Substrate-triggered Formation of a Peroxo-Fe₂(III/III) Intermediate during Fatty Acid Decarboxylation by UndA

Abstract

The iron-dependent oxidase UndA cleaves one C3–H bond and the C1–C2 bond of dodecanoic acid to produce 1-undecene and CO₂. A published x-ray crystal structure showed that UndA has a heme-oxygenase-like fold, thus associating it with a structural superfamily that includes known and postulated nonheme diiron proteins, but revealed only a single iron ion in the active site. Mechanisms proposed for initiation of decarboxylation by cleavage of the C3–H bond using a mono-iron cofactor to activate O₂ necessarily invoked unusual or potentially unfeasible steps. Here we present spectroscopic, crystallographic, and biochemical evidence that the cofactor of Pseudomonas fluorescens Pf-5 UndA is actually a diiron cluster and show that binding of the substrate triggers rapid addition of O₂ to the Fe₂(II/II) cofactor to produce a transient peroxo-Fe₂(III/III) intermediate. The observations of a diiron cofactor and substrate-triggered formation of a peroxo-Fe₂(III/III) intermediate suggest a small set of possible mechanisms for O₂, C3–H and C1–C2 activation by UndA; these routes obviate the problematic steps of the earlier hypotheses that invoked a single iron.

Graphical Abstract

The Pseudomonas enzyme UndA oxidatively decarboxylates medium-chain C_n (n = 8, 10, 12) free fatty acids to C_n-1 terminal olefins.¹ This reaction generates hydrocarbons that could be directly employed as liquid fuel or industrial feedstocks.^2–4 An initial report established that UndA requires Fe(II), dioxygen and a reducing system for multiple turnovers in vitro.¹ A mechanism to account for these requirements and for the observations of a single iron ion in the UndA x-ray crystal structure and retention of both C2 deuteria in the 1-undecene product from 2-[²H₂]-dodecanoic acid (lauric acid, LA) was proposed.¹ Addition of O₂ to the Fe(II) cofactor was envisaged to generate a Fe(III)-superoxide intermediate that would abstract a hydrogen (H•) from C3 (C_β) of the LA substrate to initiate decarboxylation. However, the C3–H bond has a homolytic dissociation energy (BDE > 95 kcal/mol)⁵ significantly exceeding that of any C–H bond currently known to be cleaved by a mid-valent metal-superoxo complex. In these known cases, a heteroatom with a non-bonded electron pair or an olefin is invariably α to the scissile C–H bond and is thought to make the initiating hydrogen atom transfer (HAT) thermodynamically feasible.^6–7 Moreover, the mid-valent-metal-superoxo reaction manifold often enables four-electron oxidation reactions,^6–7 whereas the UndA reaction is a two-electron oxidation. These inconsistencies suggested that a different mechanism could be operant.

UndA has a heme-oxygenase-like fold, similar to the Chlamydia protein associating with death domains (CADD), a protein of unknown function.^8–9 The CADD x-ray crystal structure revealed a diiron cluster reminiscent of the nonheme diiron cofactors in proteins of the topologically distinct ferritin-like structural superfamily.¹⁰ Many ferritin-like enzymes use this cofactor to activate O₂ for cleavage of strong C–H bonds,¹⁰ as must also occur in the UndA reaction. Ligands to the dinuclear cluster in CADD are conserved in UndA (Figure S1), and a recent study presented evidence, obtained by Mössbauer spectroscopy and measurement of reaction stoichiometry, that Pseudomonas syringae pv. tomato DC3000 UndA uses a diiron cofactor.¹¹ Here we show that Pseudomonas fluorescens Pf-5 UndA (78 % sequence identity to the P. syringae UndA) also harbors a diiron cluster and that binding of the LA substrate promotes very rapid O₂ activation through an intermediate that is identified by its absorption and Mössbauer spectra as a peroxo-Fe₂(III/III) complex.

We first obtained evidence by Mössbauer, extended x-ray absorption fine structure (EXAFS), and EPR spectroscopies for a coupled Fe₂(III/III) cluster in P. fluorescens UndA. Its Mössbauer features (Figure S2) are similar to those recently reported for the P. syringae UndA.¹¹ EXAFS measurements yielded an Fe-Fe separation of 3.21 Å (Figures S3 and S4). EPR spectra of a partially reduced sample exhibited an axial g < 2 signal characteristic of a mixed-valent Fe₂(II/III) cluster (Figure S5). We next sought a high-resolution x-ray crystal structure of the protein with both iron sites occupied. Anomalous difference maps from Fe(II)-soaked UndA crystals (Figure S6) revealed clear evidence for a diiron cofactor, but despite the improved Fe2 occupancy, these datasets failed to yield a complete view of the first coordination sphere in the UndA reactant complex. Specifically, persistent disorder of residues 195-200 prevented modeling of D198 (Figure S6), one of three proposed Fe2 ligands (along with E159 and H201). However, substitution with alanine of any of these three residues abolished 1-undecene production (Figure S1), underscoring the importance of site 2 coordination in catalysis.

Mixing of an anoxic solution containing UndA and Fe(II) with O₂-containing buffer resulted in slow (k_obs = 0.5 s⁻¹) development of stable absorption at ~350 nm, signifying oxidation of Fe(II) to Fe(III) (Figure 1A).^12–15 By contrast, with LA also present in the protein solution, a transient feature centered at λ_max ~ 550 nm (Figure 1B) developed to maximum intensity in < 0.01 s (black and red spectra) and decayed by ~1 s (orange spectrum). The substrate-triggered accumulation of the associated complex and its transient nature suggest that it could be an intermediate in the decarboxylation reaction.

Figure 1.

Substrate-triggered, [O₂]-dependent, rapid formation of an absorbing, on-pathway intermediate in the reaction of UndA-Fe₂(II/II) with O₂. (A and B) Absorption spectra acquired after rapid mixing at 5 °C of a solution containing 0.30 mM UndA, 0.54 mM Fe(II) (1.8 equiv), and either 0 mM (A) or 1.0 mM (B) LA with an equal volume of O₂-saturated buffer (~1.8 mM O₂). (C) Change in absorbance at 550 nm (ΔA_550nm) versus time after mixing of the same protein reactant solution as in B with an equal volume of buffer containing O₂ at the concentrations indicated in the legend. (D) Comparison of the ΔA_550nm amplitudes (differences between maximum and minimum values) and quantities of LA consumed [relative to a reaction with no added Fe(II)] in analogous reactions with varying equivalencies of Fe(II):UndA in the protein solution. Details of the reactions and the procedure for LA quantification are provided in the Supporting Information.

Early adducts between O₂ and Fe₂(II/II) cofactors have been characterized in a number of ferritin-like proteins, and many of these have visible absorption features with maxima between 420 and 700 nm.¹⁰ Most are μ-peroxo-Fe₂(III/III) complexes resulting from oxidative addition of O₂ to the cluster in a bridging mode.^16–22 The rate of formation of the 550 nm-absorbing complex in UndA depends on the concentration of O₂ (Figure 1C), confirming that no prior protein-O₂ adduct accumulates to a high level in the reaction. We obtained further evidence that the absorbing species is a diiron complex and an on-pathway intermediate by variation of the Fe(II):UndA stoichiometry (Figure 1D). The dependencies of the quantity of LA consumed (red points, left axis) and the amplitude of the 550-nm absorbance transient (blue points, right axis) are nearly identical, and both quantities approach their maximum values at an Fe(II):UndA ratio of ~ 2. The slope of the plot for LA consumed implies a stoichiometry of 0.6 LA consumed per diiron cluster, only 60 % of the theoretical value of unity. This diminished stoichiometry reflects the oxidation of Fe(II) via at least one unproductive pathway. The accumulation of the intermediate complex to a maximum of 60 % of the total added iron and its decay via (a) radical species in a fraction of events (vide infra) are consistent with this explanation.

To further characterize the 550nm-absorbing intermediate, we prepared freeze-quench samples for Mössbauer spectroscopy (Figure 2 and S7). The 4.2-K/53-mT spectrum of the 0.01-s sample exhibits diminished intensity of the high energy line (~ 2.4 mm/s) from the Fe(II) reactant complex and a new asymmetric feature at ~ 1 mm/s. The developing features could be accounted for as a pair of overlapping quadrupole doublets, each contributing 30% of the total ⁵⁷Fe absorption. The parameters of the doublets [δ₁ = 0.59 mm/s, |ΔE_Q|₁ = 1.07 mm/s (red line) and δ₂ = 0.56 mm/s, |ΔE_Q|₂ = 0.67 mm/s (blue line, see Figure S8 for analysis)] are characteristic of N/O-coordinated high-spin Fe(III), but the isomer shifts (δ) are toward the higher end of the typical range,²³ as has been seen for other μ-peroxo-Fe₂(III/III) complexes (δ ~ 0.48-0.68 mm/s).¹⁰ The 4.2-K/8-T spectrum of the intermediate can be adequately simulated using the values of δ and |ΔE_Q| determined from the 53-mT spectrum and assuming a diamagnetic ground state (Figure S9). The rapid two-electron oxidation of the Fe₂(II/II) reactant complex upon exposure to O₂ is consistent with assignment of the 550-nm absorbing intermediate as a peroxo-Fe₂(III/III) species.

Figure 2.

The UndA reaction monitored by freeze-quench Mössbauer spectroscopy. The vertical bars show 4.2-K/53-mT spectra of a sample (top) of an anoxic solution of 1.2 mM reduced, reconstituted ⁵⁷Fe-UndA and 10 mM LA (see Supporting Information for details) and of samples obtained by mixing the reactant complex with an equal volume of O₂-saturated buffer (~1.8 mM O₂ at 5°C) and freeze-quenching after 0.01 s (middle) or 5 s (bottom) at 5°C. The solid lines are simulations of the spectra of the peroxo-Fe₂(III/III) intermediate (red and blue) and Fe₂(III/III) product (purple) with parameters quoted in the text.

Samples quenched at longer reaction times have 4.2-K/53-mT Mössbauer spectra similar to that of the 0.01-s sample (Figure S7). However, the development of a spectral feature at ~ −0.2 mm/s (arrow in Figure 2) indicates decay of the peroxo-Fe₂(III/III) intermediate to a new species. Analysis of the (0.01s–5s) difference spectrum (Figure S10) yields δ ~0.52 mm/s and |ΔE_Q| ~0.97 mm/s (Figure 2, purple line) for the developing species, which we assign as the product Fe₂(III/III) cluster.

Decay of the peroxo-Fe₂(III/III) complex leads to development of absorption- and EPR-spectroscopic features characteristic of a tyrosyl radical (Tyr•) (Figure 1B and S11).^24–25 Accumulation of a one-electron-oxidized protein side chain in the two-electron-oxidation reaction converting LA to 1-undecene is unexpected, and, indeed, two lines of evidence suggest that the detected radical species is (are) not on the productive pathway. Substitution of any of the three Tyr residues closest to the active site (Tyr71, Tyr107, and Tyr197) with Phe failed to abolish the transient features (Figure S12), implying that multiple Tyr residues, rather than a single functionally essential Tyr, can be oxidized. Substitution of the cofactor proximal Trp190, posited to mediate oxidation of the Tyr(s) by a relay mechanism, with Phe did abolish the Tyr•(s), and yet this W190F variant retained the ability to consume LA (Figure S13). Intriguingly, a Fe₂(III/IV) complex accumulated (Figure S14), albeit only to 0.06 equiv, in the reaction of this variant. It is not clear whether this one-electron-reduced successor to the peroxo-Fe₂(III/III) complex is an on-pathway species resulting from substrate oxidation by a Fe₂(IV/IV) level complex (Scheme 1) or is also off-pathway.

Scheme 1.

Proposed mechanisms for oxidative decarboxylation of LA by UndA.

The conclusion that UndA uses a Fe₂(II/II) cluster to form a peroxo-Fe₂(III/III) intermediate reveals its analogy to ferritin-like nonheme diiron enzymes¹⁰ such as soluble methane monooxygenase (sMMO).¹⁶ In the sMMO reaction, the μ-peroxo-Fe₂(III/III) complex is thought to convert to an H•-abstracting Fe₂(IV/IV) complex (Q).^26–28 As recently suggested by Manley, et al.,¹¹ a similar pathway could be operant in the UndA reaction; the observation of substrate-triggered formation of the first of these two key intermediates puts such a hypothesis on firmer ground. Formation of an intermediate analogous to Q from the peroxide-level complex and HAT from C3 to Q would generate a substrate radical that could initiate 1-undecene production by one of three fates (Scheme 1, HAT): (A) oxygen rebound followed by a Grob-type multi-bond fragmentation, (B) radicaloid C1-C2 fragmentation followed by electron transfer from the •CO₂⁻ to the Fe₂(III/IV) cluster, or (C) electron transfer to the Fe₂(III/IV) cluster followed by polar decarboxylation. Alternatively, substrate decarboxylation via a Kolbe-like mechanism could be initiated by one-electron-oxidation of the fatty acid carboxylate (Scheme 1, PCET). Each of these pathways would involve a Fe₂(III/IV) intermediate, and so our detection of such a complex would potentially provide experimental support for such mechanistic hypotheses.

These hypotheses draw directly from mechanisms proposed for ferritin-like diiron enzymes, but UndA has the heme-oxygenase-like (HO-like) fold, which features a three-helix metal-binding core that is distinct from the ferritin-like four-helix core (Figure 3). Most known HO-like proteins (HO, Thi4/TenA, PqqC) do not bind metal ions directly,^29–31 but two other recently discovered examples, SznF and BesC, share with UndA a dependence on Fe(II) and O₂ for their oxidative activities.^32–33 Interestingly, prior to the work of Manley, et al.,¹¹ the functionally unassigned CADD was the only HO-like protein shown to harbor a dimetal cluster,⁸ suggesting that the metallocofactors may be generally unstable in this scaffold. Cluster lability likely results from ligand-contributing core helices with only modestly stable secondary structures, as observed for both UndA and SznF.^1,32 Although cofactor instability would seem to be detrimental to catalytic function, it could conceivably serve to obviate in situ cofactor redox recycling by a dedicated reductase protein, thereby enabling a novel modus operandi for this emerging family of nonheme diiron enzymes.

Figure 3.

Comparison of the 3-helix HO-like [CADD (1RCW), UndA (6P5Q)] and 4-helix ferritin-like [AurF (3CHH), sMMOH (1FYZ)] architectures.