Adenomatous polyposis coli (APC): a multi-functional tumor suppressor gene

Germ-line mutations in the APC gene result in familial adenomatous polyposis (FAP), which is characterized by numerous polyps in the intestines (Groden et al., 1991; Kinzler et al., 1991). Mutations in APC have been found in ∼60% of sporadic carcinomas and adenomas as well (Powell et al., 1992). Genetic studies using mutant mouse models have demonstrated that mutations in the Apc gene are responsible for intestinal tumorigenesis (Fodde et al., 1994; Oshima et al., 1995; Su et al., 1992). Homozygous Apc mutations in mice lead to embryonic lethality (Ishikawa et al., 2003; Moser et al., 1995; Oshima et al., 1995), and conditional deletion of the gene in the adult mouse disrupts homeostasis not only in the intestines but also in other tissues (Andreu et al., 2005; Gounari et al., 2005; Sansom et al., 2004). These pieces of evidence have indicated that APC is essential for development and homeostasis and that its inactivation facilitates tumorigenesis (Fig. 1, and Tables 1 and 2).

One of the most important issues in studies of APC is to identify the cellular pathways responsible for tumorigenesis when it is mutated. APC is a multi-domain protein that contains binding sites for numerous proteins, including microtubules, the Wnt/Wg pathway components β-catenin and axin, the cytoskeletal regulators EB1 and IQGAP1, and the Rac guanine-nucleotide-exchange factor (GEF) Asef1. Most (∼60%) cancer-linked APC mutations occur in a region referred to as the mutation cluster region (MCR; Fig. 1) and result in C-terminal truncation of the protein (Beroud and Soussi, 1996). Because these truncations cause loss of the domains required for binding to β-catenin and to microtubules (Fig. 1), the interaction of APC with β-catenin or microtubules has been considered to be essential for its tumor suppressor activity. β-Catenin has a dual role in cells, functioning both in cell adhesion at adherens junctions and as a signal transducer that shuttles to the nucleus in response to Wnt signalling. Binding of APC to microtubules obviously has potential consequences for cytoskeletal architecture.

Loss of APC leads to accumulation of nuclear β-catenin, which activates the T-cell factor (TCF) and lymphoid enhancer factor (LEF) transcription factors (Barker and Clevers, 2006; Korinek et al., 1997; Morin et al., 1997; Reya and Clevers, 2005), the targets of the canonical Wnt signalling pathway. In a floxed β-catenin-knock-in mouse, Cre-mediated expression of stable β-catenin in the small intestine causes formation of numerous (700-3000/mouse) polyps morphologically similar to those found in the Apc-knockout mouse (Harada et al., 1999; Oshima et al., 1995). By contrast, suppression of canonical Wnt-dependent transcription by dominant-negative TCFs inhibits the proliferation of colon cancer cells that carry mutations in APC (Tetsu and McCormick, 1999; van de Wetering et al., 2002). Activation of the canonical Wnt pathway caused by APC inactivation is thus sufficient for intestinal tumorigenesis. However, it has been proposed that mutations in APC contribute to cancer development through processes other than Wnt signalling. Below, we discuss recent advances in this area, examining potential roles of both canonical Wnt targets and effects of APC on the cytoskeleton, cell migration and chromosome instability.

Both mammals and Drosophila carry two APC genes: APC and APC2 (APCL) in mammals, and dAPC1 and dAPC2/E-APC in Drosophila (Bienz, 2002; Polakis, 1997; van Es et al., 2001). Note, however, that Drosophila dAPC1 and dAPC2 are not direct orthologues of mammalian APC and APC2, respectively (Fig. 1). Human APC has 2843 residues whereas APC2 contains 2303 residues (Fig. 1). dAPC1 and dAPC2 have 2416 and 1067 residues, respectively. Mammalian APC has an oligomerization domain, an armadillo repeat-domain, a 15- or 20-residue repeat domain important for binding to β-catenin, SAMP repeats for axin binding, a basic domain for microtubule binding and C-terminal domains that bind to EB1 and DLG proteins (Fig. 1) (Polakis, 1997).

In mammals, APC is expressed in most fetal tissues and in adult epithelial cells (Midgley et al., 1997), whereas APC2 is expressed ubiquitously and at high levels in the brain (van Es et al., 1999). Embryonic stem (ES) cells that lack Apc lose their ability to differentiate (Kielman et al., 2002), and the homozygous Apc-knockout mutation is embryonic lethal (Ishikawa et al., 2003; Moser et al., 1995; Oshima et al., 1995). An Apc2 knockout mutation has not been reported. Although the heterozygous Apc mutant mice are viable, they develop a number of intestinal polyps that resemble those in the FAP patients (Fodde et al., 1994; Oshima et al., 1995; Su et al., 1992). Accordingly, these mutant mice, often called `Min' mice for multiple intestinal neoplasia (Moser et al., 1990; Su et al., 1992), have been widely used as a model for intestinal tumorigenesis (Boivin et al., 2003; Taketo, 2006). Conditional inactivation of APC in the adult mouse intestine leads to enlargement of the crypt compartment, which is associated with an increased cell population and apoptosis (Andreu et al., 2005; Sansom et al., 2004). It also perturbs differentiation and migration of epithelial cells. Expression of APC2 in colon cancer cells that lack wild-type APC inhibits Wnt signalling to a similar extent to that achieved by APC (van Es et al., 1999), although the roles of APC2 in tumorigenesis and homeostasis have been unclear. Apc2 might be unable to compensate for loss of the tumor suppressor activity of Apc, given that loss of Apc in the intestine stimulates cell proliferation and polyp formation (Andreu et al., 2005; Fodde et al., 1994; Oshima et al., 1995; Sansom et al., 2004; Su et al., 1992).

In Drosophila, dAPC and dAPC2 function in a partially redundant manner (Ahmed et al., 2002; Akong et al., 2002a; Akong et al., 2002b) – both can inhibit Wnt signalling in human cancer cells (Hamada et al., 1999; Hayashi et al., 1997). The dAPC1 gene is highly expressed in the central nervous system and the germ-line, whereas dAPC2 is broadly expressed (Ahmed et al., 1998). Homozygous dAPC1 mutants are viable, although they show neuronal degeneration and pigment cell hypertrophy of the retina (Ahmed et al., 1998; Hayashi et al., 1997). Because maternal/zygotic dAPC2 mutants die during embryogenesis owing to excessive Wg signalling (McCartney et al., 1999), dAPC2 must be essential for embryogenesis. Zygotic mutants of dAPC2 are viable, however, and dAPC1 therefore appears to compensate for lack of dAPC2 in regulation of Wg signalling in many tissues. Both dAPC1 and dAPC2 are thus required for normal development and homeostasis in Drosophila.

APC plays a central role suppressing the canonical Wnt signalling pathway that controls cell proliferation and differentiation in the intestine (van de Wetering et al., 2002), skin (Gat et al., 1998; Huelsken et al., 2001; Nguyen et al., 2006), immune system (Cobas et al., 2004; Gounari et al., 2005; Kirstetter et al., 2006; Scheller et al., 2006), bone (Glass, 2nd et al., 2005; Hill et al., 2005) and brain (Brault et al., 2001; Kleber et al., 2005). When this pathway is stimulated, β-catenin activates TCF-dependent transcription of Wnt-target genes such as CCND1 (Tetsu and McCormick, 1999), MYC (He et al., 1998) and EphB (Batlle et al., 2002).

Several lines of evidence indicate that APC inhibits β-catenin/TCF-dependent transcription through multiple mechanisms. First, APC provides a scaffold for a destruction complex together with GSK3β and axin that promotes phosphorylation and consequent ubiquitin-dependent degradation of β-catenin (Rubinfeld et al., 1996). Second, APC promotes export of β-catenin from the nucleus, which reduces the amount of nuclear β-catenin/TCF (Henderson and Fagotto, 2002; Neufeld et al., 2000; Rosin-Arbesfeld et al., 2003). Third, APC binds to β-catenin, blocking the interaction with TCF (Neufeld et al., 2000; Rosin-Arbesfeld et al., 2003).

Finally, recent studies show that APC can inhibit β-catenin/TCF-dependent transcription through a direct interaction with a repressor complex. APC binds to βTrCP and transcriptional repressor C-terminal binding protein (CtBP) and forms a stable complex with additional co-repressors TLE-1 and HDAC1 (Hamada and Bienz, 2004; Sierra et al., 2006). Because APC interacts with βTrCP at the MYC enhancer, it may facilitate CtBP-mediated repression of Wnt-target genes (Hamada and Bienz, 2004; Sierra et al., 2006). Because the C-terminal half of APC mediates binding to CtBP, truncated APC mutants found in cancer cannot stimulate the CtBP-mediated transcriptional repression (Hamada and Bienz, 2004; Sierra et al., 2006). Such mutations thus activate β-catenin/TCF-dependent transcription by both increasing levels of β-catenin-TCF complexes in the nucleus and alleviating CtBP-mediated inhibition of the complex (Fig. 1 and Table 1).

Inactivation of APC has also been proposed to promote tumorigenesis through loss of cell adhesion (Bienz and Hamada, 2004; Birchmeier et al., 1995). APC is found at the lateral plasma membrane of mammalian epithelial cells both in vivo and in culture (Miyashiro et al., 1995; Nathke et al., 1996), and it interacts with β-catenin, which links E-cadherin to α-catenin and the actin cytoskeleton (Kemler, 1993; Rubinfeld et al., 1993; Su et al., 1993). A mutation in Apc in mouse intestinal epithelial cells can decrease the level of E-cadherin at the cell membrane and association between β-catenin and E-cadherin (Carothers et al., 2001). Moreover, expression of the full-length APC leads to increased levels of E-cadherin at the cell membrane, which enhances cell adhesion in colon cancer cells in which truncated APC is expressed (Faux et al., 2004). In addition, expression of the full-length APC in the cancer cells also causes nuclear and cytoplasmic β-catenin to translocate to the cell periphery (Faux et al., 2004). APC thus appears to affect cell adhesion by controlling the distributions of β-catenin and E-cadherin between the cytoplasm and the cell membrane (Tables 1 and 2).

In Drosophila, dAPC2 is also found in apical adherens junctions, and an inactivating mutation in dAPC2 causes mislocalization of oocytes in the egg chambers due to failure of cadherin-based adhesion (Hamada and Bienz, 2002). Although these results are consistent with the studies of mammalian cultured cells mentioned above (Faux et al., 2004), Drosophila that lack both dAPC1 and dAPC2 yield strikingly different results (McCartney et al., 2006). Complete loss of the two dAPC proteins elevates the levels of cytoplasmic β-catenin in the germ and follicle cells of ovaries, and embryonic epithelial cells. However, loss of both dAPC proteins in these tissues does not substantially affect the cortical localization of β-catenin, levels of E-cadherin, or cadherin-based adhesion (McCartney et al., 2006). More recent studies show that simultaneous deletion of both dAPC proteins in the brain does not reduce the levels of cadherins (Hayden et al., 2007). Since these results cannot be explained in terms of the predicted functions of APC, further investigation will be needed to determine its contribution to cell adhesion.

APC has been shown to regulate cell polarity and migration through control of the actin cytoskeleton (Akiyama and Kawasaki, 2006). Forced expression of APC in the small intestine induces disordered migration of epithelial cells (Wong et al., 1996). By contrast, loss of APC slows down their migration (Andreu et al., 2005; Sansom et al., 2004) and causes formation of a nascent polyp that consists of a single layer of adenoma cells (Oshima et al., 1997). In addition, the C-terminally truncated mutant of APC decreases migration of intestinal epithelial cells (Mahmoud et al., 1997). These studies suggest that expression of APC or its truncated form affects cell migration.

APC interacts with Asef, a Rac-specific GEF (Kawasaki et al., 2003; Kawasaki et al., 2000) and IQ-motif-containing GTPase activation protein 1 (IQGAP1) (Noritake et al., 2005; Watanabe et al., 2004). The interaction between APC and Asef1 was initially shown to stimulate the GEF activity of Asef1 and thus lead to Rac1 activation, cell flattening, membrane ruffling, formation of lamellipodia and increased cell migration (Kawasaki et al., 2003; Kawasaki et al., 2000). A recent analysis has shown that APC also interacts with Asef1 relative Asef2 (Hamann et al., 2007). This work also shows that both Asef proteins have GEF activities towards lipid-modified recombinant Cdc42 (which mimics post-translationally modified Cdc42) but not its unmodified form. By contrast, they fail to promote detectable GEF activity of Asefs towards lipid-modified Rac1, despite the fact that both can bind to Rac1 (Hamann et al., 2007). These results suggest that APC activates Cdc42 rather than Rac1 through its interactions with Asef1 and Asef2. Because APC interacts with Asef1 through its armadillo repeat domain, which is N-terminal to the MCR (Fig. 1), truncated APCs in most colon cancer cells should still interact with Asef1 (Kawasaki et al., 2003). Interestingly, C-terminally truncated APC stimulates Asef1 more strongly than the full-length APC does (Kawasaki et al., 2003). Expression of the truncated APC mutants could thus induce aberrant migration and invasion of cancer cells through Asef1 and Asef2. However, mutations in the Apc gene alone do not lead to invasive carcinoma, although an additional mutation in Smad4, which encodes a TGF-β-responsive transcription factor, promotes invasion (Kitamura et al., 2007; Oshima et al., 1995; Takaku et al., 1998). Therefore, aberrant cell migration caused by truncated APC is probably not sufficient to cause progression to malignancy.

Another key link between APC and the cytoskeleton is IQGAP1 (Watanabe et al., 2004), an effector of Rac1 and Cdc42 (Briggs and Sacks, 2003). IQGAP1 is a scaffold protein that controls actin filaments and microtubules (Briggs and Sacks, 2003). It recruits APC to specific sites in migrating cells, at which they form a complex with the microtubule-stabilizing protein CLIP-170, and with activated Rac1 or Cdc42 (Watanabe et al., 2004). Depletion of either APC or IQGAP1 inhibits formation of the actin meshwork and polarized migration, and also causes mislocalization of CLIP-170 (Watanabe et al., 2004). Because these results have been obtained mainly in a cell line, Vero, that was isolated from monkey kidney epithelial cells, further investigation will clarify the significance of the interaction in cells of other tissues.

Note that APC might also bind to actin directly through its C-terminal domain, which mediates binding to microtubules via EB1 (see below) (Moseley et al., 2007; Polakis, 1997). Because the interaction of APC with EB1 can disrupt its interaction with actin, EB1 could control the APC-mediated regulation of actin and microtubules (Moseley et al., 2007). Because these data are based on overexpression experiments, however, additional studies such as loss-of-function experiments are needed to establish whether there is indeed a direct link between APC and actin.

APC is found at the ends of microtubules (Nathke et al., 1996) and can bind to and stabilize them (Munemitsu et al., 1994; Smith et al., 1994). These results indicate that APC regulates the microtubule network and may play roles in microtubule-mediated processes such as cell migration and spindle formation (Nathke, 2006; Nathke, 2004). Indeed, recent studies using siRNA targeting APC have shown that depletion of APC inhibits cell migration and protrusion formation (Kroboth et al., 2007). Knocking down APC decreases overall microtubule stability and the level of post-translationally modified microtubules at the migrating edge of the cell (Kroboth et al., 2007). APC thus appears to contribute to cell migration by regulating microtubules as well as the actin cytoskeleton (Fig. 1).

Stabilization and polymerization of microtubules by APC are mediated also by interaction with EB1 (Morrison et al., 1998; Nakamura et al., 2001; Su et al., 1995). Recent results have shown that APC and EB1 function downstream of Rho and the formin family protein mDia in microtubule stabilization (Wen et al., 2004). Because APC may form a complex with EB1 and mDia at stable microtubule ends, the involvement of APC in cell migration could encompass cooperation with mDia as well (Wen et al., 2004). APC therefore seems to stimulate cell migration through several pathways (Fig. 1).

Chromosome instability (CIN) is considered to be one of the driving forces that stimulate tumorigenesis (Sotillo et al., 2007; Weaver et al., 2007). Some studies have shown that loss of APC induces mis-segregation of chromosomes (Tables 1 and 2) (Alberici et al., 2007; Aoki et al., 2007; Dikovskaya et al., 2004; Dikovskaya et al., 2007; Fodde et al., 2001; Green and Kaplan, 2003; Green et al., 2005; Hadjihannas et al., 2006; Kaplan et al., 2001; Tighe et al., 2001; Tighe et al., 2004); others have produced conflicting results (Draviam et al., 2006; McCartney et al., 2006; Sieber et al., 2002).

In mitosis, APC localizes to kinetochores, spindles and centrosomes (Dikovskaya et al., 2004; Kaplan et al., 2001; Louie et al., 2004; Olmeda et al., 2003). Truncating mutations in the Apc gene can cause CIN, which is evident from aneuploidy and chromosomal aberrations in mutant mouse embryonic stem (ES) cells and intestinal polyps (Alberici et al., 2007; Aoki et al., 2007; Dikovskaya et al., 2007; Fodde et al., 2001; Kaplan et al., 2001). In addition, depletion of APC has been shown to compromise the formation of spindles in Xenopus extracts (Dikovskaya et al., 2004). Expression of a C-terminal fragment of APC (containing the microtubule-binding region) can also cause aneuploidy in a human colon cancer cell line (HCT116) that expresses the full-length APC (Fodde et al., 2001). In addition, expression of an N-terminal fragment of APC can weaken kinetochore-microtubule interactions and induce CIN in HCT116 cells (Tighe et al., 2001; Tighe et al., 2004). APC thus seems to be involved in kinetochore function and chromosome segregation and this probably involves its interaction with microtubules (Fodde et al., 2001). Truncated forms of APC may therefore have dominant-negative effects on chromosome segregation, although the mechanism is unclear.

The taxanes, including taxol, produce unattached kinetochores by inhibiting microtubule assembly and spindle formation, which leads to activation of the mitotic spindle checkpoint and accumulation of cells at prometaphase of the cell cycle (Weaver and Cleveland, 2005). Because treatment of homozygous Apc-mutant ES cells with taxol leads to accumulation of cells in G2/M phase, the spindle checkpoint seems to work properly, which suggests that the checkpoint itself is unaffected by inactivation of APC (Fodde et al., 2001; Kaplan et al., 2001). However, further investigation has indicated that APC does have a role in the spindle checkpoint (Tables 1 and 2). Studies using siRNA directed against APC have shown that knocking down APC in U2OS and HCT116 cells induces polyploidy (Dikovskaya et al., 2007). Depletion of APC in these cells reduces levels of checkpoint proteins Bub1 and BubR1 at the kinetochores and promotes mitotic progression despite spindle damage caused by the depletion (Dikovskaya et al., 2007). Because loss of APC in these cells inhibits apoptosis as well, Dikovskaya et al. have proposed that kinetochore dysfunction and inhibition of apoptosis collaborate to promote CIN in APC-mutant cells (Dikovskaya et al., 2007).

Another analysis using siRNA to knock down APC mRNA, however, found that inhibition of APC in HeLa cells does not affect chromosome instability significantly (Draviam et al., 2006). In most of the APC-depleted cells, chromosomes congress correctly at the spindle equator and anaphase ensues with little or no delay. In addition, high levels of the mitotic spindle checkpoint proteins Mad1, Mad2, Bub1 and BubR1 bind to the kinetochore (Draviam et al., 2006). These results suggest that inactivation of APC does not interfere with the operation of the spindle checkpoint or cause dramatic changes in the extent or stability of kinetochore-microtubule interactions (Draviam et al., 2006), contradicting the results above (Dikovskaya et al., 2007). Although further studies will be needed to settle this issue, it is conceivable that the observed effects of reduced levels of APC on chromosome segregation depend on the cell type and/or the method of detection.

A role for APC in chromosome segregation has also been suggested in Drosophila. dAPC2, armadillo (fly β-catenin) and α-catenin localize to the sites of cortical spindle attachment, and a mutation in either dAPC2 or β-catenin results in the loss of nuclei (McCartney et al., 2001). However, recent studies have shown that deletion of both APC1 and APC2 has no effect on spindle assembly, spindle orientation or asymmetric division in the ectoderm during gastrulation (McCartney et al., 2006), which is consistent with the studies in HeLa cells discussed above (Draviam et al., 2006). Moreover, segregation of chromosomes is unaffected by expression of truncated APC in Drosophila (Draviam et al., 2006). Its truncated form therefore seems not to play a dominant-negative role in chromosome segregation, which also argues against the results obtained with human cancer cells (Tighe et al., 2001; Tighe et al., 2004).

We and others have found recently that activation of the canonical Wnt signalling pathway can cause CIN in mouse ES and human colon cancer cells (Aoki et al., 2007; Hadjihannas et al., 2006). Expression of stable β-catenin induces CIN in both ES cells and mouse intestinal polyps, whereas expression of dominant-negative TCF in these cells reduces the anaphase bridge index, a marker for CIN (Aoki et al., 2007; Gisselsson et al., 2000). Mutations in APC could therefore contribute to CIN through activation of the canonical Wnt signalling pathway (Fig. 1). Wnt signalling in mouse ES cells increases the fraction of surviving cells upon stimulation of the G2/M checkpoint by exposure of cells to nocodazole or colcemid. In addition, Wnt signalling suppresses the Cdc2 kinase activity that induces apoptosis at the G2/M checkpoint (Aoki et al., 2007; Tan et al., 2002). It is thus possible, through suppression of cyclin-B–Cdc2, that activation of canonical Wnt-signal-dependent transcription can inhibit apoptosis of cells even when they carry chromosomal aberrations. Hadjihannas et al. also showed that expression of conductin/axin2, a target of TCF/β-catenin-dependent transcription, induces CIN by compromising the spindle checkpoint (Hadjihannas et al., 2006). Wnt signalling could thus also induce CIN through increased levels of conductin.

Inactivation of APC has been recognized as one of the most important genetic alterations in colon carcinogenesis, and activation of the canonical Wnt signal plays an essential role in tumorigenesis caused by APC mutations. However, APC plays additional roles in other cellular processes, such as cell migration, adhesion, and chromosome segregation. Although elevated cell migration, chromosome instability and loss of cell adhesion are possible contributing factors to tumorigenesis and progression, it is still unclear whether mutations in APC accelerate tumorigenesis through such mechanisms. To understand the precise mechanisms of colon cancer progression, further studies are needed to delineate the effects of canonical Wnt signal activation and those of other APC-dependent mechanisms.

The research programs in our laboratory have been supported by grants from MESSC, Japan; OPSR, Japan; University of Tokyo-Banyu Pharmaceutical Co. Joint Fund; Takeda Foundation, Mitsubishi Foundation, and Sagawa Cancer Research Fund. K.A. is a JSPS research fellow and supported by Grants-in-Aid for Scientific Research of JSPS.