Characterization of a single-chain antibody to the beta-chain of the T cell receptor.

In this report the VH and VL genes of the anti-T cell receptor (TCR) antibody KJ16, which recognizes the TCR Vβ8.1 and Vβ8.2 regions in mice, were cloned and expressed as a single-chain antibody (scFv) in Escherichia coli. A 29-kDa protein was obtained after renaturation from inclusion bodies. The KJ16 scFv had a relative affinity for the native TCR that was slightly higher than KJ16 Fab fragments. The scFv and Fab fragments of the KJ16 antibody, together with monovalent forms of two other anti-TCR antibodies, were evaluated as antagonists of the T cell-mediated recognition of a peptide-class I complex or of a superantigen, Staphylococcus enterotoxin B (SEB) bound to a class II product. Each of the anti-TCR antibodies was efficient at inhibiting the recognition of the SEB-class II complex. In contrast, only the clonotypic antibody, which binds to epitopes on both the Vβ and Vα regions, inhibited the recognition of peptide-class I complex. We conclude that the TCR binding site for the SEB-class II ligand encompasses a larger surface area than the TCR binding site for the peptide-class I ligand.

In this report the V H and V L genes of the anti-T cell receptor (TCR) antibody KJ16, which recognizes the TCR V␤8.1 and V␤8.2 regions in mice, were cloned and expressed as a single-chain antibody (scFv) in Escherichia coli. A 29-kDa protein was obtained after renaturation from inclusion bodies. The KJ16 scFv had a relative affinity for the native TCR that was slightly higher than KJ16 Fab fragments. The scFv and Fab fragments of the KJ16 antibody, together with monovalent forms of two other anti-TCR antibodies, were evaluated as antagonists of the T cell-mediated recognition of a peptideclass I complex or of a superantigen, Staphylococcus enterotoxin B (SEB) bound to a class II product. Each of the anti-TCR antibodies was efficient at inhibiting the recognition of the SEB-class II complex. In contrast, only the clonotypic antibody, which binds to epitopes on both the V␤ and V␣ regions, inhibited the recognition of peptide-class I complex. We conclude that the TCR binding site for the SEB-class II ligand encompasses a larger surface area than the TCR binding site for the peptideclass I ligand.
Specific recognition of foreign antigens by T cells is mediated through a T cell receptor (TCR) 1 complex composed of the ␣␤ heterodimer and CD3 subunits. Under normal conditions, T cells that mature in the thymus undergo a process of selection in which self-reactive cells are eliminated (1). Despite this process, a number of diseases appear to be the result of aberrant T cell activity. These include autoimmune diseases in which the ␣␤ receptor presumably recognizes a self-antigen (2), transplantation reactions that involve alloreactive T cells (3), and "superantigen"-mediated diseases caused by a hyperactive immune response when the ␤-chain of the TCR, expressed by a large fraction of T cells, binds to the superantigen (4).
Because the TCR is the common element in all these diseases, there has been considerable interest in using the TCR as a specific target to eliminate or inhibit detrimental T cells (5). Among the molecules targeted have been CD3 (6,7) and regions of the ␣␤ heterodimer itself (8 -12). In some cases, treatment with anti-TCR antibodies in vivo has led to specific elimination of target T cell populations in animal models of human diseases. For example, most T cells involved in experimental autoimmune encephalomyelitis in mice bear the V␤8 domain, and treatment with the V␤8-specific monoclonal antibody F23.1 (8) prevented or reversed the disease (9). Another V␤8specific antibody, KJ16 (10) was used to control graft versus host disease in SCID mice injected with alloantigenic spleen cells (11) and to control the pathogenesis of collagen-induced arthritis (12). Each of these diseases is suspected to be caused by T cells that expressed primarily V␤8 regions.
In addition to their uses in the control of diseases, monoclonal anti-TCR antibodies have been used on a limited basis to map the topology of TCR binding sites (13)(14)(15). This interest stems from the fact that resolution of the TCR interactions with either peptide-MHC or superantigen-MHC at the molecular level remains to be determined. Nevertheless, several studies have used TCR mutagenesis to begin to define the regions involved in ligand binding (16 -20). For instance, Hong et al. (20) used transfectants of chimeric T cell receptors to show that the CDR1 and/or CDR2 regions of the TCR ␣ chain are important for MHC specificity and to help define the orientation of the TCR relative to its MHC ligand. Other analyses that mutagenize CDR regions have examined their role in both peptide-MHC specificity and superantigen reactivity (17).
The size of the TCR ␣␤-heterodimer is approximately that of an antibody Fab fragment, and thus topology mapping (as well as therapeutic efficacy) might be improved by reducing the size of the anti-TCR antibody to the minimal possible unit. We have previously expressed an anti-clonotypic antibody, 1B2, as a single-chain Fv (scFv; Ref. 21), and in this report we characterize the binding properties of the scFv from the anti-V␤8 antibody KJ16 (10). KJ16 scFv was expressed as a single-chain in which the V L was joined to the V H by a 26-amino acid linker (205s) (22). KJ16 scFv preparations bound to the native TCR with a relative affinity that was approximately 1.5-fold higher than Fab fragments. The biphasic nature of the binding isotherm suggested that the scFv may exist in monomer/dimer equilibrium; as with other scFv such as 1B2, the higher affinity component of the observed binding may be due to the dimeric form (21,23).
The three anti-TCR antibodies KJ16, 1B2, and F23.1 were also used to probe the topology of the TCR binding sites for either peptide-class I MHC or staphylococcal enterotoxin B (SEB)-class II MHC. KJ16 and F23.1 bind to different determinants on the V␤8 domain of the TCR (24), and the clonotypic * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. This work was supported by National Institutes of Health Grant AI35990.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  1B2 binds to determinants on both the V␤8 and V␣3 regions of the cytotoxic T cell clone 2C (21,(25)(26)(27)(28). All three of the antibodies were able to inhibit SEB-class II recognition, but only the clonotypic antibody inhibited recognition of peptide-class I. Thus, all three of these antibody epitopes reside within the superantigen binding site of the TCR. This finding is consistent with recent reports that the superantigen binding site on the TCR may extend to the V␣ domain (29 -31) and that the area on the TCR required to bind the superantigen-class II complex is larger than the area required for normal peptide-MHC recognition. Finally, because intact KJ16 has been shown to be effective in vivo in several diseases mediated by V␤8 ϩ T cells, comparisons among scFv, Fab fragments, and intact antibody for their potential to target T cell populations in vivo can now be undertaken.

EXPERIMENTAL PROCEDURES
Cell Lines-CTL clone 2C, originally derived from a BALB.B anti-BALB/c mixed lymphocyte culture (32), was maintained in RPMI 1640 containing 5 mM HEPES, 10% fetal calf serum, L-glutamine, penicillin, streptomycin, and 2-mercaptoethanol (RPMI media). 2C cells were maintained in 5-10% supernatant from concanavalin A-stimulated rat spleen cells and stimulated approximately every other week with mitomycin-C-treated BALB/c mouse spleen cells. Clone 2C expresses a V␤8V␣3 T cell receptor and recognizes the self peptide, QL9, in association with the mouse class I MHC product L d (33). Since 2C expresses a V␤8 domain, it also recognizes SEB bound to a class II MHC product (34). Daudi, a human lymphoma that expresses class II, was maintained in RPMI media. T2/L d , a human mutant B lymphoblastoidderived line transfected with L d (35) was maintained in RPMI media, 0.5 mg/ml Geneticin. KJ16 is a rat monoclonal antibody specific for the V␤8.1-2 domains of the TCR (10) and was provided by Drs. Kappler and Marrack. 1B2 is a clonotypic mouse monoclonal antibody (IgG1) specific for the 2C TCR (25). F23.1 is a mouse monoclonal IgG2a antibody specific for V␤8.1-3 and was provided by Dr. Bevan (8). CT14-G4.3 (called CT14 here) is a mouse antibody (IgG1) specific for the 10-amino acid peptide (EQKLISEEDL) from the c-Myc protein (36). KJ16, 1B2, F23.1, and CT14 hybridomas were maintained in RPMI media.
Preparation of Intact Antibodies and Fab Fragments-Intact KJ16 was prepared from tissue culture supernatant generated in a Vita-Fiber mini flow path bioreactor (Amicon). Antibody was precipitated twice in 50% ammonium sulfate, and the precipitate was resuspended in and dialyzed against 10 mM phosphate buffer, 150 mM NaCl, pH 7.3 (phosphate-buffered saline, PBS). To prepare Fab fragments, intact antibody was dialyzed against 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, and Fab fragments were generated by digestion of the antibody with papain (Sigma) for 10 min at 37°C. Fab fragments were purified from residual intact antibody by size exclusion HPLC over a Superdex G-200 column followed by protein G affinity chromatography to remove residual Fc regions, and the purified sample was dialyzed against PBS. 1B2 antibody was obtained from either ascites fluid as described (21) or from tissue culture supernatant by purification on a protein A column, and 1B2 Fab fragments were prepared as described (21). F23.1 antibody was purified from tissue culture supernatant using protein A purification. Generation and purification of F23.1 Fab fragments was as described above for KJ16 except that the antibody was digested for 5 min with papain, and protein A affinity chromatography was used to remove Fc fragments.
Construction and Expression of KJ16 scFv-Total RNA was isolated by the guanidinium thiocyanate-CsCl method. cDNA was generated by using approximately 5 g of the RNA with 35 units of reverse transcriptase (Promega) and 5 M of oligo(dT) primer. Antibody V L and V H genes were cloned using the polymerase chain reaction and degenerate primers for rat antibodies: The V L and V H genes were joined by a 26-amino acid linker, 205s, which differs from the 205 linker in that the carboxyl-terminal glycine has been replaced by alanine-serine in order to introduce a unique restriction site (NsiI) at this position. The construction included the ompA signal sequence and a 10-residue carboxyl-terminal tag derived from the c-Myc protein. The scFv construct was expressed using a hybrid O L /P R phage promoter in an Escherichia coli strain that contains a temperature-sensitive C I857 repressor gene. Induction was performed as described previously (28). Briefly, after growth in a 30-liter fermenter to an A 600 of 1.0 at 30°C, scFv expression was induced by shifting the temperature to 42°C for 1 h. Cells were pelleted, and the pellet was resuspended in TKC buffer (50 mM Tris-HCl, pH 8.0, 100 mM KCl, 10 mM CaCl 2 , 1 mM EDTA), and passed five times through a microfluidizer. Unlysed cells were removed by differential centrifugation, and insoluble inclusion body pellets were obtained by centrifugation at 21,000 ϫ g. Inclusion bodies were washed in TKC containing 0.5% Triton X-100 and then solubilized into TKC containing 6 M guanidine HCl.
Purification of scFv Preparations-KJ16 was refolded by slowly diluting the guanidine HCl-solubilized inclusion body pellets 100-fold into TKC containing 0.5 mM phenylmethylsulfonyl fluoride and 1 g/ml leupeptin. The refold mixture was allowed to stir slowly for 24 -48 h at 6 -8°C before being concentrated by filtration (A 280 Ϸ 3-5). This sample was injected onto a Superdex G-200 column size exclusion column pre-equilibrated in phosphate-buffered saline, pH 7.3. The column was run at 0.5 ml/min, and the eluent was monitored by its absorbance at 280 nm. Fractions were collected and used in an ELISA to test for binding to a TCR V␤ domain (described below).
ELISA-Two different ELISAs were used. To detect the presence of the carboxyl-terminal c-Myc tag, partially purified scFv preparations were adsorbed directly to the wells of an ELISA plate and detected using anti-c-Myc antibody as described below.
To detect direct binding of KJ16 scFv to the V␤8 region, partially purified preparations of a single-chain TCR (20 g/ml) were adsorbed to the wells of a 96-well plate (Immulon 2) overnight at 4°C. The singlechain TCR contained the variable regions of the ␣ and ␤ chains encoded by CTL clone 2C (V␣3V␤8) linked by a short peptide (28). Wells were washed and blocked using phosphate-buffered saline, pH 7.3, containing 0.1% bovine serum albumin and 0.1% Tween-20 (PBSBT) for 1 h at room temperature. Serial dilutions (in PBS) of KJ16 scFv were added for at least 1 h, wells were washed with PBSBT, and bound scFv was detected with the mouse anti-c-Myc antibody CT14. CT14 was detected with goat anti-mouse antibodies conjugated to horseradish peroxidase followed by the addition of substrate (tetramethylbenzidine; Kirkegaard and Perry Laboratories) followed by acid. Absorbances at 450 nm were quantitated using a 96-well plate reader.
Western Blot-Samples were electrophoresed through a 10% polyacrylamide gel under reducing conditions and electroblotted onto a nitrocellulose transfer membrane (Nitro ME; Micron Separations, Inc.). After blocking the membrane with 2% non-fat dry milk in 10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.05% Tween 20, protein was detected with the anti-c-Myc antibody, CT14 (used as undiluted culture supernatant). Bound CT14 was detected with goat anti-mouse antibodies conjugated to horseradish peroxidase, a chemiluminescent substrate was added (ECL; Amersham Corp.), and the blot was exposed to film.
Mass Spectrometry-Mass spectra were obtained on a TofSpec using electrospray ionization. Samples were dialyzed against 1 mM potassium phosphate buffer, pH 8.0, and concentrated to 10 -25 pmol/ml. Analysis was performed by the Mass Spectrometry Laboratory, School of Chemical Science, University of Illinois.
Flow Cytometry-For direct binding of KJ16 scFv to cell surface TCR, 3 ϫ 10 5 2C cells were incubated with a crude mixture of refolded scFv diluted 2-fold in RPMI media. After 30 min on ice, the cells were washed twice with RPMI media and incubated for 30 min with CT14 (used as undiluted culture supernatant). Binding was detected by the addition of a 1:20 dilution of 5-aminofluorescein isothiocyanate (FITC)-conjugated goat anti-mouse antibodies and flow cytometric analysis was performed with a Coulter Electronics EPICS 752 at the University of Illinois Biotechnology Center.
To determine the relative affinity of KJ16 scFv for the TCR on CTL 2C, a competition assay was performed with FITC-labeled KJ16 Fab fragments. KJ16 FITC/Fab conjugates were generated as described (38). In brief, an aliquot of 0.9 mg/ml FITC in PBS was added to 0.5 mg/ml KJ16 Fab fragments in 0.2 M sodium bicarbonate, pH 8.5, in order to yield a 50:1 (FITC:antibody) molar ratio. After incubation at room temperature for 2 h, the solution was dialyzed into PBS, pH 7.3. The number of FITC bound per antibody was calculated as follows: 3.08 (39). Various concentrations of scFv KJ16 were added to 6 ϫ 10 5 2C cells in the presence of a constant amount of FITC-labeled Fab. After a 30-min incubation on ice, the entire mixture (scFv ϩ KJ16 FITC/Fab ϩ 2C cells) was passed through a flow cytometer without washing. The amount of inhibition was measured by quantitating the fluorescence from cells by flow cytometry. The concentrations of unlabeled antibody giving 50% inhibition (IC 50 ) relative to the maximum fluorescence (in the absence of inhibitor) and the background fluorescence (in the presence of a large excess of intact antibody) were used to compare affinities. Inhibition curves were generated for KJ16 scFv and unlabeled Fab preparations in two independent experiments. Each concentration was performed in triplicate.
Cytotoxicity Assays-Target cells were incubated with 50 -100 l of 51 Cr (2.5 mCi/ml) for 1 h at 37°C. After washing, the Daudi cells or T2/L d cells were resuspended in RPMI media containing either SEB(10 g/ml) or peptide QL9 (1 nM), respectively. QL9 (QLSPFPFDL) was synthesized and HPLC purified by the Genetic Engineering Facility at the University of Illinois. Serial dilutions of antibody preparations were added to 2C cells (10 5 cells/well) and incubated for 30 min at room temperature before adding 51 Cr-labeled target cells (2 ϫ 10 4 cells/well). All assays were performed in triplicate in 96-well plates at an effector: target ratio of 5:1. After the addition of target cells, plates were incubated for 4 h at 37°C in 5% CO 2 . Supernatants were monitored with a ␥ counter, and specific cytotoxicity was determined: % specific release ϭ (experimental counts Ϫ spontaneous counts)/(maximal counts Ϫ spontaneous counts) ϫ 100.

RESULTS
Cloning and Construction of KJ16 scFv-Polymerase chain reaction products of the expected size (ϳ350 -400 base pairs) were obtained from amplification of KJ16 hybridoma cDNA with V H and V L primers (data not shown). The V L gene was cloned behind the ompA signal sequence, and the V H gene was cloned in front of a sequence that encodes a 10-amino acid carboxyl-terminal tag derived from the c-Myc protein (Fig. 1A). The c-Myc tag allows a convenient means of detecting KJ16 scFv protein with the anti-c-Myc antibody CT14 (36) (see "Experimental Methods"). Unique restriction enzyme sites flank each antibody V region in order to facilitate cloning of other scFv genes. The V L is flanked by 5Ј AatII and 3Ј SacII sites, and the V H is flanked by 5Ј NsiI and 3Ј HindIII sites. The V L and V H of KJ16 are joined by a modified 205 linker that encodes 26 amino acid residues (22). The nucleotide and predicted amino acid sequences of KJ16 scFv are shown in Fig. 1B. The calculated molecular mass of the mature protein is 28,932. Expression and Purification of KJ16 scFv-The ompA signal sequence was intended to facilitate folding and purification by targeting the protein to the periplasmic space. However, most scFv genes that include this signal sequence have yielded insoluble single-chain proteins that appear to form intracellular inclusion bodies (21,22,40,41). Initial experiments with KJ16 scFv demonstrated that the protein was also expressed in E. coli inclusion body pellets (data not shown). Small scale cultures were used to establish that solubilized inclusion body pellets were detected by the anti-c-Myc antibody in an ELISA, and a single clone was selected for large scale fermentation and characterization.
Although binding of anti-c-Myc antibody demonstrated that the recombinant protein was fully translated, it was unknown whether the protein was functional (i.e. could bind the TCR ligand). To test this, an ELISA was used to determine if the scFv could bind to a recombinant V␤8 domain absorbed to microtiter wells. The V␤8.2 domain was part of a single-chain TCR (scTCR) that could be readily detected with intact KJ16 antibody, as described previously by Soo Hoo et al. (28). The scTCR was adsorbed to wells, incubated with various dilutions of an unpurified preparation of refolded KJ16 scFv, and the bound scFv was detected with anti-c-Myc antibody (Fig. 2). KJ16 scFv bound specifically to the scTCR, as compared with bovine serum albumin used as a negative control. This assay could be used to detect activity at dilutions of the unpurified preparation as high as 1:1000 and thus was used to routinely monitor HPLC purifications of the KJ16 scFv.
KJ16 scFv was purified by HPLC chromatography through a G-200 column under nondenaturing conditions. The elution profile yielded three major peaks: one at the column void volume, a second at approximately 30 kDa, and a third at the total column volume (Fig. 3A). A V␤8-specific ELISA showed that the 30-kDa peak contained the majority of the activity (data not shown). Reduced SDS-PAGE gels (Fig. 3B) and Western blot analysis with the anti-c-Myc antibody (Fig. 3C) were used to FIG. 3. HPLC gel filtration profile, SDS-PAGE analysis, and Western blot of refolded KJ16 scFv. A, KJ16 scFv was refolded by dilution into TKC buffer and concentrated; a sample was injected onto a Superdex G-200 column equilibrated in phosphate-buffered saline, pH 7.3; and the eluent was monitored for absorbance at 280 nm. Fractions were collected and analyzed by SDS-PAGE. B, SDS-PAGE of peak fractions obtained by size exclusion HPLC. Samples were electrophoresed through a 10% polyacrylamide gel under reducing conditions, and proteins were visualized by staining with Coomassie Blue. Pre-HPLC sample represents TKC-refolded preparation before HPLC purification. Peak 1, which elutes in the column void volume, is likely covalent and noncovalent aggregates of scFv protein. Peak 2, which elutes at the expected size of an scFv monomer, is composed of two monomeric forms of KJ16 scFv. Fraction 2a contains a 28-kDa species, which could represent a proteolysed form of scFv (referred to as KJ16 scFvЈ), fraction 2c contains a 29-kDa species, which is consistent with the calculated mass of the KJ16 scFv monomer (28,932), assuming a cleaved signal sequence. The last peak to elute is at the total column volume and likely consists of proteolysed material and protease inhibitors from the refolding buffer. C, an SDS-PAGE gel similar to that shown in B was blotted onto nitrocellulose, scFv was detected with the anti-c-Myc antibody CT14, and bound CT14 was detected with goat anti-mouse antibodies conjugated to horseradish peroxidase. Both the 28-and 29-kDa species shown in fractions 2a, 2b, and 2c react with anti-c-Myc antibody. Since the 10-amino acid c-Myc tag is encoded at the carboxyl terminus, the reactivity of the 28-kDa species with the anti-c-Myc antibody CT14 suggests that the truncated scFvЈ was cleaved at its NH 2 terminus.
FIG. 2. ELISA to demonstrate KJ16 scFv binds specifically to a soluble V␤8 bearing scTCR. Serial dilutions of crude KJ16 scFv were added to wells adsorbed with scTCR (f-f) or bovine serum albumin (BSA) (q-q) at 20 g/ml for 1 h. After washing with PBS/bovine serum albumin, bound scFv was detected by adding anti-c-Myc antibody and incubating for 1 h. The plate was washed, and a secondary antibody (goat anti-mouse horseradish peroxidase-labeled antibody) was added. After 1 h and washing, substrate was added, and the absorbance was monitored at 450 nm. further characterize each peak. A major overexpressed protein of 35 kDa was present in the refolded preparations. This protein was positive by Western blotting with the anti-c-Myc antibody, confirming that it represented the recombinant KJ16 scFv. Since the first peak runs at the exclusion volume and contains a significant amount of anti-c-Myc reactive material, it likely contains covalent or noncovalent aggregates of the scFv protein. The second peak eluted at the expected size of a scFv monomer, and it contained an ϳ35-kDa species as detected by SDS-PAGE and Western blots (Fig. 3, B and C). The third peak eluted at the total column volume, and it did not contain detectable protein by SDS-PAGE and Western blot analysis; it likely consists of proteolysed material and protease inhibitors from the refolding buffer.
Closer inspection of the fractions that were contained within the second peak revealed minor size heterogeneity among the fractions (Fig. 3, B and C); all three different KJ16 scFv preparations that have been examined yielded identical results. SDS-PAGE analysis of reduced fractions from peak 2 showed that it contained two proteins with similar molecular mass. Mass spectrometry analysis indicated that the major protein in fraction 2a had a mass of 28,262 and that of fraction 2c had a mass of 29,030. Although the two proteins eluted as a single peak, the smaller 28-kDa species eluted before the larger 29-kDa species. The apparent discrepancy in the order of elution could be due to differences in the shape of the two forms, but we have not explored this possibility further. The 29,030 mass of the later species is consistent with the predicted molecular mass of the KJ16 scFv monomer (28,932), assuming a cleaved signal sequence. Thus, this species will herein be referred to as KJ16 scFv. Since the amount of the 28-kDa species increased over the lifetime of stored inclusion bodies (data not shown), the 28-kDa species could represent a proteolysed form of KJ16 scFv. This species will be referred to as KJ16 scFvЈ. The observed reactivity in a Western blot with the anti-c-Myc antibody (Fig. 3C) may suggest that the COOH terminus is intact and that the NH 2 terminus has been cleaved. If this is the case, then the scFvЈ form may not fold properly; this possibility is consistent with a shape that differs from the scFv and thus with its increased size exclusion. Without further optimization of expression, the estimated yield of HPLC-purified KJ16 scFv from a 1-liter equivalent of fermentation is approximately 300 g.
Binding Properties of KJ16 scFv-The ability of an unpurified preparation of KJ16 scFv to bind the native TCR on the surface of the V␤8 ϩ cytotoxic T cell clone 2C was tested by flow cytometry. Direct binding of the scFv to CTL 2C was detected using mouse anti-c-Myc antibody followed by fluoresceinlabeled goat anti-mouse IgG (Fig. 4A). Since scFv specifically binds the 2C TCR, we could compare the affinity of purified KJ16 scFv and scFvЈ for ligand by determining the amount of scFv required to inhibit the binding of fluorescein-labeled KJ16 Fab fragments (KJ16 FITC/Fab) to 2C cells. Approximately 3 fluorescein molecules per KJ16 Fab fragment were attached by coupling with the isothiocyanate derivative (FITC). Unlabeled scFv, scFvЈ, and Fab fragments were used as competitive inhibitors of the binding of the FITC/Fab to 2C cells. Experiments were performed at a constant but nonsaturating concentration of the FITC/Fab, and cells were examined by flow cytometry without washing to ensure that equilibrium was maintained. Previous studies using iodinated KJ16 Fab fragments have shown that KJ16 Fab fragments have a K D of 130 nM (13). 2 Analysis of the binding curves (Fig. 4B) indicated that the KJ16 scFv preparation bound at least as well as Fab fragments. The truncated scFvЈ preparation was capable of binding to the TCR, but it required greater than 10-fold more protein to yield the same inhibition as Fab fragments. This finding could be due to either a reduced affinity of the scFvЈ (compared with scFv) or to the presence of contaminating active scFv in the scFvЈ preparation.
Analysis of the binding curves suggests that the molar concentration of scFv required to inhibit 50% of the FITC/Fab from binding the 2C TCR was ϳ1.5-fold less than that of unlabeled Fab fragments, indicating that the scFv had a higher binding affinity. However, the binding curve of the scFv appeared to be contain two components: one that resembled that of the Fab fragments, and a second component that appeared to be derived from a higher affinity species. Although we have not resolved a dimeric form by size filtration, it is possible that the higher affinity component represents a dimer that exists in equilibrium with monomer. Such dimers can be observed, especially after concentration, in other scFv antibodies such as 1B2 (21,23). Nevertheless, the results show that the recombinant scFv can fold to an active form that binds the cell surface TCR at least as well as Fab fragments.
Inhibition of Ligand Binding by KJ16 scFv and Other Anti-TCR Antibodies-In principle, the scFv should act as an antagonist of T cell activity, as with intact antibody and Fab fragments. Its ability to inhibit T cell recognition could depend on several factors including its binding affinity, the location of its epitope on the TCR relative to the ligand binding site, and its size. We were in a position to compare the role of each with CTL 2C using KJ16 and two other anti-TCR antibodies, F23.1 and 1B2, that each bind to different ␤ chain epitopes. In addition, the TCR on CTL 2C recognizes two distinct ligands, the nonapeptide QL9 associated with the class I MHC product L d (33) and SEB associated with a class II product (34). Thus, we could begin to map the topology of these ligand binding sites using the different anti-TCR antibody fragments.
To examine recognition of SEB-class II ligand, the human class II ϩ cell line Daudi was used together with SEB as a target in a cytotoxicity assay with CTL clone 2C. Recognition of SEBclass II was inhibited by all forms of the KJ16, 1B2, and F23.1 antibodies (Fig. 5A). The inhibitory potential of each antibody directly correlated with their binding affinity (Table I). However, the KJ16 scFv was approximately 5-fold less effective than KJ16 Fab fragments, a finding that would not be expected from the nearly equal binding affinities of these preparations. There are at least two possible explanations for this finding. First, the scFv is less effective because it is smaller and does not sterically prevent TCR binding as well as Fab fragments. Alternatively, the scFv is less stable than Fab fragments during the course of the 4-h cytotoxicity assay at 37°C. The latter possibility may be supported by our previous finding that 1B2 scFv also appeared to be approximately 5-fold less effective at inhibiting cytolysis than would be expected from its affinity. These possibilities are currently being examined as they have direct relevance to the use of these agents in vivo.
In contrast to results with the ligand SEB-class II, inhibition of peptide-class I recognition was observed only with fragments derived from the clonotypic antibody 1B2 (Fig. 5B and Table I).
Because 1B2 binds to determinants on both the V␣ and V␤ chains, this result was not unexpected since recognition of peptide-MHC is predicted to involve both V␣ and V␤ regions of the TCR. On the other hand, recognition of peptide-MHC is not predicted to involve the region on the face of the ␤-chain that contains the KJ16 and F23.1 epitopes (42). DISCUSSION There have been many reports that use antibodies to map the topology of proteins. Most of these studies have used intact antibodies to examine surface-accessible epitopes on the protein antigen (e.g. 13,24,34). In principle, the smaller size of a scFv (ϳ30 kDa) compared with Fab fragments (ϳ60 kDa) and intact antibodies (150 kDa) may provide better resolution in topology mapping of proteins. As it becomes more routine to produce and purify scFvs, these recombinant antibodies can also be more readily prepared than Fab fragments for studies that examine protein structure. In this report, we have produced and characterized a scFv that recognizes an epitope on the V␤8 chain of the TCR. This epitope includes residue 16 in  Fab (ϫ)). KJ16 and F23.1 are antibodies that recognize different epitopes on V␤8, and 1B2 is clonotypic for the 2C TCR. A, inhibition of lysis of the MHC class II target cell Daudi pulsed with 10 g/ml SEB. B, inhibition of lysis of the MHC class I targets T2/L d -pulsed with 1 nM QL9 peptide. 2C cells were incubated with antibodies for 30 min at room temperature before the addition of 51 Crlabeled target cells at an effector:target ratio of 5:1. Percentage of inhibition was calculated by normalizing values to the amount of specific 51 Cr release that occurred in the absence of inhibiting antibody (21.5 Ϯ 3.4% for SEB-Daudi and 19.8 Ϯ 1.1% for QL9/T2/L d ). All titrations were performed in triplicate, and a separate inhibition experiment yielded similar results.  Figure 4B).
d Two independent preparations of KJ16 scFv monomer. e In cases where 50% inhibition was not achieved, the concentration that yielded the maximum inhibition is indicated. the framework region of the V␤8.2 domain (43), a residue that has recently been shown to reside on the outer face on the crystal of a V␤8.2 dimer (44). Another antibody, F23.1, has been shown to interact with residue 60 of the V␤8.2 chain (45). This residue is located closer to the CDRs of the ␤-chain than the KJ16 epitope, but both of these antibodies and their monovalent fragments are capable of cross-inhibition (24). 3 Antibody 1B2 is a clonotypic antibody as defined by the fact that it only reacts with the TCR of CTL 2C and not other T cells. This antibody has been shown to bind determinants on both the V␣ and V␤ regions of the TCR from 2C (21,(25)(26)(27)(28). 1B2 binding is not blocked by binding of F23.1 and KJ16, 3 further evidence that this epitope is located distal to the V␤ framework regions recognized by the latter antibodies.
The role of antibody affinity, epitope, and size in the ability to inhibit TCR recognition of two different ligands were examined. One ligand was the classical peptide-MHC complex that is recognized by T cells, and the other ligand was a complex of the superantigen SEB and a class II product. Numerous studies have shown that the V␤ chain is primarily involved in the recognition of the latter complex (reviewed in Ref. 4), while both the ␣ and ␤ chains are involved in peptide-MHC recognition (reviewed in Ref. 42). Our results showed that all three antibodies were capable of blocking the recognition of the SEBclass II ligand. The finding extended to both scFv antibodies KJ16 and 1B2, suggesting that the scFv approach should be useful in examining the topology of protein binding sites. The observation that the V␤8-specific antibody fragments inhibit SEB-class II recognition was not surprising given that this region of the TCR has been predicted to be involved in superantigen recognition. However, the even greater inhibition with 1B2 antibody fragments suggested that the TCR binding site for SEB-class II extends over an area that includes the V␣ region. Recently, a completely different approach was used to also implicate the ␣ chain in this recognition (29,30). Our results are consistent with this notion and the possibility that a significantly larger surface area of the TCR is involved in SEB-class II binding compared with peptide-class I binding.
Single-chain antibodies have now been generated against a variety of cell surface molecules with the goal of using them for specific cell targeting in vivo (reviewed in Ref. 46). For the delivery of toxic agents to tumor cells, it seems likely that the affinity of the scFv will be a key factor in the effectiveness of the agent. In this respect it is notable that the scFv of KJ16 appears to have an affinity that is very similar to the Fab fragments. Like scFv preparations from other antibodies, there also appear to be other forms of the scFv. The KJ16 scFvЈ (Figs. 3 and 4) appears to be a proteolytic fragment of the full-length scFv, and it has reduced or negligible binding activity. The binding studies with the KJ16 scFv (Fig. 4B) also suggested that there may be dimeric or multimeric forms that yield higher apparent affinities then the monovalent scFv. These forms have now been detected in many scFv preparations (21,23,47).
In comparison with the use of scFv antibodies to deliver toxic agents to cells, the ability of anti-TCR scFv to directly control T cells in vivo may be even less predictable. This stems from the nature of the interaction of the anti-TCR antibody with the T cell. Recently, Janeway has suggested that the mechanism of antagonism by anti-TCR antibodies may involve not only steric inhibition of ligand binding but direct conformational effects induced in the TCR (15,48). The latter effect could result in signals that either activate or anergize a T cell. Our data cannot distinguish between these possibilities in the inhibition of cytolytic activity (Fig. 5, Table I), but the proposition that anti-TCR antibodies could act simply by signaling the T cell through a monovalent interaction will have some implications in the selection of anti-TCR antibodies for use in vivo. For example, one might have selected an antibody that had the highest possible affinity as an appropriate candidate for scFv engineering and subsequent therapeutic use. However, the fact that T cells undergo negative and positive selection in the thymus and that antibodies to different epitopes may influence these processes in unpredictable ways suggests that it may be necessary to compare scFv and Fab fragments from antibodies that recognize different epitopes on the same TCR. The in vivo effects of KJ16 scFv and 1B2 scFv can now be studied in light of the in vitro effects reported here.