A complex syndrome of left-right axis, central nervous system and axial skeleton defects in Zic3 mutant mice

The failure to establish midline body pattern and left-right asymmetry properly is an important cause of congenital malformations. These defects are associated with considerable morbidity and mortality secondary to complex congenital heart defects (CHD), immune dysfunction caused by abnormalities of spleen position or number, gut malrotation, and midline central nervous system (CNS) defects. ZIC3 was the first gene identified to have a causal role in defects of human laterality. Loss-of-function mutations in ZIC3 cause HTX1, which results in heterotaxy or situs ambiguus, the aberrant positioning of internal organs with respect to each other and the left-right (LR) axis. Carrier females who are heterozygous for a ZIC3 mutation are clinically unaffected although a subset has situs inversus, the mirror image reversal of internal anatomy (Casey et al., 1993; Gebbia et al., 1997).

The molecular pathways required for correct LR axis specification and formation have been examined in different model organisms including Xenopus, chick, zebrafish and mouse (reviewed by Harvey, 1998; Burdine and Schier, 2000; Capdevila et al., 2000; Casey and Hackett, 2000). Pathways of asymmetric gene expression resulting in a cascade of lateralized signals that specify morphological asymmetry have been elucidated. This asymmetric gene expression is initiated at the node, or organizer, and occurs well before the first morphological sign of asymmetry, the rightward looping of the embryonic heart tube. These pathways appear to be conserved across species although the function of individual genes in the generation of LR asymmetry may be species specific.

There are at least three major processes involved in establishment of the LR axis: (1) node-dependent symmetry breaking; (2) establishment of the embryonic midline (notochord and ventral floorplate); and (3) nodal-dependent signaling in the left lateral plate mesoderm (LPM). Failure in any of these processes will result in disturbances of laterality, including the complete reversal of left-right anatomy (situs inversus) or randomization of organ position (situs ambiguus). Generally, situs inversus is believed to arise as a result of an early event in which either the specification of asymmetry or orientation of the LR axis with respect to anteroposterior and dorsoventral axes is disrupted. Examples include mouse mutants iv/iv (Hummel and Chapman, 1959), which is caused by deletion of left-right dynein, lrd (Supp et al., 1997; Supp et al., 1999), Kif3a (Marszalek et al., 1999; Takeda et al., 1999), Kif3b (Nonaka et al., 1999) and hepatocyte nuclear factor/forkhead homolog 4 (hfh-4) (Chen et al., 1998), all of which have abnormal node ciliary function or structure. inv mutants, in whom the gene product has an unknown function, also show situs inversus (Yokoyama et al., 1993; Mochizuki et al., 1998). Situs ambiguus is believed to result from failure to maintain asymmetric expression of genes involved in the nodal-dependent signaling pathway. This final common pathway, which appears to be conserved across species, is initiated with the expression of nodal in the perinodal region (Collignon et al., 1996; Lowe et al., 1996). This is followed by the expression of nodal and lefty2 (Meno et al., 1996; Meno et al., 1999), both TGFβ family members, and Pitx2, a homeobox transcription factor, in the left LPM. Pitx2 expression appears to be the most downstream gene in this cascade and is expressed asymmetrically in the LPM as well as in developing organs including the heart and visceral progenitors (Logan et al., 1998; Piedra et al., 1998; Ryan et al., 1998; Yoshioka et al., 1998; Campione et al., 1999; Lin et al., 1999). Although it is clear from individuals with HTX1 that ZIC3 plays a role in formation and/or specification of the left-right axis, its position within this molecular cascade is not known. The finding of both situs ambiguus and situs inversus in these individuals is unusual and suggests that Zic3 may play roles at multiple steps within the LR axis hierarchy.

The murine Zic gene was initially identified as a zinc-finger protein expressed in granule cells throughout the development of the cerebellum (Aruga et al., 1994). Further analysis in chick confirmed the role of Zic1 and Zic3 in the developing cerebellum (Lin and Cepko, 1998). Currently, four paralogous Zic family members have been described, all of which show conservation of zinc-finger motifs that place ZIC proteins within the GLI transcription factor superfamily (Aruga et al., 1996a; Aruga et al., 1996b). The expression pattern of Zic3 is suggestive of a role in body pattern formation (Nagai et al., 1997). In Xenopus, overexpression of Zic3 induces proneural genes as well as neural crest markers (Nakata et al., 1997). Further studies in Xenopus have suggested that Zic3 acts upstream of nodal in the signaling cascade of LR asymmetry (Kitaguchi et al., 2000). Other Zic family members are also known to have an effect on neural development and embryonic patterning, although no situs abnormalities have been described. Individuals with ZIC2 mutations have holoprosencephaly (Brown et al., 1998). A mouse model in which Zic2 functions as a hypomorphic allele shows a number of neural defects including holoprosencephaly, spina bifida, neurulation delay and delay in neural crest development (Nagai et al., 2000). Xenopus Zic5 (orthologous to murine and human Zic4) is also important for neural crest development (Nakata et al., 2000), whereas Zic1 appears to play an important role in skeletal patterning (Aruga et al., 1999) and cerebellar development (Aruga et al., 1998).

To address the role of Zic3 in body pattern formation and in left-right asymmetry, we have generated Zic3 null mice. Deletion of murine Zic3 leads to complex congenital heart disease, disturbances of laterality, neural tube defects and vertebral and rib defects, including partial homeotic transformations of the axial skeleton. These defects indicate that Zic3 plays an important role in both axial midline development and LR patterning. Furthermore, our results indicate that Zic3 is involved in the maintenance of nodal expression and is important both for LR axis specification and maintenance of lateralized signaling. Taken together, these results suggest a role for Zic3 in both left-right and anterior-posterior axis formation.

The mouse Zic3 locus was cloned by screening a mouse strain 129/SvEv genomic library. The targeting vector (Fig. 1A) was constructed using a 2.6 kb EcoRI-NotI fragment comprising the 5′ end of the Zic3 gene that extends 138 bp downstream of the translation initiation site, and a 3.9 kb ClaI-EcoRI fragment comprising the 3′ end of the gene, starting 311 bp downstream of the translation termination site. An IRES-lacZ-pA cassette followed by a PGKneobpA neomycin resistance expression cassette flanked by loxP sites (provided by Maki Wakamiya) was inserted between the two regions of Zic3 homology. A MC1tkpA herpes simplex virus thymidine kinase expression cassette was added to the 3′ arm of homology to enrich for homologous recombinants by negative selection with 1-2-deoxy-2-fluoro-b-D-arabinofuranosyl-S-iodouracil (FIAU) (Mansour et al., 1988).

The targeting vector was linearized at a unique SalI site outside the region of sequence homology and introduced into AB-1 embryonic stem (ES) cells as described (Behringer et al., 1994; Hogan et al., 1994). 386 G418/FIAU-resistant ES cell clones were initially screened by BamHI digestion and hybridized with a unique 3′ probe external to the region of vector homology. Expected sizes were 10.5 kb and 7.5 kb for the wild-type and targeted allele, respectively (data not shown). Six correctly targeted clones were identified. ES cells carrying the disrupted Zic3 allele were injected into C57Bl/6J (B6) blastocysts, and one ES cell clone was found to be capable of contributing to the germline of mouse chimeras. Mice were genotyped by Southern blot analysis using a SacI-XbaI fragment in the 5′ region of the gene (Fig. 1A). The expected sizes for the fragments were 3.2 kb for the wild-type and 5.8 kb for the targeted allele (Fig. 1B). Embryos were genotyped by multiplex PCR using the following primers: 5′ Zic3 TGC AGG CAT GGG ATT GAA TCC; 3′ Zic3 AAG AGC AAG TAG CTA GGA GGC; 5′ Sry TGA CTG GGA TGC AGT AGT TC; 3′ Sry TGT GCT AGA GAG AAA CCC TG; 5′ lacZ AAC TTA ATC GCC TTG CAG CA; 3′ lacZ GTA ACC GTG CAT CTG CCA GT. Expected product sizes were 386 bp for Zic3, 240 bp for Sry and 218 bp for lacZ. All primers were used in the same reaction with an annealing temperature of 58°C.

Whole-mount in situ hybridization was performed using a modification of the protocol described (Wilkinson, 1992), using Blocking Reagent (Boehringer Mannheim). Probes were labeled using a DIG RNA Labeling kit (Boehringer Mannheim). A Zic3 probe spanning 471 bp in the 3′ region of the gene was generated using PCR. The primers used for PCR were F 5′-TCTAGATTCCTTACAATGTCAGT-3′ and R 5′-AAGAAGCACTTTAACCATGAG-3′. Mouse embryos were processed to visualize β-galactosidase (β-gal) activity by X-gal staining (Hogan et al., 1994).

Embryos or neonates were harvested and staining was performed as described previously (Kochhar, 1973) using 0.015% Alcian Blue 8GX followed by 1%KOH/0.015% Alizarin Red. The skeletons were cleared in 1% KOH/20% glycerol for several days and stored in a 1:1 mixture of glycerol and 95% ethanol.

To examine the role of Zic3 in mouse development, the Zic3-coding region was deleted and replaced with a lacZ reporter gene in mouse ES cells (Fig. 1A,B). Because Zic3 is located on the X chromosome, deletion in male ES cells should generate Zic3 null cell lines. Interestingly, all of the resulting male chimeras generated with these Zic3 null ES cell lines had tail kinks (data not shown), suggesting a role in axial development (see below).

To document that embryonic Zic3 expression was ablated by the deletion of the Zic3-coding region, whole-mount in situ hybridization was performed on wild type and –/Y 9.0 days post coitum (dpc) embryos. Zic3 expression was readily detected in the CNS, tailbud and somites of the wild-type embryo but not in the –/Y mutant embryo, demonstrating that the targeted mutation had eliminated Zic3 expression (Fig. 1C, D).

Female heterozygotes derived from crosses of male chimeras with B6 mice were bred with either B6 or outbred Swiss males, and their progeny (n=217) were genotyped at 3 weeks of age (Table 1). Departure from the expected mutant to wild-type ratio indicated significant lethality in the Zic3 null males that is more dramatic on the B6/129 mixed genetic background (0.26 viability versus 0.46). Furthermore, no live-born –/Y males were recovered from crosses on the 129/SvEv inbred background (data not shown), indicating that the penetrance of the Zic3 null allele is sensitive to genetic background differences.

All viable –/Y males, regardless of genetic background, and ∼40% (17/44) of the +/– females had ‘kinked’ tails (Fig. 1E). This malformation was usually less severe in the females than in the males. These observations suggest that a lack of Zic3 causes tail abnormalities. Indeed, Zic3 has recently been found to reside within a microdeletion on the X chromosome in the classic mouse mutation bent tail (Bn) (Carrel et al., 2000; Klootwijk et al., 2000). The variability of the tail phenotype in the heterozygous females is likely to be due to X chromosome inactivation. In support of this idea, X-gal stained Zic3 heterozygous female embryos show mosaic β-gal activity in the CNS (Fig. 1F), suggesting that the lacZ knock-in allele can serve as a reporter of X chromosome activation status.

Zic3 null males were mated to heterozygous females to evaluate prenatal lethality as well as the phenotype of null females. Table 1 shows results of these crosses at weaning age (n=155) and at specific gestational stages. Ratios of genotypes as compared with wild type are shown below each line of raw data and indicate a skewing from the expected 1:1:1:1 ratio. The ratio of null males compared with wild type was 0.20, indicating a lethality rate that corresponds well to data from the heterozygous by wild-type crosses. In both sets of crosses on the B6/129 background, there was a slight decrease in the viability of heterozygotes (ratio 0.77 and 0.81). Analyses of embryos at various gestational ages failed to show a discrete time point at which lethality occurs, suggesting that the embryonic losses may occur over an extended period of time. Comparison of the ratios of null males and females at birth and at weaning age demonstrate that significant lethality also occurs after birth, with approximately 30% of null males lost in the perinatal period. In addition, the null females appear to be less affected in utero but have a larger percentage (50%) that die after birth.

To investigate the timing and cause of prenatal lethality in Zic3 deleted mice, embryos were harvested, genotyped and examined at varying gestational stages. As indicated in Table 1, there was not a definitive time point of lethality. Rather, the embryos appear to die at different gestational stages with a variety of defects. In the 65 Zic3 null embryos examined, several laterality defects, including failure and delay of embryonic turning as well as abnormal direction of turning (Fig. 2A,B) were seen. Several null embryos that survived beyond the turning stage exhibited significant rotation defects manifest by abnormal thoracic and lumbar flexion and/or abnormal looping of the distal tail including the hindlimb buds (Fig. 2E-H; compare with wild type in Fig. 2C). Other laterality defects included abnormalities of heart looping and position. At 10.5 dpc, the heart is located at the midline. However, in several null embryos, the heart was positioned to the left side of the thorax (Fig. 2D,F). This was not simply a result of abnormal embryonic turning, because abnormally turned mice with midline hearts were seen as well (Fig. 2G). This aberrant positioning is similar to that noted previously in sonic hedgehog (Shh) deficient mice (Tsukui et al., 1999). In contrast to the Shh^–/– mice, however, aberrant right lateralization of the heart was also found in a smaller number (n=2) of Zic3 null embryos.

Zic3 null embryos also demonstrated CNS defects. Neural tube closure was delayed in some Zic3 null embryos at 8.5 dpc (compare neural fold apposition in Fig. 2A with 2B). In a subset of embryos, the headfolds were not well formed (data not shown). By 9.5 dpc, neural tube defects were noted with failure of fusion present in the midbrain-hindbrain region. Several embryos with neural tube defects appeared to be dead at 10.5 and 11.5 dpc, based on lack of a heart beat (Fig. 2I). In addition, abnormalities in the size and structure of the midbrain and forebrain, including the region overlying the telencephalic vesicles (Fig. 2F), were seen. Many Zic3 null embryos without obvious gross anatomical abnormalities nevertheless showed evidence of growth delay when compared with their wild-type littermates (data not shown). These delayed embryos were found at all time points analyzed, demonstrating that embryos had a broad range of developmental defects.

Seventy-six null and heterozygous offspring were analyzed at 17.5-18.5 dpc to investigate the causes of perinatal lethality. Internal malformations were identified at necropsy in 7 (24%) Zic3 null (either –/Y or –/–) fetuses and in one +/– female. In addition, newborns that were either runted, in obvious distress or stillborn were analyzed. Internal malformations were identified in 25, including six female heterozygotes.

Table 2 summarizes the internal malformations, which include defects of the LR axis, CNS and axial skeleton. No differences were detected between null mice and heterozygous females, aside from the frequency of malformations. Of the 20 mice with LR malformations, only one, a heterozygous female, had situs inversus. Seventeen had heart defects, including dextro-transposition of the great arteries (most commonly), interrupted aortic arch, right aortic arch, atrial septal defect, ventricular septal defect and abnormal systemic venous connections (Fig. 3A-C). Heart anomalies in the null and heterozygous Zic3 mice were always found in combination with other anomalies and were never found in isolation. Abnormal lung morphology included complete reversal (n=3), bilateral left-sided (both lungs unilobate, n=2) and bilateral right-sided (both lungs with four lobes, n=9) (Fig. 3D). A right-sided stomach with reversed liver lobation was seen in six mice. A spleen was identified in all affected mutants, but it was often hypoplastic (Fig. 3E,F).

Five pups obtained by Caesarian section were exencephalic, and two newborns found dead had anencephaly (Fig. 3G). One pup had a completely open spine from the cervical to the lumbar region, one had meningocoele and one had an open neural tube defect in the lumbosacral spine. All of the exencephalic brains had very similar histological abnormalities: disorganized cerebral cortex, anteriorly displaced diencephalon, disorganized basal ganglia, thalamus and hippocampus, rudimentary olfactory nerves, and absent cerebellum (Fig. 3H,I). None of these abnormalities were identified in any Zic3 null or +/– pups with an intact skull.

Zic3 expression is detected in differentiating somites (Fig. 1C). By 10.5 dpc, Zic3 expression is restricted to the dorsal dermamyotome and subepidermal dorsal mesenchyme (Nagai et al., 1997). Twelve mice (10 null and two heterozygous) had malformations of the axial skeleton. Almost all of the anomalies were unilateral and appeared to be posterior transformations along the anteroposterior axis (Fig. 3J-N). Vertebral abnormalities included partial duplications, partial fusions and apparent homeotic transformations. Partial rib duplications and inappropriately posterior sites of rib insertion were also identified. The sidedness of the malformations appeared to be random (five right, seven left) and there was no correlation with the sidedness of LR malformations when they were present. The only bilaterally symmetric abnormalities were the absence of one pair of ribs in two pups and partial atlanto-axial (C1-C2) fusions in the midline in four pups; these six pups also had additional, unilateral malformations of the axial skeleton.

In recent years, a genetic hierarchy has been defined that regulates the formation of the LR axis (reviewed by Burdine and Schier, 2000; Capdevila et al., 2000; Casey and Hackett, 2000). Zic3 expression has been examined in embryos ≤8.0 dpc by in situ hybridization of histological sections, but no asymmetrical expression was reported (Nagai et al., 1997). We obtained similar results, with symmetric expression observed in all of the embryos studied from 6.5 dpc to 10.5 dpc (data not shown). Specifically, transcripts were detected symmetrically in (but not limited to) the primitive streak and node at all stages in which these structures were present.

To examine the relationship of Zic3 with other genes involved in LR axis specification, whole-mount in situ hybridization was performed with probes directed against Pitx2 and nodal. Pitx2 is expressed asymmetrically in the left lateral plate mesoderm beginning at the two to four somite stage and persists on the left side of the developing heart and gut through 16 dpc. Pitx2 is currently one of the most downstream regulatory genes expressed asymmetrically during mouse development, and its absence results in LR axis malformations (Kitamura et al., 1999; Lin et al., 1999; Lu et al., 1999). Two of 15 Zic3-null embryos from four to 14 somites of development had abnormal Pitx2 expression: one with diffuse left-sided expression and the other embryo with diffuse bilateral expression (head mesenchyme expression was normal in both) (Fig. 4F,G). All of the heterozygous and wild-type littermates had normal, left-sided expression (Fig. 4E).

Nodal is the earliest gene that is expressed asymmetrically during mouse development (Lowe et al., 1996). Nodal expression was studied using in situ hybridization in Zic3 wild-type, heterozygous and null embryos ranging from presomite to eight somite stages (Fig. 4 and Table 3). In wild-type embryos, nodal expression is observed in the left LPM from the two- to eight-somite stage. Among Zic3-null embryos, nodal expression appeared randomized in the sidedness of its LPM expression: three embryos expressed nodal on the left, three on the right and one bilaterally.

Nodal expression normally begins at the node earlier than in the LPM and persists longer (Collignon et al., 1996; Lowe et al., 1996). Unlike LPM expression, which often is not detected among wild-type embryos (presumably because of its brevity), expression at the node is always observed in embryos up to the six-somite stage. Correct nodal expression at the node appeared in Zic3 heterozygotes and Zic3-null embryos up to the two-somite stage, but seven out of 16 heterozygote or null embryos from two to six somites lacked expression at the node, whereas all 14 wild-type embryos examined at this stage had correct nodal expression at the node (Fig. 4 and Table 3). At the six- to eight-somite stage, no nodal expression was detected in 6/7 heterozygote or null embryos compared with 1/3 wild-type embryos. Among the seven embryos with two to six somites with absent expression at the node, LPM expression was observed on the right side in three, on the left side in two, bilaterally in one and absent in one. Right-sided or bilateral LPM expression was always accompanied by absent expression at the node.

Laterality defects are a common (one in 8500 live births) and important cause of congenital malformations (Ferencz et al., 1997). Although several genes are known to affect laterality in various model organisms, the molecular basis for the majority of human cases is not known and generalizations across species have been difficult in some cases, owing to lack of conserved gene function. Our results indicate that Zic3 null mice correctly model the defects seen in individuals with HTX1, particularly with regard to situs abnormalities and congenital heart defects, and will be an important model for investigating the mechanistic underpinnings of these disorders.

In the original HTX1 study, no phenotypically normal male offspring of carrier females were described (15 in total) (Gebbia et al., 1997). A later case report has documented a male with a mutation in ZIC3 with no obvious malformations, although other male family members had characteristic HTX1 (Megarbane et al., 2000). In our mouse model, 80% of the Zic3 null mice have phenotypic abnormalities that manifest either in the embryonic or perinatal period, with the remaining 20% apparently without obvious malformation with the exception of a ‘kinked’ tail. There is a subset of individuals with HTX1 in which the carriers show situs inversus. In the Zic3 heterozygous females, laterality defects were detected in six out of 55 animals examined and included one case of situs inversus. These results indicate that both complete and partial deficiency of Zic3 can cause disturbances of laterality. Genetic background appears to influence the penetrance of the phenotype with increased lethality noted on an inbred 129 background. The fact that some null animals survive to adulthood suggests that Zic3 is not absolutely essential during development and implies a permissive role. Our finding that more Zic3 null males died in utero, whereas more Zic3 null females died in the perinatal period suggests a sex-limited effect that will require additional investigation. The majority of perinatal deaths were attributable to CHD, suggesting that the null females may be more susceptible to this particular defect. Our results, together with those of others who have investigated laterality abnormalities in offspring of nonobese diabetic (NOD) mice (Maeyama et al., 2001) and in cryptic/ EGF-CFC knockout mice (Bamford et al., 2000), indicate the role for environmental and/or genetic modifying loci in contributing to the laterality phenotype both in mice and in humans. Because the Zic3 null mice are comparable with individuals with HTX1 in terms of the degree of affected offspring and carriers, they will be useful to begin investigations of contributing modifier loci.

Bent tail (Bn) is a spontaneous mutation of the X chromosome that has been used as a mouse model of neural tube defects (Garber, 1952). Mapping of the Bn crucial region demonstrated the mutation is associated with a submicroscopic deletion on the proximal part of the X chromosome, a region that includes the Zic3 locus (Carrel et al., 2000; Klootwijk et al., 2000). Short ‘kinked’ tails characterize the mice. Abnormalities of thoracic situs and neural tube defects, including exencephaly, were noted in these animals. In addition, one report noted cleft lip, orofacial schisis and gastroschisis in a subset of Bn mice. The original description of the Bn mice described several with forelimb abnormalities. We have not found gastroschisis, orofacial clefting or forelimb abnormalities in any of the Zic3 null or heterozygous mice examined to date. We cannot formally rule out the possibility that mice born with these defects died in the perinatal period and missed analysis. More likely possibilities include that the Bn deletion is a contiguous gene syndrome and that other genes within the submicroscopic deletion are responsible for these features. Alternatively, there may be a phenotypic variation in expression that is strain dependent and/or influenced by other loci. We conclude that deletion of Zic3 contributes to the situs abnormalities and neural tube defects in the Bn tail mice, whereas the other noted anomalies in these mice appear to be unrelated to Zic3 deficiency.

In Zic3 null and heterozygous mice, the situs abnormalities are similar to those found in humans. The non-LR malformations seen in humans (e.g. renal and anal) generally differ from those identified in mouse (exencephaly and skeletal) although olfactory nerve and cerebellar abnormalities have been observed in both mouse and human. Early pregnancy loss (prior to organogenesis in the mice and within the first trimester in humans) occurs in both humans (Casey et al., 1993) and Zic3 null mice. Further analysis of Zic3 null mice at early embryonic stages will be required to identify the cause(s) of this loss.

Murine lung anatomy is a readily identifiable marker of left-right asymmetry because the right lung normally has four lobes and the left lung has one lobe. Knockout mice have been described in which left pulmonary isomerism (one lobe bilaterally) or right pulmonary isomerism (four lobes bilaterally) is seen. Knockouts that show left pulmonary isomerism include Shh and Lefty1 (Chiang et al., 1996; Meno et al., 1998). These genes are both hypothesized to play a role at the midline, restricting Nodal, Lefty2 and Pitx2 expression to the left LPM, thereby reinforcing the left identity. Loss of these genes results in bilateral expression of the LPM signaling cascade and bilateral left sidedness. Activin receptor IIB (Oh and Li, 1997), cryptic/EGF-CFC (Yan et al., 1999) and Gdf1 (Rankin et al., 2000) knockouts, as well as Fgf8 hypomorphs (Meyers and Martin, 1999), all demonstrate right pulmonary isomerism. These genes are hypothesized to act prior to or within the asymmetric cascade of gene expression within the left LPM. Loss of gene expression leads to absence of the left sided cascade and a right-sided identity develops, presumably as a default pathway. In iv/iv or lrd mice as well as inv mice, pulmonary reversal is the predominant phenotypic lung finding. Lung anatomy is therefore not only an indicator of asymmetry, but is also related to the site of action of a gene within the hierarchy of LR axis generation. The finding of both right and left pulmonary isomerism as well as pulmonary reversal in null Zic3 mice is therefore mechanistically important. The pulmonary reversal indicates Zic3 can exert an effect at the level of axis specification and/or orientation. By contrast, the finding of pulmonary isomerism indicates a role in the propagation of the lateralized signaling cascade. To our knowledge, this is the first gene involved in LR asymmetry in which both left and right pulmonary isomerism is seen.

In addition to laterality defects, Zic3 mice also show defects of axial patterning and neural tube closure. In early Zic3 deficient embryos, failure to turn or abnormal turning is noted, suggesting defective midline patterning. Several mouse models that produce both neural defects and a disruption of axial patterning, including no turning (Melloy et al., 1998) and Sil (Izraeli et al., 1999) mutants, as well as Kif3a (Marszalek et al., 1999; Takeda et al., 1999), Kif3b (Nonaka et al., 1998) and Hnf3b (Ang and Rossant, 1994; Weinstein et al., 1994) knockout mice, show a concomitant disruption of LR axis formation, suggesting that LR asymmetry and axial patterning can be linked. The axial patterning defects in Zic3 mice are further emphasized by the skeletal abnormalities, which occur in ribs and vertebrae, both derivatives of the dorsal sclerotome. Deficiency of Zic3 also results in a bent tail, presumably caused, as in the Bn mice, by asymmetric growth of caudal vertebrae (Klootwijk et al., 2000). Therefore, lack of Zic3 appears to perturb both anteroposterior and LR patterning during embryonic development.

Previous studies in Zic1-deficient mice and Zic2 knockdown mice have demonstrated skeletal and CNS anomalies. The skeletal findings in Zic1 null mice are very similar to those noted here, whereas the Zic2 mice have limb anomalies in addition to vertebral anomalies. Humans with ZIC2 mutations have holoprosencephaly, a defect also seen in the knockdown mice together with exencephaly, anencephaly and spina bifida (Brown et al., 1998; Nagai et al., 2000). Zic1 null mice demonstrated cerebellar dysgenesis (Aruga et al., 1998). Several studies have shown significant Zic-Gli interaction in neural and skeletal patterning, and Zic and Gli proteins have been shown to physically and functionally interact through their zinc-finger domains (Koyabu et al., 2001). This interaction may be synergistic, as in the double mutation of Zic1 and Gli3 which results in increased disturbance of segmentation of the vertebral lamina (Aruga et al., 1999), or it may be inhibitory, as in the case of the Zic2-Gli2 interaction in the patterning of the neural tube along the dorsoventral axis in Xenopus (Brewster et al., 1998). The nature of Zic-Gli interactions are only beginning to be characterized and it will be important to consider them in a developmental stage specific and cell type specific context. It is currently unclear to what degree other Zic family members can compensate for Zic3 deficiency although there is precedent from the Gli2 and Gli3 knockout mice for both specific individual functions as well as redundant functions in skeletal patterning.

The generation of LR asymmetry requires LR axis specification, orientation with respect to the dorsoventral and anteroposterior axes, and propagation of asymmetric LR signals. Our results indicate that Zic3 functions at the node and in the midline and participates both in LR axis specification and maintenance of asymmetric signaling in the left LPM. Zic3-deficient embryos fail to maintain Nodal expression after the two somite stage and exhibit randomized expression of Nodal and Pitx2 in the LPM, placing Zic3 upstream of these genes within the left-right specification pathway. In contrast to our results, Fgf8 neo/- hypomorphic embryos fail to express Nodal at the node at the zero- to two-somite stage, suggesting a defect in the initiation of nodal expression (Meyers and Martin, 1999). Taken together, these results appear to place Zic3 between Fgf8 and LPM Nodal in the left-right pathway. In addition, Zic3 appears to be important in the specification of LR asymmetry based on the finding of situs inversus as well as pulmonary reversal in a subset of Zic3-deficient mice. The requirement for Zic3 in the midline is demonstrated by the axial patterning defects in the null mice as well as pulmonary isomerism. Further investigation of midline defects in these mice, including molecular characterization, will be required to determine whether this represents a primary process or is secondary to abnormalities within the node.

Gli proteins are important mediators of Shh signals that are involved in LR, CNS and somite patterning. In vitro data have shown that Zic proteins are relatively weak transcriptional activators and may function instead as transcriptional coactivators with Gli proteins (Mizugishi et al., 2001). Based on the known functional interactions between Gli and Zic proteins (Koyabu et al., 2001), the function of Zic3 may be modulated by Shh or via interactions with combinations of Gli-family transcription factors downstream of Shh signaling. Furthermore, Gli2 is induced by FGF signaling in Xenopus. It is therefore tempting to speculate that Zic3 acts in concert with Gli-family transcription factors to establish a midline developmental field required for enforcement of left and right identities via its interaction with FGF8 (responsible for left identity) and Shh (responsible for restricting left). Thus, Zic3 would be permissive but not sufficient for LR axis determination, and deficiency could result in left isomerism, right isomerism or reversal, based on the balance of competing left and right signals.

In summary, Zic3 is important in the development of left-right asymmetry and may function at multiple levels within this pathway, acting both at the node, in order to maintain Nodal expression, and in the midline. Its deficiency results in prenatal lethality in approximately 50% of mice and defects in laterality, axial patterning and neural tube formation are seen, suggestive of an important role of Zic3 in early mesodermal patterning. The phenotypes seen in Zic3 null mice are similar to those found in human HTX1 and these mice will be useful in studies to delineate the molecular mechanisms underlying the variable effects of mutation at this locus.

We thank Allan Bradley for the mouse ES and STO cell lines. This work was supported in part by grants from the March of Dimes (1-FY98-0310 to B. C.), the American Heart Association (#96015660 to B. C.) and the NIH (K08 HD01078 to B. C., K08 HL67355 to S. M. W. and PO1 HL67155 to J. W. B.). Veterinary resources at the M. D. Anderson Cancer Center were supported by the Cancer Center Support (Core) Grant (CA16672).