Transduction of Activation Signal That Follows HIV-1 Binding to CD4 and CD4 Dimerization Involves the Immunoglobulin CDR3-like Region in Domain 1 of CD4*

The role of CD4 during the human immunodeficiency virus type 1 (HIV-1) life cycle in T cells is not restricted to binding functions. HIV-1 binding to CD4 also triggers signals that lead to nuclear translocation of NF-κB and are important to the productive infection process. In addition to its cytoplasmic tail, in the ectodomain, the immunoglobulin (Ig) CDR3-like region of CD4 domain 1 seemed to play a role in this cascade of signals. We demonstrate in this work that the structural integrity of the CDR3-like loop is required for signal transduction. Substitutions of negatively charged residues by positively charged residues within the CDR3-like loop either inhibited NF-κB translocation after HIV-1 and gp120-anti-gp120 immune complexes binding to E91K,E92K mutants or induced its constitutive activation for E87K,D88K mutants. Moreover, A2.01–3B cells expressing the E91K,E92K mutant exhibited a lower HIV-1Lai replication. These cells, however, expressed p56 lck , demonstrated NF-κB translocation upon PMA stimulation, bound HIV-1Lai envelope glycoprotein with high affinity, and contained HIV-1 DNA 24 h after exposure to virus. E91K, E92K, and E87K,D88K mutant CD4 molecules were unable to bind a CD4 synthetic aromatically modified exocyclic, CDR3.AME-(82–89), that mimics the CDR3-like loop structure and binds to native cell surface CD4. This result together with molecular modeling studies indicates that the CDR3.AME-(82–89) analog binds to the CDR3-like loop of CD4 and strongly suggests that this region represents a site for CD4 dimerization. The negative charges on the CDR3-like loop thus appear critical for CD4-mediated signal transduction most likely related to CD4 dimer formation.

The CD4 molecule is an integral membrane glycoprotein that contains four extracellular domains (D1-D4) showing structural homology with immunoglobulin V regions (1,2). This molecule plays a key regulatory role in the immune system by stabilizing MHC 1 class II⅐T cell receptor complex interactions (3)(4)(5) and also by acting as a signal-transducing molecule during T cell activation (6,7), by its association with the proteintyrosine kinase p56 lck .
Beside its physiological function, CD4 is known to be the primary high affinity cellular receptor for HIV-1. The virus outer envelope glycoprotein (HIV-1gp120 env ) can bind to a protruding ridge located within the IgV-like D1 domain of CD4 at the CDR2-like region (8,9). Increasing evidence indicates that the role played by CD4 during the HIV-1 life cycle in T cells is not limited to its ability to serve as a receptor for the virus but probably plays other roles that are important to the productive infection process. Indeed, it has been found that transfected cells expressing mutated forms of CD4 that lack the cytoplasmic domain are fusion competent and yet present a defect in a latter stage of HIV-1 replication cycle which results in delayed HIV-1 production (10 -12). Further studies have demonstrated that truncation of CD4 cytoplasmic domain blocks NF-B translocation induced by HIV-1-mediated oligomerization of native CD4, leading to the conclusion that under certain circumstances HIV-1 can take advantage of the signal transduction function of CD4 to prepare the cell for postfusion events (13,14). More recently, we have shown that HIV-1 binding to CD4 at the surface of peripheral blood mononuclear cells activates tyrosine phosphorylation of a number of transducing proteins, including phosphatidylinositol 4-kinase and mitogenactivated protein kinase-2 which are possible intermediates in the activation pathway(s) that regulates the activity of the viral promoter (15) and activates NF-B and AP-1 transcription factors (16).
Although the CDR3-like loop in D1 of CD4 plays no role (17) or a minor role (18) for HIV-1 entry, this region may play an important role during HIV-1 replication. This was suspected from antiviral properties of antibodies directed to this region of D1 (17) and synthetic peptides resembling the CDR3-like loop (19). More recently, we have demonstrated that anti-CD4 mAb specific for the CDR3-like region inhibited HIV-1 promoter activity and HIV transcription in cells containing an integrated provirus(es) (20,21). Moreover, these mAbs inhibit HIV-1 induced mitogen-activated protein kinase-2 activation (22) and HIV-1-induced nuclear translocation of NF-B (23), indicating that the CDR3-like loop is involved in signal transduction triggering T cell activation. Based on the observation that CDR3-like loop-derived synthetic peptide can bind CD4, it has been proposed (24) that this region may be involved in CD4 dimerization process which may be required for CD4-dependent T cell activation. Recently, a constrained aromatically modified exocyclic (AME) analog that mimics the CDR3-like loop secondary structure, CDR3.AME-(82-89), was shown to inhibit CD4-MHC class II binding and T cell activation (25). Since the CDR3.AME analog binds to cell surface CD4 molecules it has been proposed that the AME prevents the formation of an essential homodimeric surface involving the CDR3-like region of CD4 (25).
Here we investigated further the structural requirements of the CDR3-like loop responsible for the cascade of signal transduction events that lead to NF-B translocation after HIV-1 binding to CD4 and probed the role played by the negative charges in this region by using a series of A2.01 T cell clones expressing different mutants of CD4. We demonstrate that the negatively charged residues at positions 87, 88, 91, and 92 in the CDR3-like loop play a critical role during the initiation of the activation signal transduction and in the binding of the CDR3.AME-(82-89) analog to CD4. Moreover, molecular modeling data show that the CDR3.AME-(82-89) analog can bind to the CDR3-like loop and indicate that residues 87, 88, 91, and 92 are essential for CD4 dimerization.

HIV Infection Assay
Cells (5 ϫ 10 5 ) were incubated for 30 min at 4°C in flat-bottom 96-microwell plates (cell culture top performance product (TPP)) with 100 l of HIV-1 at 10 3 ϫ TCID 50 per ml. Thereafter, cells were washed five times and cultured at 37°C in 24-microwell plates (TPP). The amount of virus produced by the cells was monitored twice a week by measuring reverse transcriptase activity in 1 ml of cell-free culture supernatant, as described previously (27).

Reverse Transcriptase-Polymerase Chain Reaction (PCR)
PCR detection of reverse-transcribed RNAs was performed according to the previously published procedure with slight modifications (17). Briefly, total RNA was extracted in guanidinium thiocyanate from 4 ϫ . PCR were carried out on 4 l of sample supplemented with an amplification mixture containing the p56-I/p56-II or TK I/TK II oligonucleotide primer pair (23), and 2 units of Taq DNA polymerase. The amplification reaction was run in a PHC2 thermal cycler (Techne, Cambridge, UK). The amplified products were electrophoresed in a 2% agarose gel, blotted for 2 h onto Hybond N ϩ membrane (Amersham Corp.), and hybridized with ␣-32 P-labeled p56 lck probe prepared from pGEX2t-lck vector provided by S. Fisher (ICGM, Paris). A similar method was used for detection of HIV-1 mRNAs using the M671/VPR3 oligonucleotide primer pair (29) and an ␣-32 P-labeled HIV-1 probe.

PCR Analysis of HIV-1 DNA
HIV-1 DNA was monitored by PCR analysis (see above) using the M667/M668 oligonucleotide primer pair (20). The PCR was performed on total DNA extracts (1 g) prepared from cells at 24 h following virus exposure. The amplified products were electrophoresed, blotted, hybridized with an ␣-32 P-labeled HIV-1 probe, and visualized by autoradiography.

Western Blot Assay
Cellular lysates were electrophoresed onto 12.5% SDS-polyacrylamide gel electrophoresis and blotted to polyvinylidene difluoride membrane (Millipore). The blot was then incubated for 1 h at room temperature with a blocking solution (PBS containing 10% milk and 0.05% Tween 20) prior to addition of mAb. After 1 h at 20°C, the blot was washed three times with PBS, 0.05% Tween 20 and incubated for 30 min with 1:5000 dilution of GAM Ig peroxidase conjugate (Immunotech). After 3 washes, bound mAb was detected by incubating the membrane for 1 min with ECL reagent (Amersham Corp.). The membrane was then exposed for 0.5-5 min to hyperfilms-ECL (Amersham Corp.).

NF-B Electrophoretic Mobility Shift Assay (EMSA)
Nuclear extracts were prepared as described previously (13). Briefly, cells (2 ϫ 10 6 ) were centrifuged, transferred into 1.5-ml Eppendorf tubes and washed 3 times in PBS by centrifugation at 2,000 rpm for 10 min at 4°C. The pellet was resuspended in 400 l of buffer A (containing 10 mM KCl, 2 mM MgCl 2 , 0.1 mM EDTA, 1 mM DTT, 0.1 mM phenylmethylsulfonyl fluoride, 4 g/ml leupeptin, and 10 mM Hepes, pH 7.8). After 15 min on ice, 50 l of a solution of 10% Nonidet P-40 was added to the sample, and cells were homogenized by vortexing and microcentrifuged at 4°C for 30 s. The pellets were resuspended in 100 l of buffer B (containing 50 mM KCl, 300 mM NaCl, 0.1 mM EDTA, 1 mM DTT, 0.1 mM phenylmethylsulfonyl fluoride, 4 g/ml leupeptin, 10% glycerol, and 50 mM Hepes pH 7.8) and incubated for 20 min at 4°C on a shaker. The nuclear extracts were microcentrifuged at 4°C for 5 min, and the supernatants were stored at Ϫ80°C until used. The NF-B mobility shift assay was performed using 2 g of protein of nuclear extract, 1 ϫ 10 5 cpm of 32 P-labeled probe (a double-stranded oligonucleotide NF-B with a HIV-1 sequence: sense strand only, 5Ј-GCTGG GGACT TTCCA GGGAG GCGT-3Ј) in buffer C (containing 100 mM KCl, 1 mM DTT, 1 M ZnSO 4 , 20% glycerol, 0.01% Nonidet P-40, and 50 mM Hepes pH 7.9), supplemented with BSA, tRNA, and poly(dI-dC) in a final volume of 20 l. After 20 min at room temperature, the mixture was run at 120 V in a 10% polyacrylamide gel. Sp-1 mobility shift assay was performed using a double-stranded oligonucleotide probe (sense strand only, 5Ј-GGAGG CGTGG CCTGG GCGGG ACTGG GGAGT GGCGA-3Ј).

Molecular Modeling
Construction of Mutants-Molecular modeling was carried out using QUANTA (Molecular Simulations, Inc., Cambridge, MA). The coordinates of CD4 were obtained from Brookhaven Protein data base (30), and the solvent molecules were removed. The mutant models were minimized while holding the remainder of the structure fixed except CDR3-like region, preserving the overall structure. The entire structure was subjected to conjugate gradient energy minimization for 2000 cycles to convergence, followed by an equilibration and production run of molecular dynamics at 300,000 for 60 ps. Molecular dynamics was performed to allow a change in the conformation in the loop, if more favorable. All energy calculations were performed at a dielectric constant of 1. Final energy values were calculated using CHARMm which is part of QUANTA. Electrostatic calculations were performed with GRASP (31). Charged amino acid groups were assigned full charges as provided in GRASP. The electrostatic calculations were performed with distance-dependent dielectric constants from 1, at the interior, to 80, at the outer surface.
Positioning of AME Cyclic Peptide-The CDR3-AME cyclic peptide adopts very similar conformation of the native CDR3 loop of CD4 (25). A dimeric model of CD4 was built with D1 as dimeric interface using an antibody VH-VL domain as template. The CDR3-AME peptide was positioned by superimposing the CDR3-AME to the CDR3 of one of the dimeric molecule. Since this region is solvated, no attempt was made to dock this peptide to the CD4 molecule. The positioning was done only for reference purpose. Cells were incubated with medium alone (to determine the background), anti-CD4 mAb Leu 3a, 13B8-2, ST40, or BL4, or anti-HLA class I mAb B9 -12-1. mAb binding was detected by a FITC-labeled GAM Ig. The fluorescence intensity was recorded using the log mode. mutants that have been recently constructed (18) and were stably expressed in the A2.01 T cell line, a CD4 negative line derived from the CD4 positive parent A3.01 (5). Fig. 1 summarizes the amino acid changes introduced in the CD4 molecule. Mutant CD4 molecules with a substitution of Gly for Glu at position 87 (E87G) were expressed in A2.01-M1 cell line; the Q89L substitution was expressed in A2.01-M2 and the double substitutions E87K,D88K in A2.01-2B and E91K,E92K in A2.01-3B. A3.01 cells that express the native CD4 molecule, A2.01-c26 which did not express CD4 although transfected with the CD4 wild type gene, and A2.01/CD4.403 cells expressing a truncated form of CD4 lacking the whole cytoplasmic domain were used as controls. The CD4 phenotype of the A3.01 and A2.01 cell lines and of the different clones was studied by indirect immunofluorescent staining and flow cytometry. Representative cytofluorometric profiles are shown in Fig. 2. A2.01-c26 was the only clone that was not stained with anti-CD4 mAb. All other clones expressed CD4, although there were slight variations in the level of expression; binding of anti-CD4 mAb BL4 was found with all these cells. Binding of anti-CD4 mAb ST40 was disrupted only for clones A2.01-2B and A2.01-3B expressing CD4 molecules with the double substitutions E87K,D88K and E91K,E92K, respectively. As shown in Fig. 3, the different clones used in this study expressed the lck mRNA (Fig. 3A), and the p56 lck protein was detected (Fig. 3B), indicating that conditions for CD4-dependent signal transduction through p56 lck were conserved.

Expression of CD4
These results, together with previous work (18), indicate that the mutations introduced into CD4 had a local influence affecting only the CDR3-like loop structure. These mutations had no obvious effect on the conformation of domain 1.
As shown above (see Fig. 3), A2.01-3B cells express p56 lck and therefore should be able to transduce CD4-dependent signals stimulating NF-B translocation. To confirm that the lack of NF-B translocation in A2.01-3B can be ascribed neither to a defect in the ability of the A2.01-3B clones to translocate NF-B nor to a lack of gp120 binding to the E91K,E92K mutant CD4 molecule, two sets of experiments were performed. First, the ability of A2.01-3B cells to translocate NF-B after stimulation with PMA was tested using EMSA. As shown in Fig. 6, PMA induced a shift of labeled oligonucleotide migration when mixed with nuclear extracts from A3.01 and A2.01-3B cells (lanes 4 and 8, respectively). In addition, NF-B translocation was found with extracts from A3.01 cells exposed to iHIV (lane 2) but was not found with extracts from A3.01 cells exposed to iHIV previously incubated with 10 g/ml sCD4 for 2 h at 37°C (lane 3), indicating that induction of NF-B translocation following A3.01 cell exposure to iHIV required interaction between HIV-1 and cell surface CD4. Next, we compared the ability of the A3.01, A2.01, and A2.01-3B clones to bind recombinant gp120 envelope glycoprotein (gp120) from HIV-1 Lai . As  -24), and A2.01-c26 (lanes 25-28) was performed using the p56-I/ p56-II oligonucleotide primer pair. The amplified products were electrophoresed, blotted, hybridized with an ␣-32 P-labeled lck probe, and visualized by autoradiography. An autoradiogram of PCR amplification of reverse transcribed-thymidine kinase (TK) RNA is shown as control. Control PCR amplification using the p56-I/p56-II oligonucleotide primer pair was performed in MT2 cells (that lack expression of lck mRNA) and CEM cells (that express the lck mRNA) (lanes 29 and 30 respectively). Lane 31 corresponds to an RNA-free sample that was prepared for PCR and treated like the extracted samples. shown in Fig. 7, gp120 binds to A2.01-3B clone to the same extent as to the A3.01 clone. The weak binding of gp120 to A2.01 cells that lack CD4 expression suggests that a fraction of recombinant gp120 binds to target cells independently of CD4 recognition and should be therefore considered as background.
We have previously reported that gp120-anti-gp120 immune complex binding to wild type CD4 expressed at the surface of peripheral blood mononuclear cells induced NF-B transloca-tion similar to that induced by iHIV-1 (16). Since gp120 was found to bind A2.01-3B (see Fig. 7), we assessed whether gp120-anti-gp120 immune complexes were able to stimulate NF-B translocation in these cells. As shown in Fig. 8, we observed a shift of labeled oligonucleotide migration when mixed with nuclear extracts from A3.01 and A2.01-3B cells exposed to PMA (lanes 5 and 10, respectively). NF-B translocation was also observed with extracts from A3.01 cells exposed to iHIV (lane 2) and gp120-anti-gp120 immune complexes (lane  lanes 2-6), medium containing iHIV-1 previously incubated with 10 g/ml sCD4 for 2 h at 37°C (lanes 3 and 7), or PMA (lanes 4 and 8) were reacted with radiolabeled, double-stranded NF-B oligonucleotide. The samples were electrophoresed and analyzed by autoradiography. WT, wild type. 4) but was not found with extracts from A2.01-3B cells exposed either to iHIV (lane 7) or gp120-anti-gp120 immune complexes (lane 9).
These results indicate that the integrity of the CDR3-like loop Glu-91, Glu-92 residues is required to elicit NF-B activation after HIV-1-, and gp120-anti-gp120 immune complexes binding to the CDR2-like loop in D1 of cell surface CD4.

HIV-1 Replication Is Impaired in Cells Expressing a Mutant CD4 with Substitutions E91K and E92K in the CDR3-like
Loop-To assess whether the defect of A2.01-3B cells to translocate NF-B after HIV-1 binding may delay HIV-1 production, as demonstrated in cells expressing mutant CD4 lacking the cytoplasmic region (13), HIV-1 replication was measured. When cells were infected at low virus input (10 2 ϫ TCID 50 ), a low and delayed viral replication was consistently observed in A2.01-3B and A2.01/CD4.403 cells compared with A3.01 cells and with the other CD4 mutant expressing cells (A2.01-M1, A2.01-M2, and A2.01-2B) (Fig. 9A). To ensure that the impaired HIV-1 production found in A2.01-3B cells could not be ascribed to a defect in virus entry and retrotranscription in these cells, a PCR assay was performed on the different cells 24 h after virus exposure. As shown in Fig. 9B, HIV-1 DNA was found in A2.01-3B cells in similar amount as in the other cell lines. Moreover, a semi-quantitative reverse transcriptase-PCR performed on A2.01/CD4 and A2.01-3B cells 72 h after virus exposure indicates that HIV-1 mRNAs were less abundant in A2.01-3B cells than in cells expressing the wild type CD4 molecule (Fig. 9C).
These results indicate that substitution of positive for negatively charged residues at positions 91 and 92 within the CDR3-like loop impairs HIV-1 replication at the stage of early transcription.
Mutations in the CDR3-like Loop Affect the Binding of the CDR3.AME-(82-89) Analog to CD4 and Disrupt the Putative CD4 Dimerization Region-In an attempt to get further insights into the mode of action of the CDR3-like loop in signal transduction regulation, we used a CD4 exocyclic analog, named CDR3.AME-(82-89). This analog mimics the CDR3 loop structure and was previously shown to inhibit CD4-MHC class II binding and antagonize CD4 function through its binding to CD4 (25). The ability of FITC-labeled CDR3.AME-(82-89) analog to bind the wild type CD4 molecule and mutant forms of CD4 was investigated. As shown in Fig. 10, the FITC-conju-gated CDR3.AME-(82-89) analog bound to A2.01/CD4 cells, and its binding level was comparable to that of mAb OKT4. Conversely, no binding to A2.01-2B and A2.01-3B cells could be detected, although mutant CD4 molecules were strongly expressed at the surface of these cells, as evidenced by OKT4 mAb binding. Interestingly, the binding pattern of the CDR3.AME-(82-89) analog was identical to that observed with the ST40 mAb, specific for the CDR3-like region of CD4 domain 1 (see Fig. 2), suggesting that the CDR3.AME-(82-89) analog binds to this region. No binding to CD4-negative A2.01 cells was detected.
The possibility that the CDR3.AME-(82-89) analog binds to the CDR3 loop of CD4 was further examined by computerassisted molecular modeling studies of the interaction between this analog and the first 2 domains of CD4. As shown in Fig. 11, A and B, the CDR3.AME-(82-89) analog (red) was predicted to bind to the CDR3-like loop (yellow) of CD4 (white), suggesting that the CDR3 loop represents a main dimerization site for CD4 molecules. In the crystal structure of CD4 (11,12), the CDR3 loop is stabilized by a disulfide bond, salt links, and solvent molecules. Whereas the Glu-87/Asp-88 residues from CDR3-like loop occur at the periphery of the putative interface between the analog and CD4, residues Glu-91/Glu-92 occur at the core of the interface. Although the interactions cannot be docked accurately since these regions in the crystal structure contain water molecules, hydrophobic residues at the bottom of the peptide loop such as isoleucine at position 4 (Ile-4) and N-terminal residues would presumably interact with Ile-83 of CD4, if we assume that the role of the peptide is to displace water. Similarly, Tyr-12 at the top of the analog would interact with Asp-88 of CD4.
To highlight the role of Glu-87/Asp-88 and Glu-91/Glu-92 residues in CDR3.AME-(82-89) analog binding to CD4 and CD4 dimerization, mutant models of CD4 were constructed. Structural analysis reveals that in the wild type CD4 molecule (Fig. 11C, a), the conformation of the CDR3 loop is rigid and reinforced by a disulfide bond at the bottom of the loop and by several salt links. Fig. 11C shows the surface charge distribution in D1-D2 domains of CD4. Negatively charged residues (red) Glu-87/Asp-88 are located at the periphery of the putative dimeric interface (shown by circle) and are neutralized by neighboring positively charged lysine residues (Lys-90 at the interface, Lys-29, and Lys-35, below the CDR3 loop). Residues Glu-91/Glu-92 are located at the core of the putative dimeric interface and are neutralized by lysine residues from CD4 N terminus and a water molecule in the crystal structure. Mutation E87K/D88K changes the charge spectrum at the dimeric interface (Fig. 11C, b). However, the core of the interface, where Glu-91 and Glu-92 provide some stability by interacting with N-terminal lysine residues, does not show significant changes. Conversely, mutation E91K/E92K changes the charge spectrum of the interface completely (Fig. 11C, c). The interface totally becomes positively charged (blue). The mutation not only alters the electrostatic property of the dimeric interface, it potentially alters the conformation of the CDR3 loop. This change occurs through repulsion between Lys-91, Lys-92, and the neighboring N-terminal lysine residues.
These results indicate that the CDR3.AME-(82-89) analog most likely binds to cell surface CD4 CDR3-like loop and that this region is involved in CD4 dimer formation. Mutations changing both charge distribution at the interface and CDR3like loop conformation would prevent CD4 dimerization. DISCUSSION We demonstrate in this study that the CDR3-like loop of CD4 domain 1 plays an essential role in the signal transduction pathway that triggers activation of NF-B after HIV-1 binding to CD4. We show that E91K,E92K substitutions within the CDR3-like loop result in the inability of this mutant molecule to transduce signals triggering NF-B activation and also lead to impaired replication of HIV-1. Moreover these mutations suppress the binding of the CD4 CDR3.AME-(82-89) analog, a peptide derivative that mimics the CDR3-like loop structure and binds to native CD4, suggesting possible role played by this region in dimer formation.
In the past few years, several studies have reported evidence indicating that T cell activation signals are delivered to target cells by HIV-1 antigens (32)(33)(34)(35)(36). Most recently, we reported that binding of HIV-1 to the CDR2-like loop of CD4 domain 1 stimulates NF-B translocation (13,16) and that this event requires the integrity of the cytoplasmic domain of CD4 (13). It is worth noting that the interaction of Leu 3a mAb with the CDR2-like loop also triggers AP-1 translocation (36). Several reports have suggested that the "early" HIV-1 transcription events are regulated by NF-B protein, and the "late" transcription events are regulated by Tat and Sp-1 (37)(38)(39)(40). Based on these observations, we have proposed that HIV-1 binding to cell surface CD4 stimulates a signaling pathway(s) that results in nuclear translocation of NF-B and thereby regulates the early transcription of HIV-1 (13). In agreement with this hypothesis, we have found that viral production was delayed in cells expressing mutant forms of CD4 that lack the cytoplasmic tail, although the rate of viral entry and retrotranscription was apparently not different from that found in cells expressing the wild type CD4 (10 -13).
We have recently observed that binding of 13B8-2 mAb to the CDR3-like loop of CD4 domain 1 inhibits mitogen-activated protein kinase-2 activation (22) and NF-B translocation induced by HIV-1 binding to CD4 (23), indicating that the CDR3like loop may play an important role in T cell activation. To investigate this possibility, we studied the effects of mutations affecting four negatively charged residues within the CDR3like region on signal transduction pathway that triggers activation of NF-B after HIV-1 binding to CD4 and on viral production. The structural integrity of the CDR3-like loop mutant molecules was established by binding experiments using a panel of CD4 mAbs (see Fig. 2). Consistent with previous results reported by one of us who tested 9 anti-D1, 1 anti-D2, and 1 anti-D4 anti-CD4 mAbs for binding to the mutants CD4 (18), we found no evidence of global distortion of the CD4 conformation and no obvious effect on the structure of D1. Moreover, these mutations do not affect the binding of gp120 to the CDR2-like loop in D1 (see Fig. 7). Only a local effect in the CDR3-like loop was found for double substitution mutants E87K,D88K and E91K,E92K, as detected by the loss of ST40 mAb binding to A2.01-2B and A2.01-3B cells.
We observed that cells expressing the CD4 mutants exhibit different ability to translocate NF-B after iHIV binding to CD4. Cells (A2.01-3B) expressing the E91K,E92K mutant CD4 molecule were refractory to iHIV-induced activation and exhibited a significant reduction in virus production after exposure to low concentrations of HIV-1 Lai . According to the mAb binding results, these effects cannot be attributable to the density of CD4 molecules on the surface of the A2.01-3B cells nor to alterations of CD4 overall conformation although local conformational changes in the CDR3-like loop might play a role (see Fig. 2). They cannot be due either to an impaired accessibility of HIV-1 to CD4 as demonstrated by efficient binding of recombinant HIV-1 Lai gp120 to A2.01-3B cells (see Fig. 6) or to a defect in p56 lck expression (see Fig. 3), or to a defect of NF-B translocation, since NF-B translocation was observed under PMA stimulation (see Fig. 5). The reduction in virus production thus apparently correlates with a defect in the capacity of the CD4 molecule to transduce an activation signal resulting in NF-B translocation. The reduced reverse transcriptase activity in those cells was more pronounced when cells were exposed to low virus input (data not shown). Altogether these results indicate that negatively charged residues Glu-91 and Glu-92 in the CDR3-like loop play a role in activation signal  1 and 6), medium containing iHIV-1 (lanes 2 and 7), medium supplemented with gp120 (lanes 3 and 8), medium supplemented with gp120anti-gp120 immune complexes (lanes 4 and 9), or PMA (lanes 5 and 10) were reacted with radiolabeled, double-stranded NF-B oligonucleotide. The samples were electrophoresed and analyzed by autoradiography. WT, wild type.

transduction.
In contrast to A2.01-3B cells, A2.01-2B cells expressing CD4 mutant E87K,D88K seemed to be constitutively activated and were shown to express high amounts of nuclear NF-B in the absence of exogenous stimuli (Fig. 4). Although this result was unexpected, it was reproducible with two different clones of A2.01-2B cells. The fact that two clones demonstrated similar activation status suggests that the constitutive NF-B translocation is probably not related to a putative integration of the CD4 construct in the vicinity of a cellular gene controlling cell activation. Chronic activation of these clones was further established by flow cytometry experiments indicating that they express elevated surface CD25 antigen. 2 Interestingly, both mutants E91K,E92K and E87K,D88K exhibited a similar loss of ST40 mAb binding, indicating that these four negatively charged residues are involved in the ST40 epitope but play distinct functional roles in transduction of signals mediated by CD4.
Requirement for CD4 dimerization in transducing signals was suspected from the observation that cross-linking of CD4 can trigger autophosphorylation of p56 lck (27). We found here that mutants E91K,E92K and E87K,D88K were unable to bind the CD4 CDR3.AME-(82-89) analog that mimics the CDR3like loop and specifically binds to native CD4 (25). These results, together with molecular modeling studies, indicate that the CDR3.AME-(82-89) analog binds to the CDR3-like loop (see Fig. 11) and that residues 87/88 and 91/92 can potentially be involved in this binding. Moreover our results strongly suggest that the CDR3-like loop constitutes a primary site for CD4 dimerization. Our model shows that residues 89 -93 form the core of the dimeric interface and residues 87/88 are at the periphery. The involvement of the CDR3-like region in CD4 dimerization has been previously proposed (24) based on the demonstration that CDR3-like loop-derived synthetic peptides can bind recombinant CD4 molecules. Further evidence in favor of this dimerization site has recently been obtained by some of us (25); it was found that CDR3.AME-(82-89) analog binding to CD4 inhibits CD4-HLA class II interaction and antigeninduced T cell activation, an effect consistent with prevention of CD4 homodimer formation and signal transduction.
In this regard, the defect of NF-B translocation after iHIV binding to CD4 and the reduction in virus production generated by the E91K,E92K mutant might result from prevention of CD4 dimerization. Indeed, calculation for the E91K,E92K mutant drastically changes the conformation of the CDR3-like loop. So it seems that these mutations not only disrupt the dimerization, it can also disrupt the CDR3-like loop. These data clearly indicate that both CD4 dimerization and CDR3like loop structure are essential for efficient signaling and suggest that binding of multivalent molecules to the CDR2-like loop would induce CD4 dimerization at the CDR3-like loop. In light of this model, molecules that react with the CDR3-like region (20,(41)(42)(43) might exert their anti-HIV properties by uncoupling CD4 from the signal transduction machinery, thereby preventing cell activation. It is very likely that the CDR3.AME-(82-89) analog would act like a "spacer" between the CD4 dimers accounting for its inhibitory effects on T cell activation in MHC class II restricted immune response (25) and for its anti-HIV-1 properties (44). Other antiviral agents might exert their anti-HIV properties by inducing conformational changes in CD4 that block CD4 dimerization. For example, the 2 N. Signoret, unpublished observations.  3 and 10) and of RNA extracted from uninfected cells (lanes triphenylmethane polymer ATA that binds to part of the gp120 binding site induces conformational changes of the OKT4E epitope that is thought to arise from the tertiary structure of D1 through the juxtaposition of residues Gln-20, Lys-21, and Glu-91, via the S-S bond fixing the stem of the CDR3-like loop of D1 (45).
The opposite effects induced by the E87K,D88K mutation that seemed to generate a constitutive activation of A2.01-2B cells, which express high amounts of nuclear NF-B in absence of exogenous stimuli, are intriguing. Although residues Glu-87 and Asp-88 are at the periphery of the CD4 dimeric interface and exposed to solvents, their replacement by lysine does not show any significant changes in the CDR3-like loop. So it is possible that any instability caused by mutations can be compensated either by conformation changes or by solvents. Such subtle changes might account for the creation of new interaction sites facilitating mutant CD4 homodimer formation and consequently, constitutive cell activation, as detected in A2.01-2B cells. According to this hypothesis, the presence of CD4 dimers at the cell surface might prevent the CDR3.AME-(82-89) analog from binding to the E87K,D88K CD4 mutant. Alternatively, subtle changes in the charge balance might be responsible for its impaired binding. Further studies will be needed to determine whether E87K,D88K mutations favor NF-B translocation or whether the two clones demonstrate high nuclear amounts of NF-B for another reason which remains to be determined.
Although CD4 can likely forms dimers (46), it is not known whether a dimeric CD4 state exists at the surface of helper T cell only under certain conditions or whether there is a monomer/dimer equilibrium. Native CD4 may be normally monomeric and oligomerize by interacting with HLA class II or HIV-1. Sweet and co-workers (47) have reported crystallographic studies of sCD4 suggesting that it oligomerizes by interaction between its D3-D4 regions. Most recently, Sakihama and co-workers (48) have reached the same conclusions, based on the observation that replacement of D3-D4 domains by D1-D2 of CD4 or the extracellular domain of CD2 prevented HLA class II binding to these chimeric membrane CD4 molecules. It is, however, not known how the D1-D2 domains are associated with D3-D4 on the cell surface. It is possible that D1 domain makes a primary contact followed by D3-D4 domain in a dimer formation. Moreover, it has been demonstrated that the membrane-proximal CD4 domain appears to be important for the overall conformation of CD4 (5), and replacement of this domain might have an indirect rather than a direct role in CD4 dimer formation. Our data and these results do not rule out that several domains of CD4, including the D1, D3, and D4, participate in CD4 oligomerization.
In conclusion, our results provide the first functional evidence indicating that negatively charged residues within the CDR3-like loop can play a role in CD4-mediated activation signals transduced in T cells after virus envelope binding to the CDR2-like loop. Moreover, they strongly suggest that the CDR3-like loop represents a main site for CD4 dimerization.  11. Illustration of CDR3.AME-(82-89) analog binding to CD4 and surface charge distribution changes induced by mutations. A and B, the figure represents the atomic structure of D1-D2 wild type CD4 (shown in white) and CDR3.AME-(82-89) analog (shown in red). A plausible interaction of CDR3.AME-(82-89) analog (red) with CDR3-like loop (yellow) of CD4 is shown as bond (A) and spherical (B) models. Locations of important amino acids are indicated by arrows: residues Glu-87 and Glu-91 are from the CDR3-like loop, residue Phe-43 is from the CDR2-like region and is known to be a binding site for HIV gp120. C, surface charge distribution in D1-D2 domains of CD4 molecule is shown. The pictures were generated using GRASP. Negative charges are shown as red and positive charges as blue. The putative CD4 dimeric interface is shown by circles. a, Charge distribution in the wild-type CD4 molecule. Negatively charged residues Glu-87/Asp-88 and Glu-91/Glu-92 are located at the periphery and at the core of the putative dimeric interface, respectively. b, mutation E87K,D88K changes the charge spectrum at the dimeric interface, moderately, but does not alter the electrostatic property of the core of the interface. c, mutation E91K,E92K changes the charge spectrum of the interface completely. The interface totally becomes positively charged. The orientation of the 3 CD4 molecules (a-c) is slightly different to show the binding region clearly.