Hedgehog-GLI signaling in Foxd1-positive stromal cells promotes murine nephrogenesis via TGFβ signaling

The mammalian kidney contains hundreds of thousands of nephrons that each connect to a collecting system composed of collecting ducts, calyces, the renal pelvis and the ureter (Nyengaard and Bendtsen, 1992). These structures are supported by renal stromal cells that originate during embryonic organogenesis (Hatini et al., 1996; Mendelsohn et al., 1999). Nephrons and collecting system elements are derived from the metanephric mesenchymal SIX2⁺ cells and Hoxb7⁺ ureteric bud cells, respectively, each of which is derived, in turn, from Osr1⁺ intermediate mesoderm (Mugford et al., 2008). Reciprocal inductive interactions between ureteric and nephrogenic cells result in iterative branching of the ureteric bud (to form the collecting system) and induction of nephron formation by mesenchymal cells adjacent to ureteric bud tips, respectively (Grobstein, 1953; reviewed by O'Brien and McMahon, 2013; Kopan et al., 2014; Dressler, 2009; Costantini and Kopan, 2010; Blake and Rosenblum, 2014). Nephrons form via sequential morphological changes in mesenchymal aggregates that progress to epithelial vesicles, comma- and S-shaped bodies and, finally, to mature nephrons.

Although the classical theory of mammalian nephron formation focuses on ureteric bud-metanephric mesenchyme tissue interactions, nephron formation is also controlled by Foxd1⁺ stromal cells that also derive from Osr1⁺ intermediate mesoderm (Mugford et al., 2008). During mouse kidney development, Foxd1⁺ cells can be identified as early as E11-E11.5 at the periphery of the metanephric mesenchyme (Hatini et al., 1996; Yallowitz et al., 2011; Hum et al., 2014). Thereafter, these cells come to surround the kidney and migrate between nephrogenic elements in the renal cortex (Yallowitz et al., 2011; Kobayashi et al., 2014). Homozygous deficiency of Foxd1 or genetic ablation of Foxd1⁺ cells disrupts renal development, leading to: abnormal formation and cellular composition of the renal capsular stroma; disorganization of renal patterning; and decreased nephron formation (Hatini et al., 1996; Levinson et al., 2005; Hum et al., 2014; Das et al., 2013). Furthermore, deficiency of stromal genes, including Pbx1, Pod1, Sall1 and Fat4 cause similar perturbations, including expansion of SIX2⁺ cells in the condensing mesenchyme (Schnabel et al., 2003; Ohmori et al., 2015; Bagherie-Lachidan et al., 2015; Das et al., 2013; Quaggin et al., 1999). Together, these studies suggest that stromal cells interact with nephrogenic precursors to control nephron formation. Yet the identity of the factors that mediate these cell-cell interactions is largely undefined.

Hedgehog signaling is a highly conserved pathway that controls cell specification, cell-cell interactions and tissue patterning during embryonic development (Jiang and Hui, 2008; Ingham and McMahon, 2001). In mammals, there are three known hedgehog (HH) ligands, although only sonic hedgehog (SHH) is known to have a role in renal development (Yu et al., 2002; Hu et al., 2006). In the absence of SHH ligand, the patched (PTCH) family of cell-surface receptors bind to smoothened (SMO) and allow for the formation of a proteolytic complex that cleaves full-length GLI transcriptional activators GLI2 and GLI3 to transcriptional repressors (reviewed by Briscoe and Thérond, 2013; Robbins et al., 2012). When SHH ligand is present, it binds to PTCH, which in turn alleviates PTCH-mediated inhibition of SMO; this proteolytic complex is inhibited, allowing for the accumulation of full-length GLI transcriptional activators (Bai et al., 2002; Park et al., 2000). Thus, the genetic inactivation of SMO leads to obligate processing of GLI transcription factors. The ratio of GLI activators to GLI repressor determines the overall cellular response to HH stimulation. In the mammalian kidney, SHH controls early morphogenetic events, including ureteric bud growth and metanephric induction by controlling formation of the GLI3 repressor (GLI3R) (Hu et al., 2006; Blake et al., 2016). In addition, GLI-dependent signaling is crucial for expression of genes required for ureteric branching and the development of pacemaker activity by the ureter (Cain et al., 2009; Cain et al., 2011).

We demonstrate that HH-GLI signaling in Foxd1⁺ stromal cells controls the formation of capsular stromal cells in a cell autonomous fashion and nephron development in a non-cell autonomous manner. We further demonstrate that HH signaling is active in Foxd1⁺ cortical stromal cells of the embryonic kidney and that CRE-mediated deficiency of Smo in Foxd1⁺ stromal cells disrupts the formation of the renal capsule and decreases nephron number by impairing the formation of renal vesicles. RNA sequencing in Smo-deficient kidney tissue identified Tgfb2 as a downstream target of Hedgehog-GLI signaling. Neutralization and knockdown of TGFβ2 in embryonic kidney explants decreased nephron formation. CRE-mediated deletion of the TGFβ receptor Tgfbr2 in Foxd1⁺ stromal cells disrupted stromal patterning and expanded Six2⁺ nephrogenic precursors in the condensing mesenchyme, partially recapitulating the Smo-deficient phenotype. In addition to these abnormalities, deletion of Tgfbr2 in both Foxd1⁺ and Six2⁺ cells causes nephron deficiency and renal hypoplasia. Together, these results implicate HH signaling as a regulator of TGFβ2 signaling between stromal cells and nephrogenic compartments.

Previous work using a Ptch1 reporter allele in vivo suggested that HH signaling activity is limited to the ureter and medullary domain of the embryonic kidney (Cain et al., 2009). We further investigated the spatial domain of HH signaling activity using two distinct approaches. In situ hybridization demonstrated expression of Gli1, a downstream HH signaling target, in the kidney cortex, including the developing kidney capsule and the nephrogenic zone (Fig. 1A). Gli1 expression was further examined in mice in which lacZ expression was activated at the onset of ureteric branching (E11.5) via a tamoxifen-inducible Gli1Cre^ERT2 allele. X-gal staining demonstrated β-galactosidase activity in the cortical stroma surrounding the ureteric bud and nephrogenic precursors, indicating HH signaling activity in this domain, consistent with previous observations in newborn mice (Fabian et al., 2012) (Fig. 1B). To investigate the role of HH signaling in the cortical stroma, we generated mice with conditional Smo deficiency targeted to the stromal lineage using Foxd1Cre-eGFP mice (Smo-deficient mutants; Foxd1Cre;Smo^loxP/−). In situ hybridization for Gli1 in E13.5 Smo-deficient mutants confirmed a decrease in HH signaling, specifically in the outer cortex (Fig. 1C). To further verify deletion of Smo, we used eGFP reporter expression driven by the Foxd1 promoter to isolate Foxd1⁺ cells using fluorescence-activated cell sorting (FACS) (Fig. S1). Quantitative RT-PCR (qRT-PCR) analysis of E13.5 mutant FACS-isolated stromal cells demonstrated an 86% decrease in Smo mRNA expression, as well as decreased expression of the HH target genes Ptch1 and Gli1 (Fig. 1D, Fig. S2A). Decreased Smo, Gli1 and Ptch1 expression, measured in FACS-isolated stromal cells, persisted at E15.5 (Fig. S2B).

FOXD1 is crucial to development of the renal capsular stroma and stromal-nephrogenic cell interactions (Hatini et al., 1996; Levinson et al., 2005; Hum et al., 2014; Fetting et al., 2014). Accordingly, we first investigated Foxd1 expression in Smo-deficient mice. In situ hybridization at multiple embryonic stages demonstrated decreased Foxd1 expression beginning at E11.5 (Fig. 1E-H). Loss of Foxd1 expression in Smo-deficient mice was confirmed by qRT-PCR in FACS-isolated stromal cells. Foxd1 mRNA in cells from E13.5 Foxd1Cre;Smo^loxP/− mice was decreased by 60% compared with that in cells from Foxd1Cre;Smo^loxP/+ mice (Fig. 1I). Together, these data suggest that HH signaling in cortical stroma is required for the expression of the key stromal gene Foxd1.

Since Foxd1Cre is a ‘knock-in’ allele that causes heterozygous expression of Foxd1, we first established that Foxd1Cre;Smo^loxP/+ mice do not exhibit renal histological abnormalities at P0 (Fig. S3). In contrast, histological analysis of Smo-deficient kidney tissue revealed mis-patterning of the renal cortex as early as E15.5 (Fig. 2). Although stromal cells and cortical structures were normally patterned at E13.5 (Fig. 2A,B, arrows), by E15.5, the capsular stromal layer was disrupted (arrow in Fig. 2D) and nephrogenic precursors were expanded in number (Fig. 2C,D, asterisk in inset). At P0, Smo-deficient mutants exhibit severe patterning defects in the nephrogenic zone with almost complete absence of the capsular stromal layer (solid arrows in Fig. 2), an increased number of nephrogenic precursor cells in the condensing mesenchyme (dotted arrows in Fig. 2), and tubules projecting into the kidney surface (Fig. 2E,F). Smo-deficient mice died within a few hours of birth, precluding postnatal analysis of kidney structure and function.

The morphology of the renal capsule was further analyzed by scanning electron microscopy (SEM). Consistent with histological analysis (Fig. 2A-B), E13.5 mutant tissue was characterized by a honeycomb-like appearance (Fig. 2G,I), as observed in control mice. However, by E16.5, mutant mice were characterized by a discontinuous renal capsule (Fig. 2H,J, see arrows). To further characterize the state of the capsule and cortical stroma, we analyzed expression of stromal markers RALDH2 and PBX1 at both E13.5 and E15.5. Immunostaining demonstrated the presence of RALDH2⁺ and PBX1⁺ stromal cells at E13.5 in Smo-deficient mice. However, by E15.5, staining of RALDH2 was decreased in intensity and there were gaps in the continuity of RALDH2⁺/PBX1⁺ cells in the renal capsule (Fig. S4). This apparent loss in stromal cells was not associated with decreased proliferation or increased apoptosis at E13.5 (Fig. S5A-F). However, at E15.5, Smo-deficient stromal cells were characterized by a small (6%), but significant, decrease in cortical stromal proliferation, and a marked increase in apoptosis (2.4-fold). The vast majority of cortical stromal cells that were TUNEL⁺ in Smo-deficient mice were located in the space between tubule and nephrogenic elements (Fig. S5G-L).

Finally, by P0, kidney size was markedly decreased in mutant mice, associated with a 48% decrease in volume (Fig. 2K-M). Consistent with this hypoplastic phenotype, the number of nephrons at E18.5 was decreased by 42% in mutant mice (Fig. 2N). Taken together, these data demonstrate that Smo deficiency in stromal cells disrupts development of the renal stromal capsule and causes renal hypoplasia with nephron deficiency.

Histological analysis suggested that the pool of cells that give rise to nephrons is expanded in mutant mice, despite a deficiency of nephrons in the mature kidney (Fig. 2). To further investigate this apparent paradox, we analyzed the population of SIX2⁺ cells that give rise to nephrons (Fig. 3). Qualitative analysis of tissue sections stained with anti-SIX2 antibody suggests the number of SIX2⁺ cells in the condensing mesenchyme surrounding each ureteric tip is increased at both E13.5, when this change was not apparent by histology, and at E15.5 (Fig. 3A,B,D,E). Quantitative analysis of mid-sagittal sections demonstrated a 1.6-fold increase in the number of SIX2⁺ cells per ureteric tip at E15.5 (Fig. 3C). To investigate whether the expanded SIX2⁺ domain was a result of increased proliferation, we used immunostaining with anti-phospho-histone H3 (PHH3) to mark SIX2⁺ cells in an active state of cell division. Smo-deficient mutants exhibited a 42% increase in SIX2⁺ cell proliferation compared with that in controls (Fig. 3D-F). These data demonstrate that stromal-HH signaling acts in a non-cell autonomous manner to restrict the number of nephrogenic cells.

Our observation of decreased nephron number in Smo-deficient mice, despite an expansion in the number of nephron progenitors, raised the possibility that nephrogenic precursor cells are inhibited while undergoing a mesenchymal-epithelial transformation. To determine this, we analyzed the expression of Wnt9b, a ureteric cell-derived factor that is required to induce SIX2⁺ mesenchyme condensation in conjunction with Wnt4 (Park et al., 2012; Self et al., 2006; Park et al., 2007). Expression analysis using in situ hybridization demonstrated no apparent difference in Wnt9b or Wnt4 expression at E15.5 (Fig. S6). Furthermore, quantitation of the number of pretubular aggregates, identified by NCAM staining (Fig. 3G,H), demonstrated no significant difference between mutant and control mice (Fig. 3I). As the number of ureteric branches is also a determinant of nephron number, we examined this process. We observed no difference in the number of ureteric branches generated in mutant mice (Fig. S7A-C) nor in the expression of genes involved in ureteric branching, including Ret, Gdnf and Wnt11 (Fig. S7D-I).

Next, we investigated the expression of Lhx1, which controls formation of renal vesicles from pretubular aggregates, an essential step in the process of mesenchymal-epithelial transformation during nephrogenesis (Kobayashi et al., 2005). In contrast to control mice, Lhx1 expression in apparent renal vesicles and comma-shaped bodies was markedly reduced in Smo-deficient kidneys at E15.5 (Fig. 3J,K). Reduction in Lhx1 is consistent with the hypothesis that signals other than Wnt4 regulate Lhx1, as Lhx1 expression is maintained, albeit at lower levels, in Wnt4-null embryos (Kobayashi et al., 2005). Quantitation of the number of renal vesicles, marked by E-cadherin, demonstrated a 41% decrease in Smo-deficient mutants (Fig. 3I). Together, these results indicate that loss of SMO-dependent signaling in stromal cells impairs the formation of nephrogenic epithelial structures to a degree consistent with the final decrement in nephron number.

HH ligands signal via GLI-dependent and -independent mechanisms. Previously, we have identified GLI3 as the GLI family member that controls renal development in cell lineages (ureteric, metanephric mesenchyme) other than stromal cells (Cain et al., 2009, 2011; Blake et al., 2016). Here, we have investigated the contribution of Gli3 in renal stroma by crossing Smo-deficient mice with mice deficient in Gli3 (Foxd1Cre;Smo^loxP/−;Gli3^−/−). Removal of Gli3 in the Smo-deficient background rescued tissue patterning and capsular defects (Fig. 4A-C,E-G). Immunostaining for SIX2 confirmed that the expanded condensing mesenchyme phenotype observed in Smo-deficient mutants was rescued by homozygous deletion of Gli3 (Fig. 4I-K). Furthermore, Gli3 deficiency partially rescued both kidney volume and nephron number, demonstrating that GLI3 is the effector by which HH non-cell autonomous effects are observed (Fig. 4M,N). GLI3 plays dual roles in HH signaling as it exists as a full-length transcriptional activator (GLI3A) and truncated transcriptional repressor (GLI3R). The relative amount of GLI3A versus GLI3R is regulated by HH-SMO signaling. Phenotypic rescue by generating Gli3 deficiency in Smo-deficient mice suggested that GLI3R is pathogenic in this context. This possibility was further tested by introducing an obligate GLI3R targeted to stroma using Foxd1Cre (Foxd1Cre;Gli3Tflag^/+;Δ699^/+). Compound mutant mice phenocopied Smo-deficient mutants as shown by renal hypoplasia, capsule deficiency and expansion of condensing mesenchyme (Fig. 4D,H,L). Taken together, these data demonstrate that HH signaling in stromal cells exerts cell-autonomous and non-cell autonomous effects by controlling levels of GLI3R.

We hypothesized that a HH-dependent secreted factor from the stroma controls adjacent nephrogenic cells, thus controlling the epithelialization of these cells. To identify possible effectors, we performed RNA sequencing on E13.5 wild-type and Smo-deficient whole kidneys. Of the dysregulated genes, the vast majority were downregulated in mutant mice (Fig. 5A). Key genes are listed in Table 1. Smo, Ptch1 and Gli1 expression were decreased in mutant kidneys, consistent with Smo deficiency and decreased HH signaling. Expression of Foxd1 was decreased consistent with in situ hybridization and analysis of RNA expression in FACS-sorted Foxd1-positive cells (Table 1, Fig. 1). Expression of Six2 was increased consistent with our previous analyses. To identify candidate genes that could mediate stromal-nephrogenic cell interactions, the population of dysregulated genes was filtered for factors expressed in the stroma and secreted, and with a known role in epithelialization. Among the genes decreased in Smo-deficient kidney tissue (Table 1), Tgfb2 best fits these criteria in that it is secreted, expressed in the cortical stroma (Pelton et al., 1991) and is reported to have a role in epithelialization (Plisov et al., 2001). Furthermore, Tgfb2-KO mice exhibit renal agenesis with incomplete penetrance, with reduced nephron endowment (Sanford et al., 1997). Downregulation of Tgfb2, as well as Smo and Foxd1, was verified using qRT-PCR in E13.5 whole kidneys (Fig. 5B). Tgfb2 expression was further investigated in FACS-isolated Foxd1⁺ cells from E13.5 Foxd1Cre;Smo^loxP/+ (control) and Foxd1Cre;Smo^loxP/− (mutant) mice. qRT-PCR analysis demonstrated a significant decrease in Tgfb2, as well as Smo and Foxd1, in mutant stromal cells (Fig. 5C). This finding was further confirmed by in situ hybridization (Fig. 5D,E, arrows). In addition, immunostaining for pSMAD2 demonstrated a decrease in nuclear staining within nephrogenic precursor cells, but not in ureteric bud cells, in Smo-deficient mice compared with controls (Fig. 5F-I). Together, these results demonstrate that SMO-dependent signaling in stromal cells controls expression of Tgfb2 and downstream signaling events.

We investigated the functional contribution of Tgfb2 to nephron formation in cultured embryonic kidney explants. Neutralizing antibody against TGFβ2 was added to culture media of E12.5 wild-type kidney explants for 3 days after which time the explants were imaged. Treated explants were smaller and appeared abnormal compared with controls (Fig. 6A). Quantification of nephrons, marked by WT1, revealed a 65% decrease in anti-TGFβ2 treated explants compared with controls (Fig. 6B). Nephron number was decreased in a dose-dependent manner. The possibility of antibody-mediated toxicity was ruled out by the lack of inhibitory effects in control explants treated with an equivalent concentration of IgG (Fig. 6A,B). Tgfb2 function was further tested by in vivo morpholino-mediated knockdown using a splice-inhibiting morpholino (MO) specific for TGFβ2 and a control (non-targeting) MO (Eisen and Smith, 2008; Hartwig et al., 2010; Kann et al., 2015; Jacobi et al., 2013; Motamedi et al., 2014; Rudigier et al., 2017). Tgfb2 knockdown by splice-inhibiting MO was verified using qRT-PCR on treated explants (Fig. S8). Treatment with a TGFβ2 splice-inhibiting MO resulted in similar effects to that observed with neutralizing antibodies. TGFβ2 splice MO-treated explants exhibited fewer nephrons (33%) compared with control MO-treated explants (Fig. 6C,D). Together, our data provide evidence of a functional, dose-dependent role for TGFβ2 signaling in the normal differentiation of nephron progenitor cells, consistent with Tgfb2-KO mouse models.

Next, we investigated which cell population is targeted by stroma-mediated TGFβ2 signaling, resulting in the deleterious effects on nephrogenesis. We generated mice in which deficiency of the TGFβ-specific receptor, Tgfbr2, was generated in Hoxb7⁺, Six2⁺ or Foxd1⁺ cells in a CRE-dependent manner. Deletion of Tgfbr2 in Hoxb7⁺ cells did not adversely affect renal development (Fig. 7A-C). Tgfbr2 deficiency in Six2⁺ cells caused a small but statistically significant decrease (18%) in nephron number (Fig. 7D,E). Tgfbr2 deficiency in Foxd1⁺ cells (Fig. S9) did not affect nephron number, but interfered with capsule formation (Fig. 7F,G, see inset in F). Remarkably, Tgfbr2 deficiency in both Foxd1⁺ and Six2⁺ cells concurrently disrupted capsule formation accompanied by fewer nephrons (Fig. 7H,I, inset in H). Although deletion of Tgfbr2 in Hoxb7⁺ or Six2⁺ cells did not affect the number of SIX2⁺ nephrogenic progenitors per UB tip (Fig. 8A-E), Tgfbr2 deficiency in Foxd1⁺ cells or in both Foxd1⁺ and Six2⁺ cells concurrently resulted in a 28% and 20% increase in the number of SIX2⁺ cells per UB tip, respectively (Fig. 8F-I). Thus, Tgfbr2 deficiency in both Foxd1⁺ and Six2⁺ cells produced a phenotype highly similar to that observed in Smo-deficient mice. Our results in vivo and ex vivo support the hypothesis that there is TGFβ2-dependent crosstalk between stromal and nephrogenic compartments, and that HH-GLI signaling in the stromal compartment serves as the master regulator of this interaction.

Although the embryonic Foxd1⁺ renal stroma has been shown to modulate both nephrogenesis and branching morphogenesis, the molecular mechanisms that control stromal cells themselves and that mediate stromal interactions with adjacent cells are largely undefined. Here, we demonstrate that stromal HH-SMO signaling acts in a cell-autonomous manner to control renal capsule formation and, in a non-cell autonomous fashion, to control epithelialization of nephrogenic precursor cells and final nephron number. The role of GLI-dependent signaling downstream of SMO was demonstrated by a rescue of the Smo-deficient phenotype by homozygous deletion of Gli3 in Smo-deficient mice. Furthermore, targeted expression of GLI3R in Foxd1⁺ stromal cells phenocopied the hypoplastic phenotype of Smo-deficient mutants. Analysis of Smo-deficient mutants revealed that HH signaling controls expression of Tgfb2. The functional contribution of TGFβ2 was demonstrated through ex vivo experiments in which neutralization and knockdown of TGFβ2 resulted in a decrease in nephrons. These results were further substantiated by genetic experiments in vivo, in which deficiency of Tgfbr2 in Foxd1⁺ and Six2⁺ cells generated deleterious effects on both the cortical stroma (disrupted capsule and disorganized cortical stroma) and nephrogenic elements (expanded nephrogenic precursor population and fewer nephrons). Together, our results indicate that stromal HH-SMO-GLI3 signaling controls renal capsule development and nephrogenesis in part through TGFβ2.

Progressive loss of the renal capsule in Smo-deficient mutants suggests a role for HH signaling in regulating the renal capsule. Development of the renal capsule is poorly understood, in part due to the lack of reliable and specific capsule markers. Kobayashi and colleagues demonstrated that the capsule is derived from Foxd1⁺ stromal progenitors (Kobayashi et al., 2014). Deficiency of Foxd1 does not abrogate the capsular stroma (Levinson et al., 2005). Rather, Foxd1 helps to specify the molecular identity of capsular stroma. In mice with homozygous deficiency of Foxd1, capsular cells are present but ectopically express genes including Bmp4 (Levinson et al., 2005). Our data define a novel molecular pathway that is required for specification of the cortical signaling niche of the kidney. Here, we show that HH-GLI3 dependent signaling acts upstream of Foxd1 and controls maintenance of capsule stromal cells in a TGFβ2-TGFβR2-dependent manner. Our findings of decreased Foxd1 in Smo-deficient stromal cells suggest that disruption of the capsule in the more mature kidney is due to failure to generate a sufficient number of cells to fully constitute a capsule, a modest decrease in stromal proliferation and increased apoptosis. The difference in capsule phenotype between Smo-deficient and Foxd1-deficient mice suggests that HH signaling controls additional mechanisms that control capsule formation. Indeed, we show that loss of HH signaling decreases Tgfb2 in stromal cells. Furthermore, our in vivo analyses show that deficiency of Tgfbr2 in stromal cells also disrupts formation of a complete capsule. These results provide a basis for further studies aimed at exploring the cortical signaling niche and its involvement in renal capsule formation.

A prominent phenotypic feature Smo-deficient mice is the increase in proliferation and number of SIX2⁺ cells in the condensing mesenchyme preceding the decrease in nephron number. An increase in the number of SIX2⁺ cells in the condensing mesenchyme is a consistent finding in mice with stromal deficiency of Fat4 or Sall1, as well as mice with germline deficiency of Foxd1, Pod1 or Pbx1 (Schnabel et al., 2003; Bagherie-Lachidan et al., 2015; Das et al., 2013; Ohmori et al., 2015; Levinson et al., 2005; Quaggin et al., 1999). These observations, together with our data, suggest a complex signaling environment between the stroma and adjacent SIX2⁺ mesenchyme.

Functional evidence to define the key signaling mediators is limited, as many stromal models have explored the genetic deletion of transcription factors, which are neither secreted nor act as receptors and likely act indirectly to control stromal-nephrogenic crosstalk. Investigation of Smo-deficient mutants demonstrated that TGFβ2, a secreted factor expressed in the stroma, is expressed downstream of HH-SMO signaling. Evidence of signaling between Hedgehog and TGFβ is limited, but has been explored in other contexts, i.e. bone chondrocyte differentiation, where SHH acts upstream of Tgfb2 (Alvarez et al., 2002). In renal development, Tgfb2-null mice exhibit incompletely penetrant renal aplasia, similar to Shh-null mice, suggesting similarities in the phenotypic mechanisms of these models (Sanford et al., 1997; Hu et al., 2006).

We show that kidney explants treated with TGFβ2 neutralizing antibody or targeted morpholino have fewer nephrons, which further implicates TGFB2 in nephrogenesis. Our analysis of genetic mouse models sheds further insight into how Tgfbr2-dependent signaling may control nephrogenesis. Deletion of Tgfbr2 in Foxd1⁺ stromal cells results in an increase in the number of SIX2⁺ cells in the condensing mesenchyme, similar to that observed in Smo-deficient mutants. Yet, deficiency of Tgfbr2 in SIX2⁺ nephrogenic cells has no detectable effect on the number of SIX2⁺ cells in the condensing mesenchyme, suggesting that Tgfbr2-dependent signaling is not sufficient to control SIX2⁺ cells (Fig. 9, red dotted arrow). Interestingly, removal of TGFβ signaling from both stromal and nephrogenic cells concurrently results in the accumulation of SIX2⁺ cells in mesenchymal aggregates and low nephron number. These data suggest that the primary function of TGFβ signaling in this niche is intrinsic to Foxd1⁺ cells; deficiency of TGFβ signaling in stromal cells abrogates the release of one or more downstream factors that control SIX2⁺ cells (Fig. 9). Published work has demonstrated that the atypical cadherin Fat4 with receptor Dchs1, as well as the secreted protein decorin, directly control crosstalk between stromal-nephrogenic cells (Mao et al., 2015; Bagherie-Lachidan et al., 2015; Das et al., 2013; Fetting et al., 2014). It is possible that HH-SMO-TGFβ signaling is upstream of these other signaling pathways that links stromal and nephrogenic cells.

Our data provide a novel functional role for TGFβ signaling in renal development. Previous research has implicated TGFβ signaling in controlling branching morphogenesis and nephrogenesis ex vivo (Bush et al., 2004; Clark et al., 2001). TGFβ2 has been shown specifically to induce nephrogenesis ex vivo, although induction appeared to be dose dependent (Plisov et al., 2001). Investigation of Tgfb2^+/− mice has yielded conflicting results with respect to increasing or decreasing branching morphogenesis and nephrogenesis, whereas Tgfb2-null mice have fewer nephrons (Sanford et al., 1997; Sims-lucas et al., 2008; Short et al., 2013; Short et al., 2010). Our ex vivo studies confirm that TGFβ2 is required for nephrogenesis in a dose-dependent manner based on neutralization and knockdown at various concentrations. Furthermore, we provide new-found evidence through our genetic mouse models that TGFβ-dependent signaling in stromal and nephrogenic cells is required to generate a full complement of nephrons. In addition, we also demonstrate that TGFβ signaling is crucial in the formation of the renal capsule and for normal stromal patterning, although the mechanism behind this process remains to be defined.

Our model (Fig. 9) posits that canonical HH-SMO-GLI signaling in stromal cells controls expression of Tgfb2 in the stroma. Our work adds to and emphasizes the complex signaling interactions that are present in the renal stroma and cortex of the embryonic kidney, and adds to the growing body of evidence that stromal cells influence nephrogenic cells through molecular crosstalk. Although HH-SMO-GLI signaling acts cell autonomously to control capsular integrity, it also acts non-cell autonomously to restrict the SIX2⁺ nephrogenic cell pool and control nephrogenesis, in large part through TGFβ signaling. This body of work will aid in the understanding of how stromal cells signal, function and ultimately control the complex process of kidney development.

All experimental protocols involving mice were approved by the Animal Ethics Committee at the Hospital for Sick Children and were carried out in accordance with the regulations of the Canadian Council on Animal Care. All animal work was conducted at The Centre for Phenogenomics (TCP, Toronto, Canada) following all animal use protocols.

To generate mice in which Smo inactivation is restricted to the stroma (Foxd1Cre;Smo^loxP/–), Foxd1Cre-eGFP mice (herein labelled as Foxd1Cre) were mated with Smo^+/– mice to generate Foxd1Cre;Smo^+/– (Humphreys et al., 2010). These mice were subsequently mated with Smo^loxP/loxP mice (Zhang et al., 2001; Long et al., 2001). Smo^loxP/+ mice were used as controls unless otherwise specified. Foxd1Cre;Smo^+/– and Smo^loxP/loxP mice were mated with Gli3^+/– to generate Foxd1Cre;Smo^+/−;Gli3^+/– and Smo^loxP/loxP;Gli3^+/– progeny, respectively (Litingtung et al., 2002). These two mouse lines were mated with one another to generate Foxd1Cre;Smo^loxP/–;Gli3^–/– mice (Maynard et al., 2002). Gli3^Δ699/+ mice were mated with Foxd1Cre mice to generate Foxd1Cre;Gli3^Δ699/+ mice, which were subsequently mated with RosaGli3TFlag^c/c to generate Foxd1Cre;RosaGli3TFlag^/+;Gli3^Δ699/+(Vokes et al., 2004; Böse et al., 2002). RosaGli3TFlag^/+ mice were used as controls unless otherwise specified. To label Gli1-expressing cells, Gli1Cre^ERT2 mice (Ahn and Joyner, 2004) were mated with Rosa26^LacZ reporter mice (Friedrich and Soriano, 1991). Pregnant dams were injected with tamoxifen at E11.5 and embryos dissected at E14.5. Foxd1Cre mice were mated with Tgfbr2^loxP/loxP mice (Levéen et al., 2002) to generate Foxd1Cre;Tgfbr2^loxP/+ mice, which were bred with Tgfbr2^loxP/loxP to generate Foxd1Cre;Tgfbr2^loxP/loxP and Foxd1Cre;Tgfbr2^loxP/− mutant mice. Hoxb7Cre mice (Zhao et al., 2004) and Six2Cre mice (Kobayashi et al., 2009) were mated with Tgfbr2^+/− mice to generate Hoxb7Cre;Tgfbr2^+/− and Six2Cre;Tgfbr2^+/− mice. These mice were mated with Tgfbr2^loxP/loxP mice to generate Hoxb7Cre;Tgfbr2^loxP/− and Six2Cre;Tgfbr2^loxP/− mutants. Foxd1Cre;Six2Cre;Tgfbr2^loxP/− compound mutants were generated using the above crosses. Mice were genotyped using the primers listed in Table S1.

Pregnant females were sacrificed using carbon dioxide. Embryonic (E) day 0.5 was set to noon the day the vaginal plug was found. Embryonic kidneys were dissected and fixed in 4% paraformaldehyde (PFA) overnight at 4°C. PFA-fixed kidneys were transferred to 70% ethanol the following day before paraffin wax embedding and sectioning. All tissue processing, sectioning and, Hematoxylin and Eosin staining was carried out by the CMHD Core Pathology Lab at TCP, Toronto, Canada.

Whole kidneys were briefly fixed in lacZ fix solution (25% glutaraldehyde, 100 nM EGTA, 1 M MgCl₂, 0.1 M sodium phosphate) and rinsed in wash buffer (0.1 M sodium phosphate buffer, 2% nonidet-P40, 1 M MgCl₂). Kidneys were then placed in X-gal staining solution at 37°C overnight in the dark. Stained kidneys were then immersed in wash buffer and post-fixed in 10% buffered formalin at 4°C. Whole kidneys were photographed using a Leica EZ4D dissecting microscope, processed for embedding in paraffin wax, then sectioned at 5 μm. Sections were counterstained with nuclear Fast Red.

Immunostaining was performed on PFA fixed, paraffin wax-embedded kidney sections. Sections (5 μm) were deparaffinized in xylene and rehydrated through an ethanol gradient. Antigen retrieval was performed in sodium citrate buffer (pH 6.0) in a microwavable pressure cooker.

Primary antibodies (Table S2) were incubated overnight at 4°C. Following PBS washes, slides were incubated in secondary antibody (Table S2), washed and mounted using VectaShield mounting medium (Vector Labs). Mounted slides were imaged using an epifluorescence microscope (Zeiss) or a Quorum Spinning Disk Confocal (Zeiss AxioVert 200 M). A minimum of three sections per kidney and at least three biological replicates were used in the assessments. Fixed explants were rehydrated in PBS before permeabilized using PBS with 1% Triton (PBST). Explants were then incubated with primary antibody overnight at 4°C. Following washes in PBST, explants were incubated in secondary antibody (1:500), washed in PBS and immediately imaged on a Quorum Spinning Disk Confocal (Zeiss AxioVert 200 M).

Following immunostaining, SIX2⁺ positive nephrogenic precursor cells were counted. Three to five representative fields of SIX2⁺ cells were counted per kidney; two kidneys were stained per animal (n=6 for control and mutants compared). Data were presented as the number of SIX2⁺ cells per ureteric bud tip. For proliferation analysis, double immunostaining for SIX2 and phospho-histone-H3 was used (n=5 for control and mutants compared). A proliferative index of the nephrogenic precursor cells was determined by comparing the number of double positive SIX2⁺ cells to the total number of SIX2⁺ cells.

Quantification of apoptosis at E13.5 was performed using TdT In Situ Apoptosis Detection Kit (R&D Systems, 4810-30-K). Methyl Green was used as a counter stain. Staining was performed using a standard protocol supplied with the kit (Trevigen) and was performed by the CMHD Pathology Core at The Centre for Phenogenomics. TUNEL⁺ cortical stromal cells identified by brown DAB staining were counted and normalized to kidney area in equivalent, mid-sagittal sections. The area was measured using AxioVision 4.6.3-SP1 (Zeiss).

Pregnant dams were injected with 10 mg/ml solution of BrdU (Roche 10280879001), 10 μl per gram of body weight. Embryos were dissected out after 2 h and kidneys collected. Embryonic kidneys were fixed and sectioned as described above. Sections were stained with an anti-BrdU antibody (Table S2). An index of cortical stromal proliferation was determined by comparing the number of BrDU⁺; PBX1⁺ cortical stromal cells with the total number of PBX1⁺ cortical stromal cells.

Section in situ hybridization was performed as previously described (Ding et al., 1998). DIG-labeled riboprobes encoding Gli1, Foxd1, Lhx1, Tgfb2, Ret, Gdnf, Wnt11, Wnt4 and Wnt9b were used.

E18.5 kidneys were fixed, embedded in paraffin wax and sectioned in the sagittal plane. In order to quantify glomeruli number and estimate kidney volume, the samples were processed according to previous procedures (Bertram et al., 1992; Cain et al., 2009). For counting glomeruli, slides were stained with PAS, where the sum of all glomeruli in the sections making up the kidney was used as an estimate of the total nephron number per kidney. For volumetric analysis, the area of tissue sections were measured with AxioVision 4.6.3-SP1 (Zeiss) and multiplied by 100 μm to calculate the volume of each representative section. Total kidney volume was calculated as the sum of volumes for each section.

Whole E16.5 kidneys were harvested and fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.3) overnight at 4°C. Fixed kidneys were given to The Advanced Bioimaging Centre at Mount Sinai Hospital to be dehydrated, dried and gold sputter coated. Prepared samples were imaged at the Bioimaging Centre with a FEI XL30 ESEM microscope.

RNA was extracted from either E13.5 whole kidney tissue or FACS isolated stromal cells using the RNeasy Micro RNA Extraction Kit (Qiagen). Following extraction, a standard amount of RNA from both control and mutant samples was reverse transcribed into cDNA using the SuperScript First Strand Synthesis System for RT-PCR (Invitrogen). Quantitative PCR was performed using gene specific cDNA primers and normalized to β-2 microglobulin expression before control and mutant samples were compared. Comparisons were performed using the ΔΔCt method. At least three biological and three technical replicates were used per gene assayed.

Individual kidneys from three independent E15.5 litters were prepared and sectioned at 5 µm increments, starting from the midline of the kidney. To minimize section bias, equivalent sections from wild-type and mutant kidneys were immunostained for E-cadherin and NCAM. At least two sections per kidney (either wild type or mutant) were counted. Six wild-type and mutant biological replicates were used in the analysis across two individual separate experiments.

Biological replicates from four separate E13.5 litters were used for RNA sequencing analysis. E13.5 kidneys were dissected out of embryos and kept in RNAlater reagent. Two kidneys were pooled for each biological replicate (n=4 for control and mutant). RNA was isolated using the RNeasy Micro RNA Extraction Kit (Qiagen); RNA purity and yield was determined using the Agilent RNA 6000 Pico Lab Chip analysis prior to cDNA library preparation. All samples had RIN>9.4. RNA samples were submitted to the TCAG for RNA Sequencing analysis on the Illumina HiSeq 2500 system.

Following cDNA library preparation, whole kidney transcript reads were aligned to the mouse genome (mm10) using TopHat. HTSeq was used to determine the absolute number of read counts for genes. All sequence alignment and read counts were performed by TCAG. Subsequent analysis of differentially regulated genes was carried out using R/Bioconductor package using DEseq following a standard pipeline of normalization followed by differential comparison of wild-type and mutant samples (Anders et al., 2013). A full list of candidate genes from RNA sequencing is available in Table S3. All RNA sequencing data has been deposited in the Gene Expression Omnibus (GEO) repository, accession number GSE103923.

For comparative analysis, control (Foxd1Cre;Smo^loxP/+ or Foxd1Cre;Tgfbr2^loxP/+) stromal cells were collected and compared with mutant stromal cells (Foxd1Cre;Smo^loxP/− or Foxd1Cre;Tgfbr2^loxP/loxP). Isolated kidneys were dissected in Dulbecco's Modified Eagle Medium (DMEM)/Hams-F12 culture media (Gibco) and digested in 2 mg/ml of collagenase B (Roche) in DEPC/Ringer's solution for 15 min at 37°C. After brief centrifugation, cells were re-suspended in PBS containing 1% heat-inactivated FCS and 1 mM EDTA. The cell suspension was filtered through a 40 μM nylon mesh cell strainer (BD Falcon). Propidium iodide (Sigma) was added at a final concentration of 0.5 mg/ml to label dead cells. FACS analysis was carried out using MoFlo Astrios BRVY (Beckman Coulter) or FACS Vantage (Becton Dickinson Immunocytometry Systems). Dead cells were excluded from live FACS plots based on propidium iodide. GFP-positive cells were sorted directly into Buffer RLT (RNeasy Micro Kit, Qiagen) and stored at −80°C.

E12.5 C57BL/6J kidneys were dissected out of embryos and placed onto transwell filters (0.4 µm, Falcon) in six-well plates immersed in 2 ml DMEM/F12 supplemented with FBS, 50 µg/ml transferrin (Sigma) and either: 1, 2 or 4 µg/ml anti-TGFβ2 neutralizing antibody (Abcam, ab10850); 4 µg/ml rabbit IgG (Cell Signaling, 2729S); 3 µM control MO (5′CTACTAATTGTCACAGCCGTCACGT) or 3 µM TGFβ2 splice inhibiting MO (5′ CCCACAACCAGAGCACAGGTTACCT). Explants were incubated at 37°C with 5% CO₂ and imaged on a light microscope (Zeiss). After treatment, explants were either fixed in ice-cold 4:1 methanol:DMSO for whole-mount immunofluorescence or flash frozen in lysis buffer for RNA extraction.

All statistical analyses were performed using GraphPad Prism software (Version 7.0). Student's t-test (two-tailed) was used to determine significance between mean differences in groups. P≤0.05 was considered to be statistically significant. Error bars indicate s.e.m.

We thank the staff at The Centre for Phenogenomics CMHD core for all histology work. We thank Dr Sunny Hartwig (University of Prince Edward Island) and Dr Jon Moulton (Gene Tools) for assistance in designing morpholinos. We also thank Dr Daniel Schramek (Mount Sinai Hospital) for the Tgfbr2 mice, Dr Carlton Bates (Children's Hospital of Pittsburgh) for the Hoxb7Cre mice, Dr C. C. Hui (The Hospital for Sick Children) for the Gli1CreERT2 and Gli3^−/− mice, Dr Andrew McMahon (University of Southern California) for the Six2Cre mice and Dr Ulrich Ruthar (University of Dusseldorf) for the Gli3^Δ699 mice. We also thank Dr Helen McNeill (Mount Sinai Hospital) for the Tgfb2 in situ hybridization probe.