Dishevelled 2 is essential for cardiac outflow tract development, somite segmentation and neural tube closure

Dishevelled (dsh in Drosophila or Dvl1 in mice)is a member of the conserved Wg/Wnt developmental pathway(Klingensmith et al., 1994;Theisen et al., 1994). Wg/Wnt proteins regulate cell fate and subsequent cell behavior(Bejsovec and Arias, 1991;Cadigan and Nusse, 1997;Moon et al., 1997). At least 18 different Wnts have been identified in humans but only three dishevelled genes (Dvl1, Dvl2 and Dvl3) have been found in both humans and mice (Sussman et al.,1994; Klingensmith et al.,1996; Tsang et al.,1996; Lijam and Sussman,1996; Yang et al.,1996; Semenov and Snyder,1997). All of the dishevelled proteins contain three highly conserved motifs. The N-terminal DIX domain is involved in regulating the Wnt signal via binding to axin (Zeng et al.,1997; Smalley et al.,1999; Kishida et al.,1999). An internal PDZ (or GLGF/DHR) domain in all Dvl proteins is present in many diverse membrane associated proteins, suggesting a role in cell-cell communication (Ponting et al.,1997). Finally, a DEP domain present at the C terminus is found in many G-coupled proteins. In addition, Dvl2 is the only Dvl protein that contains a proline rich SH3-binding domain. The significance of this domain is unknown. The DIX domain and N-terminal region of the DEP domain are necessary to activate the canonical Wnt pathway to mediate dorsoventral (DV) and anteroposterior (AP) patterning, as well as cell fate determination. By contrast, the PDZ domain and C-terminal part of the DEP domain are implicated in the planar cell polarity pathway (PCP) in fly to control the polarity of epithelial cells in the plane orthogonal to their apicobasal axis(Axelrod et al., 1998;Boutros et al., 1998;Moriguchi et al., 1999). Overexpression analysis in Xenopus has implicated these two domains,and most likely the PCP pathway, in the control of the convergent extension morphogenetic movements (Wallingford et al., 2000; Wallingford and Harland, 2001). Although functional significance can be assigned to each of these conserved motifs, the exact biochemical role ofDishevelled in transducing the Wnt signal is unknown. At present no human diseases have been linked to defects in DVL1, DVL2 orDVL3.

The similarity between dsh mutants and the wg phenotype in Drosophila (Perrimon and Mahowald, 1987) indicates that an overlap in phenotype might be expected between Wnt mutants and Dvl knockout mice. All three of the murine Dishevelled genes are widely expressed in embryonic and adult tissues suggesting that there may be functional redundancy among the three genes. These findings make it difficult a priori to predict specific classes of defects that may occur in mammals after disruption of each of these genes. Surprisingly, null mutants for Dvl1 exhibit deficits in social interaction and sensorimotor gating (Lijam et al., 1997). By contrast, mouse mutants that contain null mutations in Wnt genes result in mice with varied brain and developmental abnormalities (Cadigan and Nusse, 1997). For example, mice that are deficient inWnt1 display midbrain and hindbrain (cerebellar) abnormalities(McMahon and Bradley, 1990;Thomas and Capecchi, 1990),and Wnt1/Wnt3a double mutants have additional deficiencies in the neural crest (Ikeya et al.,1997). The neural crest is a migratory group of cells that emanate from the dorsal neural tube and give rise to various cell populations,including melanocytes, the dorsal root ganglia and the cardiac neural crest. Cardiac neural crest cells originate from the occipital neural tube and aid in septation of the outflow tracts and in aortic arch development.

The murine Dvl1 and Dvl2 genes are 64% identical and it is unknown whether Dvl2 has a similar role as Dvl1 in development. To address this question and to define further the role of individual Dishevelled genes in mammalian development, mice were generated that were deficient in Dvl2. We found that Dvl2 is essential for normal cardiac morphogenesis, somite segmentation, and neural tube closure. In addition, there is functional redundancy betweenDvl1 and Dvl2 in somite development and neural tube closure,as Dvl1/2 double mutants display more severe defects than theDvl2 single mutants.

A Dvl2 genomic clone was isolated from a mouse strain 129 genomic DNA library in λ FIX II (Stratagene) as previously described(Yang et al., 1996). The 3 kbBglII fragment containing most of intron 1 and part of exon 2 was cloned in the BamHI site of pPNT(Tybulewicz et al., 1991) in an orientation opposite to that of the direction of transcription of theneo marker to generate pPNT-BglI. The 4 kb HindIII fragment spanning exons 8 to half of 15 was first cloned in theHindIII site of pBluescript KS (Stratagene) and excised as aNotI/XhoI fragment for cloning into theNotI/XhoI sites of pPNT-BglII. The final construct,pPNT-Dvl2, has the neo gene interrupting the Dvl2-coding sequence at amino acid 71 (the BglII site in exon 2) and deleting exons 3-6 (the HindIII site in intron 6). Knockout mice were produced by gene targeting in TC1 embryonic stem (ES) cells(Deng et al., 1996). ES cell genotyping was performed by Southern blotting using a flanking probe andBamHI/EcoRI double digestion with [³²P]dCTP random primed radiolabeled probe (High Prime-Boehringer Mannheim) and five clones were correctly targeted. Three of these clones were injected and all gave high frequency germline transmission. Lines were established in inbred 129S6 and mixed lines of 129S6 × NIH Black Swiss, as previously described (Deng et al., 1996). Genotyping for subsequent studies was by PCR using genomic tail DNA from adult animals and DNA extracted from yolk sac for early gestation embryos. The PCR primers were as follows: Dvl2/15 forward primer in exon 5,5′-AGCAGTGCCTCCCGCCTCCTCA; Dvl2/16 reverse primer in exon 7,5′-CCCATCACCACGCTCGTTACTTTG; NLpgk neo forward primer,5′-AGGCTTACCCGCTTCCATTGCTCA. All animal experiments were carried out under protocols approved by the NHGRI/NIH and UCSD Animal Care and Use Committees and following NIH guidelines.

Brain tissue (0.2 mg) from wild-type and Dvl2^-/- mice was homogenized in 1 ml of RIPA buffer containing Pefabloc (Boehringer Mannheim). Homogenates were subjected to centrifugation for 1 minute at full speed in a microcentrifuge. An aliquot of the supernatants was saved for A280 measurement (for normalization of gel loading), while the remainder was mixed with an equal volume of 2× SDS gel loading buffer and boiled for 3 minutes. Aliquots (20 μl) were loaded on a 7.5% polyacrylamide SDS minigel and subjected to electrophoresis and transfer to nitrocellulose membrane for western blot analysis. A polyclonal antibody to the C terminus of Dvl1(Luo et al., 2002) and monoclonal antibodies to the C terminus of Dvl2 (2-10B5)(Song et al., 2000) and Dvl3(3-4D3) (Strovel and Sussman,1999) were used in conjunction with peroxidase-conjugated anti-mouse IgG (Sigma) for chemiluminescent detection (Amersham Pharmacia Biotech).

Embryo hearts were collected at 18.5 dpc or embryos were collected at 8.5-10.5 dpc and fixed in 3% aldehyde solution (1.5% paraformaldehyde, 1.5%glutaraldehyde) in 0.1 M phosphate buffer pH 7.5, dehydrated in a graded ethanol series, then stored in 100% ethanol until scanning. Samples were critically point dried, mounted and then coated with 300 Angstrom gold-palladium. Specimens were viewed and photographed with a Cambridge Instrument Stereoscan 360 scanning electron microscope (Scripps Institute of Oceanography Analytical Facility).

Embryos were prepared for in situ using the protocol detailed by Wilkinson(Wilkinson, 1992) with modifications. Embryos were dissected at the appropriate ages, fixed in 4%paraformaldehyde in PBS then dehydrated in a methanol series. Embryos were rehydrated in PBT (0.01% Tween-20 in PBS), treated with 6%H₂O₂ to remove endogenous peroxidases and then permeabilized with proteinase K. Hybridization was performed at 70°C overnight in hybridization solution (50% formamide, 5×SSC pH 4.5, 50μg/ml yeast tRNA, 1% SDS, 50 μg/ml heparin) using an RNA probe labeled with digoxigenin-UTP (Boehringer/Mannheim). Afterwards embryos were washed extensively in TBST (TBS with 0.1% Tween-20, Sigma) and blocked in 1% sheep serum in TBST. Transcripts were detected with anti-dig Fab′ conjugated with alkaline phosphatase followed by color reaction with X-phosphate/NBT substrate. Reactions were stopped with PBT and embryos stored in 80%glycerol/PBT until photographed.

Tissues were dissected and placed in 20 volumes of 10% buffered formalin,dehydrated, embedded in paraffin wax, sectioned at 8 μM and stained by routine methods at the UCSD Histology Core. Photographs were taken using a Leica DMR light microscope mounted with a Spot 2 camera.

Differential staining of mouse fetuses was carried out according to the procedure of McLeod (McLeod,1980). Embryonic 17 and 18 dpc fetuses were skinned and eviscerated, fixed in a 95% ethanol solution for 5 days, then treated in acetone for 4 days to remove fat. Fetuses were stained with 0.3% Alcian Blue,0.1% Alizarin Red S ethanol/acetic acid solution for 3 days followed by clearing in a graded series of glycerol/1% KOH (20-80% glycerol, 1 week each step). Skeletons were stored in 100% glycerol and photographed using a Leica dissection microscope and Spot 2 camera.

To produce mice that contain a null allele for Dvl2, part of exon 2 and all of exons 3-6 were disrupted by the insertion of PGKneo(Fig. 1A), replacing approximately 1.2 kb of the Dvl2-coding region and introducing multiple stop codons from PGKneo 3′ into exon 2. Targeting was performed in TC1 ES cells and five targeted clones were identified by Southern blot screening (Fig. 1B, lanes 2-6). This mutant allele was established in the germline from two separate clones in both mixed (129S6×NIH Black Swiss) and uniform (129S6) genetic backgrounds, demonstrated by southern blotting of newborn tail DNA(Fig. 1C). Homozygous,heterozygous and wild-type offspring were observed in F₂ litters in both backgrounds. Brain lysates from Dvl2^+/+ andDvl2^-/- adults were used for Western blot analysis. Dvl2 protein was absent from Dvl2^-/- brain lysates, and there was no compensatory increase of either Dvl1 or Dvl3 protein in mutant brains(Fig. 1D). PCR primers were developed that were effective in detecting both mutant and wild-type genotypes(Fig. 1E).

We examined mutant mice in inbred 129/SvEv or mixed 129/SvEv×NIH Black Swiss backgrounds and found that Dvl2^-/- mice can survive to adulthood and are fertile, but they were born in reduced numbers from heterozygous crosses. Of 352 offspring at weaning, there were 130Dvl2^+/+, 180 Dvl2^+/- and only 42Dvl2^-/- mice, a reduction of more than 50% of expected numbers. However, at 18.5 dpc, Dvl2^+/+,Dvl2^+/- and Dvl2^-/- embryos were recovered in the expected mendelian ratios, suggesting that death occurred in the perinatal period. This was confirmed by directly observing litters at birth, where some newborn Dvl2^-/- pups failed to thrive. Such pups appeared to have difficulty breathing, were often cyanotic, did not feed and displayed reduced mobility. These pups died within 24 hours. The perinatal lethality seemed to be unaffected by genetic influences caused by mouse strain modifiers as the survival percentages forDvl2^-/- mice were identical for both the mixed (16 observed out of 31 expected, n=125) and inbred strains (six observed out of 13 expected, n=51). Surprisingly, survivingDvl2^-/- mutants were predominantly female in either strain background (Table 1),suggesting an interaction between genotype and sex.

Dvl2^-/- mice that survived beyond 24 hours grew normally into adulthood (2+ years) but ∼25% (n=17) of the surviving mutant mice had kinked tails. A percentage of survivingDvl2 mutants also exhibited scoliosis and in rare instances displayed vestigial tail (n=3). This phenotype is very similar to the haploinsufficiency phenotypes seen in Wnt3A and T(brachyury) mutants (Yamaguchi et al.,1999). Embryo analysis in early gestation at 9.5 dpc revealed that 2-3% of Dvl2^-/- embryos had incomplete thoracic neural tube defects (spina bifida) and exencephaly (see below). Serial sections of adult Dvl2^-/- tissues showed no other morphological abnormalities (data not shown). Unlike the Dvl1^-/- mice,Dvl2^-/- homozygous mice displayed no abnormalities in social interaction or sensorimotor gating (data not shown).

Dvl2^-/- newborns had no abnormalities of the palate,trachea or lungs (data not shown). We next examined the hearts of 13, 18 and 20 dpc. Dvl2^-/- animals obtained fromDvl2^+/- crosses. Sixty-eight mice (16Dvl2^+/+, 36 Dvl2^+/- and 16Dvl2^-/- mice) were examined for cardiovascular defects. Eight Dvl2^-/- (50%) of the mutant mice displayed cardiovascular abnormalities (Table 2). An additional embryo displayed a cardiac abnormality but was not included in Table 2 as genotype information for remaining littermates could not be determined because of a technical error. Eight Dvl2^-/- embryos had structurally normal hearts, similar to the wild type(Fig. 2A,B). Double outlet right ventricle (DORV) was the most common defect and the placement of the aorta varied. Six mutant embryos had DORV in conjunction with ventricular septal defects (Fig. 2C,D,Fig. 3B). In one case of DORV the aorta emerged parallel to the pulmonary valve where the orientation of both great vessels were inverted and arose from the right ventricle(Fig. 2C,D). In fact, this was the most frequent manner in which DORV presented. One mutant had transposition of the great arteries (TGA). In addition, two of the mutant embryos had persistent or common truncus arteriosis (TA), a condition in which the outflow tract fails to divide into an aorta and pulmonary artery(Fig. 2E,F). These defects were clearly seen in serial sections of abnormal Dvl2^-/- hearts(Fig. 3). TGA was evident in one heart (Fig. 3C) where the aorta emerged from the right ventricle. The severity of these defects would be sufficient to account for the perinatal lethality ofDvl2^-/- mice. None of the Dvl2^+/- orDvl2^+/+ mice examined displayed any conotruncal abnormalities or other heart defects.

In separate matings,Dvl1^-/-;Dvl2^-/- embryos were examined for cardiac defects, to see if there was evidence for redundancy betweenDvl1 and Dvl2 in conotruncal development. Of eight double homozygous embryos examined between 14.5 and 18.5 dpc, three displayed conotruncal defects (Table 2),including DORV, TGA and PTA. These results are consistent with a primary role for Dvl2 in cardiac outflow tract development, and strongly suggest that Dvl1 and Dvl2 have non-redundant roles in heart morphogenesis.

The defects observed in the morphology of the great vessels inDvl2^-/- mice (DORV, TGA and PTA) are congenital heart defects that occur in embryogenesis due to abnormalities in outflow tract septation (Chien, 2000;Srivastava and Olson, 2000). Proper formation of the cardiac great vessels involves an early looping event occurring between 8.5-9.0 dpc followed by rotation of the developing conotruncal arteries. Between 11.5 and 12 dpc the developing conotruncus first rotates, followed by the development and fusion of the conotruncal cushions. As the endocardial cushions grow and migrate they develop chirality along the outflow tract. In the final step, the cushions fuse to form the aorticopulmonary septum. Cardiac neural crest cells are believed to migrate into the truncal cushions as they form and contribute to septation. Neural crest ablation studies have identified a subpopulation of neural crest cells at the level of the fourth and sixth aortic arch that affects proper septation termed the cardiac neural crest (Kirby et al., 1983; Creazzo et al.,1998; Kirby and Waldo,1995).

Therefore, markers for early cardiac neural crest were employed, such asPitx2 and plexin A2. Pitx2 is co-expressed withPax3, connexin 43 and the endothelin receptor A in the cardiac neural crest, and Pitx2 mutant mice display similar outflow tract defects asDvl2 (Kioussi et al.,2002). We used a Pitx2 probe for whole mount analysis at 10.5 dpc In wild-type embryos, Pitx2 is normally expressed in a number of tissues, including the fourth branchial arch and migrating cardiac neural crest (Fig. 4A,C,E).Dvl2^-/- embryos (Fig. 4B,D,F) and Dvl1/2 double mutants (data not shown) had nearly undetectable levels of Pitx2 in the branchial arches and in neural crest cells migrating to the cardiac outflow tract(Fig. 4B), as well as in the outflow tract (Fig. 4D,F). These results suggest that the outflow tract defects seen inDvl2^-/- mutant embryos are associated with cardiac neural crest defects. Subsequently, the expression of the cardiac neural crest markerPlexin A2 (Brown et al.,2001) was examined. Plexin A2 expression was greatly reduced in hearts from Dvl2^-/-(Fig. 4H) mutant embryos compared with wild-type hearts (Fig. 4G), further supporting the notion that the conotruncal defects displayed by Dvl2^-/- mutants result from neural crest abnormalities.

As part of our phenotypic analysis, we examined bone and cartilage development using Alizarin Red and Alcian Blue staining in 18Dvl2^-/- newborns. Nearly all (90%) of the Dvl2mutant mice displayed mild abnormalities of the vertebral bodies and ribs(Fig. 5,Table 3). Most defects were localized dorsally in the vertebral ribs and vertebral bodies(Fig. 5A-E). Many of the abnormal thoracic vertebrae were disorganized(Fig. 5A) and fusion of the ribs near the tubercle was evident (Fig. 5B). We determined whether the normal rib number was altered in any of the mutant mice. In all but one case we found that in the presence of either forked or fused ribs the total rib number on both the left and right sides was 13. These findings suggest that the normal number of ribs was specified, but segmentation did not occur properly. Dvl1/2 double homozygotes had more severe skeletal malformations of the type seen inDvl2^-/- mice (Fig. 5F). For example, a 14.5 dpc. double mutant embryo demonstrated numerous collapsed vertebrae and extensive fusion of the ribs along the vertebral column. These defects were not observed inDvl1^-/- mice (Lijam et al., 1997), suggesting that functional redundancy restricted this phenotype in either Dvl1^-/- or Dvl2^-/-single mutants.

A sternal defect was detected in one newborn mouse, and affected the sixth sternebra and the xiphoid process (Fig. 5H). A completely split or perforated xiphoid process was evident. Seven ribs are attached to the sternum but no malformations were detected in any of the seven sternal ribs. Overall, these data demonstrate thatDvl2 is essential for proper formation of ribs, vertebral bodies and sternum.

Somites originate as cells that pinch off from the presomitic mesoderm. As the somite condenses, it develops into a disorganized ball of cells internally surrounded by a layer of columnar epithelial cells. The basolateral wall of the epithelial cell undergoes an epithelial-to-mesenchymal transition that results in the formation of sclerotome ventrally and dermomyotome dorsally. The sclerotome and dermomyotome collectively give rise to the cartilage, bone,muscle and connective tissue (Keynes and Stern, 1988; Huang et al.,2000). In addition, anterior and posterior domains exist within the sclerotome as demonstrated with mesodermal probes that identify these distinct cell populations. To investigate somite differentiation in theDvl2 single and Dvl1/2 double mutants, we performed whole-mount in situ hybridization using a variety of somite markers.

No differences were found between wild-type and Dvl2^-/-mutants using several markers, including the caudal somite markersUncx4.1, delta 1 and lunatic fringe, as well as the rostral marker delta 3 (data not shown). To determine whether the axial defects that affect rib fusion could be due to abnormalities that precede somite division into caudal and rostral halves, we used probes that identify the presomitic mesoderm. There were no alterations in the coordinated expression ofNotch1 or Notch2 in wild-type andDvl2^-/- mutants (Fig. 6C-F). However, myogenin, a myotomal marker, was occasionally attached between two somites of Dvl2 mutants as a bridge(Fig. 6A,B), suggesting that abnormal, incomplete segmentation could be the cause of the somite defects seen in these mutants.

As Dvl1/2 double mutants display rib and vertebral defects of the same type seen in Dvl2^-/- mice, but of more severe extent and at higher frequency, we examined the expression pattern of somite markers in the double mutants, hoping to increase the chance that relevant abnormalities in marker expression would be observed. Similar to the results for myogenin in Dvl2^-/- mice, Uncx4.1 was detected in a fused band between adjacent somites, indicating incomplete segmentation of these two somites (Fig. 6I,J) in the Dvl1/Dvl2 double mutant embryo. The caudal probe delta 1 was expressed throughout the appropriate regions of the somite,but the spacing was irregular in the double mutant(Fig. 6K,L). In both of these cases, the abnormal somites appear to be split, consistent with the normal ultimate rib and vertebral numbers of the Dvl2 and Dvl1/2mutants. Lunatic fringe transcripts were present in the forebrain, placodes,neural tube and dermomyotome of the wild-type embryo(Fig. 6G). Overall, this pattern was repeated for the Dvl1/2 mutant but expression in the dermomyotome was markedly depressed (insetFig. 6H). No differences were found in Dvl1/2 double mutants using delta 3. Thus, minor abnormalities were evident in somite precursors of Dvl2 mutants,which were more severe in the Dvl1/2 double mutant. These defects appear to be due to mild abnormalities in somite segmentation.

Approximately 2-3% of Dvl2 homozygotes have incomplete thoracic neural tube defects (Fig. 7B),often associated with exencephaly, compared with normal neural tube closure in wild-type embryos (Fig. 7A). Although the penetrance was low, we reasoned that there may be functional redundancy among the Dvl genes because of their overlapping patterns of expression and structural similarity. To test for functional redundancy,Dvl1^+/-;Dvl2^+/- double heterozygous crosses and Dvl1^-/-;Dvl2^+/- crosses were performed. Litters were genotyped at weaning. NoDvl1^-/-;Dvl2^-/- mice were identified(20 expected, none observed, n=230). Embryos were dissected at various times from these crosses. Surprisingly, we found thatDvl1^-/-;Dvl2^-/- embryos(Fig. 7C) displayed completely open spinal neural tubes and exencephaly (cranio-rachischisis). The neural tube closure defects resulted in the development of the brain outside of the cranium, and the spinal cord was completely open to the base of the tail. The face was fused, as was the tail, suggesting that some regions of the neural tube closed normally. At 9.5-10.5 dpc, the neural tubes of wild-type embryos were completely closed in the midbrain, hindbrain and thoracic region(Fig. 7D), while the neural tubes of Dvl1/Dvl2 double mutants were completely open from the midbrain to the tail (Fig. 7E). Of 380 embryos dissected between 8.5 and 10.5 dpc from bothDvl1^+/-;Dvl2^+/- double heterozygous crosses and Dvl1^-/-;Dvl2^+/- crosses,40 Dvl1^-/-;Dvl2^-/- double homozygotes were observed (45 expected). Of these, 31 had open unfused neural tubes, one had an open neural tube with fused epidermal ectoderm, six had partially open neural tubes and two appeared normal.

There were clear defects in the developing spinal cord and brain resulting from closure defects. In spite of such defects, there appeared to be recognizable regional differentiation of forebrain, midbrain and hindbrain regions, even though there was marked disorganization secondary to exencephaly(Fig. 7F,G). Similarly, the dorsal and ventral regions of the mutant open spinal cord were recognizable(Fig. 7H,I).

Thus, Dvl2 is essential for neural tube closure in mice, and there are overlapping functions between Dvl1 and Dvl2 in neurulation.

We have generated mice carrying null mutations in Dvl2 and have found that 50% of Dvl2 null mutants display perinatal lethality because of defects in cardiac morphogenesis. Specifically, outflow tract septation defects were observed in approximately 50% ofDvl2^-/- mice analyzed in late gestation, including DORV with VSD, PTA and TGA. These cardiac defects are incompatible with life and would account for the perinatal lethality of Dvl2 mutants. These findings demonstrate a novel role for a dishevelled gene in cardiac development and specifically identify Dvl2 as a key mediator in conotruncal development. There was no evidence for redundancy betweenDvl1 and Dvl2, as the penetrance of cardiac defects was similar in Dvl2 and Dvl1/2 mutants. In addition, almost all(90%) of the Dvl2 homozygous newborn null mice displayed abnormalities in patterning of the axial skeleton in which the sternal ribs and the vertebrae were affected. These defects appeared to be due to perturbations in the segmentation of the somites, implicating Dvl2 in the control of somitogenesis. These segmentation defects were even more severe in Dvl1/2 double mutants, suggesting a dose dependence of Dvl gene function in somite formation. Finally, 2-3% of Dvl2 null mice exhibited spina bifida. Similar to somite segmentation, defects in neural tube development were even more pronounced in mice deficient for both Dvl1and Dvl2. Dvl1^-/-;Dvl2^-/- mice display a completely open neural tube with craniorachischisis indicating that the increase in severity of the neural tube phenotype is also Dvl-dose dependent. These results demonstrate that although there is significant functional redundancy between Dvl1 and Dvl2 in somite development and neural tube closure, a somewhat unique role in cardiac outflow tract patterning can be assigned to Dvl2.

Mice with targeted inactivation of the Dvl1 gene have been previously described (Lijam et al.,1997). These mice were found to exhibit alterations in sensorimotor gating and social interaction. Surprisingly Dvl2 does not seem to play a similar role in the regulation of social behavior or sensorimotor gating, as no abnormalities in these processes were observed. These observations suggest a unique function for Dvl1 in social behavior and sensorimotor gating.

The heart phenotype manifested by the loss of Dvl2 was specific to the outflow tract. Proper alignment and development of the aorticopulmonary septum involves migration and fusion of the endocardial cushions, as well as a correct pivoting event along the axis that forms the AV canals. Genes that affect early cardiac looping such as Nkx2.5, Mef2c, as well as the bHLH proteins Hand1 and Hand2 have been identified(Olson and Srivastava, 1996;Thomas et al., 1998;Srivastava et al., 1995;Lyons et al., 1995;Lin et al., 1997;Bruneau et al., 2000). Improper cardiac looping can lead to outflow tract defects and tetratolgy of Fallot based on studies in the mouse (Srivastava,2000; Creazzo et al.,1998; Olson and Srivastava,1996).

In the case of Dvl2 mutants, a defect in cardiac neural crest appears to be responsible for the observed outflow tract defects. Cardiac neural crest cells are essential for normal development of the outflow tract. These cells originate from the caudal hindbrain and migrate into the caudal pharyngeal arches (third, fourth and sixth). A subset of these cells continues to migrate into the cardiac outflow tract where it will organize the aorticopulmonary septum. If cardiac neural crest cells are removed prior to migration, several predictable outflow tract phenotypes are observed after development of the heart and great arteries is complete, including DORV, TGA and PTA (Kirby et al., 1983;Creazzo et al., 1998)(reviewed by Kirby and Waldo,1995; Chien,2000). Pitx2 is co-expressed with Pax3 in the cardiac neural crest, and Pitx2 mutant mice display similar outflow tract defects as Dvl2 (Kioussi et al., 2002). Plexin A2 has recently been identified as a marker for cardiac neural crest that is expressed at later times in development(Brown et al., 2001). Therefore, we examined the expression of Pitx2 and plexin A2 inDvl2 and Dvl1/2 mutant mice. Expression of these markers along the migrating cardiac neural crest were impaired in Dvl mutant mice,implicating the neural crest in the outflow tract defects displayed byDvl2^-/- mice.

We recently found that Dvl2 and Pitx2 were part of a common pathway regulating proliferation in specific tissues(Kioussi et al., 2002). We found that Dvl2 and Pitx2 genetically interact to produce cardiac outflow tract abnormalities. We further demonstrated that Pitx factors can exert essential roles in cardiac neural crest cell development based on the β-catenin-dependent transcriptional induction of Pitx2. Pitx2, in turn, acts upstream of genes required for the cell proliferation program, including cyclin D1 and cyclin D2. Components of this Wnt/Dvl/β-catenin→Pitx2 pathway are required, in a dose-dependent fashion, for physiological proliferation of specific cells within the cardiac outflow tract, in particular the cardiac neural crest, and pituitary gland. These findings strongly support our interpretation that the primary outflow tract defect in Dvl2 mutant mice is in the cardiac neural crest. In addition, they suggest that Dvl2 and Pitx2, among other proteins, could be novel components of a multigenic origin of cardiac outflow tract defects that occur in the human population. As cardiac outflow tract abnormalities account for ∼30% of all cardiovascular malformations in humans(Chien, 2000;Srivastava and Olson, 2000),this pathway could be important for human conotruncal defects as well.

These findings support a role for the Wnt pathway in cardiac morphogenesis through the control of cardiac neural crest development. There is some earlier evidence that the Wnt pathway plays an important role in cardiac morphogenesis, beginning with studies using antisense attenuation ofWnt1 and Wnt3a expression in whole embryo cultures(Augustine et al., 1993). Recent reports demonstrate that Wnt inhibition induces cardiogenesis inXenopus (Schneider and Mercola,2001) and in chick (Marvin et al., 2001), but the role of Wnt signaling in mammalian systems was unclear. So far, no Wnt pathway mutants display cardiac defects. However,neural crest abnormalities do occur in the Wnt1/Wnt3a double mutant. In addition, using a floxed β-catenin allele and the Wnt1-Cre transgenic mouse that expresses Cre in the neural crest, we were able to deleteβ-catenin completely in the neural crest. These mice displayed similar conotruncal defects as the Dvl2^-/- mice(Kioussi et al., 2002),consistent with a role for Dvl2 in the cardiac neural crest.

Downstream of Dishevelled, the Notch and Wntpathways have both been implicated in somite formation and segmentation. Members of the Notch pathway are intimately involved in controlling somitogenesis (reviewed by Muskaitch et al., 1994;Artavanis-Tsakonas et al.,1999), including Notch1(Conlon et al., 1995),Notch2, delta 3 (Kusumi et al.,1998) and lunatic fringe(Evrad et al., 1998; Zhang et al., 1998). Notch is downstream of Dsh inDrosophila (Axelrod et al.,1996; Couso and Martinez Arias, 1994). Notch1 and Notch2 expression were normal in the somites of Dvl2 mutants, indicating that development of the somites from presomitic mesoderm was unperturbed. However, somite segmentation was abnormal as defined by expression of the sclerotomal markerUncx4.1. These expression alterations were not identical to the severe defects displayed by Notch pathway mutants. Instead, rather subtle but distinct alterations in somite boundaries were evident as detected by the delta 1 and lunatic fringe. Wnt3a is also involved in paraxial mesoderm differentiation. Wnt3a mutants exhibit posterior truncation and lack a significant number of posterior somites(Takada et al., 1994). The somite defects in the Dvl2 mutant mice reflect mild defects in segmentation, rather than conversion of posterior to anterior fates, but a role for Dvl genes in posterior development cannot be completely ruled out.

Dvl2 is essential for normal neural tube closure, because a small number of embryos display thoracic spina bifica and exencephaly. Neural tube defects (NTDs) are a common class of birth defects in humans with an incidence that varies from 0.5-8 per 1000 live births(Elwood et al., 1992). Several factors are associated with NTDs, including genetic factors, teratogens and low levels of dietary folate. There is wide variation in the type and severity of NTDs in humans. Genetic epidemiology studies have suggested that NTDs have a multifactorial etiology with genetic predisposition because of many genes and a threshold beyond which environmental factors can trigger NTDs during crucial fetal periods. Several mouse mutants display NTDs. More than 50 loci have been identified, and many of the mutant alleles have been cloned(Juriloff and Harris, 1999). We have now identified Dvl2 as an additional locus important for neural tube closure. In addition, there is functional redundancy betweenDvl1 and Dvl2 in neural tube closure, as the double mutants display a completely open neural tube between the midbrain and tail, termed craniorachischisis.

Loop tail (Lp) is one other mouse mutant that displays cranio-rachischisis (Greene et al.,1998). The gene for Lp has been cloned(Kibor et al., 2001;Murdoch et al., 2001) and codes for a transmembrane protein loopin with a PDZ-binding domain.Ltap is related to the Drosophila gene Van Gogh,which is downstream of frizzled and Dishevelled in the planar cell polarity (PCP) pathway (Taylor et al., 1998; Wolff and Rubin,1998). These finding suggest the intriguing possibility that mammalian Dvls and Lp are in a common PCP-like pathway mediating neural tube closure. In Xenopus, Dishevelled is an integral part of the PCP pathway that regulates gastrulation via convergent extension(Wallingford et al., 2000),and has also been implicated in neural tube closure in Xenopus(Wallingford and Harland,2001; Wallingford et al., 2002). The Dvl1/Dvl2 phenotype may be the result in defects of convergent extension mechanisms via the PCP pathway, suggesting that Dvl1, Dvl2 and Lp are part of a common PCP pathway regulating neurulation in the mouse.

In summary, we have provided evidence that Dvl2 is important for cardiac outflow tract development via the cardiac neural crest, somite segmentation and neural tube closure. These findings, together with previously published (Lijam et al., 1997)and unpublished results, demonstrate that Dvl1 and Dvl2 are partially redundant, but also have unique functions. Somite segmentation and neural tube closure appear to be mediated by overlapping redundant functions that are dependent on the dose of these two genes. By contrast, social interaction and sensorimotor gating defects are unique to Dvl1mutants, while cardiac defects appear to be unique to Dvl2 mutants. The Dvl1 and Dvl2 mutant mice will be invaluable tools with which to continue to sort out the important pathways regulated byDishevelled genes to regulate complex behavior, cardiac outflow tract development and somite segmentation.

We are grateful to Ron Conlon, Randy Johnson, Tom Gridley and Chin Chiang who all provided probes for in situ analysis. We wish to thank Denise Larson,(NIH/NHGRI) Lisa Garrett, Theresa Hernandez (NIH/NHGRI Transgenic Core),Elaine Cheung (UCSD) and Kevin Nyguen (UCSD) for their technical assistance. We also wish to thank Charles Graham at the SIO Analytical Facility and Debbie Beiley (UCSD) for her continuous help in the preparation of this manuscript.