Mirror‐Image Packing Provides a Molecular Basis for the Nanomolar Equipotency of Enantiomers of an Experimental Herbicide

Abstract Programs of drug discovery generally exploit one enantiomer of a chiral compound for lead development following the principle that enantiomer recognition is central to biological specificity. However, chiral promiscuity has been identified for a number of enzyme families, which have shown that mirror‐image packing can enable opposite enantiomers to be accommodated in an enzyme's active site. Reported here is a series of crystallographic studies of complexes between an enzyme and a potent experimental herbicide whose chiral center forms an essential part of the inhibitor pharmacophore. Initial studies with a racemate at 1.85 Å resolution failed to identify the chirality of the bound inhibitor, however, by extending the resolution to 1.1 Å and by analyzing high‐resolution complexes with the enantiopure compounds, we determined that both enantiomers make equivalent pseudosymmetric interactions in the active site, thus mimicking an achiral reaction intermediate.

Although ac hirally pure compound is often the desired product of such schemes,t he early stages of both pharmaceutical and agrochemical inhibitor development typically start from synthetic racemates,w hich are often easier and more cost effective to synthesize than as pecific enantiomer (see Ref. [2,3] and references therein). Thep rogram of herbicide development described herein aims to identify inhibitors of imidazoleglycerolphosphate-dehydratase (IGPD;E C4 .2.1.19), an enzyme of histidine biosynthesis in plants and microorganisms.T his enzyme catalyzes the manganese(II)-dependent dehydration of (2R,3S)-2,3-dihydroxy-3-(1H-imidazol-5-yl)propyl dihydrogen phosphate (2R,3S-IGP) to 3-(1H-Imidazol-4-yl)-2-oxopropyl dihydrogen phosphate (IAP;F igure 1). [4] IGPD shows strict enantioselectivity for its substrate,with the other three diastereoisomers of IGP acting as competitive inhibitors of the enzyme, [5] thus indicating that the active site shows some flexibility to accommodate inverted chiral centers.P revious studies have shown that IGPD is inhibited by triazolylphosphonates, whose potencyi sb ased on mimicking key reaction intermediates. [6] In this paper we describe work on the lead compound, 2-hydroxy-3-(1,2,4-triazol-1-yl) propylphospho- nate (C348;F igure 1) and present as eries of structures of IGPD,from Arabidopsis thaliana,complexed with asynthetic racemate of C348 and with the enantiopure compounds.T he structural data show that both the (R)-and (S)-C348 make equivalent interactions with the active site of IGPD and bind with mirror-image packing.D ata from both in vivo and in vitro assays further shows that both enantiomers are equipotent nanomolar inhibitors of the enzyme.B ycomparing our enzyme/inhibitor complexes with that of structures of IGPD complexed with the substrate,2 R,3S-IGP,w ep rovide amolecular explanation of this finding.
Crystals of a DNc onstruct of IGPD isoform 2f rom Arabidopsis thaliana (At DNI GPD2 construct A; see Supporting Information for experimental methods) were grown in the presence of ar acemate of C348 and the structure of the enzyme/inhibitor complex was determined to 1.85 resolution. Theresulting electron density showed clear evidence for the ordering of the C-loop of IGPD2 (residues 193-206;s ee Figure S1 in the Supporting Information), with the enzyme adopting the same closed conformation as seen in the previously determined structure of an inactive mutant (E21Q) of At IGPD2 with its substrate,IGP (PDB:4MU4). [7] Difference density for C348 could be identified within the active site and the triazole ring could be modelled between the two manganese ions with the N2 and N4 atoms forming ligands to Mn1 and Mn2, respectively.T he C348 C2ÀOH group acts as an additional ligand to Mn1 and the phosphonate group is bound in ap ositively charged pocket, sur-rounded by the side chains of R99, R121, K177, S199, and K201, and by water-mediated hydrogen bonds to Q51 and H55. However,d espite the 1.85 resolution of the data, there was alack of electron density around C3 of the inhibitor (see Figure S2a). And, as aresult, whilst the major functional groups of the inhibitor could be identified, the chirality of the bound ligand was uncertain.
As horter DNc onstruct of IGPD2 (At DNI GPD2 construct B) yielded better quality diffracting crystals and produced as tructure of the IGPD2/C348 complex at 1.1 resolution (PDB:5 EKW;T able 1). However,a sw ith the electron density for C348 in the 1.85 structure,the density at 1.1 resolution was also weak around C3. Nevertheless, as mall peak could be observed in the map around C3 when contoured at 1 s,the position of which was consistent with the binding of the (S)-C348 (see Figure S2b). Refinement of this structure using tight geometric restraints revealed further difference features,i ndicating as econd position for the inhibitor,which, from the geometry,could only be interpreted as resulting from the binding of the (R)-C348 (see Figure S2c,d). This finding indicated that in the crystal, the active sites of some of the enzyme molecules were occupied by (S)-C348 and others were occupied by (R)-C348.S ubsequent refinement of this structure with occupancies of 0.6 and 0.4 for the (S)-and (R)-C348,r espectively,p roduced am odel which fully explained the electron density,thus indicating that both enantiomers of C348 were bound to an essentially invariant enzyme structure.  To confirm that both enantiomers of C348 bind to IGPD2, and to improve our interpretation of the mixed structure,we resolved the C348 racemate by HPLC to greater than 98 % enantiopurity.T he binding affinities for the enantiopure compounds,( S)-and (R)-C348,a gainst At DNI GPD2 (construct B) were measured by an in vitro enzyme assay and gave apparent K i values of (25 AE 3) and (15 AE 5) nm, respectively (Figure 2a;s ee Figure S3). Glasshouse spray studies,c omparing the herbicidal properties of the two enantiomers and the racemate,s howed they all exhibited similar efficacy in terms of the spray rate (kg/Ha) necessary to cause 50 %d amage (as visually assessed;s ee Figure S4). Whilst we had initially expected that the two enantiomers would have quite different binding affinities and herbicidal activity,t hey were found to be equipotent inhibitors of IGPD2. Thee nantiopure compounds were then co-crystallized with At IGPD2 DNconstruct Btoproduce structures at 1.15 and 1.5 resolution, for the S (PDB:5 EL9) and R forms (PDB:5 ELW), respectively.T he conformation of the enzyme and the position of the metal ions were equivalent in each complex, with the only clear difference,a part from the inhibitor conformation, being minor changes in the solvent structure within the active site (see Figure S5). In both structures high quality electron density covered all the atoms of the inhibitor (Figure 2b,c), thus confirming that the additional difference features seen in the complexes with the racemate arose as aresult of mixed binding. Theresulting models show that for each chiral form of C348,t he positions of the phosphonate group,N2and N4 of the triazole ring, and the C2ÀOH substituent superimpose almost exactly and make equivalent interactions within the active site of the enzyme (Figure 2d;s ee Figure S6a). This arrangement is achieved, despite the inversion in chirality,byacombination of torsionangle changes around the C3ÀC2 and C2ÀC1 bonds of the inhibitor,t ogether with ad ifference in the tilt of the triazole ring around the Mn1-Mn2 vector.T he two enantiomers thus trace an inverted path between the metal binding site and the phosphonate binding site,whereby they are related by mirrorimage packing (Figure 3a,b). Both enantiomers retain lowenergy conformations with approximately staggered torsion angles,w ith the largest deviation in the atomic position of equivalent atoms occurring at C3 of the backbone,t hus explaining the weak density at this position observed in the complexes of IGPD2 with the racemate.
Mirror-image packing of opposite enantiomers of ligands to enzymes and receptors has been reported previously. Examples include the binding of d/l-phenylalanine and the superinhibitors d/l-2-aminooxy-3-phenylpropionic acid (d/l-AOPP) to phenylalanine ammonia-lyase, [8] d/l-malate to citrate synthase, [9] and d/l-isocitrate to isocitrate dehydrogenase, [10] amongst others (see Figure S7). [11] In each of these examples,and in the binding of the two enantiomers of C348 to IGPD,e xamination of the structures shows that the position of three Rg roups around the chiral center is approximately maintained on the enzyme surface by flipping the direction of the hydrogen atom at the fourth position, by 1808 8,a fter inversion of chirality.T his corresponds to the approach of the free ligand to the enzyme surface in an inverted configuration and generates asmall separation in the position of the chiral center of each enantiomer on either side of the pseudo-mirror plane.S ubtle changes to the dihedral angles formed in the pendant groups optimize the fit within  Figure S6a.
the chiral environment of the active site.O nacase-by-case basis,w hether or not such as ituation is feasible depends on there being sufficient space to accommodate the two positions of the chiral center and the new position of the fourth ligand forming the chiral group (commonly ah ydrogen atom). In addition, the chemical nature of the pendant groups dictates whether changes in the dihedral angles are energetically accessible.A ny small differences in geometry resulting from the optimization of the fit to the active site will necessarily give rise to adifference in the affinity of the two enantiomers, but, as we show in IGPD,such differences can be remarkably small and both enantiomers can act as highly potent inhibitors.
As both enantiomers of C348 bind to At IGPD2, highresolution structures of Pyrococcus furiosus (Pf) IGPD (38 % sequence identity to At IGPD2) complexed with the enantiopure compounds (R:1 .8 ,P DB:5 DNX; S:1 .53 ,P DB: 5DNL) were determined. These showed the same mode of mirror-image packing as observed in At IGPD2 (see Figure S8), thus suggesting that the ability of IGPD to accommodate opposite chiral forms of C348 is ag eneral feature of the wider enzyme superfamily,rather than apeculiarity of the Arabidopsis enzyme.W eh ave previously proposed that catalysis by IGPD involves conversion between an open conformation of the enzyme,w hich binds to imidazole-IGP (PDB:4 MU3), and ac losed conformation, which binds imidazolate-IGP (PDB:4 MU4), wherein ad istinctly different binding site for the substrate-phosphate moiety is utilized. [7] Comparison of the open and closed At IGPD2/ substrate complexes with those of the At IGPD2/C348 complexes show that neither enantiomer of C348 can access the phosphate binding site observed in the open enzyme/ substrate complex, as the backbone of the inhibitor is one carbon atom shorter than that of the substrate.R ather,b oth enantiomers of C348 utilize the phosphate binding site that is associated with the closed conformation of the enzyme/ substrate complex, with the ordered C-loop;t he conformation believed to be that adopted by the enzyme during catalysis,thus suggesting that both enantiomers of C348 may mimic reaction intermediates.D uring the reaction catalyzed by IGPD,t he adoption of an sp 2 geometry at C3 is required for the formation of the diazafulvene intermediate,aprocess which is facilitated by the planar arrangement of the imidazolate ring, C3 and C2 of imidazolate-IGP.T he next step in the reaction involves production of the D 2 -enol, which also requires the adoption of an sp 2 geometry at C2. This geometry necessitates that C1 moves into the plane defined by C3, C2 and the imidazolate (Figure 3c). Comparison of the (R)-and (S)-C348 complexes with that of amodelled D 2 -enol shows that the sp 2 C2 of the D 2 -enol lies in approximately the same position as that occupied by the sp 3 C2 of both enantiomers of C348.M oreover,t he plane defined by the main functional groups of C348 is the same as that occupied by the modelled positions for the carbon backbone of the D 2enol, and essentially bisects the positions of the R and S enantiomers of the inhibitor (Figure 3d;see Figure S6b). This position implies that the two enantiomers of C348 are equipotent nanomolar inhibitors because a) the layout of the enzyme active site facilitates mirror-image packing and b) that they can both mimic the mode of binding of this achiral reaction intermediate.
X-ray analysis is ap owerful tool for studying the molecular basis of how ligands are recognized by biological macromolecules,but even when such studies are conducted at high resolution, ligand density can sometimes be difficult to interpret because of areas of weakness,t he origins of which are often difficult to explain and are commonly cited as instances of disorder.Inthis study,weak electron density was observed adjacent to the chiral center of the lead compound in a1 .85 resolution structure of our target enzyme in complex with ar acemate.B ye xtending the resolution and chirally resolving the two enantiomers we confirmed that the areas of weakness arose from the mirror-image packing of the two enantiomers of the inhibitor in the active site,a n observation not without precedence.W ithout the highresolution data this important finding might have been overlooked in IGPD2. Significantly,the mirror-image packing of the two enantiomers of C348 in IGPD2 gives rise to equipotencya nd, as in d/l-AOPP superinhibitors of phenylalanine ammonia-lyase, [8] the mimicry of ar eaction intermediate gives rise to potent inhibition. This study adds to the The plane (gray disk) was calculated in Chimera [1] based on the average positions of the N2, N4, C2ÀOH, and C1ÀPbond positions for both enantiomers of C348,w hich are colored teal and white for the for the S and R enantiomers, respectively. c) Amodel of the D 2 -enol (orange) in the imidazole-IGP (yellow)/IGPD2 complex (PDB:4 MU4). The sp 2 C2 of the enol can be accommodated without altering the position of the phosphate group. d) The modelled D 2 -enol bisects the positions of the two enantiomers of C348. ever-growing body of evidence that certain enzymes are chirally promiscuous and that mirror-image packing of ligands is amore common feature than is generally recognized in the field of drug development. Whilst the future challenge with IGPD is to exploit this understanding for the development of novel herbicides,o ur findings may also be relevant in other areas of drug discovery where the potential to develop inhibitors with opposite chirality may have been overlooked.