(Received for publication, April 30, 1997)
From the CNRS, Institut de Pharmacologie Moléculaire et Cellulaire, 660 route des Lucioles, Sophia Antipolis, 06560 Valbonne, France
Arno is a 47-kDa human protein recently
identified as a guanine nucleotide exchange factor for ADP ribosylation
factor 1 (ARF1) with a central Sec7 domain responsible for the exchange
activity and a carboxyl-terminal pleckstrin homology (PH) domain
(Chardin, P., Paris, S., Antonny, B., Robineau, S.,
Béraud-Dufour, S., Jackson, C. L., and Chabre, M. (1996)
Nature 384, 481-484). Binding of the PH domain to
phosphatidylinositol 4,5-bisphosphate (PIP2) greatly
enhances Arno-mediated activation of myristoylated ARF1. We show here
that in the absence of phospholipids, Arno promotes nucleotide exchange
on [17]ARF1, a soluble mutant of ARF1 lacking the first 17 amino
acids. This reaction is unaffected by PIP2, which suggests
that the PIP2-PH domain interaction does not directly regulate the catalytic activity of Arno but rather serves to recruit Arno to membranes. Arno catalyzes the release of GDP more efficiently than that of GTP from [
17]ARF1, and a stable complex between Arno
Sec7 domain and nucleotide-free [
17]ARF1 can be isolated. In
contrast to [
17]ARF1, full-length unmyristoylated ARF1 is not
readily activated by Arno in solution. Its activation requires the
presence of phospholipids and a reduction of ionic strength and
Mg2+ concentration. PIP2 is strongly
stimulatory, indicating that binding of Arno to phospholipids is
involved, but in addition, electrostatic interactions between
phospholipids and the amino-terminal portion of unmyristoylated
ARF1GDP seem to be important.
We conclude that efficient activation of full-length ARF1 by Arno requires a membrane surface and two distinct protein-phospholipid interactions: one between the PH domain of Arno and PIP2, and the other between amino-terminal cationic residues of ARF1 and anionic phospholipids. The latter interaction is normally induced by insertion of the amino-terminal myristate into the bilayer but can also be artificially facilitated by decreasing Mg2+ and salt concentrations.
ADP-ribosylation factors (ARFs),1 which were originally identified as cofactors for cholera toxin, are small GTP-binding proteins involved in intracellular vesicular transport (1, 2). Their functions include acting as regulators of the binding of coat proteins and adaptins to intracellular membranes (3, 4) and as activators of phospholipase D (5-7). It is still unclear whether phospholipase D mediates ARF signals to initiate coated vesicle formation (8, 9).
Like other G proteins, ARFs cycle between inactive GDP-bound and active GTP-bound conformations, and they are thought to shuttle during the activation cycle between cytosol and membranes, with the possible exception of ARF6 (10). Taking ARF1, the most abundant ARF family member, as a model, we previously assessed the importance of N-myristoylation for GTP-dependent binding to phospholipid vesicles in vitro (11, 12). We concluded that the increased affinity of ARFGTP for membranes cannot be ascribed to a myristoyl switch, i.e. a nucleotide exchange-dependent exposure of the myristate, as frequently proposed in reviews (3, 13). We believe that in ARFGDP, the myristate is accessible for interacting with phospholipids and that this interaction mostly accounts for the weak binding of ARFGDP to membranes. The tight binding of ARFGTP to phospholipids can be explained by a GTP-dependent release of the amino-terminal helix from the protein core, leading to the exposure of several hydrophobic residues that insert into the lipids in addition to the myristate (14).
Conversion of ARFGDP to ARFGTP is promoted
in vivo by a guanine nucleotide exchange factor.
Complementation studies in yeast recently led to the discovery of a
family of ARF nucleotide exchange factors that share a region of
sequence homology with Sec7, a yeast protein necessary for intra-Golgi
transport (15). We identified as a member of this family an
ubiquitously expressed ~47-kDa human protein termed Arno (for
ARF nucleotide-binding site opener)
(16). Another protein of very similar sequence but mostly expressed in
hematopoietic cells (17) turned out to be identical to cytohesin-1, a
protein described as a regulator of 2 integrin in lymphocytes (18).
Cytohesin-1 also promotes nucleotide exchange on ARF1 (16) as well as
on ARF3 (19). Arno and cytohesin-1 contain an amino-terminal
coiled-coil motif, a central Sec7 domain, and a pleckstrin homology
(PH) domain followed by a cluster of positively charged residues at the
carboxyl terminus. Optimal exchange activity on myrARF requires the
presence of negatively charged phospholipid vesicles supplemented with
PIP2. Studies with truncated mutants of Arno have
demonstrated that the Sec7 domain is responsible for the exchange
activity, whereas the PH domain binds to PIP2 (16).
In the present study, we have addressed two questions concerning the activation of ARF1 by Arno. (a) What is the role of the binding of the PH domain to PIP2: does it serve only to recruit Arno to membranes, or is it required to activate the exchange factor? (b) What is the role of the amino-terminal helix of ARF1 in the interaction ARF-Arno?
Egg phosphatidylcholine, egg
phosphatidylglycerol, and bovine brain PIP2 were purchased
from Sigma. [35S]GTPS was from NEN Life Science Products,
[3H]GDP was from Amersham Corp., and unlabeled
nucleotides were from Boehringer Mannheim.
Bovine
recombinant ARF1 was expressed in Escherichia coli and
purified by a single QAE-Sepharose chromatography as described previously (11). Myristoylated ARF1 was made by co-expression with a
yeast N-myristoyltransferase (20) and separated from the
contaminating nonmyristoylated protein by a precipitation at 35%
saturation of ammonium sulfate followed by sequential chromatography on
DEAE-Sepharose and Mono-S columns (12). The mutant [17]ARF1, which
lacks the first 17 amino acids, was purified by gel filtration with
Ultrogel AcA44 (14). Wild-type Arno,
PH Arno (a deletion mutant
lacking the PH domain), and Arno Sec7 domain were isolated by anion
exchange on QAE-Sepharose and gel filtration on Sephacryl S-100 HR
(Pharmacia Biotech Inc.) as described previously (16). As judged by
Coomassie Blue staining after SDS-polyacrylamide gel electrophoresis,
final purity was >95% for ARF proteins and was between 70 and 90%
for Arno proteins.
Unilamellar phospholipid vesicles were prepared by the extrusion method of Hope et al. (21). A film of phospholipids of the desired composition in PC (phosphatidylcholine), PG (phosphatidylglycerol), and PIP2 was formed in a Rotavapor and resuspended at 4 mg/ml in 50 mM Hepes (pH 7.5) with or without 100 mM KCl, as indicated. The suspension was vortexed for 20 min and freeze-thawed five times. Unilamellar vesicles were produced by extrusion through 0.1-µm polycarbonate filters (14).
Nucleotide Binding AssayUnless otherwise stated, wild-type
or truncated ARF (1 µM) was incubated at 37 °C with
[35S]GTPS or [3H]GDP (10 µM,
~1000 cpm/pmol) in 50 mM Hepes (pH 7.5), 1 mM
dithiothreitol, 1 mM MgCl2 with or without (as
indicated in the figure legends) 2 mM EDTA (1 µM or 1 mM free Mg2+), 100 mM KCl, and 1 mg/ml phospholipid vesicles. At the indicated times, samples of 25 µl (25 pmol of ARF) were removed, diluted into 2 ml of ice-cold 20 mM Hepes (pH 7.5), 100 mM
NaCl, and 10 mM MgCl2, and filtered on 25-mm BA
85 nitrocellulose filters (Schleicher & Schüll). Filters were
washed twice with 2 ml of the same buffer, dried, and counted.
[17]ARF1 (10 µM) and Arno
Sec7 domain (10 µM) were incubated, separately or
together, for 10 min at 25 °C in 20 mM Tris/HCl (pH
7.5), 100 mM NaCl, 5 mM 2-mercaptoethanol, 0.1 mM phenylmethylsulfonyl fluoride, and 1 mM
MgCl2, with or without 2 mM EDTA (1 µM or 1 mM free Mg2+). A 200-µl
sample of each incubation was applied to a Superose 12 HR 10/30 column
(Pharmacia) and eluted with the same buffer at a flow rate of 0.5 ml/min. Fractions of 300 µl were collected, and 60-µl samples were
concentrated and analyzed by SDS-polyacrylamide gel
electrophoresis.
As previously reported (16), optimal nucleotide
exchange activity of Arno on myrARF1 requires the presence of
PC/PG/PIP2 vesicles in the assay (Fig.
1A). The stimulatory effect of
PIP2 is most likely due to an interaction between this
phospholipid and the PH domain of Arno (16), but two possible
mechanisms can be proposed. The binding of the PH domain to
PIP2 could simply concentrate Arno on the surface of
vesicles and thereby facilitate its interaction with the membrane-bound
fraction of myrARF1GDP. Alternatively, binding to
PIP2 could induce a conformational change of Arno, leading
to an increased intrinsic catalytic activity, for example, by releasing
a negative constraint. To discriminate between these alternatives, we
examined the effect of Arno on [17]ARF1, a soluble substrate.
Indeed, this deletion mutant of ARF1 lacking the 17 amino-terminal
amino acids has lost the requirement for phospholipids in the
nucleotide exchange reaction (22) and remains totally soluble in the
GTP-bound state, as demonstrated by lipid-protein fluorescence energy
transfer measurements (14) and by a sedimentation assay with
phospholipid vesicles (data not shown). This allows monitoring of the
GDP-to-GTP
S exchange in the absence of phospholipids. Fig.
1B shows that Arno is indeed active on [
17]ARF1, which
first indicates that the amino-terminal
-helix of ARF1 is not
essential for the interaction ARF-Arno. Most notably, the exchange
activity of Arno on [
17]ARF1 was exactly the same in the absence
of phospholipids and in the presence of PC/PG/PIP2
vesicles. Moreover, a similar activity was observed with
PH Arno, a
mutant lacking the PH domain, or Arno Sec7 domain, a deletion mutant
restricted to the central catalytic domain (Fig. 1C).
Altogether, these results demonstrate that in full-length Arno, there
is no constraint on the catalytic site that would be alleviated by the
binding of the PH domain to PIP2-containing vesicles.
Accordingly, the interaction of the PH domain with PIP2 does not regulate the enzymatic activity of Arno but simply mediates its membrane association.
Arno Sec7 Domain Forms a Stable Complex with [
The observation that Arno Sec7 domain catalyzes
nucleotide exchange on [17]ARF1 (Fig. 1C) prompted us
to examine whether the interaction between these two soluble proteins
could be demonstrated by gel filtration chromatography. Fig.
2 shows that this is indeed the case.
When a 1:1 mixture of Arno Sec7 domain and [
17]ARF1 (10 µM each) was applied to and eluted from a Superose 12 column, the two proteins were found to be partly associated at 1 mM Mg2+ (Fig. 2B) and nearly totally
associated at 1 µM Mg2+ (Fig. 2D).
Addition of 10 µM GDP to the elution buffer at 1 µM Mg2+ markedly reduced the interaction
between the two proteins (Fig. 2E), which indicates that
Arno Sec7 domain forms a stable complex specifically with the
nucleotide-free state of [
17]ARF1.
Arno Catalyzes GDP and GTP
We previously reported that a
soluble nucleotide exchange factor present in a retinal extract can
promote GTPS release as well as GDP release from ARF1 (23). We
investigated whether purified Arno is also able to catalyze the
exchange reaction in both directions. This was first examined in the
absence of phospholipids with [
17]ARF1 as a substrate. As shown in
Fig. 3, Arno stimulates both GDP and
GTP
S dissociations from [
17]ARF1, but the effect is much more
pronounced on GDP release. Addition of 0.1 µM Arno to 1 µM [
17]ARF1 increased the rate of GDP dissociation
40-fold and increased the rate of GTP
S dissociation only 3-fold, and this was not affected by the nature of the displacing nucleotide (Fig.
3). Thus, Arno seems to have in solution a better affinity for
ARFGDP than it does for ARFGTP
S. This
preference is less obvious when the nucleotide exchange is measured
with myrARF1 in the presence of PC/PG/PIP2 vesicles (Fig.
4). Under these conditions, GDP and
GTP
S releases seem to be stimulated to a similar extent by 20 nM Arno, but this is likely due to the fact that the
reaction then occurs on the membrane surface, with both substrates.
Indeed, if phospholipids are omitted from the assay medium, the
stimulation of [3H]GDP release from myrARF1 by Arno
becomes detectable only above 0.1 µM of the exchange
factor (data not shown). This indicates that only the membrane-bound
fraction of myrARF1GDP is a good substrate for Arno, and
because myrARF1GDP is mostly soluble, this fraction must be
very low as compared with myrARF1GTP
S, which is totally
bound to phospholipids. Thus, differential binding of the proteins to
membranes greatly complicates the kinetic analysis. This is also
evidenced by the observation that in these conditions, GDP and GTP
S
are no longer equivalent as displacing nucleotide. Both
[3H]GDP and [35S]GTP
S dissociations from
myrARF1 were more accelerated by a catalytic amount of Arno when GDP
was used as the displacing nucleotide (Fig. 4). This suggests that
after the exchange, Arno dissociates more rapidly from
myrARFGDP than it does from myrARFGTP
S, which could be linked to the low affinity of myrARFGDP for
membranes. But on the other hand, Arno-stimulated
[35S]GTP
S release was less complete in the presence of GDP
(Fig. 4, right panel). It did not fit a first-order
kinetics, as if with time, less and less Arno was acting on
[35S]GTP
S-bound myrARF1, because of a progressive trapping
by the accumulating GDP-bound form of myrARF1. Altogether, these
results suggest that Arno has indeed a higher affinity for
ARFGDP, even though membrane-bound Arno dissociates more
rapidly from myrARFGDP than it does from
myrARFGTP, because the former species is less retained on
phospholipids.
Activation of Nonmyristoylated ARF1 by Arno Requires Low Mg2+ and Low Salt Concentrations
We previously
reported that Arno is inactive on unmyrARF1 when tested at 1 mM Mg2+ in the presence of 100 mM
KCl and azolectin vesicles (16). Replacement of azolectin, whose
composition in PIP2 is not known, by defined phospholipids
of optimal composition (PC/PG/PIP2, 65:30:5) allows the
detection of a small activation of GTPS binding to unmyrARF1 on the
addition of a 2-fold molar excess of Arno (Fig. 5, left panel), but this
effect remains marginal as compared with that on myrARF1 or
[
17]ARF1. However, this effect could be greatly amplified by
changing the composition of the assay buffer: (a) decreasing
the ionic strength by omission of KCl led to the rapid accumulation of
a substantial fraction of activated unmyrARF1 at 1 mM
Mg2+ (Fig. 5, left panel), and (b)
lowering the concentration of free Mg2+ to 1 µM further enhanced the activated fraction (Fig. 5,
right panel). We have shown that 1 µM free
Mg2+ allows optimal spontaneous GDP-to-GTP
S exchange on
unmyrARF1 (11). Thus, Mg2+ conditions that favor
spontaneous nucleotide exchange seem to also favor the activation by
Arno, whereas decreasing the salt concentration presumably facilitates
ionic interactions between the two proteins and phospholipids.
Lipid Dependence of the Activation of unmyrARF1 by Arno
With
the optimal Mg2+ and salt concentrations defined above (1 µM free Mg2+, no KCl), we reexamined the
effect of phospholipids on the stimulation of GTPS binding to
unmyrARF1 by wild-type Arno and Arno Sec7 domain. In the absence of
phospholipids (Fig. 6A),
[35S]GTP
S binding to 1 µM unmyrARF1 was
accelerated poorly by 0.1 µM Arno and accelerated more
significantly by 2 µM Sec7 domain. In fact, the initial
rate of exchange was a linear function of the exchange factor
concentration and was identical for full-length Arno and Arno Sec7
domain (Fig. 6A, inset). Thus, in the absence of
phospholipids, Arno and the Sec7 domain are equivalent toward unmyrARF
but are both very poor activators. In the presence of PC/PG (70:30)
vesicles, the stimulation by 2 µM Sec7 domain was only
slightly increased, but wild-type Arno became much more active because
0.1 µM Arno was roughly equivalent to 2 µM
Sec7 domain under these conditions (Fig. 6B). Inclusion of
5% PIP2 in the vesicles did not change the stimulation by
the Sec7 domain but further markedly increased the stimulatory effect
of Arno (Fig. 6C).
Thus, although unmyrARF1GDP is essentially soluble (12),
its activation by Arno is greatly facilitated by the binding of Arno to
PIP2-containing vesicles, which is in marked contrast with
[17]ARF1 (Fig. 1B). The lack of PIP2 effect
on the Sec7 domain was expected because the PH domain is missing in
this mutant protein. In fact, the Sec7 domain is completely soluble;
hence, the small difference in its activity observed without lipids
(Fig. 6A) or with lipids (Fig. 6, B and
C) cannot be due to an interaction between the Sec7 domain
and phospholipids. It must be therefore ascribed to unmyrARF1. We
previously reported that phospholipids stabilize the active form of
unmyrARF1 and thereby increase the binding of [35S]GTP
S
(11). To determine whether this is the explanation for the effect
observed here, we repeated the same experiment by monitoring [3H]GDP dissociation. Interestingly, the activation by
the Sec7 domain remained higher in the presence of phospholipids than
it was in their absence (Fig. 7).
Sec7-catalyzed exchange (after subtraction of the spontaneous exchange)
was accelerated 2-fold in the presence of lipids. This result
demonstrates that interaction of unmyrARF1GDP with
phospholipids facilitates its activation by the soluble truncated
exchange factor, whereas it does not affect the spontaneous nucleotide
dissociation at all. It can be concluded, therefore, that both
Arno-phospholipid and unmyrARF1GDP-phospholipid interactions play a role in the activation of unmyrARF1 by Arno.
The present study demonstrates that wild-type Arno is able to
promote guanine nucleotide exchange on [17]ARF1, a soluble truncated mutant of ARF1, in the absence of phospholipids. The exchange
activity is unaffected by the addition of phospholipid vesicles
composed of 65% PC, 30% PG, and 5% PIP2, vesicles that have been shown to bind >60% Arno (16). Moreover, a comparable activity is observed with an equivalent concentration of Arno
PH, a
mutant of Arno lacking the PH domain, or of Arno Sec7 domain, a
deletion mutant restricted to the central catalytic domain. These
results indicate that in wild-type Arno the catalytic site is likewise
accessible to the soluble substrate, whether Arno is membrane-bound or
in solution. Therefore, the binding of the PH domain to
PIP2-containing vesicles does not affect the catalytic activity of Arno but simply promotes membrane recruitment of the exchange factor.
The role of PIP2 binding has been examined for a number of
other PH domain-containing enzymes. In two cases, namely the dynamin GTPase (24) and the Akt/PKB kinase (25), the binding of
phosphoinositides has been proposed to directly regulate the catalytic
activity, even though in the latter case the physiological relevance of this stimulation has been recently questioned by the characterization of a 3-phosphoinositide-dependent protein kinase that
phosphorylates and activates Akt/PKB (26). Most frequently however, the
PH domain is thought to have a simple recruiting function, by
facilitating anchoring of the enzyme in PIP2-enriched
regions of the membrane. This has been clearly demonstrated for G
protein-coupled receptor kinases (27) and both and
isoforms of
phosphoinositide-specific phospholipase C (28, 29). Also, for other
small G protein-specific nucleotide exchange factors of the Dbl family
acting on Rho-like G proteins (30), the PH domain is believed to serve
primarily to target the exchanger to specific cellular locations.
Because ARF activates PLD1 (5, 6), and PLD1 activity is dependent on
PIP2 even after purification of the enzyme (31), it is
possible that the recruitment of Arno by PIP2 serves to
concentrate ARF in the vicinity of its PLD effector. It should be
stressed that PIP2-mediated membrane association of Arno
greatly enhances the activity of the exchange factor toward its
membrane-bound substrate myrARF1, even though the catalytic site is not
directly affected. This is likely to result from an increased local
concentration of the two proteins on the membrane surface that
facilitates their interaction.
The fact that Arno efficiently catalyzes GDP dissociation from
[17]ARF1 demonstrates that the amino-terminal helix of ARF1 is not
the site of interaction with the exchange factor, which contradicts the
prediction of Amor et al. (32) based on the crystallographic
structure of unmyrARFGDP. The formation of a stable complex
between Arno Sec7 domain and nucleotide-free [
17]ARF1 (Fig. 2)
should allow preparation of crystals for x-ray diffraction analysis and
identification of the true contacts between the two proteins.
It is noteworthy that in contrast to [17]ARF1, the equally
water-soluble full-length unmyrARF1GDP is not readily
activated by Arno in solution. This suggests that the amino-terminal
helix somehow hinders the catalyzed exchange reaction, just as it
hinders the spontaneous exchange. Indeed, we previously reported that at physiological Mg2+ levels, GDP does not spontaneously
dissociate from unmyrARF, whereas it is significantly released from
myrARF in the presence of phospholipid vesicles (12). Because the
amino-terminal helix in unmyrARFGDP is held in a
hydrophobic cleft (32, 33), we proposed that insertion of the myristate
into the lipid bilayer might slightly displace the helix away from the
protein core and thereby facilitate the opening of the nucleotide
binding site (23). Accordingly, when the myristate is lacking, the
amino-terminal helix would stick to the protein core and maintain the
nucleotide binding site in a locked conformation. In this case the only
way to obtain GDP dissociation is to reduce the affinity of ARF for GDP
by decreasing the Mg2+ concentration (11). The present
study shows that to accelerate Arno-catalyzed GDP release from
unmyrARF1, it is necessary to add phospholipids, to decrease the ionic
strength of the medium and to lower the concentration of free
Mg2+. Moreover, we show that the effect of phospholipids is
not entirely due to Arno binding to the vesicles, because a
lipid-dependent enhancement of the catalyzed exchange is
still observed when Arno is replaced by the sole Sec7 domain. Previous
studies measuring the co-sedimentation of ARF with vesicles of defined
lipid composition have shown that myrARFGDP weakly
interacts with phospholipids through both hydrophobic interaction of
the myristate with the bilayer and electrostatic interactions of
cationic residues with anionic lipids (14). The ionic interactions,
presumably due to the positive patch formed on the surface of the
protein by residues K10, K15, K16, K59, and R178 (32, 33), were not
detected between unmyrARFGDP and phospholipids at a
physiological ionic strength (14), but they should be stronger at a
lower salt concentration. Accordingly, the requirement for a low ionic
strength observed here could reflect the need to reinforce these
electrostatic interactions to compensate for the absence of the
myristate. Whether decreasing the Mg2+ concentration only
affects the affinity for GDP or also facilitates the movement of the
amino-terminal helix is not known, but altogether our observations
suggest that efficient activation of full-length ARF1 by Arno requires
a membrane surface and involves an interaction of the bilayer with the
amino-terminal portion of ARFGDP.
An unresolved issue, however, is the exact sequence of events during the exchange reaction. Does the helix interact with the bilayer before or after the binding of Arno to ARF? In the first alternative, the binding site of Arno would be unmasked consecutively to the movement of the helix. In the case of the bacterial elongation factor EF-Tu and its nucleotide exchange factor EF-Ts, it has been proposed that the exchanger breaks open the nucleotide binding site by disrupting the coordination pattern of the magnesium ion (34). If the same mechanism applies to Arno, it is possible that the nucleotide binding site of ARF has to be unlocked before the interaction with Arno. Accordingly, the nucleotide binding site of ARFGDP would take two possible conformations: (a) a closed conformation, corresponding to the crystal structure of unmyrARFGDP (32, 33), and (b) a looser conformation induced by a slight release of the amino-terminal helix or/and by chelating the magnesium. Arno would recognize only the latter conformation.
Alternatively, Arno could bind normally to ARFGDP with the amino-terminal helix packing against the protein core but would not be able to open the nucleotide binding site on its own, without the help of phospholipids to pull away the amino-terminal helix. We cannot presently discriminate between the two possibilities.
Whatever the exact mechanism, the present findings point to the importance of the amino-terminal helix of ARF1 and of the protein-phospholipid interactions in the activation of ARF1 by Arno. We propose that efficient catalysis of GDP release by Arno can only occur on a membrane surface with the two proteins interacting with phospholipids: Arno interacting with PIP2 by its PH domain, and ARF interacting with anionic phospholipids by its amino-terminal portion. Thus, even though Arno and ARFGDP can potentially interact in solution, the presence of the amino-terminal helix precludes any futile nucleotide exchange and warrants the membrane localization of the exchange, whereas the PIP2 requirement for Arno binding may serve to target activated ARF to specific locations.
We thank Gabrielle Imbs for skillful secretarial assistance.