Structural and Functional Consequences of Altering a Peptide MHC Anchor Residue1

Gilbert J. Kersh,2,3* Michael J. Miley,2* Christopher A. Nelson,* Arash Grakoui,* Stephen Horvath,* David L. Donermeyer,* John Kappler, Paul M. Allen,* and Daved H. Fremont4*
To better understand TCR discrimination of multiple ligands, we have analyzed the crystal structures of two Hb peptide/I-Ek complexes that differ by only a single amino acid substitution at the P6 anchor position within the peptide (E73D). Detailed comparison of multiple independently determined structures at 1.9 resolution reveals that removal of a single buried methylene group can alter a critical portion of the TCR recognition surface. Signicant variance was observed in the peptide P5-P8 main chain as well as a rotamer difference at LeuP8, 10 distal from the substitution. No signicant variations were observed in the conformation of the two MHC class II molecules. The ligand alteration results in two peptide/MHC complexes that generate bulk T cell responses that are distinct and essentially nonoverlapping. For the Hb-specic T cell 3.L2, substitution reduces the potency of the ligand 1000-fold. Soluble 3.L2 TCR binds the two peptide/MHC complexes with similar afnity, although with faster kinetics. These results highlight the role of subtle variations in MHC Ag presentation on T cell activation and signaling. The Journal of Immunology, 2001, 166: 33453354. cell development and effector function are controlled by signaling through the TCR, a cell surface receptor that specically recognizes peptide/MHC complexes (pMHC)5 on the surface of APCs (reviewed in Ref. 1). TCR - and -chains are assembled from separate gene segments by somatic DNA rearrangement. In theory, this mechanism of gene assembly can produce an enormous number of TCR specicities, with sufcient diversity to bind almost any structure. However, only a fraction of the TCRs produced by this method have useful specicities (2); therefore, a selection process takes place during T cell development to ensure a functional T cell repertoire (3). Developing T cells undergo positive selection in the thymus to ensure that they are able to, in some manner, recognize pMHC. Furthermore, T cells also undergo negative selection in the thymus to ensure that each TCR will not be triggered by self peptides presented on self MHC molecules. TCRs that survive selection represent a balance between high specicity to avoid autoreactivity, and some degree
*Department of Pathology and Center for Immunology and Department of Biochemistry and Molecular Biophysics, Washington University School of Medicine, St. Louis, MO 63110; and Howard Hughes Medical Institute, Division of Basic Immunology, National Jewish Center, Denver, CO 80206 Received for publication August 17, 2000. Accepted for publication December 12, 2000. 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.
1 This work was supported in part by Grants AI24157 (to P.M.A.) and AI48784 (to G.J.K.) from the National Institutes of Health. D.H.F. is a recipient of a Burroughs Wellcome Fund Career Award. 2 3

G.J.K. and M.J.M. contributed equally to the work.

Current address: Department of Pathology, 7301 WMB, Emory University, 1639 Pierce Drive, Atlanta, GA 30322.
4 Address correspondence and reprint requests to Dr. Daved H. Fremont, Department of Pathology and Immunology, Campus Box 8118, Washington University School of Medicine, 660 South Euclid Avenue, St. Louis, MO 63110. E-mail address: fremont@immunology.wustl.edu 5 Abbreviations used in this paper: pMHC, peptide/MHC complex; Abu, 2-aminobutyric acid; APL, altered peptide ligand; Nle, norleucine; Nva, norvaline; RMSD, root mean square deviation.

of exibility to recognize self pMHC (4). It becomes apparent then, that to understand how the TCR functions, it is necessary to decipher in detail how this receptor is able to recognize multiple ligands, and how multiple ligands can induce different biological responses. Altered peptide ligands (APLs) represent a useful tool for studying differential recognition by the TCR. Numerous APLs have been identied by introducing single amino acid substitutions into the peptide sequences recognized by individual T cell clones (5, 6). The majority involve conservative amino acid substitutions at positions known to be accessible to the TCR (7). Comparison of the available crystal structures for a single TCR bound to different pMHC class I complexes shows that the TCR is able to accommodate different ligands by making only minor structural adjustments to the TCR-pMHC interface (8, 9). This leaves open the question, how do structural changes in the pMHC result in differential TCR recognition? The major parameter that the TCR uses to distinguish between ligands is now believed to be the dissociation rate of the TCR-ligand complex (10, 11). Several studies taken together show that for a given TCR, the interactions leading to a partial agonist response have a shorter t1/2 than the interactions leading to a full agonist response (1). Although the equilibriumbinding afnity of partial agonist ligands is often lower, it seems not to be a determining factor in T cell response (12). The current model is that small changes in the off-rate of the receptor-ligand interaction can lead to differences in the signals transduced by the TCR, resulting in differences in the biological response. In this study, we investigate the basis by which a TCR can discriminate between two peptides differing at only a single MHC anchor residue. While this substitution does not signicantly alter binding of the peptide to the class II molecule, it reduces the 3.L2 T cell response approximately 1000-fold. In this study, we show that the bulk T cell response for the two pMHCs are essentially distinct and nonoverlapping. To examine the structural basis for this differential T cell response, we have determined and compared multiple 1.9- crystal structures of each pMHC. In contrast to
0022-1767/01/$02.00

Copyright 2001 by The American Association of Immunologists

3346
Table I. Crystallographic data collection
Cell Dimensions Data Set Hb/IEk 1 Hb/IEk 2 Hb/IEk 3 Hb Merged Hb(D73)/IEk 1 Hb(D73)/IEk 2 Hb(D73)/IEk 3 Hb(D73) Merged a () 147.11 147.02 146.62 147.13 146.99 146.68 146.86 146.91 b () 57.05 56.95 57.18 57.01 57.03 56.81 56.84 56.85 c () 116.92 116.53 117.03 116.80 117.17 116.43 116.54 116.73 () 92.02 91.83 92.01 91.97 91.93 91.65 91.59 91.76 Reections Total 196,283 162,254 204,650 445,426 134,812 131,729 151,110 364,002 Completeness (%)

STRUCTURE OF AN ALTERED CLASS II pMHC

I/ (I) Total (1001.90 ) 14.42 12.42 13.45 19.8 16.62 15.54 9.01 13.5 Last shell (1.961.90 ) 2.71 2.42 2.53 3.2 2.39 2.89 2.09 2.45

Rmerge (%) Total Last shell (1001.90 ) (1.961.90 ) 5.7 5.8 6.8 8.3 4.2 5.2 9.4 11.4 25.8 31.1 28.7 30.1 24.6 19.4 27.6 29.6

Total Observed (1001.90 ) 74,078 66,294 67,428 76,496 63,993 54,315 58,289 72,521 95.2 86.9 87.5 98.3 83.3 71.6 76.9 94.4

Last shell (1.961.90 ) 86.8 64.2 57.6 92.6 58.1 54.6 25.7 75.6

results from a similar study with class I pMHC, in which significant movements were seen in the MHC class I molecule as opposed to the peptide (13), our comparison reveals that conformational differences are localized to the peptide P6 substitution (E73D) and the adjacent P7 and P8 residues. Although small, we show that these differences are statistically signicant and localized to regions directly implicated in TCR recognition. To further investigate these structural ndings, we characterized the recognition of peptides with nonnatural P8 residue substitutions, the results of which correlate with our crystallographic interpretations.

Glutamax, and 50 g/ml gentamicin in 200 l total volume in a 96-well tissue culture plate with the indicated concentration of peptide. Culture wells were pulsed at 48 h with 0.4 Ci [3H]thymidine and harvested 18 24 h later.

Stability of the peptide/I-Ek complexes

The I-Ek-positive B cell lymphoma CH27 was prepulsed with either 31.6 M Hb(D73) or 0.316 M Hb for 2 h at 37C and at a concentration of 1.5 106 cells/ml. The cells were then washed and incubated at 37C in a single well of a 24-well plate. At 0, 1, 2, 4, 6, or 10 h later, the cells were washed and added at 2 104 per well to a 96-well plate containing the T cell hybridoma YO1.6 at 1 105 per well. Supernatants were removed at 22 h, and the U/ml of IL-2 was determined using the indicator line CTLL-2. For the experiment shown in Fig. 2B, the IL-2 produced when prepulsed CH27 were added to the YO1.6 cells immediately after washing (time 0) was as follows: Hb, 893 U/ml; Hb(D73), 2010 U/ml.

Materials and Methods

Peptides
Peptides were synthesized on a Rainin Symphony Multiplex synthesizer (Woburn, MA) using standard F-moc chemistry. The peptides were puried to homogeneity by reverse-phase HPLC, and their composition was conrmed by mass spectrometry and amino acid analysis (Washington University Mass Spectrometry Facility, St. Louis, MO). The names and sequences of the peptides used in the study are as follows: Hb, residues 64 76 GKKVITAFNEGLK; Hb(D73), Asp for Glu substitution at P6; Hb(A75), Ala for Leu substitution at P8; Hb(Nle75), norleucine for Leu substitution at P8; Hb(Nva75), norvaline for Leu substitution at P8; Hb(Abu75), 2-aminobutyric acid for Leu substitution at P8.

Peptide-binding assay
Relative binding strength of Hb and Hb(D73) peptides to I-Ek was determined essentially as described (14). Briey, a soluble form of I-Ek was made by cleavage from the surface of Chinese hamster ovary cells transfected with an expression plasmid encoding a glycophospholipid-linked form of I-Ek (a generous gift from Mark Davis (Stanford University, Palo Alto, CA)). Puried I-Ek was then incubated for 72 h with a 125I-labeled standard peptide (AAAYGKKVITAFNEGLK) in the presence or absence of various concentrations of the test peptide. pMHC were separated from unbound peptide using Bio-Spin 6 chromatography columns (Bio-Rad, Hercules, CA). The relative amount of bound standard peptide was determined using a gamma counter. The data are expressed as a percentage of the counts obtained in the absence of competitor peptide.

T cell responses
Proliferation of primed lymph node T cells was assayed in the following manner. CE/J mice were immunized s.c. with 20 nmol of either Hb or Hb(D73) peptides emulsied in CFA. Ten days later, the draining lymph nodes were removed and a single cell suspension was placed in culture using RPMI 1640 media supplemented with 1% normal mouse serum, 5 10 5 M 2-ME, 1 mM Glutamax (Life Technologies, Gaithersburg, MD), and 50 g/ml gentamicin. A total of 0.5 106 cells was placed per well of a 96-well plate in the presence of Hb or Hb(D73) peptides. Culture wells were pulsed at 48 h with 0.4 Ci [3H]thymidine and harvested 18 24 h later. The proliferation of the primary 3.L2-transgenic T cells was measured as described previously (7). Briey, 0.5 106 splenocytes were cultured in RPMI 1640 supplemented with 10% FCS, 5 10 5 M 2-ME, 1 mM

Preparation of soluble pMHCs

The method used for production and purication of the soluble ecto-domains of I-Ek with covalently bound peptides has been described previously (15). Briey, a baculovirus transfer vector was prepared that contained the genes for the I-Ed -chain and the I-Ek -chain. The I-Ek -chain was modied to contain an N-terminal covalent linkage that includes a exible linker and either the Hb or Hb(D73) peptide. The transfer vectors were used to produce high titer stocks of baculovirus from Sf9 cells. The

Table II. Crystallographic renement statistics

Rcrystal (%) Total (1001.9 ) Last shell (1.961.90 ) Rfree (%) Total (1001.9 ) Last shell (1.961.90 ) RMSD Coordinate Error Experimentala Data Set Bonds Angles () () M1 () M2 () Estimated Sigma () Luzzati () Average B-Factor (2)

Hb/IE 1 Hb/IEk 2 Hb/IEk 3 Hb Merged Hb(D73)/IEk 1 Hb(D73)/IEk 2 Hb(D73)/IEk 3 Hb(D73) Merged

21.9 22.4 22.0 22.2 22.9 22.7 22.0 23.2

29.2 38.1 32.9 29.6 34.5 32.0 35.5 36.0

25.8 26.7 26.0 25.4 26.4 27.9 27.4 27.0

32.7 38.7 36.5 31.8 36.3 35.2 32.0 40.5

0.006 0.006 0.006 0.006 0.006 0.006 0.006 0.006

1.29 1.28 1.29 1.37 1.30 1.30 1.32 1.37

0.12 0.17 0.15 0.14 0.22 0.18 0.18 0.19

0.10 0.16 0.13 0.13 0.20 0.19 0.18 0.19

0.24 0.41 0.31 0.23 0.35 0.32 0.41 0.34

0.25 0.27 0.26 0.25 0.27 0.28 0.28 0.28

34.5 35.1 35.1 37.8 39.0 39.1 38.6 41.3

a Experimental coordinate error is calculated as the RMSD of a given molecule in the asymmetric unit (M1 and M2) compared to its equivalent in the merged structure. The coordinate error for each merged structure is the average of the calculations.

The Journal of Immunology

large scale infection of Sf9 cells produced the desired molecules as soluble secreted proteins in the medium. These proteins were puried from the Sf9 supernatants using immunoafnity columns made from the I-E -specic Ab 14.4.4s. This was followed by gel ltration over a HiPrep 16/60 Sephacryl S-200 column (Pharmacia). Fractions were analyzed by ELISA, and those containing I-Ek heterodimers were pooled, concentrated, and stored in 10 mM HEPES, pH 7.5. The identities of the puried proteins were conrmed by N-terminal sequencing (Midwest Analytical, St. Louis, MO).

3347 Hb-specic T cells, but in many cases causes the peptide to act as a TCR antagonist (5, 6). In fact, for mice immunized with Hb peptide, the recall proliferative response of bulk lymph node T cells in vitro requires 100- to 1000-fold more peptide if Hb(D73) is used as the recall Ag than if the Hb is used (Fig. 1A). The difference in biological response initially suggested that the Glu73 residue might be a TCR contact residue, or alternatively that the Hb(D73) substitution hinders binding of the peptide to the I-Ek molecule. As demonstrated below, neither of these initial possibilities has proven true, and a more elegant molecular mechanism accounts for the ability of TCRs to discriminate between the two ligands. Two different methods indicate that the Hb(D73) peptide binds to I-Ek equally as well as the Hb peptide. First, Hb and Hb(D73) are equally effective at competing off a radiolabeled index peptide from a soluble form of the I-Ek molecule (Fig. 2A). This result suggests that the two peptides have a similar equilibrium-binding afnity for I-Ek. Second, we have compared the persistence of complexes at the cell surface formed from either Hb/I-Ek or Hb(D73)/I-Ek. This was done by pulsing APCs with a dose of Hb or Hb(D73) sufcient to induce an intermediate response from a T cell hybridoma and then incubating the APCs at 37C for different time periods. After washing, the amount of pMHC remaining was assayed using a T cell hybridoma. Although not a direct measure of the off-rate of the two peptides, it is expected that the rate of decrease in the response of the T cell hybridoma is in direct proportion to the off-rate of the peptide from the I-Ek molecule. The rate of disappearance of the two complexes was similar when measured by the loss of T cell recognition (Fig. 2B), suggesting that these ligands have similar stability on the cell surface. The Hb(D73) peptide can form an effective ligand when presented on I-Ek. In addition, most T cells specic for Hb(D73) are weakly responsive to Hb. For example, in animals immunized with the Hb(D73) peptide, the recall proliferative response of bulk lymph node T cells in vitro requires 100- to 1000-fold more Hb

Crystallization and x-ray data collection

Protein solutions of Hb/I-Ek and Hb(D73)/I-Ek were concentrated to an OD280 of approximately 10 in 10 mM HEPES, pH 7, and 5 mM sodium azide. Crystals were produced in hanging drops by vapor diffusion at 20C against wells lled with 15% polyethylene glycol 4000 (Fluka, Buchs, Switzerland), 15% 2-propanol, 300 500 mM ammonium acetate, and 100 mM citrate, pH 4.8. Diffraction quality crystals appeared within 48 h and were cryoprotected just before ash cooling through the addition of a mixture of 20% polyethylene glycol 4000, 20% ethylene glycol, 10% glycerol, 100 mM sodium citrate, and 100 mM ammonium acetate. All crystals belonged to the centered monoclinic space group C2 and had similar cell dimensions (Table I). Three data sets were collected at the Advanced Photon Source (APS) beamline 19-ID on a charged coupled device detector for each of the two proteins of interest (Hb/I-Ek and Hb(D73)/I-Ek). Data were indexed and processed using Denzo and Scalepack (16).

Model building and renement

The coordinates of the 2.3- rened Hb/I-Ek complex (RCSB code 1IEA) (17) were used as the initial model for the renement of the high resolution Hb/I-Ek structures. Rigid body renements were conducted with CNS (18) using the platform domains and the membrane-proximal Ig domains as separate objects. Additionally, CNS was used to execute multiple rounds of renement on the Hb/I-Ek models, which included temperature factor renement, conjugate gradient minimization, and electron density map generation. The models were rebuilt in O (19) using 2Fo-Fc, Fo-Fc, and simulated annealing omit maps. A total of six independent Hb/I-Ek models were generated from three unique data sets, each with two molecules in the asymmetric unit. Similarly, six Hb(D73)/I-Ek models were built starting with a modied Hb/I-Ek model in which the appropriate mutation in the peptide, E73D, was made. Renement was conducted as above with the addition of phased difference Fourier maps between Hb(D73)/I-Ek and Hb/ I-Ek data sets. These maps were used to probe for shifts in atomic positions that result from the E73D substitution. Final coordinates for both pMHCs were generated using merged wild-type and mutant data sets (Table II). Coordinates have been submitted to the Protein Data Bank (Hb/I-Ek and Hb(D73)/I-Ek PDB codes 1FNG and 1FNE, respectively). Each peptide/I-Ek structure contains 182 residues of I-E k ( 1- 182), 12 mouse Hb(6576) residues (P-4 to P9), 16 linker residues (P 1 to P 16) connecting to 185 of the natural I-E k ( 4- 188), and three carbohydrates N linked to Asn 78, Asn 118, and Asn 19. For all models, no attempt was made to build the peptide NH2-terminal regions (P-8 to P-4) or the COOH terminal 10 residues of either the - or -chains, as these regions appear highly disordered in the electron density maps. A sequencing error in His 177 was also corrected from the original I-Ek structure, which was built as Thr 177.

Results
Hemoglobin is a model Ag for TCR cross-reactivity hetThe murine hemoglobin molecule is composed of two erodimers. Allelic forms of both - and -chains exist in mice and allow for use of the hemoglobin protein as a foreign Ag (20). For example, CE/J mice (I-Ek, Hbbs) will produce a strong T cell response when immunized with hemoglobin protein prepared from CBA/J mice (I-Ek, Hbbd). This response was shown to be dependent on the presentation by I-Ek of a peptide derived from residues 64 to 76 of the minor form of the d allele of the hemoglobin -chain (hereafter referred to as Hb) (5). Numerous APLs have been described for Hb/I-Ek-specic T cells, and this system is ideally suited for biophysical studies of recognition of multiple ligands by the TCR. Of the many amino acid substitutions possible in the Hb sequence, one of the most interesting is the substitution of Glu73 with an Asp. This substitution (hereafter referred to as Hb(D73)) not only changes the peptide from a good to a poor stimulator of many

FIGURE 1. Lymph node T cells can distinguish between the Hb and Hb(D73) peptides. CE/J mice (H-2k, Hbbs) were immunized s.c. with either Hb (A) or Hb(D73) (B). Ten days later, the draining lymph nodes were removed and the proliferative responses to Hb and Hb(D73) were evaluated. In both cases, the response to the immunizing peptide is much stronger. f, In vitro stimulation with Hb; E, in vitro stimulation with Hb(D73).

3348

STRUCTURE OF AN ALTERED CLASS II pMHC face (8, 9, 2123). It seems likely then that the uniqueness of the Hb(D73)/I-Ek determinant does not result from a direct interaction of the Hb(D73) side chain with the TCR, but rather from an indirect change introduced upon substitution of the buried side chain. Hb(D73)/I-Ek differs from Hb/I-Ek in peptide but not MHC conformation To understand how the change from Glu73 to Asp73 could have such a profound effect on T cell response, we have determined the high resolution crystal structures of I-Ek complexed with covalently bound Hb and Hb(D73) peptides. Soluble forms of pMHCs were puried from baculovirus-infected insect cells. To increase the accuracy of the comparison, we set out to analyze three individual crystals of both pMHCs. A total of six independent data sets were collected for renement at 1.9 resolution (Tables I and II): three from Hb/I-Ek and three from Hb(D73)/I-Ek, each with two independent molecules per asymmetric unit. The resulting coordinates were used to compare the two pMHCs. This was done by superimposing the independent structures to one of the Hb/I-Ek complexes. The root mean square deviations (RMSD) of peptide and binding platform residues for each independent model are shown plotted vs the root mean squared positional displacements as calculated from the thermal parameter B (Sqrt[B/8 2]) (Fig. 3). Rened atomic B values represent a measurement of the displacement of an atom due to thermal motion or conformational disorder. The deviations of residues with low thermal displacements can be considered to be more signicant (24). Two identical structures would have a linear relationship between RMSD and Sqrt[B/8 2], as is observed for the Hb/I-Ek structures. The analysis clearly reveals that the most signicant statistical deviations between the two pMHCs occur in peptide residues. Peptide positions P6 (Glu73) and Asp73 and P8 (Leu75) have deviations far more signicant than any other residue, either in the rest of the

FIGURE 2. Hb and Hb(D73) peptides bind the I-Ek molecule similarly. A, Puried, detergent-solubilized I-Ek molecules (2.5 nmol) were loaded with a radiolabeled index peptide (25 pmol) in the presence of a titration of Hb peptide (f), or Hb(D73) peptide (E). The 100% value indicates the maximum amount of recovered radioactivity in the absence of competitor peptide. B, The I-Ek-positive B cell lymphoma CH27 was prepulsed with either Hb peptide (f) or Hb(D73) peptide (E), followed by washing to remove excess peptide. The prepulsed APCs were then incubated at 37C for the indicated time periods before being used to stimulate the T cell hybridoma YO1.6. IL-2 was measured in the YO1.6 supernatants as an indicator of the cellular response. The graph shows the decrease in response over time given as a percentage of the initial response. f, Hb peptide; E, Hb(D73) peptide. Although this hybridoma recognizes both Hb and Hb(D73), it requires 100 times more Hb(D73) peptide for full activation. The APCs were prepulsed with slightly more than the minimum amount of each peptide required for a maximal response. The concentrations used for prepulse were: Hb, 0.316 M; Hb(D73), 31.6 M. Although different concentrations of peptide were used, the rate that the responses decayed was found to be independent of the initial concentration. Using the responses obtained at 4, 6, and 10 h, the t1/2 of the complexes were calculated to be: Hb/I-Ek, 6 h; Hb(D73)/I-Ek, 5.1 h. It is possible that the decay of the response represents internalization of the pMHC, and not peptide release. If this is true, then both peptides have very slow dissociation from I-Ek, with t1/2 of greater than 5 h. We conclude that both the Hb and Hb(D73) complexes are sufciently stable on the cell surface to induce productive T cell responses.

peptide than Hb(D73) peptide (Fig. 1B). The converse of this experiment is also true; lymph node T cells from animals immunized with the Hb peptide require 100- to 1000-fold more Hb(D73) peptide than Hb peptide to respond in vitro (Fig. 1A). Clearly, these results demonstrate that the Hb(D73) peptide is able to form an effective ligand when complexed with I-Ek. Furthermore, the determinant formed by Hb(D73) must be, at least in part, unique as TCRs specic for either Hb(D73) or Hb can efciently discriminate between the two ligands. Previously, we have described the 2.3 crystal structure of a soluble form of the I-Ek molecule with a covalently attached Hb peptide (17). This structure revealed that Glu73 ts into the P6 pocket of the I-Ek molecule, adopting the role of a traditional MHC anchor residue. In contrast to the P5 (Asn72) and P8 (Leu75) residues, which have solvent accessible side chains that point away from the pMHC surface, the side chain of Glu73 is not expected to directly interact with the TCR. Nevertheless, based on the available crystal structures of TCR/pMHC complexes, residues anking the Hb(D73) substitution should be integral to the recognition sur-

FIGURE 3. The structures of I-Ek with covalently bound Hb or Hb(D73) differ signicantly at peptide positions 73 (P6), 74 (P7), and 75 (P8). For each residue, the root mean square value of the residue deviations (RMSD) from a Hb/I-Ek molecule is shown plotted vs their positional deviation (Sqrt[B/8 2]). RMSD values from the Hb/I-Ek structure are shown as lled squares in contrast to values from the Hb(D73)/I-Ek structure, which are shown as open circles. Three data sets were compared for Hb/I-Ek, and three for Hb(D73)/I-Ek. In these crystals, the asymmetric unit contains two separate pMHCs that were modeled independently. This analysis yielded 12 data points for each residue: six Hb/I-Ek and six Hb(D73)/ I-Ek. The RMSD values obtained for the P6, P7, and P8 residues of the Hb(D73) peptide are indicated. Although the normalized difference for Gly P7 are smaller than Asp/Glu P6 and Leu P8, they are signicant considering they are based on main chain movements alone. Residues in the nonpeptide-binding regions of I-Ek (the 2 and 2 domains) were not included in this analysis.

The Journal of Immunology

3349

FIGURE 4. Signicant changes in electron density between covalently bound Hb/I-Ek and Hb(D73)/I-Ek are clustered around the side chain at position 73 (P6) and along the peptide main chain between position 72 (P5) and 75 (P8). Models for the covalently linked Hb (yellow) and Hb(D73) (green) peptides have been superimposed and oriented such that the MHC-binding groove is depicted below the peptides, and residues 72 (P5) and 75 (P8) point up and away from the MHC surface. In both models, the oxygen atoms are colored red and the nitrogen atoms are colored blue. Overlaid onto these models is a representative difference electron density map (Fo,HbFo,D73) Hb contoured at 4 . The top panel shows positive density (yellow), dened as density that is present in the Hb/I-Ek structure, but not in Hb(D73)/I-Ek. Conversely, the bottom panel displays negative density (green), present in the Hb(D73)/I-Ek structure, but not in Hb/I-Ek. This difference map clearly shows the changes resulting from the Asp73 substitution: the side chain of residue 73 (P6) has moved, the main chain between residues 73 (P6) and 75 (P8) has shifted, and a water molecule has migrated in the P6 pocket. This traveling water is colored yellow in the top panel and green in the bottom panel. All the density differences between the two peptides were localized to a small region between residues 72 (P5) and 76 (P9). Similar analysis of difference electron density between the identical complexes yielded virtually featureless maps.

peptide, or in the 1 and 1 domains of I-Ek (Fig. 3). A smaller yet signicant movement is also seen for the P7 position (Gly74). The structural differences between the two pMHCs can be directly visualized by phased difference Fourier maps. For nearly isomorphous crystals, these maps reveal both positive and negative differences in electron density, which result from structural alterations. While no notable differences in electron density were observed between atoms of the I-Ek molecules, signicant differences were observed for the P6 substituted residue and its neighbors. A close-up view of one of these phased difference Fourier maps is displayed in Fig. 4. Superimposed on the electron density map are models of the two peptide structures. The Fo,HbFo,D73 map clearly indicates positive density for the Glu73 side chain (Fig. 4, top) and negative density for the Asp73 side chain (Fig. 4, bottom), indicating that the position of the carboxylate group in the P6 pocket is different for the two peptides. Among solvent accessible atoms, the major difference in density centers on the peptide main chain all along the region between the P5 and P8 residues. The structural differences also include the adoption of a distinct rotamer by Leu75 at P8 in Hb(D73)/I-Ek. Leu75 appears to have only a slight change in difference electron density because the shift in the peptide main chain moves Leu75 into a position in which it occupies approximately the same space as the alternate rotamer conformation observed in Hb/I-Ek. A summary and quantitation of the differences between the two peptides are presented in Fig. 5, with RMSD values plotted for

both side chain and main chain atoms. The side chains of residues at P6 and P8 are displaced by approximately 1.52 . The difference at P6 is due to the introduction of the new side chain at this position, and the difference at P8 is due partly to the shift in the main chain of the peptide, and partly to the new rotamer conformation adopted by the side chain of Leu75 in the Hb(D73)/I-Ek structure. Signicant main chain displacement can also be observed for the P6, P7, and P8 residues. These movements (RMSD 0.4 0.6 ) are clearly larger than those observed upon comparison of the Hb/I-Ek structures with themselves (Fig. 5, gray bars). The difference maps also show how the I-Ek-binding groove is able to accommodate the Asp73 side chain substitution. We previously observed an unusual cluster of carboxylate groups in the P6 pocket of the Hb/I-Ek structure (17) and proposed that this cluster of acidic residues would interact only at the acidic pH of the endosomal peptide-loading compartment. Once assembled, this structure would remain stable even after a shift to higher pH due to the solvent inaccessibility of the pocket. The cluster in the P6 pocket of Hb/I-Ek involves interaction of the Glu73 carboxylate group with Asp 66 of I-Ek. However, when Asp73 is present, there is still an acidic cluster, but the Asp73 carboxylate group now lies within 2.5 of Glu 11 from I-Ek (Fig. 6). A preference for Asp side chains to interact closely with Glu side chains has been previously shown (25), but it is not known whether this is the driving force

3350

STRUCTURE OF AN ALTERED CLASS II pMHC The structural changes alter the kinetics of the TCR-ligand interaction We and others have demonstrated that weakly stimulatory ligands display faster dissociation rates from the TCR than do full agonist ligands. Previously, we determined the kinetics of a particular Hbspecic TCR (3.L2) binding to soluble, covalently linked forms of both Hb/I-Ek and Hb(D73)/I-Ek (12). The Hb/I-Ek complex bound the 3.L2 TCR with a t1/2 of 10.8 0.09 s. In comparison, the t1/2 of the interaction between Hb(D73)/I-Ek and the 3.L2 TCR was only 7.5 0.22 s. Interestingly, with the Hb(D73) peptide a 3-fold increase in the association rate was observed, resulting in a higher equilibrium-binding afnity for the Hb(D73)/I-Ek complex with the 3.L2 TCR than seen with the Hb/I-Ek complex. Results similar to these have been observed in other receptor/ligand systems in which conservative mutations were introduced to the binding interface (27, 28). What is remarkable about our kinetic results is that they are attributable to the loss of a single solvent inaccessible methylene group.

FIGURE 5. The structural differences are restricted to a small region of the peptide. Changes occur in the position of main chain atoms between the P5 and P9 residues, and large differences are seen in the P6 and P8 side chains. RMSDs for both the main chain and side chain atoms were determined and averaged for both the Hb(D73)/I-Ek and Hb/I-Ek models. Error bars represent the minimum and maximum values observed for each residue. P1, P2, P3, etc. refer to positions within the peptide. The corresponding amino acids at these positions are listed underneath the gure. The glycine at P7 ( ) was not to be included in this analysis.

The P8 side chain is important for ligand recognition by the 3.L2 TCR TCR-docking models suggest that the Leu75 side chain at P8 should make important contacts with the TCR, and that substitutions at this position should affect T cell responses. We have tested this directly by examining the 3.L2 T cell response to Hb peptides substituted at P8 (Fig. 8). By using a series of P8 substitutions that differ in side chain length (Nle Nva Abu Ala), three conclusions can be drawn. First, this side chain is important for ligand recognition by the 3.L2 TCR. Ligands using Ala, Nle, or Abu at position 75(P8) stimulate weak responses (Fig. 8). Second, there is a preferred size for the side chain at this position: Ala is too small for good recognition, whereas Nle is too large, whereas Nva stimulates a strong response. Third, the position and conformation of the side chain in the unbound ligand seem to be important: the Leu in the Hb(D73) peptide has a different rotamer conformation and only weakly stimulates the 3.L2 T cells. The magnitude of the reduction is similar to having either a slightly smaller (Abu75) or larger (Nle75) side chain at P8. The results demonstrate the sensitivity of the TCR to subtle changes in size and orientation, and show that the P8 side chain is an important TCR contact for the 3.L2 T cell.

behind the positioning of Asp73. However, repositioning of the carboxylate group does appear to be the force driving the peptide mainchain shift in the Hb(D73)/I-Ek structure. In the P6 pocket, a water molecule shifts to ll the space created by replacement of the longer Glu73 side chain. This water forms new hydrogen bonds with the Asp73 side chain to stabilize the interaction (Fig. 6). It should be noted that Asp has not been found to occur at this position in any of the naturally processed I-Ek-bound peptides examined to date (26). Water molecules have been observed in the pockets of numerous pMHCs and appear to play a critical role in increasing the number of favorable interactions. This result probably represents a general mechanism used by I-Ek to bind peptides with side chains other than Glu73 at the P6 pocket. Fig. 7 presents a top view of the two pMHC surfaces, revealing the region of the molecule that interacts with TCR. The comparison of the complexes shows how little the MHC portion of the two molecules has been altered (Fig. 7A). The majority of the peptide atoms maintain their positions, but the signicant differences at P6-P8 can be observed. The different rotamer conformations of Leu75 at P8 can be seen, as well as the shift in the peptide main chain between the P6 and P8 residues. This shift is most evident for the P6 carbonyl oxygen. These shifts were normalized against each atoms positional deviation and mapped to the molecular surface of Hb/I-Ek (Fig. 7B). The magnitude of these shifts is colored coded with the largest normalized deviations mapped in bright red. Docking of these two pMHCs to TCRs from recently determined TCR/pMHC class I and II complexes (8, 2123) further highlighted the importance of these residues in TCR recognition. The P6-P8 residues are situated within the consensus TCR-accessible surface area (Fig. 7C). All the class I TCRs interacted with the P6-P8 region of the peptide and made signicant contacts with the P8 residue. More importantly, the modeled class II atomic contacts show that the TCR makes extensive contacts throughout the P6-P8 region of the peptide, including the carbonyl oxygen at P6 and Leu75 at P8 (Fig. 7C).

Discussion
In this study, we have examined TCR recognition of two pMHCs that differ only by a Glu to Asp substitution in the peptide sequence. The substituted side chain is directed into the binding groove of the MHC molecule and is unlikely to contact the TCR directly. Our goal was to understand the 1000-fold decrease in activity caused by this substitution. After ruling out the possibilities that the Hb(D73) peptide is a poor binder of I-Ek, or that Hb(D73) does not form a stable ligand, it became of interest to compare the atomic structures of the two complexes. The results are quite remarkable, in that a signicant structural variation results from the removal of a single methylene group. The position of the peptide main chain is retailored as is the rotamer of a solvent-exposed residue nearly 10 distal from the altered P6 pocket side chain. However, there are no signicant differences in the positions of any of the I-Ek Ag-binding platform residues that accommodate the APL. We will consider how these results might be important for T cell response, and discuss other factors that may also inuence the recognition.

The Journal of Immunology

3351

FIGURE 6. An altered hydrogen-bonding network accounts for the ability of I-Ek to bind the two different peptides. Here a schematic representation shows the hydrogen-bonding network at the bottom of the P6 peptide-binding pocket. The Hb peptide is given in the upper panel, while the Hb(D73) peptide is shown in the lower. Both hydrogen-bonding schemes require two of the three carboxylate groups in the P6 pocket to be protonated (Glu 11, Asp 66, and Glu73 or Asp73 of the peptide). Most of the hydrogen bonds are retained after the substitution, although some differences occur. The carboxylate group of the Asp73 peptide side chain does not interact directly with Asp 66, but instead rotates to form a tight hydrogen bond with Glu 11 ( 2.5 ). Hydrogen bonds also stabilize a water molecule that has moved to occupy the P6 pocket. In doing this, the water molecule loses a bond with Gln 9, but gains a new bond with Tyr 30.

A signicant change in position of the peptide main chain The Asp73 (P6) substitution causes an alteration in the peptide main chain between the P5 and P9 residues, which is dominantly shifted toward the peptide C terminus (Figs. 4 and 5). The root mean square displacement for these atoms is on the order of 0.4 to 0.6 . Although these differences are small, they are localized to a region of the pMHC that forms the TCR determinant (Fig. 7). These alterations are most likely due to the repacking of the P6 pocket. It is somewhat surprising that the 3.L2 T cell is so sensitive to this particular alteration, especially because substitutions at either of the nearby solvent-exposed side chains, P5 or P8, can have only modest effects. For example, a Thr for Asn72 substitution at P5 results in only a 50-fold decrease in activity (7). Similarly, the substitution of Nva for Leu75 (P8) results in only about a 10-fold decrease in activity, although Ala substitutions at either position result in ligands that do not induce any IL-2. The main chain adjustments in Hb(D73)/I-Ek occur along an approximately 15

length of the peptide, in a region that is contacted by TCR in all known TCR/pMHC complexes. We suggest that these documented differences in the free pMHCs could easily account for their distinct biological potency.

The leucine at P8 adopts a different rotamer conformation in Hb(D73) I-Ek Our results from six independently derived structures of both pMHCs clearly indicate that Leu75 at P8 adopts a different preferred rotamer conformation in the Hb(D73)/I-Ek complex compared with Hb/I-Ek. In the Hb(D73)/I-Ek complex, the alternative Leu75 rotamer is adopted to maintain the same hydrophobic contacts with Val 65 and Ala 68, which would otherwise be lost due to the repositioning of the peptide main chain. We have shown in this study that amino acid substitutions at P8 result in a modulation of T cell activity, with single methylene group alterations giving

3352

STRUCTURE OF AN ALTERED CLASS II pMHC

FIGURE 7. A, A top view showing the Hb(D73)/IEk-binding platform superimposed to that of the Hb/IEk-binding platform. Shown are residues that lose solvent-accessible surface area when either structure is docked with the available TCR coordinates from TCR/ pMHC complexes (8, 2123). Hb residues are shown in yellow, while Hb(D73) residues are displayed in green. Note the shift in positions of peptide P5-P8 main chain atoms, and the ip in the Leu75 rotamer conformation at P8. B, Atomic RMSD differences normalized by positional deviation (Sqrt[B/8 2]) between these two pMHCs are mapped to the molecular surface of Hb/I-Ek, which was rendered with GRASP (33). Areas colored in red represent signicant atomic variation that is solvent exposed. The majority of movement is localized to the P6-P8 region of the peptide, which is coincident with the TCR contact surface shown in the panel below. C, The colored regions represent the consensus solvent surface (34) predicted to be lost upon TCR ligation based on the docking of Hb/I-Ek to multiple class I- and class II-restricted TCRs. The pMHC atoms that are predicted to be in direct contact ( 4.2 ) with the class II TCR are colored blue. Atomic MHC class II TCR contacts are colored a lighter blue that the atomic peptide TCR contacts. The remaining peptide surface area is colored in yellow. Notice the majority of peptide contacts are localized to the P6-P8 peptide region.

The Journal of Immunology

3353 afnities, and yet there are signicant differences in both association and dissociation kinetics (12). The Hb(D73)/I-Ek complex associates with the TCR 3-fold faster than Hb/I-Ek, while its dissociation rate translates into a 3-s decrease in t1/2. Microscopy methods have allowed for the visualization of the immunological synapse formed between 3.L2 T cells and Hb/I-Ek (29). Similar experiments utilizing Hb(D73)/I-Ek as the ligand failed to reveal a durable activation cluster, consistent with the interpretation that this pMHC is a weak agonist for the 3.L2 TCR (unpublished results). In the context of results indicating the overall importance of dissociation rate in receptor-ligand interactions in general and TCR activation in particular (1, 11, 27), it seems entirely plausible that the 30% change in t1/2 could be the basis of the differential activities of the two pMHCs. The small kinetic deviations observed in this and related systems clearly illustrate the extraordinary sensitivity of the TCR in molecular discrimination. The positive and negative selection mechanisms of T cell development appear to work to maintain the low afnities of TCR/pMHC interactions to establish narrow kinetic thresholds. Indeed, stable high afnity TCR/pMHC complexes have been experimentally obtained by minor residue substitutions, indicating that there are no structural limitations of either the ligand or receptor for a more enduring embrace (9, 30). It may well be that TCRs are naturally selected in part for conformational exibility to allow for promiscuous, low afnity engagement of multiple partners (31). A wealth of information on the nature of receptor-ligand interactions has come from protein crystallography. However, a direct understanding of the energetic and kinetic roles of particular interfacial residues to the interaction is typically not revealed from the atomic coordinates alone (27, 32). Although the small structural variances of our two pMHCs are hard to relate to the large activity differences they elicit, their location, magnitude, and statistical signicance are consistent with previously documented alterations of TCR/pMHC biophysics and activities (8, 9). Our study further highlights the extreme sensitivity TCRs possess by demonstrating the apparent discrimination of pMHCs that differ by a single methylene group located in the buried side chain of an MHC anchor residue.

FIGURE 8. The P8 side chain is an important TCR contact for the 3.L2 TCR. Splenocytes from the 3.L2-transgenic mouse (35) were cultured in the presence of the indicated peptides, and proliferation was determined by incorporation of [3H]thymidine. The names and side chain structures of the different peptides are shown below the graph. Nonnatural amino acids were substituted at position 75 (P8) of Hb peptide to see the effects of a systematic elongation of the P8 side chain. Note that both Hb and Hb(D73) peptides have a Leu at P8, but they each prefer a different rotamer conformation. The nonnatural amino acids and their three-letter abbreviations are: 2-aminobutyric acid (Abu); norvaline (Nva); and norleucine (Nle).

Acknowledgments
We thank Mark Davis for the generous gift of soluble I-Ek molecules and Wayne Hendrickson for support during the initial stages of the project. Use of the Argonne National Laboratory Structural Biology Center beamlines at the Advanced Photon Source was supported by the U. S. Department of Energy, Ofce of Science, under Contract W-31-109-ENG-38. We thank the staff at SBC beamline 19-ID for their help. Finally, we thank Jerri Smith for secretarial support in the preparation of this manuscript.

rise to similar activity differences as observed for the P6 substitution in Hb(D73) (Fig. 8). Therefore, it seems likely that the different Leu75 rotamers we observed in the two pMHCs could partially account for the altered 3.L2 activation. Taken together, the P5P9 alterations create a novel Hb(D73)/I-Ek presentation surface that, surprisingly due to the minor nature of the chemical differences, generates distinct bulk T cell responses. The caveat to the conclusion that these structural differences alone account for the different immunological outcomes is that we have determined the structures of only the unliganded pMHCs. It is possible that signicant changes occur in one or both of these structures upon TCR engagement. Signicant changes in peptide conformation have been observed upon TCR binding to a class I pMHC (22). It is possible that TCR binding to our class II pMHC could expose the side chain at P6, allowing it to make direct contact with the TCR. Although this is not a scenario that we consider likely, it cannot be formally excluded by the present data. Considerations of T cell activation We have shown by surface plasmon resonance studies that the Hb and Hb(D73)/I-Ek complexes bind soluble 3.L2 TCR with similar

References
1. Davis, M. M., J. J. Boniface, Z. Reich, D. Lyons, J. Hampl, B. Arden, and T cell receptors. Annu. Rev. ImY. H. Chien. 1998. Ligand recognition by munol. 16:523. 2. Sebzda, E., S. Mariathasan, T. Ohteki, R. Jones, M. F. Bachmann, and P. S. Ohashi. 1999. Selection of the T cell repertoire. Annu. Rev. Immunol. 17: 829. 3. Robey, E., and B. J. Fowlkes. 1994. Selective events in T cell development. Annu. Rev. Immunol. 12:675. 4. Kersh, G. J., and P. M. Allen. 1996. Essential exibility in the T-cell recognition of antigen. Nature 222:495. 5. Evavold, B. D., S. G. Williams, B. L. Hsu, S. Buus, and P. M. Allen. 1992. Complete dissection of the Hb(64 76) determinant using Th1, Th2 clones, and T cell hybridomas. J. Immunol. 148:347. 6. Evavold, B. D., and P. M. Allen. 1991. Separation of IL-4 production from Th cell proliferation by an altered T cell receptor ligand. Science 252:1308. 7. Kersh, G. J., and P. M. Allen. 1996. Structural basis for T cell recognition of altered peptide ligands: a single T cell receptor can productively recognize a large continuum of related ligands. J. Exp. Med. 184:1259. 8. Ding, Y.-H., B. M. Baker, D. N. Garboczi, W. E. Biddison, and D. C. Wiley. 1999. Four A6-TCR/peptide/HLA-A2 structures that generate very different T cell signals are nearly identical. Immunity 11:45.

3354
9. Degano, M., K. C. Garcia, V. Apostolopoulos, M. G. Rudolph, L. Teyton, and I. A. Wilson. 2000. A functional hot spot for antigen recognition in a superagonist TCR/MHC complex. Immunity 12:251. 10. Lyons, D. S., S. A. Lieberman, J. Hampl, J. J. Boniface, Y. H. Chien, L. J. Berg, and M. M. Davis. 1996. A TCR binds to antagonist ligands with lower afnities and faster dissociation rates than to agonists. Immunity 5:53. 11. Savage, P. A., J. J. Boniface, and M. M. Davis. 1999. A kinetic basis for T cell receptor repertoire selection during an immune response. Immunity 10:485. 12. Kersh, G. J., E. N. Kersh, D. H. Fremont, and P. M. Allen. 1998. High- and low-potency ligands with similar afnities for the TCR: the importance of kinetics in TCR signaling. Immunity 9:817. 13. Smith, K. J., S. W. Reid, K. Harlos, A. J. McMichael, D. I. Stuart, J. I. Bell, and E. Y. Jones. 1996. Bound water structure and polymorphic amino acids act together to allow the binding of different peptides to MHC class I HLA-B53. Immunity 4:215. 14. Nelson, C. A., N. I. Viner, S. P. Young, S. J. Petzold, and E. R. Unanue. 1996. A negatively charged anchor residue promotes high afnity binding to the MHC class II molecule I-Ak. J. Immunol. 157:755. 15. Kozono, H., J. White, J. Clements, P. Marrack, and J. Kappler. 1994. Production of soluble MHC class II proteins with covalently bound single peptides. Nature 369:151. 16. Otwinowski, Z. A. 1997. Processing of X-ray-diffraction data collected in oscillation mode. Methods Enzymol. 276:307. 17. Fremont, D. H., W. A. Hendrickson, P. Marrack, and J. Kappler. 1996. Structures of an MHC class II molecule with covalently bound single peptides. Science 272:1001. 18. Brunger, A. T., P. D. Adams, G. M. Clore, W. L. DeLano, P. Gros, R. W. Grosse-Kunstleve, J. S. Jiang, J. Kuszewski, M. Nilges, N. S. Pannu, et al. 1998. Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54:905. 19. Jones, T. A., J. Y. Zou, S. W. Cowan, and M. Kjeldgaard. 1991. Improved methods for binding protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47:110. 20. Lorenz, R. G., and P. M. Allen. 1988. Direct evidence for functional self protein/ Ia-molecule complexes in vivo. Proc. Natl. Acad. Sci. USA 85:5220. 21. Garcia, K. C., M. Degano, R. L. Staneld, A. Brunmark, M. R. Jackson, T cell receptor structure P. A. Peterson, L. Teyton, and I. A. Wilson. 1996. An at 2.5 and its orientation in the TCR-MHC complex. Science 274:209.

STRUCTURE OF AN ALTERED CLASS II pMHC

22. Garboczi, D. N., P. Ghosh, U. Utz, Q. R. Fan, W. E. Biddison, and D. C. Wiley. 1996. Structure of the complex between human T-cell receptor, viral peptide and HLA-A2. Nature 384:134. 23. Reinherz, E. L., K. Tan, L. Tang, P. Kern, J. Liu, Y. Xiong, R. E. Hussey, A. Smolyar, B. Hare, R. Zhang, et al. 1999. The crystal structure of a T cell receptor in complex with peptide and MHC class II. Science 286:1913. 24. Blundell, T. L., and L. N. Johnson. 1976. Protein Crystallography. Academic Press, London. 25. Flocco, M. M., and S. L. Mowbray. 1995. Strange bedfellows: interactions between acidic side-chains in proteins. J. Mol. Biol. 254:96. 26. Marrack, P., L. Ignatowicz, J. W. Kappler, J. Boymel, and J. H. Freed. 1993. Comparison of peptides bound to spleen and thymus class II. J. Exp. Med. 178: 2173. 27. Wells, J. A. 1996. Binding in the growth hormone receptor complex. Proc. Natl. Acad. Sci. USA 93:1. 28. Chen, G., I. Dubrawsky, P. Mendez, G. Georgiou, and B. L. Iverson. 1999. In vitro scanning saturation mutagenesis of all the specicity determining residues in an antibody binding site. Protein Eng. 12:349. 29. Grakoui, A., S. K. Bromley, C. Sumen, M. M. Davis, A. S. Shaw, P. M. Allen, and M. L. Dustin. 1999. The immunological synapse: a molecular machine controlling T cell activation. Science 285:221. 30. Holler, P. D., P. O. Holman, E. V. Shusta, S. OHerrin, K. D. Wittrup, and D. M. Kranz. 2000. In vitro evolution of a T cell receptor with high afnity for peptide/MHC. Proc. Natl. Acad. Sci. USA 97:5387. 31. Willcox, B. E., G. F. Gao, J. R. Wyer, J. E. Ladbury, J. I. Bell, B. K. Jakobsen, and, P. A. van der Merwe. 1999. TCR binding to peptide-MHC stabilizes a exible recognition interface. Immunity 10:357. 32. Manning, T. C., C. J. Schlueter, T. C. Brodnicki, E. A. Parke, J. A. Speir, K. C. Garcia, L. Teyton, I. A. Wilson, and D. M. Kranz. 1998. Alanine scanning T cell receptor: mapping the energy of antigen recognition. mutagenesis of an Immunity 8:413. 33. Nicholls, A., K. A. Sharp, and B. Honig. 1991. Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins 11:281. 34. Connolly, M. L. 1983. Solvent-accessible surfaces of proteins and nucleic acids. Science 221:709. 35. Kersh, G. J., D. L. Donermeyer, K. E. Frederick, J. M. White, B. L. Hsu, and P. M. Allen. 1998. TCR transgenic mice in which usage of transgenic - and -chains is highly dependent on the level of selecting ligand. J. Immunol. 161:585.