The Wnt signaling pathway and its role in tumor development

Abstract

Cancer development depends on the aberrant activation of signal transduction pathways that control cell growth and survival and play important roles in normal embryonic development. This review will focus on one of the most powerful pathways, the canonical Wnt signal transduction cascade, which has been originally described in vertebrate and non-vertebrate embryogenesis and subsequently associated with the development of a multitude of different tumor types, mainly of gastrointestinal origin. In recent years, a variety of novel interacting components and functions have been identified in the Wnt pathway revealing not only the complexity of Wnt signaling but also its potency. Here we will concentrate on the role of the Wnt pathway in cancer development with emphasis placed on the molecular defects known to promote neoplastic transformation in humans and in animal models.

Introduction

Wnts are secreted factors that regulate cell growth, motility, and differentiation during embryonic development. Wnts act in a paracrine fashion by activating diverse signaling cascades inside the target cells. Wnt signaling has received considerable attention from cancer researchers in recent years because many of its components play important roles in tumor formation.

Wnt signaling diversifies into three main branches:

1. The "classical", also called canonical, Wnt pathway activates target genes through stabilization of β-catenin in the nucleus. The function of this pathway during embryonic development has been originally elucidated by experimental analysis of axis development in the frog Xenopus laevis and of segment polarity and wing development in the fly Drosophila melanogaster.

2. The planar cell polarity pathway involves RhoA and Jun Kinase (JNK) and controls cytoskeletal rearrangements. Its main role is the temporal and spatial control of embryonic development, as exemplified in the polar arrangement of cuticular hairs in Drosophila or the convergent-extension movements in Xenopus embryos. On a cellular level, this pathway regulates the polarity of cells through effects on their cytoskeletal organization. To date, there is no evidence for the involvement of the planar cell polarity pathway in tumor development, for which reason it is not discussed extensively here.

3. The Wnt/Ca²⁺ pathway is stimulated by Wnt 5a and Wnt 11 and involves an increase in intracellular Ca²⁺ and activation of Ca²⁺-sensitive signalling components, such as calmodulin-dependent kinase, the phosphatase calcineurin, and the transcription factor NF-AT. The Wnt/Ca²⁺ pathway can counteract the canonical Wnt pathway; however, it is not clear whether this pathway is conserved in mammals and whether it is implicated in tumorigenesis.

Overview of canonical Wnt signaling

The canonical Wnt signaling cascade controls cell behaviour by steering the transcriptional properties of DNA-binding proteins of the TCF/LEF-1 family (referred to below as "TCF"). In the absence of Wnt signaling, TCFs associate with corepressors to block expression of Wnt target genes. At the heart of the canonical Wnt pathway is the stabilization of cytosolic β-catenin, which activates target genes by binding to TCFs (Fig. 1).

Fig. 1.

Overview of the Wnt pathway. Binding of Wnts to frizzled receptors activates dishevelled which blocks the function of a complex assembled over the scaffold proteins axin or conductin. Note that axin and conductin are related proteins that can form similar complexes with APC, GSK3β, β-catenin, and diversin, and that APC can interact independently with β-catenin. In the absence of Wnts (-Wnt) the axin/conductin complexes promote phosphorylation of β-catenin by GSK3β (~P). Phosphorylated β-catenin becomes multi-ubiquitinated (Ub) and subsequently degraded in proteasomes. In the presence of Wnts or after mutations of APC, axin/conductin or β-catenin, phosphorylation and degradation of β-catenin is blocked which allows the association of β-catenin with TCF transcription factors (+Wnt or mutations). The TCF:β-catenin complexes bind to DNA and activate Wnt target genes together with various transcriptional repressors or activators. The branching of the canonical β-catenin pathway to the planar cell polarity and Ca²⁺ signalling pathways is indicated by dashed arrows. For details refer to the text

Wnts are secreted glycoproteins that bind to seven transmembrane receptors called frizzleds. LRP 5 and 6 act as essential co-receptors of Wnt ligands. Other secreted factors, such as WIF, cerberus, and FrzB antagonize Wnt binding to frizzled receptors, whereas Dickkopf (Dkk) inhibits the Wnt/LRP interaction. Another membrane-associated protein, kremen, binds to Dkk and induces endocytosis of LRP, thereby antagonizing Wnt signaling. Intracellularly, Wnt signaling leads to stabilization of cytosolic β-catenin. In the absence of Wnts, β-catenin is phosphorylated by casein kinases 1α and/or 1ε at serine residue 45; this in turn enables glycogen synthase kinase 3β to phosphorylate serine/threonine residues 41, 37, and 33. Phosphorylation of these last two amino acids triggers ubiquitination of β-catenin by β-TrCP and degradation in proteasomes. Phosphorylation of β-catenin occurs in a multiprotein complex containing the scaffold protein axin and/or its homologue axin2/conductin, the tumor suppressor gene product APC, GSK3β, and diversin, which links CK1e to the complex (Fig. 1).

In the presence of Wnts, the cytoplasmic component dishevelled blocks β-catenin degradation by a largely unknown mechanism. Stabilized β-catenin enters the cell nucleus and associates with TCF transcription factors, leading to the transcription of Wnt target genes. Transcriptional activation is mediated by the interaction of TCF:β-catenin complexes with various activators, such as the histone acetyl transferase CBP and the chromatin-remodeling SWI/SNF complex. Additionally, legless/bcl-9 recruits the nuclear protein pygopus to β-catenin thereby activating the TCF:β-catenin complex. A variety of Wnt/ β-catenin target genes have been identified which include regulators of cell proliferation, developmental control genes, and genes implicated in tumor progression. In the absence of β-catenin, TCFs can repress gene transcription either as naturally occurring dominant-negative variants or in association with transcriptional repressors such as groucho. In tumors, either overexpression of Wnts (which is uncommon in human tumors) or mutations in one the components responsible for the degradation of β-catenin, leads to stabilization of β-catenin and activation of target genes. Mutations occur predominantly in the tumor suppressor gene APC in about 80% of all colorectal carcinomas, at the N-terminal (phosphorylation) part of β-catenin in some colorectal carcinomas, and in a large variety of other tumours, and in axin or conductin, mainly in a small subset of liver and colon tumors and medulloblastomas.

Of note, β-catenin is also involved in the control of cell-cell adhesion by binding to cadherin cell adhesion molecules and providing a link to the actin cytoskeleton. It remains unsolved as to what degree this cell adhesion function of β-catenin also plays a role in Wnt signaling.

Wnt signaling in embryonic development

Wnt signaling is crucial to embryonic development in a variety of organisms ranging from C. elegans to mammals. These developmental roles will not be extensively discussed here, except for studies that deal with gene ablation experiments of Wnt signal components in the mouse. In general, it appears that one of the prominent biological phenomena controlled by Wnt signaling is the expansion of cells with predefined fates (Clevers 2002).

Two experimental in vivo settings have proved instrumental in the characterization of new components of the Wnt pathway. In Xenopus embryos ectopic expression of Wnts can induce the formation of a secondary body axis, resulting in double-headed tadpoles, a phenomenon related to the capacity of the Wnt pathway to induce dorsalization of the embryo. This Xenopus "double axis assay" is frequently used to study the activity and hierarchy of Wnt signaling components (Sokol 1999). If the amount of β-catenin is increased, for instance by injecting β-catenin mRNA at two-cell stage embryos, a secondary body axis forms, whereas if endogenous β-catenin is lowered, as for example by overexpression of negative regulators such as axin, ventralization and disturbance of axis formation occurs. In Drosophila the homologous wingless pathway is involved in the establishment of segment polarity, wing formation, and differentiation of the endoderm (Cadigan and Nusse 1997). Here, as in Caenorhabditis elegans, the powerful genetic methods available have contributed greatly to our understanding of Wnt signaling and have been a rich source for the identification of new components, which frequently also play crucial roles in human tumor development.

Components of the Wnt pathway

Secretion and regulation of Wnts

Wnts are secreted glycoproteins produced by different cell types (Cadigan and Nusse 1997). Wnt-1 was originally named int-1 and was first identified as a gene activated by integration of the LTR of the mouse mammary tumor virus resulting in the development of mammary tumors in mice. In Drosophila the homologous gene is called wingless, and the combination of both names led to the term Wnt. Currently there are 19 human Wnt proteins known, ranging around 40 kDa in molecular weight (Miller 2002). Not all Wnts activate the canonical Wnt/β-catenin pathway. The best studied Wnts acting in the β-catenin pathway are Wnt-1, Wnt-3a, and Wnt-8. It is suggested that the amino-terminal region of Wnts mediate interactions with Wnt receptors and that the carboxy-terminus is required to activate these receptors. Once secreted, Wnt proteins associate with glycosaminoglycans in the extracellular matrix and are bound tightly to the cell surface (Miller 2002).

Reception and transduction of Wnt signals involves binding of Wnt proteins to members of two distinct families of cell-surface receptors, namely the Frizzled gene family and members of the LDL-receptor-related protein (LRP) family. The canonical Frizzled receptor has an amino-terminal cysteine-rich domain that binds Wnt, seven transmembrane domains, and a short cytoplasmic tail containing a consensus PDZ domain-binding motif at the carboxy terminus (Bhanot et al. 1996). Ten Frizzled receptors have been identified in humans so far. The general structure of Frizzled receptors resembles that of seven-transmembrane G-protein-coupled receptors, suggesting that Frizzled proteins may use heterotrimeric G-proteins to transduce Wnt signals (Liu et al. 2001b; Malbon et al. 2001). Two members of the LRP-family, LRP-5 and LRP-6, bind Wnts and form a ternary complex with Wnt and Frizzled. Developmental studies show that LRPs are essential for transmission of the Wnt signal (Tamai et al. 2000; Wehrli et al. 2000; Mao et al. 2001a; Mao et al. 2001b).

Wnt signals are modulated extracellularly by diverse secreted proteins, most of them identified by genetic analyses, namely, Frizzled-related proteins (FrzB), WIF-1, (Wnt-inhibitory factor-1), and Cerberus in Xenopus (Miller 2002). These factors all bind to Wnt thereby antagonizing its function. Another antagonist, Dkk (Dickkopf) binds to the LRP coreceptor thereby blocking coreceptor availability (Mao et al. 2001a). There are four isoforms of Dkk. Dkk-3 does not affect Wnt-signaling, Dkk-1 and Dkk-2 both inhibit Wnt-signaling in Xenopus where they have been identified originally by their capacity to induce head formation. In mice, ablation of the Dkk-1 gene leads to a lack of head structure and duplications and fusions of limb digits (Mukhopadhyay et al. 2001). The binding of Dkk-1 to LRP does not affect non-canonical Wnt-signaling, namely, planar cell polarity. Dkk also binds to the membrane-anchored molecule kremen, which then triggers the internalization and clearing of the Dkk-LRP6 complex from the cell surface and blocks Wnt/β-catenin signaling (Mao et al. 2002). Again, only Dkk-1 and -2, but not Dkk-3, bind to kremen.

Although Wnt-1 has been identified in the experimental system of MMTV-induced tumorigenesis in mice, the evidence to date does not support a major role of Wnt proteins in human tumors, except for expression studies, which indicate in some cases overexpression of individual Wnts in tumor samples. Wnt5a was recently implicated as a factor promoting invasion and metastasis of melanoma cells; however, this activity was not associated with nuclear β-catenin signaling and might be related to the activation of Ca²⁺/PKC signalling (Weeraratna et al. 2002). Systematic studies of Frizzleds, DKK, LRPs, and kremen in tumors are lacking. Interestingly, DKK-1 is a transcriptional target of p53 and induced by various agents that lead to DNA damage (Shou et al. 2002).

Dishevelled

The coupling of Frizzleds to the downstream pathway remains poorly understood. Genetic analysis has shown that dishevelled is a positive mediator of Wnt signaling. In mammals three dishevelled genes are known, Dvl-1 to Dvl-3. Gene ablation of Dvl-1 led to abnormalities in social interaction and sensorimotor gating (Lijam et al. 1997), and, as a double mutant with Dvl-2, to an open neural tube. The Dvl-2 single mutant shows defects of cardiac morphogenesis and somite segmentation (Hamblet et al. 2002). Dishevelled activates both the canonical and the planar cell polarity pathway by signaling to the axin-based degradation complex of β-catenin and to RhoA/JNK, respectively.

Dishevelled has three potential interaction motifs, an N-terminal DIX domain that binds to axin and mediates dishevelled multimerization, a central PDZ domain, and a C-terminal DEP domain (Wharton 2003). The DIX domain is required for signaling in the canonical pathway but is dispensable for the planar cell polarity pathway. The DIX domain targets dishevelled to actin stress fibres and vesicular membranes. Mutation of the actin-binding motif of dishevelled enhances, while mutation of the vesicle interaction site disrupts Wnt/β-catenin signaling (Capelluto et al. 2002). The DEP domain might be required for JNK pathway activation.

A number of proteins have been shown to interact with dishevelled, some of which act as activators, some as antagonists; other partners can differentially modulate the β-catenin versus the planar cell polarity pathways. The recently identified molecule dapper possesses a leucine zipper and PDZ domain and acts as a dishevelled-associated antagonist of both β-catenin and JNK signaling (Cheyette et al. 2002). Frodo is highly similar to dapper in amino acid sequence but appears to play an opposite role with respect to Wnt signaling, i.e., frodo synergizes with dishevelled in axis formation (Gloy et al. 2002). It should be noted that dapper and frodo have been characterized only in Xenopus so far, and their function in mammals remains to be determined. Daam1, a formin homology protein, binds to the dishevelled PDZ and DEP domains to activate RhoA-dependent signaling that is implicated in planar cell polarity (Habas et al. 2001). Stbm (strabism) is a tetra-membrane-spanning protein that also functions in PCP signaling and is associated with dishevelled through the PDZ domain. Stbm leads to downregulation of Wnt/β-catenin signaling (Heisenberg 2002). Naked cuticle (Nkd) is involved in PCP signaling, activates JNK, and functions via binding to the PDZ domain of dishevelled. Moreover, Nkd is a downstream target of the Wnt pathway and could act in a negative feedback mechanism (Rousset et al. 2001).

Three kinases are known to associate with dishevelled. Casein kinase 1 (CK1) is a β-catenin pathway activator that interacts with both dishevelled and axin and stimulates binding of dishevelled to GBP thereby destabilizing the β-catenin degradation complex. However, CK1 was also implicated as the priming kinase for phosphorylation of β-catenin which leads to its degradation, suggesting that CK1 plays an inhibitory role in Wnt signaling (Amit et al. 2002; Schwarz-Romond et al. 2002; see below also). CK2 phosphorylates dishevelled in response to Frizzled (Willert et al. 1997). There might be a possible connection to cancer as CK2 promotes the formation of lymphomas when expressed in transgenic mice. Furthermore, CK2 is overexpressed in many human tumors, transformed cell lines, and rapidly proliferating tissues. Finally, PAR-1 is a dishevelled-associated kinase that is stimulated by Wnt and promotes Wnt/β-catenin signaling but inhibits JNK signaling (Sun et al. 2001).

The multiprotein β-catenin degradation complex

A multiprotein complex consisting of axin or its homologue conductin, GSK3β, APC, β-catenin, casein kinase1α/ε, and diversin, plays a pivotal role in the control of stability of β-catenin. The main function of this complex is to phosphorylate β-catenin and thereby trigger its ubiquitination and subsequent degradation in proteasomes. Phosphorylation can occur at amino acids Ser 33, Ser 37, Thr 41, and Ser 45 of β-catenin. Recent evidence suggests a two-step mechanism by which residue 45 is phosphorylated first by casein kinase 1 followed by phosphorylation of the other residues by GSK3β. Thus, CK1 serves as a "priming kinase" for GSK3β. The phosphorylation steps take place at the axin complex which serves to bring the kinases and the substrate β-catenin together. Phosphorylated β-catenin is then recognized by the ubiquitin ligase β-TrCP which induces its ubiquitination.

Axin/conductin

Axin and its homologue conductin, also called axin2 or axil, contain binding domains for the essential components involved in β-catenin degradation (Fig. 2). They consist of an amino-terminal "regulator of G-protein signaling" (RGS)-domain which binds to APC (Behrens et al. 1998; Hart et al. 1998; Kishida et al. 1998), separate binding domains for GSK3β and β-catenin in the centre of the proteins (Behrens et al. 1998; Ikeda et al. 1998), and a C-terminal (DIX) domain related to a segment of dishevelled, which mediates dimerization of axin (Fig. 2) (Zeng et al. 1997; Ikeda et al. 1998). On a molecular level, axin/conductin function as scaffold proteins by approximating all components needed for phosphorylation of β-catenin, i.e., GSK3β, casein kinase 1, APC, and β-catenin itself. In biochemical assays, physiological amounts of GSK3β phosphorylate β-catenin only very inefficiently in the absence of axin while addition of axin stimulates phosphorylation considerably (Ikeda et al. 1998). When overexpressed in mammalian cells, axin and conductin promote the degradation of cytosolic β-catenin (Behrens et al. 1998; Hart et al. 1998; Ikeda et al. 1998; Kishida et al. 1998), prevent the Wnt-induced accumulation of β-catenin, and block transcription of a TCF-dependent reporter gene (Sakanaka et al. 1998). Axin has been originally identified as the gene mutated in the fused mouse which is characterized by axial duplications reminiscent of the phenotype that is seen in Xenopus after activation of the Wnt signal (Zeng et al. 1997). Transgenic overexpression of axin prevents lobulo-alveolar development and induces apoptosis in mammary epithelial cells. Furthermore, axin blocks normal development of T-lymphocytes (Hsu et al. 2001).

Fig. 2.

Schematic representation of the structure of APC and axin and interaction with binding partners

Axin and conductin/axin2 share an overall identity in amino acids of 45%. We, and others, have shown that they are differently expressed in tissues and therefore might have different functions. While axin appears to be ubiquitously expressed during embryonic development and in adult tissues, axin2/conductin shows a more restricted expression pattern (Zeng et al. 1997; Behrens et al. 1998). In particular, axin2/conductin was shown to be upregulated in various tumors which exhibit activated Wnt signaling, i.e., in colorectal carcinomas and liver tumors (Lustig et al. 2002), and ovarian endometrioid adenocarcinomas (Leung et al. 2002). Moreover, axin2/conductin expression was upregulated by activation of the Wnt pathway, and the promoter of the axin2/conductin gene contains functional TCF-binding sites (Yan et al. 2001; Jho et al. 2002; Leung et al. 2002; Lustig et al. 2002). Altogether, this indicates that axin2/conductin is part of a negative feedback loop which is upregulated in response to Wnt signaling, while axin is a constitutive negative regulator of β-catenin activity.

β-catenin

β-catenin was originally identified as a binding partner of E-cadherin and α-catenin, thereby mediating the connection of cadherin cell adhesion molecules to the actin cytoskeleton (McCrea et al. 1991; Hulsken et al. 1994). β-catenin has a core domain of twelve so-called arm-repeats, which were first identified in the Drosophila homologue, armadillo, and amino- and carboxy-terminal flanking domains (Peifer et al. 1994). The arm-repeat domain binds to various partners such as cadherins, TCFs, APC, and axin/conductin in a mutually exclusive manner (Hulsken et al. 1994; von Kries et al. 2000). X-ray crystallography shows that the arm domain has a rod-like structure built up of a superhelix of α-helices (Huber and Weis 2001). The flanking domains of the arm domain interact with further proteins, such as α-catenin at the amino-terminus, and have transcriptional activation properties (see below). Moreover, the amino-terminal domain contains the serine and threonine residues critical for the degradation of β-catenin.

GSK3β

The serine/threonine kinase GSK3β (glycogen synthase kinase 3) binds to and phosphorylates several proteins in the Wnt pathway. Phosphorylation of β-catenin by GSK3β at position 33, 37, and 41 is crucial for recognition by the E3 ligase β-TrCP which then leads to ubiquitination of β-catenin and degradation in proteasomes (Kitagawa et al. 1999; Winston et al. 1999). In addition, GSK3β phosphorylates APC and axin (Rubinfeld et al. 1996; Yamamoto et al. 1999). Frat-1(= frequently rearranged in T-cell lymphomas) in the mouse or its homolog GBP (GSK3-binding protein) in Xenopus block GSK3β activity and compete with axin for binding to GSK3β (Salic et al. 2000). Frat-1 was identified by insertional mutagenesis in a screen for genes that enhanced the progression of transplanted T-cell lymphomas in mice (Jonkers et al. 1997). No contribution of FRAT-1 to human cancer is known up to now.

As a negative regulator of Wnt signaling GSK3β is a good candidate as a potential tumor suppressor. However, mutations in GSK3β, which would inactivate the kinase and thus stabilize β-catenin, have not been detected in colorectal carcinomas, possibly because GSK3β has additional functions that are essential in other pathways (Sparks et al. 1998). Recently, GSK3β similar to APC has been localized at the mitotic spindle and implicated in spindle formation (Wakefield et al. 2003). It is not known whether this function is related to Wnt signaling.

Casein kinase and diversin

For subsequent ubiquitination by β-TrCP, all four phosphorylation sites of β-catenin have to be phosphorylated. Mutations of each of these leads to deregulated β-catenin stabilization in cancer cells. The analysis of tumors for β-catenin mutations has detected a hotspot of mutations at Ser 45, a phosphorylation site not phosphorylated by GSK3β. Recent research has revealed Casein kinase 1α and 1ε as priming kinases for β-catenin phosphorylation (Amit et al. 2002). In Drosophila, depletion of CK1α inhibits β-catenin phosphorylation and degradation resulting in a naked cuticle phenotype, which indicates excessive β-catenin signaling (Liu et al. 2002). CK1ε can also be indirectly recruited to axin/conductin via interaction with diversin. Diversin is a negative regulator of the canonical Wnt pathway and directly interacts with casein kinase 1ε via its central domain and with axin/conductin via its C-terminal domain (Schwarz-Romond et al. 2002). In addition, there is an ankyrin repeat domain with an, up to now, unknown function. Diversin recruits CK1ε to the β-catenin degradation complex thereby facilitating the priming phosphorylation of β-catenin by CK1ε, which is followed by GSK3β phosphorylation. Although diversin and GSK3β use identical binding sites in axin or conductin, both can bind simultaneously to axin/conductin homo- or heterodimers. As its name implies, diversin has a dual function: while inhibiting the canonical Wnt pathway diversin promotes the Wnt/Jnk pathway. In Zebrafish, morpholinos against diversin block Wnt signaling-dependent morphogenesis (Schwarz-Romond et al. 2002). It has been shown that axin also binds directly to CK1ε, thereby facilitating phosphorylation of APC through axin (Rubinfeld et al. 2001). In other experimental settings it has been shown that CK1 can activate, rather than inhibit, β-catenin signaling; however, the mechanism remains unclear. Ser/Thr kinases of the CK1 family function in a number of cellular processes but expression patterns in cancer have not been determined so far.

APC

The APC gene is the main tumor suppressor gene implicated in the development of colorectal cancer. In 1991, APC was identified by positional cloning of the FAP (familial adenomatous polyposis) locus at chromosome 5q21 (Kinzler et al. 1991; Groden et al. 1991). FAP is an autosomal dominant inherited disease characterized by the development of hundreds to thousands of polyps in the colon, which eventually transform into cancer (Polakis 2000). The APC gene contains 21 exons encoding a 2843-amino acid protein. Exon 15 comprises >75% of the coding sequence of APC and is the most common target for both germline and somatic mutations which in the majority of cases lead to protein truncations (Polakis 2000). The first clues to the function of APC were provided by the identification of β-catenin as a partner of APC (Su et al. 1993). Further analysis showed that APC induces degradation of β-catenin when overexpressed in colorectal tumor cell lines, and identified APC as a binding partner of axin/conductin (Munemitsu et al. 1995; Behrens et al. 1998).

Domains, localization, and functions of the APC protein

APC binds to various proteins and has diverse functions in different subcellular compartments Fig. 2). The N-terminal third of the APC protein contains an oligomerization domain which is followed by an armadillo repeat region previously identified in the Drosophila homologue of β-catenin, armadillo (Peifer 1993). This domain is highly conserved and interacts with several proteins: (i) the APC-stimulated guanine nucleotide exchange factor (ASEF), which can activate Rac (Kawasaki et al. 2000); (ii) KAP3A which belongs to the kinesin superfamily and might connect APC to microtubules (Jimbo et al. 2002); and (iii) the regulatory subunit of the phosphatase 2A (PP2A) which also binds to axin (Seeling et al. 1999; Ikeda et al. 2000).

The central region of the APC protein contains three 15-amino-acids repeats and seven 20-amino acid repeats which all bind to β-catenin (Fig. 2). Unlike the 20-amino-acid repeats, the 15-amino acid repeats are not essential for downregulation of β-catenin and are retained in most APC-mutated tumors. The three so-called SAMP (serine, alanine, methione, proline)-repeats lie between the third and fourth, the fourth and fifth, and after the seventh 20-amino acid repeat and bind to the RGS-domain of the scaffolding proteins conductin or axin (Behrens et al. 1998). In most colon tumors SAMP repeats are deleted because they are located C-terminal to the mutation cluster region (MCR) (Fearnhead et al. 2001).

Several nuclear import and nuclear export signals (NES) are found in APC. In total, five NES have been identified, two of them located at the N-terminal part and three others near the centre within the β-catenin-binding domain (Henderson and Fagotto 2002). The N-terminal NES sequences seem to be well conserved and due to their location are retained in colon tumors. The functional role of the different NES sequences, and, in particular, their relative importance for nuclear export of APC, is a matter of debate. APC can shuttle between the cytoplasm and the nucleus. Treatment of cells with leptomycin B, an inhibitor of the nuclear export receptor CRM1, lead to a strong accumulation of APC in the nucleus. Interestingly, β-catenin also shuttles in and out of the nucleus and it was proposed that deletions of the central NES of APC in tumors leads to sequestering of the truncated APC and thus of β-catenin in the nucleus. However, a comparison of the NES sequence for functional export activity has shown that the central NES are of minor importance compared to the N-terminal NES sequences. Moreover, there is evidence that β-catenin can shuttle between the nucleus and cytoplasm independently of APC. All in all, there is still controversy whether the nuclear-export function of APC is crucial for its function as a tumor suppressor [see (Bienz 2002; Henderson and Fagotto 2002) for references and discussion].

Two domains in APC are implicated in association with microtubules. A basic region in the carboxy-terminal third probably mediates direct interaction with microtubules whereas a more carboxy-terminal domain binds to EB1, which is a microtubule-binding protein concentrated at microtubule tips and centrosomes (see also below). The functional relevance of the APC:EB1 interaction is not clear; however, it has been shown in vitro that EB1, together with the C-terminus of APC, can promote microtubule polymerization (Su et al. 1995; Mimori-Kiyosue et al. 2000; Bienz 2001; Nakamura et al. 2001).

The best-characterized function of APC is its role as a negative regulator of the canonical Wnt pathway via its interaction with β-catenin. Deletions of APC in colorectal tumors result in stabilization of β-catenin and activation of TCF:β-catenin-dependent gene transcription. Restoration of APC function by transient expression of wild-type APC can block β-catenin signaling in colon tumor cell lines (Korinek et al. 1997; Morin et al. 1997). Moreover, genetic ablation of one of the two homologues of APC in Drosophila, APC2, results in upregulation of Wnt/β-catenin signaling. On a molecular level, APC seems to function by interacting with β-catenin and axin/conductin via distinct domains (Behrens et al. 1998). However, the molecular mechanism by which APC promotes destruction of β-catenin is not clear, in particular, since degradation of β-catenin can be bypassed in vitro by overexpression of axin/conductin. It appears that activity of APC depends on its interaction with axin/conductin because mutations of the SAMP repeats lead to loss of APC activity in β-catenin degradation (von Kries et al. 2000), and mouse mutants with deletions in the APC gene that retain one SAMP repeat do not develop tumors (Smits et al. 1999). One possible role of APC under conditions of normal protein expression might be related to its capacity to bind β-catenin via several independent domains (i.e., the 15- and 20-amino acid repeats). Thus, APC might act as a molecular trap for efficient detection and clearance of β-catenin in the cytoplasm. Alternatively, APC might stimulate the activity of the axin/conductin-based degradation complex by an indirect mechanism, for instance, by recruitment of (yet unknown) activators. It is also suggested that truncated APC proteins in tumors which retain some of the β-catenin binding repeats can still partially block β-catenin signaling (Kielman et al. 2002).

Apart from this well-studied function of APC, much effort has been undertaken to determine the subcellular localization of APC, in particular, as discussed above, its localization in the nucleus and the association with the cytoskeleton. Endogenous APC has been located at the distal ends of microtubules in migrating mammalian tissue-culture cells and in epithelial inner ear cells (Nathke et al. 1996; Mogensen et al. 2002). Exogenously expressed wild-type APC has been shown to bind directly to microtubules through its C-terminal basic region. As a consequence, microtubules are stabilized in vitro and in vivo (Zumbrunn et al. 2001). APC has also been localized to the mitotic spindle and to kinetochores, and loss of APC in embryonic stem cells was associated with defects in chromosome segregation (Fodde et al. 2001; Kaplan et al. 2001). Thus, loss of APC in tumors could not only trigger β-catenin signaling but also lead to chromosomal instability and promote cancer progression. To further elucidate this possibility, Tomlinson and co-workers analysed chromosomal instability in 55 adenomas of FAP patients with two characterized mutational hits of APC (Sieber et al. 2002). Although APC was mutated on both alleles, 85% of all adenomas were diploid. The authors therefore concluded that APC mutations do not eventually lead to CIN and genomic instability.

Immunohistochemical studies showed that APC also associates with the lateral plasma membrane of polarized mammalian epithelial cells, which requires an intact actin cytoskeleton. APC truncations failed to associate with the plasma membrane, which implies that the putative function of APC at the plasma membrane might be lost in tumours (Rosin-Arbesfeld et al. 2001; Hamada and Bienz 2002). In addition, the C-terminus of APC has been shown to directly bind to the PDZ domain of hDLG, a human homologue of Drosophila discs large (DLG) tumour suppressor protein, also indicating that APC is associated with cell-cell adhesion sites since hDLG is distributed along lateral membranes (Matsumine et al. 1996). Furthermore, the binding of ASEF to APC results in actin polymerization at the cell periphery as exemplified by increased membrane ruffling and cell migration (Kawasaki et al. 2000).

Regulation of β-catenin levels

While it appears clear that the triggering of Wnt signaling in tumors is due to mutations of APC, β-catenin or axin/conductin (see below), the mechanism of inhibition of β-catenin degradation by physiological Wnt signaling remains an open question. Overexpression analyses and genetic experiments have identified various players that can promote or inhibit Wnt signaling and either act enzymatically or through modulation of protein-protein interactions. From these studies it is clear that either direct or indirect interaction of dishevelled with the axin-based degradation complex is of central importance. For instance, binding of dishevelled to axin results in recruitment of the GSK3β inhibitor GBP which could thereby block phosphorylation of β-catenin as well as of axin and APC (Li et al. 1999; Farr et al. 2000). Dishevelled could also interfere with dimerization of axin/conductin as it interacts with the DIX domain, or recruit these components to specific locations within the cell, for instance, to the vesicular structures which appear to be crucial for activation of the Wnt signal (Capelluto et al. 2002). An even more elusive question is how the signal from Frizzled receptors is transmitted to dishevelled [for review see (Wharton 2003)]. Until now the number of components that have been identified to interact with Frizzleds is surprisingly low. For instance, axin was shown to bind to the cytoplasmic tail of LRP; this interaction could bypass the requirement of dishevelled (Mao et al. 2001b).

Several other axin-independent mechanisms of β-catenin degradation have been proposed. In one setting, priming of β-catenin at residue 45 is performed by protein kinase A followed by phosphorylation at residues 33/37/41 of β-catenin by GSK3β (Kang et al. 2002). The Alzheimer disease-associated protein presenilin is needed for connecting these two phosphorylation events. Presenilin-deficient cells show increased levels of β-catenin and overexpression of presenilin led to a reduction in β-catenin even in axin-deficient cells. Furthermore, presenilin promotes phosphorylation of β-catenin by associating with PKA and GSK3β independent of axin which is not regulated by the Wnt pathway (Kang et al. 2002). The relationship of these findings to tumor development remains to be determined, although it is of interest that tumors develop in presenilin-deficient mice.

Recently, Siah-1 (human homologue of Drosophila seven in absentia) which is a p53-inducible mediator of cell cycle arrest, tumor suppression, and apoptosis, was found to interact with the carboxy-terminal domain of APC thereby promoting degradation of β-catenin in mammalian cells (Liu et al. 2001a). β-catenin degradation was independent of GSK3-mediated phosphorylation and did not require the F-box protein β-TrCP. Another report proposed a similar model in which Siah-1 links the F-box protein Ebi to β-catenin, thereby stimulating degradation of β-catenin independent of its phosphorylation (Matsuzawa and Reed 2001). In line with a role of a crosstalk of p53 with β-catenin, it has been shown that overexpression of β-catenin results in accumulation of p53 (Damalas et al. 1999), and that activated p53 can in turn downregulate β-catenin. In these experiments GSK3β was required for degradation of β-catenin (Sadot et al. 2001).

Interactions of β-catenin in the nucleus

As mentioned above, β-catenin can travel in and out of the nucleus apparently independent of conventional nuclear localization and export signals. In fact, in in vitro systems β-catenin behaves in a similar way to importins which are the factors involved in NLS-dependent transport of other proteins to the nucleus (Wiechens and Fagotto 2001). Dephosphorylation of the components of the degradation complex not only inhibits ubiquitination of β-catenin but also lowers the affinity of β-catenin to axin and APC, thereby resulting in the release of β-catenin from the complex (Willert et al. 1999). Moreover, dephosphorylated β-catenin preferentially enters the nucleus, whereas phosphorylated β-catenin, that can be enriched by blocking proteasome activity, is neither able to enter the nucleus nor activate TCF:β-catenin-dependent transcription (Staal et al. 2002; van Noort et al. 2002).

TCF/LEF

After its translocation into the nucleus, β-catenin binds to members of the LEF/TCF family thereby activating target genes. Two of the four members of the LEF/TCF family, TCF-1 (T-cell factor) and LEF-1 (lymphoid enhancer factor), were originally identified in screens for T-cell-specific transcription factors, and later two additional members, TCF-3 and TCF-4, were described. In Drosophila and C. elegans only one TCF gene exists, called dTCF/pangolin or pop-1. All proteins contain a highly conserved high-mobility group (HMG) box of around 80 amino-acids that binds to DNA as a monomer in a sequence-specific manner (van Noort and Clevers 2002). Apart from mediating DNA recognition, the HMG box also induces a bend in the DNA enabling the binding of various transcription factors which can act as activators or repressors of specific target genes (Giese et al. 1995). The first identified binding partner of TCF was β-catenin that binds to a conserved N-terminal region of TCFs via its armadillo repeats (Behrens et al. 1996; Molenaar et al. 1996).

Because TCF/LEF factors cannot activate transcription on their own, they need a co-activator. β-catenin possesses multiple transactivating elements that also operate independently of TCFs, and, in all organisms tested, there is a strict correlation between the ability of β-catenin to function in Wnt signaling and its ability to transactivate (van de Wetering et al. 1997). Both C- and N-terminal regions of β-catenin have been shown to provide transactivation domains (Hsu et al. 1998). Fusion proteins of LEF-1 with the C-terminal domain of β-catenin, or with the potent VP16 transactivation domain, activate LEF-1-dependent promoters to a similar extent (Vleminckx et al. 1999). Moreover, both fusion proteins induce axis duplication in Xenopus, indicating that the transactivation function of β-catenin is sufficient for Wnt signaling in this particular experimental setting. Similarly, cell transformation of chicken embryo fibroblasts could be achieved by expression of chimeras of LEF-1 and various heterologous transactivation domains (Aoki et al. 1999). TCF:β-catenin complexes can thus be regarded as bipartite transcription factors in which the DNA binding and transactivation functions are contributed by separate proteins.

It is presently unclear how the transactivation domains of β-catenin activate transcription, but associations of β-catenin with the TATA-binding protein have been shown. In addition, β-catenin might contact TBP indirectly via binding to Pontin52 (Bauer et al. 2000). This indicates that TCF:β-catenin complexes directly recruit components of the basal transcriptional machinery to the Wnt target genes. The histone acetylase p300/CBP was found to interact directly with the C-terminal domain of β-catenin and acts as a coactivator (Hecht et al. 2000). In addition, β-catenin recruits Brg-1, a component of the SWI/SNF complex, to TCF target gene promoters, facilitating chromatin remodeling as a prerequisite for transcriptional activation (Barker et al. 2001). Recently, the β-catenin-TCF-complexes have been shown to interact with two novel components, pygopus and BCL-9/Legless (Kramps et al. 2002; Thompson et al. 2002). Both are nuclear proteins involved in TCF activation. Legless (Lgs) encodes the homolog of human BCL-9 and serves to link pygopus to β-catenin. The recruitment is essential to transcriptionally activate Wnt target genes. As the murine Bcl-9 is an oncogene that is involved in the development of Non-Hodgkins lymphomas, further analysis might elucidate whether a deregulation of these complexes play a role in the development of B cell malignancies (Sarris and Ford 1999). No analysis of these components has been performed so far in solid tumors and further genetic experiments would be of great interest.

TCFs act as transcriptional repressors in the absence of Wnt signaling by complex formation with one of several co-repressors. Groucho binds to TCFs and acts as a co-repressor, and TCF-3 and TCF-4 also bind another transcriptional repressor, CtBP (Roose et al. 1998; Roose and Clevers 1999). Recently FHL2, an LIM-domain protein expressed in myoblasts, but downregulated in malignant rhabdomyosarcoma cells, has been identified as a muscle cell-specific repressor of LEF/TCF target genes by interacting with β-catenin in the nucleus (Martin et al. 2002). All members of the LEF/TCF family undergo alternative splicing. Most interestingly, LEF-1 and TCF-1 also use two different promoters leading to two different isoforms, one of them lacking the N-terminal β-catenin-binding domain, which might therefore act in a dominant-negative fashion (Roose et al. 1999).

It would be desirable to identify drugs that interfere with the β-catenin-TCF/LEF complex, thereby abolishing the activation of target gene and subsequent tumor formation. Unfortunately the armadillo groove binds TCF, APC, and E-cadherin. Breaking this complex therefore could lead to the dissociation of the cells by inhibiting E-Cadherin function.

As described above, the consensus view is that Wnt signaling induces β-catenin to enter the nucleus and combine with TCF to form a transcription factor complex in which TCF binds DNA and β-catenin activates transcription. Recently, a challenging view suggested that β-catenin may transduce Wnt signals by exporting TCF from the nucleus or activating it in the cytoplasm (Chan and Struhl 2002). These results were mainly obtained by genetic experiments in Drosophila and cannot be discussed here in full detail. It remains to be determined whether this novel interpretation of the function of TCF:β-catenin complexes is also valid in mammalian systems and can be extended to tumor formation.

In vivo functions of TCFs

Knockout mice have been generated for three TCFs. In adult mice TCF-1 is expressed in T lymphocytes. TCF-1^-/- mice are impaired in the generation of T lymphocytes but have functional peripheral T-cells and are fully immunocompetent (Verbeek et al. 1995). Closer examination of these mice showed that they develop spontaneous intestinal adenomas and mammary adenoacanthomas. Further analysis revealed high levels of nuclear β-catenin in the intestinal adenomas and the predominant expression of the TCF-1 isoform that lacks the N-terminal β-catenin interactions domain. These isoforms may therefore repress Wnt target genes and, being themselves activated by Wnt signaling, act in a negative feedback loop. Indeed when TCF-1^-/- mice were mated to min mice which carry a mutated apc allele and are prone to the development of intestinal adenomas, the number of adenomas was significantly increased in the progeny (Roose et al. 1999).

Conversely, LEF-1 appears to be involved in a positive feedback loop that may be important in colon tumors. The LEF-1 gene is also activated by TCF:β-catenin complexes but in this case activation leads to transcription of full-size LEF-1 that is capable of binding to β-catenin. These LEF-1/β-catenin complexes could further boost transcription of Wnt target genes (Hovanes et al. 2001). LEF ^-/- mice die shortly after birth showing deficiencies in epithelial-mesenchymal interactions [for description of the complex phenotype [see (Kratochwil et al. 1996)]. TCF-1/LEF-1 double knock-out mice result in an early embryonic phenotype that resembles that of the Wnt-3A knock-out (Galceran et al. 1999).

TCF-4 is prominently expressed in the gut epithelium throughout life. TCF4^-/- mice die shortly after birth and show a complete absence of the stem cell compartment in the crypts of the small intestine with a lack of actively dividing crypt cells. Thus, TCF-4 functions to maintain early progenitor cells. In fact, a direct target gene of TCF:β-catenin complexes, c-myc, was shown to repress the p21^CIP/WAF1 promoter, which leads to cell cycle progression. Conversely, activation of dominant-negative TCF-4 in colorectal tumor cells led to a G1 arrest and differentiation of colon carcinoma cells which was mediated by p21 ^CIP/WAF1 (van de Wetering et al. 2002b). That TCFs control cell differentiation was also revealed by the finding that EphB2 and EphB3 receptors were downregulated by dominant-negative TCF while their ligand ephrinB1 was upregulated. Further analysis in EphB knock-out mice showed that this differential regulation served to prevent the intermingling of cells within the intestinal epithelium (Batlle et al. 2002). Specifically, the repulsive interactions between ephrins and Eph receptors allow the proliferating and differentiating cells to become sorted out from each other and thereby generate compartment boundaries. The authors speculate that these findings might be of relevance for the development of intestinal adenomas as seen in the min mouse where mutant APC cells invaginate around the crypt-villus junction and avoid migration into the area of high ephrinB expression at the top of the crypt. Whether this is also relevant for human colon tumor formation remains to be determined. In particular, it is not clear whether tumor cells develop from the base or the top of the crypt compartment. At early stages of human colon adenomas, dysplastic cells were found at the luminal surface of the crypts whereas the cells at the bases of the same crypts appeared morphologically normal. Furthermore, mutation analysis by Vogelstein's group revealed that mutations of APC occur in these dysplastic cells at the top of the crypts whereas the base of the crypts did not contain APC alterations (Shih et al. 2001). Therefore, the authors propose a top-down mechanism for morphogenesis of colorectal tumors.

Target genes

A considerable number of target genes of Wnt signaling playing a role in development and tumorigenesis have been identified to date (see also http://www.stanford.edu/~rnusse/wntwindow.html for a complete list). Here we will focus on the most relevant genes implicated in cancer development. Most of these genes are direct targets, which means that their promoters contain TCF-binding sites and are upregulated by TCF:β-catenin complexes.

Cell cycle, apoptosis

The first target genes identified were c-myc, a proto-oncogene, and cyclin D1, an activator of cyclin-dependent kinases (He et al. 1998; Shtutman et al. 1999; Tetsu and McCormick 1999). Through both mechanisms Wnt signaling may stimulate progression through the cell cycle. Studies of the relevance of c-myc as a Wnt target have been described above. Briefly, upregulation of c-myc by TCF:β-catenin leads to repression of p21^CIP transcription through binding of c-myc to the p21^CIP promoter. Since p21^CIP can block the activity of several cyclin-dependent kinases, TCF:β-catenin may release cell cycle arrest by downregulating p21^CIP. TCF:β-catenin complexes can also directly stimulate the activity of CDKs and thereby promote cell cycle progression through upregulation of cyclin D1. Crossing APC deficient min mice with cyclin D1 -/- mice did not completely prevent tumorigenesis but resulted in the development of fewer tumors, indicating that cyclin D1 is not essential for intestinal tumorigenesis but may act as a modifier gene (Wilding et al. 2002). Other protooncogenes such as c-jun and fra-1 possess TCF recognition sites in their promoters and were shown to be regulated by β-catenin. Expression of the anti-apoptotic gene survivin was downregulated by APC and analysis of its promoter revealed TCF-4-binding sites, thus implicating it as a target gene (Zhang et al. 2001a). Indeed, survivin was shown to be mainly expressed at the base of the crypts of normal colon epithelium where active Wnt signaling occurs. Moreover, survivin is known to be upregulated in colon tumors.

Growth factors

Secreted factors regulated by Wnt signaling are of particular interest because these could act in a paracrine fashion on neighboring "silent" cells and thereby speed up tumor progression, for instance by promoting growth of these cells. The vascular endothelial growth factor (VEGF) is a pivotal pro-angiogenic protein and its levels have been shown to be elevated in colorectal cancer and correlated with poor outcome. VEGF expression was upregulated by Wnt signaling through TCF-binding sites in its promoter. However, it appeared that expression of VEGF was dependent on the concomitant presence of mutation in the ras oncogene (Zhang et al. 2001b). WISP-1 (Wnt-1-induced secreted protein 1) has been identified as a downstream target that is activated by Wnt/β-catenin signaling, although TCF/LEF-binding sites in its promoter play only a minor role. WISP-1 is a member of the connective tissue growth factor family and can transform fibroblasts through unknown pathways. Moreover, genomic DNA of WISP-1 has been shown to be amplified in colon cancer cell lines and colon cancers, indicating alternative mechanisms of activation of these factors (Pennica et al. 1998; Xu et al. 2000). BMP-4 expression was dependent on mutant β-catenin, and it was shown to be overexpressed in APC mutant cells, but its role has still to be elucidated (Kim et al. 2002). The secreted peptide gastrin which has cell-growth stimulating properties is a direct target of TCF:β-catenin. The importance of gastrin in colon epithelial cells was underlined by the fact that gastrin-deficient APC ^min-/+ mice showed a marked decrease in polyp number and decreased polyp proliferation rate, while transgenic overexpression of gastrin in APC ^min-/+ mice resulted in an increase in polyp number (Koh et al. 2000; Watson and Smith 2001). The receptor tyrosine kinase c-met, the receptor for the epithelial growth, motility, and survival factor HGF/scatter factor, was shown to be upregulated in adenomas of FAP patients and could be downregulated by dominant-negative TCF in APC-mutated cell lines (Boon et al. 2002). In addition, it has been shown previously that amplification of the c-met gene occurs in 10% of primary colon tumors and, in particular, in liver metastases of colon cancer. Naturally, the outcome of upregulation of c-met would be cell autonomous, but might allow the tumor cells to respond to HGF from surrounding mesenchymal cells.

Targets involved in tumor progression

Interestingly, a particularly strong nuclear accumulation of β-catenin occurs at the invasive front of colorectal carcinomas, indicating that β-catenin might regulate target genes involved in tumor invasion and progression. Indeed several proteases capable of degrading extracellular matrix such as matrilysin/MMP7 and MMP-26 were identified as targets (Brabletz et al. 1999; Crawford et al. 1999; Marchenko et al. 2002). The matrix metalloproteinase matrilysin has an impact on cancer progression since deficiency of matrilysin in min mice resulted in a decrease of total tumor number and size (Crawford et al. 1999). Moreover, nuclear accumulation of β-catenin together with MMP-7 expression at the invasion front was related to unfavourable outcome in colon cancer (Ougolkov et al. 2002). Cell adhesion molecules are also potential Wnt targets. The promoter of the Nr-CAM gene has several LEF/TCF-binding sites and Nr-CAM has been demonstrated to be overexpressed by either β-, or γ-catenin (plakoglobin) in colon and melanoma cell lines as well as in colon carcinomas. Moreover, Nr-CAM can transform fibroblasts and increase their motility. As Nr-CAM is a member of the Ig superfamily of adhesion receptors originally identified in the nervous system, it will be interesting to determine by which mechanism Nr-CAM exerts its effect in the colon (Conacci-Sorrell et al. 2002). CD 44 was shown to be strongly overexpressed already in aberrant crypt foci and its expression was lost in TCF-4 knockout mice (Wielenga et al. 1999). The γ2 chain of laminin is also upregulated in the invasive areas of colon cancer and its gene promoter contains functional TCF-binding sites (Hlubek et al. 2001).

Transcription factors

ITF-2 (immunoglobulin transcription factor-2), a basic helix-loop-helix transcription factor was shown to be a direct target of β-catenin and upregulated in human cancer, and, when overexpressed, ITF-2 induced transformation of cells (Kolligs et al. 2002). In addition, Id2, the dominant negative helix-loop regulator was shown to be upregulated in a TCF-dependent manner in colon cancer and embryonic carcinoma cells (Rockman et al. 2001; Willert et al. 2002). AF17 is thought to function as a transcriptional regulator and is a fusion partner of MLL in certain acute lymphoblastic leukaemia and myeloid leukaemia (Lin et al. 2001). AF17 was transcribed after β-catenin accumulation and upregulated in colon cancer samples. Expression of the promyelocytic leukaemia gene, PML, was activated through β-catenin or plakoglobin but showed to be LEF/TCF-independent (Shtutman et al. 2002). Further analysis revealed that PML forms a complex with β-catenin in the nucleus and, together with p300, cooperates in transactivation of a subset of direct target genes.

Negative feedback targets

Through the activation of certain target genes, the Wnt pathway creates negative feedback loops that can block its activity. These genes have been characterized before as negative regulators of the pathway. Expression of the βTrCP ubiquitin ligase receptor which targets the ubiquitination of β-catenin is activated by Wnt signaling, apparently at the post-transcriptional level (Spiegelman et al. 2000). In addition, this creates a cross-talk between β-catenin/Tcf and NF-κB pathways because it facilitates the ubiquitination of IκB. Similarly, the target gene conductin/axin2 is involved in β-catenin degradation, while naked cuticle (Nkd) interferes with dishevelled, and dominant-negative forms of TCF-1 block target gene activation (see also above).

Other targets

The multidrug resistance1 gene was shown to have TCF4-responsive elements in its promoter, and Hirohashi and co-workers proposed that by activation of this gene colorectal tumorigenesis could be initiated by suppressing cell death pathways (Yamada et al. 2000). Two target genes of Wnt signaling, the nuclear hormone receptor PPARdδ (peroxisome proliferator-activated receptor δ) and cyclooxygenase 2 (COX-2), are of special interest because they have been implicated in the development of colorectal carcinomas and are inhibited by chemopreventive non-steroidal anti-inflammatory drugs (NSAIDs), such as sulindac and acetylsalicylic acid (Fig. 3). COX-2 produces eicosanoids from arachidonic acid which bind to PPARδs and stimulate interaction of these transcription factors with specific promoters together with retinoid acid receptors. PPARδ is upregulated early in carcinogenesis by binding of TCF:β-catenin complexes to four TCF-response elements. COX-2 has been demonstrated to be a Wnt target as it was shown to be upregulated by induction of stabilized β-catenin through Wnt-1 in mouse mammary cell lines (Howe et al. 1999). However, β-catenin appears to stimulate COX-2 transcription indirectly through upregulation of PEA3 family transcription factors, which are potent transcriptional activators of COX-2 (Howe et al. 2001). NSAIDs block PPARδ function at two levels: first, they block eicasonoids synthesis through inhibition of COX-2 and second, they interfere with interaction of PPARδ with its target promoters. NSAIDs have been shown to inhibit colon tumorigenesis. For instance, the regular use of aspirin is correlated with a lower incidence of colorectal carcinomas. Moreover, treating APC^Δ716 mice with a selective COX-2 inhibitor reduced the polyp number (Oshima et al. 1996), and treatment of FAP patients with the COX-2 inhibitor celecoxib resulted in significant reduction in the number of colorectal polyps (Steinbach et al. 2000) [for discussion see also (Gupta and Dubois 2001)].

Fig. 3.

Wnt signaling and NSAIDs in colorectal carcinomas. APC mutations lead to stabilization of β-catenin which can activate transcription of COX-2 (via PEA3) and PPARδ. COX-2 is involved in synthesis of PPARδ ligands such as eicosanoids which stimulate binding of PPARδ to target gene promoters. PPARδ activates these promoters by collaborating with retinoic acid receptors (RXR). NSAIDs can block this activation at two levels: first, by inhibiting COX-2; and second, by interfering with DNA binding of PPARδ

Introducing wild-type APC into SW480 APC-mutated colon cancer cells revealed a new target gene called APCDD1 that promotes proliferation of colon cancer cells in vitro and in vivo (Takahashi et al. 2002). Another gene upregulated by TCF:β-catenin with a potential role in colon tumor cell growth and differentiation is ectodermal-neural cortex 1 (enc-1). One can expect that further downstream targets will emerge and be characterized in the next years [see a list of microarray data on (Willert et al. 2002)], and it will be interesting to compare in more detail the relevance of these genes for human tumors.

Mouse models for tumorigenesis through the Wnt pathway

Mouse models for intestinal tumorigenesis have been established by introducing germ-line mutations in the APC gene or by mutational activation of β-catenin. Unlike in the human disease, in all of these models tumors develop mainly in the small intestine and less frequently in the colon. The min mouse is the best characterized of these models and was generated by chemical mutagenesis that introduced a chain-terminating mutation at nucleotide 2549 in murine APC (Su et al. 1992). Heterozygous min mice develop numerous intestinal adenomas which show loss of heterozygosity for the APC gene. Other mouse models of FAP have been created by gene targeting of APC, such as APC ^Δ716 and APC 1638 N, both of which mimic the APC min phenotype in adenoma location but vary significantly in tumor number and life span (Fodde et al. 1994; Oshima et al. 1995). Extracolonic manifestations are prominent in the APC 1638 N animals; specifically, these animals develop desmoid tumors, cutaneous cysts, and retinal pigment epithelium abnormalities. Interestingly adenoma formation in the colon was achieved by generating APC mutations with a conditional targeting strategy and an adenovirus expressing cre-recombinase introduced rectally (Shibata et al. 1997). In another approach, a deletion mutant of exon3 encoding the phosphorylation sites of β-catenin was generated using the loxP/Cre system which also led to intestinal tumorigenesis (Harada et al. 1999). These results showed that activation of the Wnt signaling pathway can cause intestinal and colonic tumors and also suggest that the block of degradation of β-catenin is sufficient for tumorigenesis. Tumors were also obtained by transgenic expression of β-catenin in various other tissues such as the mammary gland, intestine, and skin (Gat et al. 1998; Chan et al. 1999; Romagnolo et al. 1999; Imbert et al. 2001; Cheon et al. 2002). Interestingly, the block of Wnt signaling by transgenic expression of a dominant-negative LEF-1 in the epidermis also resulted in formation of skin tumors (Niemann et al. 2002).

The TGFβ pathway has been implicated in suppressing tumorigenesis in the colon. As smad2 has been found to be associated with colorectal cancer, smad2/APC double-heterozygous mice were generated. In these mice the total number of intestinal tumors did not change compared to the APC heterozygous mice; however the mice died of intestinal obstructions caused by extremely large tumors (Hamamoto et al. 2002). The authors concluded that smad2 does not initiate tumorigenesis by itself but accelerates malignant progression of tumors to invasive cancers in the late stage of carcinogenesis (Hamamoto et al. 2002). Similarly, in compound mice generated by crossing of APC^Δ716 knockout mice with smad4 knockout mice, intestinal polyps developed into more malignant tumors than in simple APC^Δ716 heterozygotes (Takaku et al. 1998). In the converse approach, COX-2 null mutation reduced the number and size of the intestinal polyps dramatically. Furthermore, treating APC^Δ716 mice with a novel COX-2 inhibitor reduced the polyp number (Oshima et al. 1996).

Alterations of Wnt pathway components in cancer

With the notion that the tumor suppressor APC has a regulatory function in the Wnt pathway and the finding of stabilizing mutations of β-catenin in colorectal cancer, it became clear that Wnt signaling is not only important in the regulation of embryonal development but also in tumorigenesis (Bienz and Clevers 2000). Soon after, the pathway became implicated in a variety of other tumor types mainly through the discovery of mutations in APC, Axin and Axin2/conductin, and β-catenin. In most cases, alterations of these genes can be linked to an increase of transcriptionally active β-catenin. In general, only one of these genes is mutated in a given tumor sample reflecting their role in a common pathway. For instance, colon tumors with mutations in APC in general have a wild-type β-catenin gene, and vice versa, tumors with mutations in β-catenin are wild-type for APC. Most of the β-catenin mutations are activating mutations, mainly occurring in exon 3 at one of the 4 phosphorylation sites [see (Polakis 2000) for detailed analysis]. The most frequently mutated tumors are of gastrointestinal origin, although other tumors harbour mutations as well. In the following, we will describe some of the mutational analyses in tumors that have been performed so far.

Colorectal cancer (CRC)

CRC is one of the most common malignancies in the Western world. A series of molecular changes results in the development of adenomas, which in the end progress to carcinomas (Fearon and Vogelstein 1990). One of the central players is the APC (adenomatous polyposis coli) gene which was identified as the responsible germline mutation in FAP (familial adenomatous polyposis) patients. APC is also the predominant gene mutated in non-inherited CRC, about 80% of them carrying a mutation in this gene. Since APC mutations are detected very early in the adenoma-carcinoma sequence, the APC protein has been suggested to act as a gatekeeper of colorectal carcinogenesis, which means that functional loss of APC is a prerequisite for the further progression towards malignancy (Kinzler and Vogelstein 1996).

Germline mutations of APC

FAP was first described by Lockhart-Mummery in 1925 (Lockhart-Mummery 1925). Germline mutations are inherited in one allele of APC and, after loss-of heterozygosity, result in the development of hundreds of polyps in the colon. Thus, APC behaves as a classical tumor suppressor and follows the Knudsen two-hit hypothesis. FAP affects around 1 in 13,500 individuals. Different germline mutations with differences in penetrance, severity of polyposis, and extracolonic manifestations have been described [for extracolonic features and detailed genotype-phenotype analysis see (Fearnhead et al. 2001)]. Most of the mutations are nonsense or truncating frameshift mutations, occurring at two hotspots (around codon 1061 and 1309) at the N-terminal part and thereby removing most of the β-catenin regulatory domains. However, there are exceptions and different germline mutations cause different phenotypes. Moreover, it is important to recognize that genotype-phenotype correlations are imperfect because there is variation among and within families with identical mutations. Thus, there is no consequence yet for differential treatment of patients, which is still based on colonoscopic findings in individual patients and consists of removal of the whole colon (Friedl et al. 2001).

The CHRPE (congenital hypertrophy of retinal pigment epithelium) phenotype is the most common extraabdominal manifestation occurring in about 60% of FAP patients and can be used to identify patients at risk. CHRPE is only present in patients with mutations between codons 457 and 1444. Desmoid tumors are limited with mutations between codon 1445 and 1578. The clinical association of FAP with desmoid tumors and osteomas is called Gardner syndrome, while association of FAP with brain tumors is referred to as Turcot syndrome.

An often neglected clinical picture of FAP is AFAP, an attenuated form of FAP. It is characterized by development of only a few, often "flat", adenomas (<100) and a rather late onset of carcinoma formation. Interestingly, this attenuated phenotype is associated with germline mutations that result in ultra-short or very long APC proteins (Hernegger et al. 2002). Apart from the classical FAP, two germline missense variants responsible for inherited predisposition to colorectal cancer have been described: the APC variant I1307 K in 6% of the Ashkenazi Jews and the rarer E1317Q in Caucasians (Lamlum et al. 2000). These mutations result in APC proteins with amino acid substitutions in functionally critical areas.

Recent work by Fodde (Albuquerque et al. 2002) and Tomlinson (Lamlum et al. 1999) has shown that both the position and type of the second hit in the APC gene in FAP polyps depend on the localization of the germline mutation. For instance, when the germline mutations result in truncated proteins without any of the seven β-catenin downregulating 20-amino-acid repeats, the majority of the corresponding somatic point mutations retain one or two of these repeats, and vice versa, when the germline mutation results in a truncated protein retaining one 20-amino-acid repeat, most second hits remove all 20-amino-acid repeats (Albuquerque et al. 2002). This indicates selection for APC genotypes that might retain some activity in downregulating β-catenin signaling, probably because the complete abolishment of β-catenin is incompatible with tumor growth. Indeed, excess levels of β-catenin have been shown to induce apoptosis (Kim et al. 2000). By mathematical modeling it was suggested that the combinations of APC mutations are correlated with different growth advantages of the tumors (Cheadle et al. 2002).

APC mutations in sporadic carcinomas

About 80% of sporadic colorectal carcinomas and cell lines harbour APC mutations whereas in only 10–15% is β-catenin mutated (Ilyas et al. 1997; Sparks et al. 1998; Gayet et al. 2001). As in germline mutations, most of the mutations in the APC gene are nonsense or frame shift mutations leading to a truncated APC protein. About 60% of these mutations are clustered in a 700 bp "mutation cluster region" corresponding to the β-catenin/axin-binding domain. Mutation analysis of 58 sporadic colon cancers showed that APC mutations are less frequent in highly aggressive Dukes's D colon carcinomas than in carcinomas of the other Dukes stages (De Filippo et al. 2002).

The high prevalence of APC mutations in CRC and, in particular, their occurrence in premalignant lesions such as early adenomas makes identification of these mutations an interesting strategy for diagnosis and early detection of CRC. Several attempts with promising results have been made to detect APC mutations in the stool of patients using PCR amplification (Ransohoff 2002; Traverso et al. 2002).

Apart from APC mutations hypermethylation of the APC promoter at CpG sites was detected in CRC. This hypermethylation occurred rarely in FAP samples and more frequently in sporadic CRC at the wild-type allele, and was associated with a loss of transcription of this allele. Thus, hypermethylation may constitute an alternative mechanism for APC gene inactivation (Hiltunen et al. 1997; Esteller et al. 2000).

Mutations of β-catenin

Mutations in the β-catenin gene have been detected in about 50% of CRC without APC mutations, representing less than 10% of all CRC. It has been suggested that these mutations occur mainly in microsatellite instable tumors (Kitaeva et al. 1997; Sparks et al. 1998). Apart from the popular min mouse model for colon tumorigenesis, a variety of other colon cancer models exist. In one of these, colon tumors are induced in rats by the colon-specific carcinogen azoxymethane. Analysis of these colon tumors revealed mutations of exon 3 of β-catenin in 39.3% of early dysplastic lesions and 56.8% of colon cancers. Remarkably, as in human tumors, β-catenin mutations in the carcinomas were mainly detected at codons encoding functionally important residues involved in β-catenin degradation, while mutations in early dysplastic lesions were less focused and spread over the entire exon 3. The authors therefore speculate that the activating mutations of β-catenin are selected during colon carcinogenesis (Yamada et al. 2003).

Colorectal adenomas were the first tumors in which nuclear localization of β-catenin was demonstrated (Inomata et al. 1996). Later, it was shown that the nuclear staining for β-catenin on tissue section did not always coincide with activation of the Wnt pathway, and that it often showed a heterogenous pattern. In particular, immunolocalization of β-catenin in colon carcinomas showed a strong nuclear enrichment at the invasion front, whereas in large parts of the central tumor area β-catenin was only detected in the cytoplasm and at the membrane. This indicates that high levels of nuclear β-catenin play a role in the transition to the invasive state of tumor cells and, furthermore, point to regulatory mechanisms that control nuclear enrichment possibly independent of Wnt signaling. One can speculate that signals coming from the mesenchyme surrounding the invasive tumor cells might have a role here (Brabletz et al. 2002).

A matter of debate has been whether APC or β-catenin mutations play a role in mismatch repair deficient tumors. Mutation analysis of HNPCC (hereditary nonpolyposis colorectal cancer) tumors showed a lower frequency of APC gene mutation than in non-HNPCC tumors (21% vs >70%) (Miyaki et al. 1999). Surprisingly, β-catenin mutations showed a remarkably high frequency of 43% in these tumors, all occurring at regulatory sites. This analysis proposes that activation of Wnt signaling occurs concomitantly with mismatch-repair deficiencies. Similar results were obtained in analysis of sporadic colon tumors with high and low microsatellite instability. APC mutations were significantly more frequent in microsatellite stable than in instable tumors. However, in this analysis β-catenin mutations were only detected in two cases (Lovig et al. 2002). Mismatch repair-deficient CRC also showed heterozygous mutations in the Axin2 gene which resulted in truncated protein with a dominant-negative action (Liu et al. 2000). Frameshift mutations in error-prone nucleotide repeats were also found in the TCF-4 gene in these tumors; however, cell biological analysis suggests that these mutations had no effect on transcriptional activity, suggesting that they are not relevant for tumor formation (Ruckert et al. 2002).

Mutations in other tumors

The Wnt-pathway has been strongly implicated in the pathogenesis of hepatocellular carcinomas (HCC) and hepatoblastomas (HB). Activating mutations in the β-catenin gene were found in a high proportion (up to 70%) in hepatoblastomas which represent early childhood liver tumors, and to a lower extent (about 25%) in hepatocellular carcinomas (de La Coste et al. 1998; Koch et al. 1999; Park et al. 2001; Taniguchi et al. 2002). A somewhat higher proportion of β-catenin mutations was found in HCC associated with hepatitis C virus infection (Huang et al. 1999). In addition, mutations in the Axin and Axin2 genes resulting in truncated proteins were also found in about 10% of HCCs and HBs and in several HCC cell lines (Satoh et al. 2000; Taniguchi et al. 2002).

Gastric adenomas and adenocarcinomas have been investigated in several recent studies. Gastric adenomas might occur in FAP patients, and it appears that the loss of all APC SAMP repeats is necessary for duodenal and gastric tumorigenesis in FAP patients (Groves et al. 2002). APC mutations are also present in a number of sporadic gastric adenomas (76%) whereas there are contradictory reports about the frequency of β-catenin mutations (Groves et al. 2002; Lee et al. 2002). However, in gastric carcinomas only a few APC mutations have been detected (Lee et al. 2002). In another study, analysis of β-catenin mutations revealed activating β-catenin mutations in about one-third of the analyzed samples in both diffuse and intestinal gastric carcinomas (Clements et al. 2002). In contrast, in a different study, β-catenin mutations were found mainly in intestinal gastric cancers (27% of the cases) but not in diffuse-type gastric cancer (Park et al. 1999). Mutations of β-catenin and APC are extremely rare in esophageal adenocarcinomas (Choi et al. 2000).

Gastrointestinal carcinoid tumors showed a particularly high fraction of nuclear β-catenin positive cases (80%) and of β-catenin mutations (38%) with a high preference for mutation of Ser 37 (Fujimori et al. 2001).

In the most frequent pancreatic carcinomas of ductal origin, no mutations of β-catenin were found (Gerdes et al. 1999). However, in subtypes of the rare, low-malignant pancreatic abnormality, solid-pseudopapillary tumors, activating β-catenin mutations have been detected with high frequency [18/20 tumors analysed, (Abraham et al. 2002a)]. Other non-ductal pancreatic neoplasms such as pancreatoblastomas—also connected with the Beckwith-Wiedemann-syndrome—as well as acinar cell carcinomas, contained activating β-catenin and truncating APC mutations (Abraham et al. 2001b; Abraham et al. 2002b; Kerr et al. 2002). A single mutation of β-catenin was found in 21 cases of carcinomas of the Ampulla of Vater (Kawakami et al. 2002).

In endometrial carcinomas, β-catenin mutations were found in about 40% of cases (Mirabelli-Primdahl et al. 1999). In this study no differentiation according to histological subtypes was performed. Extensive analysis of β- and γ-catenin expression, of LOH and promoter hypermethylation of the APC gene, as well as mutation analysis in the APC, β- and γ-catenin, Axin and Axin2/conductin genes has been performed in 95 endometrioid and 33 non-endometrioid endometrial carcinomas. The most interesting fact was that β-catenin mutations in exon 3 were associated with the endometrioid phenotype (14.9%) whereas no mutations were found in the non-endometrioid endometrial carcinomas. Apart from that no functionally relevant APC or Axin/Axin2 gene mutations were detected in this series (Moreno-Bueno et al. 2002).

In ovarian carcinomas, β-catenin localization to the nucleus was found in 11 cases (16% of all analysed tumors) and activating mutations were found in seven of these 11 cases. Interestingly, the β-catenin-positive tumors appeared to have a better prognosis (Gamallo et al. 1999). In another study, 14 of 45 ovarian endometrioid carcinomas carried mutations in β-catenin. Mutations in APC, Axin, and Axin2 were also found in single cases (Wu et al. 2001).

Prostate cancer has been shown to carry mutations in β-catenin, APC and, interestingly, in βTrCP (Voeller et al. 1998; Chesire et al. 2000; Gerstein et al. 2002). Since each of these alterations occurred in about 10% of cases and were mutually exclusive, activating mutations of the Wnt pathway appear to play a role in at least 30% of prostate tumors. Interestingly, the androgen receptor was shown to bind to β-catenin (Truica et al. 2000; Yang et al. 2002) and to transport β-catenin into the nucleus (Mulholland et al. 2002). Furthermore, transcriptional activity of the androgen receptor was augmented by β-catenin (Truica et al. 2000; Yang et al. 2002), while androgens can block TCF:β-catenin mediated transcription in prostate cancer cells (Chesire and Isaacs 2002). This indicates an interplay of Wnt signaling and androgen-dependent pathways which might be of relevance for prostate cancer development and progression. In transgenic animal models, stabilized β-catenin induced lesions reminiscent of prostatic intraepithelial neoplasia (Gounari et al. 2002).

Thyroid carcinomas occur in a small fraction (1–2%) of FAP patients (Cetta et al. 2000). In sporadic anaplastic thyroid carcinomas, immunofluorescence staining showed nuclear localization of β-catenin in 15 (42%) of the 36 samples and mutations in 19 (61%) of the 31 patients analyzed. In a further study of 145 thyroid tumor samples, membrane β-catenin expression was decreased in most of the adenomas and in all 115 carcinomas. Among carcinomas, reduced membrane β-catenin was associated with progressive loss of tumor differentiation, possibly reflecting the disturbance in cell adhesion following loss of β-catenin from the cadherin complex. Exon 3 mutations and nuclear β-catenin localization were restricted to poorly differentiated or undifferentiated carcinomas. Mutations were present in about 25% and 66% of these subtypes, respectively (Garcia-Rostan et al. 1999; Garcia-Rostan et al. 2001). There are further interesting aspects of these studies: first, multiple mutations in exon 3 were observed frequently in a given tumor, and, second, the majority of the mutations were not in the hotspots of the serine and threonine residues usually implicated in downregulation of β-catenin. For instance Lys 49 is most frequently mutated in these tumors and was recently shown to be a target of acetylation of β-catenin by CBP, which could inactivate β-catenin's transcriptional activity (Wolf et al. 2002).

Wilms tumor, being one of the most common childhood renal malignancies, has been shown to harbor β-catenin mutations (Koesters et al. 1999; Kusafuka et al. 2002; Maiti et al. 2000). An intriguing finding in these studies was the preference for mutation of β-catenin at codon 45, which occurred in more than 90% of the cases.

Although the Wnt pathway was originally implicated in the development of melanomas because of detection of β-catenin mutations in melanoma cell lines (Rubinfeld et al. 1997), subsequent analyses showed that these mutations are rather rare in most cell lines and primary tumors (Omholt et al. 2001; Pollock and Hayward 2002; Reifenberger et al. 2002). Nevertheless, immunohistological expression analysis showed frequent upregulation of cytoplasmic and nuclear β-catenin in melanomas [(Rimm et al. 1999) see also below] underlining the significance of Wnt pathway activation in these tumors. Moreover, functional experiments revealed that the Microphthalmia-associated transcription factor (MITF), a factor that is involved in melanocyte differentiation, is a direct target of TCF:β-catenin complexes and can rescue suppression of melanoma growth by dominant-negative TCF (Widlund et al. 2002).

One of the tumor types that show the highest frequency of activating β-catenin mutations (95%) is pilomatrixoma which derives from hair follicles (Chan et al. 1999), consistent with a major role of the Wnt pathway in hair follicle development (Gat et al. 1998; Huelsken et al. 2001).

Various mutations of Wnt pathway components have been found in medulloblastomas without showing a preference for a specific gene. Thus, mutations have been detected to a varying extent in APC, β-catenin, and axin (Zurawel et al. 1998; Huang et al. 2000; Dahmen et al. 2001; Koch et al. 2001; Baeza et al. 2003). A high proportion of craniopharyngioma of the adamantinomatous type harbor activating mutations of β-catenin and show nuclear localization (Sekine et al. 2002).

There is evidence that the Wnt pathway plays a role in several tumors of mesenchymal origin. The fact that desmoid tumors, which represent an infiltrative form of fibromatosis, occur in a subgroup of FAP patients stimulated the analysis of β-catenin also in sporadic cases. Nuclear localization of β-catenin was seen in virtually all of the specimens and subsequent mutational studies showed that the β-catenin and APC genes were mutated in about 50% and 25%, respectively, of the cases (Miyoshi et al. 1998; Tejpar et al. 1999). Cell biological studies showed a requirement of TCF:β-catenin signalling for growth of desmoid cell cultures (Li et al. 1998), and transgenic expression of β-catenin in mice led to development of aggressive fibromatosis (Cheon et al. 2002). Accumulation of β-catenin by immunohistochemistry was also seen in soft-tissue sarcomas of various types; however, β-catenin mutations were rarely detected (Kuhnen et al. 2000; Sakamoto et al. 2002). Analysis of osteosarcomas revealed a cytoplasmic/nuclear accumulation of β-catenin in 33 cases out of 47 samples, although mutation analysis failed to detect any mutations. Correlation with clinico-prognostic factors did not demonstrate any prognostic value (Haydon et al. 2002). β-catenin mutations were frequent in one rare example of vascular tumors, namely juvenile nasopharyngeal angiofibroma, with 75% of tumors showing mutations mainly in the Ser/Thr phosphorylation sites (Abraham et al. 2001a). Interestingly this type of tumor also occurs in some FAP patients.

Although recent data indicate that Wnt signaling plays a role in lymphocyte development (van de Wetering et al. 2002a), and β-catenin is activated in some leukaemia cell lines (Chung et al. 2002), there is only limited genetic evidence to date for involvement of this pathway in haematological tumors. In a few cases of T cell lymphoma, mutations of β-catenin in exon3, but outside the phosphorylation region, were found (Hoshida et al. 2002).

β-catenin as a prognostic marker?

There have been several analyses to determine whether β-catenin is useful as a prognostic marker in cancer disease. These analyses are somewhat complicated by the fact that one has to differentiate between membrane, cytoplasmic, and nuclear β-catenin staining. An immunohistochemical analysis of 217 non-small cell lung carcinomas revealed that increased expression of β-catenin predicts a favourable prognosis of patients with resected tumors; however, in this analysis samples were scored positive when β-catenin was present in the cytoplasm or at the plasma membrane, and negative when staining was lost at the membrane. (Hommura et al. 2002). Thus, loss of cell adhesion function rather than preservation of Wnt signaling activity might have correlated with poor prognosis in this study.

Tissue microarray-based analysis showed that in 106 malignant melanomas phosphorylated β-catenin expression is associated with poor outcome (Kielhorn et al. 2003). It was demonstrated that cytoplasmic/nuclear staining of total β-catenin was more frequent in primary lesions, whereas phosphorylated β-catenin, which could only be detected in the nucleus by a specific antibody, was more common in metastatic lesions. The overall survival rate at 5 years was 51% (membranous) versus 25% (phospho-nuclear-β-catenin). Interestingly, the authors point out that staining for phosphorylated nuclear β-catenin does not go along with staining with the general β-catenin antibody, i.e., the phospho-β-catenin expression levels did not correlate with the cytoplasmic/nuclear β-catenin staining with a conventional anti-β-catenin antibody. Therefore nuclear/cytoplasmic versus membrane staining for β-catenin is not sensitive enough to predict clinical outcome (Kielhorn et al. 2003). It should be noted that the detection of phospho-β-catenin in the nucleus is somewhat surprising and contradictory to the fact that phospho-β-catenin is not the stabilized form of β-catenin and should be degraded quickly. Recently, it has been shown that phosphorylation of β-catenin precedes its degradation and ubiquitination, and dephosphorylated β-catenin rather than total β-catenin has been detected strongly in the nucleus after Wnt stimulation of cultured cells (Staal et al. 2002; van Noort et al. 2002). Another study analysing a panel of 650 colorectal cancer specimens by tissue microarrays revealed that phospho-β-catenin is associated with a better prognosis (Chung et al. 2001). Only phospho-β-catenin and tumor stage were both independently predictive for overall survival. Again, in this analysis, phospho-β-catenin was mainly located in the nucleus. Evaluation according to nuclear expression of β-catenin, as detected with a general anti-β-catenin antibody, did not show any significance regarding survival, which is in line with a previous study (Gunther et al. 1998). Again, it appears unclear why phospho-β-catenin is found in the nucleus. Using a new technology for automated subcellular localization and quantification of protein expression in tissue microarrays, called rapid exponential subtraction algorithm (RESA), β-catenin expression was analysed in colon cancer. Applying a continuous assessment of β-catenin levels by automated quantification, the authors demonstrate that high amounts of nuclear β-catenin in colon cancer correlates with poor prognosis (Camp et al. 2002).