Crystallization of a deglycosylated T cell receptor (TCR) complexed with an anti-TCR Fab fragment.

A strategy to overexpress T cell receptors (TCRs) in Lec3.2.8.1 cells has been developed using the “Velcro” leucine zipper sequence to facilitate α-β pairing. Upon secretion in culture media, the VSV-8-specific/H2-Kb-restricted N15 TCR could be readily immunopurified using the anti-leucine zipper monoclonal antibody 2H11, with a yield of 5-10 mg/liter. Mass spectrometry analysis revealed that all attached glycans were GlcNAc2-Man5. Following Superdex 200 gel filtration to remove aggregates, wild-type N15 or N15s, a C183S variant lacking the unpaired cysteine at amino acid residue 183 in the Cβ domain, was thrombin-cleaved and endoglycosidase H-digested, and the two derivatives were termed iN15ΔH and N15sΔH, respectively, and sized by Superdex 75 chromatography to high purity. N-terminal and C-terminal microsequencing analysis showed the expected unique termini of N15 α and β subunits. Nevertheless, neither protein crystallized under a wide range of conditions. Subsequently, we produced a Fab fragment of the murine TCR Cβ-specific hamster monoclonal antibody H57 and complexed the Fab fragment with iN15ΔH and N15sΔH. Both N15sΔH-Fab[H57] and iN15ΔH-Fab[H57] complexes crystallize, with the former diffracting to 2.8-Å resolution. These findings show that neither intact glycans nor the conserved and partially exposed Cys-183 is required for protein stability. Furthermore, our results suggest that the H57 Fab fragment aids in the crystallization of TCRs by altering their molecular surface and/or stabilizing inherent conformational mobility.

The TCR 1 complex consists of multiple transmembrane polypeptide chains on the surface of T lymphocytes (1)(2)(3). The disulfide-linked ␣␤ heterodimer (Ti) is the clonally unique component that possesses a recognition site for antigen in the context of a major histocompatibility complex protein (4,5). Sequence analysis of ␣ and ␤ subunits strongly supports that their ectodomains form a recognition unit reminiscent of an immunoglobulin Fab fragment (6 -8). This notion has been confirmed in crystallographic studies of TCR subunit fragments (9,10). On the other hand, the invariant CD3 components (␥, ␦, ⑀, , and ) possess lengthy cytoplasmic tails containing immune cell tyrosine-based activation motifs and are involved in signal transduction (11,12). To date, most of the attributes of TCR recognition have been studied largely indirectly because of the intimate membrane association of this complex.
To understand the process by which T cells recognize pathogens in explicit molecular terms, recent efforts have begun to focus on the structural nature of the TCR. However, many efforts to express soluble TCR ␣␤ heterodimers in both prokaryotic and eukaryotic systems have been hampered by inefficient pairing of ␣ and ␤ subunits in the absence of their respective transmembrane regions and associated CD3 components (reviewed in Refs. 13-21). We have recently developed a methodology to overcome this obstacle by adding 30-amino acid-long segments to the carboxyl termini of ␣ and ␤ extracellular domains via a thrombin-cleavable flexible linker (22). These peptide segments (Base-p1 for ␣ and Acid-p1 for ␤) were previously shown to selectively associate to form a stable heterodimeric coiled coil termed the leucine zipper (23). Homodimeric structures are not favored due to the electrostatic repulsion among amino acid side chains. Furthermore, the yield of these engineered proteins was 5-10-fold greater than that of the TCR expressed in the absence of the synthetic leucine zipper (22). Through the use of a panel of mAbs directed at native ␣ and ␤ epitopes within constant and variable regions, it was further shown that the fusion heterodimer was native.
Efforts to obtain TCR crystals from such material, however, were unsuccessful, perhaps owing to the glycosylation heterogeneity inherent in the baculovirus system and/or the intrinsic conformer heterogeneity of the TCR heterodimers under examination. Here we describe a system using Chinese hamster ovary Lec3.2.8.1 cells (24) to express TCRs that can be readily deglycosylated by endoglycosidase H. Our results demonstrate that deglycosylation of the TCR with endoglycosidase H does not alter the native TCR structure, implying that the intact glycans are not required for the protein's stability. The behavior of a C183S TCR variant lacking the unpaired cysteine at amino acid residue 183 of the C␤ domain indicates that this partially exposed and conserved residue is not necessary for TCR expression or structure. We further show that by complexing such a deglycosylated TCR derivative with a specific anti-TCR Fab fragment, it is possible to obtain TCR crystals that diffract at atomic resolution.

EXPERIMENTAL PROCEDURES
Cloning of N15 into the pEE14-GS Expression Vector-The generation of constructs containing the TCR fused to leucine zipper components, carboxyl-terminal to the membrane-proximal cysteine, has been described previously (22). In brief, two DNA fragments encoding a linker sequence and leucine zipper peptide were generated using synthetic oligonucleotides and polymerase chain reaction. The DNA fragments were then subcloned into the BamHI-EcoRI site of the pCRII vector (Invitrogen) for DNA sequencing analysis and further manipulation. The TCR-leucine zipper cDNAs for N15␣Base and N15␤Acid were generated by ligation of EcoRI-BamHI fragments from the pcN15␣ and pcN15␤ constructs (22) to the BamHI-EcoRI fragments of Base-p1 and Acid-p1 DNAs, respectively. The cDNAs of pN15␣Base and pN15␤Acid were digested with EcoRI to release the TCR inserts. The DNA fragments were subcloned into the EcoRI site of the glutamine synthetase expression vector, pEE14-GS (25), to generate the plasmids pEE14-N15␣Base and pEE14-N15␤Acid, respectively. The pEE14-N15␤Acid plasmid was then double-digested with MluI plus NaeI to release the DNA fragment containing the hCMV-MIE promoter plus N15␤Acid. The cDNA fragment was then blunted with Klenow DNA polymerase and ligated with MluI linkers. The MluI fragment of hCMV-N15␤Acid was then subcloned into the MluI site of pEE14-N15␣Base to yield the final glutamine synthetase expression construct pEE14GS-N15␤␣. All restriction enzymes and linkers were obtained from New England Biolabs Inc.
Mutagenesis of N15 to Produce N15 s -To generate the N15 s mutant TCR with the cysteine residue converted to serine at position 183 of the ␤ subunit, two independent polymerase chain reactions were performed using the full-length cDNA of N15␤Acid as a template. The first reaction employed a 5Ј-oligonucleotide amplimer (oligonucleotide 1) encoding the leader peptide sequence, 5Ј-GAAGAAGCATGTCTAACAACT-GTCCTCGC-3Ј, in conjunction with the antisense strand amplimer (oligonucleotide 2), 5Ј-CGGCTGCTCAGGGAGTAGCTATAATTGCTC-3Ј, to generate a DNA fragment corresponding to amino acids 1-187 of N15␤C183S. The second reaction used the mutagenic sense strand amplimer (oligonucleotide 3), 5Ј-GCAATTATAGCTACTCCCTGAG-CAGCCG-3Ј, paired with the antisense strand oligonucleotide (oligonucleotide 4), 5Ј-GGAATTCCTACTGAGCCAGTTCCTTTTCC-3Ј, to generate a 3Ј-DNA fragment encoding amino acids 173-285 of N15␤C183S. Subsequently, the two DNA fragments were denatured, renatured, and used as templates for a polymerase chain reaction with oligonucleotides 1 and 4 to generate a full-length cDNA fragment encoding for the N15␤C183SAcid subunit. The DNA fragment was subcloned into the pCRII vector for DNA sequencing analysis. After sequence verification, the cDNA was then cloned into the EcoRI site of the pEE14 expression vector and cotransfected with the pEE14 -N15␣Base construct into Lec3.2.8.1 cells to produce N15 s TCR.
Expression of N15 Using a Glutamine Synthetase Vector in Lec3.2.8.1 Cells-The Lec3.2.8.1 cells were grown at 37°C as a monolayer in ␣-minimal essential medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum, 2 mM glutamine, and 50 units/ml penicillin/ streptomycin. Two days before transfection, the Lec3.2.8.1 cells were trypsinized, and 1 ϫ 10 6 cells were seeded per 100-cm 2 tissue culture dish (Becton Dickinson) in GMEM-S medium (containing 1 ϫ GMEM (without glutamine and without tryptose phosphate broth), 0.27% sodium bicarbonate, 1 ϫ nonessential amino acids, 0.06 mg/ml each glutamate and asparagine, 1 mM sodium pyruvate, 1 ϫ nucleosides, and 50 units/ml penicillin/streptomycin) plus 10% dialyzed fetal calf serum. For each dish, 20 g of SalI-linearized pEE14GS-N15␤␣ DNA was used for transfection by a calcium phosphate precipitation method (Stratagene) following the manufacturer's protocol. Forty-eight hours after transfection, the cells were trypsinized, resuspended in 10 ml of GMEM-S medium containing 25 M methionine sulfoximine, and cultured onto 96-well plates. Three to four weeks later, the growing clones were assayed for secretion of soluble TCR (sTCR) heterodimers by enzyme-linked immunosorbent assay. In brief, 5 g/ml anti-C␤ mAb H57-597 (H57) was coated onto Immulon plates (Dynatech Laboratories Inc.) at room temperature for 2 h and then blocked with 1% bovine serum albumin in borate-buffered saline at room temperature for 2 h. Fifty microliters of individual clone culture supernatants was plated overnight at 4°C, mixed with 1 g/ml biotinylated anti-C␣ mAb H28 for 2 h, and developed with horseradish peroxidase-conjugated streptavidin (Sigma). The highest secreters were picked and ranked by rescreening the supernatants using an indirect capture with anti-Velcro mAb 2H11 (22) on BIAcore (Pharmacia Biosensor). Rabbit anti-mouse IgG1 (Pharmacia Biosensor) was coupled to the sensor chip by standard N-hydroxysuccinimide/N-ethyl-NЈ-(dimethylaminopropyl)carbodiimide chemistry. mAb 2H11 (mouse IgG1) was captured on the surface through its Fc segment (flow rate of 5 l/min, 30 -60 g/ml in HEPESbuffered saline). A 1:3 diluted supernatant was then passed over the anti-Velcro surface. Stable clones routinely produce TCR at Ն5 mg/liter, corresponding to 1000 -1200 resonance units, under these experimental conditions. Note that since these experiments were not done under equilibrium conditions, the calibration curve is nonlinear.
Large-scale Production of TCR Protein-The transfected Chinese hamster ovary Lec3.2.8.1 cell lines producing sTCR (N15-2B33 for N15 and N15-S29 for N15 s ) were cultured in glutamine-free medium (complete GMEM-S medium containing 10% dialyzed fetal calf serum and 25 M methionine sulfoximine). For large-scale protein production, ϳ8 ϫ 10 6 Lec3.2.8.1 transfectants were seeded per flask into 60 175-cm 2 tissue culture flasks (Falcon) in 35 ml of complete GMEM-S medium and were grown to confluence (ϳ3-4 days). Subsequently, an additional 35 ml of fresh GMEM-S medium containing 5% dialyzed fetal calf serum, 25 M methionine sulfoximine, and 4 mM sodium butyrate was added to each of 40 flasks, and cells were cultured for an additional 5-6 days, at which time the supernatants (2.8 liters) were harvested and centrifuged for 10 min at 2000 rpm. The cells from the remaining 20 flasks were harvested using 0.125% trypsin and 0.5 mM EDTA and expanded into 50 175-cm 2 tissue culture flasks, and the entire process was repeated.
Purification of sTCR Protein-The Lec3.2.8.1 supernatants of N15-2B33 and N15-S29 containing N15 sTCR-leucine zipper fusion proteins (N15 and N15 s , respectively) were filtered using a Corning filter (0.22 m), and further steps were performed at 4°C. N15 and N15 s were purified by a two-step procedure involving immunoaffinity chromatography and gel filtration. The first step used mAb 2H11 coupled at 5 mg/ml to GammaBind Plus-Sepharose beads (Pharmacia) via crosslinking with dimethyl pimelimidate, and the second step consisted of FPLC (Pharmacia) gel filtration chromatography on a Superdex 200 column controlled by a computer FPLC director program (Pharmacia). Approximately 1.5 liters of N15 supernatants was first passed over the GammaBind Plus-Sepharose bead preclear column (5-ml bed volume), followed subsequently by a mAb 2H11 affinity column (also 5-ml bed volume) using pump (Bio-Rad) control with a flow rate of ϳ0.5 ml/min. To remove any contaminating proteins bound nonspecifically to the mAb 2H11 resin, an extensive washing step was performed with a pH step gradient from pH 7.5 to pH 6.8 buffer (20 mM Tris/citric acid (pH 6.8), 0.5 M NaCl, and 5% glycerol) using at least 50 times the column bed volume. The efficiency of the washing was monitored by observing the A 280 of the flow-through supernatant, with washing continued until the absorbance fell below 0.020. Next, the bound N15 TCR protein was eluted with low pH buffer (50 mM citrate, 20 mM Tris, 0.5 M NaCl, and 10% glycerol (pH 4.0)), and peak fractions were immediately adjusted to pH 7.2 using 1 M Tris-HCl (pH 9.5), pooled, and dialyzed against TBS (20 mM Tris and 150 mM NaCl (pH 7.2)) with 10 mM EDTA overnight at 4°C. The dialyzed protein was concentrated to 2 ml with an Amicon Centriprep-10 concentrator and loaded onto a 1.6 ϫ 60-cm Superdex 200 column (Pharmacia) equilibrated with TBS.
Microchemical Analysis of Purified Proteins-Protein concentration of N15 samples was determined using the specific extinction coefficient at 280 nm (0.69), which was calculated by the method of Gill and von Hippel (26), as well as by using either the Bio-Rad version of the Bradford dye binding assay or the micro-BCA protein assay (Pierce). Amino-terminal microsequencing was carried out on 450 pmol of N15 protein on an Applied Biosystems Model 477A pulsed liquid-phase protein sequenator. For C-terminal amino acid sequencing, we employed a Hewlett-Packard C-terminal sequencer using the C-terminal routine 2.0 with the samples spotted on a Zytex membrane. Protein analysis was performed on denaturing gels using 10 or 12% SDS-PAGE by the Laemmli method (27). Prior to electrophoresis, nonreducing samples were heated at 65°C for 10 min or at 100°C for 5 min, and reducing samples were boiled for 5 min in a loading buffer containing 100 mM dithiothreitol. Gels were stained with Coomassie Brilliant Blue R-250, and the protein bands were quantitated by laser densitometry (Pharmacia Biotech Inc.). Isoelectric focusing (IEF) of native gels was carried out with the PhastGel system using precast gels (pI range 3-9) (Pharmacia), while native gradient gels were 8 -25%.
Glycan Analysis by Electrospray Ionization Mass Spectrometry-For mass spectrometry, the purified N15 protein was treated with endoglycosidase H (endo-H) (Boehringer Mannheim) following the manufacturer's instructions. The released glycan and protein was methylated, which also releases and methylates possible O-linked glycans (28,29).
The methylated glycans were chloroform-extracted, dried under vacuum, and dissolved in a methanol/water solution (6:4, v/v) containing 0.25 mM sodium hydroxide. Electrospray solutions were injected into the mass spectrometer at a rate of 0.75 liter/min through a stainless steel hypodermic needle. The voltage differences between the needle tip and the source electrode were 3-4 kV. For collision-induced dissociation studies, singly or multiply charged precursor ions were selectively transmitted by the first mass analyzer and directed into the collision cell containing argon at roughly 2 millitorrs with acceleration voltages of 20 -60 V, and the fragments were scanned in a quadruple mass analyzer after the collision cell as described (28).
Immunoreactivity Analysis-The reactivities of N15 and endo-Htreated N15 (N15 H ) with a panel of mAbs were examined using a sandwich enzyme-linked immunosorbent assay procedure. In brief, 5 g/ml H28 or H57 was coated on Immulon II plates at room temperature for 2 h and blocked with 1% bovine serum albumin in boratebuffered saline (pH 8.3) at room temperature for 2 h. Then, 50 ng of N15 or N15 H was added to each well overnight at 4°C. Subsequently, 50 l of various hybridoma cell supernatants was added for 2 h at room temperature, and the reaction was developed with horseradish peroxidase-conjugated secondary antibodies. Each comparison of a mAb reactivity with N15 and N15 H was expressed as the ratio of A 490 with N15 H to A 490 with N15.
Preparation of TCR Derivatives-Any potentially exposed free sulfhydryl group of Cys-183 in the C␤ domain was blocked either with N-ethylmaleimide or iodoacetamide. Using [ 14 C]iodoacetamide (Amersham Corp.), radioactivity was found to be incorporated into the ␤ chain in HTBS (10 mM HEPES, 10 mM Tris, and 100 mM NaCl (pH 8.0)) after Յ2 h of incubation at 27°C (data not shown). Since both chemical modifications resulted in proteins with similar crystallographic behavior, only the iodoacetamide results are presented herein. Modified N15 (iN15) and N15 s were digested with bovine thrombin (Calbiochem) at 27°C in HTBS for Յ2 h (10 -20 mg/500 units of enzyme) to yield iN15⌬ and N15 s ⌬, respectively. For iodoacetamide modification, the chemical (10:1 molar ratio) was added just before thrombin so that blockage and cleavage were carried out at the same time. Subsequently, after the protein solution was buffer-exchanged into 50 mM NaAc (pH 5.5), iN15⌬ or N15 s ⌬ was digested with endo-H for 2 h at 27°C to yield iN15⌬ H and N15 s ⌬ H , respectively. The scale was normally 5-20 mg in 2 ml (5 mg/0.1 unit of enzyme).
Thrombin-digested or double-digested (thrombin and endo-H) N15 or N15 s was purified on a Superdex S-75 column (1.6 ϫ 60 cm; Pharmacia) in 150 mM (NH 4 )Ac at 0.5 ml/min. The column was run at 4°C under the control of FPLC, and 84 ϫ 1-ml fractions were collected after 36 ml. An aliquot of the peak fractions was analyzed by nonreducing SDS-PAGE.
Generation of the H57 Fab Fragment-Forty to one-hundred milligrams of H57 antibody was purified from Cell Pharm supernatants on an 8-ml protein A column (Pharmacia). The antibody was digested with immobilized papain (Pierce) overnight at 37°C, and the undigested material and the Fc fragment were separated from the Fab fragment by the protein A kit (Pierce) according to the manufacturer's instructions. The binding buffer was removed by desalting with a Sepharose G-25 column (75 ml). The purity was checked by both reducing and nonreducing SDS-PAGE as well as native Phastgel.
Titration of TCR Derivatives and the H57 Fab Fragment to Form the Bimolecular TCR-Fab Complex by Native Gel Analysis-Based on the protein concentrations determined at 280 nm and/or by micro-BCA protein assay in an enzyme-linked immunosorbent assay format, increasing amounts of one component were titrated against constant or decreasing amounts of the other at various points (e.g. 3:1, 2:1, 1:1, 1:2, 1:3-1:6, v/v), and each mixture was incubated at 4°C for Ͼ3 h. The protein concentration was at a range of 1-10 mg/ml according to different batches. An aliquot of each was loaded onto the sample comb and analyzed on an 8 -25% native Phastgel. The individual components and the complex were visualized by Coomassie Blue staining. Preparative scale mixing of TCR and Fab components was carried out based on the ratio at which formation of a 1:1 (mol/mol) TCR-Fab complex was visualized by native gel analysis. The bimolecular complex was then concentrated to 10 -16 mg/ml, and the buffer was exchanged into 20 mM Mes (pH 6.0) and 0.025% NaN 3 at 4°C with a 10K spin unit (Filtron Technology Corp. or Millipore Corp.).
Crystallization Conditions and X-ray Diffraction-A hanging droplet (2 l) and a sitting droplet (4 -6 l) were used for crystallization by a vapor diffusion method (30). The droplet was a 1:1 (v/v) mixture of 10 -13 mg/ml TCR-Fab complex and the crystallization buffer. The same buffer of 1 ml was used as a reservoir solution for diffusion. The 1:1 ratio complex was carefully prepared by titration as described under "Results and Discussion." A sparse matrix method (31) was applied for screening crystallization conditions. Typically, ϳ100 different conditions were tried for each protein sample. The most promising condition was then refined for growing crystals of diffraction quality.
X-ray diffraction pictures were taken of a cluster of crystals (dimension of ϳ100 ϫ 50 ϫ 10 m) using an Elliot GX-13 rotating anode generator operated at 40 kV and 60 mA for CuK ␣ radiation. This was very useful for testing the diffraction behavior of various crystals obtained and permitted the identification of a suitable cryoprotectant buffer for freezing the crystals at Ϫ160°C. A 2.8-Å data set was collected on the X12C beamline at the National Synchrotron Light Source Brookhaven National Laboratory with a single frozen crystal of 150 ϫ 100 ϫ 25 m under Ϫ160°C on a MAR image plate.

Baculovirus-produced N15 Glycoproteins Are Resistant to
Endoglycosidase Treatment-Prior studies showed that expression of a representative TCR-leucine zipper fusion protein (N15) in a baculovirus system yielded the desired TCR ␣␤ heterodimers without formation of unwanted ␣␣ or ␤␤ homodimers (22). Although expression was adequate at 1 mg/liter and purity Ͼ90%, the N15 protein was unable to crystallize under a wide range of experimental conditions (data not shown). Because the ␤ chain was variably glycosylated, we attempted to remove N-linked glycans by treatment of the purified baculoN15 protein with endo-H assuming that glycan heterogeneity was detrimental to the crystallization efforts. As shown in Fig. 1A, however, even at an enzyme/substrate ratio of 5 g of protein to 1 milliunits of endo-H for a 16-h incubation, baculoN15 was resistant to endo-H digestion. This resistance of the baculoN15 glycans is not unexpected due to core fucosylation of a large fraction of the oligomannose oligosaccharides in the baculovirus expression system (32,33). BaculoN15 was also resistant to enzymatic deglycosylation with N-glycanase as well (data not shown).
Expression of the N15 TCR in Lec3.2.8.1 Cells-To obtain a form of TCR protein from which the glycan could be readily cleaved, we expressed the N15 TCR in the Chinese hamster ovary Lec3.2.8.1 cell derivatives that exclusively synthesize homogeneous high mannose glycans (24). For this purpose, the same N15 TCR ␣ and ␤ cDNAs encoding 30-amino acid-long leucine zipper sequences appended to the carboxyl termini of the ␣ and ␤ extracellular domains (via a flexible linker with a thrombin cleavage site) as used in the baculovirus vector were cloned into the pEE14-GS vector as pEE14GS-N15␣Base and pEE14GS-N15␤Acid, respectively and then to create pEE14 GS-N15␤␣ (Fig. 1C). The latter was transfected into Lec3.2.8.1 cells, and clones were selected by screening cell supernatants for immunoreactivity with anti-TCR C␣ and C␤ mAbs or the anti-leucine zipper mAb 2H11 as described under "Materials and Methods." In this way, Lec3.2.8.1 clones producing N15 ␣␤ heterodimers (N15) at 5-10 mg/liter were obtained.
As shown in Fig. 1 (A and B), the N15 protein, in contrast to baculoN15, could be readily deglycosylated after just a 30-min exposure to endo-H under the same experimental conditions used for baculoN15. SDS-PAGE analysis indicated that under nonreducing conditions, the size of the glycosylated heterodimer shifts from 80 -90 to 60 -70 kDa upon endo-H treatment. Moreover, under reducing conditions, the endo-H treatment results in a shift of the N15 ␣ band from 36 to 31 kDa and collapse of the three N15 ␤ bands at 41, 43, and 45 kDa to a single 39-kDa band. Note that protein sequence analysis of baculoN15 previously demonstrated that the closely spaced and more slowly migrating bands in the gel were N15 ␤-related and that the single rapidly migrating band corresponded to N15 ␣ (Ref. 22 and data not shown). Removal of several Nlinked glycans on each N15 subunit could account for this shift in apparent molecular mass. Potential N-linked glycosylation sites are found in the N15 ␣ subunit at Asn-21, Asn-179, and Asn-193 and in the N15 ␤ subunit at Asn-77, Asn-117, Asn-179, and Asn-228. Hence, distinct glycoforms of the ␣␤ heterodimer with variable occupancy of the N-linked sites on the ␤ chain are consistent with the SDS-PAGE analysis. However, if the glycan distribution were random and the resolution of the gel system sufficient, we would expect to detect five ␤ bands (corresponding to ␤ chains with 0 -4 occupancy of the N-linked sites). Given that there are only three ␤ bands visualized, a subset of these conformers/glycoforms is present. Such preferential sitespecific glycosylation has been reported previously for other proteins (34). Because ␣ subunit Asn-179 and ␤ subunit Asn-117 are predicted to lie in close proximity to one another on the surface of N15, 2 it is unlikely that both sites will be glycosylated on the same molecule. That three discretely sized ␤ bands but only one ␣ band exist as indicated by reducing SDS-PAGE analysis implies that glycosylation of the N15 ␤ subunit is more heterogeneous than that of the ␣ subunit.
Electrospray Ionization Mass Spectrometry Analysis of N15 Glycans-To examine whether the heterogeneity in the N15 ␤ chain resulted from differential utilization of potential N-linked glycosylation sites on individual N15 ␤ chains and/or heterogeneity in the nature of the glycans attached to the various N-linked sites, mass spectrometry analysis was performed on the endo-H-cleaved N15 glycan (Fig. 2). Electrospray mass spectrometry ionizes methylated carbohydrates by the adduction of alkali metal cations added to the solution. The m/z values of 1335 and 679 in the electrospray spectrum ( Fig.   2A) correspond to a molecular species of 1312 Da adducted with one and two sodium cations, respectively. A molar mass of 1312 Da is the expected mass of a methylated oligosaccharide with five hexose (mannose) residues and a single N-acetylhexosamine (N-acetylglucosamine) residue.
Collision-induced dissociation of methylated and alkali metal-adducted oligosaccharides was also performed (Fig. 2B) to identify high mannose branching isomers by a combination of glycosidic fragments, which convey information on the connection topology of the mannose residues, and ring opening fragments, which identify specific linkages. The lack of an ion fragment pair at m/z 445 (charge on the nonreducing end) and 912 (charge on the reducing end) excludes topologies in which two mannose residues are eliminated by cleaving a single glycosidic bond. The ion pair at m/z 709 (overlapped with a m/z 708 reducing end fragment) and 737 arises from dissociation of the core mannose pyranose ring and identifies a 6-linked branch containing three hexose residues. These structural features identify the specific Man 5 isomer shown in Fig. 2B as the only glycan on the N15 TCR. Hence, the differences in mobility of N15 ␤ subunits under reducing SDS-PAGE must be a consequence of differential utilization of N-linked sites. Given that the protein/glycan mixtures were also methylated and would have released methylated O-linked glycans, the lack of detectable O-linked glycan adducts of N15 ␣ or ␤ subunits implies that the N15 sTCR undergoes no O-linked glycosylation in Lec3.2.8.1 cells.
Unaltered Immunoreactivity of N15 H Compared with That of N15-To examine whether endo-H treatment of the N15 protein (N15 H ) altered its immunoreactivity relative to non-deglycosylated N15, a comparative enzyme-linked immunosorbent assay analysis (see "Materials and Methods") was performed using the previously described anti-C␤ mAb H57 (35), anti-V␤5.2 mAb MR9.4 (36), and N15 V␤-reactive mAbs N15R4, R7, R8, R13, R15, R22, R28, R34, R35, R46, R53, and R54 (22). Among the V␤5.2-specific mAbs, N15 R53 is clonotypic, being directed at the CDR3 region of the N15 ␤ chain. When each of these mAbs was tested, the N15 H /N15 reactivity ratio ranged from 0.52 to 1.02. This similarity argues that the immunoreactivity to mAbs is essentially unaltered upon deglycosylation with endo-H, consistent with the notion that N15 H maintains a native TCR structure comparable to N15. Thus, the intact glycans are not required to maintain TCR structure. Nevertheless, in contrast to Fab fragments, TCRs have multiple glycosylation sites, many of which are conserved, implying an important role for glycans in the function of the TCR and its physiology.
Large-scale Purification of N15 and N15 s -Having established that the N15 protein could be readily deglycosylated with endo-H, large-scale production and purification efforts were initiated. The Acid-p1/Base-p1 leucine zipper-specific mAb 2H11 was utilized for immunoaffinity purification of the N15 ␣␤ heterodimers, with the majority of the N15 protein eluting from the column with 50 mM citric acid (pH 4.0). Superdex 200 chromatography of the immunoaffinity-purified protein (Fig. 3A) revealed the majority of the N15 sTCR to be homogeneous, eluting with V e ϭ 70.3 ml. Approximately 20% of the N15 protein exists in oligomeric form eluting in two distinct peaks; the first peak contained large aggregates eluting earlier than the protein marker ferritin (440 kDa), while a second peak consisted of smaller N15 oligomers, perhaps dimers and tetramers, eluting between aldolase (160 kDa) and ferritin (440 kDa). SDS-PAGE analysis of an aliquot from the peak Superdex 200 fraction at V e ϭ 70.3 ml of two representative runs is shown under nonreducing and reducing conditions, respectively (Fig. 3B). The major protein migrates with an apparent molecular mass of 80 -90 kDa under nonreducing conditions. A band whose mobility is slower than that of the 112-kDa marker represents trace contamination with disulfide-linked N15 dimers in the purposefully overloaded gel. The several discrete bands between 36 and 45 kDa represent noncovalently linked N15 ␣␤ heterodimers. Upon reduction, all N15 protein runs as four bands: an N15 ␤ triplet at 41, 43, and 45 kDa and a single 36-kDa ␣ band.
Because we suspected that the unpaired cysteine at residue 183 within the ␤ chain constant region might create disulfidelinked N15 dimers and further complicate biochemical purification by fostering disulfide exchange, we mutated this cysteine to a serine, the corresponding residue found in Ig C H1 domains (37). cDNAs encoding the mutated TCR (N15 s ) were then expressed in Lec3.2.8.1 cells. Fig. 3A shows that the chromatographic profile of N15 s is very similar to that of N15. However, there is a substantial reduction in disulfide-linked N15 dimers observed by SDS-PAGE under nonreducing conditions (Fig. 3B). Moreover, the amount of non-disulfide-linked heterodimer as judged by nonreducing SDS-PAGE is dramatically reduced (Fig. 3B). Because attempts to crystallize N15 and N15 s proteins under a range of conditions were unsuccessful, we next generated a series of TCR derivatives and assessed whether they were more suitable crystallization candidates.
Generation and Purification of TCR Derivatives-To generate sTCR derivatives, both N15 and N15 s were utilized as sources of starting proteins. In the case of N15, the accessible SH group of ␤ chain Cys-183 was derivatized with iodoacetamide (iN15) to prevent disulfide exchange. N15 and N15 s were thrombin-cleaved to yield iN15⌬ and N15 s ⌬, respectively, or additionally digested with endo-H, resulting in iN15⌬ H and N15 s ⌬ H , respectively. From the overlay of chromatograms depicted in Fig. 4A, it is obvious that the double-digested iN15⌬ H and N15 s ⌬ H derivatives have a smaller hydrodynamic volume than the thrombin only-digested iN15⌬ and N15 s ⌬ derivatives, with V e ϭ 57.7 ml versus 55.5 ml, respectively. When the fractions across the major peaks were analyzed by nonreducing SDS-PAGE (Fig. 4B), we observed protein bands above the major one spreading from the earlier fraction into the main peak; this was most evident in the iodoacetamide-treated derivatives iN15⌬ and iN15⌬ H . Furthermore, there were nondisulfide-linked heterodimers comigrating with the disulfidelinked equivalents in the iN15⌬ and iN15⌬ H samples. On the other hand, the overlay of the chromatograms in Fig. 4A indicates that the hydrodynamic properties of the N15 and N15 s proteins were similar after comparable treatment. More important, as shown in Fig. 4B, the overall complexity of the N15 s derivatives was less, including virtual elimination of the nondisulfide-linked heterodimers.
The N termini of the N15 s ⌬ H heterodimeric protein were sequenced through amino acid residue 14; signal(s) generated at each cycle matched the predicted N15 ␣ and ␤ sequences. This result shows that there have not been any unexpected proteolytic events in the TCR subunits during thrombin cleavage and/or endo-H digestion. Furthermore, C-terminal sequencing of N15 s ⌬ H unambiguously defined arginine in cycle 1, consistent with the fact that thrombin cleaves within the flexible linker regions of each subunit after arginine and before glycine. In agreement with this result, mass spectrometry analysis of the leucine zipper peptide showed a mass (1373.3 and 1375.9 triple charged state) essentially identical to that of the expected cleavage product (1373.7 and 1376.2) with no detectable heterogeneity. The various N15 and N15 s derivatives each underwent a crystallization trial. However, none of these sTCR derivatives yielded crystals.
pI Heterogeneity of TCR Derivatives-The above result is interesting in view of the lack of N-or C-terminal heterogeneity in N15 s ⌬ H and the homogeneous size of the N15 s ⌬ H protein in the peak fractions as assessed by SDS-PAGE (Fig. 4B, panel d).
To address whether there might be heterogeneity within the TCR components by other criteria, native IEF gels were run, and the various TCR derivatives were compared on the basis of charge. As shown in Fig. 5, at least three major pI bands and several minor bands are observed within the 6.5-8 pI range for N15 and N15 s . Moreover, this charge heterogeneity is observed in purified N15 s ⌬, indicating that the heterogeneity is not a function of the leucine zipper sequence, but rather is intrinsic to the TCR itself. However, note that removal of the leucine zipper sequence results in a shift of the protein pI to the 5.0 -6.5 pH range. Although not shown, N15 s ⌬ H is identical to N15 s ⌬ in displaying the same range of pI heterogeneity. Collectively, these data show that the pI heterogeneity resides within the TCR protein itself.
Several explanations might account for this rather striking charge heterogeneity. One possibility is that the N15 TCR can exist in several stable conformational states, perhaps reflecting variability in the quarternary structure of the heterodimer. Alternatively, differential occupancy of N-linked glycosylation sites (with GlcNAc 2 -Man 5 for N15 s or with GlcNAc for N15 s ⌬ H ) on the ␤ subunits may alter the surface charge within the N15 TCR preparation. This possibility seems more remote, however, in view of the similar extent of heterogeneity when comparing N15 s ⌬ and N15 s ⌬ H , given that only one GlcNAc is attached to the N-linked sites of the latter. Furthermore, that pI heterogeneity is not unique to the N15 TCR is clear from analysis of the unrelated N26 TCR specific for VSV-8/K b (data not shown).
Complex Formation between sTCR Derivatives and the H57 Fab Fragment-Given the failure of N15 TCR proteins to yield useful crystals on their own and in view of several independent observations (38) showing that Fab fragments can stabilize proteins by offering new molecular surfaces for crystallization, we next complexed N15, N15 s , and their derivatives with the H57 Fab fragment. The hamster H57 mAb is specific for the mouse C␤ TCR constant region. To this end, the Fab fragment was generated by immobilized papain, purified, and tested by titration analysis with the TCR on an 8 -25% native Phastgel to observe a complex formation. By maintaining the Fab fragment amount constant and titrating in an increasing volume of N15 s ⌬ H , it can clearly be seen that a new band forms corresponding to the N15 s ⌬ H -Fab complex (data not shown). In representative experiments, full complexation is evident at an N15 s ⌬ H /Fab ratio of 3:1 (v/v) since neither Fab nor N15 s ⌬ H bands are detected. TCR excess occurs as the amount of N15 s ⌬ H increases relative to the Fab fragment, e.g. at a ratio of 4:1 or higher.
Crystallization of TCR Derivatives with the H57 Fab Fragment-Once the ratios of TCR to the Fab fragment giving a molar complex formation were determined, large-scale mixing of TCR and the Fab fragment was performed. The TCR-Fab complex was then concentrated to 10 -13 mg/ml, and crystallization trays were set up. Table I  FIG. 5. Native IEF gel of sTCR derivatives. N15 TCR and derivative proteins were compared on an IEF Phastgel (pH 3-9), and the IEF gel was stained with a silver stain kit. The calibration kit markers (pH 3-10) are indicated as pI standards (pI Stds.). N15, N15 s , unpurified N15 s ⌬, and Superdex 75-purified N15 s ⌬ are shown. yielded crystal growth over quite a broad pH range (5.5-7.0). These crystals tend to cluster. At low pH, the flower-like fine crystals appear first; later clusters of wedge-shaped single crystals are produced. To date, the largest crystals measure 400 ϫ 200 ϫ 30 m. These crystals demonstrate a strong polarization of light. Single crystals can diffract to beyond 2.8 Å, with mosaicity of 0.2-0.3°at Ϫ140°C. SDS-PAGE analysis and silver staining of both these crystals confirmed the presence of Fab and TCR components (data not shown). A native data set of N15 s ⌬ H -Fab [H57] crystals has been collected at Brookhaven National Laboratory from one single frozen crystal grown at pH 5.5. It belongs to p2 1 space group with a ϭ 74.5 Å, b ϭ 122.1 Å, and c ϭ 111.0 Å and ␤ ϭ 108.2°. There are likely two complexes per asymmetric unit, yielding a Matthews coefficient (V m ) of 2.48 (solvent content of ϳ50%). The structure determination is in progress.
Implications-Very recently, two crystal structures of isolated fragments of TCR molecules have been reported. One study involved x-ray analysis of an unglycosylated TCR ␤ chain derived from a murine TCR specific for a hemagglutinin peptide of influenza virus (hemagglutinin-(110 -120)) restricted by major histocompatibility complex class II I-E d (9). The other structural study involved a V␣⅐V␣ homodimer specific for a myelin basic protein nanopeptide restricted by I-A u (10). Collectively, five differences are evident between Fab fragments and TCRs. 1) V␤ and C␤ domains are in intimate contact in the crystal structure, with ϳ830 Å 2 of interface. This area is two to three times that of the corresponding region of V H -C H1 interaction and, as such, implies that the ␤ chain is more rigid than the antibody V H -C H1 chain in the Fab fragment. 2) The surface of the C␣ domain expected to contact the C␤ domain is polar, carrying a net positive charge in contrast to the known hydrophobic C H -C L and V H -V L interfaces and the predicted hydrophobic V␣-V␤ interface.
3) The complementarity determining region loops of the V␤ domain represent a new set of conformations not found in antibody hypervariable loops. 4) The C␤ domain contains an insertion with respect to an Ig constant domain between residues 219 and 232, forming a solvent-exposed loop on the external V␤ face, perhaps facilitating contacts between the TCR heterodimer and the CD3 signaling components. 5) In the crystallographically resolved TCR V␣⅐V␣ homodimer, the topology of the strands is such that the CЉ strand is translocated from the AGFCCЈ face to the BED face of the ␤-barrel.
Given the unique features of the TCR fragments, it is not unexpected that fundamental differences in behavior of the complete TCR ␣␤ heterodimer would become evident. For ex-ample, the lack of stable subunit association may be a consequence of differences at the V␣-V␤ interface and/or the C␣-C␤ interface relative to the corresponding antibody module. Our results show for the first time that heterodimer formation can be forced through the use of a leucine zipper motif and that after such pairing, the leucine zipper can be enzymatically removed, leaving the heterodimer intact. Presumably, the disulfide bond formed between cysteines on C␣ and C␤ domains stabilizes the TCR heterodimer despite a weak intrinsic association between ␣ and ␤ subunits.
The rather striking pI heterogeneity of the TCR ␣␤ derivatives even in the absence of glycan uncovered herein is most likely a reflection of different TCR conformers. Such conformational mobility may arise from floppiness within the ␣ subunit itself or at the interface between V␣ and V␤ or C␣ and C␤ modules. This potential mobility may be important in TCR signal transduction. Our results further indicate that neither intact glycans nor the invariant unpaired Cys-183 in the C␤ domain is required for stability of the TCR structure.
Finally, from our current results, we conclude that altering the TCR molecular surface and/or restricting the mobility of TCR conformers through ligation of the Fab fragment to the TCR is important for crystallization of TCRs such as N15. If this notion is correct, then ligation of a TCR with antigen/major histocompatibility complex or a superantigen may offer alternative strategies capable of accomplishing the same goal. From a practical perspective, the use of a Fab fragment specific for all TCR ␤ constant regions may be a useful way to overcome crystallization difficulties with TCRs alone in situations where superantigen binding or peptide/major histocompatibility complex binding is not feasible.