HIV-1 Nef stabilizes the association of adaptor protein complexes with membranes.

The maximal virulence of HIV-1 requires Nef, a virally encoded peripheral membrane protein. Nef binds to the adaptor protein (AP) complexes of coated vesicles, inducing an expansion of the endosomal compartment and altering the surface expression of cellular proteins including CD4 and class I major histocompatibility complex. Here, we show that Nef stabilizes the association of AP-1 and AP-3 with membranes. These complexes remained with Nef on juxtanuclear membranes despite the treatment of cells with brefeldin A, which induced the release of ADP-ribosylation factor 1 (ARF1) from these membranes to the cytosol. Nef also induced a persistent association of AP-1 and AP-3 with membranes despite the expression of dominant-negative ARF1 or the overexpression of an ARF1-GTPase activating protein. Mutational analysis indicated that the direct binding of Nef to the AP complexes is essential for this stabilization. The leucine residues of the EXXXLL motif found in Nef were required for binding to AP-1 and AP-3 in vitro and for the stabilization of these complexes on membranes in vivo, whereas the glutamic acid residue of this motif was required specifically for the binding and stabilization of AP-3. These data indicate that Nef mediates the persistent attachment of AP-1 and AP-3 to membranes by an ARF1-independent mechanism. The stabilization of these complexes on membranes may underlie the pleiotropic effects of Nef on protein trafficking within the endosomal system.

Nef is a 27-kDa, myristoylated, peripheral membrane protein that alters the trafficking of transmembrane proteins within the endosomal system and is required for the maximal virulence of HIV-1 (1). Nef misdirects CD4 from the plasma membrane and the Golgi to lysosomes, and it retains class I major histocompatibility complex (MHC) 1 in the trans-Golgi region (2)(3)(4). When fused to the transmembrane domains of integral membrane proteins such as CD4 or CD8, Nef increases the rate of internalization of these chimeras from the cell surface (3,5). These observations led to the hypothesis that Nef functions as a connector between the cytoplasmic domains of cellular proteins and the membrane trafficking machinery. This hypothesis was supported by the observations that Nef binds vesicle coat proteins including the adaptor protein (AP) complexes (6 -9). However, proteins whose trafficking is affected by HIV-1 Nef now include mature class II MHC, the immature class II MHC/invariant chain complex, the membrane-anchored cytokines TNF and LIGHT, CD28, and DC-SIGN (4, 10 -13). These data indicate a general effect of Nef on the trafficking of membrane proteins within the endosomal system. This general effect is supported by the observation that Nef induces a morphologic expansion of endosomal membranes (14,15).
Membrane trafficking within the endosomal system is mediated in large part by vesicles coated with AP complexes (16,17). Of the four members of the AP complex family, AP-1, AP-3, and AP-4 coat vesicles that mediate transport between the trans-Golgi, endosomes, and lysosomes, whereas AP-2 coats vesicles that mediate endocytosis. The complexes are involved both in the formation and budding of coated vesicles as well as in their selection of cargo. This selection requires the recognition of specific sequences in the cytoplasmic domains of transmembrane proteins (18). These sequences are tyrosine-or leucinebased motifs, which mediate direct binding to the AP complexes (19 -21).
HIV-1 Nef contains a canonical leucine-based AP-binding motif (EXXXLL) within a solvent exposed, unstructured loop near the C terminus of the protein (5,9,22). However, the role of this motif in the specific interactions of Nef with the various AP complexes is incompletely defined (5,9,(22)(23)(24). Leucines of the EXXXLL motif are required for the binding of Nef to AP-1 (5). These residues are also required for the binding of Nef to the medium subunit of AP-3 (23), but binding of Nef to intact AP-3 has not been demonstrated. The role of the glutamic acid residue in the Nef EXXXLL motif is untested, but analogous residues have been implicated in the binding of several mammalian and yeast proteins to AP-3 (25,26). Although peptides containing the EXXXLL sequence of Nef or the DKQTLL sequence of the T cell receptor ␥ chain are competitive for binding to the ␤ subunit of AP-2, the direct binding of HIV-1 Nef to intact AP-2 is extremely weak (9,24).
With the exception of AP-2, the association of the AP complexes with membranes is regulated by ADP-ribosylation factor 1 (ARF1) (27)(28)(29)(30). ARF1 undergoes a cycle of membrane association and dissociation controlled by a myristoyl-switch mechanism, which is in turn regulated by guanine-nucleotide exchange and GTP hydrolysis (27,31). ARF1 associates with membranes when bound to GTP. The GTP-bound state of ARF1 is induced by guanine nucleotide exchange proteins (GEPs), which catalyze the exchange of GDP for GTP. Certain Golgiassociated ARF1-GEPs are inhibited by the fungal metabolite brefeldin A (BFA) (32)(33)(34)(35). By inhibiting these GEPs, BFA induces the GDP-bound state of ARF1 and causes the dissociation of ARF1 and ARF1-dependent coats from membranes. ARF1 also dissociates from membranes in response to GTPaseactivating proteins (GAPs), which cause the hydrolysis of GTP to GDP on ARF (36). The physiologic action of ARF1-GAPs allows the dissociation of ARF1-dependent vesicle coats from membranes during vesicular transport (37). When membraneassociated, ARF1 generates a high-affinity binding site for the AP complexes (38). The nature of this binding site is not known. ARF1 may activate an unidentified "docking protein" for the AP complexes (38 -41), or it may induce a modification of membrane phospholipids that enables AP binding (42). Alternatively, ARF1 may itself be the docking protein, an hypothesis supported by its ability to bind directly to subunits of the AP-1 and AP-4 complexes (43,44).
To understand the basis of the pleiotropic effects of Nef on protein trafficking within the endosomal system, we tested the hypothesis that Nef regulates the association of AP complexes with membranes. We report that Nef causes the persistent attachment of AP-1 and AP-3 to juxtanuclear membranes despite experimental perturbations of the ARF1-based regulatory system. Nef neither displayed ARF-like activity nor did it modulate ARF1 or its regulators in vitro. Instead, the direct binding of Nef to AP complexes appears crucial for this membrane stabilization, because mutations in the leucine-based motif of Nef affected both the binding to AP-1 and AP-3 in vitro and the stabilization of these complexes on membranes in vivo. These findings support the hypothesis that a persistent attachment of AP complexes to membranes underlies a general effect of Nef on the endosomal system.

Plasmid Constructions
Plasmid vectors for the expression of wild-type or mutated HIV-1Lai Nef fused to the extracellular and transmembrane domains of the human CD8 ␣ chain were constructed in the pRcCMV plasmid (Invitrogen) as described (14). Nef point mutants were generated by PCRdirected mutagenesis using appropriate primers as described (14). The pCNstop plasmid used for expression of CD8Stop, which lacks a cytoplasmic tail, was provided by A. Baur (Erlangen, Germany). The plasmid vector for the expression of the Nef-GFP fusion has been described (23). Plasmids for the expression in Escherichia coli of the wild-type or mutated Nef fused to GST were constructed in the pGEX-4T2 plasmid (Amersham Biosciences) as described (6). Plasmids expressing ARF1/ T31N-HA and ARF1-HA were provided by Julie Donaldson (45). The plasmid expressing ARF1-GAP-His was provided by Victor Hsu (36,46).

Cell Lines and Transfections
HeLa cells were grown in Dulbecco's modified Eagle's medium with Glutamax (Invitrogen) supplemented with 10% fetal calf serum (Invitrogen), 100 units/ml penicillin (Invitrogen), and 100 g/ml streptomycin (Invitrogen). Cells were transfected either by electroporation as described with 12 g of plasmid DNA (14) or lipid-mediated transfer using 4 g of DNA and FuGENE reagent (Roche Molecular Biochemicals) or Lipofectin (Invitrogen).

Indirect Immunofluorescence Microscopy
After transfection with the indicated plasmids, HeLa cells were spread on glass coverslips in 24-well plates (8 ϫ 10 4 cells/well) and then stained for immunofluorescence 24 h later as described (14). When indicated, the cells were treated with BFA (Sigma or Epicentre) at 37°C before fixation; the concentration and duration of BFA treatment are indicated in the figure legends. For anti-␦-adaptin staining, cells were fixed in 4% paraformaldehyde in PBS, quenched for 10 min with 0.1 M glycine in PBS and permeabilized for 10 min with 0.1% Triton in PBS. For anti-␥-adaptin staining, cells were either fixed with methanol for 10 min at Ϫ20°C and then permeabilized as described above or fixed with 3% paraformaldehyde and permeablized with 0.1% Nonidet P-40. After permeabilization, cells were incubated for 30 min with 0.2% bovine serum albumin in PBS, then successively incubated for 30 min at room temperature with primary and secondary antibody mixtures to stain the adaptins and HA or poly-His epitope tags. Cells were then washed, blocked for 10 min with 10% mouse serum in PBS, and stained for 30 min with CD8fluorescein isothiocyanate mAb. Cells transfected with plasmid expressing Nef-GFP were stained only by indirect immunofluorescence for the indicated adaptin. Coverslips were mounted on slides using immunofluor mounting medium (ICN). Confocal microscopy was performed with a Bio-Rad MRC1000 instrument or a Zeiss microscope with a Bio-Rad laser scanning confocal attachment. Images were collected using single fluorescence excitation and acquisition; the absence of crossover between the signals from the doubly and triply labeled cells was confirmed using appropriate controls. Images were processed using Adobe Photoshop software.

In Vitro Assay for Binding between Nef and AP Complexes
HeLa cells (10 7 ) were lysed in 50 mM Tris-HCl (pH 8), 5 mM EDTA, 150 mM NaCl, and 1% Triton X-100. The cytoplasmic lysates were incubated overnight at 4°C with 2 g of GST or GST-Nef proteins immobilized on GSH-Sepharose beads (Amersham Biosciences). Beads were washed five times in lysis buffer. Bound cellular proteins were analyzed by Western blotting using anti-␥-and anti-␦-adaptin antibod-ies and chemiluminescent detection; the signals were quantified using NIH Image Software.
ARF GAP Assays-Hydrolysis of GTP on [⌬17]ARF1 was assayed by a modification of the assay described (52). GAP assays contained 150 nM [␥-32 P]GTP-loaded [⌬17]ARF1, 25 mM Hepes (pH 7.4), 5 mM MgCl 2 , 100 mM KCl, 1 mM dithiothreitol, and 0.5 mM AMP-PNP, and GAP1 as indicated in a final volume of 40 l. Reactions were incubated at 25°C in the absence of coatomer with the indicated amounts of ARF-GAP1, or in the presence of coatomer (100 nM) and 300 nM ARF-GAP1. Reactions were terminated by the addition of 0.5 ml of cold charcoal suspension (5% charcoal in 50 mM NaH 2 PO 4 ). Following centrifugation, the amount of 32 P i in the supernatant was determined.

HIV-1 Nef Co-localizes with AP-1 and AP-3 Complexes on
Juxtanuclear Membranes-The subcellular distribution of Nef in relation to AP complexes was examined by confocal immunofluorescence microscopy (Fig. 1). For these studies, we used two fusion proteins: one in which the green fluorescent protein (GFP) was appended to the C terminus of Nef (Nef-GFP (23,53)), and the other in which the entire Nef sequence was appended to the lumenal and transmembrane domains of CD8, generating a membrane protein containing Nef as the cytoplasmic domain (CD8-Nef (14,54)). These chimeras are functional for the down-regulation of CD4 and class I MHC (data not shown and Fig. 2) and have been extensively used to analyze the interaction of Nef with the endocytic machinery (3,5,8,14,24,54). In HeLa cells, Nef-GFP was concentrated in a juxtanuclear region near the cell center, the region in which AP-1 complexes were also concentrated (Fig. 1A). CD8-Nef was also concentrated in a juxtanuclear region (Fig. 1B), and its distribution overlapped extensively with AP-1 and AP-3, but not with AP-2. Notably, the distribution of Nef-GFP included a cytoplasmic component, but that of CD8-Nef did not (Figs. 1 and 3), because CD8-Nef is exclusively membrane-associated, whereas Nef-GFP, which associates with membranes via its N-terminal myristoyl group, is in part cytoplasmic. Furthermore, in most cells that expressed CD8-Nef, the intensity of staining for AP-1 and AP-3 in the juxtanuclear region was greater than in cells that did not express the chimera. This supraphysiologic recruitment of AP-1 and AP-3 by CD8-Nef was reminiscent of the recruitment of AP-1 or AP-3 to juxtanuclear membranes caused by the overexpression of transmembrane chimeras containing the cytoplasmic domains of the mannose 6-phosphate receptor, Lamp I, or Limp II (55). These data confirm that the CD8-Nef chimera provides an optimal experimental system for studying the relationship between Nef and the AP complexes.
Nef Sequesters CD4 and Class I MHC in a Condensed Transferrin Receptor-positive Compartment-Ultrastructural analysis by electron microscopy revealed that Nef expression severely affects the morphology of the endosomal compartment (14,15). To identify more precisely the membrane systems affected by Nef, we characterized the compartment in which CD8-Nef resides by co-staining for cellular markers of the endocytic pathway (Fig. 2). CD8-Nef co-localized extensively with transferrin receptor in the juxtanuclear region near the cell center. Strikingly, CD8-Nef induced a marked increase in transferrin receptor staining in this juxtanuclear region. These data indicate that Nef induces an expansion of this compartment, presumably via condensation of peripheral endosomal membranes into this region. Notably, both CD4 and class I MHC co-localized with Nef in the juxtanuclear region (Fig. 2). In contrast, little to no co-localization was observed between Nef and CD63, a marker of late endosomes and lysosomes (Fig.  2), or between Nef and sialyltransferase, a Golgi marker (data not shown). Together, these observations suggest that Nef sequesters CD4 and class I MHC in a perinuclear transferrin receptor-positive endosomal compartment at steady state.
Nef Renders the Juxtanuclear Distribution of AP-1 and AP-3 Resistant to BFA-We hypothesized that Nef might alter the endosomal system by regulating the membrane association of AP complexes. To test this, we treated Nef-expressing cells with BFA, a fungal metabolite that inhibits Golgi-associated ARF1 guanine nucleotide exchange proteins (GEPs) and causes the release of ARF1 and ARF1-dependent coat components from membranes (Fig. 3). In cells that did not express Nef, AP-1 and AP-3 became diffusely cytoplasmic because of dissociation HeLa cells were transfected with a plasmid expressing Nef with GFP appended to its C terminus (Nef-GFP). Twenty-four hours later, the cells were fixed and stained using indirect immunofluorescence for ␥-adaptin (a specific subunit of the AP-1 complex), then examined by confocal microscopy. Green, GFP; red, ␥-adaptin. B, subcellular distribution of CD8-Nef. HeLa cells were transfected with a plasmid expressing Nef as the cytoplasmic domain of the transmembrane protein, CD8 (CD8-Nef). Twenty-four hours later, the cells were fixed and stained using indirect immunofluorescence for the indicated adaptins (red), followed by direct immunofluorescence for CD8-Nef (green). The cells were visualized using confocal microscopy. ␣-Adaptin is a specific subunit of AP-2; ␦-adaptin is a specific subunit of AP-3. Scale bar, 10 m. from membranes. In contrast, AP-1 and AP-3 remained concentrated with Nef in a juxtanuclear region despite BFA treatment in cells expressing either Nef-GFP (panel A) or CD8-Nef (panel B). The juxtanuclear distribution of AP-1 and AP-3 was not preserved in BFA-treated cells that expressed a CD8 construct lacking a cytoplasmic domain, which localized predominantly to the plasma membrane (data not shown). These data indicate that Nef stabilizes the association of AP-1 and AP-3 with membranes. This stabilization was also observed following infection of CD4-positive HeLa cells with wild-type virus, indicating that it occurs at physiological levels of Nef expression (data not shown).

The Nef-mediated Association of AP-1 and AP-3 with Juxtanuclear Membranes Is Resistant to the Expression of Dominant Negative ARF1 and the Overexpression of ARF1-GAP-
The resistance to BFA suggested that the Nef-mediated membrane stabilization of the AP complexes was independent of ARF1 activity. To test this hypothesis, we first expressed a dominant negative ARF1 mutant (ARF1/T31N), either with or without co-expression of CD8-Nef (Fig. 4A). ARF1/T31N is defective in guanine nucleotide binding and causes the dissociation of ARF1-dependent coats from membranes, presumably as a consequence of sequestration of ARF1-GEPs (56). In cells expressing ARF1/T31N alone, the AP-1 staining was diffuse and faint (Fig. 4A, upper panel), similar to that observed in cells treated with BFA. In contrast, AP-1 remained concentrated in the juxtanuclear region in cells co-expressing ARF1/ T31N and CD8-Nef (Fig. 4A, lower panels). Similarly, AP-3 remained concentrated in the juxtanuclear region of cells coexpressing ARF1/T31N and CD8-Nef (data not shown).
The hypothesis that the Nef-mediated stabilization was independent of ARF1 was tested further by overexpression of ARF-GAP1 (Fig. 4B). Overexpression of ARF-GAP1 causes the release from membranes of the ARF1-dependent vesicle coat components as a consequence of the hydrolysis of ARF1-bound GTP (57). In cells overexpressing ARF-GAP1 alone, the AP-1 staining was diffuse and faint (Fig. 4B, upper panel), similar to that observed in cells treated with BFA or expressing ARF1/ T31N. In contrast, in cells co-expressing exogenous ARF-GAP1 and CD8-Nef, AP-1 remained concentrated in the juxtanuclear region (Fig. 4B, lower panel). Similarly, AP-3 remained concentrated in the juxtanuclear region of Nef-expressing cells despite the overexpression of ARF-GAP (data not shown). Altogether, these results support the hypothesis that the expression of Nef induces membrane stabilization of AP-1 and AP-3 via an ARF1independent mechanism.
Nef Does Not Modulate the GTPase Cycle of ARF1-Because mimicry or modulation of the ARF1-based regulatory system could explain the stabilization of AP complexes induced by Nef, we assessed using in vitro assays whether Nef might display ARF1-like or ARF1-GEP activities, or alternatively might stim-

FIG. 3. Nef confers resistance of AP-1 and AP-3 to the membrane-dissociating effect of brefeldin A.
HeLa cells were transfected with the plasmid that expresses Nef-GFP (panel A) or CD8-Nef (panel B), then examined by confocal immunofluorescence microscopy to detect Nef-GFP and CD8-Nef (green) and either AP-1 (␥-adaptin, red) or AP-3 (␦-adaptin, red). Where indicated in panel A, cells were treated with 5 g/ml BFA for 30 min before fixation and staining; where indicated in panel B, cells were treated with 10 g/ml BFA for 15 min before fixation and staining. Scale bar, 10 m. ulate ARF1-GEPs or inhibit ARF-GAP (Fig. 5). First, we tested whether Nef might be an ARF-like protein and bind GTP. Although no similarity between Nef and ARF1 in primary or tertiary structure is apparent, and Nef does not contain any canonical GTP-binding motifs, controversial GTP-binding properties of Nef were initially reported (58 -60). As indicated in Fig. 5A, GST-Nef did not bind GTP, even in the presence of a highly active ARF-GEP (the Sec7 domain of ARNO), whereas a truncated form of ARF, [⌬17]ARF1, could be activated by the Sec7 domain of ARNO and bound GTP efficiently under the same conditions. Second, stoichiometric amounts of GST-Nef had no effect on the kinetics of spontaneous nucleotide exchange on ⌬17-ARF1, measured by tryptophan fluorescence (Fig. 5B). Nef also had no effect on the exchange activity of catalytic amounts of the ARF1-GEPs ARNO (Fig. 5B) or cytohesin-1 (data not shown). Third, GST-Nef did not affect the activity of ARF-GAP1, either at high ARF1-GAP concentrations (Fig. 5C) or in the presence of coatomer (Fig. 5D), which strongly increases ARF-GAP1 activity (52,61).
These in vitro data indicated that the stabilization of AP complexes in vivo is unlikely to be mediated by an ARF-like or ARF-GEP-like activity of Nef or by modulation of the activity of the proteins involved in the regulation of ARF1. In addition, we did not detect any binding between Nef and ARF1 in either yeast two-hybrid or GST pull-down assays (data not shown), suggesting that Nef is unlikely to mediate the stabilization of AP complexes by recruiting ARF1 itself to membranes.
Nef Maintains the Juxtnuclear Concentration of AP-1 in BFA-treated Cells Despite the Dispersal of ARF1 into the Cytosol-To further exclude a role for ARF1 in the Nef-mediated stabilization of AP complexes, we compared the distributions of ARF1, AP-1, and Nef in BFA-treated cells. Cells were transfected with a vector expressing an HA-tagged wild-type ARF1 (ARF1-HA), in combination with the CD8-Nef expression vector, and were then examined by immunofluorescence microscopy after BFA treatment (Fig. 6). In the absence of Nef, ARF1-HA and AP-1 were concentrated in the juxtanuclear region in untreated cells, and both became dispersed throughout the cytoplasm when the cells were treated with BFA (Fig. 6, top and middle panels). However, in BFA-treated cells that expressed CD8-Nef (Fig. 6, lower panels), AP-1 remained concentrated in a punctate, juxtanuclear distribution, whereas the distribution of ARF1-HA became diffuse and cytosolic. These data definitively confirm that the membrane stabilization of AP complexes induced by Nef in BFA-treated cells is independent of ARF1.

The Stabilization of AP-1 and AP-3 on Juxtanuclear Membranes Requires the Leucine-based motif in Nef-HIV-1
Nef contains a leucine-based motif (EXXXLL) required for the association with AP complexes (see Fig. 7B, and Refs. 5 and 22). To determine whether the membrane-stabilization effect required AP-binding mediated by this motif, alanine substitutions were introduced in place of the leucine residues of the EXXXLL sequence. The resulting CD8-NefL164A/L165A mutant was tested for the ability to maintain the juxtanuclear concentration of AP-1 and AP-3 in the presence of BFA (Fig.  7A). As previously reported (14), mutation of these leucine residues caused a significant fraction of CD8-Nef to localize to the periphery of the cytoplasm and plasma membrane. Strikingly, this mutation completely abolished the ability of Nef to maintain the juxtanuclear concentration of either AP-1 or AP-3 in cells treated with BFA. We checked that the leucine-based motif of Nef was also required for direct recruitment of both AP-1 and AP-3 complexes. Recombinant GST-Nef fusions were used to analyze the interaction between Nef and the intact AP complexes from cell lysates (Fig. 7B). The dependence of the binding between Nef and AP-1 on the leucine-based motif was confirmed (5). In addition, we documented that Nef binds the intact AP-3 complex, and this binding was also leucinedependent.
We next examined the role of the glutamic acid residue in the EXXXLL motif, because the presence of an acidic residue at this location in leucine-based motifs has been associated with the ability to bind AP-3 complexes in vitro and to utilize AP-3based sorting pathways in vivo (25,26). A mutation encoding alanine substitution E160A was introduced into the CD8-Nef chimera, and the mutant was tested for the ability to stabilize the membrane association of AP-1 and AP-3 in the presence of BFA (Fig. 7A). This mutant was concentrated in the juxtanuclear region, but in non-BFA-treated cells it did not colocalize significantly with AP-3 ( Fig. 7A and data not shown). In BFA-treated cells, NefE160A failed to stabilize AP-3 on membranes, and it was less effective than wild-type Nef in stabilizing AP-1 (Fig. 7A). In GST pull-down assays, NefE160A FIG. 4. Nef overcomes the membrane-dissociating effects of the dominant negative ARF1/T31N mutant and the overexpression of ARF-GAP1. A, HeLa cells were transfected with a plasmid expressing the epitope-tagged, dominant negative mutant ARF1/T31N-HA, either with or without the plasmid that expresses CD8-Nef. Twenty-four hours later, the cells were fixed and stained using indirect immunofluorescence for AP-1 (␥-adaptin, red) and for ARF1/T31N-HA (blue), followed by direct immunofluorescence for CD8 (green). The upper row of images shows cells that were transfected only with the plasmid expressing ARF1/T31N-HA, and the lower row of images shows cells co-transfected with both the plasmids expressing ARF1/T31N-HA and CD8-Nef. Cells expressing ARF1/T31N-HA are indicated by arrows. Scale bar, 10 m. B, HeLa cells were transfected with a plasmid expressing ARF1-GAP (polyhistidine-tagged), either with or without the plasmid expressing CD8-Nef. Twenty-four hours later, the cells were fixed and stained using indirect immunofluorescence for AP-1 (␥-adaptin, red) and for ARF1-GAP (blue), followed by direct immunofluorescence for CD8 (green). The upper row of images shows cells that were transfected only with the plasmid expressing ARF1-GAP, and the lower row of images shows cells that were co-transfected with both plasmids expressing ARF1-GAP and CD8-Nef. Cells expressing ARF1-GAP are indicated by arrows. Scale bar, 10 m.
bound AP-1 with reduced efficiency (ϳ38% of wild-type activity), but it did not bind at all to AP-3 (Fig. 7B).
Together, these data indicate that the EXXXLL motif is important for the binding of Nef to AP-1 and AP-3 in vitro and for the Nef-induced stabilization of these complexes on membranes in vivo. The data also indicate that the acidic residue of the Nef EXXXLL motif is especially critical for the binding to AP-3 and the stabilization of AP-3 on membranes. DISCUSSION The pleiotropic effects of the HIV-1 Nef protein on the trafficking of cellular transmembrane proteins indicate a general perturbation in the regulation of vesicular transport throughout the endosomal system. The data herein support the hypothesis that this perturbation occurs at the level of the regulation of the membrane attachment of vesicle coat components. Specifically, Nef stabilizes the association of AP-1 and AP-3 complexes on juxtanuclear membranes by a direct binding mechanism that is independent of ARF1. First, Nef conferred resistance of AP-1 and AP-3 to the membrane-dissociating effects of BFA, dominant negative ARF1 (ARF1/T31N), and ARF-GAP1. Second, Nef lacked ARF-like or ARF-GEP-like activity, and it neither stimulated ARF-GEPs nor inhibited ARF-GAP1. Third, Nef maintained the juxtanuclear membrane-association of AP-1 despite the cytosolic dispersal of ARF1 in BFA-treated cells. These data indicate that Nef causes the persistent association of AP complexes with membranes, even when their physiologic attachment as mediated by ARF1 is inhibited.
The leucine-based motif in Nef was required for both the binding of Nef to AP-1 and AP-3 in vitro and the membrane stabilization of these complexes in vivo. This motif has been associated with most of the perturbations in protein trafficking induced by Nef, including the down-regulation of CD4 and CD28, and the up-regulation of DC-SIGN, the immature class II MHC/invariant chain complex, and the membrane-anchored cytokines TNF and LIGHT (9 -13).
The correlation between the ability of Nef to interact with specific AP complexes and to stabilize these complexes on membranes was supported further by the selective importance of the acidic residue within the Nef EXXXLL motif for binding and membrane stabilization of AP-3. The role of this glutamic acid residue for the specific recruitment of AP-3 was predicted based on the requirement of analogous acidic residues within the leucine-based sorting motifs of Limp II and tyrosinase for the binding to AP-3 in vitro (25). Similarly, the AP-3-mediated transport of the Vam3p protein to the vacuole in yeast requires an analogous acidic residue (26).
Two models can explain the persistent membrane association of AP complexes induced by Nef. First, Nef may mediate de novo attachment events. As a peripheral membrane protein that binds AP-1 and AP-3, Nef may be a constitutively active viral homologue of a putative cellular docking protein for these complexes (38 -41). Alternatively, Nef and ARF1 may be similar insofar as each is able to constitute or activate the membrane attachment sites for AP complexes. In support of such a similarity, constitutively active ARF1 (ARF1/Q71L), which is locked in the GTP-bound state and constitutively associates with membranes, induces a BFA-resistant association of AP complexes with membranes similar to that induced by Nef (Refs. 29 and 63, and data not shown). Mechanistically, both Nef and ARF1 bind AP complex subunits, so either may directly bind AP complexes to membranes (43,44). Alternatively, both Nef and ARF1 may generate AP binding sites indirectly by recruiting phospholipid-modifying enzymes to membranes (42,62).
The second model is that Nef causes the persistent attach- D, Nef does not affect GAP1 activity in the presence of coatomer. GAP activity was measured as above using [␥-32 P]GTP-loaded [⌬17]ARF1, ARF-GAP1 (300 nM), and coatomer (100 nM). GST-Nef or GST (5 M) were included together with GTP-loaded ARF, ARF-GAP1, and coatomer as indicated. Open circles, GTP hydrolysis under these conditions without coatomer; filled circles, GTP hydrolysis with coatomer but without GST or GST-Nef. Data are the percentage of ARF-bound GTP hydrolyzed. ment of AP complexes by inhibiting their release from membranes. In this model, the initial attachment of the complexes to membranes remains mediated by ARF1, but the subsequent interaction with Nef causes a decrease in the rate at which the complexes cycle off membranes. This model is particularly compatible with the experiments using BFA, in which the block to ARF1-mediated attachment of the complexes is introduced after the expression of Nef. The model is potentially less compatible with the experiments in which dominant negative ARF1 or ARF-GAP1 are expressed concurrently with Nef, but it remains formally possible. The regulation of the dissociation of AP complexes from membranes is not well understood. However, this dissociation is presumably a prerequisite to the fusion of transport vesicles with target membranes.
Whether caused by de novo attachment or a block to release, how would the membrane stabilization of AP complexes by Nef relate to its effects on protein trafficking? We observe herein that Nef expands a perinuclear endosomal compartment and sequesters CD4 and class I MHC in this membrane system (see Fig. 2). The membrane stabilization of AP-1 and AP-3 may induce the expansion of this compartment either by blocking the formation of donor vesicles on endosomal membranes or by inhibiting coat dissociation and the subsequent fusion of vesicles derived from endosomal membranes with their targets. Either effect would cause the accumulation of AP-binding cargoes within endosomal structures. This model predicts that Nef would sequester intracellularly receptors that normally return to the cell surface after endocytosis and transit through recycling endosomes. Such a block to recycling has been proposed as part of the mechanism of Nef-mediated down-regulation of surface CD4 (8,64). This model accounts for a general downregulation of AP-binding cell-surface proteins by Nef, but how can the up-regulation of certain molecules be explained (10, 12, 13)? We hypothesize that the sequestration of AP complexes by Nef leads to a competitive inhibition of access to the complexes for some cellular proteins, causing their rerouting by default to the plasma membrane (22). Consequently, Nef may either upregulate or down-regulate the surface expression of a specific protein, depending on whether the protein in question is competitive with Nef for AP complexes and is excluded from APcoated transport vesicles, or whether it is noncompetitive with Nef and is trapped in an expanded endosomal compartment.
Notably, the de novo attachment model leads to the question of why HIV-1 would induce the membrane association of AP complexes independently of the physiologic regulatory mechanism. One possibility is that such an activity enables the recruitment of AP complexes to nonphysiologic locations. For example, Nef is a virion-associated protein (65), and it may recruit AP complexes to sites of virion assembly along the plasma membrane. This could facilitate an aspect of viral morphogenesis such as the incorporation of the envelope glycopro-FIG. 6. Nef preserves the juxtanuclear membrane association of AP-1 in BFA-treated cells despite the cytosolic dispersal of ARF1. HeLa cells were transfected with a plasmid expressing wild-type ARF1 (HA-tagged), either with or without the plasmid expressing CD8-Nef. Twenty-four hours later, the cells were fixed and stained using indirect immunofluorescence for AP-1 (␥-adaptin, red) and ARF1-HA (green), followed by direct immunofluorescence for CD8-Nef (blue). Upper and middle panels, cells were transfected only with the plasmid expressing ARF1-HA and treated (middle panels) or not (upper panels) with 10 g/ml BFA for 15 min before fixation and staining. Lower panels, cells were transfected with both plasmids expressing ARF1-HA and CD8-Nef, and then treated with 10 g/ml BFA for 15 min before fixation and staining. Scale bar, 10 m. 7. The BFA-resistant membrane association of AP-1 and AP-3 requires the leucine-based AP-binding motif in Nef. A, HeLa cells were transfected with plasmids expressing either the wildtype Nef (CD8-Nef wt) or mutated CD8-Nef chimeras (NefL164A/L165A and Nef E160A), then fixed and stained 24 h later for CD8-Nef (green) and either AP-1 (␥-adaptin, red) or AP-3 (␦-adaptin, red). The cells were treated with 10 g/ml BFA for 15 min before fixation. Scale bar, 10 m. B, GST pull-down assays of the interaction of Nef with intact AP-1 and AP-3 complexes. HeLa cell lysates were incubated with equal amounts of purified GST, GST-Nef, GST-NefL164A/L165A, or GST-NefE160A previously immobilized on GSH-agarose beads. Bound proteins were resolved by SDS-PAGE and the association of AP complexes was analyzed by Western blotting with anti-␥-adaptin (AP-1) or anti-␦-adaptin (AP-3); the bands were quantified using NIH Image software and the data expressed as the percent of binding relative to wild-type Nef. Unfractionated HeLa cell lysates were run as a control for the detection of ␥and ␦-adaptin (left lanes). The Coomassie Blue-stained gel documents the equal relative loading of the GST fusion proteins on the beads. tein, the cytoplasmic domain of which contains AP-binding motifs (66,67). In support of this scenario, the leucine-based AP-binding motif in Nef is required not only for the membrane stabilization of AP complexes described here but also for the optimal infectivity of HIV-1 virions (22).