A common structural basis for the inhibition of ribulose 1,5-bisphosphate carboxylase by 4-carboxyarabinitol 1,5-bisphosphate and xylulose 1,5-bisphosphate.

Ribulose 1,5-bisphosphate carboxylase/oxygenase (Rubisco) catalyzes the carboxylation of ribulose 1,5-bisphosphate. The reaction catalyzed by Rubisco involves several steps, some of which can occur as partial reactions, forming intermediates that can be isolated. Analogues of these intermediates are potent inhibitors of the enzyme. We have studied the interactions with the enzyme of two inhibitors, xylulose 1,5-bisphosphate and 4-carboxyarabinitol 1,5-bisphosphate, by x-ray crystallography. Crystals of the complexes were formed by cocrystallization under activating conditions. In addition, 4-carboxyarabinitol 1,5-bisphosphate was soaked into preformed activated crystals of the enzyme. The result of these experiments was the release of the activating CO2 molecule as well as the metal ion from the active site when the inhibitors bound to the enzyme. Comparison with the structure of an activated complex of the enzyme indicates that the structural basis for the release of the activator groups is a distortion of the metal binding site due to the different geometry of the C-3 hydroxyl of the inhibitors. Both inhibitors induce closure of active site loops despite the inactivated state of the enzyme. Xylulose 1,5-bisphosphate binds in a hydrated form at the active site.

Both carboxylation and oxygenation reactions are dependent on prior activation of the enzyme. Activation involves the reversible derivatization of a lysine residue by non-substrate CO 2 in the active site to form a labile carbamate that is subsequently stabilized by a magnesium ion (12,13).
In the physiological reaction, substrate CO 2 reacts with the enediol to form a 6-carbon intermediate, 2-carboxy-3-keto-Darabinitol 1,5-bisphosphate (compound III) or its hydrated gemdiol form (compound IV). This intermediate is sufficiently stable that it can be isolated by acid quenching of the enzyme (14). Two stereoisomers of an analogue of this intermediate, 2-carboxy-D-arabinitol 1,5-bisphosphate (2-CABP) and 4-carboxy-D-arabinitol 1,5-bisphosphate (4-CABP, Fig. 4) are exceptionally tight binding and virtually irreversible inhibitors of spinach Rubisco. The overall dissociation constants for 2-CABP and 4-CABP are 1.9 ϫ 10 Ϫ13 and 2.8 ϫ 10 Ϫ11 M, respectively (15). Because cleavage of the gemdiol form of the 6-carbon intermediate is assumed to involve the abstraction of a proton from one of the hydroxyls, comparative affinities of 2-CABP and 4-CABP could reflect differences in the interaction of the enzyme with the hydroxyls of the intermediate that ultimately leads to cleavage (15).
We present here the structure of two inhibitor complexes with Rubisco from spinach, one with 4-CABP and the other with XuBP. The complex with XuBP, the C-3 epimer of the substrate RuBP, is meant to mimic the state of the enzyme before the entry of the gaseous substrates CO 2 or O 2 . 4-CABP is the C-3 epimer of 2-CABP with a close similarity to the predominant form of the 6-carbon intermediate, the gemdiol form of 2-carboxy-3-keto-D-arabinitol 1,5-bisphosphate (16). The structure of the 4-CABP complex therefore is expected to give information on this particular step in catalysis. XuBP and 4-CABP share the same configuration at the C-3 center (Fig. 4), a configuration different from the true substrate RuBP and intermediates resulting from it in the reaction on the enzyme. The structures could therefore give information on the stereochemical guidance of the reaction by the enzyme.

EXPERIMENTAL PROCEDURES
Both complexes were crystallized using spinach Rubisco purified as described earlier (17). XuBp and 4-CABP were obtained as a gift from G. Lorimer, DuPont.

4-CABP Complex
Cocrystallization Experiment-Crystals of Rubisco complexed with 4-CABP could be obtained by vapor diffusion from ammonium sulfate solutions under activating conditions in a way similar to procedures described for the 2-CABP complex (18). The crystallization mixture contained 100 mg/ml spinach Rubisco, 3 mM MgCl 2 , 25 mM NaHCO 3 , and 10 mM 4-CABP in 50 mM phosphate at pH 7.4. This represents a 5-fold increase in the concentrations of HCO 3 Ϫ and the inhibitor compared with crystallization of the complex of activated spinach Rubisco with 2-CABP. The concentration of MgCl 2 was kept at 3 mM to prevent excessive formation of salt crystals. For crystallization, this solution was mixed in a 1:1 ratio with a 1.7 M solution of ammonium sulfate in 50 mM phosphate buffer as in Andersson and Brä ndén (18). Crystals appeared after 1 week and were isomorphous with the crystals obtained from the activated complex with 2-CABP (24) and diffracted to approximately 1.5-Å resolution. Details of data collection, data reduction, and refinement are summarized in Tables I and II. Soaking Experiment-As an alternative to cocrystallization, 4-CABP was soaked into preactivated crystals according to published procedures (19). The activated carbamylated enzyme was stabilized in the crystal by the presence of the product 3-PGA. 3-PGA, which binds relatively weakly to the enzyme (K d in the millimolar range) (20), can then be replaced by a tighter binding ligand by soaking in the compound into the preformed crystals. 4-CABP was added to crystals of activated Rubisco at a concentration of 8 mM, with the concentrations of NaHCO 3 and MgCl 2 maintained at 50 and 10 mM, respectively, throughout the soak. Soaking was continued for about 4 days prior to data collection. The crystals are isomorphous to cocrystallized 4-CABP crystals and diffract to a resolution beyond 2.1 Å. Refinement of the various complexes was performed similarly as for the cocrystallized complex, but the level of agreement could never be brought under 0.30. To verify this unexpected result, the experiment was repeated a second time, but this time data collection was performed at 100 K where a complete data set can be collected using a single crystal. Details of data collection, data reduction, and refinement are summarized in Tables I and II.

XuBP Complex
Crystals of the complex with XuBP were grown by cocrystallization under activating conditions and vapor diffusion. An earlier report indicated that the XuBP complex might turn over at a slow rate if magnesium is used as an activator metal (8). Calcium that seems to form only non-productive complexes (21) was therefore used as a precaution. The complex was formed by incubating a 60 mg/ml protein solution buffered at pH 7.8 with 25 mM HEPES with 10 mM CaCl 2 , 50 mM NaHCO 3 , and approximately 20 mM XuBP for 10 min at room temperature. The protein solution was mixed in a 1:1 ratio with a well solution containing 15-16% polyethylene glycol 4000, 25 mM HEPES, pH 7.8, 0.2 M NaCl, 50 mM NaHCO 3 , and 10 mM CaCl 2 . Crystals grew overnight at 4°C, and final dimensions were 0.3 mm ϫ 0.6 mm ϫ 0.05 mm. Details of data collection, data reduction, and refinement are summarized in Tables I  and II. Phases were determined by molecular replacement using a combination of the CCP4 program package (22) and X-PLOR (23). The refined coordinates of the spinach Rubisco 2-CABP complex (Protein Data Bank entry 8ruc) (24) were used as a search model. For the rotation-function calculations using the program ALMN, the whole hexadecameric molecule was placed in an artificial P1 unit cell of dimensions a ϭ 220 Å, b ϭ 220 Å, and c ϭ 220 Å. Data between 10 and 5.5 Å were used, and the outer radius of Patterson integration was 35 Å. The resulting peaks were filtered by using the PC-refinement option in X-PLOR and then subjected to translation search using TSFGEN (CCP4, presently TFFC). The translation search gave one unambiguous solution that was 7.8 higher than the next peak in the list, and the initial model had an R-factor of 0.326 for reflections between 10 and 3 Å resolution (see Table  II for further details of refinement). The ligand XuBP was first introduced in the keto form at C-2, but it soon became evident that its C-2 diol better fit the electron density maps.

4-CABP Complex-
Although the crystals were obtained under presumed activating conditions, the resulting complex did not have density either for the carbamate on Lys-201 or for the activator magnesium ion ( Fig. 2A). Fig. 3 shows a comparison of the active sites in the 4-CABP and previously determined 2-CABP complexes. The inhibitor is bound to the enzyme such that the carboxylate group on the compound is in the same position as that on 2-CABP in complex with the activated enzyme. In effect the only difference is the stereochemistry around C-3 (Fig. 4). Except for the lack of activator carbamate and metal in the 4-CABP structure, the 4-CABP and 2-CABP structures are extremely similar overall, the C ␣ atoms superimposing with a root mean square (r.m.s.) deviation of 0.154 Å on all C ␣ atoms. Differences at the active site consist of relatively few side chain movements. His-294 has moved around such that it forms a contact with the terminal amino group of Lys-201. Glu-60 from the symmetry-related large subunit that reaches into the active site is also in a clearly different conformation and interacts with the hydroxyl group on C-4 of 4-CABP (the equivalent of the C-2 hydroxyl of 2-CABP). Phe-127 is rotated, a change that seems consequent to the movement of Glu-60 (19). The remaining residues are not significantly different between the 4-CABP and 2-CABP complexes.
Data from crystals where 4-CABP was soaked into preformed crystals of Rubisco containing the activator carbamate, the magnesium ion, and 3-PGA were similar in quality to the data from the cocrystallized 4-CABP crystals (Table I). Despite extensive trials at refining the complex, it was not possible to reduce the R-factor below 0.30. Electron density maps calcu- lated at this stage showed density for all protein atoms, the carbamate, the metal, and 4-CABP. Nevertheless, all attempts at further minimization failed with both data sets, suggesting that the crystals may contain a mixture of different species.
XuBP Complex-Both the overall structure and the conformation of active site side chains are extremely similar between the XuBP and 4-CABP complex. A superposition of the active sites in the XuBP and 4-CABP structures is shown in Fig. 5. C ␣ atoms superimpose with an r.m.s. deviation of 0.192 Å (C ␣ atoms for the XuBP complex superimpose with those of the 2-CABP complex with r.m.s. of 0.245 Å). In the XuBP complex, as with the 4-CABP complex, the inhibitor is bound to an active site devoid of the activator carbamoyl group and the metal ion (Fig. 2B). The XuBP molecule binds as a hydrated diol at C-2.
The sugar chain appears more puckered than in the case of either 2-CABP or 4-CABP since it has moved to fill the space occupied by the metal ion in the 2-CABP structure. The hydroxyls of the diol interact with Lys-334 and Glu-60 and with Asp-203 and Lys-175, respectively (Table III). Cocrystallization with XuBP and 4-CABP Results in Nonactivated Complexes of the Enzyme with the Inhibitors-Catalytic activity of Rubisco from all species requires the presence of the metal in the active site. Both inhibitor complexes described here were prepared under conditions that produce activated enzyme forms (19). What crystallized in both cases were, however, inactivated complexes with only the inhibitors bound in the active site. The two inhibitors induced closures of flexible loops around the active site, identical to those observed in the activated 2-CABP complex (loop six of the ␣/␤-barrel, a Cterminal segment, and a segment from the N-terminal region of the 2-fold related subunit), but the activating magnesium ion and the lysyl-bound CO 2 were lost, indicating that loop movements are independent of carbamylation or the presence of divalent metal (for further description of movements during Rubisco catalysis, see Ref. 19).

XuBP and 4-CABP Are C-3 Epimers of RuBP and 2-CABP-In
Inhibition by XuBP-Binding of XuBP could either occur by preferential binding of the inhibitor to the non-activated form

data collection and reduction for 4-CABP and XuBP complexes
For the 4-CABP complex, x-ray diffraction data to 2.3-Å resolution were collected from one crystal using a MARR imaging plate at station X31 at the EMBL outstation at DESY, Hamburg. Data were processed using the in-house modification of the MOSFLM data processing package and programs in CCP4 (22). The data set was 82.2% complete in the final resolution shell from 2.36 to 2.3 Å and I/I larger than 3 for all measurements. Data from crystals of activated Rubisco soaked with 4-CABP were collected in two separate experiments at station 9.6 of SRS Daresbury, UK on a MARR image plate detector. One data set was collected from three crystals at room temperature, and the second data set was collected from one crystal at 100 K under otherwise identical conditions. The data were processed with the DENZO data reduction suite (31) with similar results. Data set 1 was scaled and reduced with SCALEPACK (31) whereas SCALA (CCP4, Ref. 22) was used for the scaling of data set 2. For the XuBP complex, data from the XuBP complex were collected from 11 crystals at stations 9.5 and 9.6 at SRS, Daresbury in two stages. The first was collected on film from 4 crystals and the second on a MARR imaging plate from 7 crystals. Data were processed using DENZO. A total of 482,448 measurements were obtained from the films and were merged together with R ϭ 0.102 and a completeness of 61.3% between 20 and 2.3 Å. The image plate data provided a further 1,036,520 measurements merging with R ϭ 0.163 and completeness 58% between 20 and 2.3 Å. Both data sets were then merged in SCALEPACK (R merge 0.147) to provide a data set of 217,862 unique reflections 62.5% complete to 2.3 Å (90% complete to 2.8 Å). 4

TABLE II Statistics for refinement for the 4-CABP and XuBP complexes
Refinement was performed using the program X-PLOR (23). The Engh and Huber force field (32) was used in all refinement steps and the behavior of R free (33) was monitored throughout. Five percent of the data in both cases were kept aside for the R free calculations. All map calculations were carried out with programs in the CCP4 package (22). The graphics program O (29) was used for inspection of electron density maps and manipulation of the coordinates between each macrocycle of refinement. The refined coordinates of the spinach Rubisco-activated complex with 2-CABP (Protein Data Bank entry 8ruc) (24) were used as the initial model of the 4-CABP complex and as a search model for the solution of the XuBP structure. All waters, 2-CABP, and the magnesium ion were excluded, and the occupancy of all residues in the active site was initially set to zero. After initial rigid body refinement to define non-crystallographic symmetry operators, refinement was continued using strict non-crystallographic symmetry restraints (4-fold in the case of the 4-CABP complex and 8-fold for the XuBP complex). Overall anisotropic B-factor refinement and scaling were applied in the case of the XuBP complex. Water molecules were added using the water adding tools in CCP4 and O. Peaks were identified as water molecules when they had well defined electron densities greater than 3ϫ the standard deviation of the Fo-Fc difference map and were within hydrogen bonding distance (2.5-3.2 Å) of a potential donor or acceptor. Water peaks were then included in the refinement with unit occupancy. Individual isotropic thermal parameters were refined before the calculation of electron density maps. Thermal parameters were then reset before further refinement. The models were subjected to critical quality analyses using X-PLOR, PROCHECK (34) of the enzyme or by an initial binding to the activated enzyme with subsequent loss of the metal and the activator CO 2 . Inhibition studies indicate that XuBP initially binds to the activated enzyme with weak affinity, as shown by an initial competitive inhibition phase with respect to the substrate RuBP (6). After 20 min the inhibition is essentially non-competitive, implying inactivation of the enzyme. The end product of this reaction is a complex between XuBP and the non-activated form of the enzyme (9,11). Loss of either the carbamate or the metal will lead to deactivation, leaving the metal ligands free to find other binding partners. In the XuBP structure, O-3 and one of the C-2 diol oxygens form polar interactions with Glu-204 and Asp-203, respectively, interactions that would not be possible in the presence of the metal ion or with a different stereochemistry at C-3. XuBP Binds as a Diol in the Active Site of Spinach Rubisco-The electron density of XuBP is consistent with the inhibitor binding as a C-2 diol. This was also found in the XuBP complex of the enzyme from Synechococcus (25). It could be that binding of the diol is favored in this particular structure. Alternatively, XuBP could bind as a ketone to the enzyme, which then assists the hydration of XuBP to a diol. A number of residues are close enough to act as potential acids and bases, thereby facilitating diol formation. Lys-175 and Lys-177 both approach the area which the carbonyl oxygen of the keto-form might be expected to occupy in such a complex and may polarize the carbonyl bond for a nucleophilic attack by a water to form a diol. Asp-203 and Glu-60 are sufficiently close to act as proton acceptors. Protonation of Asp-203 would destabilize the magnesium coordination and favor loss of the metal, but the conspicuous movement of Glu-60 with respect to the activated form of the enzyme might suggest this as a potential base. A curious difference between the Synechococcus and the spinach enzymes is that the latter displays "fallover" (7), the gradual loss of activity with time accompanied by production of XuBP in vitro. This phenomenon has not been detected in the Synechococcus enzyme (10). Given the very high degree of structural similarity of the XuBP complexes from spinach and Synechococcus (r.m.s. difference for all C ␣ atoms of the L subunits was 0.38 Å), the explanation for the different kinetic behavior must be linked to other differences in these enzymes. An obvious possibility is to consider differences in the stability of the enediolate on the enzyme (10) as well as differences in affinities for RuBP and XuBP. Such differences could result in different likelihoods for 4-CABP Complex-The binding of this inhibitor to Rubisco is not as extensively studied, but it has so far been assumed that it can form a stable complex with the magnesium-containing, carbamylated form of the enzyme (15). 2 From the present x-ray diffraction study, it is evident that 4-CABP forms a very stable binary complex with the non-carbamylated metal-free enzyme, a stability that is reflected by the well ordered crystals that can be grown from the complex (the limit of diffraction is about 1.5 Å). In contrast to cocrystallization studies, the soaking experiments with activated crystals are inconclusive in structural 2 G. Lorimer, personal communication. terms but in line with the notion (15) that initially an activated metastable quaternary E-CO 2 -Mg-4-CABP complex is formed but that this complex slowly rearranges to a more stable nonactivated E-4-CABP binary complex, thus expelling the carbamate and the metal from the active site. The data set obtained can be interpreted as a mixture of the two states. Similar soaking experiments using 2-CABP or RuBP 3 did not show this behavior, indicating that there is no inherent problem with the soaking procedure. In the cocrystallization experiment, there is ample time for the expulsion of the metal from the complex to happen before the onset of crystallization. In the soaking experiment, this process is much slower due to a restriction in conformational flexibility of the molecules imposed by the crystal lattice. A similar behavior could hold for the XuBP complex, i.e. initial binding to the activated enzyme with subsequent expulsion of the carbamate and the metal due to the different stereochemistry of the C-3 hydroxyl, and this is corroborated by the inhibition studies by McCurry and Tolbert (6).
The Metal May Be Important in Guiding the Ligand into the Active Site-This work shows that XuBP and 4-CABP bind in the same orientation as the activated complex of 2-CABP. Earlier soaking experiments with 2-CABP into crystals of nonactivated Rubisco from Rhodospirillum rubrum (26) and to-bacco (27) showed the ligand binding with the P1 phosphate in the P5 binding site and vice versa. These findings could be explained with the metal directing the ligand in the active site in the early stages of the binding in the present study. No such guidance was possible in the earlier soaking studies (26,27).
2-CABP and 4-CABP (Fig. 4) are both close mimics of the gemdiol form of the 6-carbon intermediate, 2-carboxy-3-keto-Darabinitol 1,5-bisphosphate (Fig. 1, compound IV). In the case of 2-CABP a stable activated complex can be formed whereas the configuration of 4-CABP seems to induce strain, which leads to the expulsion of the activator metal and carbamate on Lys-201 from the active site. This strain could be used in catalysis to drive the breakage of the C-2-C-3 bond whereas in the non-productive complexes it leads to deactivation. The hydration at C-2 of XuBP could be a consequence of the loss of the metal and not the cause of it.  Fig. 4 for numbering convention of atoms.