Central Cavity of Fructose-1,6-bisphosphatase and the Evolution of AMP/Fructose 2,6-bisphosphate Synergism in Eukaryotic Organisms*

Background: AMP and fructose 2,6-bisphosphate (Fru-2,6-P2) synergistically inhibit eukaryotic fructose-1,6-bisphosphatase (FBPase). How did this property evolve? Results: Directed mutations, kinetics, and structure determinations link the central cavity of FBPase to AMP/Fru-2,6-P2 synergism. Conclusion: A central cavity in any FBPase implies AMP/Fru-2,6-P2 and AMP/Fru-6-P synergism. Significance: AMP/Fru-2,6-P2 synergism of eukaryotic FBPases probably evolved from AMP/Fru-6-P synergism of an ancestral FBPase. The effects of AMP and fructose 2,6-bisphosphate (Fru-2,6-P2) on porcine fructose-1,6-bisphosphatase (pFBPase) and Escherichia coli FBPase (eFBPase) differ in three respects. AMP/Fru-2,6-P2 synergism in pFBPase is absent in eFBPase. Fru-2,6-P2 induces a 13° subunit pair rotation in pFBPase but no rotation in eFBPase. Hydrophilic side chains in eFBPase occupy what otherwise would be a central aqueous cavity observed in pFBPase. Explored here is the linkage of AMP/Fru-2,6-P2 synergism to the central cavity and the evolution of synergism in FBPases. The single mutation Ser45 → His substantially fills the central cavity of pFBPase, and the triple mutation Ser45 → His, Thr46 → Arg, and Leu186 → Tyr replaces porcine with E. coli type side chains. Both single and triple mutations significantly reduce synergism while retaining other wild-type kinetic properties. Similar to the effect of Fru-2,6-P2 on eFBPase, the triple mutant of pFBPase with bound Fru-2,6-P2 exhibits only a 2° subunit pair rotation as opposed to the 13° rotation exhibited by the Fru-2,6-P2 complex of wild-type pFBPase. The side chain at position 45 is small in all available eukaryotic FBPases but large and hydrophilic in bacterial FBPases, similar to eFBPase. Sequence information indicates the likelihood of synergism in the FBPase from Leptospira interrogans (lFBPase), and indeed recombinant lFBPase exhibits AMP/Fru-2,6-P2 synergism. Unexpectedly, however, AMP also enhances Fru-6-P binding to lFBPase. Taken together, these observations suggest the evolution of AMP/Fru-2,6-P2 synergism in eukaryotic FBPases from an ancestral FBPase having a central aqueous cavity and exhibiting synergistic feedback inhibition by AMP and Fru-6-P.

Escherichia coli FBPase (eFBPase) has 41% sequence identity with pFBPase but a fundamentally different mechanism of regulation (30 -32). eFBPase is subject to feed-forward activation by phosphoenolpyruvate, the binding of which favors an active tetramer over an inactive or less active dimer (30 -31). AMP and glucose 6-phosphate (Glc-6-P) are synergistic inhibitors that together drive the enzyme to a T-like state (32). Fru-2,6-P 2 , although not present in E. coli, is a potent inhibitor of eFBPase (14). Unlike pFBPase, Fru-2,6-P 2 /AMP synergism is absent in the kinetics of eFBPase, and, consistent with the hypothesis of Hines et al. (14), eFBPase remains in the R-state in the presence of Fru-2,6-P 2 .
The foregoing begs the following question. If subunit pair rotation in response to Fru-2,6-P 2 is necessary for synergism, then why do subunit pairs rotate in pFBPase and not in eFBPase? A large water-filled cavity is present at the center of the pFBPase tetramer. In eFBPase, however, hydrophilic side chains fill the corresponding region. Shown here by directed mutations, kinetics, and structure determinations, the central cavity of pFBPase enables the subunit pair rotation of the tetramer in response to Fru-2,6-P 2 . Moreover, sequence information implies the existence of bacterial FBPases that probably exhibit AMP/Fru-2,6-P 2 synergism, even in organisms unlikely to produce Fru-2,6-P 2 . The FBPase from one such organism, Leptospira interrogans, exhibits not only AMP/Fru-2,6-P 2 synergism but also the AMP-enhanced binding of Fru-6-P. Fru-2,6-P 2 as a dynamic regulator of FBPase in eukaryotic systems probably evolved then from a primordial FBPase subject to synergistic feedback inhibition by AMP and Fru-6-P.

EXPERIMENTAL PROCEDURES
Materials-Fru-1,6-P 2 , NADP ϩ and AMP came from Sigma. Fru-2,6-P 2 was prepared by published protocols (33). Glucose-6phosphate dehydrogenase and phosphoglucose isomerase were purchased from Roche Applied Sciences. The FBPase-deficient E. coli strain DF 657 was from the Genetic Stock Center at Yale University. Other chemicals were of reagent grade or equivalent.
Expression and Purification of Wild-type and Mutant pFBPases-E. coli strain DF 657 was used for expression of pFBPases. The cell culture of E. coli DF657, transformed with plasmid, grew to A 600 of 1.0, at which time transcription was induced by the addition of isopropyl-␤-D-thiogalactopyranoside (final concentration 1 mM). The culture was maintained (with shaking, 37°C) for an additional 16 h before harvesting. The supernatant solution of a cell-free extract was loaded onto a Cibracon Blue-Sepharose column previously equilibrated with 20 mM Tris-HCl, pH 7.5, and 5 mM MgCl 2 . The column was washed with 20 mM Tris-HCl, pH 7.5. Enzyme was eluted with 5 mM AMP and 20 mM Tris-HCl, pH 7.5. The pH of eluted protein solution was adjusted to 8.5 before loading onto a DEAE-Sepharose column equilibrated with 20 mM Tris-HCl, pH 8.3. Purified enzyme was eluted with a NaCl gradient (0 -0.5 M) in 10 mM Tris-HCl, pH 8.3, and then dialyzed extensively against 50 mM Hepes, pH 7.4, for kinetic investigations and for crystallization experiments. Purity and protein concentrations of FBPase preparations were confirmed by SDS-PAGE (34) and by the Bradford assay (35), respectively.
Cloning, Expression, and Purification of lFBPase-The gene for lFBPase (UniProt ID Q8F421) was synthesized and cloned into a pET24b plasmid from Genscript. The integrity of the construct was confirmed by sequencing the promoter region and the entire open reading frame as above for mutant constructs of pFBPase. Expression of lFBPase followed the procedure outlined above for pFBPase. The supernatant component of the cell-free extract was subject to ammonium sulfate fractionation. The 30 -50% ammonium sulfate fraction was collected and dialyzed in 20 mM Tris, pH 8.0, before loading onto a DEAE-Sepharose column. Purified enzyme was eluted with a NaCl gradient (0 -0.5 M) in 20 mM Tris-HCl, pH 8.0. Enzyme was dialyzed extensively against 50 mM Hepes, pH 7.4, for kinetics studies. Purity and protein concentrations of lFBPase preparations were confirmed by SDS-PAGE (34) and by Bradford assay (35), respectively.
Kinetics-Phosphoglucose isomerase and glucose-6-phosphate dehydrogenase were used as coupling enzymes in assays for FBPase (1). For specific activity measurements, reduction of NADP to NADPH was monitored by absorbance at 340 nm. Other assays used the same coupling enzymes but monitored the formation of NADPH by its fluorescence emission at 470 nm using an excitation wavelength of 340 nm. Assays were performed at 22°C in 50 mM Hepes, pH 7.5. Assay solutions for wild-type and mutant pFBPases contained EDTA and KCl at concentrations of 10 M and 150 mM, respectively, whereas KCl was absent in the assay solution for lFBPase. Data for AMP inhibition with respect to Mg 2ϩ and Fru-2,6-P 2 inhibition with respect to Fru-1,6-P 2 were fit to several models using Grafit (36), with best fits of data to models reported here.
Tryptophan Fluorescence-Fluorescence measurements were made at room temperature using a 1-cm 2 quartz cuvette on an SLM Amico 8000 fluorometer. The excitation wavelength was 295 nm, and emission scans were integrated from 310 to 370 nm. Fluorescence scans (repeat of three scans for each data point) were performed after the additions of small volumes of ligand to 2 ml of a 2 M lFBPase solution. The total volume of added titrant did not exceed 1% of the initial volume. Data fitting was carried out using protocols in the literature (37).
X-ray Data Collection, Structure Determination, and Refinement-Data were collected at Iowa State University from single crystals on a Rigaku R-AXIS IVϩϩ rotating anode/image plate system using CuK ␣ radiation from an Osmic confocal optics system and a temperature of 110 K. The program d*trek (38) was used in reducing data.
Cav Ϫ FBPase with Fru-2,6-P 2 and AMP are isomorphous to R-and T-state FBPase complexes, respectively. R-state FBPase (PDB code 1CNQ) and T-state FBPase (PDB code 1EYJ) were used as initial models for molecular replacement. The resulting models underwent refinement using CNS (39) with force constants and parameters of stereochemistry from Engh and Huber (40). A cycle of refinement consisted of slow cooling from 1000 to 300 K in steps of 25 K followed by 100 cycles of conjugate gradient minimization and concluded by the refinement of individual thermal parameters. Restraints of 1.5 Å 2 on nearest neighbor and next-to-nearest neighbor main chain atoms, 2.0 Å 2 on nearest neighbor side chain atoms, and 2.5 Å 2 on nextto-nearest neighbor side chain atoms were employed in thermal parameter refinement. Water molecules were added to difference electron density of 2.5 or better until no significant decrease was evident in the R free value. Water molecules in the final models were within acceptable donor-acceptor distances to each other and/or a polar group of the protein. Stereochemistry of the models was analyzed by use of MolProbity (41).
An overview of FBPase showing the relationship of ligand binding sites and loop 50 -72 to the central cavity appears in Fig. 1. The number of water molecules in the central cavities of pFBPases, varying in quaternary state and loop conformation (PDB codes 1CNQ, 1NUY, 1YYZ, 2F3D, 2QVV, 2QVU, 1EYJ, and 1FRP), averages to 31, and that of a similar ensemble of eFBPases (PDB codes 2OX3, 2QVR, 2GQ1, and 2Q8M) averages to 14 (supplemental Fig. S1). The difference in water structures of the central cavities of pFBPase and eFBPase is not solely due to the resolution of crystal structures because cavities with the highest numbers of discrete sites for water molecules in eFBPase (2GQ1; 22 water molecules, 1.5 Å resolution) and pFBPase (1YYZ; 48 water molecules, 1.9 Å resolution) follow the trend established by the ensemble averages ( Fig. 1). Order/ disorder transitions of water molecules in the cavity could be a factor in AMP/inhibitor synergism. Only three positions in or near the central cavity differ in residue type between eFBPase and pFBPase. Ser 45 , Thr 46 , and Leu 186 in the porcine enzyme are His 37 , Arg 38 , and Tyr 180 , respectively, in eFBPase ( Fig. 1). Side chains of Ser 45 /His 37 point into the central cavity, whereas side chains of Thr 46 /Arg 38 and Leu 186 /Tyr 180 are at the edge. If the central cavity is essential to synergism, then mutations that replace residues of pFBPase with those of eFBPase should significantly reduce AMP/Fru-2,6-P 2 synergism.
Kinetics of Wild-type and Mutant pFBPases-His 45 , Arg 46 , and Cav Ϫ FBPase have substantial activity relative to wild-type enzyme ( Table 1). The K m for substrate for Tyr 186 pFBPase is 2-fold greater than that for wild-type enzyme. Mutations cause only minor changes in the K a for Mg 2ϩ . AMP is a competitive inhibitor with respect to Mg 2ϩ for all mutants. Hill coefficients of ϳ2 indicate positive cooperativity in AMP binding. The K i of AMP for Arg 46 pFBPase decreases 6-fold, whereas for other mutants, the K i for AMP inhibition is comparable with that of wild-type enzyme. His 45 , Arg 46 , and Cav Ϫ pFBPases exhibit residual activity (10 -20% of V max ) at saturating levels of AMP. This phenomenon has been attributed to active AMP complexes in intermediate quaternary states of the tetramer (24,45). Fru-2,6-P 2 is a potent inhibitor, competitive with respect to Fru-1,6-P 2 for all FBPases here. In general, cavity mutations have relatively minor effects on activity and ligand binding affinities.
Data in Fig. 2 represent the linear increase in fluorescence with time due to the production of NADPH. In previous work (14), each curve was fit by an empirical relationship with adjustable parameters for maximal and minimum velocities and a Hill coefficient and then replotted as relative velocity versus [AMP]. An alternative model-based definition of synergism (Scheme 1) provides an analytical expression valid at all concentrations of Fru-2,6-P 2 , including an infinite concentration. and disengaged (magenta) conformations. C, the central cavity of Leu 54 pFBPase (left, I R -state, AMP complex, 1.9 Å, PDB code 1YYZ) and the central region of eFBPase (right, ammonium sulfate complex, 1.5 Å, PDB code 2GQ1). Comparable regions exhibit significant differences in hydration levels. The surface renderings are of subunits C1 and C4 of the tetramer, with ball-and-stick models representing selected residues from subunits C2 and C3 and water molecules (red spheres). The icon with associated arrow indicates the region viewed and viewing direction, respectively. Image generated with PyMOL (57).

TABLE 1 Parameters from the kinetics of wild-type and mutant forms of pFBPases and wild typelFBPase
Values in parenthesis are standard deviations in the last significant digit.  AMP/Fru-2,6-P 2 Synergism in Fructose-1,6-bisphosphatase MARCH 21, 2014 • VOLUME 289 • NUMBER 12

Specific activity
In Scheme I, substrate S binds to FBPase subunits independently (48), resulting in an ES complex that contributes to activity (k cat ). Two molecules of AMP (represented by A) bind with a Hill coefficient of 2 and affect an R-to T-state transition (20). The ESA 2 complexes for some mutant pFBPases retain residual activity (k cat Ј is nonzero). Fru-2,6-P 2 (represented by B) and substrate are mutually exclusive in binding to the enzyme (5, 6); however, the binding of AMP and substrate are not mutually exclusive (43). AMP and Fru-2,6-P 2 bind to distinct sites (EBA and EBA 2 ). The Hill coefficient for AMP becomes unity in the presence of Fru-2,6-P 2 . Hence, the equilibrium constant (K ibaa ) for the binding of A to EBA equals 4K iba , as required for the independent binding of ligands to two identical sites. where a ϭ (v 0 Ϫ SV max kЈ cat / K S K issa k cat ; b ϭ (v 0.5 )(B/(K ib K iba )); c ϭ (v 0.5 )(1 ϩ S/K S ϩ B/K ib ) Ϫ SV max /K S ; and v 0.5 is one-half of the sum of the maximum (at A ϭ 0) and minimum (at A ϭ ∞) velocities for a given concentration of B. The quadratic formula provides A 0.5 at any value of B including B ϭ ∞ ( Table 2). Synergism is defined here  (Table 2).
AMP Complex of Cav Ϫ pFBPase (Protein Data Bank Code 4GWS)-Crystals of the AMP complex of Cav Ϫ pFBPase (space group P2 1 2 1 2, a ϭ 60.24 Å, b ϭ 164.29 Å, and c ϭ 79.14 Å) belong to the same space group and lattice as crystals of canonical T-state pFBPase. A C1-C2 dimer exists in the asymmetric unit. Both monomers in the model start from residue 9. Electron density for residues 60 -72 in the dynamic loop is weak and is associated with high B-parameters. Statistics for data collection and refinement are shown in Table 3.
The AMP complex of Cav Ϫ pFBPase is in the T-state, with a 15°subunit pair rotation and a disengaged loop 50 -72. One molecule each of Fru-6-P, Mg 2ϩ , P i , and AMP binds to each subunit. The root mean square deviation in C␣ carbons between AMP complexes of Cav Ϫ and wild-type pFBPase (both in the T-state) is 0.3 Å. Further details regarding the structure of the AMP complex of pFBPase can be found in the literature (16).
Fru-2,6-P 2 Complex of Cav Ϫ pFBPase (PDB Code 4GWU)-Crystals of the Fru-2,6-P 2 complex of Cav Ϫ pFBPase (space group I222, a ϭ 54.00 Å, b ϭ 81.22 Å, and c ϭ 165.06 Å) belong to the same space group and lattice as crystals of wild-type pFBPase in its R-state (PDB code 1CNQ). The asymmetric unit is a monomer. Electron density for residues 1-9 and 52-70 is absent. Electron density for residues 22-27 at the AMP binding site is weak, consistent with the previously reported R-state structure (49). Side chains of His 45 and Tyr 186 are clear in electron density maps, whereas the side chain of Arg 46 is partially disordered (Fig. 3). Electron density confirms the presence of bound Fru-2,6-P 2 . Mg 2ϩ at metal site-1 is in contact with the 1-hydroxyl group and 2-phosphoryl group of Fru-2,6-P 2 .
Superposition of the entire tetramer of the Fru-2,6-P 2 complex of Cav Ϫ pFBPase onto the wild-type R-state, wild-type Fru-2,6-P 2 complex, and wild-type T-state yields root mean square deviations in C ␣ coordinates of 0.75, 2.3, and 2.5 Å, respectively. The Fru-2,6-P 2 complex of Cav Ϫ pFBPase is near the canonical R-state, exhibiting a 2°subunit pair rotation as opposed to the 13°rotation of the Fru-2,6-P 2 complex of wildtype pFBPase (14) (Fig. 3). The 2°rotation in Fru-2,6-P 2 complex of Cav Ϫ pFBPase is similar to the 3°rotation of the I R -state induced by AMP in Leu 54 pFBPase (23). The shearing of helices H1 and H2 observed in the AMP complex of Leu 54 pFBPase is evident in the Fru-2,6-P 2 complex of Cav Ϫ pFBPase, as is the loss of the hydrogen bond between Asn 35 and Thr 14 and the movement of Ile 10 from its hydrophobic pocket (Fig. 3); however, unlike the AMP complex of Leu 54 FBPase, disruption of the Glu 192 -Thr 39 hydrogen bond across the C1-C4 subunit interface does not occur.
Fru-2,6-P 2 induces local conformational change in Cav Ϫ pFBPase similar to that induced in wild-type pFBPase (Fig. 3) (14). Loop 121-126 moves toward Fru-2,6-P 2 probably in response to hydrogen bonds formed between backbone amide groups of Ser 123 and Ser 124 and the 2-phosphoryl group of Fru-2,6-P 2 . ␣-Carbons of Glu 97 and Glu 98 move toward the Mg 2ϩ at site-1, most likely a consequence of altered metal coordination due to the absence of metal at site-2. Movements in residues 97-98 and 121-126 may block Asp 74 from forming a functionally essential hydrogen bond with the C-terminal end of loop 50 -72. Loop 50 -72 itself is disordered, and metal cations are absent from sites 2 and 3. The conformation of loop 264 -274 in the Fru-2,6-P 2 complex of Cav Ϫ pFBPase is nearly identical to that of the product complex of wild-type pFBPase (R-state), having not undergone the conformational change observed in the R-to T-state transition.
where i runs over multiple observations of the same intensity, and j runs over all crystallographically unique intensities. b R factor ϭ ⌺ ʈF obs ͉ Ϫ ͉F calc ʈ/⌺͉F obs ͉, where ͉F obs ͉ Ͼ 0. c R free based upon 10% of the data randomly culled and not used in the refinement. MARCH 21, 2014 • VOLUME 289 • NUMBER 12

JOURNAL OF BIOLOGICAL CHEMISTRY 8455
Sequence Analysis of Type I FBPases-The alignment and hierarchical clustering of 307 Type I FBPase sequences from the Uniprot database (50) employed the ClustalW Web server on the GenomeNet database. Comparisons of sequence positions occupied by signature residues responsible for AMP inhibition, tetramer stability, Glc-6-P inhibition, phosphoenolpyruvate activation, and cavity formation segregate FBPases into groups with mutually exclusive regulatory mechanisms. Position 45, for instance, is a small side-chain residue in eukaryotic FBPases but large and hydrophilic in E. coli-like FBPases (Fig. 4). Residue 112 is critical for AMP inhibition (42) and is lysine in the vast majority of eukaryotic and E. coli-like FBPases. The salt link between Lys 42 and Glu 192 is essential to tetramer stability, and position 192 is glutamate in eukaryotic FBPase but a noncharged residue type in E. coli-like FBPases (30,31). Positive charges at positions 228 and 38 are important for Glu-6-P inhibition and anion activation in eFBPase, respectively, both of which are absent in eukaryotic FBPases (30 -32).
The association of residue type with regulatory mechanism facilitates the prediction of regulatory properties of FBPases that have yet to be studied experimentally. Sequences of eukary-otic FBPases fall into blocks of relatively high similarity corresponding to fungi, mammalian muscle, mammalian liver, and plants (Fig. 5). The eukaryotic FBPases should be stable tetramers subject to synergistic inhibition by AMP and Fru-2,6-P 2 (signature residues are present for the AMP site, tetramer stability, and central cavity). A block with 64 sequences in the middle of the similarity matrix includes eFBPase. E. coli-like FBPases have residue types associated with phosphoenolpyruvate activation, a dynamic equilibrium between dimer and tetramer forms, AMP inhibition, no central cavity, and, hence, no AMP/Fru-2,6-P 2 synergism. More than half of all similarity blocks have scant or no experimental data. FBPases in the large block at the top right of the similarity matrix, for instance, should lack all known mechanisms of FBPase inhibition (AMP and/or Glc-6-P inhibition or reversible disulfide bond formation) and yet be stable tetramers.
Kinetics of lFBPase-FBPases from Leptospira are between eukaryotic and E. coli-like FBPases in the similarity matrix but resemble eukaryotic FBPases, having signature residues for AMP inhibition, tetramer formation, and AMP/Fru-2,6-P 2 synergism. A bacterial FBPase with properties of a eukaryotic FIGURE 3. Conformational changes of Cav ؊ pFBPase induced by Fru-2,6-P 2 . A, subunit pair rotations induced by Fru-2,6-P 2 in wild-type (blue, 13°, PDB code 2QVV) and Cav Ϫ pFBPase (magenta, 2°) relative to R-state pFBPase (green, PDB code 1CNQ). C1-C2 subunit pairs are aligned, and residues 10 -100 in C3-C4 subunit pairs are drawn as C␣ traces. B, central cavity in Fru-2,6-P 2 complexes of Cav Ϫ pFBPase, showing electron density (contour level, 1) for mutated residues His 45 , Arg 46 , and Tyr 186 . The image was generated with PyMOL (57). C, conformational differences at the active site for the Fru-2,6-P 2 complex of Cav Ϫ pPBPase (magenta), the product complex of wild-type pFBPase (green; PDB code 1CNQ), and the Fru-2,6-P 2 complex of wild-type pFBPase (blue). Individual subunits are aligned to eliminate differences due to subunit pair rotations. Electron density (contour level, 1) is provided for Fru-2,6-P 2 and Mg 2ϩ . D, conformational differences at the AMP site for the Fru-2,6-P 2 complex of Cav Ϫ pFBPase (magenta), the AMP complex of Leu 54 pFBPase (cyan; PDB code 1YYZ), and the R-state product complex of wild-type pFBPase (green; PDB code 1CNQ). Dotted lines indicate hydrogen bonds. The relative positions of helices H1 and H2 are similar in the Fru-2,6-P 2 complex of Cav Ϫ pFBPase and AMP complex of Leu 54 pFBPase, which have subunit pair rotations of 2 and 3°, respectively, relative to the R-state product complex of wild-type pFBPase. Icons represent the region viewed and viewing direction.
FBPase would suggest a pathway for the evolution of regulatory mechanisms, but the predicted properties of lFBPase first need confirmation by direct measurement.
Potassium activates pFBPase at millimolar concentrations (1, 2, 51) but inhibits lFBPase. Hence, assays here for lFBPase omit KCl. lFBPase catalyzes the hydrolysis of Fru-1,6-P 2 with specific activity and K m values comparable with those of pFBPase. Mg 2ϩ is an essential cofactor, but unlike pFBPase, Mg 2ϩ activation is not cooperative in lFBPase catalysis ( Table 1). Loss of Mg 2ϩ cooperativity may be due to the change in residue type (aspartate to glutamate) at position 68 (pFBPase numerology). The mutation Asp 68 3 Glu in pFBPase reduces Mg 2ϩ cooperativity (46), and eFBPase has glutamate at position 68 with no Mg 2ϩ cooperativity (52). All other Mg 2ϩ -binding residues are identical among the three wild-type FBPases under consideration here.
AMP inhibition of lFBPase is comparable with that of pFBPase. The Hill coefficient for AMP inhibition is close to 2. AMP is a competitive inhibitor with respect to Mg 2ϩ (K i ϭ 0.58 Ϯ 0.05 M) and a non-competitive inhibitor with respect to Fru-1,6-P 2 (K i ϭ K is ϭ 0.83 Ϯ 0.05 M).
Anion activators, such as PEP or citrate, activate eFBPase by promoting tetramer formation (31). To test the effect of PEP on lFBPase, Mg 2ϩ -initialized assays were used (31). PEP activation is not present for lFBPase. Instead, millimolar levels of PEP cause moderate inhibition. Assays of Glc-6-P inhibition followed the published protocol for eFBPase (32). Glc-6-P inhibition becomes evident only at millimolar levels, ϳ50-fold weaker than Glc-6-P inhibition of eFBPase. Glc-6-P inhibits pFBPase and lFBPase at comparable concentrations.
Sedimentation Equilibrium-lFBPase and pFBPase at three protein concentrations (A 280 ϭ 0.3, 0.5, and 0.7) were sedimented to equilibrium at angular speeds of 15,000, 18,000 and 21,000 rpm. Sedimentation equilibrium data were fit to single component and dimer-tetramer equilibrium models with UltraScan (53). Regardless of model, pFBPase and lFBPase give similar apparent molecular masses, 111.9 and 112.1 kDa, respectively. The determined molecular masses are less than the calculated mass of the tetramer (140 kDa).
Tryptophan Fluorescence-lFBPase has but one tryptophan (Trp 219 ), which on the basis of structural similarity is near the common binding sites for the 6-phosphoryl groups of substrate, product, and Fru-2,6-P 2 . Steady-state fluorescence emission from Trp 219 increases as levels of substrate, product, or Fru-2,6-P 2 increase, whereas increasing concentrations of AMP cause little change. Substrate, product, Fru-2,6-P 2 , and AMP have no effect on fluorescence emission from 100 M tryptophan in the absence of lFBPase. The fluorescence change caused by Fru-6-P, Fru-1,6-P 2 , or Fru-2,6-P 2 is arguably due to a change in the local environment of Trp 219 . Indeed, at saturating concentrations, the three active site ligands induce nearly identical limiting changes in fluorescence emission.
Data from titrations of 6-phosphoryl ligands in the presence and absence of AMP are fit to Equation 3, where ⌬F is the change in fluorescence upon the addition of ligand L, F o is the fluorescence in the absence of ligand, and K d is the dissociation constant. AMP does not affect the binding of Fru-1,6-P 2 ( Fig. 6 and Table 4); however, AMP at a concentra- tion that causes 50% inhibition (1.8 M) enhances the binding of Fru-2,6-P 2 by ϳ2-fold, consistent with data from kinetics. AMP enhancement of Fru-6-P binding to lFBPase is also 2-fold, suggesting that the reverse effect (Fru-6-P enhancement of AMP binding) is comparable with Fru-2,6-P 2 enhancement (15-fold; Table 2). Fru-6-P enhancement of AMP binding cannot be measured because of assay design (coupling of Fru-6-P production to NADPH formation).

DISCUSSION
Ser 45 3 His and Cav Ϫ pFBPases represent first instances of modified pFBPases that retain potent AMP and Fru-2,6-P 2 inhibition with greatly reduced AMP/Fru-2,6-P 2 synergism. Kinetic parameters for catalysis, Mg 2ϩ activation, AMP inhibition, and Fru-2,6-P 2 inhibition of mutant enzymes investigated here are similar to those of wild-type pFBPase. Reduced Fru-2,6-P 2 /AMP synergism in Cav Ϫ pFBPase is not due to the disruption of the active or allosteric binding sites but rather to an impaired quaternary response; Cav Ϫ pFBPase exhibits only a 2°s ubunit pair rotation in contrast to the 13°subunit pair rotation for the wild-type enzyme. Conceivably, Cav Ϫ mutations could stabilize the R-state (or destabilize the T-state), making the quaternary transition unfavorable; however, previously reported mutations that destabilize quaternary states (Ala 54 3 Leu and Ile 10 3 Asp, for instance) broadly impact the kinetics of pFBPase. Moreover, impairment of the R-to T-state transition is specific to Fru-2,6-P 2 . AMP drives Cav Ϫ pFBPase to a canonical T-state, which is structurally indistinguishable from the T-state of the wild-type enzyme.
The effect of Fru-2,6-P 2 must be more than the displacement of loop 50 -72 from its engaged conformation. Loop 50 -72 is disordered in the Fru-2,6-P 2 complex of Cav Ϫ pFBPase, but the enzyme remains in an R-like state. In contrast, the loop 50 -72 acquires its disengaged conformation in the Fru-2,6-P 2 complex of wild-type pFBPase, and the enzyme is in a near-T quaternary state. Structural evidence here indicates the subunit pair rotation must go beyond 2°to allow the disengaged conformer.
The solvated central cavity, absent in Cav Ϫ pFBPase, evidently enables the requisite subunit pair rotation in response to the binding of Fru-2,6-P 2 . The binding of AMP to Leu 54 pFBPase causes a 3°subunit pair rotation, retaining the engaged conformation of loop 50 -72 (23) and an organized water structure in the central cavity (Fig. 1). In contrast, in the T-state of Leu 54 pFBPase, loop 50 -72 and the water structure in the cavity are disordered. Arguably, an order-to-disorder transition of water molecules within the central cavity, caused by the movement of loop 50 -72 from its engaged conformation, provides the requisite motive force for subunit pair rotation. In Cav Ϫ pFBPase, the order-to-disorder transition of water structure is inoperative, and consequently movement of loop 50 -72 from its engaged conformation results in virtually no subunit pair rotation. The mutation of Ser 45 3 His alone is largely responsible for this effect, suggesting a critical role for Ser 45 in maintaining an organized water structure when loop 50 -72 is engaged. The volume of the histidyl side chain (137 Å 3 ) relative to that of serine (89 Å 3 ) is probably a factor; however, the side chain of serine in this specific context may offer the capacity, not shared by other amino acid types, to "seed" a network of hydrogen-bonded water molecules.
Fru-2,6-P 2 /AMP synergism, although reduced, is still present in His 45 and Cav Ϫ pFBPase. The 2°subunit pair rotation caused by Fru-2,6-P 2 in Cav Ϫ pFBPase is similar to changes induced by AMP in Leu 54 pFBPase (I R -state, 3°subunit pair rotation). Structural changes at the C1-C4 interface due to small subunit pair rotations could be the structural basis for the weak synergism that remains in Cav Ϫ pFBPase. Indeed, wildtype and mutant pFBPases here exhibit reduced AMP cooperativity in the presence of Fru-2,6-P 2 , which may come from conformational changes at the C1-C4 interface (such as the weakening of the hydrogen bond between Glu 192 and Thr 39 ) within the first few degrees of subunit pair rotation. The model above suggests that an engaged dynamic loop in the presence of bound Fru-2,6-P 2 would eliminate Fru-2,6-P 2 /AMP synergism altogether, and in fact, such is the case for eFBPase; there is no Fru-2,6-P 2 /AMP synergism, and the Fru-2,6-P 2 complex of wild-type eFBPase reveals a dynamic loop in its engaged conformation (14).
The failure of the loop 50 -72 to achieve a disengaged conformation in its Fru-2,6-P 2 complex of Cav Ϫ pFBPase is also consistent with the results of targeted molecular dynamics simulations, which indicate that the transition of loop 50 -72 from an engaged to disengaged conformation comes late in the R-to T-state transition (24). Moreover, previous work ties the conformational change in loop 264 -274 to subunit pair rotation from the I R -to I T -state (from 3 to 12°, respectively) (23,24). The Fru-2,6-P 2 complexes of Cav Ϫ and wild-type pFBPases reveal a conformational change in loop 264 -274 only for the latter complex.
The foregoing analysis supports the following hypothesis. Any tetrameric FBPase with signature residues of a central cavity should exhibit AMP/Fru-2,6-P 2 synergism. Hence, all eukaryotic organisms for which sequences of FBPases are known should exhibit AMP/Fru-2,6-P 2 synergism.
Fru-2,6-P 2 is not present in bacteria (54), so the evolution of AMP/Fru-2,6-P 2 synergism from a bacterial ancestor is problematic. E. coli FBPase employs Glc-6-P as a synergistic inhibitor with respect to AMP (32). Unlike Fru-2,6-P 2 , Glc-6-P is an allosteric non-competitive inhibitor with respect to substrate. Vastly different kinetic properties and a distinct allosteric binding site for Glc-6-P in eFBPase make Glc-6-P an unlikely precursor of Fru-2,6-P 2 through evolution. The analysis of Type I FBPase sequences, however, indicates a central cavity in lFBPase. Moreover, AMP/Fru-6-P binding synergism in lFBPase suggests an evolutionary pathway from AMP/Fru-6-P synergism in bacteria to AMP/Fru-2,6-P 2 synergism in eukaryotic organisms.