Allosteric Mechanism of Pyruvate Kinase from Leishmania mexicana Uses a Rock and Lock Model*

Allosteric regulation provides a rate management system for enzymes involved in many cellular processes. Ligand-controlled regulation is easily recognizable, but the underlying molecular mechanisms have remained elusive. We have obtained the first complete series of allosteric structures, in all possible ligated states, for the tetrameric enzyme, pyruvate kinase, from Leishmania mexicana. The transition between inactive T-state and active R-state is accompanied by a simple symmetrical 6° rigid body rocking motion of the A- and C-domain cores in each of the four subunits. However, formation of the R-state in this way is only part of the mechanism; eight essential salt bridge locks that form across the C-C interface provide tetramer rigidity with a coupled 7-fold increase in rate. The results presented here illustrate how conformational changes coupled with effector binding correlate with loss of flexibility and increase in thermal stability providing a general mechanism for allosteric control.

of activity regulation, including the binding of amino acids, phosphorylation, and the binding of oncoproteins, may provide further therapeutic approaches for targeting hPYK (6,7).
Most allosterically regulated proteins are enzymes in which the binding of an activator or inhibitor to the effector site can affect the binding of a substrate at the active site. The Monod-Wyman-Changeux model of allostery (8) suggests that oligomeric enzymes undergo symmetrical transitions (classically between the T-and R-states) 3 (36) that can be stabilized by ligand binding. However, there are now examples of allosteric control that do not show obvious conformational change (9), and a growing body of work exists to support the idea that flexible regions of molecules (which exist as an ensemble of conformers) may undergo allosteric regulation by changes in the conformer population (10). There are examples of allosteric enzymes in which the same protein has been captured in both a T-state (which has low affinity for substrate) and an R-state (which has higher affinity for substrate), and some structural insight has been obtained by the study of at least one of the allosteric states of Ͼ50 proteins (10 -12). However, a full understanding of the allosteric effect requires structural information on each of four states: (i) apoenzyme, (ii) active site complex, (iii) effector site complex, and (iv) complex with filled active and effector sites. For allosterically regulated enzymes, no such data exist (see supplemental Table S1), and in many cases even when the T-and R-state structures are available, they have been examined under different conditions (of differing pH and salt concentrations) (12), all of which can affect structural conformation. This lack of a consistent structural series has to date precluded the full systematic study of allosteric mechanisms. Here, we present the structures and thermodynamic properties of the four allosteric states of LmPYK, the pyruvate kinase from Leishmania mexicana, a protozoan parasite belonging to the trypanosomatid family.
Pyruvate kinases are homotetrameric enzymes that catalyze the final reaction of glycolysis in which phosphoenolpyruvate and ADP are converted into pyruvate and ATP. PYK monomers (50 -60 kDa depending on species) are composed of four domains: the N-terminal, A-, B-, and C-domains (Fig. 1). The C-domain houses the effector site which is located 40 Å from the active site. Adjacent C-domains form the C-C or "small" interface, and bordering A-domains form the A-A or "large" interface. The B-domain forms a mobile lid at one end of the (␣/␤) 8 -barreled A-domain, and the active site lies in the cavity between them.
Allosteric enzyme behavior of PYKs may be manifested either through the binding of the effector molecule or through the binding of the substrate phosphoenolpyruvate (13,14). In mammals and many other species fructose 1,6-bisphosphate (Fru-1,6-BP) acts as the effector, but trypanosomatid PYKs are allosterically activated by fructose 2,6-bisphosphate (15). PYK is particularly suitable for studying allosteric regulation because unlike many other allosteric enzymes, the product does not bind to the effector site and participate in allosteric regulation. The PYK allosteric mechanism has also been investigated using a large number of site point mutants (16 -19). A general conclusion from the wealth of data is that intersubunit interactions on the A-A and C-C interfaces strongly influence the allosteric effect whereas mutations affecting the intrasubunit A-C interface are less sensitive (20).

EXPERIMENTAL PROCEDURES
Expression and Purification-LmPYK was overexpressed and purified by a modified version of the published protocol (21). Sulfate molecules are commonly observed bound to the effector and active sites of LmPYK structures (22), competing with the binding of the natural glycolytic ligands. Additional purification steps were introduced to remove contaminating sulfate molecules. Briefly, LmPYK-sulfate samples were concentrated and buffer exchanged (PD-10 column; Amersham Biosciences) into buffer A (50 mM triethanolamine-HCl (TEA) buffer (pH 7.2), 20 mM KCl, 20% glycerol) using standard protocols. LmPYK samples were loaded onto a (35-ml) DEAE-Sepharose ion-exchange column at 0.5 ml min Ϫ1 , preequilibrated in buffer A. The column was washed (3.0 ml min Ϫ1 ) with 10 column volumes of buffer A. LmPYK was eluted over a 5-column volume elution gradient (0 -60%) with buffer B (50 mM TEA buffer (pH 7.2), 200 mM KCl, 20% glycerol). Fractions containing LmPYK were pooled, concentrated, and buffer exchanged into buffer C (20 mM TEA buffer (pH 7.2) and 20% glycerol). LmPYK samples were concentrated to 30-mg ml Ϫ1 , and aliquots were stored at Ϫ20°C for up to 3 months.
Site-directed Mutagenesis and Characterization-Site-directed mutagenesis of the L. mexicana PYK gene was performed on plasmid pET28a-LmPYK (17). For each mutation, two complementary oligonucleotides containing the desired mutation were synthesized. The total volume of amplification mixture was 50 l containing 100 ng of plasmid, 0.5 g of each primer, a 200 M concentration of each of the four deoxynucleotides, and 2.5 units of Pfu polymerase. PCR was performed using the following program: first 1 min 95°C; 16 cycles: 30 s 95°C, 1 min 55°C and 14 min 72°C; and a final incubation of 10 min at 72°C. 10 units of the DpnI restriction enzyme was then added directly to each amplification reaction. The reaction mixtures were incubated at 37°C for 1 h to digest the parental DNA and were used to transform Escherichia coli XL1-blue cells. The presence of the mutations and the absence of other changes in the gene were ascertained by sequencing. The mutated plasmids were introduced into E. coli BL21(DE3) for gene expression. The mutant LmPYKs were expressed and purified as described in Ref. 17 but using slightly modified conditions; after induction of expression by the addition of 1 mM isopropyl 1-thio-␤-D-galactopyranoside, growth was continued overnight at 16°C. Enzyme assays and kinetic studies were performed as described previously (17).
Crystallization and Data Collection-Purified LmPYK aliquots (30 mg ml Ϫ1 ) were diluted to 15 mg ml Ϫ1 using a buffer containing 20 mM TEA buffer (pH 7.2). For co-crystallization experiments, the appropriate ligand(s) at a final concentration of 3.3 mM were added to the protein sample. All reagents used were of the highest purity available. Both LmPYK and complex crystals were obtained at 4°C (except ATP/oxalate crystals, which were obtained at 17°C) by vapor diffusion using the hanging-drop technique (23). The drops were formed by mixing 1.5 l of protein solution with 1.5 l of a well solution, composed of 10 -16% PEG 8000, 20 mM TEA buffer (pH 7.2), 50 mM MgCl 2 , 100 mM KCl, and 10 -15% glycerol. The drops were equilibrated against a reservoir filled with 0.5 ml of well solution. Crystals grew to maximum dimensions after 1 week. Prior to data collection, crystals were equilibrated for 24 h over a well solution composed of 12-18% PEG 8000 (2% above the reservoir concentration), 20 mM TEA buffer (pH 7.2), 50 mM MgCl 2 , 100 mM KCl, and 25% glycerol, which greatly improved the resolution of diffraction and eliminated the appearance of ice rings. Intensity data were collected at the European synchrotron radiation facility in Grenoble, France, on beamlines ID23-1 and BM14, and also at the Diamond synchrotron radiation facility in Oxfordshire, United Kingdom, on beamline IO3 from single crystals flash frozen in liquid nitrogen at 100 K. Data were then processed with MOSFLM (24) and scaled with SCALA (25).
Structure Determination-Generally, LmPYK structures were solved by molecular replacement using the program PHASER (26). A monomer from the previously determined structure of LmPYK (Protein Data Bank code 1PKL) (21) divided into two ensembles, ensemble 1 (residues Pro 87 -Pro 187 , a complete B-domain) and ensemble 2 (residues 1-86, 188 -481, 489 -498), served as search models. The resulting model was divided into three rigid body domains (B-domain, 87-187; A-domain, 1-86 and 187-356; and C-domain, 358 -498) and subjected to 10 cycles of rigid body refinement using the program REFMAC (27). When appropriate, ligands and water molecules were added to the models. Models were then subjected to cycles of restrained refinement (except the LmPYK⅐Fru-2,6-BP (5 Å) model), with manual adjustments to ligands and side chains using the program COOT (28). Figures were generated using PyMOL (29). A more detailed description of the refinement processes for each of the different LmPYK crystal structures can be found in the supplemental text.
Protein Data Bank Accession Codes-The atomic coordinates of the apoenzyme (3HQN), ATP and oxalate (3HQO), effector Fru-2,6-BP plus ATP and oxalate (3HQP), and effector only (3HQQ) have been deposited in the Protein Data Bank (entry codes shown in parentheses).
Analysis of Model Geometry-The geometry of the model was assessed using MolProbity (30). Although electron density was well defined for Thr 296 (a key active site residue), it commonly exhibits geometry outwith the Ramachandran plot in many PYK structures. This is primarily due to a restricted geometry, which facilitates interactions with active site ligands.
Structural Analysis-Superpositions of PYK structures were performed using both PyMOL and CCP4 superpose. The allosteric rigid body rotations were calculated by simultaneously superposing the A-and C-domains (AC cores, residues 18 -86 and 188 -480) of the T-and R-state tetramer structures (root mean square fit of the C␣ atoms is 1.86 Å). The resulting coordinates were recorded for both structures. Using the new coordinates, CCP4 SUPERPOSE was used to superpose each individual chain of the inactive onto the active structure (the average root mean square fit of the C␣ atoms for each chain is 0.58 Å), providing both the centroid and the rotation matrix. The rotation matrices were used to calculate the angle of rotation.
Thermal Shift Assay-Solutions of 25 l of Sypro Orange (diluted 1/500 in 20 mM TEA buffer (pH 7.2); Molecular Probes), 1 l of each ligand (ligands were dissolved in 100 mM TEA buffer (pH 7.2) to 100 mM), 5 l of 10ϫ metal solution (200 mM TEA buffer (pH 7.2), 500 mM MgCl 2 , and 1 M KCl), and 1 l of 2.0 mg ml Ϫ1 protein were added to the wells of a 96-well PCR plate (Bio-Rad), and the final volume was adjusted to 50 l using 20 mM TEA buffer (pH 7.2). Buffer was added instead of test ligand/metals in the control samples. The plates were sealed with optical quality sealing tape (Bio-Rad) and heated in an i-Cycler iQ5 real-time PCR detection system (Bio-Rad) from 20 to 80°C in increments of 1°C. Fluorescence changes in the wells of the plate were monitored simultaneously with a chargecoupled (CCD) camera. The wavelengths for excitation and emission were 485 and 575 nm, respectively. The temperature midpoint for the protein unfolding transition, T m , was calculated using the Bio-Rad iQ5 software.

RESULTS AND DISCUSSION
The T-to R-state Transition: Chain Rotations around a Single Pivot Point-The different LmPYK conformers observed in the four new crystal structures (see supplemental Table S2 for data collection and refinement statistics) were crystallized under near identical (physiological) conditions (pH 7.2, 50 mM MgCl 2 , 100 mM KCl, 10 -16% PEG 8000). The structure of the apoenzyme (LmPYK in the T-state with no ligands bound; see supplemental text T1) is shown in diagram form in Fig. 2a; the complex with active site ligands (LmPYK⅐ATP⅐OX with Mg 2ϩ ATP and Mg 2ϩ oxalate) in Fig. 2b; the R-state conformer with effector site plus active site ligands (LmPYK⅐ATP⅐OX⅐Fru-2,6-BP, supplemental text T2) in Fig. 2c; and the complex with effector ligand (LmPYK⅐Fru-2,6-BP) in Fig. 2d. The transition from the inactive (T) to active (R) state involves a rigid body rotation of the A-and C-domains (AC core, residues 18 -86 and 188 -480) of 6 o around a pivot point that lies at the base of the ␣␤-barrel of domain A (Fig. 2e, supplemental Fig. S2, and supplemental text T2 and animated gif file).
LmPYK Co-crystallized with Substrates Alone Is Stabilized in the R-state-LmPYK with active-site ligands alone adopts an R-state AC core conformation nearly identical to that of the fully ligated complex (as shown by the low (0.39 Å) root mean square fit of the AC cores for LmPYK⅐ATP⅐OX and LmPYK⅐ATP⅐OX⅐Fru-2,6-BP), although the B-domains are in different conformations (Fig. 2b and supplemental text T3). The bound oxalate molecule (an analogue of the enol pyruvate moiety of phosphoenolpyruvate and pyruvate) (31), mimics the major interactions of phosphoenolpyruvate and binds the short unstable (22) A␣6Ј helix in the active site. The 6 o rigid body rocking motion (Fig. 2e) of LmPYK moves a crucially important Arg 310 side chain (mutation of the equivalent Arg in E. coli PYK results in total inactivity (32), and a mutation to either Lys or Trp results in PYK deficiency in humans (19)) into the vicinity of the active site of the adjacent subunit (Fig. 2b), where it forms two stabilizing hydrogen bonds with the backbone carbonyls (residues Arg 262 and Gly 263 ) belonging to the oxalate (phosphoenolpyruvate)-stabilized A␣6Ј helix. This series of intersubunit hydrogen bonds formed along the large A-A interface stabilizes an R-conformation (Fig. 2b). Thus, the 6 o rigid body T-to R-state transition is shown to be required for enzyme activity; however, it does not explain how effector binding increases k cat /S 0.5 by a factor of 7 (supplemental Table S4).
Effector Molecule Fru-2,6-BP Locks the LmPYK Tetramer in the R-state-To separate the allosteric effects of active site ligands from those of effector site ligands, it is important to examine the individual complexes. The structure of LmPYK⅐Fru-2,6-BP (supplemental text T4) provides an answer to the controversial question of how binding of the effector molecule can influence enzyme activity at a site Ͼ40 Å away. LmPYK⅐Fru-2,6-BP was determined at 5-Å resolution and contained 24 monomers (six complete tetramers)/asymmetric unit. The LmPYK⅐Fru-2,6-BP crystals diffracted beyond 3 Å, but to obtain adequate separation of reflections the resolution was limited to 5 Å. Despite the poor resolution of the x-ray data, noncrystallographic averaging of the six tetramers resulted in high quality electron density maps (supplemental Fig. S6a), and Fru-2,6-BP molecules were clearly identified at the effector site of all monomers in the asymmetric unit (supplemental Fig.  S6c). As in the LmPYK⅐ATP⅐OX structure, the AC cores of the LmPYK⅐Fru-2,6-BP tetramer adopt an R-state conformation (Fig. 2d), with a root mean square fit between these two tetramers of 0.48 Å. The major structural difference, however, is that Fru-2,6-BP binding results in ordered electron density for the effector loop (supplemental Fig. S6d), which is normally disor-dered in the absence of bound Fru-2,6-BP. This ordered effector loop (residues 481-487) pushes outward across the small C-C interface toward the adjacent chain in a conformation identical to that observed for the LmPYK⅐ATP⅐OX⅐Fru-2,6-BP structure. The stabilization of the effector loop results in the formation of four pairs of strongly stabilizing salt bridge interactions (Asp 482 -Arg 493 and Lys 484 -Glu 498 ) ( Fig. 3a and  supplemental Fig. S6e), as observed in the LmPYK⅐ ATP⅐OX⅐Fru-2,6-BP structure. These interactions link the adjacent C-domains and stabilize the small C-C interface, locking the LmPYK tetramer in the R-state, preprimed for highly efficient phospho transfer (Fig. 2d).
Movement of the Flexible Lid-like B-domain Is Not Allosterically Regulated-Comparisons of previously determined inactive and active PYK structures (supplemental Table S3

Removal of the Salt Bridge Locking Mechanism by Site-directed Mutagenesis Results in a Nonallosteric but Active
Enzyme-To test whether the enhanced activity stems from salt bridge formation consequent on effector binding, we examined a number of mutants (supplemental Table S4) including K484A, D482A, R349A, and E498A. The double mutant K484A/D482A should be sufficient to prevent formation of all eight salt bridges observed to form across the C-C interface upon effector binding. Using phosphoenolpyruvate as substrate (without effector), this mutant has a k cat /S 0.5 value of 1.5 ϫ 10 5 M Ϫ1 s Ϫ1 ; addition of effector made no significant difference to the enzyme efficiency (k cat /S 0.5 ϭ 1.4 ϫ 10 5 M Ϫ1 s Ϫ1 ). By contrast, addition of effector to wild-type enzyme enhanced the k cat /S 0.5 value ϳ7-fold. This result strongly supports the conclusion that the allosteric effect of Fru-2,6-BP is from stabilization of the C-C interface.
Binding of Fru-2,6-BP Plays the Most Important Role in Stabilizing the LmPYK Tetramer in Solution-The stabilizing effects of ligands were also analyzed using a thermal shift assay (34). Addition of effector molecule (Fru-2,6-BP) to the apoenzyme dramatically increases the melting temperature (T m ) from 40°C up to 59°C (Fig. 3, b and c). Addition of the active site ligands oxalate and ATP has a small additional stabilizing effect and increases the T m by 3°C to 62°C. Comparison of the x-ray structures shows that the only major differences observed between the LmPYK⅐ATP⅐OX⅐Fru-2,6-BP complex (Fig. 2c) (T m ϭ 62°C) and the LmPYK⅐ATP⅐OX complex (Fig. 2b) (T m ϭ  traces). The four stages of the unfolding process are shown: 1, LmPYK below 30°C is stable; 2, LmPYK in complex with active site ligands begins to dissociate into monomeric or dimeric species (T m ϭ ϳ40°C); 3, LmPYK in complex with active site ligands only, fully denatured; 4, LmPYK in complex with Fru-2,6-BP in the absence (blue) or presence (green) of active site ligands unfolds as a single, highly stable species, presumably due to additional salt bridges formed at the C-C interface. The presence of active site ligands conferred only a very modest increase in stability compared with the stabilization by Fru-2,6-BP. c, schematic representations of LmPYK crystal structures and their experimentally determined melting temperatures (T m ). 42/51°C, phosphoenolpyruvate T m ϭ 42/53°C) are the interactions formed by the effector loop. As discussed previously, these result in eight additional salt bridges across the C-C interface and correlate with a ϳ7-fold increase in k cat /S 0.5 upon addition of effector (k cat /S 0.5 3.2 ϫ 10 5 M Ϫ1 s Ϫ1 to k cat /S 0.5 2.4 ϫ 10 6 M Ϫ1 s Ϫ1 ; supplemental Table S4). Because there are no additional structural changes near the active site on addition of Fru-2,6-BP, this increase in enzyme activity must be related to the enhanced stability and rigidity of the protein upon effector binding.
The Allosteric Mechanism of LmPYK Requires an Extension of the Monod, Wyman, and Changeux Model, Which Incorporates Changes in Flexibility-Rigid domain movements and local flexibility are both important in the allosteric regulation of LmPYK. The principal tenet of the Monod-Wyman-Changeux model is that "oligomers undergo reversible transitions between discrete conformations which primarily affect the quaternary organization, preserve its symmetry and are accessible in the absence of ligand" (35). The proposed allosteric mechanism for LmPYK based on analysis of the AC cores of the x-ray structures presented in this paper (Fig. 2) fits this model: the transition between the T-and R-states is characterized by the symmetrical rocking motion of the AC cores, a movement that governs the all-important regulatory switch. However, careful experimental dissection of the allosteric mechanism using the structural and thermodynamic results presented here clearly shows that the T-to R-state transition is not solely responsible for controlling enzyme activity: two completely independent ligand binding events stabilize an R-state tetramer but only effector-binding enhances enzyme activity. We have shown that the rate increase is due to a reduction in enzyme flexibility. Ligand binding to the active site in the absence of effector stabilizes a less active (k cat /S 0.5 value of 3.2 ϫ 10 5 M Ϫ1 s Ϫ1 ) and less robust (ATP⅐OX complex, T m ϭ 42/51°C) R-state. However effector binding locks via salt bridges a highly active (k cat /S 0.5 value of 2.4 ϫ 10 6 M Ϫ1 s Ϫ1 ) rigidified R-state (T m ϭ 59°C).
The different binding events neatly explain how the effector molecule apparently primes the active site, situated Ͼ40 Å away, without the need for a message to be transmitted through the chain: when the tetramer is in a conformation stabilized by the effector molecule, it is also in the optimal conformation (for each of the four subunits) to bind substrate (Fig. 2e). The structural, kinetic, and thermodynamic data for the different ligated states of LmPYK explain its allosteric mechanism and provide a number of general principles that are likely to apply in many other allosteric systems. We would, for example, expect that effector-bound complexes are typically less flexible, have higher thermal stability, and have greater enzymatic activity; a generalization that is indeed consistent with disparate available published structural and enzymatic data from different allosteric proteins. In the particular case of PYK, these observations open the door to the development of "allosteric drugs" to tackle parasitic diseases and cancer.