Arf GAPs and Their Interacting Proteins

Authors

  • Hiroki Inoue,

    1. Laboratory of Cellular and Molecular Biology, Center for Cancer Research, National Cancer Institute, Bethesda, MD 20892, USA
    Search for more papers by this author
  • Paul A. Randazzo

    Corresponding author
    1. Laboratory of Cellular and Molecular Biology, Center for Cancer Research, National Cancer Institute, Bethesda, MD 20892, USA
    Search for more papers by this author

Paul A. Randazzo, randazzo@helix.nih.gov

Abstract

Membrane trafficking and remodeling of the actin cytoskeleton are critical activities contributing to cellular events that include cell growth, migration and tumor invasion. ADP-ribosylation factor (Arf)-directed GTPase activating proteins (GAPs) have crucial roles in these processes. The Arf GAPs function in part by regulating hydrolysis of GTP bound to Arf proteins. The Arf GAPs, which have multiple functional domains, also affect the actin cytoskeleton and membranes by specific interactions with lipids and proteins. A description of these interactions provides insights into the molecular mechanisms by which Arf GAPs regulate physiological and pathological cellular events. Here we describe the Arf GAP family and summarize the currently identified protein interactors in the context of known Arf GAP functions.

The ADP-ribosylation factor (Arf) GTPase activating proteins (GAPs) are a family of multidomain proteins expressed in eukaryotes. The proteins were first identified as regulators of Arf proteins. The Arfs are GTP-binding proteins that control membrane traffic and remodeling of the actin cytoskeleton. By regulating Arfs, Arf GAPs affect both membranes and the actin cytoskeleton. However, the Arf GAPs have multiple domains that can function both dependently and independently of Arf proteins to elicit structural changes and to transduce signals in cells.

The Arf family of GTP-binding proteins is a subfamily of the Ras superfamily. There are six Arf genes in the mammalian genome (five in the human genome). The six Arf proteins are divided into three classes on the basis of amino acid sequence (1). Class I includes Arf1, Arf2 and Arf3; class II includes Arf4 and Arf5; and class III includes Arf6. Of these, Arf1 and Arf6 are the most extensively studied (2–4). Arf1 has been implicated in Golgi–endoplasmic reticulum (ER) retrograde transport, intra-Golgi transport, trafficking from the trans Golgi network (TGN), transport in the endocytic pathway and recruitment of paxillin to focal adhesions (FAs). Arf6 has been found to affect endocytosis, phagocytosis, receptor recycling and the formation of actin-rich protrusions and ruffles (4,5).

The function of Arf proteins is dependent on binding and hydrolyzing GTP, thereby cycling between GTP-bound (Arf·GTP) and GDP-bound (Arf·GDP) forms of the protein. The nucleotide exchange rate intrinsic to Arf is less than 0.01/min and the intrinsic GTPase rate is not detectable (6). Consequently, the cycle requires the action of accessory proteins called guanine nucleotide exchange factors and GAPs. ArfGAP1 was the first Arf GAP purified and cloned. Examination of ArfGAP1 in vivo yielded data that were consistent with function as an Arf1 regulator at the Golgi apparatus (7). The Arf GAPs that were subsequently identified are structurally complex proteins with molecular weights between 80 and 200 kDa. In addition to regulating membrane traffic, several Arf GAPs have been found to be regulators of the actin cytoskeleton and to be elements of signal transduction pathways.

The actin structures affected by Arf GAPs are integrated with membranes and involved in changes in cell shape or cell movement such as migration and cancer cell invasion. The actin structures include FAs, invadopodia (and highly related podosomes), peripheral membrane ruffles and circular dorsal ruffles (CDRs). Focal adhesions, invadopodia and podosomes (8,9) attach the actin cytoskeleton to the extracellular matrix (ECM), mediate signal transduction and are sites of active endo- and exocytosis. Peripheral membrane ruffles are actin-rich extensions from the cell edge that are also sites of active endo- and exocytosis and signaling (10). Circular dorsal ruffles are ring-like structures projecting from the dorsal surface of the cell. Circular dorsal ruffles are comprised of labyrinths of tubulated membranes linked to polymerized actin and are sites of endocytosis (8,9,11,12).

Part of the effect of the Arf GAPs on cytoskeletal structures is mediated by the regulation of Arfs. Arf6 has been implicated in the formation of invadopodia and peripheral membrane ruffles (5,13). Arf1 has been implicated in FA dynamics (14). However, the molecular basis for Arf function at each of these sites is not well defined. Furthermore, some of the effects of Arf GAPs on the cytoskeleton have been found to be independent or only partly dependent on GAP activity; in these cases, the dominant determinant of the effect of the Arf GAP protein is association with proteins that contribute to the regulation or structure of the cytoskeletal element. These considerations together with examination of Arf GAPs as Arf effectors have provided the basis for models explaining the regulation of cell movement and new insights into the regulation of the cytoskeleton and associated signaling. In this review, we describe the Arf GAP family, catalog proteins that interact with Arf GAPs and discuss hypotheses related to the significance of the associated proteins to cell signaling and regulation of cell structure.

The Arf GAP family

The Arf GAPs have a common domain, the Arf GAP domain, comprising a zinc-binding motif. At least 24 genes that encode proteins with Arf GAP domains have been found in the human genome. Most of them have several synonyms, which can be a source of confusion in the Arf GAP literature. In Figure 1A, the synonyms are summarized. We have classified the Arf GAPs into two major types, ArfGAP1 and AZAP types, according to the overall domain structure (15) (Figure 1). The former type of Arf GAPs have an Arf GAP domain at the extreme N-terminus of the protein and the latter contain an Arf GAP domain between the PH and ankyrin (ANK) repeat domains. Each Arf GAP group is further subdivided based on additional domains (Figure 1B).

Figure 1.

Figure 1.

Synonyms and representative domain structures of Arf GAPs. A) Synonyms of Arf GAPs. Arf GAPs were classified into two large types and subdivided into seven subtypes based on their overall domain structures. Synonyms that have been used in literature and databases are listed with their NCBI Gene ID. B) Domain structures of representative Arf GAPs. One from each subtype was drawn to scale. Arf GAP, Arf GAP domain; CALM BD, CALM binding domain; CB, clathrin box; CC, coiled-coil domain; h, human; m, mouse; Pro (D/ELPPKP)8, 8 tandem proline-rich (D/ELPPKP) motifs; Pro (PxxP)3, 3 proline-rich (PxxP) motifs; r, rat; Rho GAP, Rho GAP domain.

The six genes grouped as ArfGAP1 type are divided into three subtypes – ArfGAP, SMAP and GIT. ArfGAP1 and ArfGAP3 belong to the Arf GAP subtype. In addition to an N-terminal GAP domain, ArfGAP1 has two ArfGAP1 lipid-packing sensor (ALPS) domains, which are short stretches of amino acids that can sense membrane curvature (16). The GAP activity of ArfGAP1 is stimulated by diacylglycerol and by increases in membrane curvature (16,17). The primary function of ArfGAP1 is regulation of the Golgi apparatus. Initially, ArfGAP1 was proposed to function in coat disassembly of transport vesicles from the Golgi through inactivation of Arf1 (7,18,19). Consistent with this model, the ALPS domains of ArfGAP1 provide a mechanism by which GAP activity increases with changes in membrane curvature associated with forming transport vesicles (20,21). Recent evidence indicates that ArfGAP1 may also be a component of a vesicle coat complex that promotes cargo sorting and that drives vesicle formation (22–26).

SMAP1 and SMAP2 have been recently characterized (27,28). Each contains a clathrin box for clathrin binding in addition to an N-terminal Arf GAP domain. SMAP1 and SMAP2 function as GAPs for Arf6 and Arf1, respectively. Evidence for this specificity includes colocalization with Arf6 in HeLa cells treated with aluminum fluoride (AlF4) (27,28), which is thought to stabilize an Arf–GDP–AlF4–GAP complex (29). SMAP2 colocalized with the clathrin adaptor protein-1 (AP1) and EpsinR (an AP1-binding protein involved in TGN38 trafficking), on early endosomes and TGN, which is an Arf1-dependent pathway (28). The primary function of SMAPs is thought to be as regulators of Arfs.

Two members of the GIT subfamily, GIT1 (Cat1/p95APP1) and GIT2 (Pkl/Cat2/p95APP2), have three ANK repeats, a Spa-homology domain (SHD), a coiled-coil domain and a C-terminal paxillin-binding site (PBS). The GITs have been proposed to function as regulators of both membrane traffic and FAs. Although GITs do not show a preference for particular Arf isozymes in vitro (30), they colocalize with Arf6 in the cell periphery in vivo. Moreover, functional analyses of GITs on endocytic events, including G-protein-coupled receptor internalization, suggest GITs function with Arf6 (31,32). It has also been reported that overexpression of GIT2-short, a splice variant of GIT2 lacking the C-terminal PBS, induced redistribution of β-COP, which is consistent with function as an Arf1 GAP (33).

AZAP-type Arf GAPs are characterized by a PH, Arf GAP, ANK repeat domain structural motif. Twelve genes for AZAPs are subdivided into four subtypes (ASAPs, ACAPs, ARAPs and AGAPs). ASAP-subtype GAPs comprise a Bin/amphiphysin/Rvs (BAR) domain, a PH domain, an Arf GAP domain, ANK repeats, a proline-rich (Pro) domain and an SH3 domain; ASAP3 (UPLC1/DDEFL1/ACAP4) does not have the C-terminal SH3 domain. ASAP1 (DEF1/DDEF1/centaurin β4/AMAP1/PAG2) has been implicated in the regulation of FAs, CDRs, invadopodia and podosomes (12,34,35). ASAP1 and ASAP2 (Papα/DDEF2/centaurin β3/AMAP2/PAG3) prefer Arf1 and Arf5 to Arf6 in vitro (36,37). Consistent with the in vitro results, overexpression of ASAP1 in vivo has been reported to decrease Arf1·GTP levels and to increase Arf6·GTP levels (38,39). Furthermore, ASAP1 does not inhibit Arf6-dependent membrane protrusions (40). Some evidence supports the idea that ASAP1 and ASAP2 also function with Arf6 through direct binding and slow catalysis (41,42).

ASAP1 GAP activity is more extensively analyzed than those of other Arf GAPs. The GAP activity is stimulated about 10 000-fold by PI(4,5)P2 and phosphatidic acid (PA) (36,43). This stimulation depends on a conformational change that is caused by PI(4,5)P2 binding to the PH domain (43,44). Kinetic studies support the proposal that ASAP1 functions in binary complex with Arf1 to induce GTP hydrolysis; however, a description of the details concerning the contribution of specific amino acids from ASAP1 and Arf to catalysis will require further structural studies (29,44,45).

ACAPs are similar to ASAPs in structure. They are composed of BAR, PH, Arf GAP and ANK repeats domains. ACAP1 (centaurin β1) and ACAP2 (centaurin β2) were biochemically characterized and found to have a preference for Arf6 over Arf1 and Arf5 in vitro and in vivo (40). Both are activated by PI(4,5)P2 and PI(3,5)P2 (40).

ASAPs and ACAPs contain BAR domains. The structure of the BAR domain of Drosophila amphiphysin has recently been solved. It was found to be a crescent-shaped dimer. The BAR domain of amphiphysin as well as those from other proteins including ACAP1 bind synthetic liposomes. The efficiency of binding was inversely related to the radii of the vesicles and directly related to the curvature. Based on this property, the BAR domain was proposed to be a curvature sensor. Recombinant BAR domains also cause deformation of the liposomes, resulting in tubular structures (46). The BAR–PH domain of ACAP1 was found to function primarily as a curvature sensor. Similarly, a recombinant protein comprising the BAR and PH domain of ASAP1 sensed membrane curvature. On the other hand, a recombinant protein comprising the BAR, PH, Arf GAP and ANK repeat domains of ASAP1 poorly sensed membrane curvature but efficiently induced tubulation of liposomes. The tubulating activity was regulated by Arf1·GTP and the GAP domain of ASAP1 (47). A functional relationship with the geometry of membrane surfaces may be a common characteristic of ASAPs and ACAPs, which contain BAR domains, and ArfGAP1, which contains ALPS domains.

ARAPs are the largest proteins in the Arf GAP family. GAPs in this subtype have a Rho GAP domain in addition to an Arf GAP domain. They also contain a sterile α-motif (SAM), five PH domains and a Ras association (RA) domain. Two of the five PH domains contain a PI(3,4,5)P3-binding consensus sequence. PI(3,4,5)P3 more potently stimulates the Arf GAP activity of ARAPs than do other phosphoinositides (48–51). The three members of the ARAP subtype of Arf GAPs have different Arf specificities. ARAP1 (centaurin δ1) functions with Arf1 and Arf5 (48). ARAP2 (centaurin δ1/PARX) preferentially uses Arf6, as compared with Arf1 and Arf5, as a substrate in vitro and in vivo (50). ARAP3 (DRAG/centaurin δ3) has been reported to function as an Arf6 or Arf5 GAP in vitro (49,51) and to regulate Arf6-dependent events including membrane protrusions and ruffling in vivo (49,52).

AGAPs have a GTP-binding protein-like domain (GLD) at the N-terminus. The PH domains of AGAPs are split with an 80-amino-acid insert between β strands 5 and 6. AGAP1 (GGAP1/centaurin γ2) and AGAP2 (PIKE/GGAP2/centaurin γ1) prefer Arf1 and Arf5 to Arf6 in vitro, and function at endosomes with Arf1 in vivo. The GAP activity of the AGAPs is stimulated by PI(4,5)P2 and PA (53–55).

Proteins that interact with Arf GAPs

As described, Arf GAPs are multidomain proteins. Through these domains, Arf GAPs are able to interact with a variety of proteins (Table 1) that we have categorized into three groups based on function in membrane traffic, signaling or regulation of the actin cytoskeleton. Selected interactions are mapped in Figure 2 together with the proposed Arf specificity of the Arf GAPs.

Table 1.  Arf GAPs and their interacting proteins
DesignationBinding proteinBinding site on Arf GAPFunction/remarksReference
  1. CC, coiled-coil domain; NFκB, nuclear factor kappa B; SEAP, secreted alkaline phosphatase.

ArfGAP1γ-adaptin (AP1)C-terminalUnknown62
 KDEL receptor/ERD2UnknownER protein retrieval56
 KDEL protein recruits ArfGAP1 binding 
 p24aUnknownKDEL protein sorting?94
ArfGAP3γ-COP (COPI)UnknownGolgi → ER retrograde transport?61
SMAP1ClathrinClathrin boxClathrin-dependent Tfn endocytosis27
SMAP2CALMCALM binding domainUnknown; clathrin assembly? 
 ClathrinClathrin boxEE → TGN retrograde transport28
GIT1PIXSHDFA turnover95
 Actin remodeling 
 Cell spreading, migration 
 Centrosome maturation 
 PLCγSHDPLCγ activation, IP3 production69
 MEK1SHDMEK1 activation by AngII or EGF77
 FAKSHDCell migration75
 PaxillinPBSFA turnover86
 Hic-5PBSPaxillin-like86
 HuntingtinCC + PBSHuntington disease pathogenesis 
 GPCR kinasesUnknownβ2AR downregulation78
 14-3-3zetaUnknownActin remodeling96
GIT2PIXSHDFA turnover87
 Actin remodeling 
 Cell spreading, migration 
 PaxillinPBSFA turnover86,87
 Hic-5PBSPaxillin-like86
 LeupaxinPBSPaxillin-like; podosome in osteoclast88
 GPCR kinasesUnknownβ2AR donwnregulation79
ASAP1CIN85Pro (PxxxPR)EGFR recycling66
 Tumor invasion 
 CD2APPro?Unknown37
 CortactinPro (PxxP)?Cell migration, tumor invasion34
 CrkPro (PxxP)Unknown36
 CrkLPro?FA turnover35
 ASAP1 recruitment to FA 
 c-SrcPro (PxxP)Unknown36
 FAKSH3Focal adhesion turnover83
 Cell spreading 
 Pyk2SH3Inhibition of ASAP1 GAP activity84
 POB1SH3Cell migration67
ASAP2PaxillinPro?Cell migration41
 Amphiphysin IImProTac endocytosis42
 Pyk2SH3?SEAP secretion?37
ACAP1NOD1/2BAR–PHDownregulation of NFkB90
 Integrin β1pSer around ANKIntegrin β1 recycling60
 Cell migration 
 Tfn receptorUnknownTfR recycling59
 CellubrevinUnknownUnknown59
ACAP2Vaccinia virus K1L proteinUnknownUnknown92
ARAP2RhoARho GAPFA turnover50
 Stress fiber formation 
ARAP3SHIP2SAMUnknown74
 CIN85Pro (PxxxPR)Unknown66
 Rap1RAStimulation of Rho GAP activity85
AGAP1AP3PHEndosome–lysosome trafficking54
 NO-sensitive soluble guanylyl cyclasePH–GAP–ANKUnknown92
AGAP2PLCγN-terminal (Pro)Nucleotide exchange on GLD?70
 Nuclear PI3 kinaseN-terminal (Pro)Activation of nuclear PI3K72
 Protein 4.1NN-terminal (Pro)Downregulation of nuclear PI3K72
 HomerN-terminal (Pro)Coupling of mGluR and PI3K73
 Antiapoptosis in neuronal cell 
 AktGLDActivation of Akt kinase activity81
 Tumor invasion 
 AP1PHAP1-dependent Tfn recycling55
Figure 2.

Figure 2.

Selected Arf GAPs protein complexes involved in receptor trafficking, cell migration and invasion. Arf GAPs that are related to receptor trafficking, FA turnover, cell migration/spreading or tumor invasion are mapped with the interacting proteins and expected substrate Arfs. The GAPs that use Arf1 or Arf6 as a substrate are labeled with blue or yellow, respectively. The GAPs whose substrate specificity is controversial or that may use Arf5 are labeled with green.

Membrane traffic

Transmembrane cargo proteins

Two Arf GAPs, ArfGAP1 and ACAP1, have been found to bind directly to transmembrane proteins that can be considered either cargo or cargo receptors. ArfGAP1 binds to p24 cargo proteins and to ERD2, a receptor for proteins with the ER retention signal KDEL that mediates retrograde transport of ER-resident proteins from the Golgi to the ER (56). Peptides from p24 family proteins have been found to inhibit Arf GAP activity (57). This observation was the basis of a model explaining GAP control of cargo sorting (58), which is described in detail in Nie and Randazzo (26). The effect of ERD2 binding to Arf GAP1 enzymatic activity has not been examined.

ACAP1 interacts with transferrin receptor (TfR), cellubrevin and integrin β1 (59,60). ACAP1 recognizes two distinct diphenylalanine-based sequences in the cytoplasmic tail of TfR. Disruption of the interaction impairs recycling of the receptor to the plasma membrane. These observations were the basis for the suggestion that ACAP1 may function as novel coat or adaptor protein in the recycling compartment. It has been proposed that ArfGAP1 functions in a similar capacity, directly binding cargo to carry it into membrane trafficking intermediates (22).

Membrane traffic coat proteins

Three classes of Arf GAPs have been found to bind to vesicle coat proteins or coat protein adaptors. ArfGAP1 has been found to bind to coatomer and clathrin AP1 (61,62). The work addressing the consequences of the interaction with coatomer is limited. In one study, coatomer was found to stimulate Arf GAP activity (63); it has also been reported that coatomer has only a small effect on Arf GAP activity (64). Whether the coatomer–Arf GAP interaction has an effect on the formation of vesicles has not been established. Both SMAP-type GAPs bind to clathrin (27,28). In addition, SMAP2 interacts with CALM, a clathrin assembly protein (28). The interaction was found to drive the formation of transport intermediates from the plasma membrane and from the TGN. AGAP1 and AGAP2 associate with clathrin adaptor proteins, AP3 and AP1, respectively (54,55). In both cases, AGAP was found to affect the function of the endocytic compartment containing these clathrin adaptors. Specific molecular consequences of the SMAP or AGAP – coat proteins interactions – have not been determined. One possibility that has been proposed is that the Arf GAPs function as a subunit of a vesicle coat protein in a manner analogous to the role of Sec23 in ER-to-Golgi transport mediated by COP-II vesicle coats (65).

Adaptor proteins involved in membrane traffic

ASAP1, ASAP2 and ARAP3 have been reported to bind to proteins that function as part of the endocytic machinery – CIN85, POB1 and amphiphysin IIm (42,66,67). CIN85 is an adaptor protein containing three SH3 domains and a proline-rich domain. CIN85 was first characterized as a Cbl-interacting protein for epidermal growth factor (EGF) receptor (EGFR) internalization. The SH3 domains of CIN85 interact with an atypical proline-rich motif (PXPXPR) in ASAP1 and ARAP3 (66). Overexpression of ASAP1 accelerated EGF and EGFR recycling in CHO cells and HeLa cells (47,66), although the detailed mechanism underlying this effect is still unknown. CD2AP is a protein highly related to CIN85 that has also been found to bind to ASAP1 and has been proposed to recruit ASAP1 to the plasma membrane (39).

The SH3 domain of ASAP1 (PAG2) mediates binding to a proline-rich motif in the EH-domain-containing protein POB1 (67). Treatment of cells with EGF results in the phosphorylation of POB1 and its recruitment to the EGFR (68). POB1 simultaneously binds Ral-binding protein (RalBP), which contains a Rho GAP domain (67). The complex of ASAP1, POB1 and RalBP could regulate actin remodeling by controlling RhoA·GTP levels, thereby coordinating changes in the actin cytoskeleton with membrane traffic.

Amphiphysin IIm is a splice variant of amphiphysin II, which is a protein with BAR and SH3 domains. Amphiphysin II functions in synaptic vesicle endocytosis. The proline-rich domain of ASAP2 (Papα/PAG3/AMAP2) binds to the SH3 domain of amphiphysin IIm. Both overexpression and reduced expression (achieved using siRNA) of ASAP2 inhibit internalization of Tac, the interleukin-2 receptor α subunit, in HeLa cells (42).

Signaling proteins

Lipid-modifying enzymes

Three types of Arf GAPs – GITs, AGAPs and ARAP3– interact with enzymes that regulate signaling lipids. Phospholipase Cγ (PLCγ), an important element of the PIP2 to IP3 pathway, binds to the SHD of GIT1 (69). This interaction is necessary for PLCγ activation induced by angiotensin II (AngII) and EGF in vascular smooth muscle cells and 293 cells.

AGAP2 (PIKE/centaurin δ1) is another Arf GAP that binds to lipid metabolizing enzymes. AGAP2 has been reported to bind to PLC γ through its N-terminal proline-rich sequence (70). In one report, PLCγ was found to function as an exchange factor for nucleotide on the GLD domain of AGAP2 (70), although recent kinetic analyses indicate that the affinity for nucleotide of AGAPs is such that basal exchange rates are extremely rapid, obviating the need for an exchange factor (71). One form of AGAP (PIKE) binds to and activates phosphatidylinositol (PI) 3-kinase (PI3K) (72), preventing neuronal apoptosis. The interaction with PI3K is thought to be important for the growth and invasion of glioblastoma cells (73).

ARAP3, through its SAM domain, has recently been reported to bind the PI 5-phosphatase, SHIP2 (74). The physiological consequences of this interaction have not been defined; however, given ARAP3 is recruited to the plasma membrane with PI(3,4,5)P3 and its Arf GAP activity is regulated by PI(3,4,5)P3, a plausible model is that ARAP3 negatively regulates PI3K signaling by recruiting SHIP2 to the membrane creating a negative feedback loop.

Protein kinases

Protein kinases have been found to function upstream and downstream of Arf GAPs. The interaction between GIT and p21-activated kinase (PAK) (32) is one of the best studied. It is indirect. PIX, a Rac/Cdc42 exchange factor, acts as a bridge between GIT and PAK. Interfering with the association between GIT and PAK has been found to disrupt cell motility and function of FAs (75,76). The effects of GIT on actin are thought to be mediated, in part, by the action of PAK. GITs have also been implicated as regulators of serine/threonine kinases within signaling pathways. GIT1, interacts with MEK1, extracellular signal-regulated kinase kinase. The interaction is constitutive, but GIT1 stimulates MEK activity in response to AngII and EGF activation of Src-dependent phosphorylation of GIT (77).

GIT is an upstream regulator of PAK and MEK but functions downstream of another serine/threonine kinase, GPCR kinase2 (GRK2). Both GIT1 and GIT2, excluding the splice variant GIT2-short, bind to GRK2 (78,79). GPCR kinase2 is recruited to GPCR in response to an agonist. GITs are recruited by GRK2 into this complex and mediate subsequent internalization of the G-protein-coupled receptor. This function of GIT depends on GIT Arf GAP activity. Point mutants of GIT that are deficient in GAP activity block internalization.

AGAP2 (PIKE/centaurin γ1/GGAP2) has been reported to interact with two classes of kinases. Src family proteins phosphorylate AGAP2 on two tyrosines (80). The phosphorylation prevents apoptotic cleavage of AGAP2 during programmed cell death, which is consistent with the proposed role of AGAP2 in the anti-apoptotic signaling pathway. One signal in this pathway that is downstream of AGAP2 is PI3K. AGAP2 also binds to and activates a serine/threonine kinase that is critical to this pathway, Akt (81,82). The AGAP2/Akt complex has been reported to be dependent on GTP, although given the lack of nucleotide specificity and low affinity recently reported in kinetic studies (71); the details of the molecular mechanisms regulating the complex are not obvious and need further examination.

ASAP1 binds to and functions with two protein tyrosine kinases, Src and focal adhesion kinase (FAK), including the FAK homolog Pyk2 (35,83,84). Expression of constitutively active Src results in the phosphorylation of ASAP1. The SH3 domain of Src and Src family proteins bind to a PXXP motif in ASAP1. Focal adhesion kinase binds to the SH3 domain of ASAP1. The first evidence for a role of FAK in the cellular function of ASAP1 came from studies examining FAs (83). Disruption of the interaction resulted in partial dissociation of a FA protein called paxillin from FAs and changes in cell motility. Deletion of the SH3 domain of ASAP1 also disrupted the formation of podosomes consistent with direct binding of FAK to ASAP1 having a role in this process; however, another protein partner cannot be excluded on the basis of available data. The phosphorylation of ASAP1, consequent to expression of activated Src, is necessary for the formation of podosomes and invadopodia , which are actin-rich structures on the ventral surface of cells that mediate adhesion, ECM degradation and tumor invasion. Both Src and Pyk2 have been found to directly phosphorylate ASAP1, which inhibits the GAP activity for Arf1 (84). A relationship between the effect of phosphorylation on GAP activity and podosome formation has not been examined to the best of our knowledge.

Small G proteins

Members of other G-protein families, in addition to Arf, function with Arf GAPs to regulate remodeling of the actin cytoskeleton. ARAP family Arf GAPs function with Rho and Rap family proteins in at least two capacities to regulate actin and actin-associated structures (49,52,85). ARAP1 and ARAP3 contain active Rho GAP domains that use RhoA as a substrate in preference to Rac1 and Cdc42. Effects of ARAP1 and ARAP3 on the actin cytoskeleton are dependent on the Rho GAP activity, presumably functioning to inactivate RhoA. ARAP2 has an inactive Rho GAP domain, consequent to lack of the catalytic arginine found in all other Rho GAPs; however, ARAP2 does bind RhoA·GTP through its inactive Rho GAP domain. RhoA binding to ARAP2, as well as ARAP2 Arf GAP activity, is required for FA and stress fiber formation in U118 glioblastoma cells, consistent with ARAP2 function as a RhoA effector (50).

The ARAPs also contain an RA domain immediately C-terminal of the Rho GAP domain. Rap1 binds to the RA domain of ARAP3 and regulates the Rho GAP activity. Regulation of peripheral actin ruffles that are regulated by ARAP3 depend on the ability of ARAP3 to bind to Rap1 (52,85).

Actin cytoskeleton

Adhesion and scaffold molecules

The GITs and the ASAPs have been found to regulate FAs and ASAPs have also been implicated in the regulation of invadopodia and podosomes. Consistent with this function, interacting proteins that are components of these adhesive structures have been identified. Three Arf GAPs, GIT1, GIT2 and ASAP2 (PAG3/AMAP2), have been reported to interact with paxillin and related proteins, hic-5 and leupaxin (41,79,86–88). Paxillin is an adaptor protein often used as a marker of FAs that functions in the transduction of signals mediated through growth factor receptors and integrins. Paxillin regulates FA dynamics and, as a consequence, cell migration, spreading and adhesion. The LD4 motif in paxillin binds to the SHD of GIT. Interfering with GIT association with paxillin results in altered FAs and accelerated cell migration.

Crk and CrkL are adaptor proteins that may contribute to ASAP1 function in FAs and membrane ruffles (35,36). CrkL is SH2- and SH3-domain-containing protein that binds to paxillin and Cas at FAs. Two PXXP motifs in ASAP1 mediate binding to Crk and CrkL. The interaction was found to be necessary for ASAP1 association with FAs (35).

ASAP1 (AMAP1/PAG2) has also been reported to associate with cortactin, a multidomain protein with N-terminal acidic, actin-binding, proline-rich and SH3 domains. Cortactin is found in peripheral membrane ruffles and invadopodia. The SH3 domain of cortactin was found to bind to a proline-rich motif specific to one splice variant of ASAP1 (34). Subsequent work showed that the SH3 domain of ASAP1 had a more dominant role in forming the complex than the splice-variant specific proline-rich motif. Both ASAP1 and cortactin associate with invadopodia in invasive breast cancer cell lines. Furthermore, disruption of the ability of ASAP1 to form a complex with cortactin interferes with the formation of invadopodia and structurally related podosomes (34). ASAP1 expression correlates with the invasive potential of uveal melanoma cells and mammary carcinoma cells. Based on these findings, the ASAP1/cortactin complex has been proposed to function as a part of an invasive machinery in cancer cells. In a recently described model, the ASAP1/cortactin complex links the highly tubulated membranes found in invadopodia and podosomes to polymerized and branched actin (89).

Miscellaneous proteins

Arf GAPs have been found to bind to a number of proteins that are not clearly associated with changes in actin or membranes. ACAP1 (centaurin β1) and ACAP2 (centaurin β2) have been reported to interact with bacteria-derived intracellular peptidoglycan sensor proteins, NOD1 and NOD2, and vaccinia virus protein KILT, respectively (90,91). AGAP1 associates with nitric oxide (NO)-sensitive soluble guanylyl cyclase through N-terminal GLD region (92).

Arf GAPs as regulated signaling platforms

Although the roles of Arf GAPs as adaptors or scaffolds for signaling proteins and as elements of signaling pathways have been described for GITs, ARAPs, AGAPs (PIKE) and ASAPs, the mechanisms by which the distinct activities of each domain of Arf GAPs may be integrated have not been elucidated. Recent findings in studies of ASAP1 (34,47) provide a basis to speculate about mechanisms by which integration to control a particular cellular activity is achieved. We propose a model in which ASAP1 functions as a regulated signaling platform to control the dynamics of invadopodia and podosomes. These structures are labyrinths of tubulated membranes associated with polymerized actin (8). The BAR, PH and Arf GAP domain of ASAP1 bind to PIP2 and Arf·GTP to induce membrane tubulation (47). In this way, ASAP1 could be considered an Arf effector. ASAP1 must be phosphorylated on tyrosine, dependent on the nonreceptor tyrosine kinase Src, to function at podosomes. Thus, ASAP1 integrates three signals – Arf·GTP, PIP2 and Src. ASAP1 associates with the tubulated membranes and may undergo conformational changes leading to additional protein–protein interactions (47). ASAP1 binds cortactin, which induces actin polymerization though interaction with Arp2/3 and binds to filamentous actin (34). ASAP1 would thus link polymerized actin to the tubulated membranes. ASAP1 binds FAK (83) and Src (36), tyrosine kinases important to the formation and maintenance of invadopodia and podosomes. The adaptor protein Crk bound to ASAP1 has the potential of bringing other signaling proteins into the complex (35). The association with POB1/RalBP could control Rho·GTP levels (67), also important for maintenance of invadopodia. The ASAP1-dependent complex could be rapidly controlled: GAP activity of ASAP1 is robust and, maintenance of the complex on the tubulated membranes would depend on continued generation of Arf·GTP. At other sites, with different signals, ASAP1 could bind a different group of proteins, resulting in a different output. For instance, ASAP1 in FAs does not bind cortactin. Thus, rather than a simple scaffold or adaptor, ASAP1 would function as a multiplexer, providing a unique output for a particular set of inputs. This model may generalize to other multidomain Arf GAPs, such as ARAPs or AGAPs. The idea that Arf GAPs could function as Arf effectors could extend to simpler Arf GAPs, as suggested for yeast Arf GAPs (93).

Conclusions and perspective

Arf GAPs are structurally complex proteins. Each has GAP activity, inducing hydrolysis of GTP bound to Arf. Some of the functions of Arf GAPs in the regulation of actin and membranes are attributable to the GAP activity. However, Arf GAPs affect actin and membranes through additional mechanisms. In this review, we cataloged GAP-associated proteins and described potential mechanisms by which they may contribute to the effects of Arf GAPs on actin or membranes. The Arf GAPs also directly interact with lipids, described in Nie and Randazzo (26), to affect the structure of biological membranes. One challenge at this time is to understand how protein and lipid association with Arf GAPs and GAP activity are integrated to coordinate changes in the actin cytoskeleton and membranes necessary for biological behaviors such as cell migration and pathological behaviors such as cancer cell invasion. We speculate in this review about mechanisms of integration for ASAP1, suggesting that ASAP1 functions as a regulated signaling platform similar to a multiplexer. Ongoing structural studies focused on the functional relationships between domains within single Arf GAPs will be important for understanding the molecular basis of integration of the distinct activities associated with specific Arf GAPs.

Acknowledgments

This work was supported by the Intramural Research Program of the National Cancer Institute, Department of Health and Human Services. We apologize to authors whose work may have been omitted due to restrictions in the length of the review, or Paul Randazzo’s or Hiroki Inoue’s oversight.

Ancillary