Substrate and Metal Complexes of 3-Deoxy-d-manno-octulosonate-8-phosphate Synthase from Aquifex aeolicus at 1.9-Å Resolution

3-Deoxy-d-manno-octulosonate-8-phosphate synthase (KDO8PS) from the hyperthermophilic bacteriumAquifex aeolicus differs from its Escherichia coli counterpart in the requirement of a divalent metal for activity (Duewel, H. S., and Woodard, R. W. (2000)J. Biol. Chem. 275, 22824–22831). Here we report the crystal structure of the A. aeolicus enzyme, which was determined by molecular replacement using E. coli KDO8PS as a model. The structures of the metal-free and Cd2+ forms of the enzyme were determined in the uncomplexed state and in complex with various combinations of phosphoenolpyruvate (PEP), arabinose 5-phosphate (A5P), and erythrose 4-phosphate (E4P). Like the E. coli enzyme, A. aeolicus KDO8PS is a homotetramer containing four distinct active sites at the interface between subunits. The active site cavity is open in the substrate-free enzyme or when either A5P alone or PEP alone binds, and becomes isolated from the aqueous phase when both PEP and A5P (or E4P) bind together. In the presence of metal, the enzyme is asymmetric and appears to alternate catalysis between the active sites located on one face of the tetramer and those located on the other face. In the absence of metal, the asymmetry is lost. Details of the active site that may be important for catalysis are visible at the high resolution achieved in these structures. Most notably, the shape of the PEP-binding pocket forces PEP to assume a distorted geometry at C-2, which might anticipate the conversion from sp 2 tosp 3 hybridization occurring during intermediate formation and which may modulate PEP reactivity toward A5P. Two water molecules are located in van der Waals contact with the siand re sides of C-2PEP, respectively. Abstraction of a proton from either of these water molecules by a protein group is expected to elicit a nucleophilic attack of the resulting hydroxide ion on the nearby C-2PEP, thus triggering the beginning of the catalytic cycle.

The recently determined structures of Escherichia coli KDO8PS and DAH7PS (8 -11) provide evidence that these two enzymes are not only mechanistically, but also structurally related, and that they probably originated from a common ancestor. However, although all the known bacterial and eukaryotic DAH7PSs require divalent metals for activity (12,13), two different groups or classes of KDO8PSs have been recognized with respect to their requirement for metals (14). The most extensively studied member of the KDO8PS family, the E. coli enzyme, does not require a metal. In contrast, KDO8PS from the hyperthermophile Aquifex aeolicus requires a divalent cation for activity, with the level of activation varying over a 10-fold range in the order Cd 2ϩ ϳ Mn 2ϩ Ͼ Ni 2ϩ ϳ Co 2ϩ Ͼ Ca 2ϩ Ͼ Cu 2ϩ Ͼ Mg 2ϩ ϳ Zn 2ϩ (15).
To understand the basis for the different metal requirement between these two enzymes, we have determined the crystal * This work was supported in part by United States Public Health Service Grants AI42868 (to D. L. G.) and GM53069 (to R. W. W.) and by a donation in memory of Michael Cooperman. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The atomic coordinates and structure factors ( structure of A. aeolicus KDO8PS in the presence of various combinations of substrates and metal. Results from this study support the previous hypothesis (11) that the reaction of KDO8P synthesis proceeds through the formation of a linear intermediate and point to the essential role played by a surface loop in isolating the active site from bulk solvent and by the geometry of the active site in influencing the electronic configuration of the substrates during turnover.

EXPERIMENTAL PROCEDURES
Protein Purification and Crystallization-Recombinant A. aeolicus KDO8PS was isolated from E. coli BL21(DE3) cells harboring plasmid pAakdsA, as described previously (16). A concentrated KDO8PS solution (30 mg/ml in ϳ15 mM Tris-Cl) was obtained from purified enzyme that had been dialyzed against 5 mM Tris-Cl (pH 7.5), lyophilized, and then reconstituted with distilled water. Lyophilization does not affect the activity of the reconstituted enzyme. Crystals were obtained by vapor diffusion in hanging drops. For this purpose, the concentrated enzyme solution (30 mg/ml) was mixed 1:1 with a reservoir solution containing 100 mM sodium acetate (pH 4.8 -4.9) and 5-6% polyethylene glycol 4000. The crystallization trays were incubated at 4°C for 1-2 weeks. Crystals (typically 0.12-0.15 mm 3 ) were harvested and maintained in a cryoprotectant holding solution consisting of 100 mM sodium acetate (pH 4.8 -4.9), 21% polyethylene glycol 4000, 16% glycerol, and 5% ethylene glycol.
The concentrated solution of A. aeolicus KDO8PS as well as crystals of this enzyme in the original crystallization drops have a slight pink color. As discussed by Duewel et al. (15,16), this color is probably due to the presence of substoichiometric amounts of bound iron in the enzyme produced in E. coli. For these studies, we have opted to collect diffraction data from crystals that were either completely depleted of any metal or in which the iron had been replaced with Cd 2ϩ , the ion that confers the highest activity in solution (15). Incubation of A. aeolicus KDO8PS crystals in a cryoprotectant liquor containing 100 M CdCl 2 , with several daily changes of this solution, bleaches their color completely within 2 days. We attribute this color change to the replacement of iron by Cd 2ϩ . For experiments in which it was necessary to remove any residual traces of metal from the enzyme, crystals were incubated for several days in a cryoprotectant holding solution containing 10 mM EDTA, with frequent changes of the solution. Crystals incubated with EDTA are also completely colorless.
Structure Determination and Refinement-Crystals of A. aeolicus KDO8PS belong to space group P3 1 21 with unit cell dimensions a ϭ b Х 84.3 Å and c Х 159.5 Å and diffract to ϳ1.9 Å. Data sets were collected at 100 K with an R-axis IV image plate detector or with a Bruker Hi-Star detector at the CuK ␣ wavelength (see Table I). Oscillation data were processed with HKL (17) or SAINT (Brucker, Inc.). The structure of the A. aeolicus enzyme was determined by molecular replacement using the structure of one monomer of the tetrameric E. coli enzyme (11) as the starting model. All the steps of the molecular replacement procedure were carried out using the CNS Version 1.0 suite of crystallographic programs (18). Self-rotation and cross-rotation function analyses clearly identified the presence of two subunits in the asymmetric unit, related by a 2-fold axis of local symmetry. Correct placing of the subunits was obtained with two consecutive translation searches. Model refinement was also carried out with CNS Version 1.0 using cross-validated maximum likelihood as the target function (19). Solvent molecules were added during the final stages of refinement after the protein model had stabilized.

RESULTS
Overall Structure of A. aeolicus KDO8PS-The asymmetric unit of A. aeolicus KDO8PS crystals contains a homodimer. Two crystal forms of E. coli KDO8PS previously reported by this laboratory contain instead a homotetramer of the enzyme with 222 local symmetry (9,11). A third crystal form reported by Wagner et al. (10) contains only one chain in the asymmetric unit, but application of crystal symmetry generates a tetramer essentially identical to that present in the other two crystal forms. Application of one of the operations of crystallographic symmetry (x ϭ y, y ϭ x, z ϭ Ϫz) of space group P3 1 21 produces a tetrameric assembly of A. aeolicus KDO8PS that can be superimposed on the E. coli enzyme with a root mean square deviation of only 1.22 Å for 696 aligned C-␣ atoms. In line with this finding, comparative analysis of the A. aeolicus (M r 29,734) and E. coli (M r 30,833) KDO8PSs by analytical size-exclusion chromatography reveals that the two enzymes have almost identical elution profiles (16). The enzyme from Salmonella typhimurium also demonstrates analogous chromatographic behavior (20). These observations suggest that the native form of KDO8PS might be a tetramer in all Gram-negative bacteria.
As in the case of E. coli KDO8PS, the A. aeolicus enzyme adopts a (␤/␣) 8  the N terminus, as the E. coli enzyme has an additional ␤-hairpin that seals the N-terminal end of the barrel. Consequently, the first residue of the thermophilic KDO8PS corresponds to residue 16 of the E. coli sequence. The enzyme active sites are located at the C-terminal end of the ␤-barrel of each subunit, at the interface with the adjacent subunit. Three long loops, L2, L7, and L8, control access to the active site cavity (Fig. 2). L7 and L8 were both disordered in the E. coli enzyme (11). L8 is always ordered in A. aeolicus KDO8PS; L7 becomes ordered under particular conditions (see below).
Structure of the Substrate-free Enzyme-The structure of substrate-free A. aeolicus KDO8PS was determined for the both the Cd 2ϩ and metal-free forms of the enzyme. The Cd 2ϩ form (Table I, Cd 2ϩ column) was intentionally pursued because Cd 2ϩ is the most effective activator of KDO8PS (15). In the active site of the enzyme, the Cd 2ϩ ion displays a distorted octahedral coordination: the thiolate of Cys-11 and the ⑀-nitrogen of His-185 provide two axial ligands, whereas the carboxylate moieties of Glu-222 and Asp-233 and a water molecule provide the equatorial coordination (Fig. 3). The water molecule that acts as equatorial ligand is also stabilized by a hydrogen bond to the -nitrogen of Lys-46. These features of the active site of A. aeolicus KDO8PS are particularly reminiscent of the homologous enzyme DAH7PS. The recently reported crystal structure of the E. coli phenylalanine-regulated DAH7PS (8) shows a lead ion (Pb 2ϩ ) bound in the active site and interacting with PEP. The lead ion, which was included in the crystallization for the purpose of determining the structure by multiple wavelength anomalous diffraction, but is itself an extremely poor activator of DAH7PS (8), is believed to occupy the same active site location as other metals that are good activators (13). As expected, several residues in the coordination sphere of the lead ion in DAH7PS (Cys-61, His-268, Glu-302, and Asp-326) have counterparts in the active site of A. aeolicus KDO8PS (Cys-11, His-185, Glu-222, and Asp-233).
In the substrate-free Cd 2ϩ form of A. aeolicus KDO8PS, a phosphate ion (herein designated as PO 4 -1) occupies the position corresponding to the phosphate moiety of PEP in DAH7PS (8) or to the SO 4 -1 site described in the E. coli enzyme (11) (Fig.  3). A second phosphate ion (herein designated as PO 4 -2), corresponding to the SO 4 -2 ion of the E. coli enzyme, is located ϳ10 Å from PO 4 -1 in a raised position close to the opening of the barrel and beneath the long loops L2 (residues 44 -58) and L7 (residues 187-201) (Figs. 2 and 3). Some broken density associated with L7 is visible in one of the two subunits contained in the asymmetric unit; L7 of the other subunit is completely disordered.
The structure of the substrate-and metal-free enzyme was obtained by extensive incubation of the crystals in the presence of 10 mM EDTA (Table I, -column). This structure is essentially identical to that of the Cd 2ϩ enzyme, suggesting that the presence of the metal does not alter appreciably the overall conformation of the enzyme. However, small differences are present in the positions of the side chains of the residues directly involved in metal coordination. In particular, the water molecule that was one of the ligands of Cd 2ϩ is no longer visible. L8, which provides the metal ligand Asp-233, moves slightly away, and a different rotamer is present in the side chain of Gln-188 (beginning of L7), such that the carboxylate and amide moieties of these 2 residues are now in hydrogen bond distance (Fig. 3). One of the largest changes is observed in the position of the imidazole ring of His-185, which undergoes a rotation of ϳ40°around the C-␤-C-␥ bond. Although the position of PO 4 -1 is essentially unchanged, PO 4 -2 is not present, possibly as a consequence of the lack of stabilizing interactions provided by L7. This loop is completely disordered in both subunits of the metal-free enzyme between Val-187 and Met-200.
Structure of the Enzyme in Complex with PEP-Crystals of the metal-free or Cd 2ϩ enzyme were incubated in the presence of 5 mM PEP (K m(PEP) ϭ 40 M at 60°C), and their structures were determined (Table I,  enzyme both in the presence and absence of a divalent metal. The most notable difference between the two structures is in the position of the Glu-222 side chain. If Cd 2ϩ is present, Glu-222 is a metal ligand, whereas in the absence of the metal, its side chain moves slightly away from the vacant metal site (data not shown). Interestingly, the water molecule that, in the presence of Cd 2ϩ , is one of the metal ligands is retained even in the absence of Cd 2ϩ . This water is located on the si face of PEP in van der Waals contact with the substrate C-2 and C-3. The presence of PEP must have a significant stabilizing role, as this water molecule is not visible in the structure of the substrateand metal-free enzyme. A water molecule is also in van der Waals contact with the re side of PEP (Fig. 4A). The potential role in catalysis of the water molecules located on both the si and re sides of PEP is discussed below.
A striking feature of PEP bound in the active site of KDO8PS is that the geometry of the molecule at C-2 deviates appreciably from the planarity observed in the crystal structures of free PEP in various ionization states (reviewed in Ref. 21): the magnitude of the distortion is provided by the value of the angle between the plane defined by C-2, C-1, and C-3 and the plane defined by C-1, C-3, and O-2, which is ϳ12°instead of the expected value of 0° (Fig. 4B). This deviation of the PEP molecule from perfect planarity appears to be imposed by the particular geometry of the active site and may be indicative of the fact that the enzyme modifies the electronic configuration of the substrate to favor a nucleophilic attack at C-2 PEP (see below).
Structure of the Enzyme in Complex with A5P-The structure of A. aeolicus KDO8PS in complex with A5P was determined by soaking crystals of the Cd 2ϩ enzyme in the presence of 10 mM A5P (K m(A5P) ϭ 8 M at 60°C) ( Table I, Cd 2ϩ /A5P column). As predicted by Radaev et al. (11), A5P binds with its phosphate moiety coincident with the position occupied by PO 4 -2 in the substrate-free enzyme (see above). The position of PO 4 -1 is unchanged with respect to the substrate-free enzyme. The phosphate moiety of A5P is stabilized by a salt bridge to Arg-49 and by hydrogen bonds to the backbone and side chain of Ser-50 (Fig. 5A). In this structure, L7 is not ordered and therefore does not appear to contribute significantly to A5P binding. The carbon tail of A5P is stabilized by a network of hydrogen bonds directed to the substrate hydroxyls and to the aldehyde moiety, originating from Asp-233, Asn-48, and PO 4 -1 and from several water molecules (Fig. 5, A and B). In particular, C-2-OH A5P is only 2.9 Å from Cd 2ϩ and can be considered a weak ligand of the metal. In this fashion, the substrate C-2-OH appears to replace the water molecule that is a ligand of Cd 2ϩ in other structures of the enzyme. Polar interactions of A5P with the active site residues are listed in Table II. A5P binds only to one of the two molecules of KDO8PS present in the asymmetric unit. The active site of the other molecule is filled by PO 4 -1 and PO 4 -2. Moreover, if crystallographic symmetry is applied to generate the enzyme tetramer, it becomes apparent that the two active sites in which A5P binds are located on the same "face" of the enzyme. This observation is reminiscent of the asymmetry observed in the E. coli enzyme with regard to the binding of phosphate ions at a surface site (11).
Structure of the Enzyme in Complex with PEP plus A5P-A. aeolicus KDO8PS has optimal activity at 95°C and no detectable activity at 4°C (16). As crystals of this enzyme are obtained and maintained at 4°C, it was deemed possible to bind both substrates PEP and A5P without turnover. Both metalfree and Cd 2ϩ crystals of A. aeolicus KDO8PS were incubated in the presence of 5 mM PEP and 10 mM A5P, and the structure of the enzyme was determined under these conditions (Table I, PEP/A5P and Cd 2ϩ /PEP/A5P columns, respectively). The two substrates bind in the active site at the same positions they occupy in the structures of the enzyme in complex with PEP alone or A5P alone, respectively. However, in the Cd 2ϩ enzyme, A5P binds to only one of the two active sites contained in the asymmetric unit (as observed in the structure of the Cd 2ϩ enzyme incubated with A5P alone), and PO 4 -2 fills the second active site. In the metal-free enzyme, A5P binds to both active sites (Fig. 6). In all the cases in which A5P and PEP bind simultaneously, L7 becomes ordered and isolates the active site from the external environment. This loop stabilizes the phosphate moiety of A5P via hydrogen bonds to the backbone and side chain of Ser-197 (Fig. 6, A and C). Interestingly, the electron density associated with L7 is better ordered in the metal-free enzyme than in the Cd 2ϩ enzyme. Likewise, the electron density of A5P is not well defined in the Cd 2ϩ enzyme, such that only the positions of the phosphate moiety and the bridging oxygen can be identified unambiguously (data not shown). This suggests that, in the Cd 2ϩ enzyme, some residual activity may be present even at 4°C, such that a higher degree of motion characterizes the chemical species possibly involved in catalysis. In the metal-free enzyme, the density of A5P is reasonably well defined in one of the two active sites contained in the asymmetric unit and shows that this substrate presents the re side of its aldehyde moiety to the si side of C-2 PEP (Fig.  6, C and D), in agreement with previous stereochemical studies (4 -6). In this site, there is no electron density at the position usually occupied by the water molecule that coordinates Cd 2ϩ .

TABLE II
Polar interactions between active site residues and PEP or A5P Cd 2ϩ enzyme/A5P column, distances of A5P (in parentheses in Å) from protein atoms in the structure of the Cd 2ϩ enzyme in complex with A5P (see Fig. 5); Metal-free enzyme columns, distances of PEP and A5P (in parentheses in Å) from protein atoms in the structure of the metalfree enzyme in complex with both substrates. These distances were measured in the active site in which the A5P density is best defined (see Fig. 6, C and D). PEP atoms are named as in Fig. 4B. A5P atoms are named sequentially from the aldehyde oxygen (O-1) in the direction of the bridging oxygen (OP-4).

Cd 2ϩ enzyme/ A5P
Metal-free enzyme PEP A5P In the other active site, the water molecule is present, but the density of A5P is not continuous between C-2 and C-4 (Fig. 6, A  and B). Polar interactions of PEP and A5P with the active site residues are listed in Table II. A ͉F o Ϫ F c ͉ difference map of the active sites in the metal-free enzyme shows clearly three small positive densities surrounding the sulfur of Cys-11. The most plausible interpretation of these densities is that, during the course of the prolonged incubation of the enzyme first with EDTA to remove any traces of metal and then with PEP and A5P, the sulfhydryl group (-SH) of Cys-11 was oxidized to a sulfonic group (-SO 3 ). This interpretation is supported by other studies carried out with the soluble enzyme in which it was shown that, in the absence of metal, Cys-11 becomes prone to oxidation. 2 Structure of the Enzyme in Complex with PEP plus E4P-It has been previously reported that E4P, the substrate of DAH7PS, does not serve as a substrate for A. aeolicus KDO8PS (15,16). Furthermore, in the E. coli enzyme, E4P acts as a competitive inhibitor with respect to A5P. 2 Therefore, in view of the mechanistic and structural similarities between KDO8PS and DAH7PS, it might be expected that E4P binds in the active site of KDO8PS in the same position as A5P, but that such binding does not lead to a condensation product with PEP because E4P is one carbon shorter than A5P. To verify this hypothesis, crystals of A. aeolicus Cd 2ϩ KDO8PS were incubated in the presence of 5 mM PEP and 10 mM E4P, and the structure of the enzyme was determined (Table I, Cd 2ϩ /PEP/ E4P column). Under these conditions, PEP binds to both subunits, but E4P binds only to one subunit. If a tetramer is generated by application of crystal symmetry, binding of E4P takes place in the two active sites that are visible from the same side of the enzyme. In these sites, the phosphate moiety of E4P occupies the same position assumed by the phosphate moiety of A5P. The electron density for the bridging oxygen and C-4 are well defined; however, unambiguous assignment of the configuration assumed by the remaining part of E4P in the active site is not possible. A plausible interpretation of the electron density is that the aldehyde carbonyl is hydrogenbonded to the water molecule coordinating Cd 2ϩ and to one of the hydroxyls of the phosphate moiety of PEP; C-3-OH E4P and C-2-OH E4P are also in hydrogen bond distance from the hydroxyl of Ser-197 and from the amide moiety of Asn-48, respectively (Fig. 7). As in the case of A5P and PEP, L7 becomes well ordered when E4P and PEP are both bound and isolates the active site from the external environment. DISCUSSION KDO8PS and DAH7PS catalyze similar condensation reactions between PEP and A5P and between PEP and E4P, respectively. Two mechanisms have been proposed to explain these reactions. According to the first hypothesis (22) a water molecule attacks at C-2 PEP , whereas C-3 PEP is added to the aldehyde of A5P or E4P; this process would yield a linear intermediate (Fig. 8, MECHANISM I). According to the second hypothesis (5,23), condensation of C-3 PEP with the carbonyl carbon of A5P or E4P is concurrent with an attack by C-3-OH of the monosaccharide on C-2 PEP ; this process would lead to the formation of a cyclic intermediate (Fig. 8, MECHANISM II). We speculated that two sulfate ions bound in the active site of E. coli KDO8PS occupy the positions of the phosphate moieties 2 Ronald W. Woodard, unpublished data.
FIG. 6. Binding sites of PEP and A5P. A, relative positions of PEP and A5P inside one of the two active sites contained in the asymmetric unit. Color coding is as described in the legend to Fig. 4. B, SigmaA-weighted omit map of PEP and A5P contoured at 1 with the section of the refined model already shown in A. The electron density of A5P is fragmented between C-2 and C-4. C, model for the binding of PEP and A5P in the other active site of the enzyme contained in the asymmetric unit. Notice the absence of the water (WAT) molecule that, in the metal form of the enzyme, ligates Cd 2ϩ . D, SigmaA-weighted omit map of PEP and A5P contoured at 1 with the section of the refined model already shown in C.
of PEP and A5P and posited that the distance of 13 Å between these positions is consistent with a mechanism that proceeds through the formation of a linear intermediate (11). However, direct experimental confirmation of this hypothesis was not possible since attempts to visualize the substrates PEP and A5P by incubating crystals of E. coli KDO8PS in the presence of these compounds were hampered by the high ionic strength of the holding solution, which prevented binding of either substrate. The crystal structure of DAH7PS, although providing the location for PEP in the active site, did not contain E4P, hence also could not provide direct experimental confirmation of our hypothesis.
In contrast, crystals of the A. aeolicus enzyme are stable in a holding solution of low ionic strength, which allowed the determination of the structure of the enzyme both in its substratefree form and in the presence of bound PEP and A5P (or E4P). The distance between the phosphate moieties of PEP and A5P (or E4P) in A. aeolicus KDO8PS is only 10 Å, which is the same distance observed in the structure of the E. coli DAH7PS (8) between the phosphate moiety of PEP and a lone sulfate (or phosphate) ion believed to occupy the position of the phosphate group of E4P. A distance of 10 Å between two sulfate ions bound in the active site has been reported also by Wagner et al. (10) for a crystal form of E. coli KDO8PS in which only one chain is present in the asymmetric unit. We had originally postulated that the distance of 13 Å between the two sulfate ions bound in the active site of E. coli KDO8PS (versus 10 Å in DAH7PS) reflected the fact that A5P, the substrate of KDO8PS, is one carbon longer than E4P, the substrate of DAH7PS. This element of our hypothesis is clearly not supported by the observation that, in A. aeolicus KDO8PS, both A5P and E4P are accommodated in the active site without a change in the distance between the phosphate moieties of these substrates and that of PEP. The experimentally observed positions of PEP and A5P (Fig. 6) or of PEP and E4P (Fig. 7) are, however, clearly consistent with the formation of a linear reaction intermediate because the distance between C-2 PEP and C-3-OH A5P or between C-2 PEP and C-2-OH E4P , is Ͼ5 Å. If the reaction took place via formation of a cyclic intermediate, these two atoms would have to be much closer than 5 Å for a bond to be formed between them, without invoking large conformational changes in the active site.
A. aeolicus KDO8PS, like all known DAH7PSs, requires a divalent cation for activity. Moreover, the coordination of the metal in the A. aeolicus enzyme is very similar to that observed for the metal present in the structure of E. coli DAH7PS. What is the function of the metal in these enzymes? One possibility is that it serves a structural function such as that of correctly orienting PEP, A5P (or E4P), or the reaction intermediate in the active site. With regard to this point, we have not observed significant changes in the position of PEP in A. aeolicus KDO8PS in the absence or presence of metal. However, changes in the position of PEP might occur during catalysis. For example, in the crystal structure of the E. coli DAH7PS in complex with Mn 2ϩ and the PEP analog 2-phosphoglycolate, the octahedral coordination of the metal is completed by the carboxylate moiety of 2-phosphoglycolate, which substitutes for PEP in the active site (24). The binding mode of 2-phosphoglycolate might reflect the fact that a direct coordination of PEP with the active site metal occurs during turnover. Likewise, a direct interaction between A5P and Cd 2ϩ might occur at some stage during catalysis. An example of this kind of interaction was observed in the structure of the Cd 2ϩ enzyme in complex with A5P (Fig. 5). Thus, two different conformations of A5P have been observed in the various complexed states of A. aeolicus KDO8PS. In one conformation, a water molecules coor-dinates Cd 2ϩ and is also in hydrogen bond distance from both C-2-OH A5P and C-3-OH A5P (Fig. 6). In the second conformation, C-2-OH A5P replaces water as a ligand of Cd 2ϩ (Fig. 5). Thus, although the active site holds PEP very tightly, it allows significant conformational flexibility to A5P. It is unlikely that such an active site would have been retained during evolution without a specific reason. An interesting possibility is that the two conformations of A5P reflect the affinity of the enzyme for A5P (or for the chemical groups originating from A5P) at different stages of catalysis. At the start of the reaction, water may be necessary as a simultaneous ligand of the metal and A5P to favor the formation of a bond between PEP and A5P (see below). Once the reaction intermediate is formed, then that part of the molecule that originated from A5P could be stabilized via a direct interaction with the metal. The requirement for water at the beginning of the reaction may be rationalized by noticing that the postulated mechanism of formation of a linear intermediate predicts that a water molecule attacks C-2 PEP (Fig. 8, MECHANISM I) and that this attack would be favored by abstraction of a proton from this water by a base. The water molecule that acts as one of the Cd 2ϩ ligands in A. aeolicus KDO8PS is in van der Waals contact with the si face of C-2 PEP . A water molecule is also one of the equatorial ligands of the metal in the structure of DAH7PS in complex with Mn 2ϩ and 2-phosphoglycolate (24). Coordination of water to a divalent cation is expected to lower its pK a and to favor its deprotonation to a hydroxide ion. However, if the function of the metal in A. aeolicus KDO8PS (and in DAH7PS) is that of activating a catalytic water, how can E. coli KDO8PS catalyze the same reaction without metal? In this context, it is probably important that, in the E. coli enzyme, a water molecule is present in almost the same position as the water coordinating Cd 2ϩ in the A. aeolicus enzyme. In the E. coli enzyme, this water is stabilized by hydrogen bonds to two bases, the -nitrogen of Lys-60 and the ⑀-nitrogen of His-202, that might favor its deprotonation.
If the water molecule that coordinates Cd 2ϩ is involved in FIG. 7. Binding sites of PEP and E4P. SigmaA-weighted omit map of PEP and E4P contoured at 1 with a section of the refined model. The electron density of E4P is fragmented at the C-1-C-2 bond (see "Results" for details). Color coding is as described in the legend to Fig. 4. Hydrogen bonds (O-1 E4P to a water, O-1 E4P to PEP, C-2-OH E4P to Asn-48, and C-3-OH E4P to Ser-197) are shown as transparent rods. WAT, water. attacking the si face of C-2 PEP , then the formation of the linear intermediate would be the result of a syn addition of water to C-2 PEP and of C-1 A5P to C-3 PEP . However, a syn addition to PEP is unprecedented. Therefore, it is worthwhile to examine the alternative possibility of a water attack on the re side of C-2 PEP . Such an event might indeed occur in A. aeolicus KDO8PS because a water molecule is also present on the re side of PEP in van der Waals contact with C-2 PEP (Figs. 4, 6, and 7). This water is stabilized by a network of hydrogen bonds that includes one of the hydroxyls of the phosphate moiety of PEP, Asp-81, and His-83 (Fig. 9). A chain of hydrogen bonds could be involved in transferring a proton from this water molecule via Asp-81 to His-83, which is a more likely final acceptor than the aspartic acid. Asp-81 and His-83 are conserved in all known KDO8PSs, and a glutamic acid is present at the equivalent position of Asp-81 in DAH7PSs (see multiple alignment of Radaev et al. (11)). However, a residue equivalent to His-81 that could act as a final accepting base is not present in DAH7PSs. On the other hand, the electronic environment of the active site in DAH7PS might raise the pK a of the glutamic acid carboxylate sufficiently for it to act as a base. Therefore, as water molecules potentially capable of attacking C-2 are located on both the si and re sides of PEP, a direct discrimination between a syn and an anti attack on PEP is not possible at this time.
An additional observation allowed by the high resolution of the structures reported in this study deserves special attention. In all the cases in which PEP is bound in the active site of A. aeolicus KDO8PS, it appears to be distorted to optimally fit in the binding pocket. In particular, the geometry of C-2 PEP is intermediate between trigonal and tetrahedral (Fig. 4B). This distortion might be preparing C-2 PEP for the conversion from sp 2 to sp 3 hybridization that will take place during the formation of the linear intermediate. Although the direction of the distortion is such that C-2 bulges out of the si face (Fig. 4B), it is difficult to predict on theoretical grounds whether this would favor a water attack from the si side or from the re side. An additional effect of the distortion is that the dihedral angle between the C-2ϭC-3 double bond and the CϭO double bond of the carboxylate moiety of PEP is different from 0°. It has been theorized that the relative orientations of the two planes in which the CϭC and CϭO double bonds reside control the reactivity of C-3 PEP toward C-1 A5P (25), and ab initio calculations indicate that such reactivity should be maximal when the angle between the two planes is 90°. However, nothing is known yet about how solvation effects and the active site mi-croenvironment might change the intrinsic properties and reactivity of PEP as derived from quantum chemistry calculations. Undoubtedly, additional theoretical and experimental work will be necessary to clarify these points.
One of the key observations resulting from our study is that binding of substrates controls the opening or closing of the active site to the surrounding environment. When both A5P (or E4P) and PEP are bound simultaneously, L7 is well ordered and isolates the active site from bulk solvent. If A5P alone or PEP alone binds, the loop is not ordered, suggesting that it might be alternating between an open and a closed conformation. Thus, A. aeolicus KDO8PS is structured in such a way that its active site remains open until both substrates are bound sequentially. This is particularly important in view of the fact that the active site is shaped like a funnel, with PEP bound at the bottom and A5P bound on top of PEP in proximity of the active site opening. This observation is consistent with earlier kinetic studies indicating that the reaction of KDO8P synthesis is an ordered process in which the binding of PEP precedes the binding of A5P (26). It is also important that the FIG. 8. Proposed mechanisms of KDO8P synthesis. MECHANISM I, a linear tetrahedral intermediate is formed after a water molecule attacks C-2 PEP and C-3 PEP is added to C-1 A5P . MECHANISM II, a cyclic intermediate is formed via nucleophilic attack by C-3-OH A5P on C-2 PEP followed by condensation of C-3 PEP with the carbonyl carbon of A5P. Although both mechanisms are shown as syn addition, the exact mode of addition is not known.
FIG. 9. Solvent localization on the si and re sides of PEP. PEP bound in the active site of KDO8PS is viewed edge-on, with its si face pointing up and to the left and its re face pointing down and to the right. The water (WAT) molecule located on the re side of C-2 PEP could be activated by transfer of a proton to a hydrogen bond chain ending with His-83. active site becomes isolated from bulk solvent during catalysis. This may be needed to create a microenvironment in which the pK a values of catalytic waters and specific amino acid groups are different from the values normally expected in bulk phase equilibria. In view of all of these mechanistic requirements, there must exist a way by which information on the occupancy of the two substrates in the active site controls the conformation of L7. In this context, it may be of relevance that His-185, located at the very beginning of L7, is one of the Cd 2ϩ ligands. The ⑀-nitrogen of His-185 is between 3.3 and 3.6 Å from C-2-OH A5P (in the two conformations observed for this substrate) and is also in hydrogen bond distance of the water molecule sitting on the si side of PEP. Also at the beginning of L7, on the carboxyl side of His-185, Gln-188 is hydrogen-bonded to C-4 -OH A5P (in the conformation of A5P observed in the metal-free enzyme with PEP also bound). Thus, it is possible that His-185 and Gln-188 sense disturbances in the delicate network of hydrogen bonds and hydrophobic interactions surrounding the substrates and act as transducers of information from the active site to L7.
Of particular mechanistic interest is the observation that, in the Cd 2ϩ enzyme, A5P (or E4P) binds only to one of the two active sites of the dimer contained in the asymmetric unit of the crystal. Application of crystal symmetry to generate the tetramer reveals that binding of A5P occurs at the active sites located on only one of the two faces of the enzyme. When A5P binds simultaneously with PEP, the associated ordering of L7 also takes place only on one face of the enzyme (Fig. 10). A similar asymmetry was also observed in the binding of sulfate ions to the surface of the E. coli enzyme (11,24). As structures of KDO8PS in complex with A5P (or E4P) were obtained by soaking crystals in the presence of this substrate, one would expect an equal binding to the two faces of the enzyme, if the faces were equivalent. The observation that only one face binds A5P (although both are equally exposed to solvent) strongly suggests that the asymmetry is preexistent to the addition of the substrates and that turnover cannot take place in both faces simultaneously. Thus, at any point in time, the entire population of the enzyme is committed to bind A5P only on one of its faces: this situation translates into a structural asymmetry that is always present in the crystal and, we would argue, also in the soluble enzyme, but becomes clearly manifest only upon the addition of A5P. We believe that this hypothesis is consistent with the recent determination of pre-steady-state burst of phosphate release in the reaction of KDO8P synthesis catalyzed by the E. coli enzyme, which indicated that only ϳ50% of the active sites contribute to the burst (27). This observation was originally interpreted as an indication of partial inactivation of the enzyme during the course of the study.
A key role in establishing the enzyme asymmetry is played by the metal. If Cd 2ϩ is present, a change in conformation appears to originate in the active sites located on one face of the enzyme, which prevents binding of A5P in the active sites located on the opposite face (see structure of the Cd 2ϩ enzyme in complex with PEP and A5P or E4P). However, in the absence of metal, information on the occupancy of the active sites of one face is no longer transferred to the other face. The consequence of this lack of communication is that the active sites on both faces of the enzyme act independently and bind A5P simultaneously (see structure of the metal-free enzyme in complex with both PEP and A5P). Thus, altogether, the crystallographic studies of KDO8PS point to the existence of a mechanism of alternating site (or face) catalysis that has been overlooked by previous kinetic studies. FIG. 10. Molecular surface of tetrameric Cd 2؉ KDO8PS with bound PEP and A5P. The enzyme surface is colored white, with the exception of the three loops that control access to the active site cavities, which are colored chartreuse (L2), magenta (L8), and blue (L7) (see also Fig. 2). A, face of the enzyme containing the active sites in which both PEP and A5P are bound simultaneously. L7 seals the active site from the bulk phase. B, face of the enzyme containing the active sites in which only PEP binds. L7 is disordered, and the active site cavity is visible. This figure was generated with Grasp (32).