Experimental Evidence for Radical-Polar Crossover in Competition with, Rather than Preceding, Alkene Formation by Microbial Ethylene-Forming Enzyme (EFE)

Abstract

Ethylene-forming enzyme (EFE) catalyzes a reaction that sets it apart from other iron(II)- and 2-oxoglutarate-dependent (Fe/2OG) oxygenases. In this reaction, all four oxidizing equivalents of O₂ are unleashed upon 2OG, fragmenting it to ethylene (from C3 and C4) and three fully oxidized C₁ equivalents (from C1, C2, and C5), while the would-be “prime substrate,” L-arginine, escapes unmodified. We previously proposed that ethylene formation proceeds by a radical-polar-crossover mechanism involving three unusual steps: (1) formal insertion of O₂ between C1 and C2 of 2OG, forming a succinylperoxycarbonatoiron(II) complex and appending an additional oxygen to C1; (2) radical C–O coupling between a C3-C5-derived propionate-3-yl radical and a C1-derived, Fe(III)-coordinated carbonate; and (3) polar fragmentation of the resultant (2-carboxyethyl)carbonatoiron(II) complex to ethylene, CO₂, and carbonate. Here, we used isotopic labeling to distinguish the three C₁ products and stopped-flow infrared absorption (FTIR) spectroscopy to track their formation. The results confirm the prediction that C1 is not directly converted to CO₂, implying that it must indeed become (bi)carbonate. Comparable kinetic data on the A198L variant, which produces ethylene and the abortive product, 3-hydroxypropionate, in similar quantities, reveal that these two products do not, as we had originally proposed, form in competing reactions of a common (2-carboxyethyl)carbonatoiron(II) intermediate. Rather, as suggested by a pair of computational studies separately led by Sayfutyarova and Christov, ethylene is formed in competition with radical coupling by an olefin-forming fragmentation that reduces the Fe(III) cofactor. In other words, crossover to the polar manifold thwarts, rather than enables, ethylene formation.

Graphical Abstract

Introduction

Ethylene-forming enzyme (EFE) is an unusual iron(II)- and 2-oxoglutarate-dependent (Fe/2OG) oxygenase that catalyzes two competing reactions.^1,2 It uses the well-established, canonical mechanism of the Fe/2OG-oxygenase superfamily in its minor (~ 30%) pathway, in which the co-substrate, 2OG, is decarboxylated to succinate and an oxoiron(IV) (ferryl) intermediate abstracts a hydrogen atom from C5 of L-arginine to fragment the “prime substrate” to guanidine and pyrroline-5-carboxylate (Scheme 1A, lower branch).³ In its major (70%), namesake pathway, L-arginine serves only as an essential activator, and all four oxidizing equivalents of O₂ are unleashed upon 2OG in fragmenting it to ethylene (from C3 and C4) and three fully oxidized C₁ equivalents (from C1, C2, and C5) (Scheme 1A, upper branch).^3,4 EFE has attracted attention in recent years both for its unique mechanistic features, which broaden the chemical repertoire of the Fe/2OG enzyme class, and for its potential to enable sustainable biological production of ethylene from a ubiquitous primary metabolite.^5–8

Scheme 1:

Reactions and possible mechanisms of EFE. (A) Three reactions catalyzed by EFE. (B) Mechanism of L-arginine oxidation and propionate-3-yl radical formation proposed in prior work. (C) Previously proposed mechanisms of ethylene and 3HP formation via common (2-carboxylethyl)carbonatoiron(II) precursor. (D) Revised mechanism from this work involving direct fragmentation of propionate-3-yl radical to ethylene in competition with formation of the (2-carboxylethyl)carbonatoiron(II) complex that yields only 3HP.

In prior work, we established that the ferryl complex (Scheme 1B, RO2), the defining intermediate of the Fe/2OG superfamily, accumulates only in the minor, L-arginine-oxidizing (RO) pathway and not in the major, ethylene-forming (EF) pathway.³ Evidence included demonstration that the variant protein with iron ligand aspartate 191 replaced by glutamate (D191E) reacts almost exclusively (~ 96%) through the RO pathway and accumulates approximately four times as much ferryl complex as the wild-type (WT) enzyme. In a subsequent study, we proposed that the EF pathway involves exclusively mid-valent iron species and proceeds via a radical-polar-crossover mechanism. In the proposed mechanism, the radical manifold is accessed by O–O-bond homolysis of a succinylperoxycarbonatoiron(II) complex (Scheme 1B, EF1), which forms by net insertion of O₂ between C1 and C2 of 2OG,⁹ a step that, at that time, lacked experimental precedent but had been predicted in multiple computational studies.^10–14 These studies proposed formation of the novel complex either from the precedented peroxysuccinatoiron(II) complex by recapture of the C1-derived CO₂ between the iron and its acylperoxide ligand or from the superoxoiron(III)•2OG complex by direct insertion of the O₂ unit into the C1–C2 bond. Although the mechanism of its formation remains unresolved, we provided experimental evidence for the intermediacy of the succinylperoxycarbonatoiron(II) complex by showing that, exclusively in the EF pathway, oxygen atoms from O₂ are appended to both C1 and C2 of 2OG, an outcome not previously seen from an Fe/2OG oxygenase. In FTIR measurements on the gas from acidified EFE reactions with 1-¹³C₁-2OG and ¹⁸O₂-saturated buffer, we observed the four peaks characteristic of ¹⁶O=¹²C=¹⁶O (626 in the Air Force Geophysics Laboratory (AFGL) shorthand notation), ¹⁶O=¹³C=¹⁶O (636), ¹⁶O=¹²C=¹⁸O (628), and ¹⁶O=¹³C=¹⁸O (638) (Scheme 1A), the latter two species revealing that both C2 (628) and C1 (638) of 2OG form a new bond with an atom of O₂, as they would in the posited C1–C2/O₂-insertion step (Scheme 1B, EF1) but not in the C1/O₂-replacement step (leading to RO1) that occurs in other Fe/2OG enzymes and the minor pathway of EFE.⁹ In ensuing steps, peroxide-bond homolysis from EF1 and β-scission of the resultant succinate-1-yl radical in EF2 were proposed to generate a propionate-3-yl radical adjacent to a carbonatoiron(III) complex (EF3), from which a novel C–O-bond-forming radical-ligand-transfer (RLT)¹⁵ step was envisaged to enact crossover to the polar manifold (Scheme 1C). The resultant (2-carboxylethyl)carbonatoiron(II) complex was proposed then to undergo a Grob-like polar fragmentation—with the C5-derived carboxylate acting as electrofuge and the C1-derived carbonate as nucleofuge (Scheme 1C, green arrows, upward branch)—to produce ethylene.

The key evidence for the proposed radical-polar-crossover mechanism was the observation that EFE produces a partially fragmented ω-hydroxyacid product when challenged with any of several C4-substituted 2OG analogs (methyl, hydroxy, fluoro, difluoro).⁹ In a subsequent paper, we showed that the enzyme produces a small quantity (0.5%) of the corresponding product, 3-hydroxypropionate (3HP), even from the native 2OG substrate.^{16 18}O-tracer experiments revealed that the alcohol oxygen of the ω-hydroxyacid product originates from O₂ in precisely half of events and from the C1 carboxylate of the 2-oxodiacid substrate in the other half.^9,16 We explained this result as arising from (1) facile rotation of the stably Fe(III)-coordinated carbonate (Scheme 1B, EF3, curved arrow) before attack by the propionate-3-yl radical on one of its non-coordinating oxygen atoms—of which one originates from the C1 carboxylate and the other from O₂ via the C1-C2/O₂-insertion step (Scheme 1B, top)—and (2) elimination of CO₂ from the unstable carbonate monoester to form the ω-hydroxyacid abortive product (Scheme 1C, black arrows, downward branch).

Rationale for Current Study: (1) Test for Conversion of C1 to (Bi)carbonate Rather than CO₂.

A prediction of our mechanistic hypothesis that we did not adequately test in the prior studies^9,16 is that C1 of 2OG would initially be converted to (bi)carbonate ion rather than directly to CO₂. In that work, the acidification of the reaction solution that enabled gas-phase detection of all three oxidized C₁ equivalents would necessarily have obscured the initial form of the C1-derived product. Accordingly, in the present study, we constructed a hand-mixing stopped-flow (SF) apparatus that we interfaced with a commercial Fourier-transform infrared (FTIR) spectrometer to enable acquisition of spectra on the seconds timescale, which—although not state-of-the-art for an SF-FTIR instrument—is fast enough to enable detection of CO₂ before it hydrates (Figure S1). With data obtained on this apparatus, we show here that the expected FTIR signature of CO₂ derived from C1 via the EF pathway is essentially undetectable after the signatures of the C2- and C5-derived CO₂ equivalents have fully developed (before they subsequently decay by hydration). This result establishes that C1 does not initially form CO₂ and, therefore, must indeed be converted to (bi)carbonate. In addition to confirming a key tenet of the proposed mechanism, the instrument and approach provide the experimental foundation for clarifying the nature of the ethylene-yielding step, as explained below.

Rationale for Current Study: (2) Test Hypothesis of Grob-like Ethylene-Yielding Step.

A key question that remained following our initial study⁹ was whether the (2-carboxyethyl)carbonatoiron(II) complex arising from the native 2OG substrate lies on or off the EF pathway. In the on-pathway scenario that we favored (Scheme 1C), this complex was envisioned to partition between 3HP-yielding elimination (black arrows, downward branch) and ethylene yielding Grob-like fragmentation (green arrows, upward branch). In the off-pathway scenario (Scheme 1D), the propionate-3-yl/carbonatoiron(III) RLT step (black arrows, downward branch) would compete with direct fragmentation of the propionate-3-yl intermediate (green arrows, upward branch), such that the alkene would form only when the (2-carboxyethyl)carbonatoiron(II) complex does not. In our ensuing study, we attempted to distinguish between these two possibilities by introducing sterically perturbing substitutions in the 2OG binding pocket near C3 and C4 and examining their impacts on the ethylene:3HP partition ratio.¹⁶ We anticipated that, in the on-pathway scenario, the polar-concerted fragmentation would require strict antiperiplanar alignment of the carboxylate electrofuge and carbonate nucleofuge and thus be readily disrupted by these substitutions, whereas the elimination reaction would be insensitive to the steric perturbations. For the off-pathway scenario, we anticipated that a direct radicaloid fragmentation of the propionate-3-yl intermediate would also be relatively insensitive to subtle repositioning of the nascent olefinic carbons (C3–C4 of 2OG), especially given evidence from our prior work for facile rotation about the C3–C4 bond at some intermediate stage of the reaction.⁹ We identified variants with bulk-enhancing substitutions that produce 3HP as their major product, suggesting derailment of the polar-concerted fragmentation, and others with bulk-diminishing substitutions that enable production of propylene from (4R)-methyl-2OG, a bulkier analog that is converted exclusively to the ω-hydroxyacid product (2-methyl-3-hydroxypropionate) in the major pathway of the WT enzyme. We interpreted the success of this approach—in keeping with its design logic—as evidence for the on-pathway scenario and polar-concerted ethylene-yielding step. However, two recent computational studies, by Wang and Sayfutyarova¹⁴ and by Rifayee, et al.,¹⁷ predicted that ethylene should form directly from the propionate-3-yl species, with the competing propionate-3-yl/carbonatoiron(III) RLT step committing the reaction to the 3HP product. Rifayee et al. also calculated a prohibitively high barrier for the proposed polar-concerted fragmentation of the (2-carboxylethyl)carbonatoiron(II) complex, weighing further in favor of the off-pathway scenario.¹⁷

At the outset of the current study, we recognized that the ability to measure the kinetics of formation of the three oxidized C₁ products, resolved according to the carbons of 2OG from which they originate, in reactions of variant EFEs that make similar quantities of ethylene and 3HP could enable us to resolve the discrepancy regarding the nature of the ethylene-yielding step. Accordingly, we carried out single-turnover SF-FTIR experiments, most informatively with 1-[¹³C]-2OG and ¹⁸O₂ on the 3HP-favoring A198L EFE variant, to show that appearance of the C1-derived 638 in the abortive, 3HP-producing branch of the major pathway is markedly slower than formation of the C5-derived 626 in the ethylene-yielding branch. The failure of these two CO₂ equivalents to form with identical kinetics, as they necessarily would if arising from competing fates of the common (2-carboxylethyl)carbonatoiron(II) intermediate, conclusively rules out the on-pathway scenario (Scheme 1C) in favor of the off-pathway case (Scheme 1D). Thus, a revised mechanism for EFE and its variants emerges, now supported by both computational and experimental evidence, in which the second reaction branchpoint between ethylene and 3HP production involves a competition between the propionate-3-yl/carbonatoiron(III) RLT step discovered in our prior work and a radicaloid C4–C5-bond fragmentation that resets the cofactor to its resting Fe(II) state by coupled electron transfer (Scheme 1D). This new consensus reopens the question of the precise nature of the ethylene-yielding step, which has been viewed differently in the multiple computational studies, and the mechanism by which either C4 substitutions of 2OG or bulk-enhancing amino-acid substitutions near C4 disfavor this fragmentation with respect to the competing RLT step.

Experimental

FTIR Spectroscopy.

FTIR spectra were collected at 4 cm⁻¹ resolution on a Nicolet is50 FTIR (Thermo Scientific, Waltham MA) spectrometer equipped with an LN₂-cooled MCT-A detector with a 3.6 μm long-pass filter (Edmund Optics, Barrington NJ). Experiments were conducted in cell consisting of two 32 mm, 3 mm thick CaF₂ windows with a 40 μm PTFE spacer coated in silicone grease for a liquid tight seal when compressed. One window of the cell had inlet and outlet holes (Pike Technologies, Fitchburg WI, PN:160–1142) to interface with the home-built SF device, as described in detail in the Supporting Information. Background and static control spectra were collected with 1,000 scans, while kinetic data were collected using the OMNIC 9 Macro program to dynamically change the number of scans per unit time and thereby generate a pseudo-logarithmic time scale. The first 11 spectra (3 – 40s) were acquired with 2 scans/spectrum, the next 10 spectra (46 – 102s) were acquired with 6 scans/spectrum, the next 10 spectra (120 – 274s) were acquired with 24 scans/spectrum, the next 10 spectra (316 – 692s) were acquired with 65 scans/spectrum, and the final 19 spectra (773 – 2230s) were acquired with 130 scans/spectrum.

EFE Kinetic Assays Using the Stopped-Flow FTIR Apparatus.

A detailed description of the constructed apparatus for manual rapid mixing and coupling to the spectrometer is provided in the Supporting Information. The apparatus was assembled and connected to a circulating water bath, which was operated for at least 1 h to allow the system to stabilize at 4 °C. The protein syringe was first prepared inside an MBraun (Stratham, NH) LABmaster anoxic chamber. An O₂-free solution of 1.4 mM EFE, 1.26 mM ferrous ammonium sulfate (Fe(NH₄)₂(SO₄)₂(H₂O)₆), 5 mM 2OG, 5 mM L-arginine, and 100 mM sodium HEPES (pH 7.5) was prepared in sufficient volume to afford ≥ 300 μL per required SF trial and loaded into the Hamilton syringe. An equal volume of O₂-saturated 100 mM sodium HEPES (pH 7.5) buffer was loaded into the second Hamilton syringe. The syringes were removed from the glovebox and connected to the stopped-flow apparatus. The syringe plungers were simultaneously depressed using the syringe holder device over a mixing period of ~ 0.5 s, and collection of FTIR spectra was simultaneously manually actuated.

Data Processing and Analysis.

A background spectrum, collected over 1,000 scans, was subtracted from each kinetic FTIR spectrum. The baseline was then corrected, excluding regions with CO₂ absorbance, using a home-built software package (Kazan viewer, https://github.com/AlexeySilakov/KazanViewer). Data plotting and fitting were performed in Igor 9 Pro (Wavemetrics, Lake Oswego, OR).

Results

SF-FTIR Analysis of CO₂ Products Using 2OG Isotopologues and EFE Variants with Altered Partitioning.

After rapid mixing of the anoxic EFE•Fe(II)•2OG•L-arginine complex with O₂-saturated buffer, we detected the sharp absorption feature at 2,343 cm⁻¹ from dissolved 626^18,19 in the first spectrum (Figure 1, front row), which, on our home-built apparatus, corresponds to a reaction time of ~ 3 s. Our previous study showed that a single turnover of wild-type (WT) EFE is complete within ~ 5 s, and it is therefore consistent with expectations that CO₂ production would be largely complete by the time of the first SF-FTIR measurement. Indeed, at longer reaction times, the 2,343-cm⁻¹ feature decays as the dissolved 626 produced by EFE undergoes hydration. The observed rate constant for this decay phase (~ 0.004 s⁻¹) is consistent with reported values for CO₂ hydration under these reaction conditions.²⁰ With [¹³C₅]-2OG as substrate, the absorption maximum shifts to 2,277 cm⁻¹, characteristic of 636 (Figure 1, second row). With substrate labeled only at C1 (1-¹³C₁-2OG), both 626 (from C2 and C5) and 636 (from C1) peaks develop (Figure 1, third row). In this case, 626 should arise only from the EF pathway, as confirmed by the barely detectable peak in the corresponding reaction of the D191E variant (Figure 1, fourth row), which promotes < 5% flux through the EF pathway. In the WT EFE reaction, the 636 product should, under these reaction conditions, arise from the RO pathway and possibly also the EF pathway, depending on the identity of the C1-derived product in the major reaction. The time-dependent spectra, specifically the ratios of the 626 and 636 peaks, are inconsistent with the direct production of CO₂ from C1 in the EF pathway (Figure 1, third row). In this case, the 636:2OG stoichiometry should be unity (1:1), whereas, because 626 should be produced from both C2 and C5 but only in the EF pathway (70%), the 626:2OG stoichiometry should be 1.4. Thus, the ratio of 636:626 would be 1:1.4 (0.71) for the case of direct production of CO₂ from C1 in the EF pathway. By contrast, for the case of production of (bi)carbonate from C1 in the EF pathway, the 636:626 ratio should be 0.3:1.4 (0.21), because the 636 product would then arise only from C1 in the minor RO pathway, while the 626 product would still arise from both C2 and C5 in the EF pathway. The actual experimental ratio of peak intensities of 0.18 ± 0.03 (mean and 99% confidence interval of the first 10 spectra from 3 independent trials) agrees well with the prediction of the proposed (bi)carbonate-yielding mechanism for the EF pathway but is significantly less than the value of 0.71 predicted for the case of direct production of CO₂ from C1 in both pathways. This conclusion is confirmed by the corresponding data on the 3HP-favoring A198L variant (Figure 1, back row), in which the 626 peak is suppressed by the failure of the final ethylene-yielding step (which normally produces 626 from C5) in the majority of events through the major pathway but the 636 peak is enhanced, because these failure events make both 3HP and a C1-derived CO₂. Here, the 636 peak would be enhanced to an even greater extent, if not for the fact that it is produced slowly enough in the 3HP-yielding pathway to allow competing hydration to kinetically suppress the maximum peak intensity (see below).

Figure 1:

Time-resolved FTIR absorption spectra from reactions of wild-type (WT) EFE and its D191E and A198L variants. Reactions were initiated by rapidly mixing a buffered anoxic solution of enzyme, Fe(II), L-arginine, and natural-abundance (NA) or ¹³C-labelled 2-oxoglutarate with an equal volume of the same buffer saturated with O₂. Reactant concentrations are given in Experimental section above. Spectra were acquired from ~ 3 to 2230 seconds after mixing.

Dual-Isotope Approach to Isolate the C1-derived CO₂ Equivalent in the 3HP-Yielding Abortive EF Pathway.

To gain additional resolution of the oxidized C₁ products from the three reaction pathways (RO, successful EF, and abortive EF), we repeated the SF-FTIR experiment with ¹⁸O₂. In the WT reaction with 1-[¹³C]-2OG and ¹⁸O₂, the 638 isotopologue should arise uniquely from C1 in the 3HP-yielding abortive EF pathway: only the C1–C2/O₂ insertion (not the C1/O₂ replacement) appends an ¹⁸O atom to C1, the sole site of ¹³C in the substrate, and only the 3HP-yielding branch (not the ethylene-yielding branch) affords CO₂ rather than (bi)carbonate as co-product (Scheme 1A). Moreover, under these reaction conditions, 626 must arise uniquely from C5 and 628 uniquely from C2 (which acquires an O-atom from O₂), thus potentially allowing the multiple decarboxylation steps to be resolved in time. Only the 636 isotopologue is generated by multiple pathways—the RO pathway and the half of the 3HP-yielding branch of the major-pathway in which the ¹⁸O initially appended to C1 departs with 3HP. In the reaction with WT EFE, peaks at 2,343, 2,326 and 2,277 cm⁻¹, characteristic of 626, 628, and 636,²¹ fully develop by the reaction time of the first spectrum (Figure 2A). With appropriate scaling, the kinetic traces reporting the intensities of the three peaks are almost perfectly coincident (inset). As explained above, the single EFE turnover that occurs under these reaction conditions is (nearly) complete by the first (~ 3-s) observation, and the decay of absorbance reflects merely the hydration of the three isotopologues, with the expected indistinguishable kinetics. The peak at 2,260 cm⁻¹ for 638 is absent at the early reaction times and grows in, albeit slowly and to a very modest extent (expanded view in Figure S5). This behavior again reflects the fact that 638 is not generated directly by WT EFE (except in the barely detectable fraction of events yielding 3HP). Rather, it forms after initial production of the corresponding (bi)carbonate, by slow dehydration (k_eff = 0.0044 s⁻¹)²⁰ in the events in which one of the two ¹⁶O atoms originally present in the C1 carboxylate is eliminated from the (H)C¹⁸O¹⁶O₂²⁻ (bi)carbonate product in its slow dehydation.^9–11 Importantly, the simple decarboxylation step that occurs in the RO pathway, which liberates C1 directly as CO₂, does not append an O₂-derived oxygen atom upon C1, thus yielding 636 rather than 638 in this experiment. The results of Figure 2A thus further validate the conclusion that (bi)carbonate is the immediate C1-derived product of the EF pathway.

Figure 2:

Time-resolved FTIR absorption spectra from reactions of WT (A) and A198L (B) EFE initiated by rapidly mixing a buffered anoxic solution of enzyme, Fe(II), L-arginine, and 1-¹³C₁-2OG with an equal volume of the same buffer saturated with ¹⁸O₂. Reactant concentrations are given in Experimental section above. Spectra collected from ~ 3 to 2230 seconds. Kinetic plots for scaled absorbances at 2343 (626), 2326 (628), 2277 (636), and 2260 (638) cm⁻¹ are plotted in the insets. Parameters for scaling and fitting (dashed lines) are given in Table S1. The scaled sum of absorbance at 2326 and 2260 cm⁻¹ is also plotted in the inset to panel B.

Dual-Label Experiment with 3HP-favoring A198L Variant to Assess Timing of Ethylene and C1-Derived CO₂ Production.

The alternative mechanisms shown in Schemes 1C and 1D, with the (2-carboxyethyl)carbonatoiron(II) complex either on (C) or off (D) the pathway to ethylene, are potentially distinguishable by the relative kinetics of ethylene and 3HP production. In the on-pathway case (Scheme 1C), the two products would both form with an observed rate constant, k_obs, equal to k_eth + k_3HP, because their formation would track with decay of their common (2-carboxyethyl)carbonatoiron(II) precursor. The fractional yield of each product would be the ratio of its rate constant (k_eth or k_3HP) to k_obs. In the off-pathway case (Scheme 1D), the (2-carboxyethyl)carbonatoiron(II) complex and ethylene would form contemporaneously, but, because 3HP would not be produced until the subsequent step, its formation could lag behind formation of ethylene, provided that the CO₂-elimination step would be slower than preceding steps. In either mechanism, ethylene would form along with the C5-derived CO₂ (the only 626 in the reaction with 1-[¹³C]-2OG and ¹⁸O₂) while 3HP would form along with the C1-derived CO₂ (the only 638), thus making these two CO₂ isotopologues perfect surrogates for the primary products in the dual-isotope experiment. Because we anticipated that the meager (0.5–1%) flux in the WT EFE reaction through the 3HP-yielding elimination would not afford sufficient signal to determine the relative kinetics of 626 and 638, we leveraged the A198L variant, which produces 3HP and ethylene in similar quantities, to determine which mechanistic case (on- or off-pathway, Scheme 1C or D, respectively) is operant.

The reaction of A198L EFE produces the same four CO₂ isotopologues (626, 628, 636, 638) generated by the WT enzyme (Figure 2B), but their relative quantities, as judged by peak intensities, are perturbed by the A198L substitution in a manner that can be explained by our prior study. In comparison to the spectra from the WT EFE reaction (Figure 2A), the peak of 626 at 2,343 cm⁻¹ has diminished intensity relative to the peak of 628 at 2,326 cm⁻¹, because the derailment of ethylene production in favor of 3HP through the major pathway prevents liberation of C5 as 626 but does not impair the C2–C3 fragmentation that releases C2 as 628 (with one atom of ¹⁸O from the ¹⁸O₂ substrate). The peaks of 636 at 2,278 cm⁻¹ and 638 at 2,260 cm⁻¹ are both slightly enhanced, because the CO₂ elimination that affords 3HP also produces either 636 or 638, depending on whether the O-atom appended to C1 from O₂ is ultimately conferred to 3HP or left behind on the C1-derived CO₂. Kinetically, the A198L substitution causes only a modest slowing of the reaction, which allows the tail end of the formation phases of the 626, 628, and 636 species to be visualized. The kinetic traces for the 626 and 628 products from C5 and C2 are coincident (inset), showing that ethylene formation (concomitant with 626) is not delayed relative to formation of the first CO₂ equivalent in the C2–C3-cleaving β-scission of the succinate-1-yl radical (628). This observation alone does not resolve the mechanistic question at hand, because, in either mechanism, steps downstream of the C2–C3 fragmentation could be fast enough to render formation of all three C₁ products kinetically unresolved. However, together with the kinetics of the 638 isotopologue, which arises solely from C1 via the abortive EF pathway, the data do answer the question. Formation of the 638 product is markedly delayed relative to formation of the 626 and 628 products, implying the accumulation of an intermediate that breaks down very slowly to yield the C1-derived CO₂ equivalent unique to the abortive branch of the major pathway. By fitting the kinetics of the 638 peak, we could estimate a rate constant of 0.005 ± 0.001 s⁻¹ for its formation. This rate constant should reflect the elimination of CO₂ from the (2-carboxyethyl)carbonatoiron(II) complex, and, because there is no corresponding slow phase of 626 production from C5, it must be true that 626 and ethylene are produced in a step preceding breakdown of this complex. In other words, as predicted by the recent computational analyses, the (2-carboxyethyl)carbonatoiron(II) complex is not on the pathway to ethylene.

In the dual-isotope reaction of A198L EFE, the 636 isotopologue is produced via two different pathways: the RO pathway and the fraction of the abortive pathway in which ¹⁸O remains with 3HP. Gratifyingly, the kinetic trace for 636 (Figure 2B inset, blue squares) can be adequately reproduced by summing the appropriately scaled experimental traces for either 626 or 628 (purple circles), reflecting the pathway to ethylene (which is only modestly delayed in the variant), and 638, reflecting the very slow CO₂ elimination from the off-pathway (2-carboxyethyl)carbonatoiron(II) complex. This internal consistency further corroborates our mechanistic interpretation.

Discussion

The asynchronous production of 3HP and ethylene—which we have demonstrated here by monitoring their concomitantly produced C1- and C5-derived CO₂ co-products—in the A198L variant of EFE experimentally verifies the recent computational predictions^14,17 of an earlier second branchpoint than we had proposed in our prior studies.^9,16 This new understanding negates the hypothesis that the partition-altering effects of 2OG-C4 and sterically perturbing amino-acid substitutions arise from their differential impacts on a stereo-electronically demanding ethylene-yielding Grob-like fragmentation of the (2-carboxyethyl)carbonatoiron(II) intermediate and more permissive 3HP-yielding elimination of the same complex.¹⁶ The results raise anew the issues of (1) the precise nature of the ethylene-forming step and (2) why the 2OG and amino acid substitutions favor ω-OH-acid over ethylene formation.

The two most recent computational studies on EFE^14,17 both invoked (1) direct insertion of the O–O unit of a superoxoiron(III) complex into the C1–C2 bond of 2OG to form the succinylperoxycarbonatoiron(II) complex and (2) direct fragmentation of the propionate-3-yl radical to ethylene (intermediates and barrier heights calculated by Wang and Sayfutyarova are summarized in Figure S9). However, they invoked different pathways for the final fragmentation, which involves both an electron transfer (ET) to the Fe(III) cofactor and an olefin-forming decarboxylation (DC). Rifayee, et al. considered two possible pathways for this ET-DC step,¹⁷ each beginning from a different representative configuration (“snapshot”) of the presumptive superoxoiron(III) complex obtained through molecular dynamics (MD) simulations.¹¹ From the complex dubbed WT1-RC, β-scission of the succinate-1-yl radical was found to result in barrierless fragmentation of the nascent propionate-3-yl radical, implying that the latter species is a transition state rather than intermediate. This view of C2–C3 and C4–C5 cleavage as effectively concerted seems inconsistent with our published observations of (1) random ethylene cis/trans stereochemistry in the reaction with 3R,4S-[²H₂]-2OG, which requires that an intermediate with cleaved C2–C3 bond be sufficiently long-lived for rotation about the C3–C4 bond to randomize the initially fixed orientation of the two chiral centers, and (2) equal allocation of the ¹⁸O-atom from ¹⁸O₂ between carbonate and 3HP in the abortive pathway, which requires facile rotation about the Fe(III)–(¹⁶OC¹⁸O¹⁶O)²⁻ bond prior to the C–O-coupling RLT step leading to 3HP.⁹ From the superoxoiron(III) complex dubbed WT2-RC, the authors calculated a stepwise pathway through an authentic propionate-3-yl intermediate, more consistent with available data.¹⁷ From this intermediate, decreasing the C3•···O(CO₂–Fe) distance without additional restraints led directly to the ET-DC outcome. Thus, consideration of the Rifayee, et al. study¹⁷ together with experimental observations would suggest an EF mechanism involving a stable propionate-3-yl species that undergoes ET-DC along the approach needed to enable the C–O-coupling RLT step. This pathway is consistent with the experimental results reported herein.

Rifayee, et al. also found that restraining the C4–C5 bond (original 2OG numbering) in the propionate-3-yl intermediate allowed the C–O-coupling RLT step to proceed, leading to the (2-carboxyethyl)carbonatoiron(II) complex, from which they found no energetically accessible pathway for ethylene production.¹⁷ Our new calculations, which predict an insurmountable activation barrier of ~ 67 kcal/mol for the previously proposed Grob-like fragmentation, concur with the view that this complex can yield only 3HP (Figure S9). The fact that minimal restraint of the C4–C5 bond in the quasi-stable radical was, in the Rifayee, et al. study,¹⁷ enough to bias the reaction to formation of the (2-carboxyethyl)carbonatoiron(II) complex instead of ethylene is consistent with our observations that the reaction makes detectable quantities of both ω-OH-acid and alkene products and is readily biased toward one or the other by steric perturbations. Indeed, the observed partitioning of ~ 99.5:0.5 between ethylene and 3HP requires a difference in activation barriers of only ~ 3 kcal/mol, and the A198L substitution need only shift the relative barrier heights by < 4 kcal/mol in favor of the C–O-coupling to account for the observed effects. Although accurately parsing such small differences may exceed the capacity of the computational methods, the Rifayee, et al. calculations are consistent with closely spaced transition states for the competing reactions. In terms of the rationale for the RLT/3HP-favoring effect of the A198L substitution, the authors calculated that the increased hydrophobic interactions of L198 with the C4 carbon could stabilize the propionate-3-yl intermediate with respect to the ethylene-yielding fragmentation. With decreasing C3•···O(CO₂–Fe) distance, C–O coupling to form the (2-carboxyethyl)carbonatoiron(II) complex was predicted to proceed with a barrier equivalent to that for the WT2 pathway. When the authors tried to interrogate ethylene production from this state by elongating the C4–C5 bond, C–O coupling was still calculated to have the lower barrier, and only when an additional restraint on the C3•···O(CO₂–Fe) approach prevented this coupling could the ethylene-yielding ET-DC prevail, with an activation barrier of 5 kcal/mol.¹⁷ It would appear from their analysis that increased hydrophobic interactions selectively raise the ET-DC barrier relative to the C–O coupling barrier.

Although they reached the same top-line conclusion—that the ethylene-yielding ET-DC outcome has a lower activation barrier than the C–O-coupling RLT step—Wang and Sayfutyarova¹⁴ highlighted aspects of the reaction and possible pathways for the competing fates of the propionate-3-yl/carbonatoiron(III) intermediate not considered by Rifayee, et al.¹⁷ In interrogating decay of the intermediate by decreasing the C3•···O(CO₂–Fe) distance, Wang and Sayfutyarova calculated that a modestly (~ 7 kcal/mol) downhill, low-barrier (~ 2 kcal/mol) reconfiguration of the cofactor could switch the L-arginine guanidium hydrogen bond from the D191 carboxylate ligand to the C1-derived carbonate ligand, allowing the former ligand to become bidentate and converting the five-coordinate carbonatoiron(III) complex to a six-coordinate intermediate.¹⁴ Four different potential energy scans (PESs) from this optimized six-coordinate structure led to C4–C5 cleavage, producing ethylene and the C5-derived CO₂. The pathway with the lowest calculated barrier (5.4 kcal/mol) arose from compression along the C3•···O(CO₂–Fe) coordinate. Surprisingly, this pathway also resulted in formation of a new bond between an O-atom of the C1-derived carbonate ligand and the C2-derived CO₂ produced in the earlier β-scission. Moreover, the resultant pyrocarbonatoiron(II) product complex was calculated to have Mössbauer parameters similar to those of an experimentally detected Fe(II)-containing intermediate formed through the EF pathway but not through the RO pathway,³ lending credence to the idea of CO₂-carbonate coupling associated with ethylene formation. A second PES starting from the six-coordinate carbonatoiron(III) complex along the (increasing) C4···C5 coordinate led to ethylene formation with a slightly higher barrier (8.0 kcal/mol) and without pyrocarbonate formation. Interestingly, in the analogous PES, the five-coordinate complex preceding the coordination change (with D191 H-bonded to L-Arg and its carboxylate monodentate) was calculated to undergo the C–O coupling to form the (2-carboxyethyl)carbonatoiron(II) complex, hinting at the importance of the postulated first- and second-sphere rearrangement in control of the reaction outcome. The final two scans from the rearranged six-coordinate complex, along C2···O(CO₂–Fe) and C3···Fe compressions, were calculated also to result in C4–C5 cleavage and ethylene production, but with higher activation barriers of ~ 12 and ~ 18 kcal/mol, respectively. Only the latter PES produced pyrocarbonate.¹⁴

The prediction of pyrocarbonate formation in two of four ethylene-yielding pathways implied that the retained C2-derived CO₂ might actively influence the reaction outcome, and so Wang and Sayfutyarova more explicitly interrogated this possibility, first by replacing it with water and then by removing it altogether.¹⁴ Under C3•···O(CO₂–Fe) compression, both complexes were calculated to undergo C4–C5 cleavage and ethylene formation, necessarily without pyrocarbonate formation. Activation barriers (3.1 and 1.5 kcal/mol, respectively) were even less than those calculated for the pyrocarbonate-forming transformation of the complex with C2-derived CO₂ still bound. However, the calculated energetic advantage of the ET-DC step over the C–O coupling step was diminished by replacement of the CO₂ with water and even more so by its removal (from 13.7 kcal/mol to 6.0 kcal/mol to 1.3 kcal/mol). The analysis suggests that the retained C2-derived CO₂ could raise both barriers but selectively favor the ethylene-yielding ET-DC over the C–O-coupling RLT step that commits the reaction to 3HP.¹⁴

The idea that the retained CO₂ generated in a prior fragmentation could change the downstream trajectory is intriguing also with respect to the first branchpoint. Computational studies have considered multiple mechanisms for the net C1–C2/O₂-insertion that forms the succinylperoxycarbonatoiron(II) complex committed to the major pathway,^10–14 but several of these have invoked an initial C1/O₂ replacement in common with the minor RO pathway.^10,12 In this subset of mechanisms, the resultant peroxysuccinatoiron(II) complex sits at the actual branch point, and the competition is between peroxide heterolysis to form the ferryl complex (RO pathway) and CO₂ insertion into the Fe–acylperoxide bond to form the succinylperoxycarbonatoiron(II) complex (major pathway). Flux down the RO branch could reflect preemption of the CO₂ insertion by CO₂ dissociation. In this scenario, effective retention of the C1-derived CO₂ would be imperative to efficient ethylene production. In the same manner, retention of the C2-derived CO₂ equivalent might, according to the analysis of Wang and Sayfutyarova,¹⁴ be crucial to favoring the ET-DC step over the C–O-coupling RLT step to ensure ethylene rather than 3HP production. This idea could unify both branchpoints as competitions between dissociation of CO₂, leading to an “undesired” outcome, and the retention of CO₂ needed for the chemical step that yields the desired intermediate or product. In this view, a means to retain CO₂ would be a key design feature of EFE and could theoretically be engineered to maximize ethylene yield.

Conclusion

Combined isotopic and kinetic resolution of the CO₂ equivalents generated from 2OG by A198L EFE has established that the C5-derived CO₂—which forms with ethylene—and the C1-derived CO₂—which forms with 3HP—cannot, as we proposed in prior work, arise from competing breakdown pathways of a common intermediate, namely the (2-carboxyethyl)carbonatoiron(II) complex. Rather, formation of this intermediate by the unusual propionate-3-yl/carbonatoiron(III) RLT step competes with ethylene production by ET-DC of the radical and commits the reaction to 3HP production. This experimental demonstration accords with computational studies showing both a lower barrier for the ET-DC step than for the RLT step and a prohibitively high barrier for the originally proposed ethylene-yielding Grob-like fragmentation of the (2-carboxyethyl)carbonatoiron(II) complex. The results thus re-open the question of the precise nature of the ethylene-yielding ET-DC reaction and basis for its preemption by substitutions to the 2OG substrate or amino acids near it.