Utilization of Dioxygen by Carotenoid Cleavage Oxygenases*♦

Carotenoid cleavage oxygenases (CCOs) are non-heme, Fe(II)-dependent enzymes that participate in biologically important metabolic pathways involving carotenoids and apocarotenoids, including retinoids, stilbenes, and related compounds. CCOs typically catalyze the cleavage of non-aromatic double bonds by dioxygen (O2) to form aldehyde or ketone products. Expressed only in vertebrates, the RPE65 sub-group of CCOs catalyzes a non-canonical reaction consisting of concerted ester cleavage and trans-cis isomerization of all-trans-retinyl esters. It remains unclear whether the former group of CCOs functions as mono- or di-oxygenases. Additionally, a potential role for O2 in catalysis by the RPE65 group of CCOs has not been evaluated to date. Here, we investigated the pattern of oxygen incorporation into apocarotenoid products of Synechocystis apocarotenoid oxygenase. Reactions performed in the presence of 18O-labeled water and 18O2 revealed an unambiguous dioxygenase pattern of O2 incorporation into the reaction products. Substitution of Ala for Thr at position 136 of apocarotenoid oxygenase, a site predicted to govern the mono- versus dioxygenase tendency of CCOs, greatly reduced enzymatic activity without altering the dioxygenase labeling pattern. Reevaluation of the oxygen-labeling pattern of the resveratrol-cleaving CCO, NOV2, previously reported to be a monooxygenase, using a purified enzyme sample revealed that it too is a dioxygenase. We also demonstrated that bovine RPE65 is not dependent on O2 for its cleavage/isomerase activity. In conjunction with prior research, the results of this study resolve key issues regarding the utilization of O2 by CCOs and indicate that dioxygenase activity is a feature common among double bond-cleaving CCOs.

Carotenoid cleavage oxygenases (CCOs) 3 are non-heme iron-dependent enzymes found in all kingdoms of life that participate in the metabolism of carotenoids and related compounds. Since their initial discovery in bacteria and plants, tremendous progress has been made in elucidating the biological substrates of these enzymes (1)(2)(3). Most characterized CCOs catalyze the oxidative cleavage of carotenoid carbon-carbon double bonds to yield aldehyde and/or ketone-containing apocarotenoid products (Fig. 1). Such activity was possibly the earliest to evolve in this family because apocarotenoids, particularly retinal, mediate photoreception by the ancient type 1 opsin protein family, members of which carry out many important physiological processes in phototactic and phototrophic organisms (4,5). Another group of bacterial and fungal members has evolved to cleave carbon double bonds of stilbenes such as resveratrol and lignin-related phenylpropanoids (6,7). The RPE65 class of CCOs, found only in vertebrates, underwent catalytic neofunctionalization during its split from a ␤-carotene oxygenase (BCO)-2-like ancestor to acquire the non-oxidative retinyl ester cleavage/isomerase activity central to the visual cycle that generates 11-cis-retinal required for visual pigment (type II opsin) formation (Fig. 1, A and B) (8,9).
The spin-forbidden reaction of O 2 with alkene substrates of CCOs is made kinetically favorable through its reductive activation by the non-heme iron(II) centers of these enzymes (10). Of central importance to understanding the mechanism of O 2 activation by CCOs and its reaction with substrates is knowledge of whether one or both of the oxygen atoms from O 2 is/are incorporated into the reaction products. In the former case, the non-O 2 -derived oxygen atom found in the reaction is expected to originate from an active site-bound water molecule. A number of studies using isotopically labeled O 2 and H 2 O were carried out to address this question for CCOs, but the issue still remains debatable (11)(12)(13)(14)(15). A factor complicating the interpretation of such experiments is the exchange of oxygen incorporated into the nascent products with that of bulk water during the sample workup and analysis. This makes dioxygenases appear to be monooxygenases (16,17). An initial study on the catalytic mechanism of chicken BCO1 that suggested it was a monooxygenase (11) was criticized for its use of conditions that allowed substantial solvent back-exchange (17). More recent studies of human BCO1 (12) and Arabidopsis CCD1 (14), in which solvent exchange was carefully monitored, demonstrated dioxygenase labeling patterns for these CCOs. By contrast, labeling experiments performed on resveratrol and isoeugenol-cleaving CCOs, NOV2 (13) and isoeugenol monooxygenase (15), from the non-carotenogenic bacteria Novosphingobium aromaticivorans and Pseudomonas nitroreducens, respectively, suggested a monooxygenase pattern of oxygen incorporation for both enzymes. Thus, the labeling studies performed to date seem to indicate that perhaps both monooxygenase and dioxygenase catalytic mechanisms may exist within the CCO enzyme family (18), a possibility previously suggested by a theoretical study of the Synechocystis ACO catalytic mechanism (19).
In contrast to oxidative carotenoid cleavage by CCOs, a basic consideration of the ester cleavage/isomerization chemistry catalyzed by RPE65 does not suggest an obvious role for O 2 in the retinoid isomerization reaction (Fig. 1B). Additionally, substrate and water isotope labeling studies do not indicate direct incorporation of O 2 -derived oxygen into the 11-cis-retinol product of the isomerization reaction (20 -23). However, the close structural similarity of the RPE65 and carotenoid-cleaving CCO Fe(II) centers, together with the evolutionary relatedness of these proteins, raise the possibility that certain features of their catalytic modes could be shared, including a requirement for O 2 (24). Recently, a hypothetical role for O 2 in RPE65catalyzed retinoid isomerization that does not involve its permanent incorporation into the reaction products has been proposed (25).
In this study, we carried out isotope labeling studies on highly purified and active preparations of Synechocystis ACO and Novosphingobium NOV2 with assay protocols that minimized solvent oxygen back-exchange. These experiments unambiguously showed that both enzymes, like human BCO1 and Arabidopsis CCD1, are dioxygenases. A T136A substitution in the ACO sequence that was predicted to open an occluded coordination site for solvent binding, thus potentially favoring monooxygenase chemistry, did not alter the dioxygenase-labeling pattern (19). Using ACO as an internal control, we demonstrated that O 2 is not required for RPE65-mediated conversion of all-trans-retinyl esters into 11-cis-retinol, in agreement with a recently proposed mechanism of retinoid isomerization (26). These data provide compelling evidence for a common dioxygenase mechanism among all double bond-cleaving CCOs.

Materials and Methods
Phylogeny Inference-The CCO amino acid sequences of interest were retrieved from the NCBI protein database. Atomic coordinates for CCOs of known structure, Synechocystis ACO, bovine RPE65, and Zea mays VP14 (PDB accession  Synechocystis ACO, a prototypical CCO member, specifically cleaves all-trans-8Ј-apocarotenol at the 15,15Ј-double bond position to generate all-transretinal (RAL) and 8Ј-hydroxy-15Ј-apocarotenal (C10-apocarotenal). RPE65, an atypical CCO member, catalyzes a coupled ester cleavage/isomerization of all-trans-retinyl esters (atRE) rather than oxidative carotenoid cleavage. This reaction produces 11-cis-retinol (11cROL), a key intermediate in the regeneration of visual chromophore required for vertebrate vision. NOV2 from Novosphingobium cleaves resveratrol at the interphenyl double bond to produce 3,5-DHBA and 4-HBA. Wavy red lines indicate the location of the scissile bond in each reaction. codes, 4OU9, 4RSC, and 3NPE, respectively) were obtained from the Protein Data Bank. A structure-based sequence alignment was first generated using the program MUSTANG (27). This structure-based alignment was then used as a profile for sequence-based alignment of the remaining proteins using MUSCLE (28). A standard sequence-based alignment of the proteins was also generated with MUSCLE to assess the influence of the alignment methodology on the inferred tree topology. Columns containing gaps were removed from the alignment. The CCO phylogeny was inferred based on the maximum likelihood optimality criterion as implemented in PhyML (29). The LG substitution matrix with an estimation of the ␥-parameter (four categories) and invariant sites resulted in the highest log-likelihood tree as assessed by ProtTest (30). Improvements to the starting neighbor-joining tree were carried out by both nearest neighbor interchange and surface pruning and regrafting rearrangements. Both tree topology and branch lengths were optimized. Tree robustness was assessed by analysis of 1000 bootstrap pseudo-replicates. The presented maximum likelihood tree had a log likelihood value of Ϫ14321.6 ( Fig. 1A). An identical tree topology was obtained when the procedure was repeated with the pure sequencebased alignment.
Protein Expression and Purification-Synechocystis ACO was expressed and purified as described previously (31). The ACO T136A point mutant was generated with a QuikChange sitedirected mutagenesis kit (Stratagene, Santa Clara, CA) and confirmed by DNA sequencing. T136A-ACO was expressed and purified identically to the wild-type protein. The coding sequence of NOV2 from N. aromaticivorans DSM 12444 (GI: 499765715) was synthesized and cloned into the pET3a expression vector (Novagen) without any fusion tags. The integrity of this expression plasmid was confirmed by sequencing. The plasmid was transformed into the T7 express BL21 Escherichia coli strain (New England Biolabs, Ipswich, MA) for protein expression studies. One-liter cultures containing 100 g of ampicillin/ml of LB media were grown at 37°C to an A 600 nm of 0.5-0.8 when the temperature was lowered to 28°C and additional ampicillin (100 g/ml) was added. Protein expression relied on leaky T7 promoter activity. After overnight growth (12-16 h), cells were harvested by centrifugation and stored at Ϫ80°C. NOV2 was purified in the same manner as ACO (31) by ammonium sulfate fractionation and gel filtration chromatography. Purified protein samples (Ն95% pure as judged by SDS-PAGE analysis) were flash-frozen and stored at Ϫ80°C or placed on ice for immediate use.
Enzymatic Assays and Rate Determination for Background Oxygen Exchange-ACO activity studies were conducted according to previously established methods (31). NOV2 enzymatic activity was assayed with resveratrol as a substrate in a 20 mM HEPES-NaOH, pH 7.0. NOV2 steady-state kinetics were assessed by monitoring the reduction of resveratrol absorbance at 304 nm over time using a Flexstation 3 plate reader device (Molecular Devices, Sunnyvale, CA). Reactions were run at 28°C in 200 l of 10 mM BisTris-HCl, pH 7.0, containing 0.8 g of purified NOV2 and resveratrol (Sigma, 99% purity) at eight concentrations ranging from 62.5 to 0.49 M separated by 2-fold dilutions. Absorbance measurements were taken every 15 s for a total of 10 min. Initial reaction velocities were determined from the linear portions of the reaction profiles. Absorbance was related to absolute concentrations by simultaneous absorbance recordings of resveratrol standards. For NOV2 oxygen labeling studies, resveratrol was delivered in DMSO to a HEPES-NaOH, pH 7.0, buffer system containing purified NOV2, and the reaction was carried out at 28°C with 500 rpm shaking in a Thermomixer (Eppendorf, Hauppauge, NY), as described in detail below. Following the incubation, products were directly extracted by ethyl acetate and analyzed by HPLC (see below). Product amounts were quantified by plotting peak areas of known quantities of standards. To determine the background oxygen exchange for ACO-and NOV2-catalyzed reaction products, the exchange reactions were performed in H 2 18 O. For ACO reaction products, the reaction mixture consisted of 192 l of H 2 18 O (99% 18 O atom, Sigma), 4 l of 1 M HEPES-NaOH, pH 7.0, 1 l of 10% (w/v) Triton X-100, and 3 l of 1 g/l freshly purified ACO. Then either authentic alltrans-retinal (TRC, Toronto, Canada, Ͼ 95% purity) or HPLCpurified 8Ј-hydroxy-15Ј-apocarotenal generated in-house by a large scale ACO-catalyzed reaction were added to the reaction mixture. The reaction mixture then was incubated at 28°C with 500 rpm shaking in a Thermomixer. Samples (20 l) were removed from the mixture at 0, 1, 2, 5, 15, 20, and 30 min and placed into tubes containing 200 l of hexane followed by vigorous shaking. Then the extracted mixtures were centrifuged in a bench top centrifuge at 15,000 rpm for 2 min. Organic phases were collected and analyzed by LC-MS (see below). Two additional experiments were conducted in the same manner but with 1% (w/v) BSA and/or 5% (v/v) glacial acetic acid to evaluate protein or pH-facilitated oxygen exchange. Background oxygen exchange for the NOV2 reaction products was determined in a similar manner. Briefly, the reaction mixture contained 150 l of H 2 18 O (97% 18 O atom, Sigma), 3 l of 1 M HEPES-NaOH, pH 7.0, and 2.3 l of freshly purified NOV2 at 4.4 g/l. Authentic 4-hydroxybenzaldehyde (4-HBA) or 3,5-dihydroxybenzaldehyde (3,5-DHBA) (Sigma, Ͼ95% purity for both compounds) dissolved in DMSO was added to the reaction mixture. Samples (20 l) were removed from the mixture at 0, 2, 5, 10, and 15 min, placed into tubes containing 200 l of ethyl acetate, dried under argon, and then redissolved in acetonitrile for MS analysis. As with the apocarotenoid product experiment, reactions containing 1% (w/v) BSA or 5% (v/v) glacial acetic acid were used as controls.
Isotope Labeling Study in H 2 18 O-For the labeling study in H 2 18 O, cleavage reactions were performed in the same manner as described above. Specifically, 3 l of purified ACO at 1 g/l were added to a reaction mixture containing 192 l of H 2 18 O (99% 18 O atom, Sigma), 4 l of 1 M HEPES-NaOH, pH 7.0, and 1 l of 10% (w/v) Triton X-100. The cleavage reaction was initiated by adding 5 l of 4 mM all-trans-8Ј-apocarotenol dissolved in ethanol. The reaction mixture was incubated in a Thermomixer at 28°C with 500 rpm shaking for 5 min. Then the reaction was quenched with 200 l of methanol, and 400 l of diethyl ether/hexanes (4:1, v/v) was added to extract the cleavage products. For the NOV2-catalyzed reaction, 2 l of 25 mM resveratrol in DMSO was added to a 200-l reaction mixture consisting of 194 l of H 2 18 O (97% 18 O atom, Sigma), 4 l of 1 M HEPES-NaOH, pH 7.0, and 2.3 l of 4.4 g/l purified NOV2, and this reaction was allowed to proceed for 5 min at 28°C with 500 rpm shaking. The 4-HBA and 3,5-DHBA products of the reaction were then extracted with 500 l of ethyl acetate and immediately purified by normal phase HPLC as described below. Peaks corresponding to each cleavage product were collected in glass tubes, dried under argon, and dissolved in acetonitrile. Both NOV2 products were separated by HPLC (see below). ACO and NOV2 reaction products were further analyzed by mass spectrometry (MS) as outlined below. Sample Deoxygenation and Isotope Labeling Study in 18 O 2 -Sample deoxygenation was performed in 3-ml screw-capped glass vials with a gas-tight Teflon septum. A vial containing 1 ml of 20 mM HEPES-NaOH, pH 7.0, was flushed with ultrapure argon (Airgas, Cleveland, OH) with constant stirring. This deoxygenation treatment was carried out for different times to evaluate its efficiency. Oxygen supplementation was accomplished by flushing the argon-purged buffer solution with O 2 (Airgas) for 3 min. To test ACO activity in argon-purged or oxygen-supplemented buffer, 5 l of Triton X-100 (10% w/v) and 7.5 l of freshly purified ACO at 2 g/l was injected into the treated reaction solutions with an airtight syringe (Hamilton, Reno, NV). The reaction was initiated by injecting 10 l of 4 mM all-trans-8Ј-apocarotenol in ethanol and allowed to proceed for 3 min at room temperature with 800 rpm shaking. Then 1.5 ml of methanol was injected to quench the reaction followed by addition of 1 ml of diethyl ether/hexanes (4:1, v/v) to extract the products. The organic phase was collected for LC/MS analysis. For the cleavage reaction performed in an 18 O 2 atmosphere, 30 g of purified ACO and 25 l of substrate at 4 mM in ethanol were used, and 18 O 2 was introduced into the argon-purged reaction vial. The reaction was carried out for 10 min, followed by product extraction and LC-MS analysis. Both the deoxygenation and NOV2-catalyzed reaction in the presence of 18 O 2 were conducted in the same way as for ACO. Following a 15-min argon purge of the reaction solution, 5 l of 20 mg/ml NOV2 and 10 l of 25 mM resveratrol in DMSO were injected to initiate the reaction, which was carried out for 3 min under an 18 O 2 atmosphere. Reaction products were extracted with 500 l of ethyl acetate and immediately dried by argon gas to remove the organic solvent and co-extracted water. The dried sample then was redissolved in ethyl acetate, purified by HPLC, and further analyzed by MS.
HPLC and Mass Spectrometric Analyses of Cleavage Products-HPLC-MS analyses of the ACO-catalyzed cleavage products of all-trans-8Ј-apocarotenol were performed according to previously established methods (31). Products of NOV2catalyzed resveratrol cleavage were extracted with ethyl acetate, dried under argon, redissolved in ethyl acetate, and injected onto a normal phase Zorbax Sil column (5 m, 4.6 ϫ 250 mm) (Agilent, Santa Clara, CA). Separation of the reaction products was achieved with hexanes/ethyl acetate (3:2, v/v) as the mobile phase at a flow rate of 1.4 ml/min. Peaks eluted at ϳ4.2 min (4-HBA) and ϳ5.3 min (3,5-DHBA) were individually collected, dried under argon, and dissolved in acetonitrile. Purified products from the isotope labeling experiments were directly injected into the APCI source of an LXQ linear ion trap mass spectrometer (Thermo Scientific, Waltham, MA) with acetonitrile and 10 mM ammonium formate. MS data were analyzed with the Xcalibur 2.0.7 software package. Incorporation of 18 O was quantified by comparing relative ion intensities within the isotopic distribution window.
RPE Microsome Deoxygenation and Isomerization Activity Assays-Bovine RPE microsomal membranes enriched in RPE65 were prepared as described previously (32). The concentration of RPE65 in microsomal preparations was determined by quantitative densitometry on SDS-polyacrylamide gels. Deoxygenation of RPE microsomes was carried out in the same manner as for ACO and NOV2. Briefly, 0.5 ml of RPE microsomes in a 3-ml screw-capped glass vial with a gas-tight Teflon septum was purged with ultrapure argon. Afterward, 2.5 l of 20 M all-trans-retinol in N,N-dimethylformamide was injected into the vial, and the reaction and product analyses were performed as described previously (32). O 2 supplementation was accomplished by flushing the sample with O 2 gas for 3 min before adding the substrate. The efficiency of O 2 depletion/ supplementation of RPE microsomes was assessed by monitoring the progress of ACO-catalyzed all-trans-8Ј-apocarotenol cleavage in the microsomal reaction mixture. Specifically, 7.5 l of purified ACO at a concentration of 1 mg/ml were injected into the RPE65 reaction mixture after O 2 depletion/supplementation followed by injection of the apocarotenoid substrate. Triton X-100 was omitted from these control experiments as it interferes with RPE65 activity. The reaction and analysis were carried out as indicated above.
Protein Crystallization, Structural Determination, and Analysis-Crystallization of T136A-ACO was conducted as described previously for the wild-type protein (31). Briefly, 1.5 l of purified enzyme at 10 mg/ml in 20 mM HEPES-NaOH, pH 7.0, containing 0.02% (w/v) Triton X-100 was mixed with a reservoir mixture containing 0.1 M BisTris propane-HCl, pH 6.0, 21-23% (w/v) sodium polyacrylate 2100, and 0.2 M NaCl in a 1:1 ratio. Crystallization was carried out by the hanging-drop vapor-diffusion method at 8°C. Rod-shaped crystals typically appeared within 2 weeks. Mature crystals were directly harvested and flash-cooled in liquid nitrogen before x-ray exposure. Diffraction data were collected at the NE-CAT 24-ID-E beamline of the Advanced Proton Source. The diffraction data were indexed, integrated, and scaled with the XDS package (33). Mutant T136A-ACO crystals were isomorphous to previously reported orthorhombic wild-type ACO crystals, and their structures were determined by rigid body refinement in REF-MAC5 (34) using PDB entry 4OU9 as the starting model. Manual adjustments to the structure were made with COOT (35), and restrained refinement was carried out in REFMAC5. Structures were validated using MOLPROBITY (36) and the wwPDB structure validation server (37). A summary of the x-ray data and refinement statistics is shown in Table 1. All structural figures were prepared with PyMOL (Schrödinger, New York, NY).

Results
Assessment of Apocarotenoid Solvent Back-exchange-ACOcatalyzed cleavage of 8Ј-apocarotenol generates two aldehyde products, all-trans-retinal (RAL) and a second C10-apocarotenal molecule (Fig. 1B). To quantify the solvent exchange rates of DECEMBER 18, 2015 • VOLUME 290 • NUMBER 51 both aldehyde groups, RAL and the C10-apocarotenoid were incubated in a 200-l ACO reaction mixture containing 3 g of ACO with 18 O-water (99% 18 O atom) used as the reaction solvent. The incubation was quenched, and the apocarotenoids were extracted with hexane at different time intervals (0 -30 min). The exchange rate was determined by quantifying the ratio of target compound and its 18 O-labeled counterpart that was generated by the oxygen exchange. Virtually no exchange was observed within the time course of this experiment as found previously (Fig. 2, A and B) (38). However, the exchange rates for both products were significantly increased by supplementing the reaction mixture with 1% (w/v) BSA or by lowering the pH with 5% (v/v) acetic acid, both of which can catalyze the oxygen exchange reaction (Fig. 2, A  and B). For both apocarotenoids, about 80% of the carbonyl 16 Table 2.

H 2
18 O using 1 mg of ACO (0.5% w/v protein). Here, both products showed a remarkable increase in their m/z ϩ 2 counterparts, suggesting that 16 O atoms in the aldehyde products were replaced by 18 O from water ( Table 2). These data strongly suggest that oxygen atoms in the RAL and C10-apocarotenal aldehyde groups originated from O 2 rather than water.
Apocarotenoid Labeling Studies in the Presence of 18 O 2 -To further confirm the origin of the aldehyde oxygen atoms, additional labeling studies were carried out in an 18 O 2 -enriched environment. An efficient method for removal of atmospheric O 2 in the reaction solutions was first developed. This procedure efficiently displaced dissolved O 2 in the reaction mixture within 10 -15 min, as shown by a drastic reduction in ACO-catalyzed RAL formation (Fig. 3B). Importantly, this RAL reduction could be restored by introducing O 2 back into the solution (Fig.  3B). In fact, O 2 supplementation caused an ϳ2.5-fold enhancement in ACO activity attributable to the higher concentration of dissolved O 2 in the reaction solutions.
Labeling studies in an 18 O 2 -enriched environment were performed by forcing 18 O 2 back into the reaction mixture following the deoxygenation treatment. In this case, the isotopic distributions of both reaction products were shifted by 2 Da (Fig. 2,  G and H), indicating the formation of 18 O-labeled products. Only small amounts of 16 O-labeled RAL and C10-apocarotenal (Ͻ10%) were detected under these conditions, likely generated by residual 16 O-oxygen in the reaction mixture and low levels of background solvent exchange (Table 2). Taken together, these labeling studies in 18 O-water and 18 O 2 environments unequivocally demonstrate that ACO is a dioxygenase.
Preparation of Highly Purified and Active NOV2-NOV2 is a stilbene oxygenase member of the CCO family that, among other substrates, cleaves resveratrol to form 4-HBA and 3,5-DHBA (Fig. 1B). Previously, this CCO was characterized as a monooxygenase, indicating the possibility that different groups of CCOs could adopt different oxygenation mechanisms (13). Thus, examination of the catalytic properties of NOV2 activity could reveal the determinants of monoversus dioxygenase activity in CCOs. The previous oxygen labeling study employed crude cell lysates from NOV2-expressing E. coli due to a substantial loss of activity after NOV2 was subjected to purification procedures. Excess E. coli protein in these labeling experiments could have significantly promoted solvent back-exchange in the aldehyde products, which would obfuscate the monoversus dioxygenase assignment. To circumvent this limitation, we developed an expression and purification method for NOV2, based upon the protocol used for ACO that enabled the production of a highly pure and active protein sample (Fig. 4, A and  B). NOV2 expressed under reduced temperatures via leaky T7 promoter activity accumulated to very high levels in the cytosolic fraction of E. coli. A monodisperse and Ͼ95% pure preparation of NOV2 was obtained by a simple purification scheme consisting of ammonium sulfate fractionation and gel filtration chromatography (Fig. 4, A and B). With this homogeneous enzyme sample, we carried out a steady-state kinetic study of NOV2 that revealed a k cat value similar to that of ACO (Fig. 4C) (31,39). Notably, NOV2 purified in this manner retained its activity for several weeks when stored on ice or at Ϫ80°C.
Assessment of the Solvent Back-exchange Rate for 4-HBA and 3,5-DHBA-To measure solvent back-exchange in the benzaldehyde products of NOV2-catalyzed resveratrol cleavage, authentic 4-HBA and 3,5-DHBA were incubated in a 150-l H 2 18 O reaction mixture in the presence of 10 g of purified NOV2 for different periods of time. The aldehyde oxygen of 4-HBA remained unchanged after 15 min of incubation (Fig.  5A). By contrast, the carbonyl oxygen in 3,5-DHBA was rather susceptible to solvent exchange with ϳ20% of oxygen replaced  by water-derived 18 O during a 2-min incubation (Fig. 5B). Addition of 1% (w/v) BSA into the reaction mixture significantly promoted the exchange for 3,5-DHBA but not for 4-HBA. Lowering the pH also drastically accelerated the oxygen substitution for both products with nearly 80% of the original oxygen atoms replaced by those from water (Fig. 5, A and  B). These results suggested that the oxygen back-exchange would not significantly cloud the interpretation of product oxygen labeling in the NOV2-catalyzed reaction. showed two peaks corresponding to 4-HBA and 3,5-DHBA. The identity of each product was confirmed by their absorbance spectra (Fig. 6A). Subsequent MS analysis of the products from H 2 16 O/ 16 (Fig. 5, E and F, compared with C and D) with a slightly higher content of the m/z ϭ 139 species (ϳ10%) for 3,5-DHBA that was attributable to solvent back-exchange. These labeling results suggested that the carbonyl oxygen atoms in the final products originated from molecular oxygen rather than water.

NOV2 Labeling Studies in the Presence of H
NOV2 Labeling Studies in the Presence of 18 O 2 -18 O 2 labeling experiments for NOV2 were carried out in deoxygenated buffer in the same manner as described above for ACO. Removal of atmospheric O 2 resulted in a dramatic reduction in 3,5-DHBA production by NOV2 that could be restored by reintroduction of O 2 gas (Fig. 6B). MS analysis of products generated in the 18 O 2 atmosphere during the 1-min reaction revealed a 2-dalton shift for the 4-HBA molecule compared with the mass observed from reactions carried out in standard O 2 (Fig. 5G). Similarly, the isotopic distribution patterns revealed 78.1% of 3,5-DHBA molecules become 18 O-labeled (Fig. 5H). Because 3,5-DHBA has a readily exchangeable aldehyde oxygen, it is likely that this unlabeled 3,5-DHBA species (21.9%) was generated by solvent back-exchange. Extension of the assay incubation time to 2 min only slightly reduced the percentage of 3,5-DHBA carrying an 18 O label (Table 3). However, when the sample was dried in a SpeedVac under heat (ϳ15 min) rather than under a stream of argon (ϳ2 min), the unlabeled 3,5-DHBA species substantially increased to 35% (Table 3), likely due to the elevated temperature and increased time of exposure to water. Therefore, this indicated that seemingly minor differences in sample preparation can have a significant impact on the experimental outcome. Taken together, the high incorporation rate of 18  Activity and Structure of T136A-ACO-A density functional theory (DFT) study of the ACO reaction mechanism found that Thr-136 could promote a dioxygenase-type mechanism by limiting access of water to the iron center (19). The non-heme iron center of wild-type ACO adopts an octahedron-like structure with four coordination sites occupied by His ligands and a fifth, trans to His-183, occupied by solvent. The sixth site is occluded by the methyl group of Thr-136, positioned ϳ4.6 Å away from the Fe(II) atom. This arrangement creates a hydrophobic microenvironment around the sixth coordination site that likely impedes its accessibility to water (40). DFT calculations indicated that the coordinated solvent is displaced by O 2 during catalysis, which could potentially bind in a side-on or end-on fashion. A structure-based sequence alignment revealed that most CCOs contain a hydrophobic or semi-hydrophobic residue (Thr, Val, or Ile) at this position (Fig. 7, A and B). A reduction in the bulkiness at this site could create enough space for water to coordinate iron and allow it to participate in the reaction. In fact, VP14, one of the few CCOs possessing a non-bulky Ala residue at this position (Fig. 7A), was demonstrated to bind water at this site (41). If the ACO reaction occurs through an epoxide intermediate, an iron-coordinated solvent molecule could act as a nucleophile to open the epoxide, which would give rise to a monooxygenase labeling pattern.
To test these hypotheses, we generated a T136A version of ACO, in which the sixth coordination site was expected to be water-accessible, and we examined its catalytic activity and labeling pattern. Cleavage of 8Ј-apocarotenol by this mutant enzyme was severely impaired with minimal product formation observed using the standard reaction protocol. The poor activity could nevertheless be overcome by increasing the amount of enzyme used in the assay. In reactions containing 100 g of T136A-ACO, RAL formation after a 10-min incubation was ϳ50% that formed by wild-type protein over the same period of time (Fig. 8A). Using these modified conditions, we carried out a labeling study using T136A-ACO in the presence of H 2 18 O.
We observed that the RAL and C10-apocarotenal products generated by T136A-ACO were labeled primarily with 16 O, similar to those generated by the wild-type enzyme (Fig. 8, B and C compared with D and E). In both cases, the reaction products showed increased 18 O labeling compared with those produced under standard assay conditions by wild-type protein. This elevation is attributable to an increase in nonspecific protein-catalyzed oxygen exchange caused by the large quantities of enzyme used in these experiments. A crystal structure of T136A ACO confirmed that this substitution could theoretically allow for the binding of water to the coordination site trans to His-304. The ϳ6-Å gap separating the Ala-136 methyl side chain and the iron center could easily accommodate an iron-bound solvent molecule with minimal steric hindrance (Fig. 7C). However, rather than a 6-coordinate octahedral geometry with water molecules occupying the vacant sites, we instead observed a 5-coordinate trigonal bipyramidal structure with a single water molecule bound to the iron. Additionally, there was a slight but consistent increase (average 0.26 Å) in the Fe-His bond lengths compared with the wild-type protein (Fig. 7, B and C). Together, these data demonstrate that the Thr-136 side chain is not a factor dictating the dioxygenase-labeling pattern of ACO; however, it does appear to be critical for overall catalytic activity and structural integrity of the iron center.
Influence of O 2 on RPE65 Retinoid Isomerase Activity-RPE65 is an atypical CCO that catalyzes ester cleavage and isomerization of all-trans-retinyl esters instead of oxidative cleavage of carotenoid substrates (Fig. 1B) (42). Although not necessarily expected based on the chemistry being performed, the question of whether O 2 is required for this important reaction has not been experimentally evaluated. Exploration of this possibility is certainly warranted given the close phyletic relationship of RPE65 to the oxygen-utilizing BCO1 and BCO2 enzymes (Fig. 1A), as well as the presumed structural similarity of their iron centers. To address this question, we compared RPE65 isomerase activity in native RPE microsomes before and after sample deoxygenation. To evaluate the extent of O 2 removal from the microsome-containing sample we measured, in parallel, the apocarotenoid oxygenase activity of ACO that was exogenously added to the RPE65 reaction mixture, serving as an internal control. Importantly, ACO activity in the RPE65 reaction system was comparable with that found under standard reaction conditions (Fig. 9A). Similar to what was observed under standard conditions (Fig. 3B), ACO activity was markedly impaired following argon treatment to remove O 2 with a 3.4-fold reduction in turnover number. The activity was fully rescued and in fact augmented by reintroduction of O 2 into the reaction system (Fig. 9B). ACO thus served as a reliable indicator of the oxygen concentration in the RPE65 reaction system. When RPE65 activity was measured following the same deoxygenation treatment, the amount of 11-cis-retinol formed was only marginally decreased (ϳ14%), and the reduced activity could not be restored by reintroduction of O 2 into the reaction mixture (Fig. 9B). Moreover, O 2 supplementation of the reaction mixture without prior deoxygenation actually depressed the activity by ϳ30% in contrast to the ϳ34% activity boost seen for ACO (Fig. 9B). This loss of RPE65 activity in the presence of excess O 2 could be caused by oxidation of the iron cofactor,  Table 3.
which is required to be its ferrous form to be catalytically competent (43). Importantly, the total number of substrate turnover events was 3.6 times higher for RPE65 compared with ACO, which rules out the possibility that a reduced O 2 requirement by RPE65, due to less overall catalytic activity, could explain its insensitivity to O 2 concentration (Table 4). Collectively, these results support an O 2 -independent mechanism for RPE65-catalyzed retinoid isomerization.

Discussion
The question of whether CCOs are monooxygenases or dioxygenases, although seemingly straightforward to determine experimentally, has remained contentious despite several published studies addressing the subject. Major difficulties and problems associated with these studies include their use of crude cell extracts as a source of enzymatic activity, the low activity of CCO enzymes in general, inclusion of only one of the two cleavage reaction products for analyses of isotopic labeling, and most importantly, solvent back-exchange of the aldehyde cleavage products facilitated by both high protein concentrations and long incubation times. Taken at face value, discrepancies between these studies might also indicate that both monooxygenases and dioxygenases could exist within the CCO family. With these issues in mind, we set out to examine the oxygen-labeling pattern of the well characterized prototypical FIGURE 6. HPLC analysis of NOV2-catalyzed reaction products and the influence of O 2 depletion and supplementation on NOV2 enzymatic activity. A, HPLC trace of NOV2-catalyzed reaction products generated from cleavage of resveratrol. The peak assignments were made based on their elution times, in comparison with authentic standards, and their characteristic absorbance spectra are shown to the right of the HPLC. Absorbance maxima for resveratrol, 4-HBA, and 3,5-DHBA were 304, 271, and 267 nm, respectively. mAU, milliabsorbance unit. B, NOV2 activity assays were performed either without argon pretreatment to displace O 2 (control) or after a 10-or 15-min argon purge (white bars) and compared with NOV2 samples undergoing the same treatments but subjected to 3 min of O 2 gas supplementation as a final step (gray bars). Error bars represent S.D.s calculated from measurements performed in triplicate.  (45,46). The position in the sequences homologous to Thr-136 of Synechocystis ACO (Novosphingobium NOV2, bovine RPE65, maize VP14, and human BCO1) is marked with a blue star. A conserved Glu residue that participates in second shell iron coordination is marked by a blue square. B, structure of the wild-type ACO Fe(II)-center showing a 5-coordinate partially filled octahedral geometry. C, structure of the T136A-ACO Fe(II)-center that has a structure more consistent with trigonal bipyramidal geometry because of a change in the binding position of the coordinated solvent. Notably, no extra electron density that would indicate the presence of a coordinated solvent molecule between Ala-136 and iron was observed. Blue mesh represents 2mF o Ϫ DF c electron density contoured at one root mean square deviation and computed without inclusion of the iron-bound solvent molecules in the structural models. Bond lengths (in Å) between Fe(II) and nitrogen atoms that form the 4-His coordination shell are shown. Iron and the iron-bound water molecules are depicted as orange and red spheres, respectively. B and C are walleye stereoviews. (65.9%) b a Reaction products were dried by argon stream after a 2-min reaction. b Reaction products were dried by SpeedVac method after a 2-min reaction.
CCO Synechocystis ACO as well as a distantly related stilbenecleaving bacterial CCO called NOV2 that was previously classified as a monooxygenase. Capitalizing on our recent success in generating a highly purified native ACO with robust and durable enzymatic activity, we found that this enzyme exhibits a dioxygenase pattern of O 2 incorporation into its reaction products. This result is consistent with a previous computational study on the ACO reaction mechanism, which favored a dioxetane-based mechanism that would cause both oxygen atoms of O 2 to appear in the reaction products (19). A key prediction from this theoretical study was that the ability of a vacant site in the iron coordination sphere to bind water could govern whether the reaction occurs through a dioxetane or an epoxide intermediate, which could give rise to either dioxygenase-or monooxygenase-labeling patterns in the reaction products, respectively. This site, located trans to His-304, is blocked in ACO by the hydrophobic methyl group of Thr-136. We replaced Ala for Thr at this position to potentially remove the steric and electrostatic barriers to solvent binding at the sixth coordination site. Notably, VP14 has a naturally occurring Ala residue at the corresponding position in its sequence, and the reported crystal structure contains bound solvent at the coordination site in question (41). Despite a major reduction in catalytic activity caused by the T136A substitution, the dioxygenase-labeling pattern for the enzyme was maintained. The crystal structure of T136A-ACO showed that the iron center remained five-coordinate just like the native enzyme even though the coordination geometry adopted a trigonal bipyramidal structure due to a shift in the coordinated solvent molecule. Thus, the dioxygenase activity of ACO is resistant to changes in solvent coordination potential of its iron center. Given the reported ability of the VP14 iron center to accept water in its sixth coordination site, it would be of interest to examine the oxygen-labeling pattern of this enzyme which, despite some claims to the contrary, has not yet been properly investigated.
Our labeling experiments with the stilbene-cleaving CCO, NOV2, previously described as a monooxygenase (13), clearly demonstrated a dioxygenase pattern of oxygen incorporation into its benzaldehyde reaction products. What are the causes for this discrepancy in experimental outcomes? Quantification of oxygen labeling in the prior study indicated a substantial FIGURE 8. Activity and in vitro isotope-labeling analysis of T136A-ACO. A, activity results of T136A and wild-type ACO. 100 g of purified protein were used in each reaction. Enzyme activity was assessed by following the formation of RAL over time. Note that the substrate is completely consumed by the wild-type protein within the 1st min of the reaction; therefore, product formation in this graph cannot be used to compare the activity levels of wild-type and T136A ACO. Error bars represent S.D.s calculated from triplicate measurements. B and C, mass spectra of apocarotenoid products generated by wild-type ACO in an H 2 18 O/ 16 O 2 environment. Note the increased levels of 18 O-labeled RAL (m/z ϭ 287.2) and C10-apocarotenal (m/z ϭ 169.1) generated as a result of the high protein concentration used in the assay. D and E, mass spectra of apocarotenoid products generated by T136A-ACO in an H 2 18 O/ 16 O 2 environment were highly similar to those in B and C, which indicates preservation of dioxygenase activity in the ACO point mutant.   18 O in each product. Importantly, because the samples were not deoxygenated prior to initiation of the reaction, a significant amount of product could have been labeled with atmospheric 16 O 2 thus lowering the incorporation of 18 O 2 into the products. However, this explanation for the labeling results is in conflict with the labeling pattern observed in the same study from reactions performed in H 2 18 O/ 16 O 2 in which very little (ϳ10%) 4-HBA was labeled with 18 O. These prior experiments employed crude NOV2-containing cell lysate as an enzyme source, and contaminating proteins could have exacerbated solvent back-exchange. The reactions in this study were also carried out for 15 min, a period of time that, under our reaction conditions, allowed substantial solvent back-exchange to occur for 3,5-DHBA. Thus it appears likely that the prior analyses were confounded by unappreciated solvent back-exchange. This difficulty was circumvented in the present experiments by the use of highly purified and active NOV2, which enabled the use of low protein concentrations and short reaction times. Back-exchange was further minimized by employing water-free normal phase HPLC conditions for product purification and MS analyses together with rapid solvent removal from the product-containing extracts.
Our understanding of carotenoid cleavage by CCOs would greatly benefit from a genuine high resolution CCO-substrate structure, which has yet to be reported. A major difficulty in obtaining such a complex relates to the extreme hydrophobicity of carotenoid substrates, which limits their aqueous solubility and prevents stoichiometric formation of enzyme-substrate complexes for structural studies. However, hydroxylated stilbene substrates of the lignostilbene-cleaving CCOs, such as resveratrol, exhibit much greater water solubility and could be more amenable to structural studies. Our demonstration that NOV2 is a dioxygenase indicates that this enzyme may serve as a reliable model system for studying the general mechanism of alkene cleavage by CCOs. We have described a straightforward expression and purification procedure that results in ample quantities of purified and active NOV2 for structural and spectroscopic studies of the alkene cleavage reaction.
Taken together with prior studies, our demonstration of dioxygenase activity in a primitive cyanobacterial carotenoidcleaving CCO, as well as the lignostilbene-cleaving CCO, NOV2, strongly suggest that all alkene-cleaving CCOs are dioxygenases that effect the double bond cleavage reaction by a common catalytic mechanism. It should be noted that a dioxygenase or monooxygenase-labeling pattern does not necessarily imply a specific mechanism of O 2 incorporation into the substrate. For example, DFT calculations on carotenoid cleavage by ACO have shown that a dioxygenase-labeling pattern could be achieved through either dioxetane or epoxide reaction intermediates (19). Future biophysical studies are needed to delineate the precise mode of catalysis by alkene-cleaving CCOs.
In contrast to other CCOs, we have shown here that RPE65 does not rely on O 2 to catalyze ester cleavage/isomerization of all-trans-retinyl esters. This result is not unexpected given that transformation of all-trans-retinyl esters into 11-cis-retinol does not entail any redox chemistry. Moreover, the hydroxyl oxygen atom in the 11-cis-retinol product has been directly shown to originate from water (20). However, it has recently been proposed that O 2 may play a catalytic role in retinoid/ carotenoid isomerization by bonding with C12 of the retinoid backbone to generate a temporary C 11 -C 12 single bond that would allow free bond rotation to a cis-like state followed by oxygen dissociation to restore the polyene conjugation in an 11-cis configuration (25). Even though O 2 does not stoichiometrically participate in this proposed reaction, a major reduction in O 2 concentration, as accomplished by our deoxygenation procedure, should still greatly slow the rate of retinoid isomerization, which is not what we observed. The lack of activity reduction after deoxygenation also cannot be explained by potential tight binding of O 2 to the iron center as our prior structural and spectroscopic studies of RPE65 did not reveal such a stable Fe-O 2 complex (20,44). Most importantly, a recent structural determination of RPE65 in complex with a retinoid mimetic indicates that the polyene chain of carotenoids binds at a position distant from the iron center, which excludes formation of an Fe-O 2 -retinoid complex (26). Thus, the commonality in reaction mechanisms between alkenecleaving CCOs and RPE65 appears unrelated to their O 2 requirements but rather relate to their use of iron to bind oxygen: for O 2 activation in the former and as an ester-polarizing Lewis acid in the latter.
Author Contributions-X. S., M. G., K. P., and P. D. K. conceived and designed the study. X. S., M. G., J. Z., K. A. K., J. v. L., K. P., and P. D. K. performed experiments and analyzed the data. X. S. and P. D. K. wrote the manuscript. All authors reviewed the results and approved the final version of the manuscript.