Fat4/Dchs1 signaling between stromal and cap mesenchyme cells influences nephrogenesis and ureteric bud branching

Metanephric kidney development requires interactions between three adjacent cell layers: the epithelial ureteric bud (UB), the cap mesenchyme (CM) and the stromal mesenchyme (stroma) (Fig. 1A) (Costantini and Kopan, 2010; Kopan et al., 2014). The UB forms a tubule that undergoes elongation and reiterative branching to form the collecting ducts. Growth and branching of the UB depends upon reciprocal cell signaling interactions with the neighboring CM, which coalesces around the tips of the branching epithelium. CM cells are also the progenitor cells for nephrons, which are created through a mesenchymal-to-epithelial transition to form a renal vesicle, which then undergoes further growth and morphogenesis (Costantini and Kopan, 2010; Kopan et al., 2014). The CM is surrounded by stroma. Genetic ablation of stromal cells or mutation of stromal-expressed genes, including Ecm1, Foxd1 and Fat4, has revealed that stroma influences the development of the CM and UB (Das et al., 2013; Hatini et al., 1996; Hum et al., 2014; Levinson et al., 2005; Paroly et al., 2013). Nonetheless, there remain substantial gaps in our understanding of the cellular and molecular mechanisms that orchestrate the communication between stromal cells and other cell layers required for kidney morphogenesis. Here, we examine the role of Dchs1/Fat4 signaling in this process.

Much of our understanding of Dchs1 and Fat4 comes from studies of their Drosophila homologs, Dachsous (Ds) and Fat. Ds and Fat are large cadherin family transmembrane proteins that bind to each other to regulate both Hippo signaling and planar cell polarity (PCP) (Matis and Axelrod, 2013; Reddy and Irvine, 2008; Staley and Irvine, 2012; Thomas and Strutt, 2012). Hippo signaling is a conserved signal transduction pathway best known for its influence on organ growth, which it controls by regulating a transcriptional co-activator protein called Yorkie (Yki), or in vertebrates the Yki homologs Yap and Taz (Pan, 2010; Staley and Irvine, 2012). PCP is the polarization of cell morphology and cell behavior within the plane of a tissue (Goodrich and Strutt, 2011; Wansleeben and Meijlink, 2011). PCP signaling is intrinsically bidirectional, as it polarizes each pair of juxtaposed cells. Conversely in Fat/Hippo signaling, Ds acts as a ligand that activates Fat, which functions as a receptor for Hippo signaling (Reddy and Irvine, 2008; Staley and Irvine, 2012), but there is also some evidence for a reciprocal Fat-to-Ds signal (Degoutin et al., 2013).

Analysis of Fat4 and Dchs1 mutant mice has revealed that Dchs1/Fat4 signaling is essential for the morphogenesis of multiple mammalian organs, including the kidney (Mao et al., 2011; Saburi et al., 2008, 2012; Zakaria et al., 2014). Requirements for DCHS1 and FAT4 in humans have been revealed by the linkage of mutations in these genes to Van Maldergem syndrome (Cappello et al., 2013). Mice mutant for Dchs1 or Fat4 have smaller kidneys, with fewer ureteric branches and a modest accumulation of small cysts (Mao et al., 2011; Saburi et al., 2008); hypoplastic kidneys have also been reported in Van Maldergem patients (Mansour et al., 2012). Differences between murine wild-type and Dchs1 or Fat4 mutant kidneys appear as early as embryonic day (E) 11.5, when the growth and branching of the UB in mutants lags behind that in wild-type embryos (Mao et al., 2011).

Differentiation of nephron progenitor cells (CM) into nephrons was reported to be defective in Fat4 mutants (Das et al., 2013), reminiscent of the effect of stromal cell ablation on CM differentiation (Das et al., 2013; Hum et al., 2014), and it was suggested that Fat4 participates in stromal-to-CM signaling. The inhibition of nephron progenitor cell differentiation in Fat4 mutants was attributed to increased Yap activity (Das et al., 2013), although how this might be achieved is unclear, as the molecular pathway linking Fat to Yap identified in Drosophila does not appear to be conserved in mammals (Bossuyt et al., 2014; Pan et al., 2013). Conversely, there is growing evidence that Ds/Fat PCP signaling mechanisms are conserved between insects and vertebrates, including the ability of human FAT4 to rescue PCP phenotypes in flies (Pan et al., 2013) and observations of abnormal cellular polarization in Dchs1 or Fat4 mutant mice (Mao et al., 2011; Saburi et al., 2008; Zakaria et al., 2014).

Here, we focus on the role of Dchs1 in mouse kidney development. We report that Dchs1 mutants share the expansion of CM identified in Fat4 mutants, consistent with the hypothesis that they act as a signaling pair. We also further characterize Dchs1 phenotypes in other cell types within the kidney, and show through conditional deletion that Dchs1 is specifically required within CM for the normal development of CM, UB and stroma. Analysis of genetic mosaics establishes that the subcellular localization of Dchs1 is polarized within CM cells, where it accumulates on surfaces contacting stromal cells. Our observations suggest that Dchs1 functions as a receptor for a Fat4 signal from stromal cells that influences the behavior of CM and, indirectly, that of the UB.

Immunolocalization and RNA in situ studies of E12.5 kidneys revealed that Fat4 and Dchs1 are expressed predominantly within the CM and stroma, rather than in the UB (Mao et al., 2011). Similarly, Fat4 and Dchs1 protein expression could be readily detected within the CM and stroma of E14.5 kidneys (Fig. 1B,D), whereas the UB appears to have levels of staining close to background (i.e. similar levels to those as observed in mutant kidneys, Fig. 1C,E). These observations raise the question of how Fat4 and Dchs1 influence UB branching and growth. To determine where Dchs1 is genetically required for kidney development, we selectively removed Dchs1 from each of the three main cell types within the developing kidney by using a Dchs1 conditional allele (Dchs1^f) combined with either Hoxb7-Cre (expressed in UB) (Yu et al., 2002), Foxd1-Cre (expressed in stroma) (Humphreys et al., 2010) or Six2-Cre (expressed in CM and immediate derivatives) (Kobayashi et al., 2008). The efficacy and specificity of these Cre lines was confirmed using a conditional Rosa-26 lacZ reporter (Soriano, 1999) (supplementary material Fig. S1) and by Dchs1 antibody staining (see below).

When kidneys from postnatal day (P) 0 mice with conditional deletion of Dchs1 in stromal cells or in the UB were examined, they appeared morphologically normal in terms of overall kidney size and shape (Fig. 1F-I,L,M,O,P). Conversely, conditional deletion of Dchs1 from CM resulted in smaller, rounder kidneys, reminiscent of those observed in Dchs1 mutant mice (Fig. 1J,K,N,Q) (Mao et al., 2011). Thus, Dchs1 is required in CM or its derivatives for overall kidney growth, but not in UB or stroma.

In developing wild-type kidneys, mesenchymal cells condense to form a two- to three-cell-wide cap around the UB tips; this cap is marked by strong expression of Six2 (Fig. 2A,C,E,G) (Kobayashi et al., 2008). Mutation of Fat4 results in increased numbers of CM cells around each ureteric bud tip (Fig. 2B,F) (Das et al., 2013). Dchs1 mutants exhibit a similar expansion of CM (Fig. 2D,H), consistent with inferences that Dchs1 and Fat4 function as signaling partners (Mao et al., 2011; Zakaria et al., 2014). In addition to the increase in thickness of the CM, the CM also appears disorganized in both Dchs1 and Fat4, i.e. the width and shape of the cap appears more irregular than in wild type (Fig. 2A-H). Moreover, some CM cells appear to be mislocalized. In wild type they are restricted to the distal side of the ureteric tips, near the surface of the kidney, but in mutants they can accumulate proximal to the ureteric tips (Fig. 2A-D). Increased thickness and disorganization of CM has also been noted in kidneys with stromal deletion, or mutation of Ecm1, Foxd1 or Fat4 (Das et al., 2013; Hatini et al., 1996; Hum et al., 2014; Levinson et al., 2005; Paroly et al., 2013).

To investigate the origin of these defects in CM thickness and organization, we examined CM cells in Fat4 mutants throughout kidney development, beginning at E10.5 when the ureteric epithelium first invades the metanephric mesenchyme. At E11.5, the thickness of cap mesenchyme cells appears similar between wild-type and mutant kidneys (Fig. 2I,J), whereas by E12.5 the CM of Fat4 mutants is clearly abnormal (Fig. 2K,L). Abnormalities in kidney morphogenesis nonetheless appear by E11.5 (Mao et al., 2011), and one intriguing difference revealed by Six2 staining is that in Fat4 mutants the end of the UB remains completely surrounded by CM, whereas in wild type, as the UB elongates, regions of the UB not at the tip (the UB stalk) are no longer directly in contact with Six2-expressing CM (Fig. 2I,J).

To quantify the local increase in CM, we compared the volume of Cited1-expressing cells per UB tip in Fat4 mutant and wild-type kidneys at E14.5; a threefold increase was observed in the mutants (Fig. 2Q). This was not associated with a detectable increase in CM cell proliferation (Fig. 2R-T). We also note that the local increase in CM around each UB tip does not correspond to an increase in total CM within the kidney, because there are fewer, more widely spaced, UB tips. This is illustrated by the observation that total Six2 expression, as measured by quantitative reverse-transcription PCR (Q-PCR), normalized to kidney size (using a ubiquitously expressed gene, GAPDH, as a Q-PCR standard) was similar in wild-type and Fat4 mutant kidneys (110% of wild type; Fig. 3A). The observation that the total amount of CM is similar in wild-type and mutant kidneys (when normalized for kidney size) raises the question of whether Dchs1/Fat4 signaling has a specific effect on CM, or if the increased CM at UB tips might simply reflect a similar amount of CM distributed amongst fewer tips. We do not favor this latter possibility, because the CM and UB tips normally engage in reciprocal signaling to maintain their fates (Kopan et al., 2014). Additionally, we note that Vangl2^Lp mutants also have smaller kidneys with reduced branching (Yates et al., 2010), but they do not exhibit the expansion of CM observed in Dchs1 or Fat4 mutants (supplementary material Fig. S2E,F).

To determine where Dchs1 functions to influence CM cells, we examined kidneys with conditional deletion of Dchs1 in each of the three main cell types present within the early kidney. The expansion and irregular shape of CM was visible in kidneys with conditional deletion of Dchs1 in CM (Six2-Cre Dchs1^f/−; Fig. 2M-P). The phenotype appears to be less severe than in Dchs1 null mutants, but this probably reflects perdurance of Dchs1 mRNA and protein after conditional deletion of the gene, as Six2 expression in the kidney only begins at ∼E10.5 (Oliver et al., 1995). No effect on CM thickness or organization could be detected in kidneys with conditional deletion of Dchs1 in stroma or UB (supplementary material Fig. S2A-D). Thus, Dchs1 is required within the CM to influence CM thickness and organization.

The smaller size of Dchs1 and Fat4 mutant kidneys correlates with reduced branching of the UB (Mao et al., 2011). Based on the density of UB tips in E14-E15 kidneys, branching is reduced by deletion of Dchs1 in CM, but not UB or stroma (Fig. 2P; supplementary material Fig. S2A-D). Reduced branching was confirmed by counting all tips in E12.5 kidneys from mice with conditional deletion of Dchs1 in Six2-expressing cells (Six2-Cre Dchs1^f/−), compared with heterozygous controls (Six2-Cre Dchs1^f/+) (Fig. 3B,D,E). This reduction is less than that observed when comparing null animals with wild-type siblings (Mao et al., 2011), but this could stem from a lag in loss of Dchs1 after conditional deletion, and as deletion of Dchs1 within UB or stroma had no detectable effect on kidney development, we conclude that Dchs1 function within CM influences the growth and branching of the neighboring UB.

The observations that Dchs1 and Fat4 are predominantly expressed in CM and stroma, and that Dchs1 is genetically required in CM, raises the question of how they influence branching of the UB. The UB can be subdivided into distinct cell types based on cell behavior and molecular markers. One fundamental subdivision is between tip cells and non-tip (trunk) cells. Tip cells engage in reciprocal signaling with cap mesenchyme cells to maintain their respective fates (Costantini and Kopan, 2010; Dressler, 2009; Kopan et al., 2014). For example, tip cells express Wnt ligands, including Wnt9b and Wnt11, that signal to CM, whereas CM cells express regulators of Ret and FGF signaling, including GDNF and FGFs, that signal to tip cells. Dchs1/Fat4 signaling is not required for this feedback loop per se, because Gdnf, Ret and Wnt 11 expression levels are all similar between wild-type and mutant kidneys (Fig. 3A) (Mao et al., 2011). However, the local expansion of CM in Dchs1 and Fat4 mutants could potentially increase the number of UB cells exposed to signals from CM, thereby increasing the population of ureteric cells specified as tip rather than trunk. To evaluate this hypothesis, we examined a tip cell marker, Sox9, in Fat4 mutant kidneys versus wild-type kidneys. Indeed, an increase in Sox9-expressing cells could be observed at ureteric bud tips (Fig. 3C,F-I). This expansion in Sox9-expressing cells correlates with loss of staining of a trunk cell marker, the lectin Dolichos biflorus agglutinin (DBA) (Fig. 3J,K). We also quantified the expression levels of genes expressed in tip and trunk cells by Q-PCR (Fig. 3A). Two tip genes, Six2 and Ret, were expressed at similar levels between wild-type and mutant kidneys (Fig. 3A) (Mao et al., 2011). However, E-cad (Cdh1 – Mouse Genome Informatics) levels were reduced (to 70% of wild type) in Fat4 mutant kidneys (Fig. 3A). We interpret this as stemming from the reduction in the fraction of the kidney that comprises the UB (E-cad-expressing cells), consistent with the reduced branching and more widely spaced UB tips observed by confocal microscopy. Taken together, this Q-PCR analysis implies that the fraction of the UB system comprising tip cells (i.e. Sox9- and Ret-expressing cells) is increased in mutants. Direct examination of a trunk marker (Wnt7b) revealed a modest reduction in RNA levels (to 88% of wild type; Fig. 3A). Altogether, these observations are consistent with the hypothesis that the branching defects in Dchs1 and Fat4 could arise as a secondary consequence of the expansion of CM and a promotion of tip cell fate in neighboring UB cells, which could then delay the formation of new branches.

Despite the increased thickness of the CM, the space between clusters of CM is increased in Dchs1 and Fat4 mutants (Fig. 2E-H), concomitant with the decreased branching of the UB. The space between CM is normally occupied by Foxd1-expressing stromal cells. To assess the identity of intervening cells in Dchs1 and Fat4 mutant embryos, a transgene expressing GFP under Foxd1 control was examined in E14.5 Dchs1 or Fat4 mutant kidneys. This revealed that the intervening space between CM is filled with Foxd1-expressing stromal cells in both wild type and mutants (Fig. 3L,M; supplementary material Fig. S2G,H). As the space between clusters of CM remains as narrow as in wild type after conditional deletion of Dchs1 in the UB or stroma (supplementary material Fig. S2A-D), but not after deletion of Dchs1 in CM (Fig. 2P), we infer that this stromal cell phenotype also stems from the requirement for Dchs1 in CM. This local expansion in stroma might simply reflect a similar number of Foxd1-expressing cells filling an expanded space between CM within a smaller mutant kidney.

Alternatively, it could be that Dchs1-Fat4-dependent signals specifically influence stromal cells. The continued expression of Foxd1 implies that normal stromal cell fate does not require Dchs1/Fat4 signaling. We also examined the expression of two genes that act in stromal cells to influence kidney development. Extracellular matrix 1 (Ecm1) is expressed by stromal cells in response to retinoic acid signaling, and influences UB branching at least in part by restricting Ret expression to UB tips (Paroly et al., 2013). Decorin (Dcn) expression is normally repressed by Foxd1 in stromal cells, and misexpression of Dcn contributes to the Foxd1 mutant kidney phenotype (Fetting et al., 2014). However, Q-PCR analysis revealed that both of these genes are expressed at similar levels in wild-type and mutant kidneys (Ecm1 at 84% of wild type, Dcn at 107% of wild type; Fig. 3A).

During early stages of nephrogenesis, CM cells condense to form a pre-cellular aggregate (PA), undergo a mesenchymal-to-epithelial transition to form a renal vesicle (RV), which as it begins to elongate into the nephron tubules transitions through distinctive comma-shaped body (CB) and S-shaped body (SB) stages (supplementary material Fig. S3A). Further elongation and morphogenesis establishes the nephron tubules and glomeruli. Das et al. (2013) reported that nephron differentiation was severely impaired in stroma-less kidneys or Fat4 mutants. By contrast, Hum et al. (2014) reported that stroma-less kidneys had a similar extent of nephron differentiation as did controls, but that differentiated nephron structures were mispositioned. Using Wt1 as a marker, which stains CM, distal regions of the RV, CB and SB, and podocytes, we observed that nephron differentiation still occurs in Fat4 mutants, as visualized by the appearance of Wt1-expressing podocytes (Fig. 4A,B). However, there are abnormalities in early nephron morphogenesis. The shape of the Wt1-expressing portion of the CB and SB is more variable, and often differs from that of wild type (Fig. 4C,D). Conditional deletion of Dchs1 within CM also resulted in abnormal morphogenesis, including heterogeneity in the size and shape of the forming PA, RV and CB (supplementary material Fig. S3).

In addition to strong expression in UB tips, Sox9 is also expressed during early stages of nephrogenesis, when it is visible within connecting tubules and in proximal regions of the RV, CB or SB (Fig. 4G) (Reginensi et al., 2011). Using Sox9 as a marker, in wild-type kidneys Sox9-expressing tubules that extend away from the kidney surface can be observed connecting the UB to the SSB (Fig. 4G, arrow). By contrast, in Dchs1 or Fat4 mutant kidneys, the initial orientation of these connecting tubules is frequently (23/48 scored) abnormal, extending parallel to the surface of kidney (Fig. 4H,I). The ends of these Sox9-expressing tubules are often associated with cysts (Fig. 4I), which become even more evident at later stages (Fig. 4B,F). Accumulation of small cysts in P0 kidneys of Dchs1 and Fat4 mutants has been reported previously (Mao et al., 2011; Saburi et al., 2008); these cysts were identified in aquaporin 2-expressing (marks collecting duct and connecting tubule) and Tamm–Horsfall protein (THP; Umod – Mouse Genome Informatics)-expressing (marks nephron distal tubule and loop of Henle) tubules. We have extended cyst characterization here by identifying cysts in the outer (cortical) region associated with abnormal initiation of nephrogenesis; these cysts can express Sox9 or cytokeratin 8 (CK8, a marker of the UB; Krt8 – Mouse Genome Informatics) (Fig. 4B,D,F,I). We also note that we detect cysts not only at P0, but also at E17.5 and E16.5. This is potentially significant as it was proposed that tubule dilation could be ascribed to defects in oriented cell division (Saburi et al., 2008), but the existence of oriented cell divisions earlier than P0 has been controversial (Karner et al., 2009; Yu et al., 2009). Our observations emphasize that nephrogenesis occurs in Dchs1 and Fat4 mutants, but it is morphologically abnormal, and this abnormal nephron morphogenesis can be associated with cyst formation.

Das et al. (2013) concluded that the local increase in CM thickness in Fat4 mutants stems from impaired differentiation of these nephron progenitor cells, and linked this impaired differentiation to activation of the Hippo pathway transcription factors Yap and Taz. This increased Yap/Taz activity was identified through increased nuclear localization of Yap in CM (Das et al., 2013). When we stained control kidneys, we observed a complex Yap distribution in the kidney, which appears to be consistent with previous studies (Das et al., 2013; Reginensi et al., 2013) and which includes cells with predominantly nuclear (which can be dispersed or concentrated in sub-nuclear foci), predominantly cytoplasmic, and mixed distributions of Yap (Fig. 5A; supplementary material Fig. S4A,C). The specificity of the antisera was confirmed by staining Yap mutant tissues (supplementary material Fig. S4D). However, when we examined Yap localization in either Dchs1 or Fat4 mutants, we were unable to detect an obvious difference in Yap localization between wild-type and mutant CM (Fig. 5B; supplementary material Fig. S4B). We are uncertain of the reason for this discrepancy with Das et al. (2013), but note that we examined Yap localization using both conventional antigen retrieval methods (Fig. 5A,B), as well as using the tyramide signal amplification (TSA) approach employed by Das et al. (2013) (supplementary material Fig. S4A,B).

To evaluate further the potential relationship between Fat4/Dchs1 signaling and Yap/Taz activity, we examined genetic interactions. In Drosophila, in which Fat regulates Yki activity, over-growth phenotypes of fat mutants are suppressed in yki heterozygotes (Reddy and Irvine, 2008). To assess whether a similar genetic interaction could be detected in murine kidneys, we compared Dchs1^−/− or Fat4^−/− kidneys with Dchs1^−/− Yap^−/+ or Fat4^−/− Yap^−/+ kidneys. However, neither overall kidney morphology nor the expansion of CM was visibly altered by loss of one copy of Yap (Fig. 5C-E; supplementary material Fig. S4E,F). Homozygous Yap mutants die at E8.5, but homozygous Taz mutants can survive. However, mutation of Taz did not suppress Fat4 kidney phenotypes, including smaller kidneys, reduced branching, and local expansion of CM (Fig. 5F-M; supplementary material Fig. S4G). Instead, we found that Taz mutants, which have a cystic kidney phenotype on their own (Hossain et al., 2007; Makita et al., 2008; Tian et al., 2007), and Fat4 mutants, exhibit an enhanced cyst phenotype, such that Fat4 Taz double-mutant kidneys were almost entirely filled with large cysts (Fig. 5F-I; supplementary material Fig. S4G). Genetic enhancement between Fat4 and Taz is also reflected in the fact that although we could readily obtain P0 Fat4 or Taz single mutants, Fat4 Taz double mutants did not survive past E18.5. Thus, although earlier studies have revealed that Yap and Taz do play essential roles in kidney development (Das et al., 2013; Hossain et al., 2007; Makita et al., 2008; Reginensi et al., 2013; Tian et al., 2007), we did not obtain evidence supporting a role for Dchs1/Fat4 signaling in repression of Yap/Taz activity in CM.

The determination that Dchs1 functions specifically within CM, and the implication that it participates in or influences signaling between stroma and CM, prompted us to examine its localization in greater detail. Close examination of Dchs1 protein in E14.5 kidneys revealed that it is not uniformly expressed in mesenchymal cells. Instead, its expression appears strongest along the outer edge of the CM, where CM cells abut stromal cells (Fig. 1B; Fig. 6). As Dchs1 is expressed in both stroma and CM, and indeed appears more strongly expressed in stroma, antibody staining in wild-type kidneys could not determine which cells are responsible for the strong Dchs1 protein staining near the CM-stromal interface. Thus, we conducted Dchs1 antibody staining on kidneys with conditional deletion of Dchs1 in either of these cell populations. Conditional deletion of Dchs1 using Six2-Cre led to absence of detectable Dchs1 staining in CM cells, with moderate expression remaining visible in stromal cells (Fig. 6B,D,F). Notably, the normal peak of Dchs1 staining at the stromal-mesenchymal interface was not observed. Conversely, when Dchs1 was deleted specifically from stromal cells using Foxd1-Cre, a peak of Dchs1 expression at the stromal-mesenchymal interface remained visible (Fig. 6H,J,L). This peak of expression does not completely surround CM, but rather appears at discrete locations within a thin focal plane. This accumulation of Dchs1 protein reveals that Dchs1 expressed by CM cells can accumulate at high levels where they contact stromal cells. As Dchs1 in these same cells does not accumulate to similar levels on cell membranes not in contact with stromal cells, these observations identify a molecular polarization of these outer CM cells. Polarization of Ds protein localization has been observed in Drosophila (Ambegaonkar et al., 2012; Bosveld et al., 2012; Brittle et al., 2012) and is thought to be crucial for establishment of PCP by the Ds/Fat pathway, but polarization of Dchs1 in vertebrates has not previously been reported.

As CM condenses into the RV, Dchs1 protein accumulates apically (supplementary material Fig. S3C,E). However, as E-cad also concentrates apically in these cells, this might simply reflect the emerging apical-basal polarity of these cells, rather than a specific role for Dchs1 in RV formation.

Metanephric kidney development requires coordinated signaling among stroma, CM and UB. Our analysis of Dchs1 establishes that it plays an essential role in signaling between the stroma and CM, and that it functions specifically within CM to promote the differentiation and morphogenesis of these nephron progenitors. The importance of stroma-to-CM signaling in kidney development was first revealed by the mutant phenotypes of genes expressed in stromal cells, and by analysis of stroma-less kidneys (Das et al., 2013; Hatini et al., 1996; Hum et al., 2014; Levinson et al., 2005; Paroly et al., 2013). Fat4 was implicated in this process because of the similarity of Fat4 mutant phenotypes to the stroma-less phenotype (Das et al., 2013). Our analysis of Dchs1 and Fat4 reveals that they have similar mutant phenotypes throughout the kidney, resembling those of stroma-less kidneys (Das et al., 2013; Hum et al., 2014). Thus, we infer that Fat4/Dchs1 signaling is an essential mediator of stromal-to-CM signaling. The direct participation of Dchs1 in stromal-CM signaling is further supported by the detection of Dchs1 protein accumulation at the stroma-CM interface. We note that Bagherie-Lachidan et al. (2015) report a contribution of Dchs2 to kidney development, which is revealed by the observation that the Dchs1 kidney phenotype is slightly weaker than the Fat4 kidney phenotype, whereas the Dchs1 Dchs2 double mutant phenotype is equivalent to Fat4.

Tissue-specific conditional deletion reveals that Dchs1 acts specifically within CM to influence kidney development. Through conditional deletion of Fat4, Bagherie-Lachidan et al. (2015) determined that Fat4 acts specifically within Foxd1-expressing cortical stroma. These genetic results imply that they act as signaling partners to effect signaling between stroma and CM, with Fat4 in stromal cells communicating with Dchs1 in CM. As Drosophila homologs of Dchs1 and Fat4 (Ds and Fat) bind each other (Matis and Axelrod, 2013), it is likely that this Fat4-Dchs1-dependent signaling is also mediated by direct binding, and this suggestion is supported by the accumulation of Dchs1 in CM at sites of contact with stroma. Another distinctive feature of Fat4/Dchs1 signaling revealed by our studies is its deployment in communication between distinct cell layers composed of different cell types. In Drosophila, in which Ds/Fat signaling is best known for its role in helping cells interpret and respond to molecular gradients (Matis and Axelrod, 2013; Reddy and Irvine, 2008), Ds/Fat signaling occurs among neighboring cells within an epithelial monolayer. In the mammalian kidney, by contrast, we report here that Fat4/Dchs1 signaling participates in inductive signaling between neighboring mesenchymal cell populations.

Several possible mechanisms by which Fat4/Dchs1 signaling modulates nephrogenesis can be considered based on our current understanding of these genes and their phenotypes (Fig. 7). First, Fat4 and Dchs1 might act as ligand and receptor, respectively, for an intercellular signaling pathway that controls a transcriptional program within CM necessary for nephron morphogenesis (Fig. 7A). Indeed, Das et al. (2013) proposed that Fat4-dependent signaling from stromal cells inhibits CM differentiation by activating the Hippo pathway transcriptional co-activators Yap and Taz. However, neither we nor Bagherie-Lachidan et al. (2015) could detect changes in Yap localization in Fat4 or Dchs1 mutants that would be consistent with increased Yap activation, nor did we find support for this hypothesis in our examinations of genetic interactions between Yap and Fat4 or Dchs1, or between Taz and Fat4. Additionally, we note that although Ds/Fat signaling influences the activity of the Yap homolog Yki in Drosophila, the manner in which it regulates Yki differs from the regulation of Yap reported by Das et al. (2013). In the best-studied Drosophila Fat/Hippo pathway, Ds acts as a ligand and Fat acts as a receptor, such that if this pathway were operating we would have expected Yap activation within stromal cells in Fat4 or Dchs1 mutants, rather than within CM. There is also some evidence in Drosophila for a reverse, Fat-to-Ds signal, but this reverse signal appears to activate rather than inhibit Yki (Degoutin et al., 2013), such that if this pathway were operating we would have expected decreased, rather than increased, Yap activity in Fat4 or Dchs1 mutants. These observations, together with the absence of the key Fat/Hippo pathway transducer Dachs in vertebrates (Bossuyt et al., 2014), and the inability of Fat4 to provide Hippo pathway activity in transgenic Drosophila (Pan et al., 2013), emphasize that if there is transcriptional pathway activated by Fat4/Dchs1 signaling in the kidney, it must involve a mechanism unrelated to any so far described in Drosophila.

A second potential class of models for how Fat4 and Dchs1 influence nephrogenesis invokes signaling from Dchs1 in CM to Fat4 in stroma to induce a transcriptional program that influences stromal cells, including production of additional signals that feedback and influence the CM and/or UB (Fig. 7B). Although this possibility cannot be excluded, we do not favor it, as it would require the existence of additional, as yet unidentified, stromal signals. Known stromal genes that influence kidney development (Foxd1, Ecm1, Dcn) were not significantly affected in Fat4 mutants. Conversely, the hypothesis that Fat4 in stroma signals to Dchs1 in CM does not require invoking additional, as yet unidentified, signals to explain why ablation of stroma phenocopies Fat4 and Dchs1 mutants.

In another class of models, the altered behavior of CM cells and abnormal nephrogenesis could stem from disturbances in cell polarity rather than transcriptional targets of Fat4/Dchs1 signaling (Fig. 7C). In contrast to the lack of evidence for conservation of the connection to Hippo signaling, the influence of Dchs1 and Fat4 on PCP is clearly conserved in mammals (Mao et al., 2011; Pan et al., 2013; Saburi et al., 2008; Zakaria et al., 2014). Our studies have provided further support for this conservation by using genetic mosaics to show that Dchs1 protein can be molecularly polarized in vertebrate cells. In Drosophila, Ds protein localization is polarized by differential expression of Fat in neighboring cells, and this polarization is thought to be crucial for the establishment of PCP (Matis and Axelrod, 2013). Polarization of cellular organization could modulate the responsiveness of CM to additional signals from their neighbors, and/or their ability to signal to other cells. For example, in the Drosophila leg, PCP signaling modulates Notch signaling through effects on Notch receptor endocytosis (Capilla et al., 2012). Interestingly, Notch signaling also plays essential roles in nephrogenesis, as Notch2 mutant cells are unable to form proximal tubules (Cheng et al., 2007). PCP is also well known to influence polarized cell behaviors, including cell migration and cell intercalation, which might be reflected in the mispositioning and misorientation of developing nephron structures that we observe.

In analyzing stroma-less kidneys, Das et al. (2013) reported a frequent block in differentiation, which they attributed to increased activation of Yap and Taz. However, Hum et al. (2014) emphasized the mispositioning of differentiated nephron structures, while reporting that differentiation still occurs. In characterizing Dchs1 and Fat4 mutants, we and Bagherie-Lachidan et al. (2015) frequently observe differentiated nephron structures (e.g. Wt1-expressing podocytes), and also clearly observe cells undergoing early stages of nephrogenesis. However, the location and arrangement of these cells is frequently abnormal, as evidenced by the abnormal location and arrangement of Sox9-expressing cells in the upper SB and Wt1-expressing cells in the lower SB, and the disorganization of forming RV. In some cases, cysts form in cortical regions from cells that express markers of the UB or connecting tubules, possibly because defects in nephrogenesis have left these as blind sacs.

Dchs1 or Fat4 mutations also impair UB growth and branching, even though they are not genetically required in the UB. It could be that Dchs1/Fat4 signaling promotes the production of an as yet unidentified signal from the CM or stroma that represses UB tip cell fate, resulting within mutants in both an increased exposure of CM to UB tip cell signals, and decreased UB branching. However, we favor what we consider to be a simpler hypothesis, as it does not invoke additional, as yet unidentified signals: the Dchs1-Fat4 phenotype in the UB could be accounted for by proposing that the lack of Dchs1 within the CM (or lack of interaction with Fat4) leads to a local expansion of CM, which, by virtue of the known role of CM in maintaining UB tip cell fate, then impairs UB morphogenesis. Consistent with these hypotheses, we observed an expansion of tip cell markers in mutants; a similar expansion of tip markers was recently reported in stroma-less kidneys (Hum et al., 2014). Moreover, differences in contact between CM and UB in mutants are visible as early as E11.5, around the time that branching defects first appear in Fat4 or Dchs1 mutants (Mao et al., 2011). These observations emphasize that normal kidney branching and growth requires a delicate balance between tip and trunk cell fates within the UB. Moreover, the appearance of reduced UB branching as a prominent phenotypic consequence of disturbed stromal-CM signaling highlights the intricate coordination of signaling interactions among tissue layers required for kidney morphogenesis.

Gene-targeted mutations in Fat4 and Dchs1, Vangl2^Lp, and Cre-expressing lines Hoxb7-Cre, Six2-EGFP/Cre and Foxd1-GFP/Cre, have been described previously (Copp et al., 1994; Humphreys et al., 2010; Kobayashi et al., 2008; Mao et al., 2011; Saburi et al., 2008; Yu et al., 2002). Yap heterozygous mice were generated by converting a published conditional allele (Zhang et al., 2010) with Tg(Sox2-Cre)1Amc/J (Jackson Laboratory). Taz (Wwtr1) mutant mice, Cited1-tagRFP and conditional Rosa26-lacZ mice were obtained from the Jackson Laboratory. Animals were housed in an Institutional Animal Care and Use Committee (IACUC)-approved facility using IACUC-approved protocols. Genotyping was performed as described previously (Mao et al., 2011).

E14.0 or older whole kidneys were cut longitudinally into halves with razor blades before antibody staining. For vibratome sections, kidneys were first mounted in 3% low gelling temperature agarose (Sigma, A0701-25G), and cut at 100 µm thickness, and mounted with 100-µm glass beads as spacers to prevent tissue compression. Antibody staining of whole kidneys or sections was performed as described previously (Mao et al., 2011), using the following primary antibodies: goat anti-β-catenin (AF1329, R&D Systems; 1:100), goat anti-E-cadherin (AF748, R&D Systems; 1:400), rabbit anti-Six2 (11562-1-AP, Proteintech; 1:300), mouse anti-Six2 (H00010736-M01, Abnova; 1:400), rabbit anti-Wt1 (SC-192, Santa Cruz; 1:50), rabbit anti-Sox9 (AB5535, Millipore; 1:400), chicken anti-GFP (GFP-1020, Aves lab; 1:1000), rat anti-CK8 (Troma-I, Developmental Studies Hybridoma Bank, Troma-I; 1:200), rat anti-Fat4 (gift of T. Tanoue) (Ishiuchi et al., 2009), rabbit anti-tagRFP (AB233, Evrogen; 1:800) and mouse anti-phospho-Histone H3 (9706, Cell Signaling; 1:400). For Dolichos biflorus agglutinin (DBA) staining, kidneys were fixed with methanol at −20°C, and stained using DBA-biotin (L6533, Sigma; 1:300). For tyramide signal amplification (TSA), slides were treated first for antigen retrieval, and then incubated with primary antibodies, including rabbit anti-Yap1 (4912S, Cell Signaling; 1:400). Slides were then incubated with appropriate fluorescent secondary antibodies (donkey anti-mouse 488, Life Technologies, A21202; and donkey anti-goat 647, Jackson ImmunoResearch, 705-495-147) plus anti-rabbit IgG-biotin (donkey anti-rabbit, Jackson ImmunoResearch, 111-065-152, Jackson ImmunoResearch) to detect Yap1 antibody, and biotin signals were amplified using HRP-streptavidin and Alexa Fluor 546 tyramide (TSA kit, T20933, Life Technologies). Secondary antibodies were from Jackson ImmunoResearch, and stained kidneys were imaged on a Leica SP5 confocal microscope.

For X-gal staining, P0 kidneys were fixed in 4% paraformaldehyde for 4 h at 4°C, then frozen, and sections were cut at 12 µm thickness. Slides were washed twice with PBS plus 2 mM MgCl₂, 0.01% deoxycholate and 0.02% NP-40, and incubated in freshly made and filtered staining solution {2 mM MgCl₂, 5 mM potassium ferrocyanide [K₄Fe(CN)₆-3H₂0] and 5 mM potassium ferricyanide [K₃Fe(CN)₆], 1 mg/ml X-gal} for 17 h at 37°C under darkness, washed again with PBS, and counter-stained with Nuclear Fast Red.

Total RNAs were extracted from E14.5 Fat4 wild-type and mutant kidneys using the Life Technologies PARIS kit. RNA extraction was performed on ten wild-type and 22 mutant kidneys from four litters. cDNAs were made using the SuperScript VILO cDNA Synthesis kit. Quantitative PCR was performed in a Life Technologies StepOnePlus, using SYBR Select Master Mix. Primers are listed in the supplementary material.

For measuring ureteric tip areas, Sox9-stained tips were measured using Leica LAS AF software. Because of the curved nature of kidney surface, only tips in the center area of the scanned field were measured, and the confocal z section at which the tip area was the biggest was used to determine tip area. Bud tips were counted manually from 3D confocal stacks. Statistical tests were performed using Graphpad Prism software. For measuring the total volume of Cited1-tagRFP-positive cells, confocal stacks were analyzed using Volocity (PerkinElmer) software.

We thank E. Kirichenko for assistance with genotyping; D. Pan, H. McNeill, A. McMahon, T. Tanoue and J. Yu for reagents; H. McNeill for communication of unpublished results; and F. Costantini and J. Davies for helpful advice.