The ZIC gene family encodes multi-functional proteins essential for patterning and morphogenesis

Molecular function of the ZIC proteins

The ZIC proteins are thought to regulate the transcriptional activity of target genes. As recently reviewed [14], there is now good evidence that the ZIC proteins can influence gene expression, not only as classical transcription factors through direct DNA binding via their zinc finger domains but also as co-factors by binding to other transcription factors. Recently, we and others established that the ZIC proteins can repress canonical WNT signaling target genes by binding directly to the members of the TCF (transcription factor) protein family through its zinc finger domain. This ZIC-mediated inhibition of WNT signaling occurs without the ZIC proteins contacting DNA [10, 22]. These recent studies augment previous observations that suggest that the ZIC proteins can influence the transcriptional outcome of Hedgehog signaling without contacting DNA by binding to the GLI proteins [23].

Genomic arrangement and evolution of the vertebrate ZIC genes

The sequencing, assembly, and annotation of multiple genomes over the past decade has revealed that the Zic genes have a relatively unusual arrangement within the genomes of all vertebrates examined. Figure 2 shows the arrangement of the Zic genes (as shown in the Ensembl database) in the teleost lineage (as represented by zebrafish) as well as in amphibia (represented by Xenopus) and in mammals (represented by human and mouse). It can be seen that in each case at least some of the genes exist as tandem copies, divergently transcribed from an intervening bidirectional promoter. There are few examples of these types of paralog gene clusters being retained over large evolutionary distances but notably three other well-characterized examples (the Hox, Dlx, and Fox genes) all encode transcription factors critical for embryonic patterning and morphogenesis. In the case of the Zic genes, each of the retained pairs consists of one member from each structural subclass. Based on phylogenetic analysis of Zic gene sequences and intron/exon position, Aruga et al. [6] have proposed a hypothetical model of Zic gene evolution, shown in Fig. 3. In this model, the ancestral single-copy Zic gene contained one intron and a full set of conserved structural domains. Following tandem duplication of this gene, some variation occurred such that one gene copy acquired an additional intron and the other gene copy underwent sequence divergence with the breakdown of the ZOC motif and increased divergence of zinc finger 1. Subsequently, the genome duplication events proposed to have taken place early in the vertebrate lineage, presumably led to the generation of eight Zic genes arranged in four bigene clusters with each cluster containing a Zic gene from structural subclass A and B. Gene loss would largely account for the Zic gene copy number now seen in each particular vertebrate. As can be seen in Fig. 2, the zebrafish has seven Zic genes while other characterized vertebrates have five. This discrepancy is accounted for by two factors (1) the retention in the teleost lineage of the Zic3 bigene pair (designated zic6) and (2) independent duplication of the Zic2 gene leading to two copies of this gene, one paired with Zic5 as in other vertebrates (and designated zic2a in zebrafish) and the other an unlinked copy (designated zic2b) [24]. The zebrafish zic6 is distinguished by the lack of the ZF-NC domain [24], the only of the vertebrate genes depicted in Fig. 2 to lack this domain. In addition to the differences in the teleost lineage, Fig. 2 demonstrates that in both mammalian genomes examined here the ZIC4 gene has become less compact with the addition of introns between the coding and non-coding exons and a novel intron that divides the sequence corresponding to the first coding exon in each of the other depicted genes. Examination of other mammalian genomes suggests that this pattern is not unique to mice and man but may be more widespread in the mammalian lineage.

The different features of the Zic gene arrangement in the zebrafish genome provide additional opportunities for functional divergence and there is some evidence that the function of individual Zic genes is different in zebrafish and mammals. This is exemplified by the finding that in zebrafish zic1 loss-of-function leads to the distinctive phenotype of cyclopia (the formation of a single, ventrally displaced eye) [25], whereas in mouse and man this phenotype is associated with loss of ZIC2 function [16, 26]. It has long been recognized that gene family expansion accompanied by the unusual arrangement of the Zic genes generates particular difficulties when studying their function. The genes are likely to exhibit broad overlap in gene expression and are likely to recognize the same DNA-binding motifs; they may also share protein partners and could participate in cross regulation of family members at both the gene expression and protein interaction levels.

Human and model organism genetic studies have repeatedly demonstrated that the Zic genes are required for embryonic development (as reviewed in [27–29]) and in some cases we have now been able to identify the embryonic stage and tissue in which Zic gene function is required in order to prevent the respective associated congenital defect. More study is required, however, to determine the precise molecular mechanisms of Zic gene function and regulation. Additionally, it is likely that both germ-line and somatic mutation of the Zic genes constitute as-yet-unrecognized risk factors for other diseases and disorders. As a basis for furthering our understanding of how the Zic genes contribute to human embryonic development and adult homeostasis we now review data from gene expression studies and from experiments that remove or increase Zic gene expression. We focus on experiments conducted in the two model organisms (mouse and Xenopus laevis) for which current genome annotations indicate a largely conserved gene arrangement with the human genome and for which considerable knowledge has now been amassed regarding Zic gene expression and the consequence of altered Zic gene dosage.

Spatial–temporal expression of the vertebrate ZIC genes

Xenopus zic gene expression

The Xenopus zic gene family is characterized by gene-specific but also partially overlapping expression patterns over time and is mostly limited to progenitor tissue of the future central nervous system (CNS). Prior to fertilization, zic2 transcripts have been be detected in the oocyte. Zygotic expression of all zic family members (except zic4) can be found at the onset of gastrulation and in prospective neural tissues. Very weak zic4 expression was detected at the prospective neural fold region of the stage 11.5 mid-gastrula [40]. Expression of all zic family members continued at the neural plate border during early neurulation, neural crest formation, and at the dorsal central nervous system at later neurula stages. The spatio-temporal and partially overlapping expression patterns of this family make this model organism complex but simultaneously attractive to study.

In a functional screen using a maternal cDNA library from Xenopus laevis, Houston and Wylie [41] discovered high expression levels of maternal zic2. This supported the previous findings of Nakata et al. [42] who had screened for the primary expression patterns of all zic family members. They also detected zic5 transcripts in the same cDNA extract, although this was not confirmed by Houston and Wylie.

All zic family members (except for zic4) are expressed at mid-blastula stages. During gastrulation, zic expression becomes limited to the prospective neural plate. Unlike for murine Zic expression patterns at this stage, most of the Xenopus zic genes are expressed in overlapping patterns, so only subtle spatial–temporal differences in expression can be observed. In general, zic1 and zic3 expression is more widespread compared with zic2 and zic5. The uncoupled zic3 is uniquely expressed in the more caudal lateral plate mesoderm. Zic2 and zic3 are the only members that are expressed along the complete AP axis while zic1 and zic4 expression is found in the future hyoid and branchial arches. The selective expression of zic5 is found mostly at the anterior neuroectodermal ridge [40, 42–44].

Nevertheless, all zic genes are reportedly expressed at the neural fold where they play an essential role during neural patterning [42–44]. To date, Xenopus laevis is the only vertebrate model where the expression of all zic family members at the neural plate border and at the dorsal side of the neural tube after closure is linked to neural crest cell production. Alterations in zic expression result in clear phenotypic differences in terms of the production and migration of neural crest cells. At later neurula stages, expression of all zic members continues in the dorsal CNS while their expression patterns elsewhere become further restricted. In the somites, expression patterns are mostly restricted to zic1 and zic4 transcripts although zic2 is expressed to a lesser extent and is more ventrally located. During the early tailbud stages zic1-3 are all dorsally expressed in the forebrain, midbrain, and hindbrain and are found at later stages in the telencephalon and di-/mesensephalon border, in a largely overlapping pattern. Zic5 transcripts are only detected in the midbrain and hindbrain. Clear differences are observed at the eye vesicle where zic2 and zic5 expression are mostly observed, further highlighting the coupled expression of these two head-to-head placed genes. The expression of zic2 and zic5 is limited here to the ciliary marginal zone of the retina. In contrast, zic2 expression is observed during somitogenesis and accompanied by zic1 and zic4 but not by zic5 expression. As organogenesis progresses, these expression patterns are more restricted and remain present until later stages [3, 40].

Murine Zic genes

Mutational analysis of the murine Zic genes has played an important role revealing the function of these proteins in mammalian biology and there is now an extensive catalogue of processes, tissues, and organs known to be affected by loss of normal activity of one or more Zic proteins. The majority of available mouse alleles carry severe loss-of-function mutations and are primarily useful for studying the earliest function of each of the Zic proteins. The analysis of these alleles shows that Zic gene expression is required for the development of the embryonic axes, for many processes during the development of the CNS and for the development of the visual and musculo-skeletal systems. Importantly, where tested, the murine gene mutations give rise to phenotypes associated with mutation of the orthologous human ZIC gene, suggesting that the mouse mutants serve as good models for investigating the etiology of ZIC-associated human disorders (summarized in Table 1).

Xenopus zic genes

Most of the functional studies on the ZIC gene family have been carried out in Xenopus laevis on the orthologous zic genes and zic proteins. The complete removal of gene function is difficult in Xenopus; therefore a combination of gene knockdown and overexpression studies has been used to quickly identify phenotypic defects and transcriptional alterations. These experiments have both identified novel roles for the zic gene family and confirmed functions found in other model organisms. Since many of these experimental approaches differed between independent laboratories, the interpretation and comparison of the data has proven difficult, therefore the resultant phenotypes are mostly discussed separately in this review (Table 2).

Since zic2 is maternally expressed in Xenopus laevis, it is difficult to use morpholino-based approaches to identify its function in vivo. Therefore, the host transfer method was used to eliminate maternal zic2 expression by injecting anti-sense oligonucleotides into unfertilized oocytes that were planted back into a mother animal. This method led to phenotypes including exogastrulation, anterior truncations, and axial defects [10, 41]. Although initiation of gastrulation occurred normally, a delay in blastopore closure was evident, in 30 % of the exogastrulae. The other surviving embryos exhibited deep involution of the marginal zone, reduced head size, wrinkled epidermis, and a stunted dorsal axis. Despite the fact that these abnormalities can be due to secondary effects, clear forebrain defects were evident. In addition, neural tube formation and somitogenesis were clearly disrupted. The notochord seemed to be thickened and in some cases duplicated. Analysis of nodal target gene expression revealed a general increase in nodal signaling in these embryos. Injection of the nodal antagonist cerberus short in part rescued the phenotype, indicating indeed that Nodal signaling was disturbed but was not the only signaling pathway affected. This was further confirmed by our group and others when zic2 and other zic proteins were able to repress canonical Wnt signaling [10, 22]. This finding was partially based on injection of zic3 mRNA that was found to alter the expression of organizer markers including goosecoid and siamois, which are direct Wnt target genes, and impaired notochord development. Due to the possible non-specific nature of this overexpression phenotype, a loss-of-function approach was also used. The injection of zic3 morpholinos alone did not generate obvious abnormalities but co-injection with zic2 morpholinos revealed a clear phenotype. In accordance with the maternal zic2 knockdown phenotype, the notochord was thickened and widened, A-P axis was shortened and the size of head structures were reduced [22]. Although maternal zic2 proteins cannot be directly targeted, targeting of zygotic expression by morpholino technology revealed clear phenotypes attributable to altered zygotic zic2 and zic3 expression. This again demonstrated the functional redundancy in this gene family. Other early functions associated with zic3 expression in Xenopus include regulation of left–right (L–R) patterning and convergent extension (C–E) morphogenesis. Similarly, and as previously mentioned, in humans, ZIC3 mutations result in X-linked heterotaxy, and Zic3 null mice exhibit defects in gastrulation, neural tube closure, and axial patterning [45, 48]. Xenopus zic3 morphants revealed significant impairment of C–E and in L–R patterning. The A–P axis was shortened, gastrulation was delayed, and abnormal heart and gut looping [47]. Collectively, these data support an evolutionarily conserved role of Zic3 gene function during early vertebrate embryonic development.

At the prospective neural plate region, zic3 was found to be ectopically expressed upon BMP4 signaling blockade. As for other neural inducers, it was postulated that repression of zic3 expression was relieved when neural inducing signals altered BMP4 signaling [3]. Similarly, in addition to zic3, zic1 and zic2 were shown to be downstream targets of BMP4 signaling. The expression of both genes was upregulated in the neuroectoderm and both were potent inducers of neuroectodermal differentiation, though the potency of induction was less with zic1 compared with zic2 expression [43, 76–78].

Furthermore, zic2 was shown to be expressed in stripes that alternate with those in which primary neurons differentiate in the neuroectoderm, and overlap the domains of floor plate and neural crest progenitors (stage 14). Zic2 overexpression leads to induced neural crest cell differentiation and inhibition of neurogenesis, elevating zic2 to the rank of vertebrate pre-pattern gene. The same expression pattern was found for zic5 but this was not further investigated [42, 79]. More recent studies indicate, in addition to BMP signaling, Ca²⁺ accumulation can also regulate zic expression in Xenopus and that BMP signaling blockade together with fibroblast growth factor (FGF) activation triggers zic gene activation to induce neuroectoderm formation [80].

While it has only been demonstrated for zic1 and zic3, it appears likely that zic2 expression is also under control of BMP and FGF signaling. However, mice harboring mutations in both Bmpr1a and Bmpr1b genes, which results in a clear MIH HPE phenotype, maintain Zic2 expression [81], suggesting that BMP is not required for Zic2 expression in this context. Bearing in mind that this signaling cascade is essential for neuroectodermal differentiation, it is however very probable that specific members of the ZIC gene family act through these pathways during normal neural development in these animal models [82–85]. The genetic regulation of Xenopus zic genes is also discussed in more detail in the light of vertebrate neuroectodermal differentiation [80].

Both knockdown and overexpression of zic3 resulted in clear neuroectodermal defects and lead to disturbed neurulation events. Although it is difficult to determine if the later defects are direct effects of altering zic3 expression, clear phenotypes arose after ectopic expression. Induced pigmented cells were accompanied by a general expansion of the head structures without mesodermal induction. This does not account for effects in the eyes, which were smaller compared with the wild-type embryos while markers of neural retina were upregulated. Essentially the same characteristics were found when zic1 and zic2 were overexpressed, namely a thickening of the neural tube and ectopically expressed pigmented cells. Marker-based analysis revealed that the epidermal cell fate was shifted to neural and neural crest cell fate without changing the expression of mesodermal markers [43].

Primary neurons and hindbrain cells were lost using a morpholino against zic1 expression and by injection of a dominant negative zic5 protein. Also, the induction of neural crest cells was disrupted. Similar defects were shown when XMeis3, a transcription factor that can induce hindbrain cell fates in a non-cell-autonomous manner, was knocked down, while the expression of all zic genes was maintained in this setting. This would suggest a role downstream of zic gene expression for XMeis3 function. This hypothesis was confirmed when the phenotype of the zic1 morphant was rescued following co-injection of Xmeis3 [86, 87]. Since Xmeis3 is shown to activate the hindbrain induction center that regulates early A–P cell fate in the developing CNS, its potential loss of activity can directly explain the phenotype in the zic1 and zic5 loss-of-function models.

Overexpression of zic5 in Xenopus led to a very strong neural crest cell-induced phenotype. Although zic5 is expressed only from the onset of neural crest cell induction, it induces ectopic melanin-forming cells very efficiently. Marker gene analysis shows that overexpression of zic5 also leads to the epidermal fate switched to neural and neural crest cell fate. A notable difference in the effects from overexpression of the different zic genes are to the anterior neural makers, which are unaffected in zic5 gain-of-function assays while they were disrupted in zic1-3 overexpression models. This would indicate that there are not only structural and expression-based differences, but that zic5 function appears at least in part to have been modified during evolution compared with the zic 1-3 genes [42].

Overexpression of zic4 results in an overlapping phenotype compared with the other zic gene family members. As for all zic genes, melanocyte formation was increased leading to an observed increase in neural crest cell formation. A general induction of neural tissue differentiation was evident, although no anterior neural markers (similar to zic5 overexpression) were altered. Hypoplastic eyes were observed with zic4 overexpression, which did not occur when other zic genes were overexpressed. Since zic4 expression is not observed in the Xenopus normal eye, this could be interpreted as an indirect effect. In general, the zic4 overexpression phenotype largely mimics that observed for zic5 overexpression leading again to the separation of the family into two groups that can be traced back to the chromosomal duplication where the zic4 and zic5 genes seem to have lost intrinsic characteristics that remain in the zic1, zic2, and zic3 genes [40].

Protein structure and synexpression subfamilies

Current investigations regarding Zic gene structure, expression, and function lead to an emerging picture in which gene duplication followed by loss of conserved protein domains and some alteration in expression control mechanisms has resulted in overlapping and unique functions of the orthologous and paralogous Zic family members. The most parsimonious chain of events would be that early in the vertebrate lineage, following an initial tandem gene duplication event, one copy of the tandem gene pair underwent structural changes that include loss of the ZOC motif and alteration in the amino acid composition of the first zinc finger. The subsequent genome duplications of the early vertebrate lineage therefore established two subclasses of Zic proteins. Here we define structural subclass A proteins as those that possess the ZOC motif and have retained sequence conservation across the first zinc finger (ZF1). Structural subclass B proteins are those that lack the ZOC motif and have a diverged amino acid composition of the residues immediately following the tryptophan residue of zinc finger 1. The ZOC motif is implicated in protein/protein interactions and it may therefore be anticipated that the structural subclass A Zic proteins (Zic1, 2, and 3) have some protein partners (and functions) in common but not shared by the structural subclass B proteins (Zic4 and 5). In contrast, the DNA binding ability of the Zic proteins is usually attributed to the more conserved zinc fingers (ZF3-5), suggesting that all five Zic proteins likely bind the same DNA sequences perhaps with varying affinity. Where investigated, this seems to be the case [11, 88].

This structural subclasses division persists amongst orthologous genes but there are some structural features of the Zic proteins that have undergone more rapid divergence such that differences can now be found between orthologs. This is exemplified by the location and composition of the low-complexity regions found outside of the ZOC, ZF-NC, and zinc finger domains, which can vary remarkably amongst orthologous genes (Fig. 5). Expansion mutations of the c-terminal alanine tract in ZIC2 have been associated with HPE in humans, a congenital defect known to be caused by ZIC2 loss-of-function, suggesting functional significance of this domain. An alanine repeat is found at a similar region in murine Zic2 but not in Xenopus zic2. The precise function of low-complexity regions in the Zic proteins and in general is not well characterized, but there is evidence that they are associated with protein-binding activity and that their position within a sequence may determine their binding properties and biological function [89]. Differences between the sequences of orthologs should be taken into consideration when interpreting data from apparently analogous experiments in different organisms.

The finding that the genomic arrangement of the Zic genes, primarily as bigenes, has been conserved over large evolutionary differences suggests it may have functional significance. In particular it is possible that the bigenes share regulatory regions. A corollary of this hypothesis is that the expression pattern of bigene members would be more similar to one another than to the other paralogs. It remains difficult to determine whether this is the case for practical and technical reasons. First, there are few biological systems where the expression of all five genes has been investigated in detail and, second the hybridization efficiency of the probes used to detect different Zic mRNA species via in situ hybridization varies. In particular, both Zic4 and Zic5 transcripts seem difficult to detect (i.e., they lead to low signal; see for example Zic5 signal in Fig. 4) using this technique. In order to complete surveys of Zic gene expression, it may be necessary to develop better probes or protocols for the in situ detection of Zic4 and Zic5 transcripts or to augment mRNA localization studies with additional techniques such as RT-PCR or to develop Zic isoform specific antibodies suitable for tissue hybridization studies.

Despite these caveats, the currently available data can be interrogated to assess the evidence for bigene co-expression. In Xenopus it appears that zic2 and zic5 are often co-expressed (in the oocyte, the prospective neural tissue of the gastrula and subsequently at the neural plate border and in the developing retina), but uncoupled expression can also be found. For example, in the neurula, zic2 expression is found along the axis, whereas zic5 expression is restricted to the anterior neural ridge and zic2, but not zic5, is detected in the somites. Similarly, there are many sites at which both zic1 and zic4 are expressed (the branchial arches and the somites) but also sites at which zic1 alone is detected (notably zic4 expression initiates later than all other Zic genes being first detected in the prospective neuroectoderm of the mid-gastrula). The uncoupled zic3 gene has elements in common with and unique from the bigene pair expression patterns. It seems possible that in Xenopus the bigenes do share regulatory elements but further studies are needed, for example to confirm whether some sites of zic4 and zic5 expression are currently undocumented.

In contrast to the situation in Xenopus, it seems clear from the murine expression studies that in the mouse the Zic genes can be divided into two synexpression groups with the Zic1/4 bigene forming a distinct group from that of the Zic2/5 bigene and the Zic3 gene. Notably, this division mirrors the genomic arrangement rather than the structural subclass division of the Zic genes (Fig. 6). In general, during mouse development, the regions where Zic1 and 4 are expressed are sites where all five Zic family members are transcribed. In contrast, there are several sites where only Zic2, 3, and 5 are expressed (such as during gastrulation, in the presomitic mesoderm and developing limb bud). The pattern of expression therefore seems different from that in Xenopus where, for example, at least one member of the zic1/4 bigene is expressed during gastrulation. This suggests that some Zic gene regulatory mechanisms have been significantly altered between Xenopus and the mouse, perhaps by the loss or gain of entire enhancers, rather than by the alteration of transcription binding sites within a conserved enhancer [25]. This raises the question of where the ZIC genes are expressed during human embryonic development. The strong correlation between loss-of-function phenotypes in mouse and man and the conserved genomic arrangement suggests that the expression patterns in human likely resemble those of the mouse.

Non-redundant and redundant function

In some cases, the inactivation of individual Zic genes can produce unique phenotypes (i.e., phenotypes not seen when other family members are alone mutated). Human and mouse genetics demonstrates that phenotypes in this category include ZIC2-associated classical HPE and ZIC3-associated heterotaxy indicating that the function of ZIC2 and ZIC3 in these processes is not fully redundant. This lack of redundancy may be caused a lack of co-expression. In this case, the unique phenotypes found upon individual Zic gene inactivation would reveal sites where only one Zic gene is expressed with the function of the missing member being fully compensated for (i.e., redundant) in areas of Zic gene co-expression. This idea is exemplified by the association between ZIC2 mutation and classical HPE, which has been examined by analysis of the Ku mouse mutant. In this case, the first overt molecular defect found in Zic2 ^Ku/Ku embryos is at the mid-gastrula node, a structure required for the development of the prechordal plate, the function of which is essential to prevent HPE [26]. This localized defect belies the widespread expression of Zic2 transcripts at pre- and mid-gastrula stages, which a priori may be expected to cause widespread molecular defects. Zic2 is, however, the only Zic gene expressed at the node with both Zic5 and/or Zic3 sharing other sites of Zic2 expression at these developmental stages ([32, 33] and Fig. 4). It remains to be tested whether ablation of Zic3 or Zic5 on the Ku mutant background would reveal the additional defects expected if functional redundancy, coupled with a site of Zic2-specific expression, has produced an unusually mild defect in the Zic2 mutant embryos.

In contrast to the HPE phenotype, the Zic gene murine mutational analysis suggests there are sites of full or partial Zic gene redundancy. Full functional redundancy is demonstrated when single mutants have no phenotype but the double mutant does, whereas partial redundancy is said to occur when one or both of the single mutants exhibits a phenotype that is enhanced in the double mutant. Care must be taken when performing and interpreting double mutant analyses since differences in genetic background and the use of hypomorphic alleles are both factors that can confound these experiments. Despite these caveats, there seems good evidence that the Zic genes do in some cases exhibit functional redundancy with one of the best characterized examples being the partial redundancy between Zic1 and Zic4 during cerebellum development [66, 67]. Here, null mutation of Zic1 is associated with small size and abnormal foliation of the cerebellum. Although animals null for Zic4 exhibit no detectable cerebellar phenotype the size and foliation defects are enhanced (relative to the Zic1 null) in the double mutant. Moreover, trans-heterozygous loss of both ZIC1 and ZIC4 leads to cerebellar size and foliation defects in both man and mouse. This type of non-allelic non-complementation can occur when the two genes involved are both redundant and dose-dependent [90]. This combined haploinsufficiency implies that the amalgamated gene products fall below a threshold level necessary to achieve the wild-type phenotype and is consistent with a situation in which the expression or activity of the paralogs depends upon each other’s expression. Here the null mutation in one or both of the genes causes decreased activity of the other leading to apparent haploinsufficiency.

The function of redundant family members can also be investigated by gain-of-function approaches, a strategy well suited to Xenopus embryos. All Xenopus zic family members are expressed at the neural plate border that forms the dorsal part of the neural tube and is the site of differentiation of the neural crest. The overexpression of each zic gene leads to the expansion of neural crest marker gene expression but not to any changes in expression of dorsal mesodermal markers. This highlights the ability of zic gene expression to induce anterior neural tissue and neural crest formation. In contrast with other zic genes, zic4 and zic5 were unable to induce the expression of forebrain and cement gland markers. It is interesting that this apparent functional difference mirrors the structural subclass divisions and suggests for example that the ability of zic1-3 to induce anterior markers may be dependent upon the ZOC motif.

The Zic proteins are multi-functional

Several features of Zic gene genetics suggest that they are multi-functional proteins. Firstly, some Zic genes are required for multiple steps of the development of a particular organ as exemplified by the role of Zic2 in dorsal–ventral forebrain patterning, by Zic1 and 4 in cerebellar development and by Zic2 and 3 in skeletal development. Although iterative function is a feature of many developmental proteins, the frequency of Zic gene re-use appears remarkable and is consistent with a multi-functional protein. Secondly, many transcription factors act in response to one particular signaling pathway, for example the Gli transcription factors are mediators of hedgehog signaling and the Lef/Tcf proteins are mediators of Wnt signals. In these cases, mutation of the transcription factors produces similar phenotypes to (or phenocopies) loss-of-function mutations in other pathway members such as receptors and ligands. One factor that has obscured the role of the Zic proteins is that they fail to phenocopy mutations in a particular signaling pathway and instead generally exhibit phenotypes reminiscent of a variety of signaling pathways. Additionally, Zic function is required for a myriad of biological processes and different signaling pathways are dominant in the respective biological processes. Zic gene genetics therefore suggest that the family members interact with many of the key embryonic signaling pathways and in doing so may serve to integrate and balance multiple signaling pathways. This idea is supported by biochemical experiments that demonstrate protein/protein interactions between Zic proteins and key transcriptional mediators of both the Shh and Wnt signaling pathways. These interactions appear to alter the way the pathway specific member can act, for example Zic proteins convert the Tcf transcription factors to repressors and Zic proteins can alter Gli3 repressor function [10, 23, 91].

As reviewed by Ali et al. [14] the molecular basis of Zic protein pleiotropy is beginning to be understood. The Zic genes have previously been shown to display phenotypes that result from the loss of DNA binding activity [16, 26], thus establishing that they can act as classical transcription factors. Consistent with this, the same consensus Zic DNA binding motif has now been identified by both a ChIP/Chip experiment and the UNIPROBE experiment [92, 93]. Biochemical experiments have shown that Zic proteins can also act as transcription co-factors to influence transcription of target genes without contacting DNA. In particular, they can physically interact with TCF transcription factors to inhibit Wnt signaling [10, 94] and they can also regulate Gli transcriptional activity independent of DNA binding [95]. In addition to the established ability of Zic proteins to act as transcription factors and co-factors, the expansion of the Zic gene family provides the opportunity for homo or hetero dimerization and/or dominant negative interactions between the Zic proteins and these possibilities are yet to be fully explored.

Promoting pluripotency: a unifying Zic function?

Despite our inability to link the Zic proteins with one particular signaling pathway or even one class of protein function, can a consistent biological function be found for the Zic proteins? The reason behind many of the described Zic-associated phenotypes has not been fully investigated. One function that has been ascribed to the Zic genes multiple times is that of promoting pluripotency. Zic mutant end-phenotypes are often associated with small size (of the affected organ or of the entire animal). Where investigated, the reduced size has been attributed to the premature differentiation of a progenitor population. This has the effect of ultimately decreasing the size of the organ since the progenitor differentiation occurs at the cost of increasing the size of the progenitor pool via self-renewal. Investigations in murine ES cells have also ascribed the task of promoting pluripotency to Zic3 [96, 97] and it may be that functional redundancy has obscured this role during early embryogenesis in murine Zic mutants. Although a proportion of Zic-associated phenotypes can be explained by a failure of self-renewal, not all can be; for example, in the case of the Zic2-associated neural crest cell defect it is the number of differentiated cells that is depleted, not the progenitor pool. Clearly, the Zic family members are involved in differentiation as well as the control of pluripotency. Further work is needed to determine whether many of the Zic-associated phenotypes are caused by altered fate, renewal, or other cellular properties.

Conclusions

We conclude that the Zic genes encode proteins that truly are partially redundant, multi-functional molecules that operate in a context-specific manner to drive a large number of critical events during embryogenesis. Their ancestral role in the generation of neuroectoderm and mesoderm-derived tissues seems to have been largely conserved but specific details have been altered through evolution. Our comparison here has revealed differences in protein structure, genomic arrangement, and gene expression patterns between Xenopus and mouse. These differences may explain variation in the outcome of apparently equivalent loss-of-function and gain-of-function experiments between these two organisms. They can also be used to inform ongoing investigations regarding the molecular mechanism of Zic function.