Stabilization of HIV-1 gp120-CD4 Receptor Complex through Targeted Interchain Disulfide Exchange*

HIV-1 enters cells via interaction between the trimeric envelope (Env) glycoprotein gp120/gp41 and the host cell surface receptor molecule CD4. The requirement of CD4 for viral entry has rationalized the development of recombinant CD4-based proteins as competitive viral attachment inhibitors and immunotherapeutic agents. In this study, we describe a novel recombinant CD4 protein designed to bind gp120 through a targeted disulfide-exchange mechanism. According to structural models of the gp120-CD4 receptor complex, substitution of Ser60 on the CD4 domain 1 α-helix with Cys positions a thiol in proximity of the gp120 V1/V2 loop disulfide (Cys126–Cys196), satisfying the stereochemical and geometric conditions for redox exchange between CD4 Cys60 and gp120 Cys126, and the consequent formation of an interchain disulfide bond. In this study, we provide experimental evidence for this effect by describing the expression, purification, refolding, receptor binding and antiviral activity analysis of a recombinant two-domain CD4 variant containing the S60C mutation (2dCD4-S60C). We show that 2dCD4-S60C binds HIV-1 gp120 with a significantly higher affinity than wild-type protein under conditions that facilitate disulfide exchange and that this translates into a corresponding increase in the efficacy of CD4-mediated viral entry inhibition. We propose that targeted redox exchange between conserved gp120 disulfides and nucleophilic moieties positioned strategically on CD4 (or CD4-like scaffolds) conceptualizes a new strategy in the development of high affinity HIV-1 Env ligands, with important implications for therapy and vaccine development. More generally, this chalcogen substitution approach provides a general means of stabilizing receptor-ligand complexes where the structural and biophysical conditions for disulfide exchange are satisfied.

The first step of HIV-1 2 entry into host cells involves the interaction between the viral envelope protein (Env) and the host cell receptor CD4. Functionally active (fusogenic) Env exists on the surface of virions as a trimer composed of three heterodimeric gp120-gp41 complexes (1). The latter are established through noncovalent association of monomeric gp120 with membrane-embedded gp41 (2)(3)(4), such that gp120 is presented on the exterior of the viral lipid membrane for interaction with CD4. The development of both Env-directed therapeutic compounds (5-10) and Env-based immunogens (11)(12)(13)(14)(15)(16)(17)(18)(19)(20) has been frustrated by the significant sequence heterogeneity known to exist among Envs, even from viruses that are phylogenetically closely related, which limits the cross-reactive breadth of anti-Env antibodies and therapies. Furthermore, Env is extensively glycosylated and conformationally flexible, which compounds the problem of Env ligand accessibility and reactivity. In the background of this extraordinary array of strategies evolved to enable viral escape from effective immune-mediated neutralization, and apart from a very small number of studies that have suggested the presence and relevance of CD4-independent viruses in the periphery during natural infection (21), the fundamental importance of CD4 for HIV-1 attachment and infection has dictated that at least one site on gp120, the CD4binding site (CD4bs), is structurally conserved and constitutively accessible to its cognate receptor. For this reason, much research has focused on the development of both immunogens that present stabilized forms of the CD4bs (22)(23)(24)(25), and recombinant protein or small molecule ligands targeting the CD4bs as attachment inhibitors or immunotherapeutic agents (5, 26 -30).
The challenge in this regard is substantial, however; after showing promising antiviral effects in cell culture against laboratory-adapted HIV-1 strains (9), recombinant soluble CD4 (sCD4) was shown to be ineffective against primary viral isolates (31), and very large doses of sCD4 were required to achieve modest reductions in viral loads in vivo (32). The most likely explanation for these observations, provided by pioneering structural studies on gp120 in complex with CD4 (33) and a CD4bs-directed antibody (IgGb12) (34), is that the initial interaction between CD4 and gp120 on the native Env trimer is relatively weak and that a more stable complex requires contact with surfaces that are constituted only after substantial conformational rearrangement of gp120. During natural infection, coordination of target cell CD4 molecules enables multiple contact events per trimer, and this avidity overcomes the kinetic barriers presented by the requirement for sequential sampling to a conformation competent for binding CD4 stably.
Thus, sCD4 per se, by virtue of a relatively lower overall affinity for Env, competes ineffectively with target cell-bound CD4. More recently, the development of various oligomeric CD4immunoglobulin chain fusion proteins has resulted in improvements in both the overall affinity for gp120 and pharmacokinetic characteristics of the CD4 fusion proteins. The translation of these improvements into viable therapeutic compounds with the ability to mediate potent reductions in viral loads at low concentrations remains elusive, however.
With this in mind, we examined structural models of the CD4-gp120 complex and evaluated the potential biochemical effects of several conservative mutations of critical CD4 gp120interacting residues. Our aim was to identify amino acid changes in the CD4 sequence that could be incorporated into the designs of soluble recombinant CD4 proteins, which could enhance the stability of their interactions with gp120 and correspondingly improve their competitive potencies. One particular CD4 mutation we studied was S60C, which resulted in a striking theoretical enhancement of the interaction with gp120 at this position. When modeled on a crystal structure of CD4 in complex with gp120 (Protein Data Bank code 1G9M), substitution of Ser 60 on the CD4 domain 1 ␣-helix with Cys positions a reactive thiol in the proximity of the gp120 V1/V2 loop disulfide (Cys 126 -Cys 196 ), satisfying the stereochemical and geometric conditions for redox exchange between the CD4 Cys 60 and gp120 Cys 126 and the consequent formation of an interchain disulfide bond (Fig. 1). Therefore, we proposed that the stereochemically conservative S60C CD4 mutation might result in negligible structural deviation from the wild-type sequence (35) and simultaneously promote a highly favorable interaction based on novel disulfide bond formation between the mutant sequence and the target gp120.
In this study, we present data showing that a functional recombinant 2dCD4 protein carrying the S60C mutation binds gp120 with a higher affinity than wild-type 2dCD4 and that this enhanced affinity is associated with targeted disulfide exchange. Furthermore, we show that 2dCD4 (S60C) competitively inhibits the binding of recombinant gp120 to CD4 in vitro more effectively than wild-type CD4, and potentiates a correspondingly more potent antiviral effect in cell culture. Considered with recent results by others, which have suggested critical functional (in addition to structural) roles for resident Env disulfides (36 -40), and noting the striking conservation of these disulfides in the context of a hypervariable sequence, we propose that targeted redox exchange between conserved gp120 disulfides and nucleophilic moieties (such as Cys) positioned strategically on recombinant CD4 scaffolds represents a revolutionary strategy in the development of high affinity HIV-1 Env ligands.
2dCD4 Expression, Purification, and Refolding-Recombinant pET15-2dCD4 (WT/S60C/ FD ) vectors were used to transform BL21 (DE3) E. coli; protein expression was achieved using a "leaky" (isopropyl 1-thio-␤-D-galactopyranoside-free) induction protocol by culturing freshly transformed bacteria in standard LB medium for 20 h at 20°C with vigorous shaking. The recombinant 2dCD4s were purified from inclusion bodies using a denaturing purification protocol and refolded in glutathione-containing refolding buffers. Briefly, following centrifugation of 200 ml of an overnight culture, the bacterial pellet was resuspended in 25 ml of phosphate-buffered saline (PBS, pH 7.4) containing 0.5 mg/ml chicken egg lysozyme (53,000 units/ mg, Sigma). The resuspension was incubated at 4°C for 1 h with gentle shaking, and the cells then lysed by three cycles of freeze/ thawing in liquid nitrogen and sonication. The lysate was centrifuged at 20,000 ϫ g for 30 min at 4°C, and the supernatant was discarded. The pellet was resuspended in 25 ml of solubilization buffer (PBS, pH 7.4, 8 M urea, 50 mM glycine, 2 mM ␤-mercaptoethanol, 20 mM imidazole) and homogenized using a tissue homogenizer (Omni International). The homogenate was centrifuged at 20,000 ϫ g for 45 min, and the supernatant was recovered and filtered through a 0.45-m syringe filter. A, gp120 (blue) and CD4 (green) are shown in ribbon representation, with Ser 60 on the CD4 D1 ␣-helix shown in ball-and-stick representation. The inset (enlarged in B) shows the apposition of Ser 60 with the Cys 126 -Cys 196 disulfide of the gp120 V1/V2 loop. C, molecular modeling of the CD4 S60C variant calculates the rotomeric positioning of the Cys 60 side-chain thiol to within 2 Å of the target gp120 disulfide and predicts the formation of the interchain CD4 Cys 60 -Cys 126 gp120 disulfide bond. The atomic structures of the CD4 S/C60 and gp120 Cys 126 -Cys 196 residues are shown in ball-andstick configuration (yellow, sulfur; gray, carbon; red, oxygen; blue, nitrogen; white, hydrogen).
Recombinant 2dCD4 was purified from the filtered, solubilized inclusion bodies in the presence of 8 M urea, 50 mM glycine, 2 mM ␤-mercaptoethanol, and imidazole (20 -500 mM) by standard Ni 2ϩ -affinity chromatography procedures. The purified protein (20 ml) was immediately dialyzed overnight at 4°C against 1 liter of folding buffer 1 (50 mM glycine, 10% sucrose, 1 mM EDTA, 1 mM reduced GSH, 0.1 mM oxidized glutathione (GSSG) and 4 M urea, adjusted to pH 9.6 with NaOH) and then changed into 1 liter of folding buffer 2 (50 mM sodium carbonate, pH 9.6, 10% sucrose, 1 mM EDTA, 0.1 mM GSH, 0.01 mM GSSG) for a second overnight dialysis at 4°C. The purified protein was then dialyzed exhaustively against PBS, pH 7.4 (4°C, three times with 2 liters, 4 h for the first two dialyses, and then overnight for the final dialysis), recovered, filtered through a 0.45-m syringe filter, and concentrated using Centricon ultrafiltration tubes (Millipore). The concentrated pure protein sample was aliquoted, snap-frozen in liquid nitrogen, and stored at Ϫ80°C. Using this method, we routinely purified between 0.5 and 1.0 mg of functional recombinant 2dCD4 (per 200 ml of bacterial cell culture).
Recombinant 2dCD4 Folding and Functionality Analysis-The purified, refolded 2dCD4s were analyzed by denaturing, nonreducing SDS-PAGE and CD spectroscopy to check structural disulfide bond formation and native folding patterns, respectively. For PAGE analysis, protein samples (2.0 g) were resolved on precast 12% BisTris polyacrylamide gels (Invitrogen) according to the manufacturer's instructions in the presence or absence of 50 mM 1,4-dithiothreitol (DTT), and the gels were stained with Coomassie Blue R-250. Circular dichroism spectra of 2dCD4-WT, 2dCD4-S60C, or 2dCD4 FD (10 M, in 10 mM sodium phosphate, pH 7.4, 140 mM NaCl) were collected using a Jasco J-810 spectropolarimeter with a 1-mm path length cuvette. Five scans were made at a scan rate of 100 nm/min, a bandwidth of 1 nm, and a response time of 2 s. All spectra were buffer-corrected, and normalized mean residue ellipticities (degrees⅐cm 2 /dmol) were calculated as described previously (41). All spectra were recorded at room temperature from 250 to 200 nm. The functionality of the recombinant 2dCD4s was tested by checking the efficiency of gp120 binding using surface plasmon resonance (SPR) (see below).
gp120 -2dCD4 (S60C) Complex Stabilization Analysis in Vitro-The ability of 2dCD4 (S60C) to bind gp120 and stabilize the interaction through targeted disulfide exchange was analyzed by denaturing, nonreducing PAGE and SPR. For gel shift analysis, binding reactions were set up in 20 l and contained 2 g of recombinant gp120 BaL , 6.5 g of recombinant 2dCD4-WT or 2dCD4-S60C, and 1 mM ␤-mercaptoethanol in PBS, pH 7.4. The reaction was incubated at 30°C for 2 h, before mixing with 4ϫ lithium dodecyl sulfate (LDS) sample loading buffer (Invitrogen) in the presence or absence of 50 mM DTT. The samples were then placed in a boiling water bath for 5 min and resolved on a precast 12% BisTris polyacrylamide gel. For specific detection of gp120, and free and gp120-bound 2dCD4 forms, the resolved proteins were then transferred to a nitrocellulose membrane by Western blotting and probed with an anti-gp120 antibody (ab21179, Abcam) and an anti-CD4 antibody (MT310, Santa Cruz Biotechnology), respectively. Detection was performed by standard chemiluminescence method-ologies using a secondary horseradish peroxidase-conjugated anti-IgG antibody (GE Healthcare). For SPR, all experiments were performed on a BIAcore 3000 optical biosensor (Biacore Inc., Uppsala, Sweden) at 25°C. For direct binding comparisons and kinetics, 2dCD4-WT, 2dCD4-S60C, or 2dCD4 FD were immobilized on separate flow cells within the same sensor chip using amine coupling according to the manufacturer's instructions. Briefly, a solution of 0.2 M 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride and 0.05 M N-hydroxysuccinimide was used to activate carboxyl groups on the sensor surface at a flow rate of 5 l/min for 10 min. The recombinant 2dCD4, dissolved in sodium acetate buffer (100 mM, pH 5.5) to a final concentration of 4.5 g/ml, was passed over the sensor chip surface at a flow rate of 5 l/min until the desired immobilization level (2500 response units) was achieved. Excess carboxyl groups were capped by injection of 1 M ethanolamine, pH 8.0, at a flow rate of 5 l/min for 5 min. Unbound ligand was removed by injecting glycine (10 mM, pH 2.5) at a flow rate of 5 l/min for 2 min. Flow cell four was left blank as a control for nonspecific binding and refractive index changes. Analyte solutions (gp120 BaL , 0 -500 nM) were prepared in HEPES-buffered saline supplemented with 3 mM EDTA, 1 mM ␤-mercaptoethanol, and 0.005% surfactant P20 (HBS-EP, pH 7.4) at least 30 min prior to injection and passed over the 2dCD4 surfaces at 10 l/min, with a 20-min association phase and either a 50-min (for kinetic analyses) or 10-h (for complex stability analysis) dissociation phase. To confirm disulfide linkage between 2dCD4-S60C and gp120, the surfaces from overnight dissociations were treated with increasing concentrations of DTT (10 -100 mM). When checking the specificities of the 2dCD4-gp120 interactions, mAb 17b was passed over the bound complexes at 200 nM. The 17b epitope on gp120 (a so-called CD4-induced (CD4i) epitope) is characteristically only efficiently exposed following gp120 binding to CD4, and checking the efficiency of 17b uptake provides an additional control for the quality of the gp120-CD4 interaction measured by SPR. Buffer injections and control surface binding were subtracted for all reported data, and all binding experiments were performed in triplicate, on separate chips.
2dCD4-gp120 Competition ELISA-96-Well Maxisorp microtiter plates (Nunc, Denmark) were coated with sCD4 (Progenics, 100 ng/ml in PBS, pH 7.4) for 3 h at room temperature (100 l per well). The coating solution was removed, and the plates were then blocked with 250 l of PBS containing 0.05% Tween 20 and 10 mg/ml bovine serum albumin for 16 h at 4°C. After removing the blocking buffer and rinsing once with PBS containing 0.05% Tween 20 (PBS-T), a mixture of 2dCD4 (WT/ S60C) (0 -1 M) and 1 nM of gp120 BaL was added to the wells (200 l in PBS-T), and the plates were incubated at 30°C for 2 h. When checking the effect of the reaction redox potential on relative potency of 2dCD4-mediated gp120-sCD4 binding inhibition, DTT was included in the binding reactions at 0, 10, and 100 mM. The wells were washed five times with 250 l of icecold PBS-T, and bound gp120 was then probed with 100 l of mAb 2G12 (1 g/ml in PBS-T) for 1 h at 30°C. After washing, bound gp120 was detected using 100 l of a horseradish peroxidase-conjugated monoclonal anti-human Fc antibody (Amersham Biosciences), diluted 1:2000 in PBS-T and incubated for 1 h at 4°C. Bound secondary antibodies were detected, follow-gp120-CD4 Stabilization by Interchain Disulfide Exchange AUGUST 13, 2010 • VOLUME 285 • NUMBER 33 ing washing, by chromogenic methods using TMB Ultra substrate (Pierce) and quantified spectrophotometrically by measuring A 450 levels.
Viral Inhibition Assay-The inhibitory effect of purified 2dCD4 (WT/S60C) on viral entry was assessed by an in vitro phenotypic inhibition assay (42) using an HIV-1 pseudovirus backbone (pSG3 renv ) complemented with a subtype C CAP210 envelope. Briefly, a 3-fold serial dilution of each 2dCD4 variant (in PBS, pH 7.4) was prepared in a 96-well plate (Nunc), with concentrations ranging from 22.0 to 0.03 g/ml. The diluted compounds were incubated with 200 TCID 50 of pseudovirus for 1 h at 37°C. Ten thousand TZM-bl cells (in Dulbecco's modified Eagle's medium containing 12 g/ml DEAE) were then added to each well, and the cells were incubated at 37°C for a further 48 h in a humidified CO 2 incubator. Nonvirus and virus controls were prepared in exactly the same manner, except for the omission of virus or test compounds where appropriate. Luciferase activity induced in the TZM-bl cells (measured in relative light units) was quantified using the Bright-Glo luciferase assay system (Promega) according to the manufacturer's instructions. Inhibition of viral replication was expressed as a percentage of the reduction in relative light unit values generated in the absence of any compound. Cell culture toxicity was measured using the CellTiter-Glo luminescent cell viability assay (Promega) according to the manufacturer's instructions. The significance of the difference between 2dCD4-WT-and 2dCD4-S60C-treated samples (during ELISA and virus inhibition experiments) was evaluated using a paired, two-tailed Student's t test at a confidence level of 95%.

RESULTS
Purification of Native, Functional 2dCD4-WT and 2dCD4-S60C-Because the reducing environment of the E. coli cytoplasm often renders correct folding of structural disulfide-containing proteins expressed in these hosts problematic, previous studies on the expression of functional recombinant 2dCD4 in E. coli have either made use of technologies to secrete the expressed protein to the bacterial periplasm (43) or post-purification refolding protocols (44 -45). We initially employed a periplasmic secretory vector (pET22a, Novagen) to express and purify both 2dCD4-WT and 2dCD4-S60C. Although we were able to purify an active wild-type 2dCD4 expressed by this vector from the periplasmic fraction, we were unsuccessful in this endeavor with the free thiol containing 2dCD4 variant (S60C) (data not shown), presumably because the presence of an unpaired cysteine in this polypeptide interferes with efficient periplasmic secretion and/or canonical disulfide pairing and oxidation in this milieu. We therefore decided to apply a controlled oxidative refolding protocol to the recombinant 2dCD4s affinity-purified from E. coli inclusion bodies. Because 2dCD4-S60C contains a single unpaired cysteine in addition to the single cysteine pairs in D1 (Cys 16 -Cys 84 ) and D2 (Cys 130 -Cys 159 ) of CD4, one potential problem in folding these proteins into the native 2dCD4 structure could relate to aberrant disulfide pairing during the refolding process. To qualify and quantify different disulfide-bonded intermediates generated during the oxidative refolding process and gain insights into the ability of the S60C variant to fold into a native 2dCD4 structure, we analyzed disulfide oxidation in the refolded proteins by reducing and nonreducing SDS-PAGE (46).
Under reducing conditions, both 2dCD4-WT and 2dCD4-S60C migrated as a single band of ϳ21 kDa, corresponding to the predicted size for the fully denatured, reduced 2dCD4 isoform ( Fig. 2A). When the proteins were analyzed under nonreducing conditions, the 2dCD4 proteins resolved as three distinct species that migrated between 17 and 21 kDa, most likely representing reduced protein (R) and the predominant disulfide-bonded 2dCD4 isoforms (O 1 /O 2 ) ( Fig. 2A). To gain further insights into which of the oxidized species was likely to represent the 2dCD4 isoform containing the correct disulfidebonded cysteine pairs (Cys 16 -Cys 84 and Cys 130 -Cys 159 ), we generated a folding-defective 2dCD4 mutant (2dCD4 FD ) containing a five-amino acid substitution proximal to the D1 ␣-helical loop (residues 65-69, WT sequence, -GNFPL; 2dCD4 FD sequence, -VPGLV). Under nonreducing conditions, this mutant, which was expressed, purified, and refolded using the same procedures as 2dCD4-WT and 2dCD4-S60C, resolves into three predominant species (Red, O 2 , and O 3 ), with little or no O 1 isoform present in the mixture ( Fig. 2A). When the functionality of the protein was analyzed by SPR assay, 2dCD4 FD was unable to bind gp120; in contrast, 2dCD4-WT and 2dCD4-S60C, which have the same pattern of bonding and for which O 1 is the predominant oxidized isoform, all bound gp120 efficiently (Fig. 2B).
To compare folding characteristics and further evaluate whether the variants had acquired native secondary structural elements typical of 2dCD4, we then analyzed the purified and refolded proteins by CD spectroscopy (Fig. 2C). Far-UV spectra for the 2dCD4 WT and S60C proteins were virtually identical and typical of the classical ␤-structure spectrum, mean ellipticity minima and maxima were at 212-214 and 200 nm, respectively, in line with crystallographic data showing the predominant contribution of ␤-strands to the 2dCD4 structure, and consistent with published CD data for 2dCD4 (47)(48). Conversely, the spectrum generated by 2dCD4 FD contained a large leftward shift in ellipticity minima and maxima, suggestive of irregular ␤-stranded structure and confirming, as expected, the failure of the folding-defective mutant 2dCD4 to fold into the native ␤-stranded conformation.
Taken together, these data suggest that O 1 represents the 2dCD4 isoform containing the correct disulfide pairs and that 2dCD4-S60C, like the wild-type protein, folds into a native 2dCD4 structure (with demonstrably efficient gp120 binding) through oxidation of correctly paired cysteine residues.
2dCD4-S60C Binds gp120 Stably through Disulfide Exchange-We adopted two independent approaches to study the gp120binding properties of purified 2dCD4-WT and 2dCD4-S60C. First, we used an electrophoretic mobility shift assay to detect gp120-bound 2dCD4 following resolution of the protein complexes on denaturing polyacrylamide gels. This method also enabled us to determine whether the mechanism of binding of 2dCD4-S60C could additionally involve establishment of disulfide bonds with the target gp120, by measuring the stability of bound 2dCD4 under reducing and nonreducing conditions.

gp120-CD4 Stabilization by Interchain Disulfide Exchange
ished the ability of 2dCD4-S60C to remain bound during SDS and heat denaturation, suggesting that the stability of the gp120 -2dCD4-S60C complex is related to intermolecular disulfide bond formation.
To elaborate on the gp120 -2dCD4 binding data generated by gel shift assay and to define the kinetic parameters associated with the respective interactions, we performed 2dCD4-gp120 binding analysis by SPR using a BIAcore 3000 instrument. Purified 2dCD4-WT or 2dCD4-S60C were immobilized on CM5 sensorchips and exposed to increasing concentrations of gp120 (0 -500 nM). The association rate (k a ), dissociation rate (k d ), and equilibrium (K D ) constants for each reaction were calculated by fitting the generated sensorgrams directly to a 1:1 Langmuir binding model using the BiaEval software, and representative data from three independent kinetic analyses are shown in Table 1. The calculated affinity for the interaction between 2dCD4-WT and gp120 was 53 nM, consistent with data published by others (33). In contrast, the affinity of 2dCD4-S60C for gp120 was markedly higher, with a calculated value of 6 nM. This increase in affinity is attributed exclusively to the significantly slower off-rate kinetics of 2dCD4-S60C, whereas the rates of association between the latter and wildtype 2dCD4 are almost identical. To qualify further the difference in off-rate kinetics and to determine whether this could be related, as suggested by the gel shift analysis, to a redox exchange between the gp120 and 2dCD4-S60C, we made a direct comparison of the curves generated for each protein during an extended dissociation phase and checked the effect of DTT injection on complex dissociation (Fig. 4). gp120 remained stably bound to the 2dCD4-S60C surface for at least 10 h after withdrawal of the analyte solution; during the same time period, gp120 had almost completely dissociated from the 2dCD4-WT (Fig. 4A). Injection of 10 -100 mM DTT over the surfaces resulted in a dramatic dose-related increase in the rate of dissociation of 2dCD4-S60C (Fig. 4B), with the k d value approaching that calculated for 2dCD4-WT under normal conditions (Table 1). To confirm that the sensorgrams generated for gp120 binding to the 2dCD4 surfaces represented bona fide 2dCD4-gp120 interactions, we checked whether the binding was associated with induction of the epitope for mAb 17b on gp120, which is only efficiently and stably exposed following CD4 engagement. In line with this, efficient (and near stoichiometric) binding of 17b occurred only after binding of gp120 to both 2dCD4-WT and -S60C surfaces, qualifying the sensorgrams as representative of functional 2dCD4-gp120 complex formation (Fig. 4C). Taken together, these in vitro data confirm that 2dCD4-S60C binds gp120 with superior affinity to wildtype 2dCD4 through a mechanism that involves targeted intermolecular disulfide exchange.
2dCD4-S60C Is a Strong Competitive Inhibitor of gp120-CD4 Binding and Viral Replication-We next set up sCD4-gp120 competition ELISAs and pseudovirion inhibition assays to check whether the increased affinity of 2dCD4-S60C for gp120 resulted in enhanced competitive inhibition of gp120-CD4 binding and HIV-1 replication in vitro. Recombinant gp120 was challenged with increasing concentrations of 2dCD4-WT or 2dCD4-S60C and then probed over immobilized sCD4. 2dCD4-S60C mediated a significantly more potent inhibition of gp120 binding to immobilized sCD4 than 2dCD4-WT under the conditions tested (Fig. 5A), with 50% inhibition of gp120 binding achieved by 0.025 and 0.005 g/ml of the wild-type and S60C proteins, respectively. This difference was significant across the entire range of 2dCD4 concentrations tested (p Յ 0.05). To confirm whether the observed increase in binding inhibition mediated by 2dCD4-S60C was related to thiol oxidation, we performed the competition assays in the presence of increasing concentrations of DTT. We calculated the fold-increase in gp120-binding inhibition mediated by 1.0 nM 2dCD4-S60C (relative to 2dCD4-WT) in the presence of increasing concentrations of DTT. Lowering the reaction oxidation potential with DTT reduced the relative potency of 2dCD4-S60C-mediated inhibition in a dose-dependent manner (p Յ 0.05), with the inhibitory effect of 2dCD4-S60C almost identical to that of the wild-type protein in the presence of 100 mM DTT (Fig. 5B).
We then set out to establish whether this enhancement in gp120 antagonism in vitro would translate into an improvement in viral inhibition mediated by 2dCD4-S60C. When a sub-FIGURE 3. Electrophoretic mobility shift analysis of 2dCD4-gp120 complexes. gp120 -2dCD4-WT and gp120 -2dCD4-S60C complexes were heatand SDS-denatured in the presence (panel 4) or absence (panel 3) of 50 mM DTT, resolved on a 12% denaturing polyacrylamide gel, and detected with an anti-CD4 antibody following immunoblotting. To confirm the specificity of antibody binding, the recombinant gp120 and 2dCD4 protein were resolved independently and detected with anti-gp120 (panel 1) and -CD4 (panel 2) antibodies, respectively. 2dCD4-WT 53 5.0 ϫ 10 3 2.6 ϫ 10 Ϫ4 Ϫ 5.4 2dCD4-S60C 6 5.6 ϫ 10 3 3.6 ϫ 10 Ϫ5 1.2 ϫ 10 Ϫ4 2.5 ϫ 10 Ϫ4 1.6 a Equilibrium (affinity) constant is shown. b Association rate constant is shown. c Dissociation rate constant is shown. d Dissociation rate constant in the presence of 10 mM DTT is shown. e Dissociation rate constant in the presence of 100 mM DTT is shown. f Concentration of protein required to effect a 50% reduction in viral infection is shown.

gp120-CD4 Stabilization by Interchain Disulfide Exchange
type C HIV-1 envelope-pseudotyped virion (CAP210) was challenged with the recombinant 2dCD4 variants in a cell culture inhibition assay, 2dCD4-S60C effected a 50% reduction in replication of this virus at an average concentration of 1.6 g/ml, although the corresponding IC 50 value for the wild-type protein was over 3-fold higher (5.4 g/ml) (Fig. 5C and Table 1). The observed effects were nontoxic and specifically related to the competitive effects of the 2dCD4, because VSV-G replication was not inhibited by the same challenge (Fig. 5D).

DISCUSSION
In this study, we describe the development of a recombinant two-domain CD4 protein containing a single S60C mutation (2dCD4-S60C), which is able to be folded as native 2dCD4 and binds recombinant gp120 with high affinity through the formation of an intermolecular disulfide bond. There are several potentially profound sequelae of these findings, for both anti-HIV therapy and vaccine research. In the first case, appropriate incorporation of oxidizable thiols (or other nucleophilic moieties) into the designs of existing CD4bs-targeted gp120-binding compounds may significantly enhance the affinities of such molecules for their target, correspondingly increasing the potency of their therapeutic effects. An example for which this may be highly relevant is that of the CD4-mimetic miniprotein CD4M33, which was first described by Vita and co-workers (8), and elaborated on more recently in a study that evaluated a structural analogue of CD4M33 with improved CD4 mimicry (F23) (49). The CD4M33/F23 compounds are composed of sequences resembling the CDR2-like loop of CD4 transplanted into the structurally similar region on the scorpion toxin charybdotoxin, a stabilized structure that acts as a scaffold for orienting and presenting this critical gp120-binding domain at the hydrophobic CD4-binding pocket. They exhibit slightly lower affinities for gp120 than CD4, but nonetheless they have broadly neutralizing activity against a range of laboratoryadapted and primary HIV-1 isolates. Moreover, these compounds induce structural changes in gp120 analogous to those that occur upon CD4 binding, exposing cryptic epitopes that may be vulnerable to attack by neutralizing antibodies. Thus, these compounds have potential applications as both therapies and as components of immunogens that could potentiate effective anti-HIV neutralizing antibody responses. In this regard, however, the relatively low affinity of the CD4-mimetic compounds is one of the fundamental obstacles in the development of these compounds for clinical use, and the potential utility of covalently linking F23 to gp120 has been articulated (49). As a potential therapy, lower affinity than CD4 limits the ability to compete effectively with the natural receptor in vivo at concentrations that are safe and realistically achievable. As a component of an immunogen, unstable binding between the ligand and Env target results in failure to stabilize the induced conformation in a form that constitutively exposes neutralization epitopes. Indeed, several studies have demonstrated that animals immunized with Env-CD4 complexes stabilized through chemical cross-linking develop strong neutralizing antibody responses (50,51). The overall efficacy of this strategy may be limited, however, by modification of epitopes incurred through the chemical cross-linking procedure and by the potential of these immunogens to elicit autoreactive anti-CD4 antibodies. Targeted linkage of CD4-mimetic compounds as described here may, in turn, be an alternative means of stabilizing CD4induced gp120 complexes.
Beyond representing a possible opportunity for complex stabilization through disulfide exchange, an increasing amount of FIGURE 4. gp120 -2dCD4 interaction analysis by SPR. A, recombinant gp120 (500 nM) was bound to immobilized 2dCD4 (20-min association phase) and allowed to dissociate over a 10-h period. B, at the end of this dissociation phase (inset of A), DTT (10 -100 mM) was injected over the surfaces, which resulted in specific, concentration-dependent destabilization of the gp120 -2dCD4-S60C complexes. C, quality of the interaction between gp120 and immobilized 2dCD4 was assessed by measuring CD4-induced (CD4i) epitope exposure, by injecting, following gp120 binding on the immobilized 2dCD4-WT or 2dCD4-S60C surfaces, mAb 17b at 200 nM. 2dCD4-WT, blue; 2dCD4-S60C, red; gp120 surface (no CD4 induction), gray. RU, response units. AUGUST 13, 2010 • VOLUME 285 • NUMBER 33 data points to the crucial structural, and probably functional, role of Env disulfides. It is now well established that the HIV-1 envelope undergoes extensive conformational change following engagement of the CD4 receptor, a process that is likely to require reduction of several of the nine disulfide bonds located within gp120. Indeed several reports have suggested the importance of certain redox proteins in the HIV entry process through their ability to reduce gp120 disulfides (37,52). More specifically, van Anken et al. (39) systematically studied the contribution of each gp120 disulfide to the production of correctly folded and functional Env and found that several, including the Cys 126 -Cys 196 disulfide targeted in this study, are fundamentally important for the fusogenic activity of Env. Thus, HIV gp120 disulfide bonds are not only useful electrophilic handles for compounds targeting Env, they also represent critically important functional components of the viral entry apparatus. Their disruption through targeted reduction could thus represent a powerful new therapeutic approach, particularly so because they are strikingly conserved across most (if not all) functional envelopes sequenced to date.

gp120-CD4 Stabilization by Interchain Disulfide Exchange
It should be noted that the ability of ligand-based thiols to engage in trans-disulfide exchange reactions with target disulfides on cognate receptors is theoretically dependent on several biophysical factors. In particular, the respective redox potentials of the cis-and trans-disulfides, the availability of reducing equivalents to facilitate the exchange reaction, and the overall stability of the disulfides within the context of the liganded and unliganded structures collectively define the favorability of trans-disulfide bond formation, and the corresponding propensity for reversion to cis. In the case of the gp120 -2dCD4 (S60C) interaction in vitro, the reaction between the recombinant proteins proceeds favorably in the presence of low concentrations of ␤-mercaptoethanol (1 mM), which provides the reductive force required to produce a reactive Cys 126 -Cys 196 dithiol intermediate. Once the exchange reaction has taken place, we propose the transdisulfide bond is stabilized by cooperativity with other canonical CD4-gp120 interactions. In vivo, the requirement for "priming" the target Cys 126 -Cys 196 disulfide is likely to be fulfilled by one of a number of redox enzymes that have been shown to act on gp120 disulfides (37,40,52), explaining the effectiveness of 2dCD4 (S60C) in viral inhibition assays without any reducing agent supplementation. Ongoing studies in our laboratory involve characterizing the gp120 -2dCD4 (S60C) exchange reaction thermodynamically using isothermal titration calorimetry, testing the antiviral efficacy of 2dCD4 (S60C) across a broad range of diverse HIV-1 isolates, and the generation of 2dCD4 variants containing selenocysteine (Sec, U) point substitu- FIGURE 5. 2dCD4-S60C-mediated inhibition of sCD4-gp120 binding and HIV-1 replication. A, competition ELISA. Recombinant gp120 was challenged with increasing doses of 2dCD4-WT or 2dCD4-S60C and allowed to react with sCD4 immobilized on a Maxisorp 96-well ELISA plate. Bound gp120 was detected with mAb 2G12 and a horseradish peroxidase-conjugated anti-human IgG, and the levels were expressed as a percentage of binding in the absence of any compound. The experiment was performed in triplicate on three independent occasions, and representative data from one of these is shown. B, relative potency of gp120-sCD4 binding inhibition mediated by 2dCD4-S60C, expressed as the ratio of the % gp120 bound in the presence of 2dCD4-WT:2dCD4-S60C (1.0 nM), decreases with increasing concentrations of the reductant DTT. Ratios were calculated from three independent experiments, and error bars represent positive standard deviations. C, pseudovirus inhibition assay. An HIV-1 subtype C Env-pseudotyped virus (CAP210) was challenged with increasing doses of 2dCD4-WT or 2dCD4-S60C and incubated with TZM-bl cells for 48 h at 37°C. Entry inhibition was calculated by measuring viral tat-induced luciferase activity and is reflected as a percentage of the luciferase activity in the absence of any 2dCD4. The experiment was performed in triplicate on three independent occasions, and representative data from one of these are shown. D, VSV-G inhibition and cell viability assays. VSV-G was treated in the same way as the HIV-1 pseudoviruses and controlled for the specificity of the inhibitory effects (dotted lines), whereas the viability of the TZM-bl cells treated with 2dCD4-WT and 2dCD4-S60C (expressed as a percentage of the viability in the absence of any treatment, solid lines) evaluated the relative toxicities of the 2dCD4 compounds. In all experiments, error bars are shown for each concentration of recombinant protein tested and represent positive standard deviations. The significance of the differences between the 2dCD4-WT-and 2dCD4-S60C-treated samples was calculated at each concentration tested using a paired, two-tailed Student's t test, and significant differences (p Յ 0.05) are indicated by asterisks.

gp120-CD4 Stabilization by Interchain Disulfide Exchange
tions (S60U), which should yield ligands with even higher affinities for gp120 by virtue of the extreme nucleophilicity of selenocysteine.
Structural studies have now identified the key features of HIV-1 gp120 and the mechanism of its interactions with the host cell receptor. These have shown the following: (i) prior to binding CD4, Env exists in a heavily glycosylated, conformationally flexible form that is highly resistant to antibody neutralization; (ii) CD4 binding occurs at a structurally conserved site that is vulnerable to neutralizing antibody and appropriately designed CD4-mimetic compounds and is accompanied by extensive structural rearrangements that expose additional neutralization epitopes. In turn, these pioneering studies have laid the platform for the rational design of a range of therapeutic and immunogenic compounds, which have already made significant contributions in the fields of antiviral and vaccine research. In this study, we describe a structurally conservative CD4 mutation (S60C) that effects a redox exchange reaction with the gp120 Cys 126 -Cys 196 disulfide and enhances the affinity of the gp120-CD4 interaction. We propose that this approach, in turn, represents the conceptualization of a revolutionary approach to enhancing the efficacies of gp120-targeted therapies and Env-based immunogens. Moreover, although we have discussed our findings in the context of HIV-1 therapy (and vaccine) development, it has not escaped our attention that quaternary structure stabilization by redox-active variants of natural ligands, produced in native form by strategic substitution of the structurally similar chalcogen-containing amino acids, may represent a generally applicable approach to receptor-ligand stabilization where structural and biophysical conditions for disulfide exchange prevail.