The beta-amyloid precursor protein functions as a cytosolic anchoring site that prevents Fe65 nuclear translocation.

In this study we addressed the question of the intracellular localization of Fe65, an adaptor protein interacting with the beta-amyloid precursor protein (APP) and with the transcription factor CP2/LSF/LBP1. By using tagged Fe65 expression vectors, we observed that a significant fraction of Fe65 is localized in the nucleus of transfected COS7 cells. Furthermore, the isolation of nuclei from untransfected PC12 cells allowed us to observe that a part of the endogenous Fe65 is present in the nuclear extract. The analysis of Fe65 mutant constructs demonstrated that the region of the protein required for its nuclear translocation includes the WW domain, and that, on the other hand, a small fragment of 100 residues, including this WW domain, contains enough structural information to target a reporter protein (green fluorescent protein (GFP)-GFP) to the nucleus. To evaluate whether the Fe65-APP interaction could affect Fe65 intracellular trafficking, COS7 cells were cotransfected with APP(695) or APP(751) and with GFP-Fe65 expression vectors. These experiments demonstrated that Fe65 is no longer translocated to the nucleus when the cells overexpress APP, whereas the nuclear targeting of GFP-Fe65 mutants, unable to interact with APP, is unaffected by the coexpression of APP, thus suggesting that the interaction with APP anchors Fe65 in the cytosol.

The ␤-amyloid precursor protein (APP) 1 is a widely expressed integral membrane protein that is involved in the pathogenesis of Alzheimer's diseases. In fact, the main constituent of the Alzheimer's disease senile plaques, the ␤-amyloid peptide, derives from APP as a consequence of its proteolytic processing (for a recent review see Ref. 1). The overall structure of APP resembles that of a receptor or growth factor precursor, but only a small fraction of APP resides at the plasma membrane (2,3). Part of the protein reaching the cell surface, through the secretory pathway, undergoes a juxtamembrane cleavage by the action of the ␣-secretase-ADAM10 protease (4) that results in the generation of a secreted, soluble form of APP. The rapid turnover of surface APP drives the protein to the endocytic compartment and then again to the endoplasmic reticulum-Golgi network (5,6), where it undergoes cleavage by another proteolytic activity named ␤-secretase-BACE-1 (7-10), which generates a transmembrane 12-kDa fragment. This fragment is then cleaved by the ␥-secretase activity, which leads to the formation of ␤-amyloid peptide. Recent evidence strongly suggests the identity of ␥-secretase and presenilin dimer (11,12).
The elucidation of the normal functions of APP could be important in understanding the molecular basis of Alzheimer's disease and in approaching its prevention and/or treatment, but such elucidation has been elusive up to now. One of the characteristics of this protein currently investigated is the complex network of protein-protein interactions that is centered at the short cytoplasmic tail of APP (13,14). Many proteins, in fact, interact with this C-terminal domain of APP, most of them possessing multiple protein-protein interaction domains, which in turn form complexes with other proteins, suggesting that these proteins function as adaptor proteins bridging APP to specific molecular pathways.
Three of these adaptor proteins belong to the Fe65 protein family; Fe65 was originally described as a protein highly expressed in neurons of specific areas of the mammalian nervous system and also possessing some characteristics of a transcription factor (15,16). It contains three protein-protein interaction domains, a WW and two PTB domains. The PTB2 domain, located in the C-terminal half of the molecule, is responsible for the interaction of Fe65 and of the two related proteins Fe65L1 (17) and Fe65L2 (18) with the cytosolic tail of APP (19 -21). Similarly, all three members of the Fe65 family interact with two APP-related proteins, APLP1 and APLP2 (17,18). These interactions take place at the level of the ⌽XNPXY motif (where ⌽ indicates a hydrophobic residue, and X indicates any amino acid) present in the cytodomains of APP and of its related proteins, and, differently from that observed for other PTB domains, the interaction does not require the phosphorylation of the tyrosine residue (20,22). Other molecular adaptors, possessing a single PTB domain, bind to the same ⌽XNPXY motif of APP: the proteins belonging to the X11 protein family (20,(23)(24)(25) and m-Dab1 (26,27), the mammalian orthologue of the product of the disabled gene of Drosophila. The formation of these complexes seems to affect the regulation of APP processing, considering that the expression of Fe65 leads to an increased generation of ␤-amyloid peptide in cultured cells (28,29), whereas X11-APP coexpression inhibits the proteolytic processing of APP (25,30,31). Lastly, APP seems to have other partners also, not possessing a PTB domain, that bind its cytodomain, such as a Go protein (32) and two more proteins named PAT-1 (33) and APP-BP1 (34).
This scenario appears even more complex when the molecular partners of the APP ligands are examined. The WW domain of Fe65, in fact, interacts with several proteins, one of which is Mena, the mammalian orthologue of the product of the enabled gene of Drosophila, which is a genetic suppressor of the phenotype induced in Drosophila by the abl gene mutation (35). The second PTB domain of Fe65 (PTB1) interacts with the transcription factor CP2/LSF/LBP1 (36) and, at least in vitro, with the low density lipoprotein receptor-related protein (27). On the other hand, particularly at the presynaptic level, X11 forms a trimeric complex in vivo with CASK and VELI, the orthologues of Caenorhabditis elegans Lin2 and Lin7 proteins, respectively (37)(38)(39), and interacts with Munc18 (40) and a calcium ion channel (41).
Considering that the intracellular trafficking of APP appears to be relevant for the different proteolytic fate of the molecule and, possibly, for its function(s), it is important to analyze the intracellular localization of proteins interacting with the cytodomain of APP, to evaluate the possible association of any APP partner to specific APP trafficking and proteolytic pathways. The ligands of Fe65 identified up to now suggest multiple intracellular locations of this adaptor protein, considering that Mena is mainly connected with the actin cytoskeleton remodelling system (42) and that the known functions of CP2/LSF/ LBP1 take place in the nucleus. In this paper we addressed the question of the intracellular localization of Fe65 and its relationship with that of APP. We found that transient overexpression of Fe65 in cultured cells results in its accumulation in the nucleus and that endogenous Fe65 is present in the nuclear extract from PC12 cells. The cytosol-nuclear translocation of Fe65 takes place through an unusual pathway requiring the region that includes the WW domain and is prevented by the coexpression of APP, whose cytodomain functions as a cytosolic anchoring element.
The Fe65 point mutant C655F cDNA was obtained by using the QuickChange kit (Stratagene) according to the manufacturer's procedure. For the mutagenesis, the pEGFP-C1-Fe65 expression vector was used as a template, and the following pair of oligonucleotides were used as primers: forward 5, 5Ј-CTGTGCAGGCTGACATTCATGCTCCGC-TAC and reverse 6, 5Ј-GTAGCGGAGCATGAATGCAGCCTGCACAG. Bold characters indicate the mutated residue.
The sequence and the reading frame of all the recombinant constructs were checked by nucleotide sequence by using the Sequenase kit (Amersham Pharmacia Biotech), and the expression and size of the fusion proteins was confirmed by Western blot analysis.
The CMV-Fe65-HA vector expressing wild type rat Fe65 tagged with the HA epitope was obtained by cloning in the HindIII site of pRC-CMV (Invitrogen), a polymerase chain reaction fragment obtained by using the following primers: forward 6, 5Ј-CCCAAGCTTACTAAGGCCAT-GTCTGTTCCA and reverse HA, 5Ј-CCCAAGCTTTCAAGCGTAAT-CTGGAACATCATATGGGTATGGGGTCTGGGATCCTAGAC.
Cell Culture Conditions, Transfections, and Fluorescence Microscopy-Cells were grown at 37°C in the presence of 5% CO 2 ; COS-7 cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (HyClone) and 1% penicillin/streptomycin mixture (HyClone). PC12 cells were cultured in RPMI (Life Technologies, Inc.) supplemented with 10% horse serum (Life Technologies, Inc.), 5% fetal bovine serum (HyClone), and 1% penicillin/streptomycin mixture (HyClone). For the expression of the GFP-Fe65 fusion proteins or of Fe65-HA, 1.5 ϫ 10 6 cells were transfected by electroporation at 250 microfarad and 220 V with 20 g of each construct. For cotransfection experiments 10 g each of the GFP-Fe65 constructs and of the CMV-APP expression vectors, carrying the wild type APP 695 or the APP 751 cDNAs, were used.
For fluorescence microscopy, transfected COS-7 cells were grown on coverslips. After 36 h cells were washed with PBS and fixed with paraformaldehyde (4% in PBS, pH 7.4) for 20 min at room temperature. The cells were then washed once in PBS-glycine and twice in PBS. When indicated, the cells were permeabilized with 0.1% Triton X-100. APP was stained with 6E10 or 369 antibodies, using Texas red-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories). Fe65-HA was stained with the anti-HA polyclonal antibody Y-11 (Santa Cruz Biotechnology).
The coverslips were mounted in 50% glycerol solution onto a glass microscope slide and viewed using an Axiophot microscope (Zeiss). For fluorescence observation of unfixed cells the coverslips were laid down on a drop of PBS and immediately examined. Confocal microscopy analysis was performed on a laser scanning LSM510 microscope (Zeiss) using dedicated image software.
Extract Preparation, Immunoprecipitation, and Western Blot Analysis-For the preparation of total cell extracts, monolayer cultures were harvested in cold PBS and resuspended in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 0.1 mM EDTA, 50 mM sodium fluoride, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, and 10 g/ml each of aprotinin, leupeptin, and pepstatin). The extracts were clarified by centrifugation at 16,000 ϫ g at 4°C, and the protein concentration was determined by the Bio-Rad assay according to the manufacturer's instructions. 10 g of extract were analyzed by SDS-polyacrylamide gel electrophoresis followed by immunoblotting as described (22). For the immunoprecipitation experiment, the cell extracts (2 mg) were incubated with the monoclonal antibody anti-GFP JL8 (CLONTECH) according to the manufacturer's instructions. The immunocomplexes were collected with 30 l/sample of protein A/G plus-agarose (Santa Cruz Biotechnology), resolved on SDS-polyacrylamide gel electrophoresis gel, and transferred to Immobilon-P membranes (Millipore).
PC12 cell fractionation was performed as previously described (36), and nuclei were purified on a 30% sucrose cushion by centrifugation at 9,000 ϫ g for 15 min at 4°C. Nuclear and cytosol-membrane extracts were prepared as described (36).

RESULTS AND DISCUSSION
Intracellular Localization of Fe65-To characterize the intracellular localization of Fe65 and to follow its possible intracellular trafficking, we generated two constructs driving the expression of full-length Fe65 fused to the GFP, one having GFP at the N terminus (GFP-Fe65) and the other one at the C terminus (Fe65-GFP) of Fe65. These constructs were transfected in COS7 cells, and the Western blot analysis of transfected cell extracts demonstrated that both fusion proteins are stable and of the expected size (Fig. 1A). The characteristic multiple band pattern of Fe65 (36) was also observed with GFP-Fe65 and Fe65-GFP; the latter migrates slightly faster than GFP-Fe65. The subcellular distribution of GFP-Fe65 fusion proteins in transfected cells showed that most of the fluorescent signal is localized in the area of the nucleus, with the exclusion of nucleoli (Fig. 1, B and C). Furthermore, Fe65bound fluorescence is also present outside the nuclei, and in many cells a clear definition of some edges can be seen. No difference was observed between the cells transfected with GFP-Fe65 or with Fe65-GFP. Fig. 1 also shows that there is no significant difference between the localization of the fluorescent signal in fixed cells (B and C) and in living unfixed cells (D and E). The confocal microscope analysis confirmed that the fluorescence bound to Fe65 is present within the nuclei (Fig. 2).
An identical expression pattern was seen when COS7 cells were transfected with the CMV-Fe65-HA vector, which drives the expression of wild type rat Fe65 protein tagged at the C terminus with the HA epitope, and stained with an anti-HA antibody (Fig. 1F). In addition, in this case the Fe65-HA signal is significantly represented both in the nucleus and in the cytosol.
To evaluate whether endogenous Fe65 is also present in the nucleus of nontransfected cells, we examined PC12 cells, in which the levels of the protein are significantly higher than those present in COS7 cells. Nuclei were purified from exponentially growing PC12 cells as previously described (36), and the protein extract from these nuclei was examined by Western blot. As shown in Fig. 1G, Fe65 is clearly present in the nuclear extract (N), whereas its amount in the cytosol-membrane (CM) fraction is severalfold higher than in the nuclear fraction.
The WW-containing Region Is Responsible for Fe65 Nuclear Translocation-The molecular weights, deduced from the DNA sequence, of the two GFP-Fe65 fusion proteins and that of Fe65-HA render their free diffusion through the nuclear pore unlikely; thus we addressed the question of what region(s) of Fe65 are responsible for its nuclear localization. To this aim, we generated a series of GFP-Fe65 vectors expressing various deletion mutants of Fe65. These vectors, summarized in Fig.  3A, were transfected in COS7 cells, and this resulted in the expression of fusion proteins of the expected sizes (Fig. 3H). The amino acid sequence of rat Fe65 contains only one region, from amino acids 684 to 705, with clustered basic residues (RRX 5 RRX 9 KXKR) that resembles a nuclear localization signal (NLS) and that is conserved in the human protein. On this basis a construct was generated driving the expression of a GFP-Fe65 fusion protein containing a deletion mutant of Fe65, lacking the C-terminal 23 residues from amino acids 666 to 711 (Fig. 3A). Fig. 3B shows that the cellular localization of this GFP-Fe65⌬666 -711 is identical to that of the GFP-Fe65 wild type protein, localized mainly in the nucleus and also in the cytoplasm. On the contrary, the truncation of the N-terminal 287 amino acids of Fe65 results in a dramatic change of the intracellular distribution of the GFP-Fe65⌬1-287 protein, which is completely excluded from the nucleus (Fig. 3C). The same exclusion from the nucleus was also observed in living unfixed cells (Fig. 3D). To further restrict the sequence involved in the Fe65 nuclear localization, we generated some more deletion mutants. As shown in Fig. 3E, one of these mutants, GFP-Fe65⌬1-253 (Fig. 3A), is still translocated to the nucleus. Therefore, the analysis of the GFP-Fe65 deletion mutants restricts the area in which the element responsible for the Fe65 nuclear localization resides from amino acid 250 to amino

Fe65 Nuclear Localization
acid 290, a region that includes the WW domain. These results are in agreement with the results of previous experiments based on cell fractionation, which demonstrated that Fe65 is present in both nuclear and cytosolic fractions, whereas deletion mutants of Fe65, lacking the N-terminal region, are only found in the cytosolic extracts (36).
No typical NLS is present in this region; therefore, we addressed the question of the possible presence in this region of an element that, besides being necessary for Fe65 nuclear localization, is also sufficient to target a protein to the nucleus. To do this we generated constructs encoding fusion proteins composed by two consecutive GFP units fused to various fragments of the region of Fe65 from amino acids 1 to 290 that we demonstrated to be necessary for its nuclear translocation (Fig.  3, A and H). As shown in Fig. 3F, a protein consisting of two consecutive GFP units is excluded from the nucleus, as expected from its molecular size, whereas the GFP-GFP-Fe65 protein carrying the region from amino acids 191 to 290 is translocated into the nucleus (Fig. 3G). There are several possibilities to be examined to explain the nuclear translocation in the absence of a consensus NLS. One of these is the direct interaction of Fe65 with the nuclear pore complex, as observed for other nuclear proteins lacking an NLS (44). Another possibility is that Fe65 is transported to the nucleus by a cargo system including a protein containing an NLS. An example of this "piggyback" mechanism is that of cyclin B1, which, lacking an NLS, is translocated into the nucleus in a complex containing cyclin F, which possesses two functional NLSs (45). We previously demonstrated that the WW domain of Fe65 binds several proteins, only one of which has been identified (as Mena (35), which is not a good candidate to be responsible for the translocation of Fe65 to the nucleus because it is mainly localized in the cytosol). The attempt to use the WW domain of Fe65 as bait to find its partners by the two-hybrid screening system failed because of the high background given by the GAL4-Fe65 fusion protein. This background is expected, considering the previous observation that the region of Fe65 containing the WW domain is per se able to strongly activate the transcription when fused to the GAL4 DNA binding domain (16). This means that the identification of the other ligands of the WW domain of Fe65, possibly responsible for its cytosol-nuclear trafficking, will require their purification and sequencing by classical biochemical procedures. Fig. 3G shows that, unlike the wild type GFP-Fe65 protein that is present both in the nucleus and in the cytosol (Fig. 1), GFP-GFP-Fe65191-290 is restricted to the nucleus. This suggests that a nuclear export signal present in Fe65 has been removed from the GFP-GFP-Fe65191-290 protein. One possible nuclear export signal (LX 3 LX 2 VXV) present in Fe65 from residues 180 to 189 and absent from the GFP-GFP-Fe65191-290 protein cannot be considered, because it is also absent from GFP-Fe65⌬1-253, which is located both in the nucleus and in the cytosol (Fig. 3E). Thus, the possible sequence affecting Fe65 export from the nucleus should be located in the C-terminal half of the molecule, or, as in the case of Fe65 nuclear import, it is translocated to the cytosol through interaction with an exported protein. CP2/LSF/LBP1 interacts with the PTB1 domain of Fe65 and thus could be involved in Fe65 nuclear export. However, the nuclear-cytosol transport of this transcription factor was never addressed, and its sequence does not contain any canonical nuclear export signal.

Fe65 Nuclear Localization
brane signaling triggered by the interaction of APP with soluble or cell-anchored signals. The demonstration that Fe65 does transit in the nucleus suggests that the Fe65-APP interaction could affect some nuclear functions. Therefore, we analyzed the possible relationship between Fe65 intracellular trafficking and APP expression. To address this point we cotransfected COS7 cells with GFP-Fe65 and APP 695 expression vectors. Transfected cells were stained with 6E10 antibody, which recognizes the extracellular/intraluminal domain of APP, or with 369 antibody, which recognizes the C-terminal cytosolic domain of APP. As shown in Fig. 4, in nonpermeabilized cells 6E10 antibody allowed us to stain the cells transfected with APP 695 expression vector (A), and some of these cells at the same time expressed GFP-Fe65. In these cotransfected cells Fe65-bound fluorescence is excluded from the nucleus, whereas in the cells expressing only GFP-Fe65 the fluorescence is localized in the nucleus (Fig. 4AЈ). The same experiment was done in permeabilized cells, by using the 369 antibody, which recognizes the cytosolic domain of APP. In these conditions, in addition to the already observed exclusion from the nucleus of GFP-Fe65 wild type in the cells coexpressing APP, a clear colocalization of GFP-Fe65 wild type and APP was seen at the plasma membrane, in the perinuclear cisternae, and in the area of the Golgi network (Fig. 4, B and BЈ). The cotransfection of the GFP-Fe65 vector with the APP 751 -expressing vector gave results identical to those observed with APP 695 (data not shown).
These experiments demonstrated that Fe65 is no longer translocated to the nucleus when the cells overexpress APP, thus suggesting that the interaction with APP is sufficient to tether Fe65 in the cytosol. To support this hypothesis, we generated an expression vector that encodes a GFP protein fused to the deletion mutant of Fe65 (GFP-Fe65⌬538 -711; see Fig. 3A) that lacks the PTB2 domain that is responsible for binding to APP. Fig. 5, A and AЈ show that the GFP-Fe65⌬538 -711 mutant is located in the nucleus and that in the cells coexpressing APP 695 the nuclear translocation of the Fe65 mutant is unaffected. Furthermore, this GFP-Fe65⌬538 -711 mutant does not accumulate in the perinuclear cisternae or in the Golgi area. An identical behavior was observed when the cells were transfected with a GFP-Fe65 construct bearing a point mutation changing the Cys-655 of Fe65 into a Phe (Fig. 5, B and BЈ). This Fe65 mutant is unable to interact with APP (20), as demonstrated by the coimmunoprecipitation experiment reported in Fig. 5C. In fact, APP was not coimmunoprecipitated with Fe65 in extracts from COS7 cells transfected with Fe65C655F and APP 695 , whereas APP coimmunoprecipitates with wild type Fe65.
The intracellular localization that we observed in COS7 cells cotransfected with Fe65 and APP is very similar to that previously reported in MDCK-695 cells stably expressing Fe65 (28). In fact, Fe65 and APP are clearly colocalized at the level of perinuclear cisternae and in the Golgi area. In these cells transfected Fe65 was not seen in the nucleus. The reason for this difference probably is, as also suggested by the authors of that study, that the amount of holo-APP in MDCK-695 cells is not limiting; therefore, the transfected Fe65 remains en-  a and b). In this experiment COS7 cells were cotransfected with expression vectors encoding APP 695 and GFP-Fe65 (lanes a-aЈ) or APP 695 and GFP-Fe65C655F (lanes b-bЈ). I.P., immunoprecipitation; w.b., Western blot.

Fe65 Nuclear Localization
trapped in the cytosol, where a large amount of APP is present. In any case, it is also possible that in MDCK-695 cells a fraction of Fe65 is translocated to the nucleus, but that the amount of this fraction is below the detection limit of the immunomicroscopy procedure used. According to this possibility, in nontransfected PC12 cells the amount of endogenous Fe65 present in the nuclear extract is much lower than that present in the cytosol-membrane fraction (Fig. 1G).
The inhibition of Fe65 nuclear import by interaction with APP could be a new example of regulated nuclear localization by cytoplasmic anchoring. The best studied example of this phenomenon is that of ␤-catenin. This protein plays an important role in the Wingless-signaling pathway (46); it functions as an adaptor protein binding to cadherins, membrane proteins involved in cell adhesion, and to the actin cytoskeleton (47). Furthermore, ␤-catenin is also localized in the nucleus and binds the transcription factor LEF1/TCF (48), and its nuclear translocation is regulated by cytoplasmic anchoring to cadherins (49). The similarity between Fe65-APP behavior and that of ␤-catenin-cadherins is further supported by the observation that ␤-catenin lacks a classical NLS and is therefore imported in the nucleus by an unusual pathway (50).
Another example of the extranuclear anchoring of transcription factors is given by Notch (51) and SREBP (52), whose precursors are membrane proteins that, following the proteolytic cleavage of their transmembrane helix, release in the cytosol a mature transcription factor that translocates to the nucleus. The case of Notch is particularly related to APP because of the involvement of presenilin dimers in the proteolytic cleavage of both proteins (51). This point deserves particular attention, and further experiments are needed to evaluate whether the APP processing by ␥-secretase could result in the targeting of Fe65 to the nucleus through the cleavage of the transmembrane domain of APP.
The understanding of the functions of the complex proteinprotein interaction network centered at the cytosolic domain of APP is an important step toward the discovery of APP functions and, possibly, of the molecular mechanisms regulating its amyloidogenic processing. The observations reported in this paper strengthen the hypothesis of a possible signaling mechanism linking APP to nuclear functions through Fe65 and its nuclear partner CP2/LSF/LBP1. This possibility is strongly supported by the recent finding that the CP2/LSF/LBP1 gene, located at chromosome 12, is a genetic determinant of Alzheimer's disease (53).