Structure of the T109S mutant of Escherichia coli dihydroorotase complexed with the inhibitor 5-fluoroorotate: catalytic activity is reflected by the crystal form

A single-point mutant (T109S) of E. coli dihydroorotase initially crystallizes so that the two monomers of the dimer are related by a crystallographic twofold axis. In the presence of substrate, conversion to the previously observed asymmetric dimer with substrate bound in one subunit and product in the other is observed.

DHOase belongs to the amidohydrolase superfamily, which comprises a variety of hydrolytic enzymes of the (/) 8 -barrel (or TIM-barrel) fold (Holm & Sander, 1997). The active sites of enzymes belonging to the amidohydrolase superfamily have five highly # 2007 International Union of Crystallography All rights reserved Figure 1 Cyclization of N-carbamyl-l-aspartate (l-CA-asp) to l-dihydroorotate (l-DHO).
conserved metal-binding residues, usually four histidine residues and one aspartate residue, and one or two metal ions. A common feature of the enzyme mechanism is the utilization of an activated water or hydroxide molecule bound to the metal ion(s) at the catalytic centre.
A phylogenetic analysis of the amino-acid sequences of DHOase reveals that the enzyme can be divided into two major classes that have diverged from a common ancestor of the amidohydrolase superfamily (Fields et al., 1999). Type I DHOases are the more ancient form and are found in all domains of life; they include the DHOase domain of mammalian CAD and monofunctional DHOases found in Gram-positive bacteria, including Bacillus, Lactobacillus and Streptococcus. Mammalian CAD is a trifunctional enzyme consisting of the first three enzyme activities in the pyrimidinesynthetic pathway: carbamyl phosphate synthetase (CPSase), aspartate transcarbamylase (ATCase) and DHOase (Simmer et al., 1990;Williams et al., 1990). Type II DHOases are smaller ($38 kDa compared with $45 kDa for the type I DHOases) monofunctional enzymes that are found predominantly in Gram-negative bacteria (e.g. Escherichia coli) and have a low level of sequence identity to their type I counterparts.
To date, the structures of two DHOases, those from E. coli and Aquiflex aeolicus, have been reported [PDB codes 1j79 and 1xge (Thoden et al., 2001;Lee et al., 2005) and 1xrt and 1xrf (Martin et al., 2005)]. E. coli DHOase is a monofunctional and homodimeric enzyme. Although the E. coli DHOase has been reported to contain one catalytic zinc per monomer (Brown & Collins, 1991;Washabaugh & Collins, 1984, 1986, the structure clearly showed that the active site contains a binuclear centre with a carboxylated lysine as one of the bridging ligands. In the original structure determination of E. coli DHOase, the crystals grown in the presence of racemic substrate, d,l-CA-asp, were orthorhombic with a dimer in the asymmetric unit. Interestingly, one subunit contained bound l-DHO and the other contained l-CA-asp (Thoden et al., 2001).
In our subsequent study of E. coli DHOase, we found two different conformations of a surface loop comprised of residues 105-115 (Lee et al., 2005). We also found asymmetry between the active sites in the dimer, with the product l-DHO in one subunit and the substrate l-CA-asp in the other, despite the fact that we used the product rather than the substrate to form the complex. However, we were able to resolve the positions of residues (109-112 from chain B) that were missing from the original structure and to observe two conformations of the surface loop (residues 105-115). The ability to resolve the surface loops in our structure was attributed to the use of optically pure l-DHO rather than racemic d,l-CA-asp in the crystallization medium. The loop asymmetry mirrored that of the active-site contents of the two subunits. In the substrate-bound subunit the surface loop (residues 105-115) reaches in towards the active site and makes two hydrogen-bonding interactions with the bound substrate molecule via two threonine residues (Thr109 and Thr110; 'loop-in'), whereas the loop forms part of the surface of the protein in the product-bound subunit ('loop-out'). Subsequent enzyme kinetics at low concentrations of l-DHO in the reverse degradative reaction showed positive cooperativity between the subunits.
Conformational changes during enzyme catalysis have been observed in many different enzymes and appear to be a general feature of enzymatic mechanisms (Gutteridge & Thornton, 2004;Hammes, 2002;Kempner, 1993). One of the movements involved in enzyme catalysis is the rearrangement of loops that constitute activesite lids. The conformations of these loops are tightly coupled to the catalytic state of the enzyme. In general, these movements are characterized by a closing of the active site, with the surface-loop regions moving in towards the active site of the protein, closing over the bound substrate. Catalysis takes place in the closed form and the enzyme opens again to release the product. This motion from open to closed is thought to fulfil a number of roles in enzyme reactions: (i) arrangement of the catalytic residues into the correct orientation for catalysis and/or restriction of the conformational freedom of the substrate, (ii) prevention of the escape of reaction intermediates before the reaction has completed and (iii) restriction of the entry of water and its subsequent reaction with unstable reaction intermediates.
To probe the role of the surface-loop movement of E. coli DHOase in catalysis, we generated a series of single-point mutants. The two threonine residues (Thr109 and Thr110) that interact with the bound substrate l-CA-asp in the active site of the wild-type enzyme were mutated to a number of different amino acids. The mutation of these residues produced enzymes with lower catalytic activities (unpublished work). In this paper, we report the crystal structure of one of the single-point mutants (T109S) of E. coli DHOase in complex with a product mimic, FOA (Fig. 2). We also discuss the behaviour of crystals of the mutant enzyme in the presence of l-DHO, which reveals a macroscopic effect of the loop movement on catalysis and crystallization.

Generation of T109S mutant DHOase
Mutation of Thr109 to Ser in E. coli DHOase was generated using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA) with forward primer 5 0 -CCGGCAAACGCAAGCAC-ACTAACTCCAGCCA-3 0 and reverse primer 5 0 -TGGCTGGAGT-TAGTGCTTGCGTTTGCCGG-3 0 . The wild-type DHOase in the pBS+ vector was used as a template and the mutations were confirmed by determination of the nucleotide sequences of the whole gene. E. coli strain X7014a, which lacks a gene for dihydroorotase, was obtained from the Yale E. coli Genetic Stock Center (Yale University, New Haven, CT, USA) and used for expression and purification of T109S mutant DHOase.

Crystallization
Purification procedures for T109S DHOase were adapted from the previous report for wild-type DHOase (Washabaugh & Collins, 1984). The purified protein was dialyzed into 20 mM Na HEPES pH 7.2 and 1 mM DTT. Crystals were grown by the hanging-drop vapourdiffusion method at 277 K. In the initial attempt to grow crystals of T109S DHOase, the crystallization conditions for native DHOase were applied, using 2 ml protein solution (8.2 mg ml À1 ) mixed with 2 ml reservoir solution (15-20% PEG 3350, 0.1 M MES pH 6-6.5, 75 mM MgCl 2 and 0.15 M KCl) and 0.45 ml 100 mM l-DHO (Lee et al., 2005). Crystallization conditions were modified slightly in order to increase the stability of tetragonal crystals by using reservoir solution containing 14-16% PEG 3350, 0. 0.2 M KCl and 30% sucrose. 10 mM 5-fluoroorotate (FOA, Sigma) was added to the drop to replace l-DHO. Crystals typically grew overnight. Prior to data collection, the crystals were further cryoprotected by adding 2 ml reservoir solution to the drop and were flash-cooled in a stream of nitrogen gas at 100 K.

Data collection and processing
Synchrotron data were recorded to 2.2 Å resolution from a single crystal on beamline 23-ID at the Advanced Photon Source (APS), Argonne National Laboratory using a MAR mosaic 300 CCD detector (MAR USA) with a wavelength of 0.9793 Å at 100 K. A total of 190 successive frames were collected with an oscillation angle of 0.5 , giving a total coverage of 95 and an overall redundancy of 6.1. The diffraction data were processed and scaled with HKL-2000 (Otwinowski & Minor, 1997). The unit-cell parameters are a = b = 72.6, c = 176.1 Å , = = = 90 . The data could be processed in either space group P4 or P422 with similar results (R merge = 0.040 in P4, 0.043 in P422). The higher-symmetry Laue group was preferred and chosen for molecular replacement and was subsequently confirmed by the structure refinement. The Matthews coefficient for space group P422 (V M = 3.0 Å 3 Da À1 ) suggests there is one monomer per asymmetric unit, with a calculated solvent content of 58.9% (Matthews, 1968).

Structure solution and refinement
Despite the fact that reflections were clearly visible in diffraction images to a resolution of 2.2 Å , the inherent anisotropy of the diffraction resulted in a severe limitation in the completeness of the data beyond 2.7 Å resolution ( Fig. 3 and Table 1). Therefore, only data to 2.7 Å were used for structure solution and refinement.
The crystal structure was solved by molecular replacement using Phaser from the CCP4 suite (Collaborative Computational Project, Number 4, 1994;Storoni et al., 2004). Chain A of the native DHOase structure in space group P2 1 2 1 2 1 (PDB code 1xge; Lee et al., 2005) was used as the search model. The Zn atoms, ligand molecule and solvent molecules were deleted from the search model. Residues 105-115 were given zero occupancy to avoid model bias in the phase calculation.
Structure refinement was carried out using REFMAC5 (Collaborative Computational Project, Number 4, 1994;Murshudov et al., 1997). 4.7% of the reflections were randomly selected to create an independent data set of test reflections for cross-validation throughout the refinement procedure. The initial attempt to refine the structure using the data set to 2.7 Å resulted in atomic coordi-  Table 1 Statistics of the data collection and refinement.
Values in parentheses are for the outer shell. 11.7 Generously allowed regions (%) 0.7 † Because of the severe diffraction anisotropy, data beyond these limits were excluded during refinement. ‡ Completeness was calculated using the entire spherical shell of resolution 2.7 Å . § One FOA molecule. } Estimated standard uncertainty in atomic position, based on maximum likelihood (Murshudov et al., 1997). † † Calculated using PROCHECK (Laskowski et al., 1993).

Figure 3
Diffraction pattern from a tetragonal crystal of E. coli T109S DHOase recorded on beamline 23-ID at APS. The diffraction is severely anisotropic; reflections are observed along c* (indicated by the white arrow) to 2.2 Å (black arrow). The resolution at the edge of the image belonging to 2.13 Å .

Figure 4
Analysis of the anisotropy of the data calculated using the program TRUNCATE (French & Wilson, 1978). nates with high B factors (over 100 Å 2 ). Since the data were severely anisotropic, we reasoned that the inclusion of unmeasurably weak reflections along the a* and b* directions by using a spherical resolution limit impeded the refinement. We utilized the diffractionanisotropy server (http://nihserver.mbi.ucla.edu/anisoscale; Strong et al., 2006) to analyse the data. The recommended resolution limits along a*, b*, and c* were 3.0, 3.0 and 2.2 Å , respectively, as indicated by the program TRUNCATE (Table 2 and Fig. 4; French & Wilson, 1978). Reflections falling outside the ellipsoid defined by the specified resolutions were discarded and the remaining data were scaled with anisotropic parameters (Lodowski et al., 2003;Zhang et al., 2004). A negative isotropic B-factor correction was applied to restore the magnitude of the high-resolution reflections that was diminished by anisotropic scaling. This anisotropy-corrected data set provided by the diffraction-anisotropy server was used for further refinement using REFMAC5. Several rounds of rigid-body refinement for data in the resolution range 31-3.0 Å gave R = 0.36 and R free = 0.36. The model was then subjected to restrained refinement using TLS (Winn et al., 2001) with data to 2.7 Å . The initial difference Fourier maps clearly showed that the surface loop (residues 105-115), omitted from the model, was in the 'loop-out' conformation. Further refinement with REFMAC5 was interspersed with manual model building using Coot (Emsley & Cowtan, 2004). In the final stages of the refinement, the FOA inhibitor was modelled into difference electron density in the active site and refined. The FOA parameters were generated with the assistance of the PRODRG server (Schü ttelkopf & van Aalten, 2004).
Model validation was conducted using PROCHECK (Laskowski et al., 1993), WHAT-CHECK (Hooft et al., 1996) and the MOLPROBITY server (Lovell et al., 2003). Crystals of E. coli T109S DHOase. Tetragonal crystals belonging to space group P4 1 2 1 2 (a) are very unstable and often disintegrated shortly after crystal formation (b). The small crystals grown around the remnants of the tetragonal crystals are orthorhombic crystals belonging to space group P2 1 2 1 2 1 .

Table 2
Completeness statistics of pre-and post-anisotropy correction of diffraction data.
Completeness was calculated using the entire spherical shell. The anisotropy correction removes all reflections lying outside the defined ellipsoid. The completeness measured as a percentage of the spherical shell is therefore lower following the correction. Data beyond a resolution of 2.7 Å were not used in the refinement (see Table 1).

Before correction
After correction that 87.6% of the residues fell in the most favoured region of the Ramachandran plot and 11.7% of the residues were in additionally allowed regions. Two residues (His177 and Arg20) were found in the generously allowed regions. This observation is consistent with that from the high-resolution wild-type structure. Superposition and calculation of root-mean-square differences were performed using the program LSQMAN (Kleywegt & Jones, 1997). Figures were generated using PyMOL (DeLano Scientific, San Carlos, CA, USA).

Crystal instability
The initial attempt to grow crystals of T109S DHOase in the presence of l-DHO using the crystallization conditions for wild-type crystals unexpectedly produced tetragonal crystals rather than the orthorhombic crystals characteristic of the wild-type enzyme. However, these crystals, which grew in 24 h, were extremely unstable and disintegrated shortly after formation (Fig. 5). Diffraction from these short-lived crystals was too poor to record diffraction data for refinement. However, the space group and unit-cell parameters could be determined unambiguously and were found to be identical to those of the FOA complex of T109S DHOase reported in this paper. Orthorhombic crystals isomorphous to those of the wild-type DHOase subsequently grew from the debris of the tetragonal crystals or at new nucleation sites in the same drops.
Attempts to stabilize the tetragonal crystals in the presence of l-DHO by varying the conditions failed. We reasoned that the crystal instability was caused when catalysis had generated sufficient l-CAasp to bind to the DHOase, resulting in a movement of the flexible loop (residues 105-115). Evidence for this hypothesis was ultimately  obtained by analysis of the tetragonal crystal structure (see x3.4). To stabilize the tetragonal crystals, we used FOA, a product mimic, for crystallization in place of l-DHO (Fig. 2). Since FOA cannot be hydrolysed by the enzyme, the formation of a symmetrical dimer with FOA in both active sites was possible. The crystals grown in the presence of FOA were stable and data were successfully collected. The diffraction from the crystals was very anisotropic. Reflections were observable to 2.2 Å resolution parallel to c* but only to about 3 Å perpendicular to this direction (Figs. 3 and 4).
The structure was solved by molecular replacement and refined to 2.7 Å resolution in space group P4 1 2 1 2 with a monomer in the asymmetric unit (Table 1).

Description of the structure
The final model consists of one monomer of E. coli T109S DHOase with 343 protein residues, ten water molecules, two Zn atoms and one molecule of FOA. Refinement using REFMAC5 with TLS converged with residuals R = 0.218 and R free = 0.257. Anisotropy correction of the diffraction data and the use of the TLS parameters in refinement improved the electron-density maps and refinement statistics. The molecular structure of E. coli T109S DHOase in space group P4 1 2 1 2 is closely similar to the previously solved wild-type structure in space group P2 1 2 1 2 1 [PDB codes 1j79 (Thoden et al., 2001) and 1xge (Lee et al., 2005)]. The crystals of both forms were grown under similar conditions, except for the addition of sucrose to the tetragonal crystallization conditions to provide further stabilization and cryoprotection. The core of the molecule forms a TIM-barrel fold with eight alternating -strands/-helices. Crystals of the wild-type enzyme in space group P2 1 2 1 2 1 have a dimer in the asymmetric unit, while those of T109S DHOase in space group P4 1 2 1 2 have only a monomer. There are no significant differences between the mainchain C positions of chain A from the wild-type structure and the mutant structure; the overall difference in r.m.s. distance for 343 residues is 0.23 Å , which is within the ESU of the mutant structure (0.27 Å ; Murshudov et al., 1997). Apart from the residues at the termini, the largest difference (0.64 Å ) between chain A of the wildtype structure and the mutant structure is the mutated residue Ser109 (Thr in the native structure). In contrast, superposition of chain B of the wild-type structure with the mutant structure gives an r.m.s. difference of C positions of 1.65 Å , with the greatest distance between pairs of C atoms being 14.0 Å for the same mutated residue, Ser109. In the wild-type structure the flexible loops (residues 105-115) are in different conformations: 'out' when l-DHO is in the active site (chain A) and 'in' when l-CA-asp is in the active site (chain B). In the tetragonal crystal forms the two monomers in the dimer are crystallographically identical and the loop adopts the 'out' conformation in both, consistent with the presence of a product mimic, FOA, in both active sites.
Superposition of the 'loop-in' conformation from the wild-type structure (chain B from 1xge) onto the tetragonal structure clearly illustrates the effect of the movement of the loop (Fig. 8b) on the crystal contacts. If the flexible loop moves in towards the active site, the two hydrogen bonds involving Ser112 will be lost. The loss of these contacts is apparently sufficient to destabilize the crystal Molecular packing in crystals of the tetragonal FOA complex of T109S DHOase. The flexible surface loops (residues 105-115) are coloured magenta. Symmetryrelated molecules that make direct hydrogen-bonded contacts with the central monomer (yellow) are in blue (y, x, Àz), green (y À 1/2, Àx + 1/2, z À 1/4) and cyan (Ày + 1/2, x + 1/2, z + 1/4), respectively. The Zn atoms in the active site are shown as magenta spheres. Table 3 Crystal contacts between symmetry-related molecules in the tetragonal T109S crystal.
ii † All interactions from 2.5 to 3.3 Å between potential hydrogen-bond donors and acceptors are included. ‡ (i) Ày + 1/2, x + 1/2, z + 1/4; (ii) y À 1/2, Àx + 1/2, z À 1/4. packing. Therefore, the packing of molecules in this crystal form can only be maintained when the loop is in the 'out' conformation. The presence of a product mimic, FOA, in both active sites is incompatible with the 'loop-in' conformation and it is therefore able to stabilize the tetragonal crystals, which require both loops to be 'out'.
Unlike in the tetragonal crystal, where the crystal contacts made by only two regions of the structure are responsible for the stability of the crystal packing, the molecules in the wild-type orthorhombic form make many more crystal contacts ( Fig. 6 and Table 4). The chain A molecule in the orthorhombic crystals, which has the 'loop-out' conformation, makes 12 hydrogen-bond interactions with six different symmetry-related molecules. Interestingly, these include the same crystal contact that is observed in tetragonal crystals in which residues Asp82, Ser83 and Ser112 interact with residues Lys*8, Arg*10 and Glu*292 of a symmetry-related molecule. In the orthorhombic crystals, in which the two chains have different conformations, the equivalent hydrogen-bonded contacts involving residues on chain B are incompatible with the 'loop-in' conformation. The chain B molecule in the orthorhombic crystal makes other crystal contacts with four different symmetry-related molecules via 14 hydrogen-bonding interactions. Each dimer in the orthorhombic crystals makes direct contacts with a total of ten symmetry-related molecules, whereas the equivalent dimers (two monomers) in the tetragonal crystals make direct contacts with only four symmetryrelated molecules.

Conformational change of DHOase and crystal formation
The combination of few contacts between molecules in the crystal and the involvement of the flexible loop (residues 105-115) in the crystal packing explains the instability of this crystal form when the protein was crystallized in the presence of l-DHO. Thr109 is one of protein structure communications  Table 4 Crystal contacts between symmetry-related molecules in the orthorhombic wildtype crystal.

Figure 8
Crystal contacts between symmetry-related molecules in the tetragonal FOA complex of T109S DHOase. The symmetry-related molecule (y À 1/2, Àx + 1/2, z À 1/4) is two threonine residues (Thr109 and Thr110) that are directly involved in the hydrogen-bonding interactions between the surface loop (residues 105-115) and the bound substrate l-CA-asp in the active site in the structure of wild-type E. coli DHOase (Lee et al., 2005). Mutation of this residue to serine reduces the catalytic activity of E. coli DHOase to about 30% of that of the wild-type enzyme 1 . Despite the conservative nature of the mutation, Ser for Thr, we presume that the decreased activity results from the greater conformational flexibility of the Ser side chain, which destabilizes the interaction with the substrate. As a result of the lower activity, the conversion of l-DHO to l-CA-asp is slower and initially allows a high population of DHOase with l-DHO in both active sites and both loops in the 'out' conformation. This permits the protein to crystallize in the tetragonal form with the loops of each monomer involved in crystal contacts. As l-DHO is slowly converted to l-CA-asp, the flexible loop moves in to close the active site, breaking the crystal contacts stabilized by the loop in its 'out' conformation, which then results in the disintegration of the crystals. Once sufficient l-CA-asp is available, dimers with l-CA-asp in one subunit and l-DHO in the other can form and orthorhombic crystals can grow. Wild-type enzyme converts l-DHO to l-CA-asp sufficiently rapidly that there is never a sufficient population of enzyme with l-DHO bound to both subunits to produce the tetragonal crystals.

Conclusions
We have observed the progress of the enzymatic reaction of DHOase by the formation of one crystal form in the presence of l-DHO, its subsequent dissolution as a loop moves during catalysis and the ultimate formation of a new crystal form with both substrate and product bound. The complex with l-DHO bound in both subunits could only be visualized as an inhibitor complex of E. coli T109S DHOase with lower activity. The surface loop that moves during catalysis plays a crucial role in anchoring symmetry-related molecules together in the tetragonal crystals. This explains the instability of the tetragonal crystals when crystallized in the presence of l-DHO rather than the nonhydrolysable analogue FOA.