Electrostatic Transition State Stabilization Rather Than Reactant Destabilization Provides the Chemical Basis for Efficient Chorismate Mutase Catalysis

Home , Chorismate mutase, Mutase

Electrostatic transition state stabilization rather than reactant destabilization provides the chemical basis for efficient chorismate mutase catalysis

Daniel Burschowskya, André van Eerdea, Mats Ökvista, Alexander Kienhöferb, Peter Kastb,1, Donald Hilvertb,1, and Ute Krengela,1

aDepartment of Chemistry, University of Oslo, NO-0315 Oslo, Norway; and bLaboratory of Organic Chemistry, ETH Zurich, CH-8093 Zurich, Switzerland

Edited by Arieh Warshel, University of Southern California, Los Angeles, CA, and approved October 30, 2014 (received for review May 8, 2014) For more than half a century, transition state theory has provided ability of a protein to populate this reactive ground state con- a useful framework for understanding the origins of enzyme former has been proposed to account almost entirely for the ef- catalysis. As proposed by Pauling, enzymes accelerate chemical ficiency of CM catalysis (15–17). reactions by binding transition states tighter than substrates, Structural studies have shown that CMs use an extensive thereby lowering the activation energy compared with that of array of hydrogen-bonding and electrostatic interactions to the corresponding uncatalyzed process. This paradigm has been bind the pseudodiaxial substrate conformer (18–20). Sterically challenged for chorismate mutase (CM), a well-characterized constraining chorismate in a “near attack conformation” metabolic enzyme that catalyzes the rearrangement of chorismate (NAC) (15–17, 21), in which the reacting centers are confined to prephenate. Calculations have predicted the decisive factor in to contact distances, would be expected to increase the prob- CM catalysis to be ground state destabilization rather than transition state stabilization. Using X-ray crystallography, we ability of reaction. Ground state destabilization through con- show, in contrast, that a sluggish variant of Bacillus subtilis CM, formational compression has been shown to afford large rate in which a cationic active-site arginine was replaced by a neutral accelerations for Claisen rearrangements in synthetic model citrulline, is a poor catalyst even though it effectively preorganizes systems (22). Based on molecular dynamics (MD), quantum me- chorismate for the reaction. A series of high-resolution molecular chanics/molecular mechanics (QM/MM), and thermodynamic snapshots of the reaction coordinate, including the apo enzyme, integration studies, Bruice and coworkers estimated that selec- and complexes with substrate, transition state analog and prod- tive NAC formation at the enzyme active site accounts for ∼90% uct, demonstrate that an active site, which is only complementary of the kinetic advantage provided by CM and concluded that in shape to a reactive substrate conformer, is insufficient for ef- transition state binding is unimportant for catalysis in this system fective catalysis. Instead, as with other enzymes, electrostatic sta- (15–17). In contrast, other QM/MM studies (23, 24), including bilization of the CM transition state appears to be crucial for semiempirical and higher level calculations, have suggested that achieving high reaction rates. NAC formation is the result of transition state stabilization, not the source of catalysis. Polar active site residues, most notably a cat- pericyclic reaction | catalysis | enzyme mechanism | near attack conformation | ionic arginine or lysine positioned next to the ether oxygen of the X-ray crystal structures Significance early seven decades ago, Pauling proposed that an enzyme Npreferentially binds and stabilizes the transition state of the chemical reaction it catalyzes relative to the substrate in the Chorismate mutase (CM) is a textbook model for enzyme ca- ground state (1). As a consequence, the activation energy is low- talysis. Although it promotes a simple unimolecular reaction, the origins of its 2-million–fold rate acceleration have been ered for the catalyzed versus the uncatalyzed process, giving rise to debated for decades. The relative importance of electrostatic large rate accelerations. This simple model has successfully in- transition state stabilization versus ground state destabilization formed the development of potent enzyme inhibitors (2), catalytic has been a particularly contentious issue. High-resolution crys- antibodies (3), and computationally designed catalysts (4, 5). tallographic snapshots of an engineered CM variant and its Nevertheless, due to the complexity of even the simplest enzy- complexes with substrate, transition state analog, and product matic systems, a quantitative understanding of the energetics of now provide strong experimental evidence that properly posi- catalysis has remained elusive. The relative contributions of elec- tioned active-site charges are essential in this system and that trostatic, steric, proximity, and solvent effects to transition state preorganization of the substrate in a reactive conformation stabilization and reactant destabilization, as well as the role of contributes relatively little to catalysis. A proper understanding dynamic motions and tunneling, are still heatedly debated (6–10). of the role of electrostatics in this and other enzymes is im- Chorismate mutase (CM) has become a popular model system portant for ongoing efforts to design new enzymes de novo. for studies on enzyme catalysis. It accelerates the unimolecular conversion of chorismate (1) to prephenate (3) (Fig. 1A), a key Author contributions: D.B., P.K., D.H., and U.K. designed research; D.B., M.Ö., and U.K. step in the biosynthesis of tyrosine and phenylalanine, 2 × 106-fold performed research; A.K. contributed new reagents/analytic tools; D.B., A.v.E., M.Ö., and over the spontaneous reaction in water (11). Both the enzymatic U.K. analyzed data; and D.B., A.v.E., M.Ö., A.K., P.K., D.H., and U.K. wrote the paper. and nonenzymatic processes have been shown to be concerted The authors declare no conflict of interest. Claisen rearrangements with asynchronous chairlike transition This article is a PNAS Direct Submission. states (2) (12, 13). Because chorismate exists in an extended Data deposition: The apo BsCM crystal structure and the Arg90Cit BsCM* complexes have 1a been deposited in the RCSB Protein Data Bank, www.rcsb.org (PDB ID codes 3ZOP, 3ZP4, pseudodiequatorial conformation ( ) in aqueous solution (14), 3ZP7, and 3ZO8). accessing the transition state first requires that the substrate un- 1To whom correspondence may be addressed. Email: [email protected], kast@ dergo an energetically unfavorable conformational change to org.chem.ethz.ch, or [email protected]. 1b populate a rare pseudodiaxial conformer ( )inwhichtheenol- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. pyruvyl side chain is positioned over the cyclohexadiene ring. The 1073/pnas.1408512111/-/DCSupplemental.

17516–17521 | PNAS | December 9, 2014 | vol. 111 | no. 49 www.pnas.org/cgi/doi/10.1073/pnas.1408512111 Downloaded by guest on October 1, 2021 of the Arg90Cit variant is not due to inferior transition state -O C A - - - ‡ 2 CO O C O2C 1 2 2 stabilization but rather to major structural rearrangements at the 6 O O 2 O - - - active site that adversely affect the efficiency of forming 5 CO CO CO2 4 2 2 HO 3 O 9 reactive conformers. + - Here we present structural evidence that bolsters the findings CO2 HO HO HO (1a)(1b)(2)(3) from biochemical characterization of the Arg90Cit mutant. High- B C resolution crystallographic snapshots of BsCM and its ligand Arg7 complexes provide direct insight into key species along the re- Tyr108 action coordinate. As seen with many other enzymes (8), elec-

HN + NH2 trostatic transition state stabilization rather than reactant HO Xaa90 NH 2 - O destabilization appears to be the determining factor in CM

NH O + catalysis. H2N NH2 O - X NH O HN 2 Results O Glu78 Arg63 O- HO (4) To resolve the debate about the relative importance of transition H state stabilization for CM catalysis, we set out to determine the O N Cys75 crystal structure of Arg90Cit BsCM* in its apo form and in O complex with mechanistically relevant ligands. In addition to D Arg90Cit, this variant contains a second substitution (Asp102Glu) that was introduced for semisynthetic production; this change has no affect on activity (25) but serves to distinguish the BsCM* from the WT BsCM samples. Well-diffracting crystals of WT BsCM and Arg90Cit BsCM* (Fig. S1) were obtained under very similar conditions, and the structures of both apo enzymes were solved to a resolution of 1.6 Å. The models include residues 1–117 of the 127-residue enzyme, with residues 2–115 showing well-defined electron density. R/Rfree values are 0.14/0.17 for WT BsCM and 0.17/0.22 for apo Arg90Cit BsCM*. As found previously (18), BsCM adopts a homotrimeric pseudo-α/β-barrel fold with the three active sites positioned at the subunit interfaces (Fig. S2). The overall structures of WT BsCM and Arg90Cit BsCM* are identical within error limits Fig. 1. CM reaction and enzyme active site. (A) The Claisen rearrangement (rms difference of 0.3 Å for Cα of residues 2–115, comparing of chorismate (1a and 1b) to prephenate (3) occurs via a chair-like pericyclic + trimers) and, even in the active site, the two structures super- transition state (2). (B) Schematic view of the active site of BsCM* (X = NH2 ) = impose well. For Arg90Cit BsCM* we note a somewhat greater and Arg90Cit BsCM* (X O) in complex with the conformationally con- conformational freedom for the active-site residues compared strained transition state analog 4.(C)Aσ -weighted 2F − F map of the A o c with WT BsCM, with residues Arg7, Cys75, Glu78, and Cit90 Arg90Cit BsCM* active site in complex with 4. The electron density is con- A toured at 1.5 σ; for clarity, portions of the enzyme obscuring the view of the alternating between two conformations (Fig. 2 ). These con- ligand are not shown. Residues from different subunits are shown in light formational states are largely coupled through an extensive and dark green. Dashed lines indicate H-bonding contacts. A water mole- hydrogen-bonding network, although some variation between cule, bridging the carboxylate groups of the ligand, is shown as a red sphere. the three different active sites of the protein is observed. In all (D) Stereoview of the Arg90Cit BsCM* complex (protein in green; ligand in cases, one of the conformational states closely resembles the black) superimposed on the analogous complex of WT BsCM [PDB ID code corresponding conformation in the WT BsCM structure, with 2CHT (18); protein and ligand in gray] and the newly determined apo ∼50% of the residues adopting this conformation. The virtually BIOPHYSICS AND structure of WT BsCM (white). unchanged affinity for substrate and inhibitor suggest that this COMPUTATIONAL BIOLOGY increased flexibility cannot account for the dramatically de- breaking C–O bond, were proposed to stabilize the high-energy creased catalytic efficiency of the Arg90Cit variant. Indeed, even more flexible enzymes, such as an engineered monomeric CM transition state electrostatically relative to the bound substrate. Methanococcus jannaschii Kienhöfer et al. experimentally probed the importance of from (30), which exhibits the charac- electrostatic effects in this system using a semisynthetic variant of teristics of an intrinsically disordered molten globule in the ab- sence of ligand (31, 32), exhibit near native-like activity (30, 33). the CM from Bacillus subtilis (BsCM) (25). Seeking to separate The crystal structure of Arg90Cit BsCM* in complex with ground state from transition state effects, they replaced the a transition state analog (compound 4) (34) was solved to 1.8-Å positively charged Arg90 residue at the active site of BsCM (Fig. R R B resolution with / free values of 0.18/0.23. In this complex, all 1 ) with a noncharged arginine analog, the amino acid citrulline, active-site residues adopt exclusively the WT conformation (Fig. which is not normally found in proteins. Consistent with the 1 C and D), with the exception of Arg63, which is generally hypothesis that Arg90 stabilizes a developing partial negative – disordered. As in the analogous complex with WT BsCM (18), charge in the transition state (18, 26 28), the Arg90Cit sub- the inhibitor makes numerous contacts with the protein. The stitution has little effect on the ground state Michaelis complex, side-chain carboxylate of 4 forms a salt bridge with Arg7 and also K judging from the mere threefold increase in m for chorismate hydrogen bonds with Cit90 and Tyr108; its hydroxyl group is K K (note that m approximates s for WT BsCM) (29), but it greatly hydrogen bonded to Glu78 and the backbone amide of Cys75; k k K impairs catalytic efficiency, reducing both cat and cat/ m by and its ether oxygen is within hydrogen bonding distance of more than 4 orders of magnitude (25). These changes suggest Cit90. Citrulline differs from arginine in having a neutral urea in + that the positively charged arginine contributes up to 5.9 kcal/mol place of the cationic guanidinium group (in effect, an NH2 to transition state stabilization but only 0.6 kcal/mol to the group is replaced by an oxygen atom); hence the two amino acids binding energy of the ground state. This interpretation has been are isosteric but differ in net charge. The orientation of the side- challenged by Hur and Bruice (17), however. Based on free- chain urea group was assigned based on meaningful hydrogen energy MD simulations, they have argued that the poor activity bonding geometry, which placed the citrulline NH2 group at a

Burschowsky et al. PNAS | December 9, 2014 | vol. 111 | no. 49 | 17517 Downloaded by guest on October 1, 2021 distance of 3.2 Å from the ether oxygen of 4. It is evident that Arg90Cit BsCM*, with and without the transition state analog, is very similar to WT BsCM (Fig. 1D). The modest 1.1-kcal/mol decrease in stability of the inhibitor complex compared with that of WT BsCM (25) is likely due to the slight disorder observed in the apo Arg90Cit BsCM* binding pocket. Although the chorismate to prephenate rearrangement is highly exothermic (35), the greatly impaired catalytic competence of Arg90Cit BsCM* raised the possibility of structurally character- izing the substrate complex. We therefore soaked the apo crystal in a solution containing 40 mM chorismate. To minimize conversion of chorismate to prephenate due to spontaneous thermal rearrangement or residual enzyme activity, the soaking experiments were performed rapidly (for 5 min) and at reduced temperature (4 °C), followed by flash-freezing. The structure was solved to 1.7-Å resolution, with R/Rfree values of 0.17/0.22. As anticipated, chorismate was not fully converted, allowing the substrate complex to be partially trapped. The relative occupancies of chorismate and prephenate were estimated by comparing the σA-weighted Fo − Fc difference maps, refined for different ligand ratios (Materials and Methods); final chorismate/prephenate ratios are 0.35/0.25 (chains E/D; Fig. 2B), 0.25/0.3 (chains B/A), 0.3/0.4 (chains C/B), 0.2/0.4 (chains A/C; Fig. 2D) and 0.15/0.3 (chains D/F). The electron density of the ligand mixture is highest for the side-chain carboxylate and the ring hydroxyl group, which have identical positions for substrate and product, whereas the density is lowest for the bonds being broken (C3–O7 in chorismate) or formed (C1–C9 of prephenate). Both ligands adopt conformations similar to the transition state analog 4 (Fig. 2C), with the prephenate complex closely resembling the previously published (18) prephenate-bound structure of WT BsCM (Fig. S3). This uniform mode of ligand binding explains why the enzyme’s Kd values for binding of prephenate and chorismate are similar (29). The MD simulations performed by Hur and Bruice (17, 36) predicted large conformational changes for Glu78, Arg7, Cit90, and Arg116 in the low-activity variant. Specifically, the charge imbalance resulting upon substitution of Arg90 was expected to fundamentally alter the hydrogen-bonding pattern at the active site, causing displacement of the Cit90 side chain and formation of a salt bridge between Arg7 and Glu78 rather than between Arg7 and the side-chain carboxylate of the substrate/transition state analog. As a consequence, the side-chain carboxylate of chorismate was expected to form a salt bridge with Arg116 rather than with Arg7, thereby increasing the distance between the vinyl C9 carbon and C1 of chorismate to 4.5 ± 0.2 Å, compared with 3.5 ± 0.3 Å calculated for the WT. Nevertheless, as shown in Fig. 1 C and D, we find no evidence for this altered binding mode. The crystal structures unequivocally show that the Arg90Cit substitution does not induce a major reorganization of the BsCM active site. The residues of main importance in the catalytic pocket (i.e., Arg7, Phe57, Cys75, Glu78, Cit90, and Tyr108) as- sume the same conformations as the analogous residues in WT BsCM, with and without bound ligands. Moreover, the C1–C9 distance for chorismate in the Michaelis complex ranges from 3.5 Fig. 2. Snapshots of the BsCM reaction coordinate. Stereo images of the to 3.7 Å in the occupied crystallographically nonequivalent active Arg90Cit BsCM* active site (A) without ligand and complexed with (B) sites of the variant, matching the calculated WT value very well. chorismate (in the 1b conformation), (C) transition state analog 4,and(D) This distance is slightly larger than the combined van der Waals

prephenate (3). The electron density maps (σA-weighted 2Fo − Fc maps radii of the two carbon atoms (1.7 Å each). represented by gray mesh) are contoured at 1.5 σ. Contours reflecting the different occupancies of the ligands have been added (dark gray mesh at Discussion 0.8 σ for 4 and 0.5 σ for chorismate/prephenate); real space correlation The structures of the resting Arg90Cit BsCM* enzyme (Fig. 2A) coefficients for the depicted ligands are 0.86 for chorismate, 0.85 for 4 and its complexes with substrate (Fig. 2B), a transition state and 0.88 for prephenate, consistent with well-defined structures. Inter- C D actions (<3.5 Å) with the ligands are indicated with broken lines, labeled analog (Fig. 2 ), and product (Fig. 2 ) provide experimental with the corresponding distances in ångströms. The displayed active sites snapshots of the major states of the CM reaction (see Movie S1 are between chains E/D, E/D, C/B, and A/C, respectively, which exhibit an for a morphed animation of the snapshots). The empty enzyme is rms difference over all atoms of the displayed amino acid side chains of preorganized for catalysis, binding chorismate in the rare pseu- 0.16 ± 0.03 Å. dodiaxial conformation primed for pericyclic rearrangement, the

17518 | www.pnas.org/cgi/doi/10.1073/pnas.1408512111 Burschowsky et al. Downloaded by guest on October 1, 2021 Table 1. Data collection and refinement statistics Structure Arg90Cit BsCM* Arg90Cit BsCM* with 4 Arg90Cit BsCM* with 1 + 3 WT BsCM

X-ray source MicroMax-007HF ID23-2, ESRF ID23-2, ESRF MicroMax-007HF Wavelength, Å 1.54187 0.87260 0.87260 1.54187

Space group P1 P1 P1 P21 Unit cell parameters a, Å 50.1 50.2 50.2 49.5 b, Å 51.1 50.8 51.3 165.2 c, Å 69.9 69.8 70.1 49.5 α, ° 97.8 97.4 97.5 90.0 β, ° 93.0 93.2 92.9 118.7 γ , ° 101.3 101.0 101.1 90.0 Resolution, Å 43.47–1.61 (1.71–1.61) 49.12–1.80 (1.90–1.80) 49.15–1.70 (1.79–1.70) 43.51–1.59 (1.69–1.59) I/σ(I) 21.5 (7.2) 23.3 (5.0) 15.5 (4.3) 14.4 (4.4) No. of reflections Observed 135,299 (18,781) 100,173 (11,421) 133,842 (21,371) 227,540 (19,579) Unique 76,808 (10,808) 53,585 (6,216) 70,311 (11,262) 79,445 (8,930) Multiplicity 1.8 (1.7) 1.9 (1.8) 1.9 (1.9) 2.9 (2.2) † Completeness, % 88.3 (77.0) 86.1 (61.8) 93.7 (92.8) 85.0 (59.4) ‡ Rmerge 0.020 (0.095) 0.021 (0.129) 0.034 (0.173) 0.052 (0.176) § Rmeas 0.029 (0.135) 0.029 (0.182) 0.049 (0.245) 0.062 (0.222) { R/Rfree 0.17/0.22 0.18/0.23 0.17/0.22 0.14/0.17 rms bond length, Å 0.008 0.007 0.007 0.006 rms bond angle, ° 1.203 1.051 1.043 1.171 Average B factor, Å2 Backbone 16.6 24.7 15.9 12.4 Side chain + water 20.9 31.7 23.3 16.9 All atoms 19.0 28.4 20.1 14.8 No. of atoms Protein 5,905 5,604 5,754 5,605 Ligand(s) n.a. 64 160 n.a. Solvent 450 356 563 227 Ramachandran, % Favored 98.7 99.0 99.0 98.4 Allowed 1.3 1.0 1.0 1.6 Outliers 0.0 0.0 0.0 0.0 PDB ID code 3ZOP 3ZP4 3ZP7 3ZO8

Values in parentheses are referring to the highest resolution shell. n.a., not applicable. † The three highest resolution shells are less than 90% complete, except Arg90Cit BsCM* with 1 + 3. ‡ P P P P = ^ − nh ^ = ð = Þ nh Rmerge P hpIffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffih Ih,i hP i Ih,i , whereP IhP 1 nh i Ih,i . P § = =ð − Þ nh ^ − nh ^ = ð = Þ nh RmeasP h nh nPh 1 i Ih Ih,i h i Ih,i , where Ih 1 nh i Ih,i (62). { R = jFoj − jFcj jFoj, where jFoj and jFcj are the observed and calculated structure factor amplitudes, respectively. For Rfree the BIOPHYSICS AND reflections not used in the refinement were used for the calculation. COMPUTATIONAL BIOLOGY

so-called NAC. Simply constraining the substrate to a reactive complementarity to the substrate. More important is the ability of conformation is not sufficient for efficient catalysis, however. The these enzymes to exploit electrostatic preorganization of the active kinetic parameters for Arg90Cit BsCM* indicate that populating site for the native substrate to stabilize the transition state of a NAC-like conformer contributes at most a factor of ∼100 toward promiscuous reactions. Capitalizing on this capacity could be useful the 2 × 106-fold enzymatic rate acceleration over the spontaneous for ongoing efforts to design new active sites de novo (4, 5). Claisen rearrangement. Because the missing positive charge is the Together with previous mutagenesis (26–28) and modeling salient feature of the weakly active variant, we conclude that sta- studies (23, 24, 37), our results provide a unified view of CM bilization of the transition state via electrostatic interactions is the catalysis. They highlight the importance of a cationic residue— main factor determining the catalytic proficiency of BsCM. This either an arginine or a lysine—proximal to the ether oxygen of the conclusion is supported by recent high-level QM/MM studies substrate in the transition state. All known natural CMs, including showing that the higher free-energy barrier for the Arg90Cit the structurally unrelated AroQ enzymes from Escherichia coli mutant is entirely ascribable to less effective transition state sta- (19), yeast (20), and Mycobacterium tuberculosis (41, 42) share this bilization by the neutral citrulline (24, 37). Upon completion of feature, whereas poor catalysts like Arg90Cit or the catalytic an- the reaction, the product, which is generated in an unfavorable tibody 1F7 (43) lack it. Rather than an exception to the classic conformation, is released to bulk solvent. view of enzymatic catalysis, CM appears to be an archetypical The finding that electrostatic complementarity in the transition example of the Pauling paradigm. state is a more important contributor to catalysis for BsCM than an active-site geometry tailored to a specific reactive substrate con- Materials and Methods former is potentially relevant to understanding enzyme promiscuity Production of BsCM. WT BsCM was produced in the CM-deficient E. coli strain and the evolution of enzyme function. Work on arylsulfatase (38), KA13 (44, 45) and purified according to ref. 28. Semisynthetic Arg90Cit BsCM*, alkaline phosphatase (39), and paraoxonase (40) has revealed which additionally contained the substitution Asp102Glu (*) to prevent that the dominant factor for efficient catalysis is not shape a problematic aspartimide formation, was produced according to ref. 25.

Burschowsky et al. PNAS | December 9, 2014 | vol. 111 | no. 49 | 17519 Downloaded by guest on October 1, 2021 This substitution in addition served as a “watermark”,confirmingthatthe Data Processing, Structure Determination, and Refinement. The raw diffraction Arg90Cit BsCM* variant was indeed present in the crystals. In brief, the WT images were processed using XDS (49). Scaling of the data were performed BsCM fragment (1–87) was produced recombinantly in E. coli strain KA13, using SCALA (50) and the structures were determined by molecular replacement fused to the Mycobacterium xenopi GyrA intein and a chitin-binding do- using the program Phaser (51), and programs of the CCP4 Suite (52). The WT main. It was purified and cleaved on a chitin column to yield the BsCM (1–87) BsCM crystal structure (residues 1–117) with the Protein Data Bank (PDB) ID thioester. Arg90Cit BsCM* (88–127) was produced by solid-phase peptide code 1DBF (53) was used as a search model. The final models were obtained by synthesis. Both fragments were coupled by native chemical ligation to yield iterating cycles of refinement and model building using REFMAC5 (54) and Coot full-length Arg90Cit BsCM*. The ligated enzyme was purified by ion-ex- (55), respectively (for refinement statistics, see Table 1). For cross-validation, 5% change chromatography, and its identity was confirmed using mass spec- of the reflections were selected at random. In the case of WT BsCM, the data trometry and circular dichroism spectroscopy. The correct oligomerization were almost perfectly twinned, resulting in the apparent space group C2221 rather than P2 . A perfectly twinned dataset was generated by averaging the state was confirmed with analytical size-exclusion chromatography and 1 twin-related structure factor amplitudes, and intensity-based twin refinement analytical ultracentrifugation (25). The protein was subsequently concen- was then used for REFMAC5 runs. In the twinned case, the test set was selected trated to 3 mM by ultrafiltration for crystallization. as 20 thin shells over the entire resolution range, ensuring that twin-related reflections were always in the same set. Anisotropic B factor refinement was Ligand Production. Chorismate and transition state analog 4 were prepared used for the apo structures; for the structures of Arg90Cit BsCM* bound to 4 according to previously described procedures (46, 47). The initial purity of and chorismate/prephenate, isotropic B factor refinement was used, but in- > the chorismate stock solution before soaking was estimated to be 90% by cluded translation/libration/screw-motion (TLS) anisotropy refinement in the monitoring the chorismate concentration by UV absorption measurements final runs with 18–20 TLS groups per protein chain. TLS groups were defined at 274 nm during conversion to prephenate using WT BsCM. As chorismate using the TLSMD web server (56, 57). TLS refinement runs were performed in was still converted at a low rate by Arg90Cit BsCM* and also spontaneously the PHENIX suite (58). In all cases, only parameters belonging to atoms of the rearranges to prephenate in aqueous solutions, the presence of prephenate protein model were refined for the first few cycles, and then most ordered in the crystal-soaking experiments cannot be avoided. water molecules were added in the course of two cycles at positions where there were sufficient electron density and sensible hydrogen-bonding partners. Crystallization and Crystal Soaking. Crystals of WT BsCM were grown using the In the case of 4- and chorismate/prephenate-bound structures, no water mol- hanging drop vapor diffusion technique, by adding 1 μL well solution, con- ecules or ligands were modeled in close proximity of the enzymatic pockets taining 100 mM malic acid, Mes, Tris (MMT) buffer (in molar ratios of 1:2:2, during these initial cycles. Subsequently, the ligands were modeled into the

respectively), pH 6.0 (48); 100–150 mM MgCl2; 25% (wt/vol) PEG 1000; and active sites, if at least part of the ligand electron density was clearly visible at σ σ − 0.3 mM NaN3,to1μL concentrated protein solution (3 mM WT BsCM, 10 mM 1.0 ( A-weighted 2Fo Fc map). During the refinement process, the phases Tris buffer pH 7.5, 2 mM DTT, 0.125 mM EDTA). Large rhomboid-shaped crystals improved significantly, allowing ligand modeling in additional active sites. The (to approximately 500 μm in length) grew after 4–5 d at 20 °C. Arg90Cit BsCM* occupancies of the ligands were adjusted manually, by matching the B factors crystals were grown using similar conditions. Here, 0.75 μL well solution, con- of atoms in hydrogen bonds with the enzyme. For the structures with chorismate/prephenate bound in the active site, models of both ligands were taining 100 mM MMT buffer pH 6.0, 50–150 mM CaCl2,24–26% (wt/vol) PEG added to the active sites into the same electron density, initially with equal 1000, and 0.3 mM NaN3,wasaddedto0.75μL concentrated protein solution occupancies, to represent the mixture present in the crystal. The final relative (3 mM Arg90Cit BsCM*, 20 mM Tris buffer pH 8.0, 0.6 mM PMSF, 0.3 mM NaN3). occupancies for the ligands were then estimated by comparing difference Thin, elongated rhomboid-shaped crystals (to approximately 500 μm in length) maps refined for different ligand ratios. The electron density of citrulline is in- grew after 3–4 d at 20 °C. The crystals that were not used for soaking experi- distinguishable from that of an arginine residue, unless atomic resolution data ments were briefly dipped into a cryoprotectant solution [freshly prepared are available (1.0 Å or better). Here, the watermark substitution Asp102Glu reservoir solution, with 5% (vol/vol) PEG 400 added] before flash-freezing in served its purpose, confirming the proper identity of Arg90Cit BsCM*. For a stream of dry nitrogen gas at 100 K or liquid nitrogen. For soaking, reservoir ligand-bound structures the head group orientation was determined based on solution containing in addition 5% (vol/vol) PEG 400 and a lower MMT buffer chemical sense, relying on the fact that there was only one potential bonding concentration of 27 mM was freshly prepared. Additionally, the solution con- partner with sensible H-bonding geometry close to the citrulline amide group, tained 10 mM transition state analog 4 or 40 mM chorismate, respectively, but thereby determining its orientation. In this orientation, Cit90 is positioned lacked CaCl2 for better soaking efficiency. Estimated twice the volume of the within 3.2 Å (average over four active sites) from the ether oxygen of 4 (Fig. 1 B solution was then pipetted into the drop containing the crystals. Soaking of 4 and C) and within 3.0 Å from the corresponding O7 of chorismate. Structure into the crystal was performed at room temperature for 1.5 h before freezing; validation was performed using Molprobity (59) and its implementations in chorismate soaking was performed at 4 °C for 5 min. Coot or PHENIX, structure superimpositions were performed with the CCP4- supported program Superpose (60), and images of the models were prepared X-Ray Diffraction Data Collection. For WT BsCM and Arg90Cit BsCM* (apo) with PyMOL (61). crystals, the X-ray diffraction experiments were performed using a Rigaku MicroMax-007HF instrument. The data were recorded on an R-AXIS-IV++ ACKNOWLEDGMENTS. We thank the staff at the European Synchrotron detector at a temperature of 100 K. For Arg90Cit BsCM* crystals soaked with Radiation Facility for assistance and support in using the beamline ID23-2, chorismate or 4, data were collected at the European Synchrotron Radiation Dr. R. Pulido for synthesizing transition state analog 4, and Kathrin Roderer for the production of chorismate. We further acknowledge the contribution Facility (ESRF) in Grenoble, France (Beamline ID23-2) at a wavelength of of Dr. R. Dey in the early phases of this project. This work was funded by the 0.8726 Å and a temperature of 100 K; the data were recorded on a Mar/ ETH Zurich, the University of Oslo, the Carl Trygger Foundation (Grant Rayonix 3 × 3 Mosaic 225 detector. Further information and data collection 02:158), the Swiss National Science Foundation, and the Norwegian Research statistics are listed in Table 1. Council (Grant 143645).

1. Pauling L (1946) Molecular architecture and biological reactions. Chem Eng News 10. Herschlag D, Natarajan A (2013) Fundamental challenges in mechanistic enzymology: 24(10):1375–1377. Progress toward understanding the rate enhancements of enzymes. Biochemistry 2. Schramm VL (2011) Enzymatic transition states, transition-state analogs, dynamics, 52(12):2050–2067. thermodynamics, and lifetimes. Annu Rev Biochem 80:703–732. 11. Andrews PR, Smith GD, Young IG (1973) Transition-state stabilization and enzymic 3. Hilvert D (2000) Critical analysis of antibody catalysis. Annu Rev Biochem 69: catalysis. Kinetic and molecular orbital studies of the rearrangement of chorismate to 751–793. prephenate. Biochemistry 12(18):3492–3498. 4. Hilvert D (2013) Design of protein catalysts. Annu Rev Biochem 82:447–470. 12. Wiest O, Houk KN (1995) Stabilization of the transition state of the chorismate- 5. Kiss G, Çelebi-Ölçüm N, Moretti R, Baker D, Houk KN (2013) Computational enzyme prephenate rearrangement: An ab initio study of enzyme and antibody catalysis. JAm design. Angew Chem Int Ed Engl 52(22):5700–5725. Chem Soc 117(47):11628–11639. 6. Benkovic SJ, Hammes-Schiffer S (2003) A perspective on enzyme catalysis. Science 13. Gustin DJ, et al. (1999) Heavy atom isotope effects reveal a highly polarized transition 301(5637):1196–1202. state for chorismate mutase. J Am Chem Soc 121(8):1756–1757. 7. Garcia-Viloca M, Gao J, Karplus M, Truhlar DG (2004) How enzymes work: Analysis by 14. Copley SD, Knowles JR (1987) The conformational equilibrium of chorismate in modern rate theory and computer simulations. Science 303(5655):186–195. solution: Implications for the mechanism of the non-enzymic and the enzyme- 8. Warshel A, et al. (2006) Electrostatic basis for enzyme catalysis. Chem Rev 106(8): catalyzed rearrangement of chorismate to prephenate. J Am Chem Soc 109(16): 3210–3235. 5008–5013. 9. Nagel ZD, Klinman JP (2009) A 21st century revisionist’s view at a turning point in 15. Hur S, Bruice TC (2003) Enzymes do what is expected (chalcone isomerase versus enzymology. Nat Chem Biol 5(8):543–550. chorismate mutase). J Am Chem Soc 125(6):1472–1473.

17520 | www.pnas.org/cgi/doi/10.1073/pnas.1408512111 Burschowsky et al. Downloaded by guest on October 1, 2021 16. Hur S, Bruice TC (2003) Comparison of formation of reactive conformers (NACs) 38. Luo J, van Loo B, Kamerlin SCL (2012) Catalytic promiscuity in Pseudomonas aerugi- for the Claisen rearrangement of chorismate to prephenate in water and in the nosa arylsulfatase as an example of chemistry-driven protein evolution. FEBS Lett E. coli mutase: The efficiency of the enzyme catalysis. JAmChemSoc125(19): 586(11):1622–1630. 5964–5972. 39. Hou G, Cui Q (2013) Stabilization of different types of transition states in a single 17. Hur S, Bruice TC (2003) The near attack conformation approach to the study of the enzyme active site: QM/MM analysis of enzymes in the alkaline phosphatase super- chorismate to prephenate reaction. Proc Natl Acad Sci USA 100(21):12015–12020. family. J Am Chem Soc 135(28):10457–10469. 18. Chook YM, Gray JV, Ke H, Lipscomb WN (1994) The monofunctional chorismate 40. Ben-David M, et al. (2012) Catalytic versatility and backups in enzyme active sites: The mutase from Bacillus subtilis. Structure determination of chorismate mutase and its case of serum paraoxonase 1. J Mol Biol 418(3-4):181–196. complexes with a transition state analog and prephenate, and implications for the 41. Ökvist M, et al. (2006) 1.6 Å crystal structure of the secreted chorismate mutase mechanism of the enzymatic reaction. J Mol Biol 240(5):476–500. from Mycobacterium tuberculosis: Novel fold topology revealed. J Mol Biol 357(5): 19. Lee AY, Karplus PA, Ganem B, Clardy J (1995) Atomic structure of the buried catalytic 1483–1499. pocket of Escherichia coli chorismate mutase. J Am Chem Soc 117(12):3627–3628. 42. Sasso S, et al. (2009) Structure and function of a complex between chorismate mutase 20. Sträter N, Schnappauf G, Braus G, Lipscomb WN (1997) Mechanisms of catalysis and and DAHP synthase: Efficiency boost for the junior partner. EMBO J 28(14):2128–2142. allosteric regulation of yeast chorismate mutase from crystal structures. Structure 43. Haynes MR, Stura EA, Hilvert D, Wilson IA (1994) Routes to catalysis: Structure of – 5(11):1437 1452. a catalytic antibody and comparison with its natural counterpart. Science 263(5147): 21. Bruice TC, Benkovic SJ (2000) Chemical basis for enzyme catalysis. Biochemistry 39(21): 646–652. – 6267 6274. 44. MacBeath G, Kast P (1998) UGA read-through artifacts—when popular gene ex- 22. Khanjin NA, Snyder JP, Menger FM (1999) Mechanism of chorismate mutase: Con- pression systems need a pATCH. Biotechniques 24(5):789–794. tribution of conformational restriction to catalysis in the Claisen rearrangement. JAm 45. MacBeath G, Kast P, Hilvert D (1998) A small, thermostable, and monofunctional – Chem Soc 121(50):11831 11846. chorismate mutase from the archaeon Methanococcus jannaschii. Biochemistry 23. Strajbl M, Shurki A, Kato M, Warshel A (2003) Apparent NAC effect in chorismate 37(28):10062–10073. mutase reflects electrostatic transition state stabilization. J Am Chem Soc 125(34): 46. Grisostomi C, Kast P, Pulido R, Huynh J, Hilvert D (1997) Efficient in vivo synthesis and 10228–10237. rapid purification of chorismic acid using an engineered Escherichia coli strain. Bioorg 24. Claeyssens F, et al. (2011) Analysis of chorismate mutase catalysis by QM/MM Chem 25(5-6):297–305. modelling of enzyme-catalysed and uncatalysed reactions. Org Biomol Chem 47. Smith WW, Bartlett PA (1993) An improved synthesis of the transition state analog 9(5):1578–1590. inhibitor of chorismate mutase. J Org Chem 58(25):7308–7309. 25. Kienhöfer A, Kast P, Hilvert D (2003) Selective stabilization of the chorismate mutase 48. Newman J (2004) Novel buffer systems for macromolecular crystallization. Acta transition state by a positively charged hydrogen bond donor. J Am Chem Soc Crystallogr D Biol Crystallogr 60(Pt 3):610–612. 125(11):3206–3207. 49. Kabsch W (2010) XDS. Acta Crystallogr D Biol Crystallogr 66(Pt 2):125–132. 26. Cload ST, Liu DR, Pastor RM, Schultz PG (1996) Mutagenesis study of active site resi- 50. Evans P (2006) Scaling and assessment of data quality. Acta Crystallogr D Biol Crys- dues in chorismate mutase from Bacillus subtilis. J Am Chem Soc 118(7):1787–1788. tallogr 62(Pt 1):72–82. 27. Kast P, Asif-Ullah M, Jiang N, Hilvert D (1996) Exploring the active site of chorismate 51. McCoy AJ, et al. (2007) Phaser crystallographic software. J Appl Cryst 40(Pt 4):658–674. mutase by combinatorial mutagenesis and selection: The importance of electrostatic 52. Winn MD, et al. (2011) Overview of the CCP4 suite and current developments. Acta catalysis. Proc Natl Acad Sci USA 93(10):5043–5048. Crystallogr D Biol Crystallogr 67(Pt 4):235–242. 28. Kast P, et al. (2000) A strategically positioned cation is crucial for efficient catalysis by 53. Ladner JE, et al. (2000) The 1.30 Å resolution structure of the Bacillus subtilis cho- chorismate mutase. J Biol Chem 275(47):36832–36838. rismate mutase catalytic homotrimer. Acta Crystallogr D Biol Crystallogr 56(Pt 6): 29. Mattei P, Kast P, Hilvert D (1999) Bacillus subtilis chorismate mutase is partially – diffusion-controlled. Eur J Biochem 261(1):25–32. 673 683. 30. MacBeath G, Kast P, Hilvert D (1998) Redesigning enzyme topology by directed 54. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular struc- evolution. Science 279(5358):1958–1961. tures by the maximum-likelihood method. Acta Crystallogr D Biol Crystallogr 53(Pt 3): – 31. Vamvaca K, Vögeli B, Kast P, Pervushin K, Hilvert D (2004) An enzymatic molten 240 255. globule: Efficient coupling of folding and catalysis. Proc Natl Acad Sci USA 101(35): 55. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot. – 12860–12864. Acta Crystallogr D Biol Crystallogr 66(Pt 4):486 501. 32. Pervushin K, Vamvaca K, Vögeli B, Hilvert D (2007) Structure and dynamics of a 56. Painter J, Merritt EA (2006) Optimal description of a protein structure in terms of molten globular enzyme. Nat Struct Mol Biol 14(12):1202–1206. multiple groups undergoing TLS motion. Acta Crystallogr D Biol Crystallogr 62(Pt 4): 33. Roca M, Messer B, Hilvert D, Warshel A (2008) On the relationship between folding 439–450. and chemical landscapes in enzyme catalysis. Proc Natl Acad Sci USA 105(37): 57. Painter J, Merritt EA (2006) TLSMD web server for the generation of multi-group TLS 13877–13882. models. J Appl Cryst 39:109–111. 34. Bartlett PA, Nakagawa Y, Johnson CR, Reich SH, Luis A (1988) Chorismate mutase 58. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro- inhibitors: Synthesis and evaluation of some potential transition-state analogues. molecular structure solution. Acta Crystallogr D Biol Crystallogr 66(Pt 2):213–221. J Org Chem 53(14):3195–3210. 59. Chen VB, et al. (2010) MolProbity: All-atom structure validation for macromolecular 35. Kast P, et al. (1997) Thermodynamics of the conversion of chorismate to prephenate: crystallography. Acta Crystallogr D Biol Crystallogr 66(Pt 1):12–21. Experimental results and theoretical predictions. J Phys Chem B 101(50):10976–10982. 60. Krissinel E, Henrick K (2004) Secondary-structure matching (SSM), a new tool for fast 36. Hur S, Bruice TC (2002) The mechanism of catalysis of the chorismate to prephenate protein structure alignment in three dimensions. Acta Crystallogr D Biol Crystallogr BIOPHYSICS AND reaction by the Escherichia coli mutase enzyme. Proc Natl Acad Sci USA 99(3): 60(Pt 12 Pt 1):2256–2268.

1176–1181. 61. The PyMOL Molecular Graphics System, Version 1.6.x (2010) (Schrödinger, LLC, Portland, COMPUTATIONAL BIOLOGY 37. Guimarães CRW, Udier-Blagovic M, Tubert-Brohman I, Jorgensen WL (2005) Effects of OR). Arg90 neutralization on the enzyme-catalyzed rearrangement of chorismate to 62. Diederichs K, Karplus PA (1997) Improved R-factors for diffraction data analysis in prephenate. J Chem Theory Comput 1(4):617–625. macromolecular crystallography. Nat Struct Biol 4(4):269–275.

Burschowsky et al. PNAS | December 9, 2014 | vol. 111 | no. 49 | 17521 Downloaded by guest on October 1, 2021