|Home | About | Journals | Submit | Contact Us | Français|
Arf GAP proteins are a versatile and diverse group of proteins. They control the activity of the GTP-binding proteins of the ARF family by inducing the hydrolysis of GTP that is bound to Arf proteins. The best-studied role of Arf GAPs is in intracellular traffic. In this review, we will focus mainly on the Arf GAPs that play a role in vesicle formation, ARFGAP1, ARFGAP2 and ARFGAP3 and their yeast homologues, Gcs1p and Glo3p. We discuss the roles of Arf GAPs as regulators and effectors for Arf GTP-binding proteins.
GTP-binding proteins of the ARF/SAR family are key regulators of intracellular traffic. These GTP-binding proteins are primarily controlled by 2 classes of factors: guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs). The GEFs exchange protein-bound GDP for GTP to produce the GTP liganded also called activated ARF/SAR proteins. In the activated form, the ARF/SAR proteins are membrane-associated and interact with cargo, SNARE and coat proteins. The cumulative interactions will then at one point lead to membrane deformation and transport vesicle formation. Since proteins of the ARF family do not have detectable intrinsic GTPase activity, the complex of the Arf GAP with Arf1 could be viewed as the functional GTPase, which essentially would consist of a heterodimer. The functional heterodimer is probably not very stable and may dissociate right after catalysis. Taken together, this indicates that specificity for Arf GTP-binding proteins is mostly dominated by Arf GAPs. Moreover, most Arf effector molecules interact with the GTP-bound form of Arf. Thus Arf GAPs may play a critical role in terminating Arf-effector protein interactions. Therefore it is and will remain important to understand how Arf GAPs function. Despite the good conservation of the GAP domain, the Arf GAPs share only little sequence conservation apart form the GAP domain. Thus efforts have to be undertaken in the future to elucidate the function of the other regions of Arf GAPs.
Arf GAPs were discovered less than 20 years ago, when Randazzo and Kahn  demonstrated the existence of a protein in bovine brain extracts that could strongly enhance the intrinsic GTPase activity of mammalian Arf1. A couple of month later the first Arf GAP had been purified about 15,000-fold and identified as a protein of 49 kDa . Arf GAP1 was the first Arf1-directed GAP to be cloned [3,4]. The minimal fragment of Arf GAP1 with catalytic activity was 130 amino acids in length and contained a zinc-binding motif. Interestingly, ArfGAP was stimulated by PIP2 [1,2], opening the possibility that PIP2 was a cofactor in one or more Arf1-dependent pathways. Soon afterwards, the first Arf GAP was identified in yeast, Gcs1p, which is the homologue of mammalian ArfGAP1 . Gcs1p was first identified as a protein required for entry from stationary phase into the cell cycle at low temperatures. This function is not understood to date and it remains unclear whether this is an Arf1p-dependent process . Gcs1p together with another Arf GAP, Glo3p, is essential for retrograde transport from the Golgi to the ER, demonstrating its importance in the regulation of Arf-dependent functions . Since then Arf GAPs have been identified in a large variety of eukaryotes and it is generally assumed that lmark of Arf GAPs is the GAP domain. The structure of two Arf GAP domains were solved [8,9] providing important insights into the catalytic mechanism. Interestingly, by now a number of Arf GAP domain containing proteins have been identified that in addition contain regulatory signatures for other small GTP-binding proteins, especially of the Rho family, indicating that regulatory hubs exists on which GTPase function and signaling converges.
In mammalian cells, Arf GAPs were first identified as regulators of Arf GTP-binding proteins. Despite the prevailing conception of Arfs as small GTPases, Arf proteins do not have detectable intrinsic GTPase activity and bind tightly to GTP [1,10,11]. Thus, the conversion of Arf from the GTP to GDP bound form requires a protein that induces the hydrolysis of GTP bound to Arf, called a GTPase-activating protein (GAP) [1,12].
Thirty-one genes encoding proteins with Arf GAP catalytic domains have been identified in humans . Ten subfamilies have been described based on domain structure and phylogenetic analysis (Figure 1). Six subfamilies have Arf GAP domains at the extreme N-terminus. Four subfamilies (ASAP, ACAP, AGAP and ARAP), comprising 20 proteins, have a catalytic core of PH, Arf GAP and ANK repeat domains. At least one member of 8 subtypes has confirmed GAP activity. However, 2 subtypes do not have detectable GAP activity and for other Arf GAPs, the GAP activity is dispensable for some cellular functions, leading to the hypothesis that the Arf GAPs function in capacities in addition to the negative regulation of Arf proteins [14-16]. One function may be as Arf effectors.
A number of Arf GAPs have been implicated in membrane traffic including SMAPs, ARAP1, AGAP1/2, Arf GAP1 and ARF GAP2/3. Of these, the function of Arf GAP1 in membrane traffic has been examined in the context of models of Arf-dependent membrane traffic [17-19]. In the prevailing model, the cycle of GTP binding to Arf1 and hydrolysis while bound to Arf1 is thought to be linked to coat association with and dissociation from membranes. Although numerous variations of this model have been articulated, in each the function of Arf1•GTP is essentially a glue that holds coat proteins on the membrane. GTP hydrolysis is required for the coat proteins to dissociate from the membranes. In this model, the function of Arf GAP1 is to induce hydrolysis of GTP bound to Arf1 to trigger coat dissociation and GTP hydrolysis must be delayed until after the coat protomers have assembled into a vesicle coat.
The appropriate timing of the hydrolysis of GTP on Arf1 requires the regulation of Arf GAP activity. Two regulatory mechanisms have been proposed for Arf GAP1. In one model, GAP activity is dependent on the coat protein coatomer and GAP activity is inhibited by cargo proteins [20-23]. The second and prevailing model invokes membrane curvature sensing. Coat protein polymerization results in budding from a membrane with eventual fission and a transport intermediate [24-28]. The bud and the transport intermediate have greater curvature than does the relatively flat surface from which the vesicle is supposed to bud. In this model, Arf GAP1 is able to sense membrane curvature through specific motifs called Arf GAP lipid packing sensor (ALPS) . The two ALPS motifs in Arf GAP1 are random coils in solution. When head groups of the lipid bilayer separate consequent to an increase in curvature during vesicle formation, the ALPS motifs associate with the exposed hydrophobic interior of the bilayer, forming an amphipathic helix. In this model, Arf GAP1 is recruited to the curved surface, not the flat, so its activity is restricted to the curved surface [24-27,29].
The model in which Arf GAP1 regulates Arf1 through curvature sensing is intriguing and has substantial experimental support [18,24,25] as does the model of Arf1 regulation of coat-membrane association [19,29-34]. However, the models have a number of shortcomings. First, although multiple Arf GAPs regulate membrane traffic, Arf GAP1 is the only one with ALPS domains; therefore the model cannot generalize to other Arf GAPs that function in the Golgi. Second, curvature sensing has not been easily reproduced even when using Arf GAPs with BAR domains [35,36], whose primary function is proposed to be membrane curvature sensing and induction [37-39]. For the Arf GAPs examined, the sensing or induction of curvature is highly dependent on specific in vitro experimental conditions [35,40]. Third, the role of Arf GAP as triggering coat dissociation implies a reversible process in an otherwise vectorial system. The conversion of Arf•GDP to Arf1•GTP causes coat association with membranes and the conversion of Arf1•GTP to Arf1•GDP causes coat to dissociate. Thus premature GTP hydrolysis would result in futile cycles and not in productive vesicle formation. Fourth, the models do not account for cargo sorting .
In addition to the shortcomings, 4 lines of investigation have yielded results that are either inconsistent with or difficult to reconcile with prevailing models of the function of Arf and Arf GAP in membrane traffic. The enzymology of Arf GAP1, Arf GAP2 and Arf GAP3 has recently been examined. The curvature sensitivity of Arf GAP1 reported by some laboratories has not been consistently reproduced by others , which has motivated the examination of other regulatory factors [40,41]. Arf GAP1, 2 and 3 bind to the vesicle coat protein coatomer [22,42-44]. More than a decade ago coatomer was reported to stimulate GAP activity . Until recently, the effect was not further studied other than a single paper that failed to reproduce the effect . As described in the original report Arf GAP1 was stimulated by coatomer . The effect of coatomer was more easily documented with Arf GAP2 and 3, which have lower intrinsic GAP activities and greater affinity for coatomer than Arf GAP1 [41,43,44]. The increased activity for Arf GAP1 and Arf GAP2 was consequent to a decrease in Km (Arf GAP3 was not examined) [40,41]. In these experiments, coatomer with Arf1•GTP was not sufficient to induce changes in the liposomes or the formation of vesicles. The authors concluded that interaction with coatomer, not membrane curvature, regulated Arf GAP1 and Arf GAP2 activity. Experiments in which GAP activity was measured at a single concentration of Arf and coatomer  could have missed the effect because (i) the reaction was already close to saturation, so a Km effect would not be detected and (ii) at concentrations of coatomer in excess of Arf GAP and Arf1•GTP, coatomer binds and sequesters Arf1•GTP from the GAP.
Cargo proteins bind to coatomer and to Arf GAP, and potentially could affect GAP activity by binding either protein. The effect of peptides from the cytoplasmic tail of p24 cargo proteins on Arf GAP1 and 2 was examined. Early reports indicated that the peptides inhibited activity [20,21]; however the inhibitory effect was nonspecific and coatomer independent . When peptides from several p24 family members were examined, those from p23 and p25 were found to increase GAP activity of Arf GAP1 and Arf GAP2 by increasing the kcat.These results are different than predicted by the current model in which activation of Arf GAP occurs after vesicle assembly. The cargo/coatomer complex, which activates the GAP, assembles into the vesicle coat, so hydrolysis of GTP bound to Arf likely occurs prior to or coincident with assembly of the coat protein.
A second line of investigation with results inconsistent with prevailing views of Arf function was the examination of AP-1 recruitment to the trans-Golgi network [46-48]. Arf1•GTP recruited AP-1 to a high affinity site. Following hydrolysis of GTP, Arf1 dissociated from the TGN but AP-1 remained associated with TGN at a low affinity site. The conversion from the high affinity site to the low affinity site required a cargo protein, mannose-6-phosphate receptor. Cargo, independent of Arf, drove AP-1 polymerization into a vesicle coat . In addition, the coated vesicles from the TGN contained little or no Arf1 [47,48]. Taken together these results led to the suggestion that Arf functions in the coat assembly process but not vesicle uncoating.
A third line of investigation is examination of the role of GTP hydrolysis and Arf GAP1 in the generation of transport vesicles. In the current model in which Arf GAP activity is controlled by curvature and independent of other proteins blocking GTP hydrolysis should result in accumulation of coated vesicles containing cargo. In two reports, Golgi membranes were incubated with components necessary for transport vesicle formation and either GTP or the slowly hydrolysable analog GTPγS. Although vesicle formation was detected with both forms of GTP, the vesicles formed in the presence of GTPγS had less cargo [49,50]. Blocking GTP hydrolysis by Arf in cells, either by injecting GTPγS or by expression of the GTP-locked [Q71L]Arf1, resulted in reduced cargo in COPI coated vesicles, which was the basis for the suggestion that the function of Arf was facilitating coat-cargo interaction . In vitro formation of cargo laden vesicles from Golgi membranes has been reported to require GTP hydrolysis provided specifically by Arf GAP1. Arf GAP1 also had a GAP-independent function of increasing the affinity of coatomer for cargo . The in vitro and in vivo results together support the idea that GTP hydrolysis, a GAP driven event, is involved in cargo sorting, which is process upstream of curvature sensing.
A fourth line of investigation was the examination of cells with reduced Arf GAP expression . Reduction of Arf GAP1, 2 or 3 alone had no discernible phenotype. However, if expression of all three was reduced, protein usually resident in the cis-Golgi accumulated in the ER-Golgi intermediate compartment. In addition, Golgi to ER retrograde traffic was blocked. Vacuolar structures were visualized by electron microscopy. The phenotype was similar to that seen with reduction of coat protein, which is considered an Arf effector. These results are inconsistent with the prevailing model, which predicts the accumulation of coatomer-coated vesicles on loss of Arf GAP activity. Taken together with the enzymology, examination of AP-1 recruitment and examination of the role of GTP hydrolysis for vesicle formation, these results support a model in which Arf and Arf GAP function in vesicle assembly with Arf GAPs having an effector role (Figure 2). In this model, the function of Arf is to facilitate coat-cargo association as previously suggested .
The notion that Arf GAPs are not only regulators but also effectors of Arf proteins is also supported by results from yeast. Genetic studies support a role of Arf GAPs as downstream effectors of Arf proteins. If Arf GAPs functioned purely as negative regulators of Arf, overexpression of the GAPs should either have no effect or exacerbate Arf insufficiency. Instead, four Arf GAPs (Gcs1p, Glo3p, Age1p and Age2p) expressed from a high copy plasmid suppressed a loss-of-function allele of Arf1 [53,54]. This result is consistent with the Arf GAPs functioning as Arf effectors. Subsequently studies have proved information about the possible molecular basis for the Arf effector function of yeast Arf GAPs. Both the yeast ArfGAP1 Gcs1p and the ArfGAP2/3 homologue Glo3p interact with SNARE proteins and induce a conformational change that allows recruitment of Arf1p and coatomer to the SNAREs [55-57]. This interaction does not make them effectors per se, but in light of vesicle biogenesis and primer formation, they have the qualities of an effector. It is thought that in order to form a transport vesicle, the need for such a vesicle is sensed by the formation of primer complexes  (Figure 2). Such primer would consist of a SNARE molecule a small GTP-binding protein of the ARF/SAR family and coat components [19,34]. Other cargo proteins could also participate in primer formation [58,59]. Perhaps SNARE interaction with Arf GAPs is a special case, because it could be viewed as starting point for vesicle formation and thus ensuring the inclusion of the correct set of SNAREs in each and every vesicle. At least for Glo3p it is clear that the interaction with coat proteins and cargo is vital for its function. Glo3p, which is an integral part of COPI vesicles , interacts directly not only with SNARE proteins but also with proteins of the p24 family . These interactions take place in a region called the BoCCS (Binding of Coatomer, Cargo and SNAREs) . In addition to the GAP domain and the BoCCS region, Glo3p contains a regulatory motif, the Glo3p-regulatory motif (GRM) . These 3 motifs/domains communicate with each other to link stimulation of GTP hydrolysis with vesicles formation. We propose that the GRM senses the occupied BoCCS and communicates the binding to the GAP domain, which in turn would then stimulate GTP hydrolysis on Arf1p. Unfortunately the structural knowledge about ArfGAP1- 3 is limited. The crystal structure is of only the first 136 amino acids, which does not include the BoCCS or the GRM in the case of Arf GAP2/3 (or the ALPS domain in Arf GAP1). BoCCS is an unstructured region, which could become structured upon binding of coatomer, cargo and/or SNAREs. Interestingly ArfGAP1 proteins, like Gcs1p, also contain an unstructured region, but in this case it is assumed the region reorganizes in the presence of curved membranes and forms the ALPS domain, which is supposed to sense membrane curvature. Thus both Gcs1p and Glo3p must have evolved at least two different functions, and gained some specialization during development. Gcs1p and Glo3p have been found to have overlapping, but not redundant functions in retrograde transport in from the Golgi to the ER  Moreover, in contrast to Glo3p, Gcs1p does not interact directly with coat components and it has not been found on COPI-coated vesicles [56,60]. However, it cannot be ruled out that the ALPS domain, besides sensing curvature, could act as protein interaction site. Alternatively, a yet unidentified protein-interaction site is present in Gcs1p, which allows interactions with e.g. SNAREs. All the data from yeast suggest that Gcs1p and Glo3p can function in generation of COPI-coated vesicles. This poses the question whether either of them is required for vesicle formation or whether they act sequentially in vesicle biogenesis. These possibilities are difficult distinguish experimentally, because similarly to mammalian cells, the loss of one class of GAPs is well tolerated by the cell and only deletion of both GCS1 and GLO3 is lethal. Under standard laboratory conditions, a lower efficiency of transport could be easily tolerated. Perhaps conditions that would challenge the cells in a way that highly efficient transport becomes vital, would allow us to determine the specific roles of Gcs1p and Glo3p in COPI-vesicle generation. It is notable, however, that yeast cells have at least six Arf GAPs, none of which is essential. Next to Gcs1p there is another mammalian ARFGAP1 homologue, Sps18p, which is sporulation specific. Age1p and Age2p are related to centaurin/ADAPtype Arf GAPs and Gts1p seems to be a bit of an outlier, and its function is poorly understood. From the genetic interactions so far tested, only •gcs1 •age2 and •gcs1 •glo3 are lethal. They seem to have overlapping functions in exit of the trans-Golgi network . Hence Gcs1p could play a more general role in vesicle formation. On the other hand, not all Arf1p-dependent vesicle budding events require Gcs1p function.
Although the work on Arf GAPs in yeast and mammalian cells developed in the large part independently of one another, the results from both have led to an appreciation of the diversity of Arf GAP function. It will be important in the future to get more insight at which point in the vesicle biogenesis pathway ArfGAP1 and ArfGAP2/3 function. We also need structures of the ArfGAP1 and ArfGAP2/3 to get a better mechanistic insight in their function, because despite the importance of the GAP domain, the variation in the remainder of the proteins (which is the by far the biggest bit) provides the regulatory circuit and feedback loop to precisely time GTP hydrolysis by Arf proteins. The structure determination will be quite a challenge as probably the most interesting parts are unstructured in solution.
This work was supported by the University of Basel and the Swiss National Science Foundation (A. S.) and by the intramural program of the National Institutes of Health, USA (P.A.R.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.