Conformational Integrity and Ligand Binding Properties of a Single Chain T-cell Receptor Expressed in Escherichia coli *

We recently showed that a soluble, heterodimeric murine D10 T-cell receptor (TCR) (Vα2Cα, Vβ8.2Cβ) expressed in insect cells binds both Vβ8.2-specific bacterial superantigen staphylococcal enterotoxin C2 (SEC2) and a soluble, heterodimeric major histocompatibility complex class II I-Ak·conalbumin peptide complex with a low micromolar affinity. To define further the structural requirements for the TCR/ligand interactions, we have produced in Escherichia coli a soluble, functional D10 single chain (sc) TCR molecule in which the Vα and Vβ domains are connected by a flexible peptide linker. Purified and refolded D10 scTCR bound to SEC2 and murine major histocompatibility complex class II I-Ak·conalbumin peptide complex with thermodynamic and kinetic binding constants similar to those measured for the baculovirus-derived heterodimeric D10 TCR suggesting that neither the TCR constant domains nor potential N- orO-linked carbohydrate moieties are necessary for ligand recognition and for expression and proper folding of the D10 scTCR. Purified D10 scTCR remained soluble at concentrations up to 1 mm. Circular dichroism and NMR spectroscopy indicated that D10 scTCR is stabilized predominantly by β-sheet secondary structure, consistent with its native-like conformation. Because of its limited size, high solubility, and structural integrity, purified D10 scTCR appears to be suitable for structural studies by multidimensional NMR spectroscopy.


From Procept, Inc., Cambridge, Massachusetts 02139
We recently showed that a soluble, heterodimeric murine D10 T-cell receptor (TCR) (V␣2C␣, V␤8.2C␤) expressed in insect cells binds both V␤8.2-specific bacterial superantigen staphylococcal enterotoxin C2 (SEC2) and a soluble, heterodimeric major histocompatibility complex class II I-A k ⅐conalbumin peptide complex with a low micromolar affinity. To define further the structural requirements for the TCR/ligand interactions, we have produced in Escherichia coli a soluble, functional D10 single chain (sc) TCR molecule in which the V␣ and V␤ domains are connected by a flexible peptide linker. Purified and refolded D10 scTCR bound to SEC2 and murine major histocompatibility complex class II I-A k ⅐conalbumin peptide complex with thermodynamic and kinetic binding constants similar to those measured for the baculovirus-derived heterodimeric D10 TCR suggesting that neither the TCR constant domains nor potential N-or O-linked carbohydrate moieties are necessary for ligand recognition and for expression and proper folding of the D10 scTCR. Purified D10 scTCR remained soluble at concentrations up to 1 mM. Circular dichroism and NMR spectroscopy indicated that D10 scTCR is stabilized predominantly by ␤-sheet secondary structure, consistent with its native-like conformation. Because of its limited size, high solubility, and structural integrity, purified D10 scTCR appears to be suitable for structural studies by multidimensional NMR spectroscopy.
Immune activation of T lymphocytes is normally initiated by the specific interaction between an antigenic peptide presented by the major histocompatibility complex (MHC) 1 (1) and the ␣ and ␤ (or ␥ and ␦) clonotypic chains of the T-cell receptor (TCR) complex (2,3). T-cells can also be stimulated by superantigens, although the TCR/superantigen interaction differs from that of TCR/MHC peptide in that specificity for a particular superantigen appears to be determined only by the sequence of the germ line-encoded V␤ segment of the TCR (4,5).
Because of its central role in immune recognition, detailed structural information on the TCR is of great interest. Recently, three-dimensional structures of a heterodimeric, glycosylated murine ␣␤ TCR (2C), the corresponding 2C TCR/class I MHC-peptide complex (6), and a nonglycosylated human ␣␤ TCR (A6) bound to a class I MHC-peptide complex (7) were solved using x-ray crystallography. The overall orientation of the CDR regions of both of these TCRs is similar. Moreover, the folded variable domain structures of both TCRs resemble the antigen binding region of antibodies, although differences exist in the C␣ domain and in the interdomain pairing of C␣ and C␤ (6). Crystal structures of an isolated TCR ␤-chain (8) and a V␣ homodimer (9) also showed immunoglobulin-like folding patterns. Three-dimensional crystal structures of staphylococcal enterotoxin C2 and C3 bound to the extracellular portion of the ␤-chain (V␤8.2J␤2.1C␤1) of the mouse 14.3.d TCR were also recently reported (10). Unlike the TCR/MHC interactions (6,7), only the CDR2 and to lesser extents CDR1 and hypervariable region 4 (HV4) of the V␤-chains are involved in binding to SEC.
Three-dimensional high resolution structures can also be derived from NMR-based approaches (11,12). The NMR structures have the potential advantage of being obtained in solution and can provide insights into conformational flexibility and structural dynamics (13,14). Although the NMR approach is mostly limited to proteins with molecular masses less than ϳ30 kDa that can be uniformly enriched with 13 C and 15 N isotopes and remain soluble at high submillimolar concentrations (12), deuteration in concert with 13 C, 15 N multidimensional NMR techniques can push this limit to higher molecular masses (15). Because of their limited size (ϳ30 kDa), single chain (sc) TCR molecules in which the V␣-and V␤-chains are connected by a flexible polypeptide linker (reviewed in Ref. 16) might be amenable to NMR structure determination, and various scTCR constructions have been produced previously using Escherichia coli expression systems (reviewed in Ref. 16). Despite this, the conformational integrity and ligand binding characteristics of these proteins have not been well established (Ref. 17, and for review see Ref. 16).
To define further the structural requirements for D10 TCR binding, and to obtain material suitable for NMR structural studies, we produced a soluble D10 scTCR molecule using an E.
coli expression system. Purified and refolded protein bound to SEC2 and I-A k -conalbumin peptide complex with thermodynamic and kinetic binding constants similar to those measured previously for the insect cell-derived, heterodimeric D10 TCR (18), suggesting that neither the TCR constant domains nor potential N-or O-linked carbohydrate moieties are necessary for ligand binding. In addition, based on circular dichroism and one-dimensional NMR spectroscopy, purified D10 scTCR appears to be suitable for structural studies by multidimensional NMR spectroscopy. NMR studies on the D10 scTCR may be helpful in assessing the flexibility of its various CDR regions and may provide further insight into whether the scTCR undergoes conformational changes as a result of its interactions with superantigens and class II MHC-peptide complex.
E. coli Expression Vector Encoding Maltose-binding Protein (MBP)-D10 scTCR Fusion Protein-Plasmids pFRSV-SRaD10␣ GPI TCR and pFRSV-SRbD10␤ GPI TCR, encoding D10 ␣and ␤-chain variable domain genes, respectively, were gifts from Dr. Al Bothwell. These two cDNA plasmids were used as templates from which sequences encoding D10 TCR ␣and ␤-chain variable domains were generated, using PCR. Using appropriate oligonucleotide primers, the ␣-chain gene was engineered to encode NarI and XhoI restriction sites at the 5Ј-and 3Ј-ends, respectively, as well as two stop codons at the 3Ј-end. The ␤-chain gene was engineered to encode NcoI and BamHI restriction sites at the 5Јand 3Ј-ends, respectively. The amplified genes were then ligated into plasmid p3xG"b3" using the 5Ј-and 3Ј-restriction sites, to generate plasmid p6/530. A flexible linker (the 3xG-linker, amino acid sequence GSGGGGSGGGGSGGSGA) was engineered into the plasmid between the ␣and ␤-chain genes, resulting in a gene encoding D10 scTCR in the order ␤-linker-␣ (Fig. 1). The 5Ј-end of the scTCR gene was then modified, using PCR and standard molecular biology techniques, to encode an EcoRI restriction site, followed by a thrombin cleavage site encoding amino acids LVPRG (28). This modified D10 scTCR gene was inserted between the EcoRI and SalI sites within the polylinker of the MBP fusion protein expression vector pPR998 (New England Biolabs, Beverly, MA). The resulting plasmid encoded a MBP-D10 scTCR fusion protein. The tether sequence, between the 3Ј-end of the malE gene (encoding MBP) and the 5Ј-end of the D10 TCR gene, was then deleted for the factor Xa cleavage site present in plasmid pPR998, leaving the thrombin cleavage site previously engineered at the 5Ј-end of the D10 TCR gene. The resulting plasmid encoded a MBP-D10 scTCR fusion protein with a 10-amino acid tether (NSSSLVPRGS).
Using oligonucleotide primers to generate DNA of the correct sequence by the PCR, the linker between TCR ␣and ␤-chain genes was then modified by the addition of the FLAGG sequence (IBI, New Haven, CT), encoding amino acids DYKDDDDK. The resulting 27-amino acid linker sequence was GSDYKDDDDKRSGGGGSGGGGSGGSGA. Also, to minimize potential covalent aggregation, the codon for an unpaired cysteine residue in the J␣-region was changed to encode a serine residue. Finally, to aid in protein purification, a sequence encoding six histidine residues was added to the 3Ј-end of the MBP-D10 scTCR gene. This resulted in the final MBP-D10 scTCR fusion protein expression vector, p4/219. A schematic of the resulting MBP-D10 scTCR fusion protein construct is shown in Fig. 1A, and the amino acid sequence of D10 scTCR is shown in Fig. 1B. It was subsequently discovered that the fifth codon of the ␤-chain gene of the D10 scTCR construct was mutated to encode a threonine rather than a serine residue (Fig. 1B). The mutation presumably occurred as a result of an error by Taq polymerase during a PCR procedure. The mutation appears to be benign, as the purified and refolded D10 scTCR protein was found to be in a nativelike conformation (see below).
All purification steps, with the exception of nickel-nitrilotriacetic acid (Ni-NTA) metal affinity chromatography, were performed at 4°C. Lysis supernatant was applied to a cross-linked amylose column (New England Biolabs, Beverly, MA) at a flow rate of 3 ml/min. The column was then washed with 50 mM Tris-HCl (pH 8.0) containing 0.2 M NaCl, and bound proteins were eluted with the same buffer containing 10 mM maltose. Purified fusion protein was next denatured with 6 M GdnHCl in 50 mM Tris-HCl (pH 8.0) containing 0.2 M NaCl and 10% glycerol (buffer A), and applied to a Ni-NTA metal affinity column (Qiagen, Chatsworth, CA) at a flow-rate of 3.7 ml/min. After washing the column with 2 column volumes of buffer A, a refolding gradient was initiated and maintained at a flow rate of 3.7 ml/min using fast protein liquid chromatography (Pharmacia Biotech Inc.). A 240-min linear gradient was formed from 100% buffer A to 100% buffer B containing 50 mM Tris-HCl (pH 8.0), 0.2 M GdnHCl, 20% glycerol, and 0.5 M NaCl. After washing the column with 1 column volume of buffer B, the bound material was eluted with buffer B containing 250 mM imidazole. The eluted MBP-D10 scTCR fusion protein was further purified by immunoaffinity chromatography on an immobilized 3D3 column, which specifically recognizes a D10 TCR conformational epitope (18,21,29). Finally, monomeric MBP-D10 scTCR was isolated by size exclusion chromatography using a Superdex 200PG 26/60 column (Pharmacia Biotech Inc.).
Protein concentrations were determined by spectrophotometry at 280 nm using extinction coefficients of 1.7 and 1.2 M Ϫ1 cm Ϫ1 for MBP-D10 scTCR and D10 scTCR, respectively. Quantitative amino acid analysis was performed using an Applied Biosystems (Perkin-Elmer) model 421 amino acid analysis system. To test the solubility characteristic of D10 scTCR, a small amount of purified protein was concentrated using Centricon 10 concentrator. D10 scTCR remained soluble at concentrations up to 1 mM.
Protein Analysis-SDS-PAGE (30) and isoelectric focusing (31) analyses were performed essentially as described. N-terminal amino acid sequence analyses were performed on the proteins blotted on ProBlott membranes (Applied Biosystems, Foster City, CA) as described (32).
Circular Dichroism (CD) Spectroscopy-CD analysis was performed on purified D10 scTCR (0.15 mg/ml in 2 mM HEPES (pH 7.2)). Far-UV CD spectra were recorded on a CD instrument model 62 DS (Aviv Associates, Lakewood, NJ) using a 2-mm path length cell. Data were collected using a time constant of 1 s at every 0.2 nM and with a 1 nm constant spectral bandwidth at 25°C. The CD data were analyzed for secondary structure prediction by an algorithm Prosec V3.1 provided by the manufacturer (Aviv Associates, Lakewood, NJ).
NMR Spectroscopy-One-dimensional NMR experiments were carried on purified D10 scTCR (ϳ80 M in 95% (v/v) H 2 O/D 2 O and 20 mM deuterated acetate (pH 4.5)) at 25 and 37°C on a Varian UNITYplus750 MHz spectrometer with a 1 H resonance frequency of 750.079 MHz. The spectral width was 23,995 Hz, and the carrier was placed on the H 2 O resonance which was suppressed by low power presaturation during the recycle delay (1.3 s). 8,192 complex points were obtained, and 1,024 transients were collected for signal averaging. Data processing was carried out using Felix software (Biosym Technologies, San Diego). The residual water signal was removed by the time domain convolution technique. Data were multiplied with a /2 shifted squared sine bell apodization function and zero-filled to 16,384 points prior to Fourier transformation. 1 H chemical shifts were referenced relative to the water resonance, calibrated in turn at 4.755 and 4.631 ppm at 25 and 37°C, respectively, on an external 2,2-dimethyl-2-silapentanesulfonic acid standard.
Surface Plasmon Resonance (SPR) Binding Experiments-Binding studies were performed using a commercial biosensor instrument (BIAcore TM , Pharmacia BioSensor, Uppsala, Sweden). All proteins used in these experiments were purified by size exclusion chromatography. Immobilizations were carried out in 10 mM sodium acetate at pH 5.5 for D10 scTCR or at pH 4.0 for superantigens using the Amine Coupling Kit (Pharmacia BioSensor, Uppsala, Sweden) as described (18,19). Heterodimeric, soluble D10 TCR (18) and BDC2.5 TCR (33) derived from baculovirus-infected insect cells were used as control proteins for specificity experiments. HEPES-buffered saline (HBS; 10 mM HEPES (pH 7.4), 150 mM NaCl, 3.4 mM EDTA, and 0.005% surfactant P20) was injected over the sensor chip at 5 l/min, and 20-l analyte samples diluted (ϳ25-fold) in HBS buffer were injected. Where required, the biosensor surface was regenerated with 5-10 mM HCl. All binding experiments were conducted at 25°C. The SPR signal was recorded as resonance units (RU) versus time and plotted as a "sensorgram." For kinetic rate analyses, binding studies were performed on ligands immobilized at multiple densities. The data, generated at least in triplicate, were analyzed using the non-linear curve-fitting software package (BIAevaluation 2.1, Pharmacia Biosensor) that also allows the parameters to be extracted from a single run (18, 34 -36). To determine the apparent association (k on ) and dissociation (k off ) rate constants, the association and dissociation phases of the sensorgrams were fitted on single site models, A ϩ B ϭ AB and AB ϭ A ϩ B, respectively. The apparent dissociation constant (K D ) was determined from the ratio of k off /k on .

RESULTS
Expression and Purification of D10 scTCR-The construct used to produce recombinant D10 scTCR is shown schematically in Fig. 1A. The TCR V␤ and V␣ domains were connected by a 27-residue flexible peptide linker (Fig. 1B), and the E. coli maltose-binding protein (MBP) was fused to the N terminus of the V␤ domain via a linker containing a thrombin cleavage site. In addition, six histidine residues were added to the C terminus to allow binding to a Ni-NTA affinity chromatography column. The expressed protein was purified under nonreducing conditions to preserve the disulfide bonds formed in vivo. A similar approach was utilized to refold an E. coli-expressed single chain antibody (37).
Monomeric MBP-D10 scTCR fusion protein was purified from E. coli lysis supernatants using a combination of amylose, Ni-NTA, and mAb 3D3 affinity chromatography steps. Purified fusion protein was then subjected to Superdex 200PG size exclusion chromatography to remove the small amounts of aggregates. Purity of the 70-kDa fusion protein was monitored after each step by SDS-PAGE performed under reducing ( Fig.   2A, lanes 1-4) and nonreducing ( Fig. 2A, lanes 5-8) conditions. The N-terminal sequence determined for the purified product (KIEEGKLVI) was as expected for the correctly processed maltose-binding protein (38).
Thrombin digestion of size exclusion chromatography, purified monomeric fusion protein resulted in efficient cleavage (Fig. 2B). Cleavage of the aggregated fusion protein isolated following size exclusion chromatography, by contrast, was inefficient and resulted in significant precipitation (not shown). Thrombin-cleaved D10 scTCR was purified by Ni-NTA affinity chromatography carried out under nondenaturing conditions (Fig. 2B). The overall yield of purified D10 scTCR was 0.7-1.0 mg/liter of E. coli cell culture.
Characterization of D10 scTCR-The N-terminal amino acid sequence of purified D10 scTCR was determined to be GSAVSQSP, corresponding exactly to the amino acid sequence predicted following thrombin cleavage of the fusion protein (Fig. 1B). On SDS-PAGE analysis, purified D10 scTCR migrated faster under non-reducing than under reducing conditions (Fig. 3A), suggesting the existence of intramolecular disulfide bonds that lead to a more compact structure. Furthermore, TCR family-specific mAbs B20.1 (V␣2) and MR5.2 (V␤8.1/2) recognized only the non-reduced protein on Western blot analysis (not shown). Similar results were obtained with baculovirus-expressed, disulfide-linked, heterodimeric D10 TCR. 2 Together, these results indicate that purified D10 scTCR is in a correctly disulfide-bonded conformation. The isoelectric point (pI) of purified D10 scTCR was determined to be 8.8 under native conditions (Fig. 3B), consistent with the theoretical pI of 8.9 calculated from the primary amino acid sequence. Purified D10 scTCR appears to be mono-2 S. Khandekar, unpublished observations. meric. The molecular mass was determined to be approximately 28 kDa by size exclusion chromatography performed under nondenaturing conditions (Fig. 3C). The molecular mass of purified D10 scTCR was also determined by electrospraymass spectrometry to be 27,891.5 daltons (not shown). These values are in agreement with the molecular mass of 27,893 daltons calculated from the primary amino acid sequence of the thrombin-cleaved D10 scTCR containing the C-terminal hexahistidine extension.
The conformation of purified D10 scTCR was examined by CD and NMR spectroscopy. The far UV-visible CD spectrum indicated that purified D10 scTCR contained predominantly ␤-sheet secondary structure (Fig. 4A). Deconvolution of the digitized CD data provided the following predictions for the relative percentage of secondary structural elements: 11% ␣-helix, 72% ␤-sheet, 14% random coil, and 3% turn. These results are in agreement with the recently solved crystal structures of the heterodimeric TCRs (6,7) and the immunoglobulin domain model for the V␣/V␤ TCR structure as proposed by others (17, 39 -42).
To probe further the secondary structure of the purified protein, a series of one-dimensional NMR experiments were performed. A representative one-dimensional NMR spectrum is shown in Fig. 4B. The following observations strongly suggest that purified D10 scTCR predominantly adopts a ␤-sheet secondary structure: 1) the amide signals (ϳ11.0 to ϳ6.0 ppm) show a dispersion, which is typical for folded proteins (11); 2) there are many low field shifted H ␣ resonances, which confirm that the scTCR is predominantly folded (A in Fig. 4B); 3) a few isolated signals upfield from the methyl resonances are present, which are due to large ring current shifts in a folded structure (B in Fig. 4B); 4) many downfield shifted H ␣ resonances are indicative of ␤-sheet secondary structure (42)(43)(44). The spectrum also shows clustering of a large number of sharp resonances between 8.3 and 8.0 ppm, indicating that a part of the protein may be unstructured (45) (C in Fig. 4B). The highly repetitive primary sequence of the "FLAGG/3xG"-linker and the hexahistidine tag at the C terminus may give rise to these resonances.
Binding of D10 scTCR to V␤8-specific Superantigens-Our previous SPR binding studies demonstrated that V␤8.2-specific bacterial superantigen SEC2 binds to insect cell-derived solu-   1 M (18). In contrast, SEB, another V␤8.2-specific bacterial superantigen, bound to D10 TCR very weakly (K D Ͼ 100 M), although it, like SEC2, has been shown to stimulate proliferation of the D10 T-cell clone (46). To examine superantigen binding characteristics of E. coli-derived D10 scTCR, SPR analyses were performed. First, purified D10 scTCR was immobilized to the sensor surface and bacterial superantigens SEC2, SEB, and TSST-1 were each injected over the surface at a concentration of 0.5 mg/ml (18 -20 M) in water (Fig. 5A). Similar to our previous findings (18), V␤8.2-specific superantigens SEC2 and SEB bound to immobilized D10 scTCR, whereas V␤15-specific TSST-1 did not (Fig. 5A). The negative change in base line during injection of TSST-1 is due to a change in the bulk refractive index resulting from differences in the salt concentrations of the protein solutions and the running buffer (17,18). In addition, the sensorgram shown in Fig. 5A indicates that, like baculovirus-derived D10 TCR, E. coli-expressed D10 scTCR binds to SEC2 with higher affinity than SEB due to a slower dissociation rate.
The binding of solution phase D10 scTCR to immobilized superantigens was also examined. D10 scTCR (5 to 1 M) bound to an SEC2-coated (Fig. 5B) surface in a concentration-dependent manner but did not bind to TSST-1-coated (Fig. 5C) or control (Fig. 5D) surfaces. Moreover, similar to our previous findings (18), D10 clone-specific mAb 3D3 completely blocked the sD10 TCR/SEC2 interaction (Fig. 5E). Antibody 3D3 (21) recognizes a conformational epitope formed by the juxtaposition of the V␣ and V␤ domains (18). These results therefore imply that the epitope for mAb 3D3 may overlap the binding site for SEC2. Alternatively, it may be that the excluded volume of 3D3 sterically blocks the binding of SEC2.
Kinetic constants for the D10 scTCR-SEC2 interaction were also determined using SPR measurements (34,35). For D10 scTCR binding to immobilized SEC2, k on and k off were determined to be 2.45 Ϯ 0.72 ϫ 10 4 M Ϫ1 s Ϫ1 and 1.36 Ϯ 0.15 ϫ 10 Ϫ2 s Ϫ1 , respectively. From these values, a K D of about 0.6 M was calculated. These values are similar to those determined for the baculovirus-derived, heterodimeric D10 TCR/SEC2 interaction ( Table I).
Binding of D10 scTCR to Soluble MHC Class II-Peptide Complex-Our previous studies showed that an insect cellderived, soluble, heterodimeric murine class II MHC I-A k derivative containing fused antigenic conalbumin peptide and complementary leucine zipper sequences (I-A k -CA-LZ) specifically stimulates proliferation of the D10 T-cell clone and binds to the soluble, heterodimeric D10 TCR with a K D of about 2.1 M (18). As expected, this interaction was not seen with the soluble, heterodimeric BDC2.5 TCR (18), a control TCR protein specific for a murine pancreatic ␤-cell antigen presented by murine MHC class II I-A g7 (33,47). Moreover, neither unloaded I-A k -LZ nor human HLA-DR1⅐HA peptide complex showed detectable binding to the heterodimeric D10 TCR, indicating that the bimolecular interaction between D10 TCR and I-A k -CA-LZ is specific (18).
To understand better the structural requirements for immune recognition, SPR experiments were performed to examine the binding interactions between D10 scTCR and various MHC class II-peptide complexes. Purified I-A k -CA-LZ at 5 M bound specifically to D10 scTCR immobilized on the sensor surface (Fig. 6A). The sensorgram spike at the end of the injection is due to differences in salt concentration between the protein solution and running buffer. Neither I-A k -LZ at 5 M nor human HLA-DR1⅐HA peptide complex (27) at 15 M showed detectable binding. These specificities are identical to those observed for heterodimeric D10 TCR (18).
Kinetic and equilibrium binding constants for I-A k -CA-LZ binding to immobilized D10 scTCR were also measured using the SPR biosensor. For these studies, I-A k -CA-LZ was injected over a D10 scTCR-coated surface at various concentrations (5 to 1 M) (Fig. 6B), and association and dissociation rate constants were determined to be 1.71 Ϯ 0.47 ϫ 10 4 M Ϫ1 s Ϫ1 and 2.58 Ϯ 0.38 ϫ 10 Ϫ2 s Ϫ1 , respectively. The ratio of these kinetic constants gave an equilibrium dissociation constant of 1.5 M. These values are nearly identical to those determined for the heterodimeric D10 TCR/I-A k -CA-LZ interaction (Table I) (18). DISCUSSION In the present study, milligram amounts of refolded, nativelike D10 scTCR were produced for structural and functional FIG. 4. Biophysical characterization of D10 scTCR. A, far UV CD spectrum of purified D10 scTCR. Far UV CD spectrum of purified D10 scTCR was recorded on an Aviv Associates model 62 DS CD spectrophotometer. Protein concentration, 5.2 m. B, one-dimensional NMR spectrum of purified D10 scTCR at 25°C. The dispersion of the amide resonances (ϳ11.0 to ϳ6.0 ppm) is typical for a folded protein.
The position of the water resonance which was removed using the time domain convolution technique is indicated by an arrow. A, many H ␣ resonances downfield shifted from the water resonance are indicative of ␤ secondary structure. B, isolated signals upfield from the methyl resonances due to ring current shifts are typically observed in a folded structure. C, the highly repetitive primary sequence of the "FLAGG/ 3xG"-linker and the hexahistidine tag at the C terminus may result in the clustering of amide resonances, which is indicative of unstructured regions. Some intense signals between 3.9 and 3.5 ppm presumably originate from an impurity. studies. The scTCR molecule was expressed as a fusion protein in which the E. coli maltose-binding protein was fused to the N terminus of the D10 scTCR molecule. We selected MBP (38,48) as a fusion partner for the following reasons. First, MBP can be transported to the periplasmic space of E. coli, where proper disulfide bonds can be formed (49). Second, the fusion protein can be purified in high yield utilizing amylose affinity chromatography. Third, MBP may act as a scaffold to facilitate the folding of the scTCR domains. Finally, MBP does not contain any cysteine residues that would interfere with the intramolecular disulfide bond formation within the D10 scTCR molecule.
Purified D10 scTCR possessed native-like properties based on several analytical techniques. CD and one-dimensional NMR spectra indicated that purified protein is stabilized predominantly by ␤-sheet secondary structure. Recently, similar results were reported for an E. coli-derived V␣ domain of a murine B4.2.3 TCR (42). Importantly, purified D10 scTCR remains soluble at concentrations to 30 mg/ml, further indicating a native-like conformation. This has not been reported for other E. coli-derived scTCR molecules (17,40,41,50,51). Previously, it was suggested that the solubility of E. coli-derived scTCR molecules is enhanced by the presence of Ͼ20 charged amino acid residues in the variable domains (particularly V␣) and by a small (1-2) number of potential N-linked glycosylation sites (52). The solubility of purified D10 scTCR is consistent with the first criterion (there are 23 charged amino acids in the ␣-chain) but not with the second (there are 3 potential N-linked glycosylation sites in D10 scTCR).
The ability of E. coli-derived D10 scTCR to interact with V␤8.2 TCR-specific bacterial superantigens in the absence of MHC class II molecules is consistent with our previous finding (Ref. 18, also see Table I) and also with other published reports (17,(53)(54). The kinetic and binding constants for the D10 scTCR/SEC2 interaction are similar to those of the baculovirusderived, heterodimeric (V␣C␣/V␤C␤) D10 TCR/SEC2 interaction (Table I (55). Together, these results further imply that residues in the common V␤8.2 segment alone are sufficient for binding to superantigens, at least in the absence of MHC (17,(55)(56)(57). The recently solved three-dimensional crystal structures of the ␤-chain of the 14.3.d TCR bound to both SEC2 and SEC3 further support this conclusion (10).
Kinetic and dissociation constants for the interactions between different TCRs and MHC-antigen complexes have been determined by several laboratories (58,59, also reviewed in Ref. 16). Our recent studies (18) showed that the affinity of soluble, heterodimeric D10 TCR for soluble I-A k ⅐CA peptide complex is about 2.1 M. The K D for E. coli-derived D10 scTCR/I-A k ⅐CA peptide complex was found to be nearly identical (1.5 M) to this value (Table I), further implying that the variable domains of the heterodimeric D10 TCR are sufficient for their interaction with the murine MHC class II I-A k ⅐CA peptide complex.
Malchiodi and co-workers (55) showed that an unglycosylated mutant form of the murine 14.3.d TCR ␤-chain in which four out of five potential Asn-linked glycosylation sites were eliminated through site-directed mutagenesis bound to various  forms of SEC with affinities similar to those of fully glycosylated form. Although these results suggest that N-linked glycans do not contribute to superantigen binding, the possible involvement of the fifth remaining site as well as the potential yet unidentified O-linked carbohydrate sites (60) in binding to SEC could not be ruled out. Since the E. coli-derived D10 scTCR was not glycosylated, our studies provide a direct evidence that the carbohydrates do not play a significant role in the binding of D10 TCR to SEC2 and also to murine MHC class II I-A k ⅐CA peptide complex. In addition, these results suggest that carbohydrates are also not required for the expression and conformational stability of the D10 TCR. Sequence alignments (39) and more recently the x-ray crystallographic data (6 -9) suggest that TCRs and Igs are built on the same principles. Interestingly, the interdomain contact residues that are present in the corresponding V L -V H antibody interface are also conserved between V␣ and V␤ domains of TCR (6). Moreover, similar to Ig molecules, within each V␣ and V␤ domains of 2C TCR about 24 residues form the core of the ␤-sheet sandwich (6). The CD and one-dimensional NMR spectra of purified D10 scTCR described here also indicate that scTCR is stabilized predominantly by ␤-sheet secondary structure, consistent with its native-like conformation.
Because of its size, high solubility, and structural integrity, purified D10 scTCR should be suitable for structural studies by multidimensional NMR spectroscopy. 3 Such studies may help in determining the packing of the V␣ and V␤ domains in scTCR and in assessing the flexibility of its various CDR regions in solution. In addition, the structural and dynamic studies with superantigens and class II MHC/peptide should be useful in further understanding the molecular interactions between TCR and its ligands.