G. Schneider, Department of Medical Biochemistry and Biophysics, Tomtebodavägen 6, Karolinska Institutet, S-17177 Stockholm, Sweden. Fax: + 46 832 7626, Tel.: + 46 8728 7675, E-mail: firstname.lastname@example.org or S. König, Institute of Biochemistry, Department of Biochemistry & Biotechnology, Kurt-Mothes Str. 3, Martin-Luther-University Halle-Wittenberg, 06099 Halle/Saale, Germany. Fax: + 49 3455527011, Tel.: + 49 3455524829, E-mail: email@example.com
The crystal structure of the complex of the thiamine diphosphate dependent tetrameric enzyme pyruvate decarboxylase (PDC) from brewer's yeast strain with the activator pyruvamide has been determined to 2.4 Å resolution. The asymmetric unit of the crystal contains two subunits, and the tetrameric molecule is generated by crystallographic symmetry. Structure analysis revealed conformational nonequivalence of the active sites. One of the two active sites in the asymmetric unit was found in an open conformation, with two active site loop regions (residues 104–113 and 290–304) disordered. In the other subunit, these loop regions are well-ordered and shield the active site from the bulk solution. In the closed enzyme subunit, one molecule of pyruvamide is bound in the active site channel, and is located in the vicinity of the thiazolium ring of the cofactor. A second pyruvamide binding site was found at the interface between the Pyr and the R domains of the subunit in the closed conformation, about 10 Å away from residue C221. This second pyruvamide molecule might function in stabilizing the unique orientation of the R domain in this subunit which in turn is important for dimer–dimer interactions in the activated tetramer. No difference electron density in the close vicinity of the side chain of residue C221 was found, indicating that this residue does not form a covalent adduct with an activator molecule. Kinetic experiments showed that substrate activation was not affected by oxidation of cysteine residues and therefore does not seem to be dependent on intact thiol groups in the enzyme. The results suggest that a disorder–order transition of two active-site loop regions is a key event in the activation process triggered by the activator pyruvamide and that covalent modification of C221 is not required for this transition to occur. Based on these findings, a possible mechanism for the activation of PDC by its substrate, pyruvate, is proposed.
Thiamine diphosphate (ThDP) is an essential cofactor for a number of enzyme-catalysed reactions, especially in carbohydrate metabolism. Among ThDP dependent enzymes, pyruvate decarboxylase (PDC; EC 188.8.131.52.) has been widely studied to understand the mechanism of thiamine catalysis . PDC catalyses the penultimate step in the alcoholic fermentation process, the nonoxidative decarboxylation of pyruvate to acetaldehyde. In yeast and bacteria, the catalytically active enzyme is composed of four identical or almost identical subunits. In the yeast Saccharomyces cerevisiae the major structural gene PDC1 codes for 563 amino acids . PDCs from plants form higher oligomeric complexes with subunit sizes in the range of 59–67 kDa .
All PDCs studied so far, except the enzyme from the bacterium Zymomonas mobilis[4,5], are subject to substrate activation. This process, characterized by a sigmoidal deviation from the hyperbolic v/S plot, was observed as early as 1967 in the enzyme species from wheat germs  and was described later in PDC from yeast and plant seeds [7,8]. More intensive studies using the stopped-flow technique confirmed a lag phase in product formation [7,9]. Structural analogues of the substrate pyruvate, pyruvamide and ketomalonate, respectively, are regarded as artificial activators [7,10]. Pyruvamide binds reversibly with relatively low affinity (KA = 44 ± 8 mm derived from ). Substrate activation accelerates the deprotonation rate of the C2 atom of the cofactor ThDP by at least three orders of magnitude .
Despite extensive kinetic investigations of this phenomenon the structural basis of the activation process is still not completely understood. Cross-linking of PDC from brewer's yeast in the presence and absence of pyruvate resulted in fully activated and inactivated enzyme forms, respectively, demonstrating the importance of conformational changes during substrate activation . Different conformations of native and pyruvamide-activated PDC from yeast were also suggested by small-angle X-ray solution scattering experiments with synchrotron radiation [13,14]. Irreversible activation of the yeast enzyme (demonstrated by a complete disappearance of the lag phase in product formation during catalysis) was achieved by chemical modification with the thiol specific reagent 4-hydroxymercuribenzoate (HMB) or the activator analogue 3-bromopyruvamide  suggesting the participation of a cysteine residue in substrate activation. However, these chemical modifications were accompanied by a considerable decrease in catalytic activity. Based on site-directed mutagenesis and steady-state kinetics Baburina et al. [16,17] proposed that C221 is a primary binding site of the regulatory substrate molecule and the starting point of a signal transfer pathway to the cofactor ThDP at the active site.
To date several crystal structures of PDCs have been published [18–22]. Crystal structures of PDCs from S. cerevisiae and S. uvarum were determined in the absence of any effector, and were referred to as form A PDC. Crystallization of yeast PDC in the presence of the activator pyruvamide  and ketomalonate  resulted in a new crystal form (form B PDC), with significant differences in tetramer assembly compared to form A. Due to the insufficient resolution (2.8 and 3.5 Å, respectively) of these crystallographic studies, the binding sites of the activator molecules could not be identified. We have now determined the structure of PDC from brewer's yeast strain crystallized in the presence of pyruvamide to higher resolution, which enabled the localization of the pyruvamide binding sites. Kinetic studies of substrate activation did not support a direct role of cysteines in this process. Based on these kinetic and structural data we propose a mechanism for the substrate activation of yeast PDC and the role of C221 in this process.
Materials and methods
Protein purification and crystallization
Purification and crystallization of PDC from strain WS34/70 of brewer's yeast in the presence of 320 mm pyruvamide was carried out as described previously . Pyruvamide was synthesized according to Vogel and Schinz . The batch did not contain pyruvate, as judged by 1H-NMR or by kinetic measurements using pyruvamide as a substrate of PDC. Pyruvamide, as determined by 1H-NMR and 13C-NMR, had a half-life of about 22 days at 25 °C. Crystals (form B) belonged to space group C2 with cell dimensions a = 145.1 Å, b = 119.7 Å, c = 81.3 Å, α = γ = 90°, β = 120.2°. Each asymmetric unit contained a dimer, with the twofold crystallographic axis passing through the tetramer.
Intensive crystal screening under flash-freezing conditions resulted in the identification of six crystals which diffracted to a resolution better than 2.7 Å and data were collected on a rotating anode, operated at 50 kV and 90 mA using a MAR Research image plate. Diffraction images were processed and scaled using the HKL package . The best crystal diffracted to 2.4 Å and the data set from this crystal was used in the subsequent refinement (for statistics of data collection, see Table 1).
Table 1. Statistics of data collection.
Highest resolution interval
Structure determination and refinement
The structure was determined earlier by molecular replacement and refined to an R-factor of 23% at 2.8 Å resolution . This low resolution structure of form B PDC was used as a starting model for the refinement with the higher resolution data. Refinement was carried out with X-plor and at a later stage with CNS . Data between 15 and 2.38 Å were included and a bulk solvent correction was used. Five percent of the diffraction data were set aside to monitor the progress by means of the free R-factor . Non-crystallographic symmetry restraints were imposed for most parts of the polypeptide chain, except residues where the electron density indicated obvious differences between the two subunits in the asymmetric unit. The protocol consisted of iterative rounds of refinement and model examination/rebuilding with O, until the R-free value had converged.
The sequence of the structural gene(s) of brewer's yeast PDC has not yet been determined. Therefore the sequence of PDC1 from S. cerevisiae (PO6169) was originally used to build the model . However, at several positions the electron density map was inconsistent with this sequence, but consistent with the sequence of either PDC5 (X15668) or PDC6 (X15668), the other two structural genes of PDC from S. cerevisiae. In order to be consistent with the electron density the following side chain replacements were made: position 55, R→A; 106, S→A; 143, C→A; 208, A→I; 253, S→D; 336, N→A. The replacements at positions 55, 106, and 143 were consistent with the crystal structure of form A PDC  and former versions of the sequence of PDC1.
For each pyruvamide binding site identified in the course of refinement, 2Fo-Fc and Fo-Fc maps were calculated and examined with data sets collected using different crystals. In all these maps, electron densities at the pyruvamide positions were found, thus excluding the possibility that the observed difference densities represented an artefact due to model bias or experimental errors in the X-ray data.
The statistics of the refinement and model parameters are given in Table 2. The model was analysed with procheck. The observed structure factor amplitudes and the atomic coordinates have been deposited with the Protein Data Bank (http://www.rcsb.org/pdb/), accession code 1qpb.
Table 2. Statistics of structure refinement and the final model.
Data used in the refinement (Å)
Number of reflections
Number of nonhydrogen atoms
R.m.s. bond lengths (Å)
R.m.s. bond angles (deg)
Mean B factor (Å2)
Incubations were carried out at 30 °C in 50 mm Mes pH 6.0 at an enzyme concentration of exactly 0.3 mg·mL−1. Prior to incubation, excess salt in the enzyme samples was removed by gel filtration and ultrafiltration in 50 mm Mes pH 6.0. For complete saturation with cofactors PDC was preincubated in 1 m Hepes/NaOH pH 6.8 (10 mm ThDP, MgSO4) at a ratio of 1 : 1 for 5 min at room temperature. Protein concentration was determined using the experimental extinction coefficient of 281 000 m−1·cm−1 at 280 nm  and controlled during the whole incubation period. No enzyme denaturation was detectable during incubation.
Three independent series of enzyme incubations were done: one for following the decrease of catalytic activity, a second for the estimation of the extent of substrate activation, and a third for the determination of the number of intact thiol groups by modification with HMB.
Catalytic activity of PDC was assayed spectrophotometrically at 340 nm and 30 °C in 0.1 m sodium citrate pH 6.0 at 40 mm pyruvate, according to the coupled test of Holzer et al.  with ADH/NADH as the auxiliary enzyme system. The conversion of 1 µmol of pyruvate per minute corresponds to 1 U of enzyme activity. The initial catalytic activity of the PDC used was 43–46 U·mg−1. The homogeneous state of the enzyme was verified by densitometric analysis of SDS polyacrylamide gels.
The kinetics of substrate activation were measured with a stopped-flow spectrophotometer (Applied Photophysics). One syringe contained the enzyme, and the other the substrate pyruvate and the auxiliary enzyme system in the same buffer and at the same concentrations (except that pyruvate was 1 mm) as mentioned for the measurements of the catalytic activity. Substrate conversion was followed until a sufficient part of the reaction run with steady-state rate (Fig. 4). Progress curves were analysed by fitting the data to a coupled first and zero order reaction according to the equation: absorbance = A + B · t + C · exp(–kcat · t) . Substrate activation is expressed as the ratio of the product formation rates at the beginning of the reaction and at steady state, and is determined from the fitting curve as (B – kcat · C)/B. This initial catalytic activity is about 0.5% of the steady-state rate for native yeast PDC .
The number of modifiable thiol groups was determined via modification with HMB (0.75 mm stock solution) and calculated from the absorption difference of modified and unmodified protein at 249 nm (ε = 7900 m−1·cm−1). Based on the known second order rate constants of modification, about 0.5 mm−1·s−1, the reaction should be completed within the incubation time used in the experiments (5 min).
The final model, refined to 2.4 Å resolution, comprises 8552 protein atoms of two PDC monomers, two ThDP molecules, two magnesium ions, two pyruvamide molecules and 141 water molecules. There is continuous and well-defined electron density for most parts of the protein molecule, excluding the N-terminal residue, the C-terminal residues 557–563, and the loops 104–113 of one subunit (C) and 290–304 of the other subunit (O) in the asymmetric unit (see Note), indicating disorder for these segments of the polypeptide chain. The model with good stereochemistry was refined to an R-factor of 23.2% and an R-free value of 28.6% using all the data between 15 and 2.4 Å (Table 2). In the model, 90% of the amino acid residues are located in the most favoured regions of the Ramachandran plot. One residue in each monomer, R317, is located in the disallowed region, but with well-defined electron density. This residue is also seen in an unfavourable conformation in the form A PDC structure .
The tetramer assembly of this PDC form (form B) is different from that in native PDC (form A) and these differences have been described in detail previously . The asymmetric nature of the form B PDC tetramer can be seen in Fig. 1. The four PDC monomers form a ‘dimer of dimers’ as in form A PDC  but the packing of the two dimers is completely different. In form A PDC, the four subunits are related by 222 symmetry, whereas in form B PDC, the different conformation of the two subunits (O and C) in each dimer causes significant deviations from 222 symmetry.
Each monomer is built up of three independent domains, which were denoted as Pyr domain, R domain and PP domain . Two subunits form a dimer in the asymmetric unit. The Pyr and PP domains in the dimer are related by the same transformation, while the R domains are related by a slightly different transformation, which corresponds to a subtle but significant rotation of 1.2° of the R domain in one subunit relative to the R domain in the other subunit. The R domains participate in the dimer–dimer interactions and the small difference in orientation between the two R domains translates into about a 0.8–1.2 Å displacement of residues at the dimer–dimer interface. Because of this shift of the R domains in relation to the Pyr and PP domains, they interact with different residues in the two monomers. In other words, the interfaces formed by the R domains are not equivalent.
There are two disordered loop regions in form A PDC, comprising residues 104–113 and 290–304 in each subunit . In form B PDC loop 104–113 is ordered in the O-subunit and loop 290–304 is ordered in the C-subunit . The segment 290–304 contains a short antiparallel β-sheet which is in contact with residues 104–113 from the open subunit. In fact, both loops fold over and limit access to the active site in the C-subunit. In addition, regions 104–113 from the corresponding monomers of the two dimers form major dimer–dimer interactions in the form B tetramer assembly and two of the four active sites are found in the closed conformation (Fig. 1). The corresponding loops participating in the other two active sites are disordered and therefore these active sites are exposed to solvent.
Pyruvamide binding sites
The difference electron density maps allowed the assignment of two pyruvamide binding sites in the dimer (Fig. 2). A strong positive difference electron density, indicative of a bound ligand, was found in the active site channel of one of the subunits. The shape and intensity of this difference electron density are consistent with a bound pyruvamide molecule. This pyruvamide site is located close to the thiazolium ring of the ThDP cofactor and the loop 104–113 (Fig. 2a). At the present resolution the observed electron density is compatible with four orientations of the pyruvamide molecule. The binding mode as presented in Fig. 2 is based on maximizing the number of hydrogen bonds, and avoiding unfavourable contacts. In this orientation the pyruvamide molecule forms hydrogen bonds to the carboxyl group of D28, a nitrogen atom of the side chain of H115 from the O-subunit, and the carboxyl group of E477 from the C-subunit. These amino acid residues have been identified as important catalytic residues by site-directed mutagenesis in PDC of both Z. mobilis and S. cerevisiae[33–35].
Another strong, positive difference density consistent with a bound pyruvamide molecule was found in a pocket surrounded by residues from the Pyr domain, the R domain and the loop connecting these two domains in the closed subunit (Fig. 2b). This second pyruvamide binding site is located 10 Å away from C221, the residue proposed to be covalently engaged in the formation of a hemithioketal adduct with the activator, pyruvate or pyruvamide. Two hydrogen bonds are formed between the amide group of pyruvamide and the side chain hydroxy oxygen atom of Y157 (Pyr domain) and the main chain oxygen atom of R224 (R domain). Other interactions of the bound ligand with protein residues are of van der Waals type. No difference electron density for any pyruvamide molecule was found in the corresponding location of the second monomer in the asymmetric unit. This might be due to differences in protein structure, as there are differences in packing between the Pyr and the R domains in the two subunits causing local structural changes at this site. Also, this pocket does not exist in form A PDC  due to different packing between these domains, i.e. the orientation of the R domains in the two structures differs by a 7–8° rotation, corresponding to a 2.5–3 Å displacement of atoms at the interface between the R and Pyr domain in the two PDC forms.
Kinetic studies of substrate activation
From previous studies of chemical modification of cysteines with HMB it was concluded that these residues are involved in the substrate activation process . HMB-modified yeast PDC does not show any lag phase in product formation, but has on the other hand only very low catalytic activity (about 5% of the native enzyme). Kinetic studies of the mutant C221A did not show any evidence for substrate activation and consequently, Barburina et al. [16,36] proposed that C221 is the regulator binding site. In view of the fact that no pyruvamide molecules were found close to C221 in the crystal structure (Fig. 3), and that kinetic measurements of product formation with the stopped-flow technique (in particular at temperatures below 30 °C) still demonstrate substrate activation for the variant C221A of yeast PDC (B. Seliger & G. Hübner, unpublished results), we reinvestigated the role of cysteine residues for substrate activation in yeast PDC.
Four cysteines are present in the sequence of PDC1: C69, C152, C221, and C222. C69 is buried deep inside the molecule; the other three are accessible to solvent. In yeast PDC, eight of the 16 cysteine residues in the PDC tetramer can be modified by HMB, and thus represent exposed, reactive thiols . During long-term incubation of the enzyme in aqueous solutions, the number of modifiable thiols decreases and is accompanied by a decrease in catalytic activity. These cysteine residues may be oxidized to sulfones or sulfoxides thus making them unavailable for modification by HMB. If covalent reaction of the regulator pyruvamide with a reactive cysteine residue (especially C221) is essential for substrate activation of PDC, the lag phase of product formation should disappear with progressive thiol oxidation.
We have therefore studied catalytic activity, thiol modification and substrate activation independent from each other during long-term incubation of yeast PDC. UV/visible spectra were recorded during the whole incubation period in order to detect possible enzyme denaturation which might obscure the kinetic results. While the number of HMB-modifiable cysteine residues decreases during incubation from eight to less than 0.5 per tetramer and the catalytic activity drops down to less than 10% of the original value, substrate activation is fully retained (Fig. 4). This indicates that catalytic activity of yeast PDC but not substrate activation depends on the presence of intact cysteine residues.
The present studies have revealed new insights into the structural basis of substrate activation of PDC. Activation by pyruvamide results in global and local conformational changes in the PDC tetramer. The largest rearrangement upon pyruvamide binding is the 30° rotation of one dimer relative to the other dimer, creating a new dimer–dimer interface. More localized conformational changes involve the two loops comprising residues 104–113 and 290–304. These chain segments, which are the crucial components in the new dimer–dimer interface formed in the tetramer assembly of form B PDC, undergo a disorder–order transition resulting in the closure of two active sites in the dimer. One of these loops in each monomer interacts with a pyruvamide molecule and is part of the active site in the activated molecule. Pyruvamide is structurally similar to the substrate pyruvate and it is reasonable to assume that the pyruvamide binding mode at the active site is similar to the binding of the substrate, pyruvate, in the activated complex. Indeed, pyruvamide is bound at the bottom of the active site channel, in the vicinity of the reactive C2 carbon atom of ThDP, which will attack the carbonyl carbon atom of pyruvate during catalysis. The conformation of the two loops as seen in the pyruvamide-activated complex would allow interaction with bound substrate in a similar manner. Formation of the closed subunit involves residues from loop regions of two adjacent subunits, and packing of the two dimers as seen in form B PDC is not possible unless two adjacent subunits undergo transition from open to closed form. This transition has to occur in a concerted manner and binding of the substrate is therefore not independent in two of the four active sites.
The orientation of the R domain in the form B tetramer which is different from that seen in form A PDC is important in forming dimer–dimer interactions. The presence of a pyruvamide molecule located at the domain interface formed by the Pyr and the R domain suggests that this ligand might play a role in stabilizing the present orientation of the R domain which is important for the different packing in the pyruvamide-activated enzyme. Therefore we conclude that one important feature of pyruvamide activation of PDC is the binding of the activator in the pocket formed at this interface, and that this pocket can be viewed as a ‘regulation site’.
Based on these observations, we propose a possible mechanism for the substrate activation of PDC as illustrated in Fig. 5. The basic underlying assumption of this model is that the form A PDC structure [18,19] represents the nonactivated state of the enzyme, while the structure of the pyruvamide complex of PDC is very close, if not identical, to the activated enzyme. In the nonactivated state, the tetramer consists of four equivalent active sites in the open conformation. Activation involves large conformational changes that result in a rearrangement of the subunits in the tetramer, breaking the 222 symmetry. The rearrangement is stabilized by binding of the activator at the interface between the R and Pyr domain of the closed subunits and by loop closure in two active sites. The latter has two consequences: (a) it creates a novel interface in the tetramer and (b) residues from these loops interact and stabilize binding of the substrate in the active site. As the closure of the two active sites must, for structural reasons, occur (almost) simultaneously, binding of substrate to two active sites is not an independent process. Two aspects of this mechanism are as yet unresolved. It is not clear whether a direct transition between the two symmetry-related activated species can occur without passing through the conformation of the nonactivated enzyme. It is also not yet established whether there exists a tetramer species with all subunits in the closed conformation (similar to PDC of Z. mobilis). If such a species exists, it requires a novel, not yet observed tetramer assembly, as the required conformational change cannot occur without severe disturbance of the tetramer packing in form B PDC.
The proposed mechanism of activation in PDC implies large scale conformational transitions with changes in tetramer assembly. The mechanism is consistent with the observation that PDC from Z. mobilis (ZmPDC) does not show any substrate activation. The tetramer of ZmPDC is formed by two dimers, with extensive interactions in the dimer–dimer interface, and the mode of subunit packing in the tetramer is very different from both form A and form B PDC . The tight packing of the subunits and the large interface regions between the subunits in ZmPDC suggest that this assembly is very stable and that no large conformational changes occur during catalysis. Therefore any local conformational changes at each active site of ZmPDC are independent of each other. In contrast, conformational changes at the active sites of PDC from yeast are related to changes in the tetramer assembly, which in turn control binding of the substrate.
HMB modification of yeast PDC led to an activated enzyme form . From these studies it was concluded that sulfhydryl side chains are required for activation to occur. Amino acid replacement of cysteine residues in the interface region, i.e. C221 [16,17], abolished the sigmoidal deviation of the curve in the v/S plot, but caused drastic decreases in catalytic activity, suggesting C221 as a regulation site. However, stopped-flow kinetic studies with this variant at temperatures below 30 °C show substrate activation, although to a lower extent (B. Seliger & G. Hübner, unpublished results). The present crystal structure analysis suggests an alternative explanation for these observations. The structure analysis has revealed slight, but significant changes in the interface between the R and the Pyr domain upon transition from the nonactivated to the activated state of the enzyme. It is therefore conceivable that alterations at this interface will influence the domain–domain interactions required for maintaining the proper orientation of the R domain. Chemical modification or site-directed replacement of residues at this interface could impair or, in certain cases, facilitate the conformational transitions involving the R domains, with concomitant changes in the regulatory behaviour of the enzyme. Therefore we propose that the activator indeed functions by modulating domain interactions but not by binding to C221. This conclusion is also consistent with the kinetics studies, which suggest that substrate activation is not correlated to the presence of intact, reactive cysteine residues (Fig. 4). If activator binding to a cysteine residue would be required for activation of the enzyme, the decrease in the number of accessible, reactive cysteine residues during incubation (to about 0.1 per subunit) should strongly influence the substrate activation behaviour.
This work was supported by the Swedish Natural Science Research Council, the German Federal Ministry of Education and Research, and the German Academic Exchange Service.
*Present address: Institute of Pharmacology, University of Bern, CH-3010 Bern, Switzerland.
†Present address: Dept. of Molecular Biophysics, Lund University, PO Box 124, 22100 Lund, Sweden.