Table 1 Data collection and structure determination. Crystals of the 2C–H-2K^{b}–dEV8 complex (23) are orthorhombic (*P*2_{1}2_{1}2) with two TCR-pMHC complexes in the asymmetric unit. An x-ray data set was collected on a MAR imaging plate at the Stanford Synchrotron Research Laboratory on a crystal cryopreserved (20% ethylene glycol) and flash-cooled in a gaseous nitrogen stream. X-ray diffraction data from this crystal extended beyond 2.8 Å in the *a** and *c** directions, but the overall data were truncated to 3.0 Å because of a marked anisotropy along *b**, as judged from Wilson plots. The average *I*/σ(*I*) at 3.0 Å resolution for *a** and *c** are 5.0 and 3.3, respectively. Along *b** the average*I*/σ(*I*) drops below 2.0 at 5.3 Å and below 1.0 at 4.5 Å resolution, although some measurable data extend to 3.6 Å. All of the measured intensities were used in the subsequent structure determination and refinement. The effective resolution, defined as the resolution at which the total number of collected unique reflections equals a 100% complete data set (25), is 3.2 Å when all reflections with*F* > 0, or 3.38 Å for *F* > 2σ(*F*) are considered. Data were integrated and reduced with the DENZO and SCALEPACK programs (25). Initial phase estimates were obtained by molecular replacement. Rotation and translation functions (15.0 to 4.0 Å) were calculated with AMoRe (25), with the coordinates of the unliganded 2C TCR [PDB code 1TCR (15)] and H-2K^{b}–OVA [1VAC (49)]. The Patterson maps for rotation function calculations were generated from normalized structure factors to sharpen their features and partially correct for the anisotropy. Only one solution for the TCR (peak heights of 6.5σ and 8.2σ in the rotation and translation function maps, respectively) and one solution for the MHC molecule (4.9σ and 9.4σ) were clear at this stage. A native Patterson map showed a 7σ peak at fractional coordinates (0.46, 0.33, 0.5), suggesting that a local twofold symmetry axis, parallel to a crystallographic axis, described the relation between the two molecules in the asymmetric unit. A partial translation function map, calculated after fixing the already-positioned TCR-pMHC complex and using the model in the same orientation for the three-dimensional search, was manually inspected for the existence of a peak related by this intermolecular cross vector to the original solution. A 6.5σ peak located at (0.97, 0.82, 0.87) met this requirement. The second complex in the asymmetric unit was located on the basis of this solution of the translation function. Five envelopes encompassing the variable portion and constant regions of the TCR, the α1α2 domain and peptide of K^{b}, the α3 domain of K^{b}, and β_{2}M were generated with MAMA (25), and the rotation and translation matrices relating the different domains were calculated and improved with an electron density correlation optimization algorithm (RAVE) (25). Phases were calculated from the model to 5.5 Å, refined through real-space averaging and solvent flattening with RAVE, and gradually extended to 3.0 Å in 40 cycles. Electron density maps calculated from the improved phases allowed unambiguous rebuilding of the regions that underwent structural changes and allowed the introduction of the correct dEV8 peptide sequence. The structure was refined with positional and torsion angle dynamics refinement (25), restraining the main chain and side chain atom positions of the model by means of harmonic force constants, by using the program X-PLOR 3.851 (25). The model was rebuilt with the program O (25), according to σ_{A}-weighted electron density maps (25), real-space averaged maps, and “shake” omit maps calculated after 10% of the model was omitted (25). Progress of the refinement was monitored by the decrease of*R*
_{free} (25) calculated from 4% of the unique data (1435 reflections). The anisotropic diffraction was partially corrected by introduction of an overall anisotropic tensor (*B*
_{11} = 14.6 Å^{2} ,*B*
_{22} = –11.0 Å^{2},*B*
_{33} = 14.9 Å^{2}), which increased the atomic displacement parameter array by 17 Å^{2}(the relative *B* values before the correction were V_{α} = 22 Å^{2}, V_{β} = 25 Å^{2}, Cα = 38 Å^{2}, Cβ = 26 Å^{2}, K^{b}-dEV8 = 24 Å^{2}, and β_{2}M = 26 Å^{2}). Refinement was begun with strict noncrystallographic symmetry constraints and later substituted by restraints to minimize *R*
_{free}. A flat bulk solvent correction was applied. Individual isotropic temperature factors were refined, resulting in a similar decrease in the conventional and free *R* values (3.1% drop in*R*
_{free} and 3.7% drop in*R*
_{cryst} compared with a 0.4% drop in*R*
_{free} and 1.0% drop in*R*
_{cryst} when refining two*B* values per residue), and acceptable Ramachandran plots [calculated with PROCHECK (25)]. Three ordered water molecules were located in the interface based on 3σ*F*
_{o} – *F*
_{c} residual electron density and added to the final model. The crystallization, data collection, and structure determination of H-2K^{b}–dEV8 are described in (50). {tt}