Structures of pyruvate kinases display evolutionarily divergent allosteric strategies

The transition between the inactive T-state (apoenzyme) and active R-state (effector bound enzyme) of Trypanosoma cruzi pyruvate kinase (PYK) is accompanied by a symmetrical 8° rigid body rocking motion of the A- and C-domain cores in each of the four subunits, coupled with the formation of additional salt bridges across two of the four subunit interfaces. These salt bridges provide increased tetramer stability correlated with an enhanced specificity constant (kcat/S0.5). A detailed kinetic and structural comparison between the potential drug target PYKs from the pathogenic protists T. cruzi, T. brucei and Leishmania mexicana shows that their allosteric mechanism is conserved. By contrast, a structural comparison of trypanosomatid PYKs with the evolutionarily divergent PYKs of humans and of bacteria shows that they have adopted different allosteric strategies. The underlying principle in each case is to maximize (kcat/S0.5) by stabilizing and rigidifying the tetramer in an active R-state conformation. However, bacterial and mammalian PYKs have evolved alternative ways of locking the tetramers together. In contrast to the divergent allosteric mechanisms, the PYK active sites are highly conserved across species. Selective disruption of the varied allosteric mechanisms may therefore provide a useful approach for the design of species-specific inhibitors.


Summary
The transition between the inactive T-state (apoenzyme) and active R-state (effector bound enzyme) of Trypanosoma cruzi pyruvate kinase (PYK) is accompanied by a symmetrical 8 • rigid body rocking motion of the A-and C-domain cores in each of the four subunits, coupled with the formation of additional salt bridges across two of the four subunit interfaces. These salt bridges provide increased tetramer stability correlated with an enhanced specificity constant (k cat /S 0.5 ). A detailed kinetic and structural comparison between the potential drug target PYKs from the pathogenic protists T. cruzi, T. brucei and Leishmania mexicana shows that their allosteric mechanism is conserved. By contrast, a structural comparison of trypanosomatid PYKs with the evolutionarily divergent PYKs of humans and of bacteria shows that they have adopted different allosteric strategies. The underlying principle in each case is to maximize (k cat /S 0.5 ) by stabilizing and rigidifying the tetramer in an active R-state conformation. However, bacterial and mammalian PYKs have evolved alternative ways of locking the tetramers together. In contrast to the divergent allosteric mechanisms, the PYK active sites are highly conserved across species. Selective disruption of the varied allosteric mechanisms may therefore provide a useful approach for the design of species-specific inhibitors.  ). An additional C'-domain is also found in PYKs from Geobacillus stearothermophilis [2] (figure 2) and Staphylococcus aureus [6]. The active site is nestled between the A-and B-domains and is located approximately 39 Å from the effector site which is located in the C-domain. In the tetramer, adjacent C-domains form the C-C or 'small' interface, and neighbouring A-domains form the A-A or 'large' interface. The B-domain contributes a mobile lid at one end of the (α/β) 8 -barrelled A-domain and modulates access to the active site. The overall tetrameric structure of PYK is highly conserved across distant phylogenetic groups; however, strategies for the regulation of PYK activity vary substantially between species [2,[7][8][9][10][11]. Three distinct ligand-binding sites with affinities for a range of small molecules have been identified in PYK tetramers and are particularly well characterized in mammals, trypanosomatids, yeast and bacteria (figure 1b). These sites are designated here as (i) the active site, (ii) the effector site, and (iii) the amino acid binding site. Most PYKs display cooperative binding of the substrate PEP at the active site, with characteristic sigmoid binding curves and Hill coefficients > 1 (e.g. table 1). In this way, glycolytic flux can be augmented in response to the presence of elevated levels of PEP.

Background
Modulation of PYK activity by activators and inhibitors which bind to the effector site is well characterized (table 1). Fructose 1,6-bisphosphate (F16BP) is the allosteric activator of three of the four mammalian isoenzymes: RPYK (erythrocyte), LPYK (liver) and M2PYK (embryonic or tumour). The fourth isoenzyme is the constitutively active M1PYK found in skeletal muscle. The activity of M1PYK can also be attenuated by the binding of amino acids such as proline or phenylalanine at the amino acid binding site [9,23]. PYK from Escherichia coli is also activated by F16BP, but PYK from Geobacillus stearothermophilus is allosterically activated by AMP or ribose 5-phosphate (R5P) [8] which probably bind at a position corresponding to the effector site [2]. PYK from baker's yeast Saccharomyces cerevisiae is also activated by F16BP [17] (table 1). Interestingly, PYKs from plants and archaea appear to be unresponsive to effectors such as F16BP or R5P [24,25]. There are currently no structural data for plant or archaeal PYKs.   [26]. Trypanosomatid PYKs are allosterically activated by micromolar concentrations of fructose 2,6bisphosphate (F26BP), instead of F16BP which is effective only at millimolar concentrations [27] (table 1). The X-ray structures of PYK from L. mexicana [10] and T. brucei [12] have been determined. Here, the first apoenzyme and R-state structures for T. cruzi PYK (TcPYK) are presented, thereby providing a detailed structural and enzymatic comparison for each of the 'TriTryp' family members. Analysis of the R-and T-state PYK structures from trypanosomatids, mammals and bacteria also provides an explanation of the evolutionary divergence of allosteric regulation.

Materials
ADP, PEP, oxalate (OX), F16BP, F26BP, lactate dehydrogenase (LDH; rabbit muscle), polyethylene glycol (PEG) 8000, antibiotics and buffers were obtained from Sigma-Aldrich. NADH and EDTA-free protease inhibitor mixture tablets were from Roche, glycerol was from BDH Prolabo, IPTG was from Melford and salts were from Fisher Scientific. Restriction enzymes, vector and E. coli competent cells were from Novagen. N-terminal His 6 -tagged TcPYK was prepared as described previously [28]. Briefly, the expression of His 6 -tagged TcPYK was achieved by the T7lac promoter-driven system in E. coli BL21 (DE3) cells after adding IPTG to a final concentration of 1 mM. Pure protein was obtained by immobilized metal ion affinity chromatography (IMAC) followed by gel-filtration chromatography. N-terminally His 6 -tagged human pyruvate kinases (M1PYK and M2PYK) were expressed in E. coli BL21(DE3) cells and purified using IMAC and gel-filtration chromatography as described previously [16].

Enzyme activity assay and kinetics study
The activity and kinetics of TcPYK were determined as described for T. brucei PYK (TbPYK) previously [12]. Briefly, the activity of TcPYK was measured by following the decrease in NADH absorbance at 339 nm, where one activity unit is defined as the conversion of 1 µmol substrate per minute under standard conditions. The assay was performed at 298 K in 100 µl reaction mixtures containing 50 mM triethanolamine (TEA) buffer, pH 7.2, 50 mM KCl, 10 mM MgCl 2 , 10% glycerol, 0.5 mM NADH and 5 µg (3.2 U) of LDH. The turnover number (k cat ) of TcPYK was calculated by the enzyme specific activity divided by the subunit molar mass of 56769.1 (g mol −1 ). Enzyme kinetics with regard to ADP was studied at saturating concentrations of PEP (10 mM) and variable concentrations of ADP (from 0 to 2.5 mM). Enzyme kinetics with regard to PEP was studied at saturating concentrations of ADP (2.5 mM) and variable concentrations of PEP (from 0 to 10 mM) in the presence or the absence of 1 µM F26BP, or in the presence or the absence of 4.5 mM F16BP. Enzyme kinetics with regard to the effector F26BP was studied at 2.5 mM ADP, 0.4 mM PEP and variable concentrations of F26BP (from 0 to 1 µM), while enzyme kinetics with regard to the effector F16BP was studied at 2.5 mM ADP, 0.7 mM PEP and variable concentrations of F16BP (from 0 to 4.5 mM). The Michaelis-Menten equation or its extended equations were fitted to the experimental data to calculate the values of the kinetic parameters. The parameter S 0.5 is used instead of K m when the substrate binding shows cooperativity. The parameter Ka 0.5 refers to the concentration of activator at which half the maximal activation is observed.

Thermal shift assay (differential scanning fluorimetry)
Melting temperature (T m ) was determined by a thermal shift assay using a fluorescence method as described for TbPYK previously [12]. For TcPYK and TbPYK, the assay buffer contained 50 mM TEA, pH 7.2, 100 mM KCl, 10 mM MgCl 2 , 10× SYPRO Orange dye (the dye was supplied by Invitrogen (catalogue number S6650) at 5000× concentration in DMSO, and diluted in assay buffer for use), 4 µM TcPYK or TbPYK, in the presence or the absence of 1 µM F26BP.

Dynamic light scattering
The T m values for human M2PYK and human M1PYK were also determined by monitoring changes in light scattering with increasing temperature. Five microlitres of human M2PYK or human M1PYK (10 mg ml −1 ), 5 µl of 10× metals buffer (500 mM TEA, pH 7.2, 500 mM MgCl 2 , 1 M KCl) and 1 µl of each ligand      a The outlier residue is Thr296 in chain B, a key active site residue that is commonly found in this configuration in PYK structures. b The outlier residues are Ile134 in chain A which is modelled in the poor electron density, and Thr296 in both chains.
(50 mM stocks prepared in 100 mM TEA buffer, pH 7.2), the final volume was then adjusted to 50 µl using dilution buffer (10 mM TEA (pH 7.2) and 2% glycerol). A Zetasizer Auto Plate Sampler was used to determine the Z-average molecular 'size' in terms of the hydrodynamic diameter in solution. Changes in the Z-average or particle diameter with increased temperature (293-353 K in increments of 1 K) were monitored in automated mode (typically requiring a measurement duration of 150 s; 13 acquisitions were determined for each run and repeated in triplicate). The resulting data were then analysed using the manufacturer's software provided (Malvern Instruments Ltd, Malvern, UK).

Crystallization, data collection and structure determination
Purified TcPYK aliquots (30 mg ml −1 ) were diluted to 15 mg ml −1 using a buffer containing 50 mM TEA (pH 7.2) or supplemented with 3.5 mM ATP, 3.5 µM F26BP and 3.5 mM oxalate. Single crystals of TcPYK or TcPYK complexed with F26BP, Mg 2+ and oxalate (there was no evidence of ATP from the crystal structure) were obtained at 277 K by vapour diffusion using the hanging drop technique. The drops were formed by mixing 1.5 µl of protein solution with 1.5 µl of a well solution composed of 8-18% PEG 8000, 20 mM TEA buffer (pH 7.2), 50 mM MgCl 2 , 100 mM KCl and 10-20% glycerol. The drops were equilibrated against a reservoir filled with 0.5 ml of well solution.
Prior to data collection, crystals were equilibrated for 14 h over a well solution composed of 8-18% PEG 8000 (2% above the crystallization conditions), 20 mM TEA buffer (pH 7.2), 50 mM MgCl 2 , 100 mM KCl and 25% glycerol, which eliminated the appearance of ice rings. Intensity data were collected at the Diamond synchrotron radiation facility in Oxfordshire, UK on beamline I04 (TcPYK) or I03 (TcPYK-F26BP-OX-Mg) to a resolution of 2.50 Å (TcPYK), or 2.80 Å (TcPYK-F26BP-OX-Mg). All datasets were obtained from a single crystal flash frozen in liquid nitrogen at 100 K. The TcPYK structures were solved as described previously [28].
Superpositions of PYK structures were performed using both PyMOL [29] and CCP4 superpose [30,31]. CCP4 superpose was also used to calculate the allosteric rigid body rotations of TcPYK from the superposition of T-and R-state tetramers (AC cores) as described previously [10]. Both RMS difference numbers and rotation matrices were calculated in the superposition process.

TcPYK and TbPYK have similar kinetic parameters
Values of TcPYK kinetic parameters were determined for PEP in the presence and absence of both the trypanosomatid PYK allosteric activator F26BP and the more general activator F16BP at 298 K and are summarized in table 1, together with the parameters previously reported for TbPYK [12]. In the absence of F26BP, the S 0.5 values for PEP were 1.23 mM and 1.03 mM for TcPYK and TbPYK, respectively, both exhibiting positive cooperativity (h > 1) and similar specificity constants (k cat /S 0.5 values of 137 and 141 s −1 mM −1 , respectively). In the presence of F26BP, the S 0.5 value for PEP decreased ninefold and the specificity constant increased 11-fold (average k cat /S 0.5 value of 1514 mM −1 s −1 ) for both TcPYK and TbPYK, with positive cooperativity being completely abolished (n = ∼1). By contrast, addition of the  The polypeptide chain is shown as a cartoon. Ligands and interacting residues are shown as sticks. Residue Phe213 which potentially favours the binding of oxalate in the same pose as the substrate PEP is also shown in stick format. Chain B of the protein is coloured yellow and chain D is blue. The divalent metal ion Mg 2+ and monovalent ion K + are shown as a green sphere and a purple sphere, respectively. The Fo-Fc electron densities for Mg 2+ and K + are shown as grey meshes contoured at 3.0 σ and 4.0 σ , respectively. Water molecules are shown as red spheres. Potential interatomic interactions are indicated by purple dashed lines and the relevant distances are given in ångströms. Detailed metal coordination is shown in the electronic supplementary material, figure S1. general effector F16BP decreased the S 0.5 value for PEP fourfold and increased the specificity constant fivefold, although these effects were observed only at concentrations over 1000 times higher than that of F26BP. There are no data available for the intracellular concentrations of F16BP and F26BP in T. cruzi; however, in vivo concentrations for F16BP and F26BP have been determined for the bloodstream form of T. brucei as approximately 600 µM for F16BP and approximately 10 µM for F26BP (calculated from data in [32] and assuming 175 mg total cellular protein per millilitre total cellular volume [33]). In T. cruzi, there is an interestingly large difference in the Ka 0.5 values for F26BP (0.03 µM) and F16BP (1239 µM) (table 1). The corresponding Ka 0.5 values for T. brucei are 0.01 and 50 µM. Based on the suggested relative cellular concentrations of the FBPs and their PYK affinities, it is not out of the question that F16BP may also contribute to PYK activation in trypanosomatids.

Sequence identity explains different effector recognition
The TcPYK amino acid sequence is 81% identical to TbPYK and 76% to LmPYK (electronic supplementary material, table S1), with residues involved in substrate and effector binding conserved (figure 2). This high degree of sequence conservation between trypanosomatids is reflected in their similar kinetic parameter values (table 1). TcPYK and the F16BP-regulated human M2PYK are only 47% identical (electronic supplementary material, table S1), although the residues involved in substrate binding exhibit 100% sequence conservation (figure 2). The side chains involved in effector binding are much more variable and show only 57% sequence identity.

T-to R-state transition: a common pivot region
Both T-and R-state TcPYK X-ray crystal structures were obtained under near identical (pH 7.2) conditions (see table 2 1b). In a similar way, LmPYK experiences a rigid body rotation of its AC cores during the transition between T-and R-states [10]. A superposition of the TcPYK and LmPYK AC cores gave a C α RMS fit of 0.7 and 0.6 Å for the T-and R-states, respectively.
By contrast, superposition of the AC core of active R-state TcPYK with the equivalent human M2PYK structure (PDB ID: 3bjf) gave an average RMS fit of 2.3 Å. Besides the effector site, the major difference between these structures is an extended loop (inserted loop) between the C-terminal portion of helix Aα2 and the N-terminal portion of strand Aβ3 (figure 1a). This extended loop which is also conserved in human M1PYK corresponds to six additional amino acid residues (Ser 100 -Asp-Pro-Ile-Leu-Tyr 105 ) (figure 2) and seems to play a central role in regulating M2PYK activity. Mutation of Ile103 of human M2PYK to a tyrosine causes the formation of a monomeric form of M2PYK known as cytosolic thyroid hormone-binding protein p58 [34]. In addition, phosphorylation of Tyr105 has been shown to inhibit M2PYK activity [35]. Moreover, this region also comprises part of a known amino acid binding site in M1PYK (figure 1a). The pivot point is located close to the amino acid binding site and suggests that ligands binding to this site may affect the rocking motion of the AC cores and as a result modulate enzyme activity.

Effector loop conformation is responsible for allosteric control
The hydrogen bond interactions between F26BP and trypanosomatid PYKs are remarkably similar to those found in the F16BP-M2PYK complex (figure 6) which explains why TbPYK, TcPYK and LmPYK are able to bind both effectors (table 1). However, higher concentrations (at least 1000-fold) of F16BP over the concentration of F26BP are required to induce a (weaker) allosteric response compared with F26BP. In the TcPYK-F26BP-OX-Mg structure, the 2 -and 6 -phospho groups of the F26BP molecule form a series of hydrogen bonds similar to those of F16BP in the human M2PYK-F16BP-OX-Mg structure (cf. figure 6a,b), although the furanose ring adopts a significantly different pose. Pro490 in Trypanosoma  (figure 5g,i), an interaction essential for allosteric control [10]. Instead of forming additional salt bridges, the binding of F16BP to human M2PYK stabilizes a lysine-mediated 'peg-in-hole' connection across the C-C interface [16].

4.5.
Increased thermal stability correlates with increased k cat /S 0.5 The stabilizing effects of PYK cofactors, inhibitors and metals were analysed using thermal shift assays. An increase in melting temperature (T m ) reflects an increase in structural order and reduced conformational flexibility. Addition of the effector molecule (F26BP) to apo-TcPYK or apo-TbPYK increased the melting temperature by 2 K and 8 K, respectively (figure 5d,e). The addition of F16BP to M2PYK also showed a dramatic increase in the T m from 315 to 325 K (figure 7a), close to the T m of the non-allosterically regulated M1PYK, for which all complexes had identical T m values of 325 K (figure 7b). This increase in T m upon effector binding correlates with a 10-to 16-fold increase in specificity constant (k cat /S 0.5 ) (   [36] is also shown. The two structures were superimposed by fitting corresponding Cα atoms from the effector sites of both PYK structures. Effector binding stabilizes the effector loop (Ala482-Gly488) and the interaction of the 1 -CH 2 OH group (blue circle) of F26BP additionally holds the effector loop in a conformation that favours the formation of the allosteric salt bridge (Arg494...Asp483). (b) A close-up view of the effector site of the M2PYK-F16BP-OX-Mg structure showing the binding mode of F16BP. In contrast to the TcPYK-F26BP-OX-Mg structure, the effector loop in M2PYK-F16BP-OX-Mg moves further from the C-C interface, essentially wrapping around the effector molecule. For comparison, the position of F26BP from the TcPYK-F26BP-OX-Mg structure is also shown; the two structures were superimposed using selected corresponding Cα atoms from the effector site. Interestingly, the 1 -CH 2 OH group belonging to F26BP clashes (less than 2 Å; purple circle) with the human effector loop position. Green dashed lines highlight stacking interactions. 4.6. Changes in the C-C interface account for increased enzyme stability and k cat /S 0.5 Previous site-directed mutational studies on LmPYK have shown that the increased stability is the result of a series of stabilizing salt bridges formed across the C-C tetramer interface [10]. These interactions are also observed in the TcPYK-F26BP-OX-Mg structure (figure 5i), indicating a similar mechanism of regulation. Upon addition of effector an increase in T m is observed for all trypanosomatid PYKs correlating with a 10-to 14-fold increase in k cat /S 0.5 , clearly demonstrating the relationship between rigidity and activity.
The addition of effector molecule to apo-M2PYK dramatically increased the T m from 315 to 325 K (figure 7a) with a concomitant 16-fold increase in k cat /S 0.5 (table 1). A recent analysis of the oligomeric state of M2PYK provided a novel mechanism for tetrameric stabilization, whereby M2PYK exists in a monomer-tetramer equilibrium [16,37], and binding of the allosteric effector F16BP stabilizes the pegin-hole connection across the C-C interface of the tetramer to shift this equilibrium to a predominantly tetrameric state (figure 8e,f ). Addition of F16BP reduces thermal vibration along the C-C interface (apo-M2PYK T m = 315 K, shifting to T m = 325 K for effector bound) and locks the effector loops, stabilizing  the active tetramer conformation. We suggest that the 16-fold increase in specificity constant results from stabilization of the tetramer and promotes conformational states which favour substrate binding (8.6-fold decrease in S 0.5 ) and/or catalytic activity and/or product release. In addition, restriction in its conformational states reduces the ensemble of inactive conformational states thereby favouring substrate binding. The activity of M1PYK also correlates with tetrameric stability as the X-ray structure (PDB: 3SRF) shows a tighter peg-in-hole binding across the C-C interface compared with M2PYK which locks the enzyme in a constitutively active tetramer conformation (figure 7b). Tetrameric stabilization by means of effector binding has also been observed for the bacterial PYK from G. stearothermophilis (GsPYK). GsPYK exists as a tetramer and is regulated by the allosteric activator ribose 5-phosphate (R5P) [8]. Lovell et al. have shown that the addition of the effector molecule results in a highly thermostable GsPYK conformation, demonstrating the increased rigidity of effector bound complexes. Determination of the crystal structure of GsPYK [2] revealed that the C-terminal extension forms an extra domain (domain C') in place of the trypanosomatid PYK effector loop (figure 8a). This domain is predicted to have a hinge-like movement, possibly interacting with adjacent monomers and forming stabilizing bridges across the C-C interface (figure 8b), similar to those observed for the salt bridge effector loops of trypanosomatid PYKs (figure 8d).

Evolution of allosteric control mechanisms
Trypanosomatid PYKs are allosterically regulated by the binding of F26BP, human M2PYK (and many other PYKs) by F16BP, and certain bacterial PYKs by AMP or monophosphorylated sugars. While the regulatory molecules and molecular mechanisms may vary between species the underlying principle of tetrameric stabilization in response to effector binding is conserved.
Three types of allosteric mechanisms are found in evolutionarily divergent species of PYK (figure 8). (i) Bacterial PYKs as exemplified by the GsPYK tetramer (figure 8a,b) are stabilized by the rotation of a series of additional domains (domain C') which bridge across the C-C interface thereby stabilizing the tetramer. (ii) Trypanosomatid PYKs (figure 8c,d) possess simple loops (effector loops) which again form a series of bridges across the C-C interface enhancing tetramer stability. (iii) Human M2PYK (figure 8e,f ) has evolved in a different direction whereby it is able to dissociate into inactive monomers, and the active tetramer is formed and stabilized in response to effector binding. The allosteric characteristics observed for all PYKs suggest that effector bound PYKs have higher thermal stability and greater enzymic activity. Selective disruption of these allosteric mechanisms may provide new and specific drug targets which avoid the problem of developing selective inhibitors against similar active sites. Such inhibitors/activators could be used to tackle trypanosomiasis, cancer (M2PYK [39]), anaemia (RPYK [40]) and bacterial infection (methicillin-resistant Staphylococcus aureus PYK [6]).