Independent Mutations in Mouse Vangl2 That Cause Neural Tube Defects in Looptail Mice Impair Interaction with Members of the Dishevelled Family* □ S

Mammalian Vangl1 and Vangl2 are highly conserved membrane proteins that have evolved from a single ancestral protein Strabismus/Van Gogh found in Drosoph- ila . Mutations in the Vangl2 gene cause a neural tube defect (craniorachischisis) characteristic of the looptail ( Lp ) mouse. Studies in model organisms indicate that Vangl proteins play a key developmental role in estab- lishing planar cell polarity (PCP) and in regulating convergent extension (CE) movements during embryogene- sis. The role of Vangl1 in these processes is virtually unknown, and the molecular function of Vangl1 and Vangl2 in PCP and CE is poorly understood. Using a yeast two-hybrid system, glutathione S -transferase pull- down and co-immunoprecipitation assays, we show that both mouse Vangl1 and Vangl2 physically interact with the three members of the cytoplasmic Dishevelled (Dvl) protein family. This interaction is shown to require both the predicted cytoplasmic C-terminal half of Vangl1/2 and a portion of the Dvl protein containing PDZ and DIX domains. In addition, we show that the two known Vangl2 loss-of-function mutations identified in two in- dependent Lp alleles associated with neural tube defects impair binding to Dvl1, Dvl2, and Dvl3. These find- ings suggest a molecular mechanism for the neural tube defect seen in Lp mice. Our observations indicate that Vangl1 biochemical properties parallel those of Vangl2 and that Vangl1 might, therefore, participate in PCP ent domains of Dishevelled. The procedures used for these studies were exactly as described by the manufacturer (Clontech). The system is based on the use of Saccharomyces cerevisiae strain AH109, which is engineered to have three different reporters, namely ADE2, HIS3, and lacZ (cytoplasmic (cid:1) -galactosidase and secreted (cid:2) -galactosidase) under the control of distinct GAL4 upstream regulatory sequences derived from the GAL2, GAL1, and MEL1 gene promoters, respectively. These three reporters yield strong and specific response to GAL4, and the simultaneous use of ADE2 and HIS3 as selectable markers eliminates false positives. The system also includes two plasmid vectors: pGBKT7 (Kan r , TRP1, and GAL4DBD c-Myc-tagged) and pGADT7 (Amp r , LEU2, and GAL4AD HA-tagged) that direct the production of recombinant proteins consisting of independent fusions to the DNA binding domain (DBD) or activation domain (AD) of GAL4. The two test plasmids are introduced by transformation into either AH109 (MATa) or Y187 (MAT (cid:2) ), both strains auxotrophs for leucine, tryptophan, histidine, and adenine. Individual pGBKT7 and pGADT7 constructs were trans- formed into yeast cells using a lithium acetate transformation proce-dure, and expression of the corresponding recombinant GAL4 proteins was confirmed by immunoblotting of whole cell lysates with either anti-c-Myc (9E10, BAbCO) or anti-HA (16B12, BAbCO) mouse mono- clonal antibodies. Possible reconstituted GAL4 activity was tested in diploids obtained by mating AH109 and Y187 transformants positive for expression of the test GAL4 fusions followed by plating on synthetic medium (SD) of various stringencies: low stringency lacking tryptophan and leucine (SD-Trp/-Leu), medium stringency lacking histidine, tryptophan, and leucine (SD-His/-Trp/-Leu),

Mammalian Vangl1 and Vangl2 are highly conserved membrane proteins that have evolved from a single ancestral protein Strabismus/Van Gogh found in Drosophila. Mutations in the Vangl2 gene cause a neural tube defect (craniorachischisis) characteristic of the looptail (Lp) mouse. Studies in model organisms indicate that Vangl proteins play a key developmental role in establishing planar cell polarity (PCP) and in regulating convergent extension (CE) movements during embryogenesis. The role of Vangl1 in these processes is virtually unknown, and the molecular function of Vangl1 and Vangl2 in PCP and CE is poorly understood. Using a yeast two-hybrid system, glutathione S-transferase pulldown and co-immunoprecipitation assays, we show that both mouse Vangl1 and Vangl2 physically interact with the three members of the cytoplasmic Dishevelled (Dvl) protein family. This interaction is shown to require both the predicted cytoplasmic C-terminal half of Vangl1/2 and a portion of the Dvl protein containing PDZ and DIX domains. In addition, we show that the two known Vangl2 loss-of-function mutations identified in two independent Lp alleles associated with neural tube defects impair binding to Dvl1, Dvl2, and Dvl3. These findings suggest a molecular mechanism for the neural tube defect seen in Lp mice. Our observations indicate that Vangl1 biochemical properties parallel those of Vangl2 and that Vangl1 might, therefore, participate in PCP and CE either in concert with Vangl2 or independently of Vangl2 in discrete cell types.
Neural tube closure is a complex developmental process that takes place early during embryogenesis and is a key step in formation of the central nervous system, including spinal cord (1). In humans, failure of the neural tube to close properly results in a group of syndromes collectively known as neural tube defects (NTDs), 1 which constitute the second most fre-quent cause of congenital abnormalities (1 in 1000 live births) (2). The cellular and molecular mechanisms underlying NTDs are very complex, poorly understood, and difficult to study in humans. Recent genetic studies of NTDs in model organisms have identified a number of genes and proteins that play critical roles in neural tube closure (1). In our laboratory, we have identified Vangl2 (Van Gogh-like, formerly "loop-tail associated protein," Ltap) as the gene mutated in the mouse model of severe NTD known as looptail (Lp) (3,4). Heterozygous Lp animals (Lp/ϩ) are normal except for the presence of a characteristic "kinked" (looped) tail, whereas Lp homozygotes (Lp/Lp) die at mid-gestation due to craniorachischisis, a very severe NTD that is characterized by a completely open neural tube from the midbrain/hindbrain boundary to the caudal (tail) region (3)(4)(5). Two alleles have been described for the Lp mice: naturally occurring Lp (3) and a chemically induced Lp m1Jus (4).
Vangl2 is a 521-amino acid transmembrane (TM) protein composed of four putative TM domains in the N-terminal half. The C-terminal half is predicted to be cytoplasmic and is possibly involved in intracellular signaling and/or interaction with other proteins (3)(4)(5)(6). Vangl2 mRNA is embryonically expressed in a number of tissues, including the neural tube immediately prior to, during, and after closure (3)(4)(5). The NTD phenotype of both Lp mutants (Lp, Lp m1Jus ) is associated with independent missense mutations within the Vangl2 cytoplasmic domain affecting amino acid residues otherwise conserved in the protein family: S464N (Lp) and D255E (Lp m1Jus ) (3,4). The similar phenotypic consequences of heterozygosity (Ϫ/ϩ) and homozygosity (Ϫ/Ϫ) at independent Lp alleles (4) and the strict recessive mode of inheritance of Lp-associated NTD strongly suggest that Lp mutations behave as loss-of-function mutations in a gene dosage-sensitive pathway, signifying a role for Vangl2 as a critical regulator of neural tube formation. Vangl2 orthologs have been found in flies (Drosophila), worms (Caenorhabditis elegans), zebrafish (Danio rerio), and mammals, thus defining a family of evolutionarily conserved proteins (7). Interestingly, analysis of fish, mouse, and human genomes has identified two Vangl homologous genes in each species designated Vangl1 and Vangl2.
Originally cloned in Drosophila by two groups, the primordial gene was given the dual appellation Strabismus (Stbm) or Van Gogh (Vang), based on the phenotypic appearance of certain structures in fly mutants (8,9). The gene was shown to play a role in establishing planar cell polarity (PCP), or the orientation of epithelial cells in a planar layer (10,11). In flies, disruption of the PCP pathway through inactivation of one of several so-called "core PCP genes" results in mis-orientation of normally highly organized specific epithelial structures such as ommatidia of eye, hairs on the wing cells, bristles on the legs, and others (11,12). In addition to Stbm/Vang, other core PCP proteins include the seven TM domain receptor Frizzled (Fz), an atypical cadherin Starry night/Flamingo (Stan/Fml), and two cytoplasmic proteins, Dishevelled (Dsh/Dvl) and Prickle (Pk) (13)(14)(15)(16)(17).
Extensive studies in flies have shown that the establishment of proper planar polarity relies on the formation of a multiprotein membrane complex consisting of core PCP proteins Stbm/ Vang, Pk, Dvl, and Fz. In individual cells of the fly wing, PCP proteins are initially arranged symmetrically at the apical membrane, but become asymmetrically redistributed during the establishment of PCP. A complex formed by Stbm/Vang and Pk assumes an apical-proximal localization, whereas a Fz-Dvl complex is redirected to the apical-distal portion of the cell (18 -20). This asymmetry is believed to be key in establishing planar polarity (11). At the molecular level, fly Stbm and its vertebrate orthologVangl2 have been shown to interact directly with cytoplasmic Dvl and Pk proteins leading to their recruitment to the membrane (6, 20 -23). Although the function of Pk in PCP is not well understood, the role of Dvl has been better characterized. Dvl proteins have a modular organization typical of adaptor proteins with three structurally conserved domains, the N-terminal DIX domain, the central PDZ domain, and the C-terminal DEP domain (24 -26). Dvl has been demonstrated to play key roles in both canonical Wnt and PCP pathways (reviewed in Ref. 27). The DIX domain is indispensable for canonical Wnt signaling, whereas the DEP domain is dedicated to the PCP pathway and the PDZ functions in both (25,29). 2 Mutations in the DEP domain interfere with Dvl translocation to the membrane (18,30).
Recent studies in vertebrates have revealed that Vangl2 regulates the process of convergent extension (CE, controlled by the non-canonical Wnt pathway) during gastrulation and neurulation in frogs and zebrafish (6,(31)(32)(33). During CE, dorsal mesenchymal and neuroepithelial cells become polarized in a medio-lateral direction, migrate, and intercalate. These directional cellular activities result in concomitant lengthening of the anterior-posterior axis and narrowing of the epithelial mass in a perpendicular direction (34). Failure of CE in mesenchyme produces animals with shorter, wider trunks (34,35), whereas CE distortion during neural tube formation leads to significant widening of the neural plate in the midline and, as a result, the neural tube folds are spaced too far apart to permit closure (1,36). The phenotype of the Lp mouse is compatible with manifestations of CE failure, because the body axis is shortened, the neural plate is broadened, and the neural tube does not close (3)(4)(5). Interestingly, mutations in mammalian counterparts of the Drosophila PCP proteins Flamingo (Celsr1) (37) or Dishevelled (Dvl1;Dvl2 double knockout) (38) cause phenotypes indistinguishable from that of the Lp mouse. These observations have prompted the hypothesis that the regulation of CE in vertebrates is related to the PCP pathway in flies, including a critical role for vertebrate relatives of PCP proteins such as Stbm/Vang, Dvl, Fz, and Pk (reviewed in Ref. 39).
The physiological role and biochemical properties of Vangl2 are now being unraveled, yet Vangl1 function remains a mystery. In the present study, we aimed to characterize the mammalian Vangl1 protein. To determine whether structural similarities in the Vangl family translate into conservation of function we investigated possible interactions of Vangl1 and Vangl2 with three known members of the Dvl family. We also examined the effect of Lp mutations of Vangl2 on Vangl2 interactions with Dvl proteins and demonstrated that the S464N and D255E mutations dramatically decrease Vangl2 binding to Dvl proteins. These findings suggest a molecular explanation for the loss-of-function of Vangl2 protein in Lp mouse and establish a critical role for Vangl2:Dvl interaction in proper signaling during neural tube closure.
Yeast Two-hybrid System-The commercially available yeast twohybrid system MatchMaker system 3 (#K1612-1, Clontech) was used to study possible interactions between Vangl1/Vangl2 proteins and differ-ent domains of Dishevelled. The procedures used for these studies were exactly as described by the manufacturer (Clontech). The system is based on the use of Saccharomyces cerevisiae strain AH109, which is engineered to have three different reporters, namely ADE2, HIS3, and lacZ (cytoplasmic ␤-galactosidase and secreted ␣-galactosidase) under the control of distinct GAL4 upstream regulatory sequences derived from the GAL2, GAL1, and MEL1 gene promoters, respectively. These three reporters yield strong and specific response to GAL4, and the simultaneous use of ADE2 and HIS3 as selectable markers eliminates false positives. The system also includes two plasmid vectors: pGBKT7 (Kan r , TRP1, and GAL4DBD c-Myc-tagged) and pGADT7 (Amp r , LEU2, and GAL4AD HA-tagged) that direct the production of recombinant proteins consisting of independent fusions to the DNA binding domain (DBD) or activation domain (AD) of GAL4. The two test plasmids are introduced by transformation into either AH109 (MATa) or Y187 (MAT␣), both strains auxotrophs for leucine, tryptophan, histidine, and adenine. Individual pGBKT7 and pGADT7 constructs were transformed into yeast cells using a lithium acetate transformation procedure, and expression of the corresponding recombinant GAL4 proteins was confirmed by immunoblotting of whole cell lysates with either anti-c-Myc (9E10, BAbCO) or anti-HA (16B12, BAbCO) mouse monoclonal antibodies. Possible reconstituted GAL4 activity was tested in diploids obtained by mating AH109 and Y187 transformants positive for expression of the test GAL4 fusions followed by plating on synthetic medium (SD) of various stringencies: low stringency lacking tryptophan and leucine (SD-Trp/-Leu), medium stringency lacking histidine, tryptophan, and leucine (SD-His/-Trp/-Leu), and high stringency lacking adenine, histidine, tryptophan, and leucine (SD-Ade/-His/-Trp/-Leu). Substrate 5-bromo-4-chloro-3-indolyl-␣-D-galactopyranoside (X-␣-gal) was used to visualize LacZ-producing clones directly on the plates. Protein-protein interactions were evaluated by assessing growth of diploid cells in solid medium after 72 h at 30°C.
Co-immunoprecipitation and GST Pull-down Assays-For co-immunoprecipitation experiments, lysates from transiently transfected HEK293 cells (500 l) were incubated with either 1 g of appropriate mouse monoclonal antibodies Dvl1 (3F12), Dvl2 (10B5), Dvl3 (4D3) (Santa Cruz Biotechnology), or with 1:250 dilution of rabbit polyclonal anti-Vangl1 and anti-Vangl2 antisera (see below) together with 25 l of a slurry of Protein A/G coupled to agarose beads (Invitrogen) for 16 h at 4°C on a rotating wheel. Protein A/G beads were washed twice by centrifugation at 4°C in a cold buffer consisting of 20 mM HEPES (pH 7.5), 500 mM NaCl, 10% glycerol, 0.5% Triton X-100, 1.5 mM MgCl 2 , 10 mM NaPO 4 , pH 7.5, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml of leupeptin and aprotinin, followed by two additional washes in the same buffer containing 200 mM NaCl. Immune complexes were eluted by incubation of the protein A/G beads in 2ϫ sample buffer (2% SDS, 2 M ␤ 2 -mercaptoethanol) at 100°C, 10 min. Proteins were separated by electrophoresis on SDS-containing acrylamide gels (SDS-PAGE) followed by immunoblotting with polyclonal anti-Vangl1 and anti-Vangl2 antibodies. A portion of each cell lysate (1/20) was resolved on PAGE-SDS and immunoblotted with anti-Dvl or anti-Vangl1/2 antibodies to control for input loading.
For GST pull-down assays, GST-Vangl1, GST-Vangl2, and GST (from empty pGEX4T) fusion proteins were purified from Escherichia coli Bl21cells. Cell growth, induction conditions, cell lysis, and isolation of GST fusions by affinity chromatography were exactly as described by the manufacturer of the pGEX4T plasmid (Amersham Biosciences). GST-fused proteins were stored on glutathione-Sepharose 4B beads (Amersham Biosciences) at Ϫ80°C. Equal amounts of GST-Vangl1, GST-Vangl2, and control GST proteins on beads (with respect to protein concentration) were mixed with either recombinant Dvl proteins produced by transient transfection in HEK293 cells (see above) or with Dvl proteins present in cell lysates from E13.5 mouse embryos prepared with PLC buffer. Mixtures were rotated at 4°C for 4 -16 h, washed five times with cold 50 mM Tris, pH7.4, 150 mM NaCl, 0.5% Triton X-100, 1.5 mM MgCl 2 , and 1 mM phenylmethylsulfonyl fluoride, 1 g/ml of each leupeptin and aprotinin, and eluted with 2ϫ Laemmli sample buffer. Proteins were resolved by SDS-PAGE and analyzed by immunoblotting using anti-mouse monoclonal antibodies Dvl1 (3F12), Dvl2 (10B5), Dvl3 (4D3), and GST (B-14) (Santa Cruz Biotechnology).

RESULTS
Vangl Protein Family-A search of public sequence databases reveals the presence of two Vangl genes (Vangl1, Vangl2) and proteins in mouse (Mus musculus; Q80Z96, AAK91927), human (Homo sapiens; AAH65272, Q9ULK5), and zebrafish (D. rerio; AAQ84456, NP_705960), while a single gene exists in flies (Drosophila melanogaster; NP_477177) and worms (Caenorhabditis elegans, NP_508500). Only Vangl2 homologs were found in frogs (Xenopus laevis; AAK70879). A multiple sequence alignment of these protein sequences was produced using the ClustalW program (NCBI server) and is shown in Fig. 1S (Supplementary Material). Among vertebrates, Vangl1 proteins share 67% similarity, whereas Vangl2 proteins share 74% similarity; mammalian Vangl1 and Vangl2 share 44 and 50% similarity with Drosophila Stbm/Vang and 30 and 27% with C. elegans Stbm/Vang, respectively, suggesting that the mammalian and vertebrate genes evolved by gene duplication from a common ancestor present in flies and worms, followed by evolutionary drift.
Hydropathy profiling and analysis of hydrophobic moments (TopPred program) confirm the presence of 4 putative transmembrane (TM) domains in all Vangl proteins at conserved positions corresponding to residues 114 -134, 152-172, 186 -206, and 222-242 of the human Vangl1 sequence (data not shown). The fourth predicted TM domain shows a consistently lower probability than those of TM1-3 (data not shown). Also conserved among Vangl proteins is the C-terminal ETSV tetrapeptide (ESAV in C. elegans), which encodes a PDZ binding motif previously shown to promote interaction with PDZ-containing proteins usually located at the cytoplasmic side of the plasma membrane (42,43). The presence of 4TM domains and PDZ binding motif in the same protein would specify the positions of both N and C termini of Vangl molecules at the intracellular face of the membrane. Vangl proteins show a number of putative asparagine-linked glycosylation sites (NX(S/T)) in the first 100 N-terminal residues, most notably a conserved cluster near position 60, and in the C-terminal half near position 314 -318 of the human Vangl1 protein (Supplementary

Vangl2 Mutations Impair Interaction with Dishevelled Family
Material). Their functional relevance in vivo is however uncertain, because these predicted glycosylation sites appear to be located intracellularly.
Sequence similarity is not uniformly distributed along the length of Vangl1 and Vangl2 proteins (Fig. 1). In particular, two areas near the N terminus (position 14 -53, zone I) and in the cytoplasmic half (positions 310 -340, zone III) show rather low degree of similarity (Ͻ40%), whereas two regions in the cytoplasmic half (zones II and IV) share Ͼ80% identity (Fig.  1B). Regions of high sequence conservation may be crucial for Vangl function. This proposition is supported by the observation that Vangl2 mutations S464N and D255E identified in Lp mice map to proposed zones II and IV (Fig. 1B), the regions sharing the highest degree of sequence conservation in the Vangl protein family. These observations highlight the role of the Vangl cytoplasmic domain in transducing PCP signaling, possibly via protein-protein interactions with other PCP partners. Finally, alignment of the 9 Vangl protein sequences identifies a number of invariant residues (92/520), including 17 basic, 10 acidic residues, and 3 prolines (Supplemental Material). Included into this group are the four cysteine residues of human/mouse Vangl1/2, which suggests that they might play a critical structural (disulfide bridge) or functional role that has been conserved during evolution of Vangl proteins.
Vangl1 and Vangl2 Proteins Interact with Mammalian Dvl Proteins-Dishevelled (Dvl) is a cytoplasmic protein involved in the Wnt/␤-catenin and PCP pathways (27). It is recruited to the cell membrane in response to PCP signaling (20). The C-terminal half of the Vangl2/Stbm is sufficient to bind and recruit Dvl to the plasma membrane in frog animal cap tissues (6). Three Dvl homologs have been identified in mammals, Dvl1, Dvl2, and Dvl3, but their possible interactions with Vangl1 or Vangl2 have not been studied. To determine if the structural homologies within the Vangl protein family translate into functional similarities, we tested the ability of Vangl1 and Vangl2 to interact individually with Dvl proteins in a series of proteinprotein interaction assays.
In a first set of experiments, we used a yeast two-hybrid system (Matchmaker 3.0, Clontech). Each Vangl and Dvl protein was fused to either the DNA binding domain (DBD) or the activation domain (AD) of the transcriptional activator GAL4 ( Fig. 2A). Interaction between the two test partners reconstitutes GAL4 activity detectable by activation of four reporter genes (ADE1, HIS1, and ␣and ␤-galactosidase activities). This selection system reduces the rate of false positives and simultaneously takes advantage of three levels of selection stringency, providing a semi-quantitative measure of the strength of the protein-protein interaction (see "Materials and Methods"). Several fusion constructs were produced ( Fig. 2A). For Vangl proteins, the cytoplasmic domains of Vangl1 (251-526 aa) and Vangl2 (238 -521 aa) were independently fused to GAL4-DBD. We excluded the membrane-associated regions of Vangl proteins to avoid possible interference of hydrophobic TM domains with nuclear targeting of the GAL4 fusions. Because the PDZ domain of Dvl is necessary for proper PCP in vivo and required for interaction with Vangl2 in Xenopus, we built N-terminal half-containing Dvl constructs, which include DIX and PDZ domains (Dvl1-5Ј (1-404 aa), Dvl2-5Ј (1-418 aa), and Dvl3-5Ј (1-395 aa) as well as the DEP-domain containing C-terminal portion of Dvl3 (389 -717 aa) and the DIX domain of Dvl1 (1-87 aa) fused to GAL4-AD ( Fig. 2A). GAL4-AD-Dvl and GAL4-DBD-Vangl1/2 constructs were transformed into yeast cells, and total cell extracts were tested for the presence of the respective fusion proteins by immunoblotting them with either anti-c-Myc or anti-HA antibodies directed against antigenic epitopes included in the GAL4-AD (HA tag) or GAL4-DBD (c-Myc tag) (Fig. 2B). Complementation studies in diploid cells plated on selective media (Fig. 2C) demonstrated that Vangl1 and Vangl2 proteins were indeed able to interact with the N-terminal half of each of the three Dvl proteins tested. As estimated by the number of colonies formed on selective media (not shown) or by the growth rates of such colonies (Fig. 2C), there did not appear to be major differences in the interactions of Vangl1 or -2 with Dvl1/2/3 proteins. The PDZ domain of Dvl was essential for interaction with Vangl1/2, and its deletion in construct Dvl1-DIX abrogates interaction with Vangl1/2. Likewise, the DEP domain of Dvl proteins (construct Dvl3-3Ј) did not interact with Vangl1/2.
Interactions between Vangl1/2 and Dvl proteins were further investigated by GST pull-down assays. In these experiments, the cytoplasmic domains of Vangl1 and Vangl2 were fused to glutathione S-transferase (GST) followed by expression and purification of the fusion proteins from E. coli BL21 bacteria. Purified GST-Vangl1 and GST-Vangl2 proteins were incubated with total protein extracts from mouse E13.5 embryos followed by affinity chromatography on glutathione-Sepharose beads, analysis of captured products by SDS-PAGE, and immunoblotting with anti-Dvl1, anti-Dvl2, and anti-Dvl3 antibodies (Fig. 3,  A-C). Fetal expression of each Dvl mRNA has been previously reported (38,44,45). The isoform specificity of the three anti-Dvl antibodies was ascertained by immunoblotting using total cell extracts from HEK-293 cells transfected with full-length Dvl1, Dvl2, or Dvl3 cDNAs (Fig. 3, E-G). Results in Fig. 3 (B  and C) show that both Vangl1 and Vangl2-GST fusions can retain immunoreactive Dvl2 and Dvl3 present in embryonic extracts. These interactions are specific and were not seen when control GST protein or beads were used in the assay. We consistently observed that the Vangl2 fusion captured more Dvl2 and Dvl3 than the Vangl1 fusion, despite similar amounts of Vangl1-GST and Vangl2-GST proteins in the assay (Fig. 3D). We were unable to detect endogenous Dvl1 protein in mouse E13.5 embryo lysates (data not shown). Instead, we used protein extracts from HEK293 cells transfected with Dvl1 to monitor interaction with Vangl1GST and Vangl2-GST proteins. Results in Fig. 3A show that GST-Vangl1/2 can capture Dvl1 present in such extracts. Similar interactions were observed between Vangl1/2 GST fusions and Dvl2 and Dvl3 expressed in HEK293 cells (data not shown).
Finally, interactions between full-length Vangl1/2 and Dvl proteins were analyzed by co-immunoprecipitation. Protein pairs were expressed in HEK293 cells, followed by immunoprecipitation of either Vangl1/2 or Dvls and analysis of IP products by immunoblotting with isoform-specific anti-Dvl or anti-Vangl1/2 antibodies. Typical results obtained for Dvl2 are shown in Fig. 4 (C and D); similar interactions were detected with Dvl1 and Dvl3 (data not shown). For these experiments, we generated rabbit polyclonal antisera against mouse Vangl1 and Vangl2 proteins using antigenic epitopes derived from the N termini of the proteins ( Fig. 4A; "Materials and Methods"). The isoform specificity of the two antisera was verified by immunoblotting HEK293 cell extracts expressing either Vangl1 or Vangl2 (Fig. 4B). Results in Fig. 4C show that immunoprecipitation of Dvl2 with anti-Dvl2 antibodies captures both Vangl1 and Vangl2 expressed in HEK293 cells. The effect is specific and is not seen when one of the two protein partners is absent. Conversely, immunoprecipitation of Vangl1 from HEK293 cell lysates also captured Dvl2 (Fig. 4D). Although our anti-Vangl2 antiserum readily recognizes Vangl2 by immunoblotting (Fig. 4B), it does not react with the protein under immunoprecipitation conditions (data not shown). Therefore, we did not use anti-Vangl2 antiserum to IP Dvl proteins.
Taken together, results in Figs (3,4). In both cases, the complete loss of Vangl2 protein function appears to underlie the Lp defect, because heterozygotes (Lp/ϩ) and homozygotes (Lp/Lp) for each allele show the same phenotype. In addition, the Vangl2 pathway seems sensitive to gene/  (Fig. 1S, Supplemental Material). Because the molecular basis for the loss of Vangl2 function in Lp alleles is unknown, it was of interest to determine whether these mutations affect interaction with members of the Dvl family. For this, the S464N and D255E mutations were independently built into the cytoplasmic portion of mouse Vangl2 followed by expression in S. cerevisiae yeast cells. Immunoblots of total protein extracts from yeast transformants show that both mutants are expressed in yeast cells at levels comparable to that seen for the wild type protein (Fig. 5A). This suggests that the two mutations do not affect protein stability nor do they cause rapid degradation in yeast cells. These Vangl2 transformants were mated to tester stocks expressing different fragments of Dvl proteins; effect of mutations on Vangl2/Dvl protein interaction were monitored by plating diploids on selective media of increasing levels of stringency. Experiments in Fig. 5B demonstrate that the S464N and D255E mutations impair interaction with Dvl proteins. Under condition of medium stringency (-His/-Trp/-Leu) both mutant Vangl2 proteins diploids, which were then plated on three different selection media that allow detection of protein-protein interaction (GAL4 activity) with increasing degrees of stringency: low, intermediate, and high (see "Material and Methods" for details). Controls include the DNA binding domain (DBD) and activation domain (AD) of GAL4 independently expressed in yeast. Diploids forming on the three selective media (indicated at the bottom) were allowed to grow for 72 h and were photographed. The graded stringencies of the selection media are indicated as low, intermediate, and high below each panel.  Fig. 1A) were fused to glutathione S-transferase followed by expression and purification from E. coli. Purified GST-Vangl1, GST-Vangl2, and GST control proteins were incubated with total protein extracts from HEK293 cells transfected with mouse Dvl1 (A) or with total protein extracts from mouse E13.5 embryos (B and C), followed by affinity chromatography on glutathione-agarose beads, analysis of captured products by SDS-PAGE, and immunoblotting with anti-Dvl1 (A), anti-Dvl2 (B), and anti-Dvl3 antibodies (C). GST-Vangl1, GST-Vangl2, and GST proteins purified from E. coli were also analyzed by immunoblotting with anti-GST antibody (D). The lane designated "beads" contains GST-Sepharose beads mixed with either fetal tissue extracts (B-D) or with Dvl1-transfected HEK293 cells (A). The isoform specificity of the anti-Dvl1 (E), anti-Dvl2 (F), and anti-Dvl3 (G) was verified by immunoblotting of total protein extracts from HEK293 cells transfected with Dvl1, Dvl2, or Dvl3, and from head (E13head) or trunk (E13trunk) tissues from E13 mouse embryos.

Vangl2 Mutations Impair Interaction with Dishevelled Family
FIG. 4. Reciprocal interactions between Vangl1 and Dvl2 detected by co-immunoprecipitation. A, sequence of Vangl1 and Vangl2 segments used to construct GST fusion immunogens to produce isoform-specific anti-Vangl1 and anti-Vangl2 rabbit polyclonal antisera; identical amino acids are shaded. B, the isoform specificity of the anti-Vangl1 and anti-Vangl2 antisera was verified by immunoblotting. Total cell extracts lose binding to Dvl1 only, whereas under a high stringency condition (-Ade/-His/-Trp/-Leu) the mutations abrogate binding to all three Dvl proteins. These findings suggest a molecular basis for the loss of Vangl2 protein function seen in Lp mouse mutants. Considering that stringency of selection medium provides a "semi-quantitative" measure of the binding affinities between different partners, our findings further suggest that the Vangl2/Dvl1 interaction is most mutation-sensitive, possibly reflecting a weaker affinity of Vangl2 for Dvl1 than for Dvl2/3.

DISCUSSION
The Vangl1 and Vangl2 genes encode membrane proteins that have been highly conserved during evolution and are derived from an ancestral precursor Stbm/Vang found in flies. Genetic studies in model organisms have shown that Vangl2 plays a critical role during development (31,33): mutations in mice cause defects in the formation of the neural tube (3)(4)(5) and Stbm/Vang inactivation in flies alters the organization of structures in the eye (ommatidia) and wing (hair) (8,9). These phenotypes indicate that Vangl proteins function to organize planar cell polarity and to direct convergent extension (CE) movements during embryogenesis. The precise function of Vangl proteins during these complex processes is poorly understood, and the mechanism by which Vangl2 inactivation leads to neural tube defects is completely unknown.
Unlike flies and worms, which have a single Vangl gene, most studied vertebrates (except Xenopus) have two, Vangl1 and Vangl2. In a first attempt to identify structure/function relationships in this protein family, we have carried out multiple sequence alignments of known Vangl proteins to search for conserved structural features. We identified several common domains that might underlie characteristic functions of the Vangl protein family. A first conserved feature of Vangl proteins is their four putative TM domains. Interestingly, identities of TM1, -2, and -3 as transmembrane domains are predicted with a higher degree of certainty than TM4, raising the possibility that there are only three TM domains. An invariant PDZ binding motif is identified at the C terminus and presumably interacts with PDZ-containing proteins normally found at the cytoplasmic face of the plasma membrane (46,47). The presence of multiple conserved N-linked glycosylation sites (NX(S/T)) in the N-terminal domain suggests that this domain might in fact be extracellular. Taken together, an alternative model for Vangl proteins with an intracellular C terminus, three TM domains, and an extracellular N terminus might be proposed. Distinguishing between these the two models will require additional direct topological studies of Vangl proteins by epitope mapping (48) or other methods (49).
A second observation is that the predicted cytoplasmic domain is the most conserved region of Vangl proteins. In particular, two sub-domains (II and IV) show a very high degree of similarity, suggesting that they serve in a conserved function of these proteins. This is highlighted by the fact that both known Lp loss-of-function mutations map to these sub-domains. These observations together with results in Figs. 2-5 strongly suggest that these two conserved subdomains play a critical role in the interaction of Vangl proteins with other core PCP proteins.
Finally, we identified a number of invariant residues in addition to the two residues mutated in independent Lp alleles. There are several basic residues as well as four cysteines in mammalian Vangl1/2 that are absolutely conserved across the Vangl family. This suggests that disulfide bridges might participate in a common membrane-associated organization of these proteins.
Neural tube closure involves CE, and mutations in Vangl2 and Dvl1/Dvl2 cause the same neural tube defect called craniorachischisis (3,4,38). Both Vangl2 and Dvl proteins are central to the PCP and CE pathways (31,35,39), and they have been shown to interact physically (6,20). However, it is unclear whether these properties extend to all members of the Vangl and Dvl families. Thus, we looked for possible interactions between mouse Vangl1/2 and Dvl1/2/3 proteins. Using a yeast two-hybrid system, GST pull-down, and co-immunoprecipitation assays, we observed that Vangl1 and Vangl2 could interact with all three Dvl proteins. This involves the cytoplasmic domain of Vangl1/2 and the N-terminal portions of Dvl (DIX plus PDZ domains) but does not require the N-terminal portion of Vangl1/2. Interestingly, the cytoplasmic domain of Stbm/Vang has been shown to interact with another PCP protein, Prickle (20,23); in particular, an 85-amino acid segment overlapping the poorly conserved sub-domain III has been shown to underlie the Pk interaction (23). Additional studies are required to determine if interactions with Dvl and Pk involve independent or overlapping sets of Vangl1/2 C-terminal segments. Finally, the sequence-divergent N-terminal portion of Vangl proteins is nevertheless important for function, because a 13-amino acid in-frame insertion at position 21 causes a loss-of-function in the Vangl2 homolog of the zebrafish trilobite (tri tc240a ) mutant (33). These results suggest that the N-terminal segment of Vangl proteins might be involved in other aspects of protein function, including membrane targeting, sorting to appropriate compartments, and/or possible interaction with other PCP proteins.
Independent mutations (S464N and D255E) in Vangl2 cause the severe neural tube defect observed in the two known allelic variants of the Lp mouse. S464N and D255E represent loss-offunction mutations, because: (a) these variants are specific to Lp chromosomes and are not found in 36 other phylogenetically distant mouse strains; (b) they affect residues that are either highly conserved (Ser-464) or invariant (Asp-255) in the Vangl family; and (c) the Lp defect can be corrected in transgenic mice carrying bacterial artificial chromosome clones containing an intact copy of Vangl2 (3,4,50). The NTD of Lp mice is inherited in a recessive fashion but appears to be in a gene dosagesensitive pathway, because the Lp/ϩ mouse exhibits a mild looped tail phenotype, whereas the Lp/Lp mouse has both a looped tail and severe craniorachischisis. Here we show that the two Lp mutations abrogate or strongly impair interaction between Vangl2 and all three members of the Dvl family, thereby providing a molecular basis for craniorachischisis in Lp mice. How do these results fit with the known role of Vangl and Dvl in PCP and CE? The little functional data available come from subcellular localization studies in flies during PCP signaling in the eye and in the wing (reviewed in Refs. 11 and 39). In both cases, the core PCP proteins Dsh, Pk, Stbm/Vangl, and from HEK293 cells transfected with either pCB6 plasmid or with pCB6-Vangl1 or pCB6-Vangl2 constructs were separated on SDS-PAGE and analyzed by immunoblotting with either anti-Vangl1 (left) or anti-Vangl2 (center) antisera or with an anti-c-Myc antibody directed against a corresponding antigenic epitope inserted in Vangl1 and Vangl2 (right). C, Vangl1/2 were individually expressed with Dvl2 by transient cotransfection in HEK293 cells, followed by immunoprecipitation with anti-Dvl2 antibody (IP:Dvl2), and by immunoblotting with anti-Vangl1 (left panel, W:Vangl1) or anti-Vangl2 antisera (right panel, W:Vangl2). Protein extracts used in the immunoprecipitation assay (input, 1/20 part of the volume subjected to the IP) were analyzed by parallel immunoblotting with anti-Vangl1, anti-Vangl2, and anti-Dvl2 antibodies (bottom panels). D, Vangl1 and Dvl2 were expressed by transient co-transfection in HEK293 cells, followed by immunoprecipitation with anti-Vangl1 antiserum and by immunoblotting with anti-Dvl2 (as in C). The asterisk identifies the presence of a nonspecific protein species cross-reacting with anti-Dvl2 antibody.
FIG. 5. Loss-of-function mutations in Vangl2 impair interaction with Dvl proteins in a yeast two-hybrid system. A, fusion proteins, containing the GAL4DBD and the cytoplasmic domains of either wild type Vangl2 (Vangl2wt) or Vangl2 mutant D255E (Vangl2-255) and S464N (Vangl2-464) corresponding to independent alleles of the neural tube defect-causing looptail mouse mutation, were expressed in yeast cells and analyzed by immunoblotting with anti-c-Myc antibody. The expressed proteins are marked by arrowheads; molecular weights of the proteins markers are indicated. B, the ability of wild type and mutant variants of Vangl2 to interact with different Dvl proteins was tested in a yeast two-hybrid system, as described in legend to Fig. 2. The graded stringencies of the selection media are indicated as low, intermediate, and high below each panel.
Fz cluster at the apical side of the cell at the beginning of PCP signaling and then become asymmetrically distributed at the end of signaling. Our findings suggest that Vangl2 mutations interfere with binding to Dvl proteins, implying that this physical association is critical for regulation of PCP and CE during mammalian development. It is conceivable that disruption of the Vangl:Dvl part of the tertiary multiprotein complex would affect interactions of other PCP proteins within the complex leading to the interruption of the PCP signaling event. It will be interesting to analyze a parallel effect of Vangl2 mutations on interactions with other PCP proteins.
In this study we show that, like Vangl2, Vangl1 can interact with the three Dvl proteins, raising the interesting possibility that Vangl1 might be implicated in regulation of PCP and CE during embryogenesis. Vangl1 and Vangl2 might be functionally redundant but could regulate PCP/CE in distinct cell populations individually expressing one or the other protein. Alternatively, Vangl1 and Vangl2, although highly similar, might have certain distinguishing functional features that would be simultaneously required for directing PCP/CE in the same cells. A formal clarification of the role of Vangl1, if any, in establishing PCP and CE (including neural tube closure) awaits the creation and characterization of a mouse mutant bearing a null mutation at this locus. However, we have recently established that Vangl1 and Vangl2 share a non-overlapping pattern of mRNA expression in the developing neural tube. At the time of closure, Vangl2 is broadly expressed in the neural tube except for a narrow segment of neuroepithelial cells in the floor plate (3). This narrow band of cells turns out to be the only site of Vangl1 expression in the developing neural tube (data not shown). These restricted Vangl1 and Vangl2 expression patterns might explain the inability of the former to rescue Lp phenotype. In addition, it has recently been shown in zebrafish that abundant ectopic expression of the Vangl1 homolog (microinjected RNA) could partly rescue the CE phenotype of loss-of-function Vangl2 (trilobite) mutants (28). Together, these results support a model in which Vangl1 and Vangl2 might direct PCP/CE in different cell populations.