The Arf1p GTPase-activating protein Glo3p executes its regulatory function through a conserved repeat motif at its C-terminus.

ADP-ribosylation factors (Arfs), key regulators of intracellular membrane traffic, are known to exert multiple roles in vesicular transport. We previously isolated eight temperature-sensitive (ts) mutants of the yeast ARF1 gene, which showed allele-specific defects in protein transport, and classified them into three groups of intragenic complementation. In this study, we show that the overexpression of Glo3p, one of the GTPase-activating proteins of Arf1p (ArfGAP), suppresses the ts growth of a particular group of the arf1 mutants (arf1-16 and arf1-17). Other ArfGAPs do not show such a suppression activity. All these ArfGAPs show sequence similarity in the ArfGAP catalytic domain, but are divergent in the rest of molecules. By domain swapping analysis of Glo3p and another ArfGAP, Gcs1p, we have shown that the non-catalytic C-terminal region of Glo3p is required for the suppression of the growth defect in the arf1 ts mutants. Interestingly, Glo3p and its homologues from other eukaryotes harbor a well-conserved repeated ISSxxxFG sequence near the C-terminus, which is not found in Gcs1p and its homologues. We name this region the Glo3 motif and present evidence that the motif is required for the function of Glo3p in vivo.


Introduction
ADP-ribosylation factors (Arfs) constitute a ubiquitous family of small GTPases in eukaryotes, and are now recognized as essential components in membrane traffic. There are six Arf proteins in mammals and three Arfs in the yeast Saccharomyces cerevisiae, which are classified into three and two classes, respectively. These Arf proteins regulate formation of vesicles in a variety of steps of membrane traffic by interacting with coat components and also appear to control phospholipid metabolism and cytoskeletal organization in various ways (Donaldson and Jackson, 2000;Kirchhausen, 2000;Nie et al., 2003;Schekman and Orci, 1996). Different classes of Arfs are thought to function at different places within a cell. Evidence is also present that a single Arf species may have multiple roles (Donaldson et al., 2005;Donaldson and Jackson, 2000). In yeast, among the three Arf proteins, Arf1p and Arf2p are 96% identical in amino acid sequences. They execute redundant essential functions for growth and their double knockout is lethal (Stearns et al., 1990), whereas Arf3p is dispensable.
We previously isolated eight temperature-sensitive (ts) mutants of the yeast ARF1 gene by PCR-based random mutagenesis in the arf2⌬ background. These arf1 ts mutants showed a variety of transport defects and morphological alterations in an allele-specific manner. Furthermore, intragenic complementation was observed between certain pairs of mutant alleles, both for cell growth and intracellular transport. These results demonstrated that the single Arf1 protein is indeed involved in many different steps of intracellular transport in vivo . However, the mechanism of managing these various steps of transport by Arf1p remained unsolved.
Our next interest was to understand how the single Arf1p can fulfill such divergent functions. Many Arf interactors are known to date, including guanine nucleotide exchange factors (GEFs), GTPase-activating proteins (GAPs) and effectors (Donaldson et al., 2005;Nie et al., 2003). ArfGEFs and ArfGAPs contain the Sec7 and the ArfGAP domains, respectively. In the human genome, at least 15 ArfGEFs and 16 ArfGAPs are present, and four ArfGEFs and four ArfGAPs exist in S. cerevisiae, indicating the presence of complicated networks for the regulation of Arfs. The yeast arf1 ts mutants we constructed may have different abilities to interact with such a variety of regulators and targets and thus express different downstream activities.
For the purpose of identifying the molecules that execute the pleiotropic functions downstream Arf1p, our arf1 ts mutants are very useful because the multiple roles of Arf1p can be dissected by different mutant alleles. One straightforward genetic approach is to screen for multicopy suppressors of these arf1 ts mutants. The eight arf1 ts mutants we constructed have been classified into three groups by intragenic complementation tests. arf1-16 and arf1-17 belong to the same group and show very similar phenotypes in protein transport. In both arf1-16 and arf1-17 alleles, glutamic acid at position 41 is mutated to valine (E41V), and arf1-16 has one additional mutation, D129E. These mutant cells have severe lesions in the Golgi-to-ER retrieval, not only by COPI-and Rer1p-dependent systems but also by the HDEL-receptor-mediated mechanism. They also show retardation of ER-to-Golgi transport of carboxypeptidase Y and secretion of underglycosylated invertase. Here, we report that the temperature sensitivity of arf1-16 and arf1-17 alleles is suppressed by the overexpression of Glo3p, but not by other ArfGAPs. We also describe the presence of a novel sequence motif in the C-terminal region of Glo3p, which is important for its suppression activity.

Results
GLO3 is a multicopy suppressor of arf1-16 and arf1-17 To identify genes that participate in multiple functions of Arf1p, we screened the S. cerevisiae genomic DNA library on a multicopy vector for plasmids that suppress the growth defect of arf1-16 at 35°C. In addition to authentic ARF1 and ARF2, several clones containing different genomic loci were obtained. Subcloning and sequencing analysis identified five genes, GLO3, BET1, BOS1, OLE1 and GYP6, as suppressors of arf1-16 ( Fig. 1). Among them, GLO3, encoding an ArfGAP, showed the strongest suppression activity on arf1-16. Growth of the arf1-17 allele, which has the same E41V mutation and is classified in the same intragenic complementation group with arf1-16, was also rescued by GLO3 (Fig. 1).
Of the ArfGAPs, only GLO3 can suppress arf1-16 and arf1-17 In S. cerevisiae, at least four genes (GLO3, GCS1, AGE1 and AGE2) are reported to encode ArfGAP (Poon et al., 1999;Poon et al., 2001b;Poon et al., 1996;Zhang et al., 2003). In contrast to GLO3, however, overexpression of GCS1, AGE1 or AGE2 did not suppress the ts growth of arf1-16 and arf1-17 at all. In the case of GCS1, its overexpression even exaggerated the growth defect of arf1-16 and arf1-17, but had no effect on the wild type (Fig. 1). The expression levels of GLO3 and GCS1 in arf1-16 and arf1-17 cells under these conditions were confirmed by immunoblotting of HA-tagged proteins (supplementary material Fig. S1). The temperature sensitivity of other arf1 ts mutants (arf1-11, -13, -14 and -18) was not suppressed by the overexpression of any of these ArfGAPs (supplementary material Fig. S2), indicating that the suppression was arf1-allele specific and GLO3 specific.
arf1-17p mutant protein shows a severe defect in GDP binding Because the ts growth of arf1-16 and arf1-17 was suppressed by the overexpression of one ArfGAP, we examined nucleotide-binding activities of these arf1 mutant proteins. arf1-16 or arf1-17 protein was co-expressed with Nmyristoyltransferase in E. coli, and purified by DEAE-Sepharose and gel-filtration chromatography as described in Materials and Methods. Some portions of arf1-16p and arf1-17p behaved as a much larger protein perhaps due to selfoligomerization. In the case of arf1-16p, most of the protein was eluted in fractions corresponding to a molecular mass of around 160 kDa (data not shown), therefore the purification procedure of wild-type (WT) Arf1p was not directly applicable to arf1-16p. The tendency for oligomerization was much smaller in the case of arf1-17p; 60-70% of arf1-17p was eluted in fractions of the same molecular size as WT-Arf1p (~20 kDa). So we decided to use arf1-17p for the biochemical analysis. Stoichiometric myristoylation was confirmed for both WT-Arf1p and arf1-17p. and arf1-17 (NYY17-1) cells were transformed with pNY24-GLO3 (2 GLO3), pNY24-GCS1 (2 GCS1), pNY24-AGE1 (2 AGE1), pNY24-AGE2 (2 AGE2), and the transformants were grown at 23°C. Cells were diluted to a final concentration of 1.0 at OD 600, and 10 l of 10-fold dilutions were spotted on MCD (minimal glucose casamino acid medium) (-Trp) plates. The plates were incubated at the indicated temperatures for 2 days. Integrants of wild-type ARF1 (WT, NYY0-1) and mutants transformed with pYO324 (vector) were used as controls. (B) arf1-16 and arf1-17 cells were transformed with pNY25-BET1 (2 BET1), pNY25-BOS1 (2 BOS1), pNY25-OLE1 (2 OLE1), and pNY25-GYP6 (2 GYP6), and the transformants were spotted on MVD (-Leu) plates and grown as described in A. Integrants of wild-type ARF1 (WT) and mutants transformed with pYO325 (vector) were used as controls.
The purified Arf1 proteins were then subjected to the analysis of guanine-nucleotide binding activities. Aliquots of Arf1p were mixed with [ 35 S]GTP␥S or [ 3 H]GDP and incubated at 23 or 30°C in the presence of dimyristoylphosphatidylcholine (DMPC)/cholate micelles as described previously (Kahn and Gilman, 1986). The resulting complex was trapped on a nitrocellulose filter and washed, and then the guanine nucleotide bound to Arf1p was quantified by scintillation counting. Fig. 2 shows the time course of GTP␥S-and GDP-binding to the WT and arf1-17 proteins. For WT-Arf1p, the binding kinetics was almost the same at 23 and 30°C. By contrast, arf1-17p showed a clear temperature sensitivity in nucleotide binding. The GDP binding of arf1-17p at 30°C plateaued during the first 90 minutes of incubation, the level of which was about twofold lower than the maximal value at 23°C. GTP␥S binding of arf1-17p was also lower at 30°C, but not as defective as GDP binding. These results suggest that arf1-17p has a less defective binding to GTP than to GDP at 30°C. If the ts defect of arf1-17 cells is related to such imbalance of GTP/GDP binding at high temperature, it may be that this defect is suppressed by the overproduction of ArfGAP.
Glo3p and Gcs1p show similar GAP activities toward arf1-17p in vitro The above finding led us to speculate that different affinity of Glo3p and Gcs1p to arf1-16p/17p would explain their different effects on the ts growth of the mutants. We tested the interaction between ArfGAPs and arf1-16p/17p using two methods, the two-hybrid analysis and the in vitro GAP assay. It has already been reported that an N-terminal-truncated form (⌬N17) of Arf1p with the Q71L mutation, which is stabilized Journal of Cell Science 119 (12) in the GTP-bound active form, interacts with both Glo3p and Gcs1p in a two-hybrid system (Eugster et al., 2000). We examined the interaction of arf1-16p/17p with these ArfGAPs using the same two-hybrid system. We generated two-hybrid constructs expressing N-truncated (⌬N17) versions of Arf1(Q71L)p, arf1-16(Q71L)p, or arf1-17(Q71L)p as fusions to the LexA DNA-binding domain in the bait vector (pGlida), and then tested for interaction with ArfGAPs (Glo3p and Gcs1p) fused to the B42 activation domain in the prey vector (pJG4-5) (Fig. 3). All chimeric proteins were expressed under the GAL1 promoter and their expression levels were confirmed by immunoblotting (supplementary material Fig. S3). We could observe that both Glo3p and Gcs1p interact with Arf1(Q71L)p as described previously. By contrast, neither of them showed positive interaction with arf1-16(Q71L)p or arf1-17(Q71L)p.
Next we tested the GAP activities of Glo3p and Gcs1p toward arf1-17p in vitro. Recombinant Glo3p and Gcs1p were produced in E. coli and subjected to the in vitro GAP assay using purified WT-Arf1p and arf1-17p as substrates (Fig. 4). Glo3p and Gcs1p stimulated the hydrolysis of GTP to GDP by WT-Arf1p at both 23 and 35°C. The GTPase activity of arf1-17p was also enhanced by these ArfGAPs. The stimulatory effects were almost comparable to those of the wild type at 23°C, but were significantly lower at 35°C. However, what should be noted here is that the effects of Glo3p and Gcs1p on the arf1-17 mutant protein were almost indistinguishable in this in vitro reaction.
The C-terminal domain of Glo3p is important for its suppression activity on arf1-16 How can we then explain the difference between Glo3p and   (Poon et al., 1999). The homology between Glo3p and Gcs1p is limited to their N-terminal region containing the ArfGAP domain, which is known as the minimal domain to execute the ArfGAP activity of rat Arfgap1 in vitro (Cukierman et al., 1995). By contrast, the similarity between Glo3p and Gcs1p is very low in other regions of the molecules. It has also been reported that the C-terminal non-catalytic domain of rat Arfgap1 is required for its recruitment from cytoplasm to membranes and for its interaction with the KDEL receptor (Aoe et al., 1999;Huber et al., 1998). From these aspects, we suspected that the clue to understand the difference between Glo3p and Gcs1p in vivo lies in the C-terminal domains of the proteins.
To examine this possibility, we performed domain swapping of Glo3p and Gcs1p and examined the ability of the chimeric molecules to suppress the growth defect of arf1-16 (Fig. 5). The ArfGAP domain of yeast ArfGAPs contains a functionally important C 2 C 2 H 2 zinc-finger motif (Zhang et al., 1998), and the similarity between Glo3p and Gcs1p is especially high around this motif. When the C-terminal large portion (residues 60-353) of Gcs1p was replaced by the C-terminal region (residues 65-494) of Glo3p (named Gcs1-A-Glo3p: 1-60/65-494), overexpression of the chimeric protein rescued the temperature sensitivity of arf1-16. However, the reverse chimeric construct, Glo3-A-Gcs1p (1-64/60-353), did not show the suppression activity. Furthermore, the overexpression of Glo3-A-Gcs1p compromised the ts growth of arf1-16, just like Gcs1p itself. Similarly, chimeric proteins fused at a slightly further downstream position, Gcs1-N-Glo3p (1-111/122-494) and Glo3-N-Gcs1p (1-122/113-353), caused suppression or aggravation of the ts growth, depending on which protein the C-terminus was derived from. These results support the differentiated roles of the C-terminal region of Glo3p and Gcs1p.
Tandem Glo3 motif at the C-terminus of Glo3p is essential for the suppression of arf1-16 Because Glo3p and Gcs1p showed different effects in vivo depending on their C-termini, we went on to examine whether ArfGAPs from other eukaryotic organisms show any particular similarity to either of them. We searched genome databases of Fig. 3. Interactions of mutant arf1p with Arf GAPs in the DupLEX-ATM two-hybrid system. The interactions between N-terminally truncated (N17) Arf1p in the bait vector (pGlida) and Glo3p or Gcs1p in the prey vector (pJG4-5) were tested in the host strain EGY48. The Q71L mutation had been introduced in wild-type (ARF1), arf1-16 mutant (arf1-16) and arf1-17 mutant (arf1-17) genes. Transformants were plated on MVD (-Trp, -His), grown for 2 days, replicated on MVGal (minimal galactose medium) (-Trp, -His, -Leu), MVD (-Trp, -His), and MVD (-Trp, -His, -Leu) plates, and then further incubated at 30°C for 6 days. Empty prey vector (pJG4-5) was used as a negative control. Homo sapiens, Mus musculus, Caenorhabditis elegans, Drosophila melanogaster and Arabidopsis thaliana using the whole length of amino-acid sequences of Glo3p and Gcs1p independently, and picked up ArfGAP(s) that showed high blast scores in each species. As shown in Table 1, completely separable subgroups could be defined, and we hitherto classify these ArfGAPs as Glo3-type and Gcs1-type, respectively. Interestingly, multiple alignments of the Glo3-type ArfGAPs revealed limited but very high sequence similarity at their Ctermini (Fig. 6A). We call this conserved C-terminal region the Glo3 motif. The Glo3 motif generally contains two repeats of the ISSxxxFG sequence, with the exception of D. melanogaster and C. elegans proteins, in which one of the two repeats diverges. K386 and L411 of S. cerevisiae Glo3p in this motif are also well conserved among the Glo3-type ArfGAPs.
In contrast, none of Gcs1-type ArfGAPs contain the Glo3 motif. Other types of ArfGAPs such as Age1p or Age2p of yeast do not possess the Glo3 motif either.
To examine the physiological importance of the Glo3 motif, Journal of Cell Science 119 (12) we introduced mutations in this motif of Glo3p and tested its ability to suppress the growth defect of arf1-16 ( Fig. 6B-D). Residues in either the first or second repeat of ISSxxxFG were replaced by alanines and subjected to the suppression test.
Overexpression of the glo3-11 and glo3-12 proteins, in which only one ISS sequence was mutated, partially suppressed the ts growth of arf1-16. When the both ISS sequences were mutated (glo3-13 and 14), the suppression activity was completely lost. This result indicates that the Glo3 motif is critical for the in vivo function of Glo3p.

Discussion
We have been investigating how Arf1p plays multiple roles in living cells. arf1 ts mutants we isolated show a variety of transport defects and morphological alterations in an allelespecific manner. To identify partner molecules that participate in individual functions of Arf1p, we have screened for multicopy suppressors of the arf1 ts mutants. In the present study, we have demonstrated that the overexpression of GLO3, one of the ArfGAPs, suppresses the ts growth of arf1-16 and arf1-17 mutants, which are classified into the same intragenic complementation group, whereas other ArfGAPs (GCS1, AGE1 and AGE2) do not. The overproduction of Gcs1p is inhibitory for the growth of arf1-16 and arf1-17 cells. This result raises the possibility that different ArfGAPs act differently on Arf1p to allow it to fulfill divergent functions.
In the nucleotide-binding assay, the arf1-17 protein showed lower binding of GDP than GTP at 30°C. If this imbalance of GDP/GTP binding was a cause of the mutant phenotype of arf1-17 cells, it seemed reasonable that the overproduction of the Glo3p ArfGAP suppressed it. The difference in the suppression activity between Glo3p and Gcs1p might be explained by their different affinities to the mutant Arf1p. However, the GAP activities of Glo3p and Gcs1p on the arf1-17 protein turned out to be almost identical in vitro. The key to solve this colons and stops indicate that the stronger-and weaker-score groups is fully conserved, respectively. SAT motifs (C 2 C 2 H 2 ) (Zhang et al., 1998) are underlined, and the ArfGAP domain of rat Arfgap1 (residues 1-136) (Cukierman et al., 1995;Goldberg, 1999) is shaded. (B) Construction of chimeric proteins of Glo3p and Gcs1p. (C) Suppression activities of Glo3p and Gcs1p chimeric proteins. arf1-16 cells were transformed with pNY24-GLO3 (Glo3p), pNY24-GCS1-A-GLO3 (Gcs1-A-Glo3p), pNY24-GCS1-N-GLO3 (Gcs1-N-Glo3p), pNY24-GLO3-N-GCS1 (Glo3-N-Gcs1p), pNY24-GLO3-A-GCS1 (Glo3-A-Gcs1p), and pNY24-GCS1 (Gcs1p), and the transformants were grown at 23°C. Cells were diluted, spotted on MCD (-Trp) plates, and incubated at the indicated temperatures for 2 days. Integrants of wild-type ARF1 (WT) and mutants transformed with pYO324 (vector) were used as controls. apparent disparity was revealed to lie in the non-catalytic regions of ArfGAPs.
All known ArfGAPs contain the ArfGAP domain, which executes the ArfGAP catalytic activity, but the amino-acid sequences in the rest of molecules are different from each other. When the non-catalytic C-terminal portions of Glo3p and Gcs1p were exchanged, their effects on arf1-16 were also switched, suggesting that this region is required for their individual functions in vivo. Interestingly, we found a novel motif at the C-terminus of Glo3p (residues 386-422 of Glo3p), which is conserved among Glo3-type ArfGAPs in many eukaryotes. We named this region the Glo3 motif. The characteristic feature of this motif is one or two repeats of the ISSxxxFG sequence. Point mutations of the conserved motif resulted in the loss of the suppression activity in arf1-16, indicating that the Glo3 motif is required for the proper activity of Glo3p in living cells. These Glo3-motif mutant proteins are supposed to retain the ArfGAP activity, because the N-terminal ArfGAP domain remains intact. Therefore, the activity of Glo3p is probably modulated by the third molecule(s) through the interaction with the Glo3 motif in vivo. arf1-16/17 mutant proteins presumably have a problem in interacting with such modulators.
What could be the third molecule(s) then? Glo3p has been thought to play an important role in the Golgi-to-ER retrograde transport in cooperation with the coatomer complex. Dogic et al. (Dogic et al., 1999) showed in an analysis of yeast disruptants of the ArfGAP genes that the deletion of GLO3 alone resulted in a defect of the retrieval of dilysine-tagged proteins from the Golgi. Glo3p was also reported to interact with Sec21p (␥-COP) and Sec27p (␤Ј-COP), whereas Gcs1p failed to bind to coatomer both in a two-hybrid assay and in immunoprecipitation analysis (Eugster et al., 2000;Lewis et al., 2004). The Glo3 motif at the C-terminus may be required for the interaction with the coatomer complex. In an attempt to support this possibility by a genetic analysis, however, none of the COPI genes could suppress the ts growth of arf1-16 by their single overproduction (our unpublished results).
Another class of molecules we are interested in are Bet1p and Bos1p. Recent studies have shown that the ArfGAP activity is required not only to remove the protein coat but also for concentration of cargo into vesicles (Lanoix et al., 1999;Lanoix et al., 2001;Nickel et al., 1998;Pepperkok et al., 2000). In the case of a related GTPase, Sar1p, we have also shown  (C) NYY16-1 cells (arf1-16) were transformed with pYO324 (vector), pNY14-3HA-GLO3 (CEN 3HA-GLO3), pNY24-3HA-GLO3 (2 3HA-GLO3), pNY24-glo3-m2HA (2 3HA-glo3-11), pNY24-glo3-m1HA (2 3HA-glo3-12), pNY24-glo3-m3HA (2 3HA-glo3-13) and pNY24-glo3-m5HA (2 3HA-glo3-14). Cells were grown, diluted and spotted on MCD (-Trp) plates, and incubated at the indicated temperatures for 2 days. Integrants of wild-type ARF1 (NYY0-1: WT) transformed with pYO324 were used as a control. (D) Immunoblotting of 3HA-Glo3p to examine expression levels. The transformants that were used in C were grown to an early log phase at 23°C, and incubated at 35°C for 1 hour. Total cell lysates (60 g) were prepared and analyzed by immunoblotting using the anti-HA monoclonal and anti-Sec61p antibodies. Lane numbers correspond to the transformant numbers used in C. that the regulation of GTP hydrolysis by Sec23p GAP is important for cargo selection during COPII vesicle budding (Sato and Nakano, 2005). If Glo3p promotes the sorting of a particular set of cargo molecules, such cargo could be isolated as multicopy suppressors of arf1-16. BET1 and BOS1, which have been isolated as multicopy suppressors of arf1-16/17 in the present study, encode SNARE molecules functioning during the vesicular recycling between the ER and the Golgi apparatus (Martinez-Menarguez et al., 1999;Rexach et al., 1994). BET1 and BOS1 were also isolated as multicopy suppressors of the temperature sensitivity of ⌬glo3 mutant cells (Poon et al., 1999). Furthermore, Rein et al. (Rein et al., 2002) reported that Glo3p enhances the binding of COPI and arf1⌬N17p to the cytoplasmic tails of Bet1p, Bos1p and Sec22p. If one hypothesizes that arfGAP has a role to enrich these SNAREs as cargo into COPI vesicles, then the defects of arf1-16/17 mutant proteins could be explained by the defective ability of responsible ArfGAP to select cargo. In our preliminary experiment, arf1-16p/17p showed rather elevated binding to Bet1p than the wild-type Arf1p, but this binding was not significantly affected by the addition of either Glo3p or Gcs1p (our unpublished result).
Why does Gcs1p exaggerate the ts defect of arf1-16 and arf1-17 cells? Glo3p and Gcs1p are known to play essential overlapping functions in retrograde vesicular transport (Poon et al., 1999). Indeed, glo3⌬ gcs1⌬ double deletion mutant cells are inviable (Poon et al., 1999;Zhang et al., 1998). However, there is also evidence suggesting that Glo3p and Gcs1p have additional and distinct roles. For example, the function of Gcs1p overlaps with Age2p for transport from the yeast trans-Golgi network (Poon et al., 2001b). The simultaneous overexpression of GLO3 and GCS1 resulted in weaker suppression of arf1-16/17 than that of GLO3 alone (data not shown), suggesting that these ArfGAPs are competing for something. Arf1p itself is not a likely candidate because the overexpression of other ArfGAPs, AGE1 and AGE2, have no effect on the growth of the mutants. One attractive model may be that Glo3p and Gcs1p function on the same membrane and compete with each other for the binding site on the membrane. Antonny's group showed that rat Arfgap1 and Gcs1p bind preferentially to liposomes containing conical lipids but not those of cylindrical lipids, and the preference for conical lipids correlated with its GAP activity on liposome-bound Arf1p-GTP (Antonny et al., 1997;Bigay et al., 2003). More recently, they proposed that the central region of Arfgap1 contains a sequence that acts as a lipid packing sensor and named it the ALPS motif (Bigay et al., 2005). Because this ALPS motif is well conserved in the Gcs1-type ArfGAPs but not in others, it remains an open question whether or not the Glo3-type ArfGAPs also sense lipid structures. Since Glo3p possesses overlapping functions with Gcs1p in retrograde transport from the Golgi to the ER, but not with Age1 and Age2p (Poon et al., 1999;Poon et al., 2001b), it may be reasonable to assume that Glo3p and Gcs1p function on the same membrane, e.g.
Golgi. An excess amount of Gcs1p might fill up the binding sites of ArfGAPs on the membrane where Glo3p also binds. The critical defect of the arf1-17 mutant protein could be an inability to bind specific cargo, e.g. Bet1p, which needs access to Glo3p on the membrane. OLE1, a gene encoding a fatty acid desaturase required for monounsaturated fatty acid synthesis, has also been isolated as a multicopy suppressor of arf1-16 in Journal of Cell Science 119 (12) the present study. The overexpression of OLE1 may increase the amount of conical lipids, resulting in the gain of more binding sites for ArfGAPs.
In summary, our results suggest that Glo3p and Gcs1p possess different functions in vivo and that their non-catalytic C-terminal regions are required for their individual activities. Both Glo3-type and Gcs1-type ArfGAPs are found in human, mouse, fly, worm and plants. The Glo3-type ArfGAPs may well play distinct roles from Gcs1-type ArfGAPs, not only in yeast but also in other eukaryotic organisms. The human genome contains at least 16 ArfGAP proteins. They have been divided into several subgroups (Arfgap1 type, Gits type and AZAP type) according to the overall domain structures (Randazzo and Hirsch, 2004). Human Arfgap1 and Arfgap3 are categorized as Arfgap1-subgroup proteins. In this study, we present the evidence that the Arfgap1-subgroup proteins are further classified into two groups, Gcs1 type and Glo3 type. The Glo3-type ArfGAPs have the Glo3 motif in their noncatalytic domains.
During the preparation of this manuscript, Lee et al. (Lee et al., 2005) published a paper reporting that rat Arfgap1 contains at least two distinct binding sites for coatomer, one in its catalytic domain and the other in the non-catalytic domain. They also revealed that a truncated Arfgap1 containing the first 136 residues (out of non-catalytic C-terminal domain) failed to bind cargo proteins. The ALPS motif was also found in the non-catalytic regions of rat Arfgap1 and Gcs1p (Bigay et al., 2005). Further experiments should address the physiological roles of non-catalytic domains of ArfGAPs, including the Glo3 motif. Arf1p probably uses these different kinds of ArfGAPs and thus plays multiple roles in living cells.

Isolation of multicopy suppressors of arf1-16
The S. cerevisiae genomic DNA library based on YEp13, a multicopy vector with the LEU2 marker (Yoshihisa and Anraku, 1989), was introduced into the arf1-16 mutant (NYY16-1). Transformants were grown on MVD (minimal glucose medium) lacking leucine at 23°C for 1 day and then incubated at 35°C for 3 days. Colonies that grew at 35°C were re-streaked on MVD plates and confirmed for growth at 35°C. Plasmids were recovered from the candidate clones and reintroduced into arf1-16 to verify the suppression activity. Out of approximately 50,000 transformants screened, 78 positive clones were obtained. Among strong suppressors, four clones contained either ARF1 or ARF2. In addition, we found that 12, 6, 4, 14 and 2 clones repeatedly contained genomic fragments harboring GLO3, BET1, BOS1, OLE1 and GYP6, respectively. They were subcloned into pYO325, a multicopy vector with the LEU2 gene, to confirm that the suppression activities depended on these five genes.

Two-hybrid assay
The bait plasmid pGlida, the prey plasmid pJG4-5, and the yeast reporter strain EGY48 were obtained from OriGene Technologies, Inc. (Rockville, MD, USA). Plasmid constructions and the examination of interaction between (⌬N17)Arf1(Q71L)p and ArfGAPs were performed as previously described except that the bait pEG202 was replaced with pGlida (Eugster et al., 2000).

Purification of myristoylated-Arf1 proteins
Yeast N-myristoyltransferase (NMT1) expression vector (pACYC177-ET3d-NMT1) was constructed as previously described (Haun et al., 1993) and its activity was confirmed using [1-14 C]myristic acid (data not shown). The full-length ARF1 and arf1-17 genes were cloned into NdeI-XhoI sites of pET21a, and E. coli BL21 (DE3) was co-transformed with the resulting plasmids (pET-ARF1 and pET-arf1-17, respectively) and pACYC177-ET3d-NMT1. Transformed cells were selected for both ampicillin and kanamycin resistance and grown at 37°C. After myristate (50 M) bound to BSA (6 M) was added to the culture (Franco et al., 1995), the coexpression of ARF1 and NMT1 was induced with 1 mM isopropyl ␤-Dthiogalactoside (IPTG), and the temperature was reduced to 30°C to increase the yield of myristoylation. Purification of myristoylated-Arf1 and myristoylated-Arf1-17 proteins was performed by the method previously described with some modifications (Peyroche and Jackson, 2001;Randazzo et al., 1995;Weiss et al., 1989). Briefly, the cells were harvested, resuspended in 50 ml lysis buffer [20 mM Tris-HCl (pH 8), 1 mM EDTA, 200 M GDP, 1 mM DTT and 0.4 mM PMSF], and lysed. After removal of debris, the lysate was centrifuged at 100,000 g for 1 hour at 4°C. The supernatant was recovered, and ammonium sulfate was added to 33% saturation. The mixture was incubated for >1 hour at 4°C with agitation and centrifuged at 20,000 g for 30 minutes at 4°C. The pellets were resuspended in dialysis buffer [20 mM Tris-HCl (pH 8), 1 mM MgCl2, 5 M GDP, 1 mM DTT and 0.4 mM PMSF], and dialyzed against the same buffer three times (2 hours, overnight, then 3 hours). The dialyzed supernatant was mixed with DEAE-Sepharose FF (Amersham), incubated for 10 minutes at 4°C, and centrifuged at 30,000 g for 10 minutes. The material that failed to bind to DEAE-Sepharose was concentrated to about 4 ml by ultrafiltration using a Centriprep 10 (Amicon). The supernatant was filtrated through a 0.45 m filter, and then loaded onto a 120-ml HiLoad 16/60 Superdex 200 pg column (Amersham) equilibrated in buffer A (20 mM Tris-HCl, pH 8, 1 mM MgCl 2 and 1 mM DTT).

Guanine-nucleotide binding assay
GTP␥S-and GDP-binding activities were measured by the procedure described previously with some modifications (Peyroche and Jackson, 2001;Randazzo et al., 1995). Purified Arf1p or arf1-17p (1.25 M each) was incubated in 25 mM HEPES-KOH (pH 7.5), 1 mM DTT, 1 mM MgCl 2 , 3 mM DMPC, 0.1% (w/v) sodium cholate and 10 M [ 3 H]GDP or [ 35 S]GTP␥S (10,000 cpm/pmol) at 23 or 30°C. At appropriate time intervals, aliquots of the reaction mixture were taken and the reaction was stopped by diluting samples (20 l) into 1 ml of stop buffer [20 mM HEPES-KOH (pH 7.5), 100 mM NaCl and 10 mM MgCl 2 ] at 4°C. Proteins were filtrated through a nitrocellulose membrane filter, washed four times with 2 ml of the stop buffer and then the radioactivity of the bound nucleotide was measured by scintillation counting.

GAP assays
Construction of expression vectors encoding hexahistidine (His 6 )-tagged GCS1 and GLO3 was performed as described previously (Poon et al., 2001a). The plasmid pGLO3-HIS encoding His 6 -tagged GLO3 at its C-terminus was constructed by cloning the gene into pET21a (Novagen). For GCS1, the first six codons of the GCS1 ORF were replaced by His 6 sequences, and cloned into pET21a. The resulting plasmid, pHIS-GCS1, expresses N-terminally His 6 -tagged GCS1. Both Gcs1p and Glo3p were largely insoluble and were thus prepared under denaturing conditions by guanidine extraction followed by Ni-NTA purification, as described before, except that a HiTrap Q (Amersham) column was used instead of a Resource Q column. GAP activity was assayed as described before with some modifications (Cukierman et al., 1995;Huber et al., 2001). Arf1 protein (5 M) was first incubated for 15 minutes at 23°C with 2 M [ 32 P]GTP in the presence of 50 mM HEPES-KOH (pH 7.5), 1 mM DTT, 2 mM EDTA, 1 mM MgCl 2 , 10 mM ATP, 3 mM DMPC and 0.1% cholate. The reaction was terminated by the addition of MgCl 2 to 2 mM.
The [ 32 P]GTP-loaded Arf1 proteins were diluted into the assay system containing 4.6 mM MgCl 2 , 50 mM HEPES-KOH (pH 7.5), and 10 mM ATP. Following a brief warming, the reaction was initiated by the addition 0.2 M ArfGAP proteins. At appropriate times, aliquots were transferred into microfuge tubes containing the final concentration of 250 mM EDTA. The samples were spotted onto a polyethyleneimine cellulose plate (MERCK) and developed with 1 M LiCl/1 M HCOOH (1:1, v/v). The plate was dried, fluorographed, and quantitatively analyzed with an image analyzer BAS2500 (Fuji Film, Tokyo, Japan).