ADP Ribosylation Factor 1 Activity Is Required To Recruit AP-1 to the Major Histocompatibility Complex Class I (MHC-I) Cytoplasmic Tail and Disrupt MHC-I Trafficking in HIV-1-Infected Primary T Cells^,

Abstract

HIV-1-infected cells are partially resistant to anti-HIV cytotoxic T lymphocytes (CTLs) due to the effects of the HIV Nef protein on antigen presentation by major histocompatibility complex class I (MHC-I), and evidence has been accumulating that this function of Nef is important in vivo. HIV Nef disrupts the normal expression of MHC-I by stabilizing a protein-protein interaction between the clathrin adaptor protein AP-1 and the MHC-I cytoplasmic tail. There is also evidence that Nef activates a phosphatidylinositol 3 kinase (PI3K)-dependent GTPase, ADP ribosylation factor 6 (ARF-6), to stimulate MHC-I internalization. However, the relative importance of these two pathways is unclear. Here we report that a GTPase required for AP-1 activity (ARF-1) was needed for Nef to disrupt MHC-I surface levels, whereas no significant requirement for ARF-6 was observed in Nef-expressing T cell lines and in HIV-infected primary T cells. An ARF-1 inhibitor blocked the ability of Nef to recruit AP-1 to the MHC-I cytoplasmic tail, and a dominant active ARF-1 mutant stabilized the Nef–MHC-I–AP-1 complex. These data support a model in which Nef and ARF-1 stabilize an interaction between MHC-I and AP-1 to disrupt the presentation of HIV-1 epitopes to CTLs.

INTRODUCTION

CD8⁺ cytotoxic T lymphocytes (CTLs) are important for the control of chronic viral infections. In a virally infected cell, major histocompatibility complex class I (MHC-I) molecules present peptides derived from viral proteins. Once the T cell receptor (TCR) on CD8⁺ CTLs recognizes a “non-self” signal presented by MHC-I, the CTL releases perforins and granzymes that kill the virally infected cell, preventing the further spread of the virus (reviewed in reference 3). CTLs play an important role in the control of HIV infection (for a review, see reference 9), and recent evidence indicates that individuals mounting a Gag-specific CTL response have improved parameters with regard to controlling disease (17, 24). Despite the efficacy with which CTLs control viral load early in infection, anti-HIV CTLs ultimately fail to prevent the progression of disease in most infected people.

Studies performed in vitro have shown that the HIV Nef protein protects infected cells from CTL-mediated lysis (8, 28, 54, 63). Nef has been shown to protect HIV-infected primary T cells from CTL lysis using flow cytometric killing assays (8, 28), CTL coculture assays (63), and chromium release assays (54). Although Nef limits the ability of CTLs to recognize and kill infected cells, it does not appear to abrogate the capacity of CTLs to produce inhibitory cytokines in response to infected cells (54). Recent in vivo evidence supports the hypothesis that CTLs may control HIV infection in vivo primarily by the production of inhibitory cytokines but fail to eradicate the infection because the CTLs cannot efficiently lyse the infected cell source of new virions (62).

Nef binds directly to the cytoplasmic tail of MHC-I allotypes (60) and recruits the clathrin adaptor protein AP-1. As a result, MHC-I is transported into an endolysosomal pathway from the trans-Golgi network (TGN), and MHC-1 expression on the cell surface is reduced (43). There is evidence that MHC-I delivered to endosomes via the activity of AP-1 is subsequently targeted for degradation via Nef binding to the COP-I coatomer, β-COP (46).

Nef has also been implicated in promoting the internalization of MHC-I from the cell surface (47). The relative contribution of internalization versus disruption of forward transport to overall MHC-I downmodulation varies depending on the cell type (23) and has been attributed to both ADP ribosylation factor 6 (ARF-6) (4)- and AP-1 (23, 30)-dependent pathways. The relative importance of ARF-6 versus AP-1 has not yet been determined in studies that directly compared their roles in HIV-infected primary T cells.

The adaptor protein AP-1 is a heterotetrameric complex that recognizes trafficking signals in cargo and recruits clathrin-sorting machinery to the trans-Golgi network. AP-1 is made up of μ1, β1, γ, and σ1 subunits, which collectively sort cargo containing Yxxφ or [D/E] xxxLL trafficking signals into the endolysosomal network (for a review, see reference 42). MHC-I downmodulation can be inhibited by knocking down expression of the AP-1 μ1 subunit (43, 46) or by overexpressing a dominant negative mutant of AP-1 μ1 which lacks a functional tyrosine-binding pocket (49, 61). Mutation of a tyrosine residue in the MHC-I cytoplasmic tail disrupts binding of AP-1 to the Nef–MHC-I complex in cells (43, 61) and in experiments using purified proteins (34, 49). Based on these data, the AP-1 tyrosine-binding pocket has been proposed to interact with the MHC-I cytoplasmic tail tyrosine, and this interaction is stabilized by Nef.

Binding of adaptor proteins to trafficking signals is normally regulated by the activity of ADP ribosylation factors (ARFs), which are small GTPases that control the assembly and disassembly of intracellular trafficking complexes. ARF activation and recruitment to cellular membranes is cyclical and regulated by its GTP binding state. Guanine nucleotide exchange factors (GEFs) promote the exchange of GTP for GDP. GTPase-activating proteins (GAPs) support ARF catalysis of GTP and thus are important to inactivate ARF (20; reviewed in reference 12). There are a number of different ARF proteins expressed by cells. ARF-1 is a clathrin regulatory protein that, upon binding GTP, undergoes a conformational change exposing a myristoyl group that inserts into membranes and subsequently stabilizes AP-1 (1, 50, 56) or COP-I coatomer (48, 56) binding to trafficking signals. The ARF-1 GEF inhibitor brefeldin A (BFA) stabilizes an abortive ARF-GDP-bound complex (37), thus preventing ARF-1 cycling.

ARF-6 is a related GTPase that regulates the recycling of cargo that has been internalized by clathrin-independent endocytosis (13, 40). There is evidence that ARF-6 is required for Nef-dependent MHC-I downmodulation in HeLa cells based on experiments with ARF-6 mutants that are “locked” in a GTP-bound, constitutively active state (ARF-6-Q₆₇L) (4, 64).

Here, we used T cell systems, including HIV-infected primary T cells, to demonstrate that ARF-1 activity was required for Nef-dependent MHC-I trafficking via AP-1. Coprecipitation of AP-1 with the MHC-I–Nef complex was specifically inhibited by BFA and stabilized by ARF-1 Q₇₁L (53). In contrast, we were unable to detect a clear requirement for ARF-6. These data help to clarify the relative contributions of the two pathways implicated in HIV immune evasion and highlight the AP-1/ARF-1 pathway as an important potential target for drug development.

MATERIALS AND METHODS

Cell lines and primary cell isolation.

CEM T cells and SupT1 cells expressing hemagglutinin (HA)-tagged HLA-A2 were created and maintained as previously published (61). Primary T cells were purified from buffy coats obtained from the New York Blood Center. Mononuclear cells were purified by Ficoll-Hypaque centrifugation, and primary T lymphocytes were isolated through adherence and CD8 and CD56 depletion of mononuclear cells. Following isolation, T cells were stimulated with phytohemagglutinin (PHA) at 10 μg/ml (Sigma-Aldrich). After 24 h, 50 U/ml interleukin-2 (IL-2) was added to the culture medium. Forty-eight hours after IL-2 stimulation, the T lymphocytes were used for HIV transduction. Donor MHC-I allotype (HLA-A2, HLA-BW4, or HLA-BW6) was determined at the time of CD4⁺ T cell isolation using flow cytometric techniques.

DNA constructs. (i) Construction of pMSCV IRES GFP vectors expressing ARF-1 and ARF-6.

pCB6 expressing Myc-tagged wild-type or T₃₁N ARF-1 was obtained from Didier Trono (Ecole Polytechnique Fédérale de Lausanne). The arf-1 open reading frame (ORF) was amplified by PCR using either wild-type or T₃₁N pCB6 Myc-ARF-1 as a template with the primers listed in Table S1 in the supplemental material. Murine stem cell virus (MSCV) Myc-ARF-1 Q₇₁L internal ribosome entry site (IRES) green fluorescent protein (GFP) was created through standard two-step PCR mutagenesis using wild-type MSCV ARF-1 IRES GFP as a template with the primers listed in Table S1. The PCR products were cloned into the BamHI site of MSCV IRES GFP (pMIG) (57).

pXS expressing HA-tagged wild-type, T₂₇N, or Q₆₇L ARF-6 was obtained from Julie Donaldson (National Institutes of Health). ARF-6 constructs were amplified using pXS HA-ARF-6 as a template with the primers listed in Table S1 in the supplemental material. The PCR product was cloned into BglII and EcoRI sites of pMIG.

(ii) Construction of HIV vectors expressing ARF-1 and ARF-6.

To construct HIVs that also contained both GFP and ARF ORFs, we first made a version of HIV (pNL-GI) in which a portion of the env ORF was replaced by a GFP IRES multiple cloning site cassette. PCR was used to amplify the IRES from pNL-PI (8) and to add additional restriction enzyme sites downstream of the IRES. The PCR product was ligated into the NheI and BglII sites in the env ORF of pNL4-3-deltaE-EGFP (66) just downstream of GFP. pNL-GInef ⁻ was generated by creating a frameshift mutation within the nef ORF of pNL-GI by digesting with XhoI, filling in, and religating the ends.

pNL-GI was then used to create HIV constructs expressing ARF-1 and ARF-6. To create pNL-GIarf constructs, linker primers were designed to create a XbaI site in the 5′ end and a MluI site in the 3′ end of the amplicon during PCR amplification of the ARF-1 and ARF-1 Q₇₁L from MSCV ARF-1 IRES GFP and MSCV Myc-ARF-1 Q₇₁L IRES GFP, respectively. Primers for this step are listed in Table S1 in the supplemental material. Digested PCR products were ligated into the XbaI/MluI sites downstream of the IRES element in pNL-GI.

Due to an internal XbaI site present in ARF-6, pNL-GI ARF-6 +/− nef and pNL-GI ARF-6 Q₆₇L +/− nef were engineered by designing linker primers to create a SpeI site in the 5′ end, which is compatible with XbaI overhang ligation, and a MluI site in the 3′ end of the amplicon during PCR amplification of the ARF-6 and ARF-6 Q₆₇L from MSCV ARF-6 IRES GFP and MSCV ARF-6 Q₆₇L IRES GFP, respectively. PCR primers used are listed in Table S1 in the supplemental material. Digested PCR products were ligated into the XbaI/MluI-digested parental vector. All constructs were confirmed by sequencing.

(iii) shRNA constructs.

FG12 small hairpin RNA (shRNA) lentiviral vectors were constructed as previously described (39, 46). The ShNC construct was previously described (46). The target sequence for shARF-6, beginning at position 247, was as follows: GATCCCCGGTCTCATCTTCGTAGTGGTTCAAGAGACCACTACGAAGATGAGACCTTTTTGGAAA.

Virus preparation and transductions. (i) Retrovirus.

Retroviral supernatants were prepared as described previously (36, 57). Bosc cells (36) were transfected with the MSCV constructs described above, the retrovirus packaging vector pCL-Eco (33), and pHCMV-G (36). Briefly, 5 × 10⁵ CEM or SupT1 cells were spin transduced with 1 ml of retroviral supernatants plus 8 μg/ml Polybrene at 2,500 rpm for 2 h in a tabletop centrifuge at room temperature.

(ii) Adenovirus.

Replication-defective adenovirus was produced by the University of Michigan Gene Vector Core Facility. Adenoviral transductions were performed as previously described (59). Transductions were performed using 1 × 10⁶ cells in 1 ml of RPMI 1640 containing 2% fetal bovine serum; 10 mM HEPES; and 2 mM each penicillin, streptomycin, and glutamine. The multiplicity of infection (MOI) was 200 for CEM and 100 for SupT1 (based on 293 cell infectivity, which is greater than CEM infectivity).

(iii) HIV.

HIV constructs were produced by transfecting 293 cells with each construct and harvesting the supernatant. PHA- and IL-2-stimulated CD8-CD56-depleted T lymphocytes (5 × 10⁵) were transduced using 1 ml of supernatant with 8 μg/ml Polybrene using a spin transduction protocol. Primary T lymphocytes were harvested for analysis of MHC-I downmodulation at 36 h postransduction. CEM-SS cells (5 × 10⁵) were transduced with 0.3, 0.2, or 0.1 ml of pNL4-3 supernatant with 8 μg/ml Polybrene using a spin transduction protocol. CEM-SS cells were harvested at 48 h postransduction.

Flow cytometry and antibodies.

Cells were stained for 20 min on ice in fluorescence-activated cell sorter (FACS) buffer (phosphate-buffered saline [PBS], 1% human serum, 1% fetal bovine serum [FBS], 1% HEPES, and 1% NaN₃), washed, and then stained for 20 additional minutes in secondary antibody. Mouse antibody to HLA-A2 antibody, BB7.2 (35), and mouse antibody to CD4, OKT4 (41), were purified from ascites (22) provided by the University of Michigan Hybridoma Core Facility. Mouse antibody to placental alkaline phosphatase (PLAP) antibody was obtained from Serotec. The secondary antibodies used were goat anti-mouse IgG_2B-phycoerythrin (1:250; Invitrogen) and goat anti-mouse IgG1-Alexa Fluor 647 (1:250; Invitrogen). Isotype controls were obtained from BD Biosciences. Stained cells were analyzed on a Becton Dickinson FACSCanto cytometer. Analysis was performed using FlowJo software (Tree Star Inc.).

Immunofluorescence microscopy and antibodies.

ARF-6-transduced CEM cells were adhered to glass slides with Celltak (BD Biosciences), fixed, permeabilized, and stained for indirect immunofluorescence as previously described (43). Images were collected using a Zeiss LSM 510 confocal microscope and processed using Deneba Canvas software. The following antibodies were utilized to localize proteins via microscopy: anti-HLA-A2 (BB7.2) and anti-GFP (3E6; Invitrogen). The following secondary antibodies were obtained from Molecular Probes and were used at a dilution of 1:250: goat anti-mouse IgG_2B-Alexa Fluor 546 and goat anti-mouse IgG_2A-Alexa Fluor 488.

Immunoprecipitations and Western blotting.

Immunoprecipitations were performed as previously described (61). CEM cells (15 × 10⁶) were transduced with either control or Nef-expressing adenovirus. At 48 h after adenoviral transduction, the cells were spin transduced with either Dulbecco's modified Eagle medium (DMEM) (mock) or MSCV ARF-1 IRES GFP viral supernatants. At 72 h after adenoviral transduction, all cells were incubated in 20 mM NH₄Cl for 16 h. Where indicated, samples were also incubated in 50 μM BFA for 16 h. Cells were then harvested and lysed in 1% digitonin (Wako) lysis buffer, described previously (46, 61). Lysates were normalized for total protein and GFP transduction rates, when appropriate, prior to immunoprecipitation. Input controls were 1% of the immunoprecipitated protein. After an overnight preclear at 4°C, lysates were immunoprecipitated for either HLA-A2 (BB7.2 cross-linked beads) (61) or Myc-tagged ARF-1 (9e10 cross-linked beads) (14). Immunoprecipitates were eluted and analyzed by Western blotting as previously described (59). Western blot antibodies used were anti-Nef (AG11) (6) and anti-Myc (9e10) (14), which were produced by the University of Michigan Hybridoma Core Facility and purified as previously described (22). Antibodies also used for Western blotting were anti-AP-1 γ (BD Biosciences), HA (HA.11, Covance), anti-phosphatase and tensin homolog (anti-PTEN) (Cell Signaling Technology), anti-glyceraldehyde-3-phosphate dehydrogenase (anti-GAPDH) (Santa Cruz Biotechnology), and anti-AP-1 μ1 (RY/1) (55). The secondary antibody for anti-Nef, anti-HA, and anti-Myc was goat anti-mouse IgG1-horseradish peroxidase (HRP) (Zymed Laboratories Inc.). The secondary antibody for anti-AP-1 μ1 and anti-PTEN was goat anti-rabbit-HRP (Zymed Laboratories Inc.). The secondary antibody for AP-1 γ was rabbit anti-mouse-HRP (Zymed Laboratories Inc.).

RESULTS

Functional ARF-1 is required for Nef to disrupt the trafficking of MHC-I.

Nef binds directly to the cytoplasmic tail of MHC-I allotypes (60) and recruits the clathrin adaptor protein AP-1. Bound MHC-I trafficks from the trans-Golgi network (TGN) to an endolysosomal pathway, and MHC-I is ultimately degraded in lysosomal compartments (43). However, it is not known whether Nef-dependent AP-1 binding and clathrin recruitment bypass the normal AP-1 regulatory steps that include the GTPase ARF-1. To determine if ARF-1 activity was necessary for Nef to disrupt the transport of MHC-I to the cell surface, we utilized two ARF-1 mutants, ARF-1 T₃₁N and ARF-1 Q₇₁L. ARF-1 T₃₁N resembles the inactive GDP-bound form of the molecule (10), whereas ARF-1 Q₇₁L resembles the activated GTP-bound form (53). We also tested BFA, which inhibits ARF-1 by disrupting the function of ARF-1 GTP exchange factors (GEFs).

We observed a striking inhibitory effect of dominant active ARF-1 (Q₇₁L) on Nef-dependent MHC-I downmodulation (Fig. 1A; quantified in Fig. 1C). The ARF-1 inhibitor, BFA, also significantly increased MHC-I surface expression in the presence of Nef (Fig. 1A; quantified in Fig. 1D). Overexpression of wild-type and T₃₁N ARF-1 did not significantly affect Nef-dependent MHC-I downmodulation (Fig. 1A; quantified in Fig. 1C; also see Fig. S1 in the supplemental material). However, the lack of a phenotype with T₃₁N may have been due to inadequate expression of the mutant relative to the endogenous wild type, because control experiments failed to show the expected effect of this mutant on Golgi architecture in our system (data not shown). CD4 downmodulation by Nef was unaffected by these treatments, consistent with previously published data showing that the internalization and degradation of CD4 are independent of the GTP-bound state of ARF-1 (Fig. 1A and C) (15).

Fig. 1.

ARF-1 activity is required for Nef-induced downmodulation of HLA-A2. (A) Flow cytometric analysis of HLA-A2, HLA-A2 A₃₂₃V, and CD4 surface expression plus or minus Nef in cells expressing ARF-1 constructs as indicated. Cells expressing ARF-1 vector control were GFP⁺, and these cells were gated on for this analysis. Dark gray shaded curve, vector alone; light gray shaded curve, adeno-Nef; black curve, adeno-Nef plus ARF-1 as indicated. BFA, brefeldin A. (B) Western blot analysis of Myc-ARF-1 expression levels in transduced CEM T cells. (C and D) Quantitation of panel A. The median fluorescence normalized to wild-type ARF-1 is shown ± standard deviation (SD); n ≥ 3. * denotes a P value of <0.05, ** denotes a P value of <0.01, and *** denotes a P value of <0.001.

We also observed an inhibitory effect of BFA and ARF-1 (Q₇₁L) on MHC-I surface expression in the absence of Nef (quantified in Fig. 1C and D). This is expected due to the known effects of these factors on protein export from the endoplasmic reticulum (29). Despite the inhibitory effect of these factors on MHC-I surface expression in the absence of Nef, both BFA and ARF-1 (Q₇₁L) caused a significant increase in MHC-I surface expression in the presence of Nef (Fig. 1). The opposing effects observed in the absence versus the presence of Nef support the conclusion that these inhibitors specifically affect Nef-dependent pathways.

For comparison, we also included a molecule that contained a more natural AP-1 trafficking signal, which was previously shown to bind AP-1 in the absence of Nef. This signal was made by substituting a valine for an alanine in the cytoplasmic tail of MHC-I HLA-A2 to create a Yxxφ sorting signal (Y₃₂₀SQV₃₂₃ [HLA-A2 A₃₂₃V]) (61). Compared to wild-type MHC-I HLA-A2, this molecule has relatively low surface expression due to its interaction with AP-1 (61). Interestingly, the effect of ARF-1 mutants and BFA on the surface expression of HLA-A2 A₃₂₃V in the absence of Nef was somewhat different from what occurred in the presence of Nef. The main difference was that, in the absence of Nef, there was a relatively small but statistically significant further reduction of HLA-A2 A₃₂₃V expression with ARF-1 Q₇₁L (Fig. 1C), whereas in the presence of Nef, ARF-1 Q₇₁L inhibited the Nef-dependent reduction in the surface expression of HLA-A2 molecules (Fig. 1A and C). Given that HLA-A2 A₃₂₃V is a model for trafficking due to AP-1, these differences provide suggestive evidence that the effects of Nef on MHC-I HLA-A2 are not due to AP-1 alone and probably involve additional steps that are disrupted by ARF-1 Q₇₁L.

Functional ARF-6 is dispensable for Nef-dependent MHC-I downmodulation.

ARF-6 may also be important for Nef-dependent internalization of MHC-I based on evidence that ARF-6 (Q₆₇L) inhibited Nef-dependent MHC-I trafficking in HeLa cells (4, 64). To examine the relative effect of ARF-1 and ARF-6 mutants on Nef activity in our system, we generated ARF-6 constructs analogous to the ARF-1 constructs tested in Fig. 1 and similar to those previously shown to disrupt Nef activity in HeLa cells (4, 64). We found that, in contrast to what was seen for ARF-1 Q₇₁L, there was no significant effect of ARF-6 mutants on Nef activity in T cells (Fig. 2A and B). As controls, we demonstrated that the ARF-6 mutants were expressed in the cells (Fig. 2C). Additionally, we verified that the ARF-1 and ARF-6 constructs transduced the cells with similar efficiencies (see Fig. S1 in the supplemental material). For these experiments, the ARF-1 and ARF-6 constructs were made identically as bicistronic elements that coexpressed GFP from an internal ribosome entry site (see Fig. S1), allowing GFP to serve as a measure of both transduction efficiency and relative expression levels.

Fig. 2.

ARF-1 but not ARF-6 activity is required for Nef-dependent HLA-A2 and CD4 downmodulation in T cell lines. (A) Flow cytometric analysis of HLA-A2 or CD4 surface expression in CEM-SS cells transduced with control or Nef-expressing adenovirus plus the indicated ARF-6 construct. ARF-6 proteins were expressed via bicistronic murine retroviral vectors that also expressed GFP. Similar GFP⁺ populations based on fluorescence mean intensity were analyzed for HLA-A2 or CD4. Gray shaded curve, control vector; gray outlined curve, adeno-Nef; black curve, adeno-Nef plus the indicated ARF-6 protein. (B) Quantitation of panel A. The median fluorescence normalized to wild-type ARF-6 is shown ± SD (n = 3). (C) Western blot of ARF-6 and Nef expression. (D) Confocal fluorescence microscopy of CEM-SS cells transduced with the indicated ARF-6 construct and stained with antibodies against HLA-A2. ARF-6 proteins were expressed via bicistronic murine retroviral vectors that also expressed GFP. DAPI, 4′,6-diamidino-2-phenylindole.

ARF-6 Q₆₇L was active in our experimental system based on its effect on MHC-I surface expression in the absence of Nef. Microscopic analysis revealed intracellular accumulation of HLA-A2 in CEM-SS cells expressing ARF-6 Q₆₇L but not in cells expressing wild-type ARF-6 (Fig. 2D) or ARF-6 T₂₇N (data not shown). This phenotype is consistent with previous reports that ARF-6 Q₆₇L inhibits the recycling of internalized MHC-I back to the cell surface and that ARF-6 T₂₇N does not yield a strong phenotype (5, 32).

Previous studies have suggested that Nef activates ARF-6 via a phosphatidylinositol 3 kinase (PI3K)-dependent step. Thus, levels of the PI3K inhibitor, phosphatase and tensin homolog (PTEN), could influence the relative amount of ARF-6-dependent activity we observed. Indeed, it has been suggested that the lower expression level of PTEN in the CEM-SS cell line used in some of our studies may influence PI3K-induced ARF-6-dependent effects on MHC-I, and this may have limited our ability to detect this pathway (19). To address this possibility, we used Western blot analysis to examine the levels of PTEN expression in several cell lines as well as in primary CD4⁺ T lymphocytes. We confirmed that PTEN expression varied dramatically among cell lines and that CEM-SS cells did not express much, if any, PTEN (Fig. 3A). We also performed experiments in a cell line (the SupT1 lymphoblastoid line) with very high PTEN expression to determine whether PTEN levels might be affecting our ability to detect a requirement for ARF-6 activity. Similar to our observations in CEM-SS cells, we found that expression of ARF-1 Q₇₁L significantly reduced HLA-A2 downmodulation by Nef, while expression of ARF-6 Q₆₇L did not (Fig. 3B and C). As a control, we again demonstrated that the ARF-1 and ARF-6 constructs transduced the cells and expressed the bicistronic message to similar degrees (see Fig. S2 in the supplemental material).

Fig. 3.

ARF-1 but not ARF-6 activity is required for Nef-dependent HLA-A2 and CD4 downmodulation in PTEN-expressing T cell lines. (A) Western blot of PTEN or control (GAPDH) levels in the indicated cell type. (B) Flow cytometric analysis of HLA-A2 surface expression in SupT1 cells transduced with control or Nef-expressing adenovirus plus the indicated ARF-6 construct; analyzed as described for panel A. Dark gray shaded curve, control adenovirus plus ARF as indicated; light gray shaded curve, adeno-Nef plus vector; black curve, adeno-Nef plus ARF as indicated. (C) Quantitation of panel B. The median fluorescence normalized to empty vector treatment is shown ± SD (n = 3). Values obtained with wild-type ARF-1 were significantly different from those with ARF-1 Q₇₁L (P < 0.01). Values obtained with wild-type ARF-6 were not significantly different from those with ARF-6 Q₆₇L (P = 0.46).

Finally, we compared the effects of ARF-1 and ARF-6 mutants in HIV-infected primary T lymphocytes expressing endogenous HLA-A2. For these experiments, wild-type ARF or its dominant active mutant was inserted directly into the envelope open reading frame of a GFP-expressing HIV molecular clone (NL-GIarf) (Fig. 4A). Primary T cells were infected with the virus and then assayed 36 h postinfection. We reproducibly observed a 2- to 3-fold inhibition of Nef-dependent MHC-I downmodulation when the infected cells expressed the ARF-1 mutant Q₇₁L but not when the infected cells expressed the corresponding ARF-6 mutant, Q₆₇L (Fig. 4B). These data support a model in which an ARF-1-dependent pathway rather than an ARF-6-dependent pathway is needed for Nef-dependent MHC-I downmodulation in HIV-infected primary T cells.

Fig. 4.

ARF-1 but not ARF-6 activity is required for Nef-dependent endogenous HLA-A2 and CD4 downmodulation in HIV-infected primary T cells. (A) Map of HIV engineered to express ARF-1, ARF-1 Q₇₁L, ARF-6, or ARF-6 Q₆₇L and GFP plus or minus Nef. (B) Flow cytometric analysis of endogenous HLA-A2 surface expression in primary CD4⁺ T cells infected with the indicated HIV. Cells were harvested 36 h postinfection. Numbers inside each flow plot reflect the mean fluorescence intensity of MHC-I in the GFP⁺ population. Fold downmodulation of MHC class I surface expression derived from the Nef⁺ and Nef⁻ populations of each HIV construct is shown under the bottom panels. Results are representative of three independent experiments.

As a positive control for ARF-6 Q₆₇L activity, we measured the surface expression of MHC-I in the absence of Nef. For endogenous HLA-A2, Bw4, and Bw6 MHC-I allotypes, we observed 18% (Fig. 4B), 31%, and 29% (data not shown) reductions in MHC-I expression, respectively, when ARF-6 Q₆₇L was expressed. Over three independent experiments in which HLA-A2 expression was measured in primary T cells expressing ARF-6 Q₆₇L, the average reduction in surface expression was 13% (P < 0.002). These data are consistent with the previously reported inhibitory effect of ARF-6 Q₆₇L on the recycling of internalized MHC-I back to the cell surface (5, 32).

To further examine the contribution of ARF-6 to Nef-dependent MHC-I downmodulation, we used a lentiviral system expressing shRNA to silence ARF-6 expression. We found that shRNA directed against ARF-6 had no significant impact on HLA-A2 downmodulation by Nef, as compared to negative control shRNA in T cells (Fig. 5A and B). Western blot analysis confirmed the silencing of ARF-6 with these vectors (Fig. 5C).

Fig. 5.

ARF-6 knockdown does not inhibit Nef-induced MHC-I downmodulation. (A) Flow cytometric analysis of CEM-SS cells transduced with lentivirus expressing shRNA and GFP and subsequently transduced with the indicated adenoviral vector at 3 days after lentiviral transduction. Flow cytometric analysis was performed at 3 days after adenoviral transduction. (B) Quantitation of the flow data from panel A. Relative downmodulation of the indicated molecule. Errors bars represent SD (n = 6). (C) Western blot analysis confirming specific knockdown of ARF-6. (D) Flow cytometric analysis of CEM-SS cells transduced with lentivirus expressing shRNA and GFP and subsequently mock infected or infected with NL4-3 HIV at 6 days after lentiviral transduction. Flow cytometric analysis was performed at 48 h after HIV infection. (E) Quantitation of the frequency of infection in cells from panel D. (F) Summary graph of HLA-A2 downmodulation in Gag-expressing cells as compared to mock-infected cells in panel D.

ARF-6 is known to regulate cellular entry by several microorganisms, including HIV, and ARF-6 silencing has been shown to inhibit HIV infection (16). Therefore, we used inhibition of HIV infection as a positive control for ARF-6 silencing. At lower MOIs, we indeed observed that ARF-6 silencing resulted in a lower rate of infection in CEM-SS cells expressing ARF-6 shRNA-containing lentiviral vectors (Fig. 5D and E). The magnitude of this effect was similar to that previously reported to occur with ARF silencing (16). Under these conditions, ARF-6 knockdown did not inhibit HLA-A2 downmodulation in HIV-infected (Gag⁺) cells (Fig. 5F). These results are in agreement with prior reports demonstrating that ARF-6 silencing had no effect on MHC-I downmodulation in Jurkat T cells and in HeLa cells (64).

Dominant active ARF-1 increases AP-1 recruitment to HLA-A2 A₃₂₃V and the HLA-A2-Nef complex.

To determine the mechanism by which ARF-1 affects MHC-I surface expression, we used coimmunoprecipitation assays to measure AP-1 binding to cargo plus or minus the expression of ARF-1 mutants. As a positive control, we first examined AP-1 binding to a natural Yxxφ sorting signal using HLA-A2 A₃₂₃V in the absence of Nef. As seen in Fig. 6A, lanes 2 to 5, and as previously reported (61), AP-1 coprecipitated with HLA-A2 A₃₂₃V in the absence of Nef. Analogous to what is observed in Fig. 1, the expression of wild-type (lane 3) or T₃₁N (lane 5) ARF-1 did not affect AP-1 recruitment by this assay. However, ARF-1 Q₇₁L (lane 4) dramatically increased the coprecipitation of AP-1 with HLA-A2 A₃₂₃V.

Fig. 6.

Dominant active ARF-1 stabilizes AP-1 binding to the Nef–MHC-I complex. (A) Immunoprecipitation (IP) of HLA-A2 A₃₂₃V with Western blot analysis of associated AP-1. HLA-A2 A₃₂₃V was immunoprecipitated from CEM-SS cells transduced with a bicistronic murine retroviral vector expressing GFP and ARF-1 proteins as indicated. Based on GFP expression, 50 to 70% of the cells were transduced (n = 4). (B) Immunoprecipitation of HLA-A2 with Western blot analysis of associated proteins. HLA-A2 was immunoprecipitated from CEM-SS cells transduced with control or adeno-Nef and a bicistronic murine retroviral vector expressing GFP and ARF-1 proteins as indicated. Based on GFP expression, 50 to 70% of the cells were transduced (n = 4).

Next, we used the same assay system to determine whether ARF-1 activity affected the formation of the Nef–MHC-I–AP-1 complex. Similar to what was observed with HLA-A2 A₃₂₃V, we found that ARF-1 Q₇₁L dramatically increased the coprecipitation of AP-1 with HLA-A2 in Nef-expressing cells. Conversely, we found that treatment with BFA disrupted complex formation (Fig. 6B, lanes 10 and 11). The effect of BFA was specific because it could be rescued by overexpression of ARF-1 Q₇₁L (Fig. 6B, lane 12) as previously reported (65). Similar to what we observed with HLA-A2 A₃₂₃V, the expression of ARF-1 T₃₁N had little effect on AP-1 coprecipitation (Fig. 6, lane 9), because, based on our controls, ARF-1 T₃₁N was not expressed at levels high enough to overcome endogenous ARF-1 activity (data not shown).

We were also able to detect the presence of ARF-1 in the Nef–MHC-I–AP-1 complex (Fig. 6B). ARF-1 Q₇₁L coprecipitated readily, but both wild-type and ARF-1 T₃₁N could also be detected, albeit at lower efficiency. Of note, the expression of ARF-1 mutants and BFA treatment had no significant effect on Nef coprecipitation with HLA-A2, confirming previous data that Nef can bind to the cytoplasmic tail of HLA-A2 in the absence of AP-1 (60, 61).

We also performed the reverse experiment, in which we asked whether we could observe coprecipitation of AP-1, Nef, and HLA-A2 when we immunoprecipitated ARF-1. For this study, we examined complexes formed both in the absence and in the presence of Nef. Indeed, we were able to detect AP-1 coprecipitating with ARF-1 Q₇₁L (Fig. 7). In Nef-expressing cells, we were also able to detect Nef coprecipitating with ARF-1 independent of the GTP-bound state of the ARF-1 molecule (Fig. 7, lanes 4 to 6), as previously reported (38). Additionally, in Nef-expressing cells, there was an enhancement in the amount of AP-1 that coprecipitated with ARF-1 Q₇₁L (Fig. 7, lane 6). Finally, we also detected HLA-A2 coprecipitating with ARF-1 Q₇₁L in Nef-expressing cells. These data suggest that ARF-1 Q₇₁L potently stabilizes interactions among AP-1, Nef, and MHC-I HLA-A2 and that its mode of inhibition of MHC-I downmodulation is not the disruption of complex formation but rather the formation of a static complex that sequesters necessary trafficking components. Therefore, we propose a model in which ARF-1 is necessary to form the trafficking vesicles at the trans-Golgi network that contain MHC-I, Nef, and AP-1 (Fig. 8).

Fig. 7.

MHC-I, Nef, and AP-1 coprecipitate with ARF-1 Q₇₁L. (Left) Immunoprecipitation (IP) of myc-tagged ARF-1 with Western blot analysis of associated proteins. ARF-1 proteins were immunoprecipitated from cells treated as for Fig. 3B. Based on GFP expression, 50 to 70% of the cells were transduced (n = 2). (Right) Western blot analysis of input lysate.

Fig. 8.

Model of Nef-dependent CD4 and MHC-I trafficking. LY, lysosome; LE, late endosome; MVB, multivesicular body; ER, endoplasmic reticulum; PM, plasma membrane.

DISCUSSION

HIV causes a persistent infection that evades eradication by anti-HIV CTLs. Viral persistence is mediated in part by the activity of the HIV Nef protein, which disrupts antigen presentation by MHC-I to CTLs. A number of cellular factors have been implicated in this pathway, and thus it is important to clarify the relative contribution of each to help focus the development of inhibitors.

Here, we confirm a requirement for AP-1, and we provide new evidence that the small GTPase ARF-1 is also needed for Nef's effects on MHC-I trafficking. The ARF-1 mutant that resembles activated, GTP-bound ARF-1 (ARF-1 Q₇₁L) and the ARF-1 inhibitor BFA both had dramatic effects on the formation of the Nef–MHC-I–AP-1 complex and on surface MHC-I levels in Nef-expressing cells. Both ARF-1 Q₇₁L and BFA have effects on Golgi architecture that could indirectly influence whether MHC-I, Nef, and AP-1 are in the proper location to form a complex. However, the fact that ARF-1 coprecipitates with MHC-I in Nef-expressing cells and the fact that ARF-1 Q₇₁L stabilizes the AP-1–Nef–MHC-I complex support the model that ARF-1 activity is directly needed for complex formation. These results provide insights into what is required for complex formation and further elucidate how this process occurs (Fig. 8).

The effect of ARF-1 Q₇₁L on MHC-I downmodulation may be explained by other studies examining COP-I coats, which have demonstrated that when ARF-1 Q₇₁L is expressed, ARF-1–GTP cannot be hydrolyzed into ARF-1–GDP, resulting in a static coat (45, 52). This effect compromises the cycles of recruitment and dissociation from membranes that are necessary for normal activity. ARF-1 T₃₁N was not active in our system to disrupt Golgi architecture or to inhibit Nef. However, the ARF-1 inhibitor BFA was active, and its inhibitory effect on complex formation could be specifically rescued by the expression of ARF-1 Q₇₁L. ARF-1 silencing also caused a statistically significant reduction in Nef-dependent MHC-I trafficking (data not shown). The impact of ARF-1 silencing on Nef was smaller than that achieved with ARF-1 Q₇₁L and BFA due to the fact that there are multiple Golgi ARFs with overlapping activity, and thus the silencing of single Golgi ARFs does not yield a clear phenotype (58).

Prior studies have shown that AP-1 is required for Nef-dependent trafficking and can be found in complexes with MHC-I and Nef in transformed T cell lines and in HIV-infected primary T lymphocytes (43). Work with purified proteins has revealed that Nef directly contacts the MHC-I cytoplasmic tail (60) and that a Nef–MHC-I fusion protein directly interacts with the AP-1 μ subunit (34, 49). A tyrosine in the MHC-I cytoplasmic tail and the tyrosine-binding pocket of AP-1 μ1 are both required for AP-1 to coprecipitate with HLA-A2 in Nef-expressing cells (61) and with purified proteins (34, 49). These data have led to a model in which Nef disrupts MHC-I trafficking from the TGN through an interaction with AP-1 that redirects MHC-I into an endolysosomal pathway instead of the cell surface (Fig. 8). In cells that demonstrate a greater degree of Nef-dependent MHC-I internalization (e.g., HeLa), AP-1 is also required for the Nef-dependent internalization of MHC-I (23).

Nef also downmodulates the HIV receptor CD4 to prevent viral superinfection (2) and to promote viral assembly and release (25, 44). Downmodulation of CD4 and MHC-I by Nef occurs by distinct mechanisms and requires different Nef domains (31, 46). Work from a number of laboratories has supported a model in which the separate Nef-dependent trafficking pathways of MHC-I and CD4 ultimately converge into a common β-COP-dependent pathway necessary for lysosomal targeting (15, 27, 38, 46). In this model (Fig. 8), Nef promotes accelerated internalization of CD4 in an AP-2-dependent manner (7, 18, 21, 51), and internalized CD4 is targeted into acidic compartments (38) and multivesicular bodies (MVBs) (11, 46) for accelerated degradation. Previous research has shown that ARF-1 is involved in COP-I coatomer recruitment to internalized Nef-CD4 complexes and is necessary for the localization of these complexes to acidic compartments (15). However, GTP binding and hydrolysis are not needed for this process (15).

In contrast to our results with ARF-1, we did not detect a significant requirement for ARF-6 activity for Nef to reduce MHC-I surface expression in HIV-infected primary T cells, CEM-SS T cells, and SupT1 T cells. The difference between our current results and prior publications may be related to a number of technical factors. The effect of ARF-6-Q₆₇L on Nef activity was reported for HeLa cells (4, 64), which have been shown to traffic MHC-I differently than T cells (23). Thus, it is possible that cell-type-specific differences explain the discrepancy. However, others have reported that ARF-6 silencing in HeLa cells does not affect Nef-induced MHC-I downmodulation (64). The different results obtained with silencing ARF-6 in HeLa cells versus ARF-6-Q₆₇L overexpression in HeLa cells may be explained by indirect effects of ARF-6-Q₆₇L on intracellular trafficking, as has previously been proposed (26).

In sum, we have further defined the mechanism by which Nef allows HIV-1-infected cells to evade the host immune response. Implicating ARF-1 in the downmodulation of MHC-I by Nef provides further support for the role of the AP-1-dependent pathway shown in Fig. 8. These data help to elucidate the mechanism by which Nef downmodulates MHC-I and reveal further targets for pharmaceuticals that may inhibit immune evasion.