Cdc42 regulates epithelial cell polarity and cytoskeletal function during kidney tubule development

Mammalian kidney development begins when the ureteric bud invades the surrounding metanephric mesenchyme (Carroll and Das, 2013; Costantini and Kopan, 2010; Dressler, 2006). The ureteric bud is induced to undergo iterative rounds of branching morphogenesis and gives rise to the collecting system, whereas the metanephric mesenchyme forms the nephrons that consist of the glomeruli and renal tubules. Reciprocal signaling between the ureteric bud and the metanephric mesenchyme is required for ureteric bud branching morphogenesis and metanephric mesenchyme development.

Cdc42, a member of the Rho GTPases, plays a crucial role in regulating multiple cell functions including cell shape, polarity, proliferation, invasion, migration and differentiation (Melendez et al., 2011). Cdc42 mediates these functions by controlling cytoskeletal dynamics due to its ability to interact with the Wiskott–Aldrich syndrome protein (Wasp) and p21-activated kinases (PAKs). Wasp regulates actin polymerization and filopodia formation through direct interactions with both profilin and actin, and PAKs alter the activity of the crucial actin-binding protein cofilin (Melendez et al., 2011). In addition, Cdc42 regulates epithelial cell polarity by forming complexes with the PAR protein family, atypical protein kinase C (aPKC) and cadherins (Goldstein and Macara, 2007).

Both in vitro and in vivo evidence suggest a crucial role for Cdc42 in kidney development. Cdc42 is required by Madin–Darby canine kidney (MDCK) cells to form tubules in three-dimensional (3D) culture by regulating the formation of the apical plasma membrane domain and tight junctions as well as primary ciliogenesis (Choi et al., 2013; Kim et al., 2007; Martin-Belmonte et al., 2007; Rodriguez-Fraticelli et al., 2010; Rogers et al., 2003; Zhang et al., 2001; Zuo et al., 2011). Two studies have shown that Cdc42 is essential for normal kidney development and function. In the first, Cdc42 deletion from renal tubular epithelia with the Ksp-cre (which causes cre-mediated recombination in developing nephrons and the renal collecting system) was shown to result in lethal cystic kidney disease that was attributed to abnormalities in ciliogenesis (Choi et al., 2013). In a second study, Cdc42 was deleted in the nephron progenitors using Six2-cre. These mice formed hypoplastic kidneys with a reduced nephrogenic zone and small papillae (Reginensi et al., 2013). This phenotype was attributed to defects in localization of the transcriptional activator Yap. Although both studies demonstrate roles for Cdc42 in kidney development, their proposed mechanisms of action for Cdc42 are different and out of keeping with its principal function as a master regulator of the actin cytoskeleton and a mediator of epithelial cell polarity.

There is little known about how apical-basal polarity is established during renal tubulogenesis. Afadin, an F-actin-binding and nectin adaptor protein, has been shown to be essential for the timely initiation of apical-basal polarity during epithelialization of developing nephrons (Yang et al., 2013). In addition, this adaptor protein is essential for the formation of a continuous apical surface and lumen formation (Yang et al., 2013). At the molecular level, absence of afadin in developing nephrons results in a delayed lumen formation and a reduced ability to correctly localize members of the Par complex (Par3–Par6–aPKC–Cdc42; in mammals, Par3 is also known as PARD3 and Par6 as PARD6, with each having two isoforms) to the forming epithelia, suggesting that defects in Par complex function and Cdc42 activation might be the underlying defect. Afadin is necessary for Cdc42 activation in vitro (Kawakatsu et al., 2002; Kurita et al., 2013; Mandai et al., 2013; Nakanishi and Takai, 2004), suggesting that the principal mechanisms whereby Cdc42 regulates tubulogenesis in the kidney might be by altering the actin cytoskeleton and apical-basal polarity of the epithelial cells.

Given the discrepant findings of Cdc42 function in renal epithelial tubulogenesis compared to its demonstrated function in other epithelia, and the recent findings suggesting that afadin regulates Cdc42 and renal epithelial cell polarity, we investigated the role of Cdc42 in the developing kidney, focusing on its role in establishing apical-basal polarity and in cytoskeletal architecture. We used Cdc42^flox/flox mice crossed with two different cre deletor strains; Six2-cre and Hox-B7-cre. Six2-cre deletes Cdc42 in the cap mesenchyme of the metanephric mesenchyme and was used in one of the prior studies (Reginensi et al., 2013). Hoxb7-cre deletes Cdc42 in the ureteric bud, which gives rise to the collecting system starting at embryonic day (E)10.5. We show that Cdc42 is required for both ureteric bud and metanephric mesenchyme development as the Hoxb7-cre:Cdc42^flox/flox mice develop a severe branching morphogenesis defect, and Six2-cre:Cdc42^flox/flox mice are unable to form nephrons. Both mice have serious abnormalities in epithelial cell polarization at all developmental stages examined. In particular, the Six2-cre:Cdc42^flox/flox kidney phenotype is highly reminiscent of the afadin mutants with disruption of Par complex formation and discontinuous lumens. Using Cdc42-null collecting duct cells, we showed that absence of Cdc42 caused profound adhesion, migration, and polarization defects, as well as marked reduction in N-Wasp (also known as WASL) and ezrin activation. Thus, we conclude that a primary function of Cdc42 in the formation of kidney tubules is to promote epithelial cell polarization through Par complex formation and to regulate cytoskeletal architecture through a Cdc42–N-Wasp–ezrin signaling pathway.

We crossed Hoxb7-cre mice, which express cre recombinase in the Wolffian duct and ureteric bud from E10.5 onwards, with Cdc42^flox/flox mice to define the role of Cdc42 in the developing ureteric bud. The mice were born in the predicted Mendelian ratio; however, all the Hoxb7-cre:Cdc42^flox/flox mice died by four weeks of age. Cdc42 deletion was confirmed in mutant mice by performing immunohistochemistry on E15 embryos (Fig. 1A,B) and immunoblotting on isolated papillae of newborn Hoxb7-cre:Cdc42^flox/flox with an antibody directed at Cdc42 (Fig. 1C). The kidneys of the Hoxb7-cre:Cdc42^flox/flox mice were much smaller than their wild-type controls and had histological features of end-stage renal disease characterized by severe destruction of the cortex and medulla (Fig. 1D–I). When Hoxb7-cre:Cdc42^flox/flox mice were killed prior to end-stage renal failure; there was relative preservation of the cortex compared to the severe tubular abnormalities in the medulla. The markedly decreased number of tubules found within the renal papilla of the dysmorphic dysplastic Hoxb7-cre:Cdc42^flox/flox kidneys were consistent with a major branching morphogenesis defect (Fig. 1G). Many lumens of the collecting ducts were not apparent (marked by arrows in Fig. 1H) and there were also areas of marked interstitial fibrosis between the dilated cortical collecting ducts (marked by * in Fig. 1I).

We next defined when the branching defect in the mice began by performing histological studies of kidneys at various embryonic stages. As early as E12.5, there were fewer ureteric bud structures (marked by arrows) and less metanephric mesenchyme induction in mutants compared to wild type (Fig. 2A,B). This became more evident over time. The mutant ureteric bud was very poorly formed at E15.5 (Fig. 2C,D) and E18.5 (Fig. 2E,F) with fewer nephrons at both these time points. The branching defect was confirmed by calbindin staining at E18.5 (Fig. 2G,H). To define why the kidneys of the Hoxb7-cre:Cdc42^flox/flox mice were hypoplastic, we quantified proliferation and apoptosis in the kidney. Interestingly there was no statistical difference in proliferation of the cytokeratin-positive ureteric bud cells in the Hoxb7-cre:Cdc42^flox/flox kidneys as defined by staining for phosphorylated histone H3 (pHH3) at E13.5, E15.5 or E18.5 (Fig. 2I–K). In contrast, when we assessed proliferation of Six2 positive cells at E18.5, there was a 60% decrease in the Hoxb7-cre:Cdc42^flox/flox kidneys (Fig. 2I,J,L) and this difference was also seen in newborn mice (data not shown). There was no difference in apoptosis rates between the two genotypes at any time points (data not shown). Thus, deletion of Cdc42 in the developing ureteric bud caused a severe ureteric bud branching defect and secondarily impaired cell proliferation within the metanephric mesenchyme, which presumably contributed to the reduced kidney size.

Cdc42 was deleted from the metanephric mesenchyme by crossing the Cdc42^flox/flox with the Six2-cre mouse. Six2 is expressed in the cap mesenchyme and this Cre line deletes floxed genes from all components of the developing nephron extending from the glomerular epithelium to the connecting segment (Kobayashi et al., 2008). We verified that Cdc42 was deleted in the developing metanephric mesenchyme by performing immunoblotting and immunohistochemistry on post-natal day (P)1 Cdc42^flox/flox and Six2-cre:Cdc42^flox/flox mice (Fig. 3A–C). Staining revealed clearly decreased Cdc42 expression in the cortical kidney tissue of Six2-cre:Cdc42^flox/flox mice. Because all the Six2-cre:Cdc42^flox/flox mice died within 24 h of birth, we defined when the kidney abnormalities first became evident. Although the kidneys appeared normal at E12.5 (Fig. 3D,E), decreased metanephric mesenchyme development was evident by E15 (data not shown). This defect was clearly obvious at E18.5 (Fig. 3F,G) and P1, where there were markedly decreased numbers of renal vesicles, and comma- and S-shaped bodies (Fig. 3H,I). The few glomeruli that were seen in the mutant mice were hypoplastic and dysplastic. Mutant pups died within a few hours of birth, presumably due to lack of filtration through the hypoplastic and dysplastic nephrons. Lotus Tetragonolobus Lectin (LTL)-positive proximal tubules were greatly reduced in mutants although they were evident and abundant in wild-type controls at E18.5 (Fig. 3J,K). There was 40% less proliferation of Six2-positive cells as defined by pHH3 staining cells in the metanephric mesenchyme of Six2-cre:Cdc42^flox/flox kidneys at E15.5 when compared to Cdc42^flox/flox kidneys (Fig. 3L–N) and there were no differences in apoptosis between the two genotypes (data not shown). Thus, removing Cdc42 in the metanephric mesenchyme resulted in severe hypoplasia and dysplasia of the developing renal vesicles and a profound proliferation defect.

One of the features of the kidneys of the Hoxb7-cre:Cdc42^flox/flox mice was obstruction of the tubular lumens (Fig. 1), which suggested polarity defects in the tubular epithelial cells. As a major function of Cdc42 is to regulate epithelial polarity, we defined whether there were abnormalities in expression of the highly conserved polarity proteins that form the Par complex found in the apical domain of polarized epithelial cells. To do this, we performed immunostaining of two key components of the complex, Par3 and atypical protein kinase C (aPKC) in Hoxb7-cre:Cdc42^flox/flox mice. There was decreased apical staining of aPKC and Par3 in all cells derived from the ureteric bud (E-cadherin, cytokeratin-double positive epithelia) in Hoxb7-cre:Cdc42^flox/flox mice at E15.5 (Fig. 4A–D) and E18.5 (Fig. S1A–D), suggesting that there is a polarity defect in the developing ureteric bud that persists over time. Consistent with these apical polarity abnormalities, there was markedly decreased expression of the tight junction protein ZO-1 (also known as TJP1) in the Hoxb7-cre:Cdc42^flox/flox mice (Fig. S1E,F). The cells within the developing ureteric bud of the Hoxb7-cre:Cdc42^flox/flox mice also had severe defects of terminal differentiation as they did not express the apical water channel aquaporin-2 (Aqp2), whereas the Cdc42^flox/flox mice did (Fig. S1E,F). To verify these terminal differentiation defects of the collecting duct were generalized, we investigated whether there was an altered expression level of the vacuolar-type H ATPase, which is expressed exclusively in the intercalated cells of the collecting duct. This protein was not detectable in the Hoxb7-cre:Cdc42^flox/flox mice (Fig. S2A,B). Finally, we also measured the number of cilia present in the Hoxb7-cre:Cdc42^flox/flox mice as Cdc42 has been shown to be required for cilia production (Zuo et al., 2011). There were 50% fewer cilia in Hoxb7-cre:Cdc42^flox/flox mice compared to Cdc42^flox/flox mice (Fig. S2C–E). Taken together, these data indicate that the polarized epithelial cells in the ureteric bud of the Hoxb7-cre:Cdc42^flox/flox mice have a profound polarity and terminal differentiation defect, which is associated with decreased primary cilia. These cellular abnormalities likely cause intraluminal tubular obstruction that subsequently results in dilated tubules and obstruction of the kidney.

During nephrogenesis, mesenchymal nephron progenitors undergo compaction to form a pretubular aggregate of cells that subsequently form a polarized sphere of epithelia called the renal vesicle, which contains a central lumen. In pretubular aggregates, neural cell adhesion molecule (NCAM, also known as NCAM1) marks the plasma membrane, but becomes restricted to basolateral membranes as a renal vesicle lumen is formed. Pretubular aggregates have little aPKC, and Par-3 is found in dispersed short segments of plasma membrane, called ‘pre-apical domains’ (Yang et al., 2013). As the renal vesicle lumen forms, both aPKC and Par3 become abundant at the apical surface, with Par-3 becoming redistributed to apical and lateral junctions shortly thereafter (Yang et al., 2013). We found that whereas Six2-cre:Cdc42^flox/flox kidneys do form NCAM-positive pre-tubular aggregates, these structures do not progress to a structure with a patent lumen by E15.5 (Fig. 5) and E18.5 (Fig. S3). aPKC is greatly reduced or absent from many cells, whereas NCAM remains relatively uniform in its plasma membrane localization. Similar to what is observed for aPKC, many structures at this stage lack Par-3 or contain only small foci of Par-3. As demonstrated previously (Reginensi et al., 2013), deletion of Cdc42 in Six2-positive cells resulted in a marginal reduction in total YAP (also known as YAP1) expression in the metanephric mesenchyme (Fig. S4A,B, arrow). The localization of the phosphorylated form of YAP, however, remained unchanged (Fig. S4C,D). Thus, Cdc42 is needed for the timely initiation of lumen formation during renal vesicle formation and it is associated with reduced YAP expression.

We also examined the next stage of nephrogenesis, when the renal vesicle elongates into a primordial tubule called the S-shaped body. In the Cdc42^flox/flox kidneys, the elongated S-shaped tubule fuses with the tubule of the ureteric bud and forms a continuous lumen throughout the S-shaped body and ureteric bud at this stage. By contrast, in Six2-cre:Cdc42^flox/flox kidneys there is evidence of lumen formation, but the lumens are short and discontinuous despite the apparent fusion of the S-shaped body to the ureteric bud (Fig. 6). The lumens of mutants remain discontinuous even in the later stages of the formation of the S-shaped body, suggesting it is not simply a delay in their ability to inter-connect. From these studies, we conclude that Cdc42 is required not only for timely lumen initiation, but also for continuous lumen formation during kidney morphogenesis.

Our in vivo findings suggested that deleting Cdc42 resulted in severe abnormalities in tubulogenesis in both the metanephric mesenchyme and the ureteric bud due at least in part to polarity defects. To determine the mechanism of this deficiency, we isolated collecting duct cells from Cdc42^flox/flox mice and deleted Cdc42 in vitro using adenoviral-mediated delivery of a plasmid encoding Cre recombinase. We completed serial dilutions on cell populations and identified cells that did not express Cdc42 by immunoblotting (Fig. 7A). We initially performed experiments on at least three different populations to show that the results obtained were not due to clonal selection (data not shown). When Cdc42^flox/flox collecting duct cells were placed in 3D collagen-I or matrigel gels, they were unable to form tubules and few cells were observed under these conditions (Fig. 7B,C). Mechanistically, collecting duct cells need to be able to adhere, spread and migrate through extracellular matrix to form tubules. We therefore defined which of these processes were defective in the Cdc42^−/− collecting duct cells. We identified a severe adhesion defect in the Cdc42^−/− collecting duct cells on collagen I, laminin-111 (data not shown) and -511 (data not shown) and fibronectin (Fig. 7F), although the cells had normal expression of integrin β1, α1, α2, α3, α5 and α6 subunits (data not shown). Thus, deleting Cdc42 altered cell adhesion of collecting duct cells on multiple extracellular matrix (ECM) components, suggesting this adhesion defect was not integrin specific. Consistent with the adhesion defects, these cells had a profound haptotactic migration defect on collagen 1, laminin-111 (data not shown) and -511 (data not shown) and fibronectin (Fig. 7G). They also had a major spreading defect on collagen 1 (Fig. 7D,E), laminin-111 and -511 and fibronectin (data not shown). Owing to the small number of cells seen in the Cdc42^−/− collecting duct cell tubulogenesis assay, we performed proliferation assays of these cells on multiple ECMs including collagen 1 and fibronectin under low-serum conditions. In agreement with the results seen in vivo and in the in vitro 3D assays, there was a marked decrease in Cdc42^−/− collecting duct cell proliferation that was independent of the ECM substrate (Fig. 7H). Thus, Cdc42 deletion from collecting duct cells results in a severe collecting duct cell tubulogenesis defect due to decreased adhesion, spreading, migration and proliferation.

Mechanistically, Cdc42 regulates multiple cellular processes, including polarization, actin cytoskeletal organization and cell signaling. Based on our in vivo and in vitro data, we initially investigated whether the Cdc42^−/− collecting duct cells had a polarization defect by growing them on Transwell inserts and performing immunostaining for the adherens junction protein E-cadherin, and the tight junction protein ZO-1. Much less of both these proteins localized to the cell–cell junctions of the Cdc42^−/− collecting duct cells when compared to Cdc42^fl/fl collecting duct cells (Fig. 7I–L). We next employed biochemical techniques to determine whether there was decreased expression or mislocalization of E-cadherin, ZO-1 and the apical proteins Par3 and aPKC in the Cdc42^−/− collecting duct cells. There was marked decrease of E-cadherin, ZO-1 and Par3 and a moderate decrease of aPKC in the Cdc42^−/− collecting duct cells (Fig. 7M,N). When we analyzed the relative amounts of these proteins in Triton-soluble and -insoluble cell fractions, there was less E-cadherin, ZO-1, aPKC and Par3 in the Triton-insoluble (membranous) fraction of Cdc42^−/− collecting duct cell lysates relative to Cdc42^fl/fl (control) cell lysates (Fig. 7O–Q). Taken together, these data demonstrate a polarity defect in Cdc42^−/− collecting duct cells, which is consistent with our in vivo findings.

The severe adhesion, migration and spreading defect manifested by the Cdc42^−/− collecting duct cells suggested they have abnormal integrin-dependent signaling. When we investigated this possibility, no differences between Cdc42^f/f and Cdc42^−/− collecting duct cells were found when we assessed activation of well recognized signaling proteins such as those of the Akt or ERK family that act downstream of integrins using re-plating assays on collagen 1 or laminin-111 and -511 (data not shown). Cdc42 has also been reported to be a key regulator of cytoskeletal proteins, such as N-Wasp, which activates the Arp2/3 complex, and ezrin, which binds N-Wasp (Albiges-Rizo et al., 2009). When we defined how Cdc42 regulated the activation of these proteins in response to glial-derived neurotropic factor (GDNF), a principal growth factor responsible for ureteric bud branching morphogenesis, phosphorylation of both these proteins was severely decreased in Cdc42^−/− compared to Cdc42^fl/fl collecting duct cells, despite the same amount of activation of the Ret receptor (Fig. 8A). These data suggest that the severe adhesion, spreading and migration defects observed in the Cdc42^−/− collecting duct cells, are due at least in part to the requirement of Cdc42 for N-Wasp and ezrin activation, and that this signaling defect is not due to decreased Ret activation.

We next examined whether an actin cytoskeleton defect was present in vivo by staining the developing ureteric bud at E15.5 with Rhodamine–phalloidin. Interestingly there was no obvious structural abnormality in the subapical distribution of actin in the Hoxb7-cre:Cdc42^flox/flox mice (Fig. 8B,C). The subapical actin also appeared normal in distribution in the Six2-cre:Cdc42^flox/flox mice except in the areas in which there was no discernible lumen (Fig. 6). As a lack of obvious structural defects does not necessarily indicate normal actin cytoskeleton function, we examined the distribution of ezrin. In Cdc42^flox/flox mice ezrin staining was clearly expressed in the apex of the lumen of the ureteric bud; however, this staining was decreased and diffuse at the apical surface of collecting ducts in the Hoxb7-cre:Cdc42^flox/flox mice (Fig. 8D,E), which is consistent with our in vitro data. Taken together, these data demonstrate that Cdc42 is required for apical actin cytoskeletal function in developing collecting ducts, and for N-Wasp and ezrin signaling in collecting duct cells in vitro.

Cdc42 has diverse cellular functions and plays a role in development of many organs. We show that the absence of Cdc42 in the ureteric bud causes severe branching abnormalities as well as profound polarity and cytoskeletal defects that cause malformed and obstructed tubular lumens. Similarly, deleting Cdc42 from the developing metanephric mesenchyme produces severe abnormalities in epithelial polarity and lumen formation resulting in failed renal vesicle and S-shaped body establishment. We confirmed the requirement for Cdc42 for tubule formation in vitro by utilizing Cdc42-null collecting duct cells, which also showed severe adhesion, migration, proliferation, polarization and cytoskeletal defects. Thus, we conclude that the principal role of Cdc42 in ureteric bud and metanephric mesenchyme development is to regulate epithelial cell polarity and actin cytoskeleton function.

It has previously been shown that deletion of Cdc42 in the distal tubule of the kidney using Ksp-cre induced fatal kidney failure; however, the few surviving mice developed cystogenesis associated with decreased ciliogenesis and interstitial fibrosis (Choi et al., 2013). Most of the characterization of these mice was performed within the cysts where increased cell proliferation and apoptosis associated with increased MAPK pathway activation was noted. No epithelial cell polarization defects were observed, as measured by E-cadherin or gp135 (also known as CNTN1) localization. The discrepancies between this and our studies could be attributed to the fact that Hoxb7-cre deletes Cdc42 at the initiation of ureteric bud branching morphogenesis in the newly forming bud tips at E10.5 (Yu et al., 2002). In contrast, Ksp-cre deletes at E12.5 in the stalk region of the ureteric bud, and from E15.5 onward, in more distal segments of the nephron (Karner et al., 2009; Shao et al., 2002). It is not clear whether the cysts observed in the Ksp-cre mutants are in the collecting duct or the distal tubule. It is interesting to note that similar to Choi et al. (2013), we did observe a decrease in cilia number in mutant collecting ducts. This cellular phenotype was not as penetrant as the defects in polarity and differentiation, suggesting that it might be a secondary manifestation of Cdc42 deletion.

Our morphological results of the Six2-cre:Cdc42^flox/flox in the developing metanephric mesenchyme are very similar to those reported by Reginensi and colleagues (2013), in that the kidneys were severely hypoplastic with almost no functional nephrons. They concluded that the principal mechanism whereby Cdc42 regulated metanephric mesenchyme development was by altering the cytoskeleton through an N-WASP-dependent mechanism that altered the nuclear localization of YAP. Furthermore, they did not comment on a major proliferation or polarity defect and they reported no changes in Par3 or aPKC localization. Indeed, we did find that deletion of Cdc42 in Six2-positive cells resulted in a marginal reduction in total YAP expression in the metanephric mesenchyme (Fig. S4A,B, arrow), although the expression of the phosphorylated form of YAP appeared unchanged (Fig. S4C,D). Our study identified several other major cellular defects, including abnormalities in epithelial cell polarity and proliferation, tubule lumen formation, and actin cytoskeletal function. We postulate these are the primary causes of the defects in terminal differentiation of both the nephron and the collecting duct epithelia. Our study is consistent with the findings in other branching organs, such as the pancreas (Kesavan et al., 2009) and lung (Wan et al., 2013), where epithelial deletion of Cdc42 has been shown to be required for branching, epithelial cell proliferation, tubule formation, normal apical polarity and actin cytoskeleton formation as well as terminal differentiation and cell fate specification. In addition, hepatocytes from Cdc42-null mice are unable to proliferate in response to hepatectomy (Yuan et al., 2009) and deleting Cdc42 from the colon results in a mild defect in apical junction orientation and impaired intestinal epithelium polarity (Melendez et al., 2013), clearly demonstrating a role for Cdc42 in epithelial cell proliferation and polarity.

Our observations in both the ureteric bud and metanephric mesenchyme are very similar to those demonstrated in pancreatic duct formation, where Cdc42 plays a cell-autonomous role in microlumen formation, in generation of apical polarity and in tight junction formation (Kesavan et al., 2009). Therefore, we propose that the tubular lumen and cell differentiation lesions in the Hoxb7-cre:Cdc42^flox/flox and Six2-cre:Cdc42^flox/flox mice are primarily caused by abnormalities in epithelial cell polarity due to defects in Cdc42-dependent aPKC activation through Par6 (Goldstein and Macara, 2007) and abnormal apical actin cytoskeletal function caused by abnormal N-Wasp activation as described in other systems (Rohatgi et al., 2000,, 1999). The Cdc42 Six2-cre mice have a phenotype that is similar to mice lacking afadin: both are defined by their polarity defects and discontinuous lumen formation as evidenced by a decrease in Par3 staining in the pretubular aggregates and/or renal vesicles, and the absence of a discrete lumen at the renal vesicle stage (Yang et al., 2013). The Six2-cre:Cdc42^flox/flox mutants do not pass through a true renal vesicle stage and small discontinuous lumens develop in S-shaped bodies resulting in a persistent defect in luminal continuity. There are increased levels of both Par3 and aPKC in S-shaped bodies of Cdc42 Six2-cre mutants compared with earlier stages. The mechanisms responsible for this ‘recovery’ of the localization of Par complex components are unknown, but it is likely that partially redundant signaling pathways compensate for this process in the absence of Cdc42. Similar to the Cdc42 mutants, afadin mutants also have defects in the actin cytoskeleton as they are unable to recruit the actin subapical ring to form apical surfaces. In addition, afadin mutants fail to form Par complexes. Taken together, these results suggest that Cdc42 and afadin act in a similar pathway to promote polarization and lumen formation. Based on in vitro data from others (Kawakatsu et al., 2002; Kurita et al., 2013; Mandai et al., 2013; Nakanishi and Takai, 2004), we speculate that afadin acts upstream of Cdc42 in the signaling pathway; however, this still needs to be verified.

We used Cdc42^−/− collecting duct cells to demonstrate that Cdc42 is required for in vitro 3D tubulogenesis as well as collecting duct cell adhesion, migration and proliferation on all extracellular matrices. Similar to our in vivo findings, these cells also have a major abnormality in epithelial cell polarity with respect to formation of adherens and tight junctions as well as the Par3–Par6–aPKC complex. Our results are similar to those demonstrating a role for Cdc42 in polarized mammary epithelial cell migration and branching, where Cdc42 was either overexpressed or knocked down in vitro (Duan et al., 2010), and overexpressed in the mouse mammary gland in vivo (Bray et al., 2013). In the kidney tubular epithelium, the functional characteristics of Cdc42 have been explored in the well-characterized polarized MDCK cells. Utilizing overexpression and knockdown techniques, Cdc42 has been shown to be required for lumen formation of tubules because it mediates exocytosis of large intracellular vacuoles that become the lumen and it also binds to the Par3–Par6–aPKC complex, which regulates tight junction formation and epithelial cell polarity (Martin-Belmonte et al., 2007,, 2008; Martin-Belmonte and Mostov, 2007; Rodriguez-Fraticelli et al., 2010). In addition Cdc42 has been shown to play a role in localizing the exocyst to the primary cilium of MDCK cells (Zhang et al., 2001; Zuo et al., 2011). The effects of deleting Cdc42 on MDCK cell tubulogenesis are much less severe than those seen in the Cdc42^−/− collecting duct cells, which never progress to the cyst stage in 3D culture. These data suggest very distinct roles for Cdc42 in tubulogenesis and lumen formation in collecting duct and MDCK cells, and that Cdc42 plays multiple roles in tubule formation in different cell types.

We demonstrate that the principal signaling defect observed in Cdc42^−/− collecting duct cells is their inability to activate N-Wasp and ezrin in response to growth factors. There was also a major defect of apical ezrin expression and localization in collecting ducts in vivo. Taken together, these data suggest that local recruitment of activated Cdc42 and its downstream effector, N-Wasp, can trigger actin polymerization in the apical domain of the epithelial cell (Castellano et al., 1999; Nakamura et al., 2000). This is consistent with the well-defined role of Cdc42 in localizing the Par3–Par6–aPKC complex to the apical part of a polarized epithelial cell (Joberty et al., 2000; Welchman et al., 2007).

In conclusion, we and others have shown that Cdc42 is required for normal tubulogenesis in ureteric bud and metanephric mesenchyme development of the kidney (Choi et al., 2013; Reginensi et al., 2013). Although the morphological observations are similar in all these studies, the mechanisms proposed differ. We demonstrate both in vivo and in vitro that the predominant role of Cdc42 in renal tubulogenesis, as in other branching organs like the lung (Wan et al., 2013) and the pancreas (Kesavan et al., 2009), is to regulate epithelial cell proliferation, polarity and the actin cytoskeleton. Thus, in keeping with its pleiotropic effects, Cdc42 regulates epithelial cell tubulogenesis by multiple mechanisms. This emphasizes the importance of studying its cell autonomous functions in different situations in epithelial cell biology.

All experiments were approved by the Vanderbilt University Institutional Animal Use and Care Committee. Cdc42^flox/flox mice, which were previously described (Wu et al., 2006), were crossed with the Hoxb7Cre mice (generous gift of Andrew McMahon, Department of Stem Cell Biology and Regenerative Medicine, USC Keck School of Medicine, CA) (Yu et al., 2002) or Six2-cre mice (generous gift of Andrew McMahon) (Kobayashi et al., 2008). Mice were a F4–F6 generation towards the C57/B6 background. Aged-matched littermates Cdc42^flox/flox mice were used as controls.

For morphological and immunohistochemical analysis, kidneys were removed at different stages of development and fixed in 4% formaldehyde and embedded in paraffin. Paraffin tissue sections were stained with either hematoxylin and eosin or periodic acid Schiff's (PAS) for morphological evaluation by light microscopy.

Immunofluorescence was performed as previously described (Yang et al., 2013). In brief, fixed sections (4% PFA for 2 h) were permeabilized with 0.3% Triton X-100 in PBS (PBST) and blocked with 10% bovine serum albumin (BSA) in PBST. Sections were incubated with primary antibodies overnight (4°C), then with fluorophore-conjugated secondary antibodies and mounted with Vecatshield (Vector Laboratories). Antibodies used were against Par3 (1:150, Millipore, 07-330), aPKCζ (1:400, Santa Cruz Biotechnology, C-20) and phalloidin 647 and 488 (1:200, Invitrogen, 42008A), YAP and phosphorylated YAP (1:500, Cell Signaling 4912, 4911, tyramide amplification), V-ATPase (1:500, Santa Cruz Biotechnology, 20943), cytokeratin (1:500, Sigma-Aldrich, C2562), LTL (1:500, Vector Labs, B1325), NCAM (1:500, no antigen retrieval), E-cadherin (1:300, Invitrogen), DBA (1:500, Vector Labs, B1035), ezrin (1:500, Upstate 07-130), ZO1 (Santa Cruz Biotechnology, 33725), Aqp2 (Sigma-Aldrich A7310), acetylated αtubulin (Cell Signaling, 5335P) and Six2 (Proteintech, 11562-1). Secondary antibodies were purchased from Jackson ImmunoResearch and Alexa-Fluor-conjugated antibodies from Invitrogen.

Kidney sections obtained at different embryonic stages were subjected to immunohistochemical analysis using primary antibodies against phosphorylated histone H3 (pHH3, 1:2000, Sigma-Aldrich, tyramide amplification), Six2 (1:500, Proteintech) and cytokeratin (1:500, Sigma-Aldrich). More than 250 Six2- or cytokeratin-positive cells were counted from each kidney at each stage and the percentage of pHH3 co-positive cells were calculated. For each stage and genotype a minimum of four different embryos were analyzed.

Collecting duct cells were isolated from Cdc42^fl/fl mice following the methodology described by Husted et al. (1988) and Cdc42 was deleted by infecting the cells with an adenocre virus in vitro. Deletion of Cdc42 was verified by immunoblotting for Cdc42. Clonal Cdc42-null cell lines were generated and similarity in their behavior was verified.

Collecting duct cells were grown in collagen or matrigel gels as previously described (Chen et al., 2004). Collecting duct cells (5×10³) were seeded into the gels, which were overlaid with 100 μl of medium and allowed to grow for 5 to 7 days. The gels were stained with Rhodamine–phalloidin (R415, 1:2000, Molecular Probes) and the tubules were photographed using a Zeiss Axio 510 confocal microscope (400×).

Cells were plated onto slides coated with collagen I (10 μg/ml) for 2 h after which they were fixed, permeabilized, exposed to Rhodamine–phalloidin (1:5000) and visualized under a fluorescence microscope.

Cell adhesion and migration assays on different ECM components were performed as described previously (Chen et al., 2004). Four independent experiments were performed in triplicate.

Proliferation assays were performed as previously described (Chen et al., 2004). Briefly, 5×10³cells were plated on different ECM components in 96-well plates and maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS). After 12 h, the cells were switched to DMEM (2% FBS) for 24 h and then pulsed with 1 µCi/well [3H]thymidine (PerkinElmer Life Sciences). After another 24 h, the cells were solubilized and radioactivity was measured using a scintillation counter.

Cells were grown on Transwell inserts consisting of polyvinylpyrolidone-free polycarbonate filters with 0.4 µm pores. After reaching confluency, cells were fixed in 4% formaldehyde and incubated with anti-ZO-1 (1:200; BD Transduction Laboratories) or anti-E-cadherin (1:1000; BD Transduction Laboratories) antibodies followed by the appropriate FITC-conjugated secondary antibody. Chamber slides were mounted and viewed using a confocal microscope.

The effect of glial-derived neurotropic factor (GDNF) on collecting duct cells was examined as previously described (Zhang et al., 2009). In brief, cells were trypsinized into serum-free DMEM and then plated on collagen I (10 μg/ml) for 45 min. GDNF (10 ng/ml) was added to the medium and the cells were lysed at different time points following growth factor stimulation.

For analysis on kidney tissues, the cortices or medullas were removed and lysed with T-Per buffer (ThermoScientific). Lysates were clarified by centrifugation, and 30 μg total protein was electrophoresed onto a 10% SDS-PAGE gel and subsequently transferred to PVDF membranes. Membranes were blocked in 5% milk with TBS containing 0.1% Tween 20 and then incubated with the different primary antibodies followed by the appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies.

Immunoreactive bands were identified using enhanced chemiluminescence according to the manufacturer's instructions. Antibodies against Cdc42 (2466, 1:1000), ezrin (3142, 1:1000), phosphorylated ezrin (3144, 1:1000) and phosphorylated Ret (3221, 1:1000) were purchased from Cell Signaling; anti-PKCζ antibody was from Upstate (17264, 1:1000), anti-ParD3A (sc-98509, 1:1000) antibody was from Santa Cruz Biotechnology, antibodies against N-Wasp (wp2001, 1:1000) and phosphorylated N-Wasp (wp2601, 1:1000) were from ECM Biosciences, anti-ZO-1 (33-9100, 1:1000) antibody was from Life Technologies and anti-E-cadherin antibody (610181, 1:2000) was from BD Biosciences

Triton-insoluble (actin-rich) and Triton-soluble fractions of collecting duct cells were prepared as described previously (Elias et al., 2014). Briefly, cells were incubated for 5 min with lysis buffer CS (50 mm Tris-HCl, pH 7.4, 1.0% Triton X-100, 5 mm EGTA, and 10 μg/ml protease inhibitor mixture). Cell lysates were centrifuged at 15,600 g for 5 min at 4°C to sediment the high-density actin-rich fraction. The pellet was suspended in 200 μl of lysis buffer D (0.3% SDS in 20 mm Tris-HCl buffer, pH 7.4, and 10 μg/ml protease inhibitor mixture).

The Student's t-test was used for comparisons between two groups, and analysis of variance using Sigma Stat software was used for statistical differences between multiple groups. P<0.05 was considered statistically significant.