Present address Institute for Biophysics, Department of Physics, Johann-Wolfgang-Goethe-University Frankfurt/Main, Max-von-Laue-Str. 1, 60438 Frankfurt/Main, Germany
The crystal structure of pyruvate decarboxylase from Kluyveromyces lactis
Implications for the substrate activation mechanism of this enzyme
Article first published online: 29 AUG 2006
Volume 273, Issue 18, pages 4199–4209, September 2006
How to Cite
Kutter, S., Wille, G., Relle, S., Weiss, M. S., Hübner, G. and König, S. (2006), The crystal structure of pyruvate decarboxylase from Kluyveromyces lactis. FEBS Journal, 273: 4199–4209. doi: 10.1111/j.1742-4658.2006.05415.x
- Issue published online: 29 AUG 2006
- Article first published online: 29 AUG 2006
- (Received 19 June 2006, accepted 13 July 2006)
- allosteric enzyme activation;
- conformation equilibrium;
- disordered loop regions;
- thiamine diphosphate
The crystal structure of pyruvate decarboxylase from Kluyveromyces lactis has been determined to 2.26 Å resolution. Like other yeast enzymes, Kluyveromyces lactis pyruvate decarboxylase is subject to allosteric substrate activation. Binding of substrate at a regulatory site induces catalytic activity. This process is accompanied by conformational changes and subunit rearrangements. In the nonactivated form of the corresponding enzyme from Saccharomyces cerevisiae, all active sites are solvent accessible due to the high flexibility of loop regions 106–113 and 292–301. The binding of the activator pyruvamide arrests these loops. Consequently, two of four active sites become closed. In Kluyveromyces lactis pyruvate decarboxylase, this half-side closed tetramer is present even without any activator. However, one of the loops (residues 105–113), which are flexible in nonactivated Saccharomyces cerevisiae pyruvate decarboxylase, remains flexible. Even though the tetramer assemblies of both enzyme species are different in the absence of activating agents, their substrate activation kinetics are similar. This implies an equilibrium between the open and the half-side closed state of yeast pyruvate decarboxylase tetramers. The completely open enzyme state is favoured for Saccharomyces cerevisiae pyruvate decarboxylase, whereas the half-side closed form is predominant for Kluyveromyces lactis pyruvate decarboxylase. Consequently, the structuring of the flexible loop region 105–113 seems to be the crucial step during the substrate activation process of Kluyveromyces lactis pyruvate decarboxylase.
pyruvate decarboxylase from Kluyveromyces lactis
pyruvate decarboxylase from Saccharomyces cerevisiae
Pyruvate decarboxylase (PDC; EC 126.96.36.199) is a key enzyme of carbon metabolism at the branching point between aerobic respiration and anaerobic alcoholic fermentation. It catalyzes the decarboxylation of pyruvate in plants, yeasts and some bacteria by using thiamine diphosphate (ThDP) and Mg2+ as cofactors. The catalytic cycle of ThDP enzymes is well established (Scheme 1). At first, the α-carbonyl group of the substrate is attacked by the deprotonated C2 atom of the thiazolium ring of ThDP [the ylid (I)]. In the case of pyruvate, the resulting lactyl-ThDP (II) is subsequently decarboxylated to yield the central intermediate of ThDP catalysis, the α-carbanion/enamine (III). Protonation of III yields hydroxyethyl-ThDP (IV), and the release of the second product acetaldehyde completes the catalytic cycle of ThDP.
The yeast Kluyveromyces lactis (formerly termed Saccharomyces lactis) is able to assimilate lactose and convert it to lactic acid. It is commercially utilized for the production of recombinant chymosin, a proteolytic enzyme used to coagulate milk in cheese manufacturing.
In contrast to S. cerevisiae, only one gene codes for PDC in Kluyveromyces lactis. The protein (SwissProt entry Q12629) has 86.3% identical residues and 96.4% similar residues compared to SwissProt entry P06169, the dominant PDC in S. cerevisiae. It is known from small-angle X-ray solution scattering experiments (unpublished results) that the catalytically active form of K. lactis PDC (KlPDC) is a homotetramer at micromolar protein concentrations (563 amino acid residues per subunit, total molecular mass 240 kDa). The cofactors ThDP and Mg2+ are bound tightly, but not covalently, at the interface of two monomers (Fig. 1). At pH values > 8, the cofactors dissociate from the protein, resulting in complete loss of catalytic activity. Lowering the pH to 5.7–6.3, which is also the optimum for KlPDC catalysis, can restore this activity almost completely.
In 1967, Davies  was the first to describe a sigmoidal deviation of the plot of reaction rate vs. substrate concentration for PDC from wheat germ. Hübner et al.  established a first model for this substrate activation phenomenon. Stopped-flow kinetic techniques were used to analyze the substrate activation of S. cerevisiae PDC (ScPDC). From studies with the inhibitor glyoxylic acid and the inconvertible activator pyruvamide (2-oxopropane amide, the amide analog of the substrate pyruvate), it was concluded that a separate binding site for the regulatory substrate molecule must exist. Later, Hübner and Schellenberger  showed that the enzyme is potentially inactive in the absence of substrate. With the single exception of the bacterial enzyme from Zymomonas mobilis, all PDCs studied so far are subject to substrate activation.
Lu et al. [7,8] described the structural consequences of substrate activation on the basis of the crystal structure of pyruvamide-activated ScPDC compared to that of ScPDC crystallized in the absence of any effectors , which is assumed to be the nonactivated state of the enzyme. Activation involves a rearrangement of the two dimers within the tetramer: the D2 symmetry of the nonactivated ScPDC is broken, and an open and a closed side of the tetrameric molecule is formed. Two different binding sites of the activator were located: one at the interface between the two domains within one subunit, and one directly at the active site. In the presence of pyruvamide, the loop regions 106–113 and 292–301 undergo a disorder–order transition and close over the active sites, thus possibly stabilizing the binding of substrate.
An alternative pathway for substrate activation is favored by Baburina et al. [10–12] and Li et al. [13,14], who suggest that an activator molecule, bound to residue Cys221, is the starting point for the activation transition. However, no electron density for a bound activator molecule could be detected directly at this amino acid residue in pyruvamide-activated ScPDC. Instead, pyruvamide was found to bind 10 Å away from Cys221, in a pocket formed by two of three domains of the subunit .
Here, we describe the crystal structure of PDC from the yeast K. lactis and the structural consequences of the substrate activation of this PDC species. Our model constitutes an extension to the activation model previously proposed and established for ScPDC .
Quality of the crystal structure model
The asymmetric unit contains a complete tetramer. Hence, the final model consists of four polypeptide chains arranged as a homotetramer of approximate D2 symmetry. Each monomer was modeled using the amino acid sequence deduced from KlPDC gene pdc1, corresponding to SwissProt entry Q12629. The refined model comprises residues 2–105, 114–289 and 303–562 of subunit A, residues 2–104 and 114–554 of subunit B, residues 2–104 and 116–556 of subunit C, residues 2–104 and 121–562 of subunit D, four molecules of ThDP, four Mg2+, and 1649 water molecules. The final R-factor is 0.158 (for complete data collection and processing statistics, see Table 1).
|Number of crystals||1|
|Crystal–detector distance (mm)||180|
|Rotation range per image (°)||0.5|
|Total rotation range (°)||265.5|
|Unit cell parameters (Å)||a = 78.72, b = 203.09, c = 79.78, β = 91.82°|
|Resolution limits (Å)||99.0–2.26 (2.32–2.26)|
|Total number of reflections||549 432|
|Unique reflections||114 899|
|I/σ (I)||20.2 (6.4)|
|Completeness (%)||98.5 (95.5)|
|Rmerge (%)||7.1 (21.5)|
|Rr.i.m. (%)||8.0 (24.7)|
|Rp.i.m. (%)||3.5 (11.8)|
|Overall B-factor from Wilson plot (Å2)||28.3|
|Optical resolution (Å)||1.70|
Neither the terminal residues, nor residues 105–113 in all subunits and residues 290–302 in one subunit, could be traced in the electron density map, probably because of too high flexibility of these regions. Even in subunits B–D, in which the latter region could be traced, the high flexibility of the loop is evidenced by B-factors > 50 Å2, which are clearly above the average of ∼ 22 Å2(Table 2). In the crystal structure of nonactivated ScPDC, none of the two loop regions are resolved . However, they are well defined in the structure of pyruvamide-activated ScPDC . Another flexible loop in KlPDC is the one comprising amino acid residues 344–360. This loop is located at the solvent-exposed surface of the tetramer and it connects the middle and the C-terminal domains (Fig. 2). In the crystal structure of pyruvamide-activated ScPDC, the cleft between these domains contains the binding site for the activator molecule.
|Resolution range (Å)||23.58–2.26 (2.32–2.26)|
|Total number of atoms (nonhydrogen)||18 466|
|Number of protein atoms||16 776|
|Rcryst (%)||15.8 (16.3)|
|Rfree (%)||21.4 (27.0)|
|r.m.s.d. from ideality|
|% in most favored regions||92.5|
|Average B-factor (Å2)|
The KlPDC tetramer consists of two asymmetrically associated identical homodimers (r.m.s.d. < 0.41 Å based on 7566 atoms). Although no activator is present, the KlPDC tetramer contains an open and a closed side and thus resembles more closely the tetramer structure of pyruvamide-activated ScPDC (Fig. 3) than that of the nonactivated ScPDC (Fig. 4). In going from the nonactivated form of ScPDC to the activated one, one dimer has to rotate by about 30° relative to the other. For comparison, the corresponding angle found for (nonactivated) KlPDC is ∼ 36°. The main difference between KlPDC and the activated form of ScPDC is the flexibility of the loop regions 105–113 and 290–302. Whereas these loops are completely ordered in pyruvamide-activated ScPDC, residues 105–113 are completely disordered, and 290–302 partially disordered, in KlPDC. As a consequence, KlPDC resembles nonactivated ScPDC more closely than activated ScPDC in terms of loop flexibility.
As in all other ThDP-dependent decarboxylases analyzed so far, the KlPDC subunit consists of three domains (Fig. 2). According to Muller et al. , these domains are termed the PYR domain (binding the aminopyrimidine ring of ThDP), the R domain (binding regulatory effectors), and the PP domain (binding the diphosphate residue of ThDP). All three domains exhibit their typical α/β-topology. The central six-stranded β-sheet of the PYR domain (residues 2–182) is surrounded by seven α-helices. The R domain (residues 193–341) consists of five α-helices and a central six-stranded β-sheet. A central six-stranded parallel β-sheet and eight α-helices form the PP domain (residues 360–556). A superposition of ScPDC and KlPDC monomers yields r.m.s.d. values < 0.85 Å (based on 3650 aligned atoms). The largest displacements are observed for the C-terminal helix (5.5 Å) and for most parts of the central R domain.
Structure of the active site
The general architecture of the active site of KlPDC corresponds to that of other ThDP-dependent enzymes. Figures 1–3 illustrate the binding of the cofactors ThDP and Mg2+ at the interface between two subunits. The aminopyrimidine ring of ThDP is bound at the PYR domain of one subunit. The diphosphate residue is bound to the PP domain of the other subunit at the same dimer together with the octahedral coordinated Mg2+(Fig. 5). The amino acid arrangement at the active site enforces the so-called V-conformation of ThDP . This relative orientation of the pyrimidine ring and the thiazolium ring is one of the three conformations that occur in crystal structures of isolated ThDP, but is the only one found in more than 60 crystal structures of ThDP-dependent enzymes analyzed so far. All residues at the active site in direct vicinity to the cofactors are identical to those of ScPDC. However, some side chain conformations appear to be different. His114, which is thought to be necessary for substrate and/or intermediate binding [18–22], is adjacent to the disordered loop region 105–113. Even the side chain of His114 exhibits rather poor electron density. The γ-carboxyl group of Asp28, a residue important for reaction intermediate stabilization , is shifted by about 2.5 Å towards the C2 atom of the cofactor ThDP, when compared to pyruvamide-activated ScPDC. Some minor differences (< 2 Å) can be identified for residues Asn471, Thr475 and Glu477, which are involved in the binding of the diphosphate group of ThDP, either directly or via Mg2+ coordination (Fig. 5).
Amino acid substitutions
Twenty of the 77 substitutions in KlPDC are nonconservative compared to ScPDC. Most of these residues are located at the surface of the tetramer and are thus probably not involved in the catalytic mechanism. No exchanges occur at the active site or the putative regulatory site . Three substitutions (Asn143-Ala, Ala196-Ser, and Ser318-Asn, the first residue referring to KlPDC and second residue to ScPDC) are located directly at the dimer–dimer interface (Fig. 6). These might affect the dimer–dimer interactions, but none of them are located at the monomer–monomer interface within the dimers. Two exchanges (Val104-Ile and Ser106-Ala) can be found in the flexible loop region 105–113.
The structural basis of the activation of PDC is the rotation of one dimer relative to the other within the tetramer. This rotation is accompanied by local conformational changes within the subunits due to the binding of pyruvamide between the R and the PP domains. The dimer reorientation leads to the generation of a closed and an open side in the tetramer. Consequently, new interaction areas are formed at the closed side of the molecule. The most important one is a disorder––order transition of two loop regions, which are flexible in the nonactivated state (residues 106–113 and 292–301). These loops close over the active sites and shield the catalytic centers from the solvent. In accordance with these results, Liu et al.  have suggested the involvement of two histidine residues (His114 and His115) adjacent to loop 106–113 in substrate and/or intermediate binding, based on kinetic studies of ScPDC variants. In contrast to the situation observed for ScPDC, a half-side closed quaternary structure of the tetramer of KlPDC exists already in the nonactivated state. This observation, based on the crystal structure, is corroborated by small-angle X-ray scattering experiments that reveal a more compact structure of nonactivated KlPDC (radius of gyration 3.85 nm) compared to nonactivated ScPDC (radius of gyration 3.95 nm) (unpublished results). Furthermore, solution structure models calculated ab initio from small-angle X-ray scattering data at low resolution (> 2 nm) illustrate a nonplanar dimer arrangement in the KlPDC tetramer.
The observed differences in the three-dimensional structure of both yeast PDCs manifest themselves in the quaternary arrangement only. The monomers and dimers can be superimposed with relatively low r.m.s.d. values, which is to some extent expected because of the high homology of their amino acid sequences. However, it was shown previously that differences exist between the two enzymes based on detailed kinetic studies of KlPDC substrate activation . Analyses of the microscopic rate constants for this process in various PDCs illustrated a particularly low binding affinity for the substrate at the regulatory site (Ka value) in the case of KlPDC. The half-side closed structure of the KlPDC tetramer may reflect this special kinetic behavior. In the case of KlPDC, binding of the regulatory substrate is not required for the induction of a change in the dimer assembly as in pyruvamide-activated ScPDC − this conformation is already preformed. From a structural point of view, it is in fact possible that substrate activation of KlPDC involves only a part of the processes in ScPDC, namely, binding of the regulatory substrate(s) in the cleft between the R and PP domains. The bound activator may then enhance the rigidity of the enzyme molecule and drive the disorder–order transition of the flexible loop region (residues 105–113). This loop forms a lid over the active site, making it inaccessible to solvent and thereby allowing the catalytic reaction .
The substrate activation model for KlPDC has been developed on the basis of crystal structure models only. One can argue that solution structures may differ from these models and that crystal contacts may influence interactions of neighboring molecules. However, we believe that our interpretation is supported by the similarity of the quaternary structures of KlPDC and pyruvamide-activated ScPDC, although the first has been crystallized in the absence of any allosteric effectors and the latter in the presence of high concentrations of the substrate surrogate pyruvamide. Furthermore, we have previously shown that crystal and solution structures of several ThDP-dependent enzymes are essentially identical in the absence of effectors . Differences seem to be dependent on the compactness of the enzyme molecules. The dimer arrangement in the crystal structure of tetrameric nonactivated ScPDC is rather loose (dimer interface area 1640 Å2 compared to 2700 Å2 calculated for KlPDC, and 3200 Å2 for pyruvamide-activated ScPDC). A nonactivated ScPDC model with an altered dimer assembly within the tetramer resulted from rigid body refinement  of crystallographic vs. solution scattering data. In this solution structure, the dimers of the crystal structure are rotated ∼ 15° and their distance is decreased by ∼ 5 Å. Possibly, an equilibrium between various quaternary PDC structures exists, which is shifted more towards a planar dimer orientation in ScPDC and towards the half-side closed conformation in KlPDC, perhaps due to the amino acid substitutions at the dimer interface mentioned above. Binding of the regulatory substrate may then stabilize the latter conformation in KlPDC and enable effective catalysis.
Protein expression and purification
The protamine sulfate treatment was omitted. Precipitation ranges were changed: for acetone (55–70%, v/v) and for ammonium sulfate (29.25–30.75 g per 100 mL). An additional ammonium sulfate precipitation of the protein guaranteed the removal of all traces of acetone. The resulting sediment was resuspended in a minimal volume of 0.1 m Mes with 2 mm dithiothreitol, pH 6.0, loaded on a SuperdexTM 200 column (26 × 600 mm), and eluted with the same buffer, but with 0.3 m ammonium sulfate at a flow rate of 0.5 mL·min−1. Fractions with catalytic activities above 35 U·mg−1 (1 U is defined as the consumption of one µmol of substrate per min) and with more than 95% homogeneity (judged by SDS/PAGE according to the method of Lämmli ) were combined, saturated with the cofactors, and precipitated with solid ammonium sulfate. The pellets were stored at − 20 °C after quick freezing.
Here, the range used for ammonium sulfate precipitation was 28.5–34.75 g per 100 mL. Size exclusion chromatography was performed as described above for ScPDC. Fractions were combined, precipitated with ammonium sulfate, resuspended in 20 mm Bistris, pH 6.8, with 2 mm dithiothreitol, and desalted on HitrapTM (GE Healthcare, Munich, Germany) Sephadex columns (5 × 5 mL) at 3 mL·min−1. The protein solution was loaded on a Poros20QE (Perseptive Biosystems GmbH, Wiesbaden, Germany) anion exchange column (4.6 × 100 mm) in the same buffer and eluted with an ammonium sulfate gradient of 0–500 mm at a flow rate of 2 mL·min−1. Fractions with catalytic activities above 40 U·mg−1 and with more than 95% homogeneity (according to SDS/PAGE) were combined, saturated with the cofactors, and precipitated with solid ammonium sulfate. The pellets were stored at − 20 °C after quick freezing.
KlPDC, stored as frozen ammonium sulfate precipitate, was diluted in crystallization buffer (50 mm Mes, pH 6.45, 5 mm ThDP, 1 mm dithiothreitol, 5 mm MgSO4 or 35 mm sodium citrate, pH 6.45, 1 mm dithiothreitol, 5 mm ThDP, 5 mm MgSO4). Excess ammonium sulfate was removed, and the protein was concentrated by the use of centrifugal concentrators (0.5 mL, 30 kDa cut-off). KlPDC was crystallized by hanging drop vapor diffusion in 24-well cell culture plates. Four-microliter drops of protein solution (3–15 mg·mL−1) were mixed 1 : 1 with PEG 2000/PEG 8000 (12–24%, w/v) in crystallization buffer. The best crystals were obtained at 20% (w/v) PEG and 2 mg KlPDC/mL at 8 °C. Microcrystals in Mes buffer were obtained after 3 days. Larger single crystals (0.4 × 0.02 × 0.02 mm) appeared after about 4 weeks. These crystals were stable for several months. Microcrystals in citrate buffer could be detected after 24 h. They grew to a final size of 0.6 × 0.02 × 0.02 mm over 10 days, with higher reproducibility than those grown in Mes buffer. However, these crystals were stable for 2 weeks only and disintegrated in solutions with PEG or glycerol concentrations less than 12% (w/v).
Data were collected under cryogenic conditions from a single crystal of KlPDC, grown in Mes buffer. The crystal was soaked with a cryoprotectant containing reservoir solution with 15% (v/v) glycerol for 30 s, frozen in liquid nitrogen and transferred into the cryogenic nitrogen stream at the beamline. A native dataset was recorded on beamline X11 (EMBL, Hamburg, Germany) using an MARCCD detector. The data were indexed and integrated using denzo and scaled using scalepack. The redundancy-independent merging R-factor Rr.i.m. as well as the precision-indicating merging R-factor Rp.i.m. were calculated using the program rmerge (available from http://www.embl-hamburg.de/~msweiss/projects/msw_qual.html or from MSW upon request). Intensities were converted to structure factor amplitudes using the program truncate[32,33]. Table 1 summarizes the data collection and processing statistics. The optical resolution was calculated using the program sfcheck.
Structure solution and refinement
Initial phases were obtained from the model of the ScPDC dimer (PDB entry code 1QPB) by molecular replacement with the program molrep. The search model for this procedure was generated by automated modeling of the KlPDC amino acid sequence using the swissmodel modeling server . Refinement (rigid body, TLS and restrained) was carried out against this data set using the program refmac5 . Inspection of electron density, model building and checking was done with the program coot. Several cycles of refinement and manual model building were carried out until the free R-factor and the crystallographic R-factor had converged (Table 2).
Figures were prepared with pymol (DeLano Scientific, San Carlos, CA) and ds viewerpro (Accelrys Software Inc., San Diego, CA). The coordinates and structure factors have been deposited in the Protein Data Bank, http://www.pdb.org (PDB ID code 2G1I).
Interface-accessible surface areas were calculated by using the program provided by the protein–protein interaction server (http://www.biochem.ucl.ac.uk/bsm/PP/server).
The authors thank the EMBL outstation for access to beamline X11 at the DORIS storage ring, DESY, Hamburg.
- 32Collaborative Computational Project Number 4 (1994) CCP4 suite: programs for protein crystallography. Acta Crystallogr D50, 760–763.
- 341999) sfcheck: a unified set of procedures for evaluating the quality of macromolecular structure-factor data and their agreement with the atomic model. Acta Crystallogr D55, 191–205., & (
- 362004) Coot: model-building tools for molecular graphics. Acta Crystallogr D60, 2126–2132.& (