Cell, Vol. 70, 69-79,
July IO, 1992, Copyright
0 1992 by Cell Press
Two Distinct Members of the ADP-Ribosylation Factor Family of GTP-Binding Proteins Regulate Cell-Free Intra-Golgi Transport Timothy C. Taylor,’ Richard A. Kahn,t and Paul Melancon” “Department of Chemistry and Biochemistry University of Colorado Boulder, Colorado 80309-0215 tLaboratory of Biological Chemistry Developmental Therapeutics Program Division of Cancer Treatment National Cancer Institute Bethesda, Maryland 20892
Summary We have used an intra-Golgi transport assay to identify GTP-binding proteins involved in regulation of protein traffic. Two soluble proteins of 20 kd were purified by their ability to mediate GTPyS-dependent inhibition of transport. These GTP-dependent Golgi binding factors, or GGBFs, exhibit a 3-fold difference in activity and are differentiated by their hydrophobicity, isoelectric points, and apparent size. Removal of 80% of GGBFs from cytosol abolishes GTPyS sensitivity but does not affect inhibition by aluminum fluoride. We demonstrate that GGBFs are members of the ADPribosylation factor (AM) family. Recombinant ARFl exhibits GGBF activity and myristoylation is required. The distinct biochemical properties of GGBFs indicate that members of the ARF family may have related but distinct functions in intracellular transport. Introduction A variety of genetic and biochemical approaches have established a fundamental role for various classes of GTP-binding proteins in the regulation of intracellular protein traffic. The first indication came from genetic analysis of a Saccharomyces cerevisiae mutant, sec4-8, which accumulates post-Golgi secretory vesicles at the nonpermissive temperature. Analysis of SEC4 revealed the presence of consensus sequences highly conserved among GTPbinding proteins (Salminen and Novick, 1987). Several other genes involved in protein secretion have since been shown to encode small GTP-binding proteins (Segev et al., 1988; Nakano and fvluramatsu, 1989). A biochemical dissection of intracellular transport reactions has been made possible by the development of cell-free assays that reconstitute most types of intracellular transport (Goda and Pfeffer, 1989; Rothman and Orci, 1992). The sensitivity of these assays to nonhydrolyzable analogs of GTP and to aluminum fluoride has provided further evidence that GTP-binding proteins are involved in the formation and fusion of transport vesicles (Bourne, 1988; Balch, 1990). Among the GTP-binding proteins identified thus far are two superfamilies, represented by the ras and heterotrimerit G proteins. Members of the ras superfamily, such as rabs/sec4/yptl, ADP-ribosylation factors (ARFs), and
Sarl p, form distinct subfamilies, each containing a large numberof proteinsof related sequences. However, little is known about their function. Current information suggests that each step of intracellular transport may involve one or more members of each of these subfamilies of GTPbinding proteins (Melancon et al., 1991; Pfeffer, 1992). Genetic and biochemical experiments have demonstrated that rab proteins play an essential role in several intracellular transport reactions (for review see Pfeffer, 1992). Distinct rab proteins are associated with nearly everyorganelle, suggesting that they function in the selective targeting and/or attachment of carrier vesicles to their acceptor sites (Chavrier et al., 1990; Balch, 1990; Pfeffer, 1992). rab proteins are primarily membrane associated, and polyisoprenylation at a C-terminal cysteine residue is essential for both activity and membrane association (see Pfeffer, 1992). Additional primary structure near the C-terminus appears to determine the specific subcellular localization of rab proteins (Chavrier et al., 1991). The ARF and Sarl proteins are also implicated in protein traffic. The domains involved in GTP binding and hydrolysis in these proteins are more closely related to those of heterotrimeric G proteins (involved in signal transduction) than to those of the ras superfamily (Sewell and Kahn, 1988). Furthermore, neither ARF nor Sarlp is modified with an isoprenyl group, but ARF is myristoylated at an N-terminal glycine (Kahn et al., 1988). ARF is predominantly cytosolic whereas Sarlp is tightly membrane associated (Nishikawa and Nakano, 1991). The different properties of ARF and Sarl p suggest distinct roles for these proteins in the budding process. In S. cerevisiae, Sarl p is required to form vesicles involved in endoplasmic reticulum-Golgi transport(Okaetal., 199l;d’Enfertet al., 1991) but it is absent from the vesicles themselves (R. Schekman, personal communication). Genetic studies, also in S. cerevisiae, have demonstrated that reduction in the ARF level slows down secretion of invertase (Stearns et al., 199Oa). ARF function is essential, since disruption of both ARF genes is lethal (Stearns et al., 1990b). In animal cells, ARF localizes primarily to the Golgi complex (Stearns et al., 199Oa) and is present on non-clathrincoated vesicles that accumulate in the presence of GTPyS (Serafini et al., 1991). Several ARF genes have been identified using recombinant DNA methods (Kahn et al., 1991). Evidence for the involvement of ARFs in budding comes from the demonstration that synthetic peptides derived from the N-terminus of ARFl block formation of coated vesicles from Golgi membranes (Kahn et al., 1992). Biochemical analysis of intra-Golgi transport was initiated by the development of a cell-free assay by J. E. Rothman and his colleagues (Balch et al., 1984). This assay measures the transport-coupled glycosylation of a glycoprotein (VSV-G) encoded by vesicular stomatitis virus (VSV). In brief, transport is measured as the glycosylation of VSV-G when it is transferred from mutant (“donor”) Golgi membranes (i.e., membranes lacking the enzyme N-acetyl-glucosaminyl [GlcNAc] transferase 1) to wild-type
Cell 70
(“acceptor”) membranes (i.e., membranes lacking VSV-G but containing the transferase activity). Transport reactions contain UDP-[3H]GlcNAc, a radiolabeled substrate for the transferase. The extent of 3H incorporation into VSV-G, a measure of transport, is then quantitated by immunoprecipitation at the end of the reaction. GTP-binding proteins are involved in intra-Golgi transport, since aluminum fluoride and nonhydrolyzable analogs of GTP inhibit this cell-free assay (Melancon et al., 1987). Here we establish that the GTP-binding proteins responsible for the GTPyS-dependent inhibition of the Golgi cellfree assay are soluble and not located on Golgi membranes. Two GTP-binding proteins were purified to homogeneity on the basis of their ability to confer GTPyS sensitivity on the cell-free assay. Both proteins are recognized by an antibody to ARFl. However, these ARF proteins differ up to 3-fold in their specific inhibitory activity. These results suggest that different ARFs may function in related but distinct aspects of intracellular transport, such as assembly of different types of coated structure. We further establish that rab and heterotrimeric G proteins, if they bind GTPyS under our conditions, do not contribute significantly to any GTPyS-dependent inhibition of the intra-Golgi transport reaction.
inhibition of Transport by GTPyS Requires Soluble GTP-Binding Proteins Previous work showed that the Golgi transport-assay is inhibited by a submicromolar concentration of GTPyS (Melancon et al., 1987). The extent of inhibition is determined by the ratio of soluble proteins to membrane proteins and not by the amount of cytosol present (Melancon et al., 1989). This suggests that a soluble protein, present in limiting amounts in cytosol, is required for irreversible inactivation of Golgi membranes. Other experiments established that this soluble factor associates tightly with membranes in the presence of GTPyS (Melancon et al., 1987). For this reason the factor has been termed GTPdependent Golgi-binding factor, or GGBF. The GTPySdependent inhibition was termed “GGBF activity.” Two possibilities for GGBF activity were considered: GGBF is a factor required for activation of a membraneassociated GTP-binding protein(s), or it is itself a GTPbinding protein. Membranes and cytosol were incubated separately with GTPrS to determine whether a GTP+sensitive protein resideson membranesor in cytosol. First, donor and acceptor Golgi membranes were incubated separately at low cytosol concentration with 20 uM GTPrS at 37°C for 15 min. Membranes were then recovered by centrifugation andassayed in the presenceof fresh cytosol and excess GTP. Preincubation with GTPyS had no effect on the activity of either donor or acceptor (data not shown). This indicated that either GTPyS does not bind significantly to membrane-associated GTP-binding proteins under these conditions or that such binding confers no inhibitory effect. To test whether cytosol contains a GTPyS-sensitive protein, cytosol was preincubated with GTPyS at either 0%
or 30% (see Experimental Procedures). Nucleotides were then removed from cytosol by chromatography. Cytosol samples were tested in the transport assay to determine which nucleotides were bound by cytosolic proteins during incubation in the first step. We found that preincubation with GTPyS at 30°C causes a significant decrease in transport (data not shown). This indicated that GTPyS binds a cytosolic target that inhibits the transport assay. The addition of excess GTP prevented the inhibition, and no effect of GTPyS was observed when the preincubation was carried out at 0%. The effect is guanine specific, since preincubation with ATPrS did not reduce transport significantly (data not shown). These experiments suggested that at least one cytosolic factor required for the inhibition by GTPyS, possibly GGBF, is a soluble GTPbinding protein. .: Prepa’ration of Cytosol Lacking GGBF Activity Cytosol provides a large number of factors that are essential to observe transport in the cell-free assay. An assay for GGBF was developed by preparing a cytosol that provided sufficient amounts of these transport factors yet exhibited a reduced level of GGBF activity. We took advantage of the fact that GGBF binds membranes very tightly in the presence of GTPyS. Whole CHO cell homogenates were incubated with GTPyS under conditions optimized for binding of GGBF to membranes, as described in Experimental Procedures. Following incubation, the homogenate was centrifugei to remove membranes and associated factors. The high-speed supernatant was desalted by chromatography to remove nucleotides and other small metabolites and then used in the transport assay. Figure 1 demonstrates that cytosol obtained from GTPyS-treated CHO homogenates (GTPrS-CHO cytosol) can be used to support transport reactions even though it shows only limited sensitivity to GTPyS. In Figure lA, titratlons of GTPyS-CHO cytosol and cytosol prepared from mock-treated homogenates (CHO) are compared in their ability to promote transport (open symbols). The two cytosols are indistinguishable even at the lower levels at which soluble factors are limiting. Further analysis of these cytosols showed that the kinetics of transport in reactions containing CHO and GTPyS-CHO cytosols are identical (data not shown). In contrast, titrations performed in the presence of 4 uM GTPrS show a clear difference between the two cytosols (closed symbols). Whereas an increase in CHO cytosol clearly leads to greater sensitivity to GTPrS (compare open and closed triangles), no such effect is observed with GTPyS-CHO cytosol (compare open and closed circles). This reduced GTPyS sensitivity indicates that GGBF was largely removed from the GTPyS-CHO cytosol by our procedure. These observations also suggest that GGBF, if required for the reaction, is not present at a limiting concentration in cytosol. Owing to the cumbersome nature of the data in Figure lA, GGBF activity was redefined in terms of a ratio: a GGBF ratio is calculated for each titration point by dividing the transport observed in the presence of GTPrS by that observed in the absence of GTPyS. Figure 1B shows a GGBF ratio plot of the data presented in Figure 1A. Note
Regulation 71
of Intra-Golgi
Traffic
by ARFs
A. 2000
iz 8
1500
lz g
1000
E E +
500
n
8.
1.0
0.6 g d
0.6
L 8
o.4 0.2
0.0
0
20
CYtosol
40
00
Mf)
Figure 1. Comparison of Transport Activities of Cytosols Prepared from Mock-Treated and GTPyS-Treated CHO Cell Homogenates (A) A series of transport assays were performed in the presence of increasing amounts of either mock-treated CHO cytosol (triangles) or GTPyS-CHO cytosol (circles), as described in Experimental Procedures. Reactions were carried out in the absence (open symbols) or presence (closed symbols) of 4 uM GTPTS. The extent of transport in 2 hr incubations is plotted as a function of the content of cytosol. (8) GGBF ratio plots of the data presented in (A). For each pair of points in the titrations shown in (A) (without and with GTPyS), a GGBF ratio was calculated by dividing the extent of transport measured in the presence of GTPyS by that measured in the absence of GTP$S. The four curves in (A) are represented more clearly as two GGBF ratio curves.
that a low GGBF ratio indicates a high level of GGBF activity. The sensitivity to GTPyS can be restored to transport reactions containing GTPrS-CHO cytosol when bovine brain cytosol (BBC) is added (data not shown). This established that GTPyS-CHO cytosol could be used in transport reactions to assay for GGBF in fractions from various purification procedures. This experiment also demonstrated that BBC, a readily available and abundant source of cytosol, could be used as starting material for the purification of GGBF. The modified intra-Golgi transport assay containing saturating amounts of GTPyS-CHO cytosol will be referred to as the GGBF assay. Loss of GGBF Activity Correlates with a Loss of GTP-Binding Proteins of 20 kd The removal of GGBF activity from cytosol described
above prompted us to investigate the content of GTPbinding proteins remaining in the CHO and GTPyS-CHO cytosols. Using [35S]GTPyS and a filter-binding method (Kikuchi et al., 1988), we found that the level of GTPbinding proteins present in GTPyS-CHO was only slightly reduced compared with CHO (data not shown). However, as shown in Figure 2, aclear difference was revealed upon analysis of the GTP-binding proteins by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and GTP ligand blotting. Comparison of CHO and GTPyS-CHO cytosols shows a depletion of GTP-binding proteins migrating at ~20 kd in the GTPyS-CHO. The association of the 20 kd cytosolic GTP-binding proteins with membranes occurs with a tH of 3 min at 37°C and is not significant at 0°C (data not shown). Quantitation of GTP bound in several blots indicates that 70%-80% of the 20 kd GTP-binding proteins are absent from GTPyS-CHO cytosol. In contrast, the other major small (24 kd-30 kd) GTP-binding proteins in this cytosol are apparently unaffected by the procedure. This agrees with the filter binding results mentioned above. We found that when BBC is fractionated by sizeexclusion chromatography, GGBF activity is found only in fractions containing proteins smaller than 25 kd (data not shown). Furthermore, these fractions contain GTPbinding proteins of ~20 kd exclusively. These observations, along with the fact that loss of the 20 kd GTP-binding proteins coincides with loss of GGBF activity, strongly suggest that GGBF is a 20 kd GTP-binding protein. Purification of Two 20 kd GTP-Binding Proteins Exhibiting GGBF Activity BBC was subjected to three consecutive purification steps to yield two preparations of purified GGBF (termed GGBF and GGBF*), which differ in their physical properties and in their specific GGBF activities. Fractions containing GGBF activity were identified at each step by their ability to restore GTPyS sensitivity in reactions containing GTPySCHO cytosol. Fractions were also analyzed for their content of 20 kd GTP-binding proteins by quantitating GTP ligand blots as described in Experimental Procedures.
CHO
“cTHp;J” Golgi
Figure 2. GTP Ligand Blot of CHO and CHO Golgi Membranes
Cytosol,
GTPTS-CHO
Cytosol,
CHO and GTPyS-CHO cytosols (36 ug each) and acceptor Golgi membrane proteins (18 ug) were separated by SDS-PAGE on a 15% gel and subjected to GTP ligand blot analysis as described in Experimental Procedures. The Golgi sample was run separately from the cytosols. An autoradiogram is shown. Molecular size markers are trypsin inhibitor (21.5 kd) and carbonic anhydrase (31 kd).
Cdl 72
BBC was loaded on an anion-exchange column (Cl Sepharose) and eluted with a linear KCI gradient. A single symmetric peak of activity eluted (Figure 3A) that was pooled and loaded on a hydrophobic interaction column (phenyl-Superose). Most of the GGBF activity elutes from this column in a peak corresponding to a minor fraction of the 20 kd GTP-binding proteins (Figure 3B). A minor peak of GGBF activity corresponding to the main 20 kd GTPbinding protein peak is also observed. This activity, labeled GGBF’, was pooled separately from the main GGBF
peak. Both GGBF and GGBF’ pools were further purified in parallel by size-exclusion chromatography. Figure 3C shows the size-exclusion (Superdex 75) elution profile of the GGBFpool from Figure3B. GGBFactivity elutes as a single peak. SDS-PAGE analysis of the fractions reveals a single protein of ~20 kd (Figure 4A), which binds GTP (Figure 48) and whose distribution coincides exactly with that of GGBF activity (see Figure 3C). Similar results were observed for the purification of GGBF’ (data not shown). Purified GGBF and GGBF’ fractions constitute approximately 0.02% and 0.05% of the total soluble protein from bovine brain, respectively, as estimated from the yield of protein in the purified fractions (Figure 3C) and GTP ligand blots. Attempts to sequence both forms of GGBF indicated that the N-terminus of both proteins is blocked (data not shown). Results of the purification are summarize&n Table 1. Overall recovery of GGBF activity as GGBF is < 2%. This poor yield may reflect the following: -2% of GGBF activity present in the starting material is recovered as GGBF’; denaturation of GGBF may have occurred during purification; or GGBF may have become separated from factors that potentiate GGBF activity as measured in whole cytosol.
Fraction
LD Fraction
Figure
3. Purification
of GGBF
5/8 9/12 13 14
15
16 17
B.
Activity
(A) Elution of GGBF activity from Q Sepharose. Diluted, desalted BBC in 20 mM KCI was applied to the column equilibrated with buffer lacking KCI. Proteins were eluted with a linear KCI gradient to 500 mM and 14 ml fractions were collected. GGBF activity was measured as described in Experimental Procedures. GTP-binding proteins of 20 kd were quantitated by Phosphorlmager analysis of GTP ligand blots. (B) Elution of GGBF activity from phenyl-Superose. Five milliliters of concentrated GGBF pooling from Q Sepharose in 1.1 M (NH&SO, was applied to the column equilibrated with buffer containing 1 .l M (NH&SO,. Proteins were eluted with a descending ammonium sulfate gradient and 1.5 ml fractions were collected. GGBF activity and 20 kd GTP-binding proteins were quantitated as in (A). The GGBF’ pooling is indicated by a bold asterisk. (C) Elution of protein and GGBF activity on Superdex 75. Five millititers of the GGBF pooling from phenyl-Supsrose was applied to the column and 3 ml fractions were collected. Molecular size standards are 66 kd for albumin and 25 kd for chymotrypsinogen.
Figure 4. Total Protein and GTP-Binding Fractions from GGBF Purification
Proteins
in Superdex-75
(A) A 15% SDS-polyacrlyamide gel was run and stained with Coomassie brilliant blue. Eighty microliters from each fraction (or 16 nl of each of fractions 5-6 and 9-12) was precipitated with 10% TCA and the pellets resolubilized in 30 nl of buffer. Samples were: 25 ul from the samples of the combined fractions 5-6 and 9-12 and 25 nl of samples of fractions 13-16 (the column load [LO] is also shown). Molecular size standards are phosphorylase b (97 kd), serum albumin (67 kd), ovalbumin (43 kd). carbonic anhydrase (31 kd), trypsin inhibitor (21.5 kd), and lysozyme (14 kd). (B) GTP ligand blot analysis of the fractions as in (A). Four microliters of each fraction was applied to a 15% gel and processed as described in Experimental Procedures. An autoradiogram of the region around MW of 20 kd is shown. GTP-binding proteins of a molecular size ~20 kd were the only ones present in the load and column fractions.
Regulation 73
Table
of Intra-Golgi
1. Purification
Traffic
of GGBF Protein
by ARFs
Activity
GGBF-Active Pooling
0x0
Concentration (mg/ml)
Percent Recovery’
BBC G-Sepharose Phenyl-Superose GGBFD GGBFD
2247 147 12.4 0.41 0.82
10.7 1.31 2.07 0.07 0.15
100 18 3.0 1.7 1.8
a Recovery of GGBF activity is reported as a percentage of the total volume of the pool required to observe I&, relative to the I& and volume of the BBC starting material. Activity of GGBF’ is reported only for the purified protein and is not included in the recoveries of GGBF. b The two fractions of greatest activity from Superdex 75.
GGBF Activity Is Neutralized by Brefeldin A The fungal metabolite brefeldin A prevents the inhibition of the intra-Golgi assay by GTPvS (Orci et al., 1991). Addition of 40 pM brefeldin A has only a small inhibitory effect on transport, but it negates the inhibition by GTPyS in reactions containing a high ratio of CHO cytosol to membranes. Therefore, we tested brefeldin A in GGBF assays containing an inhibitory concentration of GGBF. Figure 5 demonstrates that brefeldin A abolishes sensitivity of the GGBF assay to exogenous GGBF. Increasing amounts of brefeldin A were added to assays containing GTPyS-CHO cytosol and enough GGBF to bring the GGBF ratio from 1 .l (no GGBF) to 0.25. Following a brief incubation at 22OC, GTPyS was added to half of the reactions, and assays were carried out as before. Brefeldin A did not reduce transport in the GGBF assay (data not shown), in agreement with previous work (Orci et al., 1991). However, brefeldin A at >, 32 t.tM blocks GGBF activity maximally. This demonstrates that purified GGBF is sufficient to reproduce accurately the GGBF activity observed with whole cytosol. Different Forms of GGBF Have Different GGBF Activities Figure 6 shows a reverse-phase high performance liquid chromatography (HPLC) analysis of purified GGBFs. GGBF and GGBF’ elute as single peaks with no other significant protein peaks observed. Note that the GGBF preparation contains significant GGBF* (at a level of 25% [Figure 6AJ) and that GGBF is a 15% contaminant in the GGBF* preparation (Figure 6B). Isoelectric focusing reveals that GGBF is slightly more acidic than GGBF* with pls of 5.5 and 5.9, respectively (data not shown). These data demonstrate that neither GGBF preparation is resolved into more than two proteins and that the contaminant in each preparation is most likely the complementary GGBF. A quantitative analysis of the GGBF activity of GGBF and GGBF’ is presented in Table 2. The amount of protein required for 50% inhibition (IC&) of the GGBF assay is shown. GGBF exhibits an I& of 0.6 pg whereas that for GGBF” is 1.6 pg. This approximately 3-fold difference in specific activity represents a lower limit since GGBF contains 25% of the less active GGBF’, and much of the
arefeldin Figure
5. Brefeldin
A Prevents
A (PM)
GTPyS-Dependent
Inhibition
by GGBF
GGBF assays were carried out as described in Experimental Procedures, and each assay included 1.7 ug of purified GGBF (to give a GGBF ratio of 0.25). Brefeldin A was added to duplicate tubes at the indicated concentrations, and the tubes were incubated at 22W for 2 min. At this time, half of the reactions received GTPyS on ice to a final concentration of 4 PM. GGBF ratios are plotted as a function of brefeldin A concentration. The GGBF ratio in the absence of GGBF was 1.1.
activity of GGBF* may be contributed nation by GGBF.
by the 15% contami-
GGBF and GGBF’ Are Members of the ARF Family The size and cytosolic nature of GGBFs is similar to that of the GTP-binding protein ARF. ARF was first purified from bovine extracts as a cofactor for ADP-ribosylation of
Figure
6. HPLC
Analysis
of GGBF
and GGBF’
GGBF and GGBF’ pools from Superdex 75 chromatography were applied to an analytical reverse-phase HPLC column and eluted with linear gradient of acetonitrile as in Experimental Procedures. Two separate runs were made: GGBF (75 ug) (A) and GGBF’ (110 ug) (B). Absorbance was monitored at 260 nm. The region of the chromatograms from 54%-63% acetonitrile is shown. GGBF’ (asterisk) and GGBF elute in the gradient at apparent acetonitrile concentrations of 58% and 60%, respectively.
Cell 74
Table 2. GGBF
c
and ARF Activities
Protein
GGBF
GGBF GGBF’ Myr-ARFl ARFI
Q.6 1.6 2.4 >3.0
Activity” PB ps pg’ pg
ARF Activityb 12.9 11.5 6.7 ND
ARFl
ARF4
GGBF
llB9
’ Reported as the weight of protein to achieve IC, in GGBF assays as described in Experimental Procedures. ’ Reported in units of pmol of ADP-ribosylated G, per pmol of activated GGBF/ARF, determined as described in Experimental Procedures. ’ Assuming 50% of the ARF proteins present in the sample are myristoylated. Myr-ARFl and ARFl refer to recombinant proteins obtained from E. coli strains (Weiss et al., 1989) containing or lacking the gene for N-myristoyl transferase, respectively. ND: not determined.
B.
Figure 7. lmmunoreactivity Antibodies
G.. by cholera toxin (Kahn and Gilman, 1984). SDS-PAGE analysis indicated that ARF consisted of a doublet at 20 kd. To establish whether GGBFs are related to ARFs, we first examined the ARF content of GTPyS-CHO cytosol by immunoblot. The level of ARF was found to be reduced by approximately 70% (data not shown), consistent with the depletion of 20 kd GTP-binding proteins (see Figure 2). Figure 7 demonstrates that purified GGBFs are recognized by an ARF-specific antibody. GGBF and GGBF* were screened with antibodies against ARFl (ml D9) and ARF4 (R-891). Antibody ml D9 reacts strongly with ARFl and ARF3, which have 98% sequence identity (R. A. K., unpublished data). lmmunoreactivity wasmeasured relative to that of purified recombinant ARFl and ARF4. Figure 7A shows that both forms of GGBF react with mlD9. GGBFs react as well as ARFl , whereas ARF4 reactivity is less than 20%. In contrast, GGBFs do not react significantly (
July IO, 1992, Copyright
0 1992 by Cell Press
Two Distinct Members of the ADP-Ribosylation Factor Family of GTP-Binding Proteins Regulate Cell-Free Intra-Golgi Transport Timothy C. Taylor,’ Richard A. Kahn,t and Paul Melancon” “Department of Chemistry and Biochemistry University of Colorado Boulder, Colorado 80309-0215 tLaboratory of Biological Chemistry Developmental Therapeutics Program Division of Cancer Treatment National Cancer Institute Bethesda, Maryland 20892
Summary We have used an intra-Golgi transport assay to identify GTP-binding proteins involved in regulation of protein traffic. Two soluble proteins of 20 kd were purified by their ability to mediate GTPyS-dependent inhibition of transport. These GTP-dependent Golgi binding factors, or GGBFs, exhibit a 3-fold difference in activity and are differentiated by their hydrophobicity, isoelectric points, and apparent size. Removal of 80% of GGBFs from cytosol abolishes GTPyS sensitivity but does not affect inhibition by aluminum fluoride. We demonstrate that GGBFs are members of the ADPribosylation factor (AM) family. Recombinant ARFl exhibits GGBF activity and myristoylation is required. The distinct biochemical properties of GGBFs indicate that members of the ARF family may have related but distinct functions in intracellular transport. Introduction A variety of genetic and biochemical approaches have established a fundamental role for various classes of GTP-binding proteins in the regulation of intracellular protein traffic. The first indication came from genetic analysis of a Saccharomyces cerevisiae mutant, sec4-8, which accumulates post-Golgi secretory vesicles at the nonpermissive temperature. Analysis of SEC4 revealed the presence of consensus sequences highly conserved among GTPbinding proteins (Salminen and Novick, 1987). Several other genes involved in protein secretion have since been shown to encode small GTP-binding proteins (Segev et al., 1988; Nakano and fvluramatsu, 1989). A biochemical dissection of intracellular transport reactions has been made possible by the development of cell-free assays that reconstitute most types of intracellular transport (Goda and Pfeffer, 1989; Rothman and Orci, 1992). The sensitivity of these assays to nonhydrolyzable analogs of GTP and to aluminum fluoride has provided further evidence that GTP-binding proteins are involved in the formation and fusion of transport vesicles (Bourne, 1988; Balch, 1990). Among the GTP-binding proteins identified thus far are two superfamilies, represented by the ras and heterotrimerit G proteins. Members of the ras superfamily, such as rabs/sec4/yptl, ADP-ribosylation factors (ARFs), and
Sarl p, form distinct subfamilies, each containing a large numberof proteinsof related sequences. However, little is known about their function. Current information suggests that each step of intracellular transport may involve one or more members of each of these subfamilies of GTPbinding proteins (Melancon et al., 1991; Pfeffer, 1992). Genetic and biochemical experiments have demonstrated that rab proteins play an essential role in several intracellular transport reactions (for review see Pfeffer, 1992). Distinct rab proteins are associated with nearly everyorganelle, suggesting that they function in the selective targeting and/or attachment of carrier vesicles to their acceptor sites (Chavrier et al., 1990; Balch, 1990; Pfeffer, 1992). rab proteins are primarily membrane associated, and polyisoprenylation at a C-terminal cysteine residue is essential for both activity and membrane association (see Pfeffer, 1992). Additional primary structure near the C-terminus appears to determine the specific subcellular localization of rab proteins (Chavrier et al., 1991). The ARF and Sarl proteins are also implicated in protein traffic. The domains involved in GTP binding and hydrolysis in these proteins are more closely related to those of heterotrimeric G proteins (involved in signal transduction) than to those of the ras superfamily (Sewell and Kahn, 1988). Furthermore, neither ARF nor Sarlp is modified with an isoprenyl group, but ARF is myristoylated at an N-terminal glycine (Kahn et al., 1988). ARF is predominantly cytosolic whereas Sarlp is tightly membrane associated (Nishikawa and Nakano, 1991). The different properties of ARF and Sarl p suggest distinct roles for these proteins in the budding process. In S. cerevisiae, Sarl p is required to form vesicles involved in endoplasmic reticulum-Golgi transport(Okaetal., 199l;d’Enfertet al., 1991) but it is absent from the vesicles themselves (R. Schekman, personal communication). Genetic studies, also in S. cerevisiae, have demonstrated that reduction in the ARF level slows down secretion of invertase (Stearns et al., 199Oa). ARF function is essential, since disruption of both ARF genes is lethal (Stearns et al., 1990b). In animal cells, ARF localizes primarily to the Golgi complex (Stearns et al., 199Oa) and is present on non-clathrincoated vesicles that accumulate in the presence of GTPyS (Serafini et al., 1991). Several ARF genes have been identified using recombinant DNA methods (Kahn et al., 1991). Evidence for the involvement of ARFs in budding comes from the demonstration that synthetic peptides derived from the N-terminus of ARFl block formation of coated vesicles from Golgi membranes (Kahn et al., 1992). Biochemical analysis of intra-Golgi transport was initiated by the development of a cell-free assay by J. E. Rothman and his colleagues (Balch et al., 1984). This assay measures the transport-coupled glycosylation of a glycoprotein (VSV-G) encoded by vesicular stomatitis virus (VSV). In brief, transport is measured as the glycosylation of VSV-G when it is transferred from mutant (“donor”) Golgi membranes (i.e., membranes lacking the enzyme N-acetyl-glucosaminyl [GlcNAc] transferase 1) to wild-type
Cell 70
(“acceptor”) membranes (i.e., membranes lacking VSV-G but containing the transferase activity). Transport reactions contain UDP-[3H]GlcNAc, a radiolabeled substrate for the transferase. The extent of 3H incorporation into VSV-G, a measure of transport, is then quantitated by immunoprecipitation at the end of the reaction. GTP-binding proteins are involved in intra-Golgi transport, since aluminum fluoride and nonhydrolyzable analogs of GTP inhibit this cell-free assay (Melancon et al., 1987). Here we establish that the GTP-binding proteins responsible for the GTPyS-dependent inhibition of the Golgi cellfree assay are soluble and not located on Golgi membranes. Two GTP-binding proteins were purified to homogeneity on the basis of their ability to confer GTPyS sensitivity on the cell-free assay. Both proteins are recognized by an antibody to ARFl. However, these ARF proteins differ up to 3-fold in their specific inhibitory activity. These results suggest that different ARFs may function in related but distinct aspects of intracellular transport, such as assembly of different types of coated structure. We further establish that rab and heterotrimeric G proteins, if they bind GTPyS under our conditions, do not contribute significantly to any GTPyS-dependent inhibition of the intra-Golgi transport reaction.
inhibition of Transport by GTPyS Requires Soluble GTP-Binding Proteins Previous work showed that the Golgi transport-assay is inhibited by a submicromolar concentration of GTPyS (Melancon et al., 1987). The extent of inhibition is determined by the ratio of soluble proteins to membrane proteins and not by the amount of cytosol present (Melancon et al., 1989). This suggests that a soluble protein, present in limiting amounts in cytosol, is required for irreversible inactivation of Golgi membranes. Other experiments established that this soluble factor associates tightly with membranes in the presence of GTPyS (Melancon et al., 1987). For this reason the factor has been termed GTPdependent Golgi-binding factor, or GGBF. The GTPySdependent inhibition was termed “GGBF activity.” Two possibilities for GGBF activity were considered: GGBF is a factor required for activation of a membraneassociated GTP-binding protein(s), or it is itself a GTPbinding protein. Membranes and cytosol were incubated separately with GTPrS to determine whether a GTP+sensitive protein resideson membranesor in cytosol. First, donor and acceptor Golgi membranes were incubated separately at low cytosol concentration with 20 uM GTPrS at 37°C for 15 min. Membranes were then recovered by centrifugation andassayed in the presenceof fresh cytosol and excess GTP. Preincubation with GTPyS had no effect on the activity of either donor or acceptor (data not shown). This indicated that either GTPyS does not bind significantly to membrane-associated GTP-binding proteins under these conditions or that such binding confers no inhibitory effect. To test whether cytosol contains a GTPyS-sensitive protein, cytosol was preincubated with GTPyS at either 0%
or 30% (see Experimental Procedures). Nucleotides were then removed from cytosol by chromatography. Cytosol samples were tested in the transport assay to determine which nucleotides were bound by cytosolic proteins during incubation in the first step. We found that preincubation with GTPyS at 30°C causes a significant decrease in transport (data not shown). This indicated that GTPyS binds a cytosolic target that inhibits the transport assay. The addition of excess GTP prevented the inhibition, and no effect of GTPyS was observed when the preincubation was carried out at 0%. The effect is guanine specific, since preincubation with ATPrS did not reduce transport significantly (data not shown). These experiments suggested that at least one cytosolic factor required for the inhibition by GTPyS, possibly GGBF, is a soluble GTPbinding protein. .: Prepa’ration of Cytosol Lacking GGBF Activity Cytosol provides a large number of factors that are essential to observe transport in the cell-free assay. An assay for GGBF was developed by preparing a cytosol that provided sufficient amounts of these transport factors yet exhibited a reduced level of GGBF activity. We took advantage of the fact that GGBF binds membranes very tightly in the presence of GTPyS. Whole CHO cell homogenates were incubated with GTPyS under conditions optimized for binding of GGBF to membranes, as described in Experimental Procedures. Following incubation, the homogenate was centrifugei to remove membranes and associated factors. The high-speed supernatant was desalted by chromatography to remove nucleotides and other small metabolites and then used in the transport assay. Figure 1 demonstrates that cytosol obtained from GTPyS-treated CHO homogenates (GTPrS-CHO cytosol) can be used to support transport reactions even though it shows only limited sensitivity to GTPyS. In Figure lA, titratlons of GTPyS-CHO cytosol and cytosol prepared from mock-treated homogenates (CHO) are compared in their ability to promote transport (open symbols). The two cytosols are indistinguishable even at the lower levels at which soluble factors are limiting. Further analysis of these cytosols showed that the kinetics of transport in reactions containing CHO and GTPyS-CHO cytosols are identical (data not shown). In contrast, titrations performed in the presence of 4 uM GTPrS show a clear difference between the two cytosols (closed symbols). Whereas an increase in CHO cytosol clearly leads to greater sensitivity to GTPrS (compare open and closed triangles), no such effect is observed with GTPyS-CHO cytosol (compare open and closed circles). This reduced GTPyS sensitivity indicates that GGBF was largely removed from the GTPyS-CHO cytosol by our procedure. These observations also suggest that GGBF, if required for the reaction, is not present at a limiting concentration in cytosol. Owing to the cumbersome nature of the data in Figure lA, GGBF activity was redefined in terms of a ratio: a GGBF ratio is calculated for each titration point by dividing the transport observed in the presence of GTPrS by that observed in the absence of GTPyS. Figure 1B shows a GGBF ratio plot of the data presented in Figure 1A. Note
Regulation 71
of Intra-Golgi
Traffic
by ARFs
A. 2000
iz 8
1500
lz g
1000
E E +
500
n
8.
1.0
0.6 g d
0.6
L 8
o.4 0.2
0.0
0
20
CYtosol
40
00
Mf)
Figure 1. Comparison of Transport Activities of Cytosols Prepared from Mock-Treated and GTPyS-Treated CHO Cell Homogenates (A) A series of transport assays were performed in the presence of increasing amounts of either mock-treated CHO cytosol (triangles) or GTPyS-CHO cytosol (circles), as described in Experimental Procedures. Reactions were carried out in the absence (open symbols) or presence (closed symbols) of 4 uM GTPTS. The extent of transport in 2 hr incubations is plotted as a function of the content of cytosol. (8) GGBF ratio plots of the data presented in (A). For each pair of points in the titrations shown in (A) (without and with GTPyS), a GGBF ratio was calculated by dividing the extent of transport measured in the presence of GTPyS by that measured in the absence of GTP$S. The four curves in (A) are represented more clearly as two GGBF ratio curves.
that a low GGBF ratio indicates a high level of GGBF activity. The sensitivity to GTPyS can be restored to transport reactions containing GTPrS-CHO cytosol when bovine brain cytosol (BBC) is added (data not shown). This established that GTPyS-CHO cytosol could be used in transport reactions to assay for GGBF in fractions from various purification procedures. This experiment also demonstrated that BBC, a readily available and abundant source of cytosol, could be used as starting material for the purification of GGBF. The modified intra-Golgi transport assay containing saturating amounts of GTPyS-CHO cytosol will be referred to as the GGBF assay. Loss of GGBF Activity Correlates with a Loss of GTP-Binding Proteins of 20 kd The removal of GGBF activity from cytosol described
above prompted us to investigate the content of GTPbinding proteins remaining in the CHO and GTPyS-CHO cytosols. Using [35S]GTPyS and a filter-binding method (Kikuchi et al., 1988), we found that the level of GTPbinding proteins present in GTPyS-CHO was only slightly reduced compared with CHO (data not shown). However, as shown in Figure 2, aclear difference was revealed upon analysis of the GTP-binding proteins by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and GTP ligand blotting. Comparison of CHO and GTPyS-CHO cytosols shows a depletion of GTP-binding proteins migrating at ~20 kd in the GTPyS-CHO. The association of the 20 kd cytosolic GTP-binding proteins with membranes occurs with a tH of 3 min at 37°C and is not significant at 0°C (data not shown). Quantitation of GTP bound in several blots indicates that 70%-80% of the 20 kd GTP-binding proteins are absent from GTPyS-CHO cytosol. In contrast, the other major small (24 kd-30 kd) GTP-binding proteins in this cytosol are apparently unaffected by the procedure. This agrees with the filter binding results mentioned above. We found that when BBC is fractionated by sizeexclusion chromatography, GGBF activity is found only in fractions containing proteins smaller than 25 kd (data not shown). Furthermore, these fractions contain GTPbinding proteins of ~20 kd exclusively. These observations, along with the fact that loss of the 20 kd GTP-binding proteins coincides with loss of GGBF activity, strongly suggest that GGBF is a 20 kd GTP-binding protein. Purification of Two 20 kd GTP-Binding Proteins Exhibiting GGBF Activity BBC was subjected to three consecutive purification steps to yield two preparations of purified GGBF (termed GGBF and GGBF*), which differ in their physical properties and in their specific GGBF activities. Fractions containing GGBF activity were identified at each step by their ability to restore GTPyS sensitivity in reactions containing GTPySCHO cytosol. Fractions were also analyzed for their content of 20 kd GTP-binding proteins by quantitating GTP ligand blots as described in Experimental Procedures.
CHO
“cTHp;J” Golgi
Figure 2. GTP Ligand Blot of CHO and CHO Golgi Membranes
Cytosol,
GTPTS-CHO
Cytosol,
CHO and GTPyS-CHO cytosols (36 ug each) and acceptor Golgi membrane proteins (18 ug) were separated by SDS-PAGE on a 15% gel and subjected to GTP ligand blot analysis as described in Experimental Procedures. The Golgi sample was run separately from the cytosols. An autoradiogram is shown. Molecular size markers are trypsin inhibitor (21.5 kd) and carbonic anhydrase (31 kd).
Cdl 72
BBC was loaded on an anion-exchange column (Cl Sepharose) and eluted with a linear KCI gradient. A single symmetric peak of activity eluted (Figure 3A) that was pooled and loaded on a hydrophobic interaction column (phenyl-Superose). Most of the GGBF activity elutes from this column in a peak corresponding to a minor fraction of the 20 kd GTP-binding proteins (Figure 3B). A minor peak of GGBF activity corresponding to the main 20 kd GTPbinding protein peak is also observed. This activity, labeled GGBF’, was pooled separately from the main GGBF
peak. Both GGBF and GGBF’ pools were further purified in parallel by size-exclusion chromatography. Figure 3C shows the size-exclusion (Superdex 75) elution profile of the GGBFpool from Figure3B. GGBFactivity elutes as a single peak. SDS-PAGE analysis of the fractions reveals a single protein of ~20 kd (Figure 4A), which binds GTP (Figure 48) and whose distribution coincides exactly with that of GGBF activity (see Figure 3C). Similar results were observed for the purification of GGBF’ (data not shown). Purified GGBF and GGBF’ fractions constitute approximately 0.02% and 0.05% of the total soluble protein from bovine brain, respectively, as estimated from the yield of protein in the purified fractions (Figure 3C) and GTP ligand blots. Attempts to sequence both forms of GGBF indicated that the N-terminus of both proteins is blocked (data not shown). Results of the purification are summarize&n Table 1. Overall recovery of GGBF activity as GGBF is < 2%. This poor yield may reflect the following: -2% of GGBF activity present in the starting material is recovered as GGBF’; denaturation of GGBF may have occurred during purification; or GGBF may have become separated from factors that potentiate GGBF activity as measured in whole cytosol.
Fraction
LD Fraction
Figure
3. Purification
of GGBF
5/8 9/12 13 14
15
16 17
B.
Activity
(A) Elution of GGBF activity from Q Sepharose. Diluted, desalted BBC in 20 mM KCI was applied to the column equilibrated with buffer lacking KCI. Proteins were eluted with a linear KCI gradient to 500 mM and 14 ml fractions were collected. GGBF activity was measured as described in Experimental Procedures. GTP-binding proteins of 20 kd were quantitated by Phosphorlmager analysis of GTP ligand blots. (B) Elution of GGBF activity from phenyl-Superose. Five milliliters of concentrated GGBF pooling from Q Sepharose in 1.1 M (NH&SO, was applied to the column equilibrated with buffer containing 1 .l M (NH&SO,. Proteins were eluted with a descending ammonium sulfate gradient and 1.5 ml fractions were collected. GGBF activity and 20 kd GTP-binding proteins were quantitated as in (A). The GGBF’ pooling is indicated by a bold asterisk. (C) Elution of protein and GGBF activity on Superdex 75. Five millititers of the GGBF pooling from phenyl-Supsrose was applied to the column and 3 ml fractions were collected. Molecular size standards are 66 kd for albumin and 25 kd for chymotrypsinogen.
Figure 4. Total Protein and GTP-Binding Fractions from GGBF Purification
Proteins
in Superdex-75
(A) A 15% SDS-polyacrlyamide gel was run and stained with Coomassie brilliant blue. Eighty microliters from each fraction (or 16 nl of each of fractions 5-6 and 9-12) was precipitated with 10% TCA and the pellets resolubilized in 30 nl of buffer. Samples were: 25 ul from the samples of the combined fractions 5-6 and 9-12 and 25 nl of samples of fractions 13-16 (the column load [LO] is also shown). Molecular size standards are phosphorylase b (97 kd), serum albumin (67 kd), ovalbumin (43 kd). carbonic anhydrase (31 kd), trypsin inhibitor (21.5 kd), and lysozyme (14 kd). (B) GTP ligand blot analysis of the fractions as in (A). Four microliters of each fraction was applied to a 15% gel and processed as described in Experimental Procedures. An autoradiogram of the region around MW of 20 kd is shown. GTP-binding proteins of a molecular size ~20 kd were the only ones present in the load and column fractions.
Regulation 73
Table
of Intra-Golgi
1. Purification
Traffic
of GGBF Protein
by ARFs
Activity
GGBF-Active Pooling
0x0
Concentration (mg/ml)
Percent Recovery’
BBC G-Sepharose Phenyl-Superose GGBFD GGBFD
2247 147 12.4 0.41 0.82
10.7 1.31 2.07 0.07 0.15
100 18 3.0 1.7 1.8
a Recovery of GGBF activity is reported as a percentage of the total volume of the pool required to observe I&, relative to the I& and volume of the BBC starting material. Activity of GGBF’ is reported only for the purified protein and is not included in the recoveries of GGBF. b The two fractions of greatest activity from Superdex 75.
GGBF Activity Is Neutralized by Brefeldin A The fungal metabolite brefeldin A prevents the inhibition of the intra-Golgi assay by GTPvS (Orci et al., 1991). Addition of 40 pM brefeldin A has only a small inhibitory effect on transport, but it negates the inhibition by GTPyS in reactions containing a high ratio of CHO cytosol to membranes. Therefore, we tested brefeldin A in GGBF assays containing an inhibitory concentration of GGBF. Figure 5 demonstrates that brefeldin A abolishes sensitivity of the GGBF assay to exogenous GGBF. Increasing amounts of brefeldin A were added to assays containing GTPyS-CHO cytosol and enough GGBF to bring the GGBF ratio from 1 .l (no GGBF) to 0.25. Following a brief incubation at 22OC, GTPyS was added to half of the reactions, and assays were carried out as before. Brefeldin A did not reduce transport in the GGBF assay (data not shown), in agreement with previous work (Orci et al., 1991). However, brefeldin A at >, 32 t.tM blocks GGBF activity maximally. This demonstrates that purified GGBF is sufficient to reproduce accurately the GGBF activity observed with whole cytosol. Different Forms of GGBF Have Different GGBF Activities Figure 6 shows a reverse-phase high performance liquid chromatography (HPLC) analysis of purified GGBFs. GGBF and GGBF’ elute as single peaks with no other significant protein peaks observed. Note that the GGBF preparation contains significant GGBF* (at a level of 25% [Figure 6AJ) and that GGBF is a 15% contaminant in the GGBF* preparation (Figure 6B). Isoelectric focusing reveals that GGBF is slightly more acidic than GGBF* with pls of 5.5 and 5.9, respectively (data not shown). These data demonstrate that neither GGBF preparation is resolved into more than two proteins and that the contaminant in each preparation is most likely the complementary GGBF. A quantitative analysis of the GGBF activity of GGBF and GGBF’ is presented in Table 2. The amount of protein required for 50% inhibition (IC&) of the GGBF assay is shown. GGBF exhibits an I& of 0.6 pg whereas that for GGBF” is 1.6 pg. This approximately 3-fold difference in specific activity represents a lower limit since GGBF contains 25% of the less active GGBF’, and much of the
arefeldin Figure
5. Brefeldin
A Prevents
A (PM)
GTPyS-Dependent
Inhibition
by GGBF
GGBF assays were carried out as described in Experimental Procedures, and each assay included 1.7 ug of purified GGBF (to give a GGBF ratio of 0.25). Brefeldin A was added to duplicate tubes at the indicated concentrations, and the tubes were incubated at 22W for 2 min. At this time, half of the reactions received GTPyS on ice to a final concentration of 4 PM. GGBF ratios are plotted as a function of brefeldin A concentration. The GGBF ratio in the absence of GGBF was 1.1.
activity of GGBF* may be contributed nation by GGBF.
by the 15% contami-
GGBF and GGBF’ Are Members of the ARF Family The size and cytosolic nature of GGBFs is similar to that of the GTP-binding protein ARF. ARF was first purified from bovine extracts as a cofactor for ADP-ribosylation of
Figure
6. HPLC
Analysis
of GGBF
and GGBF’
GGBF and GGBF’ pools from Superdex 75 chromatography were applied to an analytical reverse-phase HPLC column and eluted with linear gradient of acetonitrile as in Experimental Procedures. Two separate runs were made: GGBF (75 ug) (A) and GGBF’ (110 ug) (B). Absorbance was monitored at 260 nm. The region of the chromatograms from 54%-63% acetonitrile is shown. GGBF’ (asterisk) and GGBF elute in the gradient at apparent acetonitrile concentrations of 58% and 60%, respectively.
Cell 74
Table 2. GGBF
c
and ARF Activities
Protein
GGBF
GGBF GGBF’ Myr-ARFl ARFI
Q.6 1.6 2.4 >3.0
Activity” PB ps pg’ pg
ARF Activityb 12.9 11.5 6.7 ND
ARFl
ARF4
GGBF
llB9
’ Reported as the weight of protein to achieve IC, in GGBF assays as described in Experimental Procedures. ’ Reported in units of pmol of ADP-ribosylated G, per pmol of activated GGBF/ARF, determined as described in Experimental Procedures. ’ Assuming 50% of the ARF proteins present in the sample are myristoylated. Myr-ARFl and ARFl refer to recombinant proteins obtained from E. coli strains (Weiss et al., 1989) containing or lacking the gene for N-myristoyl transferase, respectively. ND: not determined.
B.
Figure 7. lmmunoreactivity Antibodies
G.. by cholera toxin (Kahn and Gilman, 1984). SDS-PAGE analysis indicated that ARF consisted of a doublet at 20 kd. To establish whether GGBFs are related to ARFs, we first examined the ARF content of GTPyS-CHO cytosol by immunoblot. The level of ARF was found to be reduced by approximately 70% (data not shown), consistent with the depletion of 20 kd GTP-binding proteins (see Figure 2). Figure 7 demonstrates that purified GGBFs are recognized by an ARF-specific antibody. GGBF and GGBF* were screened with antibodies against ARFl (ml D9) and ARF4 (R-891). Antibody ml D9 reacts strongly with ARFl and ARF3, which have 98% sequence identity (R. A. K., unpublished data). lmmunoreactivity wasmeasured relative to that of purified recombinant ARFl and ARF4. Figure 7A shows that both forms of GGBF react with mlD9. GGBFs react as well as ARFl , whereas ARF4 reactivity is less than 20%. In contrast, GGBFs do not react significantly (
