Structural Elucidation of Chalcone Reductase and Implications for Deoxychalcone Biosynthesis*

4,2′,4′,6′-tetrahydroxychalcone (chalcone) and 4,2′,4′-trihydroxychalcone (deoxychalcone) serve as precursors of ecologically important flavonoids and isoflavonoids. Deoxychalcone formation depends on chalcone synthase and chalcone reductase; however, the identity of the chalcone reductase substrate out of the possible substrates formed during the multistep reaction catalyzed by chalcone synthase remains experimentally elusive. We report here the three-dimensional structure of alfalfa chalcone reductase bound to the NADP+ cofactor and propose the identity and binding mode of its substrate, namely the non-aromatized coumaryl-trione intermediate of the chalcone synthase-catalyzed cyclization of the fully extended coumaryl-tetraketide thioester intermediate. In the absence of a ternary complex, the quality of the refined NADP+-bound chalcone reductase structure serves as a template for computer-assisted docking to evaluate the likelihood of possible substrates. Interestingly, chalcone reductase adopts the three-dimensional structure of the aldo/keto reductase superfamily. The aldo/keto reductase fold is structurally distinct from all known ketoreductases of fatty acid biosynthesis, which instead belong to the short-chain dehydrogenase/reductase superfamily. The results presented here provide structural support for convergent functional evolution of these two ketoreductases that share similar roles in the biosynthesis of fatty acids/polyketides. In addition, the chalcone reductase structure represents the first protein structure of a member of the aldo/ketoreductase 4 family. Therefore, the chalcone reductase structure serves as a template for the homology modeling of other aldo/keto-reductase 4 family members, including the reductase involved in morphine biosynthesis, namely codeinone reductase.

The evolution of sessile terrestrial plants necessitated the acquisition of mechanisms to combat herbivore and pathogen attack (1). A common defense mechanism induced in response to such challenges is chemical in nature and involves the de novo synthesis of antimicrobial compounds, collectively known as phytoalexins (2). Chalcone Synthase (CHS) 1 and Chalcone Reductase (CHR) play key roles in the formation of an important set of phytoalexin precursors (recently reviewed in Ref. 3). In both legumes and non-legumes, CHS catalyzes the first committed step of flavonoid biosynthesis, forming 4,2Ј,4Ј6Јtetrahydroxychalcone (chalcone) from coumaroyl-CoA and three molecules of malonyl-CoA. This iterative multistep reaction generates linear di-, tri-, and tetra-ketide-CoA intermediates as well as a cyclic and prochiral coumaryl-trione intermediate ( Fig. 1). In addition to CHS, leguminous plants possess CHR, which acts on an intermediate of the multistep CHS reaction, yielding chalcone and 4,2Ј,4Ј-trihydroxychalcone (deoxychalcone) from the coupled catalytic action of these two enzymes (Fig. 1). CHR therefore sits at a critical biosynthetic branch point that has afforded legumes with the additional ability to synthesize a set of related deoxychalcone-derived phytoalexins in response to herbivore or pathogen attack including isoflavonoids, coumestans, pterocarpans, and isoflavans (2). In addition to their activity as phytoalexins, CHR products function in the induction of symbiotic root nodulation by Rhizobium bacteria leading to nitrogen fixation (4,5).
AKR superfamily enzymes are monomeric, (␣/␤) 8 -barrel, NAD(P)(H) binding oxidoreductases that metabolize a wide array of chemical substrates (23). Most AKR superfamily members possess broad substrate specificities, making the unequivocal determination of the physiological function of uncharacterized AKRs difficult. As of 2004, the more than 120 AKRs identified have been classified into 14 families (24).
Within the AKR superfamily, CHR, the alkaloid biosynthetic * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We present here the three-dimensional structure of M. sativa (alfalfa) CHR in complex with the cofactor NADP ϩ . This structure represents the first AKR4 family member to be structurally characterized and thus serves as a starting point for the structural understanding of other AKR4 family enzymes. As proof of principle in this regard, we discuss structural insights gleaned from a COR homology model derived from our CHR structure.
The CHR structure presented here provides insight into the coupled action of CHS and CHR during the CHS/CHR-catalyzed formation of deoxychalcone. The biosynthetic route to chalcone is well understood due to extensive structural and mechanistic investigations of CHS (3,29), whereas the mechanistic details of deoxychalcone formation, specifically the identity of the CHR substrate, remain unknown. CHR does not reduce chalcone (30,31), and CHS-mediated reduction of the thioester carbonyl of the diketide intermediate would prevent further polyketide extension by CHS, as was recently confirmed by Oguro et al. (30). Although the prevailing assumption has been that one of the two remaining linear CHS intermediates must be the CHR substrate, it has been recently noted that reduction of the prochiral cyclized trione intermediate formed through the CHS-catalyzed intramolecular Claisen condensation of the tetraketide intermediate would produce the same deoxychalcone product ( Fig. 1) (3). The three-dimensional structure of CHR presented here suggests the identity of the CHR substrate, as the high quality of our structural data enables us to utilize computer-assisted docking of possible CHR substrates in order to circumstantially address the identity of the CHR substrate.
Structure Elucidation-Phase determination was accomplished by molecular replacement using CNS (36). The search model was generated with Modeler (34) using the structure of human type III 3␣hydroxysteroid dehydrogenase (PDB accession number 1IHI) as a template. Given the documented variability in the A, B, and C loops of AKR family members, these loop regions were removed in our search model. It was necessary to rotate one monomer of the original molecular replacement solution ϳ90°to relieve steric clashes between symmetry mates. The 4-fold pseudo-symmetry of the CHR Triose Phosphate Isomerase (TIM)-barrel enabled this near-correct solution. The initial CNS molecular replacement model was manually adjusted and completed by model building in O (35) and refined using CNS (36) to 1.7 Å resolution ( Table I). The final structure was evaluated with PRO-CHECK (37). The CHR⅐NADP ϩ complex had 91.9, 8.1, 0.0, and 0% of residues in the most favored, allowed, generously allowed, and disallowed regions of the Ramachandran plot, respectively. Main chain and side chain structural parameters were well above average with a G factor of 0.39 (37). The final structural coordinates and structure factors were deposited to the Protein Data Bank under PDB accession number 1ZGD.
Automated Substrate Docking-Genetic Optimization for Ligand Docking (GOLD) was employed for automated docking of putative substrates into the CHR active site (38). GOLD optimizes the fitness score of many possible solutions using a genetic algorithm. Parameters controlling the precise operation of the genetic algorithm were as follows: population size (100), selection pressure (1.100000), number of operations (100,000), number of islands (5), niche size (2), crossover weight (95), mutate weight (95), and migrate weight (10). The default parameter values for van der Waals and hydrogen bonding were used throughout the docking process. Because of the known catalytic mechanism of AKR superfamily enzymes, whereby a hydride from the nicotinamide C4 is transferred to the substrate carbonyl carbon for reduction, all docking calculations imposed a constraint requiring the C6 atom of coumaryl-trione to be within 2.5-4 Å of nicotinamide ring C4 and between 3.5 and 5 Å of nicotinamide ring C3. These restraints effectively impose a distance restraint on one of the two trione carbonyls that can be reduced as there is free rotation around the C6-C7 bond. Prior to aromatization, the six-membered ring is prochiral at C6 and non-planar, adopting either a modified boat or chair conformation; therefore, the C1 and C5 carbonyls are not equivalent (Fig. 3). Imposing a carbonyl carbon-nicotinamide C4 distance constraint on either of the carbonyls would bias the calculation. Fifty docking calculations were run for each ligand, and the GOLD score was used to identify the lowest energy docking results.

RESULTS
Purification and Crystallization-Alfalfa CHR was expressed in E. coli as an N-terminal octahistidine-tagged fusion protein and purified via Ni 2ϩ -NTA affinity and size exclusion chromatography. CHR crystallized in the presence of a 14-fold molar excess of the NADP ϩ cofactor with or without 5 mM 4,6-dioxoheptanoic acid. CHR crystals diffracted to 1.70 Å. Molecular replacement with a CHR homology model based on the structure of human 3␣-hydroxysteroid dehydrogenase (PDB code 1IHI) and phase extension to high resolution yielded experimental electron density maps of excellent quality revealing a molecule of NADP ϩ bound to the CHR cofactor binding site of each molecule in the asymmetric unit. No additional electron density corresponding to 4,6-dioxoheptanoic acid was observed. Iterative rounds of model building and refinement of the two molecules of CHR in the asymmetric unit yielded a final CHR structure exhibiting well defined electron density for all but the first five residues (Table I). We investigated crystal soaking as well co-crystallization strategies of CHR with an additional set of substrate and product analogs. Several promising data sets were collected but failed to exhibit any additional electron density corresponding to a bound ligand near or within the active site encompassing the NADP ϩ cofactor/product.
Overall Structure-The three-dimensional fold of CHR is that of an (␣/␤) 8 -barrel, with eight parallel strands lining the interior of the ␣/␤-barrel and eight helices packing along the outside of the ␤-strands (Fig. 4A). The barrel is closed at the N-terminal end by a ␤-hairpin. This (␣/␤) 8 pattern is disrupted by the insertion of two additional helices following ␣8 (A1) and between ␤7 and ␤8 (A2). The additional ␣-helices pack together along the outer barrel. This pattern of an (␣/␤) 8 -barrel plus two additional ␣-helices was first observed in TIM and is often referred to as a TIM-barrel (39). The C-terminal end of the TIM barrel is open but partially shielded by five loops. In the AKR superfamily, these A, B, and C loops are critical for substrate specificity (24,40). Two shorter loops, namely ␤1␣1 and ␤2␣2, contribute key active site residues (Fig. 4B).
Despite the low sequence identity when comparing CHR to other AKRs, CHR clearly exhibits an ␣/␤ structural core nearly identical to human 3␣-hydroxysteroid dehydrogenases (1.089 Å root mean square deviation for PDB accession code 1j96, (41)) and human and porcine aldose reductases (1.299 Å root mean square deviation for PDB code 1ef3, (42)). However, the CHR three-dimensional structure bears some striking differences within the loop regions responsible for substrate recognition in the AKR superfamily (Fig. 4C). Unlike other structurally characterized AKR family members, the tip of the CHR ␤1␣1 loop points toward the B loop, forming a clamp near the diphosphate portion of the bound NADP ϩ cofactor. The A loop bends toward the central C-terminal opening of the TIM barrel. Specifically, Phe-130 and Phe-133 located at the top of the A loop and His-120 and Trp-121 residing at the base of the A loop project inward in the direction of the active site situated in the center of the TIM barrel. The B loop of CHR is considerably shorter than structurally related AKRs, resulting in a more spacious opening of the TIM barrel near the C terminus. Finally, the top portion of the C loop also bends inward toward the center of the TIM barrel. Notably, Pro-301 orients downward and faces the active site defined by the bound NADP ϩ cofactor. Pro-301, together with Trp-89, Trp-121, Phe-130, Phe-132, and Ile-298 contribute to the overall hydrophobicity of the active site.
Nicotinamide Cofactor Binding Site-The experimental electron density map clearly shows a molecule of the oxidized nicotinamide cofactor, namely NADP ϩ , bound to the active site of each CHR monomer in the asymmetric unit of the CHR crystal. As in all structurally characterized members of the AKR superfamily described to date, NADP ϩ is sequestered in an extended anti-conformation within a small tunnel at the C-terminal end of the barrel, with the nicotinamide group protruding down into the core of the TIM barrel and the adenine moiety lying between the C-terminal ends of ␤7 and ␤8 (24, 40, 43) (Fig. 5). The A-face of the nicotinamide moiety is solvent exposed, consistent with transfer of the pro-R C4 hydride of NADPH to the substrate carbonyl carbon (43). The B-face of the nicotinamide moiety interacts with the -electrons of Phe-214 through a face-to-face stacking interaction. Notably, the aromatic nature of this stabilizing interaction is conserved in other AKRs, which typically maintain a Tyr at this position (Fig. 5). The central diphosphate portion of the NADP ϩ molecule lies in the tunnel defined by the ␤1␣1 loop (residues 30 -32) and the B loop (residues 220 -223, Fig. 4B). Subtle differences in residue identities and hydrogen bonding patterns occur across AKR subfamilies. As this is the first example of an AKR4 family structure, the CHR coordinates will be valuable tools for the investigation of AKR4 family NADPH binding sites.
The CHR tunnel is unique in that the ␤1␣1 and B loops clamp down tightly over the NADP(H) molecule, forming a narrow tunnel that translates toward the adenine moiety relative to other AKRs (Fig. 4C). Perhaps this tight clamp compensates for the fact that, like the 3␣-hydroxysteroid dehydrogenase subfamily, CHR lacks the "safety belt" of salt bridges over the NADP(H) molecule found in aldose and aldehyde reductases (43,44). A quantitative examination of the thermodynamics of NADP(H) binding and an investigation of CHR kinetic parameters will be necessary to define the physiological relevance of this seemingly tight structural clamp.
Active Site Architecture-The active site is formed at the C-terminal end of the TIM-barrel with the nicotinamide moiety and site of hydride transfer lying at the base of the catalytic cavity. The active site surface is defined by residues contributed from the variable loop regions of CHR (A, B, C, ␤1␣1, and ␤2␣2 loops). Predominantly hydrophobic and aromatic residues line the unoccupied entrance to the active site cavity, molded by Pro-29, Ala-57, Trp-89, Phe-130, and Phe-132. Largely polar residues define the base of this catalytic surface and include Asp-53, Tyr-58, Lys-87, His-120, Trp-121, and Asn-167 (Fig. 6).
The AKR family "catalytic tetrad" (Asp-53, Tyr-58, Lys-87, and His-120) is strictly conserved in the CHR primary sequence and spatially in the tertiary structure. Kinetically, the reaction mechanism in the AKR superfamily proceeds by an ordered Bi-Bi kinetic mechanism whereby NADPH binds first and the NADP ϩ product dissociates last (23). In the reductive direction, the pro-R C4 hydride of NADPH is transferred to the substrate carbonyl carbon. A conserved tyrosine residue in the active site (Tyr-58 in CHR) acts as a general acid to protonate the carbonyl oxygen as negative charge accumulates. In other AKRs, the pK a of this Tyr is lowered to near physiological pH are labeled as such to facilitate discussion, yet because of the prochirality of C6 and the unknown stereochemistry of the intramolecular cyclization reaction catalyzed by CHS, the precise origin of these carbons relative to the linear tetraketide-CoA is unknown, unlike carbons C3 and C6. by interaction with a Lys residue, thus facilitating proton donation. In CHR, the ordered hydrogen bonding network encompassing the putative general acid, Tyr-058, and Asp-53-Lys-87 provides the means to alter the pK a of the phenolic hydroxyl of Tyr-58 (43). Given the spatial conservation of key AKR active site residues, CHR-catalyzed reduction is likely to proceed via a similar NADPH-mediated hydride transfer mechanism. Identification of the Putative ␤-Keto Substrate Binding Site-To experimentally investigate the prevailing assumption that the true CHR substrate is a CoA-linked linear polyketide, we attempted to measure CoA binding to CHR. Utilizing both intrinsic fluorescence titration and isothermal titration calorimetry, we were unable to detect any CoA association with CHR even at concentrations exceeding 10 mM. This is in stark contrast to CHS for which CoA association was readily detected using intrinsic fluorescence titration (45) and to cinnamoyl-CoA reductase of monolignol biosynthesis for which CoA association was readily detected using isothermal titration calorimetry. 2 Although this lack of detectable CoA binding is not proof that CHR does not utilize a linear CoA-linked polyketide intermediate, it would be surprising that a highly functionalized and negatively charged carrier like CoA does not contribute to the thermodynamic association with CHR if the CHR substrate is tethered to CoA.
Pursuing another line of investigation, we also utilized computer-assisted docking of potential CHR substrates including coumaroyl-derived diketide-CoA, coumaroyl-derived triketide-CoA, and coumaryl-trione in the CHR active site. The results of these docking calculations revealed a highly favorable binding arrangement of the non-CoA-linked cyclized coumaryl-trione. This in silico result is consistent with the fact that the structure of the cyclized coumaryl-trione mirrors the chemical architecture of the CHR active site cavity, namely aromatic/ hydrophobic-rich near the coumaryl ring binding site (entrance) and polar in character along the cyclic polyketide (base). Furthermore, the puckered coumaryl-trione ring structurally resembles the carbohydrate skeleton of known AKR family substrates such as UDP-glucose (Fig. 7). Attempts at docking CoA and CoA-linked intermediates using identical procedures failed to predict any consistent putative binding mode. Analysis of the CHR structure revealed putative linear CoAlinked polyketide intermediates would not be sterically excluded from the CHR active site. However, attempts at docking CoA and CoA-linked intermediates using identical procedures revealed a set of random docking solutions and thus no consistent putative binding mode.
The docking results suggest that the cyclic polyketide ring adopts a modified boat conformation in the CHR active site, with the (C1) and (C5) ketones oriented downward and toward the NADP(H) cofactor and the C3 ketone oriented upward. The unknown chirality of C6 necessitates the arbitrary numbering of C1, C2, C4, and C5 (Fig. 6). The oxygen of the C1 trione carbonyl is within hydrogen bonding distance of both Tyr-58, the proposed general acid, and His-120, thought to be critical for orientating the substrate for reduction. The C3 ketone oxygen is poised to form a hydrogen bond with the indole ring NH of Trp-121 via an ordered water molecule present in the active site of our cofactor-bound structure. Moreover, the C5 ketone oxygen sits in close proximity (3. Given the inherent flexibility in loop regions, it is possible that a slight loop B movement would engage the trione via a hydrogen bond with the ⑀-amino group of Lys-219. Although our structure lacks a bound polyketide substrate, the backbone of Arg-223 is proximal to the modeled trione and the side chain ␦-guanido moiety is completely disordered. In the substratebound form of CHR, it is possible that Arg-223 is ordered and provides additional hydrogen bonds to the C5 ketone oxygen, further stabilizing the proposed cyclic trione substrate. CHR Provides Insight into the Three-dimensional Architecture of COR-The elucidation of the CHR x-ray crystal structure provides a more informed starting point for homology modeling of the closely related plant AKR, Papaver somniferum codeinone reductase (COR) (25), which shares 54% amino acid identity with M. sativa CHR. COR catalyzes the penultimate step in morphine biosynthesis, reduction of codeinone to codeine (Fig. 8) (25). Notably, the CHR and COR loops are highly conserved in length and related in sequence. Because these polypeptide segments contribute side chains to the active site, this modeling result is informative and likely to be functionally relevant, as a high degree of variation exists across AKR family members in both the identity and the number of residues forming these loop regions (24). The COR homology model reveals a highly conserved NADP(H) binding site bearing mainly conservative substitutions when compared with CHR. The exception to this trend is the substitution of an Ile residue at the position equivalent to Lys-219 in CHR, which in this latter case appears to interact with the adenine phosphate oxygen (nO2) and is also poised to interact with the CHR substrate (Fig. 5). This suggests that in COR hydrogen bonding interactions of the adenine phosphate oxygen, nO2, occur solely via the backbone amide of Ile-219, as opposed to CHR where both the backbone amide of Ser-215 and the ⑀-amino group of Lys-219 form hydrogen bonds with the nO2 phosphate oxygen.
As expected, the active site of COR bears the AKR catalytic tetrad consisting of Asp-Tyr-Lys-His, and most of the residues defining the active site cavity are identical to those in CHR (Fig. 9). There are, however, several interesting differences in the modeled COR active site when contrasted with CHR. For instance, several amino acid residue changes relative to CHR effectively widen the active site in COR. Although not extensive, these critical differences include the aforementioned Lys to Ile substitution at position 219 in CHR, along with a Trp to His substitution at the position equivalent to Trp-121 in CHR and a Phe to Asn substitution at the position equivalent to Phe-132 in CHR. These substitutions result in a more predictably spacious active site cavity in COR, as the COR substrate codeinone is slightly bulkier than the proposed CHR substrate coumaryl-trione (Fig. 8). DISCUSSION The utilization of the CHR structure as a template for homology modeling of a related enzyme of plant secondary metabolism, namely COR from P. somniferum, suggests a structurally satisfying explanation for the means by which COR has acquired the ability to accommodate its bulkier alkaloid substrate (Fig. 8). Moreover, the reasonably accurate model of COR allows for the mutational analysis of the residues predicted to act as key recognition elements in codeinone binding and catalytic turnover.
Outside of the CHR/COR branch of AKR4s, the precise physiological function of many AKR4 family members remains unknown. The absence of available AKR4 family structural information prior to the elucidation of the CHR structure described here, combined with the inherent ambiguities that accompany the use of more distantly related AKRs as structural templates for homology modeling, has hindered progress toward rationally probing the nature of the true physiological substrate of AKR4s. Utilizing the closely related CHR structure as a guide, a directed mutagenic approach of the substrate and cofactor binding pockets can now be employed to perturb specificities and evaluate the effect in vivo.
The elucidation of the high resolution CHR crystal structure and subsequent modeling of potential substrates enables comparative analysis of two evolutionary strategies used by Nature for polyketide reduction. SDR family ␤-keto reductases functioning in fatty acid and type I and type II polyketide metabolism have been extensively studied at both a structural and functional level (19). Recently reported crystal structures of the actIII polyketide reductase (20) and the ␤-ketoacyl-ACP reductase (16,17) reveal that these structurally and functionally related enzymes bind the NADP(H) cofactor via a Rossmann fold and rely on a catalytic triad of Ser, Tyr, and Lys to catalyze polyketide reduction. Unlike CHR, these enzymes possess tetrameric quaternary structures and possibly display allosteric regulation of catalysis via this tetrameric arrangement (17). Based upon sequence comparisons, it is clear that CHR is structurally distinct from the SDR ketoreductases and, as discussed previously, is a member of the AKR superfamily.
Finally, elucidation of the structure of CHR affords a new experimental direction for probing the nature of the poorly understood mechanisms surrounding the CHR reaction and its substrate specificity. The channeling of substrates-products between CHS and CHR has been proposed (46); however, the structural characterization of CHS reveals that direct overlap of the two active sites is highly unlikely (29). CHS bears only one entrance to the buried active site cavity; this narrow entrance, ϳ15 Å in length, sequesters the pantetheine arm of CoA to present acyl-CoA substrates and intermediates to the buried catalytic machinery of CHS. The architecture of CHS suggests that passive diffusion of acyl-CoA intermediates or the final cyclized trione intermediate between the CHS and CHR active sites may be the only conduit for product (CHS)/substrate (CHR) transfer. Therefore, if CHR acts on a linear polyketide intermediate (Fig. 1B, red arrows), then following CHS-catalyzed extension the linear polyketide must diffuse out of the CHS active site, bind to the CHR active site for reduction, and then return to the CHS active site for further catalysis. Alternatively, CHR reduction of the fully extended and CHS-cyclized trione (Fig. 1B, green arrow) would require only a single CHS/CHR diffusion event, followed by spontaneous aromatization in solution. Notably this latter scenario also eliminates the necessity for CHS to achieve the same Claisen cyclizationpromoting productive conformation of both the reduced and non-reduced linear tetraketide intermediate. This fact is an important consideration when contemplating the reactivity of the final tetraketide intermediate prior to CHS-catalyzed Claisen cyclization (3).
Given the key role diffusion likely plays in the CHR reaction, competing reaction steps may greatly influence the final product distribution, including the ratio of deoxychalcone to chalcone formed, the upper limit of which has been experimentally observed to be ϳ1:1 even in the presence of a great excess of CHR (30). Either reloading of polyketide-CoA intermediates onto the CHS catalytic cysteine or the spontaneous aromatization of the coumaryl-trione in the surrounding aqueous solution may compete with CHR binding and reduction and thus limit the yield of deoxychalcone. The presumption thus far has been that the true substrate of CHR is a linear polyketide-CoA intermediate (47). Yet to date, including the recent analysis by Oguro et al. (30), there is no experimental evidence for or against any of the three feasible CHR substrates (Fig. 1). The dismissal of the cyclized but prearomatic coumaryl-trione as a feasible CHR substrate is perhaps due to the inherently short lifetime of this molecule in aqueous solution. Indeed, the spontaneous aromatization of this intermediate is favorable; however, we must consider that the timescale of enzyme reactions is significantly shorter than that of chemical synthesis and purification.
Though our structural results cannot exclude the possibility that a linear coumaryl-derived polyketide-CoA intermediate is the substrate of CHR, the combination of our docking results, the structural similarity between coumaryl-trione and the carbohydrate skeleton of known AKR family substrates, chemical logic, as well as our inability to demonstrate CoA binding through two complementary energetic measurements, lead us to conclude that coumaryl-trione is the most likely substrate of CHR (Fig. 1B, green arrow). The docking result of this coumaryl-trione in the CHR active site predicts a highly favorable binding mode with extensive van der Waals interactions and hydrogen bonds. The structure presented here reveals the identity of active site residues that may play a key role in formation of the active site surface and in direct interactions with the coumaryl-trione substrate. Synthesis of various substrate analogs and subsequent experimental investigation of the CHR mechanism using a combination of x-ray crystallography, kinetic analyses, and binding studies will provide more definitive proof as to the nature of the CHR substrate. The structural work reported herein is a crucial step in this direction and supports a provocative yet largely overlooked hypothesis (3) for the recognition and turnover of a prearomatic chalcone product by CHR.