Colon carcinoma cell lines HCT116, DLD-1, LS411, and LoVo were obtained from American Type Culture Collection and maintained in McCoy's 5a medium/10% FBS (HCT116), RPMI 1640/10% FBS (DLD-1 and LS411), or Ham's F12 medium/20% FBS (LoVo), respectively. The breast cancer cell line MCF-7 was kindly provided by Dr. N. Rosen (Memorial Sloan-Kettering Cancer Center) and maintained in DME high glucose supplemented with 12 nonessential amino acids and 10% FBS. MDCK II cells were cultured in DME/10% FBS medium.

Cell fractionation procedures were carried out at 4°C unless otherwise noted. Protease inhibitors were added to solutions at the following final concentrations: PMSF, 0.017 mg/ml; leupeptin, 0.002 mg/ml; aprotinin, 0.004 mg/ml; antipain, 0.01 mg/ml; benzamidine, 0.05 mg/ml; ST inhibitor, 0.01 mg/ml; iodoacetamide, 0.1 mg/ml. HCT116 and MCF-7 cells were grown to confluency in 14-cm tissue culture dishes, washed once with PBS, and harvested by scraping. Cells were pelleted at 200 g for 5 min and resuspended in hypotonic lysis buffer (10 mM Tris-HCl, pH 7.8, 10 mM KCl) and lysed by Dounce homogenization after a 20-min (HCT116) or 10-min (MCF-7) incubation on ice. Lysis was assessed by light microscopy after staining with trypan blue. The sample was centrifuged at 5,000 rpm for 30 min to pellet unlysed cells and nuclei. Isotonic salt concentrations were restored (140 mM KCl) and the supernatant was further fractionated by centrifugation at 100,000 g for 1 h at 4°C. The resulting fractions were labeled Pellet 100 (P100) and Supernatant 100 (S100).

To evaluate a potential membrane association of APC in HCT116 and MCF-7 cells, the P100 fraction was resuspended in a 55% sucrose solution (in buffer A: 100 mM NaCl, 20 mM Tris-HCl, pH 7.8, 5 mM EDTA) and loaded onto 0.5 ml of a 66% sucrose solution (cushion). 5 ml of a continuous 20–50% sucrose gradient was poured on top (in buffer A). The samples were spun in a Beckman Coulter SW 55Ti rotor at 45,000 rpm for 16 h at 4°C. 0.55-ml fractions were collected and analyzed by Bradford protein assay and Western blotting.

To characterize biochemical properties of APC after flotation, gradient fractions containing plasma membrane material (sucrose concentrations of ∼30–37%, density 1.12–1.16 g/cm³) were pooled. A portion of the membrane pool (two thirds) was treated with detergents (1% NP-40 or 1% octyl glucoside) or high salt (1 M NaCl) and spun at 100,000 g for 1 h. Samples were analyzed by Western blotting.

To further fractionate the S100 sample according to size, velocity gradient centrifugation was performed. 4.5 ml of a continuous 10–40% sucrose gradient was layered over 0.5 ml of a 66% sucrose solution (cushion), and an aliquot of the S100 fraction was loaded on top of the gradient. The sample was spun at 55,000 rpm in a Beckman Coulter SW55 rotor for 4.5 h. The marker proteins catalase (4–5S pool as monomer derived from 11.5S pool as tetramer) and thyroglobulin (19S) were spun in parallel gradients. 625-μl fractions were collected and analyzed for protein content by Bradford protein assay and Western blotting. To estimate S values >19S, a software program for simulation calculation of velocity gradients was used based on a method developed by McEwen 1967.

To detect APC, samples were separated in 3% agarose gels and transferred by capillary transfer to nitrocellulose overnight as described (Smith et al. 1993). For the detection of E-cadherin, β-catenin, dishevelled, and axin samples were loaded on 7% SDS polyacrylamide gels (for GSK-3β), on 10% SDS polyacrylamide gels, separated, and transferred to nitrocellulose by semidry transfer. Western blotting was performed as described (Vleminckx et al. 1997). The following primary antibodies were used: APC, anti-APC NH₂ terminus (Ab-1, mouse IgG1), anti-APC COOH terminus (diluted 1:1,000) (Ab-2, mouse IgG1; Oncogene Research Products), or anti-APC COOH terminus (diluted 1:500) (C-20, rabbit polyclonal IgG; Santa Cruz Biotechnology, Inc.); β-catenin, anti–β-catenin NH₂ terminus (polyclonal rabbit antiserum, diluted 1:5,000); E-cadherin, anti–E-cadherin (diluted 1:7,500) (mouse IgG; Zymed Laboratories); and GSK-3β, anti–GSK-3β (diluted 1:2,500) (mouse IgG1; BD Transduction Laboratories). Polyclonal rabbit antisera directed against murine dishevelled and murine axin (diluted 1:2,000) were kindly provided by R. Nusse (Stanford University, Stanford, CA).

To analyze confluent epithelial cells, glass coverslips were coated with poly–L-lysine (100 μg/ml) for 1 h at room temperature, and cells were allowed to adhere to the coverslips for at least 72 h in appropriate culture medium (see above). To examine cells at subconfluency, glass coverslips were coated with collagen (50 μg/ml), and cells were allowed to adhere for 24–48 h. State of confluency was assessed by light microscopy before fixation. Cells were fixed (100% methanol, 5 m at −20°C), washed three times in PBS, blocked with BSA (1% in PBS, 30 min at room temperature), and incubated with primary antibody (diluted in 1% BSA/PBS 1 h at room temperature). Samples were then incubated in appropriate secondary antibodies conjugated to Cy3 (Jackson ImmunoResearch Laboratories) or FITC (Molecular Probes) (diluted 1:750 or 1:200, respectively, in 1% BSA/PBS, 30 min at room temperature). Specimens were examined by confocal microscopy equipped with a 25× or 63× objective.

Primary antibodies used for immunofluorescence microscopy were as follows: APC, anti–human APC NH₂ terminus (diluted 1:75) (N-15, rabbit polyclonal IgG; Santa Cruz Biotechnology, Inc.); β-catenin, anti–β-catenin COOH terminus (diluted 1:200) (clone 14, mouse IgG; BD Transduction Laboratories); β-tubulin, anti–β-tubulin (diluted 1:200) (clone Tub 2.1, mouse IgG; Sigma-Aldrich). As controls for APC staining, either normal rabbit IgG was substituted for the primary anti-APC antibody or an excess of the neutralizing peptide against which the anti-APC antibody was raised, was incubated with the anti-APC antibody before immunofluorescence analysis (five times excess by weight, overnight at 4°C in 1% BSA/PBS).

To determine the subcellular localization of APC in mouse colon, specimens were briefly washed in ice cold PBS, embedded in OCT compound (Miles, Inc.) for 1 h on ice, and frozen in liquid nitrogen. Frozen samples were then cut into 12-μm sections. Sections were air dried and fixed in 100% acetone for 10 min on ice, rehydrated in PBS, and blocked with 10% donkey serum in 2% BSA/PBS for 30 min at 37°C. Samples were then incubated with primary antibody in 2% BSA/PBS for 1 h at 37°C (anti-APC antibody N-15, diluted 1:50; anti–β-catenin antibody clone 14, diluted 1:100). Slides were washed in 0.1% BSA/PBS for 5 min, washed in 0.05% Triton X-100/PBS for 5 min, and additionally washed in 0.1% BSA/PBS twice for 5 min. Samples were then incubated in appropriate secondary antibodies conjugated to Cy3 (Jackson ImmunoResearch Laboratories) or FITC (Molecular Probes) (diluted 1:750 or 1:200, respectively, in 2% BSA/PBS, 30 min at 37°C). After additional washing, sections were mounted and examined by confocal microscopy equipped with a 63× objective. Normal rabbit IgG and preincubation of the anti-APC antibody with the competing peptide against which the antibody was raised were used as controls.

To localize APC at the subcellular level, we initially used the human colon carcinoma cell line HCT116, because this cell line expresses wild-type APC. It does express a mutant β-catenin protein that carries a deletion of ser45 (Rubinfeld et al. 1997). The fractionation protocol described in Fig. 1 was used. As shown in Fig. 2 a, APC distributed approximately equally into pelletable and soluble material after the initial 100,000-g spin. This distribution pattern was consistently observed when fully confluent monolayers were used.

To determine whether the pool of APC that sedimented at high speed (P100) was associated with membranes equilibrium, density flotation centrifugation was performed. With this method, APC mostly fractionated at light densities (∼1.13 g/cm³) and comigrated with vesicles containing the integral membrane protein E-cadherin, suggesting that APC associated with membranes in HCT116 cells (Fig. 2 c). Although APC was consistently found to float with membranes, in some experiments APC could also be detected in dense fractions of the gradient (1.3 g/cm³) where most of the total protein of the P100 fraction sedimented (data not shown).

To determine the nature of the association of APC with the membrane, this fraction was treated with mild nonionic detergents or high salt concentrations (Fig. 2 d). High salt did not release APC from the membrane fraction, suggesting that APC is tightly associated with the membrane. Neither NP-40 nor octyl glucoside efficiently solubilized APC. The fraction that pelleted after detergent solubilization of membranes did not refloat during subsequent density equilibrium centrifugation (data not shown), indicating that this APC subfraction is not associated with classic lipid rafts (Simons and Ikonen 1997), but rather is present in a very large submembranous protein complex.

To determine where the membrane-associated fraction of APC localizes in HCT116 cells, we used indirect immunofluorescence microscopy with the polyclonal rabbit IgG antibody N-15 raised against an NH₂-terminal peptide of human APC. Immunolocalization was carried out on fully confluent, polarized epithelial cell monolayers. Confocal microscopic analysis of immunostained HCT116 colon carcinoma cells showed that APC predominantly localized to the apical plasma membrane and to a very small degree to the apicolateral cell border (Fig. 3, d–g). APC and β-catenin, for the most part, did not colocalize. β-Catenin staining was mostly seen laterally with very little overlap of the two proteins detectable at the apicolateral cell borders (Fig. 3 g). Essentially no APC was seen in more basal sections of the cell (Fig. 3 d). APC staining was most pronounced in apical sections of the cells (Fig. 3 f). The apical localization of APC is best observed in the z-section of analyzed monolayers (Fig. 3 g). For controls using normal mouse or rabbit IgG, no specific staining was detected (Fig. 3, a and b). Furthermore, apical staining of APC was effectively blocked by preincubation of the antibodies with the neutralizing peptide against which the anti-APC antibody was raised (Fig. 3 c).

To determine whether the apical membrane localization of APC is unique to the HCT116 colon carcinoma cell line or due to the presence of a mutant β-catenin found in HCT116 cells, we analyzed the breast cancer cell line MCF-7, which expresses wild-type β-catenin and wild-type APC. The distribution of APC during the initial 100,000-g spin was similar in MCF-7 cells compared with HCT116 cells. APC fractionated to about equal proportions into pellet and supernatant (Fig. 4 a). Furthermore, as found in HCT116 cells, APC in MCF-7 cells also sedimented at low densities characteristic of plasma membranes during equilibrium centrifugation, suggesting that APC similarly associates with the plasma membrane in this cell line (Fig. 4 b). To assess whether APC also localizes specifically to the apical plasma membrane in MCF-7 cells, we analyzed this cell line by confocal immunofluorescence microscopy (Fig. 5, a–g). APC localized predominantly to the apical membrane and in general did not overlap with β-catenin, which mainly localized to lateral cell borders (Fig. 5, d–g; negative controls for MCF-7 cells shown in a–c). Also, to evaluate APC localization in a nontransformed epithelial cell line, we examined fully confluent, well-polarized MDCK cells by immunofluorescence microscopy (Fig. 5, k–n; negative controls for MDCK cells shown in h–j). APC also localized apically in MDCK cells, whereas β-catenin was found laterally. In general, no overlap between the two proteins was observed.

To determine where APC localizes in situ, we examined samples of normal mouse colon by immunofluorescence microscopy. APC was enriched at the apical membrane of epithelial cells predominantly in upper regions of colonic crypts towards the lumen of the digestive tract (Fig. 6, a and b). In addition, some diffuse cytoplasmic staining was detected. In contrast, β-catenin was seen mainly at the lateral borders of colonocytes and no obvious colocalization with APC was found. Immunoreactivity towards the APC protein in epithelial cells decreased substantially towards the base of the crypts (Fig. 6 c). Additionally, APC was detected in nonepithelial cells in the lamina propria and throughout the submucosa and muscularis layers. APC staining was effectively blocked with the competing peptide against which the anti-APC antibody was raised (Fig. 6 d).

Since a lot of attention has been given to the association of APC with microtubules, we also performed double immunostaining for APC and β-tubulin. In fully confluent cell monolayers, no apparent colocalization of APC and microtubules was observed (data not shown). In subconfluent HCT116 cells (Fig. 7, a–c) and subconfluent MDCK cells (Fig. 7, d–f), APC was found over large regions of the cell surface and perhaps in the cytoplasm. We did occasionally detect APC at the tips of cell protrusions where microtubules project, as reported previously (Nathke et al. 1996); however, these were very rare (<5% of identified microtubule-containing protrusions) (Fig. 7 a). Much more commonly observed was APC distributed all along the surface of the cell protrusions (Fig. 7, b–f). Staining patterns for HCT116 and MDCK cells were very similar.

Truncations of APC are loss of function mutations that are responsible for loss of tumor suppressor activity. Therefore, we asked whether truncation mutations in APC affect its apical membrane localization. The distributions of APC in several colon carcinoma cell lines with known truncations of the APC protein were examined by confocal immunofluorescence microscopy. The linear representation of the full-length human APC protein is shown in Fig. 8 a. The DLD-1 cell line (Fig. 8, b–e) harbors a truncation mutation within the central region of the APC protein, a region which is known to be crucial for the tumor suppressor activity of APC through downregulating β-catenin (Munemitsu et al. 1995). The cell line LoVo (Fig. 8, f–i) carries a mutation at codon 1114 of APC, retaining the NH₂ terminus of APC including one of the three 15–amino acid repeats. The cell line LS411 (Fig. 8, j–m) carries a mutation at codon 789 which causes chain termination shortly after the seven armadillo repeats (Homfray et al. 1998). A polyclonal antibody directed against the NH₂ terminus of the APC protein was used to detect APC. In all cell lines, the mutant APC protein predominantly localized to the apical plasma membrane as observed in serial transverse sections through confluent monolayers. Again, most APC protein did not colocalize with most of the lateral membrane-associated β-catenin in DLD-1, LoVo, and LS411 cells. APC staining was effectively abolished by competition with the specific neutralizing peptide against which the anti-APC antibody was raised (Fig. 8e, Fig. i, and Fig. m). These data suggest that an NH₂-terminal fragment of the APC protein containing the oligomerization domain and the arm repeat region is sufficient to localize APC to the apical membrane.

Although most of the APC localized to the apical membrane in epithelial cells by immunofluorescence, we did identify a soluble pool of APC biochemically upon initial high speed centrifugation (Fig. 2 a). To characterize the soluble fraction of APC further, we determined the size of APC-containing protein complexes by velocity gradient centrifugation. With this method, APC sedimented in two size-fractions corresponding to sedimentation coefficients of ∼20S and 60S (Fig. 9 c and 10 c). In particular, the 60S pool of APC fractionated away from most of the total protein in the S100 fraction, most of which sedimented at a smaller size of ∼4–5S and some at ∼20S (data not shown). This distribution pattern of APC into two distinct pools was repeatedly observed in multiple experiments.

A convergence of biochemical and genetic studies have shown that APC can form complexes with several proteins in the Wnt signaling pathway, including β-catenin, axin, dishevelled, and GSK-3β. APC and its binding partners are thought to form a high molecular weight complex (called the destruction complex) that affects β-catenin signaling by catalyzing β-catenin phosphorylation, which is a prerequisite for proteasomal degradation of β-catenin. To get a better understanding of the localization of the destruction complex in epithelial cells, we determined the subcellular distribution of key components of the complex in confluent polarized epithelial cells. After high speed centrifugation of the postnuclear fraction from HCT116 cells, axin, dishevelled, and GSK-3β remained largely cytosolic, whereas APC distributed to about equal proportions into pellet and supernatant (Fig. 9 a). β-Catenin pelleted almost completely during this fractionation step, presumably because most of the β-catenin in HCT116 cells is bound to the membrane protein E-cadherin. To determine whether the small fractions of axin, dishevelled, and GSK-3β that sedimented at 100,000 g codistributed with APC and membranes, equilibrium density flotation gradients were run (Fig. 9 b). As before, most of the APC present in the P100 floated with membranes (assessed by its codistribution with β-catenin), whereas small amounts of axin and GSK-3β were detected in both the membrane and the dense fractions of the gradient. On the contrary, none of the small fraction of dishevelled protein present in the P100 floated with membranes, but rather sedimented with the bulk of the protein in gradient fractions with high densities.

With the antibodies used, two forms of axin and dishevelled were consistently detected by Western blotting. The higher molecular weight form of dishevelled distributed mostly in the S100 fraction (Fig. 9 a). For axin, small amounts of a higher molecular weight form were seen in the P100 fraction and were enriched in membrane fractions after flotation of the P100 (Fig. 9 b). The significance of these distinct immunoreactive forms of axin and dishevelled is unclear.

To determine whether the large soluble pool of the components of the destruction complex cofractionates with APC, the S100 fraction of HCT116 cells was analyzed by velocity gradient centrifugation (Fig. 9 c). Most of the soluble β-catenin, dishevelled, and GSK-3β sedimented at a size range of ∼4–5S. Only minor portions of GSK-3β and β-catenin overlapped with APC in the 20S fraction. Axin, on the other hand, was almost exclusively present in the 20S fraction of the gradient, thereby significantly cofractionating with the 20S pool of APC. None of axin, dishevelled, GSK-3β, or β-catenin codistributed with the 60S pool of APC in HCT116 cells. A summary of the overall approximate distributions of APC and components of the β-catenin destruction complex in HCT116 cells is depicted in Table .

To assess whether the distributions of these proteins were unique to HCT116 cells or dependent on the NH₂-terminal mutation of β-catenin expressed by this cell line, we also analyzed their fractionation patterns in the breast cancer cell line MCF-7 (Fig. 10). APC as well as β-catenin, axin, dishevelled, and GSK-3β distributed into subcellular pools in MCF-7 cells very similar to HCT116 cells. APC fractionated equally into P100 and S100 pools, whereas axin, dishevelled, and GSK-3β were largely soluble and β-catenin mostly pelleted (Fig. 10 a). In density flotation gradients, most complex components from MCF-7 cells behaved comparable to those from HCT116 cells (Fig. 10 b). APC and β-catenin mostly floated with membranes, whereas the small amounts of particulate dishevelled were mainly detected in dense fractions. In contrast to the HCT116 cells where the small P100 pools of axin and GSK-3β distributed in both membrane and dense fractions, the small pool of particulate axin fractionated exclusively in membrane fractions, and most of the pelletable GSK-3β sedimented in the dense fractions. Nonetheless, because the fraction of axin and GSK-3β present in the P100 before flotation is minor compared with the soluble pool of these proteins (see Fig. 10 a), their overall distributions are very similar in HCT116 and MCF-7 cells.

The fractionation pattern of the major soluble pool of the destruction complex in MCF-7 cells was also analyzed by velocity sizing (Fig. 10 c). Largely identical to HCT116 cells, APC fractionated into two peaks, a 20S and a 60S fraction, whereas β-catenin, dishevelled, and GSK-3β again distributed into the 4–5S fraction. Axin was again mainly found in the 20S pool, thus overlapping with the 20S fraction of APC. Again, none of the components of the destruction complex fractionated with the 60S pool of APC.