GIT proteins, A novel family of phosphatidylinositol 3,4, 5-trisphosphate-stimulated GTPase-activating proteins for ARF6.

ADP-ribosylation factor (ARF) proteins are key players in numerous vesicular trafficking events ranging from the formation and fusion of vesicles in the Golgi apparatus to exocytosis and endocytosis. To complete their GTPase cycle, ARFs require a guanine nucleotide-exchange protein to catalyze replacement of GDP by GTP and a GTPase-activating protein (GAP) to accelerate hydrolysis of bound GTP. Recently numerous guanine nucleotide-exchange proteins and GAP proteins have been identified and partially characterized. Every ARF GAP protein identified to date contains a characteristic zinc finger motif. GIT1 and GIT2, two members of a new family of G protein-coupled receptor kinase-interacting proteins, also contain a putative zinc finger motif and display ARF GAP activity. Truncation of the amino-terminal region containing the zinc finger motif prevented GAP activity of GIT1. One zinc molecule was found associated per molecule of purified recombinant ARF-GAP1, GIT1, and GIT2 proteins, suggesting the zinc finger motifs of ARF GAPs are functional and should play an important role in their GAP activity. Unlike ARF-GAP1, GIT1 and GIT2 stimulate hydrolysis of GTP bound to ARF6. Accordingly we found that the phospholipid dependence of the GAP activity of ARF-GAP1 and GIT proteins was quite different, as the GIT proteins are stimulated by phosphatidylinositol 3,4, 5-trisphosphate whereas ARF-GAP1 is stimulated by phosphatidylinositol 4,5-bisphosphate and diacylglycerol. These results suggest that although the mechanism of GTP hydrolysis is probably very similar in these two families of ARF GAPs, GIT proteins might specifically regulate the activity of ARF6 in cells in coordination with phosphatidylinositol 3-kinase signaling pathways.

ankyrin repeats, and a carboxyl-terminal GRK interaction domain. One major splice variant of GIT2, termed GIT2-short, lacks the carboxyl-terminal GRK interaction domain (35). Overexpression of GIT1 leads to reduced ␤ 2 -adrenergic receptor signaling and increased receptor phosphorylation, which appear to result from reduced receptor internalization and resensitization (11). These cellular effects of GIT1 require an intact ARF GAP activity, suggesting a critical role of the coupling of GIT1 and ARF in ␤ 2 -adrenergic receptor endocytosis.
Both ASAP1/DEF-1 and PAP/KIAA0400 were identified through interactions with known proteins, the Src and Pyk2 kinases, respectively (13,14). They share a similar organization, with a PH domain, central ARF GAP-like putative zinc finger domain, multiple ankyrin repeats, and a carboxyl-terminal SH3 domain. Both were active as GAPs for ARF1 and ARF5, but poor GAPs for ARF6, and both were activated by PIP 2 (13,14). PAP was localized to the Golgi complex and the plasma membrane and shown to prevent the ARF-dependent generation of post-Golgi vesicles, in vitro (14), whereas ASAP1/ DEF-1 promoted the differentiation of fibroblasts into adipocytes (15). To understand the similarities and differences among ARF GAP proteins, we compared the zinc binding, ARF GAP activity, ARF family specificity, and lipid regulation of GIT1 and GIT2 to those of ARF-GAP1.

EXPERIMENTAL PROCEDURES
Materials-TLC plates were purchased from VWR Scientific and lipids from Sigma. Sources of other materials have been published (11, 16 -18).
Preparation of Recombinant ARF Proteins-For large-scale production of recombinant ARF proteins, 10 ml of overnight culture of the appropriate transformed bacteria were added to a flask with 500 ml of LB broth and ampicillin, 50 g/ml, followed by incubation at 37°C with shaking. When the culture reached an A 600 of 0.6, 250 l of 1 M isopropyl-␤-D-thiogalactopyranoside was added (0.5 mM isopropyl-␤-D-thiogalactopyranoside final concentration). After incubation for an additional 3 h, bacteria were collected by centrifugation (Sorvall GSA, 6000 rpm, 4°C, 10 min), and stored at Ϫ20°C. Bacterial pellets were dispersed in 5 ml of cold phosphate-buffered saline, pH 7.4, with trypsin inhibitor (20 g/ml), leupeptin and aprotinin (each 5 g/ml), and 0.5 mM phenylmethylsulfonyl fluoride. Lysozyme (10 mg in 5 ml) was added. After 30 min at 4°C, cells were disrupted by sonication and centrifuged (Sorvall SS34, 16,000 rpm, 4°C, 20 min). The supernatant was applied to a column (2.5 ϫ 100 cm) of Ultrogel AcA 54 equilibrated and eluted with TENDS buffer (20 mM Tris-HCl, pH 8.0, 1 mM EDTA, 100 mM NaCl, 2 mM DTT, 250 mM sucrose, 5 mM MgCl 2 , 1 mM NaN 3 ). Fractions that had both high ARF activity and high purity were pooled and further purified on a column (1.5 ϫ 40 cm) of DEAE, eluted with a linear gradient of 50 to 500 mM NaCl (0.6-ml fractions), and by gel filtration on Ultrogel AcA 34 (1.5 ϫ 30 cm) before storage in small portions at Ϫ20°C. For large-scale production of recombinant myristoylated ARF proteins, expression was induced in bacteria containing the N-myristoyltransferase gene, in the presence of 0.5 M myristic acid and 0.06 M bovine serum albumin as described by Franco et al. (19). Myristoylated ARF6 was purified by hydrophobic interaction HPLC on a TSKgel Phenyl-5PW column (Supelco, Belfonte, PA) using the method described by Randazzo (4). A single protein peak eluted in the decreasing salt gradient and was shown to stimulate cholera toxin-catalyzed ADP-ribosylation.
Construction and Expression of GAP Proteins-Rat GIT1/6xHis, ⌬45GIT1/6xHis were purified from baculovirus-infected Sf9 cells as described previously (11), using nickel affinity and ion exchange chromatography. The human GIT2-short/6xHis (35) was transferred into the pVL1393 shuttle vector, and used to prepare recombinant baculoviruses by recombination in Sf9 cells with Baculo-Gold virus DNA (Pharmingen). The GIT2-short/6xHis protein was purified from infected Sf9 cells using ProBond metal chelate resin (Invitrogen) batchwise, followed by chromatography on a HiTrap-Q column (Amersham Pharmacia Biotech), as described for GIT1/6xHis (11). Full-length rat ARF- and myristoylated ARF6 (lane 9) were purified from Escherichia coli containing a plasmid coding for the N-myristoyl transferase. Five g of each GAP and 2.5 g of each deleted GAP or ARF was separated by SDS-polyacrylamide gel electrophoresis in a 10% gel and stained with Coomassie Blue. Position of molecular size markers is on the left. A doublet pattern for purified GIT2-short and ARF-GAP1 is often seen and likely reflects post-translational modification. GAP1 cDNA (5) was amplified from a rat brain cDNA library and subcloned into a modified pBK-CMV vector (Stratagene) using EcoRI and XhoI. The entire cDNA was then re-amplified using an antisense primer that inserted a 6xHis tag immediately before the stop codon, and subcloned as before into pBK-CMV. The pBK-⌬45ARF-GAP1 construct was prepared by amplification using a 5Ј-primer that adds an initiator Met codon immediately before codon 46. All constructs were sequenced on both strands from specific primers by using automated dye terminator chemistry with AmpliTaq FS reagents (Applied Biosystems) and an ABI 377 instrument. The ARF-GAP1/6xHis and ⌬45ARF-GAP1/ 6xHis cDNAs were subcloned into the EcoRI and XhoI sites of the pVL1392 recombination vector (Pharmingen), which was used to prepare a recombinant baculovirus using Baculo-Gold virus DNA (Pharmingen). The ARF-GAP1/6xHis and ⌬45ARF-GAP1/6xHis proteins were purified from the soluble extract of infected Sf9 cells using nickel affinity and ion exchange chromatography, essentially as described for GIT1/6xHis (11).
Assay of GTPase Activity-The indicated amounts of ARF were incubated for 30 min at 30°C in 20 mM Tris, pH 8.0, 10 mM DTT, 2.5 mM EDTA with bovine serum albumin, 0.3 mg/ml, and phosphatidylserine (PS), 30 g/ml (160 M), with 0.5 M [␣-32 P]GTP (3000 Ci/mmol) and 10 mM MgCl 2 . 10-l samples were incubated at 30°C for 5 to 30 min, with the indicated amounts of GAP protein or an equal volume of buffer (total volume 50 l), in the presence of the indicated phospholipids ( Fig.  3), before proteins with bound nucleotides were collected on nitrocellulose by vacuum filtration (18). Bound nucleotides were eluted in 250 l of 2 M formic acid, of which 3-4-l samples were analyzed by TLC on polyethyleneimine-cellulose plates developed with 1 M formic acid, 1 M LiCl. TLC plates were subjected to autoradiography at Ϫ80°C for 18 -36 h. The remaining solution was used to quantify the total amount of nucleotide bound in the assay (GDP plus GTP) by scintillation counting. The absolute quantities of GTP and GDP, with or without incubation with GAP protein, were calculated by multiplying the total bound nucleotide by the fraction determined to be GTP or GDP by TLC. As shown previously (4,10,20), in most assays the concentration of ARF-GTP was much less than the K m . Under these conditions, substrate is consumed at a first-order rate equal to V max /K m . This rate was meas-

RESULTS AND DISCUSSION
We have previously shown that GIT1 and GIT2, of which both exhibit the conserved putative zinc finger motif found in the ARF-GAP1 protein, stimulated hydrolysis of GTP bound to ARF1, without themselves binding GTP or acting as a nucleotide releaser for ARF1 (11,35). ARF-GAP1 is to date the best characterized GAP for ARFs. We compared the respective activities of purified recombinant, 6xHis-tagged ARF-GAP1, GIT1, and GIT2-short proteins (Fig. 1). Confirming our previous observations (11,35), GIT1 and GIT2 stimulated GTP hydrolysis by ARF1. The rate of hydrolysis of GTP bound to ARF1 induced by GIT1 and GIT2 was comparable to the rate of hydrolysis induced by ARF-GAP1 (Fig. 2A). The activities of , or amino-terminal deleted ARF GAPs for 10 min at 30°C before stimulation of cholera toxin-catalyzed ADP-ribosylagmatine formation was assayed for 60 min at 30°C. ARF activity is the difference between CTA-catalyzed formation of [ 14 C]ADPribosylagmatine without and with ARF1 protein (nanomole/h). Data are mean Ϯ one-half the range of values quadruplicate assays. These findings were replicated twice with two independent preparations of proteins. GIT1, GIT2, and ARF-GAP1 were constant for at least 20 min before slightly slowing down ( Fig. 2A), so for further assays we used 10-min incubations. In addition, the assays appeared linear until nearly 50% of the added GTP-bound ARF substrate was converted to GDP-bound ARF. GIT1, GIT2, and ARF-GAP1 stimulated hydrolysis of GTP bound to ARF1 in a concentration-dependent manner (Fig. 2, B and C). GIT1 was slightly more potent than ARF-GAP1 and GIT2-short was the least potent GAP (Fig. 2C). These results suggest that the intrinsic GAP activity of GIT1, GIT2, and ARF-GAP1 is very similar in the absence of additional cofactors.
Phospholipids seem to play a critical role in the control of ARF activities by ARF regulatory proteins. It was concluded that the effect of phospholipids on ARF GAP activity was to increase the GAP concentration at the membrane where GTPbound ARF resides (8). Investigations with ARF-GAP1 have shown that it is indeed stimulated by PIP 2 and dioleylglycerol (4,8), but the phospholipid dependence of GIT proteins has not been explored. We compared the effects of PIP, PIP 2 , and PIP 3 on the GAP activity of GIT1, GIT2-short, and ARF-GAP1. Confirming previous observations (4,5,20), 200 M PIP 2 dramatically stimulated the GAP activity of ARF-GAP1 from 0.64 to 2.46 nmol/min/mg, but not that of GIT1 (from 0.88 to 0.91 nmol/min/mg) or GIT2 (0.56 to 0.65 nmol/min/mg) (Fig. 3). On the other hand, 200 M PIP 3 significantly increased the hydrolysis of GTP bound to ARF1 induced by GIT1 (0.82 to 2.56 nmol/min/mg) and GIT2 (0.56 to 1.01 nmol/min/mg) but not by ARF-GAP1 (0.64 to 0.87 nmol/min/mg) (Fig. 3). PIP did not increase GTP hydrolysis by any of the ARF GAPs tested here (Fig. 3). Because dioleylglycerol, produced mainly from phos-phatidylcholine hydrolysis by phospholipase D (an effector of ARF), dramatically increased the activity of a recombinant fragment of ARF-GAP1, it was suggested that phospholipase D activity could be a major regulator of ARF GAPs (8). Dioleylglycerol had similar effects on Gcs1, an analogous ARF GAP from yeast (9). We confirm that DAG (C18:1-(3)C18:1) stimulated the GTPase activity of ARF-GAP1, but no effect on the activity of GIT1 or GIT2 was found (Fig. 3). The ARF GAPs that are activated by PIP 2 or other phosphoinositides are presumably subject to diverse kinds of regulation. From these results it seems plausible that GIT proteins and ARF-GAP1 are involved in distinct signaling pathways, which is also suggested by their distinct cellular localization. Centaurin-␣ has been described to be a potential PIP 3 -binding protein with similarity to ARF GAPs that could complement a yeast strain deficient in the yeast ARF GAP Gcs1 (21), suggesting that several ARF GAPs could be regulated by PIP 3 . In many cell types, the agonist-stimulated PI 3-kinase utilizes predominantly PIP 2 as a substrate to generate PIP 3 (22,23). The ratio of PIP 2 and PIP 3 seems to play a critical role in many aspects of vesicular trafficking (24). It will now be of particular interest to know if signaling pathways involving PI 3-kinases or other enzymes involved in the synthesis or hydrolysis of PIP 3 may contribute to intracellular regulation of GIT proteins.
It was demonstrated that the amino-terminal GATA-like zinc finger motif of ARF-GAP1 (5) and ARD1, an ARF-related protein that has an amino-terminal GAP domain (18), are critical for GTP hydrolysis. In the presence of GTP or a nonhydrolyzable analogue, all members of the ARF family serve as allosteric activators of cholera toxin (CTA) ADP-ribosyltransferase (25). As expected, addition of ARF-GAP1 and GIT1 reduced the ability of ARF1 to activate CTA in the presence of GTP, but not GTP␥S (Fig. 4). This result confirms that GIT1, like ARF-GAP1, influences the biological activity of ARF1 by promoting GTP hydrolysis. Deletion of 45 amino acids from the amino terminus of GIT1 or ARF-GAP1 (containing the conserved zinc finger motif) completely prevented inhibition of ARF-induced CTA activation (Fig. 4), suggesting that the zinc finger motif is required for GAP activity of both proteins. We can exclude that this loss of activity results from complete misfolding of the ⌬45GIT1 protein, because several additional GIT interacting proteins (GRK2, PIX, paxillin) do bind normally to ⌬45-GIT1 in co-immunoprecipitation assays.

TABLE I Zinc content of ARF GAPs assessed by inductively-coupled plasma emission spectroscopy
The indicated amounts of Hi-Trap-Q chromatography purified Histagged ARF-GAP1 (GAP1), GIT1, and GIT2-short proteins were analyzed for metal ions by inductively-coupled plasma emission spectroscopy. The detection limit for zinc was 0.1 g/ml. No significant amounts of the other metal ions were found in any samples, except for Na ϩ from the buffer and Ca 2ϩ in single samples of ARF-GAP1 and GIT2-short. Members of the ARF GAP family identified to date share a conserved domain containing a CX 2 CX 16 CX 2 C putative zinc finger (5,11,13,14). To investigate whether ARF GAP proteins actually bind to a metal through this domain, we subjected three purified ARF GAP proteins to plasma emission spectroscopy, a sensitive technique that allows for the simultaneous detection of over 20 metal ions (Table I). For each of the three ARF GAP proteins analyzed, the only metal ions detected over background were zinc and sodium (from the NaCl buffer), except for one instance, when significant calcium was also detected in ARF-GAP1/6xHis and GIT2-short/6xHis. The nanomole of zinc detected was similar to the nanomole of protein analyzed for each ARF GAP, consistent with a near one-to-one complex of zinc with protein. Recent publication of the crystal structure of the ARF-GAP1 protein amino-terminal domain bound to ARF1 reveals that one zinc ion is indeed bound by the four conserved cysteine residues of the ARF GAP domain (26). Interestingly, the metal ion does not contact the ARF protein, but appears to play a role in determining the overall structure of the GAP domain (26). Mutation of single cysteine residues within this zinc finger abrogated GAP activity of ARF-GAP1 (5) and ARD1 (18), presumably because such mutants could not bind zinc. As GIT1 and GIT2 also appear to bind zinc, we predict that zinc chelation by this CX 2 CX 16 CX 2 C motif is a common feature of ARF GAP proteins. The effect of removing the zinc to form the cognate apoproteins, or of replacing it with similar metal ions, remains unexplored. It is now well established that RasGAPs and RhoGAPs share a common mechanism of the GTPase-rate enhancement involving a critical arginine residue (27,28). A putative arginine finger motif has also been postulated within the zinc finger region in ARF-GAP1 (29) and this specific arginine residue is present also in GIT1 and GIT2. Recently, data from the crystal structure of the ARF1⅐ARF-GAP1 complex suggested that the postulated arginine in ARF-GAP1 did not make contact with the active site of the GTPase (26). The function of this conserved arginine residue in ARF GAPs remains to be established.
We tested further the GAP activities of GIT1, GIT2-short, and ARF-GAP1 on different members of the ARF family. ARF-GAP1 stimulated hydrolysis of GTP bound to ARF1, ARF2, ARF3, and ARF5, but not to ARF6 (Fig. 5A), confirming previous observations that this GAP affects the class I and class II, but not class III ARFs (4). On the other hand, GIT1 and GIT2 promoted GTP hydrolysis by all five ARFs tested (Fig. 5A). Neither ARL1 nor ARL2, members of the ARF-like family, nor ARD1, the ARF-related protein with its intrinsic GAP domain, were substrates of GIT1, GIT2, or ARF-GAP1 (Fig. 5A). Over a range of GAP concentrations where the two GIT proteins were quite active, ARF-GAP1 failed to accelerate the GTPase activity of ARF6 (Fig. 5B). These results suggest that ARF-GAP1 and GIT proteins are indeed specific GAPs for ARF proteins. More importantly, they indicate that unlike ARF-GAP1, both GIT1 and GIT2 can stimulate the efficient hydrolysis of GTP bound to ARF6.
Native ARF proteins in the cell are myristoylated on their amino termini, while the bacterially expressed recombinant ARF proteins used in the preceding assays in this study are not. The effect of myristoylation of ARF on the GAP activity of GIT1, GIT2-short, and ARF-GAP1 was also tested. No significant difference in GAP activities was detected whether or not ARF1 or ARF6 were myristoylated (Fig. 6A). These results are in agreement with previous observations made with a different ARF GAP activity purified from rat spleen (10). It is notable that myristoylated ARF6, the form found in cells, was a particularly good substrate for GIT1 and GIT2 compared with ARF-GAP1 (Fig. 6B).
Of all ARF GAP proteins characterized to date, the GIT proteins appear to be the best GAPs for ARF6, a type III ARF with several unusual features including a predominantly plasma membrane localization. ARF6 also has been localized on endosomes (30,31) and on secretory granules (32), where it is believed to participate in endocytotic and exocytotic trafficking events. Interestingly, GIT proteins were identified in a yeast two-hybrid screen through their ability to interact with GRK2. Overexpression of GIT1 leads to increased receptor phosphorylation and reduced ␤ 2 -adrenergic signaling, resulting from attenuated receptor internalization and resensitization (11). These cellular effects do not reflect regulation of GRK kinase activity, but require an intact amino terminus, suggesting a function for ARF in regulating ␤ 2 -adrenergic receptor endocytosis (11). The involvement of ARF6 in this specific recycling pathway will now have to be investigated.
In addition to their role as ARF GAPs, GIT proteins also appear to have other important cellular functions. A two-hybrid screen with GIT1 identified an interaction with ␤-PIX, a putative rac1/cdc42 guanine nucleotide-exchange factor that binds the rac1/cdc42-activated PAK kinases (35). Both GIT1 and GIT2 interact with ␣and ␤-PIX proteins, in a multiprotein complex which also contains p21-activated kinase (PAK) (35). Bagrodia et al. (33), starting with PAK kinase, recently identified the GIT⅐PIX⅐PAK complex, as did Turner et al. (34), who additionally discovered the interaction of a distinct third GIT family member (p95-PKL) with paxillin. Together, these studies document a role for GIT proteins in anchoring the PIX-PAK complex, via paxillin, in cellular focal adhesions. These results suggest that the multidomain proteins of the GIT family could be at the cross-roads of several signal transduction pathways involving multiple small GTPases in vesicular trafficking and cytoskeletal remodeling.