Structures of Mammalian and Bacterial Fructose-1,6-bisphosphatase Reveal the Basis for Synergism in AMP/Fructose 2,6-Bisphosphate Inhibition*

Fructose-1,6-bisphosphatase (FBPase) operates at a control point in mammalian gluconeogenesis, being inhibited synergistically by fructose 2,6-bisphosphate (Fru-2,6-P2) and AMP. AMP and Fru-2,6-P2 bind to allosteric and active sites, respectively, but the mechanism responsible for AMP/Fru-2,6-P2 synergy is unclear. Demonstrated here for the first time is a global conformational change in porcine FBPase induced by Fru-2,6-P2 in the absence of AMP. The Fru-2,6-P2 complex exhibits a subunit pair rotation of 13° from the R-state (compared with the 15° rotation of the T-state AMP complex) with active site loops in the disengaged conformation. A three-state thermodynamic model in which Fru-2,6-P2 drives a conformational change to a T-like intermediate state can account for AMP/Fru-2,6-P2 synergism in mammalian FBPases. AMP and Fru-2,6-P2 are not synergistic inhibitors of the Type I FBPase from Escherichia coli, and consistent with that model, the complex of E. coli FBPase with Fru-2,6-P2 remains in the R-state with dynamic loops in the engaged conformation. Evidently in porcine FBPase, the actions of AMP at the allosteric site and Fru-2,6-P2 at the active site displace engaged dynamic loops by distinct mechanisms, resulting in similar quaternary end-states. Conceivably, Type I FBPases from all eukaryotes may undergo similar global conformational changes in response to Fru-2,6-P2 ligation.

Although five nonhomologous FBPases may exist in living organisms, the most prevalent (Type I) is the sole form of FBPase in eukaryotes and the primary form in many bacteria (16 -20). The mammalian and bacterial Type I enzymes are reportedly homotetramers (subunits labeled C1-C4 by convention) and exist in distinct quaternary states R and T. Physiological effectors influence the relative stability of the R-and T-states. AMP transforms the active mammalian R-state enzyme, observed in crystal structures in the absence of inhibitors (21)(22)(23)(24), to an inactive T-state (10, 24 -26). Transition to the T-state entails a rigid body rotation of 15°of the upper (C1-C2) subunit pair relative to the lower (C3-C4) subunit pair (10,27). Mutant FBPases with destabilized T-states adopt intermediate quaternary states, I R (27) and I T , 4 in the presence of AMP. In the porcine liver enzyme, the AMP-induced transition to the T-state displaces a dynamic loop (residues 52-72) from the active site, disrupting at least one of three binding sites for essential metal cofactors (23,24,27). The conformational change in the dynamic loop is consistent with competitive inhibition of catalysis by AMP with respect to Mg 2ϩ (28).
Four molecules of Fru-2,6-P 2 bind per tetramer (29,30). Fru-2,6-P 2 is a linear competitive inhibitor with respect to Fru-1,6-P 2 , lowers the concentration of AMP necessary for 50% inhibition of mammalian FBPases (AMP/Fru-2,6-P 2 synergism), and induces sigmoidicity in the binding of Fru-1,6-P 2 (7,31), an effect not caused by AMP (3). Binding studies are consistent with kinetics; the presence of Fru-2,6-P 2 enhances AMP binding (30). Moreover, the binding of one equivalent of Fru-2,6-P 2 elicits change in UV difference spectra similar to that induced by AMP (30). The conformational changes in FBPase that underlie these observations are unclear. Specifically, how FBPase responds to Fru-2,6-P 2 alone is not known. The Fru-2,6-P 2 ⅐AMP complex is isomorphous to the AMP complex (T-state) (26), and FBPase crystal structures with Fru-2,6-P 2 , in the absence of AMP, are in the R-state. These R-state structures, however, either lack density for the 2-phosphoryl group (a likely consequence of degradation of Fru-2,6-P 2 to fructose 6-phosphate (8)) or are the result of soaking R-state crystals with the inhibitor (9). The crystallization of FBPase with Fru-2,6-P 2 from solution and clear evidence of an intact ligand would go a long way to understanding the conformational changes induced by Fru-2,6-P 2 .
Reported here for the first time are co-crystallizations of porcine FBPase with Fru-2,6-P 2 . Ligation of the active site by Fru-2,6-P 2 in the presence of Mg 2ϩ or Zn 2ϩ causes a quaternary transition to the I T -state, similar to that induced by AMP. Hence, we propose that AMP/Fru-2,6-P 2 synergism arises from a distinct quaternary state of FBPase, formed in the presence of Fru-2,6-P 2 . FBPase from Escherichia coli provides a test of hypothesis. In the presence of phosphoenolpyruvate (PEP) or citrate, the E. coli enzyme is in an active R-state very similar to that of the porcine enzyme (32). Moreover, AMP (in conjunction with Glc-6-P) causes a global conformational change to a T-like state (33). Fru-2,6-P 2 , although absent in prokaryotes (34 -36), inhibits E. coli FBPase but without AMP/Fru-2,6-P 2 synergism (37). Demonstrated here is the binding of Fru-2,6-P 2 to the active site by a structure determination and by the competitive inhibition of E. coli FBPase with respect to Fru-1,6-P 2 . Unlike the porcine enzyme, however, Fru-2,6-P 2 causes no global conformational change in E. coli FBPase, consistent with the absence of AMP/Fru-2,6-P 2 synergism.
Preparation of Porcine and E. coli FBPases-Porcine FBPase was isolated as previously described (38) with minor modifications. Pooled FBPase, eluted from Cibacron Blue-Sepharose chromatography, was adjusted to pH 7.5-8.3 before loading onto a DEAE-Sepharose column (equilibrated with 5 mM MgCl 2 and 20 mM Tris-HCl, pH 7.5). A Sephadex G-50 column (equilibrated with 20 mM KH 2 PO 4 /K 2 HPO 4 , pH 7.0) replaced the dialysis step after DEAE chromatography. Protein (10 mg/ml in 20 mM KH 2 PO 4 /K 2 HPO 4 , pH 7.0) was filtered through a 0.22-m filter and flash frozen in 200-l aliquots using a dry ice/ethanol bath and stored at Ϫ80°C. Protein concentrations were determined by the method of Bradford (39), using bovine serum albumin as a standard. Selenomethionine-substituted E. coli FBPase was isolated as previously described (20).
Data Collection-Crystals were screened, and data were collected for the bacterial enzyme at Iowa State University on a Rigaku R-AXIS IVϩϩ rotating anode/image plate system using CuK ␣ radiation from an Osmic confocal optics system at a temperature of 110 K. Data were collected from the mammalian enzyme crystals at 100 K on Beamline 4.2.2 of the Advanced Light Source, Lawrence Berkeley Laboratory. The program d*trek (40) was used to index, integrate, scale, and merge intensities, which were then converted to structure factors using the CCP4 (41) program TRUNCATE (42).
Structure Determination and Refinement-The subunit from the canonical R-state of porcine FBPase (Protein Data Bank accession identifier 1CNQ), less ligands and water molecules, was transformed by CCP4 programs PDBSET (43) and LSQKAB (44) into a C1-C2 dimer. That dimer and the program AMORE (45) were used in a molecular replacement solution of the porcine Mg 2ϩ ⅐Fru-2,6-P 2 complex. The C1-C2 dimer of the refined Mg 2ϩ ⅐Fru-2,6-P 2 complex (less ligands and water molecules) was used in turn as the initial model for the Zn 2ϩ ⅐Fru-2,6-P 2 complex. The subunit of R-state E. coli FBPase (Protein Data Bank identifier 2OWZ), less ligands and water molecules, enabled a molecular replacement solution of the E. coli Mg 2ϩ ⅐citrate⅐Fru-2,6-P 2 complex.
Structural models underwent energy minimization followed by individual thermal parameter refinement using CNS (46). Force constants and parameters of stereochemistry were from Engh and Huber (47). Restraints for thermal parameter refinement were as follows: 1.5 Å 2 for bonded main-chain atoms, 2.0 Å 2 for angle main-chain atoms and angle side-chain atoms, and 2.5 Å 2 for angle side-chain atoms. Noncrystallographic restraints were not used in the refinement. Manual adjustments in the conformation of specific residues employed the program XTALVIEW (48). Ligands (Mg 2ϩ , Zn 2ϩ , Fru-2,6-P 2 , HPO 4 2Ϫ , and citrate) and water molecules were fit to omit electron density until no improvement in R free was observed. Water molecules with thermal parameters above 55 Å 2 or more than 3.2 Å from the nearest hydrogen bonding partner were removed from the final model. Protein geometry was analyzed using the program PROCHECK (49). Superposition of Structures-CCP4 programs PDBSET (43) and LSQKAB (44) and models of dimers and tetramers of FBPase were used in pairwise superpositions. Displacements between corresponding C ␣ positions in superimposed structures were measured using XTALVIEW (48). Superpositions employed porcine FBPase structures with Protein Data Bank identifiers 1CNQ, 1EYK, 1Q9D, and 1YYZ, as canonical R-, T-, I R -and I T -states, respectively. Citrate-bound E. coli FBPase (Protein Data Bank identifier 2OWZ) represents the R-state of the E. coli enzyme. The angle of rotation between subunit pairs in various quaternary states of FBPases is sensitive to the subset of residues used in superpositions. Established residue subsets were the basis of comparison of porcine structures with each other (27) and with E. coli structures (20). Determination of the subunit pair rotation first superimposes subunits C3-C4 of a tetramer onto the porcine R-state. The angle through which subunit pair C1-C2 must then rotate in order to be superimposed onto the subunits C1-C2 of the R-state determines the subunit pair rotation.
Kinetic Experiments-Activity assays employ the coupling enzymes phosphoglucose isomerase and glucose-6-phosphate dehydrogenase and monitor the formation of NADPH by either absorbance at 340 nm or fluorescence emission at 470 nm (27). Assays (total volume, 2 ml) were conducted at 22°C in 50 mM Hepes, pH 7.5, with 100 M EDTA and 150 M NADP ϩ . 150 mM KCl was present in porcine FBPase assays only. Saturating levels of Fru-1,6-P 2 (40 or 20 M) and MgCl 2 (10 or 5 mM) for E. coli and porcine enzymes, respectively, were used to measure specific activity. E. coli FBPase (but not porcine FBPase) is sensitive to the method of assay, so two assays were employed (32). Assays were either initiated by the addition of 1.4 g of enzyme (enzyme-initiated assays) or by incubating the enzyme in assay mixtures for 1-2 h at 22°C without MgCl 2 and then initiating the reaction by the addition of metal (metal-initiated assays). Porcine FBPase assays were initiated by the addition of metal.

RESULTS
Enzyme Purity-Preparations of native and selenomethionine-substituted E. coli FBPase, used for kinetic and structural investigations, respectively, have specific activities of 35-40 units/mg and appear as single bands on SDS-PAGE. Analysis of the N-terminal residue reveals a single type (methionine or selenomethionine), indicating no N-terminal proteolysis. Purified porcine FBPase used here has a specific activity of 30 units/mg, migrates as a single band on SDS-PAGE, and has a pH 7.5/9.6 activity ratio of 3.3, characteristic of a nonproteolyzed mammalian FBPase.
Kinetics of AMP and Fru-2,6-P 2 Inhibition-Enzymes isolated here have kinetic properties as previously reported for porcine (50) and E. coli (32) FBPases, including the presence or absence of AMP/Fru-2,6-P 2 synergy reported in previous investigations (6,7,37). Enzyme-or metal-initiated assays of E. coli FBPase result in different I 0.5 values and Hill coefficients for AMP (32). Regardless, Fru-2,6-P 2 has no effect on AMP inhibition of E. coli FBPase using either assay method (Fig. 1). Lines in A-C of Fig. 1 are from fits of the Hill equation, where V and V max represent the observed and maximum velocities, I is the concentration of AMP, I 0.5 is the concentration of AMP that gives 50% inhibition, and n is the Hill coefficient. Values of n are 1.0 Ϯ 0.2 and 1.7 Ϯ 0.2 for data of A and B, respectively, as is consistent for enzyme-and metal-initiated assays (32). In Fig. 1C, the Hill coefficient for AMP inhibition of porcine FBPase declines from 1.9 Ϯ 0.2 to 1.3 Ϯ 0.1 as the Fru-2,6-P 2 Inhibition of Fructose-1,6-bisphosphatase DECEMBER 7, 2007 • VOLUME 282 • NUMBER 49 concentration of Fru-2,6-P 2 increases. Moreover, Fru-2,6-P 2 clearly enhances the inhibition of porcine FBPase by AMP. I 0.5 values decline as a function of increasing concentration of Fru-2,6-P 2 (Fig. 1D). The decrease in I 0.5 of AMP follows Equation 2, which is of the same form as the Hill equation with n set to unity.
In the above, I 0.5 is the concentration of AMP causing 50% inhibition, I 0.5min is the limiting I 0.5 value for AMP inhibition in the presence of an infinite concentration of Fru-2,6-P 2 , I 0.5max is the observed value for 50% inhibition by AMP in the absence of Fru-2,6-P 2 , and ␣ 0.5 is the concentration of Fru-2,6-P 2 that produces 50% of the synergistic effect. The parameter ␣ 0.5 represents an apparent dissociation constant for Fru-2,6-P 2 . Its value from the data of D is 0.36 Ϯ 0.04 M, within range of observed dissociation constants (0.1-2 M) reported for Fru-2,6-P 2 from kinetics and binding experiments (6,7,29,31,51). Enhancement of AMP inhibition by Fru-2,6-P 2 is simply I 0.5max /I 0.5min , ϳ4.5-fold based on the data of Fig. 1. I 0.5 for Fru-2,6-P 2 inhibition also decreases as a function of AMP concentration (7) (data not shown).
Fru-2,6-P 2 -bound Structures of Porcine FBPase (Protein Data Bank Codes 2QVU and 2QVV)-Crystals of porcine FBPase belong to the space group P2 1 2 1 2 with a ϭ 59.2, b ϭ 165.6, c ϭ 79.6 Å (Fru-2,6-P 2 ⅐Mg 2ϩ complex) and a ϭ 59.0, b ϭ 165.7, c ϭ 79.2 Å (Fru-2,6-P 2 ⅐Zn 2ϩ complex) ( Table 1). The two mammalian structures are isomorphous and contain a C1-C2 subunit pair in the asymmetric unit. A second subunit pair (C3-C4, related by crystallographic symmetry) completes the biological tetramer. Models of porcine FBPase, as usual lacking electron density for residues at the N terminus, begin with Asn 9 and go to the C-terminal residue at position 337. Only residues [63][64][65][66][67][68][69][70] are not in strong electron density. These residues are part of the dynamic loop, which is in the disengaged conformation in both the Mg 2ϩ -and Zn 2ϩ -bound complexes of the porcine enzyme. Individual polypeptide chains of the dimer in the asymmetric unit have no significant differences in conformation. C ␣ atom alignments using all residues of each polypeptide chain, except for the last six residues at the C terminus, give root mean squared deviations of 0.15 and 0.22 Å for the Mg 2ϩ and Zn 2ϩ complexes, respectively, values in line with estimated coordinate uncertainty. The largest structural divergence, other than conformational differences of C-terminal residues, occurs where electron density is weak in the dynamic loop.
Strong electron density covers Fru-2,6-P 2 bound in the active sites of the Mg 2ϩ and Zn 2ϩ porcine complexes (Fig. 2). Mg 2ϩ is in metal site 1, coordinated by Asp 118 , Asp 121 , Glu 280 , a water molecule, and the 1-hydroxyl-group of Fru-2,6-P 2 ( Table 2). Glu 97 and five water molecules coordinate a second Mg 2ϩ near metal site 3. Two of these coordinated water molecules interact with the 2-phosphoryl group of Fru-2,6-P 2 . This is the first instance of metal binding near site 3 without the dynamic loop in its engaged conformation. In the engaged conformation, the metal at site 3 interacts with Glu 97 and Asp 68 . Zn 2ϩ is only at site 1, binding Asp 118 , Asp 121 , Glu 280 , one water molecule, and the 1-hydroxyl group of the inhibitor. Electron density maps (2F o Ϫ F c ) and (F o Ϫ F c ), calculated with or without modeled ligands, reveal only a sphere of electron density at the binding locus for the 5Ј-phosphoryl group of AMP, with no interpretable density for atoms of a phosphoryl sugar.
The Mg 2ϩ -and Zn 2ϩ -bound structures of porcine FBPase exhibit no significant differences beyond coordinate error with monomer alignments, resulting in an overall root mean squared deviation of 0.14 Å between corresponding C ␣ positions. The porcine Fru-2,6-P 2 complexes differ from the canonical R-state of FBPase by changes in quaternary structure and the conformation

Statistics of data collection and refinement
Space group and unit cell parameters are described under "Results." Parenthetical values pertain to the highest resolution shell of data. More than 99% of residues are in most favored or allowed regions of the Ramachandran plot, and no residues are in disallowed regions (49).
where i runs over multiple observations of the same intensity, and j runs over all of the crystallographically unique intensities.
c R free is the R factor based upon 10% of the data randomly culled and not used in the refinement.  of the dynamic loop. The rotation angle between subunit pairs C1-C2 and C3-C4 of the Fru-2,6-P 2 complexes differs by 13°from the canonical R-state and 2°from the T-state. The quaternary state of the porcine tetramer in its Fru-2,6-P 2 complexes is close to that of the I T -state (subunit pair rotation angle of 12°), recognized first in an enzyme complex with the pseudotetrapeptide OC252 (38). A survey of FBPase quaternary states and crystal forms (Table 3) reveals the R-state and the I T -state in two space groups (I222 and P2 1 2 1 2). Hence, the quaternary state is not a likely consequence of crystal "packing forces." Subunit pair rotations of individual R-state structures vary by ϳ0.5°, and the Mg 2ϩ and Zn 2ϩ Fru-2,6-P 2 complexes reported here differ by ϳ0.75°. Moreover, the OC252 complex, AMP-bound Ile 10 3 Asp FBPase, and Fru-2,6-P 2 -ligated struc-

Fru-2,6-P 2 Inhibition of Fructose-1,6-bisphosphatase
tures have subunit rotation angles of 12-13°. On the basis of crystallographic data then, the I T quaternary state is as well defined as any quaternary state of porcine FBPase. At 1.5 Å resolution, the Fru-2,6-P 2 ⅐Mg 2ϩ porcine complex is also the highest resolution structure of an inhibited FBPase. Conformational differences between R-state (product complex) and I T -state (Fru-2,6-P 2 complex) subunits indicate factors contributing to global conformational change in the porcine tetramer. In the presence of Fru-2,6-P 2 , movements in residues 97-99 and 123-126 block Asp 74 from its R-state, loopengaged conformation (Fig. 3). Moreover, metal activators do not recognize coordination site 2 and have difficulty recognizing site 3. Asp 74 hydrogen bonds with the backbone amide of Lys 71 in the loop-engaged conformation, and its mutation to alanine dramatically reduces k cat (52). Metal coordination is significant to the stability of the engaged loop, since mutations that disrupt the integrity of the loop also cause significant increases in the K a for metal activation (53). Hence, after a catalytic turnover, Fru-2,6-P 2 probably binds to an open active site containing a metal ion only at site 1. The position of the 2-phosphoryl group relative to that of the 1-phosphoryl group destabilizes metal coordination at sites 2 and 3, and conformational relaxation of residues 123-126 and 97-99 sterically precludes the engaged conformation of the loop. The Fru-2,6-P 2 -bound R-state subunit has high energy, and a transition to the I T -state occurs when the requisite number of molecules of Fru-2,6-P 2 bind to the R-state tetramer.
Fru-2,6-P 2 (32). One monomer is present in the asymmetric unit. The tetramer (reproduced from crystallographic symmetry) is in the canonical R-state, although Fru-2,6-P 2 is in the active site, covered by strong electron density (Fig. 2). Mg 2ϩ is at metal site 1, and citrate is at the allosteric activation site. The model has all 332 residues of E. coli FBPase in good or strong electron density.
The dynamic loop (residues 42-63) in association with Fru-2,6-P 2 appears ordered for the first time in a crystal structure of a bacterial Type I FBPase (Fig. 4), adopting a conformation similar to that of the engaged loop of porcine FBPase (23). The C-terminal half of the dynamic loops for the E. coli and porcine enzymes exhibit similarities in sequence and interactions (Table 4). Glu 60 (corresponding to Asp 68 in porcine FBPase) interacts with the 2-phosphoryl group of the inhibitor through water molecules and interacts directly with Arg 271 . Additionally, Wat 462 , which interacts with Glu 60 and Glu 89 , occupies the position corresponding to metal site 3 in the mammalian enzyme. Electron density levels and donor acceptor distances of ϳ2.8 Å are not consistent with the assignment of Mg 2ϩ to this locus.
The engaged dynamic loop for the E. coli complex stands in contrast to the porcine Fru-2,6-P 2 complexes that have disengaged loops. Interactions of a disengaged loop for E. coli FBPase (if it were to exist) would differ significantly from those of the porcine enzyme (Table 4). Ala 54 of porcine FBPase, for instance, packs with Ile 10 , Ile 59 , Ile 190 , and Ile 194 , but all corresponding residues differ for the E. coli enzyme. Replacements of Thr 46 and Leu 80 in porcine FBPase with Arg 38 , and Lys 72 , respectively, would introduce electrostatic charge in predomi- with heavy lines representing the Fru-2,6-P 2 -bound enzyme and light lines representing the superimposed R-state enzyme (products and metals are omitted to improve clarity). The superposition aligns residues that bind the 6-phosphoryl group of the fructophosphoryl sugars in each structure. Residues 97-99 move toward residues 123-125, precluding Asp 74 from interactions that stabilize the loop-engaged conformation. In contrast, Asp 66 in the Fru-2,6-P 2 complex of the E. coli enzyme (bottom) is in a conformation with interactions comparable with those of loop-engaged porcine FBPase. Conformational differences in Tyr 274 (E. coli enzyme) relative to those of Tyr 279 (porcine enzyme) (boxed area) may allow Glu 89 and Glu 90 of the E. coli enzyme to adopt conformations characteristic of a loop-engaged subunit even in the presence of Fru-2,6-P 2 . Conformational differences in the tyrosyl residues may arise from the substitution of Ala 87 for Val 95 of the porcine enzyme (boxed area). This drawing was prepared with XTALVIEW (48).
nantly hydrophobic regions of the loop-disengaged conformer. Hence, a stabilized disengaged loop in the porcine system could be the driving force for an R-to I T -state transition. The probable absence of a stabilized disengaged loop in the E. coli system may explain in part the lack of a global conformational change in that system in response to Fru-2,6-P 2 .
A successful model for AMP/Fru-2,6-P 2 synergism then must connect the action of AMP at its allosteric site to that of Fru-2,6-P 2 at the active site. AMP and Fru-2,6-P 2 complexes of mammalian FBPases are deficient in essential metal cofactors (63), and hence AMP/Fru-2,6-P 2 synergy could arise from the expulsion of metal ions from the active site by independent actions of AMP and Fru-2,6-P 2 . AMP expels metal ions by promoting an R-to T-state transition and a disengaged dynamic loop as confirmed by investigations in directed mutation, crystallography, and kinetics (23,24,50). Inorganic phosphate in product complexes coordinates metal activators at sites 2 and 3 (24), but Fru-2,6-P 2 (by virtue of a different location for its 2-phosphoryl group) destabilizes metal coordi-

TABLE 4
Corresponding residue types and interactions involving the dynamic loops of porcine and E. coli FBPases nation and the engaged conformation of the dynamic loop, evidently promoting an R-to I T -state transition. UV difference spectroscopy (30) and fluorescence emission from a tryptophan reporter group (50) support similar conformational responses to AMP and Fru-2,6-P 2 by mammalian FBPases. AMP/Fru-2,6-P 2 synergism correlates with the I 0.5 values for Fru-2,6-P 2 and AMP, and these in turn depend on the type of metal cofactor (31). Mn 2ϩ (followed by Zn 2ϩ and then Mg 2ϩ ) is most effective in stabilizing the engaged conformation of the dynamic loop against AMP-and Fru-2,6-P 2 -induced conformational change (50), and Mn 2ϩ -activated FBPase also requires the highest concentrations of AMP and Fru-2,6-P 2 in order to observe inhibition and synergism (31).
In binding experiments without metal activators and substrate, one bound molecule of Fru-2,6-P 2 generates a UV difference spectrum indicative of conformational change (30). However, in kinetics, where metal and substrate are present, Fru-2,6-P 2 inhibits without cooperativity. Hyperbolic inhibition of FBPase would result if the R-to I T -state transition occurs only in response to the binding of the fourth molecule of Fru-2,6-P 2 . If so, then a single engaged dynamic loop will maintain the R-state. Indeed, substrate saturation curves for FBPase, hyperbolic in the absence of Fru-2,6-P 2 , become sigmoidal in its presence (7,31). The binding of the first molecule of substrate to an I T -state tetramer saturated with Fru-2,6-P 2 would cause a transition to the R-state and enhance catalysis for the remaining subunits. Although speculative, this model is consistent with the observed data. In principle, the number of bound molecules of Fru-2,6-P 2 that causes an R-to I T -state transition could be determined using hybrid tetramers of FBPase.
A thermodynamic model for AMP/Fru-2,6-P 2 synergism in porcine FBPase assumes three quaternary states, R, T, and I T (Fig. 5). In the presence of substrate and metal cofactors and in the absence of AMP or Fru-2,6-P 2 , the porcine enzyme is in the R-state, defined by a subunit pair rotation angle of 0°. Transition to the T-state (subunit pair rotation of 15°) by a sequential allosteric mechanism requires a minimum of two bound AMP molecules (66). By binding to an R-state subunit, AMP drives a shearing motion of helices H1 and H2, disrupting intra-and intersubunit hydrogen bonds, the end result of which is an R-state AMP complex of high energy (27). A significant drop in free energy favors the transition from R(AMP) 2 to T(AMP) 2 . The equilibrium constant governing AMP inhibition is ⌬G 1 0 in Fig. 5.
Fru-2,6-P 2 promotes a transition from the R-to the I T -state. Since AMP inhibition retains a Hill coefficient above unity in the presence of Fru-2,6-P 2 , the I T -to T-state transition may yet require two molecules of bound AMP. The equilibrium constant for the binding of AMP in the presence of Fru-2,6-P 2 , however, is now determined by ⌬G 2 0 in Fig. 5. AMP/Fru-2,6-P 2 synergism appears if ⌬G 2 0 is Ͼ⌬G 1 0 . If a three-state system as described exhibits synergism, then single mutations may have little impact on the phenomenon. For instance, mutations that destabilize interactions of AMP will raise the energy of all AMP-associated forms of the enzyme by roughly similar amounts. As a consequence, ⌬G 1 0 and ⌬G 2 0 will become smaller by equal amounts, but ⌬G 2 0 Ͼ ⌬G 1 0 still holds. Similarly, mutations at the active site destabilize all the complexes with Fru-2,6-P 2 but have no effect on ⌬G 1 0 and ⌬G 2 0 . In fact, no mutation of a mammalian FBPase has eliminated AMP/Fru-2,6-P 2 synergism without severely reducing or eliminating inhibition by either ligand.
Fru-2,6-P 2 inhibition of liver FBPase in vivo may depend on levels of stored glycogen. The glucose gavage of fasted animals (liver glycogen-depleted) results in elevated levels of serum glucose and the rapid biosynthesis of glycogen; however, about half of the glucose in such newly synthesized glycogen comes from gluconeogenesis (70). Hence, gluconeogenesis could run unabated when glycogen reserves are low, and indeed levels of total Fru-2,6-P 2 rise sufficiently to inhibit FBPase only after the restoration of liver glycogen reserves (15, 70 -72).
To the extent that excessive hepatic gluconeogenesis contributes to diabetic hyperglycemia (70,73), mammalian FBPase is a target for the development of anti-diabetic drugs (38,74,75). In principle, the reduction of gluconeogenic flux via the inhibition of FBPase could reduce serum glucose levels. Several novel inhibitors have been developed (38,74), and detailed knowledge of the structural mechanism of AMP inhibition has resulted in the rational design of a potential drug that targets the AMP site (75). This FBPase-specific inhibitor alleviates diabetic hyperglycemia in rats, providing a proof-of-principle for the use of FBPase inhibitors in diabetes therapy (76,77). On the basis of work presented here, inhibitors that destabilize the engaged conformation of the dynamic loop of FBPase or stabi-lize the disengaged conformation of that loop, will act synergistically with AMP. Indeed, the inhibitor OC252, which stabilizes the I T -state of porcine FBPase with a disengaged dynamic loop, exhibits such synergistic inhibition (38). AMP destabilizes the R-state and causes a transition to the T-state as for the porcine enzyme. ⌬G 1 0 determines AMP equilibrium binding affinity. Fru-2,6-P 2 (F26) stabilizes the R-state. AMP (perhaps in conjunction with Glc-6-P) destabilizes the Fru-2,6-P 2 R-state and causes a transition to an Fru-2,6-P 2 ⅐AMP T-state. ⌬G 2 0 determines AMP binding affinity in the presence of enzyme-bound Fru-2,6-P 2 . If Fru-2,6-P 2 stabilizes the R-and T-states equally, then ⌬G 1 0 ϭ ⌬G 2 0 , and no AMP/Fru-2,6-P 2 synergism occurs.