Mouse Crumbs3 sustains epithelial tissue morphogenesis in vivo

The human apical protein CRB3 (Crb3 in mouse) organizes epithelial cell polarity. Loss of CRB3 expression increases the tumorogenic potential of cultured epithelial cells and favors metastasis formation in nude mice. These data emphasize the need of in vivo models to study CRB3 functions. Here, we report the phenotypic analysis of a novel Crb3 knockout mouse model. Crb3-deficient newborn mice show improper clearance of airways, suffer from respiratory distress and display perinatal lethality. Crb3 is also essential to maintain apical membrane identity in kidney epithelial cells. Numerous kidney cysts accompany these polarity defects. Impaired differentiation of the apical membrane is also observed in a subset of cells of the intestinal epithelium. This results in improper remodeling of adhesive contacts in the developing intestinal epithelium, thereby leading to villus fusion. We also noted a strong increase in cytoplasmic β-catenin levels in intestinal epithelial cells. β-catenin is a mediator of the Wnt signaling pathway, which is overactivated in the majority of colon cancers. In addition to clarifying the physiologic roles of Crb3, our study highlights that further functional analysis of this protein is likely to provide insights into the etiology of diverse pathologies, including respiratory distress syndrome, polycystic kidney disease and cancer.

transmission were obtained. These mice were mated to FLP deleter mice to remove the Neo cassette. Animals carrying the resulting floxed allele (Crb3 flox ; Fig. 1a) were crossed to a germline Cre deleter mouse line (CMV-Cre). Heterozygous mice (Crb3 +/− ) devoid of the Cre recombinase transgene were and processed for Southern blotting prior to hybridization with the P1 or P2 probes [see (a)], respectively. DNA isolated from wild type (WT) ES cells was used as control. Data confirm homologous recombination at the Crb3 locus. (d) PCR reaction, amplifying the region contained in-between exon 1 and exon 4 of the Crb3 gene (F3 and R3 primers; see (a) for expected size of amplicons), performed on genomic DNA extracted from E18.5 wild type embryos, Crb3 heterozygous embryos (Crb3 +/− ) or Crb3 homozygous mutants (Crb3 −/− ). Cre-mediated recombination efficiently excised exon 2 in mice carrying the Crb3 flox allele. then mated to generate knockout animals (Crb3 −/− ). A PCR reaction amplifying the region comprised between exon 1 and exon 4 confirmed the genomic deletion in heterozygous and homozygous mutant mice (Fig. 1a,d). Consequently, mRNA molecules containing exon 2 were less abundant in Crb3 +/− heterozygous animals and undetectable in Crb3 −/− mutant mice (Fig. 2a). Similarly, we observed a loss of Crb3 protein expression in knockout animals ( Fig. 2b-g). Together, these data show that we have established a conditional Crb3 null allele.
Crb3 is essential for postnatal viability. Crb3 knockout pups were present at parturition, but show signs of respiratory distress such as arrhythmic breathing, spasms of the chest and cyanosis ( Fig. 2i and not shown). We analyzed 158 animals at post-natal day 1 (P1) and found no surviving Crb3 −/− animals. Thus, loss of Crb3 expression is associated with a fully penetrant perinatal lethality, which likely results from impaired respiratory functions. However, Crb3 +/− mice were viable and fertile, and showed no obvious phenotype. To circumvent the postnatal viability issue of Crb3 −/− mutants, we used unborn mice at embryonic day 18.5 (E18.5) to study the impact of Crb3 depletion on general development and epithelial tissue morphogenesis. Crb3 −/− embryos were represented at the expected Mendelian ratio at E18.5 in both mutant lines that we established (Fig. 2h). Loss of Crb3 expression was not associated with any gross anatomical defects (Fig. 2i), and Crb3 −/− embryos had a similar weight to their wild type and Crb3 +/− littermates (Fig. 2j). Overall, these results establish that Crb3 is not necessary for general embryonic development, but is required for postnatal survival.
Loss of Crb3 is associated with an accumulation of mucosubstances in airways. The respiratory distress associated with loss of Crb3 expression suggests lung morphogenesis defects and/or impaired lung physiology in mutant animals. We dissected lungs from wild type, Crb3 +/− and Crb3 −/− embryos at E18.5, and observed no significant difference in lung weight (Fig. 3a). Histological analysis revealed that the overall saccular lung architecture was established in the absence of Crb3 (Fig. 3b,c). However, loss of Crb3 expression caused an atrophy of luminal spaces (compare Fig. 3b,c). Moreover, airways were filled with ectopic debris in Crb3 −/− mutants (Fig. 3c, arrow). These debris were strongly stained by the Periodic Acid Schiff solution (PAS; Fig. 3d,e), which detects polysaccarides, glycoproteins and glycolipids. This suggests that mucus accumulates in airways of Crb3 knockout animals. Immunostaining of the T1α and CC10 markers revealed the presence of type I pneumocytes and Clara cells, respectively ( Fig. 3f-i). Moreover, ciliated cells were present in the lung epithelium of Crb3 −/− embryos, as shown by acetylated-tubulin staining that highlights cilia (Fig. 3j,k). The presence of these typical lung cell types suggests that Crb3 is not essential for cell lineage specification and differentiation in the epithelial compartment of lungs.
CRB3 is required for proper TJ assembly in cultured cells, as alteration of CRB3 expression causes defects in the distribution of the TJ-associated protein ZO-1 21,25,26 . However, the sub-cellular localization of ZO-1 was similar in the lung epithelium of control and Crb3 knockout animals (Fig. 3l,m), thereby suggesting that Crb3 is not necessary for TJ assembly in vivo. We next investigated the polarized architecture of epithelial cell in Crb3-deficient animals by looking at the distribution of the Crb3-associated apical protein Pals1. The apical localization of this protein was preserved in lung epithelial cells devoid of Crb3 (Fig. 3n,o). However, we observed a slight but constant ectopic localization of Pals1 at the lateral membrane in the absence of Crb3 (Fig. 3o, arrowhead). This suggests that a saturable compensatory mechanism maintains Pals1 distribution in the absence of Crb3. We also investigated the subcellular localization of Par3, which is a critical member of the Par polarity complex 11 . Members of this complex are concentrated at TJ and show physical and functional interactions with the Crb3 complex 21,32 . Par3 displayed a similar distribution in wild type and Crb3 −/− lung epithelial cells (Fig. 3p,q). Finally, E-cadherin and β -catenin decorated the lateral membrane in Crb3 mutant lung epithelial cells, as observed in their wild type counterparts ( Fig. 3r-u). E-cadherin and β -catenin staining further revealed that Crb3-deficient cells were cohesive and formed a continuous epithelial tissue. This suggests that the integrity of the lateral membrane and cell-cell adhesion are preserved in the absence of Crb3. Overall, these data show that Crb3 is required for clearance of airways, and thus for lung physiology and animal survival. In addition, our data provide evidence that Crb3 is dispensable for formation of adhesion complexes, including TJ. Finally, the apical-basal axis is established in the lung epithelium of Crb3 mutant animals, although minor polarity defects are observed.
Crb3 knockout mice develop cystic kidneys. CRB3 has been shown to play a role in ciliogenesis in cultured kidney epithelial cells, which display a primary cilium at their apical surface 22,42 . Dysfunctions of primary cilia are associated with the pathogenesis of polycystic kidney diseases (PKD), which are characterized by the presence of many fluid-filled kidney cysts and renal failure 43,44 . Thus, we investigated kidney morphogenesis in Crb3 mutant embryos at E18.5. Kidneys from heterozygous and homozygous Crb3 mutant animals were slightly smaller compared to their wild type counterparts, but the difference in kidney weight was not statistically significant (Fig. 4a). Strikingly, kidneys of Crb3 knockout embryos displayed numerous cysts lined by cells with a flat morphology, whereas normal tubules are composed of cuboidal cells (compared Fig. 4b,c, arrows). Moreover, a loose interstitial tissue that is typical of fibrosis encountered in renal diseases surrounded cysts 45 (Fig. 4c, asterisks). Crb3-depleted kidneys also contained some epithelial tubules exhibiting a relatively normal aspect (Fig. 4c, black arrowheads), and Primers used amplify exon 2, which contains the sole in-frame start codon. Crb3 expression was normalized to actin levels. (b-g) Lung, kidney and intestine isolated from wild type or Crb3 −/− E18.5 embryos were co-stained for Crb3 and E-cadherin (E-cad). Scale bar in b represents 40 μ m and also applies to (c-g). (h) Heterozygous mice of both Crb3 mutant lines that we established (754 and 951) were mated, and progenies were genotyped at E18.5. The number of mice corresponding to each genotype is written in bold, and the percentage relative to the total number of mice analyzed is in brackets. The percentage of each genotype respects the Mendelian ratios.    (18.5). Asterisks denote cysts. Nuclei were stained with DAPI. Scale bar in d represents 40 μ m and also applies to e. (f-u) Kidneys were dissected from wild type or Crb3 knockout (Crb3 −/− ) E18.5 embryos, paraffin-embedded, sectioned and stained for actetylated-tubulin ( Ac Tub; (f,g)), Pals1 (h,i), Par3 (j,k), aPKC (l, m), ZO-1 (n,o), E-cadherin (E-cad; (p,q)), β -catenin (β -cat; r, s) or Na + , K + ATPase (t,u). Scale bar in f represents 20 μ m and also applies to (g-u).
Scientific RepoRts | 5:17699 | DOI: 10.1038/srep17699 glomeruli did not show any obvious defects in the absence of Crb3 (Fig. 4b,c, yellow arrowhead). We stained kidney histological sections of Crb3 −/− embryos with Aqp1 and Aqp2, which stain proximal tubules and collecting ducts, respectively. Most cysts were positive for Aqp1 and negative for Aqp2 (Fig. 4d,e), showing that proximal tubules are especially sensitive to the loss of Crb3. Overall, this histopathological analysis revealed that loss of Crb3 expression causes a phenotype reminiscent of PKD.
As stated above, PKD is often associated with abnormalities of primary cilium function 42 . However, immunostaining of acetylated-tubulin highlighted a similar number of cilia in wild type and Crb3 knockout animals (Fig. 4f,g). In addition, normal and Crb3-deprived cells form a primary cilium of comparable length ( Supplementary Fig. 1). Cell polarity is also impaired in PKD, and recent evidence suggests that loss of epithelial polarity may contribute to cystogenesis 46 . We thus compared the distribution of polarity markers in wild type and Crb3 −/− kidney epithelium. The Crb3-associated protein Pals1 is restricted to the apical domain in wild type epithelial cells (Fig. 4h, arrowhead). Pals1 maintained an apical localization in Crb3-deficient epithelial tissues, but showed decreased levels (Fig. 4i, arrowheads). Loss of Pals1 was particularly evident in epithelial cells lining cysts (Fig. 4i, arrow). Consistent with this finding, Crb3-deficient embryos showed a decreased expression level of Pals1 ( Supplementary Fig. 2). Similarly, the apical distribution of Par3 and aPKC was preserved in tubules devoid of Crb3 ( Fig. 4j-m, red arrowhead), but these proteins were barely detectable at the apical domain of cyst cells (Fig. 4j-m, arrow). Instead, they displayed a cytoplasmic distribution. In addition, aPKC showed ectopic localization at the lateral membrane in the absence of Crb3 (Fig. 4m, yellow arrowheads). The misdistribution of Par3 and aPKC was not associated with a change in their total expression levels ( Supplementary Fig. 2), but it indicates important polarity defects in kidney epithelial cells. This contrasts with the subtle misdistribution of polarity markers observed in the lung epithelium (see Fig. 3). We thus investigated the expression and localization of Crb2, which could potentially compensate for the loss of Crb3. Crb2 mRNA was detected in lungs and kidneys, and its expression levels were comparable in control and Crb3 mutant embryos ( Supplementary Fig. 3a). Crb2 was localized to the apical membrane of lung epithelial cells, whereas the apex of the epithelium lining kidney tubules lacked obvious accumulation of Crb2 (Supplementary Fig. 3b-d). This is consistent with a previous study establishing that Crb2 mRNA expression is strong in glomeruli, but low or absent in the epithelium of kidney tubules 19,47 . We found that Crb2 expression in glomeruli was maintained in Crb3 −/− embryos ( Supplementary Fig. 3h, arrows), and that there is no compensatory upregulation of Crb2 in the kidney epithelium of Crb3 mutant animals ( Supplementary  Fig. 3h, arrowheads). Thus, the lack of a putative functional redundancy with Crb2 in kidney tubules could potentially explain why the apical domain of kidney epithelial cells is especially sensitive to loss of Crb3.
In contrast to Pals1, Par3 and aPKC, ZO-1 displayed a similar distribution in kidney epithelial cells of wild type and Crb3 mutant embryos (Fig. 4n,o). This suggests that the altered distribution of Pals1, Par3 and aPKC results from loss of apical identity or apical differentiation rather than disruption of TJ. We also analyzed the distribution of E-cadherin and β -catenin, which showed a normal lateral distribution in Crb3 knockout animals ( Fig. 4p-s). This suggests that cell-cell adhesion is preserved in the absence of Crb3, and that the identity of the lateral membrane does not require Crb3. The maintenance of baso-lateral membrane identity was further supported by the distribution of the Na + , K + ATPase pump, which was associated with basal and lateral membranes in both control and Crb3-deficient mice (Fig. 4t,u). Together, these data provide evidence that Crb3 is required to maintain epithelial cell polarity in kidneys. Specifically, Crb3 is essential for the localization of several apical proteins, whereas it has no impact on the distribution of tight and adherens junction components. These polarity defects are accompanied by formation of cysts similar to what is encountered in PKD.

Loss of Crb3 expression causes villus fusion in the intestine.
To obtain a broader perspective on the role of Crb3 in vivo, we analyzed the phenotype of intestinal epithelial cells in Crb3 −/− embryos. In contrast to the lung and kidney epithelium, the intestinal epithelium is composed of columnar cells. The intestinal epithelium is continuously self-renewing from epithelial stem cells. Proliferative cells are restricted to the inter-villus epithelium in embryos or to the crypt compartment during postnatal life. Differentiated cells are found on villi, which are finger-like structures protruding in the gut lumen. The majority of cells lining villi are absorptive cells, also known as enterocytes. Crb3 −/− embryos developed an intestine comparable in length to Crb3 +/− and wild type animals (Fig. 5a). In addition, villi were developed in Crb3 knockout embryos at E18.5, and enterocytes adopted a columnar morphology similar to wild type cells (Fig. 5b,c). Consistent with the normal architecture of enterocytes, the distribution of the apical marker Ezrin and of the lateral protein E-cadherin was not obviously affected in most cells of the intestinal epithelium of Crb3 knockout embryos (Fig. 5f-i). Similarly, the TJ protein ZO-1 showed a similar distribution in wild type and Crb3-deficient animals (Fig. 5j,k). However, in the absence of Crb3, Pals1 displayed an ectopic localization in the cytoplasm (Fig. 5l,m). Moreover, while no adhesive contacts were observed between cells of neighboring villi in wild type and Crb3 +/− mice (Fig. 5b,d and Supplementary Fig. 4), many villi were fused in Crb3 −/− mouse embryos. Villus fusion created ectopic multilayered epithelial structures (Fig. 5c,e; asterisks), in which epithelial cells were devoid of the apical markers Ezrin and Pals1 (Fig. 5g,m; arrow). In contrast, E-cadherin distribution was apolar in these cells and covered their entire surface (Fig. 5i,o; arrow). Together, these results demonstrate that Crb3 is essential for proper morphogenesis of the intestinal epithelium.
Scientific RepoRts | 5:17699 | DOI: 10.1038/srep17699 Crb3 knockout mice display high levels of cytoplasmic β-catenin in the intestine. Wnt signaling acts as the dominant mitogen for intestinal epithelial cells 6 . Ectopic activation of Wnt signaling drives colon cancer 6 . The importance of this pathway in the digestive track led us to investigate β -catenin distribution in the intestinal epithelium of Crb3 knockout mice. Consistent with our observation that loss of Crb3 had no impact of E-cadherin distribution outside of villus fusion points (Fig. 5i), β -catenin membrane association was maintained in Crb3 −/− animals at E18.5 (Fig. 6a,b). In addition, the level of β -catenin in villus epithelial cells was comparable in wild type and Crb3 knockout animals (Fig. 6a,b). However, the inter-villus epithelium of Crb3-deficient animals showed a strong accumulation of β -catenin as compared to wild type mice (Fig. 6a,b). Although some β -catenin staining was found in the nucleus of Crb3 mutant cells (Fig. 6b, arrows), β -catenin mostly accumulated in the cytoplasm of these cells. Western blotting confirmed that reduction Crb3 expression caused a dose-dependent increase in β -catenin levels (Fig. 6c). β -catenin-dependent expression of Wnt-target genes favors cell proliferation, which is confined to the inter-villus compartment at E18.5. As β -catenin expression is mostly increased in this compartment in Crb3 knockout embryos, we investigated cell proliferation by quantifying the number of Ki-67 positive cells. The number of proliferative cells was similar in the intestinal epithelium of wild type and Crb3 −/− embryos (Fig. 6d-f). This is consistent with the fact that β -catenin did not show a strong nuclear accumulation in the absence of Crb3 (Fig. 6b). Together, these data demonstrate that loss of Crb3 causes a strong increase in β -catenin expression in the inter-villus epithelium. However, Crb3 deficiency is not sufficient to cause a strong nuclear accumulation of β -catenin or cell proliferation.

Discussion
In this study, we have established a novel conditional Crb3 knockout mouse model and showed that ubiquitous loss of Crb3 expression results in respiratory distress and perinatal lethality. Moreover, our data have demonstrated that Crb3 knockout mice suffer from ectopic accumulation of mucosubstances in lungs, develop cystic kidneys and display fusion of villi in the intestine. Importantly, Whiteman et al. also reported these defects using an independent Crb3 knockout model 23 . While both models allow for a complete loss of Crb3 expression, they result from different modifications of the Crb3 locus. Moreover, we have used a pure C57BL/6 background, whereas Whiteman et al. used a mixed genetic background 23 . Finally, different Cre deleter lines were used to perform germline deletion and to obtain null animals in these studies 23 . Thus, we trust that our phenotypic analysis reveals bona fide Crb3 functions. The availability of complementary in vivo models will ensure a robust, comprehensive and compelling investigation of Crb3 roles.
It was previously shown that reduction of zebrafish Crb3a expression is associated with a shortening of auditory kinocilia 48 . Similarly, knockdown of CRB3B strongly impairs ciliogenesis in MDCK cells 22 . Surprisingly, our analysis did not reveal obvious changes in the number or length of cilia at the surface of lung and kidney epithelial cells. Lack of strong ciliogenesis defects in lungs may result from a functional redundancy with Crb2, which is expressed in lung epithelial cells (this study and references 18,23 ). In support of this premise, zebrafish Crb2b is required for renal cilium formation 48 . Although cilia are formed in the absence of Crb3, it is possible that their ultra-structure and/or functions are impaired. Indeed, Crb3 knockout mice show phenotypes that are suggestive of altered cilia physiology. First, mice devoid of Crb3 display ectopic mucosubstances in airways. This could result from dysfunction of motile cilia, which are required for clearance of the lungs in newborn mice. Of note, zebrafish Crb2b is necessary for cilium motility 48 . Secondly, Crb3 mutant animals develop cystic kidneys. Several lines of evidence support the notion that cystogenesis results from defective primary cilia 42,43 . Thus, as CRB3 localizes to the primary cilium in kidney epithelial cells 22 and that Crb3 loss-of-function is cystogenic, the contribution of Crb3 to the integrity and/or physiology of primary cilia would be worth investigating further. Intracellular signaling pathways initiated by the mechanosensory function of cilia are intimately linked to planar cell polarity (PCP) 44 , which defines the polarized organization of epithelial cells in the plane of the tissue (perpendicular to the apical-basal axis). PCP orients cell division along the longitudinal tubular axis in developing nephrons 44 , thereby ensuring unidirectional tubule elongation. PCP defects randomize the plane of cell division leading to tubule diameter expansion and cyst formation 44,49,50 . Interestingly, knockdown of CRB3B causes mitotic spindle abnormalities and cytokinesis defects 22 . Moreover, we showed that the expression level of Par3 and aPKC is reduced in Crb3-deficient renal cyst cells, and that aPKC is mislocalized in tubules prior to cyst formation. These defects were never reported before. The Par3-aPKC protein module is required to orient the plane of division in kidney epithelial cells 51 . It is thus possible that alteration of oriented cell division contributes to cystogenesis in Crb3 mutant embryos. The decreased level of the apical markers aPKC, Par3 and Pals1 also revealed apical-basal polarity defects in renal epithelial cells of Crb3 knockout animals. This is in agreement with earlier studies in tissue culture showing that CRB3 controls epithelial polarity by acting as an apical determinant 26,27 . These observations are also consistent with data from the literature showing a loss of apical-basal polarity in human PKD and animal models of these pathologies 46,52 . We have also shown that most cysts develop from proximal tubules in Crb3 −/− embryos. The morphogenesis of these tubules requires mesenchymal to epithelial transition (MET). This raises the intriguing possibility that Crb3-deficient animals develop cysts owing to improper MET during the formation of proximal tubules. This hypothesis is consistent with previous studies showing that loss of CRB3 expression is associated with an epithelial to mesenchymal transition (EMT) and loss of the epithelial phenotype 36,[38][39][40]53 . Although additional studies are required to define the mechanisms by which Crb3 maintains the homeostasis and physiology of the kidney epithelium, our data clearly establish that further deciphering of Crb3 functions is crucial for understanding PKD, and perhaps other ciliopathies.
The subcellular distribution of ZO-1 at the apical-lateral border is maintained in Crb3 mutant epithelial tissues. This suggests that Crb3 is not essential for TJ assembly in vivo. This result is in contrast with previous studies showing that alteration of CRB3, PALS1 or PATJ expression leads to TJ defects in cultured epithelial cells 21,25,[29][30][31] . The role of these proteins in TJ formation was often explored using the calcium switch assay in which cell-cell contacts are first broken by chelating extracellular calcium, and then re-established by addition of fresh culture media. This assay thus reflects early steps of TJ biogenesis. Of note, disruption of the CRB3 complex delays but do not prevent TJ formation after calcium switch 26,29,30 . Other studies have used overexpression experiments to explore the role of members of the CRB3 complex in TJ integrity 21,25,26 . These cell culture assays might have revealed a subtle role for CRB3, PALS1 and PATJ in TJ dynamics rather than an essential role for this protein complex in the establishment and maintenance of these junctions. In accordance with this premise, electron microscopy analysis uncovered that Crb3-deficient epithelial cells form TJ in lungs, kidneys and intestine 23 .
Our analysis of the Crb3 mutant phenotype has highlighted abnormal fusion of intestinal villi. During embryogenesis, the gut is first composed of a stratified epithelium surrounding a central lumen 54 . Cells facing the lumen display a typical apical-basal polarity. In contrast, underlying cells show adhesive contacts all-around their surface and lack a clear apical domain 23,55 . Then, this multilayered tissue is transformed to a simple epithelium covering the forming mesenchymal core of villi. This transition results from development and expansion of secondary lumens created by de novo formation of apical membrane free of E-cadherin-based cell-cell contacts 55 . The epithelium at fusion points between adjacent villi remains multilayered in Crb3-deficient animals. Moreover, cells composing this ectopic stratified epithelium lack apical markers, whereas E-cadherin is present on the entire surface of these cells. This suggests that Crb3 contributes to apical domain formation and restriction of adherens junction to the lateral domain in intestinal epithelial cells, thereby allowing for the formation of a monolayered and polarized epithelium. Thus, our in vivo analysis supports previous studies showing that knockdown of CRB3 in MDCK cells interferes with establishment of polarity and lumen formation 27,56 . It was also previously reported that Drosophila Crb is require for polarity in renal epithelial tubules only when morphogenetic movements start 57 . Strikingly, in genetic backgrounds where these movements are stalled, polarity is maintained in renal tubules of crb mutant fly embryos. This suggests that the primary role of Crb is to maintain the polarity phenotype during cell-cell rearrangements. Thus, the complex cell-cell rearrangements taking place when the intestinal epithelium is converted from a multi-layered tissue to a simple epithelium may explain the critical requirement for Crb3 in this tissue. Interestingly, knockout of mouse Ezrin causes a similar intestinal phenotype to the loss of Crb3 expression 55 . Moreover, the cytoplasmic tail of Crb3 interacts with the FERM domain of Ezrin in pulldown assays 23 . This suggests that the physical and/or functional interaction between these two proteins is required for proper morphogenesis of the intestinal epithelium. However, it is also possible that Crb3 and Ezrin act in parallel pathways, as we found that Ezrin is properly localized in the absence of Crb3 in region where villi are not fused. Reciprocally, Crb3 is distributed normally in most cells of the intestinal epithelium of Ezrin −/− mice 55 . Concomitant knockout of Crb3 and Ezrin is required to clarify the functional relationship linking the proteins encoded by these genes.
We also reported for the first time that lack of Crb3 expression is associated with a strong increase in cytoplasmic β -catenin levels in the intestinal inter-villus epithelium. This suggests that Crb3 modulates the Wnt signaling pathway, which promotes proliferation in the gut epithelium. In the absence of Wnt, the cytosolic pool of β -catenin is targeted for degradation by the destruction complex, which contains APC. The binding of a Wnt ligand to its transmembrane receptor inactivates the destruction complex and stabilizes β -catenin, which ultimately translocates to the nucleus and contributes to the expression of Wnt target genes 58 . Over 80% of colon cancers result from loss-of-function mutations in the APC gene and ectopic activation of the Wnt signaling pathway 6,59,60 . Although the knockout of Crb3 is not sufficient to cause nuclear enrichment of β -catenin and excessive proliferation at E18.5, long-term lack of Crb3 expression in the adult intestinal epithelium could favor tumor formation. Indeed, the strong increase in cytoplasmic β -catenin levels in Crb3-deficient cells is likely to act synergistically with genetic alterations supporting nuclear translocation of this crucial mediator of Wnt signalling. Uncoupling of β -catenin stabilization and nuclear accumulation in cancer was supported by analysis of colon cancer samples. Indeed, it was reported that bi-allelic APC mutations are associated with accumulation of β -catenin in the cytoplasm, but are not sufficient to cause detectable nuclear translocation of this protein 61 . Moreover, nuclear β -catenin accumulation is observed in an important fraction of intramucosal tumors and invasive cancers but not in adenomas carrying APC mutations 62 . These data indicate that defective degradation of β -catenin is required but not sufficient for strong expression of Wnt-target genes, and support a model where molecular alterations other than stabilization of β -catenin contributes to a full activation of Wnt signalling and development of aggressive cancers. Thus, loss of Crb3 expression may prime the development of β -catenin-dependent tumors in the intestine.
Overall, our study revealed that Crb3 is essential for postnatal survival in mice. Our results also support a role for Crb3 in maintenance of the apical domain in epithelial cells. Impaired epithelial cell polarity in Crb3 knockout mice is associated with epithelial tissue morphogenesis defects reminiscent of human pathologies such as PKD. Moreover, our study provides evidence that Crb3 limits accumulation of cytoplasmic β -catenin in the intestine, thereby suggesting that Crb3 dysfunctions could predispose to colon cancer. Thus, in addition to providing crucial insights into the function of Crb3 in vivo, the availability of our mouse model to the research community will contribute to the understanding of pathologies associated with polarity defects, including cancer and renal failure.

Methods
Generation of the Crb3 mutant allele. The targeting vector was constructed from a C57BL/6 BAC clone (RP23: 209N12). The target region contains exon 2 and has a size of 768 bp. The long homology arm extends 6.0 kb 5′ to the single LoxP site, which is inserted upstream of exon 2. The short homology arm extends 2.2 kb 3′ of the LoxP/FRT-flanked Neo cassette, which is inserted in intronic sequences downstream of exon 2. Linearized vector (NotI) was transfected by electroporation in C57BL/6N embryonic stem cells. After selection with G418, surviving clones were analyzed by Southern blots and PCR. Two positive clones (754 and 951) were microinjected into Balb/c blastocysts. Resulting chimeras were mated with B6.Cg-Tg(ACTFLPe)9205Dym/J mice (Jackson Laboratory stock number 005703) to remove the Neo cassette and obtain two lines, originating from two independent ES clones, in which exon 2 is floxed by loxP sites (Crb3 flox mice). Exon 2 contains the start codon, and there is no in-frame ATG codon downstream of exon 2. Thus, removal of exon 2 creates a null allele. Crb3 mutant mice were generated at inGenious Targeting Laboratory. Southern blotting. Genomic DNA was digested with XbaI, submitted to electrophoresis on a 0.8% agarose gel and transferred to a nylon membrane. Digested DNA was hybridized with a probe targeting the 5′ external region of the Crb3 locus (probe P1, see Fig. 1a). Primers used to generate the P1 probe were as follow: 5′ -AGT CTT GGC CCA GAA CTC CTG AG-3′ and 5′ -AGT CAG ACC CTA TCT CAG ATA TGC C-3′ . An additional strategy was employed to detect the 3′ internal region of the Crb3 gene on genomic DNA digested with NcoI. The probe used was generated with the following primer pair (probe P2, see Fig. 1a): 5′ -TCA GAA GCT CGG ACA CTC ATA GTG-3′ and 5′ AGT CCC AGG CGT TGG TAG TGA TG-3′ .
Husbandry and generation of Crb3 null animals. Crb3 flox/+ mice were mated to the B6.C-Tg(CMV-cre)1Cgn/J line (CMV-cre; Jackson Laboratory stock number 006054). CMV-cre is active in the germline. Animals carrying the floxed allele and the Cre recombinase were crossed to C57BL/6J mice. Resulting Crb3 heterozygous males and females (Crb3 +/− ) devoid of Cre recombinase were then bred to generate null animals (Crb3 −/− ). Wild type and heterozygous littermates were used as controls.
Statistical analysis. Statistical significance was assessed using the Student's t test or ANOVA analysis of variance. A P value inferior to 0.05 was considered statistically significant.