A critical role for NF2 and the Hippo pathway in branching morphogenesis

Branching morphogenesis is a complex biological process common to the development of most epithelial organs. Here we demonstrate that NF2, LATS1/2 and YAP play a critical role in branching morphogenesis in the mouse kidney. Removal of Nf2 or Lats1/2 from the ureteric bud (UB) lineage causes loss of branching morphogenesis that is rescued by loss of one copy of Yap and Taz, and phenocopied by YAP overexpression. Mosaic analysis demonstrates that cells with high YAP expression have reduced contribution to UB tips, similar to Ret−/− cells, and that YAP suppresses RET signalling and tip identity. Conversely, Yap/Taz UB-deletion leads to cyst-like branching and expansion of UB tip markers, suggesting a shift towards tip cell identity. Based on these data we propose that NF2 and the Hippo pathway locally repress YAP/TAZ activity in the UB to promote subsequent splitting of the tip to allow branching morphogenesis.

B ranching morphogenesis is critical for the development and function of most epithelial organs, and is essential for the formation of the mammalian kidney. The kidney develops through reciprocal signalling between a ureteric epithelium that forms the collecting duct, and surrounding mesenchymal nephron progenitors and stroma 1 . The bilateral symmetry and characteristic shape of kidneys indicates branching morphogenesis is highly controlled. In addition, the rotational angle between one set of branches and the next is relatively fixed implying tight regulation 2 . How this developmentally critical branching is so tightly regulated is still unclear.
The first step of kidney development occurs when the ureteric bud (UB) invades the metanephric mesenchyme. Tip identity is defined in response to metanephric mesenchyme-derived signals including Glial cell-line-derived neurotrophic factor (GDNF) and fibroblast growth factors 3 . The tip contains progenitor cells that can self-renew or be left behind to give rise to trunk cells 4 . Binding of GDNF to its receptors GFRa1 and RET, triggers tyrosine kinase signalling, which induces the outgrowth of the UB. Once the ureter has invaded the metanephric mesenchyme, GDNF/RET signalling at UB tips leads to growth and repetitive branching of the ureter to form a ureteric tree that will give rise to the collecting duct. Loss of Gdnf, Ret or Gfra1 impairs branching morphogenesis causing kidney defects ranging from renal dysplasia to complete agenesis [5][6][7] . RET signalling is essential to form and maintain tips, and promotes a feed-forward signalling loop, in which RET signalling promotes expression of Ret transcript. The tip domain swells to form an ampulla before a symmetry breaking event occurs that allows it to split into two new tips. How symmetry is broken in the ampulla to form a branch is not understood.
Studies both in flies and mammals suggest that NF2 can also regulate the Hippo pathway. The Hippo pathway is a conserved kinase cassette that regulates tissue growth by controlling the activity of YAP and TAZ (refs 11-13). YAP and TAZ are closely related transcriptional co-activators that promote the expression of pro-proliferative and anti-apoptotic genes. Upstream of YAP and TAZ are the Hippo kinases MST1/2 and LATS1/2, which negatively regulate YAP and TAZ and cause their exclusion from the nucleus. Loss of Hippo signalling leads to unrestricted proliferation in flies and mammals, and has been linked to a variety of developmental abnormalities and cancers 14,15 . NF2 can bind and recruit LATS to the plasma membrane, where it is activated by MST kinases 16 . NF2 has also been shown to bind other Hippo pathway components 8 . NF2 is one of many regulators of the Hippo pathway 17 : Cell adhesion, cell polarity, mechanical forces and the cytoskeleton have all been shown to regulate YAP localization in tissue culture 18 , suggesting that as tissues grow and develop, feedback may occur from resultant changes in the environment.
Here we uncover an unsuspected role for NF2 and the Hippo pathway in kidney branching morphogenesis. We find that Nf2 conditional mutants have severe renal hypodysplasia due to defective branching morphogenesis. Nf2 kidney hypodysplasia can be rescued by reducing Yap and Taz dosage, and phenocopied by YAP overexpression, suggesting that NF2 restricts YAP/TAZ activity to promote branching morphogenesis. Lats1/2 deletion leads to kidney agenesis that can be rescued by reducing YAP/TAZ levels, suggesting that high YAP/TAZ activity inhibits branching. Importantly, loss of Nf2, or overexpression of YAP inhibits RET signalling and expression of tip markers. Our mosaic analysis reveals that YAP overexpressing cells show cellautonomous loss of RET signalling and are excluded from UB tips. In contrast, we find that UB-specific deletion of Yap and Taz leads to cyst-like branching, associated with an expanded UB tip domain. Taken together, these data demonstrate an essential role for NF2 and the Hippo pathway in regulation of branching morphogenesis in the mammalian kidney.

Results
Nf2 deletion results in severe kidney defects. To investigate the role of NF2 during kidney development, we first stained developing kidneys with NF2 antibody, and found that NF2 protein uniformly localized along the apical membrane of the UB and nephrons throughout development ( Fig. 1a and Supplementary  Fig. 1a-h). NF2 is maintained in the adult kidney, although at a weaker level ( Supplementary Fig. 1i). The apical staining is specific, as it is lost in Nf2 mutant tissue (Supplementary Fig. 1j-m).
NF2 is required for branching morphogenesis. To examine the role of NF2 in ureter branching in greater detail, we placed T-stage (2 UB tips) kidney rudiments dissected from wild-type and Nf2 UB À / À E11.5 embryos in culture and counted the number of UB tips after 48 h. Strikingly, deletion of Nf2 resulted in loss of ureter branching (number of UB tips: 2±0.4, n ¼ 21) compared with controls (23.5 ± 4.7, n ¼ 21, Fig. 1i-k). Interestingly, while NF2 function is dispensable for the formation of the UB (marked by PAX2) and its first branching event (Fig. 1l,m), subsequent branching failed in the absence of Nf2 (Fig. 1n-q  and 2a,b).
Despite the loss of branching, Nf2-depleted UBs continued to grow, as indicated by the wider epithelial tubules observed in Nf2 UB À / À kidneys compared with controls (Fig. 1i,j and 2c,d). Quantification at E12.5 revealed a 2.2-fold increase in UB width and 1.85-fold increase in cellularity in the UB compartment of Nf2 UB À / À embryos compared with controls ( Fig. 2e,f). At E12.5, the dense packing leads to the appearance of stratification; however, confocal z-stack imaging of PAX2/E-CADHERIN staining revealed that UB cells in Nf2 mutants contact both apical and basal surfaces, consistent with a pseudostratified structure (Supplementary Movie 1 (control) and Supplementary Movie 2 (Nf2 UB À / À )).
To ascertain if altered apoptosis could explain the morphological defects seen in Nf2 mutants, we examined CLEAVED CASPASE 3 (CC3) staining in control and Nf2 UB À / À kidneys at E12.5. CC3-positive cells were rarely detectable in Nf2 UB À / À and control UB cells (marked by CALBINDIN) suggesting cell death is not the primary cause of defective branching ( Supplementary  Fig. 1t,u). Quantification revealed higher apoptotic rates in the cortex of Nf2 UB À / À compared with controls ( Supplementary   Fig. 1v) consistent with known roles of branching in sustaining survival of the mesenchyme. Quantification of EdU incorporation at E11.5 showed no change in proliferation of UB cells ( Supplementary Fig. 1w-y). We conclude that neither proliferation nor apoptosis are primarily responsible for Nf2 UB À / À branching defects.  Fig. 1d for counterstaining with E-CADHERIN to mark the UB and early nephron. (b-d) Macroscopic view of the urogenital system from wild-type and Nf2 UB À / À mutants at P0 shows severe kidney hypoplasia (arrows) in Nf2 mutants. (e) Quantification revealed a 76% decrease in kidney size in Nf2 mutants (n ¼ 14 kidneys) compared with controls (n ¼ 9). Error bars represent s.d., ***Po0.0001, Student's t-test. (f-h) Periodic acid-Schiff (PAS) staining of P0 kidneys from wild type and Nf2 UB À / À . (i,j) Ex vivo analysis demonstrates loss of ureter branching in Nf2 UB À / À compared with controls, visualized using anti-CALBINDIN antibody (pink arrows point to UB tips, yellow arrowheads to the ureter/collecting duct (CD) and asterisks mark the Wolffian duct). Branching morphogenesis during kidney development involves signalling between the UB, metanephric mesenchyme and stromal compartments 1,3 . PBX1 staining revealed normal differentiation of the stromal compartment in Nf2 mutants ( Supplementary Fig. 2c,d). Staining with condensing mesenchyme (CM) markers (SALL1, SIX2 and PAX2) revealed a decreased CM population at E12.5 and E13.5, and a depletion of CM at E18.5 in Nf2 UB À / À kidneys (Fig. 2a,b,g,h and Supplementary Fig. 2a-h). PAX2 staining on kidney explants confirmed the decreased CM population in Nf2 mutants compared with controls ( Supplementary Fig. 2i,j). Moreover, we observed a difference in the morphology of the CM domain, as SIX2 cells appeared sparser in Nf2 UB À / À kidneys compared with controls (compare SIX2 staining in Fig. 2a,b and Supplementary Fig. 2e,f).
binding to its receptors RET and GFRa1 induces both ERK/ MAPK and PI3K pathway activation. PI3K activation is critical for UB branching, and necessary for activation of downstream RET target genes ETV4 and ETV5 (ref. 20). Pharmacological inhibition of the ERK/MAPK pathway or loss of Mek1/2 also results in impaired branching but without affecting the expression of ETV4/5 (ref. 21). Thus both ERK/MAPK and PI3K signalling contribute to branching morphogenesis.
To test if the RET pathway is affected by Nf2 deletion, we performed in situ hybridization and immunostaining for Ret and its downstream targets. Although Nf2 UB À / À kidneys had fewer UB tips compared with controls, Ret was expressed at levels comparable to controls at UB tips (Fig. 2i,j). We next examined RET signalling activity. No changes in Etv4, ETV5 and phospho-AKT were observed in Nf2 UB À / À kidneys compared with controls ( Fig. 2m-p, Supplementary Fig. 2s-v) suggesting that NF2 does not primarily affect PI3K activation. Interestingly, expression of tip marker genes (Wnt11, Crlf1, Sprouty1 and Dusp6) are seen at the few tips remaining in Nf2 UB À / À kidneys (Fig. 2k,l, and Supplementary Fig. 2m-r). We next examined ERK/MAPK pathway activation with antibodies to p-ERK ( Fig. 2q-u). In controls, p-ERK staining was observed at the UB tips and in early nephrons. In Nf2 UB À / À mutants, p-ERK staining persisted in early nephrons, but UB tip staining was greatly reduced suggesting that loss of Nf2 may impair UB branching due to defective ERK/MAPK activation.
Yap/Taz haploinsufficiency rescues the Nf2 UB À / À phenotype. Since NF2 is required for tight junction formation and cell polarity in the epidermis 22 , we examined Nf2 UB À / À kidneys for CRUMBS3, ZO-1, b-CATENIN and LAMININ. All these markers revealed that UB cells depleted for Nf2 had normal apico-basal polarity at the time of defective branching ( Supplementary Fig. 3a-f).
To assess branching morphogenesis in greater detail, we dissected wild-type and Nf2 UB À / À Yap UB À / þ Taz UB À / þ kidney rudiments at E11.5 and placed them in culture for 48 h. Lowering YAP/TAZ dosage in Nf2 mutants largely rescued the branching capacity of kidney explants (Fig. 3l,m). Reducing YAP and TAZ also decreased the non-cell-autonomous tubule dilatation (Fig. 3n), and restored phospho-ERK expression (Fig. 3o-t). Remarkably, while most of Nf2 mutants died within 48 h after birth, all Nf2 UB À / À Yap UB À / þ Taz UB À / þ mice lived to adulthood (n ¼ 5 animals). These results suggest that the loss of branching morphogenesis in Nf2 mutants is due to excessive YAP/TAZ activity.
YAP expression reversibly inhibits branching. Our rescue experiments suggest that increased YAP/TAZ activities in Nf2 UB À / À inhibit UB branching. To directly test this hypothesis, we overexpressed YAP in the UB lineage by generating Hoxb7: Cre tg/ þ Rosa26-lox-STOP-lox-rtta-IRES-EGFP Yap Tg animals (called Yap UB-OE ). This approach allows constitutive expression of the wild type, HA-tagged form of YAP in the UB compartment after doxycycline treatment either by feeding pregnant dams with doxycycline food or by adding doxycycline to the kidney culture medium.
Initially, we overexpressed YAP in the UB lineage throughout kidney development by feeding pregnant dams with doxycycline (0.625 g kg À 1 ) from E11 to E18.5. Yap UB-OE animals developed severe renal hypodysplasia at E18.5 (Fig. 4a-d). To explore if YAP overexpression in the UB lineage affected branching, we cultured E11.5 Yap UB-OE kidney rudiments (T-stage) in presence of doxycycline (1,500 ng ml À 1 ), and counted the number of UB tips 24 and 48 h later. Interestingly, while control kidneys had 9±2.8 and 22.5 ± 1.3 UB tips after 24 and 48 h, respectively ( Fig. 4e-g,q), doxycycline-induced YAP overexpression in the UB led to a complete blockage of branching (2 ± 0 UB tips, Fig. 4h-j,q and Supplementary Fig. 4a). Similarly, UB branching was blocked when doxycycline was added to the medium 24 h after dissection ( Fig. 4k-m,q). Remarkably, when doxycycline was withdrawn from the medium, UB branching resumed (Fig. 4n-p,q). Thus, YAP rapidly and effectively blocks branching in a reversible manner.
Constitutive YAP overexpression resulted in UB branching arrest (Fig. 4), but UBs overexpressing YAP continued to grow, as indicated by the wider epithelial tubules observed in Yap UB-OE kidneys explants (Fig. 4i,j). Similarly, feeding pregnant dams with doxycycline diet from E11 to E13.5 was sufficient to activate YAP expression ( Supplementary Fig. 5o,q), block UB branching, and led to increased UB width and epithelial pseudostratification in E13.5 Yap UB-OE embryos (Fig. 5d,f). As with Nf2 mutants, proliferation in Yap UB-OE mutants was unaffected at E13.5 ( Supplementary Fig. 5e-g). Loss of Nf2 or YAP overexpression blocked branching and resulted in a pseudostratified epithelium. We speculate this is due to the continuation of proliferation in the absence of branching. Thus, increased YAP activity in the UB lineage leads to loss of branching.
YAP overexpressing cells are excluded from UB tips. YAP overexpression throughout the UB lineage inhibits branching morphogenesis and leads to kidney hypoplasia (Fig. 4). Interestingly, due to the occasional mosaic activity of the Hoxb7:Cre line, we observed that kidney hypoplasia correlated with the percentage of YAP overexpressing (YAP OE ) cells (Fig. 5a-g). We arbitrary divided the Yap UB-OE animals in three categories (low, intermediate and high) based on the percentage of YAP OE cells in the ureteric epithelium. Kidneys with a low percentage of YAP OE cells (o12%), were similar in size to controls, while those with a high percentage (4 86%) were severely reduced in size, likely as a consequence of branching arrest (Fig. 5g).
Next, we wondered if the tip exclusion behaviour of YAP OE cells is linked to the level of YAP overexpression. We first examined the impact of increasing doxycycline exposition (pregnant dams fed with 0.2, 0.645 (the same concentration used above) and 2 g kg À 1 doxycycline food from E11 to E18.5) on kidney branching. Remarkably, all diets led to severe kidney hypoplasia ( Supplementary Fig. 5n), likely due to the high level of YAP overexpression observed even at low dose ( Supplementary  Fig. 5o-r). Therefore, to better control the level of YAP overexpression, we examined the effect of varying the concentrations of doxycycline in cultured kidneys and analysed YAP induction after 24 h using western blot analysis. We also quantified UB branching after 48 h using CALBINDIN staining.
Decreasing the level of doxycycline to 15 ng ml À 1 resulted in branching similar to controls (Fig. 5m), with HA-positive cells observed in the entire UB compartment (UB tip and trunk, as observed from kidney explant sections in Fig. 5n-q). Thus cells with low levels of YAP overexpression (Fig. 5r) can contribute to active UB tips. We conclude that cells with high levels of YAP activity have a cell-autonomous defect in their ability to contribute to the UB tip domain and RET signalling.
Yap/Taz deletion expands UB tip leading to dilated branching. Our data clearly show that high YAP/TAZ activity is incompatible with UB branching morphogenesis. To determine if Yap and Taz expression is essential for branching morphogenesis, we examined Yap and Taz mutants using the Hoxb7:Cre. Dilated tip structures were apparent in Yap UB À / À kidney explants (Fig. 7a,b), while UB depleted of Taz branched normally (data not shown). Double Yap/Taz UB À / À kidneys were hypoplastic with decreased proliferation (Supplementary Fig. 7a-c) and severe disruptions in branching morphogenesis, resulting in cyst-like branching (Fig. 7c-f). In controls, expression of Ret and its downstream targets ETV5, Crlf1, Etv4 and Dusp6 were restricted to the UB tips, whereas in Yap/Taz mutants, their expression domains expanded (Fig. 7g À p). This suggests that in the absence of Yap and Taz, UB tip fate expands. Thus Yap/Taz deletion from the UB epithelium leads to the opposite effect from loss of Nf2 and excess YAP, namely expanded expression of RET pathway/ tip genes and cyst-like branching, that may represent unrestricted outward expansion of a tip fate.

Discussion
We have uncovered an unexpected role for NF2 and the Hippo pathway in regulating branching morphogenesis in the mammalian kidney. We found that NF2 and LATS1/2 are required in the UB lineage to restrain YAP/TAZ activities, and showed that YAP overexpression is sufficient to block branching morphogenesis. In addition, reducing YAP and TAZ levels can rescue branching defects of both Lats1/2 and Nf2 mutants. This suggests that increased YAP/TAZ activity is the primary cause of defective branching on loss of Nf2 and Lats1/2.
Using mosaic analysis, we also found that cells overexpressing YAP have a cell-autonomous defect in their ability to contribute to the UB tip domain. Thus, cells with heightened YAP levels, similar to Ret À / À and Etv5 À / À cells 3,24 , are excluded from tips. Strikingly, rare YAP OE cells found at tips failed to activate RET signalling in a cell-autonomous manner. In contrast, UB depleted for Yap and Taz showed expansion of the tip domain. Taken together, these data suggest YAP/TAZ inhibits tip fate and signalling.
We cannot distinguish if YAP directly represses RET signalling or represses tip fate, thereby affecting RET signalling. Either scenario might allow a splitting of the tip, allowing new branch formation. How might this occur? One attractive scenario for the action of YAP/TAZ in symmetry breaking is that crowding that occurs at tips 25 , impacts the levels of nuclear YAP/TAZ and thereby RET signalling. This would split the tip domain, leading to branch formation. The positive feedback of RET signalling to increased Ret transcription would then promote further growth of each tip, until a critical point of cell crowding and tension occurs at a new ampulla, which would then again regulate nuclear YAP, causing splitting of the ampulla and branch formation.
Tip cells produce secreted paracrine signals that maintain and pattern both the stromal and mesenchymal compartments. Interestingly, Nf2 UB À / À and Yap UB-OE kidneys showed rapid depletion of the CM (as early as E13.5) suggesting that increased YAP activity at the tips is detrimental to the expression of an unidentified signal that maintains the CM. We noted cell death in the Nf2 UB À / À outer cortex, likely due to the fact that CM cells receive reduced inductive signals from the smaller number of UB branches that formed.
Nf2 deletion leads to both cell-autonomous and non-cellautonomous tubule dilations at P0. The non-cell-autonomous effects may be due to secondary defects subsequent to loss of renal structure. Interestingly, lowering YAP and TAZ levels in the UB compartment of Nf2 mutants significantly reduced tubule dilatations suggesting that when NF2 inhibits YAP and TAZ in the UB compartment to promote branching morphogenesis, it may also restore inductive signals to the adjacent CM for nephron formation. It is interesting to note that while Yap or Taz haploinsufficiency partially rescues the branching defect of Nf2 mutants, Taz haploinsufficiency has only a limited ability to rescue tubule dilatations compared with Yap haploinsufficiency. This confirms that YAP and TAZ have redundant function in branching 26 , and suggests that YAP and TAZ have different functions in restricting tubule dilations.
If both NF2 and LATS1/2 inhibit YAP/TAZ activity, why does Lats1/2 UB-deletion give rise to stronger phenotype than Nf2 deletion? Phospho-YAP staining persists in Nf2 UB À / À , while it is absent in Lats1/2 UB À / À kidneys, suggesting that in the absence of Nf2, LATS1 and LATS2 kinases are still active. Our study suggests that LATS1/2 kinases might also have other roles than YAP/TAZ phosphorylation. We found that Lats1/2 UB À / À displayed obvious polarity defects, not observed in Nf2 UB À / À or Yap UB-OE (Supplementary Fig. 4b-g). The suppression of defects of both Lats1/2 and Nf2 mutants by reducing YAP/TAZ dosage is consistent with a linear NF2-LATS1/2-YAP/TAZ activity pathway, but does not exclude other models (Fig. 7q).
While our genetic analysis clearly demonstrates an essential role for tight regulation of YAP/TAZ activity during branch formation, we could not detect a clear difference in the localization of YAP in the tips compared with the trunk. We speculate that changes in YAP/TAZ localization are transient and/or only occur at low levels. We note that changes in Yorkie localization are not detectable in Drosophila discs in normally proliferating cells, suggesting that very low levels of YAP, undetectable by staining, can still be potent in transcriptional regulation. In vivo reporters of YAP/TAZ activity that can be imaged at real-time in the mouse kidney will be needed to obtain a better resolution of changes that are occurring at tips during development.
In conclusion, we have shown that NF2, LATS1/2, YAP and TAZ have critical roles in branching morphogenesis in the mammalian kidney. We propose a model in which the environment at the tip alters YAP/TAZ activity, which in turn feeds back to inhibit tip fate/signalling, break symmetry at the ampulla and promote branch formation. Since branching morphogenesis is critical for the formation of many organs, we speculate that the Hippo pathway will also play critical roles in branch formation in other tissues.

Methods
Mouse lines. Hoxb7:Cre tg/ þ , Taz flox and Yap flox , Lats1 flox , Lats2 flox , Nf2 flox and Yap Tg mouse strains were generously provided and have been described elsewhere 19,[27][28][29][30] . All mice were maintained in a mixed genetic background and both male and female mice were assessed in our study. The age of all animals is indicated in the figure legends. Husbandry and ethical handling of mice were conducted according to guidelines approved by the Canadian Council on Animal Care. Genotyping was done by PCR using genomic DNA prepared from mouse ear punches.
Histological and immunological analyses. Embryonic samples from timed matings (day of vaginal plug ¼ E0.5) were collected, fixed in 4% paraformadehyde overnight (O/N) at 4°C, serially dehydrated and then embedded in paraffin. Microtome sections of 7 mm thickness were examined histologically by periodic acid-Schiff staining. For immunofluorescent analysis, paraffin sections were dewaxed and serially re-hydrated with ethanol. Antigen retrieval was performed for 20 min in boiling Antigen Unmasking Solution (H-3300, Vector). Sections were incubated for 1 h in blocking solution (3% BSA, 10% goat serum, 0.1% Tween20 in