Convergent Theoretical Prediction of Reactive Oxidant Structures in

Article pubs.acs.org/JACS

Convergent Theoretical Prediction of Reactive Oxidant Structures in Diiron Arylamine Oxygenases AurF and CmlI: Peroxo or Hydroperoxo? Chao Wang†,‡ and Hui Chen*,† †

Beijing National Laboratory for Molecular Sciences (BNLMS), CAS Key Laboratory of Photochemistry, CAS Research/Education Center for Excellence in Molecular Sciences, Institute of Chemistry, Chinese Academy of Sciences, Beijing 100190, P. R. China ‡ University of Chinese Academy of Sciences, Beijing 100049, P. R. China S Supporting Information *

ABSTRACT: AurF and CmlI are currently the only two known diiron arylamine oxygenases. On the basis of extensive quantum mechanical/molecular mechanical (QM/MM) spectroscopic and mechanistic modelings, here we predict that the key oxygenated intermediates in AurF and CmlI, so-called P, are uniformly hydroperoxo species having similar structures. As a basis for mechanistic unification in AurF and CmlI, the proposed diferric-hydroperoxo P is calculated to be able to promote the arylamine N-oxygenation with highly accessible kinetics. This convergent μ-η0:η2 structural assignment of P’s in AurF and CmlI can rationalize many conundrums for P, including the different Mössbauer spectroscopic parameters, low O−O vibrational frequency, ambiphilic reactivity, and inertness toward C−H activation. In view of the very limited knowledge about hydroperoxo species in diiron enzymes, the novel diferric-hydroperoxo-mediated N-oxygenation mechanism revealed in this work opens up a new avenue for understanding the O2 activation mode in nature. For elucidating the structures of transient oxidants for diiron enzymes, the promising approach of QM/MM Mössbauer spectroscopic modeling is highlighted as a key problem solver in mechanistic enzymatic research.

1. INTRODUCTION Diiron arylamine oxygenases can catalyze the six-electron arylamine N-oxygenation to nitroaryl derivatives. This constitutes a key step in the biosynthesis of some antibiotics.1,2 In this regard, AurF from Streptomyces thioluteus and CmlI from Streptomyces venezuelae are the only two such diiron enzymes known that utilize O2 to carry out the arylamine N-oxygenation in producing the antibiotics aureothin and chloramphenicol (Scheme 1A), respectively.3,4 AurF and CmlI are highly similar (34%) in amino acid sequence,5 and thereby they provide an opportunity to reveal how diiron oxygenases activate O2 to promote amine N-oxygenation. To probe their mechanisms, in

particular those of the reactive oxidants from O2 activation, extensive experimental studies have been done for both AurF and CmlI.6−18 Krebs, Bollinger, and their co-workers discovered for AurF,9 while Lipscomb, Que, Münck, and their co-workers found for CmlI,14 that there exist hyperstable oxidant intermediates, denoted as P (or P′), in these two enzymes. However, there is still a great deal of uncertainty concerning the identity and structure of P. For AurF, Krebs et al. speculated several possibilities,9,10 including μ-η0:η2 (b) and distorted μ-η2:η2 (c) peroxo species, as well as protonated μ-η0:η2 (d) and μ-η1:η1 (e) peroxo species (Scheme 1B). On the other hand, for CmlI, Lipscomb et al. proposed that P is a single diferric peroxo species, but in a μ-η1:η2 (f) binding mode (Scheme 1B).14 Very recently, on the basis of an experimental nuclear resonance vibrational spectroscopic (NRVS) approach combined with density functional theory (DFT) QM model calculations, Solomon, Krebs, Bollinger, and their co-workers suggested protonated μ-η1:η1 (e) peroxo species (Scheme 1B) as the structure of P in AurF.15 This is different from the most recently updated assignment of P to μ-η0:η2 (b) peroxo species in CmlI by Lipscomb, Que, and their co-workers based on Xray absorption spectroscopy (XAS) and resonance Raman study.16 Obviously, despite the notable deviation from the

Scheme 1. (A) Chemical Transformations by AurF and CmlI and (B) Various Structures Proposed for P

Received: June 19, 2017 Published: August 28, 2017 © 2017 American Chemical Society

13038

DOI: 10.1021/jacs.7b06343 J. Am. Chem. Soc. 2017, 139, 13038−13046

Article

Journal of the American Chemical Society canonical μ-η1:η1 (a) peroxo structure (Scheme 1B) of diiron enzymes for both AurF and CmlI recognized from spectroscopic characterizations,9,14 there is still no consensus about the structure or identity of P in these two diiron enzymes, which precludes the possibility of mechanistic unification in AurF and CmlI. Concerning the identity and structure of P, for which currently no direct experimental evidence but only indirect spectroscopic results are available,9,14 there are several important conundrums to be clarified. First, the resonance Raman spectra reported by Lipscomb et al. indicated that O−O vibrational frequency, ν(O−O), is unusually low for P in CmlI compared with those of peroxo species identified previously in several other diiron enzymes.14 Second, the Mö ssbauer spectroscopic quadrupole splitting parameters (ΔEQ) of P are notably very different between CmlI (−0.23/−0.68 mm/s) and AurF (0.35/−0.66 mm/s).9,14 Whether this is an indication of different identities, or just suggesting different structures of P in these two enzymes, remains unknown. Third, as found by Lipscomb, Que, Guo, and their co-workers, the reactivity of P of CmlI must be ambiphilic, i.e., being electrophilic and nucleophilic simultaneously.17 But how could a peroxo species P be both electrophilic and nucleophilic? Considerable efforts have been made by Solomon et al. in the very recent work on AurF to reveal the origin of the electrophilic reactivity of P,15 but we know little about the source of its nucleophilic reactivity. Finally, P is not able to effect oxidation of C−H bond.14 However, previously, there was an example in diiron enzyme wherein peroxo species was considered to participate in the oxidation of weak C−H bond.19 Generally, the reason for most of the above conundrums about P was attributed by Lipscomb et al. to a peroxo binding mode not yet recognized in diiron enzymes, which was tentatively assigned to a μ-η1:η2 bridging one (f in Scheme 1B).14 Is there any alternative rationale? In view of the fundamental importance of P in diiron arylamine oxygenases AurF and CmlI, it is highly desired to distinguish the real chemical and structural forms of P from the various possibilities. In this work, on the basis of extensive quantum mechanical/molecular mechanical (QM/MM) modelings, we are able to elucidate that P’s in both CmlI and AurF are uniformly hydroperoxo species rather than the peroxo ones that were often proposed in previous works. This convergent hydroperoxo prediction for P’s in AurF and CmlI, the protonation state of which is in line with the very recent work on AurF by Solomon et al.,15 can resolve all the conundrums mentioned above and form a structural basis for mechanistic unification in AurF and CmlI. Furthermore, as additional support for the hydroperoxo assignment for P, our detailed QM/MM mechanistic explorations indicate that such a hydroperoxo P is able to bring about the N-oxygenation of arylamine with highly feasible kinetics. Importantly, compelling examples of diiron enzymes that employ hydroperoxo as the reactive oxidant for the substrate are still very rare. Therefore, the convergent hydroperoxo prediction for P’s in AurF and CmlI, if proved in future experiments of direct structural determination, will largely reshape the current mechanistic understanding of diiron arylamine oxygenases and also open new avenues to investigate the central issue of O2 activation mode in diiron enzymes.20

the basis of the recently determined X-ray structure of AurF with product p-nitrobenzoate bound (PDB code: 3CHT, chain A of the dimer).1 The initial structure of CmlI was taken from the X-ray structure (PDB code: 5HYG, monomer).2 The protonation states of titratable residues were determined by combining the pKa values predicted from PROPKA 2.021 with careful inspection of the Hbonding and metal-binding environment. After these procedures, we found that all histidine residues are singly protonated, and that all arginine and lysine residues are protonated. For the acidic residues, all the glutamic and aspartic residues were deprotonated. There are no cysteine linkages in any crystal structures. After adding the missing hydrogen atoms through the HBUILD module,22 the positions of H atoms in the protein structure were optimized by adopting the Newton−Raphson method implemented in CHARMM.23 Subsequently, a 16 Å thick solvent water layer was built around the enzyme. The inner 8 Å of the water layer was then relaxed through a procedure involving (1) MM optimization, (2) heating to 300 K, (3) equilibration for 3 ps, and (4) a second optimization. This procedure for adding water was repeated twice to ensure that less than 100 additional waters were added in one procedure. QM/MM Methodology and Software. QM/MM calculations were carried out by ChemShell,24 combining Turbomole25 and DL_POLY.26 The electronic embedding scheme27 was employed in the polarizing effect of the enzymatic environment on the QM region. To treat the QM/MM boundary, the hydrogen link atoms28 and the charge shift model29 were used. For the QM part, we used the density functional theory (DFT) method in unrestricted formalism to handle symmetry-broken open-shell electronic states. For the MM part, the CHARMM force field30 was used. Geometry optimizations were performed by using the B3LYP functional31 in combination with def2SVP basis set.32 The standard convergence thresholds used for QM/ MM geometry optimization are MaxStep = 0.18000 × 10−2 (Bohr); RMSStep = 0.12000 × 10−2 (Bohr); MaxGrad = 0.45000 × 10−3 (Hartree/Bohr); and RMSGrad = 0.30000 × 10−3 (Hartree/Bohr). To refine the electronic energy, single-point B3LYP/MM calculations with the triple-ζ basis set def2-TZVP32 were conducted, based on the geometry from B3LYP(def2-SVP)/MM calculations. Based on previous Mössbauer spectroscopic experiments,9,14 an open-shell singlet state (Stotal = 0) anti-ferromagnetically coupled from the high-spin (S = 5/2) ferric site is considered for P in this work. QM Region. The QM region in the enzyme, shown in Scheme 2, comprises a diiron cofactor with the first coordination sphere residues

Scheme 2. Residues Included in the QM Region of QM/MM Calculations for AurF/CmlI

of two irons, which covers three histidines (His139, His223, and His230 in AurF; His147, His232, and His239 in CmlI), four glutamic acids (Glu101, Glu136, Glu196, and Glu227 in AurF; Glu109, Glu144, Glu205, and Glu236 in CmlI). When necessary, the QM region also includes substrate (paminobenzoate in AurF) and an oxygen molecule. Histidines were modeled as methylimidazole, glutamic acid as CH3COO−. QM/MM Optimized Region. To keep the QM/MM geometry optimization calculations feasible, it is not possible to take the whole enzyme as the optimized region, and commonly a QM/MM optimized region close to the enzymatic active center is selected. The QM/MM geometry optimization region includes the QM region and all residues within 6 Å from the QM region. AurF: Leu31, Trp35, Arg38, Ala39, Ala40, Val41, Tyr93, Arg96, Val97, Ile98, Ala99, Thr100, Glu101, Gln102, Leu103, Ile104, Ala105, Glu106, Ala108, Phe109, Ser128, Gln131, Ala132, Ile133, Val134, Asp135, Glu136, Ser137, Phe138,

2. COMPUTATIONAL DETAILS Setup of System. The initial structure of p-aminobenzoate Noxygenase (AurF) with and without substrate bound was prepared on 13039


Article

Journal of the American Chemical Society

η1:η2 peroxo (C-P3,A-P3), μ-η1:η1 hydroperoxo (C-P4,A-P4), and μ-η0:η2 hydroperoxo (C-P5/C-P6,A-P5/A-P6), as well as some other species (Figure S2). From their QM/MMcalculated Mössbauer parameters, especially the quadrupole splitting (ΔEQ) as collected in Table 1, it is clear that one pair of μ-η0:η2 hydroperoxo species, i.e., C-P6 (ΔEQ = −0.45/−1.04 mm/s) and A-P6 (ΔEQ = 0.53/−0.92 mm/s) with similar structure, stand out for their good agreement with the experimental values (ΔEQ = −0.23/−0.68 mm/s for CmlI, 0.35/−0.66 mm/s for AurF),9,14 especially after considering the possible error size of 0.3−0.5 mm/s for the current Mössbauer ΔEQ calculations. Therefore, our results for the first time unify AurF and CmlI to imply that their P’s are identical in constitution and similar in structure. This convergent structural proposal for P’s provides a basis for mechanistic unification in AurF and CmlI. Importantly, despite the fact that both the current work based on Mössbauer spectroscopic data (ΔEQ) and the recent work of Solomon et al. based on NRVS spectroscopic data15 similarly suggest hydroperoxo chemical identity for P, the proposed structures are different. In this work a μ-η0:η2 hydroperoxo structure for P is proposed for both AurF and CmlI, whereas Solomon et al. assigned a μ-η1:η1 hydroperoxo structure similar to that of A-P4 for AurF.15 Notably, our calculated Mössbauer spectroscopic data (ΔEQ) for A-P4 in Table 1 do not support the assignment of A-P4 as the structure of P in AurF. The final resolution of this controversy about the real structure of P in AurF and CmlI depends on future direct structural measurement. Interestingly, despite their similarity in structure, including H-bonding pattern of hydroperoxo moiety (Figure 2), C-P6 and A-P6 differ in one sign of the calculated ΔEQ. For these two similar structures, being in agreement with differently signed Mössbauer experimental data of two enzymes (AurF and CmlI) simultaneously can hardly be viewed as an accidental coincidence. To check whether the MM polarization is the key factor to cause the difference, we did Mössbauer modeling based on structures of C-P6 and A-P6 after removing the MM protein environment. However, the calculated Mössbauer spectroscopic parameters38 are similar to the corresponding data shown in Table 1, which implies that protein polarization is not the origin for the different Mössbauer spectroscopic parameters. Checking the geometries of C-P6 and A-P6 in Figure 2, the largest coordination difference by up to 0.08 Å is from the Fe−O dative bond distance of the bridging Glu residue (Glu227/Glu236 in AurF/CmlI) and Fe1, for which iron the Mössbauer spectroscopic parameters make a difference between two enzymes. We attribute this structural difference, which is also exerted by protein environment via different geometric constraints on active site, to the origin of the calculated difference in Mössbauer parameters. These results hence imply that it is not the different identity of P, but the different enzymatic environments of AurF and CmlI, that lead to the observed difference in ΔEQ, which constitutes a warning counter example for simply assuming similar ΔEQ data for those enzymatic species of similar structures in active site. These enzyme-specific data of ΔEQ, also render the adopted QM/MM approach indispensable in the current case, which excludes the computational approach based on pure QM model without proper consideration of the protein environment. Correspondingly, we noted that there exists significant difference for ΔEQ between the current QM/MM results and previous QM results based on an active site model.39 The

His139, Thr140, Tyr141, Met142, His143, Gly192, Ala193, Val194, Ala195, Glu196, Thr197, Cys198, Ile199, Asn200, Leu202, Leu203, Ala204, Leu206, Ala207, Leu218, Ile219, Thr220, Thr221, Leu222, His223, Leu224, Arg225, Asp226, Glu227, Thr228, Ala229, His230, Gly231, Ser232, Ile233, Val234, Phe264, Arg302, Wat503, Wat504, Wat507, Wat509, Wat510, Wat539, Wat542, Wat548, Wat579, Wat599, Wat604, Wat610, Wat612, Wat615, Wat626, Sol1509, Sol2046, Sol7694. CmlI: Leu41, Trp45, Arg48, Ala49, Val51, Tyr101, Asn104, Thr105, Val106, Leu107, Ile108, Glu109, Gln110, Ile112, Ala113, Asn114, Ala116, Phe117, Gln139, Ala140, Met141, Val142, Asp143, Glu144, Gln145, Tyr146, His147, Thr148, Leu149, Met150, His151, Phe200, Ala201, Thr202, Val230, Ala204, Glu205, Ile206, Ile208, Asn209, Tyr211, Leu212, Thr228, Val229, Lys230, Leu231, His232, Asn233, Arg234, Asp235, Glu236, Tyr237, Cys238, His239, Ala240, Ser241, Ile242, Ser243, Gly244, Met246, Wat523, Wat524, Wat528, Sol138, Sol2335, Sol2358, Sol2526, Sol2609, Sol2775, Sol3978, Sol5883, Sol7696, Sol7723. Mö ssbauer Spectroscopic Parameters. The Mössbauer isomer shift (δ) and quadrupole splittings (ΔEQ) were calculated with the program ORCA33 using B3LYP functional at the corresponding QM/ MM optimized geometries with the inclusion of MM point charges. In these calculations, iron was described by the triply polarized core properties basis set CP(PPP),34 and the other atoms were described by the def2-SVP basis set.32 The isomer shift was evaluated from the electron density at the iron nucleus, using the calibration constants obtained previously from B3LYP/TZVP/Fe:CP(PPP) level (geometries optimized at the TPSS/TZVP level).35 For the effect of using larger, while computationally much more expensive triple-ζ basis set (TZVP) in Mössbauer spectroscopic calculations, our pervious testing results indicated that the difference between these two calculation levels (separately employing triple-ζ TZVP and double-ζ def2-SVP basis sets, in both geometry optimization and Mössbauer spectroscopic parameter calculation) is very small in isomer shift calculation.36 In the comparison between calculated and experimental data, it is notable that current DFT calculations for 57Fe Mössbauer spectroscopic parameters have certain errors, and 0.1 mm/s magnitude of standard deviation for isomer shift is normally expected.35 It was also found to be more challenging to model quadrupole splitting accurately,37 and the maximum error of 0.3−0.5 mm/s might be expected.

3. RESULTS AND DISCUSSION QM/MM Modelings for Structures of P’s in AurF and CmlI. We commenced our QM/MM modeling by resolving the chemical identity and structure of P. As shown in Figure 1, after QM/MM relaxations, we obtained a series of candidate diferric species for P in both AurF (A) and CmlI (C), which include μη1:η1 peroxo (C-P1,A-P1), μ-η2:η2 peroxo (C-P2,A-P2), μ-

Figure 1. Candidate diferric structures for P in AurF and CmlI (denoted as A-Px and C-Px, x = 1−6) obtained from the QM/MM approach. 13040


Article


Table 1. QM/MM-Calculated Mössbauer Spectroscopic Parameters (Isomer Shift, δ, and Quadrupole Splitting, ΔEQ) of Candidate Diferric Structures for P in CmlI and AurF, in Comparison with Experimental Data calcd dataa (mm/s)

calcd dataa (mm/s)

CmlI

δ

ΔEQ

AurF

δ

ΔEQ

C-P1 C-P2 C-P3 C-P4 C-P5 C-P6

0.71/0.56 0.59/0.65 0.63/0.53 0.54/0.49 0.55/0.50 0.53/0.44

1.51/1.44 −0.60/0.54 −1.50/1.46 0.61/−0.94 0.40/−0.91 −0.45/−1.04

A-P1 A-P2 A-P3 A-P4 A-P5 A-P6

0.66/0.77 0.57/0.67 0.66/0.61 0.55/0.58 0.55/0.54 0.55/0.47

1.25/2.95 −0.73/−0.65 1.70/−1.17 0.63/1.17 −0.55/0.75 0.53/−0.92

exptb

0.62, 0.54

−0.23, −0.68

exptc

0.61, 0.54

0.35, −0.66

a

For calculated data of each species, data before and after slash are for Fe1/Fe2. Fe1 and Fe2 refer to the iron coordinated with two and one His residues, respectively. The calculated ΔEQ data in best agreement with experimental data are shown in bold. bFrom ref 14. cFrom ref 9.

Figure 2. QM/MM optimized structures of A-P6 and C-P6.

diiron core apparently can hardly represent P that has been determined experimentally to be diferric species, which was also confirmed by the calculated Mössbauer spectroscopic parameters here (δ = 1.03/0.58 mm/s, ΔEQ = 2.33/−1.52 mm/s for Fe1/Fe2, Fe−Fe distance 3.02 Å). For hydroperoxo species, a diferric core can be kept during QM/MM geometry optimization. However, the Fe−Fe distance (2.94 Å) in the optimized structure is apparently shorter than that implicated from experiment (3.35 Å),16 which is very close to the 3.32 Å in C-P6. Moreover, the corresponding calculated Mössbauer spectroscopic parameters (δ = 0.52/0.54 mm/s, ΔEQ = −1.56/1.25 mm/s for Fe1/Fe2) for the oxo-bridged μ-η0:η2 hydroperoxo are different far from the experimental values (δ = 0.62, 0.54 mm/s, ΔEQ = −0.23, −0.68 mm/s). Therefore, our QM/MM modelings generally do not support oxo-bridged peroxo/hydroperoxo species as P. Concerning the protonation state of peroxo moiety of P, interestingly, our QM/MM vibrational frequency calculation indicate that H/D shift on ν(O−O) is very small (0.8 cm−1) for C-P6, which implies that the missing H/D shift of ν(O−O) in resonance Raman experiment, as observed in the recent study of CmlI,16 may not be a definitive sign to determine the protonation state of peroxo moiety. QM/MM Modelings for Reactivity of P in AurF. With the above QM/MM Mössbauer spectroscopic modeling results in support of hydroperoxo species C-P6/A-P6 as P in CmlI/ AurF uniformly, a pivotal question about their reactivity arises, i.e., are they reactive in arylamine N-oxygenation? For scrutinizing their N-oxygenation reactivity we employed A-P6 structure of AurF, for which enzyme there exists product-bound (p-nitrobenzoate) X-ray crystal structure.1 With A-P6 as the oxidant, the QM/MM-calculated most favorable N-oxygenation reaction pathway from substrate p-aminobenzoate (ArNH2) to

current QM/MM results for AurF and CmlI, along with our recent discovery that QM/MM modeling for the intermediate from O 2 activation in another diiron enzyme cADO (cyanobacterial aldehyde-deformylating oxygenase) can assist determination of its real structure,36 all point to the promising utility of QM/MM Mössbauer spectroscopic modeling in diiron enzymes. In addition to the Mössbauer parameters, the calculated vibrational frequency data of C-P6 also support its assignment as P. Compared with the canonical μ-η1:η1 peroxo species CP1, the calculated ν(O−O) of P is lowered by 90 cm−1, which is in line with the resonance Raman experimental discovery by Lipscomb et al. that ν(O−O) of P in CmlI was 60−107 cm−1 lower than observed for the other previously reported enzyme diferric-peroxo intermediates.14 In contrast, the μ-η1:η2 peroxo species C-P3 assumed as P by Lipscomb et al., was calculated to have its ν(O−O) 168.5 cm−1 lower than C-P1, which seems too low to support its assignment as P. Interestingly, a μ-η0:η2 coordination mode of the (hydro)peroxo to the diiron center was recently discovered in a mutant of the diiron enzyme toluene 4-monooxygenase (T4moH).40 Different from the above structures with carboxylate double bridges (Glu144/Glu236) for two irons, very recently Lipscomb and Que updated their structural proposal for P in CmlI to oxobridged μ-η0:η2 peroxo species.16 To check this proposal with QM/MM modeling, for CmlI we optimized both peroxo and hydroperoxo μ-η0:η2 structures with Glu144/oxo double bridges (C-P9 and C-P10 in Figure S2). Interestingly, we found that the peroxo species cannot produce diferric electronic structure. Alternatively, a mixed-valent ferrous−ferric diiron core bonding with μ-η0:η2 superoxo moiety was obtained from geometry optimization. Such a diferric-reduced species generated from internal electron transfer (ET) from the peroxo moiety to the 13041


Article


Figure 3. Most favorable N-oxygenation reaction pathway of p-aminobenzoate (ArNH2) to p-nitrosobenzoate (ArNO) with A-P6 from the QM/ MM calculations.

Scheme 3. Comparison between QM-Calculated N-Oxygenation Mechanisms Reported Previously and the QM/MM-Calculated Mechanism in This Work For AurF

p-nitrosobenzoate (ArNO, a confirmed intermediate in the enzymatic conversion of ArNH2 to ArNO1,17) is depicted in Figure 3. The QM/MM-calculated N-oxygenation mechanism initiates from a N−H activation of the ArNH2-bound A-P6 (1), through an unusual double proton-coupled electron-transfer (doublePCET) process, i.e., H2/H3 transfer respectively to O1/E196

coupled with an ET of amine N lone pair to Fe2 (Fe coordinated with H230, Figure 2), which reduces the iron from ferric to ferrous one in 2. This double-PCET step that effectively cleaves the first N−H bond of amino group in the substrate ArNH2, only needs to overcome a barrier of 14.7 kcal/mol via transition state TS1. Subsequently, a H-bonding pattern change (via TS2) and PCET process (via TS3) transfers 13042


Article


Figure 4. Preferred and kinetically favorable reaction pathway of p-nitrosobenzoate (ArNO) to p-nitrobenzoate (ArNO2) with A-P6 from the QM/ MM calculations.

generated either with or without involving ArNHOH (2eoxidation intermediate), these data appear to imply that the arylamine oxidation scenario proposed recently by Lipscomb et al. via ArNO better describes the N-oxygenation reaction among other alternative ones.17 Here, it is interesting to compare the current N-oxygenation mechanism from the QM/MM modelings of the whole AurF enzyme with the most possible ones found in the very recent work of Solomon et al. from the pure QM modelings of the enzymatic active site only.15 As comparatively shown in Scheme 3, Solomon et al. identified two kinetically most favored pathways (with reaction barriers of ca. 12−14 kcal/mol): (1) one is a stepwise mechanism with first single-ET-assisted O−O cleavage of the η1:η1 hydroperoxo structure similar to A-P4 to generate high-valent Fe(IV)-oxo species, followed by the N−O formation; (2) the other is a direct electrophilic attack on substrate initiated from the concomitant N−O/O-O formation/cleavage of μ-η0:η2 hydroperoxo structure corresponding to A-P5 (not A-P6), via a two-electron process without involving high-valent ferryl species.15 In both kinetically comparable pathways of Solomon et al., no N−H bond in arylamine substrate is cleaved before O−O cleavage or N−O bond formation (the mechanism after N−O bond formation was not explored in Solomon’s work15), which is in sharp contrast to our most favored pathway from 1 (corresponding to A-P6) shown in Figure 3 featured by N−H cleavage (with reaction barrier of ca. 14 kcal/mol) before O−O cleavage and N−O bond formation. In fact, in our QM/MM calculations, the pathways initiated by neither O−O cleavage nor N−O formation from 1 are favored kinetically, with barrier as high as 32 kcal/mol (Figures S3 and S7), in which O−O cleavage or N−O formation is accompanied by N−H activation.41 From these comparisons, it is clear that alternative to the ones reported by Solomon et al.,15 our QM/MM data in this work provide a mechanistically quite different but kinetically comparable N-oxygenation pathway. Since our work demonstrates that N−H activation is likely to be the rate-limiting step, it should be reflected by the kinetic isotope effect (KIE) of deuterated amine substrate. Therefore, we here suggest using KIE as a probe for the mechanism of arylamine oxygenation in diiron enzymes.

H2 from the distal O1 to the carboxylate residue E101, and reduces ferric-ferrous mixed-valent (Fe1III/Fe2II) core in 2 to diferrous (Fe1II/Fe2II) core in 4, concomitantly generating a superoxo fragment. This superoxo in 4 then attacks the Nradical of the substrate through TS4, with a concerted O−O homolytic cleavage, to generate ArNHO and regenerate mixedvalent (Fe1III/Fe2II) core in 5. ArNHO is transformed to ArNOH in 6 through a double proton transfer (PT) process (H1/H3 transfer respectively to O2/O1) via TS5, which effectively breaks the second N−H bond of amino group in the substrate. From 6, ArNO is generated via TS6 by a PCET process, i.e., H1 transfer to O2 coupled with an ET to Fe2, hence reducing the Fe2 from ferric to ferrous one in 7. Finally, via TS7, a PT process of H2 from the residue E101 to O2 protonates the hydroxo bridge to produce water with the diferrous core left in 8. In view of the whole pathway from ArNH2 to ArNO, the highest barrier is 14.7 kcal/mol via TS1 for breaking the first N−H bond in substrate ArNH2. This relatively low barrier indicates that the kinetics employing A-P6 as oxidant for arylamine N-oxygenation is highly feasible, which supports the assignment of this intermediate as P. Notably, PCET processes that involve ET from the substrate to the diiron core are extensively encountered in the N-oxygenation pathway (Figure 3), which echoes the opinion repeatedly proposed previously that ET from the substrate is likely to play an important role in the mechanism of AurF and CmlI.14,18 Another consistency between our QM/MM results and experimental mechanistic proposal is that the ArNO intermediate is formed with the diferrous diiron core.17 Interestingly, we found that there exists two kinetically comparable pathways from 5 to 6, one is the simple double PT process as outlined in Figure 3, the other is via hydroxylamine radical ArNHOH+• (Figure S5). The formation of the ArNHOH+• cation radical associated with ferric-ferrous mixed-valent Fe1II/Fe2III core is fully consistent with the experimental findings in CmlI that ArNHOH can reduce the diferric core of enzymes to generate P in the presence of O2,17 in which the mixed-valent Fe1II/Fe2III core is the partially 1ereduced intermediate state from the diferric core. In these two kinetically comparable reaction pathways, since our calculations suggest that ArNO (4e-oxidation intermediate) is inevitably 13043


Article


Figure 5. C−H activation mechanism of p-methylbenzoate (ArCH3) with A-P6 from the QM/MM calculations.

(HAA) reactions.43,44 Over a wide scope of non-heme mononuclear iron enzymatic and biomimetic systems reported to date, there are only few examples to suggest ferrichydroperoxo reactive in oxygenation reaction.43,45−50 Among these few cases, of exceptional relevance to the current elucidation of N-oxygenation reactivity of diferric-hydroperoxo oxidant is the recent finding of Nam, Shaik, and their coworkers, which demontrated that a mononuclear synthetic ferric-hydroperoxo complex is reactive in sulfoxidation of thioanisoles,46 as also implicated latter in a mononuclear nonheme iron-containing enzyme of superoxide reductase.49 These studies indicate that ferric-hydroperoxo and diferric-hydroperoxo species can directly perform heteroatom (e.g., S or N) oxygenation reactions, with the former found reactive for OAT reaction, whereas the latter active for oxygenation of N−H bonds. The current finding of the high N−H oxygenation reactivity of hydroperoxo species in diiron arylamine oxygenase may also have some mechanistic implication on the mononuclear non-heme arylamine oxygenase PrnD.51 Finally, it is worthwhile to point out that although the mechanism explored in this work is based on A-P6 species, similar mechanism and reactivity are also expected for the species such as A-P5 with same identiy but only different H-bonding pattern.

The mechanisms of Solomon et al. were put forward to rationalize the electrophilic reactivity of P, but since no reactivity of P toward ArNO was reported,15 it is unclear whether it can explain the nucleophilic reactivity of P. From our QM/MM calculations, the ambiphilic character of P in Noxygenation can be well rationalized by its hydroperoxo rather than peroxo form. The electrophilic reactivity of diferrichydroperoxo has been clearly demonstrated in the transformation of ArNH2 to ArNO (Figure 3). The nucleophilic reactivity, is shown in the mechanism of the ArNO-to-ArNO2 transformation (Figure 4), in which we found that PCET from hydroperoxo can easily generate superoxo species, thus making P ready for the subsequent nucleophilic attack to the N of the nitroso group. In this way, P can change between the two faces of its ambiphilic reactivity. Alternative to the ArNO-to-ArNO2 transformation in Figure 4, we also tested the mechanistic proposal of Krebs and Bollinger about the ArNHOH-toArN(OH)2 pathway.10 The results (Figure S9) indicate that this pathway is kinetically less favorable than the one in Figure 4. In addition, different from the previous proposal,10 the produced Ar-dihydroxylamine was not found to reduce the diferric diiron cluster (Figure S9). Therefore, our calculations do not support the mechanistic pathway of generating Ardihydroxylamine from Ar-hydroxylamine oxygenation by P. It is intriguing to know why C−H oxygenation is not preferred by A-P6, as evidenced experimentally for P in CmlI.14 To explore this issue, we replaced the ArNH2 substrate by ArCH3. In comparison to ArNH2, the higher barrier and smaller reaction energy of the QM/MM-calculated pathway for ArCH3 (Figure 5) indicate that the benzylic C−H cleavage is neither kinetically nor thermodynamically preferred compared with the anilino N−H cleavage, in agreement with the experimental discovery for P. Interestingly, the superoxo species 14 has to be generated before the C−H activation can occur, which implies that hydroproxo species A-P6 is unable to directly promote the C−H cleavage process. Interestingly, compared to C−H bond, N−H activation was found to be PCET prone by iron-oxo oxidant,42 which is corroborated here for diiron-hydroperoxo oxidant as shown in Figure 3. To date, except for the very recent study of AurF by Solomon et al.,15 there is still no study to support the direct reactivity of diferric-hydroperoxo species, which is a missing piece in understanding the chemistry of non-heme diiron enzymes. The high N-oxygenation but low C−H oxygenation reactivity found above for A-P6, is in sharp contrast to the known missing reactivity of heme ferric-hydroperoxo species in oxygen atom transfer (OAT) and hydrogen atom abstraction

4. CONCLUSIONS In summary, on the basis of extensive QM/MM spectroscopic and mechanistic modelings, in this work we predict that the key oxidants resulting from O2 activation in the only two known diiron arylamine oxygenases, AurF and CmlI, so-called P, are uniformly hydroperoxo species rather than the commonly believed peroxo species. Our results clearly suggest that P’s in AurF and CmlI share chemical identity and have similar structures, thereby forming a basis for mechanistic unification of these two diiron enzymes. The proposed μ-η0:η2 diferrichydroperoxo oxidant P is elucidated to be able to effect the Noxygenation reaction of amine to a nitroso group first and then to a nitro group with very accessible kinetics, through a mechanism quite different from the ones recently reported based on QM model calculations.15 This convergent assignment of P’s as μ-η0:η2 hydroperoxo complexes can successfully rationalize many conundrums about P in the research area of AurF and CmlI, including the different Mössbauer spectroscopic parameters, low O−O vibrational frequency, ambiphilic reactivity, and inertness toward C−H activation. Notably, the current convergent structural prediction of μ-η0:η2 hydroperoxo P’s in AurF and CmlI, based mainly on Mö s sbauer 13044


Article

Journal of the American Chemical Society spectroscopic analysis, is different from the recent μ-η1:η1 hydroperoxo assignment of P in AurF, derived from a different spectroscopic approach (NRVS).15 Considering the relatively high stability of P’s in AurF and CmlI, it is hoped that these structural predictions for P would be confirmed/disproved by direct experimental structural characterization in the near future. In view of the currently very limited knowledge of hydroperoxo species in diiron enzymes, the new diferrichydroperoxo-mediated N-oxygenation mechanism revealed in this work opens up a new avenue for our understanding of the O2 activation mode in nature. This work, along with the recent encouraging results in elucidating the structure of transient oxy species from another diiron enzyme, cADO,36 demonstrates the promise of QM/MM Mössbauer spectroscopic modeling in diiron enzymes, the performance of which certainly merits more investigation in the future.

■

(14) Makris, T. M.; Vu, V. V.; Meier, K. K.; Komor, A. J.; Rivard, B. S.; Münck, E.; Que, L., Jr.; Lipscomb, J. D. J. Am. Chem. Soc. 2015, 137, 1608. (15) Park, K.; Li, N.; Kwak, Y.; Srnec, M.; Bell, C. B.; Liu, L. V.; Wong, S. D.; Yoda, Y.; Kitao, S.; Seto, M.; Hu, M.; Zhao, J.; Krebs, C.; Bollinger, J. M., Jr.; Solomon, E. I. J. Am. Chem. Soc. 2017, 139, 7062. (16) Jasniewski, A. J.; Komor, A. J.; Lipscomb, J. D.; Que, L., Jr. J. Am. Chem. Soc. 2017, 139, 10472. (17) Komor, A. J.; Rivard, B. S.; Fan, R.; Guo, Y.; Que, L., Jr.; Lipscomb, J. D. J. Am. Chem. Soc. 2016, 138, 7411. (18) Platter, E.; Lawson, M.; Marsh, C.; Sazinsky, M. H. Arch. Biochem. Biophys. 2011, 508, 39. (19) Beauvais, L. G.; Lippard, S. J. J. Am. Chem. Soc. 2005, 127, 7370. (20) Wallar, B. J.; Lipscomb, J. D. Chem. Rev. 1996, 96, 2625. (21) Bas, D. C.; Rogers, D. M.; Jensen, J. H. Proteins: Struct., Funct., Genet. 2008, 73, 765. (22) Brünger, A. T.; Karplus, M. Proteins: Struct., Funct., Genet. 1988, 4, 148. (23) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.; Swaminathan, S.; Karplus, M. J. Comput. Chem. 1983, 4, 187. (24) Sherwood, P.; de Vries, A. H.; Guest, M. F.; Schreckenbach, G.; Catlow, C. R. A.; French, S. A.; Sokol, A. A.; Bromley, S. T.; Thiel, W.; Turner, A. J.; Billeter, S.; Terstegen, F.; Thiel, S.; Kendrick, J.; Rogers, S. C.; Casci, J.; Watson, M.; King, F.; Karlsen, E.; Sjøvoll, M.; Fahmi, A.; Schäfer, A.; Lennartz, C. J. Mol. Struct.: THEOCHEM 2003, 632, 1. (25) TURBOMOLE, version 6.4; TURBOMOLE GmbH: Karlsruhe, Germany, 2012. (26) Smith, W.; Forester, T. R. J. Mol. Graphics 1996, 14, 136. (27) Bakowies, D.; Thiel, W. J. Phys. Chem. 1996, 100, 10580. (28) de Vries, A. H.; Sherwood, P.; Collins, S. J.; Rigby, A. M.; Rigutto, M.; Kramer, G. J. J. Phys. Chem. B 1999, 103, 6133. (29) Sherwood, P.; de Vries, A. H.; Collins, S. J.; Greatbanks, S. P.; Burton, N. A.; Vincent, M. A.; Hillier, I. H. Faraday Discuss. 1997, 106, 79. (30) CHARMM force field: MacKerell, A. D., Jr.; Bashford, D.; Bellott, M.; Dunbrack, R. L., Jr.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T. K.; Mattos, C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; Roux, B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; Watanabe, M.; Wiórkiewicz-Kuczera, J.; Yin, D.; Karplus, M. J. Phys. Chem. B 1998, 102, 3586. (31) Becke, A. D. J. Chem. Phys. 1992, 96, 2155. (32) Weigend, F.; Ahlrichs, R. Phys. Chem. Chem. Phys. 2005, 7, 3297. (33) Neese, F. ORCA, version 2.9.1; University of Bonn, Bonn, Germany, 2006. (34) Neese, F. Inorg. Chim. Acta 2002, 337, 181. (35) Römelt, M.; Ye, S. F.; Neese, F. Inorg. Chem. 2009, 48, 784. (36) Wang, C.; Zhao, C.; Hu, L.; Chen, H. J. Phys. Chem. Lett. 2016, 7, 4427. (37) (a) Nemykin, V. N.; Hadt, R. G. Inorg. Chem. 2006, 45, 8297. (b) Bochevarov, A. D.; Friesner, R. A.; Lippard, S. J. J. Chem. Theory Comput. 2010, 6, 3735. (c) Hendrich, M. P.; Gunderson, W.; Behan, R. K.; Green, M. T.; Mehn, M. P.; Betley, T. A.; Lu, C. C.; Peters, J. C. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 17107. (d) Popescu, C. V.; Mock, M. T.; Stoian, S. A.; Dougherty, W. G.; Yap, G. P. A.; Riordan, C. G. Inorg. Chem. 2009, 48, 8317. (38) The calculated Mössbauer spectroscopic parameters after removal of the MM point charges are as follow: C-P6, δ = 0.57/ 0.46 mm/s, ΔEQ = −0.44/−0.93 mm/s; A-P6, δ = 0.57/0.47 mm/s, ΔEQ = 0.43/−0.86 mm/s, all presented for Fe1/Fe2. (39) Jayapal, P.; Ansari, A.; Rajaraman, G. Inorg. Chem. 2015, 54, 11077. (40) Acheson, J. F.; Bailey, L. J.; Brunold, T. C.; Fox, B. G. Nature 2017, 544, 191. (41) Using the corresponding N−O and O−O bond distances of the transition state of the direct electrophilic attack on substrate found in ref 15, we also tried constraint QM/MM geometry optimization from 1 (ArNH2-bound A-P6) with N−O/O−O bonds fixed. Such constraint geometry optimization for mimicking the transition state

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.7b06343. Figures S1−S9 and Tables S1−S4, absolute energies, and Cartesian coordinates of the optimized structures (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*[email protected] ORCID

Chao Wang: 0000-0001-5664-2350 Hui Chen: 0000-0003-0483-8786 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the National Natural Science Foundation of China (Nos. 21290194, 21473215, and 21521062), and Institute of Chemistry, Chinese Academy of Sciences.

■

REFERENCES

(1) Choi, Y. S.; Zhang, H.; Brunzelle, J. S.; Nair, S. K.; Zhao, H. Proc. Natl. Acad. Sci. U. S. A. 2008, 105, 6858. (2) Knoot, C. J.; Kovaleva, E. G.; Lipscomb, J. D. J. Biol. Inorg. Chem. 2016, 21, 589. (3) He, J.; Hertweck, C. J. Am. Chem. Soc. 2004, 126, 3694. (4) He, J.; Magarvey, N.; Piraee, M.; Vining, L. C. Microbiology 2001, 147, 2817. (5) Lu, H.; Chanco, E.; Zhao, H. Tetrahedron 2012, 68, 7651. (6) Winkler, R.; Hertweck, C. Angew. Chem., Int. Ed. 2005, 44, 4083. (7) Simurdiak, M.; Lee, J.; Zhao, H. ChemBioChem 2006, 7, 1169. (8) Krebs, C.; Matthews, M. L.; Jiang, W.; Bollinger, J. M., Jr. Biochemistry 2007, 46, 10413. (9) Korboukh, V. K.; Li, N.; Barr, E. W.; Bollinger, J. M., Jr.; Krebs, C. J. Am. Chem. Soc. 2009, 131, 13608. (10) Li, N.; Korboukh, V. K.; Krebs, C.; Bollinger, J. M., Jr. Proc. Natl. Acad. Sci. U. S. A. 2010, 107, 15722. (11) Fries, A.; Winkler, R.; Hertweck, C. Chem. Commun. 2010, 46, 7760. (12) Fries, A.; Bretschneider, T.; Winkler, R.; Hertweck, C. ChemBioChem 2011, 12, 1832. (13) Chanco, E.; Choi, Y. S.; Sun, N.; Vu, M.; Zhao, H. Bioorg. Med. Chem. 2014, 22, 5569. 13045


Article

Journal of the American Chemical Society of the direct electrophilic attack on substrate in ref 15 is calculated to be 37.3 kcal/mol higher than for reactant 1, being much higher than the barrier value of ca. 13 kcal/mol found in ref 15, which implies that the potential energy surface around the reaction coordinates is considerably different in QM/MM and QM calculations for AurF. (42) Usharani, D.; Lacy, D. C.; Borovik, A. S.; Shaik, S. J. Am. Chem. Soc. 2013, 135, 17090. (43) Ray, K.; Pfaff, F. F.; Wang, B.; Nam, W. J. Am. Chem. Soc. 2014, 136, 13942 and references therein. (44) (a) Watanabe, Y.; Nakajima, H.; Ueno, T. Acc. Chem. Res. 2007, 40, 554. (b) Ogliaro, F.; de Visser, S. P.; Cohen, S.; Sharma, P. K.; Shaik, S. J. Am. Chem. Soc. 2002, 124, 2806. (c) Derat, E.; Kumar, D.; Hirao, H.; Shaik, S. J. Am. Chem. Soc. 2006, 128, 473. (d) Li, C.; Zhang, L.; Zhang, C.; Hirao, H.; Wu, W.; Shaik, S. Angew. Chem., Int. Ed. 2007, 46, 8168. (e) Davydov, R.; Perera, R.; Jin, S.; Yang, T.-C.; Bryson, T. A.; Sono, M.; Dawson, J. H.; Hoffman, B. M. J. Am. Chem. Soc. 2005, 127, 1403. (f) Park, M. J.; Lee, J.; Suh, Y.; Kim, J.; Nam, W. J. Am. Chem. Soc. 2006, 128, 2630. (45) Cho, J.; Jeon, S.; Wilson, S. A.; Liu, L. V.; Kang, E. A.; Braymer, J. J.; Lim, M. H.; Hedman, B.; Hodgson, K. O.; Valentine, J. S.; Solomon, E. I.; Nam, W. Nature 2011, 478, 502. (46) Kim, Y. M.; Cho, K.-B.; Cho, J.; Wang, B.; Li, C.; Shaik, S.; Nam, W. J. Am. Chem. Soc. 2013, 135, 8838. (47) Liu, L. V.; Hong, S.; Cho, J.; Nam, W.; Solomon, E. I. J. Am. Chem. Soc. 2013, 135, 3286. (48) Zang, C.; Liu, Y.; Xu, Z.-J.; Tse, C.-W.; Guan, X.; Wei, J.; Huang, J.-S.; Che, C.-M. Angew. Chem., Int. Ed. 2016, 55, 10253. (49) Rat, S.; Ménage, S.; Thomas, F.; Nivière, V. Chem. Commun. 2014, 50, 14213. (50) (a) Wolfe, M. D.; Altier, D. J.; Stubna, A.; Popescu, C. V.; Münck, E.; Lipscomb, J. D. Biochemistry 2002, 41, 9611. (b) Neibergall, M. B.; Stubna, A.; Mekmouche, Y.; Münck, E.; Lipscomb, J. D. Biochemistry 2007, 46, 8004. (c) Bassan, A.; Blomberg, M. R. A.; Siegbahn, P. E. M. J. Biol. Inorg. Chem. 2004, 9, 439. (d) Rocklin, A. M.; Kato, K.; Liu, H.-w.; Que, L., Jr.; Lipscomb, J. D. J. Biol. Inorg. Chem. 2004, 9, 171. (51) (a) Lee, J.; Zhao, H. Angew. Chem., Int. Ed. 2006, 45, 622. (b) Lee, J.; Simurdiak, M.; Zhao, H. J. Biol. Chem. 2005, 280, 36719.

■

NOTE ADDED AFTER ASAP PUBLICATION Figure 4 was corrected on September 11, 2017.

13046


Convergent Theoretical Prediction of Reactive Oxidant Structures in

Recommend Documents