Arfs are a family of Ras-related GTP-binding proteins that function in the regulation of membrane trafficking and structure. The six mammalian Arf proteins are expressed ubiquitously and so it is anticipated that each will have a distinct localization and function within the cell. It has been assumed that much of this specificity will be defined by determining which regulators of Arfs, the GEFs (guanine nucleotide-exchange factors) and GAPs (GTPase-activating proteins) function with which Arf proteins. Although in vitro assays may indicate Arf preferences for the numerous Arf GEFs and GAPs that have been identified, in the cell the different Arfs, GEFs and GAPs are targeted to specific compartments where they carry out their functions. We have embarked on studies to define regions of the Arf1 and Arf6 proteins that determine their sites of action and specific activities at the Golgi and plasma membrane respectively. Chimaeras were made between Arf1 and Arf6 in order to identify regions of the protein that contributed to targeting and function. Whereas Arf6 is targeted to the plasma membrane through multiple regions along the protein, we have found a Golgi-targeting region in Arf1 that is sufficient to target Arf6 to the Golgi complex.
- coat protein
The Arf proteins comprise a conserved family of proteins in the Ras superfamily of low molecular mass GTP-binding proteins. Arf proteins act to regulate membrane traffic and organelle structure [1,2]. The mammalian Arf proteins have been divided into three classes based on amino acid composition (Figure 1). Human Arf 1 and Arf 3 belong to class I, human Arf 4 and Arf 5 belong to class II and Arf6 is the sole class III member. Metazoans contain at least one member from each class, whereas yeasts lack a class II member. Also shown in Figure 1 are the Drosophila melanogaster homologues for each class to demonstrate the remarkable conservation among distinct residues that make up the extreme N-terminus and C-terminal half of the protein in each class. Arfs are myristoylated at the N-terminus, a co-translational modification that is essential for membrane association and biological activity.
Arf proteins undergo a cycle of GTP-binding and hydrolysis to form the GTP-bound active and GDP-bound inactive forms of the protein. Many GEF (guanine nucleotide- exchange factors) have been identified that catalyse exchange of GTP for GDP on Arfs . An even larger number of GAPs (GTPase-activating proteins) have been found to catalyse GTP hydrolysis, important regulators since the Arfs have negligible intrinsic GTPase activity . For both the GEFs and GAPs, biochemical assays have been used to assign Arf specificity or preference for these enzymes, but in cells these regulators localize to specific membrane compartments where they presumably encounter specific Arf proteins.
Shared effectors but distinct localizations for Arf1 and Arf6
Upon activation, Arf proteins modify membrane surfaces by recruiting coat proteins and activating lipid-modifying enzymes. Cytosolic coat protein recruitment on to membranes is the most studied and best understood activity of Arfs. Coat proteins whose membrane association is dependent upon Arf-GTP include the COPI (coat protein complex I) coat, the clathrin AP1 (adaptor protein 1), AP3 and AP4, and the monomeric GGAs (Golgi-associated α-adaptin homology Arf-binding proteins) [5,6]. These coat proteins sort membrane protein cargo into transport structures. All Arfs can activate PLD (phospholipase D) [7,8] and PIP5K (phosphatidylinositol 4-phosphate 5-kinase)  to generate phosphatidic acid and PIP2 (phosphotidylinositol 4,5-bisphosphate) respectively. The formation of these acidic phospholipids phosphatidic acid and PIP2 in the membrane can lead to changes in membrane curvature and recruitment of proteins to these regions.
All Arf proteins have nearly the same effector domain regions, Switch I and Switch II (Figure 1). Indeed, in biochemical assays, all of the Arf proteins can, to varying extents: (i) recruit coat proteins to Golgi membranes , (ii) bind to GGA , (iii) activate PLD , and (iv) activate PIP5K . For each of these activities, a preference for a particular Arf may be indicated; however, due to the variable potency of recombinant Arf preparations and the conditions of the biochemical assays, one cannot be sure as to whether one particular Arf is better suited for an activity than another. In the cell, on the other hand, Arfs are targeted to particular membranes where they function.
Arf1 and Arf6 are the most studied mammalian Arf proteins whose primary localizations in cells are at the Golgi and PM (plasma membrane) respectively (Figure 2). In cells, Arf1 reversibly associates with Golgi membranes during its GTP cycle, Arf1-GDP being cytosolic and Arf1-GTP bound to the membrane. Arf1-GTP recruits COPI on to pre- and cis-Golgi structures, and AP1, AP3 and AP4, and GGA on to TGN (trans-Golgi network) and endosomal membranes . In all the cases that have been examined, Arf3 functions identically with Arf1, not surprising since these two Arfs differ in only seven amino acid residues clustered at the N- and C-termini (Figure 1). There is also some evidence that Arf5, a class II Arf, can function at the early Golgi [13,14] and at the trans-Golgi to affect GGA association . The issue of whether Arf5 or Arf4 is normally functioning at the Golgi is not known since immunological reagents that can distinguish among the Arf members are lacking; thus we do not have immunolocalization information of the endogenous Arf proteins.
Arf6, on the other hand, is neither localized to nor does it appear to affect any Arf-associated activities at the Golgi complex . Arf6 is present at the PM and to some extent on endosomal membranes where it regulates the flow of trafficking into and out of the cell and the actin cytoskeleton at the PM (Figure 2) . Arf6 co-localizes with at least one isoform of PIP5K and many of the activities of Arf6 can be attributed to its stimulation of PIP5K and generation of PIP2 [9,18,19]. PIP2 is primarily localized to the PM where it recruits a range of types of proteins and the Arp2/Arp3-actin polymerization machinery . Additionally, the Arf6 stimulation of PLD is PIP2-dependent and is critical for many Arf6-trafficking functions [21,22]. In contrast, Arf6-GTP does not appear to directly assemble any identified COP on the PM, although its influence on PIP2 levels is important for AP2 and clathrin assembly .
These two Arfs, Arf1 and Arf6, represent distinct contrasts in localizations, Golgi versus PM, and in function, coat recruitment versus PIP2 formation/actin polymerization. We have been interested in identifying amino acids in Arf1 and Arf6 that confer specific membrane targeting and Arf function. The structures of Arf1 and Arf6, determined by crystallography, are rather similar [24,25]. This allowed us to design chimaeras between Arf1 and Arf6 to examine which regions of the Arf proteins were responsible for localization and function.
Identification of a Golgi-targeting motif in Arf1
If you divide Arf proteins roughly into half, beginning at amino acid 101 in Arf1 (NEAR…) and 97 in Arf6 (DEAR….) (see Figure 1), it is clear that the C-terminal half of the proteins are more diverse than the N-terminal half. We originally made chimaeras at this junction, which is the start of helix α-3 in the Arf structure. We found that Arf-1-6 localized to the PM , whereas Arf-6-1 localized not only to the PM but also to the Golgi complex . This suggested that the C-terminal half of Arf1 was required for the chimaera to be recruited to the Golgi, whereas Arf6 had PM-targeting information in both N- and C-terminal halves of the protein.
We made additional chimaeras and narrowed down a minimal Golgi-targeting motif to a 16 amino acid sequence that included helix α-3 (Figure 1, underlined sequence). Insertion of that 16 amino acid Arf1 sequence into Arf6 caused the chimaera Arf-6-1-6 to localize to the Golgi . Further mutation within the helix α-3 sequence revealed that MXXE (Figure 1, arrows) was critical for Arf-6-1-6 localization at the Golgi. This MXXE motif is conserved in all Arf proteins that have been shown to localize to the Golgi and absent from all Arf6 homologues.
We found that a GFP (green fluorescent protein) fusion protein containing only the 16 amino acid sequence could also localize, albeit inefficiently, to the Golgi . This led us to attempt to isolate a Golgi-associated Arf1 receptor that might recognize the Arf1 helix α-3 sequences. Although affinity approaches were not successful, we considered the possibility that SNARE (soluble N-ethylmaleimide-sensitive fusion protein attachment protein receptor) proteins in the ER (endoplasmic reticulum)–Golgi system could act as receptors for Arf1-GDP. Spang and co-workers showed using yeast proteins that Arf1-GDP could bind to ER–Golgi SNARE proteins Sec22, Bet1, and Bos1 . We found that Arf1 could bind in vitro to membrin, the mammalian homologue of Bos1, and in cells, Arf1 tightly co-localized with expressed membrin . These associations of Arf1 and membrin in vitro and in cells were dependent upon the helix α-3 and MXXE 
As shown in Figure 2, Arf1 can act at multiple locations within the ER–Golgi system. Membrin, however, localizes to the early Golgi including the ERGIC (ER–Golgi intermediate compartment) and cis-Golgi  and we think membrin is recruiting Arf1 to these regions of the Golgi. Membrin and its cognate SNARE partners, Sec22, Bet1 and syntaxin 5, are believed to facilitate homotypic fusion of ER-derived COPII vesicles [30,31]. It is possible that following this fusion event, or at the ERGIC, membrin becomes available for binding Arf1-GDP. Also present on the ERGIC and early Golgi is GBF1, an Arf GEF . In this way, COPII secretory vesicles leaving the ER can subsequently acquire the COPI coat. The use of SNAREs as both components that facilitate fusion and as receptors for proteins, like Arf, that regulate transport is an intriguing idea. We observed that following activation, Arf1-GTP no longer associates with membrin . Interestingly, a recent study finds that Arf1-GTP can bind to a different set of SNAREs, GS15 and ykt6, that have been implicated in retrograde COPI-mediated traffic . This hand-off of Arf1 from one set of SNAREs to another could also have interesting consequences for SNARE trafficking and recycling.
Although the helix α-3 16-amino acid sequence of Arf1 was sufficient to target Arf6 to the Golgi, it was not entirely necessary since an Arf1 protein that contained the 16 amino acids from Arf6 could still localize to the Golgi . This second Golgi-targeting motif thus appeared to require sequences in both the N- and C-terminal ends of Arf1. This Arf (1-6-1) localized more to the trans-Golgi and TGN than did wild-type Arf1 . We do not know how Arf1 lacking helix α-3 is targeted to the TGN, but there it co-localizes with BIG1, an Arf GEF that has been localized to the TGN and associated endosomal systems .
These two modes of targeting of Arf1 to the Golgi allowed us to examine the kinetics of association of Arf1–GFP to the early Golgi and compare it with the association of Arf-1-6-1 to the TGN. Remarkably, we observed a much faster kinetic cycle of Arf-1-6-1–GFP association at the TGN when compared with Arf1–GFP association at the early Golgi . These differences probably reflect differences of GEF and GAP activities at the two regions of the Golgi. Indeed, we know that GBF1 is at the early Golgi and BIG1 and BIG2 at the TGN. ArfGAP1 is active at least on the early Golgi and there may be additional Golgi-associated GAPs as well.
Arf6 PM targeting and function
In contrast with the discrete Golgi-targeting sequence identified in Arf1, Arf6 PM-targeting information is contained in both N- and C-terminal halves of Arf6 . We are investigating the possibility that Arf6 may bind to a SNARE at the PM. However, it is notable that all Arf6 homologues from all organisms are basic proteins (pI>8.0) that contrast with the neutral character of class I and class II Arfs. This characteristic is striking and suggests that Arf6 carries a high positive surface charge that may facilitate binding to membrane surfaces, such as the PM, that have a high negative charge. Indeed, Arf6 is associated with membranes that contain PIP2 .
In studying Arf6 function, we found that two amino acids, Gln37 and Ser38, adjacent to Switch I were essential for the ability of Arf6 to generate protrusions , and may be necessary for membrane-trafficking functions of Arf6. These two residues are conserved among all Arf6 homologues, with the exception of the Saccaromyces cerevisiae homologue, which also cannot generate protrusions when expressed in mammalian cells .
A full understanding of Arf function in cells will require knowledge as to which Arf is localized where in the cell. At those locations the Arfs will encounter specific GEFs and GAPs that have their own distinct modes of localization. The amino acid conservation among Arfs is striking, and distinct clusters of class-specific residues may indicate unique interactions with regulators and effectors. Teasing out these interactions will be a challenge but will reveal the full range of Arf functions in cells.
Localization and Activation of Ras-like GTPases: Focused Meeting held at the Royal Agricultural College, Cirencester, U.K., 21–23 March 2005. Organized and Edited by A. Ridley (Ludwig Institute of Cancer Research, London, U.K.) and M. Seabra (Imperial College London, U.K.).
Abbreviations: AP, adaptor protein; COP, coat protein complex; ER, endoplasmic reticulum; ERGIC, ER–Golgi intermediate compartment; GAP, GTPase-activating protein; GEF, guanine nucleotide-exchange factor; GFP, green fluorescent protein; GGA, Golgi-associated γ-adaptin homology Arf-binding protein; PIP2, phosphatidylinositol 4,5-bisphosphate; PIP5K, phosphatidylinositol 4-phosphate 5-kinase; PLD, phospholipase D; PM, plasma membrane; SNARE, soluble N-ethylmaleimide-sensitive fusion protein attachment protein receptor; TGN, trans-Golgi network
- © 2005 The Biochemical Society