Functional interactions between Fat family cadherins in tissue morphogenesis and planar polarity

Proper coordination of tissue patterning and growth is essential for development of multicellular organisms. Genetic studies in Drosophila melanogaster revealed a crucial role for the cell adhesion molecule Fat (Ft) in regulating both tissue growth and a form of tissue organization known as planar polarity (Lawrence et al., 2008; Matakatsu and Blair, 2004; Yang et al., 2002). Effects on planar polarity are mediated by a signaling module, comprising the atypical cadherins Ft and Dachsous (Ds) and the kinase Four-jointed (Fj), that lies upstream of, or in parallel to, the ‘core’ planar cell polarity (PCP) elements, such as Frizzled (Fz), Dishevelled (Dvl, or Dsh) and Van Gogh (Vang), and tissue-specific effectors such as RhoA and Inturned (Axelrod, 2009; Goodrich and Strutt, 2011; Gray et al., 2011; Lawrence et al., 2008; Saburi and McNeill, 2005). Core PCP elements and tissue-specific effectors are conserved from flies to humans and are required for tissue organization in mice, including the polarized orientation of sensory hair cells in the cochleae and of hair follicles in the skin (Guo et al., 2004; Montcouquiol et al., 2003; Qian et al., 2007; Wang, J. et al., 2006) as well as convergent extension (CE) movements that occur during gastrulation, cochlear extension and neural tube closure (Axelrod, 2009; Goodrich, 2008; Simons and Mlodzik, 2008). How the Ft signaling module influences these events remains poorly understood.

Recently, we demonstrated that a mammalian homolog of Drosophila ft, Fat4, plays a fundamental role in vertebrate planar polarity (Saburi et al., 2008). Planar polarity is the organization of cells within the plane of a tissue, as exemplified by the orderly arrangement of sensory hair cells in the cochlea and by the coordinated movement of cells during CE. Fat4^–/– mice display distinctive planar polarity phenotypes, including misorientation of hair cells, a short anterior-posterior body axis, short cochlear ducts and broadened spinal cord. Loss of Fat4 also randomizes oriented cell division within renal tubular epithelia in the developing kidney, resulting in cystic kidney disease. Mutating one copy of Vangl2, a homolog of Drosophila Vang, significantly increased the severity of cystic tubular dilation in Fat4^–/– kidneys. Furthermore, a murine homolog of Drosophila ds, dachsous 1 (Dchs1), functions with Fat4 to control tissue patterning (Mao et al., 2011). These findings imply that the Ft/Ds/Fj signaling module might be conserved from flies to humans and that it acts in parallel to a Vangl2-dependent ‘core’ PCP pathway to ensure proper tissue morphogenesis and patterning.

The Fat family of cadherins comprises two family members in Drosophila (ft and ft2) and four in vertebrates (Fat1-4) (Tanoue and Takeichi, 2005). Fat4 exhibits the highest homology to ft, whereas Fat1 and Fat3 are more similar to ft2 [also called kugelei or fat-like (ftl)] (Castillejo-Lopez et al., 2004; Gutzeit et al., 1991). Loss of ft2 in flies results in collapse of tracheal epithelia (Castillejo-Lopez et al., 2004). In contrast to ft, ft2 mutations do not affect planar polarity in eyes and wings or induce hyperplastic overgrowth in imaginal discs. Instead, ft2 mutant flies lose planar alignment of stress fibers in ovarian follicle cells (Viktorinova et al., 2009). These observations indicate significant functional differences between ft and ft2 in flies. Similarly, Fat1^–/– mice do not display planar polarity phenotypes or hyperplastic overgrowths (Ciani et al., 2003). Instead, Fat1 mutants suffer from perinatal lethality accompanied by a failure in glomerular slit formation in kidneys. Fat2^–/– and Fat3^–/– mice also lack classic planar polarity phenotypes, with no discernible phenotypes reported in Fat2 mutants (Barlow et al., 2010). The subcellular distribution of Fat3 in the nervous system suggests a role for Fat3 in neurite fasciculation (Mitsui et al., 2002; Nagae et al., 2007) and retinal cells in Fat3^–/– mutants have altered dendritic morphology (Deans et al., 2011). Fat1 is proposed to control actin dynamics through interactions with Ena/Vasp proteins (Moeller et al., 2004; Tanoue and Takeichi, 2004). Thus, vertebrate orthologs of Drosophila ft2 might function to regulate the cytoskeleton.

Most of what is known about Fat signaling pathways has come from analysis of flies. Drosophila Ft controls planar polarity and fj expression through interactions with the transcriptional co-repressor Atrophin (Atro; Grunge – FlyBase) (Fanto et al., 2003). Loss of atro leads to planar polarity defects in eyes and wings that phenocopy loss of ft. Moreover, atro mutant flies exhibit strong genetic interactions with ft in planar polarity. There are two human orthologs of atro: atrophin 1 (ATN1) and atrophin 2 (ATN2). ATN2 encodes two alternative transcripts: ATN2L (also known as RERE) and ATN2S (Yanagisawa et al., 2000). ATN1 is a transcriptional co-activator (Shen et al., 2007), whereas ATN2L is a transcriptional co-repressor that recruits histone deacetylases (Zoltewicz et al., 2004). The human neurodegenerative disorder dentatorubral-phallidoluysian atrophy is caused by polyglutamine expansion in ATN1 (Yanagisawa et al., 2000). Atn1 null mice have no overt phenotypes (Shen et al., 2007), whereas a truncated Atn1t allele impairs longevity and spermatogenesis (Yu et al., 2009). By contrast, Atn2l loss in ENU-induced openmind (om) mice leads to early embryonic lethality, with severe defects in ventralization of anterior neural plate, heart looping and somitogenesis (Zoltewicz et al., 2004). Fat1 and atrophins co-immunoprecipitate in smooth muscle and together regulate planar alignment of the microtubule-organizing center, directing polarized cell migration (Hou and Sibinga, 2009). Intriguingly, the transcriptional co-activator Atn1 potentiates these events, whereas the transcriptional co-repressor Atn2l inhibits migration.

To gain further insight into the role of Fat signaling during animal development, we asked whether different Fat family genes in mice have overlapping or redundant functions. To date, only Fat4 has been implicated in planar polarity. One possibility is that vertebrate ft-like genes evolved distinct roles, similar to the functional separation of ft and ft2 in flies. Alternatively, the action of Fat cadherins in planar polarity might be masked by their functional redundancies. We tested these hypotheses by exploring whether the distinctive phenotypes in Fat4^–/– mutants would be modified by loss of Fat1 or Fat3. We went on to investigate the molecular basis of Fat4 signaling in diverse morphogenetic contexts by testing genetic interactions with Fjx1, atrophins and Vangl2. Taken together, our genetic studies reveal complex synergistic and antagonistic interactions among Fat family genes, and suggest that Fat4 functions in atrophin-dependent pathways to direct tissue organization and planar polarity during mouse development.

The Fat4^2E11Δflox allele (Saburi et al., 2008) was maintained by backcrossing to 129S3/SvImJ inbred mice (The Jackson Laboratory, USA). Fat1^lacZneo mice were a gift from Dr Charles ffrench-Constant (University of Edinburgh, UK). Vangl2^Lp (Vangl2^S464N) mice were generously provided by Dr Phillipe Gros (McGill University, Canada). Fjx1^lacZ were generously provided by Dr Manfred Gessler (University of Würzburg, Germany). Atn1^Δ and Atn2^om were generous gifts from Dr Andrew Peterson (Genentech, USA). Double heterozygotes were generated by crosses between heterozygous mice, and were maintained by backcrossing to 129S3/SvImJ.

Serial Hematoxylin and Eosin (H&E)-stained longitudinal sections of newborn kidney (4 μm) were prepared as described (Saburi et al., 2008). The entire slide (15-22 sections per kidney) was digitized with Aperio Scanner (Aperio Technologies), and sections were processed for quantification of luminal space with Image-Pro software (Media Cybernetics). For spatial filtering, the lower threshold was set to 50 μm². Sets of luminal space were summed and divided by total volume of kidney sections. To control for genetic background effects, all controls were obtained from siblings in the respective cross.

Inner ears were fixed in 4% paraformaldehyde for 24 hours at 4°C. After fixation, cochleae were incubated with anti-acetylated α-tubulin antibodies (Sigma) in PBS and/or anti-frizzled 6 (Fz6) antibodies (gift of Dr Jeremy Nathans, Johns Hopkins University, USA) as described (Saburi et al., 2008). Fluorescent images were processed for length analysis with Image-Pro software. The cochlea was divided into four regions defined as basal, mid-basal, mid-apical and apical turn. One continuous gain or loss of outer hair cell patches was counted as an emergence.

H&E-stained transverse sections of newborn lumbar vertebra were prepared as described (Saburi et al., 2008). Images were processed for measurements of the distance between vertebral pedicles and between vertebral body and top roof of cartilaginous annulus (see Fig. 1E) with Image-Pro software. Primary numerical data and details of statistical analyses are available at www.mshri.on.ca/mcneill.

Fat4^–/– mice exhibit a range of phenotypes consistent with defective planar polarity, including cystic kidneys, short cochleae and laterally broadened lumbar spinal cords (Saburi et al., 2008). In order to investigate whether other Fat family members and planar polarity signaling components influence Fat4 function in these tissues, we first performed a quantitative analysis of these phenotypes in Fat4^–/– mutants. Fat4^–/– kidneys exhibit multiple anomalies, including cystic dilations of renal tubules, reduced branching of the ureteric bud (UB) and a decrease in organ size (Mao et al., 2011; Saburi et al., 2008) (Fig. 1A,B). Cystic tubular dilations are prominent in Fat4^–/– kidneys, especially in the loops of Henle and collecting ducts (Fig. 1A′,B′).

In Fat4^–/– mutants, cochleae are small and malformed, with modest effects on vestibular apparatus size (supplementary material Fig. S1B,C). Morphometric analysis revealed that both cochleae (P=2.09×10^–10) and vestibules (P=4.93×10^–3) are significantly reduced in size (supplementary material Fig. S1F). However, the cochlea-to-vestibule ratio of Fat4^–/– inner ears is significantly lower than that of wild-type siblings (P=5.57×10^–8; supplementary material Fig. S1G), indicating that loss of Fat4 causes a reduction in the size of both the cochlear and vestibular apparatus, with a stronger impact in the cochleae. Indeed, the cochlea is consistently shorter in Fat4^–/– mutants than in wild type (Saburi et al., 2008) (Fig. 1C′,D′), even after normalization to the vestibule (P=1.47×10^–6; supplementary material Fig. S1G). In addition, the base of the cochleae does not form the proper curvature (supplementary material Fig. S2A,C). These observations suggest that loss of Fat4 results in reduction in cochlear length, which is likely to be due to a failure in elongation rather than to reduction in inner ear size. Although Fat4^–/– newborns are considerably shorter than wild-type newborns (Saburi et al., 2008), they are comparable in body weight to wild-type or Fat4^–/+ siblings (supplementary material Fig. S3A). Therefore, the hypoplastic Fat4^–/– inner ears are likely to result from defects in morphogenesis and not from overall growth retardation.

During cochlear development, one row of inner hair cells (IHCs) and three rows of outer hair cells (OHCs) are patterned with remarkable fidelity along the length of the organ of Corti (OOC) (Fig. 1G,H,I and supplementary material Fig. S3D). Loss of Fat4 causes a modest misorientation of hair cells, predominantly in the third row OHCs (OHC3s) (Mao et al., 2011; Saburi et al., 2008). In addition, discontinuous extra rows of OHCs emerge in mid-apical and, more rarely, in mid-basal turns of Fat4^–/– cochleae (Fig. 1J,K,L), in a fourth row, labeled OHC4s, that typically contains 3-10 cells (Fig. 1J′-L′).

Another site of Fat4 activity is the caudal spinal cord, which is broadened laterally and flanked by vertebrae that are not fully developed and remain unfused upon loss of Fat4 (Fig. 1E,F). The mediolateral vertebral length between the two pedicles is larger in Fat4^–/– mutants than in wild-type siblings [mediolateral-to-dorsoventral (ML/DV) ratio of lumbar vertebra: wild-type 1.40±0.11 (n=30) versus Fat4^–/– 1.62±0.13 (n=18); P=9.72×10^–8; compare Fig. 1F with 1E]. This is consistent with a recent report from Mao et al. (Mao et al., 2011) that vertebral column width is modestly increased in Fat4^–/– mutants exclusively in the lumbar and posterior thoracic region.

The kidney, cochlea and spinal cord phenotypes observed in Fat4^–/– mutants are similar to those of Looptail (Lp) mutant mice. In flies, both Fat signaling and core PCP activity are required for planar polarity. Since most Lp phenotypes reflect disruptions in planar polarity, Fat4 and Vangl2 might work together in vertebrate planar polarity. Most planar polarity phenotypes are more severe in Vangl2 Lp mutants than in Fat4^–/– mutants (Montcouquiol et al., 2003), with the exception of the kidney (Yates et al., 2010). Vangl2 Lp mutants display modest anomalies in kidneys in terms of the cystic tubular dilations, UB branching and reduction in organ size that are typical of Fat4^–/– mutants (compare supplementary material Fig. S4D,E; 129S3 background).

Cystic defects in Fat4^–/– kidneys increase in severity upon introduction of one copy of the Vangl2 Lp mutation (Saburi et al., 2008) in a background of [(129P2/Ola × FVB/N)F1 × BCF1]F2. Exacerbation of cystic defects caused by the Lp mutation were consistent in a 129S3 background (P=2.92×10^–3; Fig. 2C,D,P). Vangl2^Lp/+ heterozygotes do not exhibit obvious kidney anomalies on the 129S3 background (Fig. 2A,B), as demonstrated by quantifying the total dilated luminal space of tubules in the kidneys (P=0.09; Fig. 2P).

During inner ear development, the cochlear epithelium undergoes significant elongation mediated by CE movement (Yamamoto et al., 2009). In Vangl2 Lp mutants, defective PCP signaling disrupts CE and results in short and broad cochlear epithelium (Montcouquiol et al., 2003; Qian et al., 2007; Wang, J. et al., 2006). Although Vangl2^Lp/+ heterozygotes exhibit no change in cochlear length (P=0.52; Fig. 2E,F,Q), Fat4^–/– cochlear elongation defects are significantly enhanced by the presence of a Vangl2 Lp mutation (P=5.92×10^–5; Fig. 2G,H,Q). The vestibular apparatus of Fat4^–/–;Vangl2^Lp/+ inner ears is externally comparable to that of Fat4^–/– siblings (supplementary material Fig. S2J,K,M,N). Intriguingly, the distribution of Fz6 is not disrupted in the Fat4^–/– OOC (supplementary material Fig. S5A′,B′). These findings suggest that Fat4 cooperates with Vangl2 to control cochlear elongation, but independently of the asymmetric subcellular distribution of the core PCP elements.

In addition to cochlear shortening, defects in CE during inner ear development frequently cause aberrant organization of hair cells in the OOC (Montcouquiol et al., 2003; Qian et al., 2007; Wang, J. et al., 2006). To determine whether Fat4 and Vangl2 also act together to pattern the sensory epithelium, we investigated whether the extra OHCs (OHC4s) seen in Fat4^–/– cochleae increase in frequency upon introduction of a Vangl2 Lp mutation. Importantly, Vangl2^Lp/+ heterozygotes show no specific defects in the OOC (Fig. 2S). However, patches of OHC4s emerged more frequently in the mid-apical turn of Fat4^–/–;Vangl2^Lp/+ cochleae than in Fat4^–/– mutants (P=7.54×10^–3; Fig. 2M,N,O,S). Moreover, ectopic OHC4s are often continuous exclusively in the mid-apical region of Fat4^–/–;Vangl2^Lp/+ cochleae (Fig. 2S). Thus, synergy between Fat4 and Vangl2 influences both cochlear lengthening and patterning. The modest misorientation of hair cells in the Fat4^–/– OOC is not detectably exacerbated upon introduction of a Vangl2 Lp mutation (Fig. 2M-O).

One of the most dramatic phenotypes observed in Vangl2 Lp mutants is craniorachischisis, characterized by anencephaly accompanied by spina bifida (Montcouquiol et al., 2003) (supplementary material Fig. S4E). The spinal cord and vertebral column are normal in Vangl2^Lp/+ newborns, indicating that there is no Vangl2 haploinsufficiency in caudal neural tube and axial skeleton morphogenesis (ML/DV ratio of lumbar vertebra, P=0.20; Fig. 2I,J,R; data not shown). Fat4^–/–;Vangl2^Lp/+ mutants are born without detectable malformations in the cranial neural tube (supplementary material Table S1). In addition, the presence of the Vangl2 Lp mutation did not consistently exacerbate Fat4^–/– spinal cord or vertebral phenotypes (Fig. 2K,L,R; data not shown).

Many phenotypes observed in Fat4^–/– mutants are modest compared with those of core PCP mutants, raising the possibility that Fat4 functions in a redundant manner with other Fat family cadherins. To examine functional redundancies among Fat family members, we tested whether cystic defects in Fat4^–/– kidneys increase in severity upon loss of one copy of Fat1. During kidney development, Fat1 is expressed in the UB epithelium (Rock et al., 2005). Fat1^–/– kidneys do not exhibit visible dilations of renal tubules or reduction in organ size (supplementary material Fig. S4A,B), nor is there evidence for haploinsufficiency, as measured by quantification of total renal tubular luminal space (P=0.88; Fig. 3A,B,P). Strikingly, however, removal of one copy of Fat1 significantly enhances cystic tubular dilations of Fat4^–/– kidneys (P=2.83×10^–2; Fig. 3C,D,P), with defects prominent in loops of Henle and collecting ducts (Fig. 3D).

To determine whether Fat4 also synergizes with Fat1 during cochlear elongation, we quantified Fat4^–/– cochlea length in a Fat1^–/+ background. Although Fat1^–/– homozygotes occasionally exhibit aberrant round ears (supplementary material Fig. S2A,B,E,F), no defects in cochlear lengthening were observed in Fat1^–/+ heterozygotes compared with wild type (P=0.46; Fig. 3E,F,Q). However, upon loss of one copy of Fat1, Fat4^–/– cochleae became slightly shorter than in Fat4^–/– mutants (P=1.03×10^–2; Fig. 3G,H,Q).

Next, we asked whether synergy between Fat1 and Fat4 also governs patterning of OHCs. Fat1^–/+ heterozygotes show neither alterations of row numbers nor misorientations of hair cells in the OOC (Fig. 3S; data not shown). However, in Fat1^–/+;Fat4^–/– mutants the frequency of OHC4s increased in mid-basal and basal turns of cochleae (P=1.37×10^–2 for the basal turn; Fig. 3M-O,S). Thus, removal of one copy of Fat1 makes the OHC row number more sensitive to loss of Fat4. Notably, more than one extra row of OHCs was never observed in the Fat1^–/+;Fat4^–/– OOC (Fig. 3M,N,O).

Fat1 and Fat4 are expressed in both the developing neural tube and intervertebral discs (Rock et al., 2005). However, Fat1^–/– homozygotes revealed no apparent anomalies in spinal cord and vertebrae even in the lumbar region (ML/DV ratio of lumbar vertebra, P=0.21; Fig. 3I,J,R). Moreover, loss of Fat1 did not alter the severity of either spinal cord or vertebral phenotypes in Fat4^–/– mutants (P=0.16; Fig. 3K,L,R).

Murine ft-like genes are also expressed in the developing neural tube, suggesting a role for Fat cadherins in neural tube morphogenesis (Rock et al., 2005). Fat1^–/– mutants have deformed eyes and craniofacial malformations, with a low incidence of holoprosencephaly (Ciani et al., 2003). We found that Fat1^–/– mutants born from Fat1^–/+;Fat4^–/+ intercrosses in a (C57BL/6 × 129S3)F1 background often displayed exencephaly (supplementary material Table S2). On a 129S3 background, however, Fat1^–/– mutants were born at the predicted Mendelian ratio, with no external anomalies other than facial malformations (Fig. 4F and supplementary material Table S3). These observations suggest that 129S3 restricts the incidence of exencephaly in Fat1^–/– mutants, implying the existence of recessive modifier loci for Fat1 in cranial neural tube development.

At 18.5 days of gestation (E18.5), three-quarters of Fat1^–/–;Fat4^–/– mutants were exencephalic (Fig. 4A,C-E), indicating synergy between Fat1 and Fat4 in cranial neural tube development. The exencephalic Fat1^–/–;Fat4^–/– embryos display cyanosis, indicating problems in the respiratory and/or blood circulation system. Notably, the double mutants exhibit swollen inner ears with extremely shortened and widened cochleae (Fig. 4N,O). Importantly, neither Fat1^–/– nor Fat4^–/– siblings exhibited cranial neural tube defects on this background (Fig. 4A,F).

Based on our evidence for functional redundancy between Fat1 and Fat4, we asked whether Fat3 also shares functions with Fat4. Fat3^–/– kidneys are normal, with no significant change in the luminal space of renal tubules (P=0.31; Fig. 5A,B,N). Fat3^–/–;Fat4^–/– mutants are viable at birth with no external anomalies in craniofacial structure (supplementary material Table S4; data not shown). Surprisingly, cystic defects in Fat4^–/– kidneys are suppressed upon loss of Fat3 (P=2.93×10^–2; Fig. 5C,D,N), suggesting that Fat3 and Fat4 serve antagonistic functions in controlling renal tubular elongation.

To determine whether Fat3 and Fat4 act antagonistically in other contexts, we examined cochlear development in Fat3^–/– and Fat3^–/–;Fat4^–/– mutants. Fat3^–/– mutants do not display defects in cochlear lengthening (P=0.68; Fig. 5I,J,O). Fat4^–/– cochlear length is not altered upon loss of Fat3 (P=0.44; Fig. 5K,L,O). However, in contrast to the gain of subsets of OHC4s in Fat4^–/– mutants, Fat3^–/– cochleae exhibited discontinuous loss of OHC3s, with patches of three to ten missing cells, predominantly in the basal turn (P=3.84×10^–4; Fig. 5M,Q). Despite these apparently opposing functions, even more OHC3s emerged in the mid-apical region of Fat3^–/–;Fat4^–/– than Fat4^–/– cochlea (P=1.38×10^–2; Fig. 5Q). Moreover, fewer OHC3s were lost from the basal turn of Fat3^–/–;Fat4^–/– than Fat3^–/– cochleae (P=6.22×10^–3; Fig. 5Q). Thus, Fat3 and Fat4 exhibit complex interactions during hair cell patterning.

Fat3 and Fat4 are strongly expressed in both the neural tube and the intervertebral discs during development (Rock et al., 2005). Although morphogenetic anomalies are not evident in the spinal cord or vertebrae of Fat3^–/– newborns (ML/DV ratio of lumbar vertebra, P=0.10; Fig. 5E,F,P), incomplete fusion of vertebral arches of Fat4^–/– mutants increased in severity upon loss of Fat3 (P=8.68×10^–3; Fig. 5G,H,P). Indeed, Fat3^–/–;Fat4^–/– mutants exhibited severely flattened vertebral arches (Fig. 5H). Thus, Fat3 and Fat4 cooperate during the fusion of vertebral arches.

Having revealed overlapping functions for different Fat family genes, we investigated whether any of these phenotypes involve disruption of the Ft/Ds/Fj signaling module. We focused on Fjx1, the sole vertebrate ortholog of Drosophila fj. Fjx1 is expressed in developing kidneys, cochleae, intervertebral discs and neural tube (Rock et al., 2005). However, Fjx1^–/– mutants display no apparent defects in any of these tissues (Fig. 6B,F,J,P-S). Similarly, most Fat4 loss-of-function phenotypes are unaffected by loss of Fjx1, with no change in cochlear length (P=0.94; Fig. 6G,H,Q), OHC4 frequency (P=0.45; Fig. 6M,N,O,S) or mediolateral length of lumbar vertebrae (P=0.96; Fig. 6K,L,R) in Fat4^–/–;Fjx1^–/– mutants as compared with Fat4^–/– mutants. Loss of Fjx1 occasionally exacerbates cystic defects in Fat4^–/– kidneys associated with the development of duplex kidneys and ureteral duplication in a background of [(129P2/Ola × FVB/N)F1 × C57BL/6]F2 (Saburi et al., 2008), but not in the 129S3 background used in this study (P=0.62; Fig. 6C,D,P). Fat1^–/–;Fjx1^–/– kidneys appear normal at birth, as do Fat1^–/– kidneys (Fig. 4L).

Although loss of Fjx1 does not strongly enhance most Fat family phenotypes in a 129S3 background, there are strong genetic interactions in the neural tube. Fat4^–/–;Fjx1^–/– mutants are born with no anomalies in the cranial neural tube (supplementary material Table S5). Strikingly, however, half of Fat1^–/–;Fjx1^–/– E18.5 embryos exhibit anencephaly, a severe form of exencephaly accompanied by degeneration of cerebral tissue (Fig. 4B,G-I and supplementary material Tables S6 and S8). Anencephalic double mutants also display severe anemia, indicating impaired hematopoiesis (Fig. 4G,H,I). Moreover, double-mutant embryos occasionally exhibit an amorphous cranium, with a widened frontonasal prominence and midline facial clefting (Fig. 4I). Neither Fat1^–/– nor Fjx1^–/– single-mutant siblings showed craniofacial abnormalities on the 129S3 background (Fig. 4B,F and supplementary material Table S6). Even in anencephalic Fat1^–/–;Fjx1^–/– embryos, no abnormalities are observed in kidneys, spinal cord or lumbar vertebrae (Fig. 4L,M).

Drosophila Ft regulates planar polarity through interactions with Atrophin (Fanto et al., 2003). We investigated whether vertebrate Fat4 also interacts with atrophins to control tissue morphogenesis. We first tested whether Fat4 interacts with Atn1 in kidney development. Whereas Atn1^–/– homozygotes have normal kidneys (Fig. 7A,B,N and supplementary material Fig. S6C), Atn1^–/–;Fat4^–/– kidneys frequently fail to form medulla (supplementary material Fig. S6A,B) and lack loops of Henle and medullary collecting ducts (supplementary material Fig. S6D,E). Because tubular segments were poorly formed in Atn1^–/–;Fat4^–/– kidneys, we were unable to determine if loss of Atn1 affects the cystic defects of Fat4^–/– kidneys. However, cystic dilations were not significantly modified in Fat4^–/– kidneys upon loss of one copy of Atn1 (P=0.83; Fig. 7C,D,N). Thus, although kidney organogenesis can be severely impaired in Atn1^–/–;Fat4^–/– kidneys, we were unable to determine whether genetic interactions with Atn1 can enhance the cystic defects in Fat4^–/– kidneys.

We examined whether Fat4 requires Atn1 to control cochlear elongation and OHC patterning. The cochlear length is comparable in wild type and Atn1^–/– mutants (P=0.88; Fig. 7E,F,O). Simultaneous loss of Atn1 and Fat4 does not modify cochlear length in Fat4^–/– mutants (P=0.70; Fig. 7G,H,O). Similarly, neither additional rows nor misorientations of sensory hair cells are observed in Atn1^–/– cochleae (Fig. 7M,Q). Atn1^–/– mutants exhibit occasional discontinuous loss of OHC3s in the basal turn of cochleae, as in Fat3^–/– mutants, although the Atn1^–/– defects are not statistically significant (P=0.13; Fig. 7M,Q). Loss of OHC3s was not observed in Fat4^–/– cochleae (Fig. 2S, Fig. 3S, Fig. 5Q, Fig. 6S, Fig. 7Q, Fig. 8S). Loss of OHC3s in Atn1^–/– cochleae decreased in frequency upon loss of Fat4, but this was statistically insignificant compared with siblings (P=0.10; Fig. 7Q). Occasional OHC3 loss occurred even in wild-type siblings (Fig. 7Q). As 129S3 wild-type cochleae never display loss of OHC3s (supplementary material Fig. S3D), this is likely to be due to a strain-specific variant unlinked to Atn1.

Loss of Atn1 does not dramatically disrupt morphogenesis of the neural tube or vertebra (Fig. 7I,J and supplementary material Tables S7, S8), as shown by measuring the ML/DV ratio of lumbar vertebra (P=0.06; Fig. 7P). Moreover, lumbar spinal cords are not significantly broadened in Atn1^–/–;Fat4^–/– mutants (Fig. 7L; data not shown). By contrast, improper fusion of vertebral arches is slightly suppressed in Fat4^–/– mutants upon loss of Atn1 (P=2.26×10^–2; Fig. 7K,L,P), suggesting that Atn1 functions antagonistically to Fat4 during vertebral arch fusion.

We wondered whether a second atro-like gene, Atn2l, which is most similar to Drosophila atro, functions in Fat4 activity. Homozygous Atn2 om mutation leads to embryonic lethality between E9.5 and E11.5 associated with cardiac failure (Zoltewicz et al., 2004), precluding determination of whether renal tubular morphogenesis is perturbed in Atn2^om/om kidneys. Whereas kidneys develop normally in Atn2^om/+ heterozygotes (Fig. 8A,B,P), removal of one copy of Atn2l enhances cystic defects in Fat4^–/– kidneys (P=1.37×10^–2; Fig. 8C,D,P). This enhancement is similar to the exacerbation of cystic defects observed in Fat4^–/–;Vangl2^Lp/+ kidneys, suggesting that Atn2l and Fat4 function in planar polarity patterning in the kidney.

Whereas Atn2^om/+ cochleae show no defects in hair cell organization (Fig. 8S), Atn2^om/+;Fat4^–/– mutants occasionally develop a fourth row of OHCs, predominantly in the mid-apical region (Fig. 8S). This effect is similar to what occurs in Fat4^–/–;Vangl2^Lp/+ and Fat1^–/+;Fat4^–/– cochleae, but opposite to the loss of OHC3s seen in Atn1^–/– and Fat3^–/– cochleae. Loss of OHC3s was never observed in Fat4^–/– cochleae (Fig. 2S, Fig. 3S, Fig. 6S, Fig. 8S). Atn2l haploinsufficiency does not affect cochlear lengthening (Fig. 8E,F,Q) and Fat4^–/– cochlear length is not significantly modified upon removal of one copy of Atn2l (P=0.83; Fig. 8G,H,Q).

We tested whether spinal cord or vertebral anomalies in Fat4^–/– mutants are enhanced by loss of one copy of Atn2l. Atn2^om/+ heterozygotes exhibit no malformations in the spinal cord or vertebrae (Fig. 8I,J,R). Moreover, neither the widening of spinal cord nor improper fusion of vertebral arches is modified in Atn2^om/+;Fat4^–/– mutants as compared with Fat4^–/– siblings (P=0.25; Fig. 8K,L,R).

Atn2 om mutants exhibit complete failure in cranial neural tube closure, particularly in the forebrain (Zoltewicz et al., 2004). However, no Atn2l haploinsufficiency is observed in cranial neural tube formation. Although Fat1^–/–;Fat4^–/– mutants display exencephaly, reminiscent of the fully penetrant neural tube defects in Atn2 om mutants, Atn2^om/+;Fat4^–/– mutants are born at the predicted Mendelian ratio with no detectable abnormalities in the cranial neural tube (supplementary material Table S9).

Fat cadherins regulate a wide variety of developmental events in flies, including planar polarity, tissue growth and organ shape. Previously, we showed that Fat4 plays a key role in planar polarity in mammals, as ft does in flies. However, it has not been clear whether other Fat family members play similar roles and whether other components of the Fat-planar polarity pathway are also conserved from flies to human. Here, we report that mouse Fat1 and Fat3 cooperate with Fat4 in a number of developmental contexts, including kidney, cochlea, cranial neural tube and caudal vertebra. In addition, we provide genetic evidence that the roles for Fat4 in renal tubular elongation, vertebral arch fusion and OHC patterning are modified by vertebrate atrophins. Together with evidence that a Vangl2 Lp mutation enhances multiple Fat4^–/– phenotypes, these findings suggest that Fat4 acts through a conserved pathway to control planar polarity. Further studies will be needed to uncover the biochemical and molecular bases of the genetic interactions among Fat cadherins and atrophins in regulating tissue organization.

We found that proper elongation of renal tubules in the kidneys involves both Fat1 and Fat4, raising the question of how each ft-like gene mediates its effects. One intriguing possibility is that Fat cadherins act through the Hippo signaling pathway, as attenuation of Hippo signaling contributes to renal cyst formation in fat1 morphant zebrafish (Skouloudaki et al., 2009). A role for Hippo signaling in renal tubular morphogenesis is also supported by recent genetic studies in mice and humans (Happe et al., 2011; Hossain et al., 2007; Makita et al., 2008).

Fat1 physically interacts with atrophins (Hou and Sibinga, 2009). We showed here that removal of one copy of Atn2l, but not Atn1, makes renal tubular elongation more sensitive to loss of Fat4, does loss of one copy of Fat1. These observations suggest that Fat1 compensates for Fat4 through interactions with Atn2l. Because it remains to be determined whether Atn2l binds to Fat4, synergy between Fat1 and Fat4 in this context might be mediated by sharing of the effector, Atn2l, and/or by parallel action of Fat1-Atn2l to Fat4 signaling (supplementary material Fig. S7).

Additional complexities in Fat functions are revealed in cochleae, where both synergistic and antagonistic effects were observed. Fat1 and Fat4 again act synergistically here, as cochlear shortening was exacerbated and the emergence of ectopic OHC4s was enhanced in the basal turn of Fat1^–/+;Fat4^–/– cochleae. By contrast, Fat3^–/– cochleae display discontinuous loss of OHC3s near the base and loss of OHC3s can be suppressed by loss of Fat4. Thus, Fat4 interacts cooperatively with Fat1 to stimulate cochlear extension and limit OHC row number, but antagonistically with Fat3 in OHC patterning (supplementary material Fig. S7).

The cochlear phenotype in Fat4^–/– mutants is consistent with disrupted planar polarity signaling. Planar polarity is necessary for CE movements associated with cochlear elongation, and mutations in core PCP genes result in expansion in OHC row number in the apex (Etheridge et al., 2008; Montcouquiol et al., 2003; Yu et al., 2010). The core PCP mutants also exhibit disorganization of hair cells near the base, namely ectopic hair cells beyond the OHC3s (Wang, Y. et al., 2006), as seen in the Fat1^–/+;Fat4^–/– cochleae. The fact that cochlear elongation is impaired in Fat4^–/– mutants and that Fat4^–/– cochleae become shorter still in the Vangl2^Lp/+ background suggest that both Fat4 and Vangl2 control planar polarity in cochlea. Emergence of ectopic OHC4s was also exacerbated in the mid-apical turn of Fat4^–/–;Vangl2^Lp/+ cochleae. As in flies, whether Fat4 and core PCP components act in parallel or together remains unknown. Intriguingly, defects in both cochlea elongation and OHC patterning in Fat4^–/– mutants are exacerbated upon loss of one copy of Fat1, suggesting that Fat1 might influence planar polarity in cochlea. As in the kidney, impaired OHC patterning in Fat4^–/– cochleae, but not cochlear shortening, increased in severity upon removal of one copy of Atn2l, suggesting that synergy between Fat1 and Fat4 in the OOC might be mediated by Atn2l as described above (supplementary material Fig. S7).

Analysis of vertebral development revealed additional redundancies amongst Fat family cadherins. Fat4^–/– mutants display malformed vertebral arches caused by failure in fusion at the dorsal midline. A similar phenotype was reported in Dchs1^–/– mutants (Mao et al., 2011), suggesting that Dchs1 is a Fat4 ligand (Ishiuchi et al., 2009; Mao et al., 2011). The vertebral phenotype in Fat4^–/– mutants increased in severity upon loss of Fat3, suggesting that Fat3 and Fat4 act together in intervertebral discs to direct proper fusion of vertebral arches (supplementary material Fig. S7). Although little is known about the cellular events that underlie this process, incomplete fusion might be a PCP phenotype, as Dvl1^–/–;Dvl2^–/– mutants also display wide vertebral columns accompanied by classic PCP phenotypes such as looped tails, opened neural tube and abnormal cochlear extension (Hamblet et al., 2002). In support of this idea, impaired fusion of vertebral arches in Fat4^–/– mutants was occasionally exacerbated by Vangl2^Lp/+ heterozygosity.

The remarkable breadth of Fat activities is highlighted in the nervous system, where once again Fat family members have both overlapping and distinct functions. Fat1 seems to be a key player in the cranial neural tube, whereas Fat4 acts more caudally. Some of these functions might reflect functional redundancy, as Fat1^–/–;Fat4^–/– mutants suffer from an increased incidence of exencephaly compared with Fat1^–/– or Fat4^–/– mutants. However, Fat1 and Fat4 appear to have independent functions during neural tube morphogenesis. For instance, Fat1 shows unique synergy with Fjx1, as Fat1^–/–;Fjx1^–/– mutants exhibit anencephaly, whereas Fat4^–/–;Fjx1^–/– mutants display no detectable anomalies in the cranial neural tube. Drosophila Fj is a kinase that phosphorylates the extracellular cadherin domains of both Ft and Ds, modifying their ability to bind to each other (Brittle et al., 2010; Ishikawa et al., 2008; Simon et al., 2010). Thus, Fjx1 might regulate Fat and Dachsous cadherins in vertebrates, altering their physical interactions. Consistent with this model, loss of Fjx1 enhances Fat3^–/+ but not Fat3^–/– phenotypes in retina (Deans et al., 2011). Taken together, Fat cadherins may function in parallel, downstream of Fjx1, to regulate cranial neural tube development. Because loss of Fat1 may induce Fjx1 expression, as ft or Fat4 mutation does in flies or mice (Fanto et al., 2003; Saburi et al., 2008), simultaneous loss of Fat1 and Fjx1 might also disrupt a feedback loop for other Fat and Dachsous cadherins to compensate for Fat1 (supplementary material Fig. S7).

It remains to be determined whether the exencephaly and anencephaly in Fat1^–/–;Fat4^–/– and Fat1^–/–;Fjx1^–/– mutants are due to abnormal planar polarity signaling. However, synergies similar to those we found for Fat family genes are also observed in the cranial neural tube for core PCP genes (Fischer et al., 2007). Loss of Fz3 leads to failure of cephalic neural tube closure at low penetrance depending on genetic background (Wang et al., 2002), whereas Fz6^–/– mutants do not have specific deformities in the neural tube. However, simultaneous loss of Fz3 and Fz6 causes craniorachischisis with nearly complete penetrance (Wang, Y. et al., 2006), and Fz3 heterozygosity is sufficient to induce severe anencephaly in Fz6^–/– mutants, although with reduced penetrance (Stuebner et al., 2010). Alternatively, Fat4 could play a role in the caudal neural tube via the Hippo signaling pathway. Indeed, a recent study revealed a role for Fat4 in the restriction of cell numbers in the neuronal progenitor pools in the chick neural tube, via the Hippo mediator YAP (Van Hateren et al., 2011), raising the possibility that the broadened spinal cords in the Fat4^–/– mouse mutants might be caused by aberrant expansion of neural progenitors.

Fat cadherins comprise a diverse family of proteins with substantially different intracellular domains, and we have shown here that some function redundantly to pattern tissues, whereas others function antagonistically. Moreover, Fat family members exhibit tissue-specific interactions, such as Fat1 and Fat4 in kidney, cochlea, cranial neural tube and Fat3 and Fat4 in vertebra. Although several Fat loss-of-function phenotypes in mice reflect altered planar polarity signaling, such as cystic kidneys and shortened cochleae (Mao et al., 2011; Saburi et al., 2008), other phenotypes may not be due to defective planar polarity and might instead reflect altered growth regulation. Consistent with this model, only a subset of Fat4^–/– phenotypes is enhanced by disrupted Vangl2 activity. Our data also suggest that atrophins might function with Fat cadherins in vertebrates, as in flies, to regulate planar polarity. Finally, we have seen strong inbred genetic background effects in Fat loss-of-function phenotypes, not only in the cranial neural tube, which generally shows a clear strain dependency of defects (Banting et al., 2005; Choi and Klingensmith, 2009; Colmenares et al., 2002; Ikeda et al., 1999; Kaneko et al., 2007; Kooistra et al., 2012; Sah et al., 1995; Sang et al., 2011; Stottmann et al., 2006; Wright et al., 2007), but also in other tissues. These as yet unidentified modifiers might impact the distinct phenotypic readouts caused by loss of Fat, such as defective Fat-planar polarity signaling versus Fat-Hippo signaling. Although these complexities introduce new challenges, our findings emphasize the wide-ranging impact that Fat cadherins have on animal development.

We thank Michael Dean for insightful comments and discussions.