Communication between Thiamin Cofactors in the Escherichia coli Pyruvate Dehydrogenase Complex E1 Component Active Centers

Kinetic, spectroscopic, and structural analysis tested the hypothesis that a chain of residues connecting the 4′-aminopyrimidine N1′ atoms of thiamin diphosphates (ThDPs) in the two active centers of the Escherichia coli pyruvate dehydrogenase complex E1 component provides a signal transduction pathway. Substitution of the three acidic residues (Glu571, Glu235, and Glu237) and Arg606 resulted in impaired binding of the second ThDP, once the first active center was filled, suggesting a pathway for communication between the two ThDPs. 1) Steady-state kinetic and fluorescence quenching studies revealed that upon E571A, E235A, E237A, and R606A substitutions, ThDP binding in the second active center was affected. 2) Analysis of the kinetics of thiazolium C2 hydrogen/deuterium exchange of enzyme-bound ThDP suggests half-of-the-sites reactivity for the E1 component, with fast (activated site) and slow exchanging sites (dormant site). The E235A and E571A variants gave no evidence for the slow exchanging site, indicating that only one of two active sites is filled with ThDP. 3) Titration of the E235A and E237A variants with methyl acetylphosphonate monitored by circular dichroism suggested that only half of the active sites were filled with a covalent predecarboxylation intermediate analog. 4) Crystal structures of E235A and E571A in complex with ThDP revealed the structural basis for the spectroscopic and kinetic observations and showed that either substitution affects cofactor binding, despite the fact that Glu235 makes no direct contact with the cofactor. The role of the conserved Glu571 residue in both catalysis and cofactor orientation is revealed by the combined results for the first time.

Communication between the thiamin diphosphate cofactors (ThDPs) 4 at the active centers of ThDP enzymes has been reported for several years (1)(2)(3)(4)(5)(6). An intriguing pathway for such interaction had been suggested by Perham's group (2) on the basis of crystal structure studies on E1bs (the E1 component of the pyruvate dehydrogenase complex (PDHc) from Bacillus stearothermophilus). According to this suggestion, there is a hydrated tunnel of acidic residues involved in proton shuttling from the N1Ј atom of the ThDP in one active site (activated ThDP, active site is closed) to the corresponding atom in the second active site (non-activated ThDP, active site is open) over a distance of 20 Å, dubbed a "proton wire" (2). This mechanism explains how two active sites communicate with each other, with no significant conformational changes in the structure of the subunits with the exception of the active site loops (2). At the same time, for E1h (human PDHc E1), a so-called "flip-flop" mechanism was proposed for active site communication through the shuttle-like motion of the heterotetramers, although the x-ray structure did not reveal directly a structural nonequivalence of the two active centers (4). Direct kinetic evidence of the chemical non-equivalence of the E1h active sites with respect to ThDP and substrate analog methyl acetylphosphonate binding was reported in support of the flip-flop mechanism. It was demonstrated that only one of the two bound ThDPs is in the activated state (5).
It was pointed out in a note accompanying the proton wire suggestion that not all ThDP enzymes have such an acidic proton wire arrangement of amino acids apparent in their x-ray structures (3), and, in fact, a study on the E1 component of a human branched-chain ␣-keto acid dehydrogenase demonstrated that a substitution of acidic residues on the path does not affect the overall activity of that complex. It was suggested that the two active sites in this enzyme operate independently during catalysis (6). This enzyme, however, does not have an uninterrupted array of acidic residues between ThDPs. Our groups have studied the kinetic behavior and x-ray structure of E1ec (the PDHc E1 component from Escherichia coli), which, as seen in Fig. 1, indeed has such an apparent series of acidic and hydrogen bonding-capable residues, all able to participate in active center communication in an uninterrupted array of residues and water molecules between the two ThDPs (3). We therefore deemed this enzyme a satisfactory test case for the pathway suggested by the proton wire hypothesis. We report biochemical and biophysical experiments, including steadystate kinetics, spectroscopic observation of the rates of formation of both the first step of the reaction sequence (formation of the ylide) and of the first predecarboxylation intermediate, fluorescence determination of ThDP binding, and x-ray structural analyses of the E235A and E571A variants on the putative pathway. The biochemical results clearly indicate that the interaction between active centers is greatly impaired upon substitution of any of the three acidic residues (E571, E235, and E237) or of Arg 606 along the communication pathway, whereas the x-ray results reveal the structural basis for the diminished functionality in the E235A and E571A variants examined by crystallographic methods.

EXPERIMENTAL PROCEDURES
Construction of the E235A, E237A, E571A, and R606A Variants of E1ec-The pGS878 plasmid encoding E1ec, two synthetic oligonucleotide primers (Integrated DNA Technologies) complementary to the opposite strands of the DNA, and the QuikChange site-directed mutagenesis kit (Stratagene) were used for mutagenesis reactions. The following synthetic oligonucleotides (and their complements) were used as mutagenic primers (mismatched bases are underlined, and mutated codons are shown in boldface type): 5Ј-CGGTGAAAT-GGACGCACCGGAATCCAAAGGTG-3Ј (for E235A), 5Ј-GTGAAATGGACGAACCGGCATCCAAAGGTGCGATC-3Ј(for E237A), 5Ј-GAAGGGATCAACGCGCTGGGCGCAG-GTTG-3Ј (for E571A), and 5Ј-GATGTTCGGCTTCCAGGC-TATTGGCGATCTGTGC-3Ј (for R606A). The presence of mutations was verified by sequencing of the e1ec gene with specific primers (University of Medicine and Dentistry of New Jersey).
Overexpression and Purification of E1ec and Its Variants-E. coli JRG 3456 cells deficient in the e1ec gene were used for plasmid transformation. Cells were grown in LB medium containing 50 g/ml ampicillin to A 600 ϭ 0.5-0.8, and protein expression was induced by isopropyl 1-thio-␤-D-galactopyranoside (0.5 mM) at 37°C for 5 h. Purification of E1ec and its variants was according to the protocol reported previously (7).
Activity and Related Measurements-The activity of E1ec and its variants was measured in the overall PDHc reaction after recon-stitution with E2ec and E3ec components as reported previously (8). The reaction medium contained the following in 1.0 ml: 0.1 M Tris-HCl (pH 8.0), 1 mM MgCl 2 , 0.20 mM ThDP, 2 mM sodium pyruvate, 2.5 mM NAD ϩ , 0.1-0.20 mM CoA, 2.6 mM dithiothreitol at 30°C. The reaction was initiated by the addition of PDHc and CoA. Steady-state velocities were taken from the linear portion of the progress curve. One unit of activity is defined as the amount of NADH ϩ produced (mol/min/mg E1). The E1-specific activity was measured in the model reaction monitoring the reduction of 2,6-dichlorophenol-indophenol at 600 nm (8). The lag phases () in the progress curves for NADH ϩ production were determined as described previously (9).
Hydrogen/Deuterium Exchange Kinetics at C2-H of the Enzymebound ThDP by Rapid Quench 1 H NMR Spectroscopy-The deprotonation rate constants at the C2 carbon of E1ec-bound and free ThDP were determined using a hydrogen/deuterium exchange technique as described earlier (5,10). The apo-E1ec enzymes were reconstituted with an equimolar amount of ThDP and 1 mM Mg 2ϩ in 20 mM KH 2 PO 4 (pH 7.0) and mixed with D 2 O (99.9%) in a 1:1 mixing ratio at 30°C, by using a rapid quench flow device (RQF-3, Kintek Corp. (Austin, TX)) for reaction times up to 2000 ms and by manual mixing for longer reaction times. After acid quench of the exchange reaction, the isolated cofactor was analyzed by 1 H NMR spectroscopy. The singlet signal of the C6Ј-H proton (8.01 ppm) was used as an internal non-exchanging standard. To obtain the exchange rates, the relative decay of the signal intensity of C2-H (9.68 ppm) was fitted to a monoexponential or double exponential function as detailed (5).
Fluorescence Spectroscopy-The fluorescence spectra of E1ec and its variants were recorded using a Cary Eclipse fluorescence spectrometer from Varian Inc. (8). The protein was diluted to a concentration of 0.034 mg/ml (concentration of active centers ϭ 0.342 M) in 10 mM KH 2 PO 4 (pH 7.0), containing MgCl 2 (5 mM) and pyruvate (1.0 mM).  Circular Dichroism-CD spectra were recorded on an Aviv model 202 CD spectrometer and on a Chirascan CD spectrometer from Applied Photophysics (Leatherhead, UK) in a 1-cm path cell at 30°C. The E1ec or its variants were diluted to a concentration of 2.0 mg/ml (concentration of active centers ϭ 20 M) in 20 mM KH 2 PO 4 (pH 7.0) containing MgCl 2 (2 mM) and 0.20 mM ThDP for E1ec and 0.50 mM ThDP for the E235A, E237A, E571A, and R606A variants, the maximum concentration of ThDP that could be used for CD titration. For titration experiments with methyl acetylphosphonate (MAP; a substrate analog), the E1ec variants were preincubated with ThDP for about 10 min before the first aliquot of MAP was added, and the steady-state level of the CD band was reached after ϳ2-2.5 h upon the addition of the last aliquot of MAP, eliminating the effect of the ThDPdependent lag phase (lasting for about 40 -100 s) on CD measurements.
Time-resolved CD-Stopped-flow CD experiments were carried out at 302 nm on a Pi Star Ϫ180 CDF instrument from Applied Photophysics with a path length of 10 mm and bandwidth of 2 nm at 30°C. In a typical experiment, protein was diluted to a concentration of 5-10 mg/ml in 20 mM KH 2 PO 4 (pH 7.0) containing ThDP (0.50 mM) and MgCl 2 (2.0 mM). The reaction was started upon mixing enzyme with MAP (2 mM) in 20 mM KH 2 PO 4 (pH 7.0) and was monitored for varied time intervals (5-200 s). The data were treated according to a double exponential function as in Equation 1.
Crystallization and Data Collection-Crystallization of the E571A and E235A variants of the E1ec complexed with ThDP in the presence of Mg 2ϩ was done separately by the sitting drop vapor diffusion method. The best crystals were obtained at a reservoir solution with 15-20% polyethylene glycol 2000 monomethyl ether, 10% propanol, 0.2% NaN 3 , and 60 mM HEPES (pH 7.0) at 22°C. Drops were 6 -10 l, consisting of equal parts of reservoir and protein solution. For crystallization of the E571A-ThDP, the protein concentration was 25.6 mg/ml, whereas for the E235A-ThDP, it was 28 mg/ml with the ThDP concentrations at both 0.5 and 5 mM in the protein solution. In all cases, the crystals grew within 4 -6 weeks and were isomorphous to E1ec-ThDP crystals (12). The crystals typically have dimensions of 0.15 ϫ 0.20 ϫ 0.30 mm. Low temperature (Ϫ180°C) data sets were collected on the SERCAT (sector-22ID) beamline at the APS (Advanced Photon Source, Argonne National Laboratory). All data sets were processed with the HKL2000 package (13). The E571A-ThDP data set was truncated at 2.1 Å resolution, whereas the E235A-ThDP data sets were truncated at 1.96 and 1.98 Å resolution for the complexes of E235A crystallized with 0.5 and 5 mM ThDP, respectively. Structure Determination and Refinement-Because the E1ec-ThDP crystals are isomorphous with the E571A-ThDP and E235A-ThDP crystals, the atomic coordinates of the E1ec-ThDP structure, refined to 1.85 Å resolution (12), were used as the starting model. The initial model included 1602 amino acids, with the cofactors ThDP and Mg 2ϩ along with neighboring active site residues omitted. Following rigid body refinement, simulated anneal-   APRIL 9, 2010 • VOLUME 285 • NUMBER 15

JOURNAL OF BIOLOGICAL CHEMISTRY 11199
ing was performed with the CNS program (14) without imposing any non-crystallographic symmetry, with a random subset of all data (ϳ5%) set aside for calculation of R free . After refinements of the protein parts were complete, 2F o Ϫ F c composite, simulated annealing, omit maps were calculated and examined. In the case of the E571A-ThDP, the map revealed the active site residues and location of only the diphosphate group of the ThDP cofactor; hence, only those atoms were used for further refinement. For the E571A-ThDP, strong electron density clearly indicated that residues His 142 and Tyr 598 in the active site were disordered, partially occupying two rotamer conformations. The alternate conformation setup procedure in CNS (14) was followed to provide for refinement of the disorder. Similar to the E1ec-ThDP structure reported previously, the regions 1-55, 401-413, and 541-557 were completely disordered, and they remain absent in the model. The model was then refined by simulated annealing, and subsequent cycles consisted of positional and B-factor refinement. Further model building, water molecule addition, and refinement cycles resulted in an R factor of 19.7% for the E571A-ThDP complex.
For the complexes of E235A with both the high and low ThDP concentrations, the maps revealed the active site residues and locations of the ThDP and Mg 2ϩ cofactors. In the lower ThDP concentration data set, the electron density for the cofactors is very weak, and a difference map showed some negative density for the thiazolium ring in one of the active sites. In the higher ThDP concentration data set, the cofactor electron density is stronger but still shows differences in the two active sites. Average B-factors for the atoms in the two cofactors for both data sets were significantly higher than the values for the nearby protein residues, suggesting that they may be bound at partial occupancy. To fit the occupancy of the cofactors in the higher concen-tration ThDP data set structure, B-factors for the cofactors were set to the average values of 44 Å 2 from neighboring protein atoms in one active site and 35 Å 2 from corresponding atoms in the other active site, respectively. Group occupancy refinement for the cofactors was then carried out in the CNS (14) program. A similar procedure was followed for the data set at lower ThDP concentration. For both structures, the cofactors were treated as three groups (4-aminopyrimidine ring, thiazolium ring, and diphosphate) for the occupancy refinement with the results given in Table 1. For the E235A-ThDP (0.5 mM) complex, the resulting thiazolium ring occupancy was estimated to be less than 40% in one active center. As in the E571A-ThDP structure, again strong electron density clearly indicated that the nearby residue His 142 is disordered in one of the active centers partially occupying two rotamer conformations, and this disorder was treated similarly. Once again, residues 1-55, 401-413, and 541-557 were completely disordered in the E235A-ThDP structures at both ThDP concentrations, and they remain absent in the models. Further model building, manual corrections, water molecule addition, and refinement cycles resulted in R factors of 19.4% for the E235A-ThDP (0.5 mM) complex and 19.9% for the E235A-ThDP (5 mM) complex, respectively.
For all structures, inspection of electron density maps/model building and refinement was carried out with the COOT (15) and CNS (14) packages, respectively. Graphical representations of protein models were generated by using the program RIBBONS (16). The final structures were checked with the MolProbity (17) Web server. All data collection and refinement statistics are given in Table 2.

Steady-State Kinetic Analysis of the E1ec Variants Created for Residues on the Communication Path Reveals that ThDP Binding Is Affected
As seen in Fig. 1, the residues Glu 571 , Glu 235 , Glu 237 , and Arg 606 are on the path connecting the ThDPs in the two active centers of E1ec. Earlier we reported that substitution of the highly conserved glutamate 571 within strong hydrogen-bonding distance of the N1Ј atom of ThDP resulted in a significant reduction of the overall activity: E571A (1.7%), E571D (0.61%), and E571Q (1.3%) (18,19) (Table 3). Upon reconstitution of the E235A, E237A, and R606A variants with E2ec and E3ec components, the overall PDHc activity measured under the optimal conditions for E1ec was significantly reduced: E235A (11.4%), E237A (7.2%), and R606A (3.3%) (Table 3). Surprisingly, increasing the ThDP concentration from 0.20 mM, saturating for E1ec, to 1 mM or even 5 mM enhanced the activity of the variants but not E1ec: E235A (2.4-fold), E237A (1.7-fold), E571A (1.7-fold), and E571D (12-fold). The activity of the R606A variant was unchanged with 0.20 mM or 2 mM ThDP and increased by 2.6-fold with 5 mM ThDP present (Table 3). In the E1-specific reaction, a similar response upon increasing ThDP concentration was observed for the E237A and R606A variants ( Table 3), signaling that ThDP binding was affected upon these substitutions in E1ec.
From the steady-state kinetic analysis of ThDP binding, the values of S 0.5,ThDP were 133 M (E235A) and 633 M (E237A), as compared with K m ,ThDP ϭ 1.6 M for E1ec (Table 3 and Fig. 2). The K m,ThDP values for E571A, E571Q, and R606A were close to  APRIL 9, 2010 • VOLUME 285 • NUMBER 15 JOURNAL OF BIOLOGICAL CHEMISTRY 11201 that for E1ec; however, for R606A, the progress curves for NADH production displayed an activation lag phase (100 s) over the entire range of ThDP concentrations (1-100 M) (Fig.   2, bottom). The K m,pyruvate was not significantly affected by E571A (0.17 mM), E571Q (0.35 mM), E235A (0.99 mM), E237A (1.36 mM), and R606A (0.51 mM) substitutions as compared with that for the E1ec (0.29 mM). We also emphasize that con-  Table 4.  version of Glu 571 to Ala or to a charge-neutralized substituent Gln appears to make little difference in activity, K m,ThDP , or K m,pyruvate ( Table 3).

Effect of Substitutions along the E1ec Active Center Communication Pathway on Individual Reaction Steps
The individual reaction steps are shown in Scheme 1. Fluorescence Experiments Reveal That the Substitutions Affect ThDP Binding at the Second Site-The binding of the ThDP to E235A, E571A, E571D (not shown), and R606A variants using percentage quenching of intrinsic fluorescence versus [ThDP] gave a biphasic plot (Fig. 3), indicating negative cooperativity, not observed for E1ec. For E1ec, a similar plot was hyperbolic with S 0.5,ThDP ϭ 1.84 M and Hill coefficient (n H ) ϭ 1.78, indicating positive cooperativity of the two active centers (not shown). For the E235A, E571A, E571D, and R606A variants, the second active center was not saturated even at 300 M ThDP (Fig. 3). The calculated K d,ThDP1 values for the first active center were not very different from S 0.5,ThDP for E1ec; however, the K d ,ThDP2 values for the second one were 17-fold (E235A) and 134-fold (R606A) higher than that for the first active center (Table 4), clearly indicating negative cooperativity upon ThDP binding. With the large S 0.5,ThDP estimated for the E237A variant, quenching data could not be analyzed because ThDP concentrations above the usable range (Ͼ600 M) would have been needed. The results affirm that binding of ThDP to the second active center is impaired in all of the variants here discussed.

Substitutions Do Not Affect Dimerization-We
Effect of E571A Substitution on Binding N1Ј-MeThDP, an Electrostatic APH ϩ Analog-We next investigated whether or not the enlarged ThDP site in the E571A variant could accommodate N1Ј-MeThDP, a stable analog of the N1Ј-protonated 4Ј-aminopyrimidinium form of ThDP (APH ϩ in Scheme 1). Remarkably, N1Ј-MeThDP binds to E571A even better than to E1ec; quenching of intrinsic fluorescence gave K d,N1Ј-MeThDP ϭ 66 M with E571A, whereas with E1ec no saturation was evident even at a 300 M concentration of N1Ј-MeThDP (Fig. 4). CD spectra of E571A recorded in the presence of 1-300 M N1Ј-MeThDP did not display any CD bands in the 300 -370-nm region, also consistent with the APH ϩ form of ThDP, for which we do not observe a CD signature in this wavelength range (data not presented) (20).

Rate of Ylide Formation at C2 of the E1ec-bound ThDP Suggests Half-of-the-sites
Reactivity-Monitoring the rate of hydrogen/deuterium exchange at the thiazolium C2 position of ThDP provides information about the first requisite step for catalysis (Scheme 1, k obs,AP3 Yl ) (10). Such rate constants for the enzyme-bound ThDP are presented in Table 5. The experiment revealed that one-half of the enzyme-bound ThDP C2-H is exchanged within 2 s; however, complete exchange resulted after 30 s (Fig. 5, top and bottom). The k obs,AP3 Yl is 160 Ϯ 67 s Ϫ1 for the fast exchanging site (activated site) and 0.04 Ϯ 0.02 s Ϫ1 for the slow exchanging site  Table 5.  APRIL 9, 2010 • VOLUME 285 • NUMBER 15 (dormant site), clearly demonstrating half-of-the-sites reactivity, indicating that when both E1ec active centers are filled with ThDP, only one of the two is in the activated state (Table 5 and Fig. 5). For the E235A and E571A variants, the rate constants for the fast exchanging site were k obs,AP3 Yl ϭ 9.8 Ϯ 5.7 s Ϫ1 and k obs,AP3 Yl ϭ 9.3 Ϯ 3.2 s Ϫ1 , respectively, at least 1 order of magnitude smaller than that for the E1ec (Fig. 6). For these two variants, a large amount of unbound ThDP was detected with the rate constants of hydrogen/ deuterium exchange of k obs,AP3 Yl is 0.14 Ϯ 0.06 s Ϫ1 for E235A and 0.12 Ϯ 0.02 s Ϫ1 for E571A. No hydrogen/deuterium exchange was detected for a slowly exchanging site, indicating that E235A and E571A variants have only one of the two active centers filled with ThDP. For the R606A variant, the hydrogen/deuterium exchange was very slow, with k obs,AP3 Yl of 0.08 Ϯ 0.01 s Ϫ1 , and the fast phase was not observed, indicating that within the time frame of the experiment, this variant (Fig. 6, bottom) did not bind ThDP, consistent with the ThDP activation lag phase measurements (Fig. 2, bottom).  (Fig. 7). For the R606A variant, probably both active centers are filled (about 92% is calculated), however at a molar ratio of [MAP]/[R606A active centers] of 20:1 (data not shown). For E571A and E571D, such results were reported and discussed earlier (21) .

Formation of the First Covalent (Predecarboxylation) Intermediate upon the Addition of the Substrate Analog MAP-As
The rate constants for PLThDP formation were very much affected in the variants. It was demonstrated using stopped-flow CD that upon mixing of E1ec with MAP, a positive CD band at 302 nm developed (reporting formation of the 1Ј,4Ј-imino-PLThDP) and reached a maximum in ϳ1-1.5 s. The kinetic curve could be resolved into two rate constants: k 1 ϭ 5.78 Ϯ 0.31 s Ϫ1 and k 2 ϭ 1.02 Ϯ 0.09 s Ϫ1 (Fig. 8).
For the E235A variant, the maximum amplitude of the CD band was

Structural Analysis of the E571A-ThDP and E235A-ThDP Complexes
The three crystal structures studied here are all very similar (in a global sense) to the previously reported crystal structure of the E1ec-ThDP complex (12). The main-chain folds are identical, and two subunits are present in an asymmetric unit. The two independent subunits are almost identical stereochemically and are related by a 2-fold non-crys-tallographic symmetry axis, with the active sites located at the subunit interfaces.
Structure of E571A-ThDP-In E571A-ThDP, the absence of the Glu 571 side chain creates additional space near residue Tyr 598 , and electron density maps clearly indicate two disordered sites for Tyr 598 in both active centers to partially fill this space. The most significant observation, however, is the well defined cofactor electron density present in both active centers only for the diphosphate tail of the ThDP. There is no clear electron density for the thiazolium and 4-aminopyrimidine rings. The observed diphosphate group interaction, nevertheless, is similar to that observed in the E1ec-ThDP structure (12). The octahedrally coordinated Mg 2ϩ binding site contains three protein ligands, two oxygen atoms from the diphosphate and one water molecule. Regarding the two alternative conformations of Tyr 598 in the active site, one of them is the same as that observed in the E1ec-ThDP, whereas the other is a new conformation seen here for the first time (Fig. 10). An immediate consequence of this feature is the steric hindrance that would now occur between the tyrosine ring and 4Ј-aminopyrimidine ring of the cofactor, if the latter were indeed present. The distance between the tyrosine and 4Ј-aminopyrimidine ring would be only 2.3 Å if both were present simultaneously (12). Because there is no clear evidence for electron density defining the cofactor rings in these active centers, the ThDP model from the E1ec-ThDP structure was used in Fig. 10 for interpretation of the results.
In addition, residue His 142 occupies the rotamer conformation observed in the E1ec-ThDP complex, an alternative conformation, or both in the E571A active centers as shown in Fig.  10. His 142 is disordered in only one of the two active centers, partially occupying two rotamer conformations (Fig. 10b). In the other active center, this residue adopts only one conformation (Fig. 10, top). The new H142 rotamer position would collide with the thiazolium ring of the cofactor, if the latter were present in its usual location.
Structural Analysis of E235A-ThDP Complex at Different ThDP Concentrations-As observed in the E571A-ThDP complex, changes are observed in the active centers of E235A-ThDP relative to the E1ec-ThDP complex but differ in their nature. Significant differences are observed between the cofactor's degree of order (as indicated by occupancies and electron density features) as well as in the conformations of certain amino acid side chains in the active centers. The structural changes observed in the active centers clearly affect the binding of the cofactors, as indicated in Table 1. The most significant observation is the absence of well defined electron density for the cofactor's thiazolium ring in both structures ( Fig. 11 and Table  1). The average occupancy of 0.5 for the thiazolium rings clearly indicates that the rings are not firmly fixed within the active centers. Even after a 10-fold increase of ThDP concentration (from 0.5 to 5 mM), the occupancy of the thiazolium ring is increased only from 0.36 to 0.53 in one of the active centers ( Table 1). As in the E571A-ThDP structure, disorder is also apparent in the His 142 side chain of E235A-ThDP, and the occupancy difference is consistent with the His 142 alternate rotamer conformation present in one of the active centers, because the alternate conformation would collide with the cofactor thiazolium ring, thereby affecting its binding. This structural information also clearly indicates non-equivalence of active centers, because the alternate conformation for residue His 142 is present only in one active center.

DISCUSSION
The major conclusion derived from structural, kinetic, and spectroscopic studies using substitutions at the Glu 235 , Glu 237 , and Arg 606 positions, residues that along with the highly conserved Glu 571 residue are hypothesized to form a communication pathway between the two active center ThDPs (2,3), is that ThDP binding in the second active center is affected by these substitutions according to the following observations. 1) Steady-state kinetic analysis revealed that ThDP binding is affected. 2) Analysis of the quenching of intrinsic protein fluorescence by ThDP revealed negative cooperativity upon ThDP binding in the E1ec variants, in contrast to the positive cooperativity reported for E1ec, indicating that ThDP binds with great preference in one of the two active centers. 3) The hydrogen/ deuterium exchange kinetics at the thiazolium C2-H position of the enzyme-bound ThDP clearly demonstrated half-of-thesites reactivity for the E1ec with a fast exchanging (activated) site and a slowly exchanging (dormant) site. For the E235A and E571A variants, the rate constant for the fast exchanging site was at least 1 order of magnitude slower than that for the E1ec. No hydrogen/deuterium exchange was detected for a slowly exchanging site, indicating that only one of the two active centers is filled with ThDP (half-of-the-sites reactivity). 4) Half-ofthe-sites reactivity was also demonstrated by CD titration of the E235A and E237A variants with MAP, an analog of pyruvate; about 49% of the active centers were filled with PLThDP (a stable predecarboxylation intermediate analog) in E1ec variants, as compared with E1ec, in which about 92% of the active centers were filled with PLThDP. The rate constants for PLThDP formation were at least 100 times slower than for E1ec. 5) The crystal structures of the E571A-ThDP and E235A-ThDP complexes do not reveal significant changes in the folding of the two subunits but do reveal pronounced differences in the active centers (see below).
It has been suggested that the two active centers in the E1bs PDHc communicate by mostly acidic residues in a hydrated channel (2). In the E1ec (3), the two active center ThDPs are separated by ϳ20 Å and are interconnected by residues Glu 571 , Glu 235 , Glu 237 , and Arg 606 and water molecules. The only direct cofactor contacts through this chain are via the catalytically important and conserved hydrogen bonds between the ThDP N1Ј nitrogen atoms and the side chains of a conserved Glu 571 (Fig. 1). Structural studies presented in this paper show that Glu 571 is involved in proper ThDP binding and orientation as well as being directly involved in the catalytic mechanism but is not required for diphosphate binding. This is in accord with spectroscopic studies indicating the compromised binding of ThDP and its analogs N1Ј-MeThDP and ThTTDP. The site-directed substitution studies clearly indicate that the catalytic activity is greatly impaired upon substitutions of the above mentioned residues, despite the fact that some of them are quite distant from the active center and make no obvious interactions with the cofactors or expected reaction intermediates. In this regard, the crystallographic analysis reported here provides the structural basis for the reduced activity associated with the E235A and E571A variants located on the proposed pathway. It reveals a limited ability to properly bind and orient the ThDP as indicated by the reduced occupancy and apparent disorder of ThDP, especially for the thiazolium ring. It is noteworthy that structural studies of the E1ec-ThDP complex indicated no direct interaction between the cofactor and Glu 235 (Fig. 11, top), but there is a strong hydrogen bonding interaction between Glu 235 and the conserved residue Glu 571 ; the latter residue interacts directly with the cofactor. The interaction between Glu 571 and the N1Ј atom of ThDP plays a role in assisting tautomerization of the 4Ј-aminopyrimidine ring leading to the reactive 1Ј,4Ј-iminopyrimidine tautomer, and the latter tautomer is involved in activation of the thiazolium ring to form the active ylide.
The E235A and E571A structural results indicate that the interaction between these residues plays a role in cofactor binding and activation, probably by affecting the protonation state of Glu 571 (given their proximity, Glu 235 and Glu 571 cannot both be ionized at the same time) and therefore affecting its ability to stabilize a protonated N1Ј atom in the APH ϩ ionization state of ThDP (compare Figs. 1 and 10).