New insight into cofactor-free oxygenation from combined experimental and computational approaches §

Citing this paper Please note that where the full-text provided on King's Research Portal is the Author Accepted Manuscript or Post-Print version this may differ from the final Published version. If citing, it is advised that you check and use the publisher's definitive version for pagination, volume/issue, and date of publication details. And where the final published version is provided on the Research Portal, if citing you are again advised to check the publisher's website for any subsequent corrections.

New insight into cofactor-free oxygenation from combined experimental and computational approaches §

Soi Bui and Roberto A Steiner
Molecular oxygen (O 2 ), in spite being a potentially strong oxidant, typically displays very poor reactivity with organic molecules. This is largely due to quantum chemical reasons as O 2 in its ground state is a diradical ( 3 O 2 ) whilst common organic substrates are in a singlet state. For this reason catalysis involving O 2 as a reactant is typically mediated by enzymes containing redox metal and/or organic co-factors. Cofactorindependent oxygenases (and oxidases) are therefore intriguing enzymes from a fundamental viewpoint. This review looks at recent advances that have been made in understanding of this class of intriguing biocatalysts highlighting the power of an inter-disciplinary approach involving structural biology, spectroscopy and theoretical methods.

Introduction
The incorporation of one or both atoms of molecular oxygen (O 2 ) into organic substrates is catalyzed by monooxygenases and dioxygenases, respectively. These enzymes, collectively known as oxygenases [1], play a key role in the metabolism of aromatic amino acids [2], fatty acids [3,4], sugars [5], and vitamins [6], as well as in the biosynthesis of collagen [7]. Oxygenation reactions are also used for the degradation of various endogenous and exogenous organic compounds [8], whilst some bacteria employ oxygenases to facilitate the breakdown of molecules that are environmental pollutants [9]. Furthermore, oxygenases can catalyze enantiospecific reactions making them attractive for the production of chiral chemicals, although their industrial application is not without practical problems [10].
From a mechanistic perspective, the task of oxygenases is a difficult one because molecular oxygen in its normal 'resting' state (the form present in the air) is a diradical with a triplet ground-state electronic structure ( 3 O 2 ) that is not reactive toward the vast majority of singlet-state organic substrates. To deal with this limitation of quantum chemical nature, oxygenases typically rely on transition metals or redox organic cofactors for catalysis [11,12 ,13]. A remarkable group of oxygenases, however, can catalyze O 2 -incorporation reactions in a cofactorindependent manner [14]. Cofactor-free oxygenases (and oxidases) are therefore intriguing from the viewpoint of fundamental enzymology. In this review we will briefly summarize recent advances in our mechanistic understanding of this fascinating group of enzymes.

Cofactor-free oxygenases
Structural information is available for various cofactorindependent oxygenases including bacterial carbon monoxide-forming 2,4-dioxygenases that degrade quinolones [15 ], the vancomycin biosynthetic DpgC dioxygenase [16,17], Streptomyces coelicolor ActVA-Orf6 monooxygenase involved in polyketide-tailoring [18], bioluminescent coelenterazine luciferases [19][20][21], and RhCC from Rhodococcus jostii RHA1 whose exact biological role is still unknown [22]. Their three-dimensional structures show that cofactor-independent oxygenation is enabled by a variety of folds and quaternary arrangements (Figure 1). Whilst bacterial carbon monoxide-forming 2,4-dioxygenases ( Figure 1a) and Renilla coelenterazine luciferase ( Figure 1j) are monomeric enzymes belonging to the a/ b-hydrolase fold superfamily, DpgC, ActVA-Orf6 and RhCC are assembled into oligomers. Interestingly, Oplophorus luciferase (Figure 1k) that catalyzes the same reaction as Renilla monooxygenase, relies on a completely different architecture that is similar to those of fatty acidbinding proteins (FABs). The oligomeric ActVA-Orf6 (Figure 1g), that oxidizes 6-deoxydihydrokalafungin to dihydrokalafungin in the biosynthesis of the polyketide antibiotic actinorhodin, is a homodimer in which each protomer displays a ferredoxin-like fold whilst DpgC is a hexamer (Figure 1b) formed by individual chains whose C-terminal region that contains the active site shows homology to the crotonase family of enoyl-CoA isomerase/ dehydratases. DpgC is unique as there are no other examples of redox chemistry from this enzyme class. The recently identified RhCC monooxygenase (Figure 1n), is a trimer that belongs to the tautomerase superfamily characterized by a structural b-a-b fold. Like DpgC, RhCC is the first oxygenase identified within this superfamily.     Our current mechanistic understanding of cofactor-independent oxygenation is largely based on studies performed on the bacterial carbon monoxide-forming 1-H-3-hydroxy-4-oxoquinaldine 2,4-dioxygenase (HOD) from Arthrobacter nitroguajacolicus Rü 61a. HOD and its homologous 1-H-3-hydroxy-4-oxoquinoline 2,4-dioxygenase (QDO) from Pseudomonas putida 33/1 were the among first cofactor-free dioxygenases reported in the literature [23]. They are approximately 31-kDa monomeric enzymes that catalyze the O 2 -dependent cleavage of the N-heteroaromatic ring of quinoline derivatives with concomitant formation of carbon monoxide ( Figure 1b). This reaction is chemically identical to that catalyzed by the metal-dependent quercetin dioxygenase [24][25][26]. However, neither HOD nor QDO are related in sequence to the latter enzyme. HOD and QDO share $37% sequence identity and constitute a separate family of cofactor-free dioxygenases of the a/b-hydrolase fold superfamily [27]. A BLAST search reveals a number of bacterial proteins with high sequence similarity to HOD/ QDO that are likely to be involved in similar breakdown reactions ( Figure 2a). Although most of these proteins have not been investigated functionally, a recent report indicates that AqdC1 and AqdC2 from Rhodococcus erythropolis BG43 are competent for the degradation of the Pseudomonas aeruginosa quorum sensing signal molecule 2-heptyl-3-hydroxy-4(1H)-quinolone (PQS, Pseudomonas quinolone signal) [28]. Purified recombinant AqdC1 catalyses the O 2 -dependent cleavage of PQS to form Noctanoylanthranilic acid and CO with an apparent K m and k cat values of 27 mM and 21 s À1 , respectively, thus supporting a 2,4-dioxygenolytic cleavage like that catalyzed by HOD/QDO [28]. Overall, CO-forming bacterial dioxygenases appear to be able to operate on 3-hydroxy-4(1H)-quinolones bearing alkyl chains of different length at substrate position 2 (-R in Figure 1b) albeit with different effectiveness.
The structures of both HOD and QDO have been determined by X-ray crystallography [15 ] (Figure 2b). As expected on the basis of comparative sequence analysis [29] these dioxygenases belong to the a/b-hydrolase fold superfamily. This is a group of enzymes that rely on a nucleophile-histidine-acidic residue triad to hydrolyze typically C-O, C-N, C-C bonds although a large array of reaction types can be catalyzed by members of this versatile family [30 ]. Currently, CO-forming bacterial dioxygenases and the bioluminescent coelenterazine Renilla luciferase monooxygenase are the only oxygenases known to employ the a/b-hydrolase scaffold and its catalytic machinery for O 2 -dependent catalysis. HOD and QDO, like most enzymes of the a/b-hydrolase fold family, display a two-domain structure constituted by a 'core' domain and a 'cap' domain ( Figure 2b). The core domain folds in the canonical a/b hydrolase architecture consisting of a mostly parallel, eight-stranded b sheet surrounded on both sides by a helices (only the second b strand is antiparallel) with the central b sheet featuring a left-handed superhelical 908 twist. The cap domain, formed by four a helices is positioned between b6 and aD. Both domains contribute to delineate the active site cavity. However, only the core domain hosts the nucleophile-histidine-acidic residue catalytic triad, which in the case of HOD is formed by Ser101-His251-Asp126 (Figure 2a,b). A Ser-His-Asp composition is the most common form of the triad [27]. Occasionally, the Ser nucleophile is replaced by a Asp or Cys residue, as in the case of Renilla luciferase where the triad is Asp-His-Asp [19].

Substrate activation
A common step in cofactor-independent oxygenation is the initial deprotonation of the bound organic substrate [14,31]. The H + ion abstracted in various cofactorindependent-catalyzed reactions is highlighted in Figure 1 (blue circle). The structure of HOD in complex with its natural substrate QND (-R = -CH 3 in Figure 1b) obtained under anaerobic conditions reveals that the organic molecule binds in the active site between the core and the cap domains ( Figure 2c) with the substrate's O3 atom at H-bond distance to the Ne2 atom of the triad's H251 (2.6 Å ) [15 ]. This interaction affords the deprotonation of the substrate's hydroxyl group by the His-Asp subset of the triad, essential for substrate activation. Site-directed mutagenesis experiments have shown that replacement of H251 by an alanine (H251A) reduces k cat for QND dioxygenation by several orders of magnitude essentially abolishing catalytic activity [15 ,32,33]. Also, replacement of the  acidic residue of the triad (D126A) yielded a strongly impaired variant. Consistent with this, density functional theory (DFT) calculations show that the 3OH proton is transferred to His251 in the wild-type system, but it remains on the substrate in the HOD H251A and HOD D126A systems [33]. The nucleophile S101 appears to be dispensable for catalysis as S101A replacement only impacted negatively on QND binding as judged by an increase in the K m constant. The idea of a nucleophile-independent mechanism is further supported by the lack of sequence conservation for this residue whilst the histidine-acidic dyad is strictly invariant (Figure 2a). Kinetic analysis under anoxic transient-state conditions indicates that substrate deprotonation is not rate-limiting as suggested by the 15-fold lower k cat (20 8C) compared with k H (5 8C) at low pH [33]. Overall, the His-Asp dyad is essential for substrate deprotonation and its consequent activation toward O 2 .

How does the activated substrate react with O 2 ?
Mechanistic concepts borrowed from the field of flavindependent oxygenation [34,35] led to the plausible hypothesis that the activated (carb)anion possesses a sufficiently low thermodynamic potential that allows the transfer of an electron to O 2 to generate a substrate radical and superoxide anion [32]. This charge-transfer mechanism is shown for HOD catalysis on the left-hand side of Figure 3a. Following formation of the (S -O 2 À ) pair ( 3 R CT in Figure 3a) radical-radical recombination would lead to a C2-(hydro)peroxide ( 1 I 1 ). To address experimentally the nature of the compounds involved in the reaction with O 2 a recent elegant study used the cyclic hydroxylamine spin probe 1-hydroxy-3-methoxycarbonyl-2,2,5,5-tetramethyl-pyrrolidine (CMH) in the HOD-catalyzed reaction [36 ]. Using a HOD variant in which the W160 residue that forms part of the active site cavity (Figure 2c) is substituted by an alanine (HOD W160A ) the authors were able to detect significant amounts of CM-nitroxide radical by electron paramagnetic resonance spectroscopy. This variant also released the proposed peroxide intermediate (Figure 1c), which was reduced to the corresponding alcohol and characterized by NMR spectroscopy. These data were interpreted as evidence for two key intermediates in the catalytic mechanism: a substrate radical and a substrate (hydro)peroxide. However, no significant amounts of CM-nitroxide radical were detected when wild-type HOD was used in the reaction. A possible reason for this is that the wild-type protein is able to efficiently shield its active site, whereas the mutant protein may have a more open or more flexible active-site pocket thus allowing access to the CMH probe [36 ].
A more recent study put forward an alternative hypothesis for the generation of the peroxide intermediate. On the basis of a combined kinetic, spectroscopic and DFT computational work that takes advantage of available structural information Hernandez-Ortega et al. [37 ] proposed a novel reaction mechanism that does not involve a formal single electron transfer to dioxygen. The calculations indicate that the 3 R CT complex is unable to form spontaneously by electron transfer, and, consequently it is expected that any reactivity of O 2 will come from dioxygen and not from the stabilization of a radical pair. This is proposed to occur with C2 atom of the activated substrate ( 1 S À ) reacting directly with 3 O 2 to form a C-OO bond as a triplet-state intermediate peroxide ( 3 I 1 ) (direct attack mechanism in Figure 3a). The latter compound then undergoes an inter-system crossing leading to ( 1 I 1 ). The reaction energy profile for the direct attack mechanism and subsequent catalytic steps is shown in Figure 3b. Following the formation of ( 1 I 1 ), attack of the peroxide on the carbonyl function affords the formation of the endoperoxide ( 1 I 2 ), which then decomposes with the formation of significantly stabilized carbon monoxide and anthranilate derivative products ( 1 P). Overall, the formation of ( 3 I 1 ) encounters the highest barrier along the reaction trajectory (DG z = 17.4 kcal mol À1 in a solvent model), and hence is proposed to be the rate limiting step for HOD catalysis. This in agreement with transient-state kinetic studies that show that oxygen-dependent steps are rate-limiting for overall catalysis. Additionally, kinetic and computational analyses show that the W160A substitution decreases the activity dramatically (k cat / K m is 12,500 AE 600 s À1 mM À1 and 2.4 AE 0.3 s À1 mM À1 for wt HOD and HOD W160A , respectively) with energy profiles for the corresponding DFT model showing increased barriers for 3 TS 1 and 1 TS 3 [37 ]. Thus, the W160A replacement could therefore have an effect on the O 2 activation mechanism.

Peroxide intermediate
The formation of peroxide intermediates is central to the mechanism proposed for essentially all cofactor-independent oxygenases (Figure 1). Yet, their structural elucidation has proven difficult. Urate oxidase (UOX) is a tetrameric cofactor-independent oxidase [14,38] that catalyses the breakdown of urate to 5-hydroxyisourate (5-HIU) with the latter compound further degraded   to allantoin (Figure 4a). On the basis of UV-visible spectroscopic evidence, early mechanistic studies suggested that a peroxide adduct (5-PIU) is generated initially by the dioxygenation of the urate dianion (UD) [39]. In subsequent steps H 2 O 2 release and H 2 O attack at position 5 would form 5-HIU. How UOX exactly contributes to the generation of UD is currently not clear. A mechanism of general base catalysis enabled by a Thr-Lys dyad has been proposed [40] and a neutron diffraction study on the aerobic chloride-inhibited UOX-substrate complex suggested that UD might be generated from the deprotonation of the iminol tautomer of UA [41].
Recently, X-ray crystallography combined with online incrystallo Raman spectroscopy has provided direct evidence for a C5(S)-(hydro)peroxide in the initial cofactor-independent dioxygenation stage of the UOXcatalyzed reaction [42 ]. Using 9-methyl uric acid (MUA) as substrate, electron density maps at (near)atomic resolution unambiguously reveal that exposure of the anaerobic UOX-MUA complex (Figure 4b) to O 2 results in MUA conversion into its C5(S)-peroxo derivative (5-PMUA) with clear pyramidalization at C5 (Figure 4c). Non-resonant Raman spectra recorded from crystals of anaerobic UOX-MUA and UOX-5PMUA complexes show that the spectral region centered at 600 cm À1 reports on changes resulting form the oxygenation reaction (Figure 4e). Upon MUA peroxidation a distinct band develops at 605 cm À1 (blue trace) whilst the shoulder at 597 cm À1 in the UOX:MUA complex (green) disappears. Quantum mechanics/molecular mechanics (QM/MM) calculations predict a band at 600 cm À1 (experimental 605 cm À1 ) for the 5PMUA (hydro)peroxide resulting form a set of modes involving C5-Op1 bond stretching and C5-Op1-Op2 bending coupled to ring distortions.
Remarkably, the C5-Op1 bond is susceptible to selective radiolysis at very low X-ray doses (Figure 4d,f). Rupture of the C5-Op1 bond is accompanied by the loss of pyramidalization at C5 leading to a planar organic structure while a diatomic molecule 1.2-Å long, consistent with O 2 , is liberated and trapped above it (Figure 4d). Dioxygen produced in situ adopts a welldefined position with its molecular axis 3.15 Å above the flat organic molecule and both O1 and O2 atoms interacting with N254(Nd2) at distances of 3.19 Å and 3.09 Å , respectively, whilst the T57(Og1) atom is closest to O2 at 2.65 Å . This is consistent with results obtained using room-temperature O 2 pressurization (40 bars) in the presence of the AZA inhibitor [43]. Online Raman analysis shows that peroxide rupture causes the 605 cm À1 'signature' band to selectively decrease in a dose-dependent manner (Figure 4f) consistent with QM/MM calculations that assign this band to the stretching of the C5-Op1 bond.

Conclusions
Interdisciplinary approaches are central to modern enzymology. Structural biology techniques in combination with steady-state and transient-state kinetics, theoretical quantum chemistry calculations, and advanced solution and in-crystallo spectroscopic measurements have provided important insight into how cofactor-independent oxygenases work. Although not discussed here, the use of molecular dynamics simulations is also uncovering specific pathways and access points for O 2 in this class of oxygenases that are common to different types of proteins [44 ] and often dynamic in nature.
The extensive work carried out on the bacterial cofactor-free 2,4-dioxygeases HOD and studies on various other cofactor-independent oxygenases convincingly suggest that the generation of a substrate (carb)anion is a prerequisite for O 2 reactivity. Thus, these enzymes apparently utilize the intrinsic reactivity of carbanions toward electrophiles as a general catalytic concept [14].
Although the exact mechanism with which O 2 reacts with the activated substrate requires further study, the work of Hernandez-Ortega et al. [37 ] provides a novel perspective on this reaction by proposing a direct attack to generate a triplet-state peroxide that immediately collapses into a triplet-state. It will be interesting in the future to see whether this novel concept can be generalized to other similar catalytic systems. Finally, using a combination of (online and offline) Raman-assisted Xray crystallography and theoretical calculations Bui et al. [42 ] provided unambiguous evidence for a C5-peroxide intermediate in the initial dioxygenation step in UOX cofactor-independent catalysis. Additionally, selective radiolysis afforded exquisite insight into the elusive dioxygen positioning in the ternary E-S-O 2 complex. This experimental information will guide further calculations aimed at addressing O 2 reactivity in this system. The recent advances on cofactor-independent oxygenases discussed here well exemplify the necessity of synergizing experimental and computational approaches in mechanistic enzymology.