Four-electron oxidation of p-hydroxylaminobenzoate to p-nitrobenzoate by a peroxodiferric complex in AurF from Streptomyces thioluteus

Abstract

The nonheme di-iron oxygenase, AurF, converts p-aminobenzoate (Ar-NH₂, where Ar = 4-carboxyphenyl) to p-nitrobenzoate (Ar-NO₂) in the biosynthesis of the antibiotic, aureothin, by Streptomyces thioluteus. It has been reported that this net six-electron oxidation proceeds in three consecutive, two-electron steps, through p-hydroxylaminobenzoate (Ar-NHOH) and p-nitrosobenzoate (Ar-NO) intermediates, with each step requiring one equivalent of O₂ and two exogenous reducing equivalents. We recently demonstrated that a peroxodiiron(III/III) complex (peroxo--AurF) formed by addition of O₂ to the diiron(II/II) enzyme (-AurF) effects the initial oxidation of Ar-NH₂, generating a μ-(oxo)diiron(III/III) form of the enzyme (μ-oxo--AurF) and (presumably) Ar-NHOH. Here we show that peroxo--AurF also oxidizes Ar-NHOH. Unexpectedly, this reaction proceeds through to the Ar-NO₂ final product, a four-electron oxidation, and produces -AurF, with which O₂ can combine to regenerate peroxo--AurF. Thus, conversion of Ar-NHOH to Ar-NO₂ requires only a single equivalent of O₂ and (starting from -AurF or peroxo--AurF) is fully catalytic in the absence of exogenous reducing equivalents, by contrast to the published stoichiometry. This novel type of four-electron N-oxidation is likely also to occur in the reaction sequences of nitro-installing di-iron amine oxygenases in the biosyntheses of other natural products.

The enzyme AurF from Streptomyces thioluteus converts para-aminobenzoate (Ar-NH₂, where Ar = 4-carboxyphenyl) to para-nitrobenzoate (Ar-NO₂) in the biosynthesis of the antibiotic, aureothin (1, 2). It is structurally similar to the β₂ subunits of class I ribonucleotide reductases and the oxygenase components of bacterial multicomponent monooxygenases (BMMs), which all use carboxylate-bridged di-iron clusters to activate O₂ (3–5). Following some initial controversy over whether the active form of AurF also contains a di-iron cluster (6) or, instead, a dimanganese (7, 8) or manganese/iron cluster (9), two recent studies established unequivocally that the di-iron form is active (although they did not rule out the possibility that the manganese/iron form could also be active) (10, 11). The first study showed that Fe₂-AurF could convert Ar-NH₂ to Ar-NO₂ in the presence of O₂ and a reducing system (although not the native reductase, which has not yet been identified) (10). It reasserted the previously proposed reaction sequence comprising three canonical diiron-oxygenase cycles (6), each involving, first, combination of O₂ with the diiron(II/II) form of the enzyme () to form an intermediate that oxidizes the substrate (Ar-NH₂, Ar-NHOH, and Ar-NO in cycles one, two, and three, respectively) by two electrons and, second, reduction of the resultant diiron(III/III) form of the enzyme () back to the O₂-reactive (Scheme 1A). It purported to detect the Ar-NO intermediate of the second cycle [Ar-NHOH having already been detected previously (6, 7)] and discussed the nature of its formation (10). Whereas the Hertweck group proposed hydroxylation of Ar-NHOH to Ar-N(OH)₂ followed by dehydration to Ar-NO (7, 8), Zhao and coworkers reasserted their previous proposal that Ar-NHOH undergoes direct dehydrogenation to Ar-NO (10). The basis for this proposal was isotopic labeling studies, which had shown that reaction of Ar-NH¹⁶OH in the presence of results in incorporation of only one atom of ¹⁸O into the Ar-NO₂ (6). This labeling pattern is consistent with the formation of the presumptive Ar-N¹⁶O intermediate by a dehydrogenation mechanism, which would leave it devoid of any ¹⁸O label, so that the third oxidation step with would then generate Ar-N(¹⁶O)(¹⁸O) with only one atom of ¹⁸O (Scheme S1, Zhao pathway). The labeling pattern is inconsistent with formation of Ar-NO by hydroxylation to Ar-N(¹⁶OH)(¹⁸OH) followed by dehydration, which should result in a 1∶1 mixture of Ar-N¹⁶O and Ar-N¹⁸O (assuming no isotope effect nor stereochemical bias in the dehydration step) and, in the absence of exchange with solvent of the O-atom from the nitroso group, subsequently result in formation of a 1∶1 mixture of Ar-N(¹⁶O)(¹⁸O) and Ar-N(¹⁸O)₂ in the final oxidation step (Hertweck pathway).

Scheme 1.

Reactions catalyzed by AurF. (A) Previously proposed stoichiometry of the AurF-catalyzed conversion of Ar-NH₂ to Ar-NO₂ (8–10); (B) stoichiometry of the AurF-catalyzed reaction indicated by this study; and (C) reactions of with Ar-NH₂ and Ar-NHOH.

The second study to confirm the activity of Fe₂-AurF reported trapping and characterization of a long-lived peroxodiiron(III/III) complex () that is competent to oxidize Ar-NH₂ (presumably to Ar-NHOH, but this intermediate was not explicitly verified) (11). The nearly complete conversion of Ar-NH₂ to the fully oxidized Ar-NO₂ at low ratios of (< 0.3) suggested that the peroxide complex might also be capable of the two subsequent oxidations (Ar-NHOH → Ar-NO → Ar-NO₂). In this study, we have confirmed that the two-electron-oxidized substrate species (Ar-NHOH) does indeed react with . The reaction does not, however, generate Ar-NO and a diiron(III/III) cluster, the products expected from the published reaction sequence (7, 9, 10) and our previous study (11). Rather, the reaction generates and the fully oxidized Ar-NO₂, implying that a four-electron redox process occurs. In the presence of excess O₂, the so produced regenerates , priming for oxidation of another molecule of Ar-NHOH. Thus, the reaction is catalytic without exogenous reducing equivalents. The results mandate reformulation of the overall conversion of Ar-NH₂ to Ar-NO₂ by Fe₂-AurF (Scheme 1B) and reevaluation of the mechanism of the last two oxidation steps. Specifically, they suggest that the Ar-NH₂ → Ar-NO₂ conversion proceeds by a sequence of two consecutive, mechanistically analogous hydroxylations followed by an inner-sphere, proton-coupled, two-electron transfer (Scheme 1B), rather than by the alternating, mechanistically distinct hydroxylation, dehydrogenation, and hydroxylation steps previously proposed (Scheme 1A). The new mechanism, which accounts simply for the aforementioned isotope labeling experiments (Scheme S1, central pathway), invokes a four-electron N-oxidation, which is initiated by a peroxodiiron(III/III) intermediate and is, to the best of our knowledge, unprecedented.

Results

Testing for a Reaction Between and Ar-NHOH by Stopped-Flow Absorption (SF-Abs) Experiments.

Our recent study showed that mixing with O₂ results in rapid formation of , which can be monitored by its absorption at 500 nm with molar absorptivity of ∼500 M^-1 cm^-1 (11). Subsequent mixing of the intermediate with Ar-NH₂ results in a rapid decrease in absorbance at this wavelength (A₅₀₀), reflecting oxidation of Ar-NH₂ by . A₅₀₀-versus-time traces from similar, sequential-mixing, SF-Abs experiments, in which was formed by an initial mix of with O₂ and then exposed to Ar-NHOH in the second mix, show that the two-electron-oxidized substrate species also reacts efficiently (Fig. 1). Revealingly, the kinetic behavior depends markedly on the relative concentrations of AurF, O₂, and Ar-NHOH. When the initial mix to form delivers ∼2 equiv O₂ and the second mix delivers 1 equiv Ar-NHOH, A₅₀₀ decreases rapidly, reaches a minimum after ∼50 ms, and then increases to approximately the original value (red traces in Fig. 1 A and B). By contrast to this transient behavior, A₅₀₀ decreases and remains stable when ∼2 equiv O₂ is delivered in the first mix and ≥2 equiv Ar-NHOH is delivered in the second mix (Fig. 1A, blue and green traces). Similarly, A₅₀₀ decreases and remains stable if ≤ 1 equiv O₂ is delivered in the first mix and ≥1 equiv Ar-NHOH is delivered in the second mix (Fig. 1B, blue trace). These observations can all be explained by assuming that the reaction between Ar-NHOH and generates . With overall O₂/Ar-NHOH/AurF ratios of 2/1/1, one equiv O₂ remains after the first has been used to form . This remaining O₂ adds to the produced by reaction of the first equiv of with Ar-NHOH, generating a second equiv of and yielding the transient kinetic behavior. With O₂/Ar-NHOH/AurF = 2/≥2/1, sufficient substrate remains to consume the second equiv of , O₂ is at that point exhausted, and the is then stable. With O₂/Ar-NHOH/AurF = 1/1/1, O₂ is exhausted in the initial formation of , and the generated upon reaction with Ar-NHOH is again stable.

Fig. 1.

Sequential-mixing SF-Abs experiments to monitor the reaction of with Ar-NHOH. (A) A solution of (0.9 mM Fe₂) was mixed with an equal volume of reaction buffer (see Experimental Procedures for composition) containing 1.8 mM O₂ (O₂/Fe₂ = 2). This solution was allowed to react at 5 °C for 0.5 s to permit accumulation of and was then mixed with an equal volume of an O₂-free solution of Ar-NHOH (in buffer) at concentrations appropriate to give the Ar-NHOH/Fe₂ ratios indicated in the inset. (B) A solution of (0.9 mM Fe₂) was mixed with an equal volume of reaction buffer containing either 0.9 mM (black and blue) or 1.8 mM O₂ (gray and red). This solution was allowed to react at 5 °C for 0.5 s to permit accumulation of and was then mixed with an equal volume of an O₂-free solution of buffer (black and gray) or 0.45 mM Ar-NHOH (blue and red).

Evaluation of Di-iron Products in Reaction of with Ar-NHOH by Mössbauer Spectroscopy.

Mössbauer spectroscopic experiments were conducted to test the conclusion that reaction of with Ar-NHOH produces , which is stable when O₂ is limiting and transient when O₂ is in excess. The 4.2-K/53-mT spectrum of a sample prepared by reacting with limiting (0.75 equiv) O₂ reveals that is the predominant species in the sample (vertical bars in Fig. 2A). The red line plotted over the data is a “reference spectrum” of the intermediate complex, which was generated by analysis of the experimental spectrum of a sample that was prepared so as to yield a maximum fraction of the intermediate (see Fig. S1 for the reaction conditions, explanation of the analysis, and the Mössbauer parameters). The reference spectrum accounts for 62% of the total intensity of the experimental spectrum in Fig. 2A. (We estimate an uncertainty of ± 3 on this and all other percentages of total absorption area given in the text.) As expected, use of limiting O₂ results in a significant fraction (36%) of unreacted starting material (δ = 1.23 mm/s, ΔE_Q = 3.08 mm/s; blue line). A minor fraction (6%) of is also present (δ = 0.54 mm/s, ΔE_Q = 1.86 mm/s, green line). Treatment of an identical sample with one equiv Ar-NHOH for 45 ms or 400 ms prior to rapid freezing (freeze-quenching) results in marked spectral changes (vertical bars in Fig. 2 B and C, respectively, and Fig. S2). The features of decay (to 33% and 8%, respectively; red lines), and features attributable to grow in. The contribution from remains constant. The changes are best illustrated by the difference spectrum generated by subtracting 2A from 2B (Fig. 2D, vertical bars). In this presentation, the features pointing upward are associated with the decaying species (; red line), and the features pointing downward with the developing species (; δ = 1.23 mm/s, ΔE_Q = 3.00 mm/s; blue line). We note that the measured quadrupole splitting parameter, ΔE_Q, of the developing Fe^II species (3.00 mm/s) is slightly different from that of the reactant complex (ΔE_Q = 3.08 mm/s), suggesting that cycling through the state might cause a conformational change at the cluster, as was observed for the diiron-carboxylate oxidase, stearoyl-acyl carrier protein Δ⁹ desaturase (12). The summation of the appropriately weighted reference spectra for the decaying (-28%) and the developing (28%) agrees well with the experimental difference spectrum (compare solid black line and vertical bars in Fig. 2D). The difference between the spectra of the 400-ms and 45-ms samples (C-B, Fig. 2E vertical bars) is, within the experimental uncertainty, identical to the B-A difference spectrum (Fig. 2E, solid line) and again demonstrates conversion of 28% of to 28% of . The similarity of these two difference spectra strongly suggests that the process being monitored has only one kinetically significant step (two states, reactants and products), because a sequence of two or more steps (three or more states) ought not to give a constant difference spectrum from 0 to 45 ms and 45 ms to 400 ms. The kinetics of the conversion of to reflected in the Mössbauer data are reasonably consistent with those determined by SF-Abs. One minor inconsistency is that 8% of remains at 400 ms, whereas the SF-Abs data indicate that decay should be essentially complete by this reaction time (Fig. 1B, blue trace). We attribute this remaining to the reaction of a small fraction of the product with O₂ from the air during passage through the freeze-quench reaction hose and into the cryosolvent.

Fig. 2.

4.2-K/53-mT Mössbauer spectra of samples in which was reacted with Ar-NHOH or Ar-NH₂. In all cases, the red, blue, and green lines illustrate the fractional contributions of the reference spectra of , , and , respectively, to the experimental spectrum, as described in the text. (Left) A solution of (1.2 mM Fe₂) was mixed with 0.5 equivalent volume of buffer solution containing 1.8 mM O₂ (O₂/Fe₂ = 0.75). This solution was allowed to react at 5 °C for 0.11 s to permit accumulation of . (A) The solution was then directly freeze-quenched. (B and C) The solution was then mixed with one-sixth equivalent volume of an O₂-free solution of 3.6 mM Ar-NHOH (Ar-NHOH/O₂ = 1), and this solution was allowed to react for 45 ms (B) or 400 ms (C) prior to being freeze-quenched. D and E are the difference spectra B-A and C-B, respectively. The black line in D is the sum of the contributions of (-28%) and (28%). The black line in E is the difference spectrum B-A for comparison. (Middle) A solution of (1.8 mM) was mixed with two equivalent volumes of a buffer solution containing 1.8 O₂ mM O₂ (O₂/Fe₂ = 2). This solution was allowed to react at 5 °C for 110 ms to permit accumulation of . (F) The reaction was then directly freeze-quenched. (G and H) The solution was then mixed with one-sixth equivalent volume of an O₂-free buffer solution containing 3.6 mM Ar-NHOH (Ar-NHOH/O₂ = 0.5), and this reaction was allowed to proceed at 5 °C for 45 ms (G) or 2 s (H) prior to being freeze-quenched. I and J are the difference spectra G-F and H-G, respectively. The black line in J is difference spectrum C–B, scaled by a factor of -0.8 for comparison. (Right) A solution of (1.2 mM) was mixed with 0.5 equivalent volume of a buffer solution containing 1.8 mM O₂ (O₂/Fe₂ = 0.75), and the reaction was allowed to proceed at 5 °C for 110 ms to permit accumulation of . (K) The reaction was directly freeze-quenched. (L) The resulting solution was then mixed with one-sixth equivalent volume of an O₂-free buffer solution containing 0.91 mM Ar-NH₂ (Ar-NH₂/O₂ = 0.25), and the reaction was allowed to proceed for 4 s before being freeze-quenched. M is the difference spectrum L–K. The black line in M is the sum of the contributions of (-29%), (12%), and (17%).

The 4.2-K/53-mT Mössbauer spectrum of a sample prepared by reacting with ∼2 equiv O₂ per diiron cluster (Fig. 2F, vertical bars) indicates the presence of 82% (red line) and 9% each of and . The spectra of identical samples that were subsequently mixed with one equiv Ar-NHOH and then allowed to react for 45 ms (Fig. 2G) or 2 s (Fig. 2H) before being freeze-quenched (see Fig. S3 for additional spectra corresponding to different reaction times) confirm the conclusion from the SF-Abs experiments that initially decays (to 52% in Fig. 2G, red line) and is subsequently regenerated (74% in Fig. 2H, red line). The G-F difference spectrum (Fig. 2I, vertical bars) reveals the conversion of 28% of (red line) to 20% (blue line) and 6% (green line). The could result from the reaction of some of the with small amounts of either Ar-NH₂ or Ar-NO present in the Ar-NHOH substrate solution (see Fig. S4). The H-G difference spectrum (Fig. 2J, vertical bars) reveals clean conversion of 20% (blue line) to 20% (red line; note that the reformation of is reflected by inversion of the peaks). This difference spectrum is very similar to the difference spectrum 2E from the limiting-O₂ reaction (solid line in Fig. 2J, scaled by -0.8). Recovery of the is nearly complete (73% of total Fe, 89% of the initially present). This observation, together with the Mössbauer and SF-Abs data on the reaction with limiting O₂, confirm that Ar-NHOH effects a four-electron reduction of to , presumably with concomitant four-electron oxidation of Ar-NHOH to Ar-NO₂ (confirmed below).

Verification of Catalytic Oxidation of Ar-NHOH by .

Our reformulation of the AurF six-electron-oxidation sequence eliminates the previously proposed requirement for a total of four exogenous electrons in the last two steps (compare Scheme 1 A and B). should, therefore, be capable of multiple turnovers when oxidizing Ar-NHOH (Scheme 1C, right side). To verify this prediction, 15 μM was incubated with 450 μM Ar-NHOH (Ar-NHOH/Fe₂ = 30) and ∼0.9 mM O₂ (O₂/Ar-NHOH ∼ 2). After 20 min at 0 °C, the small-molecule components were separated from the protein (by passage through a molecular weight filter, requiring an additional 10 min at 4 °C) and were analyzed by reverse-phase high-performance liquid chromatography (RP-HPLC; see Experimental Procedures and SI Text for details). The chromatogram of a standard solution containing 400 μM each of Ar-NH₂, Ar-NHOH, and Ar-NO₂ (Fig. 3, red trace) shows that the compounds are efficiently separated by this procedure. Injections of the individual compounds allowed the elution peaks to be assigned as indicated by the chemical structures. (Note that, because of differences in their molar absorptivities, the peak height and area for the Ar-NO₂ product is less than that for an equivalent quantity of Ar-NHOH substrate.) The chromatogram from a control sample in which 450 μM Ar-NHOH was exposed to O₂ under the conditions of the enzyme reaction but in the absence of (green trace) shows the prominent peak of the synthetic compound and two small peaks from unknown contaminants or decay products. This control confirms that Ar-NHOH is relatively stable on this time scale in the absence of the enzyme. The chromatogram of the experimental sample containing the enzyme (blue trace) shows that > 95% (> 430 μM) of the Ar-NHOH has been consumed and ∼450 μM Ar-NO₂ has been produced (estimated by comparison of the peak area to that in the chromatogram of the standard mixture). Thus, the enzyme has accomplished 28–30 turnovers in the absence of any obvious source of reducing equivalents. Catalytic oxidation of Ar-NHOH under these conditions is inconsistent with the previous formulation of the Ar-NHOH → Ar-NO → Ar-NO₂ oxidation steps (Scheme 1A) but entirely consistent with our reformulation (Scheme 1 B and C).

Fig. 3.

Reversed-phase high-performance liquid chromatography (RP-HPLC) of the small-molecule reactants and products following incubation of with excess Ar-NHOH and O₂. (15 μM) was incubated with 450 μM Ar-NHOH and ∼0.9 mM O₂ for 20 min at 0 °C. Small molecules were separated from the enzyme and analyzed as described in SI Text (blue). A control experiment was carried out under identical conditions, except for omission of (green). A solution containing 400 μM each of Ar-NH₂, Ar-NHOH, and Ar-NO₂ was also analyzed (red).

Testing for Reduction of by Ar-NHOH.

The surprising ability of Ar-NHOH to reduce by four electrons to and the known ability of hydroxylamine compounds to reduce Fe^III complexes suggested the possibility that Ar-NHOH might also reduce to , thereby possibly permitting the six-electron-oxidation sequence to proceed without net input of any exogenous electrons and with only a single equivalent of O₂. However, prolonged incubation (10 min at 22 °C) of as-isolated with one equiv (1.1 mM) Ar-NHOH yielded less than 10% reduction to (Fig. S5), implying that two exogenous electrons and two equiv O₂ are indeed required for the complete six-electron oxidation sequence.

Reevalution of the Diiron Products from the Reaction of with Limiting Ar-NH₂.

Scheme 1C predicts that reaction of with limiting Ar-NH₂ in the absence of excess O₂ should generate equal quantities of two distinct diiron products (highlighted in boxes). The first oxidation should produce Ar-NHOH and . The Ar-NHOH so produced should then dissociate from and react with a second equivalent of to generate Ar-NO₂ and . In our previous study (11), O₂ was present in excess, which (we now understand) must have converted the (the reduced product of the Ar-NHOH → Ar-NO₂ oxidation) back to (red arrows), thereby preventing detection of the reduced enzyme form. We therefore tested the prediction of Scheme 1C by Mössbauer spectroscopy on samples prepared with limiting O₂. A sample was enriched in by reaction of with 0.75 equiv O₂. The Mössbauer spectrum of this sample (Fig. 2K, vertical bars) confirms that is the major species (72%, red line). The spectrum of an identical sample that was subsequently treated with 0.19 equiv Ar-NH₂ and allowed to react to completion (4 s) prior to being freeze-quenched shows marked differences (Fig. 2L). Analysis of the L-K difference spectrum (Fig. 2M, vertical bars) indicates that 29% of (red line) is converted to 17% (green line) and 12% (blue line). The sum of these spectral contributions (Fig. 2M, black solid line) reproduces the experimental difference spectrum well. The total loss of intensity attributable to is only 76% of the theoretical value: 0.19 equiv Ar-NH₂ should consume 0.38 equiv complex, resulting in a loss of 38% (compared to the observed 29%) of total intensity. The yield of is within experimental error of the theoretical value (17% compared to 19%), but the yield of is only ∼60% of the theoretical value (12% compared to 19%). As argued above, the observed yield of and observed consumption of are most likely diminished from their theoretical values by exposure to atmospheric O₂ in the freeze-quench procedure, which results in conversion of a fraction of the product (∼7% of the total absorption intensity, essentially the same as in Fig. 2C) back to . The important point is that both products predicted by Scheme 1C are readily detected.

Discussion

AurF catalyzes the six-electron oxidation of Ar-NH₂ to Ar-NO₂ (2). All previous studies assumed that this conversion entails three sequential two-electron oxidations (Scheme 1A) (8–10). Of the two proposed intermediates in this sequence, Ar-NHOH and Ar-NO, the former compound was unequivocally identified (6) and reasonably presumed to form in the rapid reaction of with Ar-NH₂ (11). In this study, we investigated the reaction of this first intermediate, Ar-NHOH, with . Surprisingly, this reaction couples the oxidation of Ar-NHOH by four electrons (to Ar-NO₂) to the complete reduction of to (Scheme 1B). The proposed intermediates in this reaction, Ar-NO or Ar-N(OH)₂ and a cluster, apparently do not accumulate substantially during the reaction (even in the active site during a single turnover), as implied by the near identity of the Mössbauer difference spectra at different reaction times (Fig. 2, spectra D, E, I, and J) and the ability to account for these spectral changes by summation of the spectra of only the reactant and product states ( and ).

This finding can explain several published observations and has implications for the mechanism of the Ar-NHOH → Ar-NO₂ conversion by AurF. The reaction entails, formally, transfer of an O-atom from the peroxide moiety of to Ar-NHOH and transfer of two H-atoms from Ar-NHOH to the diiron cluster. The Hertweck group proposed that this conversion might proceed by a sequence of hydroxylation of Ar-NHOH to Ar-N(OH)₂, elimination of water to form Ar-NO, and transfer of a second O-atom to yield Ar-NO₂ (Scheme S1, right branch) (7, 8). They attempted to identify the proposed Ar-NO intermediate but failed to do so, despite employing a sensitive assay. They concluded that “rapid and possibly spontaneous turnover” of this intermediate to Ar-NO₂ could rationalize their failure to detect it (7). Their result is consistent with our observation that the presumptive intermediates in the four-electron conversion of Ar-NHOH to Ar-NO₂ do not accumulate.

The Zhao group proposed a different mechanism, founded on the presumption that Ar-NO is an obligatory intermediate on the pathway to Ar-NO₂ (Scheme 1A) and the observation that oxidation of Ar-NHOH in the presence of gas results in production of Ar-NO₂ with at most one atom of ¹⁸O (6, 10). As explained by Scheme S1, the authors interpreted this result to imply that the oxidation of Ar-NHOH to Ar-NO must be a direct dehydrogenation, rather than the hydroxylation-dehydration sequence proposed by Hertweck. The presumed intermediacy of Ar-NO also understandably led the Zhao group to search for this intermediate, which they purported to identify by HPLC with initial UV absorption detection followed by mass spectrometric (MS) detection of a fragment ion of appropriate mass-to-charge ratio (m/z) to be the decarboxylation product of the Ar-NO molecular ion (10). We suggest that the Ar-NO detected by Zhao and coworkers could have arisen from nonenzymatic oxidation or disproportionation of Ar-NHOH (a true accumulating intermediate in the AurF sequence) rather than as a product of the AurF reaction. Supporting this view, we have repeatedly detected by HPLC-MS measurements on solutions of Ar-NHOH dissolved in O₂-containing buffer (with or without AurF) a species of the correct m/z to be the Ar-NO parent ion (Fig. S4). Our inference is also corroborated by the Hertweck group’s detection of the dimerized substrate species, azoxybenzol-4,4′-dicarboxylic acid, which, they proposed, could have formed nonenzymatically via an Ar-NO intermediate (7).

Whereas it is possible that the substrate, Ar-NH₂ (Ar-NHOH), could either rapidly trap a more reactive (e.g., ) complex with which rapidly interconverts or, upon binding, could trigger the rapid conversion of the peroxide intermediate to the N-oxygenating complex, we presume for the purposes of this discussion that the peroxide complex is itself the N-oxygenating species. The hydroxylation of Ar-NH₂ is likely to involve nucleophilic attack of the amine on the peroxide moiety of (11). Hydroxylation of Ar-NHOH could also proceed by nucleophilic attack of the N-atom on the peroxide O-atom. Indeed, Ar-NHOH is expected to be even more nucleophilic than Ar-NH₂ as a result of the so-called alpha effect (13). Examination of the published structure of the complex (10) suggests a relatively simple trajectory for this hydroxylation mechanism (Scheme 2). The peroxide ligand is depicted in a μ-1,1 or distorted μ-η²∶η² coordination mode, because the spectroscopic properties of are different from those of well-characterized complexes (14, 15). However, the peroxide moiety could also be protonated (i.e., in a μ-1, complex), as we previously proposed (11). Nucleophilic attack of the bound Ar-NHOH on the uncoordinated or more distally coordinated O-atom of the (hydro)peroxide moiety would cleave the peroxide O-O bond and transfer an O-atom to the substrate. The product would be the Ar-N(OH)₂ intermediate proposed by the Hertweck group, or perhaps its deprotonated form, bound to the resulting form. Transfer of two electrons and two protons from Ar-N(OH)₂ to the cluster (formally, a dehydrogenation), rather than the sequence of dehydration followed by O-atom transfer proposed by the Hertweck group, could then give Ar-NO₂ directly, with the ¹⁸O-isotope labeling pattern observed by Zhao and coworkers (Scheme 2 and Scheme S1, middle). The new proposal appears to accommodate all available experimental data. The latter steps might conceivably occur by deprotonation of the Ar-N(OH)₂ intermediate, perhaps by the μ-oxo bridge (as depicted in Scheme 2) or one of the protein carboxylate residues (16), followed by inner-sphere electron transfer steps.

Scheme 2.

Proposed mechanism of the four-electron oxidation of Ar-NHOH to Ar-NO₂ by . The scheme is derived from the X-ray crystal structure of the complex, Protein Data Bank identification code 3CHT (10).

By contrast, the first step of the Zhao pathway (6, 10), dehydrogenation of Ar-NHOH, is distinct from the aforementioned nucleophilic attack of the substrate on the electrophilic complex. Distinct reactivities of intermediates with similar structures are well documented for the mononuclear non-heme-iron enzymes, in which the Fe^IV-oxo (ferryl) unit can act either as electrophile (17), transferring its O-atom to an electron-rich substrate, or as hydrogen-atom-abstractor, initiating hydroxylation, halogenation, desaturation, and cyclization outcomes (18). One-electron oxidation of Ar-NHOH by with deprotonation and coordination of the resulting aminoxyl radical (Ar-NHO^•) to the Fe₂ cluster, which is formally an H-atom abstraction akin to those effected by the ferryl intermediates, seems a conceivable reaction pathway on the basis of precedent from inorganic chemistry (19, 20). We anticipate that this step would cleave the O-O bond and generate a high-valent (hydr)oxo-bridged -cluster, given that the related intermediate in the I100W variant of toluene/o-xylene monooxygenase converts to a complex upon transfer of an electron from the introduced tryptophan residue (21). Several decay pathways for the hypothetical complex can be envisaged. Transfer of another proton [presumably to one of the bridging oxygenic ligands or a protein carboxylate residue (16)] and another electron via an inner-sphere mechanism would yield formally Ar-NO and a cluster. In the final step, this cluster would have to serve as an O-atom donor to generate and Ar-NO₂. Alternatively, the order of these steps (electron transfer, proton transfer, and O-atom transfer) within the complex could be different.

Of the two general pathways described above, nucleophilic attack of Ar-NHOH on the peroxide moiety versus formal H-atom transfer from Ar-NHOH to the peroxide, we prefer the former pathway (Scheme 2), because it most simply accounts for the outcome of the reaction, implies more similar reactivities of the Ar-NH₂ substrate and Ar-NHOH intermediate toward , and accounts for all available data. Additional experiments will be required to distinguish among these and other mechanistic possibilities.

Experimental Procedures

Materials.

AurF was prepared as previously described (11). Ar-NHOH was synthesized according to a published procedure (22). Its purification and characterization are detailed in SI Text and Figs. S6 and S7.

AurF Turnover Assay.

Catalytic conversion of Ar-NHOH to Ar-NO₂ was demonstrated by using RP-HPLC to resolve the two compounds and UV absorption to detect them. Comparison of the chromatograms of assay and control samples to that of a standard mixture of Ar-NH₂, Ar-NHOH, and Ar-NO₂ was used to quantify the Ar-NHOH substrate remaining and Ar-NO₂ product generated (Fig. 3). The standard mixture was prepared by dissolving the solids in 50 mM HEPES buffer (5% glycerol, pH 7.5) to a final concentration of 400 μM of each. The reaction was carried out in the same buffer, which was saturated with O₂ at 0 °C by vigorous stirring on ice under 1.1 atm of the gas. Immediately before initiation of the reaction by addition of enzyme, Ar-NHOH was dissolved in the O₂-saturated buffer to a final concentration of 450 μM. (1.0 mM) was prepared by treating as-isolated AurF with one equiv sodium dithionite for 30 min at room temperature in the absence of O₂. After a brief (∼30 s) exposure to air the was added to the reaction system to a final concentration of 15 μM. The solution was stirred on ice for 20 min and then filtered through an Amicon Ultra-0.5 centrifugal filter (10,000 molecular weight cut-off; Millipore) at 4 °C (13,000 rpm, 10 min). The filtered solution was analyzed by RP-HPLC as described in SI Text. An otherwise identical sample lacking only served as the control for this reaction.