Photo-Regulation of Enzyme Activity: The Inactivation of a Carboligase with Genetically Encoded Photosensitizer Fusion Tags

Genetically encoded photosensitizers are able to produce reactive oxygen species upon illumination and are exploited in a wide range of applications, especially in the medical field. In this work, we envisioned to further apply these genetically encoded photosensitizers for the light-dependent control of single enzymes in multi-step biocatalysis. One of the challenges in the application of several enzymes in a cascade is the unwanted cross-reactivity of these biocatalysts on reaction intermediates when all enzymes are simultaneously present in the reaction. As one strategy to address this issue, we investigated whether the introduction of genetically encoded photosensitizers as fusion tags would allow the selective inactivation of enzymes after successful transformation by simply turning on light. We tested five different photosensitizers as molecular biological fusion tags to inactivate the pyruvate decarboxylase variant E469G/W543H from Acetobacter pasteurianus. Dimeric photosensitizer tags, like the flavin-binding fluorescent proteins from Bacillus subtilis and Pseudomonas putida showed the tendency to form insoluble protein aggregates in combination with the tetrameric carboligase. Enzyme activity was, to some extent, retained in these aggregates, but the handling of the insoluble aggregates proved to be unfeasible. Monomeric photosensitizer tags appeared to be much more suitable when fused to the tetrameric enzyme. In the dark, the singlet oxygen photosensitizing protein (SOPP3)-tagged carboligase retained 79% of its activity as compared to the unfused enzyme. Upon blue light exposure, the SOPP3 tag showed the best specific inactivation and enabled complete inactivation of the carboligase within 30 min. SOPP3 is thus seen as a promising photosensitizer tag to be applied in future multi-step enzyme cascades to overcome the challenge of cross-reactivity.


INTRODUCTION
Photosensitizers have the unique ability to locally produce short-lived reactive oxygen species (ROS) when irradiated with light of a specific wavelength (Jacobson et al., 2008). When a photosensitizer is bound to a particular ligand or antibody, the highly reactive ROS enable targeted inactivation of proximate biological structures as proteins, lipids, DNA and carbohydrates in a process reported as chromophoreassisted light inactivation (CALI) (Liao et al., 1994). Upon excitation of a photosensitizer in the presence of molecular oxygen, ROS generation can occur at the functional chromophore via two types. In the type I mechanism, an electron is transferred to oxygen leading to the formation of the superoxide radical anion, which can further react to hydrogen peroxide (H 2 O 2 ) and the hydroxyl radical. The type II pathway involves an energy-transfer to molecular oxygen resulting in the production of singlet oxygen ( 1 O 2 ) (Abrahamse and Hamblin, 2016). Because of its energetically unfavorable state, 1 O 2 is highly reactive resulting in a lifetime of~4 µs and a diffusion range of~200 nm (Moan, 1990;Skovsen et al., 2005;Ogilby, 2010). In contrast, hydrogen peroxide, shows a lifetime of~1 ms and thus can diffuse over distances up to 5 µm (Reth, 2002). Originally, dyes as malachite green (Jay, 1988), fluorescein (Surrey et al., 1998) or curcumin (Bernd, 2014) were used to perform CALI. In a more recent approach, genetically encoded photosensitizers are used for the light-induced inactivation of proteins or cells. These types of photosensitizers can be produced within a living system and do not have to be added exogenously (Bulina et al., 2006a). So far, CALI has been used in various applications as in photodynamic therapy (PDT) for the destruction of cancer cells or diseased tissue (Robertson et al., 2009), for the modulation of specific protein activities (Surrey et al., 1998) or the light-driven inactivation of microbial cells (Endres et al., 2018;Hilgers et al., 2019;Hally et al., 2020). By fusing photosensitizers to target proteins, the efficient application of CALI was demonstrated in diverse approaches in vivo and in vitro (Takemoto, 2021). Another promising application for CALI would be the area of multi-step biocatalysis. With multiple enzymes used in a cascade, the synthesis of complex molecules from simple and optimally renewable starting compounds, is possible (Lopez-Gallego and Schmidt-Dannert, 2010;Erdmann et al., 2017). As enzymes often show very high chemo-and stereoselectivity with regard to product formation, they are very atom efficient catalysts, especially for the formation of chiral pharmaceutical compounds and fine chemicals (Ricca et al., 2011). However, with respect to the substrate range, many enzymes do not only accept one substrate but a number of natural and synthetic substrates. This behavior, also called promiscuity, gives enzymes immense strength for synthetic purposes (Bornscheuer and Kazlauskas, 2004). At the same time, this promiscuity can be a challenge in syntheses with multiple enzymatic steps, as an enzyme might not only bind its desired substrate, but could also accept one of the intermediates or the final product when present at the same time in the same reaction vessel. Undesired side-activity decreases the atom efficiency and increases the complexity of the reaction (Hult and Berglund, 2007). To overcome the issue of cross-reactivity, the enzyme activity has to be regulated in space, by separating the enzymatic steps in compartments, or in time, by adding the enzymes sequentially and removing the respective enzyme after successful transformation, which can be challenging and cost-intensive (Erdmann et al., 2017;Gruber et al., 2017;Claaßen et al., 2019). CALI could represent a suitable tool to selectively turn off enzyme activity in multistep biocatalysis by light when appropriate genetically encoded photosensitizers are fused to the target enzymes. Through genetic fusion, the photosensitizers are located closest to the target enzyme. This allows the CALI principle to be carried out in the most efficient way, since the high phototoxicity of the short-ranged ROS formed, especially singlet oxygen, can be optimally utilized (Lin et al., 2013). Detailed characterization of ROS species formed can be found in the literature (Pletnev et al., 2009;Takemoto et al., 2013;Wingen et al., 2014;Westberg et al., 2015Westberg et al., , 2017Endres et al., 2018;Hilgers et al., 2019;Onukwufor et al., 2020). Light is an attractive external stimulus, as it provides an orthogonal, noninvasive regulation approach with high spatial and temporal resolution (Willner and Rubin, 1996;Cruz et al., 2000;Mayer and Heckel, 2006;Lu et al., 2014). With CALI, the enzymes of a multi-step cascade could be inactivated after successful transformation, thus, cross-reactivity could be avoided (Claaßen et al., 2019). A disadvantage is that enzyme inactivation by CALI is not reversible, decreasing the economic efficiency of this concept. ROS exposure causes the proteins to suffer oxidative damage which results in amino acid modification, cleavage of the polypeptide chain and conversion of the proteins to derivatives (Stadtman, 2006;Sharma et al., 2012;Vatansever et al., 2013). Here, we tested CALI on the pyruvate decarboxylase variant E469G/W543H from Acetobacter pasteurianus (ApPDC-EGWH). This carboligase variant is a (S)-selective enzyme with the ability to produce phenylacetylcarbinol (PAC) from the inexpensive substrates benzaldehyde and pyruvate (Sehl et al., 2017). PAC, in turn, is a versatile compound which serves as a building block for pharmaceutically active molecules as nor(pseudo)ephedrines (Sehl et al., 2013(Sehl et al., , 2014. The variant ApPDC-EGWH has demonstrated high conversion of 3-hydroxybenzaldehyde in particular (Figure 1), which is a starting substrate for the formation of substituted tetrahydroisoquinolines (Erdmann et al., 2017;Sehl et al., 2017). To demonstrate the applicability of genetically encoded photosensitizers as light-responsive regulators for enzyme activities in biocatalysis, we created and purified five ApPDC-EGWH fusion enzymes carrying different photosensitizer domains. First, we compared the carboligation activity of the fusion enzymes in the dark to the untagged ApPDC-EGWH. Here, we aimed to exclude a negative influence of the tagged photosensitizers on the enzyme activity when not exposed to light. Subsequently, the inactivation potential of each fusion enzyme was determined, to evaluate which photosensitizer tag is most suitable for selective carboligase inactivation and therewith potentially suppressing cross-reactivity in a multi-step synthesis approach.

Vector Construction and Transformation
A pET28a vector (Merck, Darmstadt, Germany) was used for the construction of photosensitizer fusion genes. Details on the ApPDC-EGWH and the LbADH (alcohol dehydrogenase from Lactobacillus brevis) construct in a pET21a vector can be found in (Sehl et al., 2017) and in (Kulig et al., 2012), respectively and in the Supplementary Material. Information on the photosensitizers used as fusion tags (SOPP, SOPP3, EcFbFP, Pp2FbFP and SuperNova) are given elsewhere (Wingen et al., 2014;Endres et al., 2018;Hilgers et al., 2019) and in the Supplementary Material. Primers with specific overhangs for the cloning of the respective genes including sequences encoding N-terminal His 6 -tags were designed and ordered from Eurofins Genomics (Ebersberg, Germany). The expression plasmids were constructed with the Gibson Assembly ® Cloning Kit (NEB, Frankfurt Germany) and verified by sequencing (LGC Genomics, Berlin, Germany). E. coli BL21(DE3) cells (Merck, Darmstadt, Germany) were transformed with the respective expression plasmids by adding 1 µl plasmid solution (100 ng μL −1 ) to 100 µl bacterial solution (OD 600 =~12, in 80 mM CaCl 2 , 20% glycerol). For pET28a-SOPP3-ApPDC-EGWH only, E. coli Tuner ™ (DE3) cells (Merck, Darmstadt, Germany) were used. After 30 min incubation on ice, a heat shock was performed at 42°C for 40 s and the cells were stored on ice for 2 min. 900 μl S.O.C. medium (Thermo Fisher Scientific, Waltham, MA, United States) was added and the cells were incubated for 1 h at 37°C and 700 rpm in a thermomixer (comfort 5335r, Eppendorf, Hamburg, Germany). Finally, the cells were plated on lysogeny broth (LB) agar (with 50 μg ml −1 kanamycin, only for cells transformed with pET21a-ApPDC-EGWH 100 μg ml −1 ampicillin was used) and incubated overnight at 37°C.

Protein Production
Proteins were produced in shaking flasks with a filling volume of up to 15%. A single colony from the respective overnight plate was transferred to 50 ml LB medium (with 50 μg ml −1 kanamycin, for cells transformed with pET21a-ApPDC-EGWH 100 μg ml −1 ampicillin was used) and the preculture was cultivated overnight at 37°C and 150 rpm (Infors HT, Bottmingen, Switzerland). For the main cultures, LB medium was used except for the production of ApPDC-EGWH, LbADH and Pp2FbFP-ApPDC-EGWH which was done in auto induction (AI) medium. Details on the cultivation media are provided in the Supplementary Material. During production, cells containing light-sensitive proteins were handled under exclusion of light. LB cultivations lasted 24 h (induction with 0.5 mM IPTG, for E. coli Tuner ™ (DE3) cells with 0.1 mM IPTG) and AI cultures were incubated for 48 h. Cultivations were started at 37°C (decreased to 20°C after 2 h) and 150 rpm (Infors HT, Bottmingen, Switzerland). After harvesting (Avanti J-20 XP, Rotor JLA-8.1000, Beckman Coulter, Brea, CA, United States; 30 min, 8,000 rpm, 4°C), cells containing light-sensitive proteins were stored under exclusion of light.

Protein Purification and Storage
For the purification and storage of light-active fusion enzymes, light exposure was limited to an absolute minimum. Therefore, columns used for purification as well as the lyophilizer and respective solutions were wrapped with aluminium foil and windows were covered. The untagged ApPDC-EGWH was purified and stored under normal light conditions (electric and sunlight). Frozen cells containing respective enzymes were thawed (30% (w/v)) in TRIS buffer ( (Bradford, 1976), the respective buffer was used as blank.
After freezing the protein solution overnight, it was submitted to lyophilization (Alpha 1-4 LD Plus, Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany) for 3 d. After lyophilization, the light-sensitive fusion constructs were stored under exclusion of light. In case of the insoluble EcFbFP-ApPDC-EGWH and Pp2FbFP-ApPDC-EGWH, the pellets obtained after cell disruption and centrifugation were subjected to two washing steps with TRIS buffer (50 mM, pH 7.5, 2.5 mM MgSO 4 , 0.1 mM ThDP) and lyophilization was applied as described above.

Activity Tests with Photosensitizer Fusion Enzymes
Initial rate determination (without light exposure) of the photosensitizer-tagged and the untagged carboligase was performed using the reaction from 3-hydroxybenzaldehyde to (S)-3-hydroxyphenylacetylcarbinol ((S)-3-OH-PAC) ( Figure 1). The transformations were performed on 1 ml scale in 50 mM HEPES buffer (pH 7.5) containing 2.5 mM MgSO 4 , 0.1 mM ThDP, 40 mM 3-hydroxybenzaldehyde and 200 mM pyruvate (sodium salt). The enzymes SOPP-ApPDC-EGWH (5.0 mg ml −1 ), SOPP3-ApPDC-EGWH (0.3 mg ml −1 ), EcFbFP-ApPDC-EGWH (24.8 mg ml −1 ), Pp2FbFP-ApPDC-EGWH (2.47 mg ml −1 ) and SuperNova-ApPDC-EGWH (2.0 mg ml −1 ) were applied in a purified, lyophilized formulation. Protein concentrations where determined using the Bradford assay (Bradford, 1976). Samples were incubated for a duration of 30 min at 22°C and 850 rpm in a thermomixer (comfort 5335r, Eppendorf, Hamburg, Germany). Initial rate determinations were performed under dark conditions in 1.5 ml amber-colored reaction tubes. The activity test of SOPP3-ApPDC-EGWH (0.5 mg ml −1 ; purified and lyophilized) at different light conditions (details are given in Table 2) was done for the same reaction conditions except the exposure time was set to 60 min. Incubation in illuminated setups was done using a stirred 1.5 ml quartz glass cuvette (Hellma GmbH & Co. KG, Müllheim, Deutschland; 600 rpm). The light intensities of the blue and orange LED and of the normal light conditions were determined at room temperature using the PM100D energy meter with an S302C sensor (Thorlabs, Newton, NJ, United States). The sensor was positioned at the exact position of the reaction setup (instead of the cuvette) and aligned with the respective light source. Intensities of the blue and orange LED and of the normal light conditions were calculated using the pre-installed sensor software by setting 450 nm or 610 nm as the calibration wavelength, respectively. A detailed description of the light intensity measurement is given in the Supplementary Material (Supplementary Figure S3). Unless stated otherwise, experiments were carried out as three technical replicates. The enzymatic reaction in samples taken during the carboligase reactions was quenched 1:20 (v/v) in 100% acetonitrile. Product formation was analyzed via HPLC which is described in the following. The activity is given in U µmol enzyme −1 , which is defined as the amount of enzyme in µmol, which catalyzes the formation of 1 μmol (S)-3-OH-PAC per minute. The activity test with soluble crude cell extract and the insoluble protein fraction of the EcFbFP-ApPDC-EGWH production ( Figure 5A) was done for the reaction from benzaldehyde to (S)-phenylacetylcarbinol ((S)-PAC). For the activity test with soluble crude cell extract and the insoluble protein fraction of the Pp2FbFP-ApPDC-EGWH production ( Figure 5B), the above mentioned reaction from 3-hydroxybenzaldehyde to (S)-3-OH-PAC was Frontiers in Catalysis | www.frontiersin.org March 2022 | Volume 2 | Article 835919 applied. In both cases, the amount of lyophilized soluble and insoluble protein was equivalent to 500 mg wet cell weight.

Inactivation of Photosensitizer Fusion Enzymes
Inactivation experiments (with blue/orange light exposure) were carried out in a stirred 1.5 ml quartz glass cuvette (Hellma GmbH & Co. KG, Müllheim, Deutschland; 600 rpm). Inactivation assays were performed for the above mentioned reaction from 3hydroxybenzaldehyde to (S)-3-OH-PAC, except that 50 mM HEPES buffer (pH 6.5) was used. The enzymes SOPP-ApPDC-EGWH (3.0 mg ml −1 ), SOPP3-ApPDC-EGWH (1.0 mg ml −1 ), EcFbFP-ApPDC-EGWH (3.6 mg ml −1 ), Pp2FbFP-ApPDC-EGWH (4.0 mg ml −1 ) and SuperNova-ApPDC-EGWH (1.0 mg ml −1 ) where applied in a purified, lyophilized formulation. Protein concentrations where determined using the Bradford assay (Bradford, 1976). A single blue LED (λ max = 450 nm, royal blue, XP-E2 SMD-LED, Star-PCB, Cree, Durham, NC, United States;~60 mW cm −2 ) was applied for the illumination of fusion enzymes. Orange light (λ max = 610 nm, amber, NCSA219B-V1 SMD-Star-PCB, Nichia, Anan, Japan;~40 mW cm −2 ) was used only for the SuperNova-tagged fusion enzyme. Further specifications on the LEDs are given in the Supplementary Material (Supplementary Table S7). The illumination setup was cooled with ice to compensate for a large temperature increase, nevertheless, during 30 min of LED illumination the temperature gradually increased from 20°C to 24°C. For the sake of comparison, the corresponding dark control was exposed to the same temperature changes. The enzymatic reaction in samples taken during the carboligase reactions was quenched 1:20 (v/v) in 100% acetonitrile. Product formation was analyzed via HPLC as described in the following. Control experiments with the ApPDC-EGWH in blue and orange light are shown in the Supplementary Material (Supplementary Figure S2A,B).

Inactivation Effect of Photosensitizer Fusion Enzymes on an Additional Alcohol Dehydrogenase in the Same Reaction Solution
Initial rate assays with the LbADH were performed using the transformation from benzaldehyde to benzyl alcohol   (Figure 9.). The enzymatic reactions were performed on 0.5 ml scale in a 50 mM potassium-phosphate buffer (pH 7) containing 1 mM MgCl 2 , 10 mM NADPH, 10mM benzaldehyde and 0.01 mg ml −1 LbADH. Lyophilized LbADH after purification was obtained from elsewhere (Kulig et al., 2012). Protein concentrations where determined using the Bradford assay (Bradford, 1976). Samples were incubated for 2 min at 22°C and 850 rpm in a thermomixer (comfort 5335r, Eppendorf, Hamburg, Germany). Experiments were carried out as three technical replicates. For light experiments with carboligase fusion enzymes, 0.1 mg ml −1 LbADH and 1 mg ml −1 of the respective purified and lyophilized fusion enzymes were illuminated for 10 min in 50 mM potassium-phosphate buffer (pH 7, with 1 mM MgCl 2 ) prior to the enzymatic reaction, using blue or orange LEDs as described above. Then, the initial rate of the respective illuminated LbADH was determined. The enzymatic reaction in samples taken during LbADH reactions was quenched 1:9 (v/v) in a solution of 50% acetonitrile and 50% water (v/v). Product formation was analyzed via HPLC which is described in the following.

HPLC Analytics
Samples were analyzed using a 1260 Infinity Quaternary LC system (Agilent Technologies, Inc., Santa Clara, CA, United States) equipped with an 1260 diode array detector and a Chiralpak IE column (85325, Daicel, Osaka, Japan). For carboligase samples, analysis was carried out isocratically with a flow rate of 0.7 ml min −1 at 20°C for 10 min using a solvent mixture of acetonitrile and water (25:75; v/v). 5 μL Samples were injected and the absorption of (S)-3-OH-PAC was detected at 220.4 nm with a retention time of 6.2 min, while (R)-3-OH-PAC was detected with a retention time of 6.85 min. The absorption of 3-hydroxybenzaldehyde was detected at 220.4 nm with a retention time of 8.8 min. For LbADH samples, analysis was carried out isocratically with a flow rate of 1 ml min −1 at 20°C for 10 min using a solvent mixture of acetonitrile and water (50:50; v/v). 10 μL Samples were injected and the absorption of benzyl alcohol was detected at 215 nm with a retention time of 4.1 min. Calibrations with (S)-3-OH-PAC, 3-hydroxybenzaldehyde and benzyl alcohol are shown in the Supplementary Material (Supplementary Figure S1).

Construction of Photosensitizer Enzyme Fusions
In a first step, the ApPDC-EGWH was fused to five genetically encoded photosensitizers resulting in corresponding photosensitizer-enzyme fusion constructs (Figure 2.). Mainly, flavin mononucleotide-binding fluorescent proteins were tested as fusion tags, which originate from blue light receptors of light-oxygen voltage (LOV) domains from plants or bacteria (Table 1) (Drepper et al., 2007;Shu et al., 2011). These are the singlet oxygen photosensitizing proteins SOPP and SOPP3 (Westberg et al., 2017), the flavin-binding fluorescent protein (FbFP) from Bacillus subtilis codon-optimized for Escherichia coli (EcFbFP) and the FbFP from Pseudomonas putida (Pp2FbFP; Wingen et al., 2014). In addition, one member of the green fluorescent protein (GFP)-like photosensitizer family (SuperNova) was used (Takemoto et al., 2013). SOPP3 is one of the strongest genetically encoded singlet oxygen producers currently known (Westberg et al., 2017), but it has been shown that SOPP3 is conditionally able to produce very high amounts of H 2 O 2 as well . While EcFbFP produces mainly H 2 O 2 and only small quantities of 1 O 2 , Pp2FbFP is producing higher amounts of 1 O 2 than H 2 O 2 (Endres et al., 2018). For SuperNova it is suggested that predominantly H 2 O 2 via the type I pathway is produced (Pletnev et al., 2009;Hilgers et al., 2019). Other sources report, that both types of ROS, but in comparison to KillerRed (the dimeric form of SuperNova) more 1 O 2 and less H 2 O 2 are produced by SuperNova (Bulina et al., 2006a(Bulina et al., , 2006bTakemoto et al., 2013). Thus, the selected photosensitizers show a variety in ROS production with different amounts of produced 1 O 2 (ranked according to literature: SOPP3 > SOPP > Pp2FbFP > EcFbFP > SuperNova) and H 2 O 2 (ranked according to literature: SOPP3 > SOPP > EcFbFP > Pp2FbFP > SuperNova). Considering the position of the C-and N-termini in the crystal structure of the ApPDC (PDB ID: 2VBI), the photosensitizer tags were fused to the N-terminus of the enzyme. In ApPDCs, four monomers form an active tetrameric enzyme (Rother neé Gocke et al., 2011). As the C-termini of the monomers in the quaternary structure of the ApPDC are in close proximity to each other, it was predicted, that the introduction of tags might impair proper folding of the tetrameric enzyme. Five fusion constructs coding for SOPP-ApPDC-EGWH, SOPP3-ApPDC-EGWH, EcFbFP-ApPDC-EGWH, Pp2FbFP-ApPDC-EGWH and SuperNova-ApPDC-EGWH were cloned and subsequently produced in E. coli.

Production of Photosensitizer Enzyme Fusions
Production of soluble photosensitizer fusion enzymes proved to be difficult and was not possible for all constructs (Figure 3). In E. coli BL21(DE3), only low production levels of SOPP-ApPDC-EGWH were observed after SDS-PAGE analysis, as compared to the untagged ApPDC-EGWH, while for the SOPP3 fusion variant, at first, soluble protein could not be detected at all. Changing the expression host to E. coli Tuner ™ (DE3) cells resulted in a reasonable production of SOPP3-ApPDC-EGWH. The fusion enzyme SuperNova-ApPDC-EGWH revealed to give the best production of all soluble fusion enzymes. In contrast, for the EcFbFP and Pp2FbFP fusion enzymes mainly production of insoluble protein could be detected. Several approaches to increase the solubility were investigated, amongst others (data not shown): (i) Changing the expression host to E. coli Tuner ™ (DE3), (ii) testing different expression media (LB, AI, terrific broth) with (iii) additives as cofactors, betaine or FIGURE 4 | Schematic representation of the formation of insoluble three-dimensional protein aggregates by the combination of a dimeric photosensitizer with a tetrameric enzyme. Green: dimeric photosensitizers (EcFbFP or Pp2FbFP); blue: tetrameric ApPDC-EGWH. sorbitol, (iv) varying cultivation parameters as the total cultivation time (24, 48, 72 h), the expression temperature (15°C, 20°C) and the IPTG concentration used for induction (0.1, 0.25, 0.5 mM). Even (v) the co-expression of chaperones, did not shift the expression of both fusion enzymes towards a higher degree of soluble protein production. As none of the optimization strategies enhanced the soluble production of the fusion enzymes, we believe that not a misfolding prevented the production of soluble fusion proteins, but we rather produced insoluble protein aggregates (Figure 4). Because soluble protein production was not observed for the EcFbFP-ApPDC-EGWH and Pp2FbFP-ApPDC-EGWH constructs (Figure 3.), the soluble part of the crude cell extracts and the respective insoluble protein fraction, were subjected to qualitative activity assays under dark conditions to search for possible retained activity ( Figure 5). The activity test with soluble crude cell extract and the insoluble protein fraction of the EcFbFP-ApPDC-EGWH production ( Figure 5A) was done for a test-reaction setup from benzaldehyde and pyruvate to (S)-phenylacetylcarbinol ((S)-PAC) and CO 2 . However, for the activity test with soluble crude cell extract and the insoluble protein fraction of the Pp2FbFP-ApPDC-EGWH production ( Figure 5B), the reaction from 3-hydroxybenzaldehyde and pyruvate to (S)-3-hydroxyphenylacetylcarbinol ((S)-3-OH-PAC) and CO 2 was applied. These activity tests show preliminary results from very early project phases where no respective calibration was available, thus, the relative peak areas are presented. In the case of the soluble crude cell extracts, a small amount of activity was detected for the soluble EcFbFP fusion construct ( Figure 5A), while no activity was observed for the soluble Pp2FbFP construct ( Figure 5B). However, in case of the insoluble protein fractions, both constructs showed significant enzyme activity over the course of 6 h. It has been reported, that insoluble proteins could maintain activity to a certain extent when aggregated, if the active protein structure is preserved (Krauss et al., 2017). Intentional protein aggregation/immobilization can occur by introducing a suitable tag via molecular biological fusion, which can result in the formation of catalytically-active inclusion bodies (CatIBs). Although their activity is significantly reduced compared to soluble protein preparations, advantages are gained with respect to long term stability and production protocols (Diener et al., 2016). Most likely, a similar principle is relevant for the constructs presented here. Native EcFbFP and Pp2FbFP are dimers, whereas the other photosensitizers used in this study are monomers. When a dimeric photosensitizer fuses to a protein, aggregation is possible. If the enzyme to which the naturally dimeric photosensitizer is fused is a monomer, the photosensitizer could cause a fusion between two photosensitizer-enzyme constructs, as this would achieve the preferred quaternary structure of the photosensitizer dimer. Most likely this has no impact on the solubility. But for enzymes with four or more subunits, the fusion to a dimeric photosensitizer might lead to the formation of an insoluble three-dimensional protein aggregate as proposed in Figure 4. This principle is e.g. known from enzyme immobilization, when enzymes are cross-linked to each other forming a mesh (Sheldon and van Pelt, 2013). Multimerization of proteins via the fused photosensitizer tag was observed before for KillerRed (Bulina et al., 2006a;2006b). Thus, protein engineering was applied to the photosensitizer and the result was a monomeric variant of KillerRed, called SuperNova (Takemoto et al., 2013). This is the reason why SuperNova was also applied in this study. Nevertheless, EcFbFP-ApPDC-EGWH and Pp2FbFP-ApPDC-EGWH are active and the formation of insoluble protein aggregates could be promising if immobilization is intended. As the design of monomeric variants of EcFbFP and Pp2FbFP exceeds the scope of this work, the insoluble protein aggregates of EcFbFP-ApPDC-EGWH and Pp2FbFP-ApPDC-EGWH are evaluated in this study.

Activity of Photosensitizer-Tagged Enzymes and the Effect of Different Light Conditions on it
Besides the proof of activity for the dimeric photosensitizer forming aggregates (Figure 5), the initial rates of all photosensitizer fusion enzymes were measured to determine the effect of the different photosensitizer tags on the enzyme activity. The aim for the fusion constructs was to achieve as high activity as possible compared to untagged ApPDC-EGWH under dark conditions to have powerful fusion constructs for the application in syntheses. An optimal fusion construct would show similar (or even higher) activity than the untagged enzyme without light exposure but very quick inactivation upon light exposure. Initial rate experiments were performed under dark conditions to avoid any potential ROS formation by remaining light sources in the lab. To correct for the different weight of the fusion proteins, the enzyme activity in U µmol enzyme −1 was calculated to enable a comparison on a molar basis ( Figure 6). Thus, the displayed activities correspond to an amount of 1 µmol of each fusion enzyme. The initial rate was determined for the mentioned carboligation from 3-hydroxybenzaldehyde to (S)-3-OH-PAC. Since the correct amount of 3-hydroxybenzaldehyde is difficult to detect, as the substance may adhere to the reaction vessel or pipette tips, we decided to determine enzyme activities based on product formation, as this is a more accurate measure. Compared to the activity of the untagged ApPDC-EGWH, the SOPP variant only appeared to retain roughly a third of its activity. This loss in activity was not unexpected for the fusion enzyme, as immobilization or tagging of enzymes has been reported to have a negative effect in some cases due to reduced enzyme flexibility and active site accessibility (Spahn and Minteer, 2008). Interestingly, the initial rate activity of the fusion construct SOPP3-ApPDC-EGWH was much closer to that of the unfused carboligase. With 79% remaining activity, this was a very promising fusion candidate for further inactivation experiments. As SOPP and SOPP3 are structurally very close (Westberg et al., 2015(Westberg et al., , 2017, a similar effect of the tags on the carboligase was expected and generally observed. The activities of the EcFbFP and Pp2FbFP variants are low as expected after the experiments shown in Figure 5. Additionally, they are less reliable, as the insoluble protein aggregates affected the formation of inhomogeneous reaction solutions, which complicates the sampling during the experiments. As a high residual activity is the aim, fusions with multimeric photosensitizers are thus not recommended when the target enzyme has four or more subunits as in case of the ApPDC-EGWH. Unexpectedly, the fusion of SuperNova had a beneficial effect on the carboligase as an activity increase of 25% was observed. This is interesting as SuperNova (29 kDa) is roughly twice the size compared to the four LOV-based photosensitizers (13-19 kDa). Fortunately, all the fusion proteins tested in this study were active to a certain extent while the carboligase fusions to SOPP3 and SuperNova showed the highest activities. In addition, it should be emphasized that in all activity measurements only the (S)-stereoisomer of 3-OH-PAC was detectable. This demonstrates a highly selective production of the (S)-enantiomer by all photosensitizer fusion enzymes (enantiomeric excess of >99%), which is in accordance with literature of the unfused carboligase (Sehl et al., 2017). Conclusively, the photosensitizer tags do not negatively affect the enzymes stereoselectivity. As the photosensitizer fusion proteins are engineered to be sensitive to light, the effect of naturally occurring light (e.g. electric light in the laboratory and sunlight) on the protein activity was determined. Therefore, five general light conditions in different setups were tested (Table 2; Figure 7). After 60 min reaction time, product formations observed under dark and shady conditions revealed similar results within the experimental error. However, for both the experiments under low light and full light conditions, the product formation was drastically reduced. As expected and demanded the blue light fraction of the sunlight appeared to have a significant detrimental effect on the photosensitizer fused Blue LED (λ max = 450 nm), 1.5 cm distance~60 FIGURE 7 | Activity test of SOPP3-ApPDC-EGWH (0.5 mg ml −1 ; purified and lyophilized) at different light conditions (details are given in Table 2) and 60 min of exposure time. General reaction conditions: 40 mM 3hydroxybenzaldehyde, 200 mM pyruvate (sodium salt) and 50 mM HEPES buffer (pH 7.5) with 2.5 mM MgSO 4 and 0.1 mM ThDP. Incubation on 1 ml scale at 22°C and 850 rpm in a thermomixer (dark conditions) or in a stirred quartz-glass cuvette (illuminated setups). The error bars describe standard deviation of technical triplicates.
Frontiers in Catalysis | www.frontiersin.org March 2022 | Volume 2 | Article 835919 proteins, even with the electric lights switched off. Upon illumination with a blue LED (λ max = 450 nm,~60 mW cm −2 ), product formation diminished already after 30 min under nonoptimized conditions. For the first time, this demonstrates the potential of genetically encoded photosensitizers for lightmediated inactivation of the fused enzyme in biocatalysis approaches. Furthermore, as the sunlight appeared to already have a strong effect on the photosensitizer, leading to enzyme inactivation, all of the protein purifications and experimental preparations were performed under shady conditions. This includes coverage of the windows and electric lights switched off during experiments. In a next step, all photosensitizer fusion constructs were illuminated, to comparatively evaluate the inactivation potential of each construct.
Since we observed virtually complete inactivation of SOPP3-ApPDC-EGWH after 30 min of blue light irradiation in the previously described experiment with different light conditions, we chose a reaction time of 30 min for the inactivation of all fusion enzymes. However, the amount of product present in the reaction system after 30 min of inactivation time was formed mainly at the beginning of each reaction where the inactivation is not yet far advanced. In order to identify the degree of inactivation at the end of each reaction, the increase in product between the end point of each reaction after 30 min and the previous measurement point after 20 min reaction time was determined. The degree of inactivation was then calculated by comparing this value from the experiment with blue light exposure with the value from the respective experiment in dark conditions. The results show each photosensitizer to inactivate the attached ApPDC-EGWH to a certain extent ( Figures 8B-F). The degree of inactivation is equally high for SOPP-ApPDC-EGWH (94%) and for SOPP3-ApPDC-EGWH (92%) after 30 min of light exposure, which confirmed the inactivation seen for SOPP3-ApPDC-EGWH before (Figure 7). SOPP and SOPP3 are both reported to be strong ROS producers (Westberg et al., 2015;Westberg et al., 2017;Endres et al., 2018;Hilgers et al., 2019). For EcFbFP-ApPDC-EGWH, an inactivation tendency was observed, though the handling issues of the insoluble protein aggregates make it hard to draw a precise conclusion. The fusion enzymes Pp2FbFP-ApPDC-EGWH (38%) and SuperNova-ApPDC-EGWH (42%) both show less effective inactivation kinetics. A lower inactivation performance of Pp2FbFP-ApPDC-EGWH was expected, as the insoluble protein aggregates minimize the light transmission into the reaction   (Endres et al., 2018). Moreover, as the enzyme showed low activity, a high amount of protein lyophilisate was required, this limited light transmission, repeatedly. Regarding SuperNova, previous work showed this tag to exhibit a low type I photosensitizing activity which led to a less efficient inactivation of bacterial cells as compared to other photosensitizers . In additional tests, we confirmed, that the seen inactivation is purely mediated by the photosensitizer tags, as neither the untagged carboligase nor any of the reaction components (especially (S)-3-OH-PAC and 3-hydroxybenzaldehyde) appeared to be sensitive to blue or orange illumination (Supplementary Figure S2A,B). In summary, SOPP-ApPDC-EGWH and SOPP3-ApPDC-EGWH displayed the best inactivation performance close to full inactivation while the aggregate-forming fusion proteins and the SuperNova construct performed less than half as efficient. Genetic fusion of the photosensitizer to the target enzyme resulted in successful inactivation of the target enzyme by effectively utilizing the locally produced ROS and especially the high phototoxic effect of 1 O 2 (Sharma et al., 2012). Thus, by applying a single LED setup for illumination, the time necessary to achieve complete inactivation of the carboligase was 30 min. To enable complete inactivation as quickly as possible, preferably within a few minutes, further studies should be devoted to optimizing the inactivation setup. Conceivable setups are those, in which the light intensity can be regulated or LEDs are positioned around the sample instead of exposing only from one side.

The Effect of Illuminated Photosensitizer Fusion Enzymes on Other Enzymes in a Reaction System
The unspecific inactivation potential of the carboligase fusion enzymes was determined by illuminating photosensitizer fusion enzymes for 10 min in presence of the alcohol dehydrogenase from Lactobacillus brevis (LbADH) in the same reaction vessel, before the relative initial rate of the LbADH was measured. Specifically, in this experiment we were interested in whether free ROS can affect other enzymes in the same reaction system. Thus, the ADH represents another compound (or enzyme) in a reaction solution which could be passively affected by free ROS produced by the carboligase fusion enzymes upon illumination. During the previous inactivation experiment almost full inactivation was achieved after 30 min of blue light exposure (92% for the SOPP3-tagged carboligase), but a high degree of inactivation (74%) was already present after 10 min. Therefore, for the evaluation of the unspecific inactivation an approach lasting 10 min was chosen. In our opinion 10 min is a suitable inactivation time that can be achieved in an optimized inactivation setup. The LbADH activity was determined for the conversion of benzaldehyde to benzyl alcohol (involving Mg 2+ and NADPH; Figure 9). The activity of the LbADH in dark conditions without addition of a carboligase fusion enzyme served as control. Similar to the ApPDC-EGWH, neither the LbADH alone nor respective reaction components (especially benzaldehyde and benzyl alcohol) were sensitive to blue or orange illumination (Supplementary Figure S2C,D). In the presence of Pp2FbFP-ApPDC-EGWH and SuperNova-ApPDC-EGWH, the LbADH activity is decreased by~10% (Figure 10). It is in accordance to the results shown in Figures 8E-F, that Pp2FbFP and SuperNova in fusion with the ApPDC-EGWH have only low CALI potential after 10 min of light exposure and beyond. The LbADH inactivation mediated by SOPP-ApPDC-EGWH, SOPP3-ApPDC-EGWH and EcFbFP-ApPDC-EGWH (20-30%) is slightly higher, which might be caused by higher production levels of H 2 O 2 which was reported for SOPP, SOPP3 and EcFbFP before (Endres et al., 2018;Hilgers et al., FIGURE 10 | Relative activity of the LbADH (0.01 mg ml −1 ; purified and lyophilized) after illumination at a concentration of 0.1 mg ml −1 in combination with SOPP-ApPDC-EGWH, SOPP3-ApPDC-EGWH, EcFbFP-ApPDC-EGWH, Pp2FbFP-ApPDC-EGWH or SuperNova-ApPDC-EGWH (1 mg ml −1 each; purified and lyophilized). Illumination conditions: 50 mM potassium phosphate buffer (pH 7) with 1 mM MgCl 2 for 10 min with a single blue (λ max = 450 nm,~60 mW cm −2 ) or orange LED (λ max = 610 nm, 40 mW cm −2 ; only for SuperNova-ApPDC-EGWH) in a stirred quartz-glass cuvette. General reaction conditions: 10 mM benzaldehyde, 10 mM NADPH and 50 mM potassium phosphate buffer (pH 7) with 1 mM MgCl 2 . Incubation on 1 ml scale at 22°C and 850 rpm for 2 min in a thermomixer (dark conditions). The activity of the LbADH in the dark and without addition of any photosensitizer fusion enzymes served as control. The error bars describe standard deviation of technical triplicates. Frontiers in Catalysis | www.frontiersin.org 2019). H 2 O 2 can travel longer distances than 1 O 2 as it has a halflive of~1 ms (Reth, 2002). Thus, the undesired inactivation is depending on the H 2 O 2 amount produced and probably also on the sensitivity of the present enzymes in the reaction system. Nevertheless, the passive inactivation of the ADH by the ROS produced by the fusion enzymes under exposure to light is considerably lower than is the case with direct inactivation of the fusion enzymes. Compared to the previous inactivation experiment ( Figure 8C), where SOPP3-ApPDC-EGWH was already inactivated by 74% after 10 min of blue light illumination, the passive inactivation of the LbADH after 10 min of joint exposure with SOPP3-ApPDC-EGWH is significantly lower at about 30%. This is most likely due to the fact that photosensitizers always produce different types of ROS and it has been shown in the literature that the phototoxicity of photosensitizers is greatly enhanced by the production of 1 O 2 in particular . In order to further evaluate the selectivity of the inactivation mediated by the fusion enzymes, it should be highlighted, that the photosensitizer fusions were used in 10-fold excess compared to the LbADH to see the effect in this experiment. This high surplus would normally not take place in a multi-step synthesis approach. As mentioned earlier, an exposure time of 10 min should be undercut in future optimizations. This would further reduce H 2 O 2 formation and therewith undesired enzyme inactivation.

CONCLUSION
In this work we aim for the construction of active photosensitizer fusion enzymes which show a high degree of inactivation after illumination. We envision these fusion proteins to be a suitable tool to prevent cross-reactivity in multi-step enzyme cascades, which will be shown in another publication. Hence, we fused the photosensitizers SOPP, SOPP3, EcFbFP, Pp2FbFP and SuperNova to the carboligase variant ApPDC-EGWH. SOPP3 proved to be the most promising of the tags tested, for the reasons of soluble expression of the fusion protein, high enzyme activity of the fusion construct in the dark and efficient inactivation upon specific light exposure. The initial rate of the SOPP3 fusion enzyme is similar to the untagged ApPDC-EGWH in the dark. Of all tags tested, only the SuperNova fusion enzyme showed an increase in activity in the dark as compared to the unfused carboligase, however, the SuperNova tag demonstrated only limited CALI activity. SOPP3-ApPDC-EGWH, along with the SOPP-fused variant revealed the best inactivation performance and they could be inactivated by over 90% within 30 min of blue light illumination. Furthermore it was observed, that dimeric photosensitizers as EcFbFP and Pp2FbFP are not suited as tags for enzymes consisting of four or more subunits as the formation of insoluble protein aggregates minimizes the enzyme activity and clearly complicates the handling of the fusion enzymes. Efficient use and photo-induced inactivation of SOPP3-ApPDC-EGWH in an enzyme cascade is currently conducted and optimized in our lab. Additionally, on the long run this method can enable selfcontrolled regulation of activity by switching on/off a blue light source, e.g., if an in-line monitoring of substrate decrease and product increase is present. Further, selective enzyme regulation might be available in micro-scale chips (Gruenberger et al., 2013;Burmeister et al., 2019) and whole-cell biocatalysis (Jakoblinnert and Rother, 2014;Wachtmeister and Rother, 2016), which is currently challenging to achieve.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
TG, JS, and SS designed and performed the experiments. TG wrote the paper, TD, MvS, FH, NB, and DR helped to draft the manuscript, gave advice for certain techniques and supervised the research. All authors contributed to manuscript revision, read, and approved the submitted version.

FUNDING
This project was funded by the European Research Council in frame of ERC starting grant 757320 "Light Controlled Synthetic Enzyme Cascades".