Fat1 interacts with Fat4 to regulate neural tube closure, neural progenitor proliferation and apical constriction during mouse brain development

Neural tube closure defects (NTDs) are very common and severe congenital malformations that affect ∼1/1000 pregnancies (Mitchell, 2005). The neural tube forms by thickening of the dorsal surface ectoderm, which folds up and joins at the midline. Neural tube closure requires tight coordination of numerous processes, including polarized apical constriction, proliferation and apoptosis (reviewed by Copp, 2005). Neural tube closure in the cranial region is specifically dependent on actin dynamics. It is believed that the actinomyosin cytoskeleton underlying the apical surface of the neuroepithelium is important for either dorsal lateral hinge point formation or the convergence of neurofolds, or both. This is illustrated by the fact that many actin-regulating proteins, such as Mena (Enah – Mouse Genome Informatics) and Vasp, when deleted from the developing mouse embryo lead to a defect in cranial neural tube closure (reviewed by Copp, 2005; Copp and Greene, 2010). Genes regulating neural progenitor proliferation, neuronal differentiation and apoptosis are also essential for proper neural tube closure, particularly in the cranial region. In addition, mutations in planar cell polarity (PCP) genes lead to severe NTDs, in which the neural tube fails to close along the entire length of the neural tube (Tissir and Goffinet, 2013).

After closure of the neural tube, the mammalian cortex is formed from radial neuroepithelial cells, and later radial glial precursors, the predominant cortical precursor cells during corticogenesis. Radial glial precursors are bipolar cells with apical adherens junctions that line the lateral ventricle, forming an epithelial structure. These radial precursors can divide symmetrically or asymmetrically, producing either an intermediate neuronal progenitor, or a neuron that can then migrate along the radial precursor basal process to form the new neuronal layers of the cortex (Fig. 1A).

Fat cadherins are extremely large cell adhesion molecules, with more than 30 cadherin repeats. In Drosophila there are two Fat family cadherins: Fat and Fat-like (Kugelei – FlyBase). Fat regulates PCP, and restricts tissue growth through the Hippo pathway (reviewed by Sopko and McNeill, 2009; Staley and Irvine, 2012) and via regulation of metabolism (Sing et al., 2014). It has been proposed that Fat binding to another atypical cadherin, Dachsous, leads to dimerization and processing of Fat, ultimately activating an intracellular cascade (Feng and Irvine, 2009; Sopko et al., 2009). Although less studied, Fat-like has been shown to regulate the organization of actin stress fibers in the ovarian follicle cell (Viktorinova et al., 2009). Four mammalian Fat genes have been identified, designated Fat1 to Fat4. Fat is most similar to Fat4, and Fat-like is most similar to Fat1-3. Whereas Fat2 is expressed only in the adult, Fat1, 3 and 4 and their ligands dachsous 1 and 2 are expressed during embryogenesis and are found throughout the developing central nervous system (Rock et al., 2005).

Fat1 mutants die around birth, probably from failure in glomerular slit formation in the kidneys (Ciani et al., 2003). In addition to kidney defects, Fat1 mutants have deformed eyes and craniofacial malformations, with low incidence of holoprosencephaly, which has not been well described or characterized. Fat1 is also involved in muscle formation, and linked to facioscapulohumeral dystrophy (Caruso et al., 2013). At the cellular level, Fat1 interacts with Mena/Vasp proteins to regulate cell migration and actin dynamics in cell culture (Moeller et al., 2004; Tanoue and Takeichi, 2004; Hou and Sibinga, 2009). However, the relevance of this interaction to development has not been explored.

By contrast, more is known about Fat4. Embryos mutant for Fat4 exhibit PCP defects, suggesting that the polarity function of Fat is conserved (Saburi et al., 2008; Mao et al., 2011). The Fat4 intracellular domain can rescue most of the PCP defects of fat mutant flies (Pan et al., 2013). Although structure/function analyses suggest that the Drosophila growth regulatory domain is not conserved in Fat4 (Matakatsu and Blair, 2012; Zhao et al., 2013; Bossuyt et al., 2014), Fat4 restricts the growth of neuronal progenitors in the developing chick neural tube (Van Hateren et al., 2011). Moreover, Fat4 mutations are associated with Van Maldergem syndrome, a recessive multiple malformation syndrome in humans that includes periventricular neuronal heterotopia (Cappello et al., 2013). Fat4 localizes at the subapical domain of neuronal cortical precursors in the mouse, where it regulates the height of the subapical region (Ishiuchi et al., 2009).

Here, we show that loss of Fat1 leads to exencephaly in mouse embryos and to an increase in radial precursor proliferation in the cortex. Knockdown of Fat1 in the developing cortex in utero also leads to overproliferation of radial precursors. At the cellular level, we show that Fat1 is important for actin accumulation at the apical junctions of radial precursors and is needed for apical constriction. We confirm that Fat1 and Fat4 interact genetically in exencephaly, and show additive effects on proliferation in knockdown experiments. Our proteomic analyses reveal that Fat1 and Fat4 bind dramatically different sets of actin-regulating and junctional proteins. As our co-immunoprecipitation experiments suggest that Fat1 and Fat4 form cis-heterodimers, we propose a model in which Fat1 and Fat4 dimer formation coordinates distinct pathways at apical junctions to regulate brain development.

Fat1 is present in the developing central nervous system with strong expression in the neocortex (Ciani et al., 2003). We analyzed in detail Fat1 expression in the developing cortex throughout neurogenesis from E11.5 to E18.5 (Fig. 1B,C; supplementary material Fig. S1A-C). Interestingly, this analysis revealed that Fat1 expression is strongest in the germinal region known as the ventricular and subventricular zone (VZ and SVZ) of the developing cortex at mid-neurogenesis (E13-14) (Fig. 1B,C; supplementary material Fig. S1B). Towards the end of neurogenesis, at E18.5, Fat1 expression can also be seen in the upper neuronal layer of the cortex (supplementary material Fig. S1C). Taking advantage of Fat1-lacZ mice, we performed β-galactosidase staining to show that Fat1 expression colocalizes strongly with the radial precursor markers Pax6 and nestin in the neocortex (Fig. 1D; supplementary material Fig. S1D). This expression pattern suggests that Fat1 is involved in cortical development, possibly regulating cortical precursors.

To elucidate the role of Fat1 in cortical development, we analyzed Fat1 mutant embryos. Neural tube closure normally concludes at ∼E10 with cranial fold and caudal neuropore closure. At E11.5, we find that ∼40% (6/17) of Fat1^−/− embryos in a CD-1 genetic background show open and separated anterior neural folds (Fig. 1E). This defect is specific to the cranial region, since Fat1^−/− embryos have a closed posterior neural tube. Cranial NTDs lead to exencephaly, characterized by an expansion of the brain outside the skull, which is visible as early as E13-14 (Fig. 1F). At E14.5, ∼40% (16/37) of CD-1 Fat1^−/− embryos are exencephalic.

A role for Fat1 in corticogenesis was investigated further by analyzing Fat1 mutant brains at later time points. We first characterized the overall morphology of Fat1 exencephalic brains at mid-neurogenesis (E14-15), when defects in proliferation and differentiation of cortical precursors can be easily assessed. Development of the mouse cortex follows a well-established pattern: radial precursors line the lateral ventricles and form the VZ, and can give rise to more radial precursors, neurons or intermediate neuronal progenitors (Fig. 1A). Above the VZ, intermediate neuronal progenitors divide away from the ventricular surface and populate the SVZ to eventually produce neurons (Englund et al., 2005). Newborn neurons migrate through the intermediate zone (IZ) to form the neuronal layers of the cortical plate (CP).

The morphology of Fat1^−/− exencephalic brains was analyzed by staining coronal sections for the dorsal cortical precursor marker Pax6, the pan-neuronal marker βIII-tubulin, the intermediate neuronal progenitor marker Tbr2 (Eomes – Mouse Genome Informatics), the layer 2-4 neuronal marker Satb2 and the ventral interneuron marker Dlx2 (Fig. 2A; supplementary material Fig. S1E,F). This analysis revealed that the third ventricle is open in Fat1^−/− exencephalic brains and that the thalamic region is expanded to the point of being displaced on top of the cortical hemispheres (Fig. 2A; supplementary material Fig. S1E,F). In contrast to the ‘exposed’ third ventricle of the thalamus, both the lateral ventricles of the cortex are present and appear closed. However, coronal cortical sections stained for Pax6 and Tbr2 revealed a dramatic lateral expansion of the VZ (Fig. 2A). Lateral ventricle length was quantified by measuring the length of the apical VZ lining the dorsal portion of the lateral ventricle, marked by Pax6, from different rostral-caudal coronal sections of Fat1 mutants and control brains. This analysis revealed a ∼2-fold increase in ventricular length (Fig. 2B). This was accompanied by a ∼2-fold increase in the VZ area stained with the radial precursor marker Pax6 (Fig. 2C).

It has previously been shown that lateral expansion of the telencephalon is associated with aberrant proliferation of radial precursors (Chenn and Walsh, 2002). To directly test whether proliferation is altered in Fat1 exencephalic mutants, we examined phospho-histone H3 (pHH3), a marker of mitosis (Fig. 2D). Immunostaining of cortical coronal sections for pHH3 revealed a ∼2-fold increase in the number of pHH3-positive cells per unit ventricular length in Fat1 exencephalic cortices (Fig. 2E), indicating increased proliferation. We note that this analysis probably underestimates the increase in proliferation in Fat1 mutants, since these exencephalic mutants also show increased ventricular length.

Our analysis of Fat1 mutant embryos suggested that Fat1 is involved in the control of radial precursor proliferation and maintenance, but these alterations could be secondary to the NTDs. To examine this possibility, we compared Fat1 exencephalic mutants with Vangl2 mutant embryos, which also exhibit NTDs. Vangl2 is a core PCP pathway component and its mutation leads to classical PCP defects, including an extreme NTD called craniorachischisis, in which the entire neural tube fails to close from the most anterior to posterior regions of the embryo (supplementary material Fig. S2A) (Kibar et al., 2001; Lake and Sokol, 2009). Coronal sections through E14-15 embryos showed that the lateral ventricles of Vangl2 mutants are mostly similar to those of control siblings. In a few cases, Vangl2 mutant ventricles showed some elongation, but never as dramatic as observed in Fat1 mutants (data not shown). Moreover, pHH3 immunostaining of Vangl2 mutants did not reveal any significant changes in the number of mitotic cells (supplementary material Fig. S2A,B). Thus, it is likely that little, if any, of the overproliferation observed in Fat1 mutants can be attributed to NTDs.

To determine the role of Fat1 in radial precursors in the absence of potential genetic compensation mechanisms in Fat1 mutant mice, we used in utero electroporation to acutely knockdown Fat1 in the developing cortex (Fig. 3A). The efficiency of the Fat1 shRNA was first confirmed in cell culture, where it strongly decreased the expression of a tagged murine Fat1 (supplementary material Fig. S3A). Fat1 shRNA was electroporated with a nuclear EGFP expression vector into E13/14 cortices and embryos were harvested 3 days later. This approach transfects radial precursors that form the VZ (Gauthier et al., 2007). Many of these precursors differentiate into neurons over the subsequent 3 days, and migrate through the IZ to the CP (Fig. 3A). Analysis of coronal cortical sections showed that Fat1 knockdown caused mislocalization of EGFP-positive cells, with more cells in the SVZ and IZ compared with control electroporated brains (Fig. 3B,C). We examined whether this change in cell localization reflected a change in precursor number and proliferation by immunostaining similar sections for the radial precursor marker Pax6 and the proliferation marker Ki67 (Fig. 3D,G). Knockdown of Fat1 caused a significant increase in Pax6-positive radial precursors, accompanied by a modest but significant increase in proliferation (Fig. 3E,H). In addition, significantly more of the Pax6-positive cells were localized in the IZ, with correspondingly fewer in the VZ, in Fat1 knockdown brains (Fig. 3F). This coincided with an increase in the proportion of proliferating radial precursors, as monitored by double labeling for Pax6 and Ki67 (Fig. 3I; supplementary material Fig. S3B). Intriguingly, the proliferating radial precursors were localized to the IZ (Fig. 3J; supplementary material Fig. S3B). In contrast to radial precursors, the proportion of Tbr2-positive intermediate neuronal progenitors was unchanged upon Fat1 knockdown (Fig. 3K,L). However, a significant percentage of these coexpressed Pax6 and Tbr2 (Fig. 3M; supplementary material Fig. S3C). Thus, Fat1 knockdown leads to an increase in proliferating radial precursors and an increase in the proportion of intermediate progenitors that still express Pax6.

To determine whether this increase in radial precursor maintenance was accompanied by a decrease in neuronal differentiation, we stained similar sections for the pan-neuronal marker βIII-tubulin and the neuronal layer 2-4 marker Satb2, which marks the majority of neurons that are born during the time frame of our in utero electroporation experiments (Tsui et al., 2013) (Fig. 3N; supplementary material Fig. S3D,E). Knockdown of Fat1 caused a significant reduction in the proportion of neurons positive for EGFP, βIII-tubulin and Satb2 (Fig. 3O,Q) and a mislocalization of these neurons below the CP (Fig. 3P,R; supplementary material Fig. S3D,E). Thus, Fat1 knockdown alters radial precursor proliferation, differentiation and neuronal migration.

Interestingly, previous work showed exencephaly in Fat1^−/− embryos only when Fat4 was also deleted (Saburi et al., 2012). This difference is likely to be due to the different genetic backgrounds of the mouse strains used to breed our Fat1 and Fat4 mutant mice. Indeed, crossing Fat1^+/−; Fat4^+/− mice onto a number of different mice strains confirmed this hypothesis (supplementary material Fig. S4A). We found that the exencephaly phenotype is primarily dependent on loss of both Fat1 alleles, with enhancement from additional loss of Fat4 alleles occurring in a few genetic backgrounds. In outbred CD-1 mice, for example, deletion of Fat1 alone results in exencephaly in ∼20-30% of progeny, with no effects from additional loss of Fat4, whereas the frequency of the phenotype increased from 20% (1/5) to 57% (4/7) upon loss of one allele of Fat4 in a mixed 129/B6 background (supplementary material Fig. S4A). Thus, depending on the genetic strain of the mice, Fat4 and other unknown genetic modifiers play redundant, compensatory roles with regard to the cranial NTD caused by loss of Fat1.

We analyzed Fat1^−/−; Fat4^−/− or Fat1^−/−; Fat4^+/− exencephalic embryos, in the mixed 129/B6 or CD-1 genetic background, by immunostaining of E14-15 coronal cortical sections. We henceforth refer to these as Fat1-Fat4 mutants. Fat1-Fat4 mutants have similar brain morphology to Fat1 mutants (supplementary material Fig. S4B,C; data not shown). Cortical ventricle length was measured by immunostaining similar sections for the cortical precursor markers Sox2 and Pax6, and quantifying the length of the apical VZ lining the lateral ventricle (supplementary material Fig. S4B,C). This analysis showed that Fat1-Fat4 exencephalic lateral ventricles are twice as long as those of control littermates (supplementary material Fig. S4D). Moreover, similar to what we observed in Fat1 mutant exencephalic cortex, a ∼3-fold increase in pHH3 staining was measured in Fat1-Fat4 mutant cortices (supplementary material Fig. S4E,F).

To further analyze cortical precursor proliferation, we injected pregnant Fat1-Fat4 mutant mice at E13.5 with BrdU, a thymidine analog that incorporates into the DNA of cells undergoing S phase. Embryos were harvested at E14.5 and immunostained for BrdU and Ki67 (Fig. 4A). Analysis of BrdU-positive, Ki67-negative cells revealed that the proportion of cells that exited the cell cycle within 24 h following the BrdU injection was significantly decreased in Fat1-Fat4 exencephalic cortices, indicating that more cells remained in the cell cycle (Fig. 4B). To address which cell type is affected, sections were stained for BrdU and the radial precursor marker Sox2 (Fig. 4C). In Fat1-Fat4 exencephalic mutants, there is a significant increase in the proportion of BrdU-labeled cells that are Sox2-positive precursors (Fig. 4D). Co-staining of BrdU with the intermediate neuronal progenitor marker Tbr2 revealed no change in this population of progenitors, suggesting that Fat1 specifically affects proliferation of radial precursors (Fig. 4E,F). Together, these data suggest that the exencephaly observed upon loss of Fat1-Fat4 is due, in part, to an increase in cortical radial precursor proliferation, leading to a laterally expanded cortex.

To examine whether, as suggested by the knockout data, Fat1 and Fat4 interact genetically to regulate radial precursor proliferation, we performed in utero electroporations with Fat4 shRNA. Analysis of cortical sections 3 days post-electroporation showed that, as observed with Fat1 knockdown, Fat4 knockdown leads to the mislocalization of EGFP-positive cells (supplementary material Fig. S5A,B) and an increase in Pax6-positive radial precursors (supplementary material Fig. S5C,D), with a greater number in the IZ and fewer in the VZ compared with controls (supplementary material Fig. S5E). This was accompanied by a mislocalization and decreased number of Satb2-positive neurons (supplementary material Fig. S3D and S5F-H). The proliferation of EGFP-positive cells was not significantly altered, as assessed by Ki67 staining (supplementary material Fig. S5I,J).

To determine whether Fat1 and Fat4 interact genetically to regulate cortical development, we performed concomitant knockdown of Fat1 and Fat4. Analysis of cortical sections 3 days following in utero electroporation of both Fat1 and Fat4 shRNAs revealed a mislocalization of EGFP-positive cells that was similar to that seen with single knockdowns (Fig. 5A,B). Immunostaining for Pax6 showed that knockdown of both Fat1 and Fat4 caused an increase in the proportion of EGFP-positive, Pax6-positive radial precursors (Fig. 5C,D), which were also mislocalized, with a greater number of these cells in the IZ than the VZ compared with controls (Fig. 5E). This increase was greater than that observed upon independent knockdown of either Fat1 or Fat4 alone (shFat1 or shFat4 versus shFat1-4, P<0.05; Ctrl versus shFat1-4, P<0.001). The increase in radial precursors was accompanied by an increase in proliferating, EGFP-positive, Ki67-positive cells (Fig. 5F,G), which was also significantly greater than that seen upon knockdown of Fat1 or Fat4 individually (shFat4 versus shFat1-4, P<0.01; shFat1 versus shFat1-4, P<0.05; Ctrl versus shFat1-4, P<0.001). Coincident Fat1 and Fat4 knockdown also decreased the proportion of EGFP-positive, Satb2-positive neurons, with a greater proportion of these neurons found below the CP (Fig. 5H-J). Thus, Fat1 and Fat4 interact genetically to regulate the proliferation and maintenance of radial precursors.

The apical domain of radial precursors is composed of endfeet with adherens junctions that span the lateral ventricles. Apical junctions have been shown to play a role in symmetric versus asymmetric divisions in radial precursors (reviewed by Fietz and Huttner, 2011; Shitamukai and Matsuzaki, 2012). We therefore analyzed the junctional proteins ZO-1 (Tjp1), Par3 (Pard3), Mupp1 (Mpdz), Pals1 (Mpp5) and β-catenin in cortical sections of Fat1 and Fat1-Fat4 mutant exencephalic embryos (Fig. 6A; supplementary material Fig. S6A; data not shown). Immunostaining for these proteins showed no apparent change in any of these apical markers, indicating that apical junctions are still present in the exencephalic brains.

Apical junctions of radial precursors are connected to a rich actin network that maintains normal apical constriction at the ventricular surface. Since Fat1 has been linked to F-actin regulation (Moeller et al., 2004; Tanoue and Takeichi, 2004), we analyzed the integrity of the apical actin network of these cells by staining sections with Rhodamine-phalloidin. Interestingly, we observed a decrease in F-actin staining at the apical junctions of Fat1-Fat4 exencephalic mutants (Fig. 6B). Quantification of phalloidin intensity at apical cell junctions revealed a 20% decrease in actin accumulation in Fat1-Fat4 exencephalic mutants compared with control siblings (Fig. 6C).

To analyze the morphology of radial precursor apical endfeet and check for associated apical constriction defects, we analyzed thick cortical coronal sections (10 μm). Sections were immunostained for ZO-1 and/or Par3 to delineate the apical domain of radial precursors, and z-stack confocal images were reconstructed to obtain an orthogonal view of the apical surface (Fig. 6A). In Fat1^−/− and Fat1-Fat4 exencephalic cortices, the apical surface area of radial precursors increased by ∼20% compared with control siblings, suggesting that Fat1 regulates the apical constriction of radial precursors, possibly through regulating actin abundance at the apical junctions (Fig. 6D,E; supplementary material Fig. S6B,D).

To test whether this apical expansion phenotype is specific to loss of Fat1, and not a secondary defect due to NTDs, we analyzed Vangl2^−/− embryos, which have a severe NTD (supplementary material Fig. S6C). Vangl2 mutants showed no increase in the apical surface area of cortical precursors compared with control littermates (supplementary material Fig. S6D). This indicates that the apical constriction defect seen in Fat1 and Fat1-Fat4 exencephalic mutants is not simply a consequence of cranial NTDs but is specifically caused by mutations in Fat cadherins.

The additive effects we observed with loss of Fat1 and Fat4 in vivo, and with Fat1 and Fat4 knockdown in utero, suggest that Fat4 and Fat1 interact genetically to regulate cortical precursor maintenance and neural tube closure. All mammalian Fat homologs have very similar extracellular domains; however, their intracellular domains are more divergent, suggesting that they control different intracellular pathways. Alignment of the Fat1 and Fat4 intracellular domains, in particular, revealed only 29% sequence identity, suggesting the additive effects we observed are not solely due to redundancy.

To determine which proteins might mediate the functions of Fat1 and Fat4 in controlling apical constriction and proliferation, we used affinity purification coupled to mass spectrometry to determine the interactome of Fat1 and Fat4. To specifically look for intracellular targets and because Fat cadherins are huge proteins with enormous extracellular domains (>400 kDa), we used Flag-tagged deletion constructs containing the intracellular domains of both proteins (Fat1-ICD-Flag and Fat4-ICD-Flag). Fusion proteins were stably expressed in HEK293 cells, affinity purified using anti-Flag-conjugated beads and subjected to mass spectrometry to identify associated polypeptides. For both Fat1 and Fat4 we focused on a list of potential interactors identified with a total of at least 20 high-confidence peptides (supplementary material Fig. S7).

Interestingly, Fat1 and Fat4 interactomes shared very little overlap, suggesting that they have distinct sets of intracellular partners. Some of these proteins are already known to bind Fat1 (Mena, Vasp and Homer) or Fat4 (Lix1l, Mupp1), validating the biological significance of our results (Moeller et al., 2004; Tanoue and Takeichi, 2004; Ishiuchi et al., 2009). The finding that Fat1 binds Mena and Vasp is exciting, as loss of Mena/Vasp causes NTDs (Menzies et al., 2004), providing a potential mechanism for neural tube defects in Fat1 mutants.

We confirmed these biochemical interactions using immunoprecipitation followed by western blot. This analysis showed that Fat1-ICD-Flag binds strongly to endogenous Mena and Vasp (Fig. 7A; data not shown). We also immunoblotted Fat1-ICD-Flag immunoprecipitates with antibodies to the Fat4 ICD interactor Mupp1. Importantly, Fat1-ICD-Flag does not bind Mupp1, nor is it found in the Fat1 interactome. This analysis confirmed the specificity of our Fat1 and Fat4 interactomes. We also confirmed a strong interaction between Fat4-ICD-Flag and the apical junction protein Mupp1 (Fig. 7A).

Fat4-ICD-Flag immunoprecipitates were also subjected to western blot to test the interaction with a newly identified Fat4 interactor, Par3. Immunoblotting for Fat4 ICD immunoprecipitates with antibodies to endogenous Par3 confirmed the interaction with Fat4 (data not shown). We also immunoblotted Fat4-ICD-Flag immunoprecipitates with antibodies to Mena (Fig. 7A) and found that Fat4-ICD-Flag does not bind to Mena. Together, these data indicate that Fat1 and Fat4 bind a different set of downstream effectors and identify novel interactors for both the Fat1 and Fat4 ICDs.

The identification of the unique interactomes of Fat1 and Fat4 could help explain the divergent roles of Fat1 and Fat4 but does not explain their common functions. In Drosophila, Fat forms cis-homodimers (Feng and Irvine, 2009; Sopko et al., 2009). We examined whether Fat1 and Fat4 also form cis-heterodimers, potentially explaining their common functions and genetic interactions. We transfected HEK293 cells with tagged Fat1 and Fat4 constructs lacking the majority of their extracellular domains but containing the transmembrane and cytoplasmic domains (Fat1-ΔECD-HA and Fat4-ΔECD-Flag). Lysates were immunoprecipitated with a Flag antibody, and western blots were performed with an HA antibody to determine if Fat1 co-immunoprecipitated with Fat4 (Fig. 7B). Importantly, we found that Fat1-ΔECD-HA binds strongly to Fat4-ΔECD-Flag (Fig. 7B), but not to another transmembrane protein used as a negative control. We also found that Fat1-ΔECD-HA does not co-immunoprecipitate with a truncated Fat4 composed solely of the ICD, without the transmembrane domain (Fat4-ICD-Flag), suggesting that Fat4 interacts with Fat1 via its transmembrane domain. Thus, Fat1 and Fat4 can bind to one another, forming cis-heterodimers that could bring together diverse protein partners to regulate neural progenitor proliferation, apical constriction and neural tube closure.

Here, we provide the first characterization of Fat1 function in neural development. We demonstrate that the atypical cadherin Fat1 is crucial for cranial neural tube closure and proliferation of cortical precursors. We show that Fat1 regulates apical constriction and actin accumulation of radial precursors in the developing mouse brain. We also demonstrate that Fat4 can interact with Fat1, and shares some functions with Fat1 during brain development. However, we find that the interactomes of Fat1 and Fat4 are distinctly different. We propose a model in which Fat1 and Fat4 interact at subapical junctions, where they form complexes with actin regulators and apical junction proteins, respectively. We suggest this is important for maintaining radial precursor apical constriction and for regulating the balance between self-renewal and asymmetric neurogenic divisions.

Through our unbiased interactome analysis of Fat1, we show that Fat1 specifically forms a complex with several actin-regulating proteins and we observe a decrease in actin accumulation in Fat1 mutant radial precursors. Interestingly, it has previously been shown that deletion of two actin-regulating interactors, Mena and Vasp, leads to the same cranial neural tube defect as we observe in Fat1 mutants (Menzies et al., 2004). These data suggest that Fat1 regulates neural tube closure by regulating the activity of the Mena/Vasp complex. This also fits with previous data showing that Fat1 regulates actin dynamics and actin-dependent cell motility in cell culture through Mena/Vasp (Moeller et al., 2004). Observation of Mena/Vasp localization through immunostaining did not reveal any changes in the accumulation or localization of these proteins, suggesting that Fat1 is not necessary to simply anchor these actin-regulating proteins at the apical junctions. We speculate that Fat1-Fat4 binding promotes cross-talk between actin-regulating and junctional proteins, leading to the constriction of apical junctions and suppression of growth.

Interestingly, a recent study in zebrafish retina revealed that an increase in the apical surface area of neuroepithelial cells, through inhibition of Shroom3 and Llgl1, could affect neurogenesis (Clark et al., 2012). Moreover, mouse Shroom mutants exhibit NTDs similar to those we show for Fat1 (Hildebrand and Soriano, 1999). Thus, apical constriction might be an important mechanism to regulate the proliferation of various neuroepithelia.

We believe that the radial precursor maintenance and overproliferation phenotype observed in Fat1 and Fat1-Fat4 exencephalic mutants is the direct result of Fat1 and Fat4 deletions and not a secondary defect due to abnormal neural tube closure. First, Fat1 and Fat4 are both expressed at mid-neurogenesis in the frontal cortex, specifically in radial precursors, suggesting they play a role in brain development after neural tube closure (our data; Ishiuchi et al., 2009). Second, acute loss of Fat1 and/or Fat4 in radial precursors by in utero electroporation resulted in enhanced radial glial precursor proliferation that occurs without neural tube closure defects. Third, overproliferation was not observed in the core PCP pathway Vangl2 mutant, which also has severe NTDs.

Recent studies suggest a role for the Yorkie homolog Yap (Yap1) in controlling neural progenitor maintenance downstream of Fat4 (Van Hateren et al., 2011; Cappello et al., 2013). In the developing cortex, Yap is expressed in radial precursors and not in differentiated neurons (Cappello et al., 2013). Immunostaining of Yap in Fat4 and/or Fat1 mutants revealed an expansion of Yap-positive cells that follows the expansion of radial precursors (data not shown). However, we did not observe any changes in Yap localization, nuclear:cytoplasmic ratio, or accumulation that could be indicative of Yap inhibition by the Hippo pathway. Similarly, western blot analyses of embryonic cortex extracts revealed no changes in the phosphorylated inactive forms of Yap or its homolog Taz (Wwtr1 – Mouse Genome Informatics) (data not shown). These data suggest that changes in Yap activity are not responsible for the overproliferation seen in Fat1^−/−; Fat4^−/− mutants.

Radial precursors have a unique morphology, with apical domains that span the lateral ventricles and long basal processes that extend to the pia matter of the cortex. Cell divisions occur at the apical surface of the ventricles, and it is believed that apical junctions and cell division asymmetry play a role in regulating self-renewal versus neurogenic divisions (Bultje et al., 2009; Marthiens and ffrench-Constant, 2009). Our data indicate that Fat cadherins are important in controlling the balance between self-renewal and neurogenic division and do not change the ratio of intermediate neuronal progenitors.

Interestingly, we observe the presence of basally located Pax6-positive radial precursors upon Fat1 and Fat4 knockdown, suggesting that Fat1-4 are involved in anchoring radial precursors at the apical ventricular surface. However, we also see an increase in apical mitosis, suggesting that apical anchoring is not the sole means by which they regulate neuronal radial precursor proliferation.

We note that our in utero electroporation studies also suggest that there are defects in radial migration, which might be linked to defects in the cytoskeleton. More work is needed to determine the basis of these potential migration defects and whether these basally localized Pax6-positive radial precursors could be outer radial glial cells (oRG) (Wang et al., 2011).

Loss of Fat1 and Fat1-Fat4 leads to reduced apical constriction and decreased actin accumulation at cell junctions. Fat cadherins bind both junctional and cytoskeletal proteins, which could explain the reduced apical constriction. We favor the hypothesis that defects in apical constriction contribute to radial precursor overproliferation. Although the link between apical constriction and proliferation in the control of neuroepithelium remains to be elucidated, we speculate that enlargement of apical surfaces increases cell exposure to extracellular proliferation cues from the ventricle. Alternatively, it is possible that lack of junctional constrictions might perturb the asymmetric distribution of cellular components during mitosis and create a bias toward symmetrical divisions. More studies will be needed to elucidate these mechanisms.

We note that only a subset of Fat1^−/−; Fat4^−/− double-mutant embryos show exencephaly and neural progenitor proliferation defects (from 20-57%, depending on the genetic background). We do not see any defects in apical surface area, progenitor proliferation or neural tube closure in non-exencephalic embryos (Fat1^−/−, Fat4^−/− or Fat1^−/−; Fat4^−/−; supplementary material Fig. S8; data not shown). Since we have shown proliferation defects in in utero knockdown experiments and because other neural tube defects do not cause the constriction or proliferation defects, we do not believe that exencephaly causes radial precursor overproliferation or changes in apical junctions. Instead, we think that, in a subset of Fat mutant embryos, the weakening of junctions leads to variable apical constriction defects that cause overproliferation and neural tube closure defects.

Fat1 and Fat4 are relatively divergent, as indicated by the low sequence identity of their ICDs, and this is consistent with our unbiased identification of completely different interactomes for the two proteins. However, our genetic data show that Fat4 and Fat1 share functions in neural tube closure and proliferation control, and our biochemical analysis suggests Fat1 and Fat4 interact in cis. We propose that Fat1 and Fat4 binding regulates radial precursor development. Because the Fat4 interactome includes junctional proteins (Par3, Mupp1), the two bound cadherins are in an ideal position to regulate crosstalk between apical junctions and actin dynamics. Such cooperation is needed throughout development, and especially in brain development where the neuroepithelium has to go through drastic morphological changes to form the complex layered neural cortex, and to titer self-renewal with complex differentiation.

Mouse lines are described in the supplementary Methods. Husbandry and ethical handling of mice were conducted according to guidelines approved by the Canadian Council on Animal Care.

Embryonic samples from timed matings were fixed in 4% paraformaldehyde overnight at 4°C, serially dehydrated and embedded in paraffin. For immunofluorescent analysis, paraffin sections were dewaxed and rehydrated via an ethanol series. Antigen retrieval was performed by boiling the sections for 20 min in Antigen Unmasking Solution (H-3300, Vector Laboratories). After blocking for 1 h in 3% BSA, 10% goat serum and 0.1% Tween 20 in PBS, sections were incubated overnight with primary antibodies diluted in 3% BSA, 3% goat serum, 0.1% Tween 20 in PBS. For details of antibodies, see the supplementary Methods.

Cryosections were prepared as described (Barnabé-Heider et al., 2005). Rhodamine-phalloidin (1:400; Invitrogen) was incubated for 1 h on cryosections.

Fluorescent images were taken with a Nikon C1 Plus Digital Eclipse confocal microscope. For apical surface reconstruction, 0.15 µm z-stack confocal sections were taken and 3D reconstruction obtained with NIS-Elements software (Nikon).

For staining of coronal sections of embryonic brain, a protocol similar to that of Nagy et al. (2007) was followed with minor modifications, as detailed in the supplementary Methods.

Digoxigenin-UTP-labeled Fat1 riboprobes were generated and samples processed according to standard protocols (Komatsu et al., 2014).

Analysis of cell cycle exit by BrdU incorporation is described in the supplementary Methods.

In utero electroporation was performed as described (Gauthier et al., 2007) with E13/14 CD-1 mice, injecting a 1:3 ratio of EGFP plasmid:shRNA (4 μg DNA total); for details of the EGFP nuclear expression plasmid and Fat1, Fat4 and control shRNAs (including validation in HEK293 cells), see the supplementary Methods. Five 50 ms pulses of 40-50 V with 950 ms intervals were delivered per embryo. Brains were dissected 3 days post-transfection and fixed in 4% paraformaldehyde at 4°C overnight, cryopreserved and sectioned coronally at 16 µm. The same anatomical level per embryo was analyzed using an Olympus IX81 inverted fluorescence microscope equipped with an Okogawa CSU X1 spinning disk confocal scan head. A total of three to four sections were analyzed per embryo, and at least three embryos per condition. Stitching of 20× objective images was performed using Volocity (PerkinElmer) software. VZ, SVZ and CP layers were delineated using Hoechst staining.

Cell lines, the lysis protocol and western blot for Fat4 and Mupp1 and for Fat4 and Fat1 co-immunoprecipitation analyses are described in the supplementary Methods.

Affinity purification and identification by mass spectrometry of Fat1 and Fat4 interaction partners were performed as detailed in the supplementary Methods.

Data are expressed as mean±s.e.m. and were tested for statistical significance with Student's t-tests unless otherwise indicated, in which case they were analyzed by Student-Newman-Keuls post-hoc ANOVA. P<0.05 was considered significant. Tests were performed using Excel (Microsoft) or Prism5 (GraphPad).

We thank Dr Charles ffrench-Constant and Dr Phillipe Gros for Fat1 and Vangl2 mutant mice, respectively; Dr Gertler for providing Mena antibody; and Dr Gessler for the Fat1 in situ probe plasmid.