FGFR2 signaling enhances the SHH-BMP4 signaling axis in early ureter development

The urothelium is a stratified epithelium that lines the inner surface of the urinary drainage system. In the ureter and bladder, it consists of three major cell types that are organized in radial layers of variable thickness. Large binucleated superficial (S-) or umbrella cells border the lumen and primarily account for the essential barrier function of the tissue. They are sealed by tight junctions and are covered by crystalline plaques of specialized surface proteins: uroplakins (UPKs). Underneath, much smaller intermediate (I) cells form one to several layers depending on species and organ, and serve as precursors for S and basal (B) cells. These small B cells abundantly express keratin 5 (KRT5) and provide an anchor to the basal lamina and the surrounding fibro-muscular wall (Dalghi et al., 2020; Wang et al., 2017).

This urothelial cytoarchitecture derives from highly coordinated proliferation, patterning and differentiation processes that act on epithelial progenitors (the cloacal epithelium in the bladder, the distal ureteric bud in the ureter) starting around embryonic day (E) 10.5 of mouse development (Wang et al., 2017; Yamany et al., 2014). After an initial phase of proliferative expansion, the mono-layered epithelial primordia stratify concomitant with the expression of the transcription factor ΔNP63, indicating I-cell differentiation. After 2 days, the adluminal layer starts to express S-cell markers. B-cell differentiation occurs 2 additional days later (Bohnenpoll et al., 2017a; Gandhi et al., 2013).

Urothelial development in the bladder and ureter does not occur in a cell-autonomous fashion but requires signaling input from adjacent mesenchymal cells, which, in turn, depend on signals from the epithelial primordium to develop into a layered fibro-muscular wall (Balsara and Li, 2017; Bohnenpoll and Kispert, 2014; Cunha et al., 1991; Wang et al., 2017). To date, members of three classes of secreted proteins have been characterized as essential mesenchymal signals for urothelial development: bone morphogenetic protein 4 (BMP4), retinoic acid (RA) and fibroblast growth factors (FGFs). Analysis of conditionally mutant mice showed that Bmp4 is required in the ureteric mesenchyme (UM) for stratification and cytodifferentiation of the adjacent ureteric epithelium (UE) (Mamo et al., 2017). Expression of Bmp4 in the UM depends on the transcription factor FOXF1, which, in turn, requires input from an epithelial sonic hedgehog (SHH) signal. SHH acts through this FOXF1-BMP4 axis to control not only epithelial differentiation but also survival, proliferation and smooth muscle cell (SMC) differentiation of surrounding mesenchymal cells, thereby coupling the development of the two tissues (Bohnenpoll et al., 2017c; Yu et al., 2002). Bmp4 expression receives an additional input from the canonical (CTNNB1-dependent) branch of WNT signaling triggered by WNT ligands from the UE (Trowe et al., 2012). RA has been found to prevent the differentiation of B and S cells in ureter explant cultures (Bohnenpoll et al., 2017b). In the bladder, loss of RA signaling led to a single-layered epithelium with B cells, indicating that, in this context, RA signaling is required for S-cell specification (Gandhi et al., 2013).

Embryos deficient for Fgf7 exhibit a thinning of the bladder urothelium, particularly of the I-cell layers. In vitro, FGF7 stimulated I-cell proliferation and delayed their differentiation into S cells (Tash et al., 2001). Furthermore, FGF10 was shown to act as a mitogen for urothelial cells (Bagai et al., 2002; Zhang et al., 2006). How these mesenchymal FGF signals are transmitted to the epithelium and what targets their signaling pathway has is unknown, as is the interaction of FGF signaling with other signaling systems in this context.

Here, we set out to analyze the role of FGF signaling in urothelial development using the murine ureter as a model. We provide genetic evidence that FGFR2 signaling enhances SHH-BMP4 signaling activity, which is essential for epithelial and mesenchymal proliferation, and differentiation in this organ.

Previous work described expression of Fgf7 and Fgf10 in the early bladder mesenchyme, while expression of Fgfr1 and Fgfr2 was found in the adjacent cloacal epithelium (Cancilla et al., 1999; Dudley et al., 1999; Peters et al., 1992; Tash et al., 2001). To determine whether the expression of these FGF signaling components is conserved in ureter development, we performed in situ hybridization analysis on transverse sections of the trunk region of E12.5 to E18.5 wild-type embryos (Fig. 1). At E12.5, both Fgfr1 and Fgfr2 were expressed in the UE and UM, Fgfr2 more strongly and enhanced in the UE. Expression of both genes continued at lower levels at E14.5 and disappeared in both tissues until E18.5. Expression of Fgf7 and Fgf10, the encoded proteins of which predominantly bind to the epithelial (IIIb) isoform of FGFR2 (Igarashi et al., 1998; Jans, 1994; Ornitz and Itoh, 2015), occurred weakly in the UM, particularly at E14.5 (Fig. 1A). We did not detect specific expression of other FGF ligand genes in the UM or the UE from E12.5 to E16.5 (Fig. S1). Importantly, Spry1 and Spry2, target genes of the FGF signaling pathway (Hanafusa et al., 2002), were strongly expressed in the UE at E12.5 and at E14.5 (Fig. 1B). These findings suggest that mesenchymal FGF7 and FGF10 predominantly activate epithelial FGFR2 signaling in early ureter development.

To explore the specific role of Fgfr1 and/or Fgfr2 in the UE, we used a conditional gene inactivation approach with floxed alleles of Fgfr1 and Fgfr2 (Hoch and Soriano, 2006; Yu et al., 2003), and a Pax2-cre line that mediates recombination in the nephric duct, the precursor of the UE and of the renal collecting duct system (Bohnenpoll et al., 2017a; Trowe et al., 2011). We mated Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/+ males with Fgfr1^fl/fl;Fgfr2^fl/fl females and analyzed the genotype distribution at different time points of embryogenesis. At all stages, Pax2-cre/+;Fgfr1^fl/fl;Fgfr2^fl/+ embryos were found at approximately one-half of the expected frequency, and Pax2-cre/+;Fgfr1^fl/fl;Fgfr2^fl/fl embryos at a quarter, indicating that homozygous loss of Fgfr1 accounts for lethality before E12.5, which is further enhanced by removal of Fgfr2 function (Table S1). Notably, expression of Spry1 and Spry2 was strongly reduced at E12.5 and E14.5 in the UE of embryos with loss of two alleles of Fgfr2, indicating that FGFR1 does not contribute in a major fashion to FGF signaling in this tissue (Fig. S2).

Morphological inspection of whole urogenital systems at the end of embryonic development, at E18.5, revealed that conditional loss of two and more alleles of Fgfr1 and Fgfr2 led with variable severity and penetrance to sex-independent hydroureter formation (Fig. 2A; Table S2A). Approximately 40% of Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/+ (n=32) and 30% of Pax2-cre/+;Fgfr1^fl/fl;Fgfr2^fl/+ urogenital systems (n=11) presented with mild unilateral hydroureter, whereas Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/fl (n=26) and Pax2-cre/+;Fgfr1^fl/fl;Fgfr2^fl/fl (n=8) urogenital systems had an increased occurrence (∼85%) of strong bilateral hydroureter. In the last two genotypes, we detected one case each of ureter/kidney agenesis. Loss of both alleles of Fgfr2 (Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/fl; Pax2-cre/+;Fgfr1^fl/fl;Fgfr2^fl/fl) was additionally affected with uni- or bilateral dilatation of the epididymis, while kidney size and ureter length was strongly reduced in Pax2-cre/+;Fgfr1^fl/fl;Fgfr2^fl/fl urogenital systems only (Fig. 2A; Table S2A). Histological analysis confirmed hydroureter formation upon loss of two or more alleles of Fgfr1 and/or Fgfr2; however, this did not translate into hydronephrosis in any of the genotypes (Fig. 2B, Fig. S3A).

To test for patency of the ureter and its junctions, we injected ink into the renal pelvis of isolated urogenital systems and observed its flow to the bladder upon mild hydrostatic pressure. In most of the embryos with conditional loss of two or three alleles of Fgfr1 and/or Fgfr2 (Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl//+; Pax2-cre/+;Fgfr1^fl/fl;Fgfr2^fl/+; Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/fl), the ureteric lumen was contiguous and the distal ureter inserted normally in the dorsal bladder neck. In 60% of Pax2-cre/+;Fgfr1^fl/fl;Fgfr2^fl/fl urogenital systems (n=5), the ink did not reach the bladder, either due to insertion of the distal ureter into the urethra (1 out of 5) or due to ureteropelvic junction obstruction (n=2) (Table S2B). Histological analysis of the ureter-bladder connection of these specimens confirmed these findings (Fig. S3B). We conclude that loss of Fgfr2 is associated with strong hydroureter formation. Additional loss of Fgfr1 contributes to kidney hypoplasia and to increased physical obstruction along the ureter and its junctions.

We next used immunofluorescence analysis of marker proteins to judge cytodifferentiation of the epithelial and mesenchymal tissues of the ureter (Fig. 2C, columns 1-5). Expression of CDH1, a marker of the lateral-basal membrane of epithelial cells, was found in all mutants but the epithelium appeared mono-layered in mutants with loss of two alleles of Fgfr2. Expression of KRT5, ΔNP63 and UPK1B combinatorially marked B cells (KRT5⁺ΔNP63⁺UPK1B⁻), I cells (KRT5⁻ΔNP63⁺UPK1B⁺) and S cells (KRT5⁻ΔNP63⁻UPK1B⁺) (Bohnenpoll et al., 2017c) in the control and in mutants with loss of one allele of Fgfr2. In mutants with a complete loss of Fgfr2, the mono-layered epithelium expressed the S-cell marker UPK1B, whereas KRT5- and ΔNP63-expressing B and I cells were largely absent. Expression of markers for SMCs (ACTA2 and TAGLN) and the tunica adventitia (Fbln2) occurred in the mesenchymal wall of all mutants; expression of a marker of the lamina propria (ALDH1A2) was absent in ureters with complete loss of Fgfr2 function (Fig. 2C,D, columns 1-5).

Loss and/or reduction of ΔNP63, KRT5 and ALDH1A2 expression was also detected in a rare undilated Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/fl ureter, confirming the dilatation-independent nature of these changes (Fig. 2, column 6). We conclude that loss of Fgfr2 in the UE compromises differentiation of I and B cells but also affects the development of lamina propria fibrocytes in the UM.

To define both the onset as well as the progression of cellular defects in ureters with complete loss of Fgfr2, we analyzed earlier embryonic stages (Fig. 3). We used Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/fl (from now on termed Fgfr2cKO) embryos for this and all subsequent assays, as they exhibited the same ureteric cytodifferentiation defects as Pax2-cre/+Fgfr1^fl/fl;Fgfr2^fl/fl embryos but presented in a normal Mendelian ratio. Fgfr2cKO ureters exhibited a clear histological division of the UM into an inner layer with rhomboid-shaped condensed cells and an outer layer with loosely organized fibroblast-like cells from E12.5 onwards, as in the control. However, both the UE and the UM were hypoplastic (Fig. 3A). The UE appeared less stratified at E14.5 and subsequent stages, and did not activate expression of ΔNP63 and KRT5. Expression of UPK1B occurred normally from E15.5 onwards. Expression of SMC markers was delayed by 1 day (Fig. 3B, Fig. S4 for higher magnification images of histological and CDH1 staining).

Given the obvious tissue hypoplasia in Fgfr2cKO ureters from E12.5 onwards, we analyzed whether changes in apoptosis and/or proliferation may be causative. In fact, the TUNEL assay detected apoptotic cells in the outer region of the UM at E12.5 (Fig. 3C). Moreover, the BrdU incorporation assay revealed strongly reduced proliferation in the UE of mutant embryos at this stage (Fig. 3D,E; Table S3). Hence, epithelial FGFR2 signaling plays a crucial role in epithelial proliferation, stratification and I-/B-cell differentiation, and (indirectly) in mesenchymal apoptosis and differentiation.

We next performed transcriptional profiling by microarray analysis of E13.5 Fgfr2cKO and control ureters to identify molecular changes that may underlie the cellular defects in these mutants. Using an intensity threshold of 100 and fold changes of at least 1.5, we identified 97 genes that were consistently upregulated and 49 genes that were downregulated in Fgfr2cKO ureters (Fig. 4A; Tables S4 and S5). Functional annotation by DAVID did not find enrichment of meaningful terms in the list of upregulated genes (Table S6). However, in the pool of downregulated genes terms associated with SHH/SMO activity and AP1 signaling were over-represented (Table S7).

In fact, among the most downregulated genes were Hhip, Ptch1 and Foxf1, which have previously been found to depend on SHH signaling in the ureter (Bohnenpoll et al., 2017c), Shh itself and AP1/immediate early genes (Fosb, Egr1, Fos and Egr2). Moreover, Hoxb8, a gene linked to proliferation control (Guo et al., 2019; Wang et al., 2019), was strongly reduced as was Aldh1a3, a gene encoding an RA-synthesizing enzyme, and Elf5 (Fig. 4A), an epithelial target of RA signaling in the ureter (Bohnenpoll et al., 2017b). Spry1 and Spry2, direct targets of FGF signaling (Hanafusa et al., 2002), were reduced confirming our previous analysis (Fig. S2).

We used in situ hybridization analysis to validate these changes in E12.5 and E14.5 Fgfr2cKO ureters. We detected reduced expression of Hoxb8, Aldh1a3 and Elf5 in the UE at E12.5 (Fig. 4B,C). Expression of Shh was strongly reduced in the UE, as was expression of Ptch1 and Foxf1 in the UM at E12.5 and E14.5 (Fig. 4D). Expression of Fosb, Egr1 and Fos was not detected in the control or was unchanged in the mutant (Fig. S5).

Given the strong downregulation of the Shh-Foxf1 axis, we also analyzed expression of the effector gene of this pathway: Bmp4 (−1.2 in the microarray) (Mamo et al., 2017). Bmp4 expression was clearly reduced in the UM both at E12.5 and E14.5. Moreover, expression of Id genes (Id2, Id3 and Id4), direct transcriptional targets of BMP signaling (Hollnagel et al., 1999; Liu and Harland, 2003), was reduced both in the UE and UM of Fgfr2cKO embryos at E14.5 (Fig. 4E). As BMP4 signaling is mediated by different cytoplasmic effector proteins in the developing ureter (Mamo et al., 2017), we analyzed their activated, i.e. phosphorylated, forms by immunohistochemistry. We found reduced expression of P-SMAD1/5/9 in the UM, and of P-AKT in the UE at E14.5, while P-ERK1/2 and P-P38 expression was unaffected (Fig. 4F; Fig. S6).

In agreement with the microarray data, other signaling systems involved in ureteric proliferation and differentiation were either not or only marginally affected. Expression of mesenchymal RA signaling components (Aldh1a2) and targets (Rarb) (Mendelsohn et al., 1991) was unchanged (Fig. S5). Expression of Wnt7b was weakly reduced in the UE at E12.5; Wnt9b expression in the UE was normal; expression of Axin2 was weakly reduced in the UM (Fig. 4G, Fig. S5).

Finally, RT-qPCR analysis confirmed significantly reduced expression of Spry1, Shh, Ptch1, Foxf1, Bmp4 and Elf5 in E14.5 Fgfr2cKO ureters (Fig. 4H; Table S8). We conclude that the Shh-Foxf1-Bmp4 axis is strongly affected by loss of epithelial Fgfr2 in the early ureter, whereas RA and WNT signaling are only partly and weakly compromised.

To test the individual contribution of reduced SHH, BMP4, RA and WNT signaling activity to the proliferation and cytodifferentiation defects of Fgfr2cKO ureters, we performed pharmacological pathway rescue experiments in ex vivo cultures (Fig. 5). As we were not able to confirm expression (changes) of AP1 components and targets in the UM of mutants (Fig. S5), we excluded this pathway from further investigation.

Fgfr2cKO ureters explanted at E13.5 and cultured for 4 days in minimal medium (DMEM only) exhibited a short ureter with epithelial tissue hypoplasia. Moreover, the ratio of ΔNP63⁺ to CDH1⁺ cells dropped to around 50% (control: 90%), the thickness of the SMC layer (marked by TAGLN) was reduced and expression of the lamina propria marker ALDH1A2 was nearly absent, highly reminiscent of the changes observed in vivo. Addition of 2 µM of the SHH signaling/SMO agonist purmorphamine (Li et al., 2008) or of 100 ng/ml BMP4 increased the percentage of ΔNP63⁺ cells in Fgfr2cKO ureter explants almost to control level and rescued epithelial hypoplasia. In the case of reactivation of the SHH signaling pathway, mesenchymal aspects of the Fgfr2cKO phenotype were also rescued: expression of ALDH1A2 was restored as was the thickness of the SMC layer. Addition of BMP4 rescued the mesenchymal phenotype partly. Addition of 1 µM RA enhanced the loss of ΔNP63⁺ and ALDH1A2⁺ cells, and did not rescue the tissue hypoplasia of the mutant ureter. The WNT agonist BIO (Sato et al., 2004) rescued the epithelial hypoplasia but left all other parameters in Fgfr2cKO ureters unaffected (Fig. 5; Table S9). Hence, SHH and BMP4 signaling are functional mediators of FGFR2 activity in the UE.

Previous work has shown that the genetic ablation of SHH and BMP4 signaling leads to tissue hypoplasia and a complete lack of cytodifferentiation (Bohnenpoll et al., 2017c; Mamo et al., 2017). To explore the consequences of a partial reduction of these signaling activities for ureter development, we explanted E13.5 wild-type ureters and cultured them for 4 days in DMEM supplemented with increasing concentrations of the SMO antagonist cyclopamine (Chen et al., 2002; Cooper et al., 1998) or noggin (NOG), which sequesters BMP4 from its receptor (Zimmerman et al., 1996). In both cases, we detected a dose-dependent decrease of stratification and of ΔNP63⁺ cells, a reciprocal increase of luminal ΔNP63⁻ cells that were lined by UPK expression at low and medium doses, and an ablation (cyclopamine) or reduction (NOG) of the SMC layer. At the highest doses of NOG, UPK expression was partially reduced. Expression of ALDH1A2 was strongly affected by mild reduction of SHH signaling, whereas reduction of BMP4 signaling had a weaker but dose-dependent effect (Fig. 6A-E; Fig. S7 for higher magnification images of histological and CDH1 staining; Table S10). We conclude that reduction of SHH or BMP4 signaling largely recapitulates the phenotypic changes in Fgfr2cKO ureters.

Previous work described the role of FGFR signaling in the development of numerous components of the urinary system, but its role(s) in ureter development has remained unexplored (Walker et al., 2016). Based on their expression in the early UE, we used a conditional gene targeting experiment to analyze the specific function of Fgfr1 and Fgfr2 in this tissue. Our phenotypic characterization of compound and double mutants revealed that FGFR2 function maintains the structural integrity of the ureter by controlling different cellular programs in both the epithelial and mesenchymal tissue compartment of this organ.

Owing to our breeding strategy, we recovered only Fgfr1-Fgfr2 compound mutants for phenotypic analysis. Although we cannot formally exclude a (minor) contribution of heterozygous loss of Fgfr1 to the observed phenotypic changes of the ureter in embryos with homozygous loss of Fgfr2, we are convinced that control of early ureter development is exerted almost exclusively by FGFR2. First, Fgfr2 is much more strongly expressed than Fgfr1 in the UE from E12.5 to E14.5. Second, complete loss of Fgfr1 with combined loss of one allele of Fgfr2 did not result in changes in FGF signaling, i.e. Spry1 and Spry2 expression, in the UE, whereas complete loss of Fgfr2 did. Third, complete loss of Fgfr2 but not of Fgfr1 resulted in severe ureteric cytodifferentiation defects. Fourth, FGF7 and FGF10, the two ligands with expression in the UM, predominantly signal through the epithelial isoform of FGFR2 (Igarashi et al., 1998; Jans, 1994; Ornitz and Itoh, 2015), whereas specific expression of FGF ligands that preferentially signal through FGFR1 was not detected in the early ureter. Fifth, previous studies using a Hoxb7cre line for recombination in the ureteric bud lineage did not detect defects in the urogenital system of Hoxb7cre/+;Fgfr1^fl/fl embryos, whereas Hoxb7cre/+;Fgfr2^fl/fl embryos exhibited renal hypo(dys)plasia due to reduced branching morphogenesis, and thinning of the early ureter and hydroureter, which are highly reminiscent of the phenotypic changes observed in our Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/fl embryos (Sims-Lucas et al., 2011; Zhao et al., 2004).

Although we detected embryonic lethality in our compound mutants, Hoxb7cre/+;Fgfr1^fl/fl, Hoxb7cre/+;Fgfr2^fl/fl and Hoxb7cre/+;Fgfr1^fl/fl;Fgfr2^fl/fl mice exhibited a normal Mendelian distribution at embryonic and adult stages (Sims-Lucas et al., 2011; Zhao et al., 2004). The Pax2-cre line used in our conditional gene targeting experiments also recombines outside the nephric duct epithelium (and its derivatives), particularly strongly in the midbrain-hindbrain region and the branchial arches at E9.5 (Kuschert et al., 2001). These expression domains are likely to give rise to vessels in the brain but also to the second heart field from which the atria, the right ventricle and the outflow region are derived (Kelly et al., 2001; Mjaatvedt et al., 2001). Given the known role of FGF signaling in the second heart field region and in vessel development (Park et al., 2008; Yang et al., 2015), deletion of Fgfr1 and/or Fgfr2 might contribute to embryonic lethality due to cardiac/circulatory insufficiency.

The study by Zhao et al. characterized the function of Fgfr2 in renal development, but it did not explain the thinning of the early ureter and hydroureter formation in these mice (Zhao et al., 2004). Our study shows that these defects relate to an independent function of FGFR2 signaling in the development of the distal aspect of the ureteric bud. Epithelial Fgfr2 was required for proliferation, stratification and I-/B-cell differentiation in the UE but its loss also had an impact on apoptosis, and SMC as well as lamina propria differentiation in the UM. Although work in the bladder assigned FGF7 and FGF10 a role as mitogens for I cells (Bagai et al., 2002; Tash et al., 2001; Zhang et al., 2006), our findings indicate a broader function for epithelial FGFR signaling – using a relay system – to coordinate the development of both tissue compartments in the early ureter.

Our ink injection experiments revealed that in Fgfr2cKO urogenital systems physical obstruction occurs only in 20% of the specimens, indicating that delayed SMC differentiation contributes or causes hydroureter formation, as observed in other mouse models (Weiss et al., 2019). Luminal occlusion in distal ureter regions due to epithelial hypoplasia and/or distal ureter maturation defects due to delayed ureter budding may contribute to this defect, as reported for Hoxb7-cre/+;Fgfr2^fl/fl mice (Sims-Lucas et al., 2011; Zhao et al., 2004). Our Fgfr1/Fgfr2 double mutants exhibited an increased incidence of hydroureter formation due to ureteropelvic junction obstruction, a blind-ending distal ureter and ectopic urethral connectivity. They also showed strong renal and ureter hypoplasia. We assume that these defects reflect the combined function of Fgfr1 and Fgfr2 in UB formation, and branching morphogenesis of the collecting duct system from the proximal UB tip region, as previously reported (Sims-Lucas et al., 2011; Zhao et al., 2004).

Fgfr2cKO ureters exhibit a spectrum of phenotypic changes in both the epithelial and mesenchymal compartment that are similar in nature but reduced in severity compared with those seen when the SHH-FOXF1-BMP4 signaling axis is lost. Moreover, the temporal window of epithelial FGFR2 signaling activity aligns with the expression profile of Shh and Bmp4 in the ureter; both are strongly downregulated after E14.5 (Bohnenpoll et al., 2017c; Mamo et al., 2017; Yu et al., 2002). Expression of Shh as well as of Ptch1, Foxf1 and Bmp4, which represent the Shh effector level, was reduced in Fgfr2cKO ureters. Activation of SHH/SMO signaling by purmorphamine and addition of BMP4 largely rescued the proliferation and differentiation defects in Fgfr2cKO ureters. Finally, reduction of SHH and BMP4 signaling in wild-type ureters recapitulated the phenotypic changes observed in Fgfr2cKO ureters in a dose-dependent manner. Together, this provides compelling evidence that FGFs act via FGFR2 to enhance Shh expression and consequently SHH-FOXF1-BMP4 activity in the early ureter (Fig. 7).

It is important to note that the FGF7/10-FGFR2-SHH-BMP4 signaling axis has previously been characterized in the early development of other organs, including the urethra, the limb, the palate and the eyelid (Huang et al., 2009; Petiot et al., 2005; Revest et al., 2001; Rice et al., 2004). The primordia of all of these organs, as well as of the ureter and bladder are characterized by a composite design with epithelial and mesenchymal tissue compartments. The coordinated development of these compartments is assured by reciprocal signaling systems in which the FGF7/10-FGFR2-SHH-BMP4 module seems of outstanding relevance.

Our microarray analysis identified Hoxb8 as the most downregulated gene in the Fgfr2cKO ureter. As Hoxb8 has been implicated in proliferation control (Guo et al., 2019; Wang et al., 2019), its reduced expression in the UE may contribute to urothelial hypoplasia. Importantly, Hoxb8 does not depend on SHH signaling in the ureter (Bohnenpoll et al., 2017c), and can be ectopically induced in neural tissues of the chick by FGF treatment (Bel-Vialar et al., 2002). This suggests that FGFR2 signaling regulates a set of genes independently of the SHH-FOXF1-BMP4 signaling axis.

Fgfr2cKO ureters displayed a mono-layered urothelium consisting of S cells. This phenotype is highly reminiscent of that seen in the bladder and ureters of mice with conditional loss of ΔNP63 in the respective epithelial primordium (Cheng et al., 2006; Pignon et al., 2013; Weiss et al., 2013). Failure to activate ΔNP63 in Fgfr2cKO ureters, therefore, likely accounts for the lack of stratification and B-cell differentiation in the mutant urothelium.

Expression and lineage tracing analysis uncovered that S and B cells are terminally differentiated cell types that arise from a common progenitor by an I-cell intermediate. The I cells were recognized as ΔNP63⁺ cells lacking high expression of UPKs and KRT5 (Bohnenpoll et al., 2017a; Gandhi et al., 2013). Differentiation of S cells in absence of ΔNP63 shows that stratification is not a prerequisite for S-cell differentiation, and suggests that S-cell differentiation is normally inhibited by ΔNP63 in I cells.

Mice with conditional loss of Smo or Bmp4 in the UM do not activate ΔNP63 in the urothelium, and lack stratification and B- and S-cell differentiation (Bohnenpoll et al., 2017c; Mamo et al., 2017). In Fgfr2cKO ureters, Shh and, consequently, Bmp4 expression is reduced but not lost, suggesting that ΔNP63 expression and stratification requires higher levels of SHH and BMP4 signaling than S-cell differentiation. This notion is supported by the restoration of ΔNP63 expression in Fgfr2cKO ureters by purmorphamine and BMP4 treatment, on the one hand, and a relatively higher decrease in I cells compared with S cells by increasing doses of cyclopamine and NOG in wild-type ureters, on the other hand. Administration of BMP4 to early kidney explants leads to UPK expression in collecting duct cells (Mills et al., 2017; Wang et al., 2009), indicating that BMP4 is required and sufficient to activate S-cell differentiation. It is conceivable that ectopic induction of I-cell differentiation and of ΔNP63 expression, respectively, require higher levels of BMP4 and/or additional positive signals, similar to the situation in other epithelia (Terakawa et al., 2016). Alternatively, concurrent repression of an inhibitor may allow induction of ΔNP63.

Interestingly, the epithelium covering the renal papilla is monolayered and consists of S cells only. It is conceivable that out of the reach of FGF7/FGF10 signals, the SHH-BMP4 axis is not sufficiently augmented to activate ΔNP63 and to drive stratification at this site.

Our expression analysis showed that Fgfr2 is strongly downregulated after E14.5, excluding a role for FGFR2 signaling in later (fetal) development and homeostasis. This is consistent with the mature urothelium being quiescent. However, under conditions of injury or infection, proliferation of I cells and differentiation into S and B cells resume to repair the urothelium within days. Interestingly, recent reports revealed that FGF7 and FGFR2 function is reused in this program (Girshovich et al., 2012; Narla et al., 2020). Whether an FGF7-FGFR2 module employs the SHH-BMP4 signaling axis in regeneration similar to the embryonic situation is an interesting question for future research.

Mice with loxP sites flanking exon 4 of the Fgfr1 locus (Fgfr1^tm5.1Sor; synonym: Fgfr1^fl) (Hoch and Soriano, 2006) and mice with loxP sites flanking exons 7 to 10 of the Fgfr2 locus (Fgfr2^tm1Dor; synonym: Fgfr2^fl) (Yu et al., 2003) were obtained from the Jackson Laboratory. Tg(Pax2-cre)1AKis (synonym: Pax2-cre) mice were previously generated in the lab (Bohnenpoll et al., 2017a; Trowe et al., 2011). All mice were maintained on a NMRI outbred background. Embryos for expression analysis of genes encoding FGF signaling components as well as for loss-of-function experiments were obtained from NMRI mice; embryos for phenotype analysis were generated by mating Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/+ males with Fgfr1^fl/fl;Fgfr2^fl/fl females. Cre-negative littermates were used as controls. For timed pregnancies, vaginal plugs detected in the morning after mating were designated as embryonic day (E) 0.5 at noon. Urogenital systems and embryos were dissected in PBS, fixed in 4% paraformaldehyde (PFA) in PBS and stored in methanol at −20°C. For genotyping by PCR, genomic DNA prepared from yolk sacs or ear clip biopsies was used.

Mice were housed in rooms with controlled light and temperature. The experiments were carried out in accordance with the German Animal Welfare Legislation and approved by the local Institutional Animal Care and Research Advisory Committee and permitted by the Lower Saxony State Office for Consumer Protection and Food Safety (AZ 33.12-42502-04-13/1356, AZ42500/1H).

Ureters for explant cultures were dissected in L-15 Leibovitz medium (F1315, Biochrom), explanted on 0.4 µm polyester membrane Transwell supports (657610, Greiner Bio-One) and cultured at the air-liquid interface with DMEM/F12 (21331020, Gibco) supplemented with 1×penicillin/streptomycin (15140122, Gibco), 1×NEAA (11140035, Gibco), 1×pyruvate (11360070, Gibco) and 1×glutamax (35050038, Gibco) in a humidified incubator with 5% CO₂ at 37°C. Pathway activating components were dissolved as follows: recombinant human BMP4 (100 ng/µl in 4 mM HCl/0.1% BSA; PHP171, Abd Serotec), purmorphamine (2 µM in DMSO; 540220, Merck), retinoic acid (RA; 1 µM in DMSO; R2625, Sigma-Aldrich), 6-bromoindirubin-3′-oxime (BIO, 5 µM in DMSO; S7198, Selleckchem), cyclopamine (1-10 µM in DMF; S11465, Selleckchem) and NOG (2.5-10 µg/ml in double distilled H₂O; Z0320525, Genescript). Medium containing DMSO or components was refreshed every second day.

Kidney size and ureter length was measured using the segmented line tool from Image J (Schindelin et al., 2012). Ureter length was measured from the pelvic region to the bladder insertion site. Kidney size was calculated by measuring the cranial to caudal length and the medial to lateral width.

Embryos, urogenital systems and ureter explants were fixed in 4% PFA, embedded in paraffin wax and sectioned at 5 µm. Sections were stained with Hematoxylin and Eosin according to standard procedures.

Immunofluorescence staining as well as immunohistochemistry was performed on 5 µm paraffin wax-embedded sections using the following primary antibodies and dilutions: monoclonal mouse-anti-BrdU (1:250; WH0007348M2, Sigma-Aldrich) polyclonal rabbit-anti-KRT5 (1:200; PRB-160P, BioLegend), polyclonal rabbit-anti-ΔNP63 (1:100; clone Poly6190, 619001, BioLegend), monoclonal mouse-anti-P63 (1:200; clone 4A4, ab735, Abcam), monoclonal mouse-anti-UPK1B (1:200; clone1E1, WH0007348M2, Sigma-Aldrich), polyclonal rabbit-anti-TAGLN (1:200; ab14106, Abcam), polyclonal mouse-anti-ACTA2 (1:200; A5228; clone 1A4; Merck), polyclonal rabbit-anti-ALDH1A2 (1:200; ab75674, Abcam), polyclonal rabbit-anti-CDH1 (1:200, a kind gift from Dr R. Kemler, MPI, Freiburg, Germany), monoclonal rabbit-anti-P-SMAD1/5/9 (1:100; 13280, Cell Signaling), monoclonal rabbit-anti-P-P38 MAPK (1:100; 4631, Cell Signaling), monoclonal rabbit-anti-P-AKT (1:100; 9271, Cell Signaling) and monoclonal rabbit-anti-P-ERK1/2 (1:100; 9102, Cell Signaling). Fluorescent staining was performed using the following secondary antibodies: biotinylated goat-anti-rabbit IgG (1:200; 111065033, Dianova), biotinylated goat-anti-mouse IgG (1:200; 115-065-166, Jackson ImmunoResearch), Alexa488-conjugated goat-anti-rabbit IgG (1:400; A11034, Molecular Probes) and Alexa555-conjugated goat-anti-mouse IgG (1:400; A21422, Molecular Probes). The signals of ΔNP63, P63 and ALDH1A2 were amplified using the Tyramide Signal Amplification system (NEL702001KT, Perkin Elmer). For co-staining with primary antibodies of the same host (ΔNP63 and KRT5 or CDH1), the staining was performed sequentially and the epitope of the first antibody was blocked with goat-anti-rabbit FAB fragment (1:50; 111007003, Dianova). The signals of P-SMAD1/5/9, P-AKT, P-ERK1/2 and P-P38 were amplified using the DAB amplification system (#NEL938001EA, Perkin Elmer). For antigen retrieval, paraffin wax-embedded sections were deparaffinized, pressure-cooked for 15 min in antigen unmasking solution (H3300, Vector Laboratories), treated with 3% H₂O₂/PBS for blocking of endogenous peroxidases, washed in PBST (0.05% Tween-20 in PBS) and incubated in TNB Blocking Buffer (NEL702001KT, Perkin Elmer). Sections were then incubated with primary antibodies at 4°C overnight. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI, 6335.1, Carl Roth).

In vivo cell proliferation rates of E12.5 and E14.5 ureters were assayed by detection of incorporated 5-bromo-2′-deoxyuridine (BrdU) on 5 µm sections (Bussen et al., 2004). Five to 25 sections of each specimen were analyzed. The BrdU labeling index was defined as the number of BrdU-positive nuclei relative to the total number of nuclei detected by DAPI counterstaining in histologically defined compartments of the ureter. Apoptosis in tissues was assessed by the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay using ApopTag Plus Fluorescein In Situ Apoptosis Detection Kit (S7111; Merck) on 5 µm paraffin wax-embedded sections.

Section in situ hybridization on 10 µm paraffin wax-embedded sections using digoxigenin-labeled antisense riboprobes was performed as previously described (Moorman et al., 2001).

RNA extraction and RT-PCR analysis for gene expression was performed on pools of 20 ureters each of E14.5 control and Pax2-cre/+;Fgfr1^fl/+;Fgfr2^fl/fl embryos. We isolated total RNA using TRIzol (#15596-018, Thermo Fisher Scientific) and synthesized cDNA from total RNA applying RevertAid H Minus reverse transcriptase (#EP0452, Thermo Fisher Scientific) as described previously (Thiesler et al., 2021). The NCBI tool Primer3 version4.1 was used to design specific primers (Table S11) (Untergasser et al., 2012; Werneburg et al., 2015). RT-quantitative (q)PCR of mouse genes was performed in 10 µl 1:2 diluted BIO SyGreen Lo-ROX mix (PCR Biosystems) with 400 nM primers and 1 ng/µl cDNA applying a QuantStudio3 PCR system fluorometric thermal cycler (Thermo Fisher Scientific). Data were processed by QuantStudio data analysis software (version 1.5.1, Thermo Fisher Scientific) using the comparative threshold cycle (ΔΔC_T) method.

Two independent pools each of control and mutant ureters were used for microarray analysis. Pool sizes were as follows: 50 ureters each from male and female E13.5 cre-negative and Pax2-cre/+;Fgfr1^{fl/(+ or fl)};Fgfr2^fl/fl embryos. Total RNA from each pool was extracted using peqGOLD RNApure (30-1010, VWR international) and subsequently processed by the Research Core Unit Transcriptomics of Hannover Medical School. Agilent whole Mouse Genome Oligo v2 (4×44K) Microarrays (G4846A; Agilent Technologies) were used for transcriptome analysis. Normalized expression data were filtered using Microsoft Excel. Functional enrichment analysis for up- and downregulated genes was performed with DAVID 6.8 web-software (david.ncifcrf.gov), and terms were selected based on P-value. Microarray data have been deposited in GEO under accession number GSE178093.

Statistical analysis was performed using the unpaired, two-tailed Student's t-test (GraphPad Prism version 7.03 and Microsoft Excel). Values are indicated as mean±s.d. P<0.05 was considered significant.

Sections were photographed using a DM5000 microscope (Leica Camera) with a Leica DFC300FX digital camera or a Leica DMI6000B microscope with a Leica DFC350FX digital camera. All images were then processed using Adobe Photoshop CS4.

We thank Rolf Kemler for the anti-CDH1 antiserum. Microarray data used in this publication were generated by the Research Core Unit Genomics (RCUG) at Hannover Medical School. Nurullah Aydoğdu was supported by the Hannover Biomedical Research School (HBRS) and the MD/PhD program Molecular Medicine.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200021.