Probing the mechanism of the Archaeoglobus fulgidus inositol-1-phosphate synthase.

myo-Inositol-1-phosphate synthase (mIPS) catalyzes the conversion of glucose-6-phosphate (G-6-P) to inositol-1-phosphate. In the sulfate-reducing archaeon Archaeoglobus fulgidus it is a metal-dependent thermozyme that catalyzes the first step in the biosynthetic pathway of the unusual osmolyte di-myo-inositol-1,1'-phosphate. Several site-specific mutants of the archaeal mIPS were prepared and characterized to probe the details of the catalytic mechanism that was suggested by the recently solved crystal structure and by the comparison to the yeast mIPS. Six charged residues in the active site (Asp225, Lys274, Lys278, Lys306, Asp332, and Lys367) and two noncharged residues (Asn255 and Leu257) have been changed to alanine. The charged residues are located at the active site and were proposed to play binding and/or direct catalytic roles, whereas noncharged residues are likely to be involved in proper binding of the substrate. Kinetic studies showed that only N255A retains any measurable activity, whereas two other mutants, K306A and D332A, can carry out the initial oxidation of G-6-P and reduction of NAD+ to NADH. The rest of the mutant enzymes show major changes in binding of G-6-P (monitored by the 31P line width of inorganic phosphate when G-6-P is added in the presence of EDTA) or NAD+ (detected via changes in the protein intrinsic fluorescence). Characterization of these mutants provides new twists on the catalytic mechanism previously proposed for this enzyme.

myo-Inositol-1-phosphate synthase (mIPS) 1 catalyzes the conversion of D-glucose-6-phosphate to L-myo-inositol-1-phosphate, the first step in de novo biosynthesis of myo-inositol. Compounds containing inositol are critical components of signal transduction pathways (1), cell walls in some pathogenic bacteria (2), and various stress responses (3,4). In some hyperthermophilic archaea and bacteria, inositol is used for the synthesis of unique solutes accumulated to balance external osmotic pressure and protect intracellular macromolecules from heat and/or salt shock (for reviews of osmolytes in hyperthermophiles see Ref. 5). The solute di-myo-inositol-1,1Ј-phos-phate (DIP) has been detected in Archaeoglobus fulgidus (6), Methanococcus igneus (7), Pyrococcus sp. (8,9), and several bacteria of the Thermotogales family when the cells are grown at temperatures above 80°C (10). The DIP synthesized by M. igneus is chiral and composed of L-I-1-P units that are synthesized by mIPS (11). Because the archaeal mIPS enzymes commit significant cell resources to DIP synthesis, the production of DIP must be tightly regulated.
The mIPS gene has been identified in the A. fulgidus genome, and the gene product was heterologously expressed in Escherichia coli (12). The recombinant A. fulgidus mIPS is an extremely heat-stable tetramer of 44-kDa subunits (much smaller than the eukaryotic homolog that is a tetramer of 60-kDa subunits). The catalytic activity of the enzyme was also shown to be absolutely dependent on divalent metal ions (Zn 2ϩ , Mn 2ϩ , or Mg 2ϩ ), and more specifically metals are required for the cyclization step of the reaction (12).
The crystal structure of this mIPS has recently been solved to 1.9 Å (13). Despite significant differences in protein length and sequence, the general architecture of the archaeal enzyme is similar to that of the eukaryotic mIPS from Saccharomyces cerevisiae (14) and bacterial mIPS from Mycobacterium tuberculosis (15). The archaeal mIPS did not have divalent cations present in the structure, although in two of the four subunits there was one metal ion whose properties were consistent with K ϩ that interacted with NAD ϩ . A metal ion at this position corresponds to the Zn 2ϩ tentatively identified in structures of M. tuberculosis mIPS (15) and more recently in the yeast IPS (16). However, electrostatic field calculations carried out for the A. fulgidus mIPS suggested that a second metal ion-binding site might be located close to the site occupied by K ϩ (13). Modeling of linearized G-6-P into the active site provided some insight into the role of metal ions, but the energy minimum has the linear G-6-P in an extended conformation, one where C-1 and C-6 are quite distant from one another. This modeling (13), combined with a direct comparison of structures of the archaeal and yeast/bacterial mIPS enzymes, led us to propose a mechanism for the archaeal enzyme that involved the following steps: (i) loss of the inositol C-5 hydroxyl proton to Lys 274 and relay of that proton to Asp 225 (a step that facilitates oxidation at C5); (ii) oxidation of C-5 by the reactive C-4 of the nicotinamide ring; (iii) enolization at the C-5-C-6 bond where Lys 367 acts as a base to withdraw the pro-R proton from C-6 and stabilization of this species by Lys 274 and Lys 367 ; (iv) additional stabilization of the negative charge on C-1 from Lys 306 and Lys 278 , which could both contribute to final hydroxyl regeneration at C-1; and (v) reduction of the inosose via a relay system of Asp 225 and Lys 367 .
To probe the catalytic roles of the individual residues proposed to directly participate in catalysis, we have mutated eight different active site residues to alanine. These include Asp 225 , Lys 274 , Lys 278 , Lys 306 , and Lys 367 , where direct roles in catalysis have been proposed, as well as Asn 255 , Leu 257 , and Asp 332 that might aid in binding of substrate, cofactor, or divalent metal ion to the enzyme. The placement of these residues in the active site of A. fulgidus mIPS is shown in Fig.  1. Characterization of enzyme activity, the ability to carry out the initial oxidation of G-6-P, and ligand binding for each of these mutants provide definitive roles for several of these residues but also suggest a modified catalytic mechanism for the archaeal mIPS.

MATERIALS AND METHODS
Construction of Mutants-A. fulgidus mIPS mutants were constructed using a QuikChange site-directed mutagenesis kit from Stratagene. Overexpression and purification of the mutants were carried out according to the procedures for the wild type mIPS described previously (12). SDS-PAGE on 12% polyacrylamide gel was used to assess the purity of the mutants, and absorbance at 280 nm was used to determine the concentration of the mutants (the extinction coefficient at 280 nm is 50210 M Ϫ1 cm Ϫ1 as calculated from the sequence). All of the mutants were stored at 4°C in 50 mM Tris acetate buffer, pH 7.5.
mIPS Specific Activity Assay-Conversion of G-6-P to I-1-P was monitored via 31 P (202.3 MHz) NMR spectroscopy (12) using a Varian INOVA 500 system with a broadband probe. Assay mixtures (5 mM G-6-P, 1 mM NAD ϩ , 1 mM MgCl 2 in 50 mM Tris-HCl, pH 7.5, with 20% D 2 O) were incubated with mIPS (typically 0.5 g for wt enzyme) at 90°C for 15 min, quenched by the addition of 2 mM EDTA (which stops the mIPS reaction), and kept on ice until used for NMR analyses. 31 P NMR spectra were acquired with a 10115 Hz sweep width, 32384 datum points, 90 o pulse width, 1000 transients, and a 2-s delay time between acquisitions.
NADH Production-Wild type or mutant mIPS (20 -40 M) was incubated with 0.4 mM EDTA, 5 mM G-6-P, and 300 M NAD ϩ at 90°C for 30 min (12). Immediately following the incubation, absorption spectra were recorded at room temperature with a Beckman DU640 spectrometer using a 1-cm cuvette. The absorbance of the protein/ligand mixture at 340 nm was used to identify mutants that produced NADH (the NADH absorption maximum at 340 nm has an extinction coefficient of 6.2 mM Ϫ1 cm Ϫ1 ). 31 P NMR Assay for P i and G-6-P Binding-Enzyme-induced line broadening of the 31 P NMR resonance (at 202.3 MHz on the same spectrometer) of inorganic phosphate, P i , was used to monitor its binding to mIPS enzymes. P i (0.01-10 mM for wild type enzyme) was titrated into a solution containing 50 -70 M enzyme in 50 mM Tris, pH 7.5, 20% D 2 O, with 0.4 mM EDTA, 25°C. 31 P spectra were acquired with the number of transients varied so that S/N Ն 5; the P i line width was measured as the width at half-height minus the line broadening applied to the free induction decay (typically 2-5 Hz depending on concentration of P i ). Instead of directly examining line width changes for G-6-P and other potential ligands binding to mIPS active site, the effect of the ligands on the weight-averaged P i resonance in the presence of mIPS was measured. Displacement of P i by G-6-P sharpens the P i resonance. 31 P spectra were recorded for the mixture of 50 -70 M enzyme, 0.2-0.5 mM P i , and 0.4 mM EDTA in Tris buffer before and after the addition of G-6-P. The binding of glucose-6-sulfate, mannose-6-phosphate, glucose, glucose-6-sulfate, and glycerol phosphate to mIPS was also evaluated by monitoring the line width of P i in the presence of the protein and in the absence and presence of added substrate analog.
Intrinsic Fluorescence Assay for NAD ϩ Binding-A Jobin y Van Fluoromax 3 spectrofluorimeter was used to monitor the binding of NAD ϩ to IPS and IPS mutants (ϳ1 M in 10 mM EDTA with 10 mM KH 2 PO 4 , pH 7). Steady state fluorescence measurements were taken at room temperature with an excitation wavelength of 290 nm and 2-mm excitation and emission slit widths. The mIPS emission was scanned from 300 to 400 nm in the absence and presence of 1-20 M NAD ϩ . Controls, run in parallel, of mIPS where only buffer was added were used to account for small volume changes associated with dilution of the mIPS when the NAD ϩ was added. The addition of NAD ϩ caused a decrease in the mIPS intrinsic fluorescence, which was plotted as ⌬I fl ϭ I o Ϫ I (in arbitrary units). The concentration of NAD ϩ for 50% of the maximum fluorescence change was used to characterize the apparent K D for NAD ϩ .
CD Analysis of Secondary Structure-CD spectra in the far UV were acquired using an AVIV 202 circular dichroism spectrometer. Enzyme samples (ϳ1 M) in both 10 mM KH 2 PO 4 , pH 7, or H 2 O, pH 7 were analyzed. Wavelength scans were taken from 300 to 190 nm at room temperature with a bandwidth of 1.0 nm and an averaging time of 1 s.

Mutant Expression and Stabilization by P i -
The eight mutants expressed were purified to Ͼ90% homogeneity as monitored by SDS-PAGE. The secondary structure of these proteins was monitored by the CD spectra in the far UV region. The CD spectra of all proteins in 10 mM potassium phosphate at pH 7 were comparable (12.5% ␣-helix, 40.2% ␤-sheet, 22.4% ␤-turn, and 29.4% random coil as obtained by the CD spectra deconvolution software; Neural Networks, version 2.0d). That fact suggested the overall secondary structure content of the proteins under these conditions is not altered by the different mutations. Given the unstructured regions around the active site in the yeast crystal structure in the absence of substrate (14), it is likely that the P i binding promotes structural stability and correct folding of the A. fulgidus IPS active site region. Older preparations of wild type mIPS often lose activity that correlates with precipitation of the protein. As little as 5 mM potassium phosphate added to precipitated enzyme is able to resolubilize the protein and restore activity. Thus, P i may aid in limited refolding of the active site regions as well as stabilize a specific folded conformation of the archaeal mIPS.
The P i bound to the purified mIPS was characterized further by 31 P NMR spectroscopy. A spectrum of mIPS (3.0 mg/ml) exhibited a single 31 P resonance at 3.0 ppm (compared with a chemical shift of 0.2 ppm for P i at pH 1). The line width of the bound resonance was 13 Ϯ 1 Hz. A comparison of the chemical shift of this bound resonance to a titration curve for P i taken under the same conditions was consistent with the bound P i having a Ϫ2 charge at the enzyme active site. This is consistent with the crystal structure, which shows a large number of lysine side chains in the phosphate-binding pocket (13). Activity of Mutants-31 P NMR spectroscopy provides an excellent way to monitor mIPS activity (12). In the 31 P spectrum, the ␣ and ␤ isomers of G-6-P (substrate) at ϳ4.4 ppm are well separated from the I-1-P (product) resonance at 4.0 ppm. The NAD ϩ cofactor at 1 mM, which is well in excess of what is needed to saturate the wild type enzyme, is characterized by an AB quartet at Ϫ9.0 ppm. The increase in I-1-P resonance intensity compared with G-6-P (at a concentration of 5 mM, well above the K m ) was used to estimate the mIPS specific activity. Mg 2ϩ (1 mM) present in the assay mixtures was also considerably in excess of the K D of 24 M established for the recombinant enzyme (12). Under these conditions, mIPS exhibited a specific activity of 13.2 mol min Ϫ1 mg Ϫ1 at 90°C. The only mutant that exhibited any conversion of G-6-P to I-1-P was N255A, although the reaction was significantly reduced. The estimated specific activity for N255A was 0.012 mol min Ϫ1 mg Ϫ1 , 0.09% that of recombinant wild type mIPS. All of the other mutants had specific activities much less than 0.001 mol min Ϫ1 mg Ϫ1 because no product was detected over long periods of time and with an increased amount of protein added to assay mixtures.
Generation of NADH by mIPS Mutants-Although most of the mutants did not produce I-1-P, it is possible that they could still carry out the initial oxidation of G-6-P and even the cyclization step to form the inosose compound. The addition of EDTA to the assay mix inhibits the A. fulgidus mIPS cyclization of 5-keto-G-6-P to myo-2-inosose-1-phosphate (12), because divalent metal cations are absolutely required for the aldol condensation reaction. As shown in Fig. 2, with wild type protein, stoichiometric amounts of NADH were produced after incubation of protein with excess G-6-P and NAD ϩ in the presence of EDTA at 90°C. Although depletion of divalent cations dramatically slows down this first step of the reaction (12), this assay can be used to see whether any of the mutants are able to carry out the first step (of the many steps and two intermediates) in I-1-P production. Of the eight mutants, NADH was produced in stoichiometric amounts to protein in only two of the mutants, K306A and D332A. Most of the mutants (e.g. L257A in Fig. 2) could not form NADH under these incubation conditions. The only mutant that exhibited very low catalytic activity, N255A, did not exhibit stoichiometric NADH production, perhaps suggesting that one or both of the ligands were weakly bound under these assay conditions. The ability to form NADH indicates that the involvement of Lys 306 and Asp 332 in catalysis must occur after the oxidation step. For the other residues, the lack of NADH formation suggests that they must be involved in substrate and cofactor binding in the first part of the reaction, stabilization of linearized G-6-P in the active site, deprotonation of the C-5 hydroxyl group, and/or oxidation of C-5.
Detection of G-6-P Substrate and Substrate Analog Binding-The five mutants that were unable to form NADH might, alternatively, be unable to bind substrates (NAD ϩ and G-6-P).
Because P i bound to mIPS could be detected in the 31 P spectrum of the isolated protein, this method could be applicable for detecting G-6-P and other (containing phosphate group) potential ligands binding to the enzyme. However, the 31 P spectrum of ϳ70 M wt mIPS mixed with 1 mM G-6-P in the presence of EDTA showed no broadening of the G-6-P phosphorus line width (which was Ͻ2 Hz in the buffer alone under these conditions). If the K D for G-6-P is comparable with the K m for this substrate, most of the enzyme would be saturated with G-6-P, leading to a significant fraction of enzyme-bound G-6-P. This observation has three possible explanations: (i) G-6-P does not bind without divalent cations bound to the protein, (ii) the bound G-6-P is not in fast exchange with free G-6-P (a possibility that must be considered because during the reaction the intermediates are tightly bound and not released into solution), and (iii) with residual NAD ϩ present, 5-keto-G-6-P might be formed and tightly bound to the enzyme so that it is not released into solution.
In contrast to what was observed with G-6-P, excess P i added to the protein with EDTA present was significantly broadened, indicating that both free and bound P i were in exchange. The P i bound to the enzyme (ϳ180 kDa) would have a long correlation time because of the large size of the complex. The 13 Hz line width observed for P i bound to the purified protein is consistent with P i bound to a large complex. When additional P i was titrated into the mIPS (68 M) solution, the P i line width increased and then narrowed as the P i concentration increased further (Fig. 3) was the difference in line width for free P i at a given concentration and P i when enzyme was present, E o was the total enzyme concentration, and L o was the total P i concentration) was used to estimate the enzyme bound line width, ⌬ b , and K D for P i binding to recombinant mIPS. Values of 19 Ϯ 5 Hz and 0.05 Ϯ 0.02 mM for ⌬ b and K D were obtained. That ⌬ b was larger than the 13 Hz line width measured for P i bound to mIPS in the absence of any added P i might suggest that the ligand binding is in an intermediate exchange regime.
With a concentration of P i and mIPS that leads to substantial line broadening, we can monitor whether, in the presence of EDTA (and at room temperature), substrate can displace P i from the mIPS active site. Any solute added that binds to the active site and displaces the P i should narrow the P i resonance because the bound solute would reduce the population of enzyme-bound P i that is averaged into the observed line width. When G-6-P (1 and 2 mM) was added to mIPS (56 M) and 0.5 mM P i , the observed P i line width (⌬ obs ) narrowed significantly, from an initial value of 7 Ϯ 1 to 3 Ϯ 0.5 Hz with 2 mM G-6-P (Fig. 4). This indicates that G-6-P and P i bind to the same location in the active site. Thus, P i binding can be used indirectly to assess whether other substrate-like molecules bind to the mIPS protein as well as to mIPS mutants.
Previously, we examined 2-deoxy-glucose-6-phosphate, glucose-6-sulfate, mannose-6-phosphate, glucose, and glycerol 3-phosphate as potential substrate analogs for the wt A. fulgidus mIPS. Although none of these molecules were substrates (12), they could still bind to the protein as inhibitors. In the presence of A. fulgidus mIPS, the P i line width decreased significantly only when increasing concentrations (up to 2 mM) of glycerol-3-phosphate were added to the solution (Fig. 4). Hence, this small organophosphate must bind to the mIPS active site to displace P i . The other phosphate-or sulfatecontaining compounds, where the cyclic form is in equilibrium with a linear form, cannot bind tightly to the mIPS, because they do not lead to a narrowing of the P i resonance. The addition of glucose also failed to decrease the P i line width. However, this last result must be viewed with some caution because although it could indicate that this moiety does not bind to mIPS, it might also mean that both glucose and P i can occupy the active site.
The line width of P i in the presence of mIPS could be used to see whether, for mutant mIPS enzymes, (i) P i can bind to the enzyme and (ii) G-6-P can displace the bound P i . In the eight mutants generated, binding of P i (as monitored by an increased line width in the presence of protein) was observed for all but L257A (Table I). The lack of P i line-broadening in the presence of L257A suggests that its active site is sufficiently perturbed so that it no longer binds P i tightly (and by inference G-6-P); if the K D increased from 0.05 mM (wild type enzyme) to 2 mM, we would be unlikely to detect any line width changes under the conditions used. As shown in Fig. 5 for several of the mutants, observed P i line widths reflecting P i binding to mIPS varied depending on the mutant. This could reflect an altered mobility in the P i ⅐mIPS complex or an altered K D that affects the off-rate of the complex. For example, the much larger P i line width in the presence of D332A suggests that removal of this aspartate has increased the affinity of P i for the active site (possibly by relieving like charge repulsion) where exchange between enzyme-bound and free P i occurs on an intermediate time scale. In contrast, the smaller line width for P i in the presence of a comparable concentration of K274A could indicate that P i does not bind as tightly to this mutant.
The increase in P i line width in the presence of mIPS protein was used to check for G-6-P binding to the mutant proteins. Using 0.5 mM P i and 50 -70 M mIPS to generate a P i line width broadened by exchange (with ⌬ obs ranging from 3.6 to 15.6 Hz compared with free P i of 1 Hz; Table I and Fig. 5), we then added 5 mM G-6-P to see whether the mIPS substrate could bind to the mutant and displace P i . As shown in Table I, for most of the mutants, G-6-P could displace bound P i as evidenced by the decreased line width observed for P i . D225A, K274A, K278A, K306A, and D332A all exhibited narrower P i line widths when G-6-P was added. However, N255A and K367 showed little change in the line width of the bound P i when 5 mM G-6-P was added. This indicates that these two mutations have impaired G-6-P binding that is much weaker than P i binding to the wild type protein.
Binding of NAD ϩ -A. fulgidus mIPS activity was shown to

FIG. 4. Line width of 0.5 mM P i in the presence of 68 M mIPS in the absence (black bars) and presence of 1 mM (gray bars) and 2 mM (white bars) substrate or
analog. G-6-P, glucose 6-phosphate; d-G-6-P, 2-deoxy-G-6-P; G-6-S, glucose-6-sulfate; G, glucose; M-6-P, mannose-6-phosphate; gly-3-P, glycerol-3-phosphate. The asterisk indicates analogs that efficiently displaced P i from the mIPS active site. The different ⌬ obs values in the absence of analogs reflect different mIPS preparations and/or experiments done on different days. be significantly increased when exogenous NAD ϩ was added to the protein (12). Although some of the subunits could have NAD ϩ bound already, adding NAD ϩ saturates all the active sites with this cofactor. NAD ϩ binding to vacant sites decreases the intrinsic fluorescence intensity but not the wavelength of maximum emission of the archaeal mIPS protein. This change in fluorescence intensity can be used to quantify NAD ϩ binding to the proteins by measuring the intrinsic fluorescence intensity at 334 nm compared with that for the protein where only buffer was added over an NAD ϩ concentration range of 1-20 M. There are two tryptophans near the active site of A. fulgidus mIPS (13) that could have altered characteristics upon NAD ϩ binding. From the change in fluorescence intensity as a function of added NAD ϩ , we used the concentration for 50% of the maximum change to define an apparent K D for NAD ϩ binding (see Fig. 6 for fluorescence changes for several of the mutants). Although the wild type, N255A, L257A, K306A, and K367A were characterized by K D values Ͻ10 M (Table II), D225A, K278A, and D332A showed considerably weaker binding of the cofactor. NAD ϩ did not induce any changes in the intrinsic fluorescence of K274A in the cofactor concentration range studied. This could indicate a dramatically reduced affinity of the mutant protein for NAD ϩ . However, if mIPS quenching upon NAD ϩ binding is caused by a conformational change that places Lys 274 in the vicinity of one of the Trp residues, removal of this lysine side chain could abolish the sensitivity of the fluorescence to NAD ϩ binding. DISCUSSION Understanding the function of key residues responsible for substrate binding and reaction intermediate formation is critical toward formulating a catalytic mechanism of an enzyme. In a complex enzyme such as mIPS with two ligands required for catalysis to commence and two intermediates, characterization of mutants solely by loss of catalytic activity is inadequate. Additional evidence for significant conformational changes along the pathway and sequestration of the intermediates from dissociation into solution are needed to draw more detailed conclusions about the role of a given residue in catalysis. Indeed, several active site mutants of yeast and M. tuberculosis mIPS have been generated based on the crystal structures (17,18) and shown to be inactive. Exactly at what point in the mechanism the activity has been impaired was difficult to deduce from those experiments. At best the mutagenesis indicated that a particular residue is critical for catalysis, but whether or not it is involved directly in catalysis or indirectly (ligand binding or structural role) cannot be gleaned from lossof-function studies only.
Our prior studies have shown that A. fulgidus mIPS requires  a Line width observed for 0.5 mM P i with 0.07-0.08 mM mIPS. b Decrease in P i line width upon the addition of 5 mM G-6-P. The initial P i line width in the presence of mIPS under these conditions was typically in the range of 4 -7 Hz, except for D332A, where the initial line width under these conditions was ‫61ف‬ Hz, and L257A, which only exhibited a very narrow P i resonance (and in essence serves as a control for free P i ).  (12).
b There was no change in the fluorescence intensity of this mutant when NAD ϩ was added to the protein.
divalent cations for the aldol cyclization step, so the catalysis can be halted after the first step (oxidation at C-5) by chelating the cations with EDTA (12). This enables us to identify residues important for at least the first step of the reaction by monitoring the production of 5-keto-G-6-P and NADH. The addition of assays for substrate binding in mutants that cannot carry out the initial redox step provides a better test of specific defects in archaeal mIPS mutants. A summary of the A. fulgidus mIPS mutants generated, and their responses in the diverse assays and initial roles proposed based on the crystal structure, is presented in Table III. The residues mutated were chosen based on the A. fulgidus mIPS crystal structure with P i and NAD ϩ present and its structural homologies to the yeast and M. tuberculosis mIPS enzymes (13).
N255A was the only mutant to show any formation of I-1-P. The specific activity was less than 0.1% that of the wt enzyme. Because this mutant can convert G-6-P to I-1-P, it must form NAD ϩ . However, in the presence of EDTA and excess substrates very little NADH was detected (certainly not stoichiometric with protein). This suggests that binding of one or both ligands to the mutant enzyme has been impaired. NAD ϩ binding to the protein, as assessed by changes in the intrinsic fluorescence of the protein, exhibited a K D similar to wild type protein; similarly P i also bound. However, a 10-fold excess of G-6-P did not compete well with the bound P i and release it from the enzyme (which would produce a narrow P i resonance). The replacement of Asn 255 by Ala significantly weakens G-6-P binding to the protein.
K306A and D332A are essentially inactive (no I-1-P was detected under the assay conditions) leading to an upper limit on the activity of the mutant enzyme that is 0.01% that of wild type enzyme. However, both mutants produced NADH stoichiometrically to protein in the presence of EDTA, implying that they perform the first and maybe the second step in the catalytic mechanism. It is likely these residues are not involved in the proper positioning of the substrate necessary for the initial reduction and proton transfer to NAD ϩ . One would assume that the proton transfer is reversible, utilizing the same residues for both the reduction and oxidation of NAD ϩ /NADH. Additionally, it has been suggested from the structure that the K ϩ can be displaced by a divalent cation (13). Although a second metal ion was not detected in the A. fulgidus mIPS crystal structure, the surface electrostatic potential suggested a second metal ion-binding site close to Asp 332 . This suggested that Asp332 could serve as a ligand for the second metal ion (Zn 2ϩ , Mn 2ϩ , or Mg 2ϩ ) that is necessary for catalysis (12). The kinetic results for both mutant enzyme K306A and D332A are consistent with these roles. Interestingly, D332A binds NAD ϩ much more weakly than wild type protein, which might suggest that the second ion is critical for correct orientation of the cofactor.
The other five mutant mIPS were blocked before the first redox step. In the crystal structure, Leu 257 is thought to be coplanar with the closed ring of the product and could help to stabilize binding of glucose moieties. Clearly, L257A does not bind G-6-P or even P i well, suggesting a significant reorganization of active site residues. Thus, Leu 257 has a primary role in substrate binding possibly by aiding the organization of the active site. A plausible hypothesis is that it participates in folding of a short helical fragment of residues 259 -264 that is critical for active site structure. Lys 274 was proposed to abstract a proton from the hydroxyl group of G-6-P C-5 (13). The various assays are consistent with that role if one assumes that the complete lack of K274A quenching with added NAD ϩ means that that residue plays a role in the fluorescence change upon cofactor binding. For D225A, P i and G-6-P binding are similar to wild type protein, whereas NAD ϩ binding is much weaker (an apparent K D of 44 M compared with 1 M for recombinant mIPS). It is likely that the main role of this residue must be in stabilization of NAD ϩ binding.
K367A can bind P i , but it has lost or at least significantly reduced its affinity for G-6-P, which would explain its inability to produce NADH. Lys 367 was initially proposed as the base in the enolization step that occurs after formation of 5-keto-G-6-P (13). Because the first step does not occur with this mutant, this residue must have a primary role in an earlier event, likely substrate binding. The remaining mutation, K278A, generated protein that could bind G-6-P but had greatly reduced affinity for NAD ϩ . The crystal structure of A. fulgidus mIPS shows that Lys 278 is not close to the active site of the nicotinamide ring. Nonetheless, the ability to bind NAD ϩ productively and generate NADH has been lost.
Several of these results are surprising. For example, one of the residues whose removal seemed to have the largest effect on NAD ϩ binding (e.g. K278A) is not close enough to directly participate in stabilization of bound NAD ϩ . Therefore, the effect of Lys 278 on NAD ϩ binding must work through some secondary effects such as electrostatic or Van der Waals' stabilization of the active site. The impaired substrate (but not P i ) binding to K367A and the lack of conversion to NADH with substrate and EDTA present indicate that the primary defect for Lys 367 must occur prior to the first oxidation step. This is in opposition to what the crystal structure suggested. Clearly, the mutagenesis results suggest this lysine plays an initial role in binding the acyclic G-6-P. However, Lys 278 is also proposed to be critical for binding the substrate via interactions with the phosphate, and although substrate binding may be weaker in K278A, it still is observed (albeit indirectly in the 31 P NMR experiment where bound P i is displaced). The interesting question is how to interpret the lack of G-6-P binding by K367A.
There is a major problem in all the structural work with mIPS. The yeast enzyme is the only one with a substrate analog bound, and it is oriented in an extended conformation that does not have substrate C-1 or C-6 near one another (17). Inhibitors of the enzyme that are analogs of acyclic G-6-P bind quite tightly and are potent inhibitors of the yeast mIPS, whereas cyclic compounds are very poor inhibitors (19). However, to generate the inosose ring, C-1 and C-6 need to be near one another. For this, one needs an enzyme-bound pseudocyclic form of G-6-P, undoubtedly a high energy conformation. The small amount of acyclic G-6-P in solution will not exist in a pseudo-cyclic conformation but will adopt extended confor- mations. When it binds to the enzyme it would have to interact with the correct combination of Lys and Asp groups that line the active site. Modeling studies with the A. fulgidus mIPS end up generating an extended acyclic G-6-P orientation that cannot be what exists on the enzyme along the catalytic pathway. Indeed, cyclic G-6-P cannot be modeled into the active site of the archaeal enzyme; a similar result was observed with the yeast enzyme. Nonetheless, binding a cyclic molecule must be possible because inosose-1-P is formed and held tightly in the active site prior to reduction to I-1-P and release into solution.
There must be a sizeable conformational change that brings C-1 and C-6 near one another for C-C bond formation. This conformational change sequesters the active site from solution so that bound G-6-P is not in exchange with free G-6-P. Assuming that the K D for substrate binding is comparable with the K m (ϳ0.1 mM) and the likely 31 P chemical shift difference is small ( 31 P chemical shifts are dominated by the ionization state of the phosphate and the shift for P i suggests it has a net charge of Ϫ2 equivalent to what it is in solution at pH 7.5-8), the exchange between bound and free G-6-P forms should not be in the slow exchange regime if the two forms are in equilibrium. Because no line broadening of G-6-P was observed in the presence of mIPS, whereas P i was effectively broadened by the protein, free and bound G-6-P must not be in equilibrium, at least on the NMR time scale. The slow NMR step could be interconversion of cyclic and acyclic G-6-P bound to the enzyme. A large conformational change in the protein could further retard the exchange of free and bound solute. With a large change involving the active site and its accessibility to solution, it may be difficult to predict what ligand/protein interactions really exist in the active site. In support of a large conformational change, we have noted that protein solubility decreases from at least 10 mg/ml to Ͻ2 mg/ml when G-6-P is incubated with the protein, excess NAD ϩ , and EDTA and heated to 85°C to generate bound NADH and 5-keto-G-6-P. 2 So how do the mutagenesis results fit into the proposed mechanism for mIPS? Most are consistent with what one would predict from the crystal structure. The two real anomalies are K367A and N255A. There is the complete loss of activity for K367A and the nearly complete loss of activity of N255A. Although one could explain decreased activity for both these mutants, what is more surprising is the dramatically reduced binding of G-6-P to each enzyme, whereas P i still binds to both. By analogy to the yeast enzyme (14,16,17), Asn 255 could interact with the 2-OH of linearized G-6-P (although it is further away in the A. fulgidus enzyme), and Lys 367 would be necessary for orienting the 5-hydroxy of G-6-P (or the keto group in 5-keto-G-6-P). There are several other residues in the active site that could hydrogen bond to acyclic G-6-P. It may not be oriented quite right, but one would expect linearized G-6-P to still bind to K367A. Our indirect assay of G-6-P binding uses a large excess of G-6-P (5 mM) added to displace P i (total concentration 0.5 mM) from the active site of mIPS. For a Ͻ0.2-Hz decrease in line width, the G-6-P would have to bind much more weakly to the protein. However, there is another potential explanation. The 5-OH is not available in the cyclic G-6-P form because that oxygen is the ring oxygen; this might strongly suggest that Lys 367 aids in promoting and stabilizing the acyclic form and generating the bound pseudo-cyclic G-6-P by protonating the ring G-6-P ring oxygen. Lys 367 may also be involved in proton transfer reactions involving the C-5 oxygen at later steps in the reaction, but its removal halts catalysis at a much earlier step. Asn 255 could either directly interact with the C-1 hydroxyl or aid in proton transfer to an acidic side chain. Deprotonation of the C-1 oxygen would stabilize the aldehyde. If both of these side chains work concurrently, one might imagine generating a pseudo-cyclic conformation of G-6-P bound to the enzyme that is poised for formation of the C-C bond in the second chemical step of the reaction. The absence of Lys 367 in this model would preclude stabilizing G-6-P in a pseudo-cyclic orientation and G-6-P would not bind well, although P i should have no difficulty binding to the mIPS mutant.
The mechanism suggested for this mIPS is presented in Fig.  7. The first step is the stabilization of a bound acyclic form of G-6-P, which we propose is carried out in large part by Lys 367 and Asn 255 . Asn 255 is shown as having a direct interaction with the substrate C-1 oxygen, but it could be part of a network that interacts with this part of the substrate. The first chemical step would become the simultaneous deprotonation of the C-5 hydroxyl by Lys 274 (or Lys 367 ) and oxidation of C5 with direct hydride transfer to NAD ϩ . Subsequently, the withdrawal of the pro-R proton from the C-6 position (by the phosphate group or one of the many lysine residues that has become deprotonated) promotes enolization of the C-5-C-6 bond. The developing negative charge is stabilized by Lys 274 and/or Lys 367 . At the other end of the active site Lys 306 in concert with Asp 332 protonates the C1 oxygen in concert with the aldol condensation. The C-C bond formation is only completed in the presence of the second metal ion. Therefore the second metal ion must bind close or directly coordinate one of the neighboring OH at positions C-1 or C-2. Finally, the reduction of the oxygen atom at C-5 is achieved by hydride transfer and one of the nearby lysine residues protonates the oxygen to generate I-1-P.