A Novel Phosphoglucose Isomerase (PGI)/Phosphomannose Isomerase from the Crenarchaeon Pyrobaculum aerophilum Is a Member of the PGI Superfamily STRUCTURAL EVIDENCE AT 1.16-Å RESOLUTION*

The crystal structure of a dual specificity phosphoglucose isomerase (PGI)/phosphomannose isomerase from Pyrobaculum aerophilum (PaPGI/PMI) has been determined in native form at 1.16-Å resolution and in complex with the enzyme inhibitor 5-phosphoarabinonate at 1.45-Å resolution. The similarity of its fold, with the inner core structure of PGIs from eubacterial and eukaryotic sources, confirms this enzyme as a member of the PGI superfamily. The almost total conservation of amino acids in the active site, including the glutamate base catalyst, shows that PaPGI/PMI uses the same catalytic mechanisms for both ring opening and isomerization for the interconversion of glucose 6-phosphate (Glc-6-P) to fructose 6-phosphate (Fru-6-P). The lack of structural differences between native and inhibitor-bound enzymes suggests this activity occurs without any of the conformational changes that are the hallmark of the well characterized PGI family. The lack of a suit-able second base in the active site of PaPGI/PMI argues against a PMI mechanism involving a trans -enediol intermediate. Instead, PMI activity may be the result of model against these data using REFMAC5 (28), the model was improved manually using XTALVIEW (29). Further refinement cycles consisted of the addition of solvent atoms using ARP/ wARP, refinement with REFMAC5, and model building with XTALVIEW. Refinement of side chain alternative conformations and aniso- tropic temperature factors was included toward the end of the refinement process. The secondary structure of the final model was calculated using DSSP (30). Inhibitor-bound Structure— Crystals were soaked for 5 days in mother liquor solutions containing 26% glycerol and 5 m M of the PGI inhibitor 5-phosphoarabinonate (PAB) (31). Synchrotron data were col- lected at the SER-CAT beamline in the same way as the high resolution native data. PAB bound at the active site was visualized by refinement of the native model against the data collected from the ligand-soaked crystal, followed by examination of the (cid:1) F o (cid:1) (cid:2) (cid:1) F c difference electron density map. After fitting of the PAB molecule, the structure was refined by using the same protocol used for the native structure and using the same free R assignments as the native data. Structure su-perimpositions were performed using the CCP4 program LSQKAB (32). Coordinates and structure factors for both structures have been deposited with the Protein Data Bank under code 1TZB for native and 1TZC for complex with PAB.

Phosphoglucose isomerase (PGI, 1 EC 5.3.1.9) catalyzes the interconversion of D-glucose 6-phosphate to D-fructose 6-phosphate via an aldose-ketose isomerization reaction. This equilibrium reaction is part of glycolysis and gluconeogenesis but also impacts other pathways in sugar metabolism such as the pentose phosphate pathway. The enzyme from bacterial and mammalian sources has been well characterized. Crystal structures show the enzyme to be a tight homodimer in which the two active sites are located at the domain interface and are formed by elements from both subunits (1)(2)(3). These structures support a catalytic mechanism for isomerization in which a glutamate (e.g. Glu-357 in rabbit PGI) acts as a base catalyst to remove a proton from C-1 or C-2 (depending on the direction of the reaction), forming a cis-enediolate intermediate. Because the open chain forms of its substrates are expected to be present in vivo in trace amounts (4), PGI also catalyzes a ringopening reaction. This reaction is acid-catalyzed by a histidine (e.g. His-388 in rabbit PGI) (5,6), and a lysine (Lys-518) also appears to assist this reaction by abstracting a proton from C-1 of Glc-6-P (or C-2 of Fru-6-P) (7).
Sequences homologous with PGI cannot be recognized readily within the genomes of Archaea. In some species, such as the euryarchaeons Pyrococcus furiosus and Thermococcus litoralis, PGI activity appears to be catalyzed by a novel enzyme that is structurally and mechanistically distinct from the PGI superfamily (8 -10). Crystal structures of this protein from P. furiosus show it to contain a cupin fold (11) at the heart of which is a metal ion that is believed to mediate a hydride shift mechanism of catalysis (12). An interesting facet of this structure is the lack of any obvious amino acids that might catalyze ring opening, leading to suggestions that, in the extremely high temperatures in which P. furiosus thrives, the proportions of sugars in their straight chain form is sufficient to support the demands of metabolism (12).
In some aerobic crenarchaeota, genes are present in which sequence similarity to some, but not all, of the highly conserved active site motifs of PGI can be detected, suggesting that, unlike the euryarchaeota, these archaeal species may contain PGIs that are distantly related to eubacterial and eukaryotic PGIs (13). The proteins from three of these genes, from Aeropyrum pernix, Thermoplasma acidophilum, and Pyrobaculum aerophilum, have been characterized and show PGI activity (13,14). Most interestingly, these enzymes also exhibit phosphomannose isomerase (PMI) activity and can catalyze the interconversion of mannose 6-phosphate (Man-6-P) (the C-2 epimer of Glc-6-P) to Fru-6-P at an equal rate as Glc-6-P to Fru-6-P (13, 14) (Fig. 1). The lack of any recognizable pmi gene in these species suggests that this PMI activity may have a function in vivo. Together with homologues from Sulfolobus species, Thermoplasma volcanicum, and Aquifex aeolicus, these enzymes appear to comprise a novel PGI/PMI family within the PGI superfamily (13).
The mechanism of PMI activity in these PGI/PMIs is inter-esting because the specificity of "conventional" PGIs for Glc-6-P and Fru-6-P is essentially absolute (15). Although the conventional enzyme can interconvert the anomeric forms of Man-6-P (16), it will not isomerize this substrate to Fru-6-P or epimerize it to Glc-6-P (except at extremely low and nonphysiological rates (17)). PMI activity within a PGI can be explained by one of two mechanisms (17). One mechanism is to use a second base catalyst in the active site and the reaction proceeds via a trans-enediol intermediate. The other mechanism is to rotate the C-2-C-3 substrate bond after proton abstraction from C-2 and prior to proton donation to C-1 (in the Glc-6-P to Fru-6-P direction), and this would presumably require a larger active site to accommodate the rotation.
To determine whether these dual PGI/PMIs do belong to the PGI superfamily and to elucidate the structural basis for both enzyme activities, we have determined the structure of PGI/ PMI from P. aerophilum in native form at 1.16-Å resolution and in complex with the PGI inhibitor 5-phosphoarabinonate at 1.45-Å resolution. These structures reveal an unexpectedly high degree of similarity with eubacterial and eukaryotic PGIs, but they also show a subtle difference in the active site architecture that may be responsible for the altered specificity.

EXPERIMENTAL PROCEDURES
Structure Determination-The crystallization of PGI/PMI from P. aerophilum (PaPGI/PMI) has been described previously (18). Briefly, the crystals belong to space group P2 1 with cell dimensions a ϭ 55.1 Å, b ϭ 100.8 Å, c ϭ 55.8 Å, and ␤ ϭ 113.2 o , and an initial native data set extending to 1.6 Å was reported (18). The crystals were cryo-protected over a period of several hours by passage through a series of mother liquor solutions (25% polyethylene glycol 8000 (w/v) and 0.22 M ammonium sulfate, buffered with 0.1 M Tris-HCl, pH 8.5) each containing increasing amounts of glycerol in 2% increments up to a maximum of 26%. Heavy atom derivatives were prepared by exchanging this solution with an equivalent solution containing a heavy atom salt. Several such compounds were tested. The method for specific iodination of aromatic residues has been described previously (19). The crystals were then flash-frozen to Ϫ180°C in situ using a cryostream (X-Stream 2000: Rigaku-MSC). Diffraction data were recorded with an RAXIS-IVϩϩ imaging plate system (Rigaku-MSC) mounted on a Rigaku RU-H3R copper rotating anode generator, operating at 50 kV and 100 mA. The x-ray beam was conditioned with Confocal Maxflux TM optics (Osmic, Inc.). For these data sets, the crystal-to-detector distance was 120 mm, and a typical exposure time was 3 min per 1.0°oscillation image. These data were processed using CrystalClear (20). Derivatives were identified by calculation of Patterson maps using PHASES (21), and phasing calculations at 2.0 Å were performed using autoSHARP (22,23) followed by solvent flattening with phase extension to 1.8 Å using SOLOMON (24).
Model Building and Refinement-The experimental phases were the starting point for automated model building using the program ARP/ wARP (25,26). To obtain the best data set for refinement, new native data were collected at the SER-CAT beamline ID22 at the Advanced Photon Source (Argonne National Laboratory.). These data were acquired on a MAR225 CCD detector with exposure times of 1 s per image, a crystal-to-detector distance of 100 mm, and an oscillation angle of 0.5°. To ensure high redundancy of the data, 360°were collected. Processing was performed using the HKL2000 software package (27). After refinement of the initial model against these data using REFMAC5 (28), the model was improved manually using XTALVIEW (29). Further refinement cycles consisted of the addition of solvent atoms using ARP/ wARP, refinement with REFMAC5, and model building with XTAL-VIEW. Refinement of side chain alternative conformations and anisotropic temperature factors was included toward the end of the refinement process. The secondary structure of the final model was calculated using DSSP (30).
Inhibitor-bound Structure-Crystals were soaked for 5 days in mother liquor solutions containing 26% glycerol and 5 mM of the PGI inhibitor 5-phosphoarabinonate (PAB) (31). Synchrotron data were collected at the SER-CAT beamline in the same way as the high resolution native data. PAB bound at the active site was visualized by refinement of the native model against the data collected from the ligand-soaked crystal, followed by examination of the ͉F o ͉ Ϫ ͉F c difference electron density map. After fitting of the PAB molecule, the structure was refined by using the same protocol used for the native structure and using the same free R assignments as the native data. Structure superimpositions were performed using the CCP4 program LSQKAB (32). Coordinates and structure factors for both structures have been deposited with the Protein Data Bank under code 1TZB for native and 1TZC for complex with PAB.

RESULTS
Structure Determination-The structure was solved by multiple isomorphous replacement using three derivatives, two gold and one iodine (Table I). Although three iodine peaks were visible in the Patterson map, this derivative had low phasing power and likely contributed little to the final phases. Phasing with SHARP at 1.8-Å resolution produced an experimental electron density map of excellent quality (Fig. 2a) and permitted automated model building of the entire structure except the methionines at each N terminus. After refinement against high resolution data extending to 1.16 Å, the final model has an R factor of 15.0% and R free of 16.4%, with excellent stereochemistry (Table II). This model comprises two subunits (of 300 and 301 residues, respectively), 625 water molecules, as well as several molecules of sulfate and glycerol from the crystallization solution. A molecule of sulfate occupies the substrate phosphate-binding site in the active site of each subunit. Only the C-terminal glutamine of subunit B is not visible in the electron density due to apparent flexibility in the vicinity of the C terminus. The N-terminal methionine appears to be absent from the protein because the amino group of Ala-2 forms an electrostatic interaction with a neighboring aspartate, and there is no room for an additional residue. The extremely high resolution of the structure also permitted the modeling of 47 side chains and some main chain regions with alternative conformations: 22 in subunit A and 25 in subunit B. The final 2(͉F o ͉ Ϫ ͉F c ͉) electron density map at 1.16-Å resolution is shown in Fig. 2b.
Structure Description-The structure of PaPGI/PMI is a tight dimer of essentially identical subunits; the two subunits superimpose with an r.m.s. deviation of 0.71 Å for all atoms. The structure of one subunit and the dimer are shown in Fig. 3. The subunit comprises two domains, each of which is built around a parallel ␤ sheet, five-stranded in the N-terminal domain and four-stranded in the C-terminal domain. The N terminus is located approximately between the two domains and extends across the face of the N-terminal domain before forming a ␤ hairpin structure (␤1 and ␤2). Thereafter, the N-terminal domain is comprised of alternating ␣␤ segments with ␣-helices connecting ␤ strands except for the connection between ␤4 and ␤5, which is not helical. After ␣7, the chain crosses over to the C-terminal domain where the same pattern of alternating ␣␤ structure occurs. The C terminus is helical (␣15) and is also located between the two domains.
The dimer is compact and globular in shape, with no significant extended structural features (Fig. 3b). Two short helical segments, ␣11 and ␣12, undergo domain swapping and are an integral part of the opposite subunit. At the dimer interface are numerous ionic interactions that may contribute to the thermostability of the enzyme.
Comparison with Conventional PGI-Even though the sequence similarity between them is barely detectable (13), PaPGI/PMI shares a common fold with conventional PGIs. The structure was superimposed with that of rabbit PGI (rPGI) (6), using an algorithm based on secondary structure matching (33), and shows the structural elements that are common to both proteins (Fig. 4). The overall domain structure is the same, but it is immediately obvious that PaPGI/PMI is far smaller and overlaps mostly with the protein core of rabbit PGI. This superimposition permits a structure-based alignment of the PGI sequences from rabbit and P. aerophilum, which shows the relationship between the two proteins more clearly (Fig. 5). A large part of the absent structure in PaPGI/ PMI corresponds to the N-terminal end of rabbit PGI, consisting of seven ␣ helices and two ␤ strands, which together form the outer surface of the protein from rabbit. The absence of ␤1 and ␤2 (rPGI nomenclature) leaves a parallel four-stranded ␤ sheet in PaPGI/PMI because in rabbit PGI these two strands are anti-parallel and so create a mixed parallel/anti-parallel six-stranded sheet in that enzyme. The C terminus is also shortened; the final helix (␣24 in rPGI) is absent, and prior to that, ␣15 is only half the length of its equivalent in rPGI, ␣23. The latter helix is important because in conventional PGI it moves toward the active site after ligand binding and contains a lysine (Lys-518) that is critical for catalysis (34). Finally, in rabbit PGI, the structure that forms a "hook" (␣20 and ␣21),

TABLE I
Statistics of the phasing using three derivatives in the multiple isomorphous method at 2.0-Å resolution followed by phase extension to 1.8 Å The native data used here were reported previously (18). The figure of merit after phasing (at 2.0 Å) was 0.53. Derivative abbreviations are as follows: AuBr2, gold(III) bromide; AuKBr1, potassium tetrabromoaurate(III); and iod4, iodine.
where F P , F PH , and F H are the protein, derivative, and heavy atom structure factors, respectively, and E r.m.s. is the residual lack of closure (͉F PH Ϫ F P ͉ Ϫ ͉F H ͉). iso, isomorphous data; ano, anomalous data. and extends to mediate intersubunit contacts in conventional PGIs (6), has no counterpart in PaPGI/PMI. In addition to these major differences, many of the connections are shorter in PaPGI/PMI.
Complex with 5-Phosphoarabinonate-To establish the identity of the active site of PaPGI/PMI, the structure was deter-mined in complex with PAB, a well known inhibitor of PGI activity (31), at 1.45-Å resolution (Table II). PAB bound in an identical manner to both active sites in the dimer, and the view of molecule A is shown in Fig. 6. At one end of the active site, the phosphate group is oriented by three serines (Ser-48, Ser-87, and Ser-89) and one threonine (Thr-92). In the middle of the  inhibitor, the C-4 hydroxyl (equivalent to the ring oxygen in the substrate) is within hydrogen-bonding distance of Lys-298 and His-219 (the latter residue belonging to the adjacent subunit in the dimer); the C-3 hydroxyl is contacted by the amide of Gly-47 and the C-2 hydroxyl by the carbonyl group of His-219. In PAB, a carboxylate group replaces the C-1-C-2 region of the substrate such that O-1␣ is equivalent to C-1 and the carbon at position 1 is equivalent to C-2 of the substrate. Glu-203 is approximately equidistant from both of these atoms, showing that this residue is best placed to abstract and donate protons to the C-1 and C-2 positions of the substrate.
This structure permits a direct comparison with that of rabbit PGI in complex with the same inhibitor (35). The active sites of the two structures were superimposed by using the coordinates of PAB (Fig. 7). This shows that the majority of amino acids forming the active site are conserved between FIG. 4. A comparison of PGI/PMI from P. aerophilum and rPGI (6). a, stereo view showing the superimposition of the backbones of one subunit from each structure (the A subunit from both). The PaPGI/PMI is colored red, and rPGI is colored blue. b, a ribbon representation of the same subunit from rabbit PGI in which those regions corresponding to the PaPGI/PMI structure are colored red and the remainder is colored blue. The figure was produced using PYMOL (40), MOL-SCRIPT (41), and Raster3D (42).

FIG. 5.
A structure-based sequence alignment of PaPGI/PMI with rPGI. The PGI sequences from P. aerophilum and rabbit are denoted P.a. and O.c. (for Oryctolagus cuniculus), respectively. The secondary structure assignments for each structure are also shown and are color ramped blue-to-red in the N-to C-terminal direction. Residues within the active site are highlighted in red, including the highly conserved residues responsible for catalysis. This figure was produced using the software SecSeq (D. Brodersen, unpublished).
conventional PGI and PaPGI/PMIs. The cluster of threonines and serines that forms the sugar phosphate-binding site in conventional PGI (3) is conserved in PaPGI/PMI as Ser-48, Ser-87, and Thr-92 with just a threonine to serine change at position 89. Residues that are important for catalysis in conventional PGI are also conserved in PaPGI/PMI; Glu-357 in rPGI is represented by Glu-203 in PaPGI/PMI, His-388 by His-219, and Lys-518 by Lys-298. There are some differences, however, most notably a proline (Pro-134) in PaPGI/PMI in place of Gly-271, which lead to an alteration in the conformation of ␤7 to ␣6 loop in comparison to the same loop in rPGI, and Thr-291 in place of Gln-511. The homology evident between the two active sites confirms PaPGI/PMI as a member of the PGI superfamily (13). In addition, the lack of any residue in   7. A comparison of the structure of PGI/PMI from P. aerophilum in complex with PAB with a structure of rabbit PGI in complex with the same inhibitor (35). In this stereo view, only the active sites of each structure (the B subunit in both cases) are shown. Residues surrounding the ligand, and the ligand itself, are colored yellow and green for the P. aerophilum and rPGI structures, respectively, and are numbered according to their respective sequences. The histidine residues (219 and 388) are colored orange (PaPGI/PMI) and cyan (rPGI) because in both structures this residue belongs to subunit A of the dimer. The figure was produced using PYMOL (40). the vicinity of the carboxylate group of PAB, other than Glu-203, that might act as a base catalyst, shows that the PMI mechanism of this enzyme is unlikely to use a trans-enediol intermediate (discussed below).
Conformational Changes Upon Ligand Binding-To determine whether conformational changes occur in PaPGI/PMI in response to the binding of ligands at the active site in the same manner as PGI from eubacterial and eukaryotic sources (e.g. in rabbit PGI (6, 36)), the native structure and its complex with PAB were superimposed. The r.m.s. deviations calculated between all main chain atoms in the structures is 0.22 Å. Examining the superimposed structures reveals almost no structural differences between the native and PAB-bound structures (Fig.  8). The exception is a slight shift in the C-terminal helix in subunit B, which is due to an improvement in the ordering of this region compared with the wild-type structure. In particular, the C-terminal residue Gln-302 is now visible and hydrogen bonds a water molecule that is close to the PAB inhibitor. Other than this, the positions of all of the residues within the active site region are essentially unchanged. Moreover, given the very close overlap of residues in rabbit PGI and PaPGI/PMI when both complexed to PAB, it is clear that the native state of PaPGI/PMI is equivalent to the ligand-bound "closed" form of rabbit PGI. DISCUSSION A major goal of this work was to determine whether PaPGI/ PMI belongs to the superfamily of PGI, as suggested by sequence similarity with some of the motifs that comprise conventional PGI (13,18). Our crystal structure of this enzyme confirms this is indeed the case. The core structure of the enzyme has the same fold, and the main differences arise from extensions in conventional PGIs, at both termini and by the insertion of residues that corresponds to the hook structure, that together form an additional "layer" around the protein.
Thus, PaPGI/PMI might represent a minimal PGI fold and that, during the course of evolution, the protein has increased in size by additions of ϳ100 residues at the N terminus, 30 residues between ␣2 and ␤1, 25 residues which form the hook structure, and 35 residues at the C terminus, as well as more gradual increases in the size of connecting loops. The similarity, however, is most pronounced at the active site. In conven-tional PGI this is formed by six motifs, and although only two of these could be recognized in the sequence for PaPGI/PMI with a tentative assignment for two others (13,18), the structure shows that all six are present in PaPGI/PMI. Of the residues that comprise the immediate substrate-binding pocket, only two differ between PaPGI/PMI and mammalian PGIs. The evolutionary constraint to maintain these specific residues for the mechanism of isomerization must therefore be very restrictive, but at the same time, the differences between the two enzymes shed light on how this enzyme can also function as a phosphomannose isomerase.
Mechanism of Phosphoglucose Isomerase Activity-Although much of the catalytic mechanism of PGI has been elucidated from crystal structures of mammalian and bacterial PGIs (1,3,6,35), the very high resolution of the structures presented here reveals it in greater detail. In this mechanism, a glutamate acts as a base catalyst and, in the aldose to ketose direction, abstracts a proton from C-2 and donates it back to C-1 (3,6,35). A separate proton moves between the carbon-bound oxygens, i.e. O-1 and O-2. The intermediate in this reaction is a cisenediolate, and its negative charge is stabilized by an arginine. The presence of Glu-203 in the same position as Glu-357 in rabbit PGI and Arg-135 in place of Arg-272 (see Fig. 7) confirms that an identical mechanism for PGI activity operates in PaPGI/PMI. As the structure of PaPGI/PMI in complex with PAB shows, the architecture of the enzyme is well suited for this mechanism because the O-⑀1 carboxylate oxygen of Glu-203 is approximately equidistant from O-1␣ and C-1 (2.7 and 3.1 Å, respectively), which are equivalent, respectively, to C-1 and C-2 of the substrates. First, this shows that proton abstraction/donation for the phosphoglucose isomerase reaction can take place from either carbon without any rearrangement in either the substrate or the active site. Second, it suggests that 3 Å is the ideal distance for proton abstraction to occur and that the precise binding interactions between substrate and active site residues serve to optimize this distance. Finally, it also illuminates why the reaction direction is determined by the relative concentrations of the two substrates in the cellular medium and not by an intrinsic property of the enzyme because, once a proton has been abstracted from C-2 of Glc-6-P, it could easily be re-donated back to the same atom, and Glc-6-P will be the product.
PaPGI/PMI Does Not Require Conformational Changes for Activity-A feature of crystal structures of PGIs from eubacterial and eukaryotic sources is a number of conformational changes that appear to occur upon the binding of inhibitor or substrate molecules (3, 6, 36 -38). In rabbit PGI, for example, these comprise the closure of residues around the sugar phosphate, an inward shift of a 3/10 helix that carries His-388, and the movement of helix ␣23, which brings Lys-518 closer to the active site (6,36). Because these shifts bring important residues, either for catalysis or substrate binding, closer to the active site, they appear to be essential for the catalytic functioning of PGI. Thus, the native state of conventional PGI is open in structure but, by an induced fit mechanism in response to substrate binding, it transitions to a closed form required for catalysis. By contrast, when PaPGI/PMI binds 5-phosphoarabinonate there are virtually no changes in the structure of the enzyme, and this is because the native enzyme is already in the closed form that is more akin to a conventional PGI bound by a ligand. Although this could be induced in part by the sulfates occupying the phosphate-binding site in each active site of the native enzyme, it would not explain the closed state of the remaining two regions that shift in conventional PGIs. The lack of any movement in the archaeal enzyme in response to the binding of ligand may be due to its inherent thermostability, which would tend to restrict any flexibility in the protein. Nevertheless, the question as to why conformational changes are apparently required in one branch of the PGI superfamily, but not another, is an interesting one.
Mechanism of Sugar Ring Opening in PaPGI/PMI-Absent from the discussion so far is the issue of sugar ring opening. PGIs from mesophilic organisms contain a conserved histidine that acts as an acid catalyst in ring opening by donating a proton to the ring oxygen, forming a hydroxyl at C-5 (5,6). They also contain a lysine that may assist this reaction by concomitantly abstracting a proton from the C-1 hydroxyl (7). Both of these residues are conserved in PaPGI/PMI, as His-219 and Lys-298, respectively; hence, a similar mechanism for ring opening must also exist in P. aerophilum. The mere existence of a ring-opening apparatus, however, raises an interesting issue. Its presence in an organism that thrives at temperatures close to 100°C (39) suggests that the proportion of sugar substrate present in the straight-chain forms is not significantly different from the trace amounts detected at 25°C (4), and this has implications for other archaeal species. One of the most intriguing aspects of the crystal structure of cupin-type phosphoglucose isomerase from P. furiosus (PfPGI) was the apparent lack of any residues that might catalyze the opening of the sugar ring (12). Given this, it was tentatively suggested that at such extreme temperatures a greater proportion of straight-chain sugars might be present, thus precluding the requirement for a ring-opening step in the PGI reaction (12). In the light of the PaPGI/PMI structure presented here, in which the catalytic machinery for ring opening is overtly present, other hypotheses for ring opening in the reaction catalyzed by PfPGI must be considered.
How PGI Catalyzes PMI Activity-At first sight, it would not appear difficult for a PGI to catalyze PMI activity; both reactions are aldose-ketose isomerizations and both involve the interconversion of the transfer of a proton between C-1 and C-2 (aldose to ketose direction). But in actual fact, the reversal of the configuration at C-2 between Glc-6-P and Man-6-P creates a mechanistic challenge because the proton must be removed from the opposite side of the substrate in Man-6-P, compared with Glc-6-P (15). This means that for a PGI to interconvert Man-6-P to Fru-6-P, a single base cannot easily abstract and re-donate a proton with the same ease as for the Glc-6-P to Fru-6-P conversion. One solution to this problem is for the active site to contain a second base (17) so that protons can be abstracted/donated on both faces of the substrate. However, this is ruled out in PaPGI/PMI by the lack of any other residues in the active site with the potential to act as a base within reach of the C-1-C-2 region of the substrates. Clearly then, another mechanism must operate for the PMI activity of PaPGI/PMI. An alternative is the rotation of the C-2-C-3 substrate bond after proton abstraction and before proton re-donation (17), and this would permit the same base to be used for both abstractions on both faces. Conventional PGIs appear to achieve their high specificity for Glc-6-P over Man-6-P (for isomerization) by preventing such a rotation through the presence of a glutamine (Gln-511 in rabbit PGI) that blocks the C-1-O-1 group as it rotates (7). In PaPGI/PMI, however, the equivalent residue to Gln-511 is Thr-291, and the smaller size of this residue leaves more room within this critical region of the active site. This, presumably, would permit a rotation about the C-2-C-3 bond of mannose 6-phosphate and lead to a change of configuration of the substrate at C-2. The same situation is also likely in other PGI/PMIs, in which the threonine is replaced by a valine or a leucine (13). A more detailed view of how such a rearrangement occurs within the active site of PaPGI/PMI, however, must await the determination of its structure in a complex with one or more of its substrates, Glc-6-P, Fru-6-P, and Man-6-P. Such a structure may also explain how eubacterial and eukaryotic PGIs can rotate the C-2-C-3 bond during the anomerization of Fru-6-P but at the same time prevent this during isomerization, which would form Man-6-P.