Data collection and refinement statistics.

aHigh-resolution cutoff was based on the CC1/2 value obtained from the program Aimless in CCP4 suite [32]. bNumbers in parentheses refer to data in the highest resolution shell. cRmerge = Σ|Ih−h>|/Σ Ih, where Ih is the observed intensity and h> is the average intensity. dRwork = Σ ||Fobs|-k|Fcal||/Σ|Fobs|. eRfree is the same as Robs for a selected subset (10%) of the reflections that was not included in prior refinement calculations. fOrdered residues: apo structure (Met1 to Tyr337 in subunit A and Ile3 to Ala317 in subunit B), a binary complex (Met1 to Tyr337 in subunit A and Ile3 to Lys315 in subunit B), and a ternary complex (Met1 to Asn338 in subunit A and Ser4 to Lys315 in subunit B). gOutliers identified using a program MolProbity [33]: three residues in apo form, Ser333 for subunit A, Asn157, Asn192 for subunit B; two residues in the binary complex, Asn157 for subunit A, Asn157 for subunit B; four residues in the ternary complex, Arg113, Asn157 for subunit A, Met143, Asn157 for subunit B.