βPIX controls cell motility and neurite extension by regulating the distribution of GIT1

Cell migration is driven by protrusive activity at the leading edge of the cell, where continuous remodelling of actin and adhesive sites are required. Experimental evidence indicates that membrane trafficking may contribute to the extension of the cell border required for protrusion (Hopkins et al., 1994; Bretscher and Aguado-Velasco, 1998), but the underlying mechanisms remain to be determined. GIT proteins are recently identified ADP-ribosylation factor (ARF) GTPase-activating proteins (GAPs) (Turner et al., 2001). Given the essential role of ARF GTPases in membrane trafficking (Nie et al., 2003), one hypothesis is that these proteins coordinate ARF-mediated membrane trafficking with cell adhesion and cytoskeletal organization during cell motility (de Curtis, 2001). The vertebrate GIT family includes two proteins of 95 kDa: GIT1, also called p95-APP1 or CAT1, and GIT2, also called PKL or CAT2 (Premont et al., 1998; Bagrodia et al., 1999; Turner et al., 1999; Di Cesare et al., 2000). Smaller splice variants of GIT2 have been identified (Premont et al., 2000). These ArfGAPs interact directly with a number of proteins including the PIX/Cool exchange factors for Rac1 and Cdc42 (Bagrodia et al., 1998; Manser et al., 1998), the focal adhesion protein paxillin (Turner et al., 1999), the focal adhesion kinase (FAK) (Zhao et al., 2000), and the synaptic adaptor proteins Piccolo and liprins (Kim et al., 2003; Ko et al., 2003). In vitro and in vivo studies suggest that GIT proteins specifically regulate the activity of Arf6 in cells (Albertinazzi et al., 2003; Vitale et al., 2000). Arf6 is a GTP-binding protein of the Arf family involved in membrane trafficking between the cell surface and endosomes (D'Souza-Schorey et al., 1995; Donaldson, 2003; Peters et al., 1995; Radhakrishna et al., 1996), which specifically colocalizes with GIT1/p95-APP1-derived constructs at the endocytic compartment of transfected fibroblasts (Di Cesare et al., 2000). Moreover, given the interaction of this protein with a complex including the Rac and Cdc42 exchange factor PIX and PAK, a serine threonine kinase acting downstream of Rac and Cdc42 (Bokoch, 2003), GIT1 has also been involved in the regulation of these small GTPases (Bagrodia et al., 1998; Manser et al., 1998). The role of the GIT proteins in regulating membrane trafficking during cell motility and the mechanisms regulating their subcellular localization remain to be determined. We have previously found that mutants of GIT1/p95-APP1 lacking a functional ArfGAP domain, but preserving the Spa2 homology domain (SHD) required for the interaction with PIX, accumulate at large cytoplasmic structures positive for the small GTP binding protein Rab11 (Matafora et al., 2001), a functional marker of perinuclear recycling endosomes. This accumulation correlates with the inhibition of neurite extension in cultures of primary neurons transfected with the ArfGAP mutants of GIT1/p95-APP1, thus implicating this ArfGAP in the coordination of membrane trafficking with cytoskeletal reorganization during growth cone motility (Albertinazzi et al., 2003).

In this study we show that altering the levels of βPIX affects GIT1 localization in the cell, leading to accumulation of GIT1 at transferrin (Tf)-positive endocytic structures similar to those obtained by expression of ArfGAP-deficient mutants, including the SHD PIX-binding domain. We also found that these mutants interfere with the cellular response to motogenic stimuli. Analysis of several βPIX mutants has allowed us to identify the requirements for βPIX-induced accumulation of GIT1, and the possible involvement of PAK as an intermediate required for the accumulation of the βPIX/GIT1 complex. Mutations affecting GIT1 accumulation are able to reverse the inhibition of neurite extension induced by ArfGAP mutants of GIT1. Altogether, our results show that βPIX is an important regulator of GIT1 subcellular localization, and that alteration of GIT1 localization interferes with the endocytic compartment and cell motility.

We previously showed that expression of ArfGAP-deficient mutants of GIT1/p95-APP1 induced the accumulation of these mutants at structures specifically positive for endogenous markers of the endocytic recycling compartment. By contrast, overexpressed full-length GIT1 was largely cytosolic. We also showed that the GIT1 mutants colocalize with other components of the GIT1 complex at these structures, as shown by colocalization of βPIX, PAK and paxillin (Matafora et al., 2001). Since these results have shown that the PIX-binding SHD domain of GIT1 is required for the recruitment of the complex at the endocytic structures, we explored the role of βPIX in the intracellular distribution of full length GIT1. We found a strong effect of the co-expression of βPIX on the localization of GIT1 both in chicken embryo fibroblasts (CEFs) and COS7 cells (Fig. 1). In fact, whereas overexpression of either protein resulted in largely diffuse distribution of these proteins, co-expression of GIT1 and βPIX induced accumulation of both proteins at large intracellular structures. These structures were very similar to those described when mutants of GIT1 with an affected (p95-K39) or deleted (p95-C2) ArfGAP domain are expressed in either fibroblasts or primary neurons (Matafora et al., 2001; Albertinazzi et al., 2003). Overexpression of proteins can induce the formation of proteinaceous structures called aggresomes that share some common features (Garcia-Mata et al., 2002). To test whether the structures induced by co-overexpression of GIT1 and βPIX were classical aggresomes, we performed staining with anti-vimentin antibodies, to look at the organization of intermediate filaments in the transfected cells, which are usually heavily affected by aggresome formation (Garcia-Mata et al., 2002). Intermediate filaments appeared perfectly conserved in transfected cells when compared to non-transfected cells (Fig. 2A), indicating that the structures induced by co-expression of GIT1 and βPIX are not aggresomes. Analysis of the distribution of endogenous TfR showed that this transmembrane endocytic protein was partially retained in the βPIX/GIT1-positive structures (not shown), confirming previous findings in cells expressing ArfGAP mutants of GIT1 and showing alterations in the endocytic compartment (Matafora et al., 2001; Albertinazzi et al., 2003). Partial accumulation of endocytosed Tf was observed in the βPIX/GIT1-positive structures after 1 hour of Tf uptake (Fig. 2B). Some labelled Tf remained trapped in these structures after 2 hours of chasing, whereas the rest of the cytoplasm was cleared of the internalized Tf that had undergone recycling (Fig. 2C). Intriguingly, overexpression of βPIX alone inhibited Tf uptake, with no evident accumulation of Tf in the large cytoplasmic structures (Fig. 2D).

To test for the localization of the endogenous TfR, 200 nm ultrathin cryosections were prepared and processed for immunofluorescence. Accumulation of endogenous TfR and overexpressed GIT1 was observed confined within the same large structures in cells cotransfected for GIT1 and βPIX (Fig. 3A-C). We then performed immunogold labelling on 60 nm ultrathin cryosections to further characterize these large structures. Using an anti-FLAG antibody to identify the transfected GIT1, we observed gold labelling at tubulovesicular endosomes (Fig. 3D), and electron dense structures (Fig. 3E-G). Double labelling of the sections with anti-FLAG antibodies (15 nm gold), in combination with anti-TfR (10 nm gold) showed an extensive colocalization of the two antigens within the electron dense structures (Fig. 3E-G). Observation at higher magnification revealed that the colocalization frequently occurred on membranes included within the electron dense structures (Fig. 3G). These data show that co-overexpression of GIT1 and βPIX induced the specific aggregation of proteins with membranes from the TfR-positive endosomal compartment. The analysis, by immunoelectron microscopy, of cells transfected with the ArfGAP-depleted protein p95-C2 showed that the same type of structures were induced (data not shown). These data suggest that perturbation of the endogenous GIT1-βPIX complexes can cause the alteration of the recycling compartment. This alteration is obtained either by increasing the levels of the complex by overexpressing both proteins, or by expressing an ArfGAP mutant able to interact with βPIX. In fact, previous studies have shown that a point mutation inactivating the ArfGAP activity of GIT1 is also able to induce aggregates of the endocytic recycling compartment, whereas mutants lacking the SHD PIX-binding domain fail to cause membrane clustering (Matafora et al., 2001).

Overexpression of either GIT1 or βPIX results in the formation of structures similar to those obtained by co-expression of the two proteins in a smaller number of cells. The analysis of these cells by immunogold labelling has shown that either protein induced cytoplasmic structures similar to those observed in cotransfected cells (Fig. 4). In cells transfected with βPIX, the electron-dense cytoplasmic structures showed intense labelling for βPIX, whereas labelling of the same structures for the endogenous TfR was rare (Fig. 4A,B). By contrast, the electron-dense structures in GIT1-transfected cells were strongly labelled for both overexpressed GIT1 and endogenous TfR (Fig. 4C,D).

To assess possible effects of the perturbation of the intracellular localization of GIT1 with cell motility, we analyzed the effects of expression of ArfGAP-deleted p95-C2 on EGF-induced cell motility in A431 cells expressing high levels of EGF receptor. A431 cells were transfected to express each of different GFP-tagged constructs of GIT1 (Fig. 5A), and followed by time-lapse videomicroscopy before and during stimulation with EGF. A431 cells respond to EGF by extending membrane ruffles and lamellipodia, and the process is mediated by the activation of the GTPases Rac1 and Cdc42 (Kurokawa et al., 2004). In A431-transfected cells, EGF induced a strong recruitment of ArfGAP-defective p95-C2-GFP at cytoplasmic structures that appeared to form at the cell periphery (Fig. 5B, arrowheads) and concentrated with time in the perinuclear region of the cells (Fig. 5B). As a control, we used the p95-C-GFP mutant that differs from p95-C2-GFP in the lack of the SHD domain (Fig. 5A). By contrast to cells expressing p95-C2-GFP, EGF treatment did not induce formation of p95-C-GFP-positive structures in A431 (Fig. 5C,E), whereas EGF stimulation in cells expressing the full length p95-GFP induced only a weak accumulation of GFP-positive structures in a smaller fraction of transfected cells (Fig. 5E). These results indicate that the clustering of endocytic membranes was strongly potentiated by the lack of the ArfGAP domain, and required the presence of the SHD βPIX-binding domain. In cells transfected with the ArfGAP-deleted p95-C2 mutant, these structures were formed by the fusion of smaller cytoplasmic structures appearing soon after stimulation with EGF (Fig. 5D). Similar results were observed in normal fibroblasts stimulated with platelet-derived growth factor (data not shown).

Accumulation of endocytic structures in p95-C2-GFP-transfected cells correlated with the inhibition of EGF-induced extension of lamellipodia and ruffles (Fig. 6A), and with strong cell retraction, that became evident 10-20 minutes after EGF addition (Fig. 6C). Inhibition of EGF-induced membrane protrusion and retraction (Fig. 6B,D, respectively) were not obvious in cells overexpressing either the full length GIT1 protein, or the p95-C mutant lacking the PIX-binding domain. p95-C-GFP-expressing cells showed lamellipodia formation following EGF stimulation (Fig. 5C). These results indicate that the ArfGAP mutant p95-C2 maintained the ability to accumulate at endocytic structures via the PIX binding domain. The observed effects on ruffling and cell retraction may be explained by the fact that the formation of these structures sequesters endocytic membranes (Fig. 2B,C and Fig. 3), and may therefore interfere with membrane recycling under conditions of acute stimulation of membrane internalization by EGF (Haigler et al., 1979). Stimulation of p95-C2 expressing cells with EGF could induce retraction also in the presence of Y-27632, an inhibitor of the kinase ROCK that mediates RhoA-induced contractility (Fig. 6E,F). Therefore, the results indicate that Rho-mediated contractility is not involved in the EGF-induced retraction of p95-C2 expressing cells.

These data indicate that binding to βPIX was required for the recruitment of GIT1 at vesicles, since the lack of the SHD domain from p95-C-GFP resulted in a diffuse cytoplasmic distribution of the protein (Fig. 5C).

One intriguing result from the previous set of experiments is that EGF stimulation resulted in weaker accumulation of full length p95-GFP-positive structures when compared to the truncated p95-C2 construct (Fig. 5E). The difference is even more striking when considering that in the p95-GFP-transfected cells showing formation of cytoplasmic structures, these were generally much less evident than in p95-C2-GFP transfected cells. This finding was somehow surprising, since both proteins contain the SHD PIX-binding domain. One possible explanation is that the full length GIT1 protein is in a conformational state that can be modified either by PIX binding, or by truncation of the protein amino-terminal region. This hypothesis is supported by the data showing that overexpression of GIT1 alone in CEFs results in the moderate formation of GIT1-positive structures in a minor fraction of cells compared to cells cotransfected with GIT1 and βPIX (Fig. 7E). The hypothesis of a conformational switch between an inactive and active form of GIT1 is supported also by previous data showing PIX-dependent stimulation of paxillin binding to GIT1 (Zhao et al., 2000).

Based on these observations, we further explored the mechanisms of PIX-mediated recruitment of GIT1 at endocytic structures. As mentioned earlier, co-expression of βPIX with full length GIT1 markedly enhanced the recruitment of both proteins at endocytic structures compared to cells overexpressing either protein (Fig. 7B,C). Several βPIX mutants were tested to look for the region(s) of βPIX required to recruit the βPIX/GIT1 complex to the cytoplasmic structures (Fig. 7A). All the PIX mutants used for this analysis preserved the ability to interact with GIT1 (Fig. 7G). In the βPIX-PG-ΔLZ mutant, the combined mutation of the SH3 domain and the deletion of the carboxy-terminal leucine zipper required for βPIX dimerization strongly prevented the accumulation of GIT1 (Fig. 7D). Mutation of the SH3 domain alone was not sufficient to prevent the localization of the complex at the cytoplasmic structures, whereas monomeric βPIX with a normal SH3 domain only partially affected the formation of these structures (Fig. 7E). On top of their ability to interact with each other, PIX and GIT proteins can form homodimers (Kim et al., 2001; Kim et al., 2003; Paris et al., 2003; Feng et al., 2004). Therefore, larger complexes between the two proteins can be formed. Large endogenous complexes containing both GIT1 and PIX have been identified (Paris et al., 2003). Moreover, it has been shown that large complexes between GIT1 and βPIX can be induced by overexpression of the two full length proteins, whereas large complexes are prevented by co-expression of GIT1 with the βPIX-PG-ΔLZ monomeric mutant, although this mutant can be co-immunoprecipitated with GIT1 (de Curtis and Paris, 2005). Based on these results, we propose that βPIX regulates the intracellular localization of GIT at intracellular membranes by the assembly of large oligomeric complexes.

PAK kinases act as downstream effectors for Rac and Cdc42 and are implicated in actin reorganization. The amino-terminal portion of PAK1 contains a proline-rich region (amino acid 184-204, the PIX binding domain; Pbd) that binds the SH3 domain of PIX (Manser et al., 1998). We prepared a PAK-Pbd construct (amino acid 150-250 of PAK1; Fig. 8A) to test whether PAK binding to PIX is involved in the formation of βPIX and GIT1-containing endocytic structures. PAK-Pbd competed for the binding of endogenous PAK to βPIX (Fig. 8B). PAK-Pbd appeared to have only a minor effect on the interaction between βPIX with GIT1, as determined by quantification of the bands of GIT and PIX coprecipitating with either PAK or PAK-Pbd: there was a less than 20% decrease in the ratio between coprecipitating GIT/PIX in co-IPs from PAK-Pbd-expressing cells (Fig. 8C). In triple-transfected CEFs, whereas wild-type PAK colocalized with the endocytic structures induced by βPIX and GIT1 (Fig. 8D,E,H), co-expression of PAK-Pbd with GIT1 and βPIX prevented the formation of these structures (Fig. 8F-H). These results indicate that PAK-Pbd prevents clustering of endocytic membranes by preventing binding of endogenous PAK to βPIX, and suggest that PAK is involved in the recruitment of the GIT complex at endocytic vesicles. According to this hypothesis, overexpression of PAK1 and βPIX induced the formation of the endocytic structures, where endogenous GIT1 was recruited (Fig. 8I).

Our results indicate that binding of βPIX to GIT1 is required for the recruitment of the complex to TfR-positive structures (Fig. 3). A role of the GIT1 complex in growth cone motility has recently been demonstrated in primary retinal neurons. Expression of p95-C2 induces inhibition of neurite extension and accumulation of the GIT1 complex at Rab11-positive structures (Albertinazzi et al., 2003). Similar results were obtained by expressing the monomeric p95-C2-LZ mutant (Fig. 10J). Here, we have used retinal neurons to further explore the role of βPIX and GIT1 in neurite extension. Expression of dimeric or monomeric βPIX mutants per se did not affect the formation of long neurites (Fig. 9). We then analyzed the effects of the co-expression of the βPIX constructs with either dimeric or monomeric p95-C2 on neurites and on the subcellular localization of the βPIX/GIT1 complexes (Fig. 10). When co-expressed with dimeric p95-C2, all βPIX mutants colocalized with p95-C2 at the endocytic structures (Fig. 10A-C,F,H), and resulted in inhibition of neurite extension (Fig. 10J). We then tested the co-expression of βPIX mutants with monomeric p95-C2-LZ. Interestingly, monomeric PIX-PG-ΔLZ (with mutated SH3/PAK-binding site) and p95-C2-LZ showed a diffuse distribution, and resulted in neurons with normal neurites (Fig. 10D,J). The ability of monomeric βPIX-PG-ΔLZ to prevent p95-C2-LZ-induced neurite inhibition and the formation of endocytic structures was specific, since co-expression of other monomeric βPIX mutants such as βPIX-ΔPH-ΔLZ (Fig. 10I,J) and βPIX-ΔLZ (not shown) with monomeric p95-C2-LZ resulted in the formation of endocytic structures, and strong neurite inhibition. These findings suggest that heterodimers formed by monomeric βPIX and p95-C2 partners are still able to induce the large structures when an intact SH3 domain is present in βPIX. Therefore, the SH3 domain of βPIX and βPIX-mediated hetero-oligomerization may be implicated in the regulation of GIT1 localization and function during neuritogenesis. βPIX-PG-ΔLZ, unable to bind PAK, binds monomeric p95-C2-LZ and would act as a dominant negative mutant on the formation of the endocytic structures. Accordingly, also monomeric βPIX-C-ΔLZ prevented p95-C2-LZ-induced formation of cytoplasmic structures and neurite inhibition (Fig. 10G,J). This construct included just the region of βPIX required for GIT1 binding (Fig. 7A), and could act as a dominant negative, similar to βPIX-PG-ΔLZ.

The correlation between the formation of endocytic structures and neurite inhibition (Fig. 10K) supports the hypothesis that blocking membrane recycling by mutant βPIX/GIT1 complexes directly affects neurite extension.

In this study we have shown the inhibitory effects of an ArfGAP-deficient mutant on growth factor-induced motogenic responses. Growth factors such as EGF and PDGF provide motogenic stimuli, by enhancing the ruffling activity and lamellipodia formation at the periphery of the cell (Ullrich and Schlessinger, 1990). In A431 cells stimulated by EGF, actin dynamics at the cell periphery is mediated by the activation of Rho family GTPases. Rac and Cdc42 (Kurokawa et al., 2004), and is accompanied by massive membrane internalization (Haigler et al., 1979). Membrane recycling back to the cell surface must accompany EGF-induced membrane internalization to avoid rounding up of the cell. The observed accumulation upon EGF stimulation of endocytic structures in the perinuclear region of cells expressing an ArfGAP-deficient GIT1 is consistent with the hypothesis that this mutant interferes with the recycling of membranes back to the cell surface. According to this hypothesis, we observed inhibition of lamellipodia formation and cell retraction in transfected cells accumulating perinuclear vesicles in response to EGF. Given the proposed role of GIT1 as an Arf6 GAP in vivo (Albertinazzi et al., 2003; Vitale et al., 2000), we would like to speculate that the observed accumulation of internalized membranes upon EGF stimulation in cells expressing the ArfGAP-deficient GIT1 mutant could be a consequence of hindering the Arf6-mediated trafficking between the endosomal compartment and the plasma membrane (Peters et al., 1995; Radhakrishna et al., 1996).

The results indicate that a protein binding to the SHD domain of GIT1 is required for the formation of the endocytic structures upon EGF stimulation. Three proteins have so far been identified as binding partners of the SHD domain of GIT proteins: the exchange factor PIX (Bagrodia et al., 1999; Turner et al., 1999), the tyrosine kinase FAK (Zhao et al., 2000), and the presynaptic cytomatrix protein Piccolo (Kim et al., 2003). Piccolo is unlikely to be involved in the observed effects, since it is expressed in the nervous system, and is involved in the organization of synaptic sites (Cases-Langhoff et al., 1996; Fenster et al., 2000). Moreover, we have not been able to reproduce the interaction between FAK and GIT1 in our system (data not shown). Therefore, PIX remains the most likely known candidate for the regulation of GIT1 recruitment at membranes in our system. This hypothesis is supported by the finding that most of the endogenous PIX is found in complex with the endogenous GIT in COS7 cells (Botrugno et al., 2006).

The role of βPIX in the recruitment of GIT1 at membranes is indicated by the induction of association of GIT1 at endocytic structures induced by overexpression of βPIX. Both the SH3 domain and dimerization of βPIX are involved in the formation of the large cytoplasmic structures. Inhibition of the formation of these structures by the PAK-Pbd polypeptide interfering with binding of endogenous proteins to the βPIX SH3 domain suggests that PAK is required in the process, although the involvement of other PIX-SH3 binding partners can not be ruled out at this point (Feng et al., 2004).

Multimeric complexes, including dimeric GIT1 and βPIX, have been previously reported (Kim et al., 2003; Paris et al., 2003; Premont et al., 2004). The role of these complexes is not clear. Since GIT1 is a regulator of Arf6 and has been found to specifically localize at endocytic structures derived from the recycling compartment, one hypothesis is that it is part of a protein coat assembled at endocytic membranes to regulate Arf6-mediated membrane recycling. In this respect, the large cytoplasmic structures induced by overexpression of GIT1 and βPIX are probably caused by the dysregulation of the cellular levels of the two proteins, leading to overproduction of multimeric components that would artifactually, but specifically, sequester membranes with endogenous transferrin receptors.

Interestingly, immunoelectron microscopy analysis has shown here that the alteration of the levels of endogenous βPIX/GIT1 complexes leads to the formation of electron-dense aggregates including both GIT1 and the transmembrane endogenous TfR. This finding, together with our recent finding that most endogenous βPIX is found stably associated with membranes together with endogenous GIT1 in cells (Botrugno et al., 2006) suggests that levels of the βPIX/GIT1 complexes need to be finely regulated in the cell, and that alterations in this direction may lead to clustering that may compromise cell function. Accordingly, it has recently been shown that GIT1 interacts with huntingtin and enhances huntingtin aggregation by recruitment of the protein into membranous structures (Goehler et al., 2004). Interestingly, GIT1 was localized to neuronal inclusions in Huntington disease patients. Moreover, biochemical analysis has shown that a carboxy-terminal fragment of GIT1 is selectively accumulated in brains of patients with Huntington disease, and this may be a significant factor in the pathogenesis of the disease.

Plasmids pFLAG–p95 (full length avian GIT1), pFLAG–p95-N, pFLAG–p95-C, pFLAG–p95-C2 and pFLAG-LacZ (Di Cesare et al., 2000), pFLAG-p95-C2-LZ, pXJ40-HA-βPIX-C and PXJ40-HA-βPIX-C-LZ (Paris et al., 2003), pXJ40-HA-βPIX-PG, pXJ40-HA-βPIX-PGΔLZ (de Curtis and Paris, 2005), pCMV6m–Pak1 encoding Myc-tagged Pak1 (Bokoch et al., 1998), and pXJ40–HA–βPIX encoding haemagglutinin (HA)-tagged βPIX polypeptide (Manser et al., 1998) were obtained as previously described. βPIX mutants were derived from pXJ40-HA-βPIX. The pXJ40-HA-βPIX-ΔLZ plasmid was prepared by digesting pXJ40-HA-βPIX with HindIII and BglII, and by ligation of the paired oligonucleotides deltaLZ1 (AGCTTACTGCACAAGTGCAAAGACGAGGCAGACCCTGAACTCAAGTTCACGCAAAGAGTCTGCTCCACAAGTGCCCGGGTAGA) and deltaLZ2 (GATCTCTACCCGGGCACTTGTGGAGCAGACTCTTTGCGTGAACTTGAGTTCAGGGTCTGCCTCGTCTTTGCACTTGTGCAGTA) to introduce a stop codon. The procedure to obtain plasmid pXJ40-HA-βPIX-ΔPH-ΔLZ, was the same as the one described to obtain pXJ40-HA-βPIX-ΔLZ, but starting from pXJ40-HA-βPIX-ΔPH. The pXJ40-HA-βPIX-ΔPH plasmid was obtained by PCR on pXJ40-HA-βPIX with the oligonucleotides PIXΔPH5 (GGGGTACCTCTGTGAGCAACCCCACC) and PIXΔPH3 (GGGGTACCACTGCCCAACGTCTTTATG). The pCMV6M-MYC-PAK-Pbd plasmid coding for amino acid 150-250 of PAK1 was obtained by PCR with primers PIXbd5 (CGGGATCCGCTGAGGATTACAATTCTTCTAATG) and PIXbd3 (CGGAATTCCTAAGACATTTTAGGCTTCTTCTTCTGC), and insertion of the PCR fragment into the pCMV6M-MYC vector digested with BamHI and EcoRI. GFP (green fluorescent protein)-labelled constructs p95-GFP, p95-C2-GFP and p95-C-GFP were obtained by cloning the full length avian GIT1, the carboxy-terminal fragments p95-C2 (amino acid 229-740) and p95-C (amino acids 347-740) into the pEGFP-N1 vector (BD-Biosciences-Clontech, Rookville, MD).

A431 and COS7 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% foetal bovine serum (Hyclone, Logan, UT). Chicken embryo fibroblasts (CEFs) from E10 chick embryos were cultured in DMEM with 5% foetal bovine serum, 1% chicken serum. CEFs and COS7 cells were transfected with Dosper (Roche, Mannheim, Germany) and lipofectamine (Invitrogen AG, Basel, Switzerland), respectively; A431 cells were transfected with FuGENE 6 (Roche, Mannheim, Germany).

Neural retinal cells were prepared from E6 chick neural retinas and cultured on 0.2 mg/ml poly-D-lysine and 40 μg/ml laminin-1 (Albertinazzi et al., 1998). For transfection by electroporation, retinal cells were resuspended in cytomix pH 7.6 (120 mM KCl, 0.15 mM CaCl₂, 10 mM K₂HPO₄/KH₂PO₄, 25 mM HEPES, 2 mM EGTA, 5 mM MgCl₂) to a final concentration of 65×10⁶ cells per ml. 200 μl cell suspension were placed in a Gene Pulser Cuvette (Bio-Rad) and 50 μg of plasmid were added (40 μg of each plasmid in cotransfection experiments). After an incubation on ice for 5 minutes, cuvettes were subjected to two sequential pulses at 0.4 KV and 125 μF, resuspended in 10 ml serum-free retinal growth medium, and plated on poly-L-lysine- and laminin-coated coverslips. Cells were cultured for 20-24 hours at 37°C, 5% CO₂, and fixed for immunofluorescence staining. Quantification of the effects of the overexpression of the constructs in retinal neurons was made by examining transfected, neurofilament-positive neurons, or cotransfected neurons. For each type of transfection, at least 50 neurons were examined morphologically from at least two distinct experiments (total of 100 neurons/experimental condition). Long neurites were equal to or longer than three cell body diameters, short neurites were shorter than three cell body diameters.

Antibodies used in this study included: polyclonal antibodies anti-FLAG (Sigma Aldrich, St Louis, MO), anti-HA-11 (BabCO, Richmond, CA); monoclonal antibodies anti-FLAG M5 (Sigma Aldrich), anti-HA 12CA5, anti-Myc 9E10 (Primm, Milano, Italy), anti-PKL/GIT1 (BD Transduction Laboratories), anti-transferrin receptor (TfR) (Zymed), anti-vimentin (clone V9, Sigma Aldrich). The polyclonal anti-PIX antibody was raised against the amino-terminal portion of βPIX. A cDNA fragment corresponding to amino acid residues 1-391 of rat βPIX was obtained by PCR, and cloned into the pGEX4T3 vector (Amersham Biosciences AB, Uppsala, Sweden). The fusion protein between glutathione S-transferase (GST) and residues 1-391 of rat βPIX was expressed in E. coli and purified for injection into rabbits. The antibody specifically recognized endogenous PIX and overexpressed βPIX, as detected by western blotting and immunoprecipitation (data not shown). Holo-Tf and Y-27632 were from Sigma-Aldrich. EGF was from Upstate Biotechnology, NY.

Cells were lysed in lysis buffer (20 mM Tris-HCl at pH 7.5, 150 mM NaCl, 0.5% Triton X-100, 0.5 mM PMSF, 1 mM sodium orthovanadate, 10 mM sodium fluoride) for 15 minutes on ice. For immunoprecipitation, equal amounts of protein were incubated for 2-3 hours at 4°C with the indicated antibodies coupled to protein A-Sepharose (Amersham Biosciences, Piscataway, NJ). After washing with lysis buffer with 0.1% Triton X-100, samples were boiled in sample buffer, blotted onto nitrocellulose membranes (Schleicher & Schuell BioScience, Dassel, Germany) and probed with the indicated antibodies. Proteins were visualized with ¹²⁵I-coupled secondary antibodies or protein A, and exposed to Amersham Hyperfilm-MP (Amersham). Protein determination was by Bio-Rad Protein Assay (Bio-Rad, Munich, Germany).

Transfected cells cultured on glass coverslips (BDH) were fixed with 3% paraformaldehyde, permeabilized with 0.1% Triton X-100, and processed for immunofluorescence. For vimentin staining, cells were fixed with methanol at –20°C. Fluorescent images were collected using the Image-Pro^® Plus software package (Media Cybernetics, L.P., Silver Spring, MD). Confocal microscopy was performed on the Leica TCS SP2. For multichannel imaging, fluorescent dyes were imaged sequentially in frame-interlace mode to eliminate cross talk between the channels. FITC and Alexa Fluor 488 were excited with a 488-nm ArKr laser line, TRITC, Alexa Fluor 568 and Alexa Fluor 546 were excited with a 543-nm HeNe laser line. Secondary antibodies for immunofluorescence were from Molecular Probes (Eugene, OR), and Jackson Immunoresearch Laboratories (West Grove, PA). Images were processed using AdobePhotoshop^® 6 (Adobe Systems Incorporated, Seattle, WA). For quantification of the GIT1-positive structures, at least 200 cells were examined for each experimental condition. Values given are the means from two to four different experiments.

For immunogold labelling, COS7 cells transfected with pFLAG–p95 and/or pXJ40-HA-βPIX were fixed 48-72 hours after transfection with 2% paraformaldehyde/1% acroleine in PBS, for 2 hours at room temperature, and processed for ultrathin cryosectioning as previously described (Confalonieri et al., 2000). Single and double immunogold labelling was performed on 60 nm thick ultrathin cryosections as described previously (Slot et al., 1991), using polyclonal antibodies to FLAG (Sigma Aldrich, St Louis, MO) either alone, or in combination with mAbs to TfR (Zymed Laboratories, San Francisco, CA). Alternatively, 200 nm thick cryosections were placed on glass slides and double labelled for FLAG and TfR by immunofluorescence, and counterstained with DAPI to identify nuclei.

A431 cells grown on glass coverslips were serum starved for 2-3 hours, briefly washed with buffer A (137 mM NaCl, 3 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 20 mM Hepes-NaOH, pH 7.4, freshly added 20 mM glucose and 2 mg/ml bovine serum albumin), and incubated for 1 hour with 60 μg/ml of Alexa Fluor 488-labelled Tf (Molecular Probes) in buffer A. Cells were cooled on ice, and excess Tf was removed by washing with ice-cold PBS and low-pH buffer (150 mM sodium chloride, 10 mM acetic acid, pH 3.5). The two washes were repeated twice. The cells were then either fixed or chased at 37°C for 2 hours in buffer A with 6 mg/ml holo-Tf. After the chase, cells were washed twice with ice-cold PBS and fixed. After fixation with 3% paraformaldehyde, cells were processed for indirect immunofluorescence, and permeabilized with 0.05% saponin during the incubation with antibodies.

To analyse the behaviour of transfected cells upon stimulation with epithelial growth factor (EGF), A431 cells were cultured on glass coverlips for 24 hours and transfected for 18 hours with pEGFP-GIT1, pEGFP-p95-C2, pEGFP-p95-C, or pEGFP plasmids. Cells were serum starved for 4-8 hours before Y-27632 and/or EGF stimulation. For time-lapse videomicroscopy, each coverslip was observed with a Zeiss Axiovert 135 TV microscope equipped with an Orca II CCD digital camera (Hamamatsu, Hamamatsu City, Japan), or with a Olympus IX 70 Deltavision equipped with a Cool SNAP HQ digital camera (Photometrix, Kew, Australia). Photographs were taken at 10- to 120-second intervals, both before and after the addition of 100-200 ng/ml of EGF and/or 10 μM Y-27632 in buffer A. Images were analysed by the Image Pro Plus software (Media Cybernetics, Silver Spring, MD).

The financial support of Telethon-Italy (grant no. GGP05051), FIRB (grant no. RBNE01WY7P), and of the European Union (MAIN Network of Excellence, FP6-502935) to I.dC., and of MIUR (Minister for University and Research) to C.T. is gratefully acknowledged. We thank Barbara Sporchia for helping in the preparation of the plasmids for the PIX mutants, and Cesare Covino and Mariacarla Panzeri (ALEMBIC, Universita' Vita e Salute San Raffaele, Milano) for help with time-lapse experiments and electron microscopy, respectively. Electron microscopy studies were also performed at the Telethon Facility for EM, Genova, Italy (Grant GTF03001 to C.T.).