Letter pubs.acs.org/JPCL
Calculated Mechanism of Cyanobacterial Aldehyde-Deformylating Oxygenase: Asymmetric Aldehyde Activation by a Symmetric Diiron Cofactor Chao Wang,†,‡ Chongyang Zhao,†,‡ Lianrui Hu,†,‡ and Hui Chen*,† †
Beijing National Laboratory for Molecular Sciences (BNLMS), CAS Key Laboratory of Photochemistry, Institute of Chemistry, Chinese Academy of Sciences, Beijing 100190, People’s Republic of China ‡ University of Chinese Academy of Sciences, Beijing 100049, People’s Republic of China S Supporting Information *
ABSTRACT: Cyanobacterial aldehyde-deformylating oxygenase (cADO) is a nonheme diiron enzyme that catalyzes the conversion of aldehyde to alk(a/e)ne, an important transformation in biofuel research. In this work, we report a highly desired computational study for probing the mechanism of cADO. By combining our QM/MM results with the available 57Fe Mössbauer spectroscopic data, the gained detailed structural information suggests construction of asymmetry from the symmetric diiron cofactor in an aldehyde substrate and O2 activation. His160, one of the two iron-coordinate histidine residues in cADO, plays a pivotal role in this asymmetric aldehyde activation process by unprecedented reversible dissociation from the diiron cofactor, a behavior unknown in any other nonheme dinuclear or mononuclear enzymes. The revealed intrinsically asymmetric interactions of the substrate/O2 with the symmetric cofactor in cADO are inspirational for exploring diiron subsite resolution in other nonheme diiron enzymes.
A
Scheme 1. Reaction of cADO-Catalyzed Biosynthesis of Fatty Cn−1 Alk(a/e)ne and the Proposed Mechanism of cADO,13,23 for Which No Direct Structural Evidence Is Available
liphatic hydrocarbons are indispensible chemicals not only as fuels for human society but also as hydrophobic substances and pheromones for organisms.1−4 In nature, aliphatic hydrocarbons represent rare examples of biological products bearing no functional groups, which can be generated as the fatty acid metabolites.5−10 Recently, a two-step alk(a/ e)ne biosynthesis pathway from physiologically abundant Cn fatty acids (n = 14, 16, 18) was identified in cyanobacteria,11 whereby Cn fatty aldehyde is reductively generated first, followed by its conversion to Cn−1 fatty hydrocarbon with formate as the C1-derived coproduct12−19 (Scheme 1). The biological catalyst for this second step of more complicated decarbonylation transformation is a nonheme diiron enzyme called cyanobacterial aldehyde-deformylating oxygenase (cADO).17 Due to this key transformation affording aliphatic hydrocarbons, cADO has drawn increasing attention for its potential use in biosynthesis of renewable, drop-in biofuels.11,20,21 From a structural perspective, cADO belongs to a superfamily of ferritin-like nonheme diiron oxygenases and oxidases, which share a 2-His-4-carboxylate coordination environment of the diiron active center.11,22 However, compared to other oxygenases and oxidases in this enzyme superfamily, cADO is unique in that, despite utilizing O2 oxidant, the aldehydedeformylating reaction catalyzed thereof is redox-neutral (Scheme 1). Hence, elucidating the unusual reaction mechanism of cADO is highly desirable, which is helpful to extend our knowledge for understanding nonheme diiron oxygenases and oxidases. To probe the reaction mechanism of © 2016 American Chemical Society
cADO, to date, only a few transient reaction intermediates have been trapped and characterized by Bollinger, Krebs, Booker, and their co-workers via 57Fe Mössbauer spectroscopy.23 On the basis of the isomer shift (δ) and quadrupole splitting (ΔEQ) Received: September 9, 2016 Accepted: October 24, 2016 Published: October 24, 2016 4427
DOI: 10.1021/acs.jpclett.6b02061 J. Phys. Chem. Lett. 2016, 7, 4427−4432
Letter
The Journal of Physical Chemistry Letters
spectroscopic data,23 we for the first time provide computational structural information helpful for probing the mechanism of cADO. Below, in the presentation of our key discoveries, the irons binding His76/His160 residues are denoted as Fe1/Fe2, respectively, to distinguish two irons. We commenced our modeling with the diferrous A-like structure such as A1 (structure as in Figure 1). The calculated isomer shift and quadrupole splitting data (Table 1) agree well
data of Mössbauer spectroscopic parameters in comparison with previous typical data for compounds of biological interest, 24,25 it was concluded that the key transient intermediate before the first 1e reduction (see green labeling in Scheme 1) has a diferric FeIII/FeIII core, in which both FeIII subsites are at high-spin state (S = 5/2) and coupled into a singlet state (Stotal = 0).23 However, without direct structural evidence, the structure assignment of this important intermediate is still largely tentative and ambiguous. Indeed, this ambiguity has brought about different readings of the trapped transient intermediate in the literature.18,26,27 In addition to this key transient intermediate trapped before the first 1e reduction (green labeling in Scheme 1), detailed structures of the other species trapped in the experiment after the first 1e reduction, are currently also obscure. Apparently, the missing structural information for reaction intermediates in cADO has led to considerable confusion in the community, which thus warrants urgent clarification that otherwise would hamper deeper understanding of the reactivity nature of this enzyme. One appealing and also puzzling structural issue associated specifically with diiron nonheme enzymes is the site differentiation between the two central iron subsites. Different from the mononuclear heme and nonheme enzymes, dinuclear nonheme enzymes generally have an inherent structural issue of equal or unequal iron subsites. Consequently, it is quite difficult to comprehend the differentiation of the two iron subsites when the coordination environment around the diiron core is nearly symmetric. As shown in Figure 1, for cADO, the diiron
Table 1. Computational and Experimental Mössbauer Spectroscopic Parameters of Key Species Explored in This Work calcd dataa
exptl data (ref 23)
species
δ (mm/s)
ΔEQ (mm/s)
δ (mm/s)
ΔEQ (mm/s)
A1 B1 B2 C1 C2 D1 E1 F1 G1 G2 H1 H2 I1 I2 K
1.01/1.04 0.65/0.68 0.72/0.67 0.51/0.58 0.59/0.60 0.52/0.48 0.51/0.51 0.52/0.51 0.53/0.58 0.49/0.53 0.51/0.54 0.55/0.54 0.52/0.53 0.51/0.60 0.15/0.59
2.99/3.24 1.36/1.55 1.65/−1.43 0.41/0.99 −1.15/−1.15 −1.96/1.83 −1.71/−1.55 −1.71/−1.54 1.97/0.35 0.69/−1.04 −1.03/−1.97 −1.89/−1.02 −1.34/−1.39 −1.49/−1.29 −0.62/1.47
∼1.3
∼3.0b
0.48, 0.55
0.49, 1.23
0.51, 0.55
1.08, 1.80b
0.52
−1.25
a
For computational data of each species, data before and after the slash are for Fe1 and Fe2, respectively. bThe reported experimental data are |ΔEQ| with the signs undetermined.
with the experimental ones, lending basic credence to our QM/ MM modelings. Here, we have to point out that we cannot distinguish whether there are water molecules and/or an aldehyde substrate coordinated to iron in the diferrous A-like structure because water-present and water-absent models show very similar Mössbauer spectroscopic parameters (see Figure S3 and Table S3 in the Supporting Information (SI)). In the comparison between calculated and experimental data, it is notable that current DFT calculations for 57Fe Mössbauer spectroscopic parameters have certain errors, and a 0.1 mm/s magnitude standard deviation for the isomer shift is normally expected.28 It was also found to be more challenging to model quadrupole splitting accurately.29−32 Given the above agreement between the calculated and experimental Mössbauer spectroscopic parameters, we proceeded to explore other intermediates. Currently, the most tantalizing structural conundrum to be clarified in cADO is the identity of the transient intermediate trapped in a Mössbauer spectroscopic experiment (δ = 0.48/0.51 mm/s, ΔEQ = 0.49/ 1.23 mm/s).23 As suggested by Bollinger et al., there are two possible assignments (shown in Scheme 1): one is the peroxyhemiacetal C-like species, and the other is the peroxo B-like species.23 To probe the possible structure of the trapped reaction intermediate in a Mössbauer spectroscopic experiment, using saturated C18 fatty aldehyde (one of the natural substrates of cADO) as the substrate, we did extensive QM/MM theoretical modelings combined with Mössbauer spectroscopic parameter calculations.
Figure 1. cADO protein and its nearly symmetric diiron cofactor (Protein Data Bank entry 2OC5;33 all hydrogens are not shown for clarity).
coordination environment appears nearly of C2 symmetry in the absence of O2 and substrate (the symmetry of the diiron cofactor guarantees that two Fe sites bear the same local coordination environment and hence are equal). However, the site differentiation for the trapped transient reaction intermediate of cADO revealed by Mössbauer spectroscopy was found to be quite pronounced.23 Mössbauer spectroscopy can detect site resolution via two quadrupole doublets, but by itself, it can hardly render a one-to-one assignment of its one quadrupole doublet signal to one specific iron subsite. Currently, it is still an open question why and how in cADO this site differentiation occurs during reaction. Obviously, this puzzle and other confusions concerning the mechanism of cADO all fundamentally owe to missing direct structural information and evidence for trapped transient reaction intermediates. Given this status, calculation and modeling represents as a key approach that may shed light on the mechanism of cADO. In this work, by combining our computational data from the quantum mechanical/molecular mechanical (QM/MM) approach with available Mössbauer 4428
DOI: 10.1021/acs.jpclett.6b02061 J. Phys. Chem. Lett. 2016, 7, 4427−4432
Letter
The Journal of Physical Chemistry Letters
Alternatively, is it possible that the peroxo B-like species is actually the transient intermediate trapped in the experiment? To check this possibility, we also did extensive searches for the peroxo structures in cADO. In total, as depicted in Figure 3, we
Concerning the peroxyhemiacetal C-like candidate, after QM/MM optimization, we obtained two peroxyhemiacetal species, denoted as C1 and C2, as depicted in Figure 2. C1 with
Figure 2. Two obtained C-like peroxyhemiacetal species with different bindings of the aldehyde substrate and dioxygen fragment (all hydrogens are not shown for clarity); C1 with dioxygen binding to both irons and the substrate to Fe2; C2 with dioxygen binding to Fe2 and the substrate to Fe1.
Figure 3. Two different peroxo species with different peroxo binding modes (all hydrogens are not shown for clarity); B1: μ−η1:η1-peroxo; B2: μ−η1:η2-peroxo. The distance of Fe2 to O of carbonyl in the aldehyde substrate is labeled.
aldehyde substrate binding to Fe2 is more stable than C2 with aldehyde binding to Fe1 by 5.0 kcal/mol. Importantly, with two key distinguishing structural features near the diiron core, C1 is certainly not the enantiomer of C2. One feature is about the dioxygen fragment, which forms a μ−η2:η0 coordination bridging the two irons in C1 but only coordinates Fe2 in C2. The second notable structural feature is from His160, which in C1 completely dissociates from Fe2, therefore breaking the symmetric environment of the coordinate residues for the diiron core (note that in a structure search, the initial setups of the two His residues for both C1 and C2 are coordinated with the irons). To our best knowledge, this His−Fe dissociative behavior of a coordinate histidine residue has never been found in any other nonheme dinuclear or mononuclear enzymes and opens up a completely new possibility in structural evolution of chemical reactions by nonheme enzymes. All of these computational predictions for structures of cADO await future experimental confirmation. Interestingly, in C1 and C2, when we exchanged the “left−right” (i.e., “Fe1−Fe2”) coordination modes of the dioxygen fragment, aldehyde substrate, and two His residues as geometric perturbations for initial setup, after geometry optimizations, we finally got the unique structures of C2 and C1, respectively. These results strongly imply that the interactions of the substrate/O2 with the diiron cofactor are intrinsically asymmetric. Then, a key question arises, that is, which one of C1 and C2 is more likely to be the transient intermediate trapped in the Mössbauer spectroscopic experiment? As summarized in Table 1, our calculated Mössbauer spectroscopic parameters indicate that C1 reproduces the experimental spectroscopic data much better than C2 does, especially for quadrupole splitting data ΔEQ that is directly related to the geometric coordination environment. Furthermore, C2 does not exhibit subsite differentiation in the calculated Mössbauer spectroscopic parameters. Hence, our above results support C1 rather than C2 as the transient intermediate trapped in a Mössbauer spectroscopic experiment. Furthermore, on the basis of our result, we now know that the Fe1 subsite is responsible for the quadrupole doublet with δ1 = 0.48 mm/s and ΔEQ1 = 0.49 mm/s, while the Fe2 subsite is responsible for the quadrupole doublet with δ2 = 0.55 mm/s and ΔEQ2 = 1.23 mm/s. It is for the first time that a one-to-one assignment for iron subsite resolution in cADO is reported.
obtained two peroxo structures, denoted as B1, B2. In B1, peroxo adopts a cis end-on μ−η1:η1 structure, while μ−η1:η2peroxo exists in B2. B1 is more stable than B2 by 15.1 kcal/mol. Importantly, as shown in Table 1, the calculated Mössbauer spectroscopic parameters (ΔEQ especially) for neither B1 nor B2 match the experimental data for the transient intermediate trapped in the Mössbauer spectroscopic experiment, which demonstrates that B-like peroxo species are not likely to be the trapped transient reaction intermediate in experiment.23 It is notable that in both B1 and B2, the aldehyde substrate does not bind the diiron cofactor. This structural information on peroxo species in cADO is different from the previous assumption.12−18,23 Concerning the transformation from the peroxo Blike species to a peroxyhemiacetal C-like intermediate, our reaction profile calculations demonstrate that instead of more stable B1, it is the less stable peroxo intermediate B2 that can directly transform to C1 by peroxo nucleophilic attack on C O in the aldehyde with a reaction barrier of 8.0 kcal/mol (see Figure S1 in the SI). C1 is slightly more stable than B1 by 1.2 kcal/mol. Considering that B1 has to change to B2 before it can transform to C1 intermediate, the effective barrier from B1 to C1 is 23.1 kcal/mol. After revealing the possible involvement of peroxyhemiacetal intermediate C1 in the reaction mechanism as the trapped diferric peroxide transient intermediate, the second structural conundrum to be clarified in cADO is about the identities of the other species after the first 1e reduction (green labeling in Scheme 1). Experimentally, two species were trapped in the Mössbauer spectra. One is characterized by two quadrupole doublets with δ1 = 0.51 mm/s, |ΔEQ1| = 1.08 mm/s and δ2 = 0.55 mm/s, |ΔEQ2| = 1.80 mm/s, which formed shortly (0.56 s) after reduction of the diferric peroxide transient intermediate.23 The other species is characterized by a single quadrupole doublet with δ = 0.52 mm/s, ΔEQ = −1.25 mm/s, which was formed after a long time (30 min) decay of the diferric peroxide transient intermediate without adding reductant.23 This species notably also has an essentially identical Mössbauer spectrum to that of aerobically isolated cADO, that is, the resting state.23 What are the possible structures of these species? To probe the subsequent structural evolvement resulting from reduction of C1, we scanned the reaction profile along the reaction coordinate of the breaking O−O bond. An 4429
DOI: 10.1021/acs.jpclett.6b02061 J. Phys. Chem. Lett. 2016, 7, 4427−4432
Letter
The Journal of Physical Chemistry Letters intermediate D1 corresponding to D is reached from 1ereduced C1 by overcoming a barrier of 9.9 kcal/mol (note that the calculated absolute energetics of intermediates after 1e reduction, for example, D- and E-like species, are not comparable to the ones before the 1e reduction, like C1). Following the reaction mechanism proposed in Scheme 1, the next C−C cleavage step to generate alkyl radical R• was calculated to have a barrier of 8.1 kcal/mol from D1 to E1. Compared to the substantially larger barrier of 23.1 kcal/mol for formation of C1 from B1, these two relatively low barriers (
