Tumor-Associated NH₂-Terminal Fragments Are the Most Stable Part of the Adenomatous Polyposis Coli Protein and Can Be Regulated by Interactions with COOH-Terminal Domains

Introduction

Mutations in the adenomatous polyposis coli (APC) gene contribute to colorectal tumorigenesis and are commonly associated with sporadic colon cancers ( 1, 2). Germ line mutations in APC result in an autosomal dominantly inherited disease called familial adenomatous polyposis ( 1), that is characterized by the development of multiple colorectal adenomatous polyps that tend to progress to carcinomas if left untreated ( 3). The biological functions of the APC protein are diverse, complex, and not yet fully understood. APC is required for the Wnt-regulated degradation of β-catenin, an important regulator of transcriptional activation and cell adhesion ( 4– 6), plays a role in chromosome segregation during mitosis ( 7– 10), and is implicated in regulating cytoskeletal dynamics, in particular microtubules ( 11– 15), but has also been described in filamentous actin (F-actin)–associated pools ( 16– 18). Consistent with such diverse functions, a large number of binding partners have been identified for APC and the intracellular distribution of the APC protein is extremely complex: APC is associated with microtubules and F-actin and is also present in cytoplasmic and nuclear pools ( 19– 23).

Little is known about the relationship between the diverse functions of APC. Some regions of the APC molecule mediate a number of distinct interactions raising the question how the functions of APC that result form these interaction are coordinated. For instance, the armadillo repeat–containing region in the NH₂-terminal domain of APC binds to KAP3 to provide a link to kinesins and thus an indirectly link between APC and microtubules ( 24), it interacts with Asef to provide indirect links to F-actin ( 25), and it can bind to the B56 subunit of PP2A which is important for the scaffolding function of APC in the Wnt pathway (ref. 26; Fig. 1C ). Thus, one region of APC alone seems involved in three distinct functions of APC. It is unlikely that all these interactions occur simultaneously raising the possibility that the interplay between intermolecular and intramolecular interactions is involved in targeting APC to a distinct function.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Limited proteolysis of APC. A, APC was immunoprecipitated from cell lysates and digested with increasing amounts [0 (lane 0), 0.03 (lane 1), 0.06 (lane 2), 0.12 μg (lane 3)] of endopeptidase Asp-N, Lys-C, and chymotrypsin as indicated. Samples were either treated with buffers alone (ctrl) or with lambda phosphatase (dephospho) before digestion. Fragments were separated by SDS-PAGE transferred to nitrocellulose and detected using antibodies against the NH₂-, middle (M), or COOH-terminal domain of APC. Distinct fragments are designated by numbers on the blot. The migration of molecular weight markers 220, 160, 120, and 100 kDa are indicated adjacent to each blot. B, schematic summary of the proteolytic fragments produced by the indicated proteases. Epitopes for the antibodies used to detect the fragments are indicated in vertically striped bosex above the sequence schematic of APC that shows the position of N-APC, M-APC, and C-APC fragments. Numbers next to each fragment correspond to the numbers in (A). Shaded or stippled regions indicate that a number of cleavage sites reside in this region. Arrows mark the approximate position of the cleavage sites and are shaded as indicated next to each protease name (right). C, output from the GlobPlot algorithm predicting the propensity for well-ordered structures. A downward slope indicates increased order; an upward slope represents increased disorder. Hatched bars indicate the position of recognized protein domains and mark the coiled coil domains (first two hatched bars before residue 400), armadillo repeats (residues 400-720), and two coiled coil domains at around 1580 and 1700). Please note that these last two are unique to human APC, whereas all others were detected in all APC sequences identified thus far. The estimated cleavage sites for the different proteases are indicated by arrows (top). The arrows are shaded to match the shading used to for each proteases as marked in (B). The position of the armadillo repeats and the microtubule binding site are shown below the graph, as are some of the proteins known to interact with NH₂-, middle-, or COOH-terminal regions of APC (below). Small grey boxes across the top of the graph show regions predicted to be particularly disordered.

Mutations in APC lead to loss of COOH-terminal regions and the expression of NH₂-terminal fragments of varying length. The idea that such NH₂-terminal fragments have dominant functions is supported by recent reports showing that expression of NH₂-terminal APC fragments leads to mitotic defects ( 9, 27). Whether such dominant effects of the NH₂-terminal fragments of APC that are expressed in tumors are the result of loss of functions due to lack of COOH-terminal regions in such fragments or gain of functions due to fewer regulatory interactions is not known.

To begin to address the molecular organization of the APC molecule so that these questions can be explored in detail, we used limited proteolysis as a means to identify regions within the APC molecule that were particularly sensitive to proteases. The proteolytic fragments we most frequently recovered in these experiments were large NH₂-terminal fragments that were similar to truncated fragments that result from the most common mutations in APC in cancer.

Our data further suggest that tumor-associated NH₂-terminal fragments of APC can be regulated by associating with COOH-terminal regions and that the ability of APC to bind microtubules efficiently is linked to this activity. We found that NH₂- and COOH-terminal regions of APC bind to each other in vitro and when coexpressed in cells. Furthermore, introducing COOH-terminal APC fragments into tumor cells that normally only contain a truncated, NH₂-terminal APC fragment (called “endogenous” APC protein), resulted in the recruitment of this NH₂-terminal APC fragment to microtubules. In addition, under these circumstances, protein interactions of the endogenous NH₂-terminal APC fragment changed. We propose that this interaction may be responsible for the regulation of functions done by the NH₂-terminal domain in the context of the full-length protein. Thus, the dominant function of NH₂-terminal APC fragments ( 9) could be at least partially explained by loss of such regulation in tumors as a result of truncation mutations.

Materials and Methods

Cell culture and transfection. HeLa, SW480, and DLD1 cells (respectively expressing wild type or truncated APC spanning residues 1-1309 and 1-1418) were routinely cultured at 37°C in 5% CO₂ in DMEM (Sigma, Poole, United Kingdom) with 10% fetal bovine serum. Cells were transfected using Polyfect 2000 (Invitrogen, Paisley, United Kingdom) for SW480 or Fugene 6 (Roche, Penzberg, Germany) for DLD1 cells using the protocol provided by the manufacturer. Insect cells, Sf21, containing recombinant baculovirus were cultured at room temperature in a spinner bottle in SF-900 II medium (Invitrogen) supplemented with 50 units/mL penicillin and 50 μg/mL streptomycin. Infection with baculovirus was initiated at 0.5 × 10⁶ cells/mL and cells were harvested ∼4 days after transfection at a density of 4 × 10⁶ cells/mL. Dictyostelium discoideum AX₂ cells, stably expressing constructs encoding different APC fragments were grown in shaking cultures in HL5 medium at room temperature ( 28).

Plasmid constructs. Constructs for protein expression in human cell lines encoding for NH₂-terminal (residues 1-1018 of human APC), middle-terminal (residues 1014-2038), and COOH-terminal (C-APC, residues 2038-2843; or CΔ1-APC, residues 2068-2843; CΔ2, like C-APC but lacking residues 2168-2451; CΔ1,2, like CΔ1 but lacking residues 2168-2451; CΔ2+ like C-APC but lacking residues 2168-2590) fragments of APC, were cloned into pEGFPC3 to introduce a green fluorescent protein (GFP) tag at the NH₂ terminus of each fragment as described previously ( 29).

Baculovirus constructs using the pFB-Nhis10HA vector encoding the same APC fragments carried an NH₂-terminal 10 His-HA-Flag tag but not GFP.

For expression in Dictyostelium AX₂ cells, these APC fragments were subcloned into the PB17SEYFP vector to introduce a YFP-tag at the COOH-terminal end of the APC fragments.

Antibodies. Antibodies against APC that were used for this study were mouse monoclonal anti-N-APC (ALi) reactive with residues 45 to 140 (as described for ab58 by abcam) ¹ mouse monoclonal anti-C-APC raised against an MBP fusion with the COOH-terminal region of human APC (as described for ab120, abcam) ² affinity purified rabbit polyclonal serum raised against and reactive with residues in the middle of human APC ( 11), rabbit polyclonal anti-N-APC antiserum raised against residues 3 to 347 ( 30). Antibodies against Kap3 and B56a were obtained from Transduction Laboratories (Lexington, KY).

Limited proteolysis. HeLa cell lysates prepared in lysis buffer (50 mmol/L Tris, 100 mmol/L NaCl, 5 mmol/L EDTA, 5 mmol/L EGTA, 40 mmol/L β-glycerophosphate, 0.5% NP40) and corresponding to 1 mg of total protein were immunoprecipitated with protein G-agarose (20 μL), and antibodies (5 μg) directed against the NH₂- or COOH-terminal domain of APC. Samples were split in half and either dephosphorylated with λ-phosphatase (New England Biolabs, Beverly, MA) or treated with control buffers. The agarose beads were then washed twice with lysis buffer and thrice with the protease buffer recommended by the manufacturer. Samples were then treated with a panel of proteases, including chymotrypsin (Sigma), Endo Asp-N (Roche), and Endo Lys-C (Roche) using 0.03, 0.06, or 0.12 μg per sample. Proteins were digested at room temperature for 30 minutes and proteolysis was stopped by the addition of 4× SDS sample buffer (Invitrogen) and immediate boiling.

Microtubule binding. Purified porcine brain tubulin was labeled with tetramethyl-rhodamine as described previously ( 29). Tubulin polymerization was carried out using unlabeled bovine brain tubulin and rhodamine labeled tubulin at a ratio of 7:1 and a final concentration of 2 mg/mL. Tubulin polymerization was induced by increasing the concentration of Taxol stepwise from 0.002 mmol/L to 0.02 to 0.2 mmol/L with 5-minute incubations at 37°C between each addition of Taxol (Universal Biologicals; Sigma); 4 μg of purified APC4 ( 29), which spans residues 1034 to 2843 of human APC, was incubated with either 5 μL polymerized microtubules or the corresponding amount of unpolymerized tubulin at room temperature for 15 minutes. Then 0.01 μg chymotrypsin was added for an additional 30 minutes at room temperature.

To measure binding of APC fragments to microtubules, 50 μg of cell lysate from Dictyostelium cells expressing different recombinant APC fragments was incubated with Taxol-stabilized microtubules at room temperature for 15 minutes.

To separate proteins bound to microtubules from unbound pools, reaction mixes were layered onto 150 μL of 40% glycerol in BRB80 [80 mmol/L K-pipes, 1 mmol/L MgCl₂, 1 mmol/L EGTA (pH 6.8)] and spun at 80,000 rpm in a TLA-100 rotor for 10 minutes at 30°C. The supernatant was removed and precipitated with 10 volumes of methanol. The pellet was directly dissolved in SDS-PAGE gel loading buffer.

In vitro interaction of adenomatous polyposis coli fragments. Cell lysate (0.5 mg) from Dictyostelium cells expressing recombinant N-APC, M-APC, or C-APC fragments fused to YFP were added to lysates from Sf21 cells containing the same fragments fused to HA. Concurrently, 40 μL of an anti-HA-antibody matrix (Roche) was preblocked with 0.4 mg/mL ovalbumin (Sigma) at 4°C for 2 hours. The matrix was then washed with lysis buffer thrice before mixed cell lysates were added for 2 hours at 4°C with rotation.

Immunofluorescence microscopy. Cells were seeded at 5 × 10⁴ cells per well into 6-well plates containing collagen-coated coverslips (Sigma). Cells were allowed to adhere and grow for 24 hours before transfection with different APC constructs. Forty-eight hours after transfection, cells were fixed with −20°C methanol for 5 minutes. The coverslips were treated with 4% donkey serum in PBS and 0.1% Triton X-100 for 1 hour. Endogenous, NH₂-terminal APC fragments were detected with either a rabbit anti M-APC polyclonal antibody (diluted 1:500; ref. 11), or a polyclonal anti-N-APC antibody (1:1,000; ref. 30) for 1 hour at room temperature. Tubulin was detected with a mouse monoclonal antibody (DM1a, Sigma; diluted 1:1,000). After washing with PBS, 0.1% Triton, the appropriate fluorescently conjugated secondary antibodies were applied (diluted 1:150; Jackson ImmunoResearch, West Grove, PA) and DNA was counterstained with 4′,6-diamidino-2-phenylindole (Sigma). Fluorescence microscopy was done using a DeltaVision restoration microscope (Applied Precision, Issaquah, WA).

Immunoprecipitation. HeLa cells grown to 70% to 80% confluency on 15-cm dishes were washed twice with cold PBS and scraped into 1 mL of ice-cold lysis buffer containing a cocktail of protease inhibitors (10 μg/mL Leupeptin, Pepstatin A, and Chymostatin A), 50 mmol/L sodium fluoride, 100 μmol/L sodium orthovanadate. After 5 minutes incubation on ice, insoluble material was pelleted by centrifugation and discarded. 0.5 mg protein from supernatant was incubated at 4°C for 2 hours with gentle agitation with 30 μL protein-G-agarose to which 8 μg of specific primary antibody had been prebound. Precipitated complexes were washed thrice with lysis buffer. Finally, the complexes were resuspended in 80 μL SDS-PAGE loading buffer containing 5 mmol/L DTT and boiled for 5 minutes before application to SDS-PAGE gels.

Antibodies and Western blotting analysis. Immunoprecipitates (20 μL) or total cell lysate (50 μg) were applied to 4% to 12% SDS/polyacrylamide gels (Novex/Invitrogen, Carlsbad, CA), transferred to Protran membranes (0.1 μm pore size, Schleicher and Schuell, Dassel, Germany), and probed with primary antibodies as described ( 29). Primary antibodies were diluted in block solution: N-APC 1:1,000; M-APC 1:2,000; C-APC 1:1,000, GFP 1:1,000 (Roche). Detection was done with horseradish peroxidase–conjugated either anti-rabbit or anti-mouse antibodies followed by enhanced chemiluminescence (Pierce, Tottenhall, United Kingdom).

Results

Proteolysis of adenomatous polyposis coli protein reveals the NH2-terminal third as the most proteolysis-resistant portion of the molecule and shows that phosphorylation may alter its conformation. Because phosphorylation is important for some of the functions of APC and is known to occur in cells, we first set out to test whether phosphorylation altered the protease accessibility of residues within the APC protein. We compared the digestion pattern of phosphorylated and nonphosphorylated APC. Full-length APC was immunoprecipitated from HeLa cells with monoclonal antibodies to either the NH₂- or COOH-terminal region of APC. The proteolytic digestion pattern of untreated (phosphorylated, “ctrl”) or dephosphorylated APC after treatment with a panel of proteases was then determined using domain-specific antibodies. The proteases we used included Asp-N, Lys-C, and chymotrypsin. We chose this panel of proteases because they have different specificities: Endopeptidase Asp-N cleaves at the NH₂-terminal side of aspartates and cysteic acids, chymotrypsin cleaves residues adjacent to large aromatic residues, and endopeptidase Lys-C cleaves COOH-terminal to lysine residues.

Figure 1 shows that the electrophoretic mobility of APC was visibly increased after phosphatase treatment confirming that APC is normally phosphorylated in cells. Phosphorylated APC was digested less efficiently by Asp-N and chymotrypsin but not Lys-C ( Fig. 1A) as apparent by the presence of large fragments in control but not dephosphorylated samples.

The full-length APC protein can be divided into N, M, and C parts, each comprising approximately one third of the length of the protein; the N-APC fragment (residues 1-1000), contains the armadillo repeats, the M-APC region (residues 1000-2000) contains binding sites for β-catenin, GSK, and Axin, and the C-APC region (residues 2000-2843) contains binding sites for microtubules, EB1, and PDZ domains. Untreated, phosphorylated APC was digested into a number of specific fragments. Based on their estimated size and reactivity with antibodies to the three regions, we were able to assign the approximate region where cleavage had occurred. For example, Asp-N produced a large fragment of about 220 kDa that reacted with N-APC and M-APC antibodies ( Fig. 1A and B, fragment 1) indicating that it was generated by cleavage within M-APC, near the boundary between M-APC and C-APC ( Fig. 1B). Alternatively, a ∼160 kDa fragment that only reacted with antibodies against M-APC and C- APC was also produced by this protease ( Fig. 1A and B, fragment 3) and this fragment is likely to represent the complementary COOH-terminal portion. Increasing the amount of Asp-N resulted in the formation of an 80- to 90-kDa fragment that only reacted with antibodies against the NH₂ terminus of APC suggesting that it represents an NH₂-terminal fragment ( Fig. 1A and B, fragment 2). Based on its size, it is likely that this fragment was the result of cleavage near the end of the armadillo repeats. Digestion with Lys-C produced a similar NH₂-terminal fragment ( Fig. 1A and B, fragment 2*) suggesting that the end of the armadillo repeats constitutes a distinct boundary between domains and is more readily accessibly than adjacent residues towards the NH₂ terminus. In addition, Lys-C produced a number of larger NH₂-terminally derived fragments of 110 to 160 kDa ( Fig. 1A and B, fragment 4). Only the larger ones of these reacted slightly with the M-APC antibody suggesting that they, too, were NH₂-terminal fragments that resulted from cleavage near the junction between the N- and M-regions of APC. Chymotrypsin also produced a number of similar NH₂-terminal fragments ranging from 90 to 130 kDa ( Fig. 1A and B, fragment 4*). The amount of these NH₂-terminal fragments was much reduced when APC was dephosphorylated before digestion with chymotrypsin. These experiments revealed that the following regions of the APC molecule are unlikely to lie in well-ordered, globular structures but reside in regions with little structure or are on the exterior of the molecule: the end of the armadillo repeat region around residue 800, the beginning and middle of the M-APC region (residues 1000-1400) and around residue 1700.

One striking observation was that the NH₂-terminal region was most resistant to proteases ( Fig. 1A) and relatively large fragments of about 90 to 160 kDa reactive with an antibody against this domain were found in digestions with all the proteases we used. Fragments that reacted with antibodies raised against the COOH-terminal domain could not be detected after any of the proteases treatments, suggesting this domain lacks extended regions of ordered structure and was easily digested ( Fig. 1A). The middle region behaved similar to the COOH-terminal domain and smaller fragments derived exclusively from this region were rarely observed. To exclude the possibility that the antibody to which APC was bound during the proteolysis had a protective effect on the NH₂ terminus, we also used a COOH-terminally directed monoclonal antibody to isolate APC protein in parallel experiments. The digestion patterns obtained with the anti-COOH-terminal APC antibody was indistinguishable from those shown in Fig. 1A (data not shown) making it unlikely that binding to a specific antibody, affected the conformation of APC or provided protection to the specific regions near the antibody-binding site. Furthermore, the relative abundance of predicted proteolytic sites for the proteases used showed no correlation with the observed cleavage pattern. In fact, there are many predicted cleavage sites within the NH₂-terminal domain (Supplementary Material).

Predicted structural domains and proteolysis-resistant domains coincide. To examine how the position of cleavage sites correlated with other structural features of the APC molecule we applied the Globplot algorithm to the human APC sequence. This algorithm gives an overall view of probabilities for regions within proteins to be disordered or globular ( 31) ³) and it also identifies known protein domains. A striking feature of the prediction was the high probability for ordered structure spanning the entire NH₂-terminal region as indicated by the long downward slope for this region of the molecule ( Fig. 1C). Similar predictions were obtained using the sequences of all known APC proteins in other species (data not shown). This prediction strongly supported our observation that such NH₂-terminal domains of APC were stable towards proteases. The algorithm also predicted the presence of two coiled-coils in the middle and end of the M-APC region. However, this was specific for human APC and was not detected in APC sequences from other species (data not shown). One of the predicted cleavage sites for Asp-N fell between these two coiled-coils ( Fig. 1C) consistent with the idea that the protease-accessible region lies between two well-ordered domains. The other estimated cleavage sites reside mostly in the middle of the molecule where only short stretches of well-ordered domains are predicted. Interestingly, the position of estimated cleavage sites in the middle of APC is similar to positions predicted to fall between small sections of order and disorder ( Fig. 1C).

Binding to microtubules protects the COOH-terminal part of adenomatous polyposis coli against proteases. The structural analysis using GlobPlot also revealed that the direct microtubule binding site in the COOH-terminal third of APC ( 15), is likely to be completely disordered as indicated by the upward slope predicted for this section of the molecule ( Fig. 1C). The previous finding that the structure of a purified protein fragment that represents this domain on its own was not altered significantly by heating to 70°C ( 15) is consistent with this prediction. We concluded that its disordered state is one reason this domain is so susceptible to degradation by proteases, which led us to speculate that this particular domain may assume structure when bound to its ligand, microtubules. To test this hypothesis, we compared the digestion of a purified COOH-terminal fragment of APC, comprising residues 1034 to 2843 of human APC (previously called APC4, spanning M-APC and C-APC; ref. 29), in the presence of microtubules or an equal amount of unpolymerized tubulin. When bound to microtubules, this fragment was not digested significantly and could be recovered in the microtubule-associated fraction after protease treatment ( Fig. 2 ). In the presence of an equal amount of unpolymerized tubulin, this fragment was almost completely degraded ( Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Binding to microtubules renders APC more stable. A fragment of APC spanning residues 1014 to 2843, APC4, was incubated with unpolymerized tubulin (Tu) or Taxol-stabilized microtubules (Mt), using equivalent amounts of tubulin. Samples were then exposed to chymotrypsin for 15 minutes. Microtubules and associated proteins were isolated by sedimentation through a glycerol cushion. Sedimented and soluble material was subjected to PAGE. APC fragments were detected using a COOH-terminal antibody. Tubulin was visualized on a Coomassie gel (bottom). Lane 1, APC4 with unpolymerized tubulin (Tu) or microtubules (Mt) before digestion (using 1/30 of the entire sample). Lane 2, the same samples immediately after digestion with chymotrypsin for 15 minutes (using 1/3 of the entire sample). Lane 3, soluble material after sedimentation through the glycerol cushion (using 1/2 of the sample). Lane 4, pelleted material from the sedimentation (using 1/2 of the sample). No protease inhibitors were added during the centrifugation step and this caused continued proteolysis in the supernatant to result in little protein recovered in this fraction [compare the total before centrifugation (lane 2 Tu) and the supernatant after centrifugation (lane 3 Tu)].

Intramolecular interactions of adenomatous polyposis coli in vitro. Our limited proteolysis experiments revealed the NH₂-terminal domain of APC as the region most resistant to proteases. On the other hand, the COOH-terminal domain of APC was easily digested by proteases but was significantly more resistant when bound to microtubules. Interestingly, phosphorylation decreases binding of APC to microtubules ( 29) but also increased its resistance to proteases ( Fig. 1A). These data suggest that both phosphorylation and protein interactions play an important role in regulating the conformation and thus the function of specific domains within the APC protein. We wanted to establish whether intramolecular interactions could be part of this regulation. To measure interactions between different domains of APC, recombinantly expressed APC fragments were mixed and their interaction tested. We used two different systems to express epitope-tagged versions of N-APC, M-APC, and C-APC. Fragments tagged with YFP were expressed in D. discoideum and the same fragments tagged with HA were expressed in Sf21 cells using baculovirus. Using antibodies to the HA-tag, fragments from Sf21 cells were immobilized and binding of YFP-tagged fragments present in Dictyostelium lysates was determined by immunoblotting material that bound to the immobilized fragments on the HA-affinity matrix. Using this in vitro approach, we confirmed that N-APC fragments interact with each other ( Fig. 3A ; refs. 32, 33) and we also made the intriguing observation that C-APC fragments seemed to bind to each other and thus may be capable of forming multimers ( Fig. 3A). Most notably, we found that the NH₂- and COOH-terminal third of APC bound to each other: soluble N-APC expressed in Dictyostelium cells (N_d) bound to immobilized C-APC ( Fig. 3A). Similarly, soluble C-APC from Dictyostelium cells (C_d) bound to immobilized N-APC ( Fig. 3A). The M-APC fragment was also retained by the C-APC fragment. However, the corresponding retention of the C-APC fragment by the M-APC fragment was not observed ( Fig. 3A). Furthermore, the M-APC fragment was detected in significant amounts in the controls leading to the conclusion that binding of the M-APC fragment was largely due to nonspecific interactions. The extremely high expression level of this protein in the Dictyostelium lysates may have contributed further to this nonspecific binding.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Intramolecular interactions of APC. A, lysates from D. discoideum expressing N-APC, middle-APC, or C-APC fragments (Nd, Md, Cd, respectively) fused to YFP were added to lysates from Sf21cells expressing the same panel of APC fragments with an HA-tag (Nb, Mb, Cb, respectively). Proteins were isolated using anti-HA-agarose, subjected to PAGE, and transferred to nitrocellulose. Bound APC fragments derived from Dictyostelium were detected with an anti-GFP antibody, which detects the YFP-tag. Control (Ctrl) samples represent proteins that bound to the HA-agarose resin when the Sf21 cell lysates were omitted (i.e., the “bait was not included in the experiment). B, SW480 cells were transfected with N-APC, M-APC, C-APC tagged with GFP or GFP alone. Expression of each protein was detected in 50 μg of whole cell lysates using an anti-GFP antibody. C, cell lysates from SW480 cells expressing N-APC (N), middle-APC (M), or C-APC (C) were immunoprecipitated with an anti-APC antibody directed against the NH₂ terminus of APC to isolate the endogenous NH₂-terminal fragment. Proteins were subjected to SDS-PAGE, transferred to nitrocellulose, and probed with antibodies against C-APC or M-APC. (This antibody detects the endogenous fragment of APC that is expressed in these cells and immunoprecipitates with the NH₂-terminally directed antibody). Only the exogenously expressed C-APC fragment bound to the endogenous N-APC in these cells. The additional bands detected with the C-APC antibody are nonspecific background bands and are detected in immunoprecipitates of untransfected (Ctrl) cells. Migration of molecular weight markers (left) and migration of the exogenous COOH-terminal fragments are indicated (arrow, right). D, lysates from Dictyostelium cells expressing either N-APC (N), middle-APC (M), or C-APC (C) were mixed with each other in the indicated combinations and then incubated with Taxol-stabilized microtubules. Microtubules and associated proteins were sedimented through a glycerol cushion and proteins in the pellet (P) and supernatant (S) were subjected to PAGE, transferred to nitrocellulose, and probed with an anti-GFP antibody to detect the different APC fragments. Migration position of the different APC fragments is indicated (arrows, right). Only a small amount of C-APC is detected because this protein is not expressed at high levels in Dictyostelium; furthermore, a proportion has lost its YFP tag. Nonetheless, the presence of C-APC leads to a reproducible shift of N-APC fragments to the microtubule-associated pool, whereas the proportion of M-APC that can be detected in the pelleted material does not change in the presence of C-APC.

C-APC coimmunoprecipitates with N-APC in cells. To validate the in vitro results, which suggested an interaction between the NH₂- and COOH-terminal domains of APC, we expressed our panel of APC fragments in tumor cells, which express a truncated form of APC spanning residues 1 to 1338, (N-APC plus a portion of M-APC). Expression of all three fragments could be detected with GFP antibodies; however, expression levels varied: M-APC was expressed most strongly and N-APC was only weakly expressed. Using domain-specific antibodies confirmed these proteins as the corresponding APC fragments (data not shown). Lysates from cells expressing these different APC fragments were immunoprecipitated with a monoclonal antibody directed against the NH₂-terminal region of APC to isolate the endogenous NH₂-terminal fragment in these cells. Bound proteins were then detected using antibodies against M-APC and C-APC ( Fig. 3C). Only exogenous C-APC but not M-APC specifically coimmunoprecipitated with the endogenous NH₂-terminal fragment ( Fig. 3C), although C-APC was expressed at significantly lower levels than M-APC. These data confirmed the ability of NH₂- and COOH-terminal domains of APC to interact in the context of cells.

C-APC recruits N-APC to microtubules. To further confirm the interaction between different APC fragments and to test whether these interactions affect the ability of APC to bind microtubules, lysates from Dictyostelium cells expressing different APC fragments were combined and then added to Taxol-stabilized microtubules. Proteins bound to microtubules were isolated by sedimentation through a glycerol cushion and probed with antibodies against GFP (which recognizes the YFP tag common to all three proteins); thus, all three APC protein fragments could be detected simultaneously. These experiments confirmed that only the COOH-terminal fragment sedimented efficiently with microtubules; neither N-APC nor M-APC on their own were enriched in the microtubule fraction as indicated by the finding that >90% of these proteins remained in the supernatant ( Fig. 3D). However, when lysates from N-APC- and C-APC-expressing cells were mixed, there was a significant shift of N-APC into the microtubule-associated fraction consistent with our findings that N-APC and C-APC can bind to each other. M-APC was not recruited to microtubules in the presence of C-APC as indicated by the absence of the full-length M-APC fragment in the pelleted, microtubule-associated material ( Fig. 3D). Importantly, the fraction of C-APC that bound to microtubules was not altered by the presence of N-APC. These results further support an association between NH₂- and COOH-terminal APC domains that is not disrupted by the presence of microtubules and does not affect the ability of C-APC to bind microtubules.

The microtubule-binding region of C-APC is required for the interaction between C-APC and N-APC. To determine the residues involved in the interaction between the NH₂- and COOH-terminal APC fragments, the interaction between the endogenous N-APC in SW480 cells and a number of C-APC fragments that differ in their ability to bind microtubules was determined ( Fig. 4 ). Deleting residues up to residue 2067 (“CΔ1”, this removes the third SAMP repeat) did not affect the relative amount of C-APC that was bound to N-APC so that the amount of this fragment recovered by immunoprecipitating N-APC was proportional to that present in the corresponding whole cell lysate ( Fig. 4A and B). However, deleting most (“CΔ2”) or all (“CΔ2+”) of the major microtubule-binding site almost completely abolished binding of C-APC to N-APC ( Fig. 4B), although similar amounts of all the C-APC fragments were expressed ( Fig. 4A). These data suggest that the region rich in positively charged amino acids between residues 2220 to 2400 is involved not only in the interaction of APC with microtubules but is also necessary for the interaction between of the NH₂- and COOH-terminal domains of APC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

The major microtubule-binding site of APC is required for the interaction between N-APC and C-APC SW480 cells were transiently transfected with GFP-tagged, COOH-terminal APC fragments lacking different regions as indicated. A, whole cell lysates were probed with a polyclonal antibody directed against the COOH-terminal domain of APC (residues 2813-2827) that is common to all the fragments so that expression levels could be compared directly. The following residues were contained in the various constructs: C, 2038-2843; CΔ1, 2068-2843; CΔ2, same as C but missing 2168-2451, CΔ1,2, same as CΔ1 but missing 2168-2451, CΔ2+, same as C but missing 2168-2590 ( 29). B-C, cell lysates from cells transfected with the indicated COOH-terminal APC fragments were immunoprecipitated with a monoclonal antibody directed against the NH₂-terminal region of APC to recover the endogenous NH₂-terminal APC fragment expressed in these cells. B, immunoprecipitated material was probed with the polyclonal anti-C-APC serum or (C) the monoclonal anti-N-APC antibody as indicated to confirm that equal amounts of N-APC were recovered from each cell lysate. Arrows point to the relevant bands. Only COOH-terminal APC fragments containing the major microtubule-binding domain bound to the NH₂-terminal fragment. Deleting a small region spanning residues 2038 to 2067 (CΔ1) did not compromise the interaction between N-APC and C-APC, whereas deleting most (CΔ2, CΔ1,2) or all of the dominant microtubule binding site (CΔ2+) rendered C-APC unable to bind endogenous N-APC, although comparable amounts of theses fragments were expressed. The migration of molecular weight markers of 220, 160, 120, 100, 90, 80, 70, 60, 50, 40, 30, and 20 kDa is shown (right).

Recruitment of NH₂-terminal adenomatous polyposis coli fragments in tumor cells to exogenous COOH-terminal fragments. To investigate whether interactions between COOH-terminal and NH₂-terminal APC fragments could affect NH₂-terminal APC fragments in tumor cells, we expressed C-APC in DLD1 cells. These cells only express truncated APC comprising residues 1 to 1418. The localization of the endogenous APC was compared in control and transfected cells. In untransfected cells or those transfected with GFP alone, the endogenous APC fragment was diffusely distributed throughout the cell ( Fig. 5 ). Most notably, when cells were transfected with C-APC, the endogenous N-APC fragment was recruited to microtubules, which were decorated with C-APC ( Fig. 5). A change in the distribution of APC was not observed in GFP-transfected control cells ( Fig. 5). These data confirmed that N-APC and C-APC domains can interact in the context of cells and showed that the presence of the COOH-terminal APC domains can alter the distribution of the NH₂-terminal fragment. Similar to the results from in vitro binding experiments, these data show that binding of C-APC to microtubules seems dominant because N-APC is recruited to microtubules that are decorated with C-APC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Microtubule associated C-APC fragment recruits N-APC fragments that are expressed in tumor cells. DLD1 cells were transfected with GFP-C-APC or GFP. Cells were fixed and then stained with antibodies to detect the endogenous NH₂-terminal APC fragment (B, F, K, O) and tubulin (C, G, L, P). GFP (I and N) or GFP-C-APC (A and E) were directly visualized. E-H and N-Q are enlargements of the regions indicated by the stippled boxes in (A-D) and (I-M). Bar, 15 μm (A-D and I-M) and 5 μm (E-H and N-Q). The increase in the endogenous N-APC observed in the cell in the middle that is transfected with higher amounts of C-APC was not always observed. Importantly, even in cells expressing low amounts of C-APC as shown in the enlargement (F), the redistribution of endogenous N-APC to microtubules was detected.

Interaction between N-APC and C-APC causes changes in the protein interactions of N-APC. To determine how the binding of C-APC to N-APC affected protein interactions of N-APC, we investigated the pattern of proteins that could be coimmunoprecipitated with the NH₂-terminal fragment in tumor cells in the absence and presence of C-APC. We found that the amount of Kap3 protein that could be detected bound to N-APC was significantly reduced in SW480 cells when they expressed C-APC compared with untransfected cells or those transfected with GFP, whereas the relative amount of B56α that bound to N-APC was increased, albeit slightly ( Fig. 6 ). The total amount of any of these proteins did not change and the amount of N-APC that was immunoprecipitated was similar under all conditions ( Fig. 6). These observations confirm that the interaction between C-APC and N-APC can indeed contribute to the regulation of the functions of the NH₂-terminal domain.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Changes in the protein interaction of N-APC in the presence of C-APC. A, equal amounts of protein from cell lysates from untransfected SW480 cells or those transfected with C-APC or GFP were prepared and subjected to PAGE and immunoblotted with antibodies against GFP, C-APC, Kap3, or B56α as indicated to reveal that the expression levels of these proteins was the same in all cells. The same lysates were immunoprecipitated with a monoclonal (B) or polyclonal (C) antibody against N-APC, and coprecipitated material was probed with antibodies against N-APC, Kap3, and B56α. Although identical amounts of the endogenously expressed N-APC was recovered in all cases, the amount of Kap3 that could be detected bound to the N-APC fragment was greatly reduced whereas the amount of B56a was slightly but reproducibly increased in cells also expressing C-APC. The migration of molecular weight (kDa) markers is shown (left).

Discussion

Our studies provide novel information about the structural organization and regulation of the APC molecule in cells. We obtained similar results using either immunoprecipitated full-length APC protein ( Fig. 1) recovered with antibodies to either NH₂- or COOH-terminal regions, or the largest APC fragment we can currently purify to homogeneity, which contains two thirds of the molecule, the middle and COOH-terminal regions of APC ( Fig. 2).

We found that most of the NH₂-terminal third of APC was resistant to proteases, whereas the middle and COOH-terminal regions were readily digested. This led us to conclude that the NH₂ terminus of APC is well folded, whereas the middle and COOH-terminal regions are more disordered. The approximate mapping of one of the proteolytically sensitive sites to around residue 1700 ( Fig. 1) was confirmed by the only successful attempt to obtain NH₂-terminal sequences from proteolytic APC fragments. In this preliminary experiment, two fragments starting with residues between 1750 and 1950 were detected. Our inability to obtain NH₂-terminal sequence information from many of the other fragments is consistent with the fact that the NH₂ termini of many proteins are blocked and further supports our result that NH₂-terminal fragments were most resistant to proteolysis.

These conclusions about the relative degree of organization of N-, middle-, and COOH-terminal regions of APC were strongly supported by predictions using GlobPlot, an algorithm that calculates the tendency within a protein for disorder and order/globularity (ref. 31). ³ These predictions were consistent for the sequences of all known APC proteins, further supporting their validity.

In particular, these predictions revealed the microtubule-binding site as a region singularly unstructured, consistent with our observation that this part of the APC molecule is extremely sensitive to proteolytic digestion. This led us to hypothesize that the COOH-terminal domain, particularly the region around the microtubule-binding site, undergoes a significant conformational change when APC binds to microtubules. Indeed, binding of APC protein to microtubules but not unpolymerized tubulin protected it against proteolysis ( Fig. 2).

We also found that phosphorylation rendered the NH₂-terminal APC region more resistant to degradation suggesting that phosphorylation of APC affects the folding of the molecule so that in its dephosphorylated state, APC is more unfolded. An alternative possibility is that proteins bound to APC were lost as a consequence of dephosporylation. However, not all proteases we used were sensitive to the phosphorylation to the same degree even when they cleaved in a similar region suggesting that overall conformational changes or protein interactions were not affected by phosphorylation.

Our observation that cleavage readily occurred between residues 800 and 1300 was particularly interesting because this is the region of APC that is most commonly mutated in cancer ( Fig. 3C). Indeed, the proteolytic fragments we detected correspond to fragments expected from the most common APC truncation mutations in cancers. ⁴ These observations suggest that selection for APC mutations may correlate with the stability of the resulting fragment and may relate to the structural arrangements of this particular region.

The site of the initial mutation in the APC gene seems to dictate whether the “second hit” is another truncation mutation or whether the remaining wild-type APC allele is simply lost ( 34, 35). This is consistent with the idea that APC mutations that confer the necessary advantages to cancer cells are not only selected for loss of functions but that the balance between lost COOH-terminal function and retained NH₂-terminal functions is also important ( 35). Indeed, we found that COOH-terminal fragments of APC bind to the NH₂-terminal fragments. Based on this finding, we proposed that normally, in the full-length APC protein, different functions are carefully balanced and that the interaction between NH₂- and COOH-terminal regions are an important feature of this regulation ( Fig. 7 ; ref. 36). In our model, loss of the COOH-terminal region due to tumor-associated APC truncation mutations, results in loss of this regulation and disruption of this balance, which could produce a dominant effect of the remaining NH₂-terminal domain that is expressed in tumor cells ( Fig. 7).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Model of how truncation mutations in APC may result in loss of the balance between different functions. A, in full-length APC, functions carried out by different domains are balanced. The interaction between the NH₂- and COOH-terminal regions may provide a key element in this balance. B, truncation mutations of APC like those found in cancer, eliminate COOH-terminal regions that normally contribute to the regulation of the molecule. Consequently, the balance is upset and functions performed by NH₂-terminal region that are normally carefully controlled can proceed unchecked.

The NH₂-terminal domain of APC contains armadillo domains, which have been identified as the binding site for a number of other proteins ( Fig. 1C). It is unlikely that all of these interactions can occur simultaneously and it is not clear which of these interactions may be most important for a dominant effect of NH₂-terminal fragments in tumor cells. Importantly, our data suggest that interactions between NH₂- and COOH-terminal domains may contribute to the regulation of the NH₂-terminal domain. Most notably, the amount of Kap3 bound to NH₂-terminal APC fragments in tumor cells was reduced when C-APC was also present. Kap3 can link APC to microtubule-based motors and thus constitutes in indirect link between APC and microtubules. Together with our finding that the direct microtubule-binding site in C-APC was required for the interaction of NH₂- and COOH-terminal regions of APC, these data suggest the following: first, that regulatory interactions between the NH₂- and COOH-terminal domains of APC are tightly linked to the interaction of APC with microtubules and second that they contribute to the intricate balance that normally exist between direct and indirect interactions of APC with microtubules.

One possibility is that the NH₂-terminal APC fragments commonly expressed in tumor cells affect cell migration and that this activity is normally regulated by its binding to COOH-terminal domains of APC in the full-length protein. However, testing whether cell migration is altered when COOH-terminal APC fragments are expressed in tumor cells with NH₂-terminal APC to force this interaction is complicated by the ability of such COOH-terminal fragments to bind microtubules and increase their stability ( 29). Another possibility is that regulation of the NH₂-terminal domain by the COOH-terminal domain of APC could be exerted by sequestering NH₂-terminal fragments to specific sites like microtubules. This idea is supported by our result that recruitment of C-APC to microtubules in cells was dominant and was not affected by binding to N-APC. This is consistent with our in vitro data showing that binding of C-APC to microtubules was not reduced when N-APC is present, although the same domain mediates both of these interactions.

Our data also showed that the amount of B56α bound to N-APC in tumor cells was slightly but reproducibly increased when NH₂-terminal fragments were bound to C-APC in cells. It is possible that this was simply a direct consequence of the loss of Kap3 which could have “made room” for B56α. Binding of B56α to APC is important for its role in Wnt signaling, whereas binding of APC to Kap3 is thought to play a role in the interaction of APC with microtubules and its targeting to microtubules ends. Although the increase in B56 bound to N-APC was small, these results raise the possibility that regulation of β-catenin and cytoskeletal organization by APC in tumor cells may be coordinated by interactions between different domains of APC.

In summary, we conclude that regulatory interactions between NH₂- and COOH-terminal regions of APC contribute to normal APC function and that the loss of such regulatory interactions in tumor cells may contribute to the dominant effect of tumor-associated NH₂-terminal fragments. Identifying the dominant effect that is most important to control in tumor cells expressing an “unrestrained” NH₂-terminal APC fragments could provide the basis for the design of therapeutics that specifically target APC mutant cells.