- Top of page
- Materials and methods
Previous studies in yeast have revealed the presence of four proteins with a conserved, cysteine-rich, ARF GAP domain that share the ability to suppress the conditional growth defect of the arf1-3 mutant. Three of these proteins have been shown previously to be ADP-ribosylation factor (ARF) GTPase-activating proteins (GAPs). We now demonstrate that the fourth also exhibits in vitro ARF GAP activity and correlates the suppressor and ARF GAP activities for all four. Because the four ARF GAP proteins are quite diverse outside the ARF GAP domain, a genetic analysis was undertaken to define the level of functional cross-talk between them. A large number of synthetic defects were observed that point to a high degree of functional overlap among the four ARF GAPs. However, several differences were also noted in the ability of each gene to suppress the synthetic defects of others and in the impact of single or combined deletions on assays of membrane traffic. We interpret these results as supportive evidence for roles of ARF GAPs in a number of distinct, essential cellular processes that include cell growth, protein secretion, endocytosis and cell cycling. The description of the specificities of the ARF GAPs for the different responses is viewed as a necessary first step in dissecting biologically relevant pathways through a functionally overlapping family of signalling proteins. Copyright © 2003 John Wiley & Sons, Ltd.
- Top of page
- Materials and methods
ADP-ribosylation factors (ARFs) are multi-functional, 21 kDa regulatory GTPases. All ARFs, from Giardia to human, are greater than 60% identical and share the ability to rescue the lethal double mutant, arf1−arf2−, in the yeast, Saccharomyces cerevisiae (Stearns et al., 1990; Kahn et al., 1991; Lee et al., 1992). ARFs have a number of biochemical activities, including the ability to recruit coat proteins (e.g. COP-I and AP-1) to membranes, co-factor activity for the bacterial ADP-ribosylating toxins (cholera and Escherichia coli toxins), activation of phospholipase D, and stimulation of PI(4)P kinase (for reviews, see Boman and Kahn, 1995; Moss and Vaughan, 1998). These activities are required, among other things, for maintenance of the morphology of, and membrane traffic through, the Golgi.
In S. cerevisiae ARFs have been implicated as regulators of a large number of essential cellular functions including vesicular traffic, maintenance of organelle morphology, respiration, sporulation, mitotic growth and entry into the cell cycle (Stearns et al., 1990; Ireland et al., 1994; Kahn et al., 1995; Rudge et al., 1998; Blader et al., 1999). Although the biochemical activities of ARFs are less well studied in yeast (e.g. no ARF-sensitive phospholipase D or lipid kinase activities have been detected in yeast; Rudge et al., 1998), a number of genetic analyses have proved to be instrumental in the elucidation of ARF function, e.g. early studies of arf1::HIS3 uncovered synthetic defects with ypt1-1, sec21-1, and sec7-1 (Stearns et al., 1990).
Genetic analyses, along with sequence homology searches, identified GCS1 as the first yeast ARF GAP and delineated a cysteine-rich ‘zinc finger’ as the ARF GAP domain (Cukierman et al., 1995; Poon et al., 1996). GCS1 (along with GLO3, AGE1 and AGE2) was also recovered in a screen for high-copy suppressors of the loss of function mutant, arf1-3. These four suppressor of arf1-3TS (SAT) genes (Zhang et al., 1998) encode proteins that contain ARF GAP domains and to date all but Age1p have been demonstrated to possess in vitro ARF GAP activity (Poon et al., 1996, 1999, 2001). The cloning of these genes as suppressors of the loss of Arf1p function was interpreted as evidence that they possess effector functions in addition to ARF GAP activity (Zhang et al., 1998). In support of this proposal, interacting or overlapping roles in vesicular transport for GLO3 and GCS1 (Golgi to ER) and GCS1 and AGE2 (TGN to endosome) have been recently described (Dogic et al., 1999; Poon et al., 1999, 2001). To provide additional characterization of the ARF GAP family, we now examine Age1p for ARF GAP activity and test for interactions between all four of these genes. The results reveal that the four SATs comprise a family of ARF effectors with overlapping biological roles and shared ARF GAP activity. Because the dissection of ARF signalling pathways has been slowed by the plethora of effects of ARF we also initiated tests for functional redundancy and specificity among the four ARF GAPs that should facilitate future characterization of diverse ARF pathways.
- Top of page
- Materials and methods
Four genes were previously cloned (Zhang et al., 1998) as high copy suppressors of arf1-3TS (SAT). While two of these genes (GCS1 and GLO3) had already been demonstrated to have functions related to the ARF GTPases (Ireland et al., 1994; Poon et al., 1996), the other two (SAT1/AGE11 and SAT2/AGE2) had not. More recently, Age2p has also been shown to possess ARF GAP activity and play a role in membrane traffic (Poon et al., 2001). As all four proteins share sequence relatedness, in the ARF GAP domain, we tested Age1p protein for ARF GAP activity.
Age1p has ARF GAP activity
Full-length Age1p and Age2p and a series of deletion mutants of Age1p (Figure 1) were engineered to include a myc epitope at their C-termini and expressed in yeast on high copy (2µ) plasmids. Total cell lysates from these strains were then assayed for protein expression (Figure 2) and ARF GAP activity (Figure 3). As seen in Figure 2, blotting with the myc antibody readily detected each of these constructs, except Age1Δ2p and Age1Δ8p.
Figure 1. Linear depiction of the Age1p protein and eight truncation mutants. The hatched area is the ARF GAP domain. The numbers above each truncation indicate residue numbers with reference to the full-length protein and the numbers to the right indicate the total length of the protein
Download figure to PowerPoint
Figure 2. Immunoblot analysis of Age1p and truncations. Total yeast lysates from cells expressing Age1p-myc, Age2p-myc or the indicated truncation mutants of Age1p-myc were probed with anti-myc mouse monoclonal antibody (9E10). Note that deletion of the N-terminal 163 residues led to increased expression and that no protein was detected for the Age1Δ2p-myc or Age1Δ8p-myc proteins
Download figure to PowerPoint
Figure 3. ARF GAP activity in yeast lysates expressing Age1p-myc, truncations of Age1p-myc, or Age2p-myc. Yeast lysates (10 µg) were assayed for ARF GAP activity using purified, recombinant yeast Arf1p (0.45 µM) or human (Δ17)ARF1 (0.45 µM). Age1p-myc, Age1Δ1p-myc, Age1Δ2p-myc, Age1Δ3p-myc, Age1Δ4p-myc, Age1Δ5p-myc or Age2p-myc were expressed by introduction of 2µ plasmids into strain RT364. Assays were performed for 10 min at 30° C, as described under Materials and methods
Download figure to PowerPoint
ARF GAP activity was assayed in total cell lysates. As seen in Figure 3, cell lysates of untransformed yeast exhibit low background levels of GAP activity, which presumably results from endogenous ARF GAP proteins. However, when Age1p, Age1Δ1p, Age1Δ4p or Age1Δ5p were overexpressed, a marked increase in ARF GAP activity was evident. Strains carrying the Age1Δ3p mutant were found to have detectable, although reduced, activity, while those expressing the Age1Δ2p mutant never showed ARF GAP activity, consistent with the inability to detect a stably expressed protein. Strains carrying the plasmid directing expression of full-length Age2p also resulted in increased ARF GAP activity, most easily detected using the human (Δ17)ARF1 mutant that binds GTP to high stoichiometry (see Figure 3).
Each of the deletion mutants of Age1p was also tested for the ability to suppress the conditional (TS) phenotype of arf1-3 as well as the more severe TS of arf1-3 age1− (data not shown). The Age1Δ1p, Age1Δ4p, Age1Δ5p and Age1Δ7p mutants all showed wild-type levels of SAT activity indicating that residues 164–399 are sufficient for suppression. The Δ3 mutant had reduced activity but increasing protein levels by deleting the amino-terminal 163 residues to form the Δ7 mutant could restore full suppression. Although it is possible that the N-terminus has a negative effect on SAT activity, this seems unlikely since the Δ4 mutant retains full activity. In contrast, suppression was not restored to the inactive Δ2 mutant by the same N-terminal deletion to form the Δ6 mutant in spite of the increase in protein levels. This suggests that a region (i.e. residues 294–399) outside of the cysteine-rich ARF GAP domain is needed for both suppression of the TS of arf1-3 and GAP activity. The correlation between SAT and ARF GAP activities seen with all of the constructs also suggests that the two activities are likely functionally related.
These results indicate the presence of at least four regions in Age1p: (a) the N-terminal extension (residues 1–163) which may act as a negative regulator of protein stability; (b) the cysteine-rich ARF GAP domain (residues 186–220); (c) a region C-terminal to the ARF GAP domain (residues 294–399) that is necessary for full GAP and SAT activity; and (d) residues 399–482, which are dispensable for these activities. The only one of these four regions that has high homology to numerous entries in protein databases is the ARF GAP domain. However, the fourth region includes a short stretch of homology to GPI-linked NAD+-arginine ADP-ribosyltransferase(s) (Accession No. Q60935):
Although the functional importance of this homology is unknown, the fact that ARFs bind to both Age1p and bacterial ADP-ribosyltransferases (cholera and E. coli heat-labile toxins) is notable.
Mammalian GAP has SAT activity
Because of the extensive structural and functional conservation between yeast and mammalian ARFs we tested for suppression of arf1-3 by cDNAs encoding full-length and truncated (1–136) rat ARF GAP1, and the human clones KIAA0050 and KIAA0148. While all four of these proteins contain the cysteine-rich ARF GAP domain, only the full-length and truncated ARF GAP 1 has been shown directly to possess ARF GAP activity (Cukierman et al., 1995; Goldberg 1999). Each protein was expressed in yeast strains RT364 (arf1-3 arf2−) and YZC243 (arf1-3 arf2−age1−) on high copy plasmids and tested for suppressor activity. Both full-length and truncated mammalian ARF GAP had SAT activity in each strain while neither of the KIAA-encoded proteins suppressed the TS (data not shown). The ability of the truncated mammalian ARF GAP to suppress arf1-3 is further evidence of functional conservation in ARF signalling in yeast and higher eukaryotes and given the minimal nature of the truncation mutation, is further support for the correlation between ARF GAP and SAT activities, noted above.
Characterization of the ARF GAP activity of Age1Δ5p
Full-length Age1p and Age2p were expressed in bacteria but each was insoluble. Of the deletion mutants (Figure 1), the most abundant, stable and soluble protein expressed in bacteria was Age1Δ5p, so it was chosen for further characterization of the in vitro GAP activity. Age1Δ5p had ARF GAP activity when purified yeast Arf1p or Arf2p (Figure 4, top panel) were substrates. This ARF GAP activity is not dependent on myristoylation of the N-terminus as the acylated myr-yArf1p and non-acylated yArf2p were each good substrates. In contrast, mutation of glutamine 71 to leucine, previously shown to be a lethal, dominant activating mutant, is sufficient to prevent the hydrolysis of bound GTP by the GAP. Age1Δ5p also catalysed the hydrolysis of GTP bound to human ARF1 (data not shown) or ARF3 (Figure 4) but not yArl1p, human ARL2 (Figure 4), or yArf3p (data not shown). Thus, the ARF GAP activity of Age1p is specific for ARF proteins, yeast or human, and does not extend to the yeast or human ARLs tested. Because some mammalian ARF GAPs are highly dependent on phosphatidylinositols, particularly PIP2 (Randazzo and Kahn, 1994), we tested for the lipid dependence of Age1Δ5p. There was no effect of adding a mixture of acidic lipids, which included phosphatidylinositols, on the GAP activity of Age1Δ5p (data not shown).
Figure 4. ARF GAP activity of purified Age1Δ5p-His6 against a number of ARF and ARL proteins. A single round of GTP hydrolysis was measured as described under Materials and methods. (α-32P)-GTP was loaded into the nucleotide binding site of ARF or ARL proteins before the ARF GAP assays were performed in the presence or absence of recombinant Age1Δ5p-His6, as indicated, at 30° C for 10 min. The ratio of ARF or ARL to Age1Δ5p-His6 was 1 : 5
Download figure to PowerPoint
Increased expression of AGE1 also suppresses lethality from excess Arf1p
The four S. cerevisiae ARF GAPs could rescue the temperature sensitive growth of arf1-3arf2− cells at 37° C with a rank order of suppressor activity of AGE1 > GLO3 > AGE2 > GCS1 (Zhang et al., 1998). In contrast, none of them allowed mitotic growth of arf1−arf2− cells when overexpressed. However, toxicity seen with increased expression of ARF1, using the stronger GAL1 promoter, was suppressed. Strain RT256 carries a low copy number plasmid with the ARF1 coding region under control of the GAL1 promoter and grows at normal rates on plates containing glucose as carbon source but fails to grow on galactose (Figure 5). The same high-copy number plasmids that suppressed temperature sensitivity of arf1-3 were transformed into this strain and growth on glucose- and galactose-containing media was assessed. Strains expressing AGE1, but not GLO3 or SPS18, were able to grow on Gal+ plates. AGE2 and GCS1 also suppressed the toxicity associated with excess ARF1 roughly equally, but less well than AGE1 (see Figure 5). Thus, increased expression of the ARF GAPs can suppress killing resulting from either increased or decreased ARF activity and they do so with a different rank order potency. Interestingly, AGE1 was the strongest suppressor of each phenotype.
Figure 5. Genes encoding ARF GAPs also suppress lethality resulting from excess ARF1. RT256, carrying p GAL1–ARF1 on a CEN plasmid, was transformed with high-copy plasmids bearing AGE1, AGE2, GCS1, SPS18 or GLO3, as indicated. Serial dilutions of cell suspensions were spotted onto selective plates (SD–Leu–Ura) with either glucose or galactose as carbon source. The plates were incubated at 28° C for 5 days before photographing. The rank order of SAT activity can be seen as AGE1 > GCS1 > AGE2 > GLO3 = SPS18
Download figure to PowerPoint
Genetic interactions among members of the ARF GAP family
Further genetic interactions between the ARF GAPs and with arf1− were tested by creating strains with all combinations of deletions. S. cerevisiae ARF1 and ARF2 encode proteins that are >96% identical and which lack discernible differences in biochemical activities. ARF1 accounts for ∼90% of total cellular ARF in this yeast (Stearns et al., 1990). Thus, the deletion of ARF1 is viewed as a lowering of total ARF to a level at which phenotypes are detectable but cell survival is not compromised. The different synthetic defects observed between the ARF1 deletion and each of the ARF GAPs is, therefore, evidence of functional differences among the ARF GAPs. These differences may entail alterations in specific activities, protein interactions or localization. These synthetic defects may also be interpreted as further evidence that the GAPs do not operate in intact cells as simple terminators of ARF signalling, but rather have distinctive effector functions.
Gene deletions were carried out in our genetic background, or in a few cases after backcrossing knockouts from other laboratories into our strains at least three times. The gcs1::HIS3 allele used was generously provided by Ira Blader and Anne Theibert and leaves the N-terminal 75 residues intact. All experiments with this allele were later repeated using a GCS1 allele in which the entire open reading frame was replaced and the same results were obtained (data not shown). Growth was examined on rich (YPD) plates at 16° C, 28° C and 37° C or at 28° C on YPD plates containing 40 mM sodium fluoride. We also monitored the integrity of the secretory pathway by examining the processing of the secreted form of invertase, and of the endocytic pathway and vacuole by examining uptake of the vital stain FM4-64.
As expected, none of the four genes encoding ARF GAPs are essential as strains carrying these single deletions are each viable (Table 3; Ireland et al., 1994; Zhang et al., 1998; Poon et al., 1999). Indeed, the growth of deletants of AGE1 or AGE2 was indistinguishable from the parental strain at each temperature tested. As previously reported (Poon et al., 1996) deletion of GCS1 yields a strain with a weak CS phenotype and sensitivity to 40 mM fluoride. The glo3− strain exhibited weak TS and strong CS phenotypes.
Table 3. Growth and genetic interactions of yeast strains carrying deletions in one or more ARF GAP
|Strain||Deletion||Growth at 28° C||Growth at 37° C||Growth at 16° C||Growth YPD/F40||Phenotype|
|TT104||Δarf1||+++||+++||++||−||FS, weak CS|
|RT587||Δglo3||+++||++||+||+++||CS, weak TS|
|RT588||Δgcs1||+++||+++||++||−||FS, weak CS|
|RT590||Δarf1Δglo3||+++||+||−||−||CS, FS, weak TS|
|Δarf1Δgcs1||−|| || || ||SL|
|RT593||Δage1Δgcs1||+++||+++||++||−||FS, weak CS|
|Δage2Δgcs1||−|| || || ||SL|
|Δglo3Δgcs1||−|| || || ||SL|
|RT596||Δarf1Δage1Δglo3||+++||++||−||+||CS, FS, weak TS|
|Δarf1Δage1Δgcs1||−|| || || ||SL|
|RT597||Δarf1Δage2Δglo3||+++||+||−||−||TS, CS, FS|
|Δarf1Δage2Δgcs1|| *|| || || ||Predicted SL|
|Δarf1Δglo3Δgcs1|| *|| || || ||Predicted SL|
|RT598||Δage1Δage2Δglo3||+++||+++||−||−||CS, FS, no longer TS|
|Δage1Δage2Δgcs1||−|| || || ||SL|
|Δage1Δglo3Δgcs1|| *|| || || ||Predicted SL|
|Δage2Δglo3Δgcs1|| *|| || || ||Predicted SL|
The combination of single ARF GAP deletions with arf1::HIS3 (ARF2 is wild-type in all cases shown in Table 3) revealed synthetic defects with each of the ARF GAPs. Enhancement of the cold sensitivity was observed in all three viable GAP/arf1− knockouts. arf1−glo3− cells were more TS and the extent of invertase processing was more compromised (data not shown) relative to either deletion alone. Synthetic lethality with arf1− was only observed with gcs1−. It has been previously reported (Poon et al., 1996) that the combination of arf1− and gcs1− leads to much slower growth than either single knockout; a difference with our results that we attribute to differences in strain background.
Crosses were also performed to obtain each combination of double disruptions among the ARF GAPs. Two of these double deletions were synthetically lethal; gcs1−glo3− and gcs1−age2− (see Table 3; Poon et al., 1999). The consequences of combined deletions of the other combinations were less severe but at least one synthetic defect was observed in several cases. The exceptions to this were the age1−age2−, which had wild-type phenotypes for all tested parameters, and gcs1−age1− which grew well at 28° C and 37° C and had the same weak CS and strong fluoride sensitivity seen with gcs1− alone. The strongest synthetic enhancement seen was the strong TS of age1−glo3−. Synthetic defects between genes that encode proteins with a common biochemical activity are often interpreted as evidence of functional redundancy of the gene products. In that case, we see that GCS1 and both GLO3 and AGE2 would have a level of functional redundancy; as would GLO3 with AGE1 and AGE2, which is consistent with published results (Poon et al., 1999, 2001). The fluoride sensitivity varied between different age2−glo3− segregants, presumably because of strain background differences, and are thus inconclusive. Somewhat surprisingly, given the effects of combining single GAP deletions with the arf1−, we detected no more synthetic defects when the double GAP deletions were combined with the ARF1 deletion (see Table 3).
Finally, the triple ARF GAP deletions were constructed and tested, although three of the four possible combinations were predicted to be lethal, based on the results from the double deletions. Surprisingly, the only viable triple deletion (age1−age2−glo3−) showed no evidence of compromised growth rate at elevated temperature, even though the double age1−glo3− was TS. Although it is likely that this difference arose from background differences in our strains, it is possible that deletion of AGE2 results in suppression of the TS seen in age1−glo3− strains. The survival of the triple null strain may also point to the importance of GCS1 for mitotic growth or could be a reflection of its ability to overlap functionally with more ARF GAP functions than any of the other three proteins. It is also proof that Gcs1p alone can carry out all of the essential functions of the four ARF GAPs. This triple deletion will likely be useful in future studies requiring a simpler system in which to study ARF GAP functions.
Protein expression levels of the four ARF GAPs
One possibility that could explain the apparent importance of GCS1 relative to the other ARF GAPs is if there is simply more Gcs1p present in cells. To determine the expression levels of the four GAPs, two new alleles of each were created in which the genomic copy of the ARF GAPs were fused at their C-termini with either GFP or a tandem, triple HA tag. Yeast lysate protein preparations from each strain were analysed by immunoblot, using the appropriate tag-specific antibody. Results were the same for the HA- and GFP-tagged (data not shown) strains and revealed large differences in the cellular levels of the ARF GAPs. Gcs1p is expressed to the highest level and Glo3p is present at the next highest level, estimated at ∼50% that of Gcs1p. In contrast, tagged Age1p and Age2p were barely detectable in yeast lysates and required much longer exposures to be visualized (data not shown). Indeed, the HA antibody was more sensitive than that for GFP, so we could not detect Age1p–GFP or Age2p–GFP by immunoblotting. Thus, the rank order of protein expression levels in cells was determined to be Gcs1p > Glo3p ≫ Age2p ∼ Age1p.
Effects of deletions on invertase processing, uptake of FM4-64, and vacuolar morphology
The covalent addition of carbohydrate groups to the secreted form of invertase as it transits the secretory pathway is used as a marker for the integrity of that pathway. Deletion of ARF1 has previously been shown to result in a partial loss of invertase processing (Stearns et al., 1990). The hallmark of the most severe defect is that seen with sec18-1 cells at the restrictive temperature, which results in accumulation of invertase in the ER as the core glycosylated form. Each of the strains carrying single or combined deletions in ARF1 and the ARF GAP-encoding genes were assayed for invertase processing and many were found to have defects (summarized in Table 4). The subtlest was with age1−, in which the mobility of the invertase is often only slightly faster than in wild-type cells. The most severe defects were seen in strains carrying deletions in GLO3, either when combined with arf1− or when combined with age1− or age2−. We interpret these findings as evidence that Glo3p likely plays a more important role than the other ARF GAPs in the early parts of the secretory pathway. Support for this conclusion was also described previously (Poon et al., 1996; Dogic et al., 1999).
Table 4. Summary of effects of ARF1 and ARF GAP deletions on assays of invertase processing, uptake of FM4-64, and vacuolar morphology
|Strain||Invertase processing||Uptake of FM4-64 at 30 min||Uptake of FM4-64 at 60 min||Vacuole morphology|
|sec18-1 at 30° C||++++||N.D.||N.D.||N.D.|
|sec18-1 at 37° C||+||N.D.||N.D.||N.D.|
We also assayed each of our strains for the ability to internalize FM4-64 and, at later times, for the morphology of the vacuole. Uptake was scored at 30 and 60 min and the morphology of the vacuole was determined after the 60 min time point, as described in Materials and methods. Uptake of FM4-64 was delayed in a number of strains, particularly in those with a deletion of GCS1 or when any of the ARF GAP genes were deleted in combination with ARF1 (see Figure 6, Table 4). Among the double ARF GAP deletions there is a correlation between the absence of Age1p and a delay in uptake of the dye. The most severe defects were found when deletion of GLO3 was combined with those for ARF1 and either AGE1 or AGE2. Somewhat surprisingly, the changes in vacuolar morphology were not tightly correlated with the delay in endocytosis (Table 4). For example, we observed aberrant vacuoles in the arf1− strain, in which uptake of FM4-64 was normal, and normal-looking vacuoles in several strains in which uptake was still not completed after 60 min. In two other strains, gcs1− and age1−arf1−, uptake was only delayed and the vacuole was vesiculated. In three of the six strains exhibiting aberrations in vacuolar morphology there was a deletion in AGE1. To generalize from the trends seen, we would predict that GLO3 has its most dramatic impact on the early secretory pathway, GCS1 and AGE2 on the endocytic pathway, and AGE1 on traffic between the vacuole and other sites. These assignments are clearly only estimates, as the nature of membrane traffic is such that secondary or indirect effects often occur.
Figure 6. FM4-64 staining in the ARF1 and ARF GAP deletion strains. Uptake of FM4-64 was assayed as described under Materials and methods. Cells were incubated with 40 µM FM4-64 on ice for 30 min. The cells were then incubated with shaking at 30° C for 60 min. (A) PSY315 (WT); (B) YZC133 (age1−); (C) RT586 (age2−); (D) RT591 (age1−age2−); (E) RT596 (age1−glo3−arf1−); and (F) RT597 (age2−glo3−arf1−). Note that single, large vacuoles predominate but occasional bilobed structures are observed in wild-type and similar strains. Many single, large vacuoles are still seen in cells that are deficient in uptake (E and F)
Download figure to PowerPoint
- Top of page
- Materials and methods
Although six yeast genes encode proteins containing cysteine-rich ARF GAP-like domains, two of them (SPS18 and GTS1) lack SAT activity and the histidines that are required for GAP activity in Age1p (Zhang et al., 1998). Thus, the four proteins now shown to possess ARF GAP activity may represent all of the ARF GAPs in this organism and form a subset of the ARF effectors in yeast that likely have roles in a number of processes previously implicated to be under regulation by ARFs, most notably regulation of the secretory and endocytic pathways. By way of contrast, there are 22 predicted proteins in the human proteome with Arf GAP domains, most of which have been documented to have the activity.
Research on ARF functions has been slowed by the complexity of ARF signalling and the multiple, essential roles for ARFs. By examining in detail this subset of functionally and structurally related ARF GAPs, we simplify the analysis to such an extent that pathways and a clearer vision of the interrelationships can emerge. The importance and conservation in evolution of SAT and ARF GAP activities were also demonstrated with a mammalian ARF GAP that retained each activity. Thus, like ARF itself, at least some functions of ARF GAPs are well conserved through evolution. The analysis of each ARF GAP in yeast is described below.
A number of results suggest that Gcs1p may play a more critical or general role in cellular signalling than the other ARF GAPs. GCS1 is the only ARF GAP that supports cell viability when the remaining three ARF GAPs are deleted. Also, the GCS1 deletion is synthetically lethal with arf1−, and with two of the three other ARF GAPs. We interpret the former as evidence that Gcs1p is involved in at least one essential function of the ARFs. Furthermore, the evidence of multiple essential genetic interactions between GCS1 and other components described here point to a more central role for Gcs1p. It appears that GCS1 plays a more prominent role in endocytosis and traffic to the vacuole than in the early secretory pathway as the single mutant, gcs1−, causes a delay in the uptake of FM4-64 and multivesiculated vacuoles but little or no change in the processing of invertase (see Table 4). The defect in endocytosis observed in our strain background was more subtle than that reported previously for gcs1− (Wang et al., 1996), in which uptake was still deficient after 5 h. All cells that displayed the multivesiculated vacuolar morphology had deletions in either ARF1 or GCS1. We favour the simplest interpretation that predicts a role for GCS1 in vacuole traffic and morphological integrity. Its higher expression levels relative to the other three ARF GAPs could account for the predominance of GCS1 among the ARF GAPs. However, the explanation is probably not this simple, as overexpression of the other ARF GAPs from high-copy plasmids does not rescue all of the phenotypes of the gcs1− strain or of various combination deletions (data not shown).
Deletion of GLO3 revealed genetic interactions with each of the other ARF GAPs and with arf1−. Our results on glo3− strains agree completely with those of Poon et al. (1999) with regard to the CS of the single deletion, the synthetic defects when combined with arf1− and synthetic lethality with gcs1−. The most severe phenotype related to glo3− was synthetic lethality with gcs1−, but conditional phenotypes were also seen when combined with age1− (TS) or age2− (FS). While glo3− is a weak TS and stronger CS, the double glo3−age1− is a strong TS. Interestingly, increasing the gene dosage of GCS1 was able to suppress the resulting TS, but not the CS of glo3−age1− cells. We take this as evidence of a close functional relationship between GLO3 and GCS1, as GLO3 on a 2µ plasmid can also suppress the synthetic lethality of the double gcs1−age2−. A synthetic defect in the processing of invertase was seen for glo3− when combined with the arf1 null. This was the most severe effect seen on invertase processing and suggests a more important role for GLO3 in the early secretory pathway than the other ARF GAPs. The importance of GLO3 to normal processing of invertase and role in the early secretory pathway have been described previously (Dogic et al., 1999; Poon et al., 1999) and provided evidence of a more specific role for Glo3p in retrieval of ER proteins from the Golgi, or retrograde traffic.
Given that AGE1 was the most potent of the suppressors of both the arf1-3 TS and the lethality resulting from overexpression of ARF1, it was surprising to find that the deletion of AGE1 had relatively subtle consequences to cells. Though the CS seen when combined with arf1− was weak, a more severe CS and TS were found with the double age1−glo3−. Short of the synthetic lethalities observed with combined GCS1 knockouts, this was the most severe phenotype arising from a double deletion in ARF GAPs and points to functional overlap between AGE1 and GLO3. In addition, though the single deletion of AGE1 showed no defects in membrane traffic, a consistent pattern of defects were observed (see Table 4) when age1− was combined with null alleles of other ARF GAPs. We infer from this a role for Age1p in the endocytic pathway that is subtle—likely a result of low protein levels.
The functional similarities between GCS1 and GLO3 have been commented on before (see above; Poon et al., 1999) but the relationship between AGE2 and GCS1 seems just as close. As already mentioned, the synthetic lethality between age2− and gcs1− can be suppressed by increasing the copy number of GLO3. There is also a synthetic super sensitivity to fluoride between age2− and glo3−. Thus, the triplet of GCS1, GLO3 and AGE2 may have more extensive functional overlap than any with AGE1.
Because it had the highest activity as suppressor of arf1-3 and lethality from overexpression of Arf1p and it had not yet been characterized biochemically, we performed a more extensive characterization of Age1p. We found a good correlation between SAT activity and ARF GAP activity in a series of deletion mutants of Age1p. The importance of ARF GAP activity to SAT activity is also evident from the limited homology among the four SAT proteins outside of the ARF GAP domain, and from the finding of SAT activity in a fragment containing a minimal mammalian ARF GAP (1–136) domain. Analysis of the deletion mutants of Age1p led us to propose the existence of four functional regions of the Age1p protein. The N-terminal extension is longer in Age1p than other ARF GAPs. We believe the ≈160 residue N-terminal region is likely to make specific protein contacts and may serve an inhibitory role in the regulation of the SAT and/or ARF GAP activities (C. J. Zhang and R. A. Kahn, unpublished observations). The ARF GAP domain is the most highly conserved and only functionally defined region of the protein. The third domain is also important to both SAT and ARF GAP activities and may be required for proper folding of the GAP domain or the formation of critical contacts with other proteins.
The greater stability and solubility of the Age1Δ5p truncation mutant allowed its characterization as an ARF GAP and tests for specificity amongst members of the ARF family in in vitro assays. Evidence of overlap between mammalian ARFs and ARLs in binding some of the same effectors (Van Valkenburgh et al., 2001) adds to the importance of distinguishing between potential ARF and ARL GAPs. Our results clearly indicate that Age1Δ5p is active across a wide range of ARF proteins, both human and yeast, but is completely inactive when presented with a similar range of ARLs.
The evidence of functional overlap between the ARF GAPs is somewhat surprising, given that the sequence homologies between the ARF GAPs are quite limited outside of the ARF GAP domain. When taken together with the observation that a mammalian ARF GAP fragment of 136 residues also possesses both ARF GAP and SAT activities it seems probable that the ARF GAP domain is the key feature in both activities. However, we (see above) and others (Ireland et al., 1994; Poon et al., 1996) have noted that residues well outside the ARF GAP domain are important for activities of the Age1p and Gcs1p proteins, respectively. One model that can explain these observations is similar to that first proposed by Goldberg (1999) after noting the potential for activated ARF to bind both a GAP and an effector simultaneously. We propose that ARFs may function to regulate the assembly of multi-subunit complexes, e.g. coat complexes, through the recruitment of several key components including the means of its signal termination, the ARF GAP. In binding both to ARF and to other components in the complex, the ARF GAP may serve dual roles as effectors and terminators of the ARF signal. Such a model would help explain our results, which clearly point to effector functions for the ARF GAPs. In any case, it is clearly insufficient to think of ARF GAPs as solely terminators of ARF signals. The phenotypes described herein for deletions of the four ARF GAPs in yeast provide many opportunities for further genetic screens that will likely prove helpful in continuing the dissection of this complex signalling network and provide further insight into mechanisms.