We report the first direct analysis of the endogenous β-catenin phosphorylation activity in colon cancer SW480 cells. By comparing parental SW480 cells that harbor a typical truncated adenomatous polyposis coli (APC) form, cells expressing full-length APC and APC-depleted cells, we provide the formal demonstration that APC is necessary for β-catenin phosphorylation, both for priming of the protein at residue serine 45 and for the subsequent phosphorylation of residues 33, 37 and 41. Truncated APC still sustains a surprisingly high phosphorylation activity, which requires the protein to bind to β-catenin through the APC 20-amino-acid (20AA) repeats, thus providing a biochemical explanation for the precise truncations found in cancer cells. We also show that most of the β-catenin phosphorylation activity is associated with a dense insoluble fraction. We finally examine the impact of full-length and truncated APC on β-catenin nuclear transport. We observe that β-catenin is transported much faster than previously thought. Although this fast translocation is largely insensitive to the presence of wild-type or truncated APC, the two forms appear to limit the pool of β-catenin that is available for transport, which could have an impact on β-catenin nuclear activities in normal and cancer cells.
The Wnt–β-catenin pathway is a major signaling route that controls embryonic patterning and tissue homeostasis. Its deregulation is involved in many cancers. The pathway is in particular over-activated in virtually all colon cancer because of mutations of the adenomatous polyposis coli (APC) tumor suppressor gene, which could actually represent the initiating event for this type of cancer (Polakis, 2007). The pathway revolves around β-catenin, which, among many other functions, is responsible for the transduction of Wnt signals into gene regulation through its ability to act as a transcriptional coactivator (Valenta et al., 2012). β-catenin appears to be controlled in the cytoplasm by a complex based on the scaffold protein axin. In the absence of Wnt signaling, β-catenin is inactivated by the axin complex (also called the ‘β-catenin destruction complex’). Soluble β-catenin is captured by axin and is sequentially phosphorylated, first by casein kinase 1 (CK1) on serine residue 45. This phosphorylation serves as a priming event for the subsequent action of glycogen synthase 3 (GSK3) on three consecutive residues, threonine 41, serine 37 and serine 33. N-terminally phosphorylated β-catenin is then ubiquitylated and rapidly degraded. Upon activation of the pathway by the binding of Wnt ligand to Frizzled and LRP5–LRP6 receptors, the axin complex is inhibited by a mechanism that remains poorly understood (Li et al., 2012; Roberts et al., 2012; Taelman et al., 2010). This results in the accumulation of soluble β-catenin that can enter the nucleus, where it interacts with transcription factors of the TCF/LEF1 family to regulate a series of target genes (Valenta et al., 2012).
Behind this seemingly simple picture of the Wnt pathway, the actual mechanisms that regulate β-catenin remain highly controversial (Hernández et al., 2012; Li et al., 2012; Roberts et al., 2012; Taelman et al., 2010). The role of APC in particular is unclear, and the consequences of the mutations found in cancer cells are still poorly defined. It is, however, well established that the recruitment of the two kinases CK1 and GSK3 and their substrate β-catenin within a single complex strongly increases the efficiency of the reaction. It is thus commonly accepted that axin, CK1 and GSK3 constitute the minimally required ‘core complex’ for β-catenin degradation. Many studies have shown that APC is also essential, because β-catenin accumulates when APC is mutated or depleted (Munemitsu et al., 1995). That the function of APC is associated with the activity of the axin complex is strongly suggested by the ability of APC to associate with both β-catenin and axin (Fagotto et al., 1999; Hart et al., 1998; Hinoi et al., 2000; Kishida et al., 1998; Rubinfeld et al., 1993; Su et al., 1993). APC binds directly to β-catenin through two different types of short repeats in the APC protein, called 15- and 20-amino-acid (15AA and 20AA, respectively) repeats. The affinity of the 20AA repeats for β-catenin is strongly increased by the phosphorylation of APC (Ha et al., 2004; Rubinfeld et al., 1996). APC also binds directly to axin, through short ‘SAMP’ motifs (Behrens et al., 1998), and indirectly through the Armadillo (Arm) repeats (Roberts et al., 2011). APC has therefore been considered to be a bona fide constituent of the destruction complex (Ha et al., 2004; Hinoi et al., 2000; Xing et al., 2003). In one model, axin and APC are thought to act as coordinate scaffolds that ensure the specificity of β-catenin phosphorylation and its regulation by the Wnt pathway. In vitro experiments using pure recombinant proteins have indeed demonstrated that APC further increases the efficiency with which the axin–GSK3 complex phosphorylates β-catenin (Hinoi et al., 2000). The presence of both low- and high-affinity β-catenin-binding sites in the APC protein led to a refined version of this model, which states that different sites are used depending on the β-catenin levels (Ha et al., 2004). The fact that phosphorylated 20AA repeats compete with axin for binding to β-catenin (Ha et al., 2004) suggested a different model, in which APC helps phosphorylated β-catenin to dissociate from axin, creating a catalytic cycle of binding and release of the substrate (Kimelman and Xu, 2006). Others have suggested that APC acts either upstream of the phosphorylation reactions, by gathering or even transporting cytosolic β-catenin to the complex (Bienz, 2002), or downstream of the phosphorylation reactions, by recruiting the ubiquitin ligase βTrCP (FBXW11) to the complex (Li et al., 2012; Su et al., 2008). A final interesting suggestion is that APC might function both in β-catenin phosphorylation and in its subsequent release from the complex (Roberts et al., 2011).
This uncertainty partly stems from the fact that the central process in this pathway, the regulation of β-catenin phosphorylation, has only been studied in vitro using purified proteins or inferred from observations of steady-state levels, rather than by direct measurement of the endogenous kinase activity. The in vitro data, although demonstrating the role of axin and APC in making β-catenin phosphorylation more efficient, also left open a relatively wide range of possible reactions. For instance, GSK3 could phosphorylate β-catenin even in the absence of APC, CK1 or axin (Hinoi et al., 2000; Yost et al., 1996), and it has been shown that APC has some enhancing effect on β-catenin phosphorylation even in the absence of axin (Hinoi et al., 2000). Whether APC acts on the priming reaction has also not been tested. A direct transposition of these in vitro data to the in vivo situation is far from straightforward because we still know very little about the actual concentrations, activities, associations and localization of the endogenous components. This raises the question of whether all of the components are required in vivo for the entire process, or whether different partial complexes might be in charge of distinct stages within the process. Different complexes might also be active under different conditions [for instance when there are low basal β-catenin levels or high β-catenin levels during Wnt stimulation (Ha et al., 2004)], or even in different cellular compartments.
Relating the in vitro and in vivo situations would require an investigation of specific reactions under endogenous conditions. Measurement of the levels of endogenous phosphorylated β-catenin and the detection of phosphorylated β-catenin at particular cellular locations is clearly not sufficient, and can be interpreted in opposite ways [e.g. local enrichments could be considered as sites of stabilized β-catenin (Faux et al., 2010)]. A few studies have measured GSK3 activity in the context of the Wnt pathway, but used substrates that were irrelevant to the pathway (Stambolic et al., 1996; Taelman et al., 2010). Such measurements almost certainly included GSK3 activity that was independent of the axin–APC complex. It is even possible that the complex might exclude or at least be poorly accessible to substrates other than β-catenin. Thus, none of the available data provide adequate information about the actual function of the pathway. In addition, it is still unclear whether APC is an intrinsic component of the machinery or merely a modulator. Axin has been suggested to be the limiting factor in the pathway, based on measurements of the relative concentrations in Xenopus egg extracts (Lee et al., 2003) [but see work by Tan et al. for a contrasting view (Tan et al., 2012)], and axin overexpression has been found to rescue the reduced level of β-catenin signaling in APC-mutated cancer cells, suggesting that APC might be dispensable when axin levels are sufficiently high (Behrens et al., 1998; Cliffe et al., 2003; Faux et al., 2008; Hart et al., 1998; Nakamura et al., 1998).
Considering the many unknowns in APC biology, it comes as no surprise that the precise effects of the mutations found in colon cancers are similarly unclear. The overwhelming majority of the mutations that have been identified so far in both sporadic (Kinzler and Vogelstein, 1996) and familial colon cancers lead to early protein termination, and thus result in a truncated protein. Although recessive, these mutations are not random – many are located in a small region in the middle of the APC coding sequence, indicating that during cancer development there is selection for cells that maintain the expression of a protein with an intact N-terminal half (Furuuchi et al., 2000). APC is a very large (>300 kDa) and complex protein, comprising multiple domains that interact with a variety of cellular components, and it has been implicated in several different cellular processes, from transcriptional regulation to mitotic-spindle positioning and cell migration (Näthke, 2004). However, the loss of function that is a result of the deletion of its C-terminus in cancer cells has been definitively linked to the Wnt pathway (Polakis, 2000). Most truncations remove all but the first of the seven 20AA repeats, as well as the axin-binding SAMP repeats (Fig. 1A; Kohler et al., 2008). The loss of most of the high-affinity β-catenin-binding sites and/or loss of the ability to bind to axin were thus prime suspects for the abnormal accumulation of soluble non-phosphorylated β-catenin (and for the activation of β-catenin transcriptional targets) that is observed in colon cancer cells. These models have been challenged, however, and alternative hypotheses have been proposed, including downstream effects on nuclear localization, retention or transcription (Bienz, 2002; Krieghoff et al., 2006; Sierra et al., 2006).
As a first step to attempt to clarify some of these issues, here, we report the results of a kinase assay to monitor endogenous activities for β-catenin S45 priming and for S33, S37 and T41 (S33/S37/T41) phosphorylation. This study proposes to address the following basic questions: is APC required for β-catenin phosphorylation, and, if so, for which of the two kinase reactions, and what is the effect of APC truncation on β-catenin phosphorylation?
Another important and still poorly explored question is the subcellular location of the active complexes that are responsible for β-catenin phosphorylation. Besides the absence of direct information on enzymatic activity, even the localization of the components of the complex has not been solidly established. The detection of endogenous proteins by immunofluorescence has suffered from the lack of specific antibodies (for example, see Brocardo et al., 2005), whereas exogenously expressed constructs tend to aggregate (for example, see Fagotto et al., 1999; Faux et al., 2008). In addition, soluble cytosolic components are known to leak during fixation (Liu and Fagotto, 2011). By contrast, studies using cell fractionation have been plagued by the systematic co-purification of nuclear and plasma-membrane insoluble fractions, and by the omission of adequate markers to validate the identity of the fractions (Liu and Fagotto, 2011). We have recently established a fractionation protocol that cleanly separates the major cellular components that might be involved in β-catenin regulation (Liu and Fagotto, 2011). Here, we use this protocol in combination with our kinase assay to compare the activity of the various compartments.
The second key process that has been investigated in this study is β-catenin nuclear transport. There have been conflicting reports about a possible role of APC in the nuclear localization of β-catenin. APC has been proposed to mediate β-catenin export, carrying it by a ‘piggy-back’ mechanism, and it was suggested that the nuclear accumulation of β-catenin in colon cancer cells was due to the failure of truncated APC to perform this function (Bienz, 2002). However, the nuclear localization of truncated APC was later contested (Henderson and Fagotto, 2002), and it might be an artifact caused by unspecific antibody staining (Brocardo et al., 2005). An alternative mechanism was proposed in which β-catenin freely shuttles through the nuclear pore (Fagotto et al., 1998; Koike et al., 2004; Sharma et al., 2012). Kinetic analysis of transport shows that the overexpression of APC or other binding partners, such as axin, actually decreases the nuclear import of β-catenin (Krieghoff et al., 2006), supporting the hypothesis that these proteins influence the distribution of β-catenin through its sequestration in particular compartments (Krieghoff et al., 2006; Roberts et al., 2011). However, the retention of β-catenin in the cytoplasm might be due to the artificially elevated levels of APC, and endogenous APC might act differently. The effect of truncated APC on β-catenin nuclear translocation also remains unexplored. In this study, we therefore analyze β-catenin transport in cells expressing physiological levels of wild-type APC or truncated APC and in cells depleted of APC.
Characterization of components of the Wnt pathway in SW480 and SW480APC cells
In this study, we compared parental SW480 cells with an SW480 cell line stably expressing full-length APC (SW480APC) (Faux et al., 2004). We verified that SW480APC cells expressed relatively low levels of full-length APC (Fig. 1B, arrow). Several lower-molecular-mass fragments were also detected, including a major band, which, according to its migration, probably represented the endogenous truncated APC (Fig. 1B, arrowhead). Note that APC levels are controlled by proteasomal degradation – both wild-type and truncated forms are targets for ubiquitylation (Choi et al., 2004). Thus, one might not necessarily expect identical levels of the truncated form in the absence and presence of wild-type APC. Note also that the band representing truncated APC protein in SW480APC cells might include a cleavage product of the full-length protein, because similar fragments are commonly observed in various cell lines that contain wild-type APC (Kishida et al., 1998; Liu and Fagotto, unpublished). The signal was globally decreased in small interfering (si)RNA-transfected cells, demonstrating that all bands were related to APC (Fig. 2C). We also compared the levels of the major components of the axin complex (examples in Fig. 1B; quantification in supplementary material Fig. S1). Axin, GSK3 and CK1α were expressed at approximately similar levels, with the exception of GSKα, which was expressed at slightly higher levels in SW480APC. By contrast, steady-state levels of β-catenin were lower in SW480APC cells, consistent with the original report (Faux et al., 2004).
We used our cell-fractionation protocol to examine the subcellular distribution of these components. This protocol yields five fractions (Fig. 1B′,B″): cytosol (Cs), nucleosol (Ns), nuclear insoluble fraction (Ni), membranes (M) and dense insoluble material (X). The latter fraction is mainly composed of cytoskeletal components, with a minor contribution from nuclear material (Liu and Fagotto, 2011). Most components of the destruction complex were distributed in similar patterns in parental and APC-rescued cells (Fig. 1B′,B′′; quantification in supplementary material Fig. S1B). The bulk of axin, GSK3 and CK1α were found in the cytosol. By contrast, CK1ε was strongly enriched in fraction X. Full-length APC was mostly cytosolic, with a second significant pool in fraction X and low levels in the nucleosol. In both cell lines, truncated APC was enriched in the cytosol, with smaller pools in nuclear and dense insoluble fractions (Fig. 1B′). The data for β-catenin (Fig. 1B′) showed that cytosolic levels were high in parental SW80 cells, consistent with previous reports (Munemitsu et al., 1995), but that levels were significantly lower in SW480APC cells. The dense insoluble fraction contained the second largest β-catenin pool in both cell lines. Nucleosolic levels of β-catenin were also slightly lower in SW480APC cells. This characterization showed that fraction X constituted the second major subcellular pool for all the components of the destruction complex. This fraction is generally discarded in cell-fractionation experiments because it has long been considered as ‘cell debris’, and it has never been analyzed in the context of the Wnt pathway. Note that Triton X-100 was present in the last step of the fractionation, thus fraction X constituted a bona fide ‘detergent-insoluble’ fraction. We performed APC immunoprecipitation for the Triton-soluble fraction of the cells (which did not include fraction X). The results showed robust co-precipitation of all components of the destruction complex in both cell lines (supplementary material Fig. S2).
β-catenin phosphorylation in SW480 and SW480APC cells
We established a specific in vitro kinase assay to monitor the endogenous activity responsible for the N-terminal phosphorylation of β-catenin. Recombinant β-catenin was used as substrate, at a concentration of 100 nM, which corresponds to the estimated cytosolic levels in non-stimulated cells (Lee et al., 2003). At the dilutions used in this assay, any contribution from endogenous β-catenin present in the cell extracts was negligible (supplementary material Fig. S3A). Priming at residue S45 and subsequent phosphorylation at sites S33/S37/T41 were detected with specific antibodies (anti-pS45 and anti-pS33/S37/T41) by quantitative immunoblotting (see Materials and Methods). Note that pre-absorption of the anti-pS33/S37/S41 antibody was essential, because this polyclonal antibody showed strong cross-reactivity with non-phosphorylated β-catenin (see Materials and Methods). Antibody concentrations were optimized to obtain a linear response over the relevant range of signal intensities.
Comparison of kinase activities in crude extracts from SW480 and SW480APC cells readily yielded an unambiguous result: the activity was significantly higher in SW480APC cells, both for S45 and for S33/S37/T41 (Fig. 1C). The differences (approximately fivefold for S45 and approximately twofold for S33/S37/T41) were highly reproducible (Fig. 1C′), highlighting the robustness of the assay and the consistency of endogenous activities. Similar differences between the two cell lines were also observed when β-catenin concentration was raised to 1 µM (supplementary material Fig. S3B).
Although not absolutely required for phosphorylation by GSK3 in in vitro experiments, S45 priming is nevertheless considered essential in vivo. Because rescue with full-length APC enhanced priming significantly more than it enhanced S33/S37/T41 phosphorylation, we wanted to determine whether APC was directly required for the latter reaction, or whether increased S33/S37/T41 phosphorylation was simply a consequence of accelerated priming. We isolated the S33/S37/T41 phosphorylation step by using a ‘constitutively primed’ phosphomimetic β-catenin variant, S45D. We found that the kinase activity towards S45D β-catenin was higher in SW480APC cell extracts (Fig. 1D,D′). We conclude that full-length APC is required for the full activity of both phosphorylation steps. The fact that S33/S37/T41 phosphorylation of wild-type and S45D β-catenin was enhanced to a similar degree in SW480APC cells indicated that priming was not limiting in SW480 cells.
The observed twofold increase in overall β-catenin phosphorylation appeared surprisingly modest if one assumed that APC was absolutely required for the reaction. Various explanations could account for this rather mild enhancement: (1) Consistent with in vitro experiments, axin could be sufficient for β-catenin phosphorylation. APC would then only improve the efficiency of the reaction. (2) Axin could be limiting in these cells, and APC expression could enhance phosphorylation only up to the maximal rate allowed by axin. (3) The APC mutation of SW480 cells might not constitute a complete loss of function, and the resulting truncated APC might still supply part of APC function in β-catenin phosphorylation.
To discriminate between these possibilities, we depleted both wild-type and truncated APC by RNA interference (Fig. 2A). If truncated APC does not show complete loss of function, one would expect that its depletion would further decrease β-catenin phosphorylation. Otherwise, depletion would have no effect, and might even increase activity, as the truncated fragment could potentially have an inhibitory effect on this process. Transfection of siRNA against APC led to a ∼50% depletion of truncated APC in SW480 cells and a ∼80–90% depletion of full-length APC in SW480APC cells (Fig. 2C). The phosphorylation activities towards S45 and S33/S37/T41 were further reduced in both SW480 cells and in SW480APC cells. The decrease was roughly proportional to the reduction in the levels of truncated and full-length APC, respectively. The steady-state levels of endogenous β-catenin were also consistently increased (Fig. 2C). We inferred that the presence of APC (or at least of its N-terminal half) was absolutely required for β-catenin phosphorylation.
We then asked whether axin was limiting, in which case one would expect that higher levels of axin might boost β-catenin phosphorylation in SW480APC cells, and perhaps even compensate for the absence of full-length APC in SW480 cells. However, mild axin overexpression (2.3-fold±0.4 in parental SW480 cells and 2.5-fold±0.9 in SW480APC cells, means±s.d., Fig. 2C′) failed to stimulate β-catenin phosphorylation (Fig. 2B). By contrast, S45 phosphorylation in SW480 cells was slightly but reproducibly decreased. We conclude that, contrary to common assumptions, axin is not limiting, or at least not in SW480 cells, for the specific reaction of β-catenin phosphorylation. These results further support the notion that APC is crucial for β-catenin phosphorylation and loss of its activity cannot be compensated for by increasing axin levels.
Taken together, these experiments led to two important conclusions: they showed that APC is required for both phosphorylation steps and they also demonstrated that truncated APC still has substantial activity.
Direct binding to APC is required for β-catenin phosphorylation
We verified that the role of APC in β-catenin phosphorylation required the direct binding of β-catenin to APC. For this purpose, we used recombinant β-catenin variants with point mutations that specifically impaired binding to APC. To distinguish between binding to the 15AA and 20AA repeats of APC (Fig. 1A), we tested three separate mutations (Fig. 3), β-catenin-APCΔ15(R386A), which is defective in binding to 15AA repeats, and β-catenin-APCΔ20(K345A) and β-catenin-APCΔ20(W383A), both of which are defective in binding to the 20AA repeats (von Kries et al., 2000). Loss of binding to a specific type of repeat was confirmed by in vitro pull down (supplementary material Fig. S4). These mutant substrates were tested for in assays for S45 and S33/S37/T41 phosphorylation, in both SW480 and in APC-rescued cells. In all cases, the activity of the mutants was lower than that of wild-type β-catenin. The difference was relatively mild for the mutant that lacked binding to the 15AA repeats, but was stronger for the two other mutants. Double mutation of K345 and W383 led to a slight but not statistically significant decrease in the S33/S37/T41 phosphorylation activity compared with the single mutants. We conclude that both types of interaction are required for full activity, with a stronger requirement for the 20AA repeats. These results also highlight the importance of the single remaining 20AA repeat in the truncated APC of SW480 cells (see Discussion).
β-catenin phosphorylation occurs mainly in an insoluble fraction
We determined the subcellular distribution of β-catenin phosphorylation activity using our cell-fractionation protocol. The distribution of the kinase activity was largely similar in SW480 and SW480APC cells, for all three measured activities (i.e. pS45 priming and S33/S37/T41 phosphorylation on wild-type β-catenin as well as S33/S37/T41 phosphorylation on constitutively primed S45D β-catenin) (Fig. 4). Surprisingly, the dense insoluble fraction ‘X’ was by far the most active pool, accounting for ∼60–70% of the total kinase activity against β-catenin. Comparatively, the cytosol showed only a modest activity (10–20%), despite the fact that it contained the largest pools of APC, axin, CK1α and GSK3 (Fig. 1B′,B″). The other significant pool of β-catenin phosphorylation activity was the nuclear insoluble fraction, which matched and even surpassed the cytosol in the case of S45D phosphorylation (Fig. 4B,D). Nucleosol and membrane fractions showed low to negligible activity.
Effect of APC and APC truncation on β-catenin nuclear transport
To directly investigate the effect of truncated and wild-type APC on β-catenin nuclear transport, we performed fluorescence recovery after photobleaching (FRAP) experiments on SW480 cells that were transfected with YFP–β-catenin (Fig. 5). The import and export kinetics in parental SW480 cells were roughly similar to those measured in HEK293 and NIH 3T3 cells (Krieghoff et al., 2006; Sharma et al., 2012). However, the use of the spinning-disk confocal microscope allowed us to obtain information about the initial phase of recovery, which had not been studied so far. We measured surprisingly fast transport kinetics, both for import and for export (Fig. 5; Table 1). The resulting recovery kinetics clearly fitted a two-phase association model, with similar kinetics in both directions (Table 1). The first phase of translocation was extremely rapid (K∼0.1/s, half-life<10 s), in fact it was almost as fast as for GFP, which was used to monitor the free diffusion of a small protein (Fig. 5D,E; supplementary material Fig. S5; Table 1). The second phase was an order of magnitude slower (K∼0.01/s, half-life>1 min). Neither import nor export seemed to reach full recovery after 5 min, but approached a plateau at ∼60–80% of the pre-bleach fluorescence levels. These observations suggested that, as far as nuclear import is concerned, β-catenin may be partitioned into three potential cytoplasmic pools, one free to diffuse through the nucleopores, the second subjected to partial retention, and a third pool that is apparently unavailable for translocation on this time scale. Thee equivalent pools would also exist in the nucleus. Note that β-catenin was transported more efficiently than Cherry–NLS, which was used as a reporter for classical importin-mediated import (supplementary material Fig. S5G,H).
Recovery curves (nuclear to cytoplasmic ratio for import, and cytoplasmic to nuclear ratio for export, see Fig. 5) were fitted with a two-phase association algorithm. The table presents, for each condition, the mean (Av.) values calculated from the fitting of each individual cell measurement, as well as the fitting of compiled data from the whole set of experiments. Both methods gave very similar values. Kfast and Kslow represent kinetics constants for the two phases. The plateau is given as the percentage of the pre-bleach value, and corresponds to the ‘mobile’ faction. Fast (%) indicates the relative contribution of the fast phase to the overall curve. t-test compared to parental SW480 cells. Low p values indicating significant differences are highlighted in bold. s.e., standard error. Significant differences are indicated in bold.
We also compared the transport kinetics of APC-rescued and APC-depleted cells (Fig. 5B–E; Table 1). For import, the kinetics of the fast phase was the same under all conditions. However, we observed differences in the contribution of the slow phase (percentage recovery − percentage fast phase), which was larger in APC-rescued cells and smaller in APC-depleted cells. The kinetics of this phase was also significantly slower in the presence of full-length APC, with a half-life shifted from ∼1 min to ∼5 min. These results indicate that APC acts as a reversible retention component, with the full-length form retaining β-catenin more strongly than the truncated form. The expression of full-length APC had no effect on nuclear export, but APC depletion stimulated the process by increasing the fast-moving fraction.
We analyzed the effect of inhibition of classical CRM1 (XPO1)-mediated export by using the drug leptomycin B (LMB) (supplementary material Fig. S4A–D). Because nucleocytoplasmic shuttling is a very fast process, the nuclear accumulation of shuttling proteins is expected to be observed within 1 h of LMB treatment. For β-catenin, however, a 4-h treatment had no detectable effect, either on import or on export. In both parental and APC-rescued cells, the recovery curves perfectly superimposed with those from control cells. In fact, APC distribution under these conditions was not significantly affected (supplementary material Fig. S6). However, longer treatments (8 h) did show an effect on import (but not export): the rate of import was increased, with the initial recovery phase becoming even faster than for APC-depleted cells, approaching the kinetics of free GFP (supplementary material Fig. S5, insert in supplementary material S5E).
We also compared the relative nuclear and cytoplasmic steady-state distribution of β-catenin by measuring the relative fluorescence of YFP–β-catenin in transiently transfected cells. We found that the nuclear signal was generally close to the cytoplasmic signal (median ∼1.2, supplementary material Fig. S7A), although it varied from cell to cell. The ratio was largely similar for parental, APC-rescued and APC-depleted cells and was not affected by LMB treatment. We also verified that variations between cells were not related to levels of expression (supplementary material Fig. S7B). Note that a very similar ratio was measured for free GFP, which is considered to freely equilibrate between both compartments. Note also that even GFP appeared to have an immobile nuclear fraction (supplementary material Fig. S5), which probably accounts for its nuclear∶cytoplasmic ratio being slightly higher than 1.
In this study, we explore systematically three crucial aspects of APC biology. We demonstrate the requirement for APC in β-catenin phosphorylation, we identify that the activity is largely restricted to an insoluble compartment, and we formally verify that β-catenin nuclear transport is independent of APC, which rather acts as its main retention factor. This study also confirms that a typical cancer-related truncated form of APC is not a null mutant in terms of Wnt regulation, but can still promote significant β-catenin phosphorylation activity. In addition, truncated APC still plays a significant role in β-catenin retention in the context of nuclear transport.
Direct comparison of the function of truncated and full-length APC
The APC-rescued SW480 cell line produced by Burgess' group has provided a powerful tool to examine APC function in Wnt signaling. Note that Faux et al. (Faux et al., 2004) observed that the re-introduction of full-length APC had effects on cell behavior, and in particular on cell adhesion. We have performed a series of verifications, which did not reveal any overt differences in the levels or subcellular distribution of the components of the axin complex. We only detected a slight increase in the amount of soluble GSK3α, which does not impact on the interpretation of our results, because the kinase activities are concentrated in the dense insoluble fraction. Even assuming the existence of other small differences, they would not account for the dramatic increase in β-catenin phosphorylation that is measured in APC-rescued cells. A second issue that is relevant to cancer cells and, as a matter of fact, to all immortalized cell lines, is the likely occurrence of additional unknown mutations. In this study, this caveat was circumvented by the comparison with cells in which APC was depleted by siRNA treatment, and all our results turned out to be extremely coherent. They can all be fully explained by the sole contribution of APC.
APC is required for β-catenin phosphorylation
Although there is abundant evidence in the literature for a requirement for APC in β-catenin phosphorylation, this has never been proven in vivo. Here, we provide the first formal demonstration of this key point. Our results support those models in which APC is a core component of the axin complex (Ha et al., 2004; Hinoi et al., 2000; Roberts et al., 2012; Roberts et al., 2011; Xing et al., 2003). The involvement of APC in β-catenin phosphorylation is further confirmed by the finding that APC is required for both S45 priming and for the subsequent S33/S37/T41 phosphorylation, and that it requires β-catenin to bind directly to APC. Although axin seems to compensate for APC truncations and rescue ‘normal’ β-catenin in SW480 cells when highly overexpressed (Behrens et al., 1998; Hart et al., 1998), APC seems to be absolutely necessary under more physiological conditions. Note that although measurements from Xenopus egg extracts suggest that axin is limiting (Salic et al., 2000; Lee et al., 2003), axin and APC are expressed at similar levels in SW480 and SW480APC cells, and axin is even more abundant in other mammalian cells (Tan et al., 2012). In Drosophila embryos, β-catenin regulation is equally sensitive to APC and axin levels (Roberts et al., 2012; Roberts et al., 2011).
β-catenin phosphorylation occurs in an insoluble fraction
The weak β-catenin phosphorylation activity in the cytosol was another surprise of our study. The cytosol seemed to contain an excess of all the components that are required to build active complexes. CK1ε was the only component absent from this fraction, but CK1α is generally considered to be at least as effective for β-catenin priming. Dilution is unlikely to explain this low activity, because in our assays cytosolic fractions were, in fact, more concentrated than the crude extracts. Furthermore, according to our immunoprecipitation data (supplementary material Fig. S2), all interactions seemed to be relatively unaffected by dilution. Note also that the conditions for cytosol extraction were milder than those used for immunoprecipitation (the buffer contained a very low concentration of digitonin as opposed to a high concentration of Triton X-100). From these results, we infer the existence of large cytosolic axin and APC pools that are either poorly active or inactive in β-catenin phosphorylation. We suggest that cytosol complexes might either participate in a dynamic, perhaps regulated, equilibrium with fully active insoluble complexes, and/or fulfill other functions, such as JNK signaling for axin or cytoskeletal regulation for APC.
Several studies have investigated the nature of APC complexes (Mahadevaiyer et al., 2007; Maher et al., 2009; Penman et al., 2005; Reinacher-Schick and Gumbiner, 2001). A systematic study by cell fractionation showed that a substantial fraction of APC was sedimentable and detergent insoluble (Reinacher-Schick and Gumbiner, 2001). Other evidence for the association of APC with dense structures came from the images of APC-positive granules either in cell protrusions (Mili et al., 2008) or associated with the plasma membrane (Reinacher-Schick and Gumbiner, 2001). The junctional localization of APC, axin and GSK3 was also reported in SW480 cells (Maher et al., 2009).
Unfortunately, it is difficult to compare our data with any of the previous biochemical investigations, because they all used ‘post-nuclear’ supernatants from an initial centrifugation. This standard step of all classical fractionation protocols removes nuclei and unbroken cells. However, the discarded pellet contains a large fraction of the cytoskeleton and of the dense plasma membranes, and substantially overlaps with our fraction X (Liu and Fagotto, 2011). Here, we demonstrate that this dense insoluble fraction contains most of the β-catenin phosphorylation activity. Because this fraction was likely missing from all previous analyses (including Bilic et al., 2007; Li et al., 2012; Taelman et al., 2010), the nature and regulation of the axin-based β-catenin destruction complex(es) need to be revisited. Note that a fraction of E-cadherin is also recovered in fraction X, which is likely to correspond to the detergent-insoluble cytoskeleton-associated junctional pool. This probably also explains the relatively high β-catenin levels in this fraction. However, it is probable that this pool is independent of the axin–APC complex, because interactions of β-catenin with APC and cadherin are mutually exclusive (Hülsken et al., 1994).
The effect of C-terminal truncations and the relative role of the 15AA and 20AA repeats
The reason for the strong selection of the cancer-related APC truncations has been thoroughly discussed (for example, see Kohler et al., 2008). The initial theory proposed that the loss of the 20AA repeats caused β-catenin stabilization. This view was later challenged, and was largely replaced with a model in which the loss of the axin-binding SAMP motifs had the central role. Recent evidence has raised questions about this view, showing in particular that APC can interact with axin independently of SAMP motifs. It has also become clear that both the type and number of β-catenin-binding repeats are important in regulation of the Wnt pathway (Roberts et al., 2011). These results have reinstated the original model as the most accurate explanation of cancer-specific APC truncations.
In any case, the fact that most truncations leave more than half of the APC protein intact indicates that the distal region of this fragment bears an important function and that there is a strong selection in cancer cells to preserve it. However, whether this residual function is directly related to β-catenin phosphorylation has remained an open question. It has even been suggested that the truncated forms might have acquired some dominant activity. Our data confirm the model of Peifer and colleagues (Roberts et al., 2011), and unequivocally establish that, in terms of β-catenin phosphorylation, APC truncation results in a bona fide loss of function, yet it produces not a null allele but rather a relatively weak one that retains a surprisingly high activity.
Thanks to our specific assay, we have been able to further dissect the requirements for the 15AA and 20AA repeats, and confirm several previous assumptions. We demonstrate that both types of APC repeats are needed for full activity of the APC protein. The 20AA repeats are, as predicted, particularly important. This is not only true for cells expressing full-length APC, but even for parental SW480 cells, thus providing a clear explanation for the maintenance of one 20AA repeat in truncated APC. In addition, we find that binding to the 15AA repeat, which was thought to have little influence on β-catenin degradation (Roberts et al., 2011; Roberts et al., 2012), was, in fact, quite important (Fig. 3). The relative contribution of the two types of repeats correlated well with their relative number in full-length and truncated APC – the 15AA repeats played a particularly important role in S33/S37/T41 phosphorylation in parental SW480 cells, whereas the 20AA repeats seemed to fulfill most of the function of full-length APC. These results do not exclude other specific defects caused by APC truncation (we did observe differences in the cytoplasmic retention of β-catenin), but, in principle, the observed decreased rate of β-catenin phosphorylation appears to account for its stabilization and the resulting over-activation of the pathway.
APC and β-catenin nuclear transport
Although APC (and axin) was proposed to mediate the nuclear export of β-catenin (Bienz, 2002), there is strong evidence that β-catenin can freely diffuse into and out of the nucleus (Fagotto et al., 1998; Kose et al., 1997; Wiechens and Fagotto, 2001; Henderson and Fagotto, 2002; Sharma et al., 2012), and that APC and other binding partners of β-catenin all negatively affect β-catenin translocation (Krieghoff et al., 2006). Our observation of very fast transport kinetics further corroborates the notion of selective diffusion, and the analysis of transport in APC-depleted cells confirms that APC is not required for translocation. Note also that the speed of fluorescence recovery that is observed in APC-depleted cells (where the apparent retention of β-catenin in the cytoplasm is very low) approaches that of freely diffusible GFP. This is remarkable, considering the differences in size and shape of the two proteins (28 kDa and globular for GFP versus >90 kDa and rodlike for β-catenin).
Another definitive argument against a role for APC, or any other potentially shuttling protein, such as axin, in β-catenin export is raised by the lack of measurable changes in β-catenin distribution after 4 hours of LMB treatment. In fact, APC does not appear to be a freely shuttling protein, because this 4-hour treatment is not sufficient to cause any significant nuclear accumulation of APC. We obtained similar results previously with axin (Wiechens et al., 2004). It thus seems clear that the ability of axin and APC to be re-exported has no short-term implications for β-catenin. This property probably serves in the long term to maintain the correct distribution of these important scaffold proteins in the various cellular compartments. Obviously, complete block of export over a longer period would impact on this distribution, and eventually also on the pool of β-catenin retained on either side of the nuclear membrane.
In terms of the impact of APC on β-catenin retention, our results perfectly confirm previous results by Behrens and co-workers (Krieghoff et al., 2006). Our specific contribution is to demonstrate that APC has measurable effects on β-catenin transport even when it is expressed at physiological levels. The fact that retention is significant in APC-expressing cells and almost nil in the absence of APC suggests that APC is a major, possibly the main, factor controlling the pool of β-catenin that is available for transport in these cells. Consistent with the data of Peifer and colleagues (Roberts et al., 2011; Roberts et al., 2012), truncated APC was still able to retain β-catenin in the cytoplasm, although as argued above, its high residual activity in stimulating β-catenin phosphorylation might have at least as much impact on the regulation of β-catenin signaling in cancer cells. Note that cadherins constitute another component that sequesters β-catenin at the plasma membrane. Because the association is very strong, we expect that it would contribute to the immobile fraction over the time scale of our experiments. We do not believe that cadherins had any significant impact on our FRAP measurements, for the simple reason that we used single spread cells grown at low density. Under these conditions, cadherin levels are extremely low in SW480 cells (data not shown). This is consistent with the live images, where GFP–β-catenin was generally not detectable at the membrane, even in cells expressing very low levels of this construct where cytoplasmic pool could not mask a membrane signal.
In conclusion, APC truncations are certainly not ‘null mutants’ in terms of β-catenin regulation, and they still fulfill an unexpectedly large part of APC function, fully consistent with the necessity for cancer cells to regulate β-catenin activity, as proposed in the ‘just-right’ hypothesis (Furuuchi et al., 2000).
MATERIALS AND METHODS
SW480 cells and SW480APC cells were kind gifts from Antony Burgess (Ludwig Institute, Melbourne, Australia). Cells were cultured in RPMI 1640 medium, supplemented with 10% fetal bovine serum, 1.5 mg/ml genetecin and 1% penicillin-streptomycin.
The antibodies used in the study were: mouse anti-APC (ALi 12–28, Santa Cruz Biotechnology), rabbit anti-APC (C-20, Santa Cruz Biotechnology), rabbit anti-APC [M-APC, a gift from Inke Näthke, University of Dundee (Näthke et al., 1996)], affinity-purified rabbit anti-axin (Wiechens et al., 2004), anti-β-catenin (H102, Santa Cruz Biotechnology), mouse anti-β-catenin (6F9, Sigma), rabbit anti-phospho-β-catenin (Ser33/Ser37/Thr41, Cell Signaling), rabbit anti-phospho-β-catenin (Ser45, Cell Signaling), mouse anti-GSK3α/β (05-412, Millipore), rabbit anti-casein kinase 1α (sc-28886, Santa Cruz Biotechnology), goat anti-casein kinase 1ε (sc-6471, Santa Cruz Biotechnology), mouse anti-casein kinase 1ε (sc-365259, Santa Cruz Biotechnology), mouse anti-GAPDH (6C5, Applied Biosystems), mouse anti-RanBP3 (BD Biosciences), rabbit anti-LRP6 (C-10, Santa Cruz Biotechnology), goat anti-γ-tubulin (sc-7396, Santa Cruz Biotechnology), mouse anti-γ-tubulin (ab11316, Abcam), rabbit anti-pericentrin (ab4448, Abcam), rabbit anti-β-actin (ab25894, Abcam) and rat anti-α-actinin (BT-GB-276S, Babraham Bioscience Technologies).
Plasmids and recombinant β-catenin construction
Myc-tagged full-length mouse axin (Zeng et al., 1997). YFP–β-catenin was constructed by adding the eYFP sequence followed by a five-glycine linker upstream of Xenopus β-catenin in pCS2+MT. CherryNLS was constructed by adding a classical nuclear localization sequence (KKKRK) to the C-terminus of Cherry fluorescent protein subcloned into the pCS2-vector. The His-tagged β-catenin and YFP–β-catenin mutants [S45D, APCΔ15(W386A), APCΔ20(K345A), APCΔ15(W383A) and APCΔ20(K345A/W383A] were produced by site-directed mutagenesis based on pCS2-YFP-β-catenin and pET-His-β-catenin (Wiechens and Fagotto, 2001) using the QuikChange II XL Site-Directed Mutagenesis Kit, according to the manufacturer's protocol. All constructs were confirmed by sequencing. GST-APCr15 and r20 were constructed using oligonucleotides coding for the sequences LDTPINYSLKYSDEQ (the first 15AA repeat of human APC) and EDTPICFSRCSSLSSLSSAED (the first 20AA repeat). APC-specific siRNA (sc-29702) and control siRNA (sc-37007) were purchased from Santa Cruz Biotechnology. Cells were transfected with plasmids or siRNAs using Lipofectamine 2000 (Invitrogen), according to the manufacturer's protocol.
Cell homogenization and fractionation
For preparation of total homogenates, cells were cultured in 6-cm plastic dishes and were harvested by scraping in 300 µl of osmolysis buffer (20 mM HEPES-NaOH pH 7.4, 0.2 mM EDTA), homogenized with 40 strokes in a tight-fit Dounce homogenizer, before addition of an equal volume of high Na+ buffer (400 mM sucrose, 300 mM NaCl, 20 mM HEPES-NaOH pH 7.4, 0.2 mM EDTA). This was followed by 40 additional strokes. Our cell-fractionation protocol was as described previously (Liu and Fagotto, 2011). The separation yielded the following five fractions (the volume of each is also given): cytosol (3 ml), nucleosol (0.5 ml), nuclear insoluble (1.35 ml), membranes (1.35 ml) and dense insoluble material (fraction ‘X’). The latter was recovered at the bottom of the Percoll gradient and was resuspended in 450 µl of low Na+ buffer (150 mM NaCl, 10 mM HEPES-NaOH pH 7.4, 0.1 mM EDTA). The Percoll gradient contained 0.6% Triton X-100 (Liu and Fagotto, 2011).
Samples were separated by SDS-PAGE according to the regular protocol, except for the detection of APC and pericentrin, which were resolved on a 4% gel without a stacking gel. The blots were developed using a chemiluminescence detection reagent (WBKLS0500, Millipore), and images were acquired with a 12-bit digital camera (Alpha Innotech MultiImage system). The data were quantified using the Gene Tools software (Syngene). Dilution series were used to verify the linearity of the signal. Note that in several cases, a large number of samples had to be blotted simultaneously, which required that several gels were run in parallel. To ensure perfectly equal conditions during transfer, incubation with antibodies and development, two to three gels were transferred onto one single nitrocellulose membrane. In all cases where collages are presented, they show conditions from a single membrane, with identical exposure time and contrast.
In vitro kinase assay
Substrates were recombinant His-tagged β-catenin proteins. Proteins were expressed in E. coli BL21 (DE3), purified on an Ni-NTA agarose column and exchanged into kinase buffer (150 mM NaCl, 20 mM HEPES-NaOH). Reactions were carried out in a total volume of 50 µl, containing 100 nM recombinant β-catenin substrates, 1 mM ATP, 1 mM MgCl2, 10 mM creatine phosphate and 10 U creatine kinase, and the following amounts of sample: total cell lysates, 20 µl; cell fractions, 40 µl (undiluted for Cs, Ns, Ni and Mem fractions, diluted 1∶3 for fraction X). The volumes were adjusted using kinase buffer. The reaction was started by the addition of the samples, was carried out at 37°C and was stopped by the addition of 4× Laemmli sample buffer plus 20 mM EDTA with incubation at 98°C for 3 min. The relative levels of phosphorylated S45 and S33/S37/T41 were determined by quantitative immunoblotting using the relevant phospho-specific antibodies. Several commercial anti-phospho-β-catenin antibodies were tested. Except for the anti-phospho-S45 antibody (Cell Signaling), all antibodies showed strong reactivity towards non-phosphorylated β-catenin. This reactivity was eliminated by preabsorbing anti-phospho-S33/S37/T41 with ∼50 µg/ml recombinant β-catenin for 60 min before incubation with the membrane. Band intensities were quantified as above, and the relative activities were calculated after background subtraction. For samples from crude extracts, results were expressed as the ratio between the signal intensity for phosphorylated β-catenin and β-actin, which was used as loading control.
Confocal microscopy and fluorescence recovery after photobleaching
Cells were grown on Fluorodish culture dishes and were transfected with YFP–β-catenin, eGFP (Clontech) or CherryNLS. Cells were maintained in a FCS2 live-cell chamber at 37°C under 5% CO2. Images were acquired by using a Quorum WaveFX spinning-disk confocal system (QuorumTechnologies), with a ×40/NA1.25 HCX PL APO CS oil objective. For photobleaching experiments, samples were photobleached with a solid-state 405-nm laser (475 mW), using a mosaic digital diaphragm (Andor Technology, Belfast, UK). Either the nucleus or the cytoplasm was bleached for 1 s at 100% laser power. The samples were imaged continuously with a separate 488-nm laser line. Between 5 and 20 frames from a single z-plane were collected every 200 ms before, and immediately following, bleaching, followed by frames taken at 2-s intervals. The average nuclear and cytoplasmic intensities were measured using Metamorph or ImageJ softwares. After background subtraction, the nucleus to cytoplasm ratios (or the cytoplasm to nucleus ratios for export experiments) were calculated. The pre-bleach ratio was set to 100%, and the ratio in the first post-bleach image was set to 0. The recovery curves shown are the averages of at least 8–15 cells from at least three independent experiments. Curve fitting and statistical calculations were computed using GraphPad Prism 6.0 and Excel softwares.
We thank Maree Faux and Anthony Burgess (Ludwig Institute, Melbourne, Australia) for the generous gift of SW480APC cells, and Laura Canty (McGill, Montreal, Canada) for providing the Cherry–NLS construct. We acknowledge the support of the McGill University Biology department Cell Imaging and Analysis Network (CIAN) for confocal microscopy.
L.W. and F.F. conceived and planned the experiments. L.W., X.L., E.G. and F.F. performed the experiments. C.W. supervised L.W.'s PhD thesis. L.W., X.L. and F.F. interpreted the results and wrote the paper.
L.W. was the recipient of a Shandong University Joint Ph.D. training program studentship. This work was supported by a grant from the Canadian Cancer Research Society to F.F.
The authors declare no competing interests.