The Enigmatic P450 Decarboxylase OleT Is Capable of, but Evolved To Frustrate, Oxygen Rebound Chemistry

Abstract

OleT is a cytochrome P450 enzyme that catalyzes the removal of carbon dioxide from variable chain length fatty acids to form 1-alkenes. In this work, we examine the binding and metabolic profile of OleT with shorter chain length (n ≤ 12) fatty acids that can form liquid transportation fuels. Transient kinetics and product analyses confirm that OleT capably activates hydrogen peroxide with shorter substrates to form the high-valent intermediate Compound I and largely performs C–C bond scission. However, the enzyme also produces fatty alcohol side products using the high-valent iron oxo chemistry commonly associated with insertion of oxygen into hydrocarbons. When presented with a short chain fatty acid that can initiate the formation of Compound I, OleT oxidizes the diagnostic probe molecules norcarane and methylcyclopropane in a manner that is reminiscent of reactions of many CYP hydroxylases with radical clock substrates. These data are consistent with a decarboxylation mechanism in which Compound I abstracts a substrate hydrogen atom in the initial step. Positioning of the incipient substrate radical is a crucial element in controlling the efficiency of activated OH rebound.

Graphical abstract

OleT, a cytochrome P450 (CYP) enzyme, has recently attracted a great deal of attention because of the unusual chemistry that it catalyzes. Using hydrogen peroxide as a cosubstrate, OleT removes the carboxylate from C_n chain length fatty acid (FA) substrates to produce C_n−1 terminal alkenes¹ and carbon dioxide.² This decarboxylation reaction is of considerable commercial interest as a potential means of intercepting microbial fatty acid biosynthesis and increasing the heating value of drop-in hydrocarbon biofuels.^3,4 The scope of OleT-catalyzed decarboxylations, which has been expanded to include structurally diverse substrates that include a range of benzylic⁵ and dioic⁶ carboxylic acids, also suggests that it may serve as a useful biocatalytic platform for the synthesis of important commodity chemicals. The OleT-catalyzed C–C_α scission reaction is a significant departure from the function of most CYPs. This is perhaps best highlighted by comparison to those that are critical for xenobiotic detoxification, where an oxygen atom derived from atmospheric dioxygen is inserted into chemically unreactive substrates, generating a metabolite that is more water-soluble than the parent compound.

Analysis of the hydrocarbon content of Jeotgalicoccus species has demonstrated that long chain fatty acids (C_n, where n ≥ 20) are the native substrates for OleT in vivo.¹ In the metabolism of eicosanoic acid (C₂₀ FA or EA), the enzyme displays nearly exclusive (≥95%) chemoselectivity for 1-nonadecene production in vitro.^2–4 Isotope tracer experiments have demonstrated that both substrate carboxylate oxygens are retained in the CO₂ product.² The lack of detectable products intermediary with respect to alkene formation from single-turnover reactions eliminates possible mechanisms whereby sequential oxygenations give rise to the atypical metabolite, such as those observed for other C–C bond-cleaving CYPs.^7–9 The OleT conversion of a fatty acid to a terminal alkene thus requires, in sum, the removal of a substrate hydride (H⁻). Although the OleT C–C_α bond scission reaction would then appear to be quite different from mechanisms for archetypal P450 hydroxylations, which proceed via substrate hydrogen atom (H^•) abstraction by the iron(IV)–oxo π cation intermediate (Compound I or P450-I) and subsequent rapid ^•OH radical recombination through a process termed “oxygen rebound,”^10–12 several lines of evidence strongly connect the two mechanisms. The primary sequence of OleT and the X-ray crystal structure determined by Munro, Leys, and colleagues¹³ have revealed that OleT belongs to the P450 fatty acid peroxygenase (CYP152) family.¹⁴ CYP152s include the well-characterized CYPs BSβ and SPα that utilize the substrate carboxylate as a general acid to efficiently utilize H₂O₂ as an oxidant for the hydroxylation of FAs at positions adjacent to the terminal carboxylate, affording alcohol products with variable regioselectivity.¹⁴ Although the active intermediate responsible for FA hydroxylation has not been directly observed in BSβ or SPα, a significant steady-state kinetic isotope effect (KIE) for hydroxylation¹⁵ strongly suggests a scenario involving H^• abstraction by P450-I and ensuing ^•OH rebound.

Transient kinetic studies by our laboratory have shown that the initial stages of OleT decarboxylation may result from a mechanism similar to those proposed for CYP aliphatic hydroxylations. Single-turnover, rapid mixing studies of OleT bound to perdeuterated EA (EA-d₃₉) with H₂O₂ result in the generation of a Compound I intermediate that we term Ole-I that accumulates to a high level.² The optical features of Ole-I suggest that it is electronically similar to the high-valent oxo complexes observed in rapid mixing studies of other CYPs^16–18 and the structurally dissimilar thiolate ligated heme enzyme aromatic peroxygenase (APO)¹⁹ with peroxyacids, all potent hydroxylation catalysts that are not known to perform desaturative decarboxylations. On this basis, we originally proposed a radical mechanism for 1-alkene production whereby Ole-I abstracts H^• from the Cβ position to form a substrate radical, resulting in formation of the iron(IV)–hydroxide, Compound II (Figure 1). Subsequent one-electron oxidation of the substrate by Compound II, coupled to transfer of a proton, would restore the aqua–ferric resting state of the enzyme and generate a substrate carbocation. Such a species could then readily eliminate CO₂ and form the Cα=Cβ bond. This type of mechanism is similar to those proposed for desaturations performed by some CYPs,^20,21 and O₂ activating non-heme mononuclear²² and dinuclear iron enzymes^23–25 and synthetic complexes²⁶ that closely structurally resemble monooxygenation catalysts. In support of this scheme, we have recently measured a large (k_H/k_D > 8) substrate ²H KIE for the decay of Ole-I and have isolated an additional intermediate during EA decarboxylation. Concomitant with the decay of Ole-I, a new intermediate is formed that has optical spectroscopic properties^18,27 and a reactivity toward phenols^27,28 that are consistent with its assignment as an iron(IV)–hydroxide species (Ole-II).²⁹ In the decarboxylation of EA, Ole-II decays slowly (~10 s⁻¹ at 5 °C), suggesting that the long chain substrate is positioned in such a way that the rapid oxygen rebound characteristic of CYPs is thwarted, allowing alkene formation to occur.

Figure 1.

Proposed bifurcative mechanism for the decarboxylation (top) and hydroxylation (bottom) reactions catalyzed by cytochrome P450 OleT with long and short chain (C_n) fatty acid substrates. The rate constants for Compound I and II decay are derived from earlier stopped flow studies with eicosanoic acid (C_n, where n = 20).^2,29 For long chain length (CL) substrates (C_n, where n ≥ 20), C_n−1 alkene formation has been shown to be the predominant (≥95%) pathway.^2–4,30 Shorter CL substrates have been observed to result in mixed chemoselectivity, producing variable percentages of fatty alcohols and C_n−1 alkenes.^1,3,4,30,31

A more precise understanding of the factors that control the branchpoint between monooxygenation and decarboxylation would allow OleT and related enzymes to be rationally engineered for fungible and targeted fuel production. A demonstrated limitation of OleT^1,3,4,30,31 and other CYP152s³² is the production of a variable and unwanted proportion of C_n alcohol products, in addition to the C_n−1 alkene, upon reduction of the substrate chain length (CL) below C₂₀. Within the framework of the proposed mechanistic scheme in Figure 1, it is unknown whether the unusually long stability of Ole-II, and the presumptive substrate radical, stems from underappreciated electronic changes in ferryl intermediates that are not readily apparent from the optical data, positional restraints of the substrate imposed by the protein, or structural contributions from the secondary coordination sphere. To better gauge the factors that contribute to this bifurcation, we have examined the binding and reactivity of short chain length FAs (C₈–C₁₂), substrates that would be metabolized to the most desirable hydrocarbon “drop-in” biofuels. Although shorter CL substrates bind to the enzyme with a substantially lower affinity, they are efficiently metabolized and support the efficient generation of Ole-I. The production of alkenes with these substrates is accompanied by the formation of alcohol products with complete incorporation of oxygen from H₂O₂. An analysis of products from the OleT metabolism of norcarane and methylcyclopropane radical clock probes provides a direct measure of the proficiency of radical recombination. The measured radical lifetimes are in sharp contrast with kinetic data for Ole-II decay for both short and long chain FAs and are similar to those measured for many other P450 hydroxylases. These studies provide compelling evidence that OleT alkene formation proceeds by a radical mechanism that bifurcates after hydrogen atom transfer (HAT) and alludes to the importance of substrate active-site chemistry, rather than the electronic structure of iron intermediates, in controlling reaction partitioning in this group of enzymes. The competition of oxygen rebound to form an alcohol and electron transfer to produce a carbocation is guided by exquisite coordination of the FA substrate.

MATERIALS AND METHODS

Chemicals and Reagents

Chemicals, unless otherwise indicated, were purchased from Sigma-Aldrich. Isopropyl β-D-thiogalactopyranoside and all buffers used in this study were purchased from Research Products International (Mt. Prospect, IL). Imidazole and Luria broth were purchased from bio-WORLD (Dublin, OH). Isotopically labeled H₂¹⁸O₂ (90% ¹⁸O) was purchased from ICON Isotopes (Summit, NJ). Perdeuterated fatty acids were purchased from CDN Isotopes (Pointe-Claire, QC). Protiated fatty acids, N,O-bis(trimethylsilyl)-trifluoroacetamide (BSTFA), and trimethylchlorosilane (TMCS) (99:1) were purchased from Supelco (Bellefonte, PA). Radical clock substrates were synthesized according to published procedures.^33,34

Heterologous Expression and Purification of P450 OleT

Wild-type cytochrome P450 OleT from Jeotgalicoccus sp. ATCC 8456 was overexpressed in Escherichia coli and purified as previously described.^2,29 Adventitiously bound fatty acids were removed in a procedure similar to that previously reported that includes treatment with a 5–10-fold molar excess of H₂O₂ and desalting on a PD10 column. Here, we have utilized a subsequent step involving incubation of the protein for 15 h at 4 °C with BioBeads SM2 resin (Bio-Rad). The protein was separated from the resin with a syringe filter and verified for complete conversion to the low-spin substrate-free form by optical spectroscopy. Protein concentrations were determined using an extinction coefficient (ε₄₁₇ = 109 mM⁻¹ cm⁻¹) determined by the pyridine hemochromagen method.³⁵

Guaiacol Oxidation Studies

The oxidation of guaiacol to the polymerized tetraguaiacol product was used as a probe to determine the optimal chain length and isotopic composition of fatty acid substrates that efficiently activate OleT yet still permit sufficient entry of small molecules into the active site. An 800 μL reaction mixture contained 200 mM KP_i (pH 7.5), 200 mM NaCl, 2 μM OleT, 2 mM guaiacol, and 800 μM fatty acid and was allowed to incubate for 1 h while being stirred at 4 °C. The reaction was initiated with 1 mM H₂O₂, and tetraguaicol formation was monitored at 470 nm using a Hewlett-Packard 8453 spectrophotometer. Initial rates were determined using a tetraguaiacol extinction coefficient of 26.6 mM⁻¹ cm⁻¹.

Determination of Substrate Dissociation Constants

A 500 μL reaction mixture containing 5 μM OleT in 200 mM KP_i (pH 7.5) and 200 mM NaCl was titrated with sequential additions of a 10 mM fatty acid stock dissolved in ethanol with a Hamilton gastight syringe. The amount of ethanol never exceeded 5% (v/v). In contrast to similar plots for the binding of EA,^2,13 the relatively high dissociation constants of C₈ and C₆ fatty acids did not necessitate fitting to a quadratic Morrison expression.³⁶ In these cases, the additive absorbance changes at 392 and 417 nm were fit as a function of substrate concentration [S] using Origin software with the following hyperbolic expression to determine dissociation constants (K_d).

$$\mathrm{\Delta }{\text{Abs}}_{392\text{nm}-417\text{nm}}=\frac{\mathrm{\Delta }{\text{Abs}}_{max}[\mathrm{S}]}{K_{\mathrm{D}}+[\mathrm{S}]}$$

In the case of C₁₀ and C₁₂ FAs, which have K_D values within an order of magnitude of the enzyme concentration utilized, the following Morrison expression was utilized for K_D determination.

$$\mathrm{\Delta }\text{Abs}=\mathrm{\Delta }{\text{Abs}}_{max}\left\{\left[K_{\mathrm{D}}+[{\mathrm{E}}_{\text{tot}}]+[{\mathrm{S}}_{\text{tot}}]-\sqrt{{(K_{\mathrm{D}}+[{\mathrm{E}}_{\text{tot}}]+[{\mathrm{S}}_{\text{tot}}])}^2-4[{\mathrm{E}}_{\text{tot}}][{\mathrm{S}}_{\text{tot}}]}\right]/\left(2[{\mathrm{E}}_{\text{tot}}]\right)\right\}$$

Using spectral deconvolution procedures to determine the fraction HS as a function of S_tot resulted in similar K_D values. The addition of small molecules (for example, guaiacol, benzene, and norcarane) to FA-bound OleT did not induce any additional spectroscopic changes in the enzyme.

Product Analysis of OleT with Short Chain Fatty Acid Substrates

A sealed vial contained a 2 mL reaction mixture of 5 μM OleT and 500 μM fatty acid (prepared from a 20 mM stock in ethanol) in 200 mM KP_i (pH 7.5). Each reaction was terminated by the addition of 1 mL of 12 M HCl and each mixture extracted with 4 mL of chloroform using a gastight Hamilton syringe. Dual internal standards consisted of an alkene (125 nmol of hexadecene) and a FA (500 nmol). C₁₂ FA was used as the standard for reactions of C₈, C₁₄ FA for reactions of C₁₀, and C₁₀ FA for reactions of C₈. Vials were centrifuged, and the organic phase was dried using N₂ and subsequently derivatized with 200 μL of BSTFA and TMCS (99:1) at 68 °C for 30 min. The quantity of fatty acid metabolized and fatty alcohol products, both nonvolatile, can be determined empirically for all CL substrates by comparing the peak areas of monotrimethylsilylated (TMS) fatty acids and di-TMS product alcohols to internal standards with response factors that have been determined in previous studies.³² For reactions using C₁₂ FA as a substrate, all of the products are nonvolatile, and we have found that the sum of all products accurately accounts for ≥98% of the substrate metabolized. This is not the case for reactions of OleT with C₁₀ FA and C₈ FA because of the volatility of the C_n−1 alkene products, nonene and heptene, respectively, that are produced. For these substrates, the presence of alkenes was verified by headspace gas chromatography (GC) analysis, but products were quantified by subtraction of the product alcohols from the total amount of substrate metabolized. Peaks were identified on the basis of their retention time, mass spectral fragmentation patterns, and comparison to authentic standards. To minimize the level of overly oxidized products (e.g., ketones and di-OH FAs), low concentrations of H₂O₂ relative to FA were used (2:1 on a molar basis). Gas chromatography–mass spectrometry (GC–MS) analysis of C₁₀ and C₈ reactions showed no evidence of these products within detection limits. Headspace GC analysis employed similar reaction conditions in a septum-sealed vial. A reaction mixture from which H₂O₂ was omitted was used as a control. Nonane (112 nmol) was added as an internal standard, and 250 μL of the reaction headspace was injected onto a GC Hewlett-Packard 5890 GC system using a Hamilton syringe needle. The gas chromatograph was equipped with a J&W Scientific DB-5MS column (30 m × 0.25 mm, 0.5 μm). The following oven temperature program was used: 300 °C injector temperature, held at 50 °C for 3 min, heated at a rate of 5 °C min⁻¹ to 100 °C, held at 100 °C for 3 min.

Insertion of Oxygen into Fatty Alcohol Products

A 2 mL reaction mixture containing 200 mM KP_i (pH 7.5), 1 mM FA, and 5 μM OleT was initiated by the addition of hydrogen peroxide [2 mL, 500 μM H₂¹⁶O₂ or H₂¹⁸O₂ (90% ¹⁸O)] over the course of 1 h. Upon completion, the mixture was extracted with an equal volume of chloroform and 100 nmol of dodecanoic acid was added as an internal standard. The combined mixture was vortexed for 15 min and centrifuged at 4 °C for 20 min. The organic phase was removed and derivatized as described above. Samples were analyzed on a Hewlett-Packard 5890 gas chromatograph with an Rtx-5 fused silica column (30 m × 0.25 mm inside diameter, film thickness of 0.25 μm) using a temperature gradient from 70 to 300 °C at a rate of 10 °C/min. The column fed directly into a Waters VG 705 magnetic sector mass spectrometer and was ionized using a 70 eV electron impact energy.

Stopped Flow Spectroscopy

Transient kinetic experiments were conducted using methods we have previously described^2,29 with minor modifications. Because of the lower affinity of C₁₂ and C₁₀ FA for OleT, it was necessary to add these substrates in a large molar excess to the enzyme to generate a fully saturated enzyme:substrate (E:S) complex. The OleT:DA and OleT:LA complexes were generated by addition of 500 μM FA to 20 μM OleT in 200 mM KP_i (pH 7.5) and allowed to incubate for 2 h at 4 °C. The measured Compound I decay rate constants exceed those for substrate association steps by at least an order of magnitude. As a result, the spectral changes monitored in short time regimes approximate single-turnover conditions. For stopped flow experiments that included a saturating norcarane concentration, 5 μL of neat norcarane was added to 2 mL of the OleT:FA reaction mixture. Samples were centrifuged to remove any remaining nonsolubilized fatty acid and subsequently loaded into one syringe of an Applied Photophysics SX20 stopped flow spectrophotometer. E:S complexes were rapidly mixed with H₂O₂ in 200 mM KP_i, and the reaction was monitored by a photodiode array (PDA) or at 370 or 440 nm with a photomultiplier tube (PMT). Ole-I decay rates were obtained by fitting the PMT time courses at 370 or 440 nm to the following summed two-exponential expression in Pro-Data software.

$$A_{t,\mathrm{obs}}=A_{\infty }+{\displaystyle \mathbf{\sum }_{i=1}^n{a}_i{\mathrm{e}}^{-t/t_i}}$$

where A_t,obs is the observed absorbance, 1/t_i is the reciprocal relaxation time (RRT) (inverse seconds), a_i is the amplitude of phase i, t is the time (seconds), and A_∞ is the final absorbance. The observed RRTs were analyzed as a function of H₂O₂ using the following kinetic scheme where Fe³⁺(HS) and Fe³⁺(LS) are the high-spin and low-spin forms of the enzyme, respectively, and Ole-I and Ole-II have the usual designations:

$${\text{Fe}}^{3+}(\text{HS})+{\mathrm{H}}_2{\mathrm{O}}_2\underset{k_{-1}}{\overset{k_1}{\rightleftarrows }}\text{Ole}-\mathrm{I}\underset{k_{-2}}{\overset{k_2}{\rightleftarrows }}\text{Ole}-\mathrm{II}\stackrel{k_3}{\to }{\text{Fe}}^{3+}(\text{LS})$$

We have previously found in reactions of EA that only the fast, larger amplitude phases (RRT₁) at 370 and 440 nm demonstrate appreciable isotopic sensitivity. This is also the case for DA and LA. The RRT₁ at 370 nm, which corresponds to Ole-I decay, is identical in value to the fast RRT at 440 nm, or Ole-II formation. The RRT₁ (1/τ_obs) versus [H₂O₂] plot for deuterated substrates was fit using the following hyperbolic expression

$$\frac{1}{{\tau }_{\text{obs}}}=\frac{k_2[{\mathrm{H}}_2{\mathrm{O}}_2]}{\frac{k_{-1}}{k_1}+[{\mathrm{H}}_2{\mathrm{O}}_2]}+k_{-2}$$

where the apparent $K_{\mathrm{D}}^{{\mathrm{H}}_2{\mathrm{O}}_2}$ is provided by k₋₁/k₁ and the rate constant for C–D bond cleavage by Ole-I is provided by k₂ (the asymptote of the hyperbolic plot). The y-intercept (k₋₂) is zero because of the irreversibility of C–D bond cleavage. The slower and smaller amplitude RRT at 370 or 440 nm (RRT₂), which is insensitive to H₂O₂, provides the decay rate constant of Ole-II (k₃).

Radical Clock Studies

A 2 mL reaction mixture containing 200 mM KP_i (pH 7.5), 200 mM NaCl, 1 mM perdeuterated decanoic acid (DA-d₁₉) prepared in ethanol, 25 μM OleT, and 2 μL of neat radical clock substrate (norcarane or methylphenylcyclopropane) was allowed to stir and incubate at room temperature for 15 min prior to the addition of hydrogen peroxide (2 mL, 5 mM) via a syringe pump over the course of 1 h. After completion of the reaction, 400 μL of CDCl₃ (stabilized with Ag) was added to each sample for product extraction. The mixture was vortexed vigorously for 1 min and centrifuged at 16200 × g. The organic layer was removed from the bottom of the conical centrifuge tube and placed in a GC–MS vial for immediate analysis. Samples that were not immediately analyzed were stored at −80 °C. Products were analyzed on an Agilent 7890A GC system (Rtx-5 G27 30 m column with fused silica) with an Agilent 5975C mass spectrometer with EI/CI capabilities (Princeton), or on an Agilent 5977E GC/MSD bundled system with an HP-5MS column (Barnard). For methylphenylcyclopropane products, the sample was injected at an oven temperature of 35 °C, held for 2 min, and heated at a rate of 10 °C/min to a final temperature of 225 °C. For norcarane-derived products, two slightly different methods were employed. For method A (Princeton), the initial oven temperature was 30 °C, which was held for 6 min and subsequently increased at a rate of 10 °C/min to 225 °C. For method B (Barnard), the initial oven temperature was 35 °C, with a 5 min hold time, followed by a 10 °C/min gradient to a final oven temperature of 225 °C. The rate constants for the oxygen rebound reactions were determined by multiplying the ratio of ring-closed to ring-opened products by the intramolecular radical rearrangement rate for each probe: k_rearr = 2 × 10⁸ s⁻¹²⁰ and 3 × 10¹¹ s⁻¹³⁷ for norcarane and methylcyclopropane, respectively. The radical lifetime is the reciprocal of this value.

RESULTS

The Metabolism of Short Chain Length FAs Results in Alcohol Products from Radical Recombination

The chemical outcome of OleT has been shown to vary significantly with the fatty acid CL. Published studies that have probed the OleT metabolism of longer CL (C_n, where n ≥ 16) fatty acids are largely consistent and generally show a high degree of chemoselectivity of the enzyme.^1–3,13 However, there is significant ambiguity in the metabolic profiles of short to mid CL substrates (C₈–C₁₂) despite their importance for the production of the most desired biofuels. Ultraviolet–visible (UV–vis) titration studies (Figure S1) examined the binding and efficiency of low- to high-spin-state conversion for this set of substrates. The data, summarized in Table 1, show a systematic decrease in both the affinity and propensity for high-spin conversion as the chain length of the FA is shortened. This is consistent with the general trend previously reported for C₁₂–C₂₂ saturated substrates by Belcher et al.¹³ The measured dissociation constants are several orders of magnitude lower than those reported for EA (<1 μM), most likely because of the extensive loss of hydrophobic contacts with the enzyme upon reduction of the acyl chain.

Table 1.

Binding Parameters of Short Chain Fatty Acids and OleT

fatty acid	KD (μM)	maximal high-spin conversion (%)
C12:0	2.8 ± 0.4	61
C10:0	25 ± 8	36
C8:0	134 ± 11	25
C6:0	164 ± 15	19

Reported turnover data for OleT and shorter chain FAs have shown significant variability, which may be attributed to differences in the turnover conditions utilized, the overall efficiency of substrate conversion, and, possibly, the analytical methods employed.^3,4,31 For example, turnover studies of C₁₂ FA (or lauric acid, LA), using either a surprisingly efficient NAD(P)H/O₂/redox partner (CamAB) system and headspace GC detection³ or a photochemical turnover system and liquid:liquid extraction,⁴ have found that alkenes are exclusively generated or not produced in detectable levels, respectively. Here, we have utilized turnover conditions in which H₂O₂ is slowly introduced into the reaction mixture. We have shown that both OleT and another CYP152 enzyme that we have termed CYP-MP retained catalytic activity with these methods and produced high levels of products, facilitating accurate product quantitation.³² Using the H₂O₂ perfusion approach, the shorter CL FAs were nearly completely metabolized within 2 h. A representative chromatogram of the reaction headspace of the C₁₀ FA reaction is shown in Figure S2 and revealed the presence of a significant amount of alkene, consistent with previous results using the CamAB/O₂ turnover system. GC–MS analysis of quenched samples was utilized to detect and quantify any possible nonvolatile products. Representative chromatograms of the C_n TMS-derivatized alcohol products and C_n−1 alkenes are shown in Figure S3. Product distributions are summarized in Table 2. For all substrates, 1-alkene was the dominant reaction product formed. However, analysis of the GC–MS chromatograms following derivatization also revealed the presence of a significant fraction of fatty alcohols for some substrates. This was most apparent for the reaction of C₁₂, which showed that alcohols, mostly localized to Cβ but also present at other positions, comprised nearly 30% of the total product formed. Although the distribution of products (alkenes vs alcohols) varied among the CLs tested here, the regiospecificity of hydroxylation showed a strong bias (>70%) toward the Cβ position for all substrates.

Table 2.

Metabolic Profile of Short Chain Fatty Acids with OleT and Slow H₂O₂ Addition

fatty acid	% metabolized	product distribution (%)
		1-alkene	alcohols	alcohol regiospecificity
C12:0	98.3 ± 0.2	61.4 ± 7.6a	38.6 ± 7.6	7.0 ± 1.3% α
				71.5 ± 15.5% β
				21.5 ± 4.4% γ
C10:0	100	90.0 ± 0.8b	10.0 ± 0.8	100% β
C8:0	100	88.4 ± 0.4b	11.6 ± 0.8	89.0 ± 5.3% β
				11.3 ± 2.3% α

^aDetermined from direct quantitation.

^bDetermined from subtraction of product alcohols from the total FA metabolized.

Given the unusual stability of Ole-II in the reaction of EA and, by extension, a presumptive substrate Cβ radical, oxygen isotope labeling studies probed whether the observed product alcohols specifically resulted from an authentic oxygen rebound event or from a process involving diffusive escape of these more weakly bound substrates after the radical is formed.^38,39 Reactions using C₁₂ FA, which formed the most alcohol products, were performed with H₂¹⁶O₂ or H₂¹⁸O₂ as an oxidant in sealed vials, quenched, and analyzed by MS following trimethylsilylation (TMS). A comparison of the demethylated [m – 15] fragment ions from the TMS-derivatized alcohols is shown in Figure 2. When the reaction was performed with H₂¹⁸O₂, a diagnostic shift of the fragment ion from m/z 345 to 347 was observed, consistent with complete (≥95%) integration of ¹⁸O from the oxidant. No appreciable exchange with solvent was observed within the available limits of detection. This indicates that the substrate radical, once formed, is retained within the OleT active site and that oxygenation via radical recombination occurs in a canonical fashion.

Figure 2.

Gas chromatography mass spectra of the trimethylsilylated Cβ alcohols produced by OleT using lauric acid as a substrate and (A) H₂¹⁶O₂ or (B) H₂¹⁸O₂ as an oxidant.

Short Chain Substrates Efficiently Generate Ole-I

The binding of fatty acid is thought to be obligatory for the efficient activation of H₂O₂ by CYP152 enzymes whereby the substrate carboxylate serves as a general acid to facilitate proton rearrangement for the generation of P450-I.^40,41 In transient kinetic studies of OleT, we have found that Ole-I is detected in reactions with H₂O₂ only when the enzyme is bound to FA.² The efficient turnover number, high chemoselectivity, and Cβ selectivity of shorter CL FAs with OleT suggest that these substrates should occupy a position very similar to that of EA within the active site. Stopped flow studies directly interrogated the reaction of C₁₀ FA [decanoic acid (DA)] with OleT. Rapid mixing of the ternary OleT:DA-d₁₉ complex with H₂O₂ resulted in the formation of readily detectable levels of Ole-I within the dead time of the instrument (Figure 3A). Over the course of 300 ms, Ole-I decayed to the low-spin Fe³⁺–OH₂ enzyme. Although the spectroscopic features of Ole-I generated with DA-d₁₉ are identical to those we have observed with EA, the level of accumulation is markedly lower for the shorter substrate (<20% for DA vs >60% for EA). In principle, this could arise from an enhanced reactivity (C–D abstraction rate) of Ole-I with DA-d₁₉, compromised H₂O₂ activation, or a more complex kinetic behavior. To address these possibilities, the absorbance changes of Ole-I were monitored at 370 nm using a photomultiplier tube (PMT). When using high H₂O₂ concentrations, the time course revealed a biphasic behavior that could be fit to a two-summed exponential expression with well-resolved reciprocal relaxation times (RRT₁ ~ 60 s⁻¹, and RRT₂ ~ 8 s⁻¹) (Figure S4). The RRTs are similar to the decay rate constants of Ole-I and Ole-II in the reaction of EA-d₃₉ (80 and 10 s⁻¹, respectively). This suggests that the poor accumulation of Ole-I did not result from a faster reaction of Ole-I with DA. The H₂O₂ dependence of both RRTs was examined, allowing for a full dissection of the kinetic rate constants for the DA reaction and efficiency of H₂O₂ activation. RRT₁, which corresponds to the rate of Ole-I decay, revealed a hyperbolic dependence on H₂O₂ concentration. RRT₂, the decay rate constant for Ole-II, was independent of H₂O₂ (Figure 3B). Using nonlinear kinetic fitting procedures, we measured an Ole-I decay rate constant (k₁ = 62 ± 4 s⁻¹), the asymptote of the plot, and an apparent K_D for H₂O₂ ${(\mathrm{K}}_{\mathrm{D}}^{{\mathrm{H}}_2{\mathrm{O}}_2}=81\pm 23\mu M)$. The Ole-II decay rate constant (k₂) was 9 ± 1 s⁻¹. Thus, the activation of OleT and the kinetic rate constants for Ole-I and Ole-II decay are similar to those of short (DA) and long CL (EA) substrates. This is consistent with the similar chemical nature [i.e., C–D bond dissociation energy (BDE)] of both classes of substrates and the largely similar reaction outcome.

Figure 3.

Transient kinetic studies of the reaction of OleT bound to perdeuterated decanoic acid (DA-d₁₉) and H₂O₂ at 4 °C. (A) Photodiode array spectra show that Compound I decays over the course of 300 ms to the ferric low-spin enzyme. (B) The decay of OleT Compound I (Ole-I) was fit to a biexponential expression and plotted vs H₂O₂ concentration, revealing the rate constants of Ole-I and Ole-II decay (~60 and 10 s⁻¹, respectively).

OleT Compound I-Catalyzed Oxidation of Alternative Substrates

To expand the inventory of substrates that could be probed in reactivity studies of Ole-I, particularly for the radical clock measurements described below, we have employed the “decoy” strategy from the pioneering work of Watanabe, Shoji, and co-workers.^41–43 This has proven to be a highly adaptable approach for many CYP152 enzymes and permits BSβ and SPα to oxidize substrates that do not contain a carboxylate. We directly tested whether the binding of a short CL FA would support the formation of Ole-I yet still permit sufficient access of a secondary molecule for oxidation. Enzyme–substrate (E–S) complexes were prepared with EA-d₃₉ or deuterated C₁₂ FA (lauric acid, LA-d₂₃) and rapidly mixed with 10 mM H₂O₂ in the presence or absence of the radical clock norcarane. When using EA-d₃₉, the initial 1 ms spectra recorded after mixing revealed the presence of an admixture of species. These included Ole-I, with a Soret maximum at 370 nm and an additional band at 690 nm, Ole-II, and the Fe³⁺–OH₂ product state (Figure 4A, red spectrum). The inclusion of norcarane did not significantly alter the relative composition of these species (Figure 4A, black spectrum). The formation and decay of Ole-II were followed at 440 nm using PMT detection under both conditions. The time courses are nearly superimposable (Figure 4B). In the presence of norcarane, the measured rate constants for Ole-II formation (Ole-I decay) and Ole-II decay were 83 ± 3 and 10 ± 2 s⁻¹, respectively. These values are identical to those we have measured for the reaction of EA-d₃₉ in the absence of a secondary molecule.^2,29 This demonstrates that in the presence of a long chain FA, norcarane cannot efficiently access the heme iron to react with Ole-I.

Figure 4.

Transient kinetic studies of the reaction of OleT bound to perdeuterated eicosanoic acid (EA-d₃₉) or lauric acid (LA-d₂₃) with H₂O₂ at 4 °C in the presence (black) or absence (red) of the radical clock norcarane. (A and C) Photodiode array spectra taken at 1 ms after mixing. (B and D) Time course of OleT Compound II formation and decay monitored at 440 nm.

Rapid mixing of the LA-d₂₃-bound enzyme with H₂O₂ also revealed the presence of Ole-I at levels of accumulation that were similar to those of reactions of DA-d₁₉ (Figure 4C, red trace). In the presence of norcarane, however, the 1 ms spectra revealed a significantly different distribution of species (Figure 4C, black trace). The absorption maximum corresponding to Ole-I at 370 nm was significantly lower and accompanied by the presence of higher levels of the Fe³⁺–OH₂ product form of the enzyme with a Soret peak at 417 nm. When the time course at 440 nm was examined, it was also apparent that Ole-II also failed to accumulate when norcarane was included in the reaction mixture (Figure 4D). However, an examination of the RRTs for Ole-II formation and decay indicated that the kinetic parameters were largely insensitive to the presence of the small molecule. The observed RRT values for Ole-II formation (RRT₁) were 105 ± 10 and 66 ± 3 s⁻¹ and the decay rate constants of Ole-II (RRT₂) 10.4 ± 0.5 and 12.5 ± 0.6 s⁻¹ for reactions in the absence and presence of norcarane, respectively. Provided that the activation of H₂O₂ is insensitive to the presence of small molecules such as norcarane, a decrease in the level of accumulation of both OleT ferryl intermediates suggests a branched oxidation mechanism from Ole-I. This implies that the products of radical clock oxidation characterized below derive from the same Ole-I oxidant that initiates the decarboxylation reaction sequence. Additionally, the lower level of accumulation of Ole-II with norcarane suggests that reactions with the radical clock substrate are substantially different from those with FAs and are likely accompanied by a much faster rebound step.

The “decoy” approach was further optimized with a series of short CL FAs using the oxidation of guaiacol to the polymerized tetraguaiacol product as a facile colorimetric probe. C₈ and C₁₀ FAs were the most proficient for producing tetraguaiacol (Figure S5). We next tested whether changing the isotopic composition of DA would further facilitate the metabolism of a secondary probe substrate. The guaiacol oxidation rate improved approximately 5-fold when using the more inert DA-d₁₉. This is consistent with the transient kinetic data for norcarane, whereby both the FA and secondary substrate, guaiacol in this case, are competitively oxidized by Ole-I. Enhanced guaiacol oxidation with short CL FAs may result from direct entry of the small molecule into the active site, as with norcarane, or by opening a more distal site in the enzyme pocket for guaiacol to bind for ensuing long-range electron transfer. Intriguingly, perfluorinated DA failed to promote guaiacol oxidation, most likely because of an unfavorable binding configuration or alteration of the carboxylate pK_a.

Considering the efficient incorporation of oxygen into product alcohols, we sought to time the radical recombination reaction. Radical clock substrates have been widely used for many decades to probe the reaction mechanisms of C–H oxidizing enzymes and biomimetic model systems.^{20,33,44–60} However, these probes have not been examined with H₂O₂-utilizing CYPs. The efficient oxidation of guaiacol using DA-d₁₉ served as a template for testing whether a similar approach could facilitate the oxidation of norcarane (bicyclo[4.1.0]heptane) and methylphenylcyclopropane (MCP). The observed product yields for radical clock substrates were highest when using DA-d₁₉. A representative GC chromatogram of the chloroform-extracted major reaction products of norcarane in the presence of DA-d₁₉ is shown in Figure 5. The peaks were well-resolved and easy to analyze on the basis of MS data and our earlier assignments. The distribution of major species is shown in the inset of Figure 5 and summarized in Table 3.

Figure 5.

Total ion chromatogram of the products from the oxidation of the radical clock norcarane using perdeuterated decanoic acid (DA-d₁₉) to trigger the formation of OleT Compound I using H₂O₂. The product ratios of major products are shown in the inset.

Table 3.

Product Distribution Profiles of Mechanistically Significant Products from the Oxidation of Norcarane and Methylphenylcyclopropane in the Presence of DA-d₁₉

norcarane	product yield (%)							radical lifetime (ps)
	cation	radical	E	G	F	I	C
trial 1	2.6	0.75	51.5	38.9	1.9	0.55	3.8	42
trial 2	1.4	0.47	47.6	33.4				29

methylcyclopropane	product yield (%)		radical lifetime (ps)
	ring-opened	ring-closed
trial 1	59.9	40.1	2.2
trial 2	58.1	41.9	2.4

The oxidation of norcarane by OleT produced an informative array of hydroxylation and desaturation products. The major products were exo- (G) and endo-2-norcaranol (E), as has been generally observed for other P450 hydroxylases. Significant amounts of the desaturation product, 2-norcarene, and its hydroxylated derivative, 4-hydroxy-2-norcarene (C), were also identified. These products are thought to derive either from a substrate cation through deprotonation or from a second C–H abstraction adjacent to the incipient radical.⁵³ However, no 3-norcarene was detected. In addition, both the radical-rearranged product, 3-hydroxymethylcyclohexene (D), and the cation-rearranged product, 3-cycloheptenol (A), were also observed in readily detectable amounts. Notably, very similar amounts of the cation-rearranged product were observed in studies of the fungal heme-thiolate hydroxylase APO.⁶¹ On the basis of the averaged ratio of rearranged products to norcaranol (G + E), a radical lifetime of ~35 ps was calculated. Using the same approach with DA-d₁₉ as a decoy, the oxidation of MCP also afforded both unrearranged and rearranged alcohol products (Figure S6). While mechanistically less informative, the product ratio is also consistent with the presence of a short-lived (picosecond) substrate radical.

DISCUSSION

We have examined the ability of OleT to convert short chain fatty acids (C₆–C₁₂), those most valuable for the development of biosynthetic strategies for fuel production, into alkenes. These compounds were evaluated as substrates for OleT using equilibrium binding, steady-state, and transient kinetic methods. The binding was markedly weaker for this set of substrates and was accompanied by a lower degree of high-spin conversion of the heme iron. Nonetheless, these shorter chain substrates efficiently activate H₂O₂ to generate Ole-I, and OleT can capably produce 1-alkenes with this panel of substrates. This is in contrast to some of the other recently identified enzymes that catalyze hydrocarbon synthesis that may have a more narrowed substrate scope.²²

Importantly, however, various amounts of Cβ hydroxylation (10–28%) accompanied the desaturative decarboxylation of shorter chain length fatty acids. For LA, hydroxylation was also observed at Cα (7%) and Cγ (22%) positions. Although they are minor products, the incorporation of labeled oxygen from H₂¹⁸O₂ into the alcohol of the hydroxyacid demonstrates that OleT is capable of prototypical aliphatic oxidations. The radical clock substrates norcarane and methylphenylcyclopropane (MPC) were also hydroxylated in the presence of deuterated FAs. This result indicates that the same Ole-I high-valent intermediate can catalyze two very different reactions, decarboxylation and hydroxylation.

A key question this work set out to answer is what controls the bifurcation between hydroxylation and decarboxylation in OleT. The Compound I intermediate generated by CYPs and porphyrin models is a very versatile reagent for chemical transformations. Depending on the balance between the nature of the axial ligand, steric hindrance near the active site, and the electronic structure of the substrate, this ferryl porphyin π-cation radical is capable of initial hydrogen atom abstraction that can be followed by either •OH rebound, heteroatom rebound, abstraction of a second hydrogen atom to form an internal double bond, or transfer of an electron from the substrate to form a substrate cation from which C–C bond cleavage reactions such as decarboxylation can occur.^12,62 A number of studies have sought to clarify the active-site requirements that allow OleT to steer the reaction coordinate toward decarboxylation. Particular focus has centered on potential proton donation pathways (e.g., His85)^1,30,63 and hydrogen bonding interactions with solvent⁶⁴ for tuning the barrier for oxygen rebound. Although we cannot exclude a possible contribution of these interactions for tuning iron–oxo reactivity, on the basis of the results presented herein, we propose that the position of the fatty acid substrate, bound to key residues in the active site, is critical for steering the reaction at several stages of catalysis. First, substrate binding allows proton rearrangement of the bound peroxide and facilitates the formation of Ole-I. Without the substrate carboxylic acid, the peroxide bond is not cleaved and Ole-I cannot efficiently form. Binding is also accompanied by a conversion to the high-spin state. Thus, this switch also provides a vacant coordination site for peroxide to bind. As a consequence of incomplete spin-state conversion upon binding shorter FAs, the relevant ferryl intermediates accumulate to lower levels but have very similar reactivities. The importance of the substrate for the generation of Ole-I is also supported by the data from the oxidation of radical clocks. Norcarane is a substrate for OleT only when a carboxylic acid is introduced.

Once Ole-I is generated, it abstracts a hydrogen atom from the FA β-carbon. However, after the formation of Ole-II, we propose that the substrate is held too firmly in place by the substrate binding pocket to move the additional angstrom or two required to allow the enzyme to donate ^•OH back to the substrate. This scenario is reminiscent of Fe²⁺- and α-ketoglutarate-dependent halogenases such as SyrB2, which can transfer either a ^•Cl or ^•OH to substrates based on exquisite (subangstrom) positioning of the carbon radical.⁶⁵ For SyrB2, orientation of the amino acid substrate relative to the rebound species is reinforced by its attachment to a phosphopantetheine arm. The substrate binding pocket of OleT must be exquisitely arranged to orient the Cβ to frustrate the rebound step and allow time for oxidation of the substrate radical. The crystal structure of EA-bound OleT¹³ and the salient protein–ligand interactions generated in Ligplot⁶⁶ are shown in Figure 6. The FA orientation is maintained through electrostatic interactions with Arg245 and an extensive series of hydrophobic contacts with the enzyme at C₈ and below. An ~45° bend in the acyl chain originates at C₁₂. Molecular dynamics studies have suggested that interactions with several of these residues, including Leu78, Phe173, Pro246, and Val292, significantly contribute to the substrate binding free energy and may serve to maintain the Cβ hydrogen at a relatively long distance (as much as 5.2 Å for C₈) from the oxoferryl species.⁶⁷ Although some anchoring at the bend of the fatty acid is lost with short CL substrates, our data show that they can slide into the correct orientation to ensure Cβ abstraction and are sufficiently immobile to enforce decarboxylation. The importance of substrate positioning has been reiterated in recent mutagenesis studies, which have demonstrated that substitution of Arg245 results in compromised activity, likely because of poor activation of H₂O₂, and switches the enzyme to the hydroxylation of shorter CL (C_n, where n ≤ 16) FAs.³⁰ Thus, an attractive means of engineering OleT for the decarboxylation of shorter substrates may be to further constrict the active-site channel and reinforce contacts at various positions along the acyl chain.

Figure 6.

(A) Crystal structure of eicosanoic acid-bound OleT (Protein Data Bank entry 4L40).¹³ (B) LigPlot⁶⁶ diagram that shows the residues involved in stabilizing the bound fatty acid in the active-site pocket.

As we have shown in reactions of Ole-II with phenols,²⁹ and in reactions with Aae-APO-II and a wide range of substrates,²⁷ iron(IV)–hydroxides are competent oxidants that can abstract an additional electron from the substrate, forming a cation at the Cβ position. From there, elimination of CO₂ and formation of an alkene are highly favorable. An example of this chemistry occurs in isopentenyl pyrophosphate biosynthesis.⁶⁸ An alternative proposal has suggested that once Ole-II is formed from Cβ-H abstraction, it could abstract an electron from the substrate carboxylate to afford a diradical, which could then rearrange to form an alkene and CO₂. While this chemistry is plausible, it would require a tremendous amount of flexibility of the substrate in the pocket during catalysis to ensure that both abstraction steps were favorable. The substrate movements that would be required are inconsistent with product profiles for FAs, which show that the enzyme is highly regioselective.

CONCLUSIONS

In summary, we have confirmed that OleT can catalyze both decarboxylase and monooxygenase reactions. If an FA is bound and an alternative substrate is offered, OleT can behave like a typical CYP monooxygenase and hydroxylate aliphatic substrates that do not contain a carboxylate. A similar result is observed when an active-site carboxylate is introduced into the protein framework⁶⁹ to mimic the strategem used by APO.⁷⁰ The radical lifetimes measured for OleT with norcarane, which is a probe sensitive enough to be used as a diagnostic tool for classifying monooxygenases,⁷¹ and methylphenylcyclopropane, are indistinguishable from those that have been measured with other CYPs. After initial hydrogen atom abstraction, a caged substrate radical is formed that undergoes fast (picosecond) oxygen rebound.⁷² Slightly larger than average amounts of cationic-derived product, relative to the amounts seen with other CYPs, were observed with norcarane. This is consistent with a model in which Ole-II is an effective one-electron oxidizing agent. The amount of desaturated products from norcarane is comparable to what has been seen with other enzymes and, as we have argued in other places, a reflection, in part, of the chemistry of norcarane.⁵³ Collectively, these results point toward substrate identity and coordination, rather than electronic structure, as a key element in dictating the reaction outcome.

It is useful to reflect on the available (bio)chemical strategies for the synthesis of deoxygenated biofuels. Many mechanisms for generating biofuels rely on expensive reducing equivalents to release water or catalyze similar types of C–C bond cleavage. In contrast, the approach illustrated by OleT, and shared with the non-heme iron enzyme UndA,²² is an oxidative approach that consumes a toxic reactive oxygen species (H₂O₂) and releases CO₂. For commercialization, the OleT chemistry (or inorganic chemistry modeled on it) would ideally be coupled to CO₂ capture or utilization. OleT also offers inspiration for synthetic model chemistry: to design and synthesize a model compound or protein that generates a high-valent intermediate capable of C–H abstraction but then holds the substrate at bay and forces it to decarboxylate.