Regulated Binding of Adenomatous Polyposis Coli Protein to Actin*

Adenomatous polyposis coli (APC) protein is a large tumor suppressor that is truncated in most colorectal cancers. The carboxyl-terminal third of APC protein mediates direct interactions with microtubules and the microtubule plus-end tracking protein EB1. In addition, APC has been localized to actin-rich regions of cells, but the mechanism and functional significance of this localization have remained unclear. Here we show that purified carboxyl-terminal basic domain of human APC protein (APC-basic) bound directly to and bundled actin filaments and associated with actin stress fibers in microinjected cells. Actin filaments and microtubules competed for binding to APC-basic, but APC-basic also could cross-link actin filaments and microtubules at specific concentrations, suggesting a possible role in cytoskeletal cross-talk. APC interactions with actin in vitro were inhibited by its ligand EB1, and co-microinjection of EB1 prevented APC association with stress fibers. Point mutations in EB1 that disrupted APC binding relieved the inhibition in vitro and restored APC localization to stress fibers in vivo, demonstrating that EB1-APC regulation is direct. Because tumor formation and metastasis involve coordinated changes in the actin and microtubule cytoskeletons, this novel function for APC and its regulation by EB1 may have direct implications for understanding the molecular basis of tumor suppression.

APC 3 is a large multidomain (310 kDa) protein implicated in colorectal cancer (1,2). Mutations associated with tumorigenesis almost invariably delete the carboxyl terminus of APC, suggesting that key functions reside in this region of the protein.
Whereas truncations of APC result in deregulated Wnt signaling and cell proliferation, the earliest detectable defect in the intestinal epithelium from Apc mutant mice is not increased cell proliferation but rather changes in cell architecture and cell migration (3,4). This suggests that loss of cytoskeletal regulation by the carboxyl terminus of APC may represent a key step in tumor formation.
APC protein displays a well established localization at clusters of microtubule plus ends in cells (5)(6)(7)(8). In addition, APC has been detected at actin-rich plasma membrane surfaces and intercellular junctions by both immunofluorescence in highly polarized cells in culture (9 -11) and immunoelectron microscopy in mouse intestinal cells (12). APC also has been observed to associate with IQGAP at the actin-rich leading edge of migrating cells (13), and loss of Drosophila APC function disrupts formation of the transient actin furrows in the syncitial blastoderm (14). However, for a variety of reasons, APC association with the actin cytoskeleton has remained controversial. First, difficulties with antibody specificity have led to skepticism about non-microtubule-associated APC localization patterns (1,8,9). Second, molecular mechanisms connecting APC protein to the actin cytoskeleton have not been clearly defined, and no functional data have yet linked APC protein directly to the regulation of actin dynamics and/or organization.
In this study, we used purified proteins to demonstrate that APC protein directly binds to and bundles actin filaments and can physically cross-link actin filaments and microtubules in vitro. Consistent with these biochemical activities, we show that the same APC fragments localize to actin and microtubule structures in microinjected cells. In addition, these interactions are regulated by APC association with EB1, a microtubule plusend tracking protein that binds to and partially co-localizes with APC in vivo (15,16). Specifically, EB1 inhibited APC binding to actin filaments (F-actin) in vitro and association with actin structures in vivo. These data provide a mechanistic and functional explanation for previous observations of APC association with the actin cytoskeleton, further suggesting that APC protein may coordinate the functions of the actin and microtubule cytoskeletons in vivo.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-GST⅐APC-basic, GST⅐ APC-C1, GST⅐APC-C, GST⅐EB1-C, and GST⅐EB1-C(KR) were expressed and purified as described previously (6). For expression of APC-C in yeast, DNA encoding human APC residues 2130 -2843 was PCR-amplified and subcloned into the BamHI-NotI sites of a URA3 2 vector that contains an insertion of the GAL promoter and an amino-terminal His 6 tag (17). The resulting plasmid (pBG680) was transformed into the Saccharomyces cerevisiae strain BJ2168 (18), and His 6 ⅐APC-C was expressed, isolated onto nickel-nitrilotriacetic acid beads (Ni-NTA, Qiagen, Valencia, CA), and eluted as described. Eluted His 6 ⅐APC-C was further purified on a Superose-6 gel filtration column (GE Healthcare/Amersham Biosciences) equilibrated in IPD buffer (50 mM imidazole (pH 8.0), 1ϫ PBS (20 mM sodium phosphate buffer, 150 mM NaCl, pH 7.4), 1 mM DTT). Peak fractions were aliquoted, snap-frozen in liquid N 2 , and stored at Ϫ80°C.
For bundling assays in the presence of microtubules, GST⅐APC-basic-His 6 and GST⅐APC-C-His 6 were generated. DNA encoding human APC residues 2134 -2843 for APC-C and 2167-2674 for APC-basic was PCR-amplified with 3Ј-primer harboring His 6 tag and replaced the same fragment in pBG690 using PpuMI and HindIII sites for APC-C and in pBG693 using EcoRI-HindIII sites for APC-basic. The resulting plasmids (pBG812 for APC-C and pBG813 for APC-basic) were transformed into BL21pLysS, and protein expression was induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside at 37°C for 4 h. Both GST⅐APC-C-His 6 and GST⅐APC-basic-His 6 were purified first onto nickel-nitrilotriacetic acid beads, eluted with 100 mM imidazole, 0.6 M NaCl, 5% glycerol, pH 8.0, and dialyzed against 20 mM Tris-HCl, 100 mM NaCl, 5% glycerol, 0.5 mM DTT. Proteins were further purified onto glutathione-agarose beads (GE Healthcare/Amersham Biosciences) and eluted with 50 mM glutathione. Eluted proteins were concentrated with centricon-YM10 (Millipore, Bedford, MA), and the buffer was exchanged to 20 mM Tris-HCl, 100 mM NaCl, 5% glycerol, 0.2 mM DTT, pH 8.0, and snap-frozen in liquid N 2 to be stored at Ϫ80°C.
Rabbit skeletal muscle actin was purified as described (19). Unlabeled tubulin was purified from bovine brain as described (20), and fluorescein isothiocyanate-labeled tubulin was obtained from Cytoskeleton (Denver, CO).
Actin Filament Bundling Assays-In all cases, actin filaments were manipulated using precut pipette tips to prevent filament shearing. Low speed centrifugation assays were performed as described (21). Briefly, 18 l of preassembled F-actin and 12 l of IPD buffer or proteins in IPD buffer were mixed and incubated for 10 min at room temperature. Reactions were centrifuged for 3 min at 16,000 ϫ g, and the supernatant and pellet fractions were analyzed by SDS-PAGE and Coomassie staining.
For light microscopy (in Fig. 1D), 5 l of preassembled 6 M F-actin was mixed with 5 l of IPD buffer or APC-C in IPD buffer. After incubation at room temperature for 10 min, 3 l was removed and diluted in 15 l of rhodamine-conjugated phalloidin (1.5 M) and incubated for 5 min at room temperature. Aliquots were subsequently diluted 25-fold in fluorescence buffer containing 50 mM KCl, 1 mM MgCl 2 , 100 mM DTT, 20 g/ml catalase, 100 g/ml glucose oxidase, 3 mg/ml glucose, 0.5% methylcellulose, 10 mM imidazole, pH 7.0, and 3 l was added to a coverslip coated with poly-L-lysine for observation.
For electron microscopy, 2 M preassembled F-actin was incubated with 0.5 M APC-C for 10 min at room temperature. 3 l of this reaction was spotted onto a carbon film grid glowdischarged for 45 s in air with current of 20 mA. Samples were then negatively stained with 2% aqueous uranyl acetate and examined in a Morgagni 268 transmission electron microscope (FEI, Hillsboro, OR). Images were acquired using an AMT charge-coupled device camera (Advanced Microscopy Techniques Corp., Danvers, MA) at 80,000 kV accelerating voltage.
Cell Culture and Microinjection-Cell culture was performed as described previously (6,(22)(23)(24). NIH3T3 cells were grown in 10% calf serum and serum-starved at confluency for 48 h where indicated. APC proteins for microinjection were diluted to 2 mg/ml in HKCL buffer (10 mM Hepes, pH 7.2, 140 mM KCl), and we estimate that ϳ10% of the cell volume was routinely injected. In experiments including both APC and EB1, EB1 proteins were microinjected at 1.5-fold molar excess to APC. Following microinjection of GST⅐APC-basic ( Fig. 4), cells were incubated for 1 h and then fixed in either 4% paraformaldehyde or Cytofix reagent (BD Biosciences). Cells microinjected with APC-C and EB1 in the absence of nocodazole ( Fig. 8) were incubated for 1 h and then fixed in Cytofix. To observe APC-C localization to the actin cytoskeleton ( Fig. 9), cells were microinjected with GST⅐APC-C (or GST⅐APC-C and GST⅐EB1-C). 20 min after incubation, cells were treated with 20 M nocodazole, incubated for 2 h, and subsequently fixed in Cytofix.
Immunofluorescence-For detection of microinjected GST⅐ APC-basic and GST⅐APC-C in the absence of other GST-fused proteins, cells were stained with monoclonal antibody to GST (1:100, 26H1, Cell Signaling Technology, Beverly, MA). Depending on the experiment, cells were also stained with rat monoclonal antibody YL1/2 to detect dynamic Tyr-microtubules (1:10, European Collection of Animal Cell Cultures, Salisbury, UK) or with Alexa Fluor 647-conjugated phalloidin (Invitrogen) to detect F-actin. To detect GST⅐APC-C co-microinjected with GST⅐EB1-C, cells were stained with anti-APC rabbit polyclonal antibody C-20 (1:50, Santa Cruz Biotechnology, Santa Cruz, CA). Secondary antibodies with minimal crossspecies reactivity (Jackson ImmunoResearch Laboratories, West Grove, PA) were used as described (22,23). Epifluorescence microscopy was performed as described (6,23).

RESULTS AND DISCUSSION
The APC Basic Domain Bundles Actin Filaments-The carboxyl-terminal third of human APC protein (APC-C; residues 2130 -2843; Fig. 1, A and B) contains its microtubule-binding basic domain followed by a short EB1-binding domain. This region of the APC protein is sufficient for localization to microtubule plus ends in vivo (7). Further, tumorigenic mutations that truncate this region of APC disrupt its localization to the actin-rich lateral membrane of cultured cells (9). To investigate possible APC-actin interactions, we purified the human APC-C polypeptide using a yeast expression system and tested its ability to bind directly to F-actin in vitro. In low speed centrifugation (16,000 ϫ g) assays, F-actin pellets specifically when filaments are organized into higher order structures by actinbinding proteins, e.g. the bundling protein coronin/Crn1 (Fig. 1C). The addition of purified APC-C to 3 M F-actin induced pelleting of F-actin at low speeds, an effect that was concentration-dependent (Fig. 1C). To visualize the F-actin structures formed, reactions containing 3 M F-actin mixed with 0.5 M APC-C or control buffer were fixed, stained with Alexa488-conjugated phalloidin, and examined by light microscopy. Single actin filaments were observed for reactions containing control buffer, whereas tight bundles of filaments were formed by APC-C (Fig.  1D) and by a known bundling protein, Crn1/coronin (25). 4 We also observed APC-C co-pelleting with F-actin. Thus, APC protein both binds directly to F-actin and bundles filaments.
We next used electron microscopy to examine F-actin bundled by APC-C. In control reactions lacking APC-C, single actin filaments were observed ( Fig. 2A). By contrast, F-actin mixed with APC-C was organized into tight, ordered bundles (Fig. 2, B and C). The ordered nature of these bundles demonstrates that the APC-actin interaction is highly specific. This specificity is further supported by the concentration-depend-ent nature of binding and by the ability of two different APC ligands to competitively displace its binding to actin (see below). Additionally, these effects were not due to protein aggregation, because APC-C (and all other APC constructs used in this study) was soluble and monodispersed by gel filtration analysis (see Fig. 6). 4 APC protein functions can be regulated by phosphorylation (e.g. GSK3␤ decreases its interactions with microtubules (26)). Because the APC-C used for the experiments above was expressed and purified from S. cerevisiae, we investigated the possibility that post-translational modification might contribute to its interactions with actin. Two assays showed that APC-C was phosphorylated (Fig. 3, A and B). First, treatment with -phosphatase altered APC-C migration on SDS-PAGE. Second, a fluorescent phosphoprotein dye detected APC-C in protein gels, and the signal was abolished by treatment with -phosphatase. Therefore, we compared the ability of 1 M yeast-expressed APC-C to bundle actin filaments after dephosphorylation with -phosphatase or mock treatment. Dephosphorylated APC-C bundled actin more efficiently than mocktreated APC-C (100 versus 50% F-actin in the pellet; Fig. 3B), suggesting that phosphorylation of APC-C diminishes its abil-   ity to bundle F-actin. This conclusion was supported further by a close examination of the gels analyzing low speed pelleting reactions (Fig. 1C), which revealed that the fraction of APC-C that co-pelleted with bundled F-actin had a distinct migration from the fraction that remains in the supernatant.
To avoid potential complications from using post-translationally modified APC protein, we next isolated APC polypeptides as GST fusion proteins from Escherichia coli. APC-C purified from E. coli bundled F-actin efficiently 4 (also see Fig. 7C). Similar results obtained for APC-C purified with His 6 or GST affinity tags suggested that neither tag affected APC function.
To better understand what region of APC-C contains the F-actin bundling activity, we purified two different subfragments of APC-C from E. coli, the basic domain (APC-basic) and the EB1binding domain (APC-C1) (Fig. 1B). In low speed centrifugation assays, APC-basic but not APC-C1 bundled F-actin in a concentration-dependent manner (Fig. 3, C and D). These results demonstrate that the APC basic domain, which also binds to microtubules (27), is sufficient to bind to and bundle actin filaments in vitro. It is intriguing that APC interactions with both microtubules and F-actin are mediated by the basic domain, as polybasic oligomers are sufficient to interact with both microtubules (28 -31) and actin (32,33).
APC Basic Domain Associates with the Actin Cytoskeleton in Cells-To address APC-basic association with the actin cytoskeleton in cells, we microinjected NIH3T3 fibroblast cells with purified GST⅐APC-basic and examined localization by immunofluorescence using anti-GST antibodies. In most cells, GST⅐APC-basic associated with both microtubules and actin stress fibers (Fig. 4, A-D). However, in a small percentage of cells, GST⅐APC-basic localized almost exclusively to actin stress fibers (Fig. 4, E-H). In serum-starved cells, which have few actin stress fibers, GST⅐APC-basic was mostly localized on microtubules (Fig. 4, I-L). GST alone did not localize to actin stress fibers or microtubules (Fig. 4, M-P). These in vivo results correlate with APC binding to and bundling F-actin in vitro and provide a direct molecular mechanism and function for APC association with the actin cytoskeleton. Key to this interaction is the APC basic domain, which our data show is sufficient for F-actin bundling in vitro and association with the actin cytoskeleton in cells.
Microtubules Regulate F-actin Bundling by APC-The basic domain of APC protein is known to bind, stabilize, and bundle microtubules in vitro (26,27,34) and contributes to APC association with microtubules in cells (7,26,27). Our finding that this same region of APC protein interacts with F-actin raises the possibility that APC-basic either forms independent nonoverlapping interactions with F-actin and microtubules or, alternatively, forms competitive overlapping interactions. To address these possibilities, we tested the ability of APC protein to bundle F-actin in the presence of microtubules in low speed centrifugation assays. For these quantitative analyses, we isolated APC-basic and APC-C from E. coli using a two-step affinity purification strategy employing tags on either end of APC (see "Experimental Procedures") to remove any degraded proteins. In the absence of microtubules, half-maximal actin filament bundling was observed at 100 nM APC-basic (Fig. 5A). In reactions containing F-actin and 100 nM APC-basic, the further addition of increasing concentrations of Taxol-stabilized microtubules inhibited F-actin bundling. Half-maximal inhibition of F-actin bundling was reached in the presence of 0.2 M microtubules, and inhibition was nearly complete at 1 M microtubules (Fig. 5B). These data support the model that APC protein interactions with F-actin and microtubules are competitive and/or overlapping. In addition, note that in the presence of 2 M microtubules, where no F-actin was detected in the pellets, microtubules pelleted (supplemental Fig. S1), indicating that APC-basic bundles the microtubules. Similar results were obtained using APC-C. 4 We also used light microscopy to visualize the effects of APC-C on the organization of fluorescently labeled F-actin and/or microtubules (Fig. 5C). Consistent with the low speed pelleting results, APC-C induced formation of visible F-actin bundles in the absence of microtubules and induced formation of microtubule bundles in the absence of F-actin. Interestingly, at equal concentrations of F-actin and microtubules (2 M each), APC-C organized bundles of microtubules that were not associated with F-actin. This suggests that APC-C may have a higher affinity for microtubules than F-actin and confirms our conclusion from low speed pelleting assays that microtubules inhibit APC-C interactions with F-actin. Intriguingly, including a lower concentration of microtubules (0.5 M) led to visible cross-linking of F-actin and microtubules, demonstrating that APC protein has the capacity under some conditions to physically link actin and microtubule polymer systems. It is presently unclear what mechanism allows APC to bundle actin filaments and microtubules and cross-link the two systems. This may be accomplished either by APC having multiple binding sites (for actin and microtubules) in one polypeptide chain or, alternatively, by APC oligomerization. Analytical gel filtration analysis of purified APC-C is consistent with the latter possibility (Fig.  6), but it is also possible that this reflects APC-C having a highly extended conformation. Clarification of this mechanism by additional biophysical tests will require isolating larger quantities of APC-C than we were unable to obtain using the methods described here. Thus, it will be important to develop high yield APC-C purification schemes in the future.
EB1 Directly Inhibits F-actin Bundling by APC-We next investigated whether the interactions of APC-C with F-actin might be regulated by EB1, an in vivo binding partner of APC. EB1 is a microtubule plus-end tracking protein that binds directly to the carboxyl terminus of APC (15,35), adjacent to  the APC basic domain (Fig. 1A). This interaction contributes to the formation of stable microtubules in cells (6) and has been suggested to promote microtubule formation in vitro (36). The carboxyl terminus of EB1 (EB1-C) binds to APC (see Fig. 7A) and to the p150 Glued subunit of dynactin (15,(37)(38)(39), whereas the amino terminus associates with microtubules and the formin Dia2 (6). Although APC protein and EB1 clearly interact and function together in vivo, they display only partial co-localization in cells, and APC-EB1 interactions are not required for APC protein localization to and stabilization of microtubules (16,26,40,41). Thus, EB1 and APC protein may share some cellular functions (e.g. on the ends of some microtubules), but they also appear to have independent functions.
In low speed centrifugation assays, the addition of EB1-C abolished APC-C bundling of F-actin (Fig. 7C). Two complementary experiments showed that this inhibition requires direct interactions between EB1-C and APC-C. First, the EB1 construct EB1-C(KR), which contains point mutations that disrupt APC-binding interactions (K220A, R222A) (6,42), had no effect on APC-C bundling of F-actin (Fig. 7C). In control reactions lacking APC-C, wild type and mutant EB1-C proteins had no effect on F-actin pelleting. Second, EB1-C had no effects on actin bundling by APC-basic (Fig. 7D), which lacks the EB1 binding site (see Fig. 1A). Thus, EB1 directly inhibits APC protein bundling of F-actin.
EB1 Directly Inhibits APC Association with the Actin Cytoskeleton in Cells-Finally, we examined the ability of EB1 to regulate APC interactions with the actin cytoskeleton in cells. In microinjection experiments, APC-C associated with microtubules in NIH3T3 fibroblast cells (Fig. 8, A-C). However, upon disruption of the microtubule cytoskeleton with nocodazole, APC-C co-localized with the actin cytoskeleton and predominantly decorated stress fibers, similar to APC-basic (Fig. 9, A-C). Next, we co-microinjected EB1-C and APC-C into nocodazole-treated cells and detected APC-C localization by immunofluorescence using the C-20 antibody, which recognizes the APC carboxyl terminus. Co-microinjection of EB1-C led to a diffuse cytoplasmic localization of APC-C (Fig. 9, D-F). In contrast, co-microinjection of EB1-C and APC-C into cells with an intact microtubule cytoskeleton (not treated with nocodazole) did not affect APC-C localization to microtubules (Fig. 8, D-F). These results suggest that binding of EB1 specifically disrupts APC protein association with F-actin and not microtubules in cells, similar to the effects we observed in vitro. Further,   these effects in cells similarly require direct EB1-APC interactions, as co-microinjection of APC-C and EB1-C(KR) did not alter APC-C localization to the actin cytoskeleton (Fig. 9, G-I). We note that the effect of EB1-C in this experiment may not be absolute, as a minor degree of co-localization between APC-C and F-actin may remain. However, the robust distinction between APC-C localization in the presence of wild type EB1-C (Fig. 9D) versus mutant EB1-C(KR) (Fig. 9G) shows conclusively that EB1 directly inhibits APC association with the actin cytoskeleton in cells.
Taken together, these results suggest that EB1 directly inhibits APC protein association with F-actin and may thereby promote association with microtubules. In this capacity, EB1 may function as a regulatory switch, altering APC protein between microtubule plus ends and the actin cytoskeleton. Transient APC-EB1 interactions may promote APC functions at microtubule plus ends, for which APC protein has high affinity. Other molecular cues, including localized microtubule disassembly at the cortex, may promote APC protein interactions with the actin cytoskeleton. Consistent with this view, EB1 and APC only partial overlap in their cellular localizations (16), and EB1 is required for APC function at only a subset of microtubules in cells (6,16). We have also shown that APC protein has the ability to cross-link actin and microtubule polymers in vitro. Thus, APC may be involved in linking functions of the microtubule and actin cytoskeletons, such as at cortical sites of actin-microtubule overlap.
Conclusions-Our results identify a new activity for APC protein in binding directly to and bundling actin filaments, providing a functional explanation for APC protein localization to actin-rich cellular regions. Further, we have shown that EB1 directly controls APC protein associations with microtubules versus F-actin. This work raises a number of questions about the APC mechanism of interaction with F-actin and its role in regulating actin dynamics and organization in cells. Future experiments will be directed at mapping APC residues that interact with actin and identifying sites of direct APC-actin association in vivo. Coordination of the actin and microtubule cytoskeletons by APC protein is likely to be regulated on many additional levels, including the APC ligands Dia (6) and IQGAP (13), ␣-catenin/␤-catenin (43,44), and possible intramolecular interactions between the amino and carboxyl termini of APC (45). Because the APC protein carboxyl terminus is deleted in most tumorigenic APC mutations, it is tempting to speculate that this cytoskeletal function is involved in APC tumor suppressor function. In particular, we note that APC protein has been localized to actin-rich intercellular junctions (10,11), which can be found in the intestinal epithelium. The ability of APC protein to bundle F-actin and to link the actin and microtubule cytoskeletons may be required for maintenance of cellcell adhesion, and disruption of these activities may contribute to uncontrolled cell migration and/or Wnt signaling, with implications for tumor formation. Alternatively, loss of APC interactions with actin and/or microtubules may impair cell migration from the proliferative zone of the intestinal epithelium, leading to the accumulation of cells and tumor formation.