This is the accepted version of the article: Solé Ferré, Jordi; Brummund, Jan (InnoSyn B.V.); Caminal i Saperas, Glòria; [et al.]. «Enzymatic Synthesis of Trimethyl-[U+03F5]-caprolactone : Process Intensification and Demonstration on a 100 L Scale». Organic process research and development, Vol. 23, Issue 11 (November 2019), p. 2336-2344. DOI 10.1021/acs.oprd.9b00185 This version is avaible at https://ddd.uab.cat/record/217009 under the terms of the license 1 Enzymatic synthesis of trimethyl-ε-caprolactone: process intensification and 2 demonstration at 100 liter scale 3 4 Jordi Soléa, Jan Brummundb, Gloria Caminalc, Gregorio Álvaroa, Martin Schürmannb and Marina 5 Guilléna. 6 a 7 Bioprocess and Applied Biocatalysis Group, Department of Chemical Biological and Environmental 8 Engineering, Universitat Autònoma de Barcelona, 08193 Bellaterra, Catalonia, Spain. b 9 InnoSyn B.V., Urmonderbaan 22, 6167 RD Geleen, The Netherlands c 10 Institut de Química Avançada de Catalunya (IQAC), Barcelona, Spain 11 12 Corresponding authors: 13 Marina Guillén 14 Department of Chemical, Biological and Environmental Engineering, Universitat Autònoma de 15 Barcelona, 08193 Bellaterra, Catalonia, Spain 16 Telephone: +34 93 581 4793 17 Fax: +34 93 581 2013 18 marina.guillen@uab.cat 19 20 Martin Schürmann 21 InnoSyn B.V., Urmonderbaan 22, 6167 RD Geleen, The Netherlands 22 Telephone: +31 683 649 402 23 martin.schuermann@innosyn.com 1 24 Abstract 25 Optimization and scale up of the Baeyer-Villiger oxidation of 3,3,5-trimethyl-cyclohexanone to 26 trimethyl-ε-caprolactones (CHL) was studied in order to demonstrate this technology on 100 L pilot 27 plant scale. The reaction was catalyzed by a cyclohexanone monooxygenase from Thermocrispum 28 municipale (TmCHMO) that utilizes the costly redox cofactor NADPH, which was regenerated by a 29 glucose dehydrogenase (GDH). 30 As a first stage, different cyclohexanone monooxygenase formulations were tested: cell free extract, 31 whole cells, fermentation broth and sonicated fermentation broth. Using broth resulted in the highest 32 yield (63%) and required the least biocatalyst preparation effort. Two commercial glucose 33 dehydrogenases (GDH-105 and GDH-01) were evaluated resulting in similar performances. 34 Substrate dosing rate and biocatalyst loadings were optimized. At 30 mL scale, the best conditions 35 were found when 30 mM h-1 dosing rate, 10% (v/v) cyclohexanone monooxygenase broth and 0.05% 36 (v/v) of glucose dehydrogenase (GDH-01) liquid enzyme formulation were applied. These same 37 conditions (with oxygen instead of air) were applied at 1 liter scale with 92% conversion, achieving a 38 specific activity of 13.3 U g-1 cell wet weight (cww), a space time yield of 3.4 g CHL L-1 h-1 and a 39 biocatalyst yield of 0.83 g CHL g-1 cww. 40 A final 100 liter demonstration was performed in a pilot plant facility. After 9 hours, the reaction 41 reached 85% conversion, 12.8 U g-1 cww, a space time yield of 2.7 g L-1 h-1 and a biocatalyst yield of 42 0.60 g CHL g-1 cww. The extraction of product resulted in 2.58 kg isolated final product. The overall 43 isolated CHL yield was 76% (distal lactone 47% and proximal lactone 53%). 44 45 Keywords: Baeyer-Villiger monooxygenase (BVMO); branched lactone synthesis; pre-industrial scale; 46 applied biocatalysis; biocatalyst yield; space time yield. 2 47 Introduction 48 Historically, the chemical pathway followed for the production of aliphatic polyesters comprises the 49 Baeyer-Villiger oxidation of (branched) cyclic ketones to obtain (branched) lactones that serve as 50 monomer building blocks. This route presents several drawbacks: limited regioselectivity, use of 51 pollutant organic solvents, halogenated oxidants and formation of undesired by-products 1. 52 The biocatalytic alternative, on the other hand, overcomes most of these disadvantages. Baeyer- 53 Villiger monooxygenases operate in mild aqueous conditions, make use of molecular oxygen as oxidant 54 molecule and frequently present high regioselectivity 2,3. Different examples of large scale applications 55 of these enzymes can be found in the literature 4,5. For example, Hilker et al. scaled up the oxidative 56 reaction of racemic bicyclo[3.2.0]hept-2-en-6-one to kilogram scale using a resin-based in situ 57 substrate-feeding/product-removal (SFPR) strategy 6. The main disadvantages of these biocatalysts 58 though, are its low stability, poor organic solvent tolerance, substrate and product inhibition and the 59 use of the costly and unstable NADPH cofactor 7–9. 60 In 2016, a cyclohexanone monooxygenase from Thermocrispum municipale DSM 44069 (TmCHMO; EC 61 1.14.13.22) was discovered 10. It exhibits an improved thermostability with a melting temperature of 62 48˚C, improved organic solvent tolerance and a broad scope towards branched cyclic ketones 63 compared to other CHMOs. Furthermore, two different fusion constructs have been obtained in order 64 to enable in situ regeneration of the cofactor. First, TmCHMO was fused to an alcohol dehydrogenase 11 65 and it was used to oxidize cyclohexanol to ε-caprolactone . Second, TmCHMO was fused to a 66 phosphite dehydrogenase that utilizes phosphite as a sacrificial substrate 12,13. 67 In this work, TmCHMO was investigated in combination with a glucose dehydrogenase (GDH), as 68 cofactor regeneration enzyme 14, for the oxidation of 3,3,5-trimethylciclohexanone (TMCH) to produce 69 trimethyl-ε-caprolactone (CHL) (Scheme 1). The two region-isomers of this product have gained 70 interest in recent years for the synthesis of aliphatic (co)polyesters with low glass transition 3 71 temperatures 15–17. Due to this property, these polymers can be used as biodegradable plasticizers or 72 encapsulating agents for coating formulations 16. 73 The sacrificial substrate used by GDH is glucose (Glc) that, on its turn, forms gluconolactone which 74 spontaneously hydrolyses to gluconic acid that causes an acidification of reaction medium. Therefore 75 the pH was controlled by addition of base and the amount of base added, provided a good indication 76 of the reaction course. 77 The combination of TmCHMO and a commercial GDH-105 to synthesize trimethyl-ε-caprolactones 78 from 3,3,5-trimethyl-cyclohexanone was recently studied and published by Delgove et al. 18. In that 79 work, reaction engineering regarding biocatalyst loading, air flow, stirring rate, substrate dosing rate 80 and methanol concentration was performed up to 500 mL scale and conversions >99% were achieved 81 with 240 mM substrate concentration. Since the reaction suffers from substrate inhibition, a constant 82 substrate feeding strategy was applied. 83 This work has been focused on further optimization of the main limiting parameters of the 84 biotransformation to finally scale up the optimized reaction to 100 liter reaction volume in the 200 L 85 pilot plant reactor of InnoSyn B.V. 4 86 87 Scheme 1. Biocatalyzed oxidation of the branched substrate 3,3,5-trimethyl-cyclohexanone (1) to a mixture of β,β,δ-trimethyl- 88 ε-caprolactone (1a) and β,δ,δ-trimethyl-ε-caprolactone (1b) with the two-enzyme system consisting of cyclohexanone 89 monooxygenase (TmCHMO) and glucose dehydrogenase (GDH) to regenerate the NADPH using D-glucose as a sacrificial 90 substrate. 91 92 Materials and methods 93 Chemicals 94 D-(+)-glucose (> 97.5%) and ethyl acetate (> 99.9%) were purchased from VWR Chemicals (Radnor, 95 USA). Dicalite 4208 was purchased from Dicalite Europe (Gent, Belgium). β-Nicotinamide adenine 96 dinucleotide phosphate disodium salt (> 93%), was obtained from SyncoZymes (Pudong Xinqu, China). 97 All the other reagents were purchased from Sigma-Aldrich and were of analytical grade if not stated 98 elsewhere. Commercial glucose dehydrogenase GDH-105 was purchased from Codexis® (Redwood 99 City, USA) in powder form. Commercial glucose dehydrogenase GDH-01 was supplied by InnoSyn B.V. 5 100 (Geleen, The Netherlands) as liquid enzyme formulation. All solutions like buffers or stocks were 101 prepared freshly prior to use. Lactones 1a and 1b (> 99%) (Scheme 1), that served as analytical 102 standards were provided by University of Maastricht and synthesized according to a chemical Baeyer- 103 Villiger oxidation procedure reported in the literature 19. 104 Recombinant production of TmCHMO in Escherichia coli. 105 Cyclohexanone monooxygenase from T. municipale (TmCHMO) was recombinantly produced in 106 Escherichia coli in a 10 L scale fed-batch, high cell-density fermentations with glucose as growth limiting 107 C-source employing an E. coli K12 derivative and a pBR322 derived expression vectors. 500 mL pre- 108 culture was used to inoculate 10 kg main culture medium with 100 µg mL-1 neomycin. Pre-culture was 109 prepared in standard Luria-Bertani (LB) medium supplemented with 100 µg mL-1 neomycin. The 110 fermentation was performed using mineral medium supplemented with 20 g L-1 yeast extract. 1.5 d 111 after inoculation of the fermenter as inducer, pre-sterilized L-arabinose was added to the fermenter 112 to final concentration 0.02 % (w/w). After about 100 h the biomass was harvested. 113 Fermentation broths were harvested either by centrifugation (wet cells) and the supernatant 114 discarded or the fermentation broth (413 gcww L-1) was used as biocatalyst as such (broth). The 115 sonicated broth was prepared by sonication of the broth with an ultrasound probe for 20 minutes (10 116 seconds on, 10 seconds off) with cooling on ice. Whole cell formulations were prepared by adding 2 117 weight equivalents of 100 mM potassium phosphate (KPi) buffer (pH 7.0) to 1 equivalent of E. coli wet 118 cells (333.3 gcww L-1). Cell free extracts (CFE) were prepared by disrupting the whole cell formulation by 119 homogenization using a Microfluidics M-110P (Westwood, MA, USA) homogenizer. 120 Glucose dehydrogenase activity measurement 121 GDH activity was measured spectrophotometrically ( λ = 340 nm) following the formation of NADPH (ε 122 = 6.22 mM-1 cm-1) 20. Certain amount of sample (1/20 of the total volume) was added to a solution 123 containing D-(+)-glucose (200 mM dissolved in sodium phosphate buffer 100 mM pH = 8) and NADP+ 6 124 (400 µM). Basal formation of NADPH by unspecific enzymes present in the GDH preparation was 125 determined by this same test but avoiding the addition of substrate and adding buffer instead. This 126 formation rate is subtracted from the measurement with glucose. One unit of activity (U) is defined as 127 the enzyme required to convert 1 µmol of NADP+ per minute at those given conditions (30°C and pH = 128 8). Plastic micro cuvettes BRAND® UV (Sigma Aldrich) were used for the assay. 129 Reactor setup (30 mL, 1 L and 100 L) Figure 1: Applied 1 liter and 100 liter reactor setup. A) stirrer, B) pH electrode, C) reflux-condenser, D) air/oxygen supply controller, E) jacketed 1000 mL reactor, F) automatic titration device, G) gas washing bottle H) oxygen sensor in gas outlet I) oxygen sensor in reaction mixture J) substrate dosing pump. 130 Biocatalytic reactions were performed in cylindrical jacketed reactors (Figure 1) that consisted of a 131 glass vessel (30 mL and 1 L) or glass-lined stainless steel vessel (100 L); a pH controller (Metrohm Titrino 132 plus 877) filled with 1 M NaOH solution (30 mL) or 5 M NaOH solution (1 L and 100 L); a propeller 133 stirrer, a thermostat (MGW-LAUDA RC6) set at 30˚C, a condenser at 6˚C, a compact mass flow regulator 134 (GCR Red-y) to keep a constant air/oxygen flow; and substrate dosing pump (Harvard Pump11). For 135 the 1 L and 100 L reactions, reactors also contained an oxygen sensor (SDL 150 – Extech Instruments) 136 in the gas outlet, an oxygen sensor for the liquid phase and a constant nitrogen flow in the headspace 137 to maintain the oxygen concentration below 8% and prevent formation of potentially explosive 138 mixtures 21. 139 Determination of the reaction progress for biocatalyzed reactions using GC-FID 7 140 Samples (150 µL) were taken periodically from the reactor, weighed and dissolved up to 10 mL with a 141 solution of acetonitrile containing 0.5 g L-1 of hexadecane that served as Internal Standard (IS). The 142 mixture was centrifuged to remove insoluble biomass and the supernatant was analyzed by GC-FID. 143 Concentration of substrate and products were determined using calibration curves. 144 GC-FID method and equipment 145 Samples containing 3,3,5-trimethyl-cyclohexanone and trimethyl-ε-caprolactones were analyzed using 146 a 7890A gas chromatograph (Agilent Technologies, USA) equipped with a HP-5 column (30 m, 0.32 147 mm, 0.25 µm df, Agilent Technologies). The column temperature was maintained at 60°C for 2 minutes, 148 increased up to 300°C at 10°C min-1 and it was held at final temperature for 2 minutes. The injector 149 temperature was kept at 200°C; for the flame ionization detector, the temperature was 300°C. 150 Hydrogen was used as a carrier gas at a flow rate of 40 mL min-1 and air at 450 mL min-1. The retention 151 times observed were: 8.5 min for the substrate 1, 11.9 min for lactone 1b, 12.1 min for lactone 1a and 152 15.8 min for the IS. 153 Isolation and purification of substrate and product from the 100 L reaction 154 The isolation and purification procedure followed here is a modified version of the protocol described 155 by Delgove et al. 18. Once the reaction was finished, 50 liter of ethyl acetate were added to the reactor 156 and stirred at 40˚C overnight. After that, 6 kg Dicalite 4208 were added and the mixture was stirred for 157 0.5 h. Subsequently the whole reactor content was filtered and the two phases were separated. The 158 filter was washed three times with 25 liters of ethyl acetate each. The ethyl acetate that was used to 159 carry out the first washing of the filter cake was mixed with the aqueous phase, which had been 160 separated from the first extraction, to perform a second extraction (40˚C and 0.5 h). After that, a third 161 extraction (40˚C and 0.5 h) was carried out with the ethyl acetate resulting from the second and third 162 washings of the filter. At the end all resulting organic phases were mixed and reintroduced into the 163 reactor to start the distillation of the solvent. The distillation was carried out at 42˚C and 170 mbar. 8 164 165 Results and discussion 166 Process development on 30 mL scale 167 Control experiments for substrate and product solubility and background titration 168 Before starting to optimize the reaction, some tests were performed to assess substrate and product 169 solubility and background titration of the system. 170 To study substrate solubility, 240 mM TMCH were added to the reactor in presence of 25 mM KPi 171 buffer and 10% (v/v) methanol. The final substrate and methanol concentrations were set based on 172 the previous work of Delgove et al. 18. The measured TMCH concentration was between 100 and 130 173 mM. Apart from this, even when substrate was dosed, the analyzed concentration over time was 174 always below the added concentration, leading to the conclusion that substrate required some 175 additional time after dosing to dissolve in the reaction system completely. 176 Regarding the product solubility, CHL was considered to be rather insoluble even in presence of 10% 177 (v/v) methanol. In order to prove this fact, the conditions reported by Delgove et al. were mimicked 178 here and samples were taken and analyzed by GC 4. 25 mM KPi pH 8, 5% (v/v) TmCHMO CFE and 0.1 179 mg mL-1 GDH-105 were the conditions used. The solubilized CHL concentration did not exceed a 200 180 mM threshold. Higher methanol concentration would be needed to solubilize the CHL content if the 181 whole substrate were converted (240 mM). 182 Moreover, control experiments were carried out to study a potential NaOH titration by the pH 183 controller without any target reaction taking place (background acidification and titration). Among 184 these background acidification effects, the most important considered were: 1) by-products of 185 potential bacterial growth, 2) by-products of enzymes present in TmCHMO cell-free extracts (CFEs) and 186 3) spontaneous degradation of NADPH to NADP+. All of them were studied; the first by using broth of 187 an E. coli strain expressing a non-target enzyme (neither CHMO nor GDH), the second by using only 9 188 TmCHMO CFE and the third, by using only the cofactor regenerating GDH liquid enzyme formulation 189 without TmCHMO. Titration values were: 10.9%, 0.5% and 1.8%, respectively (percentages expressed 190 with respect to the total substrate added). As a conclusion, the NaOH added by the pH controller was 191 considered to be as a good estimation of the total CHL produced. 192 Variation of the applied TmCHMO enzyme formulation 193 The reaction conditions, which were published recently, served as starting point for further 194 investigation 18. The reactor contained 240 mM final substrate concentration which was dosed as stock 195 solution in methanol at a dosing rate of 15 mM h-1, finally 10% (v/v) methanol, 375 mM of D-glucose, 196 0.25 mM of NADP+, 25 mM KPi buffer, pH 8.0, stirring at 500 rpm, 16 mL min-1 air flow, 5% (v/v) 197 TmCHMO CFE and 0.1 mg mL-1 GDH-105. The only implemented changes were stirring rate (1200 rpm 198 instead of 500 rpm); and co-solvent methanol that was dosed together with the substrate instead of 199 adding it to the reactor from the beginning. 200 The obtained results can be seen in the graph below (Figure 2, left columns, CFE). 40% CHL yield was 201 obtained. The ratio of product regio-isomers, 45 ± 2% distal lactone and 55 ± 2% proximal lactone, was 202 sustained in all reactions. Since the final product (polymer) consist on a mixture of the two regio- 203 isomers and both polymerize in a similar manner, small differences in the monomer’s ratio, are not 204 expected to be critical 13,15. 205 In Figure 2 – left (CFE), the difference between the three presented values (conversion, yield and 206 titration) can be attributed to the poor solubility of the components. In order to reduce this difference, 207 for the following reactions (30 mL), the whole reactior content was dissolved in acetonitrile at the end 208 of the reaction. By doing this, both substrate and product could be properly solubilized and a final 209 representative sample could be taken. The results obtained from the analysis of this final dilution are 210 the ones presented in all bar charts in this article. 10 211 The first parameter that has been investigated was the TmCHMO formulation (Figure 2). As described 212 in the experimental part, the cell free extract formulation used by Delgove et al. is the one that requires 213 more unit operations to be obtained. Thus, in order to reduce costs, three other formulations were 214 investigated. Apart from the cell free extract, also whole cells (WC), fermentation broth and sonicated 215 fermentation broth were tested. 216 The results showed that the application of broth and sonicated broth resulted in highest yields and 217 highest rates (Figure 2, right columns). The difference with the results that were achieved with the 218 other two TmCHMO formulations (CFE and WC) can be mainly explained by the higher biomass 219 concentration that these two formulations contained. While the broth and sonicated broth had 413 220 gcww L-1, the CFE and WC presented 333 gcww L-1. 221 Since fermentation broth requires the least unit operations for its production (no cell disruption, 222 centrifugation and/or washing steps), it was selected over the other three formulations. 223 In order to reduce costs, once the final TmCHMO formulation was selected (broth), the commercial 224 GDH-105 from Codexis® (0.1 mg mL-1) was substituted for GDH-01 produced in InnoSyn B.V. facilities 225 (0.5% (v/v) of LF). The pH of the reaction had to be adjusted to 7. The yields obtained were very similar, 226 63% for GDH-105 and 67% for GDH-01. 11 227 228 Figure 2. Comparison between four different formulations of TmCHMO regarding conversion (black bar), yield (grey bar), 229 titration (white bar) and maximum rate (black circles, right axis). The rate was calculated by linear regression on the reaction 230 kinetics (from 1 to 4 hours). Reactions were performed in 30 mL reactors with an enzyme loading of 5% (v/v) of the respective 231 TmCHMO formulation and 0.1 mg mL-1 of GDH-105; temperature 30°C; stirring rate 1200 rpm; air flow 16 mL min-1; pH 8; 232 [TMCH] 15 mM h-1 (240 mM final); Methanol 0.625% (v/v) h-1 (10% (v/v) final); [D-glucose] 375 mM; [NADP+] 0.25 mM; 233 titration solution 1 M NaOH. 234 Variation of the applied substrate dosing rate 235 Substrate availability in the reactor was a limiting factor for the first 6 hours of reaction (Figure 3, left). 236 As it can be observed, at the beginning of the reaction, there was no substrate accumulation (dosing 237 rate 15 mM h-1). Increasing the dosing rate could result in higher reaction rates that in turn would 238 represent shorter reaction times for constant total substrate amounts. With shorter reaction times, 239 enzyme deactivation as well as product degradation (e.g. potential hydrolysis or methyl-ester 240 formation) would be reduced. As a result, reaction yield would increase. 12 241 Doubling the dosing rate to 30 mM h-1 (Figure 3, right) reduced the reaction time from 18.2 h to 13.5 h 242 (according to the titration progress), it increased the maximum rate from 0.54 mmol h-1 to Figure 3. Reaction in 30 mL reactor with 15 mM h-1 substrate dosing rate (left graph) and 30 mM h-1 (right graph). The graphs show the reaction course of TMCH (white circles), CHL (black circles), NaOH addition (black line), substrate dosing (discontinuous line) and mass balance (combined discontinuous spot-line-spot). Conditions: biocatalyst loading of 5% (v/v) of TmCHMO Broth and 0.5% (v/v) of GDH-01; temperature 30°C; stirring rate 1200 rpm; air flow 16 mL min-1; pH 7; 240 mM final TMCH concentration; Methanol 0.625% (v/v) h-1 (10% (v/v) final); [D-glucose] 375 mM; [NADP+] 0.25 mM; titration solution 1 M NaOH. 243 0.68 mmol h-1 and the yield from 67% to 79%. However, substrate accumulation is detrimental for the 244 reaction since the BVMO suffers from substrate inhibition. That is why the following experiments were 245 aimed at reducing the limiting factors provoking this accumulation. The final mass balance, on the 246 other hand, did not change and stayed around 90% in both cases. 247 Variation of the applied TmCHMO loading 248 Once the substrate dosing rate was increased, the considered limiting factors were: TmCHMO loading, 249 GDH loading and the oxygen supply rate. Using pure oxygen instead of pressurized air was the feasible 250 change to overcome oxygen limitations in the reaction. However, this change was not possible at 30 251 mL scale due to safety issues (nitrogen purge has to be installed to keep oxygen levels in a sufficiently 252 low range) 21. Regarding GDH, on the other hand, a 0.5% (v/v) loading of GDH-01 represents 9.6-fold 253 more Units of activity compared to the experiments with GDH-105 (0.1 mg mL-1). Therefore, this higher 254 presence of GDH activity meant that, GDH was most probably not limiting the reaction. For these 255 reasons, the first investigated process parameter was the TmCHMO loading using a variation from 5.0% 256 (v/v) to 7.5% (v/v) and up to 10% (v/v) (Figure 4). 13 257 258 Figure 4. Comparison between three different TmCHMO loadings regarding conversion (black bar), yield (grey bar), titration 259 (white bar) and maximum rate (black circles, right axis). The rate was calculated by linear regression on the reaction kinetics 260 (from 1 to 4 hours). Reactions in 30 mL reactors with an enzyme loading of 5%; 7.5% and 10% (v/v) of TmCHMO Broth and 261 0.5% (v/v) of GDH-01; temperature 30°C; stirring rate 1200 rpm; air flow 16 mL min-1; pH 7; [TMCH] 30 mM h-1 (240 mM final); 262 Methanol 1.25% (v/v) h-1 (10% (v/v) final); [D-glucose] 375 mM; [NADP+] 0.25 mM; titration solution 1 M NaOH. 263 The maximum rate achieved by 10% (v/v) TmCHMO loading reaction was not twice as high as by using 264 the 5% (v/v) loading which meant that other factors were also limiting the reaction. Thus, oxygen was 265 most probably the cause. If the biocatalyst yield was taken into account, 5% (v/v) loading should be 266 used. It presented 79% yield, only 17% less than with the 10% (v/v) TmCHMO loading. However, the 267 goal was to get as close as possible to full conversion to facilitate an efficient down-stream processing 268 to isolate the product. Finally, 10% (v/v) TmCHMO broth was selected as biocatalyst loading for further 269 investigations. 270 Variation of the applied GDH loading 271 As aforementioned, GDH was most probably not limiting the reaction. However, the added amount of 272 0.5% (v/v) liquid enzyme formulation had not been optimized. A broad range of GDH-01 loadings, from 273 1% (v/v) to 0.01% (v/v) was screened (Figure 5). As it can be seen in the graph below, the maximum 14 274 rate was maintained and full conversion was achieved with 0.05% (v/v) loading. This loading represents 275 3.8 U mL-1 in the reactor, similar activity as offered with 0.1 mg mL-1 of GDH-105 (3.9 U mL-1). It could 276 be concluded that GDH-01 is as stable as GDH-105 under the applied conditions. 277 278 Figure 5. Comparison between seven different GDH loadings regarding conversion (black bar), yield (grey bar), titration 279 (white bar) and maximum rate (black circles, right axis). The rate was calculated by linear regression on the reaction 280 kinetics (from 1 to 4 hours). Reactions in 30 mL reactors with and enzyme loading of 10% (v/v) of TmCHMO Broth and 281 1%, 0.5%, 0.25%, 0.1%, 0.05%, 0.01% or 0% (v/v) of GDH-01 liquid enzyme formulation; temperature 30°C; stirring rate 282 1200 rpm; air flow 16 mL min-1; pH 7; [TMCH] 30 mM h-1 (240 mM final); Methanol 1.25% (v/v) h-1 (10% (v/v) final); [D- 283 glucose] 375 mM; [NADP+] 0.25 mM; titration solution 1 M NaOH. 284 The results for the reaction with 0.01% (v/v) loading were: 0.14 mg of total protein (0.08 mg of GDH- 285 01) added and 6.4 mmol (1153.6 mg) D-glucose consumed (NaOH added). Even though this reaction 286 did not result in full conversion, it presented the highest biocatalyst yield, 13970 g Glc g-1 GDH-01. It 287 must be taken into account though, that part of the total titration, is a consequence of bacterial 288 growth. 289 When no GDH-01 was added (right columns), 11.2% yield and 5.3% titration were still obtained. NADPH 290 regeneration system from the E. Coli cells present, probably allowed certain biotransformation of the 291 TMCH. Regarding the substrate consumption, since the TMCH solubility (100 to 130 mM) did not allow 15 292 reliable measurements of its concentration, the conversion value was misleading and it was obviated 293 in this case. 294 The GDH-01 loading that was selected for the subsequent studies did not change, it remained at 0.5% 295 (v/v). It was not advisable to go too low with GDH concentration, because this could lead to a decreased 296 reaction performance if other critical process parameters are further optimized. 297 Scale-up to 1 L scale 298 Scale-up of the optimized conditions and application of pure oxygen 299 As explained before, in order to use pure oxygen instead of air, a nitrogen flow had to be constantly 300 applied in the headspace of the reactor to keep the oxygen concentration below 8% (to avoid a 301 potentially explosive atmosphere). This installation was applied in a 1 liter set up where previously 302 optimized conditions were applied. 10% (v/v) loading of TmCHMO Broth, 0.5% (v/v) loading of GDH- 303 01 liquid enzyme formulation, 30 mM h-1 substrate dosing rate and 20 mL min-1 oxygen flow were 304 selected for the initial experiment. Oxygen measurements enabled certain oxygen control for the 305 operator, assuring that oxygen was always available in the reactor by adaption of the oxygen supply 306 rate respectively the nitrogen purge. Furthermore, oxygen mass balance (oxygen consumption) served 307 as a good estimation of the reaction progress together with the NaOH titration. 308 The reaction course graph can be seen below (Figure 6). No substrate accumulation was observed until 309 the fifth hour, meaning that the substrate feed rate of 30 mM h-1 was dominating the reaction rate, 310 which increased around 20% compared with the 30 mL reaction. The graph shows that the added 311 NaOH that was required to keep the pH stable increased until 260 mmol while the product stayed at 312 182 mmol, close to the solubility limit of CHL. The oxygen consumption, on the other hand, is well 16 313 aligned with the substrate consumption. 99% of the oxygen uptake was used to convert the substrate. Figure 6. Reaction in 1 liter reactor. The graphs show the reaction course of TMCH (white circles), CHL (black circles), NaOH addition (black line), substrate dosing (discontinuous line), mass balance (combined discontinuous spot-line-spot) and oxygen consumption (dotted line). Enzyme loading of 10% (v/v) of TmCHMO Broth #1 and 0.5% (v/v) of GDH-01 liquid enzyme formulation; temperature 30°C; stirring rate 400 rpm; O2 flow 20 mL min-1; pH 7; [TMCH] 30 mM h-1 (240 mM final); Methanol 1.43% (v/v) h-1 (10% (v/v) final); [D-glucose] 375 mM; [NADP+] 0.25 mM; titration solution 5 M NaOH. 314 315 With 1 liter reaction, solubilizing the whole content in acetonitrile was not feasible, as it was in 30 mL 316 scale. Thus, without a proper workup of the reaction mixture an adequate product analysis was not 317 possible. This is why the product yield in the course of reaction was not taken into account at this 318 stage. At the end, 92% conversion and 111% titration were reached, both values close to those 319 obtained at 30 mL (93% and 101% respectively). Reaction time was slightly reduced compared with 320 the 30 mL reaction, from 12 h to 10 h, according to the titration curve. 321 Utilization of pure oxygen instead of air allowed the reaction to be limited only by the applied substrate 322 dosing rate (30 mM h-1). It confirmed the hypothesis that oxygen supply rate was limiting the reaction 323 at 30 mL scale. At the same time, the fact that no substrate accumulation was observed, enabled 324 further increase of the substrate dosing rate in the following experiments. 17 325 Final conditions under application of the fermented material for the demonstration 326 In order to obtain sufficient biocatalyst material for the planned reactions, a new 15 L TmCHMO 327 fermentation batch was performed. Characterization and comparison of the two batches was carried 328 out at 30 mL scale using 5% (v/v) loading of TmCHMO broths. A comparative table of the new (#2) and 329 the old (#1) batches can be found below (Table 1). 330 Table 1. Comparative table of results obtained with the different batches of TmCHMO. Enzyme load 5% (v/v) of TmCHMO 331 Broth 0.5% (v/v) of GDH-01; temperature 30°C; stirring rate 1200 rpm; air flow 16 mL min-1; pH 7; [TMCH] 30 mM h-1 (240mM 332 final); Methanol 1.44% (v/v) h-1 (10% (v/v) final); [D-glucose] 375 mM; [NADP+] 0.25 mM; titration solution 1 M NaOH. Biomass added Biocatalyst Yield Batch Biomass (gcww L-1) CHL (mmol) CHL (mg) (mgcww) (kg CHL kg-1cww) #1 412.7 619.1 6.33 989.2 1.60 #2 361.9 542.9 5.10 798.5 1.47 333 334 TmCHMO fermentation batch #1, used in all the experiments discussed in previous sections, was the 335 most productive one, presenting the highest biocatalyst yield and the highest biomass concentration. 336 However, due to lack of sufficient material (#1) to perform larger scale reactions, TmCHMO material 337 of batch #2 was used in the oncoming experiments. 338 A final reaction at 1 liter scale with TmCHMO fermentation batch #2 was performed (Figure 7) adding 339 0.2% (v/v) GDH-01 instead of 0.5% (v/v). The broth used in this reaction, as well as in the pilot plant, 340 was γ-ray radiated, which enabled the demonstration of this reaction in a pilot plant set-up without 341 GMO permit 22. 18 342 Figure 7. Reaction in 1 liter reactor using TmCHMO fermentation batch #2. The graphs show the reaction course of TMCH (white spots), CHL (black spots), NaOH addition (black line), substrate dosing (discontinuous line), mass balance (combined discontinuous spot-line-spot) and oxygen consumption (dotted line). Conditions: enzyme loading 10% (v/v) of TmCHMO Broth and 0.2% (v/v) of GDH-01; temperature 30°C; stirring rate 400 rpm; O2 flow 30 mL min-1; pH 7; [TMCH] 48 mM h-1 (237 mM final); Methanol 2.47% (v/v) h-1 (10% (v/v) final); [D-glucose] 375 mM; [NADP+] 0.25 mM; titration solution 5 M NaOH. 343 344 The yield after 7 h was 78% and 184 mmol of product were generated. Additional GDH-01 liquid 345 enzyme formulation was added to make sure this enzyme was not limiting the reaction. The lower 346 yield could then be attributed to the lower productivity of TmCHMO fermentation broth that had been 347 observed in the characterization (Table 1). Uncoupling (unproductive use of NADPH equivalents from 348 glucose, observed via NaOH consumption) was in line with the previous reaction at 1 liter scale: 20% 349 over-titration with batch #2. The cause, as previously explained, could be found on the acidification 350 caused by remaining bacterial growth or resting metabolism. The oxygen consumed, in contrast to 351 titration, was well aligned with the product synthesis. At the end, 98% of the totally consumed oxygen 352 was used for the target reaction. 353 Once the reaction was finished, 50 mL of methanol were added to solubilize the whole substrate and 354 product amount. After stirring for 10 minutes, a final sample was taken. The analyzed amounts are 19 355 represented as the two last isolated points in the graph. As it can be observed, the mass balance of 356 those samples is close to 100%, demonstrating a solubility issue of the reaction components when only 357 10% (v/v) methanol was used. 358 For the pilot plant run (100 L scale), full conversion was prioritized over higher product concentrations 359 and substrate accumulation was to be avoided due to the substrate inhibition that TmCHMO suffers. 360 Furthermore, low substrate concentration in the final mixture would facilitate the separation of the 361 formed CHL from the remaining TMCH. Therefore the applied substrate concentration was reduced to 362 180 mM and dosing rate was decreased to 36 mM h-1. 363 Demonstration on 100 L pilot plant scale 364 Reaction 365 The 200 liter reactor presented the same configuration and sensors as the 1 liter reactor, plus one 366 temperature probe inside the reactor. As stated above the final substrate concentration was reduced 367 to 180 mM and substrate dosing rate was lowered to 36 mM h-1. 368 As it can be seen in Figure 8, the reaction course was similar to that obtained at 1 liter scale with the 369 same TmCHMO batch (#2). Nevertheless, the reaction almost stopped after approximately 7 hours at 370 a yield of 82% (15.12 mol). This equals about 15% less product concentration compared to the 1 liter 371 reaction. A second portion of both enzymes was added aiming to increase the final conversion. At the 372 end (9 h), 85% of substrate was converted to trimethyl-ε-caprolactones. Meaning a total product 373 amount of 15.6 mol and a concentration of 24.4 g L-1. The uncoupling was higher compared to the 1 374 liter experiment of TmCHMO batch #2 (Figure 7). In this case 47% over-titration was obtained. An 375 excess of oxygen in the reactor could have contributed to this higher NaOH addition23. Oxygen 376 efficiency was lower in this reaction, 81% of the consumed oxygen was used in the target reaction. 20 377 Figure 8. Reaction in 100 liter reactor. Enzyme loading 10% (v/v) of TmCHMO Broth batch #2 and 0.2% (v/v) of GDH-01; temperature 30°C; stirring rate 150 rpm; O2 flow 3 L min-1; pH 7; [TMCH] 34.4 mM h-1 (183.35 mM final); Methanol 2.47% (v/v) h-1 (10% (v/v) final); [D-glucose] 375 mM; [NADP+] 0.5 mM; titration solution 5 M NaOH. 378 379 A comparative summary of the different reactions that were carried out with TmCHMO batches can 380 be seen below (Table 2). 381 Table 2. Comparative table of the reactions performed at 1 liter scale using the enzyme material of two different TmCHMO 382 batches and the 100 liter reaction using TmCHMO fermentation batch #2. The substrate dosing rate and oxygen flow were 383 the only conditions that differed between reactions. The dosing rate was 30 mM h-1 for batch #1, 48 mM h-1 for batch #2 (1 L) 384 and 34.4 mM h-1 for batch #2 (100 L). The oxygen flow was 20 mL min-1 for batch #1, 30 mL min-1 for batch #2 (1 L) and 3 L 385 min-1 for batch #2 (100 L). 386 *due to the lack of reliable product analysis, these values correspond to consumed substrate (conversion) and the space time 387 yield was calculated as if the reaction was stopped after 10 hours. Ferm. Reaction Biocatalyst yield Yield Space time [Product] Specific activity batch no. volume (L) (kg CHL kg-1cww) (%) yield (g L-1 h-1) (g L-1) (µmol min-1 g-1cww) #1* 1 0.83 92 3.4 34.4 13.3 21 #2 1 0.79 78 4.1 28.8 16.5 #2 100 0.60 85 2.7 24.4 12.8 388 389 Reaction with TmCHMO batch #1 resulted in the highest biocatalyst yield, as on 30 mL scale. However, 390 since the dosing rate was only 30 mM h-1, the reaction lasted 10 hours (according to titration) and 391 substrate addition dominated the reaction rate. That is why both, space time yield and specific activity, 392 were lower compared to the other experiments. On the other hand, batch #2 (1 L) presented two 393 significant variations: a dosing rate of 48 mM h-1 and an oxygen supply of 30 mL min-1. These 394 implementations allowed the reaction to proceed at higher rate compared with batch #1 and that is 395 why it resulted in the highest space time yield and specific activity among the three. Comparing the 396 two reactions with TmCHMO batch #2, a decrease in activity and yield can be observed at 100 liter 397 scale. This can mainly be explained by the lower substrate dosing rate as well as a poorer temperature 398 control (less surface area with respect to total volume) and inefficiency of the pH controller which 399 could not dose fast enough to maintain a stable pH (variation of ± 0.15). 400 Downstream processing 401 Once the reaction was finished, the procedure described before was followed to extract all remaining 402 substrate and product from the aqueous phase. 403 The final organic phase contained 2.55 kg (16.6 mol) of substrate and product, in the same ratio as it 404 was analyzed at the end of the reaction, 85% CHL product and 15% remaining TMCH substrate. 405 However, when the reaction was finished, 18.3 mol (CHL + TMCH) were present, meaning that around 406 9.3% yield was lost during the whole downstream procedure. The final water phase analyzed only 407 contained 0.2% of the compounds. Therefore, the filter cake was also analyzed and the samples taken 408 retained 15 g kg-1 (30.7 kg total cake weight) of the substrate and product, even after the applied three 409 washing steps. Hence, the 9.1% yield loss was related to the filtration step. The washing steps were 410 not as effective as anticipated due to small cracks in the filter cake. 22 411 Finally, the final organic phase containing all extracted material, was re-introduced in the reactor and 412 heated up to 42˚C. The applied pressure from the top of the condenser was 170 mbar. Ethyl acetate 413 could be evaporated at approximately 25 L h-1 and it did not contain any traces of substrate or product. 414 The concentrated final material (2.58 kg) contained 84.5% CHL (47% 1a and 53% 1b), 11.9% TMCH and 415 3.6% impurities. The overall isolated CHL yield of the reaction was 76%. 416 417 Conclusions 418 Optimization and scale up has been conducted in the Baeyer-Villiger oxidation of 3,3,5-timethyl- 419 cyclohexanone resulting in a final reaction that fulfills most of the target metrics evaluated in the 420 project for industrial feasibility: 95 - 100% conversion, 10-20 g L-1 of product concentration and 0.26 - 421 1.9 g g-1wcw biocatalyst yield. 422 TmCHMO biocatalyst formulation did not require any further down-stream operation units after 423 fermentation. GDH-01 can be produced in E. coli in high cell density fed-batch fermentations (InnoSyn 424 B.V.) and the required amount for an efficient process is comparatively low (0.05% (v/v) liquid enzyme 425 formulation). 426 Scale up of the reaction to 1 liter allowed the use of pure oxygen. The oxygen supply rate proved to be 427 a limiting factor for the reaction when air was used at 30 mL scale. Substrate dosing rate was increased 428 4-fold, which in consequence reduced the reaction time to 25% of the initially required time. All the 429 other parameters were not affected by the scale-up. 430 The biomass concentration of the TmCHMO broth and the enzyme content (activity) of this biomass 431 were crucial parameters to be taken into account. The use of a different TmCHMO batch (#2) reduced 432 the yield from 92% to 78% at 1 L scale. Fermentation processes had variable activity yields (not 433 discussed in this publication) and requires further work to determine whether the high activity of 434 fermentation #1 or the lower one of fermentation #2 represent the rule or the exception of the average 23 435 lies in the middle. This will result in a higher reproducibility and predictability of the biocatalytic 436 reaction batches and thus an improved reproducibility of the overall process. 437 Compared with previous results published by M Dolgove et al., it has been demonstrated that 438 TmCHMO broth could be used instead of CFE, reducing by this the associated costs of broth processing. 439 The cofactor regeneration could be performed by 0.05% (v/v) GDH-01 achieving full conversion. The 440 substrate dosing rate could be increased from 15 mM h-1 up to 48 mM h-1. According to titration, the 441 reaction time was reduced more than half maintaining conversions higher than 95%. The space time 442 yield was tripled, from 1.35 up to 4.1 g L-1 h-1. The reaction could be scale up to 100 liter. 443 As a conclusion, the feasibility of an industrial application of this reaction system was successfully 444 demonstrated, even though the final conversion at 100 liter scale was slightly lower. Major constraints, 445 arising if larger scales were considered, are not expected. The BVMO oxidation of cyclic ketones of 446 medium ring size such as cyclohexanone derivatives were demonstrated on pilot plant scale and are 447 ready for replication in industrial environments. 448 449 Acknwoledgments 450 The research for this work has received funding from the European Union project ROBOX (grant 451 agreement n° 635734) under EU’s Horizon 2020 Programme Research and Innovation actions H2020- 452 LEIT BIO-2014-1. This document reflects only the author's view and the Agency is not responsible for 453 any use that may be made of the information it contains. The Department of Chemical, Biological and 454 Environmental Engineering of Universitat Autònoma de Barcelona constitutes the Biochemical 455 Engineering Unit of the Reference Network in Biotechnology and the research group 2017 SGR 1462, 456 Generalitat de Catalunya. Authors thank COST Action CM 1303–Systems Biocatalysis for financial 457 support and Marie Delgove et al. from University of Maastricht for providing analytical standards. Jordi 458 Solé acknowledges also UAB for funding his Ph.D. grant. 24 459 460 References 461 (1) ten Brink, G.-J.; Arends, I. W. C. E.; Sheldon, R. A. The Baeyer−Villiger Reacbon: New 462 Developments toward Greener Procedures. Chem. Rev. 2004, 104 (9), 4105–4124. 463 https://doi.org/10.1021/cr030011l. 464 (2) Hollmann, F.; Arends, I. W. C. E.; Buehler, K.; Schallmey, A.; Bühler, B. Enzyme-Mediated 465 Oxidations for the Chemist. Green Chem. 2011, 13 (2), 226–265. 466 https://doi.org/10.1039/c0gc00595a. 467 (3) Sheldon, R. A.; Woodley, J. M. Role of Biocatalysis in Sustainable Chemistry. Chem. Rev. 2018, 468 118 (2), 801–838. https://doi.org/10.1021/acs.chemrev.7b00203. 469 (4) Alphand, V.; Carrea, G.; Wohlgemuth, R.; Furstoss, R.; Woodley, J. M. Towards Large-Scale 470 Synthetic Applications of Baeyer-Villiger Monooxygenases. Trends Biotechnol. 2003, 21 (7), 471 318–323. https://doi.org/10.1016/S0167-7799(03)00144-6. 472 (5) Baldwin, C. V. F.; Wohlgemuth, R.; Woodley, J. M. The First 200-L Scale Asymmetric Baeyer- 473 Villiger Oxidation Using a Whole-Cell Biocatalyst. Org. Process Res. Dev. 2008, 12 (4), 660–665. 474 https://doi.org/10.1021/op800046t. 475 (6) Hilker, I.; Wohlgemuth, R.; Alphand, V.; Furstoss, R. Microbial Transformations 59: First 476 Kilogram Scale Asymmetric Microbial Baeyer-Villiger Oxidation with Optimized Productivity 477 Using a Resin-Based in Situ SFPR Strategy. Biotechnol. Bioeng. 2005, 92 (6), 702–710. 478 https://doi.org/10.1002/bit.20636. 479 (7) Leisch, H.; Morley, K.; Lau, P. C. K. Baeyer-Villiger Monooxygenases: More Than Just Green 480 Chemistry. Chem. Rev. 2011, 111, 4165–4222. https://doi.org/10.1021/cr1003437. 481 (8) Bučko, M.; Gemeiner, P.; Schenkmayerová, A.; Krajčovič, T.; Rudroff, F.; Mihovilovič, M. D. 25 482 Baeyer-Villiger Oxidations: Biotechnological Approach. Appl. Microbiol. Biotechnol. 2016, 100 483 (15), 6585–6599. https://doi.org/10.1007/s00253-016-7670-x. 484 (9) Wu, J. T.; Wu, L. H.; Knight, J. A. Stability of NADPH: Effect of Various Factors on the Kinetics of 485 Degradation. Clin. Chem. 1986, 32 (2), 314–319. 486 (10) Romero, E.; Rub, J. Ø.; Mattevi, A.; Fraaije, M. W. Characterization and Crystal Structure of a 487 Robust Cyclohexanone Monooxygenase. Angew. Chemie - Int. Ed. 2016, 55, 15852–15855. 488 https://doi.org/10.1002/anie.201608951. 489 (11) Aalbers, F. S.; Fraaije, M. W. Coupled Reactions by Coupled Enzymes : Alcohol to Lactone 490 Cascade with Alcohol Dehydrogenase – Cyclohexanone Monooxygenase Fusions. Appl. 491 Microbiol. Biotechnol. 2017, 101, 7557–7565. https://doi.org/10.1007/s00253-017-8501-4. 492 (12) Pazmiæo, D. E. T.; Snajdrova, R.; Baas, B.; Ghobrial, M.; Mihovilovic, M. D.; Fraaije, M. W. Self- 493 Sufficient Baeyer – Villiger Monooxygenases : Effective Coenzyme Regeneration for 494 Biooxygenation by Fusion Engineering **. Angew. Chemie - Int. Ed. 2008, 47, 2275–2278. 495 https://doi.org/10.1002/anie.200704630. 496 (13) Delgove, M. A. F.; Elford, M. T.; Bernaerts, K. V; De, S. M. A. Application of a Thermostable 497 Baeyer – Villiger Monooxygenase for the Synthesis of Branched Polyester Precursors. J. Chem. 498 Technol. Biotechnol. 2018, 93 (8), 2131–2140. https://doi.org/10.1002/jctb.5623. 499 (14) Weckbecker, A.; Hummel, W. Glucose Dehydrogenase for the Regeneration of NADPH and 500 NADH. Microb. Enzym. Biotransformations 2005, 17 (8), 225–238. https://doi.org/10.1385/1- 501 59259-846-3:225. 502 (15) Zhou, J.; Ritter, H. Ring-Opening Polymerization of the Trimethylcaprolactone Isomers and 503 Investigations of Their Polymerizability. Macromolecules 2008, 41, 1663–1666. 504 https://doi.org/10.1021/ma701934r. 26 505 (16) Delgove, M. A. F.; Luchies, J.; Wauters, I.; Deroover, G. G. P.; De Wildeman, S. M. A.; 506 Bernaerts, K. V. Increasing the Solubility Range of Polyesters by Tuning Their Microstructure 507 with Comonomers. Polym. Chem. 2017, 8 (32), 4696–4706. 508 https://doi.org/10.1039/C7PY00976C. 509 (17) Zhou, J.; Ritter, H. Copolyesters as Non-Toxic Plasticizers. Polym. Int. 2011, 60 (8), 1158–1161. 510 https://doi.org/10.1002/pi.3122. 511 (18) Delgove, M. A. F.; Elford, M. T.; Bernaerts, K. V; Wildeman, S. M. A. De. Toward Upscaled 512 Biocatalytic Preparation of Lactone Building Blocks for Polymer Applications. Org. Process Res. 513 Dev. 2018, 22 (7), 803–812. https://doi.org/10.1021/acs.oprd.8b00079. 514 (19) Delgove, M. A. F.; Luchies, J.; Wauters, I.; Deroover, G. G. P.; Wildeman, S. M. A. De; 515 Bernaerts, K. V. Increasing the Solubility Range of Polyesters by Tuning Their Microstructure 516 with Comonomers †. Polym. Chem. 2017, 8, 4696–4706. https://doi.org/10.1039/c7py00976c. 517 (20) Solé, J.; Caminal, G.; Schürmann, M.; Guillén, M.; Alvaro, G. Co-Immobilization of P450 BM3 518 and Glucose Dehydrogenase on Different Supports for Application as a Self-Sufficient 519 Oxidative Biocatalyst. J. Chem. Technol. Biotechnol. 2018, 94 (1), 244–255. 520 https://doi.org/10.1002/jctb.5770. 521 (21) Zlochower, I. A.; Green, G. M. The Limiting Oxygen Concentration and Flammability Limits of 522 Gases and Gas Mixtures. J. Loss Prev. Process Ind. 2009, 22 (4), 499–505. 523 https://doi.org/10.1016/j.jlp.2009.03.006. 524 (22) Schmid, A.; Dordick, J. S.; Hauer, B.; Kiener, A.; Wubbolts, M.; Witholt, B. Industrial 525 Biocatalysis Today and Tomorrow. Nature 2001, 409 (January), 258–268. 526 (23) Holtmann, D.; Fraaije, M. W.; Arends, I. W. C. E.; Opperman, J.; Hollmann, F. The Taming of 527 Oxygen : Biocatalytic Oxyfunctionalisations. Chem. Commun. 2014, 50, 13180–13200. 27 528 https://doi.org/10.1039/C3CC49747J. 529 28