Deciphering the Reaction Pathway of Mononuclear Iron Enzyme-Catalyzed N≡C Triple Bond Formation in Isocyanide Lipopeptide and Polyketide Biosynthesis

Abstract

Despite the diversity of reactions catalyzed by 2-oxoglutarate-dependent nonheme iron (Fe/2OG) enzymes identified in recent years, only a limited number of these enzymes have been investigated in mechanistic detail. In particular, several Fe/2OG-dependent enzymes capable of catalyzing isocyanide formation have been reported. While the glycine moiety has been identified as a biosynthon for the isocyanide group, how the actual conversion is effected remains obscure. To elucidate the catalytic mechanism, we characterized two previously unidentified (AecA and AmcA) along with two known (ScoE and SfaA) Fe/2OG-dependent enzymes that catalyze N≡C triple bond installation using synthesized substrate analogues and potential intermediates. Our results indicate that isocyanide formation likely entails a two-step sequence involving an imine intermediate that undergoes decarboxylation-assisted desaturation to yield the isocyanide product. Results obtained from the in vitro experiments are further supported by mutagenesis, the product-bound enzyme structure, and in silico analysis.

Graphical Abstract

INTRODUCTION

Natural product biosynthetic pathways are frequently characterized by reactions that result in the functionalization of unactivated C–H bonds.^1,2 Because the Lewis acid/base chemistry common to most enzymes is insufficient for these types of transformations, metalloenzymes typically supporting radical intermediates are often employed. The utilization of heme- and nonheme-iron metalloenzymes to effect functional group installation via C–H bond activation represents a widely adapted strategy among natural product biosynthetic pathways.^1,2 Representative examples catalyzed by 2-oxoglutarate-dependent nonheme iron (Fe/2OG) enzymes include aziridine installation, endoperoxidation, epoxidation, halogenation, hydroxylation, and olefination.^1,3–5 In spite of the discoveries of these enzymatic transformations, only a few reactions, such as halogenation and epoxidation, whose mechanisms are closely related to hydroxylation have been incorporated into synthetic applications.^6–8

To underpin the chemical principles and to unleash the potential of Fe/2OG enzymes, it is pivotal to elucidate reaction mechanisms of those chemically challenging, but useful, reactions. In this regard, Fe/2OG enzyme-catalyzed N≡C triple bond formation represents an important yet understudied transformation. The isocyanide, i.e., isonitrile, group is a zwitterion that contains a unique N≡C triple bond connection. The reactive isocyanide moiety has been extensively studied and used as an important synthetic equivalent in organic synthesis.^9–12 While these functional groups are also widespread among natural products,^13–21 there is relatively little information available regarding their biosynthesis. On the basis of the biosynthetic gene cluster analysis, however, several Fe/2OG enzymes have been proposed to catalyze isocyanide group formation in isocyanide lipopeptides (Figure 1A and Figure S19).^22,23 Analogous to other Fe/2OG enzymes, an iron(IV)-oxo species, the conserved key intermediate of these enzymes, has been implicated in triggering hydrogen atom transfer (HAT) to initiate isocyanide formation.²⁴ However, the intermediate product and the subsequent steps that lead to N≡C triple bond installation have yet to be determined.

Figure 1.

(A) Representative examples of isocyanide-containing lipopeptide and polyketide natural products. A conserved strategy involving Fe/2OG enzymes has been proposed to install the N≡C group from glycine in these compounds. (B) The conversion of the glycine moiety into isocyanide is catalyzed by an Fe/2OG enzyme in each pathway, including AecA (for aerocyanidin), AmcA (for amycomicin), ScoE (for INLP), and SfaA (for SF2768). The overall reaction is a two-step process in which the first reaction includes an iron(IV)-oxo species-triggered hydrogen atom transfer (HAT) to produce the carboxyaldimine intermediate (3). In the second reaction, two possible routes can be envisioned. The reaction may proceed through a HAT at the vinyl C—H position of 3 followed by decarboxylation. Alternatively, a pathway involving an oxygen atom transfer (OAT), ring-opening, dehydration, and decarboxylation is also conceivable. Notably, an in silico analysis suggests that the pathway involving HAT is favorable. Under in vitro conditions, 6 is produced as a side-product. (C) Carboxyaldimine 3 readily decomposes to 8 and glyoxylate (as the hydrate form) which can be detected by ¹H and ¹³C NMR, respectively.

In this work, two previously unidentified (AecA and AmcA) along with two known (ScoE and SfaA) Fe/2OG isocyanide synthases catalyzing the oxidative decarboxylation of carboxyaldimines (e.g., 3) to form isocyanides (e.g., 4) are investigated and led to formulation of a conserved reaction pathway responsible for isocyanide group installation. As illustrated in Figure 1B, following the formation of the iron(IV)-oxo species with concomitant reduction of O₂ by 2OG, catalysis likely proceeds through aldimine oxidation via HAT from the sp²-hybridized imine carbon of 3. The resulting imidoyl radical may then be further oxidized via single electron transfer followed by decarboxylation to yield the isocyanide 4. Alternatively, inspired by a recent discovery that an olefin group can effectively redirect the reactivity of leucine 5-hydroxylase from hydroxylation to epoxidation,⁸ the oxidation of 3 may alternatively proceed via oxygen atom transfer (OAT) followed by ring-opening, dehydration, and decarboxylation to afford 4 (3 → 7 → 4, Figure 1B). Compared with synthetic approaches and other enzymatic approaches used to install isocyanide groups,^25–29 the Fe/2OG enzyme-catalyzed formation of isocyanide functionalities is unique and may be useful in engineering new synthetic biological approaches for the introduction of isocyanide synthons.

RESULTS AND DISCUSSION

In Vitro Reconstitution of Two Annotated Fe/2OG Enzymes, AecA and AmcA, in Aerocyanidin and Amycomicin Biosynthesis.

Several Fe/2OG enzymes including ScoE, SfaA, and Sav607 encoded by isocyanide-containing lipopeptide biosynthetic gene clusters have been identified to catalyze isocyanide group installation using 1 as the substrate.^{22–24,30,31} In an attempt to probe the possible occurrence of a conserved mechanistic paradigm for N≡C triple bond formation, we began our study to investigate the reaction of two annotated Fe/2OG enzymes, AecA and AmcA, found in the polyketide antibiotics aerocyanidin and amycomicin biosynthetic pathways, respectively.^32,33

The likely involvement of these Fe/2OG enzymes in the construction of the isocyanide group is suggested by their gene sequence homology to those of ScoE, SfaA, and Sav607.³² To verify their catalytic functions, the aecA and amcA genes were heterologously expressed in Escherichia coli, and AecA and AmcA were isolated as N-His₆-tagged proteins. Subsequent experiments showed that, under in vitro conditions, both of them can catalyze isocyanide formation (4) using 1 as the substrate at the expense of O₂ and 2OG (Figure S1). Specifically, reaction solutions containing reconstituted enzyme, 2OG, and 1 with final concentrations of 0.1 mM reconstituted enzyme, 2.5 mM 2OG, and 1 mM 1 (200 μL in 100 mM Tris-HCl (tris(hydroxymethyl)aminomethane hydrochloride), pH 7.5) were exposed to air at 4 °C. The reactions were analyzed using liquid chromatography coupled mass spectrometry (LC-MS). These results are consistent with other Fe/2OG isocyanide forming enzymes found in isocyanide-containing lipopeptides^29,34 and set the stage to test our hypothesis that AecA and AmcA may utilize an approach similar to the previously identified Fe/2OG enzymes to effect N≡C triple bond formation. Namely, a conserved process may be employed to forge the isocyanide group by these homologous enzymes in isocyanide-containing lipopeptide and polyketide natural products.

Identification of an Imine as a Common Intermediate en Route to Isocyanide Formation.

It has been well-documented that Fe/2OG enzyme-catalyzed hydroxylation is initiated by the formation of an Fe(IV)-oxo species which abstracts a hydrogen atom (HAT) from the substrate.³⁵ The resulting radical then reacts with Fe(III)–OH to complete the hydroxylation.^5,36 The latter step is also known as oxygen-rebound.³⁷ Although an Fe(IV)-oxo has been identified as the key species to initiate HAT in isocyanide formation,²⁴ it remains to be determined how the oxygen-rebound is assimilated to effect isocyanide formation.^24,34 Alternatively, based on a computational analysis,^38,39 it has been proposed that carboxyaldimine (3) produced during the first reaction may serve as the intermediate substrate to trigger isocyanide formation (3 → 4). To establish the reaction mechanism and to investigate whether a conserved pathway is evolved in the biosynthesis of isocyanide-containing lipopeptides and polyketides, we chemically synthesized 3 and tested it with SfaA, the previously characterized isocyanide forming enzyme involved in lipopeptide SF2768 formation,^23,40 and AecA/AmcA identified in this study.

During the first few attempts under the conditions of pH 7 (100 mM Tris-HCl), the formation of the product (4) could only be occasionally detected by LC-MS. To seek for possible causes of the inconsistent results, the stability of 3 was monitored by ¹H NMR. A solution containing freshly prepared 3 dissolved in phosphate buffer (50 mM 3 and 100 mM KPi, pH = 6.0 or 8.0) was subjected to ¹H NMR analysis. It was found that, under slightly acidic conditions (pH 6.0), ~70% of 3 was degraded within 6 min to an amine (8) which was assigned by a comparison with the synthetic standard (Figure S2). Upon changing the buffer pH from 6 to 8, the stability of 3 was increased by ~2–3 fold. These observations suggested that the early inconsistent results may be ascribed to the instability of the substrate under the in vitro assay conditions in which 3 undergoes hemiaminal formation followed by decomposition to produce 8 and glyoxylate (3 → 2 → 8, Figure 1C). To minimize such a complication, we opted to freshly prepare 3 prior to each in vitro assay and carried out the enzymatic reactions at pH 8. Indeed, significant isocyanide (4) formation from 3 was consistently detected in the AecA, AmcA, and SfaA reactions (Figure 2A). While we cannot rule out the possibility that 3 can be hydrolyzed to 2 which is then used in 4 formation (i.e., 3 → 2 → 4), the detection of 8, but not 2 in the studies of 3 degradation (Figure S2), indicates that 2 is unlikely an onward intermediate as it readily decomposes to 8 and glyoxylate. Instead, these findings provide experimental evidence supporting the intermediacy of imine (3), which may be derived from 2, and its role as the common intermediate en route to N≡C triple formation (i.e., 2 → 3 → 4, see Figure 1B) in Fe/2OG isocyanide synthases.

Figure 2.

(A) LC-MS analysis of the AmcA, AecA, and SfaA catalyzed reactions. Using 3 as the substrate, AmcA, AecA, and SfaA can all effect isocyanide (4) formation. In the absence of 2OG, no obvious production of 4 can be detected. (B) Optical absorption spectra showing the binding of 1 and 3 to the SfaA/Amc·Fe(II)·2OG complex. Difference absorption spectra reflecting the binding of 2OG (20 mM) to the Sfa·Fe(II) complex (1 mM SfaA and 1 mM Fe²⁺) are shown as a gray solid line. In the presence of 1 (20 mM) and 3 (20 mM), the corresponding spectra are shown as red and blue solid lines, respectively. In addition, difference spectra reflecting the binding of 2OG (20 mM) to the AmcA·Fe(II) complex (1 mM AmcA and 1 mM Fe²⁺) in the presence of 1 (20 mM) and 3 (20 mM) are displayed as red and blue dashed lines, respectively.

In addition, the intermediacy of 3 is also corroborated by the observation of the binding of 3 in the active site of these enzymes. For the majority of Fe/2OG enzymes, the bidentate binding of 2OG to the Fe(II) center leads to generation of a weak optical absorption band (ε ~ 100–200 M⁻¹ cm⁻¹) centered around 500–550 nm.^35,41,42 This optical feature results from the metal-to-ligand charge transfer (MLCT) of Fe(II) and 2OG in the enzyme active site. Sample solutions containing reconstituted enzyme, 2OG, substrate (1 or 3) with final concentrations of 1 mM of reconstituted enzyme, 20 mM 2OG, and 2 mM of substrate (100 mM Tris-HCl, pH 7.5) were prepared in a glovebox. The solutions were centrifuged at 10,000 rpm for 5 min before optical absorption measurements. The difference spectra shown in Figure 2 were generated by subtracting the spectra of the enzyme·Fe(II)·2OG·substrate complex from the spectra of enzyme·Fe(II)·2OG. An addition of primary substrate to the enzyme·Fe(II)·2OG ternary complex slightly increases the intensity and changes the overall shape of the MLCT band. This optical feature has been used to assess the binding of the substrate in several Fe/2OG enzymes.^42–44 In here, upon an addition of 2OG to the SfaA·Fe(II) complex, a weak and broad MLCT band centered at ~530 nm was developed (Figure 2B). An addition of the native substrate (1) significantly increases the intensity of this band, which is now centered at ~540 nm along with two new features at 490 and 600 nm. In comparison, an addition of 3 also increases the intensity of the MLCT band of the SfaA·Fe(II)·2OG complex with the concomitant development of two similar absorption shoulders. Analogously, adding 1 and 3 into the AmcA·Fe(II)·2OG ternary complex resulted in similar changes (Figure 2B). Although the optical features induced by 3 are not as prominent as those of 1, this observation clearly demonstrates that both 1 and 3 can bind to the iron center in the active site of SfaA and AmcA and induce a change of the MLCT band.

Mutagenesis Studies Suggest That a Conserved Strategy Is Employed by Fe/2OG Enzymes to Effect Isocyanide Formation.

Previously, the structures of ScoE bound with substrate (1) and the hydroxylation product (2) have been reported by us and others.^24,34,45 Based on those crystallographic data, several residues including Y101, K193, G128, Y96, Y135, and R310 were identified to interact with 1 and 2. Specifically, Y96 shows a hydrogen-bonding interaction with the nitrogen atom of 1 and 2 with a distance of 2.6 and 2.5 Å, respectively. Moreover, Y101, together with G128 and K193, interacts with the C6-carboxylate of 1 and 2 (Figure S3). A sequence alignment analysis reveals that Y96 and Y101 in ScoE are conserved in isocyanide forming enzymes including AecA, AmcA, Sav607, ScoE, and SfaA (Figure S4). Hence, Y66/Y71 in SfaA and Y63/Y68 in AecA which correspond to Y96/Y101 in ScoE were selected for mutagenesis studies. In vitro assays using 1 as the substrate to investigate the roles of the Y66F and Y71F variants in SfaA along with the Y63F and Y68F variants in AecA showed that none of them can catalyze 4 and 6 formation. These findings are consistent with the previous studies performed with ScoE⁴⁵ and suggest that these conserved tyrosine residues are important for isocyanide formation.

To further delineate their possible functions, in situ ¹³C NMR assays using [5-¹³C]-1 as the substrate were conducted. Solutions containing reconstituted enzymes, 2OG and [5-¹³C]-1 with a final concentration of 0.2 mM reconstituted enzyme, 5 mM 2OG, and 1 mM substrate were prepared (600 μL, in 100 mM Tris-HCl pH 7.5). The reaction mixtures were left at 4 °C for 14 h. After overnight incubation, 30 μL of d₆-DMSO was added to the solution prior to a ¹³C NMR measurement. Both SfaA variants produce a species with a ¹³C NMR peak at δ = 91 ppm in the presence of 2OG and O₂ (Figure S5). In comparison with the substrate ([5-¹³C]-1, δ = 45 ppm), the chemical shift change (45 → 91 ppm) could originate from formation of 2. However, it may also result from glyoxylate (as the hydrate form) which can be produced through the decomposition of 3 or 2. Attempts to chemically synthesize 2 were unsuccessful. Those unsuccessful attempts are probably associated with the instability of the hemiaminal. Thus, we opted to take the ¹³C NMR spectra of glyoxylate to verify the origin of this peak. ¹³C NMR spectra of glyoxylate dissolved in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) and Tris buffer were recorded (10 mM glyoxylate in 50 mM Tris-HCl, pH 7.4 or 50 mM HEPES buffer, pH 7.5). As expected, two peaks associated with the carboxylate and the carbonyl (as the hydrate form) carbons of glyoxylate (δ = 172 and 89 ppm) were observed in HEPES and also in Tris (Figure S6). On the other hand, dissolving glyoxylate in Tris buffer shows two additional peaks (δ = 175 and 91 ppm), which are associated with the cyclized hemiaminal adduct between glyoxylate and Tris (Figure S6). Notably, the peak at 91 ppm has a C–H coupling constant of 163 Hz (observed under C–H coupling mode). The ¹³C chemical shift and the C–H coupling constant of the peak at 91 ppm assigned to the glyoxylate-Tris adduct are the same as those observed in the enzymatic reactions (Figures S5 and S6). While we cannot completely rule out the possibility that the C5 of 2 might have a chemical shift and C–H coupling constant identical to those of the glyoxylate-Tris adduct, these results indicate that the peak produced in the enzymatic reactions most likely represents glyoxylate. Taken together, our observations demonstrate the importance of the conserved tyrosine residues in isocyanide formation, as no production of 4 was noted in the tyrosine mutants, and thus again suggest that a conserved strategy is likely employed by these Fe/2OG enzymes to effect isocyanide biosynthesis. Furthermore, an observation of 2 in the ScoE active site²⁴ and production of glyoxylate as the only product using the SfaA Y66F variant suggest that the conserved tyrosine, e.g., Y96 in ScoE, could potentially serve as the general base to facilitate the formation of 3 from 2. In the absence of this tyrosine, e.g., Y96 in ScoE, Y66 in SfaA, and Y63 in AecA, only glyoxylate instead of the isocyanide product (4) is formed based on the ¹³C NMR results. In these cases, glyoxylate formation is not derived from 3, but from 2 directly.

Capture of the New Product in the Enzyme Active Site.

With the goal of observing possible intermediate- and product-bound protein structures, we separately incubated SfaA and ScoE with 1 and 4 and carried out the structural characterizations by X-ray crystallography. Although cocrystallization attempts using 4 were unsuccessful, we were able to trap a previously unidentified product (6) in the active site by exposing the ScoE·Fe·1·2OG complex to O₂. ScoE (10 mg/mL) was mixed with 3.0 mM 2OG and 4.5 mM 1 in the buffer that contains 25 mM Tris, pH 8.0, 150 mM NaCl, and 3 mM dithiothreitol. The solution was kept on ice, exposed to oxygen for 30 min, and mixed with the reservoir solutions. Crystals appeared after 3–4 days at 16 °C. We solved the X-ray crystal structure of ScoE in complex with 6 using molecule replacement. A superposition of the 1.99 Å resolution ScoE structure with the published structure of substrate-bound ScoE (PDB ID: 6L6X)²⁴ shows that these structures are highly similar with a root-mean-square deviation (RMSD) of 0.130 Å for 269 Cα atoms (Figure S3). Importantly, the ScoE·Fe·6 complex reveals the formation of 6 in which the C5–O bond has a distance of 1.2 Å (Figure 3). While Å cannot be used as the substrate to afford 4 as demonstrated in the previous studies,²⁴ trapping 6 in the ScoE active site suggests that 6 is produced during the reaction.

Figure 3.

Active sites of ScoE. (A) Active sites of ScoE with 1 colored in gray, PDB ID: 6L6X. (B) Active sites of ScoE with 6 colored in forest green, PDB ID: 7SCP. 2Fo-Fc (gray mesh, contoured at 1.0 σ) electron density map for 1 and 6. Dashed lines illustrate hydrogen-bonding interactions or a salt bridge involving the compound and residues (less than 3.0 Å). The active site residues are shown in wheat-colored stick format, and iron is shown as an orange sphere.

Identification of the Multiple Reaction Products in Situ Suggests Plausible Pathways of Isocyanide Formation.

To confirm 6 as the reaction product, we carried out in situ NMR assays of SfaA and AmcA employing [5-¹³C]-1 as the substrate following the procedure used in the studies of SfaA variants. As shown in Figure 4A, two peaks (δ = 152 and 166 ppm) were discernible in the presence of O₂ and 2OG. One peak (δ = 152 ppm) is associated with the formation of 4 in which the broad feature is characteristic of the N≡C moiety. The other peak (δ = 166 ppm) was confirmed to arise from 6 by spiking the synthetic standard of 6 into the reaction mixture (Figure 4A). The capture of 6 in the active site of ScoE and in the reactions of SfaA and AmcA demonstrates that 6 is an enzymatic product. We also carried out an ¹⁸O-isotope tracer experiment using ¹⁸O₂ and H₂¹⁸O. A reaction mixture containing 50 μM AecA, 4 mM ascorbic acid, 5 mM 2OG, 0.8 mM Fe(II), and 1 mM 1 (50 mM HEPES, pH 7.5) was prepared anaerobically. The reaction was initiated by introducing ¹⁸O₂ using a balloon. For the H₂¹⁸O experiment, an ¹⁸O-containing buffer solution was used with a volumetric ratio of H₂¹⁸O and H₂¹⁶O of ~75 to 25. As shown in Figure 4B, an incorporation of molecular oxygen into the C5 carbonyl oxygen of 6 was observed. This observation weighs against the pathway that includes 2, derived from decomposition of 3, as an onward intermediate: in this case, a water molecule should have been incorporated into 6 (3 → 2 → 6). In contrast, the result of the ¹⁸O-isotope tracer experiment supports the pathway 3 → 3′ ʒ 5 → 6.

Figure 4.

(A) ¹³C NMR spectra of SfaA and AmcA catalyzed reactions using [5-¹³C]-1 demonstrate the formation of 4 and 6 with the ¹³C chemical shifts of 152 (red) and 166 (blue) ppm, respectively. The peak at 166 pm overlaps with the carbonyl carbon of synthetic 6. (B) An LC-MS analysis of the AecA reaction under various conditions (¹⁸O₂ vs H₂¹⁸O) demonstrates the incorporation of the molecular oxygen (O₂) into 6 (m/z: 174.1 → 176.1).

The production of 6 may arise from 3 as described above Alternatively, it could result from a side-reaction, e.g., 1 → 2 → 6. When 3 was used as the substrate, both AmcA and AecA could effectively produce 4 and 6 (Figure 2A and Figure S7). Thus, both 1 and 3 can lead to 4 and 6 production. Along with the ¹H NMR studies suggesting that 2 is unstable under the given conditions (pH 6–8) (Figure S2), these observations suggest possible reaction pathways for isocyanide installation. As depicted in Figure 1B, subsequent to carboxyaldimine (3) formation, HAT of the vinyl C–H results in radical species 3′. The reaction may then proceed through electron-transfer coupled decarboxylation (3′ → 4) or hydroxyl rebound/tautomerization (3′ → 5 → 6) to afford 4 and 6, respectively. Alternatively, inspired by a recent discovery that an olefin group can redirect the reactivity of an Fe/2OG enzyme from hydroxylation to epoxidation,⁸ an oxygen atom transfer from the Fe(IV)-oxo species onto the imine moiety of 3 leads to an oxaziridine (7) which may undergo ring-opening, dehydration, and decarboxylation to generate 4. In contrast, a proton removal of 7 could prompt the opening of the oxaziridine followed by tautomerization to afford 6. Importantly, our calculations (vide infra) suggest that OAT is energetically unfavorable, and a strong base is required to facilitate the deprotonation of 7, thus disfavoring the latter pathway. While the mechanistic details of the conversion of 3 to 6 are still varied, the above results are at least consistent with 3 as a common intermediate to produce 4 and 6.

Molecular Dynamics Simulations and Quantum Mechanical/Molecular Mechanical Calculations Suggest a Plausible Reaction Mechanism for Isocyanide Formation.

To gain further insight into the plausible mechanism of 4 and 6 formation, combined molecular dynamics (MD) simulations and quantum mechanical/molecular mechanical (QM/MM) calculations were carried out (Figure 5 and Figure S8). To set up the system for MD simulation, the protonation states of titratable residues (His, Glu, Asp) were assigned based on the pK_a values predicted by PROPKA⁴⁶ and a visual inspection of local H-bonding networks. The general AMBER force field (GAFF⁴⁷) was used for the substrate (3) while the partial atomic charges were quantified by the RESP method,⁴⁸ using the HF/6-31G* level of theory. The protein structure (PDB ID: 7SCP) is used for this study. The system was minimized using the combined steepest descent and conjugate gradient method. All QM/MM calculations were performed using ChemShell,⁴⁹ combining Turbomole⁵⁰ as the QM code and DL_PPLY⁵¹ as the MM code. As demonstrated in several Fe/2OG enzyme QM/MM calculations,^52–59 the second sphere residues may also play important roles. Y96, Y101, G128, Y135, and R310 are thus included in QM computations. As suggested by the QM(UB3LYP/B1)/MM-optimized geometries of key species involved in the reaction (Figure 5C), the vinyl C—H of 3 is well-positioned to the presumptive Fe(IV)=O species, suggesting that 3 can undergo a HAT to form a radical intermediate (3′) through the σ-pathway which has been suggested in Fe/2OG enzymes^60–63 and agrees with de Visser’s study.³⁹ On the other hand, that HAT undergoes the π-pathway is suggested by Liu’s QM/MM study.³⁸ Our calculations show that this HAT step has an energy barrier of 9.9 kcal/mol which is similar to the previous studies on ScoE.^38,39 Subsequently, the reaction can proceed through either a radical-induced C—C cleavage (decarboxylation) or an oxygen-rebound with energy barriers of 10.5 and 11.5 kcal/mol, respectively (3′ → TS3′ and 3′ → TS3′b). The similar energy barriers between C—C cleavage and oxygen-rebound suggest that these two pathways are competitive, but the decarboxylation pathway is slightly favored. Such a finding is in line with our experiments and Liu’s studies.³⁸ However, the decarboxylation is found to be a preferred pathway in de Visser’s studies.³⁹ During the decarboxylation, our calculation suggests that the resulting Fe(III)—OH species may act as an electron sink to facilitate CO₂ departure. Notably, the energy barrier of the OH-rebound is slightly higher than the calculated values of several Fe/2OG hydroxylases.^64,65 It might be partially associated with the hydrogen-bonding network among Fe(III)—OH, Arg310, and a water molecule. In addition, the conversion of 5 to 6 was also evaluated using the hybrid cluster-continuum model in which the reverse keto–enol tautomerization is kinetically and thermodynamically favorable (Figure S9). The calculation result is consistent with the experimental observations. Notably, although using an olefin as a chemical handle to direct the reaction from HAT to OAT has been reported previously,⁸ the pathway involving an OAT step is energetically unfavorable (Figure S10); thus, we did not carry out further calculations.

Figure 5.

(A) Plausible reaction pathways and intermediates leading to 4 and 6 formation. (B) QM(UB3LYP/B2)/MM relative energies (kcal/ mol) for 4 and 5 formation from 3. The zero point energy (ZPE) and dispersion corrections are included in the relative energies. The spin densities are listed in Table S1. (C) QM(UB3LYP/B1)/MM-optimized geometries of 3 and 3′. The distances between atoms listed in this figure are given in Å. Other geometries are shown in Figure S8. (D) Isosurface of spin density (green) for the imidoyl radical of 3′.

Using an Fe(IV)=O species to cleave a Csp²—H bond is rare. In comparison with a typical olefin moiety that does not include neighboring groups, we speculate that the imine moiety and the carboxylate group of 3 may stabilize the resulting radical and thus lower the targeted C—H bond strength. Herein, we calculate the C—H bond strength of ethene (CH₂=CH₂) and 3 to delineate the possible influence of the nitrogen atom and the carboxyl group. Our QM calculations show that the Csp²—H bond strength is 108.6 kcal/mol for ethene which is consistent with the reported value.^66,67 On the other hand, the Csp²—H bond strength of 3 is only 87.1 kcal/mol. Specifically, the resulting substrate radical can be delocalized onto the neighboring N atom and the carboxylate group (Figure 5D) and as such stabilizes the imidoyl radical.

CONCLUSION

The reactivity of two annotated Fe/2OG enzymes responsible for isocyanide formation in aerocyanidin and amycomicin was established. AecA and AmcA are new members of the increasingly recognized subgroup of Fe/2OG enzymes that catalyze the oxidation of glycine to isocyanide and CO₂. Furthermore, using a multifaceted approach including protein X-ray structure, mechanistic probe design, in situ NMR, and computational analysis, a plausible pathway performed by a group of Fe/2OG enzymes to enable isocyanide installation was revealed. This four-electron oxidation takes place in a two-step sequence in which the first reaction results in the oxidation of the glycine moiety of the substrate to a carboxyaldimine (1 → 3). This reaction begins with hydrogen atom transfer (HAT) followed by hydroxyl rebound (e.g., 1 → 2) as inferred from the observation of an Fe(IV)-oxo species and 2 in the active site of ScoE.²⁴ This led to the proposal that the resulting carbinolamine (2) can undergo subsequent dehydration to yield the carboxyaldimine 3. The present results demonstrate that the Fe/2OG isocyanide synthases can indeed catalyze oxidative decarboxylation of 3 to afford isocyanide 4 in the presence of molecular oxygen and 2OG. While the details of the isocyanide formation using 3 remain to be fully elucidated, the carboxyaldimine 3 also undergoes a side-reaction to afford 6, which can be observed in the enzyme active site in crystallo. An observation of O₂ incorporation onto 6 suggests that a hydroxyl rebound step following HAT from 3 might take place (3 → 3′ → 5 → 6), which is supported by computation models. While a pathway involving an OAT step cannot be ruled out, it is proposed that 3′ represents a bifurcation point in the catalytic cycle. Whereas single electron transfer and decarboxylation yield the isocyanide (3′ → 4), susceptibility to hydroxyl rebound can result in partitioning to the formation of 6 (3′ → 5 → 6).

Using an imine group generated in the first reaction to stabilize the resulting radical species and to further direct the reactivity from OH-rebound to decarboxylation exemplifies an elegant approach employed by Fe/2OG enzymes to forge N≡C triple bond formation. Our study not only demonstrates that this conserved strategy is employed in isocyanide-containing lipopeptide and polyketide natural products but also establishes the plausible reaction pathway to effect isocyanide group formation. Furthermore, these findings imply a common mechanistic paradigm for the Fe/2OG enzyme-catalyzed N≡C triple bond formation and may introduce a new synthetic biological strategy for the preparation of the isocyanide group.