Determination of the crystal structure of an "open" unliganded active mutant (T141D) form of the Escherichia coli phosphate receptor for active transport has allowed calculation of the electrostatic surface potential for it and two other comparably modeled receptor structures (wild type and D137N). A discovery of considerable implication is the intensely negative potential of the phosphate-binding cleft. We
MATERIALS AND METHODST141D mutant E. coli was prepared as described previously (5) and extensively dialyzed at pH 8.5 to remove any phosphate. Crystals were grown in a hanging drop containing 7 mg/ml protein and a 1:4 dilution of the well solution: 20% (w/v) polyethylene glycol 6000, 50 mM KCl, 100 mM NH4Cl, and 20 mM CH3COOK, pH 4.5. The crystal has the spacegroup of P21 and unit cell dimensions of a = 73.22 A, b = 39.59 A, c = 113.61 A, and ,B = 92.80. The asymmetric unit contains two molecules, which were identified as A and B. Intensity data were collected at 4°C and processed as described previously (2, 5). As the crystal decayed rapidly in the x-ray beam, only data to 2.4 A were collected (14,157 unique reflections with 46% complete in the last 2.51 to 2.40 A shell). As the crystal was not isomorphous with any previous crystals of PBP, we used molecular replacement technique as implemented in XPLOR (6) to determine the structure. We succeeded in obtaining solutions to the rotation and translation searches only by using domain II fragment of the wild-type structure (2) as a search model. The rotation solutions at Euler angles and translation in fractions of the unit cell for one domain II fragment are (01 = 278.23, 02 = 63.76, and 03 = 195.91) and (x = 0.033, y = 0, and z = 0.117), respectively, and for the other domain 11 (01 = 98.29, 02 = 117.87, and 03 = 163.93) and (x = 0.458, y = 0.017, and z = 0.383), respectively. Displaying the two models of the domain II fragment and symmetry related fragments using the program CHAIN (7) revealed no bad intermolecular contacts and a large empty area where domain I of both models could be placed. Guided by this observation, we successfully pieced together the two complete PBP models in the asymmetric unit. Each of the domain II models was overlapped with the identical domain of the complete structure of the ligandbound PBP. Domain I of both models was then rotated with respect to domain II about a hinge between the two domains connecting the a-carbons of residues 86 and 254 to place the domain in the empty space. This required about 250 rotation relative to the bound structure for both models to generate the open unliganded form with no bad contacts between the dimer and symmetry related molecules. These models were suitable for XPLOR rigid body refinement against the data to 3 A, with the two domain II fragments fixed and the two fitted domain I fragments treated as rigid bodies. This refinement brought the crystallographic R-factor down from an initial value of 0.495 to 0.349. A simulated annealing refinement further considerably reduced the R-factor...