(Received for publication, September 9, 1996, and in revised form, December 10, 1996)
From the Laboratory of Biological Chemistry, Division of Basic Sciences, NCI, National Institutes of Health, Bethesda, Maryland 20892
The relationship between ADP-ribosylation factor (Arf) 1 and phosphoinositides, which have been independently implicated as regulators of membrane traffic, was examined. Because both Arf-dependent phospholipase D and Arf1 GTPase-activating protein (GAP) require phosphatidylinositol 4,5-bisphosphate (PIP2), Arf1 complexed with PIP2 has been proposed to interact with target proteins. This hypothesis was tested using Arf1 GAP as a model system. Arf1 was shown to bind to PIP2 in Triton X-100 micelles with a Kd of 45 ± 13 µM. Arf1 also bound phosphatidic acid but with 10-fold lower affinity. PIP2 binding was specifically disrupted by mutating lysines 15, 16, and 181 and arginine 178 to leucines. Decreased PIP2 binding resulted in an increased EC50 of PIP2 for activation of Arf GAP. None of the mutations that decreased PIP2 binding affected binding to or activation of GAP by phosphatidic acid, which acts at a functionally distinct site. These data support the hypothesis that PIP2 binding to Arf1 promotes interaction with Arf GAP. The implications of lipid-directed protein-protein interactions for membrane traffic are discussed.
ADP-ribosylation factors (Arf)1 are a family of GTP-binding proteins first identified and purified as a cofactor for cholera toxin-catalyzed ADP-ribosylation of the heterotrimeric GTP-binding protein Gs (1). This pathophysiologic activity has been useful for functionally defining members of the Arf family (2); however, the physiologic role of Arf is thought to be as a regulator of membrane traffic (for reviews, see Refs. 3-6). In eukaryotes, membrane traffic is mediated by protein-coated vesicles (3). With GTP bound, Arf1 recruits proteins of two classes of coats, coatomer and clathrin assembly protein-1, to Golgi-enriched membranes (7-10). This event facilitates the assembly of the coat and drives the budding of the vesicles (3). The hydrolysis of GTP by Arf1 is necessary for the uncoating and fusion of the vesicles (3-6). Arf1 has also been purified as an activator of phospholipase D (PLD) (11, 12) which has been proposed to mediate Arf-stimulated coated vesicle assembly (13).
Phosphoinositides have been independently implicated as regulators of membrane traffic (14). In yeast, mutants of phosphatidylinositol transfer protein and phosphatidylinositol 3-kinase are deficient in protein secretion and vacuolar sorting (15, 16). Phosphoinositides play a role in membrane traffic in mammalian cells as well (14). For example, exocytosis in PC12 cells requires a number of soluble proteins including phosphatidylinositol transfer protein and phosphatidylinositol 4-phosphate 5-kinase (PIP kinase) (17, 18). Similarly, phosphatidylinositol transfer protein has been purified as a required soluble factor for the formation of secretory vesicles in adrenal medulla (19).
Arf1 is a possible site at which phosphoinositides may play a role in membrane traffic. The interaction of Arf1 with two different GTPase-activating proteins (GAP)2 (20, 21) and PLD (12, 22) has been shown to be PIP2-dependent. On further characterization, the Arf1 GAP activity from bovine brain was found to be coordinately regulated by PA and PIP2 (21). Although PA did not activate GAP activity by itself, it lowered the EC50 of PIP2 required for activation by 90%. These data functionally define two lipid binding sites (21, 23). GAP activity depends on occupancy of both sites. Site 1 is specific for PIP2. Site 2 may be occupied by either PIP2 or PA.
Although two lipid binding sites have been functionally defined for the Arf1-GAP interaction, the proteins on which the binding sites reside have not been identified. Arf1 has been speculated to contain an acid lipid binding site. The effect of PIP2 on nucleotide dissociation from Arf1 has been taken as evidence for PIP2 binding to Arf1 (23). Furthermore, the crystal structure of Arf1 has revealed that basic residues from positions 10, 15, 16, 59, 178, and 181 cluster to form a solvent-exposed patch of positive potential (24, 25). This polarized surface is reminiscent of pleckstrin homology domains (26, 27) and could be a site of binding to PIP2. These data, with the functional data, have led to the suggestion that Arf1, in addition to GTP, must bind PIP2 to interact with target proteins such as GAP or PLD (21, 23).
The hypothesis that PIP2, by binding to Arf1, directs Arf1-protein interactions was tested using Arf1 GAP as a model system. PIP2 binding to Arf1 was demonstrated using a direct assay. To specifically disrupt PIP2 binding to Arf1, basic residues that comprise the positively charged surface-exposed patch on Arf1 were mutated to neutral residues. The mutant Arf1 proteins had reduced affinity for PIP2 and required greater concentrations of PIP2 for productive interaction with Arf GAP than did the wild type protein. The mutations had no effect on either PA binding or PA-dependent GAP activation. These data support the hypothesis that Arf1 with PIP2 bound is the substrate for GAP. The ability of specific lipids to direct Arf1-protein interactions could play a role in spatially and temporally separating events in membrane traffic.
Triton X-100, phosphatidylcholine (PC),
1-stearoyl-2-arachidonoyl, phosphatidic acid (PA), phosphatidylserine
(PS), and PIP2 were from Sigma.
[-32P]GTP, [
-32P]GTP, and
[35S]GTP
S were from DuPont NEN. AcA44 was from
Biosepra Inc. (Marlborough, MA). Bovine brains were obtained from a
local slaughter house. Arf1 GTPase-activating protein (Arf1 GAP)
was purified as described (21). Recombinant Gs
was
expressed in and purified from bacteria as described (28).
G
was purified from bovine brain as described (29).
Cholera toxin was purchased from Calbiochem and was preactivated by
incubation in 25 mM potassium Pi, pH 7.4, 20 mM dithiothreitol for 15 min at 37 °C.
Site-directed mutagenesis was performed by
polymerase chain reaction amplification of the open reading frame of
human Arf1 using synthetic oligonucleotides incorporating the desired
mutations. NdeI and BamHI restriction sites were
added at the 5 and 3
ends to allow subsequent subcloning into pET3C
bacterial expression vector (30, 31). The sequence of the entire open
reading frame was determined by dye-terminator sequencing using an ABI
instrument (model 373A) and reagents from Perkin-Elmer. Wild type and
mutant proteins were expressed in BL21 (DE3) (30, 31) and purified (32)
as previously described.
Binding reactions contained 1 µM Arf1 and 10 µM
[35S]GTPS (specific activity, 1000-5000 cpm/pmol) in
exchange buffer of 25 mM HEPES, pH 7.4, 100 mM
NaCl, 0.5 mM MgCl2, 1 mM EDTA, 1 mM dithiothreitol, 3 mM
L-
-dimyristoylphosphatidylcholine, and 2.5 mM sodium cholate. Samples (10 µl) were taken 12 times
ranging from 2 to 120 min, and bound nucleotide was determined by rapid
filtration on BA85 nitrocellulose (Schleicher and Schuell) (33). Data
were fit to a first order rate equation using the Marquardt
algorithm.
Arf-dependent cholera
toxin-catalyzed ADP-ribosylation of Gs was determined as
described previously (34). Incubations containing 25 mM
HEPES, pH 7.4, 100 mM NaCl, 0.5 mM
MgCl2, 1 mM EDTA, 0.25 µM
concentration of the indicated Arf protein, 2.2 µM
recombinant Gs, 20 µM G
,
40 µg/ml cholera toxin, 5 µM NAD (approximately 20,000 cpm/pmol), 3 mM
L-
-dimyristoylphosphatidylcholine, and 2.5 mM cholate were for 20 min at 30 °C.
GAP activity was determined as previously described (21). Lipids were added in mixed micelles with Triton X-100. The incubations contained 8 µg/ml GAP, 0.1% Triton X-100, and the indicated concentrations of PIP2 and PA. As discussed in Ref. 21, the concentration of Arf1·GTP was much less than the Km. Under this condition, substrate is consumed with a first order rate equal to Vmax/Km. This rate is expressed here as the fraction of Arf1·GTP hydrolyzed/min.
Arf1 Binding to Triton X-100 Micelles and Phospholipid LiposomesPhospholipids in chloroform/methanol, 95:5, were taken
to dryness under a stream of nitrogen. To prepare liposomes, the lipids were then resuspended in 25 mM Tris, pH 8.0, to a final
lipid concentration of 13 mM. The suspension was sonicated
for 5 min in a G112SP1 ultrasonic cleaner (Laboratory Supplies Co.,
Hicksville, NY). To prepare micelles, the lipids were solubilized in
1% Triton X-100. Arf1 was loaded with [-32P]GTP by
incubation with 10 µM [
-32P]GTP
(25,000-50,000 cpm/pmol) in 20 mM potassium
Pi, pH 7.4, 100 mM NaCl, 0.5 mM
MgCl2, 1 mM EDTA, and 0.1% Triton X-100 for 20 min at 30 °C, and the MgCl2 concentration was adjusted
to 2 mM. Arf1·GTP was diluted 1:19 into either mixed
micelles of phospholipids and 0.1% Triton X-100 or 650 µM phospholipid in 25 mM HEPES, pH 7.4, 100 mM NaCl, 3.5 mM MgCl2, 1 mM ATP, 1 mM GTP, and 1 mM
dithiothreitol, incubated at 30 °C for 2 min, and then chilled.
Samples (100 µl) were then applied to 1 ml of AcA44 (Biosepra) in a
Bio-Rad polyprep disposable column. The column was centrifuged at
200 × g for 2 min, and the amount of radiolabeled
nucleotide in the eluate was determined. 67 ± 5% of the
phospholipids but less than 10% of the Arf1 (in the absence of acid
phospholipids) applied to the column were recovered in the eluate.
Phospholipid composition did not affect lipid recovery. To determine
the fraction of Arf1 recovered, total protein-bound radiolabeled
nucleotide was determined in samples from the same incubations
(33).
Protein concentrations were estimated by the Amido Black assay using bovine serum albumin as a standard (35). Total lipid phosphorus was determined as described (36).
The effects of
PIP2 on nucleotide dissociation from Arf1 has been taken as
evidence for PIP2 binding to Arf1 (23). To directly demonstrate this interaction, spin columns constructed of AcA44 (an
acrylamide-agarose gel filtration matrix with a
Mr cut off of 130,000) were used to rapidly
separate Arf1·GTP bound to micelles from free protein (Fig.
1). Less than 10% of Arf1·GTP bound to micelles of
Triton X-100. The binding of Arf1 was increased by including
PIP2 in the micelles with a half-maximal effect occurring at 45 ± 13 µM. Maximal binding was 60 ± 8%
of the total Arf1. Arf1·GTP also bound to PA- and PS-containing
micelles but with lower affinity than to PIP2 (Fig. 1).
Arf1 binding to PIP2 was independent of lipid presentation.
Binding to mixed liposomes of PC and PIP2 was dependent on
PIP2 concentration with half-maximal binding occurring at
30 ± 11 µM PIP2.
Arf1 Binding to Acid Phospholipids Was Decreased by Neutralizing Positively Charged Solvent-exposed Amino Acids
Using the crystal
structure of Arf1·GDP as a guide, 5-point mutants of Arf1 were
generated by polymerase chain reaction mutagenesis. Each mutation
neutralized either one or two of the basic residues that contribute to
the polarized patch on the surface of Arf1 (Table I).
All mutants were expressed in Escherichia coli and purified
by the same methods used to purify recombinant Arf1. Two quantitative
functional assays were used to determine if the mutations had global
conformational effects. First, GTPS binding was determined (Table
I). Of the mutants, only [K59L]Arf1 did not bind nucleotide. The
reason for this has not yet been determined, but the lack of binding
precluded the use of the mutant for this study. The other four mutants,
[K16L]Arf1, [K15L,K16L]Arf1, [R178L]Arf1, and
[R178L,K181L]Arf1, were similar to wild type protein in both rate and
extent of binding. Thus, the mutations had not disrupted the nucleotide
binding pocket. Second, activity as a cofactor for cholera
toxin-catalyzed ADP-ribosylation of Gs was determined. This
activity requires that Arf1 bind GTP (1), cholera toxin (38, 39), and
Gs (1, 39) and, therefore, was felt to be a test of
preservation of global conformation. All mutants that could bind
nucleotide were indistinguishable from the wild type protein (Table
I).
|
Neutralizing the residues from the positively charged cluster on Arf1
specifically affected PIP2 binding. The PIP2
dependence for binding to Triton micelles was determined for wild type
Arf1 and the four mutants that bound nucleotide. The concentration of
PIP2 for half-maximal binding was 2-fold greater for
[K16L]Arf1 and [R178L]Arf1 and more than 4-fold greater for
[K15L,K16L]Arf1 and [R178L,K181L]Arf1 than for wild type Arf1 (Fig.
2 and Table II). The relative affinities
were independent of phospholipid presentation. Similar results were
obtained using liposomes of PC and PIP2 (not shown). In
contrast, the mutations did not detectably affect binding to PA (Fig.
3 and Table II).
|
Mutations That Decreased PIP2 Binding to Arf1 Also Affected the PIP2 Dependence of Arf1 GAP Activity
Two
phospholipid binding sites have been functionally defined for Arf1-Arf1
GAP interactions. One site binds PIP2 specifically. The
second site binds both PIP2 and PA. To examine site 1, PIP2 dependence in the presence of a fixed concentration of
PA (700 µM was chosen because it is saturating for
functionally defined site 2, see Fig. 5) was determined for the wild
type protein and the mutants (Fig. 4). The
EC50 of PIP2 in the GAP-catalyzed reaction correlated with the relative affinities of the mutants for
PIP2. The EC50 of PIP2 was
approximately 2-fold greater for [K16L]Arf1 and [R178L]Arf1 and
more than 5-fold greater for the mutants with two residues neutralized,
[K15L,K16L]Arf1 and [R178L, K181L]Arf1, than for wild type Arf1
(Fig. 4 and Table II).
To examine the effect of the mutations on the functionally defined site 2, the PA concentration dependence, at a fixed concentration of PIP2 (50 µM, which is greater than the EC50 for site 1 but less than the EC50 for functionally defined site 2 using wild type Arf1), was determined for the wild type protein and each mutant. If the mutations had affected site 2, then the EC50 of PA should be affected. If the mutations had an isolated effect on functionally defined site 1, then the EC50 of PA would be unaffected but the maximal velocity would be decreased (because site 1 would not be completely occupied at the fixed concentration of PIP2 used). The latter was the case (Fig. 5 and Table II). The maximum velocity observed with both [K16L]Arf1 and [R178L]Arf1 was approximately 70% that of the wild type protein. The maximum velocity observed with the mutants with two residues neutralized, [K15L,K16L]Arf1 and [R178L,K181L]Arf1, was 15-30% that of wild type. However, the EC50 for PA was the same for the mutants as for the wild type protein. Thus, the mutations had an isolated effect on the functionally defined PIP2 binding site 1. Taken together, these data support the hypothesis that Arf1 with PIP2 bound is the substrate for GAP.
Both Arf1 and phosphoinositides are required for efficient vesicular transport (14). Potential relationships between these required factors have been discussed previously (22, 23, 40). Here, we show that Arf1 binds PIP2. Furthermore, disrupting PIP2 binding to Arf1 is shown to increase the EC50 of PIP2 for Arf1 GAP activation. These data support the hypothesis that PIP2 binding to Arf1 directs the interaction with GAP and suggest a model in which Arf1 may "sense" lipids, interacting with specific proteins depending on the lipid environment. This mechanism could provide a means of ordering events in membrane traffic.
Arf1 binds to phospholipid vesicles containing acid phospholipids. Prior to this report, the effects of PIP2 on nucleotide exchange on Arf1 suggested that Arf1 could bind phosphoinositides (23). Here, a spin column-based assay was used to directly demonstrate Arf1 binding to PIP2-containing micelles and liposomes. Arf1 bound to other acid phospholipids but with lower affinity. The specificity of Arf binding to PIP2 was further tested by determining if any mutations in Arf1 had an isolated effect on PIP2 binding. Arf1 contains a surface-exposed cluster of basic amino acids that could play a role in binding acid phospholipids such as PIP2 (26, 27). Consistent with specificity for PIP2, neutralizing residues from this cluster affected binding of PIP2 but not PA.
Two lipid binding sites have been functionally defined for Arf1 GAP activation. One site was proposed to be on Arf1. Thus, Arf1·PIP2 would be acting as the substrate for Arf1 GAP. The identification of Arf1 mutants with reduced affinities for PIP2 allowed this to be tested. As predicted by the model, the concentration of PIP2 required for activation of GAP was found to depend on the affinity of the particular Arf1 mutant for PIP2. These data demonstrate that Arf1 is the previously functionally defined PIP2 binding site. The role of PIP2 was more than simply targeting Arf1 to the membranes. Both PA and PS promote Arf1 binding to vesicles, but neither activated GAP in the absence of PIP2. Thus, PIP2 must be inducing a conformational change necessary for the interaction with GAP.
The second functionally defined lipid binding site is structurally
distinct from site 1. This site, which could be occupied by either
PIP2 or PA, was unaffected by the mutations in Arf1 and,
therefore, could be either a distinct site on Arf1 or on GAP. As was
the case for PIP2, the role of PA is not simply recruiting Arf1 to the phospholipid surface. PA did not increase the fraction of
Arf1 bound to vesicles to an extent greater than PIP2 alone did. This phenomenon is similar to that seen with the kinetics of
PIP2-phospholipase C- (41). PA converts a sigmoid
PIP2 dependence to a hyperbolic dependence and lowers the
Km 10-fold. However, PA does not affect the
association constant of the enzyme with the detergent micelles used in
the study. Therefore, PA is acting as an allosteric modifier for both
phospholipase C and GAP. The nature of this allosterism may be of
general interest in understanding phosphoinositide metabolism since
another enzyme of this pathway, PIP kinase, is also activated by PA
(42, 43).
The ability of phospholipids to direct Arf1-protein interactions may order events in membrane traffic. Several activities have been ascribed to Arf1 including activity as a cofactor for cholera toxin, vesicle coat protein recruitment (5), PIP kinase activation (44), and PLD activation (11, 12, 40). The lipid requirements of two of these have been examined. PLD requires PIP2 but not PA, the product of the reaction catalyzed by PLD. Activity as a cofactor for cholera toxin is inhibited by PIP2.2 Thus, three different Arf1-protein interactions have different lipid requirements. Since Arf1 also affects lipid composition of membranes by activating both PIP kinase (44) and PLD (11, 12), these observations have led to suggestions that Arf1 may help order events in membrane traffic by responding to and affecting changes in phospholipid composition (21-23, 40). With low concentrations of PIP2 and PA present, Arf1 might initiate assembly of protein-coated vesicles. In the vesicle, Arf1 activates PIP kinase. This produces PIP2 which allows Arf1 to activate PLD. In turn, PLD generates PA which together with PIP2 activates GAP and, consequently, terminates the assembly signal. A combination of biochemistry and genetics using mutant proteins similar to those described in this paper should allow this and similar models to be tested.
I thank John K. Northup, Ronald Felsted, and Julie Donaldson for helpful discussions, Jenny Clark, Stacey Sturch, and Sandor Berghoffer for technical assistance, and Albert Fornace and Richard Kahn, for support.