Kinetics of T-cell Receptor Binding by Bivalent HLA-DR·Peptide Complexes That Activate Antigen-specific Human T-cells*

Monovalent major histocompatibility complex-peptide complexes dissociate within seconds from the T-cell receptor (TCR), indicating that dimerization/multimerization may be important during early stages of T-cell activation. Soluble bivalent HLA-DR2·myelin basic protein (MBP) peptide complexes were expressed by replacing the F(ab) arms of an IgG2a antibody with HLA-DR2·MBP peptide complexes. The binding of bivalent HLA-DR2·peptide complexes to recombinant TCR was examined by surface plasmon resonance. The bivalent nature greatly enhanced TCR binding and slowed dissociation from the TCR, with a t 1 2 of 2.1 to 4.6 min. Soluble bivalent HLA-DR2·MBP peptide complexes activated antigen-specific T-cells in the absence of antigen presenting cells. In contrast, soluble antibodies to the TCR·CD3 complex were ineffective, indicating that they failed to induce an active TCR dimer. TCR/CD3 antibodies induced T-cell proliferation when bound by antigen presenting cells that expressed Fc receptors. In the presence of dendritic cells, bivalent HLA-DR2·MBP peptide complexes induced T-cell activation at >100-fold lower concentrations than TCR/CD3 antibodies and were also superior to peptide or antigen. These results demonstrate that bivalent HLA-DR·peptide complexes represent effective ligands for activation of the TCR. The data support a role for TCR dimerization in early TCR signaling and kinetic proofreading.

T-cell receptor (TCR) 1 recognition of MHC-bound peptides is important for the induction of antigen-specific immune responses. Surface expression of the TCR requires coordinate expression and assembly with four associated proteins termed CD3 ␥, ␦, ⑀, and (1,2). CD3 ⑀ associates with either ␥ or ␦, resulting in the formation of ␥⑀ and ␦⑀ heterodimers. It is well established that two CD3 ⑀-containing complexes are incorporated into a TCR⅐CD3 complex (3,4). Even though the CD3 ␥, ␦, and ⑀ chains are highly homologous, analysis of knockout mice has demonstrated that each subunit is important for development of ␣␤ T-cells (5)(6)(7).
The assembly of the TCR⅐CD3 complex follows discrete steps (1). An initial step is the association of the TCR ␣ chain with CD3 ␦⑀, which can then pair with a CD3 ␥⑀-associated TCR ␤ chain to form a disulfide-linked TCR ␣␤ heterodimer (8). Positively charged residues in the transmembrane regions of both TCR ␣ and ␤ are essential for assembly with these CD3 subunits. The TCR ␣ chain carries two positively charged amino acids (arginine and lysine), while the TCR ␤ chain transmembrane segment carries a single lysine residue. Site-directed mutagenesis of these residues abolishes assembly and transfer of an 8-amino acid transmembrane segment of TCR ␣ to an unrelated protein is sufficient for association with CD3 ␦⑀. Each CD3 subunit carries a negatively charged residue in the transmembrane region and site-directed mutagenesis experiments have demonstrated that these residues are required for TCR/CD3 assembly (9 -12). As the final step in the assembly of the TCR⅐CD3 complex, the CD3homodimer associates, allowing transport to the cell surface (13).
Each of the CD3 ␥, ␦, ⑀ chains carries a single immunoreceptor tyrosine-based activation motif. While TCR chain is essential for TCR surface expression, its large cytoplasmic domain, which carries three immunoreceptor tyrosine-based activation motifs, is dispensable for signaling (14). chain knockout mice are impaired in T-cell development, but T-cell development can be rescued by transgenic expression of a truncated chain (15). Therefore, the TCR⅐CD3 complex consists of two signaling units, of which only CD3 ␥⑀ and CD3 ␦⑀, but not CD3 -, are essential (14). T-cell activation results in the recruitment of CD4 and its associated p56 Lck kinase (16).
The binding of several TCRs to their MHC/peptide ligands has been examined by surface plasmon resonance. In all cases, the TCR dissociated rapidly from the MHC-peptide complex (17)(18)(19)(20). For example, the t1 ⁄2 of a TCR specific for I-E k and a moth cytochrome c peptide was ϳ12 s, while a t1 ⁄2 of 10.8 s was determined for a TCR specific for I-E k and a hemoglobin peptide (17)(18)(19). The MHC class I restricted 2C TCR had a slightly longer t1 ⁄2 of ϳ27 s (20). When partial agonist or antagonist peptides were examined for these TCRs, even shorter half-lives were observed (17)(18)(19). These data suggest that dimerization or a higher degree of multimerization is required for effective TCR signaling.
Interestingly, all TCR-associated components represent dimers (␥⑀, ␦⑀, -). In addition, the extracellular domain of CD4 crystallized as a dimer in which the membrane-proximal D4 domain constituted the dimerization interface. This dimer was observed in three different crystal forms, indicating that it may not represent a crystallization artifact (21). For these reasons, dimerization of the TCR represents an attractive hypothesis that would account for important biochemical and biological properties of the TCR⅐CD3 complex. However, soluble bivalent antibodies to the TCR or the CD3 complex do not induce activation in the absence of antigen presenting cells. Such antibodies only induce activation when they are further cross-linked by addition of a secondary antibody, by binding to a solid support or by binding to antigen presenting cells that express Fc receptors. In fact, soluble anti-CD3 antibodies were found to act as competitors for stimulation of T-cell clones by immobilized anti-CD3 (22,23). In order to address this question, we expressed an immunoglobulin fusion protein in which the antibody F(ab) arms were replaced by HLA-DR⅐peptide complexes. The biological properties of these bivalent HLA-DR⅐peptide complexes were distinct from those of TCR/CD3 antibodies. Soluble bivalent HLA-DR⅐peptide complexes induced T-cell activation, even in the absence of antigen presenting cells. In the presence of dendritic cells, bivalent HLA-DR⅐peptide complexes activated antigen-specific T-cells at lower concentrations than antigen, peptide, or TCR/CD3 antibodies. Bivalent HLA-DR⅐peptide complexes had a much slower dissociation rate than those reported for monovalent MHC-peptide complexes. The biochemical and biological properties of bivalent HLA-DR2⅐peptide complexes indicate that TCR dimerization is sufficient for triggering T-cell activation.

Expression and Purification of Bivalent HLA-DR2⅐MBP Peptide
Complexes-Drosophila Schneider cell transfectants were generated using the S2 cell line (24) and propagated in Schneider medium (Sigma) supplemented with 5-10% fetal calf serum, 100 IU/ml penicillin, and 100 g/ml streptomycin.
The Fc segment of murine IgG2a was attached to the previously described DR␣-Fos construct (25) by overlapping PCR. This construct represented the signal peptide and extracellular domain of DR␣, a 7-amino acid linker (Val-Asp-Gly-Gly-Gly-Gly-Gly), the Fos dimerization domain, a 3-amino acid linker (Ala-Ala-Ser), and the Fc segment of murine IgG2a spanning the hinge region, C H 2 and C H 3 domains (26). The DR␣-Fos construct was amplified by PCR with the following oligonucleotides: forward primer 5Ј-AATAATGAATTCATGGCCATAAGTG-GAGTC-3Ј and reverse primer 5Ј-CCTCTGGGCTCAGATGCTGCAT-GGGCGGCCAGGATGAACT-3Ј. The forward primer carried an EcoRI restriction site while the reverse primer created the overlap between Fos and the hinge region of IgG2a. The murine IgG2a segment was amplified by reverse transcriptase-PCR using RNA extracted from spleens of NOD mice using the following oligonucleotides: forward primer 5Ј-GCAGCATCTGAGCCCAGAGGGCCCACAATC-3Ј and reverse primer 5Ј-TGTAGAGAATTCTCATTTACCCGGAGTCCGGGAG-AA-3Ј. The reverse primer encoded the 3Ј end of the C H 3 domain and a stop codon and carried an EcoRI restriction site. The DR␣-Fos and Fc segments were joined by overlapping PCR and inserted into the EcoRI site of the pRmHa-3 vector under the control of the strongly inducible metallothionein promoter (27). The orientation and sequence of the construct was confirmed by DNA sequencing. The DR␤-Jun construct with the covalently linked MBP(85-99) peptide was described previously (25) and cloned into the EcoRI site of the pRmHa-3 vector.
The transfection procedure reported by Cherbas et al. (28) was used to generate clones from the Drosophila melanogaster S2 cell line that secreted soluble HLA-DR2 molecules. Briefly, 3 ϫ 10 6 cells were transfected with 2 g of each plasmid carrying DR␣ and DR␤ constructs as well as 0.2 g of vector pH8CO (29). This vector carried the gene for dihydrofolate reductase, which conferred resistance to methotrexate. After 2-3 weeks of culture in selecting medium containing methotrexate at a concentration of 200 nM (Sigma), transfected cell lines were cloned by limiting dilution and tested for expression. One of the positive clones was scaled up in roller bottles. Expression was induced by addition of copper sulfate (1 mM) and supernatants were harvested 4 -5 days following induction and concentrated ϳ10-fold by ultrafiltration.
DR2 molecules were purified from supernatants by affinity chromatography with the HLA-DR specific mAb L243 (ATCC number HB 55). The antibody was coupled to cyanogen bromide-activated Sepharose 4B beads (Amersham Pharmacia Biotech); glycine-blocked Sepharose 4B beads were used for a preclearing column. The samples were passed over the preclearing and L243 columns at a flow rate of 0.15 ml/min at 4°C. The L243 column was then washed with 30 ml of 20 mM sodium phosphate, pH 7.0, and the protein was eluted with 20 ml of 50 mM CAPS, pH 11.5. The eluted protein was collected in 1-ml fractions, which were neutralized by addition of 200 mM sodium phosphate, pH 6.2. Fractions that contained protein were pooled and further purified by HPLC using a POROS protein A column (PerSeptive Biosystems, Cambridge, MA) with a bed volume of 0.85 ml. Samples were loaded at a flow rate of 0.5 ml/min and the column was washed with phosphatebuffered saline, pH 7.6. Bound protein was eluted by injection of 5 ml of 50 mM CAPS, pH 11.5. The peak was collected by monitoring the absorbance at 280 nm and the collected material was immediately neutralized by addition of 200 mM sodium phosphate, pH 6.2. The protein was dialyzed against phosphate-buffered saline and the final protein concentration was determined by BCA assay (Pierce Chemical Co., Rockford, IL). The final yield was ϳ200 g/liter of culture.
Expression and Purification of a Soluble MBP-specific TCR-A recombinant, soluble TCR specific for the HLA-DR2⅐MBP peptide complex was expressed in the baculovirus system using TCR cDNAs isolated from T-cell clone Ob.1A12 (30). The TCR ␣ and ␤ ectodomains were expressed with C-terminal Fos and Jun dimerization domains, in order to promote the assembly of the TCR heterodimer. The TCR chains were cloned into the pAcAB3 baculovirus transfer vector, which carries two p10 promoters in opposite orientation. This allowed expression of soluble TCR with a single recombinant baculovirus. The TCR ␣ segment that represented the signal peptide and the ectodomain as well as part of the linker was amplified with the following oligonucleotides: forward 5Ј-AAAAAATCTAGATCTATGGAAACTCTCCTGGGAGT-3Ј and reverse 5Ј-GTGGATCCGCGCGGAACCAGGTCTGCTGACGAACAGGA-ACTTTCTGGGCTGG-3Ј. The segment representing part of the linker and the Fos dimerization domain was amplified with the following oligonucleotides: forward 5Ј-CTGGTTCCGCGCGGATCCACTACAGC-TCCATCATTAACTGATACACTCCAAGC-3Ј and reverse 5Ј-AAAAAAT-CTAGATCTTCAATGGGCGGCCAGGATGA-3Ј. The overlapping PCR was done with the forward primer for TCR␣ and the reverse primer for the Fos segment. The TCR␣-Fos chain was digested with BglII and cloned into the BamHI site of pAcAB3.
The TCR ␤-Jun construct was made in the following steps. A free cysteine residue in C␤ was mutated to serine by overlapping PCR of the 3Ј segment of C␤ and a segment representing the V␤ and the 5Ј end of C␤. The V␤-C␤ segment was amplified with the following oligonucleotides: forward 5Ј-AAAAAATCTAGATGATCAATGCTGCTGCTTCTGC-TGCT-3Ј and reverse 5Ј-GCTGCTCAGGCTGTATCTGGA-3Ј. The 3Ј segment of C␤ was amplified with the following oligonucleotides: forward 5Ј-TCCAGATACAGCCTGAGCAGC-3Ј and reverse 5Ј-AAAAA-AGGATCCGCGCGGAACCAGGTCTGCTGACGAACAGTCTGCTCTA-CCCCAGG-3Ј. The overlapping PCR was done with the forward primer for V␤ and the reverse primer for the 3Ј segment of C␤. The Jun segment was amplified with the following oligonucleotides: forward 5Ј-AAAAAAGGATCCACTACAGCTCCATCACGCATCGCCCGGCTCG-AGGA-3Ј and reverse 5Ј-AAAAAATCTAGATGATCATCAATGGTTCA-TGACTTTCT-3Ј. The TCR V␤-C␤ segment and the Jun segment were digested with BamHI, ligated, and amplified by PCR with the forward primer for V␤ and the reverse primer for Jun. Also, a free cysteine residue at position 13 was mutated to serine by an additional PCR step. The TCR␤-Jun chain was digested with BclI and cloned into the BglII site of the pAcAB3 vector that already carried the TCR␣-Fos construct. The sequence and orientation of both chains was confirmed by dideoxy sequencing. A recombinant baculovirus was generated using a genetically modified baculovirus (BaculoGold, Pharmingen, San Diego, CA). TCR expression was examined by Western blot analysis with mAbs specific for TCR␣ and TCR␤ (antibodies ␣F1 and ␤F1, Endogen, Woburn, MA).
TCR was expressed on a large scale in High Five cells infected with the recombinant baculovirus and purified from concentrated supernatants using the W4F5.B mAb (ATCC HB-9282) coupled to cyanogen bromide-activated Sepharose 4B beads. Supernatants were passed over a preclearing column and the W4F5.B column at a flow rate of Ͻ30 ml/h. The column was washed with 40 ml of 20 mM sodium phosphate, pH 7.0, followed by 40 ml of 20 mM sodium phosphate, 150 mM NaCl, and 40 ml of 20 mM sodium phosphate, pH 7.0, 500 mM NaCl. Prior to elution, the column was also washed with 20 ml of 50 mM CAPS, pH 10.0. These wash steps were necessary to reduce nonspecifically bound proteins. The protein was eluted with 20 ml of 50 mM CAPS, pH 11.5, and fractions were immediately neutralized by addition of 1 M sodium phosphate, pH 6.0. Fractions that contained protein were pooled and dialyzed against 20 mM Tris, pH 8.0. The protein was further purified by anion exchange HPLC using a POROS PI column (PerSeptive Biosystems) with a bed volume of 1.7 ml. The column was equilibrated with 50 mM Tris, pH 8.0, 0.05% sodium azide and the protein was eluted with a linear gradient of 0 -100% 1 M NaCl at a flow rate of 3 ml/min. The TCR peak was concentrated and the buffer was exchanged to 25 mM Tris, pH 7.5, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 1 M leupeptin using a Centricon Plus-20 ultrafiltration unit (Amicon, Beverly, MA).
BIACORE Analysis of Binding of Bivalent HLA-DR2⅐MBP Peptide Complexes to Recombinant TCR-The binding of DR2/MBP-IgG to MBP-specific TCR was measured by surface plasmon resonance using a BIACORE 1000 (Biacore AB, Uppsala, Sweden). HLA-DR2/MBP-specific Ob.1A12 TCR and HLA-DR1/HA-specific Y22 TCR (negative control) were immobilized using EDC/NHS coupling chemistry in 50 mM sodium acetate, pH 5.5, and 50 mM sodium acetate, pH 5.0, respectively. For those experiments comparing binding to both TCRs, Ob.1A12 immobilization was lowered to match the maximum immobilization achievable with the Y22 TCR. Binding of DR2/MBP-IgG was measured at pH 7.4 using HBS as running buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% Polysorbate 20, filtered and degassed) with a flow rate of 5 l/min. Multimerization of DR2/MBP-IgG with recombinant protein A (Pierce) was achieved by incubation for 30 min at 37°C. The molar ratio of DR2/MBP-IgG to protein A in these experiments was 2:1. As a result of the relatively slow dissociation rate of the multimerized complex, an additional experiment was performed at a flow rate of 2 l/min to further examine the association and dissociation phases of binding.
Fluorescence-activated Cell Sorter Analysis of T Cells Clones with Bivalent HLA-DR2⅐MBP Peptide Complexes-T-cells (2.5 ϫ 10 5 per tube) were incubated for 1 h at 4°C with DR2/MBP-IgG at a concentration of 50 g/ml in 100 l of phosphate-buffered saline, 3% fetal calf serum, 0.1% sodium azide. Cells were then washed and incubated for 30 min with an Alexa 488-conjugated anti-mouse IgG antibody (Molecular Probes, Eugene, OR). Cells were again washed and fixed in 1% formaldehyde, phosphate-buffered saline. Mouse IgG2a (Pharmingen, San Diego, CA) was used as a negative control. Cells were also stained with an anti-CD3 antibody that was labeled with fluorescein isothiocyanate (Caltag, Burlingame, CA). This antibody was used at a concentration of 1 g/ml. Samples were analyzed in a fluorescent-activated cell sorter. The instrument was a EPICS XL, Coulter Corp., Miami, FL, which analyzes and displays the distribution of cell types (31).
Direct fluorescent labeling of HLA-DR2/MBP was achieved through partial oxidation of carbohydrate moieties, followed by conjugation to fluorescein hydrazide. The glycoprotein (ϳ3.5 M) was dialyzed into 100 mM sodium acetate, pH 5.5, then reacted with 10 mM sodium m-periodate for 30 min at room temperature in the dark. Excess periodate was subsequently removed in a second dialysis step. Fluorescein hydrazide (Molecular Probes) was dissolved in dimethyl sulfoxide and added to the oxidized glycoprotein at a final concentration of 200 M. The mixture was reacted for 2-3 h at room temperature in the dark. Insoluble material was removed by centrifugation and the sample was injected onto a POROS-protein A column (PE biosystems, Framingham, MA) using phosphate-buffered saline as a running buffer. The labeled glycoprotein was eluted with 50 mM CAPS, pH 11.5. The sample was neutralized to ϳpH 7-8 by addition of 1 M sodium phosphate, pH 6.0. The protein concentration was determined by Coomassie Protein Assay (Pierce). T-cells were incubated with these molecules (25 g/ml) for 1 h at 4°C. Cells were then washed and fixed as described above. Mouse IgG2a labeled with the same procedure was used as a negative control.
For T-cell stimulation assays, bivalent HLA-DR2⅐peptide complexes or antibodies (200 ng/well) were bound to a 96-well flat bottom plate (Maxisorb, Nunc) by overnight incubation at 4°C in 50 l of 100 mM bicarbonate buffer, pH 9.6. A mouse IgG2a antibody was used as a negative control while an anti-CD3 antibody served as a positive control. Prior to addition of T-cells wells were washed twice with sterile phosphate-buffered saline. Assays were set up in triplicates with 10 5 T-cells per well in RPMI supplemented with 10% human serum, 100 IU/ml penicillin, 100 g/ml streptomycin, 10 mM HEPES, and 2 mM glutamine. After 48 h of culture, 3 H-labeled thymidine was added (1 Ci/well) and cells were harvested after 16 -18 h onto glass fiber filters (Wallac, Gaithersburg, MD). Incorporated radioactivity was quantitated in a ␤-scintillation counter (Wallac).

Activation of T-cells with Soluble
Molecules-T-cell proliferation assays with soluble HLA-DR2⅐peptide complexes or soluble antibodies were set up in triplicates in 96-well U bottom plates with 10 5 T-cells/ well. T-cell proliferation was determined after 48 h of culture by [ 3 H]thymidine incorporation as described above. The following TCR and CD3 antibodies were tested in these experiments: anti-CD3 antibodies OKT3 (34), UCHT1 and HIT3a (Pharmingen, San Diego, CA), X35 and Cris-7 (Biodesign, Kennebunk, Maine) as well as TCR V␤2.1 antibody MPB2D5 (Biodesign). A mouse IgG2a antibody specific for TNP (Pharmingen) was used as a negative control since it carried the same Fc segment as the bivalent HLA-DR2⅐MBP peptide complex.
Recombinant MBP was expressed in E. coli with a N-terminal (His) 6tag and purified by metal-chelate and ion-exchange chromatography. Dendritic cells were obtained by stimulating the adherent cell population of blood mononuclear cells from a HLA-DR2 ϩ normal donor for 4 days with granulocyte macrophage-colony stimulating factor and IL-4 (35). Dendritic cells were then washed and irradiated with 3000 rad. Experiments were set up in triplicates with 3.5 ϫ 10 4 dendritic cells and

Expression and Characterization of Bivalent HLA-DR2⅐
Peptide Complexes-We previously reported that soluble HLA-DR2⅐peptide complexes can be expressed by replacing the hydrophobic transmembrane segments as well as cytoplasmic domains of DR␣ and DR␤ with leucine zipper dimerization domains from the transcription factors Fos and Jun (36). The sequence of the MBP(85-99) peptide that is recognized by T-cell clones specific for the DR2⅐MBP peptide complex was covalently linked to the N terminus of the mature DR␤ chain (25), as reported for murine MHC class II molecules (37). In order to generate bivalent, antibody-like molecules, the Fc segment of murine IgG2a (hinge, C H 2 and C H 3 domains) was attached in-frame to the 3Ј end of the DR␣ chain construct (Fig.  1). The IgG2a sequence was chosen since it binds with high affinity to protein A, which has four IgG-binding sites (38). This strategy therefore allowed expression of soluble, bivalent molecules that could be further multimerized with protein A.
These constructs were cloned into the pRmHa-3 vector under the control of the copper-inducible metallothionein promoter and the plasmids were transfected into Drosophila Schneider cells. Transfectants were cloned by limiting dilution and grown on a large scale in roller bottles. The protein was purified from concentrated supernatants by affinity chromatography using mAb L243, which is specific for the assembled HLA-DR heterodimer. Following elution and neutralization, the molecules were further purified by HPLC using a protein A column. SDS-PAGE under reducing conditions demonstrated two bands corresponding to DR␣-IgG and DR␤/MBP. The identity of these bands was confirmed by Western blot analysis using polyclonal antibodies specific for the Fos and Jun dimerization domains (Fig. 2, lanes 2 and 3) and HLA-DR (Fig. 2, lane 4).
The biological activity of the DR2/MBP-IgG protein was examined in a T-cell proliferation assay using T-cell clones specific for the DR2⅐MBP peptide complex as well as control clones with other MHC/peptide specificities (Table I). For these experiments, the DR2/MBP-IgG protein, a mouse IgG2a antibody (negative control) or an anti-CD3 antibody (positive control) were immobilized in a 96-well plate. T-cell activation was assessed by [ 3 H]thymidine incorporation following 48 h of culture. The DR2/MBP-IgG protein activated T-cells specific for the DR2⅐MBP peptide complex (clones Ob.1A12 and Ob.2F3), but did not activate four control clones specific for other MHC/ peptide combinations. All of the T-cell clones were activated by immobilized anti-CD3 that was used as a positive control. These results indicated that the DR2/MBP⅐IgG complex was properly folded and that it was recognized by TCRs specific for the DR2⅐MBP peptide complex.
Binding Kinetics of Multivalent HLA-DR2⅐Peptide Complexes to Purified TCR-The binding of DR2/MBP⅐IgG to the T-cell receptor was directly examined by surface plasmon resonance (BIACORE). For that purpose, a TCR specific for the DR2⅐MBP complex (derived from clone Ob.1A12) was expressed as a soluble protein in the baculovirus system (Fig. 3). The purified recombinant TCR was immobilized on a BIACORE chip by standard EDC/NHS chemistry and soluble DR2/MBP-IgG was run over this surface at a flow rate of 5 l/min. These experiments demonstrated specific, dose-dependent binding, with soluble DR2/MBP-IgG at concentrations of 0.3, 0.6, 1.2, and 2.4 M (Fig. 4A). Several experiments were performed to determine the specificity of this interaction (Fig. 4B). First, DR2/MBP-IgG was run over an unmodified dextran surface. Second, a control surface was created with a recombinant TCR specific for the influenza hemagglutinin(306 -318) peptide bound to HLA-DR1 (kindly provided by J. Hennecke and D. Wiley). Third, binding was assessed for mouse IgG2a, which carries the same Fc segment as the DR2/MBP-IgG protein. All of these controls confirmed that binding of the DR2/MBP-IgG protein to immobilized TCR was specific. We also attempted to examine the binding of monovalent DR2/MBP (expressed without the Fc segment of IgG2a) in this assay. Only a weak signal that was similar to a control protein was observed (data not shown). These results demonstrate that the bivalent nature greatly increased the binding of the DR2/MBP-IgG protein to an immobilized TCR.
The contribution of the valency of binding to the dissociation rate was further assessed by making complexes of DR2/MBP-IgG with recombinant protein A. Protein A has four IgG-binding sites and binds IgG with a 2:1 stoichiometry (IgG-protein A) (38). As shown in Fig. 4C, multimerization with protein A greatly enhanced binding of DR2/MBP-IgG to the immobilized TCR and further slowed dissociation from the TCR. The following complexes were examined: 1) DR2/MBP⅐IgG multimerized with protein A; 2) DR2/MBP⅐IgG without protein A; and 3) mouse IgG2a multimerized with protein A. Binding of DR2/ MBP⅐IgG and the DR2/MBP⅐IgG⅐protein A complex was specific since little binding was observed with the complex of mouse IgG2a and protein A.
Based on these experiments, the half-life (t1 ⁄2 ) was calculated for the dissociation of DR2/MBP⅐IgG as well as the complex of DR2/MBP⅐IgG and protein A from the TCR. To a certain extent,  (36). The Fc segment of murine IgG2a was fused in-frame to the 3Ј end of the DR␣-Fos segment. The Fc segment spanned the hinge region as well as C H 2 and C H 3 domains of mouse IgG2a. The hinge region was included in this construct in order to provide mobility of the two DR/peptide arms relative to the Fc segment. The sequence of the MBP(85-99) peptide was covalently linked to the N terminus of the mature DR␤ chain through a 16-amino acid linker, as described previously (25). This design allowed the expression of soluble, antibody-like molecules in which the antibody F(ab) segments were replaced by HLA-DR2⅐MBP peptide complexes.

TABLE I
Activation of T-cell clones specific for the HLA-DR2⅐MBP peptide complex by immobilized molecules T cell activation in response to immobilized molecules was tested in order to determine if bivalent HLA-DR2 ⅐ peptide complexes were properly folded. DR2/MBP-IgG, mouse IgG2a (negative control) and an anti-CD3 mAb (mAb HIT3a, positive control) were immobilized in a 96-well plate by overnight incubation of 200 ng/well in 100 mm bicarbonate, pH 9.6. Wells were washed with phosphate-buffered saline and T-cells (10 5 per well, triplicates) were added. T-cell proliferation was  (11)(12)(13)(14)(15)(16)(17)(18)(19) bound to HLA-A2 (clone 2G4) (30,32,33). the calculated t1 ⁄2 was dependent on the concentration of DR2/ MBP⅐IgG that was injected over the TCR surface and ranged from 2.1 to 4.6 min for concentrations of 0.3 to 2.4 M, respectively (Fig. 4A). The t1 ⁄2 for the bivalent DR2/MBP⅐IgG complex was 2.9 min, compared with 43.1 min for the complex with protein A when both were injected at a concentration of 1.9 M and a flow rate of 5 l/min (Fig. 3C). At a slower flow rate (2 l/min instead of 5 l/min) and a slightly higher level of TCR immobilization, the t1 ⁄2 for the complex of DR2/MBP⅐IgG and protein A was even longer (Ͼ200 min) (Fig. 4D). At this flow rate, there may have been a certain degree of rebinding, even though the association rate was relatively low for the DR2/ MBP⅐IgG⅐protein A complex.

Staining of Antigen-specific Human T Cells by Bivalent HLA-DR2⅐Peptide
Complexes-Binding to the TCR on the surface of antigen-specific human T-cell clones was examined using two different approaches. As a first approach, T-cells were incubated with unlabeled molecules, followed by incubation with an anti-mouse IgG antibody conjugated to Alexa 488 (Fig. 5). As a second approach, bivalent HLA-DR2⅐peptide complexes were fluorescently labeled on carbohydrate groups following oxidation of carbohydrates with sodium periodate (data not shown).  (lanes 1 and 2) and TCR ␤ (lanes 3 and 4). Purified TCR was separated by SDS-PAGE, transferred to a polyvinylidene difluoride membrane and probed with mAbs ␣F1 and ␤F1, which are specific for TCR ␣ and ␤, respectively. Bound antibodies were detected by enhanced chemiluminescence using an horseradish peroxidase-conjugated antimouse IgG antibody. The band representing the TCR ␣␤ heterodimer was stained with both TCR ␣ and ␤ chain-specific antibodies under nonreducing conditions (lanes 2 and 4).  staining was observed with two control clones (Hy.1B11 and Go.P3.1, C and D). In both sets of experiments, mouse IgG2a was used as a control. The t1 ⁄2 of 2.1-4.6 min that was measured for bivalent HLA-DR2⅐peptide complexes by BIACORE explains why the T-cell staining was relatively weak compared with staining by the anti-CD3 antibody.
We also attempted staining with molecules that were multimerized with labeled protein A, since the BIACORE experiments had indicated that multimerization delayed dissociation from the TCR. However, the labeling reduced the IgG binding capacity of protein A and prevented effective multimerization. The complexes with labeled protein A were directly tested for TCR binding by BIACORE. The delayed dissociation that characterized the binding of unlabeled protein A complexes was not observed with labeled protein A. The labeling with fluorescein isothiocyanate probably modified a lysine residue that is important for IgG binding and thereby reduced the IgG binding capacity of protein A.

Activation of MBP-specific T-cells by Soluble, Bivalent HLA-DR2⅐MBP Peptide Complexes but Not by Soluble TCR or CD3
Antibodies-MHC class II-peptide complexes may be useful for the induction of antigen-specific T-cell responses. The ability of soluble DR2/MBP IgG molecules to activate antigen-specific T-cells was therefore compared with antibodies against the TCR or the TCR-associated CD3 complex (Fig. 7). These experiments were performed with T-cell clones in the absence of antigen presenting cells, since binding to Fc receptors can create a multivalent surface. The two T-cell clones specific for the DR2/MBP⅐peptide complexes were activated by soluble DR2/MBP⅐IgG, but not by a soluble TCR antibody (directed against the V␤2.1 segment expressed by these clones) or soluble antibodies directed against the CD3 complex. Six TCR/CD3 antibodies were tested, including a TCR V␤2.1 antibody and five different antibodies against the CD3 complex. None of these antibodies activated the two HLA-DR2/MBP-specific Tcell clones when used in soluble form. However, all of these antibodies were biologically active since they activated the T-cell clones when immobilized on a solid support. These experiments were performed in media supplemented with 10% human serum. In serum-free media, soluble TCR/CD3 antibodies induced a moderate degree of T-cell stimulation, presumably because a fraction of the antibody could attach to the plate. The fact that the soluble TCR/CD3 antibodies were ineffective indicated that cross-linking of the TCR⅐CD3 complex by soluble, bivalent antibodies was not sufficient for activation of T-cells specific for the DR2/MBP peptide complex.
T-cell activation induced by bivalent HLA-DR2⅐peptide complexes was enhanced by cross-linking of CD28 or by addition of recombinant IL-2 (Fig. 8). In contrast, anti-CD3 antibodies did not activate these T-cell clones, even when costimulation was provided with the anti-CD28 antibody (data not shown). This indicated that the failure of anti-CD3 antibodies to induce T-cell activation was not due to a lack of costimulation.
One of the five T-cell clones that was examined in this set of experiments showed a moderate degree of T-cell proliferation to three of the anti-CD3 antibodies (mAbs HIT3a, Cris-7, and X35). This T-cell clone was specific for a desmoglein 3 peptide (residues 190 -204) bound to HLA-DR4 (DRA, DRB1*0402). The level of [ 3 H]thymidine incorporation that was induced by these soluble anti-CD3 antibodies was ϳ10 -20% relative to the stimulation observed with immobilized antibodies (data not shown).
The stimulatory capacity of bivalent DR2/MBP⅐IgG peptide complexes was also compared with the complex of DR2/ MBP⅐IgG with protein A (Fig. 9). Similar dose-response curves were observed, indicating that further multimerization did not enhance the ability of the bivalent molecules to induce T-cell activation. These results demonstrate binding of soluble, bivalent MHC-peptide complexes to the TCR is sufficient to induce T-cell activation.
Effective T-cell Stimulation by Bivalent HLA-DR2⅐Peptide Complexes in the Presence of Dendritic Cells-Numerous studies have demonstrated that soluble anti-CD3 antibodies are strong mitogens for T-cells in the presence of antigen presenting cells (22,34). In such experiments, anti-CD3 antibodies are multimerized by binding to Fc receptors on the surface of antigen presenting cells. Dendritic cells are potent antigen presenting cells and express type I and type II Fc␥R (CD64 and CD32); ligation of Fc␥R has been shown to induce maturation of dendritic cells (39,40). The ability to induce T-cell activation was therefore compared for DR2/MBP-IgG, recombinant MBP, MBP(85-99) peptide as well as a panel of anti-CD3/TCR antibodies.
Dendritic cells were differentiated from blood mononuclear cells of a normal HLA-DR2 ϩ donor by culture in granulocyte macrophage-colony stimulating factor and IL-4 and used as antigen presenting cells (34). Recombinant MBP and MBP(85-99) peptide induced strong T-cell stimulation at concentrations of 1 to 10 nM, while much higher concentrations were required for anti-CD3 antibodies. In contrast, picomolar concentrations of DR2/MBP-IgG induced T-cell activation (Fig. 10). Similar results were obtained with a second MBP-specific T-cell clone (clone Ob.1A12, data not shown). Three anti-CD3 and a TCR V␤2.1 antibody were tested in these experiments. The anti-CD3 antibody shown in Fig. 10 (mAb UCHT1) induced the strongest stimulation of the TCR/CD3 antibodies that were tested (mAbs UCHT1, OKT3, X35, and MPB2D5). When dendritic cells were fixed with 1% formaldehyde, a similar degree of T-cell stimulation was induced by DR2/MBP-IgG molecules (data not shown), indicating that T-cell activation was not due to uptake and processing of the MBP peptide bound to DR2/ MBP-IgG molecules. These results demonstrate that bivalent MHC-peptide complexes are superior to antigen, peptide, or TCR/CD3 antibodies in activating antigen-specific T-cells. DISCUSSION Soluble, bivalent HLA-DR2⅐peptide complexes were found to effectively activate antigen-specific human T-cells. In contrast, soluble bivalent TCR or CD3 antibodies were ineffective, indicating that cross-linking of the TCR by a soluble, bivalent antibody is not sufficient for activation. In the presence of dendritic cells that express Fc receptors, activation was induced both by bivalent HLA-DR⅐peptide complexes and TCR/ CD3 antibodies. A striking difference in biological activity was observed since Ͼ100-fold lower concentrations of bivalent HLA-DR⅐peptide complexes than of TCR/CD3 antibodies were required for T-cell activation. Staining of human T-cell clones was much weaker with bivalent HLA-DR2⅐peptide complexes than with anti-CD3 antibodies, probably because the antibodies had a higher affinity for the TCR⅐CD3 complex. Nevertheless,  bivalent HLA-DR2⅐peptide complexes were more potent in activating antigen-specific T-cells than TCR/CD3 antibodies. These results demonstrate that binding of bivalent HLA-DR2⅐peptide complexes to the TCR is sufficient to induce T-cell activation.
Three major models of TCR activation, namely aggregation, multimerization, and dimerization, have been proposed (reviewed in Refs. 23, 41, and 42). The aggregation model is based on a large number of studies which demonstrated that soluble TCR or CD3 antibodies do not induce T-cell activation, unless further cross-linking is performed (reviewed in Ref. 23). Higher order cross-linking could be induced by an additional antibody against a different TCR/CD3 epitope, a secondary antibody, immobilization of TCR/CD3 antibodies on a solid support or by binding of antibodies to cells expressing Fc receptors. The comparison of soluble, bivalent HLA-DR2⅐peptide complexes and of a large panel of TCR/CD3 antibodies indicates that soluble antibodies fail to induce a biologically active TCR dimer that is induced by bivalent HLA-DR⅐peptide complexes.
The multimerization model proposes that an active TCR ligand has a valency of greater than two (reviewed in Ref. 41). This model is based on T-cell activation experiments with biotinylated MHC-peptide complexes that were multimerized with streptavidin. In these experiments, tetrameric MHC-peptide complexes had a higher activity than dimeric or trimeric complexes. The streptavidin scaffold may provide limited mobility for bound MHC-peptide complexes and may therefore require a higher number of MHC-peptide complexes than the immunoglobulin scaffold used in this study. An important feature of the immunoglobulin scaffold is the hinge region which allows free mobility of the two F(ab) arms relative to the Fc segment.
The dimerization model proposes that TCR dimerization is required to initiate signaling (reviewed in Ref. 42). A key question is whether such a TCR dimer is formed during the encounter of appropriate MHC-peptide complexes or whether an inactive dimer is pre-assembled during biosynthesis. It is well established that two CD3 ⑀ subunits are present in the TCR complex and that CD3 ⑀ forms CD3 ␥⑀ and CD3 ␦⑀ heterodimers (3,4). The CD3 ␥⑀ and CD3 ␦⑀ heterodimers may pair with two TCR heterodimers and form a (TCR␣␤) 2 ⅐CD3 complex in which the CD3 component consists of CD3 ␥⑀, CD3 ␦⑀, and CD3 -. This model is consistent with the requirement for CD3 ␥, ␦, and ⑀ for TCR surface expression, a large body of biochemical data on the pairing of TCR chains with CD3 subunits and functional studies on the requirements for TCR signaling (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(42)(43)(44)(45). Dimeric murine MHC class II-peptide complexes also activate antigen-specific T-cells, indicating that both human and murine T-cells are activated by dimerization of the TCR (44,45). In addition, a soluble bivalent TCR was shown to specifically bind to cells that displayed the appropriate MHC class I-peptide complex (46).
Crystallographic and functional studies of the erythropoetin receptor demonstrated a preassembled receptor dimer in which the individual membrane-spanning and intracellular domains were too far apart to permit signaling by the receptor-associated JAK2 kinases. Ligand binding induced a major conformational change of the extracellular domain that reduced the distance between the two transmembrane segments from ϳ73 to 39 Å, allowing the associated JAK2 kinases to come into contact and autophosphorylate (47,48). The crystal structure of the extracellular domain of tumor necrosis factor-R1 also demonstrated a preassembled dimer in the absence of tumor necrosis factor (49). In addition, fluorescence imaging studies have indicated that the IL-2 receptor and the epidermal growth receptor may occur as preassembled receptors on the cell surface. The formation of preassembled, inactive receptors is therefore observed in a number of transmembrane receptors (50,51).
The model of a preassembled (TCR␣␤) 2 ⅐CD3 complex would account for the observation that soluble, bivalent TCR/CD3 antibodies are ineffective in inducing signaling. According to this model, antibodies could bind to different TCR complexes but fail to induce an active configuration of a pre-assembled (TCR␣␤) 2 ⅐CD3 complex. Conversely, bivalent HLA-DR⅐peptide complexes could bind to such a pre-assembled (TCR␣␤) 2 ⅐CD3 complex and induce an active configuration. A preassembled (TCR␣␤) 2 ⅐CD3 complex would also explain why T-cells are very sensitive to low densities of the appropriate MHC-peptide complex on the antigen presenting cell, even though the affinity of a monovalent TCR for a MHC-peptide complex is very low (17,20).
Initial triggering of TCRs results in transport of TCR-MHC⅐peptide complexes to a small junction between the antigen presenting cell and the T-cell, which has been termed "immunological synapse" (52,53). Within 5 to 30 min of T-cell activation, TCRs are concentrated in a small cluster that is surrounded by a ring of ICAM1. Interestingly, ICAM1 is excluded from the central cluster. An open question regarding the formation of such synapses is the precise nature of the substrate that triggers transport. Since monovalent TCR-MHC⅐peptide complexes have a very short half-life (17)(18)(19)(20), it is difficult to envision how a monovalent TCR could drag a MHCpeptide complex into the cluster. The dissociation rate observed in the BIACORE experiments for the bivalent HLA-DR⅐peptide complex indicates that a (TCR-MHC⅐peptide) 2 complex would be more suitable for transport over such distances (t1 ⁄2 of bivalent HLA-D2⅐peptide complexes of 2.1 to 4.6 min). The formation of immunological synapses as the final product of TCR activation results in a stable, highly multivalent TCR-MHC/ peptide interface and sustained signaling (52,53). Therefore, the dimerization and multimerization models may describe distinct phases in T-cell activation, namely the initial TCR triggering and transport of MHC/peptide⅐TCR complexes into such clusters and the later phase of sustained TCR signaling.
The effective activation of T-cells by bivalent HLA-DR⅐peptide complexes may also be of practical value since such molecules are more effective than peptide, antigen, or TCR antibodies. Such complexes may therefore be useful for immunization against defined viral proteins or tumor antigens. Of particular importance may be their ability to bind to Fc receptor expressing antigen presenting cells, such as dendritic cells.