Novel Allosteric Activation Site in Escherichia coli Fructose-1,6-bisphosphatase*

Fructose-1,6-bisphosphatase (FBPase) governs a key step in gluconeogenesis, the conversion of fructose 1,6-bisphosphate into fructose 6-phosphate. In mammals, the enzyme is subject to metabolic regulation, but regulatory mechanisms of bacterial FBPases are not well understood. Presented here is the crystal structure (resolution, 1.45Å) of recombinant FBPase from Escherichia coli, the first structure of a prokaryotic Type I FBPase. The E. coli enzyme is a homotetramer, but in a quaternary state between the canonical R- and T-states of porcine FBPase. Phe15 and residues at the C-terminal side of the first α-helix (helix H1) occupy the AMP binding pocket. Residues at the N-terminal side of helix H1 hydrogen bond with sulfate ions buried at a subunit interface, which in porcine FBPase undergoes significant conformational change in response to allosteric effectors. Phosphoenolpyruvate and sulfate activate E. coli FBPase by at least 300%. Key residues that bind sulfate anions are conserved among many heterotrophic bacteria, but are absent in FBPases of organisms that employ fructose 2,6-bisphosphate as a regulator. These observations suggest a new mechanism of regulation in the FBPase enzyme family: anionic ligands, most likely phosphoenolpyruvate, bind to allosteric activator sites, which in turn stabilize a tetramer and a polypeptide fold that obstructs AMP binding.

Donahue et al. (18) was the first to classify prokaryotic FBPases by sequence homology. Investigators now suggest five FBPase types (19). Types I-III are primarily in Bacteria, Type IV in Archaea, and Type V in thermophilic prokaryotes from both domains. Eukaryotic FBPases are homologous to prokaryotic Type I enzymes. Crystal structures are available for Types IV and V FBPases and for mammalian and chloroplast FBPases, but no structure of an FBPase from Bacteria is available.
FBPase is a target for potential drugs in the treatment of Type II diabetes (20 -22). Novel inhibitors have been developed (20,21), and in one case, knowledge of structure and mechanism contributed to the rational design of a new inhibitor (22). An important property of mammalian FBPase, synergism in Fru-2,6-P 2 /AMP inhibition, however, is not understood in terms of a specific molecular mechanism. Moreover, no example is known of a mutant or wild-type mammalian FBPase that lacks synergism without significant disruption of AMP or Fru-2,6-P 2 inhibition. Fru-2,6-P 2 and AMP individually are potent inhibitors of E. coli FBPase, but exhibit no synergism (15). The sequence identity (41%) of the E. coli and porcine enzymes suggest similar folds. Other studies indicate the existence of an allosteric AMP binding site and a tetrameric subunit organization for the E. coli enzyme (14,16). Hence, differences in the conformational responses of E. coli and porcine FBPases to AMP and/or Fru-2,6-P 2 could reveal the molecular basis of AMP/Fru-2,6-P 2 synergism.
Reported here is the high resolution structure of Type I FBPase from E. coli. The structure reveals a homotetramer between the canonical Rand T-states of porcine FBPase (11). Sulfate anions appear at the interface between top and bottom subunit pairs of the tetramer. Sulfate and phosphoenolpyruvate (PEP) enhance activity of the E. coli enzyme by at least 300%. Anionic ligands likely mimic a physiological effector that stabilizes E. coli FBPase as an active tetramer, a hypothesis consistent with previous reports of dilution-linked inactivation (4,14). The true physiological activator is probably PEP, an inhibitor of E. coli fructose-6-phosphate 1-kinase (23), and an activator of FBPases in other prokaryotic organisms (18, 24 -26). Sequence alignments reveal a number of organisms with Type I FBPases that probably possess the anion binding site of the E. coli enzyme. The anion binding site, however, is not present in organisms that employ Fru-2,6-P 2 in the regulation of FBPase activity.

EXPERIMENTAL PROCEDURES
Materials-Fru-1,6-P 2 , NADP ϩ , PEP, leupeptin, ampicillin, phenylmethylsulfonyl fluoride (PMSF), DEAE-Sepharose and phenyl-Sepharose came from Sigma and kanamycin sulfate from Invitrogen. VENT DNA-polymerase and all restriction enzymes were purchased from New England BioLabs. Glucose-6-phosphate dehydrogenase and phosphoglucose isomerase were from Roche Applied Sciences. The Iowa State University DNA Sequencing and Synthesis Facility produced deoxyoligonucleotides and DNA sequences. All other chemicals were of reagent grade or equivalent.
Cloning E. coli FBPase-The fbp gene that encodes Type I FBPase in E. coli (27) was cloned from genomic DNA isolated from strain XL1-Blue (Stratagene) by primer extension using polymerase chain reactions (PCR). Forward (5Ј-GTATCGACCATATGAAAACGT-TAGGTGAATTTATTG-3Ј) and reverse (5Ј-CAGAATTCT-TACGCGTCCGGGAACTCACGG-3Ј) primers introduced NdeI and EcoRI sites, respectively (sites underlined). The initial round of PCR employed VENT DNA polymerase. The amplified product was the template for a second PCR, using the same primers and the BIO-X-ACT Short DNA polymerase (Midwest Scientific), which leaves a 3Ј-A overhang. The resulting oligonucleotide was ligated into the pGEM-T Easy Vector (Promega) and the resulting plasmid used to transform electrocompetent XL1-Blue cells. Transformants were selected by growth on LB-agarose supplemented with 50 g/ml ampicillin, whereas recombinants were selected by blue/white screening. Plasmid DNA was isolated from cultures grown from single colonies. The presence of the insert was verified by digestion with EcoRI and gel electrophoresis.
The internal NdeI site within the fbp gene was eliminated by PCR using the mutagenic primer 5Ј-CTTTGTCGGCAACGACCACATG-GTTGAAATGTCGAACGC-3Ј and its reverse complement to introduce a silent mutation (base change underlined). Parental methylated DNA was digested with DpnI, and DH5␣ cells were transformed with the resulting nicked plasmid. Plasmids from single colonies were isolated, digested with NdeI, and separated by gel electrophoresis, using non-mutated plasmid as a control. The desired mutation was confirmed by DNA sequencing. The resulting construct was amplified by PCR, and then digested with NdeI and EcoRI. Simultaneously the bacterial protein expression vector pET-24b (Novagen), carrying the kan r selectable marker, was digested with NdeI and EcoRI and the cut plasmid isolated by gel electrophoresis. The excised oligonucleotide and vector DNA were purified, ligated, and transformed as previously described, using the cut and uncut vectors as negative and positive controls, respectively. The final plasmid (pECFBP) was verified by DNA sequencing.
Expression and Purification of Native and Selenomethionine-substituted E. coli FBPase-Separate preparations of native and selenomethionine-substituted enzyme were used for the kinetic and structural investigations, respectively. FBPase-deficient E. coli strain DF657 (tonA22, ompF627(T 2 R), relA1, pit10, spoT1, ⌬(fbp)287; Genetic Stock Center at Yale University, New Haven, CT) was transformed by the pECFBP plasmid and plated onto LB-agarose supplemented with 30 g/ml kanamycin sulfate. A single colony inoculant (100 ml LB-kanamycin) grew (with shaking, 37°C) for 12 h. Two percent inoculum was added to 4 liters of LB-kanamycin. The culture grew (with shaking, 37°C), to an A 600 of 1.0, at which time transcription was induced by the addition of isopropyl-␤-D-thiogalactopyranoside (IPTG, final concentration, 0.85 mM). The culture was maintained (with shaking, 37°C) for an additional 12 h. Selenomethionine-substituted protein was produced as described previously with all growth steps at 37°C. 3 Modifications to the protocol of Kelley-Loughnane et al. (16) led to purified protein in a single day. All buffers employed after cell lysis were degassed (stirring under vacuum) and contained 0.2 mM EDTA and 5 mM dithiothreitol (DTT). Cells were collected by centrifugation (3,000 ϫ g) and resuspended in 75 ml of lysis buffer (40 mM KH 2 PO 4 / K 2 HPO 4 , pH 7.0, 0.1 mM PMSF, 1 mM EDTA, 5 g/ml leupeptin, and 1 mM ␤-mercaptoethanol) for disruption by French press. Insoluble debris was removed by centrifugation (30 min, 33,000 ϫ g) prior to ammonium sulfate fractionation. Precipitated protein was collected by where i runs over multiple observations of the same intensity and j runs over all the crystallographically unique intensities. c R factor ϭ ⌺͉͉F obs ͉ Ϫ ͉F calc ͉͉/⌺͉F obs ͉, where ͉F obs ͉ Ͼ 0. d R free based upon 10% of the data randomly culled and not used in the refinement.
centrifugation from 40 -70% levels of saturated (NH 4 ) 2 SO 4 , and redissolved in 40 mM NaH 2 PO 4 , pH 7.2, and (NH 4 ) 2 SO 4 (40% of saturation). Protein was loaded onto a phenyl-Sepharose column, washed with the same buffer, and eluted by a step gradient in (NH 4 ) 2 SO 4 (20% followed by 14% of saturation). Combined fractions of highest specific activity were reduced in volume to 30 ml using an Amicon concentrator (YM 30,000 membrane). Protein was passed through a desalting column (Sephadex G-50 equilibrated with 40 mM NaH 2 PO 4 , pH 7.2), loaded onto a DEAE-Sepharose column, (equilibrated with 0.1 M NaCl, 40 mM NaH 2 PO 4 , pH 7.2) and then eluted with a linear gradient in NaCl (0.1-0.5 M). Pooled fractions of highest specific activity were desalted (Sephadex G-50 column, equilibrated with 5 mM Tris malonate, pH 7.4). Enzyme (0.75 mg/ml) used for kinetic studies was frozen in this buffer, whereas enzyme used for crystallization experiments was concentrated to 15 mg/ml and filtered through a 0.22 M filter. Protein was 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 (29), using bovine serum albumin as a standard. Protein purity was monitored by sodium dodecylsulfatepolyacryamide gel electrophoresis (30). N-terminal sequencing and mass determinations of purified proteins were performed at the Iowa State University Protein Facility, using an Applied Biosystems Voyager System 6075 matrix-assisted laser desorption/ionization timeof-flight mass spectrometer (MALDI-TOF MS).
Crystallization of E. coli FBPase-Crystals were grown by hanging drop in vapor diffusion VDX-plates (Hampton Research). Droplets contained 2 l of a protein solution (15 mg/ml of the selenomethioninesubstituted FBPase, 20 mM DTT, 0.2 mM EDTA, and 5 mM each of Fru-1,6-P 2 , Fru-2,6-P 2 , and MgCl 2 ) and 2 l of a precipitant solution (2.0 M (NH 4 ) 2 SO 4 ) and were equilibrated over 500 l of precipitant solution. Equal dimensional crystals (0.2-0.4 mm) grew within 3 days at 22°C. Crystals were immersed for 20 s in the precipitant solution, supplemented with DTT, ligands, and 27% (v/v) glycerol, before freezing directly in a cold stream of nitrogen.
Data Collection-Crystals were screened for diffraction 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. High resolution data were collected at 100 K on Beamline 4.2.2 of the Advanced Light Source, Lawrence Berkeley Laboratory. The program d * trek (31) was used to index, integrate, scale, and merge intensities, which were then converted to structure factors using the CCP4 (32) program TRUNCATE (33).
Structure Determination and Refinement-A molecular replacement solution was obtained using the program AMORE (34) and the model for R-state porcine FBPase (PDB accession identifier 1CNQ), from which residues 52-72 and all ligand and water molecules had been removed. Residues for the E. coli enzyme were manually fit to omit density, using the program XTALVIEW (35). The resulting model using CNS (36) underwent simulated annealing from 3000 to 300 K in steps of 25 K, followed by 100 cycles of energy minimization and thermal parameter refinement. Force constants and parameters of stereochemistry were from Engh and Huber (37). 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. Ligands and water molecules were fit to omit electron density until no improvement in R free was evident. Water molecules with thermal parameters above 50 Å 2 or more than 3.2 Å from the nearest hydrogen bonding partner were removed from the final model.
Comparison of Porcine and E. coli FBPases-Construction of dimer and tetramer models from the monomer, as well as pair-wise least squares superpositions of E. coli FBPase and porcine liver FBPase were accomplished using the CCP4 programs PDBSET (38) and LSQKAB (39). The canonical R-and T-states of porcine FBPase used in superpositions have PDB identifiers 1CNQ and 1EYK, respectively, whereas I Tand I R -states have identifiers 1Q9D and 1YYZ, respectively. The angle of rotation of subunit pairs in various quaternary states of FBPases is sensitive to the subset of residues used in the least-squares fit (40). By convention, C ␣ atoms of porcine FBPase residues 33-49, 75-265, 272-FIGURE 2. Stereoview of the anion activation site. Electron density is from an omit map (contour level 1, cutoff radius 1.5 Å) covering sulfate 344 and residues in its vicinity. The arrow in Fig. 1 indicates the viewing direction for this figure. Parts of this drawing were prepared with XTALVIEW (35).  Kinetic Experiments-Assays for FBPase activity employed the coupling enzymes phosphoglucose isomerase and glucose-6-phosphate dehydrogenase and monitored formation of NADPH by either absorbance at 340 nm or fluorescence emission at 470 nm (40). Assays (total volume, 2 ml) were performed at 22°C in 50 mM Hepes, pH 7.5, with 0.1 mM EDTA and 150 M NADP ϩ . Saturating levels of Fru-1,6-P 2 (40 M) and MgCl 2 (10 mM) were used in assays to measure specific activity and ligand activation. Reactions were initiated by the addition of 1.4 g of enzyme (specific activity determinations) or by incubating (1 h) the enzyme in assay mixtures without MgCl 2 , and subsequently initiating the reaction by the addition of metal. For activation assays, the coupling enzymes were dialyzed against a 1000-fold volume of 50 mM Hepes, pH 7.5, to remove ammonium sulfate that would otherwise be present at ϳ3 mM in all assays.
Sequence Alignments-Multiple sequence alignments employed the program ClustalW (41). Pair-wise alignments, which served as the basis for calculating sequence identity, employed the program ALIGN (42).

RESULTS
Enzyme Purity-Native and selenomethionine-substituted preparations of E. coli FBPase represented more than 30% of the soluble protein in the cell lysate. Preparations of E. coli FBPase, isolated by following a multiday protocol, were proteolyzed near the N terminus (confirmed by Edman degradation) and had relatively low specific activities. Kinetic parameters for porcine FBPase are sensitive to truncations near the N terminus (43), and in order to avoid any modification of the N-terminal residues of E. coli FBPase, we developed a streamlined protocol that provided enzyme free of proteolysis (confirmed by N-terminal sequencing and by electron density in the crystal structure). The first residue in E. coli FBPase is an unblocked methionine (first residue of the translated DNA sequence). Retention of the N-terminal methionine is consistent with lysine at position two of the sequence; long side-chains at sequence position two generally prevent E. coli methionyl aminopeptidase from excising the N-terminal methionine (44). Final enzyme preparations are at least 95% pure as evidenced by sodium dodecylsulfate-polyacrylamide gel electrophoresis and have specific activities of 40 units/mg. E. coli FBPase has six methionine residues per monomer. Mass analyses of native and selenomethionine-substituted proteins indicate a mass increase of 261 Ϯ 30 Da, corresponding to more than a 90% replacement of methionine with selenomethionine. Non-denaturing gel electrophoresis revealed a ladder of protein bands indicative of high molecular weight aggregates. Others have observed aggregation of E. coli FBPase (14,16), a condition we sought to avoid in crystallizing the enzyme. The inclusion of 20 mM DTT in protein solutions prevented ladder formation and improved crystal growth.  Escherichia coli Fructose-1,6-bisphosphatase JULY 7, 2006 • VOLUME 281 • NUMBER 27

JOURNAL OF BIOLOGICAL CHEMISTRY 18389
Kinetic Parameters-Native and selenomethionine-substituted enzymes were kinetically indistinguishable: k cat , K m for Fru-1,6-P 2 and K a for Mg 2ϩ are 24 s Ϫ1 , 1.7 Ϯ 0.1 M and 1.0 Ϯ 0.1 mM, respectively. The Hill coefficients for both Fru-1,6-P 2 and Mg 2ϩ are 1.1 Ϯ 0.1. These values are similar to those reported previously for E. coli FBPase (4,14,16), and differ from the porcine enzyme principally in the Hill coefficient for Mg 2ϩ , which for porcine FBPase is 1.9 Ϯ 0.1 (45).
Structure of E. coli FBPase (PDB identifier 2GQ1)-Type I FBPase from E. coli has the same overall fold and tetrameric subunit organization as other Type I FBPases for which structural information is available ( Fig. 1) (11,46,47). Crystals belong to the space group I222 (a ϭ 45.6, b ϭ 81.3, c ϭ 170.1 Å) with one subunit of the tetramer in the asymmetric unit. Data also were collected from a native protein crystal in-house to a resolution of 2.2 Å. The native and selenomethioninesubstituted structures exhibited no differences beyond coordinate uncertainty. Hence, we report here only the results of the high resolution, selenomethionine-substituted structure (Table 1).
Each subunit has 7.5 bound sulfate ions. The "half-sulfate" (sulfate 341 ) is on the crystallographic 2-fold axis that relates subunits C1 and C2 (and subunits C3 and C4) (Fig. 1). Sulfate 343 is at a lattice contact between adjacent tetramers, and sulfate 346 is at a locus that maps onto the AMP binding site of porcine FBPase (48). Sulfate 348 and sulfate 345 are at the 1-phosphoryl-and 6-phosphoryl-binding sites of Fru-1,6-P 2 , respectively, as defined by structures of porcine FBPase (11,49). An additional sulfate ion (sulfate 342 ) binds at a locus occupied by the sidechain of Asp 68 of the porcine R-state structure (49). Sulfate 344 binds at the C1-C4 interface, roughly at the location of the carboxyl group of Glu 192 in porcine FBPase (The corresponding residue is Val 186 in the E. coli enzyme), and 8.5 Å away from a symmetry-related partner (Fig.  2). Sulfate 347 binds on the surface of the subunit at a position of no known significance. No other ligands are present, even though fructosyl phosphates were present in the crystallization experiments.
The deposited model is in good electron density; however, residues 42-63, which correspond to the dynamic loop of the porcine enzyme, are without electron density and have been omitted. The dynamic loop in porcine FBPase takes on either an engaged or disordered conformation in the absence of AMP (48,49). The disordered loop in the E. coli structure may be a consequence of the high concentration of sulfate.  Fig. 1.) The quaternary state of E. coli FBPase is most similar to the I R -state of the porcine enzyme (40); however, no conclusion is possible regarding ligand-induced changes in the quaternary state of the E. coli enzyme.

TABLE 2 Conservation of the anion activation and AMP sites
Type-I FBPase sequences from more than 100 species of eukaryotes and bacteria were aligned using the program ClustalW (41). Listed below are results for a representative set of sequences for five residues, referenced by number to the porcine and E. coli FBPase sequences. Residues similar in type to the signature residues of E. coli FBPase and Glu 192 of porcine FBPase are in bold. Names of species that have FBPases with at least three of four signature residues are in bold. The overall sequence identity to E. coli FBPase was calculated by ALIGN (42). 40  (Fig. 3). Moreover, residue types flanking Phe 15 differ significantly in the E. coli and porcine enzymes. Hence, the local fold of the polypeptide chain sterically blocks AMP. AMP ligation, as observed in mammalian FBPase, would necessitate a large conformational change in the E. coli enzyme that would likely perturb helices H1 and H2. We speculate in the discussion, that such a conformational change is part of a mechanism of allosteric regulation of catalysis. Many residues are conserved between porcine and E. coli FBPases, but a number of critical residues are missing from the C1-C4 interface, most notably residues corresponding to Asn 35 , Thr 39 , and Glu 192 of the porcine enzyme (Fig. 4). Glu 192 in porcine FBPase forms critical hydrogen bonds between subunits C1 and C4 in the R-and T-states, interacting with Thr 39 and Lys 42 . Glu 192 is valine in E. coli FBPase, and Thr 39 a leucine. Sulfate 344 binds in place of the carboxyl group of Glu 192 , filling a positively charged cavity between subunits C1 and C4. Eight residues define this anion-binding site, six of which differ in type from corresponding residues in porcine FBPase. Porcine residues Thr 14 , Asn 35 , Cys 38 , Thr 39 , Ser 88 , and Glu 192 are Gly 5 , Ser 27 , Lys 30 , Leu 31 , Arg 80 , and Val 186 in E. coli FBPase. Only Lys 42 and Thr 12 in the porcine enzyme are conserved as Lys 34 and Thr 3 in E. coli FBPase. Five residues, Thr 3 , Gly 5 , Lys 30 , Lys 34 , and Arg 80 , contribute interactions to sulfate 344 (Fig. 4). In the R-state porcine enzyme, Asn 35 hydrogen bonds across the C1-C4 interface through water molecules and interacts directly with Thr 14 of the same subunit. These interactions are missing in E. coli FBPase; the corresponding residues (Ser 27 and Gly 5 , respectively) are considerably smaller and allow sulfate 344 to interact with backbone amide 5 of the first turn of helix H1. The presence of several water molecules in the cavity suggests a binding site for a larger ligand.

Species
Activation of E. coli FBPase-The specific activity of E. coli FBPase decreases with dilution in the standard assay mixture, whereas that of porcine FBPase is constant (data not shown). The decline in specific activity of the E. coli enzyme in response to dilution occurs in minutes and is not temperature-dependent, but is affected by certain salts. PEP and ammonium sulfate activate E. coli FBPase in assays initiated by the addition of Mg 2ϩ (enzyme is at equilibrium prior to the initiation of assays). The activity of E. coli FBPase increases more than 300% in the presence of 20 mM (NH 4 ) 2 SO 4 (Fig. 5). Activation is caused by the sulfate anion; NH 4 Cl only causes inhibition of the E. coli enzyme (Fig. 5); Na 2 SO 4 and KH 2 PO 4 exhibit similar activation effects (data not shown). Under identical conditions, porcine FBPase is also activated to a lesser extent by (NH 4 ) 2 SO 4 , but the observed activation is caused by NH 4 ϩ (50, 51), and can be reproduced by the addition of NH 4 Cl. Ammonium sulfate concentrations above 20 mM progressively inhibit the E. coli enzyme, presumably because of the binding of sulfate anions to the active site. PEP is a far more potent activator of E. coli FBPase, causing ϳ300% activation at a concentration of 2 mM (Fig. 5). Fifty percent of maximum activation occurs at a concentration of ϳ40 M PEP. Sequence Alignments of Type I FBPases-Sequence alignments of more than 100 Type I bacterial and eukaryotic FBPases appear in an abbreviated format in Table 2. FBPases fall into three groups: (i) sequences that have the signature residues of the anion binding site and blocked AMP site (residues corresponding in type and position to Gly 5 , Phe 15 , Lys 30 , and Arg 80 of E. coli FBPase), (ii) sequences that have a glutamate corresponding to Glu 192 of porcine FBPase, and (iii) sequences that have neither signature residues of an anion binding site nor a residue corresponding to Glu 192 of the porcine enzyme. Notably, no sequence falls into a fourth category, matching both Glu 192 of porcine FBPase and signature residues of the anion binding site. Hence, glutamate at position 192 and the anion binding site may be mutually exclusive. Approximately 65% of bacterial sequences lack glutamate at positions corresponding to 192 of porcine FBPase, and of these approximately one-half have the signature residues of an anion-binding site. Only one eukaryotic organism, the pathogenic protozoa Toxoplasmagondii, has residues corresponding in position and type to the anionbinding site. Curiously, all species known to produce Fru-2,6-P 2 always have FBPases with glutamate at positions corresponding to 192 of porcine FBPase. No evidence supports the existence of Fru-2,6-P 2 in T. gondii.
With the exception of T. gondii, organisms that have FBPases with signature residues of the anion-binding site are heterotrophic bacteria, and like E. coli, are members of the phylum Proteobacteria and class Gammaproteobacteria. Most are pathogens, a bias due perhaps to the disproportionate number of FBPase sequences available for pathogenic organisms. The anion-binding site may be present in all known Type I FBPases from the orders Enterobacteriales, Vibrionales, and Pasteurellales, all of which include important pathogens.

DISCUSSION
The structure here indicates a subunit arrangement for the E. coli enzyme, similar to that of mammalian FBPase tetramers. Enzyme concentrations in crystallization experiments, however, are ϳ20,000-fold higher than those of assays. The E. coli enzyme remains a tetramer at levels 20-fold less than those of crystallization experiments (14,16), but its apparent mass decreases slightly with concentration (14), suggesting the possibility of a dissociated state of the enzyme.
Porcine FBPase undergoes spontaneous subunit exchange (45,52,53). In the presence of Fru-1,6-P 2 or Fru-2,6-P 2 , exchange reactions involve only dimers of the tetramer (45,53). As the active site is shared between subunits C1 and C2, dimer exchange probably entails separation at the C1-C4 interface. Protein-protein hydrogen bonds between subunits C1 and C4 are few in number in the E. coli enzyme, suggesting that in the absence of an anion at the C1-C4 interface the E. coli tetramer may be unstable relative to the mammalian tetramer. At low protein concentrations, and in the absence of specific anionic ligands, the bacterial enzyme may dissociate into less active (or inactive) dimers. Sulfate ions form extensive connections between subunits C1 and C4, and presumably activate the enzyme by stabilizing a tetramer.
If indeed E. coli FBPase requires an anion to assemble into a fully active tetramer, then enzyme activity should be sensitive to concentrations of anions and protein. Large losses in activity have been observed after dialysis steps that remove ammonium sulfate (4), or by dilution of the enzyme (4,14). Most striking is the sensitivity of E. coli FBPase to the method of assay. The enzyme loses as much as 70% of its activity upon a 1000-fold dilution/incubation in the standard assay, a behavior not exhibited by porcine FBPase. This loss of activity is even more severe (in excess of 80%) after dialysis of the (NH 4 ) 2 SO 4 from coupling enzymes used in assays.
PEP relieves AMP inhibition (14,15) and activates FBPases from E. coli (Fig. 5) and other prokaryotic organisms (18, 24 -26). PEP levels in E. coli change more than 10-fold between glycolytic and gluconeo-genic growth (54), hence the suggestion by Babul et al. (14) that PEP is a physiological regulator of E. coli FBPase. If PEP is a physiological regulator of FBPase, then it must bind to the enzyme, but where? Observations of similar phenomena in Type III FBPases have led others to suggest competition between PEP and AMP for the same site (25), but such a mechanism is unlikely for E. coli FBPase: AMP-binding residues of porcine FBPase are present in E. coli FBPase and yet PEP has no effect on AMP inhibition of mammalian systems (2). The binding of PEP to the active site can only cause inhibition. So the anion-binding site at the C1-C4 interface remains as the most probable site for PEP association. A PEP-bound structure of E. coli FBPase will require new crystallization conditions that avoid high salt concentrations, but using the electron density of the sulfate anion to anchor its phosphoryl group, one can easily model the remaining atoms of PEP into the void of the C1-C4 anion binding pocket (Fig. 6).
A binding site for PEP, distinct from AMP, however, creates a new challenge: What mechanism causes the apparent mutual exclusivity in the binding of PEP and AMP? Unlike the R-state of porcine FBPase, which can bind AMP (40,53), the sulfate-ligated state of E. coli FBPase cannot bind AMP because Phe 15 occupies the binding pocket for the nucleotide. Important recognition elements for AMP are at the C-terminal side of helix H1, whereas important recognition elements for the sulfate anion (the proposed binding site of PEP) are at the N-terminal side of helix H1 (Fig. 7). As AMP binds to E. coli FBPase, the conformational perturbation on helix H1 probably disrupts hydrogen bonds to the ligand at the C1-C4 interface. In fact, one of the effects of AMP ligation of porcine FBPase is the conformational change in helix H1 that displaces Ile 10 (which is at the N-terminal side of helix H1) from a hydrophobic pocket (40). Hence, certain features of the allosteric mechanism of AMP inhibition in mammalian FBPases may be evolutionary adaptations of an ancient mechanism of allosteric regulation prevalent in bacteria.
Sequence alignments of Type I FBPases from eukaryotic and bacterial sources reveals a putative activation site common to heterotrophic bacteria from the class Gammaproteobacteria, specifically the orders Entereobacteriales, Vibrionales, and Pasteurallales. These organisms may need to adapt to changing environments and therefore may require a mechanism for rapid activation/inactivation of their gluconeogenic pathway. As concentrations of AMP are uniform for E. coli growing under glycolytic and gluconeogenic conditions (17), FBPase would experience nearly constant inhibition due to AMP. Under glycolytic conditions of growth, concentrations of PEP are low, and AMP would effectively inhibit FBPase. Under gluconeogenic conditions, PEP concentrations are high, displacing AMP and activating the enzyme. Hence, for E. coli and organisms that have the putative activation site, FBPase is normally inactive and subject to dynamic allosteric activation.
Glu 192 of porcine FBPase is conserved among eukaryotic organisms, and is prevalent in Bacteria as well. Approximately 35% of the bacterial FBPases investigated here have glutamate at positions corresponding to 192 of porcine FBPase and lack the signature residues of an anion activation site. The side-chain of Glu 192 probably accomplishes the task of an anion in stabilizing an active tetramer. Hence, the introduction of Glu 192 probably resulted in a constitutively active enzyme, thus requiring a new strategy of regulation. In bacterial FBPases that have glutamate at position 192, that mechanism of regulation is unclear. In eukaryotic organisms, a permanently activated FBPase is subject to dynamic inhibition by Fru-2,6-P 2 . At some unknown point in evolution dynamic inhibition of FBPase by Fru-2,6-P 2 became synergistic with AMP.