CM in Vangl2 mutants, which also show reduced branching (Yates et al., 2010; M.B.-L., unpublished). Thus, we favour the hypothesis that loss of Fat4 affects signalling from the stroma to Dchs1/2 in the CM to alter progenitor self-renewal.

Our transcriptome analysis of Fat4 mutants revealed previously unsuspected changes in genes involved in Notch and FGF-dependent signalling. Unravelling the contributions of these signalling pathways to Fat4-dependent regulation of progenitors will require double mutant/transgenic analyses of each of these pathways with Fat4. Genetic studies are also needed to understand the precise contribution of progenitor-regulating genes that have altered expression in Fat4 mutants. Determining how Fat4 and Dchs1/2 control Six2-dependent progenitor differentiation/renewal will provide insights into how progenitor self-renewal is controlled, and may explain the renal hypoplasia observed in patients that harbour mutations in FAT4 and DCHS1 (Cappello et al., 2013; Mansour et al., 2012).

All mouse work was carried out in accordance with Ethics Canada, approved by the Animal Care Committee, and followed The Toronto Centre for Phenogenomics standard operating protocols.

Fat4^−/− and Fat4^flox/flox (Saburi et al., 2008), Fat4^EGFP/EGFP (Wu et al., 2008), Dchs1^−/− (Mao et al., 2011), Pax3^Cre/+ (Engleka et al., 2005), Pax2-Cre (Ohyama and Groves, 2004), Foxd1^Cre/+ (Humphreys et al., 2010; Kobayashi et al., 2014), Six2-Cre^TGC (Kobayashi et al., 2008), Hoxb7-Cre (Zhao et al., 2004), Rarb2-Cre (Kobayashi et al., 2005), Yap^flox/flox (Reginensi et al., 2013), Rere^−/− (Zoltewicz et al., 2004), Fat1^−/− (Ciani et al., 2003), Fjx1^−/− (Probst et al., 2007) and Vangl2^Lp/+ (Strong and Hollander, 1949) mouse mutant alleles have all been previously described. The Dchs2^−/− mouse mutant (international nomenclature Dchs2^tm1.2FHel) will be described fully elsewhere (F.H., unpublished). Briefly, a conditional allele (Dchs2^tm1.1FHel) was first generated by flanking the coding part of the last exon, encompassing the transmembrane and intracellular domains of Dchs2, with loxP sites. Mice carrying a constitutively deleted allele (Dchs2^tm1.2FHel) were produced by crossing the former with a germline-active deleter-cre mouse line [Tg(CMV-cre)1Cgn]. Rere^flox/flox mice will be described fully elsewhere (H.P.Z., D.A.S., unpublished). Briefly, a recombineering strategy was used to generate a targeting vector in which the second coding exon of Rere was flanked by a 5′ loxP site and a 3′ FRT-flanked neo cassette and loxP site. In vivo excision of the neo cassette was achieved by crossing male mice carrying a correctly targeted Rere allele to female mice expressing the FLPe variant of the Saccharomyces cerevisiae FLP1 recombinase gene in their germ line. On exposure to Cre, the second coding exon of the Rere flox allele is excised, resulting in a shift in the reading frame and generation of a premature stop codon. Presence of a vaginal plug on noon of the day when found was considered E0.5.

Standard Periodic acid-Schiff (PAS) staining was carried out according to (Reginensi et al., 2013).

EM was carried out on E12.5 kidneys fixed in 0.1 M cacodylate buffer (with 4% paraformaldehyde and 2% glutaraldehyde). Kidneys were fixed again in osmium tetroxide and embedded in Quetol-Spurr resin, and resin sections were stained with uranyl acetate and lead citrate and imaged under a FEI CM100 transmission electron microscope.

Paraffin-embedded sections of newborn kidneys were cut at 7 µm thickness, stained with WT1 antibody and used to count the number of glomeruli per section, which was averaged for each kidney. A total of three wild-type and three mutant mice were compared.

Digoxigenin-UTP labelled Fat4, Fat1, Dchs1, Dchs2, Fgfr2, Fgf10, Hes1 and Hes5 riboprobes were generated and samples processed according to Reginensi et al. (2013).

IF was carried out on embryos embedded in paraffin according to standard protocols. Briefly, embryos were fixed overnight in 4% paraformaldehyde, washed in PBS, followed by serial dehydration with ethanol prior to embedding. Tissues were sectioned at 7 µm, deparaffinised, rehydrated, and boiled for 22 min with Antigen Unmasking Solution (H-3300, Vector Laboratories). Slides were blocked for an hour, probed overnight with primary antibodies and washed with PBS. FITC-, Alexa 488-, Cy3- or Cy5-conjugated secondary antibodies and Hoechst 33542 were added, and slides mounted with Vectashield. The following antibodies were used: SIX2 (ProteinTech; 11562-1-AP), pYAP (S127-Cell Signaling; 4911S), YAP (63.7; Santa Cruz; sc-101199), CITED1 (ThermoScientific; RB-9219-P0), PBX1b (Santa Cruz; sc-191852), WT1 (Dako; M3561), CDH1 (ECAD; BD Biosciences; 610181), CDH1 (ECAD) (rat; Invitrogen; 131900), PAX2 (Covance; PRB-276P), FOXD1 (generously provided by Dr Andy McMahon, University of Southern California, Los Angeles, USA), neural cell adhesion marker (NCAM; Sigma; C-9672), Sal-like 1 (SALL1; Abcam; AB31526), JAG1 (Cell Signaling; 2620), GFP (Abcam; ab13970) and IgG (Cell Signaling; 2729S). Frozen embryos were embedded in OCT and sectioned at 10 µm according to standard protocols. A tenascin C antibody from Sigma (T3413) was used. A Nikon D-Eclipse C1 confocal microscope was used to image all IF samples.

Paraffin-embedded embryos cut at 7 µm were used to assess the number of SIX2⁺ cells in embryonic kidneys. IF was carried out using SIX2 and ECAD antibodies to identify the CM and UB, respectively. SIX2⁺ cells that surrounded a ureteric bud were quantified in ImageJ 1.43u (NIH, USA) using a macro that sets an autothreshold and runs the watershed option, thereby analysing particles that have clear outlines and were circular.

pYAP (5127-Cell Signaling) and GAPDH (Sigma) were used in immunoblots to assess pYAP protein levels in whole kidneys. Briefly, kidneys from E13.5 embryos were dissected, flash-frozen and stored at −80°C until further use. Tissues were homogenised in RIPA buffer using 23.5-G syringes, electrophoresed on 10% SDS-polyacrylamide gels and transferred onto PVDF membranes. Samples were blocked in 5% skimmed milk in TBS-Tween for 1 h at room temperature and incubated with primary and secondary antibodies. A Fluor-S MultiImager Max imager system was used to visualise samples following chemiluminescence detection.

E12.5 embryos were processed for paraffin embedding and 7-µm sections were used for EdU incorporation assay and TUNEL analysis. For EdU analysis, 10 mg/ml of EdU was injected into pregnant dams 15 min prior to dissection of embryos. The Click-iT EdU Alexa Fluor 488 imaging kit from Life Technologies (C10337) was used to fluorescently label proliferating cells. TUNEL analysis was carried out according to Reginensi et al. (2013) using the Roche In Situ Cell Death Detection Kit (TMR red; 12 156 792 910).

Gene expression analysis was performed using RNA-Seq data from three wild-type and three Fat4 mutant samples by determining values for ‘corrected reads per kilobase per million reads’ (cRPKM), essentially as previously described (Labbé et al., 2012) and using the vast-tools pipeline (Irimia et al., 2014). Briefly, cRPKM values represent the number of unique-mapping reads per kilobase of transcript, after length-correction to account for multiple mapping positions. The corrected length is then used to divide the raw read counts per million mapped reads for each gene. The RNA-Seq datasets comprised 100 million paired-end, 101-mer reads. To assess significant fold differences between samples, an unpaired, uncorrected Student's t-test was computed with a 0.05 P-value cutoff to determine differentially expressed transcripts. All data have been submitted to Gene Expression Omnibus under the accession number GSE70452.

All data are expressed as mean values with error bars representing standard error of the means (s.e.m.). An unpaired two-tailed Student's t-test was used to determine differences between two groups. To examine for significant differences among groups, data were subjected to a one-way analysis of variance (ANOVA). When statistical differences were found by one-way ANOVA, a Tukey's multiple comparison post-hoc test was conducted to delineate significance between groups. A fiducial limit of P≤0.05 was used throughout. All statistical analyses were conducted using GraphPad Prism 5.0a software.

We thank Ken Irvine (Rutgers University) for the Dchs1^−/− mice; Jeff Wrana (Lunenfeld-Tanenbaum Research Institute) for the Yap^flox/flox mice; Carl Bates (University of Pittsburgh) for the Hoxb7-Cre mice; Andy McMahon (University of Southern California) and Akio Kobayashi (University of Washington) for the Six2-Cre, Six2^+/− and Foxd1^Cre/+ mice; and Mario Cappecchi for the Fat4^EGFP/+ mouse line (University of Utah). We are grateful to Doug Homyard for all electron microscopy analyses.