Molecular mechanisms regulating the subcellular localization of p95-APP1 between the endosomal recycling compartment and sites of actin organization at the cell surface

Rac GTPases regulate actin dynamics at the edge of migrating cells (Ridley, 1996; Hall, 1998), and there is now increasing evidence for a role played by the endocytic cycle in extending the front of eukaryotic cells (Bretscher and Aguado-Velasco, 1998b). Recycling transferrin receptors in migrating fibroblasts have been found distributed over the surface of leading lamella (Hopkins et al., 1994), and a contribution of membrane recycling from the endocytic pathway in the formation of Rac-induced ruffles has been described (Bretscher and Aguado-Velasco, 1998a). Moreover, Rho family GTPases have been directly implicated in endocytic trafficking, and in coordinating actin dynamics with trafficking at the cell periphery (Lamaze et al., 1996; Murphy et al., 1996). Progress in this direction comes from studies on ADP-ribosylation factor 6 (ARF6), a member of the ARF family of GTPases. ARF6 is specifically localized in the endosomal compartment and at the plasma membrane, and has been implicated in the regulation of membrane traffic between these compartments. The function of ARF6 on membrane recycling is supported by the effects of its overexpression on transferrin uptake and recycling to the cell surface (D’Souza-Schorey et al., 1995; Peters et al., 1995), and by the overlap of the ARF6-positive intracellular compartment with the transferrin receptor-positive recycling compartment (D’Souza-Schorey et al., 1998). A functional connection between ARF6 and Rac comes from the observation that the egress of recycled membrane to peripheral sites stimulated by ARF6 results in a localized polymerization of cortical actin, protrusive activity, and an apparent stimulation of membrane turnover by macropinocytosis (Radhakrishna et al., 1996; D’Souza-Schorey et al., 1997). This functional connection is further supported by the colocalization of Rac1 and ARF6 at the plasma membrane and on recycling endosomes, and by the block of Rac1-stimulated ruffling by the GTP binding-defective N27-ARF6 mutant (Radhakrishna et al., 1999), which has led to the suggestion that the influence of ARF6 on Rac1-mediated lamellipodia depends in part on the regulation of its trafficking to the plasma membrane.

It is still unclear how ARF6-mediated vesicle recycling is incorporated into the extension process. Recent findings on proteins that share an ARF-specific GTPase activating protein (ARF-GAP) domain suggest that they are involved in the coordination between membrane trafficking and actin reorganization during cell locomotion (de Curtis, 2001). One of these proteins, p95-APP1, has been recently identified in our laboratory as part of a multi-protein complex (p95-complex) interacting with GTP-bound Rac GTPases (Di Cesare et al., 2000). P95-APP1 is a member of a recently discovered family of multi-domain proteins including GIT1 (Premont et al., 1998), p95PKL (Turner et al., 1999), and CAT2/GIT2 (Bagrodia et al., 1999). These proteins are characterized by the presence of an amino-terminal ARF-GAP domain, and by the ability to interact directly via a Spa2 homology domain (SHD) with the Rac/Cdc42 exchanging factor PIX (Zhao et al., 2000), and with the focal adhesion protein paxillin, which binds to a carboxy-terminal paxillin binding domain (Turner et al., 1999). Our previous study on p95-APP1 has shown that both wildtype and truncated p95-APP1 induce actin-rich protrusions mediated by Rac and ARF6 (Di Cesare et al., 2000). In particular, we found that p95-C, the C-terminal portion of p95-APP1 including the paxillin binding domain, localizes at sites of actin reorganization at the plasma membrane, strongly promoting the formation of actin-rich protrusions. By contrast, the N-terminal portion of p95-APP1, including the ARF-GAP domain and the three ankyrin repeats, colocalizes with N27-ARF6 in an endosomal compartment. By further dissection of this multi-domain protein, we found that the truncated p95-C2, including both PIX- and paxillin-binding domains, accumulates via PIX around large vesicles. These vesicles are distinct from the smaller endocytic structures where the N-terminal truncated polypeptides including the ARF-GAP domain and the ankyrin repeats accumulate by a PIX-independent mechanism. Together, these observations have led us to hypothesize that the p95 complex is implicated in the regulation of membrane recycling between endosomes and the plasma membrane, and it is required to organize new integrin-mediated adhesions at sites of protrusion.

In this study, we have further analyzed the requirements for the subcellular distribution of the p95-complex by using distinct p95-APP1-derived constructs, and we have characterized the endocytic compartments where distinct truncated forms of this multi-domain protein localize. Moreover, we have identified some of the mechanisms that may be involved in the regulation of the cycling of the p95 complex between the endocytic compartment and the plasma membrane. Our results show that both the ARF-GAP domain and Rac activation regulate the subcellular distribution of the p95-complex.

The polyclonal antibodies against Rac1B (Albertinazzi et al., 1998), PIX (Daniels et al., 1999), ARF6 (Gaschet and Hsu, 1999), Rab11 (Sonnichsen et al., 2000), Early endosome antigen 1 (EEA1) (Simonsen et al., 1998), and the 6C4 mAb against lysobisphosphatidic acid (LBPA) (Kobayashi et al., 1998) have been previously described. Other antibodies included the anti-Flag mAb M5 (Kodak, New Haven, CT), a pAb anti-Flag (Santa Cruz Biotechnology Inc., Santa Cruz, CA), the anti-Myc mAb 9E10 (Sigma-Aldrich, Milan, Italy), the anti-HA-Tag mAb 12CA5, the anti-paxillin mAb (Transduction Laboratories), the mAb against LEP100 (Fambrough et al., 1988), and the anti-transferrin receptor mAb (Zymed). Secondary antibodies for immunofluorescence were from Boehringer and Jackson-ImmunoResearch Laboratories, platelet-derived growth factor (PDGF) was from Upstate Biotechnology.

The pFlag-p95, pFlag-p95-C, pFlag-p95-C2, pFlag-p95-N4, pFlag-N17-Rac1B, pFlag-V12-Rac1B and pcDNA-N27-ARF6 plasmids were described elsewhere (Di Cesare et al., 2000). The cDNA fragments corresponding to p95-C3, p95-C5 and p95-N5 were cloned into the pFlag-CMV-2 vector (Kodak), to obtain the plasmids pFlag-p95-C3, p95-C5 and pFlag-p95-N5, respectively. The pFlag-p95-K39 plasmid was obtained by site-directed mutagenesis with the QuickChange^TM site-directed mutagenesis kit (Stratagene GmbH, Heidelberg, Germany), starting from the pFlag-p95 plasmid, and using the primers 5′-AGTGCTGCAGCGTGCACAAGAGCCTGGGCCGCCACAT-3′ and 5′-ATGTGGCGGCCCAGGCTCTTGTGCACGCTGCAGCACT-3′. The pXJ40-HA-βPIX plasmid coding for the HA-tagged β-PIX polypeptide, and the pCMV6m/Pak1 plasmid coding for Myc-tagged Pak1 have been described previously (Manser et al., 1998; Bernard et al., 1999).

The Pfu DNA polymerase was from Stratagene, Klenow fragment of DNA polymerase was from Amersham Pharmacia Biotech, and restriction enzymes were from Boehringer. [α-³⁵S]dATP, ¹²⁵I-anti-mouse Ig, and ¹²⁵I-protein A were from Amersham Pharmacia Biotech. Other chemicals and FITC- and TRITC-conjugated phalloidin were from Sigma-Aldrich.

Chicken embryo fibroblasts (CEFs) obtained from embryonic day 10 chicken embryos were prepared and cultured as described (Albertinazzi et al., 1998). For immunofluorescence, CEFs grown on coverslips were transfected either with the Ca²⁺ phosphate technique or with Dosper (Boehringer) as described (Albertinazzi et al., 1998). For biochemical analysis, transient expression of proteins was achieved by transfection of CEFs by the Ca²⁺ phosphate technique. Cells were used for biochemical or morphological analysis 24 hours after transfection. For stimulation with PDGF, COS7 cells were grown in DMEM with 10% serum, transfected with FuGENE (Roche) according to manufacturer’s procedures, starved for 3-9 hours in medium without serum, and incubated with 50 ng/ml PDGF. Treated cells were immediately fixed and analyzed for immunofluorescence as described below.

Transfected and non-transfected cells were extracted with lysis buffer (0.5% Triton X-100, 150 mM NaCl, 20 mM Tris-Cl, pH 7.5, 1 mM sodium-orthovanadate, 10 mM sodium fluoride and 10 μg ml^–1 each of antipain, chymostatin, leupeptin and pepstatin). Extracts were clarified by centrifugation. 200-300 μg of protein from lysates of transfected cells were pre-cleared by incubating them for 2 hours with rotation at 4°C with 25 μl of Protein-A Sepharose beads (Amersham Pharmacia Biotech). Beads were washed four times with 1 ml of lysis buffer. Unbound material was added to Protein A-Sepharose beads with pre-adsorbed antibodies, and incubated for 2 hours at 4°C with rotation. Pellets from immunoprecipitations were washed four times with 1 ml of lysis buffer, and analysed by SDS-PAGE and immunoblotting with the indicated antibodies. Filters were then incubated with 0.2 μCi/ml of either ¹²⁵I-protein A or ¹²⁵I-anti-mouse Ig (Amersham Pharmacia Biotech), and exposed to Hyperfilm-MP (Amersham Pharmacia Biotech).

Transfected cells were fixed with 3% paraformaldehyde and processed for indirect immunofluorescence, as described (Cattelino et al., 1995). Fixed cells were incubated for 1 hour at room temperature with primary antibodies. Cells were subsequently incubated for 40 minutes with fluorescently-labelled secondary antibodies. Fluorescently-labelled phalloidin was added during the incubation with the secondary antibody. Samples were observed using a Zeiss-Axiophot or Zeiss-Axiovert microscope equipped with a 63× oil immersion objective, and a Hamamatsu C4742-98 camera (Hamamatsu Photonics K. K.). Fluorescent images were collected using the Image-Pro® Plus software package (Media Cybernetics, L. P.), and processed using Adobe PhotoShop 5.0. For the quantitation of the cytoplasmic protrusions present in transfected and non-transfected cells, only cytoplasmic protrusions longer than half the major axis of the cell body were considered. In each experiments, three sets of 100 cells were examined for each condition. Each value represents the average from counts obtained from three independent experiments. For each experimental condition, a total of about 900 cells were examined.

Several constructs derived from the multi-domain protein p95-APP1 were used in this study to analyze the function of p95-APP1 in fibroblasts (Fig. 1A). The p95-C2 polypeptide includes the SHD PIX-binding domain and the paxillin-binding domain (Fig. 1A). The specific interaction of p95-C2 with paxillin was shown here by coimmunoprecipitation from CEFs transfected with pFlag-p95-C2 (Fig. 1B, lanes 1-6). Immunoprecipitation with the anti-Flag antibody was used to show the interaction of p95-C2 with β-PIX from CEFs cotransfected with both pFlag-P95-C2 and pXJ40-HA-βPIX (Fig. 1B, lanes 7-9). These data indicate that p95-C2 can bind to both paxillin and β-PIX in these cells. As recently shown by us (Di Cesare et al., 2000), p95-C2 induced the formation of large vesicular structures in transfected CEFs, where both p95-C2 and β-PIX are localized (Fig. 2F,G). Coexpression of p95-C2 enhanced the recruitment of β-PIX around these structures, which were usually not as evident in CEFs transfected with pXJ40-HA-βPIX only. In this latter case, β-PIX had a diffuse localization, with concentration at the cell periphery (Fig. 2H), although localization of β-PIX at large vesicles was also detectable (Fig. 2N). Given the interaction of PAK1 with β-PIX, we looked at the localization of PAK1 in cells transfected with either β-PIX or p95-C2. In cells cotransfected with pFlag-P95-C2 and pCMV6m/Pak1, PAK colocalized with p95-C2 at large vesicles (Fig. 2A-D), in contrast to cells transfected with pCMV6m/Pak1 only, where PAK1 showed a diffuse staining (Fig. 2E). Moreover, in CEFs transfected with pXJ40-HA-βPIX and pCMV6m/Pak1, Pak1 and β-PIX colocalized both in ruffles at the periphery of the cells, and at large cytoplasmic vesicles (Fig. 2I,J). The large PIX-induced vesicles were distinct from the endocytic structures positive for p95-N4, the truncated polypeptide including the N-terminal ARF-GAP domain and the first ankyrin repeat of p95-APP1 (Fig. 2M-O). In cells coexpressing PAK and p95-N4 we found no evident colocalization of PAK1 with p95-N4-positive structures (Fig. 2K,L). Together these data indicate that the p95-C2/β-PIX/PAK1 complex localizes at large vesicles.

To characterize the compartment induced by the expression of p95-C2 in fibroblasts, we compared by indirect immunofluorescence the distribution of p95-C2 with that of endogenous markers for distinct endosomal compartments. These markers included the Rab5 effector EEA1, specific for early Rab5-positive endosomes (Simonsen et al., 1998), Rab 11 (Ullrich et al., 1996) and the transferrin receptor (Dunn et al., 1989; Mukherjee et al., 1997) as markers for the recycling endosomal compartment, LBPA for the late endosomes (Kobayashi et al., 1998), and LEP100 for the lysosomes (Fambrough et al., 1988). Interestingly, we found that p95-C2 colocalized specifically only with markers for the recycling compartment (Fig. 3F,G,I,J), while no detectable colocalization was observed with earlier (Fig. 3A-E) or late endocytic markers (Fig. 3L-Q). By comparison with the distribution of Rab11 (Fig. 3H) and of the transferrin receptor (Fig. 3K) in non-transfected cells, it was evident that p95-C2 expression induced an alteration of the morphology of the recycling compartment. Therefore, the large vesicles observed by p95-C2 expression correspond to morphologically altered recycling compartment.

We then compared the distribution of the markers of the endocytic pathway with the endocytic structures identified by the expression of the N-terminal fragment p95-N4 (Fig. 1A). No overlap was detected between the small scattered structures positive for p95-N4 and the early EEA1-positive compartment (Fig. 4A-C), nor with the later LBPA-positive compartment (Fig. 4L-N), or the LEP100-positive lysosomes (Fig. 4O-Q). In contrast to p95-C2, no obvious overlap was visible between p95-N4 and the endogenous transferrin receptor (Fig. 4J-K). In the case of Rab11, colocalization with p95-N4 was sporadic (Fig. 4D-I), indicating that the p95-N4-positive compartment is largely distinct from Rab11-positive recycling endosomes. Comparison with non-transfected cells showed that p95-C2 and p95-N4 did not alter the localization of EEA1, LEP100 or LBPA (not shown). Together, our observations suggest that p95-N4 and p95-C2 accumulate at distinct intracellular locations, which may correspond to different stages of the endocytic cycle.

To look at the connection between the enlarged, Rab11-positive structures and ARF6, we have compared the distribution of the N27-ARF6 mutant, known to localize into intracellular endosomal structures, with the distribution of distinct p95 mutants. We found that N27-ARF6 largely overlapped with p95-N4 (Fig. 5A,B), the N-terminal portion of p95 including the ARF-GAP domain and the first ankyrin repeat, which only sporadically colocalized with Rab11 (Fig. 4D-I). N27-ARF6 localized also at large p95-C2-positive vesicles (Fig. 5C,D), while in cells with smaller p95-C2-positive vesicles, N27-ARF6 converged into distinct elongated structures, where no detectable p95-C2 was observed (Fig. 5E-G). These results indicate that the expression of p95-C2 protein influenced both the Rab11- and ARF6-positive endocytic compartments, and led to coalescence of the two compartments once large vesicles have formed as a consequence of p95-C2 expression. Given the proposed role of both ARF6 and Rab11 into membrane recycling, these data point to a functional connection between these two types of GTP-binding proteins during vesicular transport to the cell membrane.

The finding that ablation of the N-terminal portion of p95-APP1 induced the redistribution of the resulting p95-C2 protein to large endocytic vesicles suggests that the ARF-GAP domain of p95-APP1 regulates the intracellular distribution of the protein. To test this hypothesis we looked at the effects of the expression of the p95-C3 polypeptide (Fig. 1A) in which just the ARF-GAP domain was deleted. As for p95-C2, the expression of p95-C3 induced the formation of large vacuoles in most transfected cells, which often appeared as clusters of coalescent vesicles (Fig. 6Aa). In some cases these vesicles covered a large area of the cytoplasm (Fig. 6Ac). As for p95-C2, both PAK (Fig. 6Ae-f) and β-PIX (Fig. 6Ag) colocalized around the large vesicles with p95-C3 (not shown). The characterization of the large vesicles indicated that the p95-C3-labelled structures corresponded to a transferrin receptor-positive (not shown), Rab11-positive (Fig. 6Ah-i) endosomal compartment. The accumulation of p95-C3 around this compartment was specific, since the subcellular distribution of both the early endocytic marker EEA1 (Fig. 6Aj), and of the late endosomal marker LBPA (not shown) did not show any evident overlap with the p95-C3-positive structures.

All proteins known to have ARF-GAP activity contain an arginine five residues after the fourth cysteine constituting the zinc finger which is critical for ARF-GAP activity. ASAP1, PAP, ARF-GAP1 and ACAPs containing a mutation in this residue, have little or no detectable GAP activity (Mandiyan et al., 1999; Randazzo et al., 2000; Szafer et al., 2000; Jackson et al., 2000). This arginine corresponds to Arg-39 in p95-APP1. To test whether the effects observed upon p95-C3 expression could be attributed to inactivation of the ARF-GAP activity, and not to the effects of the ablation of a portion of the protein, we expressed the epitope-tagged p95-K39 protein, in which Arg-39 was substituted by a lysine. The p95-K39 protein was still able to interact with β-PIX and paxillin, as shown by coimmunoprecipitation from transfected fibroblasts (Fig. 6B), and induced a phenotype very similar to that obtained with the p95-C3 protein in a fraction of the transfected cells (Fig. 6C). In fact, several transfected cells showed the localization of p95-K39 at large vesicles (Fig. 6Ca), which were Rab11-positive (not shown). Both PIX (Fig. 6Cc) and PAK (Fig. 6Cd) colocalized with p95-K39 around the large vesicles, indicating that the p95-K39/β-PIX/PAK complex was concentrated at this morphologically altered recycling compartment. Together, these data implicate the ARF-GAP domain of p95-APP1 in the regulation of the subcellular distribution of the p95-APP1/β-PIX/PAK complex.

We have previously reported that the expression of the p95-C polypeptide, including the C-terminal paxillin-binding domain, but not the SHD PIX-binding domain (Fig. 1A) induces protrusions in fibroblasts (Di Cesare et al., 2000). This effect is accompanied by the redistribution of paxillin away from focal adhesions, which are still detectable by using anti-integrin antibodies (Di Cesare et al., 2000). To test the requirement of paxillin-binding for enhanced protrusive activity, we have prepared the pFlag-p95-N5 plasmid coding for the p95-N5 polypeptide, in which the C-terminal paxillin-binding domain was deleted (Fig. 1A). In transfected cells, p95-N5 was often observed associated to cytoplasmic vesicles (Fig. 7Ca), not visible in cells transfected with p95-C (Fig. 7Cb). In cells transfected with p95-N5 the morphology of the Rab11 compartment was clearly affected (Fig.7Ch,i), although the colocalization of Rab11 with p95-N5-positive vesicles was only partial, and not as striking as in cells transfected with the ARF-GAP mutants. In several cases, rather than a complete colocalization of the two proteins around the vesicles (Fig. 7Cc-e), several Rab11-positive spots were observed around the large p95-N5 positive vesicles (Fig. 7Cj,k). However, staining with anti-EEA1 antibodies indicated lack of colocalization of p95-N5 with EEA1-positive vesicles (Fig. 7Cf,g). In general, cells expressing p95-N5 had less prominent protrusions compared with cells expressing p95-C (Fig. 7Ca,b, respectively). Quantitation of the percentage of cells with protrusions (Fig. 7A), and of the number of protrusions per cell (Fig. 7B) indicated a significant decrease in the protrusive activity of p95-N5-transfected cells, thus implicating the requirement of paxillin binding for the stimulation of the protrusive activity. Similar findings were observed by the expression of the p95-C5 mutant (Fig. 1A), derived from p95-C from which the paxillin-binding region had been deleted (Fig. 7Cl,m): also in this case no enhancement of protrusive activity could be observed in the transfected cells (Fig. 7A,B).

We therefore analyzed the effects of the p95-APP1-derived constructs on paxillin distribution (Fig. 8). As previously shown by us, p95-C induced the redistribution of paxillin from focal adhesion into a diffuse cytoplasmic staining with concentration at protrusions (Fig. 8A,B). By contrast, we observed no effects on paxillin distribution by p95-N5 (Fig. 8E,F). In particular, paxillin could not be detected around the p95-N5-positive vesicles observed in the transfected cells. As expected, the expression of the p95-C5 mutant did not alter the distribution of paxillin, which remained in focal adhesions (Fig. 8G,H). However, as already shown for p95-C2 (Di Cesare et al., 2000), both p95-C3 (Fig. 8C,D) and p95-K39 (Fig. 8I-K) could induce a dramatic redistribution of paxillin from focal adhesions to the large p95-positive vesicles. In these cells, paxillin was hardly detectable in focal adhesions when compared with the neighbouring non-transfected cells. In the case of β-PIX overexpression, an intermediate phenotype could be observed (Fig. 8L,M). In fact, some paxillin recruitment together with PIX could be observed in the cells showing PIX-positive vesicles, although a clear signal of paxillin in focal adhesions could still be detected.

These data show that p95-APP1 is able to relocate paxillin within the cell, and that the ability of the different p95-derived constructs to do so correlates with the presence of the C-terminal paxillin-binding site.

The p95-APP1 protein was first identified by us as an indirect interactor of activated Rac GTPases. To test the hypothesis that activation of Rac at the cell surface may regulate the distribution of p95-APP1 within the cell, we coexpressed the truncated p95-C2 protein with constitutively activated Rac. In contrast to the strong concentration of p95-C2 at large recycling vesicles observed in cells transfected with pFlag-P95-C2 only (Fig. 9A), p95-C2 was found redistributed to the cell periphery in cotransfected cells (Fig. 9B). The cotransfected cells showed the appearance of large lamellipodia (Fig. 9B,C,E) when compared with cells expressing p95-C2 alone, or with cells coexpressing p95-C2 and an inactive form of the GTPase (Fig. 9D,F). P95-C2 was often found concentrated with V12-Rac at the edge of lamellipodia (Fig. 9C,E). Interestingly, N17Rac colocalized with p95-C2 at the large intracellular vesicles (Fig. 9D,F). Similar findings were obtained by coexpressing p95-C3 and V12Rac. In contrast to the concentration of the p95-C3 protein at large intracellular vesicles (Fig. 6A), in the cotransfected cells, which had evident, large lamellipodia, the localization of p95-C3 was often diffuse, with some concentration at peripheral areas together with V12Rac (Fig. 9G,H). Fewer large vesicles were still evident only in some cotransfected cells (Fig. 9I). In these cases, colocalization at the large vesicles of p95-C3 and V12Rac was obvious. Interestingly, paxillin also co-distributed with p95-C2 and p95-C3 at the cell periphery in cells co-expressing V12Rac (Fig. 9L-O). Similarly, both PAK (Fig. 9R,S) and PIX (Fig. 9T,U) co-distributed with p95-C3 and p95-C2 at the cell edge in cells expressing V12Rac. Under these conditions, Rab11 showed a diffuse punctuate pattern (Fig. 9P,Q).

A similar colocalization with V12Rac at the edge of cotransfected cells was also observed with full length p95-APP1 and with p95-N5 (not shown). by contrast, V12Rac was unable to induce the redistribution at the cell periphery of the N-terminal p95-N4 polypeptide, which lacks the PIX-binding domain (Fig. 9J,K), thus implicating the PIX/PAK complex in the recruitment of p95-APP1 by V12Rac. Moreover, the coexpression of p95-N4 did not affect V12Rac-induced ruffling or lamellipodia formation. These data indicate that the p95-APP1/PIX/PAK complex may undergo redistribution at sites of Rac activation.

To test whether the activation of endogenous Rac is able to induce the recruitment of ARF-GAP-deficient p95 polypeptides at the cell periphery, we have looked at the distribution of p95-C2 in COS7 cells treated with PDGF. In transfected cells we found that p95-C2 localized at large vesicles (Fig. 10A). In cells treated with PDGF (Fig. 10B-E), as well as in a fraction of untreated cells, in which ruffles were still evident (not shown), p95-C2 co-localized with actin at ruffles and lamellipodia. This result indicates that p95-C2 may localize to sites of endogenous Rac activation.

We have identified the Rab11-positive recycling compartment as the endocytic compartment where p95-derived polypeptides localize upon expression in fibroblasts. Accumulation of p95-APP1 and of other components of the p95 complex in this compartment depends on the deletion or mutation of the ARF-GAP domain of p95-APP1. We have also demonstrated that Rac activation induces the localization of the p95 complex at the periphery of the cell. Together, our results suggest that p95-APP1 is implicated in the recycling of endocytosed membranes to the cell surface, at sites of Rac-mediated protrusive activity, and implicate the ARF-GAP domain of p95-APP1, paxillin, and Rac activation in the regulation of the subcellular localization of the p95-complex.

The C-terminal p95-C fragment of p95-APP1 induces the formation of protrusions, while this effect is not observed in p95-C2-expressing cells (Di Cesare et al., 2000). While p95-C is evidently localized at sites of protrusion, p95-C2 is predominantly concentrated at large intracellular vesicles. These striking differences originate from the presence (in p95-C2) or absence in (p95-C) of the SHD domain. As for GIT1 (Zhao et al., 2000), we found that the SHD domain is required for the interaction of p95-APP1 with PIX, a putative exchanging factor for Rac and Cdc42 (Oh et al., 1997; Bagrodia et al., 1998; Manser et al., 1998). Our findings support a role of PIX in the localization of the p95-complex to Rab11-positive large vesicles. Rab11 is a small GTP-binding protein required for transferrin recycling through the pericentriolar recycling endosomes (Ullrich et al., 1996; Ren et al., 1998), and represents a functional marker for this compartment. The large Rab11-positive vesicles seem to form specifically as a consequence of PIX-mediated accumulation of p95-APP1 in the endosomal compartment. The specificity is supported by the observation that only markers for the endosomal recycling compartment accumulate at these vesicles, while earlier and later endosomal compartments are not affected. The deletion (p95-C3) or mutation (p95-K39) of the ARF-GAP domain also induce accumulation to Rab11-positive structures, showing that the alteration of the ARF-GAP domain of p95-APP1 is responsible for the formation of the abnormal recycling compartment. In vitro ARF-GAP activity of the highly homologous GIT1 on ARF6 has been recently demonstrated (Vitale et al., 2000). The strong homology between avian p95-APP1 and mammalian GIT1, and the colocalization of the N-terminal portion of p95-APP1 with N27-ARF6 in the endocytic compartment suggest that this protein is a candidate ARF6 regulator in vivo. The fact that we have not been able to detect any ARF-GAP activity of p95-APP1 on ARF6 in vitro indicates that the ARF-GAP activity of p95-APP1 may be finely regulated in the cell. The conserved arginine corresponding to arginine 39 of p95-APP1 dramatically reduces the ARF-GAP activity of a number of ARF-GAPs in vitro (Mandiyan et al., 1999; Randazzo et al., 2000; Szafer et al., 2000; Jackson et al., 2000). We would like to speculate that like ARF1 in the Golgi (Roth, 1999), ARF6 regulates vesicle formation during recycling between endosomes and the plasma membrane. According to our model (Fig. 11), the effects of the ARF-GAP mutants on the Rab11-recycling compartment is due to the inability of the mutated p95-APP1 to induce GTP hydrolysis, necessary for ARF-mediated vesicle budding from recycling endosomes. The overexpressed ARF-GAP mutants, by interacting via PIX with the endosomes, would compete with the endogenous p95-APP1, and act as dominant negative forms, inhibiting p95-APP1 function in vesicle formation. The abnormal recycling compartment would be generated by the accumulation of internalized membranes, not compensated by membrane recycling in cells overexpressing the ARF-GAP mutants.

The p95-C construct missing the PIX-binding domain induces depletion of paxillin from focal adhesions, and its accumulation at the cell periphery, where it enhances protrusive activity. Polypeptides with further deletion of the paxillin binding region (p95-C5) do not influence the protrusive activity, nor paxillin distribution. By contrast, constructs including the PIX-binding region and the paxillin binding region induce a striking relocalization of paxillin, PIX and PAK1 to the Rab11-positive large vesicles. These data support a model according to which, paxillin, once removed from focal adhesions, travels with the p95-complex, via the recycling compartment, to the cell front, where it participates in the formation of new focal complexes (Fig. 11).

The absence of the PIX-binding region from the N-terminal p95-N4 construct leads to accumulation of this protein in a distinct population of endocytic vesicles. In contrast to the large Rab11-positive vesicles, the small p95-N4-positive vesicles are only weakly overlapping with the population of vesicles stained with anti-Rab11 antibodies. This localization of p95-N4 is mediated by the first ankyrin repeat (Di Cesare et al., 2000), and is independent of the SHD-mediated localization of p95 to the large vesicles. We have previously shown that p95-N4 and N27-ARF6 colocalize in these vesicles. Further work will be required to understand whether the two identified structures correspond to two altered, functionally distinct endosomal subcompartments. This hypothesis is supported by the finding that PIX and the transferrin receptor are excluded from the small p95-N4-positive vesicles, while both are present into the large Rab11-positive vesicles. The transferrin receptor cycles between the plasma membrane and the recycling endosomes, and can be induced to accumulate intracellularly by ligand internalization (Marsh et al., 1995). In non-transfected CEFs, a clear surface distribution of the transferrin receptor is observed. The concentration of the receptor in the large Rab11-positive vesicles in cells transfected with p95-C2 or p95-C3 indicates that recycling to the cell surface is altered in these cells, leading to accumulation into an abnormally enlarged recycling compartment.

The accumulation of p95-N4 in the endosomal compartment may be explained by the inability of the truncated N-terminal protein to be relocated to the plasma membrane, and implicates the C-terminal portion of p95-APP1 in targeting to the plasma membrane. One possible mechanism for targeting the p95-complex to the plasma membrane is Rac activation. In fact, overexpression of a constitutively active form of Rac induces accumulation of p95-C3 and p95-C2 to the plasma membrane together with Rac, while activated Rac does not affect the distribution of p95-N4. Therefore, PIX binding is necessary for Rac-mediated localization of the p95 complex to the cell membrane. Since PIX binds to PAK1, the localization of the p95-complex at the plasma membrane may be mediated by the binding of PAK1 to Rac (Fig. 11), although we cannot exclude the direct interaction between PIX and Rac. One could envisage that the amount of the p95-complex recruited at the plasma membrane would be dependent on the amount of active Rac at the cell surface. Under steady state conditions, the amount of endogenous GTP-Rac would not be sufficient to recruit efficiently the overexpressed p95-C2-complex to the cell surface. The p95-C2-complex would then accumulate at Rab11 endosomes via PIX. However, overexpressed V12Rac would be able to recruit a significant amount of the complex to the plasma membrane by interacting with PAK, by competing with the PIX-mediated binding of the complex to Rab11 endosomes. This idea is supported by the finding of p95-C2 at sites of endogenous Rac-mediated actin organization at the cell periphery. The results of the expression of p95-N5 and p95-C5 proteins have revealed the importance of the C-terminal paxillin binding domain on protrusive activity and paxillin relocalization, and argue in favour of a contribution of paxillin to the localization of the p95 complex at the cell periphery. In fact, the lack of the paxillin binding site leaves paxillin predominantly in focal adhesions in cells transfected with p95-N5. The p95-N5 with an intact ARF-GAP is affecting the morphology of the Rab11 compartment, although the p95-N5 protein localizes to vesicles which only partially overlap with the Rab11 compartment. Our findings imply a complex regulation of the subcellular distribution and trafficking of the p95 complex between endosomes and plasma membrane, which implicate in the process not only a functional ARF-GAP domain, but also the PIX/PAK complex and the focal adhesion protein paxillin. The lack of the paxillin binding site could affect the distribution of the protein, resulting in its partial association to a vesicular compartment that may correspond to a distinct intermediate during the recycling process. Future work on the complex network of events involved will help to further elucidate the proposed model. Interestingly, a number of ARF-GAPs including GIT2, a member of the same family of p95-APP1 (Mazaki et al., 2001), and PAPa, a member of the centaurin family (Kondo et al., 2000), have also been implicated in the regulation of paxillin distribution in the cell.

Movement of vesicles to areas of the plasma membrane involved in protrusive activity may represent a mechanism to help the forward movement of the front of migrating cells (Bretscher, 1996). The data presented in this study provide further support to a role of p95-APP1 in coordinating membrane traffic from recycling endosomes to sites of Rac-mediated actin reoganization. By using its multi-domain structure, p95-APP1 may bring together functions related to the regulation of recycling vesicular traffic, with components necessary to a productive protrusive activity driven by actin polymerization.

We are grateful to Marino Zerial for the anti-Rab11 antibody, to Harald Stenmark for the anti-EEA1 antibody, to Ed Manser and Louis Lim for the pXJ40-HA-βPIX plasmid and the anti-PIX antibody, to Victor Hsu for the anti-ARF6 antibody, and to Gary Bokoch for the pCMV6m/Pak1 plasmid and the anti-Pak1 antibody, and to Chiara Albertinazzi for helping with the transfection of COS cells. The mAb LEP100 developed by D. M. Fambrough was obtained by the Developmental Studies Hybridoma Bank, The University of Iowa, IA. Special thanks to Ruggero Pardi for critical reading of the manuscript. The financial support of Telethon-Italy (Grant n.1171 to I.d.C.) and of AIRC (Italian Association for Cancer Research) are gratefully acknowledged. C. Albertinazzi was supported by a fellowship from FIRC (Italian Federation for Cancer Research). Supported by the University Excellence Center on Physiopathology of Cell Differentiation.