Engineering C–C Bond Cleavage Activity into a P450 Monooxygenase Enzyme

Abstract

The cytochrome P450 (CYP) superfamily of heme monooxygenases has demonstrated ability to facilitate hydroxylation, desaturation, sulfoxidation, epoxidation, heteroatom dealkylation, and carbon–carbon bond formation and cleavage (lyase) reactions. Seeking to study the carbon–carbon cleavage reaction of α-hydroxy ketones in mechanistic detail using a microbial P450, we synthesized α-hydroxy ketone probes based on the physiological substrate for a well-characterized benzoic acid metabolizing P450, CYP199A4. After observing low activity with wild-type CYP199A4, subsequent assays with an F182L mutant demonstrated enzyme-dependent C–C bond cleavage toward one of the α-hydroxy ketones. This C–C cleavage reaction was subject to an inverse kinetic solvent isotope effect analogous to that observed in the lyase activity of the human P450 CYP17A1, suggesting the involvement of a species earlier than Compound I in the catalytic cycle. Co-crystallization of F182L-CYP199A4 with this α-hydroxy ketone showed that the substrate bound in the active site with a preference for the (S)-enantiomer in a position which could mimic the topology of the lyase reaction in CYP17A1. Molecular dynamics simulations with an oxy-ferrous model of CYP199A4 revealed a displacement of the substrate to allow for oxygen binding and the formation of the lyase transition state proposed for CYP17A1. This demonstration that a correctly positioned α-hydroxy ketone substrate can realize lyase activity with an unusual inverse solvent isotope effect in an engineered microbial system opens the door for further detailed biophysical and structural characterization of CYP catalytic intermediates.

Graphical Abstract

INTRODUCTION

The cytochrome P450 (CYP) superfamily of heme-thiolate monooxygenase enzymes has evolved to efficiently insert an oxygen atom into the unreactive C–H bonds of organic molecules.¹ The majority of these enzymes do this by sourcing electrons from nicotinamide cofactors and O₂ from the atmosphere. These exceptionally versatile enzymes can also catalyze a myriad of other oxidation and reduction reactions including desaturation, sulfoxidation, epoxidation, heteroatom dealkylation, and C–C bond formation and cleavage reactions.^2–4 They have also been used as templates to design biocatalysts which can perform other reactions that do not occur in nature.^5–8 This enzymatic versatility makes it unsurprising that P450s have been identified in animals, archaea, bacteria, plants, fungi, and viruses^9,10 with over one million P450-encoding genes having been discovered.^9,11

These near ubiquitous P450s rely on a shared chemistry that undergirds the utility of these enzymes. CYPs progress through a catalytic cycle (Scheme 1) featuring a ferryl–oxo π-cation porphyrin radical (Compound I) that is usually responsible for substrate hydroxylation.^12–14 Starting the catalytic cycle from water-bound, six-coordinate, low-spin, oxidized iron, binding of a substrate frequently displaces the ligating water to form a five-coordinate high-spin iron.^12,13,15 This change in the iron’s spin-state results in an observable “Type I″ blue shift in the Soret from ~417 to ~390 nm.¹⁶ Continuing along the cycle, a reduction step generates iron(II), which is required for oxygen binding. This forms a ferrous–dioxygen complex or ferric–superoxide complex.^12,13,17 Supplying an additional electron yields the peroxoanion complex that, by sequential delivery of two protons, is prepared for the O–O bond cleavage.^12,13,17 Heterolytic O–O bond cleavage yields Compound I and water.^12,13,17 In hydroxylation reactions, Compound I abstracts a hydrogen atom to form a carbon-centered radical and an iron(IV)-hydroxy species. Oxygen rebound produces the hydroxylated product and ferric iron.^14,17

Scheme 1.

CYP Catalytic Cycle

Although this mechanism of catalysis through Compound I explains the abundance of CYP-catalyzed reactions including C–H hydroxylation, CC epoxidation, N-oxidation, and aryl C–C coupling, alternative intermediates are proposed to be involved in catalyzing other reactions.^12,13 For example, tryptophan nitration by a bacterial CYP involves reacting the ferric superoxide intermediate with nitrous oxide to generate the nitrating ferric-peroxynitrite species.¹⁸ The peroxoanion and hydroperoxy intermediates have also been implicated in catalysis, especially C–C breaking reactions.^12,13

These C–C breaking, or lyase, reactions often involve significant rearrangements of the substrate architecture, multistep oxidative transformations, and substrate fragmentation.¹⁹ This reshaping of the carbon skeleton greatly transforms the molecules’ bioactivities, providing access to compounds used for hormone signaling, enzyme cofactors, defense against predation, and beyond. The scope and importance of these C–C breaking reactions are exemplified in mammalian steroid biosynthesis wherein no fewer than four such reactions occur (Scheme 2):²⁰ 14α-demethylation of lanosterol (CYP51A1), side-chain cleavage of cholesterol to form pregnenolone (CYP11A1), 17α-hydroxylation/lyase reaction of pregnenolone or progesterone to form androgens (CYP17A1), and generation of estrogens by cleaving the C10–C19 bond and aromatizing the A-ring (CYP19A1).^21,22 The latter two reactions are of significant interest as targets for drug inhibition and can, for example, slow the growth of tumors in hormone-driven cancers.^23–25

Scheme 2.

A Variety of CYP-Mediated C–C Bond-Breaking Reactions Observed in Steroid Hormone Metabolism

The lyase reactions for these P450 systems require different patterns of oxidation around the carbon–carbon bond to be cleaved. Three major types of C–C bond cleavage reactions have been observed (Scheme 2). These involve cleavage of the C–C bond between a carbonyl group adjacent to a quaternary carbon, a carbonyl group adjacent to an alcohol moiety (e.g., α-hydroxy carbonyls or acyloins), or two adjacent hydroxy groups (diols).^26,27 These all occur at some point in mammalian steroid biosynthesis. At the beginning of the cholesterol biosynthetic pathway, CYP51A1 first hydroxylates lanosterol at the C14 position to form an alcohol; this position is further oxidized to generate an aldehyde through gem-diol formation. The aldehyde carbon center then undergoes C–C cleavage to form a D-ring alkene and formic acid (Scheme 2).^19,28 CYP11A1 carries out sequential oxidation steps to produce a vicinal-diol center of which the bond between the hydroxylated centers is cleaved by CYP to form pregnenolone and an isocaproaldehyde (4-methylpentanal; Scheme 2).^19,29,30 Pregnenolone is a substrate of CYP17A1, whereby this P450 will first hydroxylate its C17 center to generate an α-hydroxy ketone moiety which is then cleaved by CYP17A1 to form a new ketone center at C17 and acetic acid (Scheme 2).^28,31 CYP19A1 operates to aromatize the A-ring with two hydroxylations at the C-19 carbon to generate an aldehyde, followed by a third monooxygenase reaction that cleaves the C10–C19 bond.

The P450 oxidants and reaction mechanisms involved in these C–C cleavage reactions are proposed to vary.³² Detailed reviews of the C–C formation and cleavage reactions by the CYPs have recently appeared.^31,33 CYP51A1 is proposed to catalyze the C–C cleavage reaction of a 14α-formyl-oxy species via a Baeyer–Villiger-like mechanism.^27,28 This arises from the addition of a P450 ferric-peroxo intermediate to the electrophilic aldehyde center.^19,28 An alternative reaction mechanism involving homolytic cleavage to form a radical intermediate that will undergo decomposition to form the cleavage product has also been suggested.³⁴ For CYP11A1, both the hydroxylation and C–C cleavage steps are thought to be catalyzed by Compound I, although the exact pathway has not been ascertained.^19,35–37 In CYP17A1, it is proposed that the ferric-peroxo form of CYP17A1 attacks the C═O center of the α-hydroxy carbonyl group to form a peroxo adduct.^33,38 This peroxo adduct can then undergo Baeyer–Villiger-type oxidative cleavage reactions to form the final ketone product and acetic acid.^27,39 Evidence for the involvement of the ferric peroxo species in CYP17A1 comes from the observation of an unusual inverse solvent isotope effect in the lyase reaction for both 17-hydroxy pregnenolone and 17-hydroxy progesterone.⁴⁰ The trapping of the peroxoanion at low temperatures and the formation of a hemiketal transition state characterized by optical and resonance Raman spectroscopy have bolstered this hypothesis.^41,42

These mammalian steroid-metabolizing CYPs are integral membrane proteins, and mechanistic and structural investigations have struggled with difficulties in obtaining a high level of heterologous protein production, purification of the P450s and their redox partners, eliciting crystal growth, and dealing with a low turnover and highly uncoupled systems. Realizing the detailed and historic biophysical success using microbial CYPs, particularly P450cam^43,44 and P450BM3,⁴⁵ we sought to reconstitute the C–C lyase reaction of CYP17A1 in a model P450 system. This could allow detailed low-temperature studies and, hopefully, eventual structural isolation of intermediates.

The bacterial enzyme CYP199A4 from Rhodopseudomonas palustris strain HaA2 has proven to be a useful system for the study of the CYP mechanism and function.^46–50 It displays high catalytic activity for the oxidation of para-substituted benzoic acids such as 4-methoxybenzoic acid.^49,51,52 The enzyme catalyzes the O-demethylation of this substrate into 4-hydroxybenzoic acid, and the X-ray crystal structures of the enzyme and its amino acid variants have been determined. These structures have been used to elucidate binding modes and to rationalize the observed activities.^46–49,51 The CYP199A4 enzyme oxidizes its substrates almost exclusively at the substituent para to the carboxylate group and hence is an ideal model system to investigate different aspects of the mechanisms of P450 reactions.⁵² For example, we have used the CYP199A4 enzyme to investigate catalytic epoxidation, sulfoxidation, and aromatic oxidation by CYP enzymes.^53–55 In addition, it can hydroxylate and desaturate the alkyl substituents of substrates such as 4-n-propylbenzoic acid.⁵⁶

Herein, we report the synthesis and use of suitable para-substituted benzoic acids to probe the mechanisms of hydroxylation and C–C cleavage (Figure 1). These substrates exploit the high degree of selectivity of CYP199A4 for oxidation at functional groups at the para-position and providing an α-hydroxy ketone moiety that resembles that of the 17-hydroxy steroids. We hypothesized that these would serve as models for CYP17A1 catalysis if oxidation by CYP199A4 generated C–C cleavage metabolites. Since active-site positioning of the heme-bound peroxoanion and the carbonyl is critical, we include the X-ray structures of the protein bound to substrates that undergo C–C cleavage and molecular dynamics (MD) simulations.

Figure 1.

Benzoic acid substrates with a carbonyl-containing substituent investigated with CYP199A4. JCM1 and JCM2 contain an α-hydroxy ketone moiety at the para-position. 4-Acetylbenzoic acid, which contains only a ketone moiety, is used as a control.

RESULTS

Synthesis of α-Hydroxy Ketone Substrates.

A two-step route to a racemic 4-(1′-hydroxy-1′-methyl-2′-oxopropyl)benzoic acid (JCM1) was proposed with metal-halogen exchange using an isopropylmagnesium chloride–lithium chloride reagent and methyl 4-halobenzooate (Scheme 3) to produce an aryl Grignard in situ. This nucleophile, when reacted with neat 2,3-butadione, would yield an α-hydroxy ketone. The nucleophilic addition to 2,3-butadione worked better with methyl 4-iodobenzoate than the 4-bromo equivalent and resulted in the desired methyl ester of the α-hydroxy ketone. Ester hydrolysis with lithium hydroxide produced the target JCM1 as a racemic mixture at 95% purity; reverse-phase flash chromatography yielded racemic JCM1 in 99+% purity.

Scheme 3.

Synthesis of α-Hydroxy Ketones onto the para-Benzoic Acid Scaffold

A three-step synthesis was devised to make 4-(1′-oxo-2′-hydroxy-2′-methylpropyl)benzoic acid (JCM2), which switches the positions of the hydroxyl and ketone moieties compared to JCM1 (Scheme 3). In the first step, a Wittig olefination with isopropyltriphenylphosphonium iodide and methyl 4-formylbenzoate was used to generate the predicted alkene in 41% yield. Ruthenium tetroxide was generated in situ and used to oxidize the alkene to the corresponding desired α-hydroxy ketone (25% yield). Ester hydrolysis was carried out as described above to generate JCM2 in 87% yield.

Both JCM1 and JCM2 were expected to bind in the active site of CYP199A4 and be oxidized at the para-substituent. This oxidation could result in hydroxylation or C–C bond cleavage metabolites (Scheme 4). In addition, 4-acetylbenzoic acid, which contains a ketone group but not the α-hydroxyl functionality, was chosen as a control substrate. Assays with this compound would interrogate how the addition of an electron-withdrawing, polar group to the substrate alters binding and activity compared to the α-hydroxy ketones. This substrate was not expected to undergo lyase activity but could be hydroxylated at the terminal methyl (Figure 1 and Scheme 4).

Scheme 4.

Potential Metabolites That Could Arise from CYP199A4-Catalyzed Oxidation of JCM1, JCM2, and 4-Acetylbenzoic Acid

Wild-Type CYP199A4 Is Minimally Active toward α-Hydroxy Ketones.

The addition of JCM1 and JCM2 did not significantly alter the UV–vis spectrum of the wild-type (WT) CYP199A4 enzyme (Table 1 and Figure S1). This lack of a spectroscopic shift indicates that CYP199A4 likely does not readily bind JCM1 or JCM2, or if it does, it does not displace the heme-bound water ligand. In contrast, the addition of the ketone analogue, 4-acetylbenzoic acid, to CYP199A4 resulted in a Type I difference spectrum with a characteristic peak at ~390 nm and a trough at ~417 nm (Figure 2). However, the substrate-induced spectroscopic shift was minimal (Figures 2 and S2). 4-Acetylbenzoic acid bound to CYP199A4 with significantly lower shift to the high-spin ferric state and binding affinity (K_d of 140 ± 5 μM) than 4-n-propylbenzoic acid (≥ 95% high-spin, 0.54 μM) and other comparable alkyl-substituted benzoic acids (Table 1).^50,56

Scheme 7.

Synthesis of 4-(2′-Oxoacetyl)benzoic Acid from 4-Acetylbenzoic Acid

Figure 2.

Spin-state shift (a), dissociation constant determination (b), and HPLC analysis of the oxidation of 4-acetylbenzoic acid with CYP199A4 with and without using alcohol dehydrogenase (ADH) to regenerate NADH (c). As the substrate absorbs strongly at ~250 nm and a shoulder of this band interferes with the increase in absorbance at 390 nm, only the decrease in absorbance at 420 nm was used in the determination of the dissociation constant for 4-acetylbenzoic acid.

The rate of NADH oxidation by WT CYP199A4 in the presence 4-acetylbenzoic acid was 407 μmol (μmol CYP199A4)⁻¹ min⁻¹ (henceforth abbreviated as min⁻¹). The oxidation of 4-acetylbenzoic acid proceeded with a product formation rate of 260 min⁻¹ and a coupling efficiency of 65% (Table 1). A single major product was observed in the turnover of 4-acetylbenzoic acid (Figure 2). After generation of this product using a whole-cell oxidation system and isolation and purification by prep-scale HPLC, this compound was identified by ¹H NMR as 4-(2′-hydroxyacetyl)benzoic acid, where a characteristic singlet integrating for two protons at 4.82 ppm was observed (Figure S3). This metabolite also has a mass spectrum consistent with a single hydroxylation (Figure S3). Thus, although the addition of 4-acetylbenzoic acid to CYP199A4 resulted in a small shift to the high-spin ferric state, the methyl group must be suitably positioned for oxidation.

The CYP199A4-catalyzed oxidation of JCM1 and JCM2 by WT CYP199A4 was investigated using a variety of experimental conditions. The NADH oxidation rates observed185 and 33 min⁻¹, respectivelywere lower than those obtained with 4-acetyl- and 4-alkylbenzoic acids (Table 1). The amount of any potential oxidized metabolites observed by HPLC or GC–MS analysis was also low (Figure S4 and Scheme 4). The oxidation reactions of JCM2 contained a small amount of terephthalic acid, the expected metabolite from sequential C–C bond cleavage reactions. However, low levels of 4-formylbenzoic acid were present as an impurity, so control reactions were carried out to assess if the metabolite, terephthalic acid, arose from enzymatic activity. In these control reactions, we observed that relatively low levels of hydrogen peroxide were able to generate terephthalic acid from JCM2 (Figure S5), discounting enzyme-catalyzed C–C bond cleavage.

The synthesized JCM1 (95%) contained a small amount of 4-acetylbenzoic acid impurity which complicated the analysis (Figure S6). Using 99+% pure JCM1 ensured that all 4-acetylbenzoic acid had been removed, and in vitro oxidations with WT CYP199A4 and control reactions confirmed that in the WT, the level of the JCM1 substrate did not decrease during these reactions and that the enzyme was not generating any metabolites from C–C bond cleavage.

Site-Directed Mutagenesis of CYP199A4 to Catalyze C–C Bond Cleavage Reactions.

We hypothesized that mutation of the CYP199A4 enzyme may enable the JCM1 and JCM2 substrates to alter their position within the active site of the enzyme to better enable catalytic C–C bond cleavage. The F182L and F298V mutations were chosen as, in previous work, they have demonstrated alterations in the active-site orientation of other substrates, as confirmed through X-ray structures and MD simulations.^56,57 These changes enabled the oxidation of substrates which were not oxidized by the WT enzyme or resulted in an altered product distribution.

We tested the F182L and F298V variants of the CYP199A4 enzyme. With the F182L variant, a ~60% shift to the high-spin form upon the addition of JCM1 was observed (Figure 3, Table 1). This spectroscopic shift is a good indication that JCM1 binds and displaces the water ligand in this protein (Figure S7). A greater increase in the shift to a high-spin ferric species (~80%) was also observed in this variant with 4-acetylbenzoic acid (Table 1 and Figure 3). A less extensive shift to the high-spin form of the F182L variant (~25%) was observed after the addition of JCM2 (Table 1 and Figure S8). With the F298V variant, a red shift in the Soret band (a Type II difference spectrum) was observed with JCM1 and no significant changes were observed with JCM2 (Figures S8 and S9).

Figure 3.

Spin-state shift analysis of F182L-CYP199A4 with JCM1 (a) and 4-acetylbenzoic acid (b) as substrates.

Encouraged by these results, we tested the oxidation of these substrates with both variant enzymes. The NADH oxidation rate of the F182L in the presence of JCM2 was exceptionally fast at >1400 min⁻¹ (Table 1 and Figure S10). However, the analysis of the oxidation and control reactions (without enzyme) established that JCM2 could react with hydrogen peroxide to generate terephthalic acid (Figures S10 and S11). Given these observations, we decided to focus on JCM1 as a candidate for investigating C–C bond cleavage by CYP199A4.

Despite the larger substrate-induced spin shift upon the addition of JCM1 to the F182L variant of CYP199A4, the measured rate of NADH oxidation was slower than for the WT enzyme (Table 1 and Figure S14). HPLC analysis of the enzyme-catalyzed oxidations of JCM1 demonstrated that there was significantly more of the 4-(2′-hydroxyacetyl)benzoic acid metabolite generated in the reactions with F182L compared to the F298V variant (Figures 4, S15). This product could arise from a C–C bond cleavage reaction to generate 4-acetylbenzoic acid that was subsequently hydroxylated or potentially from hydroxylation preceding C–C bond cleavage. Other metabolites were also generated that were assigned by GC–MS to hydroxylation products of JCM1 (Figure S16). There were also metabolites which were considered as arising from further oxidation of 4-(2′-hydroxyacetyl)benzoic acid via control reactions with 4-acetylbenzoic acid (Figures 4 and S17–S19). One of these was identified as an overoxidation metabolite, yielding a glyoxal ($ in Figure 4 and Scheme 5, Figures S20 and S21). These additional metabolites also included low amounts of terephthalic acid, which could be derived from a second C–C bond cleavage reaction of 4-(2′-hydroxyacetyl)benzoic acid or a subsequent oxidation metabolite. Importantly, the formation of these metabolites was accompanied by a decrease in the amount of substrate, and JCM1 did not react with hydrogen peroxide in the absence of the enzyme. Overall, these data suggest that JCM1 was undergoing an enzyme-catalyzed C–C bond cleavage reaction to generate 4-acetylbenzoic acid which was then further oxidized by the enzyme to the observed metabolites (Scheme 5). As 4-acetylbenzoic acid is a better substrate for the enzyme than JCM1, it is preferably oxidized when formed and its accumulation is not observed in downstream analysis.

Figure 4.

HPLC analysis of the oxidation reactions of purified JCM1 by the F182L variant of CYP199A4 using an excess of NADH (1 mM). Included for comparison is a control reaction without any P450 enzyme present. The peak labeled with the symbol “$” was identified as an aldehyde over-oxidation metabolite (Scheme 5).

Scheme 5.

Products Identified and Proposed Additional Metabolites Which Arise from the CYP199A4-Catalyzed Oxidation of JCM1

C–C Cleaving Reactions Subject to the Inverse Kinetic Solvent Isotope Effect.

Having established that the F182L variant of CYP199A4 catalyzes a C–C cleavage reaction with JCM1, we set out to investigate the mechanism of this reaction by comparing the activity of the enzyme in H₂O and D₂O. In principle, if the lyase reaction is catalyzed by an earlier intermediate of the P450 catalytic cycle that does not require the protonation steps needed to form Compound I, an inverse kinetic solvent isotope effect (KSIE) is expected. This phenomenon has been reported for the human CYP17A1.^40,42 Oxidation reactions were carried out for JCM1 and 4-acetylbenzoic acid in deuterated and protonated solvents. If the 4-(2′-hydroxyacetyl)benzoic acid produced in this reaction involves Compound I, then these reactions should result in a normal solvent isotope effect for this substrate, i.e., more product in H₂O than D₂O.

Interestingly, the oxidation of JCM1, which involved C–C bond cleavage followed by C–H bond hydroxylation, occurred with higher product formation and coupling efficiency in D₂Oan inverse isotope effect (Table 2). There was also a decrease in the amount of the metabolite proposed to arise from the hydroxylation of JCM1 relative to the metabolites which arise from the C–C bond cleavage step (Figure S22). When the levels of metabolites that arise from C–C bond cleavage (4-(2′-hydroxyacetyl)benzoic acid) were compared to those arising only from hydroxylation, the ratios between the H₂O and D₂O reactions were 0.77 (range 0.68–0.82) and 1.56 (range 1.37–1.71), respectively (Figure S22). This is indicative of an inverse KSIE and normal KSIE for the C–C cleavage and hydroxylation pathways.^40,42

Table 2.

Kinetic Parameters Determined for JCM1 and 4-Acetylbenzoic Acid with the F182L Variant of CYP199A4^d

substrate	NADHa	PFRb	C (%)c
JCM1-F182L H2O	68 ± 1	1.3 ± 0.2	1.9 ± 0.3
JCM1-F182L D2O	74 ± 9	1.8 ± 0.4	2.4 ± 0.2
4-acetylBA-F182L H2O	488 ± 9	414 ± 0.1	85 ± 2
4-acetylBA-F182L D2O	317 ± 11	308 ± 18	97 ± 2
4-acetylBA-WT H2O	265 ± 20	213 ± 21	80 ± 2
4-acetylBA-WT D2O	291 ± 16	210 ± 7	74 ± 2

^aNADH oxidation rate.

^bPFR: product formation rate.

^cCoupling efficiency, i.e., the percentage of NADH consumed that led to the formation of the oxidized metabolite. For JCM1 and 4-acetylbenzoic acid, this was based on the amount of 4-(2‘-hydroxyacetyl)benzoic acid and related metabolites in these assays (see Figure S22).

^dRates are in (mol CYP)⁻¹ min⁻¹ which is abbreviated to min⁻¹ in the text.

In contrast, the hydroxylation of 4-acetylbenzoic acid with the F182L mutant of CYP199A4 occurred more slowly in D₂O but proceeded with a slightly increased coupling efficiency (Table 2). With the WT CYP199A4 enzyme, the rate of NADH oxidation was marginally faster in D₂O, but the coupling efficiency was lower, resulting in a similar product formation rate (PFR). These results strongly suggest that an intermediate of the catalytic cycle other than Compound I is facilitating the C–C bond cleavage reactions of CYP199A4.

Crystal Structure of JCM1 Bound to the F182L Mutant of CYP199A4.

We were able to crystallize the CYP199A4 F182L variant in the presence of both JCM1 and JCM2 substrates and to determine the X-ray crystal structures for each. The crystal structures were resolved at 2.05 Å (JCM1; PDB code 8G35) and 2.16 Å (JCM2; PDB code 8G36) resolution, respectively (Figure 5 and Table S1). The overall structural fold of the CYP199A4 enzyme in both structures was similar to those obtained previously with the CYP199A4 enzyme (Figure S23, rmsd values of 0.492 and 0.597 Å, respectively, for the structure of WT CYP199A4 with 4-methoxybenzoic acid, PDB 4DO1). The amino acids in the substrate-binding pocket and the active-site water molecule which interacts with the carboxylate group were in similar positions, unless otherwise noted.⁵¹

Figure 5.

(a) Active-site structure of F182L-CYP199A4 (green sticks) complexed with the (S)-isomer of JCM1 (yellow sticks). A feature-enhanced map as a gray mesh contoured to 1.0 (1.5 Å carve) is shown around the substrate, residue L182, and active-site waters. (b) Crystal structure of F182L-CYP199A4 modeled with (R)-JCM1 (cyan sticks) and discernible water molecules present within the active site. A composite omit map (gray mesh, 1.0 σ at 1.5 Å carve) and an F_O–Fc map (red or green mesh, contoured at 2.5 σ) are shown. See the Supporting Information for further details. (c) Active-site complex of F182L-CYP199A4 and (S)-JCM1 (green sticks) superimposed with that of WT-CYP199A4 and 4-methoxybenzoic acid (cyan sticks). F298 has shifted to accommodate the additional steric bulk of (S)-JCM1. Panels (d) and (e) show distances in Å between different residues of the oxygen-binding groove of F182L and WT-CYP199A4, respectively. It can be seen that the groove has widened in the F182L mutant, allowing additional water molecules into the active site.

The electron density maps revealed the location of a ligand within the active site of each structure (Figures 5 and S24 and S25). The position of the benzoic acid moieties of both were essentially the same as those of other substrate-bound X-ray crystal CYP199A4 structures (Figure 5).^{49,51,55–58} The electron density for the ligand in the active site of the structure with JCM2 did not match that of the added substrate. It was modeled as terephthalic acid with additional water molecules within the active site (Figure S24 and Table S2). The presence of terephthalic acid instead of JCM2 highlights the sensitivity of JCM2 to the C–C bond cleavage reaction which seems to occur over the time frame of crystal growth, storage, or during data collection.

The conformations of the active-site residues of the F182L mutant in the presence of JCM1 are similar to those of WT CYP199A4, except that the F298 residue orients away from the substrate and faces the plane of the heme group to accommodate the additional steric bulk of the substrate (Figures 5 and S25). A similar movement of the F298 residue is also observed when CYP199A4 binds sterically demanding substrates.^49,54,56 The structure of F182L-CYP199A4 with JCM1 required the electron density of the bound molecule to be modeled using the (S) and (R) enantiomers of JCM1. In the structures, the active site contained additional regions of electron density that were modeled as water molecules (Figure 5, Table S3, and Figure S25 and S26). F_O–F_C difference maps were generated to assess negative and positive electron density around the bound molecule. These maps show where the electron density and the model disagree (Figure 5). Negative densities (red) show atoms in the model where there is lower electron density of the modeled substrate compared to that in the crystal structure, whereas positive densities (green) indicate electron density which has not been accounted for in the model. When the (R)-enantiomer of JCM1 was modeled into the structure of F182L-CYP199A4, the F_O–F_C maps showed a distinct region of positive density above the heme (Figure 5). This positive density is likely to be a heme-bound water molecule. An attempt was made to model in the heme-bound water, but several refinement cycles using “phenix.refine” displaced the modeled water molecule away from the heme. It is likely this water molecule, if present, would clash with the heme-facing methyl group of the (R)-JCM1. There is also negative density present on the substrate’s α-hydroxy group, indicating that insufficient electron density was present in that region.

When the (S)-enantiomer was modeled, a similar region of positive density was observed above the heme (Figure S25a), which was successfully assigned as a water molecule (W289; Figures 5 and S25 and S26). The substrate’s α-hydroxy group hydrogen bonds to this aqua ligand and W267 (Figure S26). The I-helix of the F182L mutant, when compared to the WT enzyme, contained an additional water molecule (W41) close to the oxygen-binding groove (Figure S27). This water molecule forms a hydrogen-bonding interaction with T253 which occupies a different orientation when compared to the structures of the WT enzyme (Figure S27). The side chains of residues A248, L250, and T252 also showed a change in conformation in the F182L mutant when compared to the WT enzyme. Residue T252 in the F182L mutant is displaced away from the heme-center to accommodate the additional water molecules (Figures 5 and S27). The movement of the T252 residue also caused a widening of the oxygen-binding groove in the F182L mutant (Figure 5).

When the occupancies of both enantiomers of JCM1 were refined within the same location in the active site using separate alternative locations (altloc identifiers) for each enantiomer, their occupancies were 11% for the (R)-enantiomer and 89% for the (S)-enantiomer (Figure S28). The (S)-enantiomer is therefore likely to be the preferred substrate for the enzyme in solution. The smaller occupancy of the (R)-enantiomer is consistent with the minor hydroxylation of JCM1 at the β-methyl observed in our assays (Figures S16 and S18). The greater occupancy of the (S)-enantiomer likewise correlates with the C–C lyase activity that results in the major metabolite observed in the enzyme-catalyzed reactions.

When JCM1 and the active-site waters were superimposed into the substrate-binding site of the WT-CYP199A4 (PDB: 4DO1), the α- and β-methyl groups of JCM1 clashed with residues F182 and F298 (Figures S29 and S30). The additional water molecules that were present in F182L-CYP199A4 (W107, W41, and W267) also had unfavorable interactions with residues T252 and F182 (Figures S29 and S30). This provides a rationale for the inability of the WT enzyme to bind and oxidize these substrates and demonstrates how the additional space in the active site afforded by the F182L mutation accommodates the binding of JCM1. We also note that these structures demonstrate that substrates with more polar carbonyl groups at the para-position of the benzoic acid can be accommodated close to the heme in the active site of these mutants of CYP199A4.

MD Simulations of the CYP199A4 Oxy-Ferrous JCM1 Complex.

The structure of the F182L oxy-ferrous CYP199A4-JCM1 complex was assessed in silico by docking and MD simulations into an oxy-ferrous model of CYP199A4. Because no crystal structures of the heme oxy-complex were available for CYP199A4, the first step was the creation of a CYP199A4 oxy-complex model. Aligning the hemes from the structures of the CYP101A1 oxy-complex⁴³ and a high-resolution substrate-bound CYP199A4 structure (PDB: 5UVB) enabled a reasonable placement of the oxy-complex into this enzyme.⁵⁶ The 4-cyclopropylbenzoate substrate in the original structure was removed before separately docking two different enantiomers of JCM1, (R)-JCM1 and (S)-JCM1, into the active site using methods described for the CYP homology models.⁵⁹ After equilibrating the respective α-hydroxy ketone-bound structures, MD runs of over five μs for both enantiomers were performed.

First, in both models, the substrate is displaced 1–2 Å further away from the Fe center compared to the ferric crystal structure, thus making room for the bound di-oxygen (Figure S31 and Table S4). Second, in the F182L models, the active site is accessible to water, consistent with the water observed in the ferric crystal structure of JCM1. Figure S32 shows the volume occupied by water for a minimum of 25% of the 5 μs simulation bound to (R)-JCM1 and (S)-JCM1. It is worth noting that the water accessibility in the presence of the (R) and (S) enantiomer differs. This is a direct result of a stable hydrogen bond formed between the hydroxyl group on (R)-JCM1 and the backbone carbonyl of A248 (Figure S33). In the (S)-JCM1 simulation, the α-hydroxy ketone group is free to rotate about the bond connecting it to the para-position of the benzoic acid moiety, thus causing the α and β-methyl groups to displace water and resulting in the lower water occupancy.

In addition to these substrate and water displacements, the opening of the oxygen groove, as seen in the ferric crystal structure, is certainly evident in the (R) enantiomer simulation. This more open groove is stabilized by the hydrogen bond with the substrate’s hydroxyl group and the backbone carbonyl of A248. Quite surprisingly, in the simulation of (S)-JCM1, the oxygen-binding groove narrows compared to that shown in Figure 5d and closely resembles the WT structure (Figure S34). The absence of F182 or the presence of the hydrogen bond between the substrate and A248 prevents the narrowing of the oxygen groove. Thus, the oxy-CYP199A (S)-JCM1 simulation suggests a slightly more dynamic substrate and active site, possibly allowing for catalysis of the lyase reaction. The presence of the hydrogen bond in the (R)-JCM1 model would certainly prohibit the formation of the lyase transition state proposed by Olsen (Scheme 6).⁶⁰ The lyase activity of F182L CYP199A4 toward the (R) and (S) enantiomers is being studied further.

Scheme 6.

Potential Mechanism of C–C Bond Cleavage of JCM 1 by the Ferric Peroxyanion Intermediate of CYP199A4 Based on the Results Reported Here and the Calculations of Olsen and Co-workers⁶⁰

DISCUSSION

The oxidative cleavage of carbon–carbon bonds by the cytochromes P450 is widely distributed but perhaps most investigated in the metabolism of steroids. In many cases, the chemical structure of the substrate allows for a hydrogen abstraction to rationalize the scission event, the hallmark of Compound I involvement. Such is the case in the formation of pregnenolone from cholesterol via the scission of the vicinal diol in 20,22-dihydroxycholesterol (Scheme 2). This reaction could occur either by the removal of the two hydroxyl hydrogens, formation of a dioxetane intermediate, or abstraction of the 22-hydrogen and radical scission. In the case of aromatase (CYP19A1), an aldehyde is formed at carbon 19, and the remaining hydrogen is available for abstraction. The removal of this hydrogen is thermodynamically driven by the aromatization of the A-ring, and a Compound I mechanism is strongly supported.^61,62 In the case of CYP17A1, which cleaves an α-hydroxy ketone, no easily identifiable nearby C–H bond for hydrogen abstraction is present. The suggestion that abstraction of the 17-hydroxy hydrogen commences the reaction does not carry the thermodynamic advantage of aromatization seen in CYP19A1 and has been discounted by QM/MM calculations.^60,63 Given the apparent unique challenges afforded by an α-hydroxy ketone structure, we undertook the synthesis of a potential substrate for the bacterial CYP199A4 that contained such a substructure. This enzyme prefers benzoic acid substrates, and we envisioned that the interaction of the aromatic carboxylate para to the metabolic site might position a suitable functional group near a heme peroxoanion for catalysis. This positioning would allow dissection of the intermediates in the P450 active site responsible for eliciting a C–C bond breakage.

The cleavage of C–C bonds of α-hydroxy carbonyls has been proposed to occur via the ferric peroxo intermediate rather than Compound I.⁴¹ We designed a substrate, JCM1, that provides an α-hydroxy ketone positioned in the active site above the heme of the bacterial enzyme CYP199A4. We then found a mutant of CYP199A4 which would position the α-hydroxy ketone in the active site favorably for peroxoanion attack and production of 4-acetylbenzoic acid. We observed products arising from the C–C cleavage of the α-hydroxy carbonyl substrate JCM1 with the F182L mutant of CYP199A4. Subsequent metabolism of the acetyl methyl moiety arising from initial scission of the α-hydroxy carbonyl can form an alcohol and then an aldehyde and potentially terephthalic acid through another C–C cleavage reaction.

The cleavage of the hydroxy ketone side chain of JCM1 to yield 4-acetylbenzoic acid could proceed via the Compound I- or a peroxoanion-mediated mechanism. Compound I is generated over several steps through reduction of the ferrous dioxygen state to form a peroxoanion (Scheme 1). This intermediate then undergoes protonation to form the hydroperoxo, and then, a second protonation at the distal oxygen facilitated cleavage of the O–O bond, producing water and the high-valent ferryl-porphyrin cation radical. These two proton transfer events are slowed in the deuterated solvent, as has been described in several P450 systems.⁴⁰ In addition to productive Compound I formation, most CYPs are uncoupled, wherein the reducing equivalents from the nicotinamide cofactor yield hydrogen peroxide or, via additional reduction of Compound I, a second water molecule. If, on the other hand, a peroxoanion is the key intermediate involved in C–C bond scission (Scheme 6), then the committed step to catalysis is before any proton transfer. Subsequent proton transfer steps to generate hydrogen peroxide or additional waters then represent an uncoupling of the catalytic step. If this uncoupling is slowed by the use of D₂O instead of H₂O, then the unproductive steps are slowed, the peroxoanion will build up, and a faster product forming reaction is observed in the deuterium solvent: an inverse solvent isotope effect. This is indeed the case in the generation of the androgens by human CYP17A1⁴⁰ and in the C–C cleavage of JCM1 by F182L CYP199A4. We show that WT and F182L CYP199A4 catalyzed oxidation of 4-acetylbenzoic acid with a small normal solvent isotope effect (KSIE). Yet, when the substrate is the α-hydroxy ketone JCM1, the observed overall solvent isotope effect is inverse: more product is observed in D₂O than H₂O. This increase of product formation from JCM1 in the presence of D₂O strongly supports the hypothesis that C–C bond scission can proceed using the peroxoanion intermediate but other oxidations by this enzyme proceed with a normal solvent isotope effect as expected for a Compound I-mediated process. CYP17A1 lyase activity has been demonstrated to have inverse KSIEs ranging from 0.39 to 0.8, whereas a normal KSIE of up to 1.8 has been observed for the second electron transfer step of P450cam.^40,42 It is important to note that these KSIEs can be completely or partially masked in the overall catalytic cycle if other steps are rate-limiting.

The study of these and other hydroxy carbonyl or model compounds, such as diols or aldehydes, as substrates for CYP199A4-catalyzed oxidations combined with crystallography could establish the potential of this enzyme to catalyze carbon–carbon bond-breaking reactions. We note that another system where carbon–carbon bond cleavage might be anticipated is in the oxidation of fatty acids by P450BM3, but when using this system, no products could be detected, which resulted from carbon–carbon bond cleavage of the hydroxy carbonyl moiety.^64,65 The P450 enzymes which have evolved to catalyze C–C bond cleavage reactions, such as P450BioI (CYP107H1)^66–68 and CYP17A1,^20,41,69 may have very specific requirements which are not replicated in the WT CYP199A4 or P450BM3 systems with their normal substrates. For example, P450BioI requires an acyl carrier protein-bound substrate for optimal positioning of the relevant C–H bonds and C–C bond cleavage.⁶⁶ CYP17A1 also requires specific interactions with other partner proteins, such as cytochrome b5, for optimal lyase activity.^{20,41,69–71}

CONCLUSIONS

In summary, we have shown that by the judicious design of the substrate, together with its correct positioning in the active site, peroxoanion-mediated cleavage of an α-hydroxy ketone ensues. This confirmation that C–C cleavage reactions can be modeled outside of the P450s that catalyze them as part of their native biological functioning provides a platform with which to interrogate the peculiarities of these transformations in greater detail.

EXPERIMENTAL SECTION

Synthesis of α-Hydroxy Ketones JCM1 and JCM2.

All reagents and solvents were purchased from commercial suppliers and used without further purification unless otherwise noted. Hexanes were vacuum-distilled to remove high-boiling impurities. When dry solvents were required, they were dried for two or more days over activated 3 Å molecular sieves and stored under dry nitrogen gas. Thin-layer chromatography utilized silica gel 60 doped with a fluorophore excited at 254 nm. Plates were visualized by fluorescent quenching and KMnO₄ staining. Flash column chromatography was conducted with 40–63 μm silica gel. ¹H NMR and ¹³C NMR spectra were recorded on a 500 MHz spectrometer at the University of Illinois School of Chemical Sciences NMR Spectroscopy Lab. NMR samples were made as solutions in CDCl₃ or CD₃OD using the residual solvent peak as the internal standard; δ values are given in ppm, and coupling constants (J) are given in hertz (Hz). Mass spectra were obtained from a high-resolution ESI mass spectrometer at the University of Illinois School of Chemical Sciences Mass Spectrometry Lab.

Synthesis of 4-(1′-Hydroxy-1′-methyl-2′-oxopropyl)benzoic Acid (JCM1).

Synthesis of Methyl Ester JCM1.

The procedures at milligram and gram scales were identical, and the values for both are included (Scheme 3). In an oven-dried round-bottom flask under dry nitrogen gas, methyl 4-iodobenzoate (399 mg, 1.53 mmol, 1.1 equiv; 3.51 g, 13.4 mmol, 1.1 equiv) was dissolved in dry THF (to 1.2 M). This flask was cooled to −20 °C in an ice bath before isopropylmagnesium chloride–lithium chloride (1.3 M in THF; 1.2 mL, 1.53 mmol, 1.1 equiv; 10.3 mL, 13.4 mmol, 1.1 equiv) was added dropwise and left to stir for 2 h. Neat 2,3-butadione (119 mg, 1.39 mmol, 1.0 equiv; 1.05 g, 12.1 mmol, 1.0 equiv) was added dropwise before removing the ice bath and warming the reaction mixture to room temperature. The reaction was stirred for 1 h under dry nitrogen gas, quenched with 0.1 M HCl (3; 25 mL), diluted with water (15; 100 mL), and extracted with EtOAc (4 × 15; 4 × 60 mL). The combined organic layers were washed once with brine (20; 100 mL) and dried over MgSO₄ before removing the solvents by rotary evaporation. Flash chromatography (40% EtOAc in hexanes) gave the desired methyl ester (R_f, 0.36; 148 mg, 44% yield; 565 mg, 29% yield). ¹H NMR (499 MHz, CDCl₃): δ 8.06–8.00 (m, 2H), 7.56–7.50 (m, 2H), 4.56 (s br, 1H), 3.91 (s, 3H), 2.08 (s, 3H), 1.79 (s, 3H). ¹³C NMR (126 MHz, CDCl₃): δ 23.6, 24.3, 52.4, 76.9, 77.2, 77.4, 80.1, 126.2, 130.0, 130.1, 146.5, 166.8, 209.0. HR-ESI-MS m/z: [M + H]⁺ calcd. for C₁₂H₁₅O₄, 223.0970; found, 223.0971.

Hydrolysis of Methyl Ester to Make JCM1.

The methyl ester of JCM1 (128 mg, 0.578 mmol) was dissolved in THF (to 0.11 M). A 0.5 M solution of LiOH (2.89 mmol, 5.0 equiv) in water was added all at once, and the solution was stirred until all the starting material was consumed (2 h). The solution was diluted with water (15 mL) and washed thrice with EtOAc (15 mL). The resulting aqueous layer was acidified to a pH of ~1.5 with 3 M HCl before extracting with EtOAc (3 × 15 mL). The resulting orange organic layers were combined and dried over MgSO₄, and the solvent was removed by rotary evaporation to yield 95% pure JCM1 (91 mg, 0.34 mmol, 59% yield). A sample from another reaction was subjected to reverse-phase flash column chromatography over C₁₈-functionalized silica to remove contaminating 4-acetyl benzoic acid (which were differentiated based on their λ_max values of 238 nm for JCM1 and 248 nm for 4-acetylbenzoic acid). After activating the column with 4 column volumes (CV) of 0.2% trifluoroacetic acid (TFA) in MeOH and equilibrating with 4 CV of 0.2% TFA and 10% MeOH in water, a 10–60% MeOH in water with 0.2% TFA gradient over 12 CV separated these compounds, as determined by the UV spectra of the fractions. Fractions containing only JCM1 were combined and lyophilized to yield the title α-hydroxy ketone (15 mg, 21% yield) as a white solid of 99+% purity. ¹H NMR (500 MHz, CD₃OD): δ 8.04–7.98 (m, 2H), 7.63–7.57 (m, 2H), 2.09 (s, 3H), 1.65 (s, 3H). ¹³C NMR (126 MHz, MeOD): δ 24.4, 26.2, 48.5, 48.7, 48.8, 49.0, 49.2, 49.3, 49.5, 81.6, 126.5, 130.8, 131.1, 149.4, 169.5, 211.6. HR-ESI-MS m/z: [M − H]⁻ calcd. for C₁₁H₁₁O₄, 207.0657; found, 223.0661.

Synthesis of 4-(1′-Oxo-2′-hydroxy-2′-methylpropyl)benzoic Acid (JCM2).

Wittig Reaction of Methyl 4-Formylbenzoate.

Isopropyltriphenylphosphonium iodide (2.906 g, 6.722 mmol, 1.1 equiv) was dissolved in warm (~90 °C) 1,4-dioxane. To this was added potassium carbonate (1.265 g, 9.153 mmol, 1.5 equiv), and the reaction was left to heat under reflux with stirring for 15 min. Methyl 4-formylbenzoate (1.002 g, 6.104 mmol, 1.0 equiv) was added to the orange reaction mixture and left to heat under reflux for 8 h, monitoring the reaction by thin-layer chromatography (TLC; 10% EtOAc in hexanes). After the starting material was consumed, the volatiles were removed by rotary evaporation before the resulting solid was resuspended in water (20 mL). Vacuum filtration removed a fine, white, crystalline substance that was rinsed with 40 mL of EtOAc. The filtrate was separated in a separatory funnel, the aqueous layer was washed with EtOAc (2 × 40 mL), and the combined organic layers were washed twice with brine (20 mL) before drying over MgSO₄. To this was added Celite (4.0 g) for dry loading, and the volatiles were removed under reduced pressure. Silica gel flash chromatography (10% EtOAc in hexanes) gave the methyl 4-(2-methylprop-1-en-1-yl)benzoate (473 mg, 41%; R_f, 0.31). Spectra were in accordance with literature values.^{72 1}H NMR (500 MHz, CDCl₃): δ 8.01–7.95 (m, 2H), 7.31–7.25 (m, 2H), 6.29 (s, 1H), 3.91 (s, 3H), 1.93 (d, J = 1.5 Hz, 3H), 1.89 (d, J = 1.4 Hz, 3H). ¹³C NMR (126 MHz, CDCl₃): δ 19.74, 27.24, 52.12, 124.68, 127.48, 128.73, 129.53, 138.18, 143.60, 167.25.

Alkene Oxidation to Produce Methyl Ester α-Hydroxy Ketone JCM2.

⁷³ RuCl₃ (411 μL of a 0.10 M aqueous solution, 0.024 mmol, 0.01 equiv) and NaHCO₃ (505 mg, 6.01 mmol, 2.5 equiv) were dissolved in water (2.4 mL, 1 mL/mmol alkene) before adding acetonitrile (14.2 mL, 6 mL/mmol alkene) and EtOAc (7.1 mL, one of two portions at 3 mL/mmol alkene). To this solution was added Oxone (7.608 g, 12.38 mmol, 5.0 equiv), resulting in bubbling, and the solution was stirred until bright yellow. The alkene (462 mg, 2.43 mmol) was dissolved in the second portion of EtOAc (7.1 mL) before being added to the Ru-containing solution. The reaction was monitored by TLC (10% EtOAc in hexanes) until all the starting material was consumed, 13 min. After quenching the reaction by pouring into a mixture of brine (30 mL) and saturated Na₂SO₃ (30 mL), the organic layer was collected and the aqueous layer was extracted with EtOAc (3 × 30 mL). The combined organic layers were dried over MgSO₄ before adding Celite for dry loading and removing the volatiles by rotary evaporation. The sample was loaded onto an equilibrated silica gel column (25% EtOAc in hexanes), and flash chromatography commenced with 25% EtOAc (100 mL) before switching to 30% EtOAc (150 mL) and 35% EtOAc (250 mL) sequentially. These fractions containing less than 0.1% methyl 4-formylbenzoate by GC analysis were combined to yield the α-hydroxy ketone as a white solid (136 mg, 0.612 mmol, 25% yield; R_f, 35% EtOAc, 0.29). Fractions containing methyl 4-formylbenzoate were also combined as a white solid of 95% purity by GC (198 mg, ~0.85 mmol, combined 60% yield) for further purification. ¹H NMR (500 MHz, CDCl₃): δ 8.14–8.07 (m, 2H), 8.07–8.00 (m, 2H), 3.95 (s, 3H), 1.62 (s, 6H). ¹³C NMR (126 MHz, CDCl₃) δ 28.31, 52.64, 129.52, 129.68, 133.66, 137.95, 166.27, 204.76. HR-EI-MS m/z [M + H]⁺, calcd. for C₁₂H₁₅O₄, 223.09704; found, 223.09657.

Hydrolysis of Methyl Ester to Produce JCM2.

The methyl ester of JCM2 (120 mg, 0.540 mmol) was dissolved in THF to 0.1 M (5.4 mL), and LiOH·H₂O (113 mg, 2.69 mmol, 5.0 equiv) was dissolved in water to 0.5 M (5.4 mL). These solutions were mixed and stirred at room temperature until the starting material disappeared by TLC, 25 min. The reaction was diluted with water (10 mL) and brine (20 mL), turning it slightly cloudy, before acidification to a pH of ~1.5 with 3.0 M HCl, which increased the cloudiness. Extraction with EtOAc (4 × 15 mL) removed the cloudiness; the combined organic layers were dried over MgSO₄, and the solvent removed by rotary evaporation to yield the carboxylic acid JCM2 as a white solid (98 mg, 0.471 mmol, 87% yield). ¹H NMR (500 MHz, MeOD): δ 8.19 (dd, J = 8.5, 2.6 Hz, 2H), 8.12–8.05 (m, 2H), 1.51 (s, 6H). ¹³C NMR (126 MHz, MeOD): δ 28.20, 78.50, 130.29, 130.83, 134.92, 140.60, 168.94, 206.10. HR-ESI-MS m/z: [M − H]⁻ calcd. for C₁₁H₁₁O₄, 207.0657; found, 207.0660.

Protein Production and Biochemical Assays. Protein Expression and Purification.

CYP199A4 and its redox partners (HaPux and HaPuR) were expressed in BL21(DE3) Escherichia coli (E. coli) and purified according to previously reported methods.^46,48,50,52

Spectroscopic Binding Assays.

To measure the spin-state shift, aliquots (0.1–5 μL) of a substrate stock solution (100 mM) in EtOH or DMSO were successively added to P450 (500 μL, ~1–2 μM)in Tris–HCl buffer (50 mM, pH 7.4). The UV–vis spectrum was recorded after each addition, and the additional substrate was added until there was no further shift in the Soret band at approximately 417 nm. The K_d of 4-acetylbenzoic acid to WT CYP199A4 was determined using non-linear regression.

NADH Consumption Assays.

In vitro NADH turnovers were performed at 30 °C and contained CYP199A4 (1 μM), HaPux (5 μM), HaPuR (0.5 μM), and bovine liver catalase (100 ng μL⁻¹) in oxygenated Tris–HCl buffer (50 mM, pH 7.4, in a total volume of 1.2 mL). The absorbance at 340 nm was set to zero, and the mixture was incubated at 30 °C for 2 min before NADH was added (~320 μM, an absorbance of ~2.0). The rate of NADH background oxidation was measured before initiating the reaction by the addition of the substrate. To start the reaction, the substrate (500 μM) was added from a stock solution (100 mM in EtOH/DMSO), and NADH depletion was monitored at 340 nm. The rate of NADH consumption by the P450 enzyme in units of μM_NADH μM_P450⁻¹ min⁻¹ was calculated from the slope of the graph of Abs_{340 nm} versus time using an extinction coefficient of ε_{340 nm} = 6.22 mM⁻¹ cm⁻¹ and reported as min⁻¹. All experiments were performed three times with the mean and standard deviation reported. Control reactions were also performed in which either the P450 or NADH was omitted from the turnover mixture (replaced with the same volume of buffer).

Turnover of JCM1 and JCM2 with CYP199A4.

Reactions were carried out (total volume of 600 μL) in Tris–HCl buffer (50 mM, pH 7.4) in 1.5 mL Eppendorf tubes. The reactions contained the following components: P450 (1 μM), HaPuR (0.5 μM), HaPux (5 μM), alcohol dehydrogenase (ADH, 4.5 μL of 0.029 g mL⁻¹ suspension, Roche), bovine liver catalase (100 ng μL⁻¹, Sigma-Aldrich), EtOH (12 μL), and substrate (150 or 50 μM). NADH (320 μM) was added last to start the reaction. Reactions were incubated at 25 °C and 250 rpm. Alternative conditions included the following: excess NADH, 1 mM NADH; mutant CYP199A4 F182L and F298V turnovers, only 50 μM substrate used. Negative controls included no P450/redox partners and no ADH. The remaining volume was made up by Tris–HCl buffer.

Isolation and Identification of 4-(2′-Hydroxyacetyl)benzoic Acid.

An E. coli-based whole-cell oxidation system, as reported previously,^52,54 was used to oxidize 4-acetylbenzoic acid for metabolite identification. The supernatant (200 mL) was acidified, extracted with ethyl acetate (3 × 100 mL), washed with brine (100 mL), and dried with anhydrous MgSO₄. The products were separated from the unreacted substrate using preparative reverse-phase HPLC. The solvent was removed by freeze-drying. The ¹H NMR spectrum was acquired on an Agilent DD2 spectrometer operating at 500 MHz. This purified product was also derivatized with N,O-bis(trimethylsilyl)trifluoroacetamide/trimethylsilyl chloride (BSTFA/TMSCl) and analyzed by GC–MS (see Figure S3).

Synthesis and Analysis of 4-(2′-Oxoacetyl)benzoic Acid.

This potential further oxidation metabolite of 4-(2′-hydroxyacetyl)benzoic acid was synthesized from 4-acetylbenzoic acid using an adaption of previously reported methods (Scheme 7).^74,75

A solvent system of 1,4-dioxane:water (10:1, 1 mL) was added to selenium dioxide (14.6 mg, 0.13 mmol), and the mixture was stirred at 50°C until dissolution. 4-Acetyl benzoic acid (19.5 mg, 0.12 mmol) was added, and the mixture was stirred at 90°C for 18 h. Once cooled to room temperature the suspension was filtered through cotton wool with DCM and EtOAc, and the filtrate concentrated under a stream of nitrogen gas. The residue was resuspended in EtOAc (5 mL), washed with distilled water (5 mL), dried over MgSO₄, filtered, and concentrated under a stream of nitrogen gas to yield an off-white powder containing a 20:80 mixture of the 4-(2′-oxoacetyl)benzoic acid and starting material (7.3 mg, 37% mass recovery). The metabolite decomposed during purification attempts but could be identified by NMR of the reaction mixture, which had a signal due to the aldehyde proton. HPLC coelution experiments demonstrated that this metabolite matched that formed during the enzymatic turnover of JCM1 and 4-acetylbenzoic acid ($; Figure 4 and Scheme 5). ¹H NMR (500 MHz, CDCl₃): δ 8.05–8.10 (m, 2H), 8.27–8.33 (m, 2H), 9.67 (s, 1H). HR-ESI-MS m/z: [M − H]⁻ calcd. for C₉H₅O₄, 177.1082; found, 177.0116.

X-Ray Crystallography.

Crystallization experiments were performed with WT, T252E, and F182L mutants of CYP199A4. Immediately prior to preparation of crystal trays, the protein was purified via elution through a HiPrep Sephacryl S-200 HR size-exclusion column (60 cm × 16 mm; GE Healthcare) with Buffer T at a flow rate of 1 mL min⁻¹ The purity of the protein was assessed based on the Reinheitszahl value, RZ = A₄₂₀/A₂₈₀, whereby RZ = 2 was collected and combined.

The combined fractions were then concentrated via ultrafiltration using a Microsep Advance centrifugal device (10 kDa MWCO, Pall Corporation) to a concentration of approximately 30–35 mg mL⁻¹. The substrate was then added to a final concentration of 1 mM from a 100 mM stock of EtOH and DMSO to the concentrated protein. Crystallization trays were prepared using the following optimized buffer conditions previously reported:⁷⁶ 0.2 M magnesium acetate, 100 mM Bis-Tris buffer (adjusted with acetic acid to pH 5.0–5.75), and 20–32% w/v polyethylene glycol (PEG) 3350.

Protein crystallization was achieved using the hanging-drop vapor diffusion method in 24-well trays. An equal volume of crystallization buffer was mixed with hanging drops of 1.2–2 μL of protein and was equilibrated with a reservoir of the same buffer (500 μL) at 16 °C. Red plate-like crystals were obtained after half a day to one week. Single crystals were mounted onto Micromounts of Microloops (MiTeGen LLC, New York, USA). Mounted crystals were soaked and dragged in Parabar 10312 oil (Paratone-N, Hampton Research, California, USA) before flash-frozen in liquid N₂.

X-ray diffraction data were obtained (360 images per crystal) at the Australian Synchrotron using beamline MX1⁷⁷ with an exposure time of 1 s, oscillation angle of 1°, wavelength of 0.9537 Å, and temperature of 100 K. Diffraction images were indexed and integrated using iMosfilm.⁷⁸ Aimless⁷⁹ from the CCP4 suite of programs⁸⁰ was used to carry out scaling, merging, and R_free labeling (5 % of reflections, randomly selected). The phase was solved using molecular replacement in Phaser⁸¹ using a high-resolution structure of WT CYP199A4 (1.54 Å, PDB: 5UVB) as the search model. The ligands and solvent molecules were from the search model prior to phasing to eliminate model bias. The weighted 2mF_o-DF_c map and F_o–F_c difference map were obtained and used to rebuild the model in WinCoot and determine the substrate bound.⁸² Structural refinements were carried out over multiple cycles using Phenix Refine, available in the Phenix suite of programs.⁸³ If the bound substrate was a racemic mixture, both enantiomers were modeled within the substrate-binding site separately and underwent several cycles of refinement to determine the identity of the bound enantiomer. If the conformation of the enantiomer was still unclear, the occupancies of both enantiomers were refined in the same location using different alternative conformation labels (altLoc identifiers) in the same PDB coordinate file.

Composite omit or feature-enhanced maps that reduce model bias were generated in Phenix to allow inspection of the ligand-binding site and reveal the location of all substrate atoms.^84,85 Detailed data collection and structural refinement statistics are provided in the Supporting Information The coordinates for the crystal structures were deposited into the wwPDB (Worldwide Protein Data Bank).⁸⁶ Individual PDB accession codes have been presented.

Substrate Docking and Molecular Dynamics Simulations.

The 4-cyclopropylbenzoate-bound structure of CYP199A4 (PDB: 5UVB) was selected for docking because it was among the highest in resolution (1.54 Å) and included a substrate with a branched hydrocarbon para to the carboxylate. After importing the structure into the Molecular Operating Environment (MOE), the CYP199A4 structure was aligned with the backbone of the oxygen-bound Thr252Ala CYP101A1 (PDB: 2A1O).⁴⁴ This alignment was manually adjusted to align the heme of CYP199A4 and the oxy-ferrous heme of the CYP101A1 structure. The CYP101A1 backbone and waters, CYP199A4 heme, and 4-cyclopropyl benzoate were all removed from the structure before equilibrating the oxy-ferrous CYP199A4 model to 0.01 kcal mol⁻¹ A⁻² using the Amber force field ff19SB.⁸⁷ (R)-JCM1, (S)-JCM1, and JCM2 were independently docked into the active site of this equilibrated structure using MOE according to previously reported methods.⁵⁹ Briefly, the cavity above the heme comprising the active site was selected using the Site Finder function of MOE, the compounds were independently docked, and those with their alcohols near the proximal oxygen and ketones near the distal oxygen were selected for a brief minimization to 0.01 kcal mol⁻¹. The resulting dock, the pose of which was best positioned for catalysis using the constraints established from QM/MM modeling of CYP17A1, was subjected to MD simulations for each compound.⁶⁰ The P450 MD simulations were set up using AmberTools 20 using Amber force field ff19SB.⁸⁸ Heme parameters were obtained from Shahrokh et al..⁸⁹ Substrate parameters were generated from the respective mol2 files using the general Amber force field.⁸⁷ The F182L mutation was generated using AmberTools prior to minimization and equilibration.⁹⁰ Systems were solvated using an octahedral periodic boundary with a 12 Å spacing. Potassium and chloride ions were added to neutralize the system, bringing the KCl concentration to 0.15 M. MD simulations were performed using GROMACS 2020.⁹¹ The solvated systems were first energy-minimized using the steepest decent algorithm, followed by sequential equilibration runs, constraining the protein backbone before slowly releasing the constraints on the side chains and substrates. Equilibration runs employed a modified Berendsen thermostat and a Berendsen barostat for temperature and pressure coupling. For production runs, temperature and pressure couplings were achieved using a modified Berendsen thermostat and a Parrinello-Rahman barostat, respectively. Electrostatics were calculated using Particle Mesh Ewald. For production runs, the time step was 2 femtoseconds with coordinates saved every 10 ps. Runs continued until the accumulated simulation time reached ~5 μs (for both enantiomers of JCM1).

Table 1.

Binding and Activity Parameters Determined for the Carbonyl Substrates Metabolized by CYP199A4 in This Study

substrate	HS (%)	NADHa	PFRb	C (%)c
JCM1-WT	<5	185 ± 5
JCM1-F182L	60	68 ± 1	1.3 ± 0.2	1.9 ± 0.3d
JCM2-WT	10	33 ± 4
JCM2-F182L	25	1430 ± 130
4-acetylBA-WT	~10	265 ± 20	213 ± 21	80 ± 2e
4-acetylBA-F182L	80	488 ± 9	414 ± 0.1	85 ± 2
4-n-propylBA-WT	≥ 95	688 ± 24	594 ± 72	86 ± 8

^aAbbreviations: NADH oxidation rate, min⁻¹.

^bPFR, product formation rate (if blank no product from enzyme activity was observed).

^cC, overall coupling efficiency, i.e., the percentage of NADH consumed that leads to the formation of the metabolized substrate. For JCM1 and 4-acetylbenzoic acid, this was based on the amount of 4-(2‘-hydroxyacetyl)benzoic acid and related metabolites in these assays.

^dThe amount of H2O2 generated accounted for 8 ± 1% of the NADH consumed.

^eThe amount of H2O2 generated accounted for 5 ± 1% of the NADH consumed. Shown are the spin-state shifts observed on substrate binding (%HS ± 5%). Rates are expressed as mol (mol CYP)⁻¹ min⁻¹ which is abbreviated as min⁻¹. 4-n-Propylbenzoic acid with WT CYP199A4 data is shown for comparison.⁵⁶