Directed evolution of unspecific peroxygenase in organic solvents

Fungal unspecific peroxygenases (UPOs) are efficient biocatalysts that insert oxygen atoms into nonactivated C–H bonds with high selectivity. Many oxyfunctionalization reactions catalyzed by UPOs are favored in organic solvents, a milieu in which their enzymatic activity is drastically reduced. Using as departure point the UPO secretion mutant from Agrocybe aegerita (PaDa‐I variant), in the current study we have improved its activity in organic solvents by directed evolution. Mutant libraries constructed by random mutagenesis and in vivo DNA shuffling were screened in the presence of increasing concentrations of organic solvents that differed both in regard to their chemical nature and polarity. In addition, a palette of neutral mutations generated by genetic drift that improved activity in organic solvents was evaluated by site directed recombination in vivo. The final UPO variant of this evolutionary campaign carried nine mutations that enhanced its activity in the presence of 30% acetonitrile (vol/vol) up to 23‐fold over PaDa‐I parental type, and it was also active and stable in aqueous acetone, methanol and dimethyl sulfoxide mixtures. These mutations, which are located at the surface of the protein and in the heme channel, seemingly helped to protect UPO from harmful effects of cosolvents by modifying interactions with surrounding residues and influencing critical loops.

(cyclo)aliphatic and heterocyclic hydroxylations; aromatic and aliphatic olefin epoxidations; sulfoxidations; N-oxidations; deacylations (C−C bond cleavages); ether cleavages (O-dealkylations); N-dealkylations; and halide oxidations/halogenations . However, the substrates of UPOs are often poorly water soluble, requiring reactions to be performed in organic solvents which, depending on their chemical nature and polarity, may negatively affect the enzyme activity and/or stability, impeding an efficient catalysis. Indeed, the loss of the essential water molecules of the protein shell that occurs when an enzyme is immersed in cosolvents may provoke conformational modifications leading to protein unfolding and denaturation; along with this, the potential competitive inhibitory effect of cosolvents is also an important disturbing factor (Doukyu & Ogino, 2010;Dutta Banik et al., 2016;Klibanov, 2001;Serdakowski & Dordick, 2008;Stepankova et al., 2013).
Protein engineering by directed evolution provides the means to tailor biocatalysts that tolerate organic solvents, with successful case studies having been reported for laccases, P450s, esterases, lipases or proteases, to name just a few (Chen & Arnold, 1993;Moore & Arnold, 1996;Song & Rhee, 2001;Takahashi et al., 2005;Wong et al., 2004;Zumarraga et al., 2007). In previous studies, we evolved the UPO from the basidiomycetous fungus Cyclocybe (Agrocybe) aegerita (AaeUPO) for heterologous functional expression in yeasts, as well as for diverse applications ranging from the synthesis of agrochemicals to that of pharmaceutical compounds (Hobisch et al., 2020;Molina-Espeja et al., 2017 and references herein). The goal in the current study was to obtain AaeUPO mutants by directed evolution that were active and stable in organic cosolvents. To improve the enzyme activity in organic cosolvents, acetone, acetonitrile (ACN) and dimethyl sulfoxide (DMSO) were used during the screening (organic cosolvents of different polarities and chemical nature), while selective pressure was controlled through the gradual enhancement of cosolvent concentrations. Beneficial mutations introduced by conventional directed -adaptiveevolution in Saccharomyces cerevisiae (i.e., random mutagenesis and in vivo DNA shuffling coupled to the selection of the fittest) were recombined with a set of neutral mutations taken from a former genetic drift study in which we found several UPO mutants with a noticeable activity improvement in cosolvents (Martin-Diaz et al., 2018). The final variants of this directed evolution process were characterized biochemically, and they show a notable improvement in activity and stability in the presence of high concentrations of organic cosolvents. Ultimately, the effects of these beneficial mutations are considered in the context of the enzyme' structure.

| Directed evolution strategy
The departure point for this study was PaDa-I, an evolved secretion mutant of AaeUPO that carries the mutations F12Y-A14V-R15G-A21D-V57A-L67F-V75I-I248V-F311L (the mutations underlined lie in the signal peptide). While the secretion and activity in aqueous media of PaDa-I improved, its general activity in the presence of high concentrations of organic solvents is poor, like that of the wildtype (wt) AaeUPO (Molina-Espeja et al., 2014). To improve the activity of PaDa-I in organic solvents, two generations of directed-adaptiveevolution were carried out by random mutagenesis (epPCR) and in vivo DNA shuffling. Additionally, a final cycle of in vivo site-directed recombination (SDR) was performed aimed at recombining neutral mutations previously discovered by genetic drift that increased activity in the presence of organic solvent (Martin-Diaz et al., 2018) ( Figure 1).
To promote activity in diverse types of organic solvents, mutant libraries were screened in the presence of cosolvents of different chemical nature and polarities, with decreasing logP values of −0.24, −0.34, and −1.3, respectively: acetone, ACN and DMSO. The selective pressure was regulated by gradually augmenting the concentration of the organic solvents in each round of directed evolution (see Section 5 for details), establishing a screening threshold based on the percentage of cosolvent at which the parental type retained 1/3 of its activity in aqueous medium. The main selection criterion used was referred to as the tolerance to the organic cosolvent (i.e., the activity retained in cosolvents), represented as the ratio of the activity in the presence of organic solvent to that in the absence of organic solvents (expressed as a percentage). After screening over 7000 clones in two generations of directed evolution, the best selected variant was the 18F6 mutant that carried the new mutations T197A-T198A-V244A-K290R. This variant had moderately improved C 50 values (the concentration of cosolvent at which the enzyme maintains 50% of its corresponding activity in aqueous solution) of 14%, 10%, 9.5%, and 7% (vol/vol) for acetone, ACN, methanol and DMSO, respectively, as opposed to 10%, 7%, 8.4%, and 2% (vol/vol) of the parental type PaDa-I. These numbers correlate with a C 50 increase over PaDa-I variant of 1.41-, 1.44-1.1-, and 3.65fold for acetone, ACN, methanol and DMSO, respectively.
In a previous work, we carried out a directed UPO evolution experiment by neutral genetic drift, an engineering strategy that allows to introduce neutral mutations, that is, mutations that are neutral in terms of the natural function of the enzyme, but whose gradual accumulation can open new adaptive routes enabling promiscuous activities and stabilities to be displayed. In this neutral genetic drift campaign performed on PaDa-I, we identified several neutral mutations (F191L, S226G, Q254R, S272P, A317D) that enhanced UPO´s activity in cosolvents, despite the fact it was not applied any selection pressure to increase activity in cosolvents but to maintain native activity in aqueous solution (Martin-Diaz et al., 2018). With the aim of fostering epistatic/synergetic effects, this set of neutral mutations was introduced into the 18F6 variant and recombined by in vivo SDR. SDR is based on the homologous DNA recombination apparatus of Saccharomyces cerevisiae, allowing mutant libraries to be rapidly constructed and screened by precisely evaluating the effect of mutations/reversions at the positions targeted in a combinatorial manner (Viña-Gonzalez & Alcalde, 2020) ( Figure S1). Approximately, 70% of the clones of the SDR mutant library (library size 3500 clones) were functional variants in the presence of cosolvents, possibly a consequence of the beneficial effects of neutral mutations on the whole mutagenic population. We selected the 13 most promising mutants from this process for a preliminary characterization (Table 1). The best mutant of this set was the WamPa variant, with C 50 values of 25%, 27%, 14%, and 16% (vol/vol) for acetone, ACN, methanol and DMSO, respectively, as opposed to the values of 10%, 7%, 8.4% and 2% (vol/vol) for the parental PaDa-I. The observed increase in the C 50 numbers seems to be inversely proportional to the logP of the cosolvents tested (−0.24, −0.34, −0.69, and −1.3 for acetone, ACN, methanol and DMSO, respectively); yet, we cannot establish a clear relationship between the water miscibility of cosolvents and the activity of the enzyme in their presence, taking into account that methanol and DMSO can act as substrate and inhibitor of UPO, respectively. We suspected that such relevant increased of activity in cosolvents had to be related with unique combinations of beneficial mutations from the directed evolution experiment in cosolvents with the neutral ones from the genetic drift campaign. After sequence analysis, we verified that the mutational scaffold of the 18F6 variant (T197A-T198A-V244A-K290R) was maintained in all the final clones, whereas the five neutral mutations were found in different combinations and the mutations S226G, A317D, Q254R, S272P, and F191L were overrepresented in 11,9,8,8, and 3 variants, respectively ( Figure 2). Surprisingly, WamPa included an additional beneficial mutation, G318R, introduced due to an error in the PCR amplification. Indeed, the same sequence but without the G318R mutation is present in the 1G9 mutant, which had a lower C 50 than WamPa but better than that of the parental PaDa-I variant (Table 1 and Figure 2).

| Biochemical characterization of evolved UPOs
The parental PaDa-I, the 18F6 mutant from second generation and the three best variants from the SDR library (WamPa, 27D5, and 20C4) were produced and purified to homogeneity (Reinsheitszahl F I G U R E 1 Evolution pathway for activity in the presence of organic cosolvents. New mutations are shown as stars, while the previous accumulated mutations are depicted as rectangles. Mutations introduced along the adaptive evolution process are in blue and those in red were obtained by neutral genetic drift. The mutation in green was introduced by a mis-step during PCR amplification [Color figure can be viewed at wileyonlinelibrary.com] value [Rz]~2). The activity of these pure enzymes was determined in the presence of organic cosolvents and compared to that in aqueous medium. To rule out any possible bias towards the substrate used in the high-throughput screening (HTS) assay (i.e., 2,2′-azino-bis(3ethylbenzothiazoline-6-sulfonic acid [ABTS], the colorimetric substrate for peroxidase activity), their activity on veratryl alcohol was also measured (a substrate for peroxygenase activity) ( Table 2).
Regardless of the substrate, WamPa exhibited a high activity in diverse organic cosolvents retaining approximately 7%, 30%, 12%, and 25% of its activity on ABTS in 30% acetone, 30% ACN, 30% methanol and 15% DMSO, as opposed to the 2%, 1%, 6%, and 4% activity of the parental PaDa-I. The 20C4 and 27D5 mutants also outperformed the PaDa-I variant in organic cosolvents, yet to a lesser extent than WamPa. All the variants showed similar tolerance (retained activity in organic solvents) with veratryl alcohol as a substrate, yet these values were generally higher than those obtained with ABTS as a consequence of the amount of enzyme needed for each activity assay (with K m values of 8 and 0.05 mM, for veratryl alcohol and ABTS, respectively). UPO variants were also very stable at high concentrations of organic cosolvents ( Figure 3). After 24 h in the presence of 60% (vol/vol) of the organic cosolvents, the majority of the variants retained at least 50% of their activity, and in some cases hyperactivation was detected over short incubation times.

| MUTATIONAL ANALYSIS
The WamPa variant carries nine beneficial mutations that improved its activity and stability in organic solvents (T197A, T198A, S226G, V244A, Q254R, S272P, K290R, A317D, and G318R: the mutations underlined were derived from the directed-adaptive-evolution pathway whereas the rest are neutral mutations included by SDR) (Figure 1). It is notable that three of the mutations introduced by adaptive evolution (T197A, T198A, and K290R) were also found in the aforementioned neutral genetic drift campaign (Martin-Diaz et al., 2018). Rather than serendipity, finding the same mutations using different experimental strategies (i.e., neutral drift vs. adaptive evolution) indicates that both approaches overlap with each other when targeting common biochemical traits. To rationalize the effect of directed evolution, the mutations of the WamPa variant were mapped onto the crystal structure of PaDa-I (PDB accession number 5OXU) ( Figure 4). Most of the changes were distributed at the enzyme's surface and remarkably, T A B L E 1 C 50 values for the different variants of the evolution route

Mutant
Library creation Generation The improvement is defined as the ratio of the activity of the corresponding mutant under the conditions specified to that of the parental PaDa-I under the same conditions. d Tolerance in organic cosolvents (i.e., retained activity in cosolvents) is defined as the ratio of the activity in the presence of organic cosolvents to that in the absence of organic cosolvents, given as a percentage. N.m. nonmeasurable-due to the strong background generated by acetone 30% (vol/vol).

| Surface mutations
The neutral mutations S226G, Q254R, S272P (introduced by SDR) and K290R mutation (introduced by adaptive evolution) were all located at the surface of the protein. The introduction of proline and arginine residues is typically linked with more rigidity and stabilization due to the distinctive cyclic structure of proline's side chain and arginine's positively charged guanidinium group (Doukyu & Ogino, 2010;Lehmann et al., 2020). In particular, the Q254R substitution replaces a neutral (polar) residue with a basic (chargeable) one and according to our model, this mutation could establish a new salt bridge with Asp273 from an adjacent loop that might strengthen this region (Figure 5c,d). The K290R mutation seems to break an Hbond with the Pro324 located in an adjacent helix, concomitantly forming a new H-bond with Asn286 (Figure 5e,f). This local rearrangement seems to produce tighter packing of the surroundings, marked as labile by B-factor analysis. The S226G mutation was previously thought to improve thermostability in a structure-guided evolution project (Mate et al., 2017). This substitution is located at a heme Fe 3+ distance of 12 Å and no effects were observed in the modeling, although replacing a larger, polar alcohol residue (hydroxymethyl group, CH 2 -OH) by the smallest nonpolar residue (hydrogen, H) will surely increase hydrophobicity and may imply tighter packing (Figure 5g,h).   The NucleoSpin plasmid kit was purchased from Macherey-Nagel.

| MATERIALS AND METHODS
Oligonucleotides were synthesized by Metabion. All chemicals were of reagent-grade purity.
To prepare minimal medium for plates, 100 ml 20% glucose was used instead of 20% raffinose, and 20 g/L of bacto agar were used. Selective

| DIRECTED EVOLUTION
Three rounds of directed evolution were performed. pJRoC30 was linearized with BamHI and XhoI restriction enzymes. The linearized vector was cleaned, concentrated, loaded onto a low-melting-point preparative agarose gel, and purified using the Zymoclean gel DNA recovery kit.
Then, the introduction of genetic variability was carried out as described below for each generation, and the PCR products were purified using the  Table S1.

| First generation
The PaDa-I variant was used as parental type. Genetic variability was introduced by error-prone PCR (epPCR) using Taq  Bio-Rad) were: 95°C for 2 min (1 cycle); 94°C for 45 s, 55°C for 30 s, and 74°C for 90 s (28 cycles); and 74°C for 10 min (1 cycle). PCR products were mixed with linearized plasmid as described above and subjected to DNA shuffling and cloning upon transformation in yeast.

| Second generation
The best clone obtained in the first generation, 10D1, was subjected to ep-PCR. The mutagenic rates, the PCR conditions, and the thermal-cycling program employed were the same as those described for the first generation. PCR products were mixed with linearized plasmid as described above and subjected to DNA shuffling and cloning upon transformation in yeast.

| Third generation
The best mutant obtained in the second generation, 18F6, was subjected to SDR in vivo with a palette of neutral mutations (Martin-Diaz et al., 2018). SDR was conducted as reported elsewhere with minor modifications (Viña-Gonzalez & Alcalde, 2020), Figure S1: Six 10 | DNA SEQUENCING UPO genes were sequenced by GATC-Eurofins Genomics (Germany).

| PROTEIN MODELING
The