The allosteric regulation of pyruvate kinase.

Pyruvate kinase (PK) is critical for the regulation of the glycolytic pathway. The regulatory properties of Escherichia coli were investigated by mutating six charged residues involved in interdomain salt bridges (Arg(271), Arg(292), Asp(297), and Lys(413)) and in the binding of the allosteric activator (Lys(382) and Arg(431)). Arg(271) and Lys(413) are located at the interface between A and C domains within one subunit. The R271L and K413Q mutant enzymes exhibit altered kinetic properties. In K413Q, there is partial enzyme activation, whereas R271L is characterized by a bias toward the T-state in the allosteric equilibrium. In the T-state, Arg(292) and Asp(297) form an intersubunit salt bridge. The mutants R292D and D297R are totally inactive. The crystal structure of R292D reveals that the mutant enzyme retains the T-state quaternary structure. However, the mutation induces a reorganization of the interface with the creation of a network of interactions similar to that observed in the crystal structures of R-state yeast and M1 PK proteins. Furthermore, in the R292D structure, two loops that are part of the active site are disordered. The K382Q and R431E mutations were designed to probe the binding site for fructose 1, 6-bisphosphate, the allosteric activator. R431E exhibits only slight changes in the regulatory properties. Conversely, K382Q displays a highly altered responsiveness to the activator, suggesting that Lys(382) is involved in both activator binding and allosteric transition mechanism. Taken together, these results support the notion that domain interfaces are critical for the allosteric transition. They couple changes in the tertiary and quaternary structures to alterations in the geometry of the fructose 1, 6-bisphosphate and substrate binding sites. These site-directed mutagenesis data are discussed in the light of the molecular basis for the hereditary nonspherocytic hemolytic anemia, which is caused by mutations in human erythrocyte PK gene.

Pyruvate kinase (ATP:pyruvate 2-O-phosphotransferase (EC 2.7.1.40); PK) 1 catalyzes the last step of glycolysis, where the phosphoryl group of phosphoenolpyruvate (PEP) is transferred to ADP to form pyruvate and ATP. PK requires monovalent (K ϩ ) and divalent cations (Mg 2ϩ or Mn 2ϩ ) for its activity. The reaction is essentially irreversible under physiological conditions and is critical for the control of the metabolic flux in the second part of glycolysis. Moreover, the substrate PEP and the product pyruvate are involved in a variety of metabolic pathways. Such a central position in the cellular metabolism is reflected in the regulatory properties of PK, which is a typical allosteric protein (1). The activity is controlled by several physiological effectors, including H ϩ , Mg 2ϩ , Mn 2ϩ , and K ϩ (2). Furthermore, the enzyme displays sigmoidal kinetics toward the substrate PEP and it is activated by an heterotropic effector whose nature depends on the organism (1,3). The mammalian isoenzymes R (expressed in erythrocytes), L (in liver), and M2 (in kidney and lung) are regulated by fructose 1,6-bisphosphate (FBP) (4), whereas in trypanosomatid protozoans the allosteric effector is fructose 2,6bisphosphate (5). Most bacterial PKs are activated by FBP, although in some cases the effector is a monophosphorylated sugar such as ribose 5-phosphate. Thus far, the only known PK displaying hyperbolic kinetics is the mammalian M1 isoenzyme that is present in muscle, brain, and heart. However, the M1 protein is thought to be a highly specialized descendant of the allosteric PKs, locked in an active R-like conformation (6).
PK has been characterized from a number of prokaryotes and eukaryotes and in most cases has been found to exist as a tetramer of identical subunits, each consisting of approximately 500 residues ( Fig. 1A) (1). Crystal structures are available for the enzyme from cat and rabbit muscle (7,8), yeast (9), Escherichia coli (10), and Leishmania mexicana (11). These proteins share a very similar architecture. Each subunit consists of three domains (Fig. 1B): the A domain with the classic (␣/␤) 8 topology, the B domain with a somewhat irregular fold, and the C domain with an ␣ϩ␤ organization. The eukaryotic proteins contain an additional small N-terminal domain, which is absent in the E. coli enzyme. The active site is located on the C-terminal side of the A domain (␣/␤) 8 barrel, facing the cleft between the A and B domains. The complex between yeast PK and FBP has revealed that the FBP binding site is entirely located in the C domain, at more than 40 Å from the catalytic center (9). Four identical subunits are assembled to form the tetrameric enzyme with 222 symmetry. The subunits mainly interact through the A and C domains located along the molecular twofold axes (Fig. 1A).
Progress in the knowledge of the allosteric regulation was obtained by comparison of the T-state E. coli PK with the non-allosteric muscle M1 enzyme in the active R-like confor-mation (10). The study demonstrated that each individual domain approximates a rigid body. On transition from the T-to the R-state, the domains of the functional tetramer modify their relative orientation by up to 29°. These movements are coupled to a conformational change in the active center, which upon transition to the T-state undergoes a distortion of the PEP binding site. The recent structure determination of the yeast R-state (9) and the L. mexicana T-state (11) proteins helped to refine this model, by allowing a better discrimination between the genuine allosteric conformational changes and those due to the inherent structural divergence between the mammalian and bacterial proteins. However, the complete and detailed elucidation of the mechanism for the allosteric transition must await the determination of the structure of the same PK crystallized in both active and inactive conformations.
The "domain rotation model" implies that PK must be equipped with structural elements that enable the coupling of the domain movements to conformational changes in the active site. In this context, the interfaces between domains and subunits are predicted to be critical for the enzyme regulation (3,6,9,12). The intrasubunit interface between the A and C domains may take part in the transmission of the allosteric signal between PEP and FBP binding sites. Likewise, the contact region between twofold related A (A/AЈ interface along the vertical axis in Fig. 1A) and C domains (along the horizontal axis in Fig.  1A) may fulfil the role of coupling changes in the quaternary structure with modifications of the active center. To test these predictions, we have undertaken a site-directed mutagenesis study on the E. coli FBP-dependent PK. These experiments were guided by the knowledge of the three-dimensional structure of the enzyme crystallized in the T-state. Our approach was to mutate charged amino acid residues involved in interdomain salt bridges in the A/C and A/AЈ interfaces and to evaluate the effect of the substitutions on the enzyme activity and regulation. In addition, we replaced charged residues involved in the binding of FBP to elucidate the mechanism of responsiveness to the heterotropic activator. A total of six mutant proteins were prepared and analyzed by kinetic, thermal inactivation, and crystallographic methods. These data essentially agree with the proposed model for the allosteric transition and highlight the complexity of the enzyme, which has the remarkable ability to transmit a regulatory signal across long distances spanning different domains and subunits.

EXPERIMENTAL PROCEDURES
Materials-Restriction and DNA modifying enzymes were purchased from New England Biolabs. FBP, PEP, lactate dehydrogenase, and NADH were from Roche Molecular Biochemicals. Oligonucleotides were synthesized by Life Technologies, Inc. Other chemicals were reagent grade and obtained from Sigma or Aldrich.
Site-directed Mutagenesis-For mutagenesis the double-stranded plasmid pGV5A (13), a pBluescript II KS expression vector carrying the pykI gene, was employed. The procedure followed the protocol of Deng and Nickoloff (14), using the commercially available Chameleon™ double-stranded site-directed mutagenesis kit (Stratagene). The method consists of the denaturation of double-stranded plasmid DNA and resynthesis of one strand using two primers; one, the selection primer, changes a restriction site, whereas the second, or mutagenic primer, introduces the desired mutation. Following digestion with the enzyme recognizing the old restriction site to linearize the unmutated parental plasmid, the surviving plasmid molecules are transformed into a repairdeficient E. coli strain (XLmutS). The primers used for constructing the point mutations were the following (the underlined sequences indicate the mutated bases): R271L, AAATGTATCCGTGCACTTAAAGTCGTT-ATCACT; R292D, AACCCACGCCCGACTGACGCAGAAGCCGGTG-AC; D297R, CGCGCAGAAGCCGGTCGCGTTGCAAACGCCATC; K382Q, GCTACTCAGGGCGGTCAATCTGCTCGCGCAGTA; K413Q, CAGTTGGTACTGAGCCAAGGCGTTGTGCCGCAG; R431E, TCTACT-GATGATTTCTACGAACTGGGTAAAGAACTGGCT.
The selection primer used for changing the vector unique restriction site AflIII into StyI had the following sequence: CGCAGGAAAGACCT-TGGGAGCAAAAGGCC. Plasmids isolated from cultures of individual transformants were screened by StyI digestion. More than 50% of the plasmids containing the StyI site also contained the desired point mutation, as confirmed by DNA sequencing of the entire pykI gene. Sequencing of double-stranded DNA was performed by the dideoxynucleotide chain-termination method using Thermosequenase™ kit (Amersham Pharmacia Biotech) (15).
Overexpression and Purification-E. coli BL21 (DE3) cells transformed with the plasmid pGV5A or its mutant derivatives were grown overnight in a Luria-Bertani medium containing 100 g/ml ampicillin. The cells were then diluted (1/30 v/v) in a freshly prepared growth medium, and cultures were grown at 37°C until the optical density at 600 nm reached a value in the range 0.8 -1.2. At that point expression was induced by the addition of isopropyl-␤-D-thiogalactopyranoside to a final concentration of 0.6 mM. After 3 h of induction, the cells were harvested by centrifugation. Recombinant enzyme production was calculated by dividing the total activity of the culture by the specific activity of the purified enzyme after deduction of endogenous PK activity.
Recombinant wild-type and mutant forms of PK were purified as described previously (13). All buffers contained 1 mM EDTA and 2 mM ␤-mercaptoethanol. Cells collected by centrifugation were sonicated in a solution consisting of 10 mM Tris, pH 8.5, 100 mM KCl, 10 mM MgCl 2 , and 30 g/ml pancreatic DNase. After DNA digestion, cell debris was removed by centrifugation and the supernatant was loaded onto a DEAE Sephacel (Amersham Pharmacia Biotech) column (4 ϫ 32 cm) equilibrated with 10 mM Tris, pH 8.5, 100 mM KCl. The protein was eluted with 1200 ml of a linear gradient of KCl (0.1-0.5 M KCl in 10 mM Tris, pH 8.5) and subjected to ammonium sulfate precipitation (70% saturation). Next, it was applied to a Octyl-Sepharose CL-4B (Amersham Pharmacia Biotech) column (4 ϫ 20 cm) equilibrated with 50 mM Tris, pH 7.5, 100 mM KCl, and 1.17 M ammonium sulfate. Elution was performed with 1000 ml of a linear gradient of decreasing ammonium sulfate (1.17-0 M) and increasing ethylene glycol (0 -50% v/v) concentrations. The enzyme preparation was divided into two aliquots, and each of them was applied to a Sephacryl S-200 (Amersham Pharmacia Biotech) column (1.9 ϫ 100 cm) equilibrated in 10 mM Tris, pH 7.5. The protein was eluted with the equilibration buffer. Protein concentration was determined according to Lowry (16), using bovine serum albumin as standard.
The level of expression of most of the mutants was comparable to that of the recombinant wild-type enzyme (30 -100 mg/liter of culture). The endogenous activity of the control cells transformed with pBluescript II KS was less than 150 units/liter of culture, that is less than 1 mg of endogenous enzyme/liter of culture. The final preparations were judged to be more than 95% pure from SDS-polyacrylamide gel electrophoresis analysis (17). In most cases, each preparation produced at least 20 -25  (19).
c Percentage of residues in most favored, allowed, generously allowed and disallowed regions of the Ramachandran plot as checked with the program PROCHECK (22). mg of purified enzyme/liter of culture. The endogenous PK activity was undetectable in the preparations of pure mutant proteins, as proved by the complete absence of activity in the preparations of the R292D and D297R mutants (see "Results").
Enzyme Activity Assay-PK activity was determined at 25°C by the lactate dehydrogenase-coupled spectrophotometric assay (18). The standard reaction mixture contained: 10 mM Hepes, pH 7.5, 10 mM MgCl 2 , 50 mM KCl, 2 mM PEP, 2 mM FBP, 2 mM ADP, 0.12 mM NADH, and 22 units of crystalline lactate dehydrogenase in a final volume of 1 ml. The reaction was started by adding enzyme solution. One unit is the amount of enzyme catalyzing the oxidation of 1 mol of NADH/min under the above conditions.
Kinetic Analyses-Enzymatic activity was assayed at 25°C using various concentrations of PEP, ADP, and FBP under conditions identical to those above except for substrates and effector. Kinetic parameters were determined as follows: for PEP at fixed concentration of 2 mM ADP in the absence and in the presence of 2 mM FBP; for ADP at 2 mM PEP and 2 mM FBP; for FBP at 1 mM PEP and 2 mM ADP. In all cases, the enzyme activity was assayed at 10 different concentrations of substrate or effector. All measurements were performed in triplicate, and the plot of Lineweaver-Burk was used to determine V max and apparent K m values. The Hill plot was used to determine the apparent S 0,5 (the substrate concentration giving one-half of V max ) and n H (Hill coefficient) values.
Thermal Stability Assay-Thermal stability studies were carried out at 55°C in the absence and in the presence of 2 mM FBP. The enzyme (100 -200 g/ml) was incubated in 10 mM Tris, pH 7.5, 50 mM KCl, 1 mM EDTA, 2 mM ␤-mercaptoethanol, and 2 mg/ml bovine serum albumin. Samples were removed at intervals and rapidly cooled in ice, and the enzyme activity measured as described previously.
Structure Determination-Crystals of the R271L and R292D mutants were obtained using the hanging drop vapor diffusion method at conditions (12% w/v polyethylene glycol 4000, 20 mM KCl, 20 mM MgSO 4 , 50 mM MES, pH 6.5) identical to those used for the wild-type enzyme (10). The absence from the crystallization medium of PEP, FBP, or other activating molecules is consistent with the protein being crystallized in the inactive T-state. The diffraction data for the R271L and R292D mutants were collected at the x-ray diffraction beam-line of Elettra (Trieste, Italy) at 100 K. Before freezing, the crystals were exposed for a few seconds to a solution containing 15% w/v polyethylene glycol 4000, 25% v/v polyethylene glycol 400, 20 mM KCl, 20 mM MgSO 4 , and 50 mM MES, pH 6.5. The data were processed using MOSFLM (written by A. G. Leslie) and programs of the CCP4 (19) suite. Crystals of the mutants are isomorphous to those of the wild-type protein and belong to space group P2 1 2 1 2 1 . Crystallographic refinement was performed with the program REFMAC (20), whereas model building was done using the program O (21). Tight non-crystallographic restrains were applied throughout the refinement following the same protocol used for the native structure (10). The free R-factor was calculated employing the same reflections used for the free R-factor calculations in the refinement of the wild-type protein. Table I gives a summary of the data collection and refinement statistics. Analysis of the model was done with the O (21), Procheck (22) and programs of the CCP4 suite (19). The drawings were generated with Molscript (23).

RESULTS
Site-directed mutagenesis was employed to investigate the allosteric regulation of E. coli PK. The residues subjected to mutagenesis (Fig. 1B) are located in the intrasubunit interface between the A and C domains (A/C interface, R271L and K413Q), in the interface between the A domains of twofold related subunits (A/AЈ interface, R292D and D297R), and in the FBP-binding site (K382Q and R431E). The mutant enzymes were overexpressed in E. coli cells and purified to homogeneity. Their kinetic ( Fig. 2 and Tables II and III)  tivation (Table IV) properties were characterized as described under "Experimental Procedures." Moreover, two mutants, R271L and R292D, were analyzed by x-ray crystallography.
The A/C Interface: R271L and K413Q-An intriguing feature of the PK structure is the large 40-Å separation between the FBP binding site, located in the C domain (9), and the  catalytic center, in the A domain (Fig. 1B). The enzyme must be equipped with structural components that are able to transmit the allosteric signal over such a long distance across the two domains. The crystal structure of the T-state E. coli PK shows that the interface between the A and C domains comprises several polar amino acids. Among them, the positively charged side chains of Arg 271 and Lys 413 are engaged in interdomain salt bridges. These residues were chosen as target for sitedirected mutagenesis to probe the role of interdomain interactions in the allosteric transition and signal transduction between catalytic and allosteric sites. Lys 413 belongs to the loop connecting helix C␣4 and strand C␤3 of the C domain (Fig. 1B). Its side chain amino group makes a salt bridge with the carboxylate of Asp 336 , which is located on the surface of the A domain (␣/␤) 8 barrel (distance between Lys 413 -NE and Asp 336 -OD1 of 2.7 Å). The K413Q mutation introduces a side chain, which, although uncharged, is expected to produce little perturbation on the protein threedimensional structure. However, thermal inactivation studies revealed that the loss of the positive charge had a marked effect on the enzyme function. Incubation of the K413Q mutant at 55°C leads to a rapid enzyme inactivation, both in the presence and absence of FBP (Table IV). Moreover, several kinetic parameters are considerably affected by the mutation (Tables II  and III). The Hill coefficient for FBP binding decreases from a value of 1.96 in the wild type to a value of 1.0 in the mutant, implying that cooperativity in FBP binding has been abolished. Furthermore, the saturation curve against PEP is shifted toward the left and has a less pronounced sigmoidal character, when compared with the wild-type protein (Fig. 2A). These features consistently indicate that the mutation induces a partial activation of the enzyme by altering the R 7 T equilibrium in the direction of the R-state.
R271L is the second mutant designed to probe the role of the A/C interface in the PK regulation. Arg 271 lies on a loop connecting helix A␣6 and strand A␤7 of the A domain (␣/␤) 8 barrel (Fig. 1B). In the T-state, this residue forms an interdomain salt bridge with Glu 356 , a side chain belonging to the C domain (Fig.  3A). Arg 271 was mutated to leucine and the mutant analyzed by x-ray crystallography in its inactive T-form at 2.8-Å resolution ( Table I). The overall three-dimensional structure of R271L is virtually identical to that of the wild-type protein. The root mean square deviation for the equivalent C␣ atoms of the mutant and wild-type tetramer is 0.32 Å. Leu 271 adopts a conformation very similar to that of Arg 271 in the wild type. The hydrophobic leucine is not suited for a polar interaction with the charged Glu 356 carboxylate. As a result, Glu 356 side chain moves away from Leu 271 , reorienting its conformation and becoming fully solvent-exposed (Fig. 3B). Apart from this small local rearrangement, no other changes in the enzyme threedimensional structure could be identified.
Despite these small structural perturbations, the R271L mutation markedly affects several functional properties of the protein. The mutant is characterized by a 5-fold increase in the S 0.5 for FBP (Table III) and by a shift in the right direction of   the PEP saturation curve in the absence of the allosteric activator ( Fig. 2A). Moreover, the Hill coefficient for PEP in the presence of FBP is 1.6, rather than the value of 1.0 measured for the wild-type protein (Table II). Apparently, the mutation leads to a reduced responsiveness to FBP and to a bias the allosteric equilibrium toward the T-state, reflected in the shift of the saturation curve for PEP. The involvement of Arg 271 in the allosteric transition is further supported by the heat inactivation properties exhibited by the mutant. In the absence of FBP (e.g. T-state), the thermal inactivation is faster in the R271L protein than in the wild type, whereas, in the presence of the activator (R-state), the activity of the mutated protein displays a slightly slower decay (Table IV). In other words, the mutation makes the T-state more susceptible to heat inactivation in contrast to the R-state, which becomes more resistant. The crystal structure shows that the mutation effectively removes an interdomain salt bridge, thus providing a rationale for the T-state reduced stability (Fig. 3, A and B). On the other hand, on transition to the R-state, the conformation and local environment of Arg 271 may change significantly so that the salt bridge involving Arg 271 and Glu 356 may not be present in the R-form. Thus, the difference between the thermal inactivation properties of the T-and R-states may result from different conformations and/or environments of Arg 271 side chain in the active and inactive states.
The A/AЈ Interface: R292D and D297R-Comparison between the T-state E. coli PK structure (10) and the R-state conformations of the rabbit M1 isoenzyme (8) and yeast PK in complex with FBP (9) provided strong evidence for a pivotal role of the residues at the interface between the A domains of twofold related subunits (A/AЈ interface). In this region of the protein are located Arg 292 and Asp 297 , two residues that are strictly conserved among all known PK sequences (1,9,11). In the T-state, Arg 292 and Asp 297 form an intersubunit salt bridge (Fig. 4A). On transition to the R-state, the subunit and domain rotations are associated to a reorganization of the interface area with Arg 292 rotating away in the direction of the protein surface. Such a rearrangement would be coupled to a movement of loop 6 of the A domain (␣/␤) 8 barrel, allowing the active center to attain the conformation that is competent in PEP binding (10).
Two mutant enzymes were generated, R292D and D297R (Fig. 1B), with the purpose of destabilizing the T-state conformation, possibly inducing a bias in the allosteric equilibrium toward the R-state. The two mutants turned out to be totally inactive. R292D was crystallized in conditions identical to those used for the T-state wild-type protein and its threedimensional structure solved at 2.8-Å resolution ( Table I). The overall structure of the R292D is very similar to that of the wild type (root mean square deviation of 0.39 Å for 1784 equivalent C␣ pairs of the tetramer). This fact is of particular significance; it shows that E. coli PK has the remarkable ability to tolerate a charge reversal for a residue located at the heart of the subunit interface. A local rearrangement of a few side chains in the A/AЈ contact region is at the basis of this feature. In the mutant, the side chain of Asp 292 points toward the surface concomitant to the shift of Arg 244 , which moves closer to Asp 297 , so that these two charged residues form an intrasubunit salt bridge (Fig. 4B). Remarkably, such a network of interactions is very similar to that observed in the R-state conformation of the yeast and M1 PK proteins (8,9). In these structures, Asp 297 and Arg 244 (E. coli PK numbering) interact with each other, while the Arg residue at position 292 is oriented toward the surface in the same way as Asp 292 in the R292D mutant. The R292D mutation seems to elicit a local conformational change similar to that occurring in the allosteric transition. Such a hybrid situation of a local R-like conformation present in a T-state enzyme is reminiscent of what has been observed in the crystal structure of L. mexicana PK (11). This protein was crystallized in the T-state with eight crystallographically independent subunits in the asymmetric unit. Remarkably, the conformation of the residue homologous to Arg 292 differs among the eight independent monomers, varying from the R-state conformer (Arg 292 pointing toward the surface) to the T-state conformation (Arg 292 engaged in intersubunit salt bridge with Asp 297 ). Thus, both L. mexicana PK and the R292D mutant are able to retain the T-state quaternary and ternary structure despite the Arg 292 and surrounding residues being engaged in a sort of "R-type organization." In addition to the reorganization in the A/AЈ interface, the R292D protein displays another change with respect to the wild-type protein. In all four crystallographically independent subunits, residues 282-289 and 315-320 could not be located in the electron density map, reflecting a disordered conformation (Fig. 5). These amino acids form loops 7 and 8 of the A domain (␣/␤) 8  to the R292D replacement and the movement of Arg 244 . The interactions between Met 282 and the two interface arginine residues are apparently needed to anchor loop 7 in a fixed conformation. In their absence, the loop becomes free to adopt a disordered conformation. Disordering of loop 7 would in turn exercise a kind of "domino" effect on the adjacent loop 8 (Fig. 5), which also becomes disordered in the R292D mutant.
The FBP Binding Site: Mutants K382Q and R431E-The three-dimensional structure of yeast PK (9) crystallized with FBP revealed the exact location of the allosteric effector binding site. This is formed by a pocket on the C-domain at 40-Å distance from the active site within the same subunit. The structure shows that 1Ј-phosphate is in contact with the loop connecting strand C␤1 and helix C␣3 (residues 377-383 in E. coli PK), whereas the 6Ј-phosphate interacts with the loop C␤3-C␣5 (residues 420 -430). From comparison between the crystal structures, it can be inferred that the FBP site is only partially conserved in the E. coli protein. The residues interacting with the 1Ј-phosphate are highly homologous in sequence and conformation to those of the eukaryotic protein. In contrast, the loop binding the 6Ј-phosphate is truncated by six residues in the E. coli enzyme with respect to the yeast protein.
In order to probe the FBP binding site in E. coli PK, we designed two mutants, targeting residues in the 1Ј-and 6Јphosphate binding sites, respectively.
The crystal structure of the yeast PK (9) shows that the 6Ј-phosphate of FBP is engaged in a salt bridge with Arg 459 . This amino acid is at the beginning of helix C␣5 (Fig. 1B), a region that is poorly conserved in the E. coli protein. Inspection of the E. coli structure suggested that Arg 431 could fulfil a similar role to that of Arg 459 in the yeast protein. This hypothesis provided the rationale for the investigation of the R431E mutation. The mutant exhibits functional properties generally similar to those of the wild type (Tables II and III). A small decrease in the S 0.5 for both FBP and PEP and a shift in the titration curve for PEP indicate that the mutation has a small activating effect by altering the allosteric equilibrium in direction of the R-state. These observations indicate that Arg 431 is unlikely to directly interact with the 6Ј-phosphate of FBP. Rather, the marginal effects on the enzyme kinetics seem to suggest that this residue may take part in the allosteric response to FBP binding.
In yeast PK, Thr 406 is directly in contact with the 1Ј-phosphate. This amino acid is homologous to Lys 382 of the E. coli protein and belongs to a stretch of conserved residues (9). On this basis, a mutant was produced in which Lys 382 was replaced by an uncharged glutamine side chain. Kinetic charac-terization of the resulting K382Q mutant reveals a drastic alteration of the enzyme functional properties (Tables II and  III). First of all, the S 0.5 for FBP has a 70-fold increase and the mutant, unlike the wild-type protein, retains pronounced sigmoidal kinetics even in the presence of the allosteric activator (Fig. 2B). Clearly, the responsiveness to FBP is severely affected by the mutation, in perfect agreement with the idea that Lys 382 directly interacts with the effector. In addition, the mutant is characterized by a 6-fold and 20-fold decrease in the K cat /S 0.5 for PEP in the absence and presence of FBP, respectively (Table II). Such impairment in the activity of the enzyme suggests that the mutant is unable to undergo the full transition to the active R-state. Thus, Lys 382 is critical not only for FBP binding but also for the allosteric equilibrium. The threedimensional structure shows that Lys 382 is in contact with the loop connecting the C-terminal strands C␤4 and C␤5 of the C domain (Fig. 1B). The conformation of this loop has been proposed to be affected by FBP binding and to have a role in the mechanism of allosteric transition (11). The dramatic alterations in the regulatory and catalytic properties of the K382Q mutant nicely confirm this proposal and suggest that Lys 382 may provide a structural element coupling FBP binding to the transition to the R-state. DISCUSSION The control of PK activity is of paramount importance for the cell, given the central role of this enzyme in the cellular metabolism. Comparisons between the available structures of PK molecules from different species agree in indicating that the allosteric transition involves mutual rotations of both the domains within each subunit and the subunits within the tetramer. A key prediction of this model is that residues at the domain interfaces would have the critical function to relay the allosteric signal from and to the catalytic and regulatory sites. We performed site-directed mutagenesis studies targeting charged residues involved in interdomain salt bridges at the interface between the A and C domains and between the A domains of twofold related subunits. These experiments provide strong evidence for the critical role of the domain interfaces in the enzyme regulation.
The mutants R292D and D297R affecting the A/AЈ interface are totally inactive. The crystal structure of R292D reveals that the mutation induces a local reorganization of a few interface residues. Remarkably, this turns out to be coupled to the disordering of two loops, which are part of the active site. This observation is of particular significance; it proves the existence of a direct link between the catalytic center and the residues at the A/AЈ interface. In particular, given their strategic location at the heart of the intersubunit interface, Arg 292 and neighboring residues are able to couple the subunit rotations occurring in the T 7 R transition to the alterations in the active site geometry. However, the crystal structure also shows that R292D retains the T-state quaternary structure. This implies that a local rearrangement in the A/AЈ interface is not sufficient to trigger the full transition to the R-state, despite being instrumental to the transmission of the allosteric signal. Apparently, only the binding of either PEP and/or FBP is able to elicit all the conformational changes that result in the attainment of the active R-state conformation.
The two mutants targeting the A/C interface display altered kinetic and regulatory properties. In the K413Q protein there is partial enzyme activation, in contrast to the stabilization of the inactive T-state conformation, which characterizes the R271L variant. Additionally, thermal inactivation studies show that the interface amino acids greatly contribute to the enzyme stability. In particular, the data on the R271L protein indicate that this contribution may vary in the T-and R-states depending on the specific interactions established by the residue in two forms. Such a differential effect on stability may affect the allosteric equilibrium, thereby representing an additional factor that influences the enzyme regulatory properties.
The K413Q and R271L mutants do not exhibit significant alterations in the kinetic parameters for ADP binding (Table  II). This is consistent with the idea that the two mutants do not alter the active site geometry. Rather, the mutations appear to genuinely affect the enzyme allostery. From this point of view, the mutagenesis data on the E. coli protein agree remarkably well with studies performed on Bacillus stearothermophilus PK (24,25). In this enzyme, Trp 433 , which is homologous to Lys 413 of E. coli PK, was mutated to Tyr, producing a protein with altered regulatory properties that are qualitatively similar to those of the K413Q E. coli mutant. Moreover, the effects of the W433Y mutation in the B. stearothermophilus enzyme were reversed by introducing a second mutation affecting another residue (Glu 356 , homologous to Asp 307 in E. coli PK) of the A/C interface. Clearly, the theme emerging from these mutagenesis studies on the two prokaryotic enzymes is that the A/C interface is a critical component for the enzyme regulation.
The R431E and K382Q mutations were designed to probe the FBP binding site in the E. coli protein. The K382Q mutant exhibits a huge change in the regulatory properties. This suggests that effector binding in E. coli PK will be generally similar to that observed in the crystal structure of the yeast enzyme (9), despite a relatively low level of sequence homology in the region forming the regulatory site. This observation points to the more general concept that eukaryotic and prokaryotic PK proteins are likely to share the essential features in their regulatory apparatus. The best evidence for this notion is given by the mutations in the human erythrocyte PK that cause the hereditary nonspherocytic hemolytic anemia. More than 100 pathological mutations have been identified (26). Most of them cluster in a few well defined regions, which include the domain interfaces and the FBP binding sites (3,26). For instance, the R479H mutation (27) in PK affects a residue that is homologous to Lys 382 of the E. coli protein. In the latter, the K382Q substitution severely affects the protein functionality, providing a rationale for the pathological effect of the R479H mutation. Likewise, the R510Q (28,29) substitution in the human PK affects a residue homologous to Lys 413 of the E. coli protein. The properties of the E. coli K413Q variant indicate that the human R510Q mutation will have an adverse effect on both stability and regulation, ultimately affecting the correct physiological function of the enzyme.
The portrait emerging from site-directed mutagenesis, crystallography, and genetics of PK is that of a very complex pro-tein. The architecture of PK consists of an assembly of domains and subunits in which allosteric and catalytic sites are able to communicate with each other across relatively long distances. Various protein regions, including domain interfaces and flexible domain linkers, couple changes in the tertiary and quaternary structures to alterations in the geometry of the active and allosteric sites. The full understanding of such sophisticated machinery represents a fascinating subject awaiting future studies.