Engineering Collariella virescens Peroxygenase for Epoxides Production from Vegetable Oil

Vegetable oils are valuable renewable resources for the production of bio-based chemicals and intermediates, including reactive epoxides of industrial interest. Enzymes are an environmentally friendly alternative to chemical catalysis in oxygenation reactions, epoxidation included, with the added advantage of their potential selectivity. The unspecific peroxygenase of Collariella virescens is only available as a recombinant enzyme (rCviUPO), which is produced in Escherichia coli for protein engineering and analytical-scale optimization of plant lipid oxygenation. Engineering the active site of rCviUPO (by substituting one, two, or up to six residues of its access channel by alanines) improved the epoxidation of individual 18-C unsaturated fatty acids and hydrolyzed sunflower oil. The double mutation at the heme channel (F88A/T158A) enhanced epoxidation of polyunsaturated linoleic and α–linolenic acids, with the desired diepoxides representing > 80% of the products (after 99% substrate conversion). More interestingly, process optimization increased (by 100-fold) the hydrolyzate concentration, with up to 85% epoxidation yield, after 1 h of reaction time with the above double variant. Under these conditions, oleic acid monoepoxide and linoleic acid diepoxide are the main products from the sunflower oil hydrolyzate.


Introduction
The oxirane ring of epoxides has been termed the "Lord of the chemical rings" [1] because of its high reactivity in the industrial production of bio-based chemicals and intermediates, including binder ingredients and resins. Epoxy resins comprise a group of cross-linkable materials, which polymerize with co-reactants (curing agents) into a matrix that can be used in a wide range of applications. Epoxy resins and curing agents usually contain more than one reaction site per molecule, to allow multiple crosslink reactions between them. Vegetable oils are one of the most important renewable feedstocks for a bio-based chemical industry [2][3][4]. The epoxides produced from oil fatty acids are possible ingredients for industrial resins (e.g., for board production), as long as they meet the required reaction selectivity and crosslinking properties.
Fatty acid epoxidation is industrially performed by the Prileschajew reaction [5] via percarboxylic acids, traditionally generated by strong acids [6], but also using formic acid [7] or ion-exchange resin [8]. Attempts to use milder conditions include chemoenzymatic lipase-H 2 O 2 reactions on oils, free fatty acids, and their methyl esters [9][10][11][12][13][14], which maintain the drawbacks due to the use of peracids and direct enzymatic epoxidation. Some plant peroxygenases [15], cytochrome P450 monooxygenases [16], and fungal unspecific Some plant peroxygenases [15], cytochrome P450 monooxygenases [16], and fungal unspecific peroxygenases (UPOs) [17] catalyze the direct epoxidation of (poly)unsaturated fatty acids. The latter enzymes present advantages related to their self-sufficiency (being independent of auxiliary proteins/modules and sources of reducing power) and secreted nature (being more stable than intracellular enzymes with monooxygenase activity) [18,19]. UPOs were known as aromatic peroxygenases [20], but, after first reports on their action on alkanes, fatty acids, and alcohols [21,22], numerous examples have shown their wide versatility on aliphatic compounds, including epoxidation reactions [23] and the name changed to unspecific peroxygenases (EC 1.11.2.1). The mechanism of this epoxidation reaction is illustrated in the catalytic cycle shown in Figure 1. This reaction is characterized by the presence of a modified compound II-substrate complex (Cpd II*), absent from the general peroxidase/peroxygenase catalytic cycle, that facilitates epoxide cyclization [19]. Recent studies on structure-function relationships in fatty-acid oxygenation by UPOs have revealed that the different peroxygenation patterns (enzyme regioselectivity) are ruled by the structure of the heme channel [24][25][26]. A good accessibility of the double bond of unsaturated fatty acids to the oxo group of the heme compound I (Cpd I formed after UPO activation by peroxide) is the key of the oxygenation reactions. In this way, different UPOs are able to produce different sets of oxygenated products (epoxy, hydroxy, and hydroxy-epoxy derivatives included) [17,21,27].
The UPO from the ascomycete Collariella virescens (syn. Chaetomium virescens; CviUPO) has been heterologously expressed in Escherichia coli and the recombinant enzyme (rCviUPO), isolated without any purification tag [28]. The obtained amounts are suitable for structure-function and (analytical scale) reaction optimization, making it a good starting point for future UPO studies. Moreover, rCviUPO shows good conversion of unsaturated fatty acids and, in contrast to other UPOs, generates epoxides as main products [25]. rCviUPO was already engineered by site-directed mutagenesis [25] in an initial attempt to mimic the heme environment of the UPO of related Chaetomium globosum, which efficiently epoxidizes unsaturated fatty acids [17]. Then, it was found that conversion of Recent studies on structure-function relationships in fatty-acid oxygenation by UPOs have revealed that the different peroxygenation patterns (enzyme regioselectivity) are ruled by the structure of the heme channel [24][25][26]. A good accessibility of the double bond of unsaturated fatty acids to the oxo group of the heme compound I (Cpd I formed after UPO activation by peroxide) is the key of the oxygenation reactions. In this way, different UPOs are able to produce different sets of oxygenated products (epoxy, hydroxy, and hydroxy-epoxy derivatives included) [17,21,27].
The UPO from the ascomycete Collariella virescens (syn. Chaetomium virescens; CviUPO) has been heterologously expressed in Escherichia coli and the recombinant enzyme (rCviUPO), isolated without any purification tag [28]. The obtained amounts are suitable for structurefunction and (analytical scale) reaction optimization, making it a good starting point for future UPO studies. Moreover, rCviUPO shows good conversion of unsaturated fatty acids and, in contrast to other UPOs, generates epoxides as main products [25]. rCviUPO was already engineered by site-directed mutagenesis [25] in an initial attempt to mimic the heme environment of the UPO of related Chaetomium globosum, which efficiently epoxidizes unsaturated fatty acids [17]. Then, it was found that conversion of polyunsaturated omega 6 fatty acids (i.e., those with a double bond at the sixth position from the omega end) by the F88L variant of rCviUPO promoted diepoxidation [29]. This was explained by a wider heme access channel, facilitating epoxidation of two double bonds.
Increased epoxidation selectivity was also reported for the I153T variant of the Marasmius rotula enzyme [26], the first UPO heterologously expressed in E. coli (rMroUPO) [30]. However, rMroUPO has the disadvantage, compared to rCviUPO, of its much lower expression in E. coli as a soluble enzyme.
In the present work, we aimed to improve the conversion yield and selectivity of plant lipid epoxidation by rCviUPO via two different strategies: (i) enzyme engineering to achieve two epoxy groups per molecule, and (ii) analytical scale optimization of the reaction parameters to epoxidize hydrolyzed vegetable oil.

Production of Native Enzyme and Site-Directed Variants
The CviUPO sequence [31] was optimized for E. coli expression (using the software Optimizer) [32], synthesized, cloned, and produced as previously described [28]. The nonmutated recombinant (hereinafter native) enzyme was purified by two chromatographic steps in an Äkta (GE Healthcare, Chicago, IL, USA) fast liquid chromatography system. The first step was cation exchange chromatography with a HiTrap SPFF column (GE, Healthcare, Chicago, IL, USA) in 10 mM Tris (pH 7.4). The proteins, eluted as a single peak (recorded at 420 nm) with a gradient of the same buffer supplemented with 1 M NaCl, were concentrated in an Amicon 3K device (Sigma-Aldrich, Saint Louis, MO, USA). The second step (to ensure protein purity) was size-exclusion chromatography with a Superdex 75 column (10/300 GL; GE Healthcare, Chicago, IL, USA) in 10 mM Tris (pH 7.4) with 0.15 M NaCl.
A molecular model of the CviUPO structure was obtained at the Swiss-Model server (http://swissmodel.expasy.org, accessed on 1 March 2022) [33] with the structure of the first-reported [34] and crystallized [35] UPO of Agrocybe aegerita (AaeUPO) as a template (PDB entry 2YP1). This model was used for the design of four mutated variants with progressively enlarged heme access channels. The F88A and T158A single mutations were introduced in the CviUPO gene cloned in pET23a using the Expand Long Template PCR kit from Roche (Basel, Switzerland) and the following oligonucleotides as primers (direct sequences with the mutated codons underlined): F88A: 5 -ACT TAC ACC GTT CAG CAG CGT ATC GCG AGT TAC GGT GAA ACG-3 , T158A: 5 -AAC CGC CAT AAC CTG GCG GAA CAT GAT GCA TCT C-3 . For double mutation, the vector containing the CviUPO gene with the first mutation was used as a template. The PCR products were digested with DpnI and transformed into E. coli DH5α for propagation. The gene of the 6Ala sextuple variant-including the L64A, I61A, F88A, T157A, T158A, and T165A mutations-was synthesized by ATG-biosynthetics GmbH (Merzhausen, Germany). All the variants were produced in E. coli as active cytosolic enzymes, and purified as described above for the native rCviUPO.
Enzyme purity was confirmed under denaturing conditions by 12% polyacrylamide gel electrophoresis (PAGE) in the presence of 0.1% sodium dodecyl sulfate (SDS) and 1% mercaptoethanol [36]. Proper folding and binding of the cofactor were verified by analyzing the UV-visible spectrum of the resting state of the enzymes in 10 mM Tris (pH 7.4) using a Cary 60 spectrophotometer (Agilent, Santa Clara, CA, USA). Additionally, formation of the characteristic complex between reduced heme-thiolate enzymes (ferrous form) and carbon monoxide (CO) was assessed in 0.2 M phosphate (pH 8) after addition of Na 2 S 2 O 4 and CO flushing. UPO concentrations were calculated using the rCviUPO molar extinction coefficient of E 420 = 114.2 mM −1 ·cm −1 [28].
Kinetic curves for oleic acid oxidation (estimated chromatographically) were obtained by varying the concentration from 12.5 µM to 1.6 mM substrate, in 50 mM phosphate (pH 7) containing 20% acetone (v/v). Reactions (1 mL) were triggered by the addition of 24 mM H 2 O 2 and stopped after 1 min by vigorous shaking with 100 µL of 0.1 M sodium azide. The oxygenated products were extracted, analyzed by gas chromatography-mass spectrometry (GC-MS) as described below, and their total abundance was used for the calculation of kinetic constants. All reactions were carried out in triplicate.
Curve-fit and data analysis for kinetic constant estimation were carried out using Sigma Plot 11.0. Michaelis-Menten constant (K m ) and turnover number (catalytic constant, k cat ), and their standard errors were obtained by non-linear fitting the k obs values to Equation (1) (Michaelis-Menten model) except for (i) ABTS oxidation by rCviUPO and T158A using 1 mM H 2 O 2 ; (ii) ABTS oxidation by F88A (using either 1 mM or 24 mM H 2 O 2 ); (iii) benzyl alcohol oxidation by F88A and F88A/T158A using 1 mM H 2 O 2 ; (iv) oleic acid oxidation by the F88A, F88A/T158A, and 6Ala variants; (v) H 2 O 2 reduction (in presence of 2.5 mM ABTS) by rCviUPO and T158A variant, where enzyme inhibition was observed, being therefore adjusted to Equation (2) (with the k i inhibition constant being the concentration producing half maximum inhibition); and (vi) benzyl alcohol oxidation by rCviUPO and T158A using 1 mM H 2 O 2 that was adjusted to Equation (3) (Hill model with n H providing a measurement of the cooperativity of the substrate binding to the enzyme, and K 0.5 being the substrate concentration for half saturation).
Oil hydrolyzate was assayed for epoxide production as a more economical and sustainable substrate than pure fatty acids. Sunflower oil (supplied by Cargill in the frame of the project SusBind, https://susbind.eu, accessed on 1 March 2022) was saponified and the hydrolyzed fatty acids extracted at acidic pH as previously described [37]. Initial reactions were performed for 30 min using 0.1 mM hydrolyzate, 0.25 µM enzyme, and 1 mM H 2 O 2 in 50 mM phosphate buffer (pH 7) containing 20% acetone (conditions similar to those used with individual fatty acids). In further reactions, up to 10 mM oil hydrolyzate was used, and the effect of several variables was studied including: pH 5.5/7.0, 20/30% acetone cosolvent, 0.25/0.50 µM enzyme (resulting in 100-400 substrate/enzyme (S/E) molar ratio), 1-100 mM H 2 O 2 (resulting in 1.0-6.8 equivalents per fatty-acid double bond) added with a syringe pump, and 30/60 min reaction time. Given the results obtained with individual fatty acids, the F88A/T158A variant was selected for hydrolyzate reactions, compared with native rCviUPO. Epoxidation yields were calculated taking into account the epoxidation degree, the number of unsaturations, and the reaction conversion for each substrate. In all cases, control experiments were carried out under the same conditions (H 2 O 2 included), but in the absence of enzyme.

GC-MS Analyses
Products (and unreacted substrates) from the above reactions were liquid-liquid extracted with methyl tert-butyl ether (Sigma-Aldrich, Saint Louis, MO, USA), which was evaporated under a N 2 stream. N,O-Bis(trimethylsilyl)trifluoroacetamide (Supelco, Bellefonte, PA, USA) was used to prepare trimethylsilyl derivatives. The GC-MS analyses were performed with an Agilent (Santa Clara, CA, USA) GC-MS QP2020 Ultra equipment using a fused-silica DB-5HT 30 m capillary column from J&W Scientific (Folsom, CA, USA). The oven was heated from 120 • C (1 min) to 300 • C (15 min) at 5 • C min −1 . The injector and transfer line were kept at 300 • C. Compounds were identified by mass fragmentography and comparison of their mass spectra with those of authentic standards. Quantifications were obtained from total-ion peak areas (after deconvolution when partial overlapping was observed) and molar response factors of the same or similar compounds.

Design and Catalytic Characterization of rCviUPO Variants
Previous results with rCviUPO variants [25] led us to study the further broadening of its heme channel. For simplicity, the residues surrounding heme access were mutated into alanines; two simple (F88A and T158A), one double (F88A/T158A), and one sextuple (I61A/L64A/F88A/T157A/T158A/T165A) variants were designed (Figure 2). The variants were expressed as soluble active enzymes, and purified by a combination of cation exchange and size-exclusion chromatography [28]. In all cases, electrophoretic homogeneity was attained, as revealed by SDS-PAGE ( Figure S1). Moreover, correct incorporation of the heme-thiolate group was confirmed by spectrophotometric analysis of the enzyme resting state, with the reduced enzyme complex with CO ( Figure S2 main panels and insets, respectively) exhibiting characteristic maxima around 420 and 440 nm, respectively. ity was attained, as revealed by SDS-PAGE ( Figure S1). Moreover, correct incorporation of the heme-thiolate group was confirmed by spectrophotometric analysis of the enzyme resting state, with the reduced enzyme complex with CO ( Figure S2 main panels and insets, respectively) exhibiting characteristic maxima around 420 and 440 nm, respectively. First, the optimal pH for oxidation of four peroxidase/peroxygenase substrates was determined for the native rCviUPO and variants ( Figure S3). Kinetic curves and constants were first obtained with 1 mM H2O2 as the enzyme-oxidizing substrate (Figure S4A-D  and Table S1, respectively). Catalytic constants for H2O2 were determined using 2.5 mM ABTS as co-substrate. It was observed that the native rCviUPO and its T158A variant are inhibited by high H2O2 concentrations ( Figure S4E), with inhibition constants (ki) of 5.6 mM and 1.8 mM, respectively ( Table 1). The rCviUPO variants showed decreased catalytic efficiency for the enzyme-oxidizing substrate, due to 6-to 15-fold lower affinity for H2O2 than the native rCviUPO, with all the Km values > 1 mM (native rCviUPO included). Therefore, the kcat values for the different enzyme-reducing substrates would be often undervalued under the above conditions. First, the optimal pH for oxidation of four peroxidase/peroxygenase substrates was determined for the native rCviUPO and variants ( Figure S3). Kinetic curves and constants were first obtained with 1 mM H 2 O 2 as the enzyme-oxidizing substrate (Figure S4A-D and  Table S1, respectively). Catalytic constants for H 2 O 2 were determined using 2.5 mM ABTS as co-substrate. It was observed that the native rCviUPO and its T158A variant are inhibited by high H 2 O 2 concentrations ( Figure S4E), with inhibition constants (k i ) of 5.6 mM and 1.8 mM, respectively ( Table 1). The rCviUPO variants showed decreased catalytic efficiency for the enzyme-oxidizing substrate, due to 6-to 15-fold lower affinity for H 2 O 2 than the native rCviUPO, with all the K m values > 1 mM (native rCviUPO included). Therefore, the k cat values for the different enzyme-reducing substrates would be often undervalued under the above conditions.  (2) due to observed inhibition ( Figure S4E) that, despite the good adjustments (R 2 values of 0.988 and 0.997, respectively), has limited precision in predicting kinetic constants.
For this reason, second sets of kinetic curves and constants were obtained using 24 mM H 2 O 2 ( Figure S4F -I and Table S2, respectively) for better enzyme saturation. Under these conditions, increased catalytic efficiency of the simple and double variants could be observed. For F88A/T158A, up to 75-, 10-and 4-fold increases (with respect to native rCviUPO) were found for naphthalene, veratryl alcohol/ABTS, and benzyl alcohol, respectively. The broadest channel variant 6Ala showed improved affinity with all substrates, but often also lower reaction rates, resulting in lower catalytic efficiency.
Concerning oleic acid oxygenation by native rCviUPO and variants (Table 2), smaller differences in kinetic constants than those found for the other substrates were generally observed. Among them, it is worth mentioning the following: (i) a 50% increase of catalytic efficiency by the F88A/T158A variant, due to moderate improvements in both affinity and turnover; and (ii) over 5-fold reduction in catalytic efficiency by the F88A and 6Ala variants. The structural changes introduced in the new variants generated some catalytic improvements in oxygenation ability, particularly with the F88A/T158A variant compared with the native rCviUPO, as shown using naphthalene and oleic acid. These results indicate that this heme channel variant is a potentially interesting epoxidation biocatalyst, as described below.
Concerning the products, chromatograms in Figure 3 illustrate the most interesting reactions: (i) preferential epoxidation of oleic acid by the 6Ala variant (B) without the fatty acid hydroxylated derivatives (at the allylic position) formed by the native enzyme (A), and (ii) preferential diepoxidation (>80% of products) of linoleic and α-linolenic acids (D and F, respectively) by the F88A/T158A variant instead of the mono-epoxidation produced by the native enzyme (C and E, respectively). Finally, encouraged by the recently reported enantioselectivity of fatty acid monoepoxidations with UPOs [24,25], reactions with oleic acid were carried out at a larger scale, and the epoxyoleic products were purified prior to chiral HPLC analysis. The 6Ala variant was selected due to its high regioselectivity to the epoxide, together with the F88A/T158A variant as the best UPO in terms of the overall epoxidation yield with all substrates (and native rCviUPO for comparative purposes). The results obtained ( Figure S8) revealed only low enantioselectivity with the three enzymes (ee 0-40%), although an inversion of the configuration of the main enantiomer was produced between the F88A/T158A and 6Ala variants (70% S/R and 60% R/S, respectively).

Optimization of Hydrolyzed Sunflower Oil Epoxidation
After testing the rCviUPO variants on individual unsaturated fatty acids, the goal was to accomplish the epoxidation of hydrolyzed vegetable oil. For this purpose, a sunflower oil hydrolyzate was used as substrate, and several parameters-including substrate, enzyme, and H 2 O 2 concentrations, among others-were optimized.
Firstly, reactions under the same conditions used with pure fatty acids were performed on hydrolyzed sunflower oil (0.1 mM) using native rCviUPO and its F88A/T158A variant ( Figure 4A,B, respectively). Both enzymes were able to convert the oil unsaturated fatty acids with similar regioselectivity, forming (i) monoepoxides from oleic acid and linoleic acid (at 9 and 12 positions) as main products, (ii) two diepoxy isomers (anti and syn) from linoleic acid, more abundant in the F88A/T158A reaction; and (iii) small amounts of hydroxy-and hydroxy-epoxy derivatives of oleic acid and linoleic acids, together with some minor products.
The 6Ala variant was selected due to its high regioselectivity to the epoxide, together with the F88A/T158A variant as the best UPO in terms of the overall epoxidation yield with all substrates (and native rCviUPO for comparative purposes). The results obtained ( Figure S8) revealed only low enantioselectivity with the three enzymes (ee 0-40%), although an inversion of the configuration of the main enantiomer was produced between the F88A/T158A and 6Ala variants (70% S/R and 60% R/S, respectively).   Then, differences in the epoxidation yields were observed when varying the enzyme dose, using different substrate/enzyme (S/E) molar ratios. As shown in Table 4 (entries 1-6), increasing the rCviUPO dose did not result in improved epoxidation yield, but doubling the amount of F88A/T158A (0.5 µM, S/E ratio 200) yielded higher amounts of diepoxides from linoleic acid (the epoxidation of oleic acid was retained); consequently, a better epoxidation yield (72%) was attained. Nevertheless, a 1 µM dose of the double variant (S/E ratio 100) produced virtually the same epoxidation profile. Therefore, optimal 400 and 200 S/E ratios were fixed for further scale up with rCviUPO and its double variant, respectively. extension of the reaction time up to 60 min was needed to reverse the epoxidation performance with the F88A/T158A variant at the initial reaction level (entry 10). In fact, the reaction time becomes critical in UPO epoxidation of fatty acids when the substrate concentration is close to the solvent saturation [28].    However, when the substrate loading was increased to 5 mM (Table 4 entries 7-10), its conversion rate decayed with both enzymes within 30 min reactions. Therefore, an extension of the reaction time up to 60 min was needed to reverse the epoxidation performance with the F88A/T158A variant at the initial reaction level (entry 10). In fact, the reaction time becomes critical in UPO epoxidation of fatty acids when the substrate concentration is close to the solvent saturation [28].
Finally, with 10 mM substrate loading (Table 4 entries [11][12][13][14][15][16][17], the importance of the substrate solubilization was revealed by comparing the epoxidation yields with 20% and 30% acetone cosolvent. With both enzymes, 20% acetone resulted in low conversion and epoxidation yields (longer times did not improve the results). However, the complete solubilization of substrates using 30% acetone improved the epoxidation yield, especially with the double variant (entry 14) forming mono and diepoxide amounts similar to those observed in reactions with a lower substrate concentration.
With the enzyme dose, amount of co-solvent, and reaction time already optimized, assays were conducted to determine the lowest dose of H 2 O 2 that enables the production of the highest epoxidation yield. As depicted in Table 4, the highest epoxidation yield (85%) was attained using only 1.7 equivalents of this oxidizer (entry 16), compared with the 6.8 (used in most previous assays) and 3.4 (entry 15) equivalents, while adding a stoichiometric amount of H 2 O 2 (entry 17) led to a decrease in epoxidation yield (to only 44%).
The pattern of oxygenation products produced under optimized conditions ( Figure 4C,D) was similar to that obtained previously, with linoleic acid diepoxides and oleic acid monoepoxide as the main products in the F88A/T158A reactions. No triepoxides were detected in any of the reactions, due to the low α-linolenic acid content of sunflower oil [37]. The above results reveal the remarkable potential of the rCviUPO double variant for epoxidizing sunflower oil hydrolyzate, demonstrating a much higher conversion yield than the native enzyme.

Conclusions
Opening the rCviUPO heme access channel allowed us to improve the epoxidation activity of this enzyme, selected for protein engineering in E. coli. In this way, the variant with the widest channel, after a sextuple mutation, yielded the highest epoxidation of oleic acid to the corresponding monoepoxide (with less than 5% of other oxygenation products). In contrast, a double variant (F88A/T158A) produced the best conversion of both linoleic and α-linolenic acids with diepoxides, of interest as crosslinking molecules, representing > 80% of products. Under the same conditions used for individual fatty acids, the double variant also converted hydrolyzed sunflower oil with a higher epoxidation yield than the native enzyme. Moreover, process optimization permitted us to increase (×100) the hydrolyzate concentration, epoxidizing 85% of double bonds after 1 h of reaction time with the mutated double variant. UPO engineering, strongly limited in the past by difficulties in heterologous expression of these enzymes in adequate hosts, is a requirement for developing the large repertoire of reactions of industrial relevance suggested in recent reviews [23,[39][40][41][42][43][44][45] on an enzyme family of the highest interest for selective oxyfunctionalization reactions [46].  (Table S1 and  Table S2, respectively); SDS-PAGE of purified native rCviUPO and variants ( Figure S1); UV-visible spectra of resting states and CO complexes of rCviUPO and variants ( Figure S2); Effect of pH on the oxidation of different substrates ( Figure S3); Kinetic curves for different UPO reducing substrates and H 2 O 2 ( Figure S4); GC-MS analyses of oleic, linoleic, and α-linolenic acids reactions with CviUPO and four heme channel variants ( Figure S5, Figure S6 and S7, respectively); and Chiral HPLC analysis of oleic acid epoxide ( Figure S8).