In humans, 19 WNT genes have been identified and the chromosomal locations of each is known (see Table 1) [1,2,3,4,5,6]. Several human WNT genes are located very close to each other in the genome [7,8]; these include WNT6 and WNT10a, which are located immediately adjacent to one another on chromosome 2 (about 6.4 kilobases (kb) apart), and WNT1 and WNT10b, which are located adjacent to each other on chromosome 12 (about 8.1 kb apart). WNT6 and WNT10a are transcribed in opposite directions, whereas WNT1 and WNT10b are expressed from the same strand of DNA. Several additional pairs of WNT genes are also clustered within the human genome, including WNT2 and WNT16 (about 4 megabases (Mb) apart), WNT3a and WNT14 (about 250 kb apart), and WNT3 and WNT15. In the mouse, there are at least 18 Wnt genes and the locations of all but two of them have been determined [1,2,3,5,6]. As in humans, the mouse Wnt1/Wnt10b, Wnt6/Wnt10a, and Wnt3/Wnt15 gene pairs are each located on the same chromosomes, and in the case of the Wnt1/Wnt10b and Wnt6/Wnt10a pairs the close proximity of these genes has been conserved from mouse to human. Interestingly, in the Drosophila genome, the paralogous genes wingless (wg), DWnt6 and DWnt10, are located immediately adjacent to one another on the second chromosome and are all transcribed in the same orientation. Thus, it is possible that there was an ancient cluster of Wnt genes consisting of Wnt1, Wnt6 and Wnt10 in a common ancestor of vertebrates and arthropods. In vertebrates, this cluster may have been duplicated with subsequent loss of Wnt1 from one cluster and Wnt6 from the other.

The majority of human WNT genes contain four coding exons, with exon 1 containing the initiation methionine (Figure 1a) [8]. WNT genes that differ from this pattern include WNT14, with three exons, WNT2, WNT5b, and WNT11, with five exons, and WNT8b with six exons. Several WNTs - WNT2b/13, WNT8a/d, and WNT16 - have alternative amino or carboxyl termini, which result from the use of alternative 5' or 3' exons.

The deduced evolutionary relationships of 18 of the 19 known human WNT genes are shown in Figure 2. The majority of Wnt proteins share about 35% amino-acid sequence identity, although members of a subgroup (those with the same numeral, such as WNT3 and WNT3a) share increased sequence identity (from 58% to 83%) and some overlapping sites of expression. Members of subgroups are not closely linked within the genome, however, suggesting that they were generated by gene-translocation or genome-duplication events, not by local duplication events.

Wnt genes have been identified in vertebrates and invertebrates, but appear to be absent from plants, unicellular eukaryotes such as Saccharomyces cerevisiae and from prokaryotes. To date, in vertebrates, 16 Wnt genes have been identified in Xenopus, 11 in chick, and 12 in zebrafish [5]; in invertebrates, Drosophila has seven Wnt genes, Caenorhabditis elegans five and Hydra at least one [5]. The apparent evolutionary relationships between selected invertebrate and vertebrate Wnt genes are shown in Figure 2b. In vertebrates, the orthologs in different species are highly similar in sequence. For example, human WNT1 and mouse Wnt1 are 98% identical, and human WNT3a and Xenopus Wnt5a are 84% identical at the amino-acid level. Phylogenetic analyses of vertebrate and invertebrate Wnts demonstrate orthologous relationships between several human and Drosophila Wnts (Figure 2b). The sequence identity between orthologous proteins in humans and flies ranges from 21% between human WNT8a/d and Drosophila DWnt8 to 42% sequence identity between human WNT1 and Drosophila Wingless (Wg). The evolutionary relationship between the five C. elegans Wnt genes and human WNT genes is less apparent, making it difficult to determine which C. elegans Wnt genes may have orthologs in the human genome.

Wnt proteins have an amino-terminal signal sequence, can act in a cell non-autonomous manner, and are present in the secretory pathway, indicating that they are secreted proteins [11]. In addition, genetic analyses of Wg signaling in Drosophila uncovered mutations in the porcupine gene that show a lack of Wnt activity due to the retention of Wg protein in the endoplasmic reticulum [12,13,14]. The porcupine gene is predicted to encode a protein with eight transmembrane domains and has a perinuclear localization in transfected cells [14]; overexpression of porcupine does not increase levels of secreted Wg but does change the pattern of Wg glycosylation [14]. In worms, mom-1 encodes a porcupine homolog and, when mutated, phenocopies mutants of mom-2, which encodes a Wnt, suggesting that the function of porcupine is conserved [15,16]. Although size chromatography suggests that Wg is secreted as a multimer, it remains unclear whether Wnt proteins in general are secreted as monomers, oligomers, or as part of a multi-protein complex [17]. Wnt proteins are glycosylated, but mutation of some or all of the predicted glycosylation sites in mouse Wnt1 does not abolish its activity in cultured cells [18]; these modifications may thus be unimportant for Wnt function.

Once secreted, Wnt proteins associate with glycosaminoglycans in the extracellular matrix and are bound tightly to the cell surface [19,20]. Although Wnts are found in tight association with the plasma membrane, it is possible to collect active Wnt from the medium of cultured cells [21,22]. Beyond this information, the localization of Wnt proteins in vertebrates is poorly understood. Examination of the localization of Wg in Drosophila, however, has provided critical insights into the subcellular distribution of Wnt proteins and the importance of this distribution for signaling activity. In the embryonic epidermis, Wg is found inside cells that secrete Wg and in association with the plasma membrane of secreting cells and non-secreting cells several cell diameters from the Wg source [23]. Wg is also prevalent in vesicles and multi-vesicular bodies of non-Wg-producing cells anterior to the source of Wg, suggesting that Wg is endocytosed [23,24]. This idea is supported by examination of shibire embryos, which have a mutation in dynamin, a critical component of the endocytic machinery; these mutants have defects in Wg distribution, and Wg signaling activity is compromised [25]. Similarly, expression of a dominant-negative form of shibire also reduces Wg activity [26]. Endocytosis may also help to limit the distribution of Wg signal. In contrast to cells anterior to the Wg source, cells posterior to Wg-producing cells have much lower levels of Wg in endocytic vesicles, and this asymmetry in distribution mirrors the observation that Wg acts over a much shorter range towards the posterior than towards the anterior. This difference in Wg distribution appears to be due to rapid degradation of endocytosed Wg in posterior cells [27]. The spatially restricted pattern of Wg degradation is regulated by signals through the epidermal growth factor (EGF) receptor that hasten the destruction of Wg in posterior cells [27].

Association of Wg with specific membrane microdomains also appears to play a role in controlling the distribution of Wg signals during Drosophila development. In imaginal discs, Wg is found in specialized membrane vesicles called argosomes, which are thought to be derived from lipid raft microdomains [28]. Incorporation of Wg into argosomes requires heparan sulfate proteoglycans, suggesting that proteoglycans play a role in sorting Wg to specialized membrane microdomains in Wg-producing cells or, alternatively, may play a role localizing Wg in distinct endocytic compartments in receiving cells.

Polarized distribution of wg transcripts in embryonic epithelial cells is also required for optimal signaling activity. High-resolution in situ hybridization analyses demonstrate that wg transcripts are localized apically in the embryonic epidermis and that this distribution is mediated by two cis-acting elements found in the 3' UTR of the wg mRNA [29]. Mutation of these elements results in uniform localization of wg transcripts and impaired Wg protein distribution and signaling. The asymmetric distribution of wg transcripts is dependent on dynein-mediated microtubule transport [30].

Loss-of-function mutations in 9 of the 18 mouse Wnt genes have been generated, and the phenotypes of mutant embryos demonstrate the diverse functions of Wnt genes during embryogenesis (Table 2). For example, knocking out Wnt1 results in a dramatic loss of a portion of the midbrain and deletion of the rostral cerebellum [74,75]. Inactivation of Wnt4 results in the absence of kidneys [76], masculinization of mutant females (absence of the Müllerian duct and continued development of the Wolffian duct) [77], and defects in mammary gland morphogenesis during pregnancy [78]. Targeted knockout of Wnt7a also has pleiotropic effects, including ventralization of the limbs [79], female infertility due to failure of Müllerian-duct regression [80], and a delay in the morphological maturation of glomerular rosettes in the cerebellum [81].

Overexpression and antisense 'knockdown' analyses in Xenopus have shown that the Wnt/β-catenin pathway is required for the specification of dorsal cell fates [82]. A debate is ongoing, however, over whether a maternal Wnt ligand is required to activate this pathway in dorsal cells. In support of a role for a Wnt ligand, a recent study has shown that XFzd7 is important for establishing dorsal cell fates [83], thereby implicating a Wnt ligand in this process. Furthermore, targeted knockout of Wnt3 in mice results in defects in axis formation and gastrulation, suggesting a conserved role for Wnts in regulating the establishment of the dorsal-ventral axis in vertebrates [84]. On the other hand, overexpression of a dominant-negative form of Xwnt8 in oocytes does not suppress formation of dorsal cell fates, arguing against the requirement for a maternal Wnt in axis specification [10]. Further studies are necessary to resolve the role of Wnts in vertebrate early axial development.

In flies, Wnt signaling has a variety of functions during development. The wg gene is required for cell-fate choices in the ventral epidermis during embryogenesis, as well as for many other functions, and DWnt2 is required for testis and adult muscle development [17]. In C. elegans, genetic analyses have defined a number of roles for Wnts, including establishment of polarity and endodermal cell fates in the early embryo and regulation of cell migration, among many others [85]. A comprehensive list of Wnt genes and their mutant phenotypes in vertebrates and invertebrates can be found at the Wnt gene homepage [5].

In addition to the many roles for Wnt signaling during development and in adult tissues, it is also involved in tumorigenesis in humans [59,64]. Although mutation or misexpression of a Wnt gene has not been linked directly to cancer in humans, mutation of several intracellular components of the Wnt/β-catenin pathway is thought to be critical in many forms of cancer. Most notably, patients with familial adenomatous polyposis (FAP) develop multiple intestinal adenomas early in life and have germline mutations in the APC gene. In addition, mutation of APC is associated with more than 80% of sporadic colorectal adenomas and carcinomas. More than 95% of germline and somatic mutations of the APC gene are nonsense mutations that result in the synthesis of a truncated protein lacking the region of APC that is important for its function in the destruction complex. Significantly, these truncations in APC remove binding sites for β-catenin and Axin, as well as putative phosphorylation sites for GSK3; as a result, the mutant APC protein cannot efficiently promote degradation of β-catenin. Mutations in the third exon of the human β-catenin gene (CTNNb1) that make it refractory to phosphorylation-dependent degradation and lead to inappropriate accumulation of β-catenin have also been identified in a large number of primary human cancers (see [64] for a table of β-catenin mutations in human cancers). Interestingly, mutations in CTNNb1 and APC are rarely found in the same tumor; for example, in colon cancer, in which the vast majority of tumors have mutations in APC, the overall frequency of CTNNb1 mutations is relatively low, but colorectal tumors lacking APC mutations are much more likely to have mutations in CTNNb1. Recently, Axin has also been shown to act as a tumor suppressor; mutations in the Axin1 gene have been found in human hepatocellular cancers [86]. Importantly, mutations in Axin1 and CTNNb1 found in hepatocellular carcinomas also show mutual exclusivity similar to that seen for APC and CTNNb1 in colon cancers. Together, these data strongly argue that mutations resulting in the stabilization of β-catenin can promote cancer in many tissue types.