The Coiled Coil Region (Amino Acids 129–250) of the Tumor Suppressor Protein Adenomatous Polyposis Coli (APC)

The APC (adenomatous polyposis coli) tumor suppressor protein has many different intracellular functions including a nuclear export activity. Only little is known about the molecular architecture of the 2843-amino acid APC protein. Guided by secondary structure predictions we identified a fragment close to the N-terminal end, termed APC-(129–250), as a soluble and protease-resistant domain. We solved the crystal structure of APC-(129–250), which is monomeric and consists of three α-helices forming two separate antiparallel coiled coils. APC-(129–250) includes the nuclear export signal NES-(165–174) at the C-terminal end of the first helix. Surprisingly, the conserved hydrophobic amino acids of NES-(165–174) are buried in one of the coiled coils and are thus not accessible for interaction with other proteins. We demonstrate the direct interaction of APC-(129–250) with the nuclear export factor chromosome maintenance region 1 (Crm-1). This interaction is enhanced by the small GTPase Ran in its activated GTP-bound form and also by a double mutation in APC-(129–250), which deletes two amino acids forming two of the major interhelical interactions within the coiled coil. These observations hint to a regulatory mechanism of the APC nuclear export activity by NES masking.

The tumor suppressor gene APC 1 is inactivated in the majority of human colorectal cancers. One major function of the APC protein is the regulation of the level and intracellular localization of the signaling proto-oncoprotein ␤-catenin (1). Together with other proteins the APC protein increases the turnover rate of cytosolic ␤-catenin and thereby decreases the intracellular ␤-catenin level (2)(3)(4)(5). Stabilizing mutations in ␤-catenin or truncating mutations in APC lead to an increase of the intracellular ␤-catenin concentration (6 -8). As a consequence, ␤-catenin can enter the nucleus where it activates transcription in complex with a transcription factor of the T-cell transcription factor/lymphoid enhancer binding factor family (9,10). In addition to regulating the cytosolic ␤-catenin level, the APC protein can also regulate the intracellular localization of ␤-catenin and its nuclear signaling activity. The APC protein has a nuclear export function and may shuttle ␤-catenin between the nucleus and the cytoplasm (1). The nuclear accumulation of APC, induced by the inhibition of the exportin Crm-1 (chromosome maintenance region 1), leads to the accumulation of ␤-catenin in the nucleus (12)(13)(14). The APC protein contains several leucine-rich motifs, which might function as pertinent nuclear export signals (NESs), suggesting that APC interacts directly with Crm-1 (15). Two leucine-rich motifs, which were proposed as NES motifs, are located at the N terminus including residues 68 -77 and 165-174, respectively (12,14,15).

MATERIALS AND METHODS
Subcloning, Mutagenesis, Expression, and Protein Purification-The full-length APC cDNA was kindly provided by Paul Polakis (Genentech). The fragments APC-(2-250) and APC-(129 -250) were PCR-amplified. The double mutation L189A/I196A was introduced into APC-(129 -250) by site-directed mutagenesis using the QuikChange method (Stratagene) and verified by sequencing. The gene fragments were ligated into pGEX-6P-2 and recombinantly expressed as glutathione S-transferase fusion proteins in the Escherichia coli strain BL21. Bacteria were harvested and lysed by sonification, and proteins were purified by affinity chromatography. The glutathione S-transferase fusion part was cleaved off by incubation with 2 units of PreScission protease (Amersham Biosciences)/5 mg of protein at 4°C overnight, and the cleaved protein was further purified by gel filtration on a Superdex S200 column. The purification yielded 40 mg of protein of 99% purity from 1 liter of bacterial culture. To solve the phase problem APC-(129 -250) was expressed in a methionine-deficient E. coli strain in the presence of selenomethionine (Se-Met). Mass spectroscopic analysis confirmed the presence of three Se-Mets in purified APC-(129 -250)-Se (not shown). Crm-1 was subcloned in pQE60 (Qiagen), expressed as a His 6 -tagged protein, and purified on a Ni 2ϩ affinity column.
Biochemical Analysis-The trypsin-resistant domain was identified by incubating APC-(2-250) with trypsin at a 1:1000 ratio at 30°C. The reaction was followed by the time-dependent analysis of aliquots of the reaction by SDS-PAGE over 120 min. An aliquot of the 30-min fraction was analyzed by matrix-assisted laser desorption/ionization time of flight analysis. The N-terminal sequences of the trypsin-resistant protein fragments were determined by sequential degradation from the N-terminal end and the analysis of the amino acids by capillary high performance liquid chromatography. The potential formation of multimers of APC-(2-250) and APC-(129 -250) was followed by gel filtration on a Superdex S200 36/99 column of the glutathione S-transferase-free purified protein. The presence of secondary structural elements was analyzed by CD spectroscopy on a J-710 spectropolarimeter (Jasco). The interaction between APC, Crm-1, and Ran was analyzed in a pull-down assay (16). His 6 -tagged Crm-1 was immobilized to magnetic Ni 2ϩ beads (Qiagen) and precipitated using a magnetic separator. The pull-down assay was performed in the presence or the absence of the following proteins: APC-(129 -250) wild type or double mutated APC-(129 -250)-* This work was supported by Grant MU 1091/8-1 from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
L189A/I196A and Ran in its inactive GDP-bound form or in its active form bound to the GTP analogue GppNHp, respectively. Aliquots of the precipitates were analyzed for the presence of APC-(129 -250) by Western blot using a novel polyclonal antibody. This antibody had been raised against APC-(2-250) in rabbit. In a control Western blot the detection sensitivity of this antibody was demonstrated (Fig. 4A). The equal loading of all proteins was controlled by the Bradford method right before the assay. The equal concentrations of Crm-1 and Ran were also shown by bands at the corresponding sizes of comparable intensities on a Coomassie-stained gel, which was run in parallel as a control.
Crystallization and Structure Determination-Crystallization conditions for APC-(129 -250) were screened using several commercially available crystallization screens (Hampton Research) in hanging drop vapor diffusion set-ups at 4°C, 12°C, and 20°C with 1 ml of reservoir solution. The best conditions were 14 -20% polyethylene glycol 3350, 100 mM potassium iodide, 10 mM strontium chloride at 4°C with 20 -30 mg/ml protein for both the native and selenomethionine-labeled proteins. The crystals were cryoprotected in a solution containing 20% polyethylene glycol 3350, 10% xylitol, and 5% sucrose. Data sets were collected at 100 K to a resolution of 2.0 Å (native) and 2.5 Å (selenomethionine) on a MAR CCD detector at the ID14 -1 beam line (European Synchrotron Radiation Facility, Grenoble, France) using a wavelength of 0.934 Å and processed with XDS (17). The positions of the three selenomethionines were found by SOLVE (18) and used for SAD phasing including density modification (solvent flipping) with CNS (19). The excellent initial electron density allowed unambiguous tracing of the main chain and showed very good density for the side chains. The model was built with O (20) and initially refined against the 2.5-Å Se-Met data set using the SAD phase information. Subsequently the model was refined against the native data set with a resolution of 2.0 Å, and water molecules were added. The final model has excellent geometries with R cryst ϭ 25.7% and R free ϭ 30.9% and comprises residues 130 -173 and 179 -239. Since no signs for twinning or pseudosymmetry or other problems with the data could be detected, the high R free (even after extensive refinement) is most likely due to the disordered regions (17 of 122 residues, corresponding to 14%), and we thus conclude that the structure is correctly refined. The very good experimental electron density ensures that the structure is correctly built.

RESULTS
Structural Domains in the APC Protein-To identify structural domains of the APC protein we analyzed the 2843-amino acid APC sequence with the PHD secondary structure prediction program (21). Remarkably, less than 25% of the amino acids within the APC sequence are organized within secondary . The spectrum shows a major peak at 13,949 m/z. structural elements (not shown). The APC protein includes only two stretches of more than 250 amino acids with a portion of defined structural elements higher than 60%. First, the N-terminal 250 amino acids form the stretch with the highest portion of structural elements. This domain includes the Nterminal 55 amino acids, which form a coiled coil region that is responsible for the dimerization of the APC protein (22)(23)(24). The N-terminal 250 amino acids also include the two leucinerich motifs, residues 68 -77 and 165-174, which were proposed as NES motifs (12,15). Second, the helical region from amino acids 407 through 775 covers seven so-called ARMADILLO repeats, which show sequence homology to the repeat domain of the Drosophila protein armadillo (25,26). Thus, the large APC protein seems to represent a long string of single protein binding elements and of only a few structured domains, each with a distinct function, which are connected via variable stretches, rather than a compact globular protein with surface regions of which each is responsible for one of the numerous APC functions.
Based on these outcomes, we further analyzed the N-terminal region of the APC protein. At this stage of our study, there were only preliminary structural data available for this region (22,23). We purified the recombinantly expressed protein frag-ment APC-(2-250) and found that APC-(2-250) forms oligomeric complexes (Fig. 1A). The elution profile of the analytical gel filtration indicated the formation of tetramers and trimers, although we cannot rule out the additional formation of dimers. The dimerization of the very first 55 amino acids in solution has already been proven (22) and was consistent with the crystal structure, which was solved in the meantime and which also showed indications for the formation of tetramers of the first 55 amino acids (24). Remarkably the gel filtration showed the first indication for the presence of a stable domain within APC-(2-250), namely a degradation product of ϳ16.8 kDa (Fig.  1A). The results of the CD spectroscopic analysis of APC-(2-250) were in line with the presence of primarily helical segments, which were already proposed by the secondary structure prediction (Fig. 1B). Despite several attempts and the testing of many different screening conditions, we failed to crystallize APC-(2-250). To identify protease-resistant structural domains we performed limited trypsin digestion of APC- . This digestion resulted in one main fragment of 13,949 Da as determined by mass spectroscopy (Fig. 2). The N-terminal sequencing of this fragment revealed the sequence EST-GYLEEL, which is identical to the APC protein sequence from position 129. The C-terminal end of the fragment was identified where F obs denotes the observed structure factor amplitude and F calc denotes the structure factor amplitude calculated from the model. 10% of reflections were used to calculate R free . d Phasing power, as defined in CNS (19). e r.m.s., root mean square.
by virtual design of trypsin-digested APC fragments beginning at position 129 and comparison of the masses of these fragments to the mass of the experimental fragment using public software (expasy.ch). The designed trypsin fragments APC-(129 -246) and APC-(129 -250) with the masses of 13,952 and 14,492 Da, respectively, coincided best with the experimentally obtained fragment. Next APC-(129 -250) was subcloned, expressed, and purified. Remarkably APC-(129 -250) eluted as monomer and did not show any indication of dimerization in the analytical gel filtration (Fig. 1C). This result makes it unlikely that a putative dimerization domain within the region from amino acids 125 through 156 might be responsible for the different phenotypes of the inherited cancer predisposition familial adenomatous polyposis as suggested previously (24). CD spectroscopic analysis demonstrated the thermal stability of APC-(129 -250) and the presence of helical elements (Fig. 1D).
The Structure of APC-(129 -250)-The crystal structure of APC-(129 -250) was solved at 2-Å resolution (Table I). Of the 122 amino acids of APC-(129 -250), 99 amino acids are organized within three ␣-helices (Fig. 3): helix I from amino acids 132 to 172, helix II from amino acids 179 to 204, and helix III from amino acids 208 to 239. Residues 205 and 206 form a kink between helices II and III, whereas residues 129, 174 -178, and 240 -250 are not visible in the electron density and are assumed to be flexible in the crystal. The three helices form two separate antiparallel coiled coils (Fig. 3B). The first coiled coil is formed by residues 134 -148 of helix I and residues 217-231 of helix III. The second coiled coil is formed by residues 158 -172 of helix I and residues 186 -200 of helix II. The helices are interacting with each other via apolar residues at positions a and d of heptad repeat motifs (abcdefg) n , which are commonly found within coiled coils (27,28). In addition to the apolar interhelical interactions, there are six ionic interhelical interactions between charged residues of helices I and III. Remarkably there are no ionic interhelical interactions within the second coiled coil region indicating a weaker or less specific interaction between helices I and II compared with the interaction between helix I and helix III. The potential higher flexibility of this region is consistent with increased temperature factors in this region.
The three helices frame a hydrophilic centric zone (Fig. 3C). This zone is bordered by amino acids 147-155 and 210 -217 and the hydrophobic interactions between the residues Arg-217/ Leu-148 and Lys-155/Met-210, respectively. The centric zone contains a water molecule and the charged side chains of Glu-151, Glu-152, Arg-213, and Arg-217, which reach into this zone.
The Interaction of APC-(129 -250) with Crm-1-The nuclear export of proteins and RNA molecules is mediated by the functional protein family of exportins and the small GTPase Ran (29). The cargo to be exported from the nucleus forms a trimeric complex with an exportin protein and the active RanGTP. This complex is transported out of the nucleus where it dissociates. Preliminary data indicated that a peptide with the sequence corresponding to the NES-(165-174) of the APC protein can directly interact with Crm-1 (15). After mutation of Leu-172 and Leu-174 to Ala the interaction of the NES-(165-174) peptide with Crm-1 is abolished. When this double mutated peptide was fused to an indicator protein, this indicator protein accumulated in the nucleus (15). These studies indicated that this specific region of APC is a bona fide NES, which is active in vivo, and prompted us to investigate the direct interaction between APC-(129 -250) and Crm-1. In an affinity pull-down assay, APC-(129 -250) interacted directly with Crm-1 (Fig. 4B). The addition of Ran in its active GTP-bound form to the assay led to the increase of the precipitated amount of the Crm-1bound APC-(129 -250) in comparison to the amount of bound protein in the presence of RanGDP. Thus, active Ran enhances the interaction between APC-(129 -250) and Crm-1. In general, the interaction of Crm-1 with NES sequences is thought to be determined by hydrophobic residues, especially leucines, in relatively short sequence stretches (30 -32). The leucines are arranged mainly on one side of an amphipathic helix as exemplified by the NES in p53 (33). An alignment of the NES-(165-174) of APC-(129 -250) with two other known NES sequences is shown in Fig. 5A.
Remarkably the conserved hydrophobic residues Leu-165, Ile-169, and Leu-172 of NES-(165-174) are buried within the coiled coil of helices I and II (Figs. 3B and 5B). This leaves two possibilities to explain the interaction of this APC fragment with Crm-1: either Crm-1 interacts with a part of APC-(129 -250) different from the NES-(165-174) or the leucines of the NES-(165-174) are "unmasked" to be available for the interaction. On first sight and from the structural point of view, the first possibility is more likely since it is difficult to imagine that the coiled coil structure might unravel to expose the hydrophobic residues of the NES-(165-174). However, we also have indications for the second possibility (see below). The first possibility would be similar to the situation in the Rev protein of human immunodeficiency virus, type 1 where Rev is able to interact with Crm-1 both with the NES motif and in an NESindependent way. In this case, only the NES-mediated interaction was leptomycin B-sensitive and Ran-dependent (34). The finding that a small amount of APC-(129 -250) is precipitated by Crm-1 in the presence of inactive RanGDP is an experimental indication for an NES-independent interaction between APC-(129 -250) and Crm-1 (Fig. 4B, lane 4).
The second possible explanation for an APC/Crm-1 interaction via the NES-(165-174) is supported by several findings. The second coiled coil is formed by the five apolar interactions of the amino acids Tyr-158, Leu-162, Leu-165, Ile-169, and Leu-172 of helix I with the amino acids Arg-186, Leu-189, Ala-193, Ile-196, and Met-200 of helix II, respectively (Fig. 3B). The prevention of these interactions by site-directed mutagenesis should lead to a weaker coherence between the helices I and II and should thus facilitate the unmasking of the NES-(165-174). To verify this assumption, we tested the Crm-1 interaction with a double mutated form of APC-(129 -250) in the pull-down assay. In the mutant APC-(129 -250)-L189A/ I196A (Fig. 4B, mut) two of the main interhelical interactions are abolished, which should improve the unmasking of the NES-(165-174) and the APC-(129 -250) binding to Crm-1. Indeed, the mutation increases the amount of Crm-1-bound APC-(129 -250) in comparison to the amount of Crm-1-bound wild type APC-(129 -250) (Fig. 4B). The amount of the precipitated double mutant was increased more than 2-fold compared with the non-mutated control (Fig. 4C). Another sign for a specific interaction and thus for an unmasking of the NES-(165-174) upon interaction with Crm-1 and Ran is the enhancement of the binding of APC-(129 -250) to Crm-1 by the active form of Ran, RanGppNHp, but not by the inactive RanGDP. Furthermore, two findings derived from the three-dimensional structure of APC-(129 -250) demonstrate a high flexibility of this region of the coiled coil and thus indicate that detachment of the helices I and II could indeed be a possible mechanism. First, in contrast to the coiled coil between helices I and III, there are no ionic interactions between helices I and II. Helix II, which faces the NES-(165-174), has higher temperature factors than the rest of the structure, indicating that this helix is quite flexible and might be able to detach from the NES-(165-174). Second, the unusual "kink" between helix II and III would allow a movement of helix II without disturbing the second stable coiled coil moiety, thus giving a potential explanation for this unusual structural feature. DISCUSSION Here we present the three-dimensional structure of the largest APC protein domain solved so far. The APC-(129 -250) structure comprises three helices, which form two single coiled coils surrounding a hydrophilic center. The two coiled coil regions are separated by an unusual kink between helix II and helix III. A comparison via a DALI search (35) showed that there is no other coiled coil structure yet known with a prominent kink like the one in APC-(129 -250). The NES-(165-174) is part of one of the coiled coils. APC-(129 -250) was shown to interact with Crm-1, although the conserved residues of the NES-(165-174) are masked by interhelical interactions. Both active Ran and a double mutation in APC-(129 -250), which is assumed to make the separation of the two helices easier, can improve the binding of APC-(129 -250) to Crm-1. These results support the assumption that the masked residues of the NES-(165-174) might become accessible via a conformational change induced by the interacting partners Crm-1 and RanGTP.
The  in APC and the NES Motifs in Other Proteins-Leucine-rich stretches, which might exhibit nuclear export activity, have been identified in many different proteins. In most cases their activities have been proven only by using the short motifs fused to indicator proteins. There is only a small number of NES motifs, whose activities in the corresponding full-length proteins have been definitely proven in vivo. Of these NES motifs, only a very few have been structurally analyzed. That one has to be very careful when assigning NES motifs is shown by the structures of the leucine-rich motifs in 14-3-3 (amino acids 217-227) and in IB␣ (amino acids 265-274, the C-terminal NES), which are similar to the structures of the NES-(340 -350) in p53 and of the NES-(165-174) in APC. Recent studies have shown that these motifs have no nuclear export activity in the corresponding full-length proteins 14-3-3 and IB␣, respectively (16, 36 -38).
Structurally the NES-(165-174) in APC is mainly helical and localized at the end of helix I of APC-(129 -250) (Fig. 3, A and  B). A similar helix-terminal position has already been found for the NES motifs in p53 and STAT-1 (Fig. 5) (33, 39 -41). According to our knowledge, the only unambiguous NES, whose structure is known, is the NES-(340 -350) in p53. The NES-(308 -318) in STAT-1 is not rigorously confined; however, its nuclear export activity has not been disproved as in the case of 14-3-3 and IB␣. Therefore, the NES-(308 -318) in STAT-1 is included in the superpositioning (Fig. 5B). By analogy to the NES-(340 -350) in p53, we also found the unidirectional orientation of the first three conserved hydrophobic residues of the NES-(165-174) in APC, Leu-165, Ile-169, and Leu-172 (Fig.  5B). Leu-172 is in a slightly different position compared with this position in other known NES structures. This position seems to be less well conserved since p53 shows a slightly different position and STAT-1 lacks this position. Leu-174 of APC-(129 -250) is not visible in the electron density.
That an active NES can be masked has been shown already for two other proteins (42). First, in the tetrameric form of p53, the NES-(340 -350) is masked. Second, the NES-(308 -318) in STAT-1 is buried in a coiled coil structure similar to the NES-(165-174) in APC. The tyrosine phosphorylation of STAT-1 leads to the activation of the nuclear export activity (39). Thus, it was postulated that the phosphorylation leads to the unmasking of the NES-(308 -318) in STAT-1 (40). The Crm-1 binding might be enhanced by a conformational change, which leads to the free accessibility of a formerly masked NES. This conformational change might be induced by a third interaction partner. As an example, the interaction of the functional Nterminal NES of IB␣ with Crm-1 is activated by Ran in its active GTP-bound form (16). By analogy, active Ran enhances the interaction between Crm-1 and APC-(129 -250) (Fig. 4, B  and C). We also show that the interaction between APC-(129 -250) and Crm-1 can be improved by the deletion of side chains, which are important for the masking of the NES-(165-174) because they form contacts between the NES and helix II. These results make an unmasking mechanism induced by Crm-1 and active Ran very likely. Nevertheless we are aware that our in vitro results of the affinity pull-down assay have yet to be confirmed by data from in vivo experiments. Tumor-relevant Mutations in APC-(129 -250)-According to the APC gene mutation data base (p53.curie.fr), more than 5% of all mutations identified thus far in the APC gene are located in the gene region coding for APC-(129 -250). Nearly all of these mutations are stop mutations, which have been identified in the germ line of familial adenomatous polyposis patients. Only two mutations lead to amino acid changes indicating that the amino acids at these positions play an important role in the tumor-suppressing activity of the APC protein. First, the Trp to Leu mutation at position 157 is located at the middle of helix I (43). In the wild type APC protein Trp-157 points to the outside, perpendicular to the axis of helix I, and forms a prominent protrusion (not shown). Trp-157 is surrounded by a large hydrophobic patch stretching toward Ser-171. Thus, the interaction of helix I with other domains of the APC protein or with other proteins mediated by Trp-157 might be important for the tumor-suppressing activity of the APC protein. Second, the Ser to Ile mutation at position 171 is located within the NES-(165-174) (44) with the Ser side chain pointing toward the outside. A recent study showed that the serine phosphorylation of the NES-(14 -25) of p53 is important for the regulation of its nuclear export activity (41). Provided that the NES-(165-174) is important for tumor suppression and that it is regulated by phosphorylation, one could assume that the Ser to Ile mutation at position 171 might result in the decrease or the loss of the tumor-suppressing activity of the APC protein. Our finding that the software NetPhos (11) identified Ser-171 as a putative phosphorylation site supports this notion.