Abstract

Aluminum fluoride (AlFx) is known to activate directly the α subunit of G-proteins but not the homologous small GTP-binding proteins. However, AlFx can stabilize complexes formed between Ras, RhoA or Cdc42 and their corresponding GTPase-activating proteins (GAPs). Here, we demonstrate that Arf1GDP can be converted into an active conformation by AlFx to form a complex with the Arf-GAP ASAP1 in vitro and in vivo. Within this complex ASAP1, which GAP activity is inoperative, can still alter the recruitment of paxillin to the focal complexes, thus indicating that ASAP1 interferes with focal complexes independently of its GAP activity.

Keywords: ADP-ribosylation factor; GTPase-activating protein; ASAP1; Aluminum fluoride; Focal complex

Abbreviations: AlFx, aluminum fluoride; Arf, ADP-ribosylation factor; FC, focal complex; GAP, GTPase-activating protein

Article Outline

1. Introduction

2. Materials and methods

2.1. Production and purification of recombinant Arf proteins

2.2. Preparation of phospholipid vesicles

2.3. Sedimentation assay

2.4. Transfection, immunoprecipitation and immunoblot

2.5. Cell-substratum assay and immunofluorescence

3. Results

4. Discussion

Acknowledgements

References

1. Introduction

The ADP-ribosylation factor (Arf) family of small GTP-binding proteins comprise six members among which Arf1 and Arf6 are the most extensively studied. Arf1 functions in ER-to-Golgi and intra-Golgi vesicular transport by recruiting the COPI coat onto Golgi membranes, while Arf6 is involved in peripheral membrane trafficking and actin cytoskeleton organization [1]. Because of their very low intrinsic nucleotide exchange and GTPase activities, Arf proteins are regulated by guanine exchange factors (GEFs) and GTPase-activating proteins (GAPs), respectively.

The Arf-GAP family includes at least 15 members characterized by a conserved zinc finger motif within the catalytic domain. They are multidomain proteins that can be categorized based on their structural domain organization [2]. ASAP1 is a PIP2-dependent Arf-GAP that has been proposed to coordinate membrane trafficking with actin remodeling. It has been found to bind to the non-receptor tyrosine kinases p60src and focal adhesion kinase (FAK) and, more recently, to the adaptor protein CD2-associated protein (CD2AP) [3], [4] and [5]. In vitro assays indicate that ASAP1 is more active on Arf1 and Arf5 than on Arf6 [3]. Further, a cell-based GAP assay suggested that ASAP1 would only be active on Arf1 but not Arf6 [6]. In various fibroblastic cell lines ASAP1 is a cytosolic protein with a small fraction detected at the membrane periphery where it colocalizes with focal adhesions (located underneath the cell body) and focal complexes (FCs, located at the cell periphery) markers such as paxillin, β1-integrin and FAK. At the immunofluorescence microscopic level, ASAP1 seems to be absent from and not to disrupt the Golgi apparatus and the endocytic compartments [7]. Noteworthy, exogenous expression of ASAP1 was shown to interfere with FCs formation, cell spreading, PDGF-induced ruffling and cell migration. The use of a catalytic point mutant (R497K) suggested that the observed effects were dependent upon the Arf-GAP activity [4], [5] and [7]. In contrast, expression of the protein ASAP1 deleted of the GAP domain suggested that the GAP activity was required to interfere with cell migration but not with cell spreading nor the formation of the peripheral FCs [6].

The molecular mechanism by which GAPs stimulate GTP hydrolysis has been modeled with the large heterotrimeric G-proteins by using aluminum fluoride (AlFx). The exact nature of the active species (AlF₃ or AlF₄) of aluminofluoride complexes is still being investigated [8], therefore we will use the term AlFx. It was shown that addition of AlFx could activate the GDP-bound α subunit of heterotrimeric G-protein, and trigger the conversion to the GTP-bound conformation. AlFx mimics the terminal γ-phosphate of the GTP and is positioned within the nucleotide pocket of the complex in a conformation that resembles that of a transition state. A conserved arginine residue within the catalytic pocket is required to stabilize the AlFx bound to the Gα-GDP [9]. However, for the small GTP-binding proteins, experiments revealed that AlFx was inefficient in directly stimulating the conversion to the GTP-bound conformation [10]. Structural studies demonstrated that the required arginine residue was absent from the nucleotide pocket of the small GTP-binding proteins. However, all the GAPs for small GTP-binding proteins possess an invariant arginine residue. It was therefore proposed that the GAP might be able to stabilize AlFx in the nucleotide pocket of the small GTP-binding protein forming a trimolecular complex [11]. Biochemical experiments demonstrated that AlFx stabilized a complex, most likely a transition-state, between Ras and two corresponding Ras-GAPs. This was later confirmed when the structures of the co-crystals of Ras-GDP-AlFx with p120^GAP and RhoA-GDP-AlFx with p50RhoGAP were solved [12] and [13]. Within this complex the GAP activity is inoperative but the GAP can still act as an effector bound to the active conformation of its cognate small G protein.

In cell-free assays, the recruitment of the β-COP subunit of the COPI coatomer onto a preparation of Golgi membranes was shown to be enhanced in the presence of AlFx [14] and [15]. Similarly, AlFx was shown to induce the formation of large peripheral membrane protrusions only in cells expressing exogenous Arf6 but not in untransfected cells [16] and [17]. However, because AlFx had no direct effect on several small G proteins tested [10], including Arf, it was concluded that the effects of AlFx accounted for the involvement of heterotrimeric G proteins. Here, we demonstrate that AlFx can stabilize in vitro and in vivo a complex between Arf1 and ASAP1, and that in vivo the complex Arf1-AlFx-ASAP1, in which the GAP activity is ineffective, delays the recruitment of paxillin to FCs. Our results support the role of ASAP1 as an Arf1 effector altering the formation of FCs and cell spreading which is independent of its GAP activity.

2. Materials and methods

2.1. Production and purification of recombinant Arf proteins

For the production of myristoylated Arf1 BL21 (DE3) bacteria were co-transformed with pET-3c-Arf and pBB131 coding for the yeast N-myristoyl-transferase [18], and purified as previously described [19].

2.2. Preparation of phospholipid vesicles

Sucrose-loaded liposomes of defined composition were prepared by extrusion through a 0.4 μm pore polycarbonate filter Isopore™ (Milllipore, Molsheim, France) [20]. The composition was 50% egg phosphatidylcholine, 19% liver phosphatidylethanolamine, 5% brain phosphatidylserine, 7.5% liver phosphatidylinositol, 16% cholesterol and 2.5% phosphatidylinositol biphosphate. Lipids were from Avanti Polar Lipids (Birmingham, AL).

2.3. Sedimentation assay

Myr-Arf1 was loaded with GDP ± AlFx (10 mM NaF and 100 μM AlCl3) for 30 min at 37 °C in the absence or presence of liposomes (2 mg/ml). When indicated PZA was added and incubated 15 min at 25 °C. Liposomes were then separated by ultracentrifugation and the amount of proteins bound to liposomes was quantified after SDS–PAGE and sypro orange (Bio-Rad, France) staining by densitometric analysis (LAS 3000, Fujifilm, France).

2.4. Transfection, immunoprecipitation and immunoblot

BHK cells grown in 10-cm petri dishes were transfected with the indicated plasmids using Fugene 6 according to the manufacturer’s instructions (Roche Diagnostics, Mannheim, Germany). 24h post-transfection the cells were exposed or not to AlFx (10 mM NaF and 100 μM AlCl3) for 30 min at 37 °C, lyzed in 0.5% Triton X-100, 20 mM HEPES, 125 mM NaCl, 1 mM MgCl₂, 1 mM GDP, 0.5 mM DTSSP (3,3′-dithiobis[sulfosuccinimidylpropionate]) (Pierce, Perbio Science, France) and proteases inhibitors (Complete™ cocktail tablets, Roche Diagnostics). When indicated the lysis buffer was supplemented with 10 mM NaF and 100 μM AlCl3. Lysates were centrifuged, precleared and incubated overnight at 4 °C with anti-FLAG agarose beads (Sigma–Aldrich). The immunoprecipitates were washed five times, resolved by SDS–PAGE and the indicated proteins detected after immunoblot by chemiluminescence (ECL™, Amersham France). Mouse monoclonal anti-FLAG M2 and anti-HA (clone 12CA5) antibodies were from Sigma–Aldrich and Roche Diagnostics, respectively.

2.5. Cell-substratum assay and immunofluorescence

REF52 cells were cultured in DMEM supplemented with 10% fetal bovine serum, 10 μg/ml penicillin and 0.25 μg/ml streptomycin. To follow the formation of FCs, 16 h after transient transfection with full-length FLAG-ASAP1 construct using Fugene 6 (Roche Diagnostics), the cells were trypsinized and replated on fibronectin-coated coverslips (2.5 μg/cm²) for different periods of time. When indicated the cells were exposed to AlFx at 37 °C during the last 2 h of the incubation period by addition of 30 mM NaF and 50 μM AlCl3. At the end of the incubation, the cells were fixed in 4% PFA and processed for immunofluorescence analysis as described previously [21]. Polymerized actin, FCs or ASAP1 were labeled with phalloidin (Sigma–Aldrich), anti-paxillin (Becton–Dickinson Transduction Laboratories) and anti-FLAG antibodies (Sigma–Aldrich), respectively. Confocal microscopy analysis was carried out with a Leica TCS-SP microscope equipped with a mixed-gas argon/krypton laser (Leica Microsystems).

3. Results

To determine if AlFx can stabilize the formation of a complex between Arf1 and ASAP1 we performed a sedimentation assay using purified proteins and lipid vesicles. This assay is based on the biochemical properties of Arf1, which is soluble in its inactive GDP-bound form and lipid membrane-associated in its active GTP-bound form [22] and [23]. Thus, when the myristoylated Arf1 (myr-Arf1) loaded with GDP or the PZA fragment, the minimal active fragment of the Arf1-GAP ASAP1 [3], were incubated with liposomes they were found to be mostly soluble (Fig. 1). Only 27% of myr-Arf1 and 34% of PZA were recovered in the pelleted membrane fraction. The addition of AlFx did not change the sedimentation properties of either protein. However, when mixed together in the presence of AlFx, both PZA and myr-Arf1 translocated into the pellet within the same proportions 74% and 83%, respectively (Fig. 1). These results illustrate directly for the first time the formation of a complex between the activated form of Arf1 (Arf1GDP-AlFx) and an Arf-GAP. Taking into account that 680 nM (34% of 2 μM) of PZA spontaneously associated with the liposomes, 800 nM of PZA (74% of 2 μM minus 680 nM) was recruited by approximately 830 nM of myr-Arf1GDP-AlFx (83% of 1 μM) assuming that the majority of the membrane-bound Myr-Arf1GDP (27%) had been converted to an active conformation. Thus, the quantification indicates that the stoichiometry of the complex Arf1:Arf-GAP stabilized by AlFx is 1:1.

We then set out to demonstrate a physical interaction in vivo between ASAP1 and Arf1 by co-immunoprecipitation. BHK cells were transfected with the plasmid constructs encoding for HA-Arf1 and FLAG-ASAP1. 24 h post-transfection, the cells were exposed or not to AlFx, solubilized in the presence of the cross-linker DTSSP, and then submitted to immunoprecipitation with an anti-FLAG antibody coupled to agarose-beads. As shown in Fig. 2, very low levels of Arf1 could be found co-immunoprecipitated with ASAP1 in the absence of AlFx. However, the amount of Arf1 co-immunoprecipitated with ASAP1 was significantly increased in the presence of AlFx. As a negative control cell lysates were immunoprecipitated with an irrelevant antibody. In conclusion, cells exposed to AlFx sustain an Arf1GDP-AlFx-ASAP1 complex. Together with our in vitro experiments this result indicates that AlFx stabilizes efficiently the interaction between Arf1GDP and ASAP1.

Since the inhibitory effects of wild-type ASAP1 overexpression on the formation of FCs and cell spreading are well described, the addition of AlFx could be functionally assessed. FCs are typically enriched in paxillin that decorate the ends of actin cables oriented towards the periphery of the cell. As shown in Fig. 3A, when REF52 cells are plated on fibronectin they form FCs as early as 4 h, with fully mature FCs visible all around the cells at 6 h accompanied with a progressive spreading of the cells also achieved by 6 h. In cells overexpressing FLAG-ASAP1, we observed that FCs were totally absent at 4 h and the presence of rounded cells. FCs could be detected at 6 h and were mature only after 8 h in wellspread cells (Fig. 3A). This is consistent with previous reports showing that overexpression of ASAP1 delays but does not prevent the formation of FCs. ASAP1 could either act solely by hydrolyzing the GTP of Arf1 or act as well as an effector of the activated Arf1GTP. We reasoned that if ASAP1 were acting as an effector for Arf1GTP, the stabilization by AlFx of a complex between ASAP1 and Arf1 in an active conformation should lead to a greater delay in the formation of the FCs in cells overexpressing ASAP1. In Fig. 3B, control or FLAG-ASAP1 transfected REF52 cells have been plated on coverslips coated with fibronectin and let sit for a total period of 8 or 12 h. AlFx alone had no obvious effects on the formation of FCs in control cells. Similarly, after 8 h incubation in normal media ASAP1-transfected cells formed mature FCs and were well spread, indicating that free ASAP1 does not affect FCs formation under these conditions (Fig. 3A and B). However, in ASAP1-transfected cells exposed to AlFx the FCs were virtually absent and the actin cytoskeleton was forming a net vertex surrounding the cells with no apparent small filopodia oriented outwards (Fig. 3B). Interestingly, when incubated for 12 h the cells ultimately formed FCs. This delay is consistent with those that others and we had observed in cells overexpressing wild-type ASAP1 (see Fig. 3A). Thus, the finding that the appearance of FCs is even further delayed is consistent with the role hypothesized for AlFx stabilization of Arf1/ASAP1 in vivo. Furthermore, since the GAP activity of ASAP1 is inoperative within the complex, our results indicate that the GAP activity of ASAP1 is not required for ASAP1 to interfere with FCs formation.

4. Discussion

AlFx activates the α subunit of heterotrimeric G proteins by mimicking the γ-phosphate of GTP upon its insertion, together with GDP, within the nucleotide-binding pocket [9]. However, following the same experimental conditions that led to Gα activation, AlFx was found inactive on Ha-ras, Rap1A, Rab1A, Rab1B, Rab3B, Arf1 and Arf6 ([10] and our unpublished results). Thus, it was concluded that small GTP-binding proteins were insensitive to AlFx, and as such AlFx could be used to discriminate biological functions associated to either heterotrimeric or small GTP-binding proteins. For this reason, even today, in experimental models where Arf proteins are found to be activated following AlFx treatment it is suggested that it is an indirect effect of the prior stimulation of an heterotrimeric G-protein. However, almost 10 years ago AlFx was found to stabilize a complex between RasGDP-AlFp120^GAP allowing for the crystallization of the complex [11] and [12]. Similar results were reported for Cdc42 and RhoA suggesting that Arf proteins could also be sensitive to AlFx in the presence of an Arf-GAP [13] and [24].

Here, we show for the first time that AlFx can stabilize a complex between Arf1GDP and PZA, the minimal active fragment of the Arf1-GAP ASAP1 [3]. We and others had previously shown that Arf1GDP is soluble while its active GTP-bound form associates to lipid membranes [22] and [23]. We took advantage of this property and performed sedimentation experiments to monitor the association of myr-Arf1 to liposomes. When myr-Arf1 and PZA proteins were mixed together in the absence of AlFx their supernatant/pellet distributions were unaffected. However, when AlFx was added to the reaction, myr-Arf1 was translocated to the membrane indicating that it had been converted to an active conformation resembling that of an Arf1GTP. In addition, myr-Arf1 recruited PZA onto the liposomes in a dose-dependent and saturable manner. The quantitation demonstrates that the interaction obeys to a stoichiometry of one molecule of Arf1 for one molecule of PZA. These studies indicate that AlFx mimics the presence of the γ-phosphate of the GTP in myr-Arf1GDP, converting Arf1 in an active GTP-bound conformation that interacts with PZA in association to lipid membranes.

Our results help to explain the original data showing that AlFx could stimulate β-COP recruitment to Golgi-membranes in an Arf-dependent manner [25]. Furthermore, they are consistent with two recent reports. First, AlFx was used to demonstrate that a complex between the purified coatomer, including the integral ArfGAP1, and Arf1GDP was possible solely onto small synthetic liposomes of a given size due to the presence of a curvature-sensing domain within ArfGAP1 [26]. Second, in NRK cells, overexpressed ArfGAP1 was co-precipitated with endogenous β-COP only after the cells were pretreated with AlFx [27]. These results can easily be explained by the formation of a direct complex between Arf1 and ArfGAP1 stabilized by AlFx as demonstrated for Arf1 and ASAP1 in our experiments. Similarly to Ras and RhoA, one can envision to use AlFx to co-crystallize Arf1 or Arf6 with their cognate GAPs to obtain information regarding the molecular mechanism of GTP hydrolysis.

Overexpression of ASAP1 has been shown to retard the formation of FCs, cell spreading and cell motility in various cell systems. Whereas in all cases the GAP activity appears to be required to alter the cell motility, there are contradictory results as to whether a delay in cell spreading and FCs formation are dependent on the GAP activity. While the catalytically inactive mutant, R497K, appeared much less effective than the wild-type ASAP1 to disturb FCs formation, in contrast the ASAP1 mutant deleted of the entire GAP domain, ΔZA, inhibited cell spreading and FCs formation to a much greater extent than the wild-type protein. It was proposed that the point mutant R497K that can still bind Arf1, in contrary to ΔZA, would act as a dominant negative preventing Arf1 to bind to the effector required to alter FCs formation. However, this is in opposition with the inhibitory results observed by overexpression of the wild-type protein that could also act to sequester Arf1. Instead, we propose a model whereby ASAP1 would simply act as an effector of Arf1GTP. The binding to Arf1GTP would result in a conformational change of ASAP1 to reveal the effector domain responsible for the observed inhibitory effects. Consistent with this model, we propose that the GAP domain deleted mutant ΔZA would always be in an opened conformation with the effector domain free to interact with its specific partners. In contrast, the point mutation R497K would decrease the affinity of ASAP1 for Arf1GTP thereby reducing the probability to adopt an opened conformation. In the presence of AlFx, ASAP1 bound to Arf1 in an active conformation (Arf1GDP-AlFx) would be blocked in an opened conformation revealing the effector domain responsible for the inhibitory effects. Our model is in agreement with another study suggesting that lipid binding to the PH domain of ASAP1 would cause a conformational change necessary to allow for the binding of the substrate Arf1GTP [28]. Taken together, the results are consistent with our model whereby ASAP1 acts as an effector of Arf1GTP to interfere with FCs formation and cell spreading. These effector functions are independent of the GAP catalytic activity per se, however they are dependent on the formation of the Arf1GTP/GAP complex and the multiple protein–protein interacting domains of the GAP.

Acknowledgments

We are grateful to Dr. Randazzo for providing us with purified PZA and the ASAP1 expression vector. We are indebted to Dr. K. Singer for stimulating discussions and critical review of the manuscript. We also thank Mariagrazzia Partisani for her technical skills. This work was supported by a grant from the Association pour la Recherche sur le Cancer to F.L. and M.F., and from the Cancéropole PACA projet AxeIII to P.C.. S.K. is the recipient of a fellowship from the Ministère de la Recherche et de l’Education.