Inhibitors of protein-disulfide isomerase prevent cleavage of disulfide bonds in receptor-bound glycoprotein 120 and prevent HIV-1 entry.

We previously reported that monoclonal antibodies to protein-disulfide isomerase (PDI) and other membrane-impermeant PDI inhibitors prevented HIV-1 infection. PDI is present at the surface of HIV-1 target cells and reduces disulfide bonds in a model peptide attached to the cell membrane. Here we show that soluble PDI cleaves disulfide bonds in recombinant envelope glycoprotein gp120 and that gp120 bound to the surface receptor CD4 undergoes a disulfide reduction that is prevented by PDI inhibitors. Concentrations of inhibitors that prevent this reduction and inhibit the cleavage of surface-bound disulfide conjugate prevent infection at the level of HIV-1 entry. The entry of HIV-1 strains differing in their coreceptor specificities is similarly inhibited, and so is the reduction of gp120 bound to CD4 of coreceptor-negative cells. PDI inhibitors also prevent HIV envelope-mediated cell-cell fusion but have no effect on the entry of HIV-1 pseudo-typed with murine leukemia virus envelope. Importantly, PDI coprecipitates with both soluble and cellular CD4. We propose that a PDI.CD4 association at the cell surface enables PDI to reach CD4-bound virus and to reduce disulfide bonds present in the domain of gp120 that binds to CD4. Conformational changes resulting from the opening of gp120-disulfide loops may drive the processes of virus-cell and cell-cell fusion. The biochemical events described identify new potential targets for anti-HIV agents.

namely the binding of the HIV-1 envelope glycoprotein gp120 (Env) to the primary receptor CD4, the attachment of CD4-bound gp120 to coreceptors CCR5 or CXCR4 (1), and ultimately, the fusogenic activation of envelope glycoprotein gp41 (2)(3)(4). It is generally accepted that virus binding to CD4 causes major conformational changes in gp120 (5,6), but the molecular mechanisms triggering these changes have not been fully characterized. Immunologically detectable conformational changes in gp120 occur upon its binding to the soluble ectodomain of CD4 (sCD4). They expose new gp120 epitopes, one of which overlaps with the binding site of gp120 to coreceptor CCR5 (7,8). Whether and how such changes relate to the activation of gp41 is not known. No cell-mediated triggering factors that for instance would compare with the acid pH-induced conformational changes, which drive the entry of influenza virus, have been identified in HIV-1 entry. Here we describe how a known membrane-associated oxidoreductase, protein-disulfide isomerase (PDI, EC 5.3.4.2), may cause the major structural changes in HIV-1 envelope that lead to virus entry.
PDI is a well characterized 57-kDa oxidoreductase (9) that forms, breaks, and rearranges disulfide bonds in nascent proteins reaching the endoplasmic reticulum (10). Its redox function is based on the presence in its two active sites of a Cys-Gly-His-Cys (CXXC) motif. When the cysteines of CXXC are oxidized, PDI interacts with two cysteinyl thiol groups of a neighboring peptide to form a disulfide bond. When CXXC bears two cysteinyl thiols, it cleaves neighboring disulfide bonds. In the endoplasmic reticulum, PDI acts predominantly as an oxidase (10 -12). At the cell surface, however, PDI acts as a reductase (13)(14)(15)(16)(17)(18)(19)(20)(21), and one of its functions is to catalyze thiol-disulfide interchanges. An interchange generates two free thiols in a target protein and an oxidized CXXC motif (22). PDI is inherently a promiscuous enzyme. Its property to bind to different peptides is essential to its function both inside the cell and at the cell surface. The surface functions of PDI are many and include interacting with the receptor of diphtheria toxin to facilitate the transport of the toxic fragment into the cytoplasm (13,14), reducing a disulfide bond in the ectodomain of the thyrotropin receptor (17), binding to extracellular thrombospondin (18), and binding nitric oxide in a transnitrosation reaction required for the transport of nitric oxide into megakaryocytes and for the maturation of platelets (19). PDI may reach the surface in association with other endoplasmic reticulum proteins transported to the cell membrane through a pathway that is subject to regulation (54). PDI is shed from the cell surface (23,24) and rapidly replaced (14). Cell surface and endoplasmic PDI are biochemically and immunologically identical (23,24), and changes in the expression of intracellular PDI are reflected in the expression of surface PDI (19,20).
The surface-associated reductive activity of PDI was first demonstrated by monitoring the cleavage of a disulfide-containing poly(D-lysine) conjugate attached nonspecifically to fibroblasts (25). Disulfide reduction was inhibited by membraneimpermeant DTNB and pCMBS that block free thiols (25) and by monoclonal anti-PDI antibodies (13). We showed subsequently that monoclonal anti-PDI antibodies and two small molecular PDI inhibitors prevented HIV-1 infection and that prevention was observed only when the agents were present at the time of virus-cell interaction (26). This suggested that inhibitors interfered with virus entry. This finding led to the hypothesis that surface PDI associates with the HIV receptor CD4, and through that association, it reaches and reduces a gp120-disulfide bond situated in the region of gp120 that binds to CD4. This hypothesis implied that the cleavage of a loopforming disulfide bond might cause the conformational changes in gp120 that are required to activate the fusogenic function of gp41. The following sequence of experiments was carried out to support this hypothesis. 1) We confirmed that HIV-1 target cells are capable of reducing disulfide bonds in proteins that bind to their surface (Fig. 1A).
2) The physical association of PDI and CD4 was demonstrated by coprecipitation of soluble CD4 with affinity-labeled soluble PDI (Fig. 1C) and by coprecipitation of cellular PDI with affinity-labeled cellular CD4 (Fig. 1D). 3) We showed that soluble PDI is capable of reducing disulfide bonds in soluble gp120 and that the PDI-induced opening of a disulfide loop caused conformational changes in gp120 (Fig. 2). 4) We showed that gp120 bound to the primary HIV receptor CD4 on target cells also undergoes a disulfide reduction that is inhibited by PDI inhibitors (Fig. 3). 5) We showed that the same inhibitors at the same concentrations also prevent HIV entry into target cells (Fig. 4).

EXPERIMENTAL PROCEDURES
Reagents-PDI was isolated from calf liver (13) or purchased from Pierce. The thiol-specific biotinylating agents and biotin-specific reagents used were as follows: membrane-impermeant biotinyl 3-maleimido propionamidyl-3-6-dioxaoctane diamine (MPDOD, Molecular BioSciences, Boulder, CO); Biotin-BMCC, Ultralink Immobilized Neu-trAvidin, and HRP-NeutrAvidin (all from Pierce); phenylarsine oxide (PAO), T 3 , and DTNB; and thrombin and anti-CD4-FITC mAb Q4120 (Sigma). Para-amino-PAO (aPAO) was prepared according to a published procedure (28). To increase water solubility of T 3 , its ␣-amino group was acetylated as follows: 100 mg of T 3 and 50 mg of N-acetic acid N-hydroxysuccinimidyl ester were dissolved separately in 0.5 ml of dimethylformamide. The two solutions were mixed and allowed to react overnight at room temperature. The reacted mixture was added to 30 ml of 1 ϫ 10 Ϫ3 M HCl and kept for 30 -45 min at 4°C to precipitate N-acetylated T 3 (AT3). After centrifugation, the precipitate was washed 3 times with 20 ml of water and dried at room temperature under vacuum. The purity of AT3 determined by thin-layer chromatography was Ͼ90%, and unreacted T 3 was the only byproduct. The inhibitory potency of AT3 was routinely tested on the reductive activity of the cell surface. AMD3100 was a gift from AnorMED (Langley, British Columbia, Canada). Several hybridomas producing anti-PDI antibodies, in particular mAb RL77 used here for immunodetection, were provided by C. S. Kaetzel (29). RL77 and other anti-PDI mAbs that had been used to neutralize PDI activity (13,26) lost that inhibitory function when obtained at a later date from the same source (29). C. C. Broder provided gp140 polyclonal anti-gp140 Ab R2143 and mouse anti-gp120 mAb D19. Anti-CCR5-FITC mAb 2D7 was from BD Biosciences, and anti-mouse Ig(HϩL)-FITC was from Southern Biotechnology (Birmingham, AL). Anti-CD4 mAb Leu3a, anti-HLA-DR-FITC mAb 243 as well as FACScan and CellQuest software were from BD Biosciences. The following reagents were obtained through the National Institutes of Health (AIDS Research and Reference Reagent Program, Division of AIDS): antibodies, anti-CXCR4 mAb 12G5 (from J. Hoxie), CD4 antiserum T4-4 (from R. Sweet), CD4 hybridoma SIM-2 (from J. Hildreth), and 1G5 (from E. Aguilar-Cordova and J. Belmont); cell lines, U87.CD4 and U87.CD4.CCR5 (from H. K. Deng and D. Littman), J.1.1 (from T. Folks and S. Butera), HL2/3 (from B. Felber and G. Pavlakis), and PM1 (from P. Lusso and M. Reitz); viruses and virus-containing plasmids, HIV-1 JR-FL (from I. Chen), pSV-aMLV-Env and pHIVgpt (from N. Landau and D. Littman), and pNL4-3 (from A. Adachi, H. E. Gendelman, S. Koenig, T. Folks, R. Willey, A. Rabson, and M. A. Martin).
Reductase Activity of Soluble PDI and of the Surface of Target Cells-Purified PDI was assayed by measuring its ability to cleave the disulfide of [ 125 I]tyramine-SS-poly(D-lysine) in solution (13,25). Glutathione (GSH) (0.1 mM) was added to ensure that the CXXC motif of PDI was in a reduced state. No such addition was required to measure the activity of cell surface PDI, indicating that PDI is present on the cell surface in a reduced redox state. To assay the reductive activity of the cell surface, labeled [ 125 I]tyramine-SS-poly(D-lysine) was bound to cells at 0°C for 1 h. The cells were then washed and incubated in PBS (30 min at 37°C). After PDI had been inactivated by 80 mM N-ethylmaleimide, cells were lysed with 0.5% Triton X-100, and the 25% trichloroacetic acid-soluble radioactivity released during incubation was determined.
Coprecipitation of Soluble CD4 by Biotinylated Soluble PDI-Soluble PDI biotinylated on its thiols with membrane-impermeant MPDOD was purified by ultrafiltration and mixed with an equimolar amount of non-biotinylated soluble CD4 (1 h at 25°C). After precipitation with immobilized avidin, separation by SDS-PAGE, and transfer to nitrocellulose membrane, PDI and CD4 were immunodetected on the same blot with mAb RL77 and CD4 antiserum. In control experiments, CD4 was immunoprecipitated with anti-CD4 mAb Leu3a and, upon probing with HRP-NeutrAvidin, was not biotinylated.
Coprecipitation of Cellular CD4 by Surface Proteins Biotinylated on their Thiols-The surface of 5 ϫ 10 7 U937 cells was labeled with 1 ml of 0.5 mM membrane-impermeant thiol-specific MPDOD (0°C for 30 min) to tag the thiols of PDI and other surface proteins. Cells were washed with cold Hanks' buffered saline solution and lysed in 200 mM Tris, pH 8.3, containing 0.6% CHAPS, 5 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, and 20 g/ml aprotinin. Biotinylated proteins were isolated with immobilized avidin and resolved by SDS-PAGE and transferred. The membrane was probed for PDI with mAb R77, stripped, and reprobed for CD4 with CD4 anti-serum. Cellular CD4 and PDI have a similar electrophoretic mobility. This probing identified both proteins.
Cleavage of gp120-Disulfide Bonds by Soluble PDI-Recombinant gp120 IIIB (100 g/ml) was treated with thrombin (200 g/ml, 120 min at 37°C) to make a proteolytic cut near the tip of the V3 loop. Subsequent reduction of the V3-disulfide bond is known to cleave gp120 into a 70-kDa N-terminal and a 50-kDa C-terminal fragment (33). Reduction was carried out with 5% 2-mercaptoethanol (control) or with PDI (50 g/ml in 1 mM GSH). Following separation by SDS-PAGE under nonreducing conditions, gp120 and its fragments were detected with polyclonal anti-gp140 antibody R2143.
Detection of Generated Thiols in gp120/gp140 Bound to Target Cells-Recombinant gp120 IIIB , gp120 JR-FL , or oligomeric gp140 IIIB(BH8) (10 g/ml) (34) were allowed to bind to 5 ϫ 10 7 matched target cells (1 ml, 90 min at 14°C) in Hanks' buffered saline with or without PDI inhibitors. After washing, cells were labeled (40 min, at 0°C) with Biotin-BMCC (0.5 mM) to biotinylate newly generated thiols, washed, and lysed (20 min at 4°C) in PBS containing 0.25% Nonidet P-40 (or 0.6% CHAPS), 20 g/ml leupeptin, 20 g/ml aprotinin, and 20 M phenylmethylsulfonyl fluoride. When using gp140, the envelope glycoprotein was immunoprecipitated with rabbit polyclonal anti-gp140 antibody R2143 and immobilized protein G, resolved by SDS-PAGE, transferred to nitrocellulose membranes, and probed for biotin using HRP-NeutrAvidin. Western blots were stripped and reprobed with mouse anti-gp120 mAb D19 to assess Env binding to CD4. When using gp120 IIIB , thiol-biotinylated surface proteins were isolated with immobilized avidin. Western blots were then probed for Env with Ab R2143 and probed for PDI with mAb RL77. They were stripped and reprobed for CD4 with CD4 antiserum. In other experiments, the CCR5-tropic gp120 JR-FL was incubated with U87.CD4 cells that did or did not express the CCR5 coreceptor. The newly generated protein thiols were biotinylated, isolated, and probed for envelope as in experiments using gp120 IIIB .
Virus Production-HIV-1 NL4-3 and HIV-1 JR-FL were generated by infection of SupT1 cells and PM1 cells, respectively, with cell-free viruses. Alternatively, SupT1 cells were transfected with pNL4-3. The pseudotyped HIV-1 was generated by transfection of COS-7 with pSV-A-MLV-Env and with pHIV-gpt (15 g each) according to Page et al. (35). HIV-1 titers were measured using an HIV-1 p24 antigen capture enzyme-linked immunosorbent assay (Coulter Immunotech, Hialeah, FL). Harvested viruses were filtered and stored at Ϫ80°C. The virus stocks were routinely tested for contamination with proviral or plasmid DNA by PCR and, if necessary, were treated with RNase-free DNase. Ecotropic MLV pseudotyped murine stem cell virus expressing enhanced green fluorescent protein was generated according to Hawley et al. (36).
Envelope-mediated Cell-Cell Fusion Assay-1G5 cells that contain the integrated LTR-luciferase reporter gene (38) were incubated for 30 min in the presence or absence of PDI inhibitors (1 ϫ 10 6 cells/well in 24-well plates). An equal number of J1.1 cells expressing Tat and Env were then added, and the concentration of inhibitors was adjusted. The mixed cells were allowed to fuse (6 h at 37°C). The cells were lysed, and luciferase activity was measured using a commercial kit (Promega, Madison, WI). Alternatively, the fusion of P4 cells with HL2/3 cells expressing Env and Tat (39) was assayed for activity of transactivated ␤-galactosidase as described above.
HIV-1 Binding Assay-HIV-1 NL4Ϫ3 containing HLA-DR in its envelope (harvested from infected HLA-DR ϩ cells) was allowed to interact with HLA Ϫ SupT1 target cells and was labeled with FITC-anti-HLA-DR mAb as described by Ugolini et al. (40). The cells were treated with l00 M AT3 or Me 2 SO (solvent control) or with anti-CD4 mAb Leu3A. The binding of the labeled virus to the cells was measured by flow cytometry using CellQuest software.

The Surface of HIV-1 Target Cells Has a Reductive Function
That Is Suppressed by PDI Inhibitors-To demonstrate the reductive function of the cell surface, in particular its ability to cleave disulfide bonds in membrane-bound peptides, we measured the release of acid-soluble [ 125 I]tyramine-SH from surface-bound [ 125 I]tyramine-SS-poly(D-lysine) (25). The cationic conjugate was bound to the anionic surface of U937 cells (60 min at 0°C), and the cells were then incubated for 30 min at 30°C. The release of radioactivity during incubation was inhibited in dose-dependent fashion by three PDI inhibitors that act by different mechanisms. The membrane-impermeant reagent DTNB forms mixed disulfides with thiol groups, PAO and its derivative aPAO form coordination bonds through their As ϩ3 with the vicinal thiols of the CXXC motif of proteins such as PDI (41), and T 3 or AT3 inhibit PDI by binding to sites other than CXXC (42). These agents inhibited the cleavage of both cell-bound conjugate (Fig. 1A) and soluble conjugate exposed to purified PDI (Fig. 1B). In Fig. 1A, the IC 50 (in M) for DTNB, aPAO, and AT3 are 4.9, 5.8, and 70, respectively. In Fig. 1B, the IC 50 for DTNB, PAO, and AT3 are 6.8, 12, and 86, respectively. The closeness of these two sets of values measured in two different systems is consistent with the view that in Fig. 1A the conjugate was reduced by surface-bound PDI.
Soluble CD4 Is Coprecipitated by Affinity-labeled Soluble PDI-To determine whether PDI binds to CD4, purified soluble PDI was biotinylated on its thiols with MPDOD. Equimolar amounts of non-biotinylated soluble CD4 were added and mixed, and biotinylated PDI was isolated with immobilized avidin. The two proteins were separated by SDS-PAGE and probed for CD4 with CD4 anti-serum and for PDI with mAb RL77. Both proteins were identified on the same blot (Fig. 1C). Control experiments in which sCD4 was isolated with mAb Leu3a and probed with HRP-NeutrAvidin revealed no biotinylation, confirming that sCD4 does not contain unpaired cysteines (43) and indicating that disulfide bonds of sCD4 had not been reduced by PDI.
Cellular PDI Is Coprecipitated by Affinity-labeled Cellular CD4 -To test whether a PDI⅐CD4 complex could be isolated from cell lysate, human 293 cells were engineered to express CD4, CD4-Spep, or Spep coreceptors. CD4-Spep was affinitypurified from the lysate with S-protein-agarose and was found to copurify a significant amount of cellular PDI (Fig. 1D, lane  1). When using cells expressing untagged CD4 but Spep-tagged coreceptors, the PDI⅐CD4 association was maintained and demonstrated through the binding of CD4 to coreceptor Spep. After affinity purification of the coreceptors, the eluted proteins were resolved by SDS-PAGE and the Western blots were probed for PDI, stripped, and reprobed for CD4. Again, both proteins were identified (Fig. 1D, lanes 2 and 3). When this experiment was repeated using CD4-negative but CCR5-Spep-expressing cells, PDI was not copurified by the tagged coreceptor (Fig. 1D, lane  4), indicating that PDI does not bind directly to coreceptors.
Surface Proteins Biotinylated on Their Thiols Coprecipitate Small Amounts of Cellular CD4 -PDI is a prominent cell surface protein and is the only vicinal thiol-containing surface enzyme to have been positively identified (44). Knowing from Fig. 1D that a PDI⅐CD4 complex can be isolated from target cell lysate, we tested whether thiol-biotinylated surface proteins could isolate CD4. U937 cells were treated with the membraneimpermeant and thiol-specific biotinylating agent MPDOD. Tagged proteins were isolated from cell lysate with immobilized avidin and resolved by SDS-PAGE. The Western blots were probed for PDI, stripped, and reprobed for CD4. Both proteins were identified, although only a small amount of cellular CD4 was coprecipitated by that procedure (Fig. 1E), suggesting a sparse association of CD4 with surface thiol proteins (see "Discussion").
Soluble PDI Reduces Disulfide Bonds in Soluble gp120 -To test whether disulfide bonds in recombinant gp120 were accessible to reduction by soluble PDI, we focused on the disulfide bond forming the V3 loop. It is known that thrombin causes a single proteolytic cleavage near the tip of the V3 loop and that this cleavage if followed by a disulfide reduction at the base of the loop splits gp120 into a 70-and 50-kDa fragment (33). We confirmed this fragmentation using reduction with 2-mercaptoethanol (2ME) (Fig. 2, lane 3). When 2ME was replaced by PDI (in 1.0 mM GSH), a similar albeit partial fragmentation was obtained (lane 5) showing that PDI cleaves the V3-disulfide bond. GSH alone used as control had no effect. When gp120, treated as in lane 5, was biotinylated to label newly generated thiols, probing with HRP-NeutrAvidin revealed more than two bands, indicating that disulfide bonds other than V3 were also cleaved (data not shown). Lanes 1, 2, and 4 establish that proteolytic cleavage alone (lane 1), or reduction alone with 2-ME (lane 2) or PDI (lane 4) fail to fragment gp120. Interestingly, the reduction of gp120 with PDI prior to thrombin treatment generated additional proteolytic fragments (lane 6), suggesting that PDI caused conformational changes in gp120 that exposed additional thrombin-sensitive sites. These experiments were repeated in the presence of excess sCD4 to test whether conformational changes induced by sCD4 would influence the reduction of the V3-disulfide bond. The reduction was neither enhanced nor prevented. Thus, Fig. 2 demonstrates the reduction of loop-forming disulfide bonds in gp120 by soluble PDI and some ensuing changes in the gp120 conformation. There is no evidence that fragmentation through the V3 loop occurs in cell-bound gp120.
Purified PDI and soluble gp120 do not form a detectable complex. Nevertheless, PDI is capable of reducing disulfide bonds in soluble (Fig. 2).
Disulfide Bonds in gp120 and gp140 Are Reduced upon Binding to the CD4 Receptor of Target Cells-To test whether disulfide bonds of gp120 are reduced upon binding to cell surface CD4, we used the fact that fresh preparations of gp120 do not have unpaired cysteines (27) and hence no thiols susceptible to biotinylation as verified in control experiments. Therefore, the generation of gp120 thiols is a reliable indicator of disulfide reduction. Target cells were incubated (90 min at 14°C) with recombinant Env and biotinylated with thiol-specific Biotin-BMCC. Env proteins were isolated and probed. U937 cells were incubated with recombinant oligomeric gp140 IIIB (34). After biotinylation, the cells were lysed and gp140 was isolated with rabbit anti-gp140 Ab. Proteins were resolved by SDS-PAGE, transferred, and probed for biotinylation with HRP-NeutrAvidin (Fig. 3A, top), stripped, and reprobed with mouse anti-gp120 mAb (Fig. 3A, bottom). In the absence of PDI inhibitors, gp140 was distinctly biotinylated (Fig. 3A, top, lane 1). In the The Spep proteins were purified from lysates of 293 cells with S-protein-agarose. PDI was probed with mAb RL77, and the stripped blots were reprobed for CD4 with mAb Sim-2, because the cellular form of CD4 has an electrophoretic mobility similar to that of PDI. E, coprecipitation of a small amount of cellular CD4 with biotinylated surface proteins. Thiols at the surface of U937 cells were biotinylated with membrane-impermeant MPDOD, and proteins were isolated from the cell lysates with immobilized avidin and resolved by SDS-PAGE. Western blots were probed for PDI, stripped, and reprobed for CD4.
presence of inhibitors (50 M PAO or 100 M AT3), biotinylation was markedly reduced (top, lanes 2 and 3). The inhibition of biotinylation increased with the duration of exposure to inhibitors and inhibitor concentration (data not shown). Inhibitors did not affect binding of gp140 to surface CD4 as shown by the identical intensities of gp140 bands in all lanes (Fig. 3A, bottom). Alternatively, U937 cells were incubated with gp120 IIIB in the presence or absence of aPAO. Protein thiols generated in gp120 (or present on other surface proteins) were biotinylated for isolation of biotinylated proteins with immobilized avidin. The proteins were resolved by SDS-PAGE and probed on Western blots for gp120 (Fig. 3B, top) and PDI (bottom), stripped, and reprobed for CD4 (middle). The marked gp120 biotinylation seen in lane 1, top, was significantly reduced by aPAO (lane 2). Probing for CD4 (middle, lanes 1-3) showed coprecipitation of non-biotinylated CD4 by a biotinylated surface protein, most probably PDI (as shown also in Fig. 1E). The slightly increased isolation of CD4 in the absence of aPAO (middle, lane 1 versus lane 2) was confirmed in several experiments and suggests in lane 1 that CD4 was coprecipitated by both biotinylated PDI and by biotinylated gp120, possibly as a gp120⅐CD4⅐PDI complex. The faint band of gp120 seen in lane 2 (top) is also consistent with the formation of such a complex, because it may represent non-biotinylated gp120 coprecipitated by the CD4⅐PDI complex. Probing for PDI using mAb R77 (bottom) showed comparable biotinylation in all lanes. The biotinylation of PDI in lane 2 indicates that covalent bonding through the maleimido group of Biotin-BMCC displaced the coordination bonding of aPAO to PDI thiols. Note that the aPAO and AT3 concentrations that inhibit cleavage of disulfides in gp120 (Fig. 3, A and B) are within the range of concentrations that inhibit the cleavage of [ 125 I]tyramine-SS-conjugate by soluble PDI (Fig. 1B) or by the cell surface (Fig. 1A).
To test whether gp120 from an HIV-1 strain with a different cytotropism is similarly reduced and whether coreceptor expression is required for Env reduction, CCR5-tropic gp120 JR-FL was incubated with either U87.CD4 cells (Fig. 3C, lanes 1 and  2) or U87.CD4.CCR5 cells (lanes 3 and 4). In the absence of inhibitors, thiol biotinylation was comparable in CCR5-tropic and CXCR4-tropic gp120 (Fig. 3C, compare lane 3 with B, lane  1, top). The generation of thiols was also comparable when gp120 JR-FL was bound to cells that did not or did express CCR5 (Fig. 3C, lane 1 versus lane 3). Inhibitors reduced biotinylation in both cell lines (lanes 2 and 4). Thiols were likewise generated when CXCR4-tropic gp120 IIIB was bound to cells whose CXCR4 coreceptor had been blocked by its natural ligand SDF-1 or by bicyclam (data not shown) (45). Therefore, reduction does not require the binding of gp120 to a functional coreceptor. In summary, recombinant gp120s of different coreceptor specificities as well as oligomeric gp140 undergo distinct disulfide reduction upon binding to CD4 ϩ cells whether or not the cells express coreceptors. In all cases, reduction is distinctly decreased by PDI inhibitors.
PDI Inhibitors Prevent HIV-1 Entry-To determine whether cell surface reductase activity is critical for virus entry, we measured the effect of PDI inhibitors on the accumulation of The thiols generated in gp120 IIIB during incubation were biotinylated, and biotinylated gp120 was isolated with Ultralink avidin. Western blots were probed for gp120 (top), CD4 (middle), and PDI (bottom). Top panel, lane 1, biotinylated gp120 from cells incubated in the absence of inhibitors; lane 2, decreased biotinylation in gp120 isolated from cells exposed to aPAO; lane 3, control cells incubated in the absence of gp120. Middle panel, lane 1, CD4 coprecipitated by both biotinylated PDI and biotinylated gp120; lanes 2 and 3, CD4 coprecipitated by biotinylated PDI only (as in Fig. 1E). Bottom panel, PDI bands have comparable intensities in all lanes. C, cleavage of disulfide bonds in gp120 JR-FL bound to U87.CD4 cells (lanes 1 and 2) or to U87.CD4.CCR5 cells (lanes  3 and 4). Gp120 JR-FL thiols were biotinylated, isolated, and probed for gp120 as in B. Cells that do or do not express CCR5 show comparable gp120 biotinylation (lanes 1 and 3) and comparable inhibition of biotinylation in the presence of aPAO (lanes 2 and 4).  3 and 4) in the absence of AT3 (lanes 1 and 3) or the presence of AT3 (lanes 2 and 4). Strongstop DNA was assayed as in Fig. 3A. Entry of aMLV-pseudo-typed HIV-1 is not inhibited by AT3 (lane 4). minus-strand strong-stop DNA, which is the first product of reverse transcription synthesized after virus entry. Target cells were treated with increasing concentrations of inhibitors for 30 min, and HIV-1 NL4Ϫ3 was added for 1 h. The cells were washed and incubated in complete medium for 5 h, and their total DNA was extracted and amplified by PCR. PDI inhibitors aPAO and AT3 caused a dose-dependent inhibition of strong-stop DNA accumulation in SupT1 cells (Fig. 4). Parallel amplification of a cellular ␣-tubulin gene was performed to ensure that equal amounts of DNA were in each sample. The autoradiograms of Fig. 4A were quantified, and the decreases in accumulation were plotted as percent inhibition (Fig. 4B). The dose-inhibition relationships are identical for aPAO and AT3. The inhibitors had no effect on the accumulation of DNA when added after infection but were inhibitory when added shortly prior to and/or during infection (Fig. 4C). This indicates that infection requires that the cell surface reductase be active at the time of virus-cell interaction. AT3, aPAO, and DTNB caused dose-dependent inhibitions of HIV-1 entry in a broad range of cells including P4, PM1, H9, 1G5, and macrophage-depleted peripheral blood monocytic cells (Table I) (data not shown). They also inhibited infection by the CCR5-tropic MSLV HIV-1 JR-FL in both PM1 and primary monocyte-derived macrophages (Table  I). These data complement those of Fig. 3C by showing that the effects of inhibitors on both virus entry and gp120 reduction are independent of gp120 cytotropism.
PDI Inhibitors Do Not Impair Entry of an Amphotropic MLV Envelope-pseudo-typed HIV-1-Pseudo-typed virions containing the envelope glycoproteins of amphotropic MLV (aMLV) and the HIV-1 core (35) were used to examine whether or not the inhibitors were specific for HIV-1. SupT1 cells were infected with aMLV envelope-pseudo-typed HIV-1 in the presence of AT3 or aPAO (both 100 M). Neither inhibitor prevented the infection (Fig. 4D, lane 4, shown for AT3), whereas the entry of native HIV-1 was inhibited (lane 2). Similarly, PDI inhibitors did not prevent infection of 3T3 fibroblasts by ecotropic murine leukemia virus-pseudotyped MSLV expressing green fluorescent protein (data not shown) (36). These results indicate that the reductive activity of the cell surface is not required for entry by two murine leukemia virus envelope proteins but is essential for HIV-1 entry. The normal infection by aMLV-pseudo-typed HIV-1 in Fig. 4D, lane 4, also indicates that PDI inhibitors do not interfere with the reverse transcription of HIV-1.
The following control experiments were performed to confirm the finding that PDI inhibitors prevented HIV-1 entry. 1) Uninfected cells were included as negative controls, and DNA from 8E5/LAV cells (containing a single integrated copy of HIV-1 LAV ) were used as positive controls in the PCR amplification experiments.
2) The accumulation of amplified reverse transcript was tested in standard 8E5/LAV lysates and was within the linear range of the PCR assay. 3) Inhibition of HIV-1 entry was not caused by virion inactivation, as inhibitor-pretreated virions retained Ͼ90% infectivity after inhibitor removal by ultracentrifugation. 4) Inhibitors did not affect the activity of HIV-1 reverse transcriptase in an in vitro assay. 5) Under the conditions used in the experiments of Fig. 4, DTNB, aPAO, and AT3 were not cytotoxic as measured by trypan blue exclusion and MTT cleavage assays. 6) Neither AT3 nor aPAO inhibited the expression of CD4, CXCR4, or CCR5 on target cells (Fig. 5A). 7) Inhibitors did not decrease HIV-1 binding to target cell CD4 in a whole virion binding assay (Fig. 5B), consistent with the finding that they do not impair binding of gp120 to target cells (Fig. 3A, bottom).
PDI Inhibitors Prevent Infection of P4 Cells by HIV-1 NL4-3 -To confirm that our data on the inhibition of HIV-1 entry correlate with inhibition of infection, we examined the effect of inhibitors on the ability of HIV-1 NL4-3 to transactivate the LTR-LacZ reporter gene present in P4 cells. Both AT3 and aPAO inhibited virus-induced ␤-galactosidase activity in a dosedependent fashion (Fig. 6A). This assay requires a 23-h incubation of P4 cells at 37°C after exposure to inhibitors for 30 min prior to and 60 min during infection. As aPAO is not entirely membrane-impermeant, its effects on intracellular PDI (or other proteins containing vicinal thiols) may have caused some additional inhibition. This view is supported by results obtained with GSAO, a membrane-impermeant derivative of aPAO (45) that yielded a dose-inhibition relationship with negligible inhibition at the lowest doses (Fig. 6A, dotted  curve). Other controls showed that at the concentrations used, neither inhibitor affected the activity of HIV-1 Tat as Tatmediated viral transcription was normal in P4 cells pretreated with inhibitors and scrape-loaded with Tat (data not shown).
PDI Inhibitors Prevent Envelope-mediated Cell-Cell Fusion-The effects of PDI inhibitors were also detected at the level of Env-mediated cell-cell fusion in a system that measured transactivation of a LTR-luciferase reporter gene in 1G5 cells upon their fusion with latently infected J1.1 cells expressing HIV-1 Tat (38). The exposure of the fusion partners to inhibitors for 6.5 and 6 h, respectively, caused a significant inhibition of cell-cell fusion (Fig. 6B). This was confirmed in an additional assay measuring the transactivation of the LTR-LacZ reporter gene in P4 cells upon their fusion with HL2/3 cells that express Tat (data not shown) (39). These results are in keeping with the recent report by Fenouillet et al. (46) that the PDI inhibitors used previously to inhibit HIV-1 infection (26) prevent syncytia formation among HIV-1-infected cells. Finally, it is worth noting that the dose-related inhibitions of viral entry, of infection, and of cell-cell fusion (Figs. 4B and 6,  A and B) occur at concentrations (50 -100 M) that inhibit disulfide reduction in gp120 bound to CD4 (Fig. 3, A-C) and that inhibit the reductive activities of the cell surface (Fig. 1A) and of soluble PDI (Fig. 1B).

strains into coreceptor-matched target cells is inhibited in dose-dependent fashion by PDI inhibitors
Entry and its inhibition were measured as in Fig. 4, A and B. Each IC 50 is derived from 3 to 6 dose-inhibition curves. ND, not determined. a The CXCR4 and CCR5-tropic strains were HIV-1 NL4 -3 and HIV-1 JR-FL (1 g of p24/10 6 cells), respectively. b Higher concentration of freshly dissolved DTNB required to reach effects comparable with those of aPAO and AT3 may reflect the lower specificity of DTNB and/or the instability of mixed disulfides between DTNB and PDI thiols.

DISCUSSION
The major finding that PDI inhibitors prevent HIV-1 entry strengthens the hypothesis that surface PDI plays a role in HIV-1 infection. The well established reductive function of surface PDI (13)(14)(15)(16)(17)(18)(19)(20)(21) together with the putative presence of disulfide bonding close to the domain of gp120-CD4 interaction (26) suggested that PDI might cleave disulfide bonds in a gp120 molecule bound to the cell surface. The data of Fig. 3, A-C, show that gp120 indeed undergoes disulfide reduction upon binding to its surface receptor. This reaction is prevented by concentrations of PDI inhibitors that also prevent HIV-1 entry, implying that disulfide reduction in gp120 is required for entry. Although the participation of an unidentified surface reductase has not been formally ruled out, all available evidence points to PDI as the enzyme responsible for reducing gp120.
The selectivity of aPAO and AT3 for PDI is demonstrated by the fact that T 3 (the precursor of AT3) was used to isolate PDI from the surface of mammalian cells (47), whereas GSAO (the membrane-impermeant derivative of aPAO) was used to identify surface PDI in mammalian cells lysates (44). DTNB, a nonspecific thiol blocker that has an inhibitory effect comparable with that of aPAO (Fig. 1, A and B) (Table I), has an inhibitory effect similar to that of anti-PDI antibodies in systems measuring HIV-infection (26) or measuring the fusion of HIV-1-infected cells (46). It indicates that strict specificity of binding to PDI is not required for an inhibitor like DTNB to block the specific oxidoreductive function of PDI. Definitive evidence that PDI is involved in HIV-1 entry will be sought by using engineered target cells that express inactive surface PDI or that overexpress and/or underexpress active PDI. It is known from other studies that changes in expression of intracellular PDI are paralleled at the cell surface (19,20).
An additional finding is critical to help understand how surface PDI may reach gp120-disulfide bonds, namely how PDI binds to the HIV receptor CD4 (Fig. 1, C and D). The interaction does not prevent the binding of CD4 to gp120 at the cell surface, suggesting that the ectodomain of CD4 has separate binding sites for the two proteins. As gp120 binds to the outermost CD4 domain (D1), it appears probable that PDI binds to the innermost domains (D3 or D4). The proximity of these binding sites enables CD4 to bring PDI to gp120. Because the peptide-binding domains of PDI do not overlap with its active site (42), the latter remains free to interact with gp120. Func- FIG. 5. PDI inhibitors have no effect on CD4 and coreceptor expression or on virus binding to CD4. A, lack of effect of AT3 on the surface expression of CD4, CXCR4, and CCR5. FACS analysis of PM1 cells labeled with FITC-anti-CD4, FITC-anti-CCR5 mAb, or FITClabeled secondary Ab-tagging anti-CXCR4 mAb. The expression of the three proteins is identical in the presence (dashed line) or absence (solid line) of inhibitor. B, lack of effect of AT3 on the binding of HIV-1 to SupT1 cells. HLA-DR Ϫ SupT1 cells were incubated in the presence or absence of AT3, exposed to HLA-DR ϩ HIV-1 NL4Ϫ3 , and tagged with fluorescein-labeled anti-HLA-DR mAb. The fluorescence intensity (FL1-H) is measured by FACS and expresses virion binding. Cells not exposed to virus or exposed to virus but treated with mAb Leu3A to block virus binding show comparably low fluorescence intensity (controls, two peaks on left). Cells exposed to virus in the presence or absence of AT3 show comparably high fluorescence (test, two curves on right). The four curves from left to right represent as follows: solid black, no virus; dashed gray, virus added to Leu3a-treated cells; dark gray, virus added to untreated cells; light gray, virus added to cells treated with AT3. tional evidence for a direct enzymatic interaction of PDI and gp120 is provided by the reduction of cell-bound gp120 and its inhibition by PDI inhibitors (Fig. 3, A-C). PDI catalyzes thioldisulfide interchanges both inside the cell and at the cell surface. In the complex environment of the endoplasmic reticulum, the oxidoreductive function of PDI is assisted by chaperones. When PDI acts on a disulfide-containing conjugate in solution or at the cell surface (Fig. 1, A and B) or when purified PDI acts on soluble gp120 (Fig. 2), no chaperones are involved. By contrast, when reducing receptor-bound gp120, surface PDI does require the assistance of an auxiliary protein, namely CD4, but not to enhance its catalytic activity but to reach its substrate (Fig. 3). In this case, CD4 can be viewed as fulfilling a chaperone function. The association of CD4 and PDI at the cell surface is established by the work of Fenouillet et al. (46). They show that both proteins are immunodetected in the same surface area of target cells and provide evidence for their immunocolocalization. The sparse but definite colocalization they describe (46) is consistent with the finding of Fig. 1E and Fig. 3B, lane 3, which suggests that only a small fraction of surface CD4 is coprecipitated by surface PDI. The low incidence of PDI⅐CD4 association at the cell surface may be a limiting factor of infection.
As to the accessibility of gp120-disulfide bonds to PDI-mediated reduction, it is demonstrated by the ability of soluble PDI to cleave disulfide bonds in recombinant gp120 (Fig. 2). We had previously suggested that the disulfide bonds forming the V3, V4, and V4/C4 loops were situated in the area of gp120-CD4 interaction (26). The presence of disulfide bonds in this area has been confirmed by structural data published by Kwong et al. (48). These data identify 26 gp120 amino acid residues that make contact with CD4 (see Fig. 2D in Ref. 48). We noticed that 22 of them are situated in the proximity of six gp120-disulfide bonds (Table II), three of which are shown in the gp120 surface representation of the CD4 binding site of gp120 (Fig. 7). The spatial relationship of these three disulfide bonds to gp120-CD4 contact points helps understand how a CD4-bound PDI molecule may reach a gp120-disulfide and may account for the enzymatic gp120 reduction demonstrated by Fig. 3, A-C. It is not yet known which one of these bonds undergoes the initial thiol-disulfide interchange, a reaction that generates two thiols (22) on an open gp120 loop. Here the unusual concentration of disulfide bonds in the domains represented by Table II and Fig.  7 assumes special importance, because these bonds may become substrates for secondary non-enzymatic exchanges propagated by the two thiols initially generated by PDI. Conceivably, disulfide bonds in closely associated proteins may also participate in these secondary exchanges. Such crowding of FIG. 7. Presence of disulfide bonds in the face of gp120 that interacts with CD4. The representation of the gp120 surface that makes contact with CD4 was generated from data published by Kwong et al. (45). The paired cysteines forming the V1/V2 loop (cysteines 126 -196), V1/V2 stem (cysteines 119 -205), and the V4 loop (cysteines 385-418) are shown in yellow. Amino acid residues of gp120 that make contact with CD4 are shown in red (several are identified). The representation was generated by GRASP based on Protein Data Bank code 1GC1.
FIG. 8. Working model of HIV-1 entry based on PDI-induced conformational changes in gp120 and gp41. A, state prior to infection. Top, HIV-1 with envelope glycoproteins gp120 and gp41; bottom, cell surface PDI is associated with the primary receptor and is close to the coreceptor. B, binding of gp120 to CD4. The initial CD4-induced conformational changes in gp120 expose the gp120 coreceptor binding sites (red line). Disulfide bonds in gp120 are still intact (S-S). C, PDI-induced reduction of gp120-disulfide bonds (HSϩSH) causes major conformational changes in gp120 that result in the transition of gp41 to its fusion active state. The conformational changes may also contribute to the binding of gp120 to coreceptors that help anchor gp120 to the target cell during the process of fusion. D, initiation of fusion by fusionactive gp41 and shedding of gp120 (arrow). E, completed fusion and translocation of HIV core. disulfide bonds in a gp120 domain that was shown to play a central role in Env-cell interactions (8,48) therefore has the potential of being an epicenter of conformational changes. The crystallographic data identifying that specific domain were obtained with a gp120 core that had interacted with a fragment of sCD4 (48). Therefore, it is intriguing to speculate that the relatively stable initial conformational changes induced by CD4 in that domain may expose disulfide bonds to reduction by CD4-associated PDI and usher a second set of more profound and dynamic PDI-induced structural rearrangements. The cleavage of structure-stabilizing disulfide bonds in proteins is well accepted as a cause of conformational changes. Evidence that such changes occur in gp120 is provided by the observation that PDI-induced reduction of gp120-disulfide bonds exposes new thrombin-sensitive proteolytic sites (Fig. 2,  lane 6). The functional significance of disulfide bonds in native gp120 is documented by mutational studies in which the replacement of V3 and V4 cysteine residues rendered the site of gp160 that connects gp120 to gp41 inaccessible to proteolytic cleavage and abolished virion infectivity (49). We propose that post-binding opening of disulfide loops at the time of infection destabilizes the native Env and activates subsequent Env functions that drive envelope-mediated fusion and HIV-1 entry. These two processes start with the interaction of CD4 with HIV-Env that in the case of cell-cell fusion is expressed at the surface of one fusion partner. Both require prior cleavage of gp120-disulfide bonds made possible by the association of PDI with CD4. Our proposal is consistent with the data of Fenouillet et al. (46) who show that PDI inhibitors prevent the fusion of HIV-1-infected cells and the formation of syncytia, a hallmark of HIV-1 infection. Their data are supported and extended by our use of two new inhibitors with recognized affinity for PDI and by our testing of Env-mediated fusion in two different assays. Importantly, our data show that PDI inhibitors do not affect the entry of HIV-1 when the virus is pseudotyped with an aMLV envelope, indicating that an interaction of PDI with the viral envelope is not necessary for the entry of aMLV but is specifically required for HIV-1 entry.
The significance we attribute to the CD4 association with PDI may appear inconsistent with the observation that certain CD4-independent HIV-1 strains are infective (50,51). However, mutations that render such strains infective may alter the native Env structure and cause conformational changes that mimic those initiated by PDI⅐CD4. It has also been shown that CCR5 ϩ but CD4 Ϫ cells when mixed with Env-expressing cells undergo limited fusion in the presence of sCD4 (52). A participation of PDI in this fusion is not excluded, because it is known that PDI is extensively shed from the surface of mammalian cells (20,23,24) and HIV target cells (14). Therefore, soluble PDI may be present in the medium of the mixed cell populations used in the fusion assay. We showed that soluble PDI forms complexes with sCD4 (Fig. 1C), which may bind to surface-expressed Env. Such tripart complexes may allow PDI to cleave Env-disulfide bonds, thereby initiating the structural changes that lead to limited Env-mediated fusion of CD4 Ϫ cells.
How conformational changes in gp120 activate the fusogenic properties of neighboring gp41 remains an open question. A recent publication (53) provides insight into the way gp41 may interact with gp120 of HIV-1. The report identifies the gp41 sequence that makes contact with gp120. It suggests by analogy with corresponding structures in human T-cell lymphotrophic virus, type I that changes originating in the domain of virus-receptor interaction are transmitted to the gp120-gp41 contact site and activate the fusogenic potential of gp41 (53). An awareness of the participation of PDI in envelope remodeling should prompt further investigations of the PDI-induced structural changes and promote a better understanding of the molecular rearrangements in gp120 that lead to the transition of non-fusogenic gp41 to its fusion-active state. A role of PDI in HIV-1 entry consistent with our results is depicted in the working scheme of Fig. 8.
Besides describing a new function of surface PDI, our data identify a novel step in HIV-1 entry and suggest new targets for anti-HIV-1 agents. The usefulness of agents that inhibit PDI activity may be limited by the increasing number of physiological functions attributed to surface PDI (13)(14)(15)(16)(17)(18)(19)(20)(21). Other approaches suggested by our findings include preventing the interactions of PDI with CD4 and of PDI with critical gp120disulfide bonds. In addition, conformational changes triggered by disulfide bond reduction may uncover novel PDI-induced gp120 epitopes of potential use for vaccine development.