Performance of cross‐linked enzyme crystals of engineered halohydrin dehalogenase HheG in different chemical reactor systems

Halohydrin dehalogenase HheG is an industrially interesting biocatalyst for the preparation of different β‐substituted alcohols starting from bulky internal epoxides. We previously demonstrated that the immobilization of different HheG variants in the form of cross‐linked enzyme crystals (CLECs) yielded stable and reusable enzyme immobilizes with increased resistance regarding temperature, pH, and the presence of organic solvents. Now, to further establish their preparative applicability, HheG D114C CLECs cross‐linked with bis‐maleimidoethane have been successfully produced on a larger scale using a stirred crystallization approach, and their application in different chemical reactor types (stirred tank reactor, fluidized bed reactor, and packed bed reactor) was systematically studied and compared for the ring opening of cyclohexene oxide with azide. This revealed the highest obtained space‐time yield of 23.9 kgproduct gCLEC−1 h−1 Lreactor volume−1 along with the highest achieved product enantiomeric excess [64%] for application in a packed‐bed reactor. Additionally, lyophilization of those CLECs yielded a storage‐stable HheG preparation that still retained 67% of initial activity (after lyophilization) after 6 months of storage at room temperature.

which the enzyme is embedded within a polymer structure, often a hydrogel, based, for example, on alginate, alkoxides, or ionic liquids (Arruda & Vitolo, 1999;Liu et al., 2005;Vidinha et al., 2006), and (iii) carrier-free immobilization.The latter is based on covalent crosslinking of enzyme molecules without the need for an additional carrier (Roessl et al., 2010).Here, enzymes can be cross-linked as soluble (Ashjari et al., 2020) or lyophilized enzyme (Akkas et al., 2020), as well as in the form of enzyme aggregates (Sheldon, 2007) or enzyme crystals (Häring & Schreier, 1999).Cross-linked enzyme crystals (CLECs) and cross-linked enzyme aggregates (CLEAs) have been investigated and widely used in biocatalytic reactions so far (Fernández-Penas et al., 2021;Mahmod et al., 2016;Perveen et al., 2022;Rehman et al., 2016;St. Clair & Navia, 1992;Vaghjiani et al., 2000).While immobilization on a noncatalytic carrier can decrease enzymatic activity and reduces volumetric activity in particular, carrier-free enzyme immobilizes offer much higher activity-to-volume ratios (Eş et al., 2015;Pastinen et al., 2000).Even though CLEAs have been used more widely than CLECs due to their easy preparation based on unpurified enzyme (Mahmod et al., 2016), the required enzyme precipitation before cross-linking, however, can be also accompanied by enzyme denaturation caused by severe force effects and stress on the protein structure (Kartal et al., 2011;Schoevaart et al., 2004).In contrast, the native enzyme structure is largely retained during crystallization, minimizing enzyme inactivation during CLEC formation, but the crystallization process itself is often challenging.Nevertheless, CLEC production and biocatalytic application have been reported for a variety of enzymes (Jegan Roy & Emilia Abraham, 2004), such as peroxidases, dehydrogenases, and lipases (Ayala et al., 2002;Fernández-Penas et al., 2021;St. Clair et al., 2000).
Halohydrin dehalogenases (HHDHs) are bacterial lyases that catalyze the reversible dehalogenation of vicinal haloalcohols with the formation of the corresponding epoxides (Schallmey & Schallmey, 2016;van Hylckama Vlieg et al., 2001).In the reverse reaction, which is chemically more interesting, they are also capable of opening epoxide rings with a range of N-, C-, O-and S-nucleophiles resulting in the formation of novel C-C, C-N, C-O, or C-S bonds (Hasnaoui-Dijoux et al., 2008;Ma et al., 2022;M. Majerić Elenkov et al., 2006;M. M. Majerić Elenkov et al., 2008;Molinaro et al., 2010;Wang et al., 2022;Zhou et al., 2023).Among a large number of recently discovered new HHDHs (Schallmey et al., 2014;Wang et al., 2022;Xue et al., 2018Xue et al., , 2019Xue et al., , 2020;;Zhou et al., 2023), halohydrin dehalogenase G (HheG) from Illumatobacter coccineus, along with other G-type HHDHs, displays also exceptional activity in the conversion of sterically more demanding cyclic as well as acyclic vicinal di-substituted epoxides (Calderini et al., 2019;Koopmeiners et al., 2017;Solarczek et al., 2022).Despite this unique substrate profile, HheG's industrial potential is limited by its insufficient stability as indicated by an apparent melting temperature of only 38°C (Solarczek et al., 2019).Previous attempts to immobilize HheG in the form of CLECs-since HheG crystallizes readily after simple one-step purification via affinity chromatography-yielded mechanically stable CLECs (Kubiak et al., 2022).CLEC activity, however, was dependent on the employed chemical cross-linker.While glutaraldehyde impaired HheG's activity tremendously, probably caused by cross-linking of also catalytic residues, the use of cysteine-or lysine-specific cross-linkers after selective incorporation of respective amino acids at desired cross-linking sites, yielded highly stable and active CLECs (Staar, Henke, et al., 2022;Staar, Staar, et al., 2022).Thus, crystals of HheG variants D114C and V46K cross-linked with bis-maleimidoethane (BMOE) and dithiobis-(succinimidyl propionate) (DSP), respectively, exhibited good stability toward high temperature, extreme pH conditions and the presence of organic solvents (Staar, Henke, et al., 2022;Staar, Staar, et al., 2022).Additionally, the successful reuse of both CLEC preparations over 21 days was demonstrated, confirming their superior operational stability compared with soluble enzyme.Nevertheless, as batch crystallization is the preferred crystallization method for CLEC preparation on industrial scale due to easier scalability (Margolin & Navia, 2001), HheG variant D114C seems to offer a higher potential for industrial application based on its favorable batch crystallization characteristics compared with HheG V46K (Staar, Henke, et al., 2022;Staar, Staar, et al., 2022).Apart from HheG, immobilization of HHDHs has only been reported for HheC from Agrobacterium radiobacter AD1 so far using different methods.
Thus, this enzyme has been immobilized as CLEAs using glutaraldehyde as cross-linker (Bogale et al., 2020;Liao et al., 2018), on magnetic biochar (Jiang et al., 2022), and two different synthetic organic resins (Zhang et al., 2018;Zou et al., 2018), each for application in biocatalytic dehalogenation processes.Moreover, co-encapsulation of HheC together with haloalkane dehalogenase DhaA31 from Rhodococcus rhodochrous NCIMB 13064 and epoxide hydrolase EchA from A. radiobacter AD1 in a polyvinyl alcohol-based hydrogel for the biodegradation of toxic 1,2,3-trichloropropanol has been reported (Dvorak et al., 2014).Additionally, Gul et al. (2020) investigated the immobilization of HheC on a glass fiber membrane for the detection of haloalcohols in wastewater.In contrast to our CLEC approach, however, those examples (except for the preparation of HheC CLEAs) refer to carrier-bound immobilization methods and require the addition of a noncatalytic carrier material.
In this study, we aimed to scale up CLEC production and to evaluate the performance of the resulting HheG D114C CLECs regarding their productivity as well as enantioselectivity in different chemical reactor systems to facilitate a future industrial application of this enzyme.Moreover, lyophilization was studied as a means of CLEC formulation with the aim to yield a storage-stable CLEC preparation.

| Bacterial strains and plasmids
E. coli BL21(DE3) Gold was used for heterologous protein production as outlined before (Koopmeiners et al., 2016).Further, expression vector pET-28a(+) (Merck) was used with the respective hheG D114C gene (GenBank accession number of the codon-optimized wild-type gene: KU501255) under control of the T7 promoter, resulting in the addition of an N-terminal His 6 -tag to the heterologously produced protein.
One-step purification of soluble HheG D114C from those cell pellets via the enzyme's N-terminal hexahistidine-tag was performed using immobilized metal affinity chromatography (IMAC) as described before (Staar, Henke, et al., 2022).The only difference was the lack of sodium sulfate in buffer A and B due to otherwise observed precipitation during purification.Eluted target protein was desalted using a HiPrep 26/10 desalting column (GE Healthcare) and storage buffer (10 mM Tris•SO 4 , 4 mM EDTA, pH 7.9, 10% (v/v) glycerol, 5 mM dithiothreitol).Purified protein was stored at −20°C until further use.

| 3 L fed-batch approach
For a scale-up of HheG D114C production, fed-batch fermentation was performed using a Labfors 5 bioreactor (Infors).3 L TB medium supplemented with 50 µg mL −1 kanamycine were inoculated with 10% (v/v) preculture and protein production was induced with 0.2 mM IPTG.Protein production was performed for 19 h at 22°C with pH ranging between pH 7.1 and 7.9.During fermentation, the stirrer speed (up to 800 rpm) and airflow (up to 10 mL min −1 ) were varied to keep a constant level of dissolved oxygen (pO 2 = 90%).
After 11 h, the feed (12% (v/v) glycerol, 600 mM (NH 4 ) 2 SO 4 ) was started with a linearly increasing feed flow rate from 0.18 to 0.54 mL min −1 in the first 5 h.Afterward, a constant feed flow rate of 1 mL min −1 was applied for another 3 h.Optical density at 600 nm (OD 600nm ) was measured hourly.Every 3 h, a respective volume of fermentation broth was taken corresponding to an OD 600 nm of 10 in 1 mL, and analyzed via sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).At the end of the fermentation, the cells were harvested by centrifugation (2 × 30 min at 3315g) and the resulting cell pellet was divided into six equal portions, which were stored at −20°C.
Purification of HheG D114C from those cell pellets was performed as mentioned above using a 20 mL HisTrap FF column (Cytiva).Each three-cell pellets were combined and resuspended in 30 mL buffer A (50 mM Tris•SO 4 , 25 mM imidazole, pH 7.9) before cell disruption and purification via IMAC.Eluted target protein was desalted using two HiPrep 26/10 desalting columns and storage buffer.Purified protein was stored at −20°C until further use.

| 2 mL batch crystallization
Crystallization in 2 mL scale was performed under gentle shaking in 15 mL tubes (Sarstedt) for 24 h at 8°C as described before (Staar, Staar, et al., 2022).One milliliter protein solution with a concentration of 24 mg mL −1 was mixed 1:1 with crystallization buffer (10 mM HEPES, pH 7.0, 8% (w/v) PEG4000).After 24 h, crystals were harvested by centrifugation for 3 min at 400g.The amount of crystallized protein was calculated from the remaining protein absorbance at 280 nm in the supernatant.

| 20 mL batch crystallization
Crystallization in 20 mL scale was performed under identical conditions as the 2 mL batch crystallization but using 50 mL tubes (Sarstedt).Here, 10 mL of a 24 mg mL −1 concentrated protein solution was mixed 1:1 with crystallization buffer and incubated for 24 h at 8°C.

| 50 mL batch crystallization
For 50 mL batch crystallization, stirring of the protein solution was used instead of shaking.25 mL of a 24 mg mL −1 concentrated protein solution was mixed 1:1 with crystallization buffer (final protein concentration of 12 mg mL −1 ).Crystallization was performed in a 100 mL duran bottle (Schott) using a magnetic stirrer at 200 rpm stirrer speed and 8°C.Crystals were harvested after 24 h by centrifugation for 3 min at 400g.Additionally, 50 mL batch crystallization was also performed using the same crystallization conditions but with less protein (6 mg mL −1 final protein concentration) to evaluate the necessity of a high protein concentration for efficient HheG D114C crystallization.
2.5 | Cross-linking 2.5.1 | Cross-linking in 2 and 20 mL Cross-linking of HheG D114C crystals using the cysteine-specific cross-linker BMOE was performed as described before (Staar, Henke, et al., 2022). 2 mM BMOE in either 2 or 20 mL volume of crystallization buffer containing 10% (v/v) DMSO was used for cross-linking.Harvested crystals from the batch crystallization were submerged in cross-linking solution and incubated under gently shaking for 24 h at 8°C.Afterward, CLECs were harvested by centrifugation for 3 min at 400g and washed with a respective volume of 50 mM Tris•SO 4 , pH 7.0.CLECs were harvested again by centrifugation and resuspended in 50 mM Tris•SO 4 , pH 7.0 to a concentration of 2 mg mL −1 for further use.

| Cross-linking in 50 mL
Cross-linking in 50 mL scale was performed using the same composition of the cross-linking solution (2 mM BMOE in 10 mM HEPES, pH 7.0, 8% (w/v) PEG4000, and 10% DMSO).Harvested crystals from the crystallization were submerged in 50 mL cross-linking solution and incubated for 24 h at 8°C with stirring at 200 rpm.Afterward, CLECs were harvested by centrifugation for 10 min at 400g and washed with a respective volume of 50 mM Tris•SO 4 , pH 7.0.CLECs were harvested again by centrifugation and resuspended in 50 mM Tris•SO 4 , pH 7.0 to a concentration of 35 mg mL −1 for further use.

| CLEC lyophilization
Lyophilization was performed using a Beta 2-8 LSCplus lyophilizer (Martin Christ).Each 2 mL of a 35 mg mL −1 CLEC suspension were transferred to five 4 mL glass vials (Thermo Fisher Scientific).The vials were placed in the precooled (4°C) lyophilizer.Within 1 h, shelf temperature was decreased to −35°C and kept there for 5 h.Afterward, vacuum was set to 0.1 mbar.Sublimation was performed for 12 h.Then, the shelf temperature was increased to 20°C again while vacuum was decreased to 0.08 mbar, and this condition was held constant for 2 h.After the lyophilization process, the glass vials with lyophilized CLECs were closed with a lid and stored at room temperature.One reference vial containing 2 mL 50 mM Tris•SO 4 , pH 7.0 was lyophilized in parallel to obtain the weight of lyophilized buffer.
To evaluate the storage stability of the lyophilized CLECs, they were kept at room temperature for 6 months.Every month, the residual activity was determined.For this, 1 mg stored CLECs were resuspended in 50 mM Tris•SO 4 , pH 7.0.Residual activity of 100 µg CLECs in the azidolysis (40 mM azide) of cyclohexene oxide (20 mM) in 1 mL 50 mM Tris•SO 4 , pH 7.0 at 22°C and 900 rpm was determined.After 1 h of reaction, 400 µL sample was taken, extracted, and analyzed via achiral gas chromatography (GC).
Reactions were performed in triplicates.

| Particle size distribution
Particle size analysis of HheG D114C CLECs after crystallization and cross-linking prior or after lyophilization was performed via laser diffraction using a Mastersizer 3000 (Malvern Panalytical) as described before (Staar, Staar, et al., 2022).A 2 mL volume of a concentrated CLEC solution (8.5 mg mL −1 for D114C CLECs after crystallization and cross-linking, 35 mg mL −1 after lyophilization and resuspension) was used for wet particle size distribution analysis.
Particle size distributions were determined in triplicate and the average is displayed as density distribution as well as 10%, 50%, and 90% cumulative undersizes (d 10 , d 50 , d 90 ).

| Chord length distribution
To evaluate the mechanical stress of the CLECs in a stirred batch process, online chord length distribution was determined.An Easy-Max 102 Thermostat system (Mettler-Toledo) with an inclined blade stirrer at 500 rpm stirrer speed and a 50 mL CLEC solution containing 11.1 mg of CLECs in 50 mM Tris•SO4, pH 7.0 were used.Chord length distribution was measured online using a ParticleTrack G400 (Mettler-Toledo) over a time of 24 h.Percentiles d 10 , d 50 , d 90 were determined and plotted over the mechanical stress time.

| (Fed-)Batch reactor
Batch and fed-batch reactions were carried out in 100 mL Duran bottles (Schott) using always a 50 mL reaction volume and 50 mM Tris•SO4, pH 7.0 buffer.
The influence of the stirring speed on CLEC activity was determined first.For this, reactions were performed using each 5 mg CLECs, 20 mM cyclohexene oxide, and 40 mM azide for 30 min at room temperature.Different stirring speeds (100-1200 rpm) were tested.Conversion after 30 min was determined via achiral GC.The highest conversion was set to 100% relative activity.Batch reactions were performed using each 2.5 mg CLECs, 150 mM cyclohexene oxide, and 165 or 300 mM sodium azide at room temperature and 500 rpm stirring speed for 2 h with sampling every 15 min.Fed-batch reactions were performed using each 2.5 mg CLECs and 150 mM cyclohexene oxide at room temperature and 500 rpm stirring speed for 2 h.Sequential fed-batch was performed varying either the azide concentration in the feed (feeding every 30 min each 1 mL of a 1.25, 2.5, 3.75, or 5 M azide stock solution, corresponding to azide concentrations of 25, 50, 75, or 100 mM in the reaction volume; 53 mL final reaction volume) or the number of feed pulses (feeding every 15, 20, 30, or 60 min each 1 mL of a 2.5 M azide stock, corresponding to 50 mM azide concentration in the reaction volume and resulting in 57, 55, 53, or 51 mL final reaction volume, respectively).In both cases, the first pulse was always set at t = 0 min.Semicontinuous fed-batch reactions were performed varying either the azide concentration in the feed or the feed flow rate using a KDS100 Legacy syringe pump (Thermo Fisher Scientific).Varying the feed flow rate, a 2.5 M azide stock solution was fed to the reaction at a flow rate of 1, 2, 3, or 4 mL h −1 (resulting in final reaction volumes of 52, 54, 56, or 58 mL, respectively), while 1 mL from that stock was added at t = 0 min.Varying the azide concentration in the feed, a 1.25, 2.5, 3.75, or 5 M azide stock solution was fed to the reaction at a flow rate of 2 mL h −1 (54 mL final reaction volume).Each 1 mL from the respective stock solution was added at t = 0 min.Reactions were performed in duplicate and analyzed by achiral GC.

| Fluidized-bed reactor (FBR)
An FBR was constructed using a PureCube Cartridge 1 mL FPLC column (Cube Biotech) filled with 4 mg of HheG D114C CLECs.The flow was applied upwards using a KDS100 Legacy syringe pump.A schematic representation of the FBR setup is given in Supporting Information: ] (Equation 1) and the enantiomeric excess of the formed product.Afterward, different concentrations of cyclohexene oxide (20, 50, 75, 100, 150 mM) and azide (22, 55, 82.5, 110, 165 mM) were analyzed at the same different flow rates.The continuous-flow reaction rate was determined via conversion after 10 min of each run using achiral GC, while the product enantiomeric excess was determined by chiral GC.Moreover, the long-time performance of the FBR was investigated using 20 mM cyclohexene oxide and 40 mM azide at a flow rate of 0.25 mL min −1 .Sampling was performed every 24 h and analyzed by achiral GC.

| Residence time determination
Residence time of the FBR and PBR was determined at a flow rate of 4 mL min −1 .For this, 30% (v/v) acetone in 50 mM Tris•SO 4 , pH 7.0 was injected on both columns as a pulse.UV absorbance at 280 nm over time was monitored until the UV absorbance stayed constant.
The plotted UV absorbance over time was fitted using the Boltzmann equation in Origin2021 Pro.The residence time was determined as the inflection point of the fitted curve.

| GC analysis
To determine conversion or product enantiomeric excess of biocatalytic reactions, GC analysis was performed on a GC2010 plus (Shimadzu) after extraction of reaction samples.Four hundred microliters reaction samples were mixed 1:1 with tert-  S4.

| Scale up of CLEC production
Previously, CLEC production of HheG D114C in batch has only been performed in up to 2 mL scale (Staar, Staar, et al., 2022).Thus, upscaling of batch crystallization and cross-linking was attempted to obtain larger CLEC amounts for biocatalytic application.For this, the HheG D114C production in E. coli BL21(DE3) gold was performed in a Labfors 5 bioreactor in a total volume of 3.5 L using a fed-batch procedure.This resulted in an optical cell density at 600 nm of 15.7 after 19 h of incubation (Supporting Information: Figure S1) with constant overproduction of HheG D114C over the complete incubation time (Supporting Information: Figure S2).In total, 613 mg of pure HheG D114C was isolated after IMAC purification via the N-terminal hexahistidine tag.The corresponding yield of 175 mg protein L production volume −1 is comparable to the yield achieved in 500 mL shake flask production [204 mg protein L production volume −1 (Staar, Henke, et al., 2022)].Using this purified enzyme, protein crystallization in batch was performed in 20 mL scale with gentle shaking as well as 50 mL scale under stirring for each 72 h at 8°C.The corresponding crystallization kinetics are displayed in Figure 1.For comparison, the crystallization data of HheG D114C in 2 mL scale, taken from Staar, Staar, et al. (2022), is also displayed.
The obtained crystallization data were fitted by the Avrami equation for crystallization under isothermal conditions (Avrami, 1939) to evaluate the corresponding crystallization kinetics.
For batch crystallization in 2 and 20 mL scales (both with shaking), resulting Avrami parameters are comparable (Supporting Information: Table S1), indicating similar crystallization velocities.Only the total amount of crystals differs, as expected, with 180.4 mg crystals (75% crystallization yield) obtained in 20 mL scale and 16.5 mg (69% crystallization yield) in 2 mL scale.As industrial crystallization processes are performed in stirred vessels (Schmidt et al., 2005), a stirred batch crystallization process of HheG D114C in 50 mL scale was also investigated.With a lower velocity constant [k = 0.135 h −1 ] and a higher half-time [t 0.5 = 3.0 h] compared with the shaken ones, the stirred crystallization process is slower while reaching a higher amount of total crystals obtained (339.4 mg, 57% crystallization yield).For comparison, the stirred batch crystallization was also performed using a twofold lower protein concentration (6 mg mL −1 , Supporting Information: Figure S3), resulting in an even slower crystallization process [k = 0.013 h −1 , t 0.5 = 10.4 h] as well as a lower total amount of obtained crystals (114.6 mg, 38% crystallization yield) compared with the 50 mL batch crystallization with higher protein concentration.Thus, a higher HheG concentration is indeed required for an efficient crystallization process.
Next to the crystallization kinetics, also the crystal size of the obtained CLECs was analyzed.Particle size distribution of the crystals resulting from the stirred 50 mL batch crystallization revealed a median value (d 50 ) of 2.24 µm, which is almost the same as for the shaken 20 mL batch crystallization (d 50 = 2.26 µm) and similar to the shaken 2 mL crystallization (d 50 = 2.02 µm (Staar, Staar, et al., 2022), Supporting Information: Table S2).If desired, a further enhancement of the stirred crystallization process in terms of the obtained crystal amount or crystal size should be possible by varying the impeller energy input, the protein and precipitant concentration, or through the use of a temperature ramp during crystallization (Hebel et al., 2013;Schmidt et al., 2005).
After subsequent cross-linking of the obtained crystals with BMOE, a total of 353 mg HheG D114C CLECs were obtained in 50 mL batch scale with stirring, corresponding to a yield of 0.58 mg CLECs per mg of applied soluble enzyme and 100.9 mg CLECs per liter expression volume.  of particle size distribution (Supporting Information: Table S3) revealed no decrease in particle size.Moreover, microscopic analysis confirmed that the CLECs retained their hexagonal shape during the stirring process (Supporting Information: Figure S7).The only observed difference was that some CLECs sticked together, forming some kind of aggregates after 24 h of stirring, which was also seen by a higher d 90 percentile (15.3 µm before and 20.8 µm after mechanical stress) in the particle size distribution.The small CLEC size (d 50 = 2.02 µm) could be a major factor for their mechanical stability during the stirring process.Lee et al. (2002) investigated the mechanical stability of CLECs of yeast alcohol dehydrogenase I in a rotating disk shear device.They found that smaller-sized rod-shaped CLECs (median size of 4.6 µm) did not break during the mechanical stress time, while hexagonal-shaped CLECs of bigger size (median size of 12 µm) did show a decrease in particle size with increasing mechanical stress (Lee et al., 2002).Moreover, no CLEC breakage of yeast alcohol dehydrogenase I CLECs was observed at 10,000 rpm in the rotating disk shear device (which corresponds to 27,000 rpm in a stirred vessel using a Rushton turbine) (Vaghjiani et al., 2000).Hence, CLECs in general seem to tolerate much higher shear forces in stirred processes than the ones applied in our STR approach at 500 rpm stirring.

| STR
As our HheG D114C CLECs tend to be mechanically stable in a stirred process, their biocatalytic productivity in the azidolysis of cyclohexene oxide was subsequently evaluated in an STR.As a high chemical background reaction can occur in the reaction of cyclohexene oxide with azide, especially at higher azide concentrations (Staar, Henke, et al., 2022), next to classical batch reactions with moderate (165 mM) and high (300 mM) azide concentrations, also fed-batch reactions were investigated with the aim to minimize the chemical background.As the relative CLEC activity did not change significantly with altered stirring speed (Supporting Information: Figure S8), a stirrer speed of 500 rpm was selected for all STR experiments.In batch reactions, each 2.5 mg of HheG D114C CLECs and 150 mM cyclohexene oxide were applied in 50 ml total volume in 50 mM Tris•SO 4 buffer, pH 7.0.In one experiment, 1.1 equivalents of azide (165 mM) were added, while in the other one 2 equivalents (300 mM) were used.Fed-batch reactions were divided in sequential as well as semicontinuous fed-batch approaches.Additionally, the azide concentration, the number of azide pulses as well as the azide feed flow rate were varied (for details see Section 2).Conversions in the different approaches (batch and fed-batch reactions) over a time span of 2 h were determined.In general, higher STY were obtained with higher azide concentrations, a higher number of azide pulses or a higher feed flow rate (see Supporting Information: Figure S9 for details).At the same time, however, also the chemical background conversion increased significantly.Corresponding STY as well as STY ratios of CLEC-and chemically-catalyzed conversions for the best reaction conditions (highest STY) of each (fed-)batch approach are summarized in Table 1.Based on this data, the highest STY [3.24 g product h −1 L reaction volume ] was achieved in the batch reaction with 300 mM (2 eq) azide with a STY ratio of the CLEC-catalyzed to the chemical background reaction of only 1.9.The latter indicates that a significant amount of the formed product in that CLEC reaction actually stems from the (unselective) chemical background reaction, which yields only racemic product and is therefore not desired for practical application.However, there was no fed-batch condition (with azide feeding) reaching a higher STY.Moreover, all listed conditions using high azide concentrations displayed unsatisfactory STY ratios of CLEC-catalyzed to chemical background reaction.The batch reaction with lower azide concentration [165 mM, 1.1 eq] features a better ratio regarding chemical background, but also a lower STY compared with all fed-batch conditions listed in Table 1.
Hence, the amount of applied azide determines the attainable STY of the CLEC-catalyzed reaction, but at the same time increases also the chemical background reaction in (fed-)batch applications.In addition to the reactions listed in Table 1, there was one further condition (sequential fed-batch with 50 mM azide pulse every 20 min) achieving a good STY [2.82 g product h −1 L reaction volume −1 ] at a much better STY ratio of 3.1.Unwanted chemical background azidolysis in HHDH-catalyzed epoxide ring opening reactions with azide has been reported before, for example, in the kinetic resolution of p-nitrostyrene oxide catalyzed by HheC (Lutje Spelberg et al., 2001).
In that case, step-wise addition of azide over 24 h in a 60 mL reaction T A B L E 1 Space-time yields (STY) and STY ratios [CLEC:negative] for the best reaction conditions (highest STY) of each (fed-) batch reactor system.was performed to keep the azide concentration between 0.5 and 1 mM (the apparent K M of HheC toward azide was determined to be around 0.2 mM).This enabled the preparation of (R)−2-azido-1-(4nitrophenyl)ethan-1-ol with 97% enantiomeric excess, as the unselective chemical background conversion was low.In our case, however, a much higher azide concentration has to be applied in the reaction as the K 50 value (Hill kinetics) of HheG toward azide is around 9 mM, so 40-times higher compared with HheC (Koopmeiners et al., 2017).Thus, chemical background azidolysis will always be a challenge that needs to be addressed in HheG-catalyzed reactions in batch, even though the actual amount of chemically formed product will also be dependent on the applied epoxide.

| FBR
As the chemical background azidolysis of cyclohexene oxide turned out to be a major limitation in (fed-)batch reactions, continuously operated reactor systems were investigated as an alternative.
Moreover, continuous reactor systems have been reported to be more productive than stirred vessel systems in enzyme-catalyzed reactions (Planchestainer et al., 2017;Szelwicka et al., 2019;Tomin et al., 2010).In this regard, an FBR containing a rather low HheG D114C CLEC amount [4 mg] was constructed and operated using a KDS100 Legacy syringe pump with an upwards-oriented flow.To evaluate the productivity of this FBR in the azidolysis of cyclohexene oxide, continuous-flow reaction rates [g product h −1 g CLEC −1 ] at different substrate concentrations and over different flow rates were determined (Figure 2).Cyclohexene oxide and azide were premixed in the substrate solution, while the ratio of epoxide and nucleophile was kept constant at 1:1.1 to avoid a high chemical background.As expected, the resulting continuous-flow reaction rate increased with increasing flow rate and increasing substrate concentration, reaching a maximum of 18.9 g product h −1 g CLEC −1 at 100 mM substrate concentration and a flow rate of 4 mL min −1 .A further increase in substrate concentration to 150 mM did not enhance the reaction rate further and is probably limited by the applied CLEC amount.Moreover, a constant product enantiomeric excess of 60% for the formation of 2azido-cyclohexan-1-ol was obtained over all flow rates at 20 mM epoxide concentration (Supporting Information: Figure S10), which is in agreement with the respective product enantiomeric excess achieved by CLECs of HheG D114C (Staar, Henke, et al., 2022).
Next to productivity, also the operational stability of the FBR was studied.For this, the FBR was operated over a time span of 21 days using 20 mM cyclohexene oxide and 40 mM azide at a flow rate of 0.25 mLmin −1 .Every 24 h, the conversion was measured to calculate the corresponding continuous-flow reaction rate as well as the cumulative turnover number (Figure 3).As a result, a rather constant continuous-flow reaction rate of around 1 g product h −1 g CLEC −1 was obtained over the complete operation time resulting in a linearly increasing cumulative turn over number (TON), which reached a value of 1.22 • 10 5 mol product per mol of enzyme after 21 days.
Thus, no loss in productivity could be observed over the 3-week operation time, illustrating the high operational stability of the CLECcontaining FBR.Such a long-time performance of our CLECs has also been demonstrated for an application in batch mode with daily recycling of the CLECs (Staar, Henke, et al., 2022).Taking into account that previously a half-life of 64 days at 22°C has been determined for our HheG D114C CLECs (Staar, Henke, et al., 2022), significantly longer continuous operation times and thus even higher TON values should be possible with our system.

| PBR
Even though high continuous-flow reaction rates were achieved using the FBR, productivity in this system is still limited by the applied CLEC amount [4 mg in our case], as catalyst loading is generally lower in FBRs than in PBRs (Matoh et al., 2019).Thus, also a PBR containing our HheG D114C CLECs was constructed with the aim to enhance productivity further.
A first PBR containing a total of 150 mg CLECs, however, led to complete clogging after only a few reaction runs, which is probably caused by the small particle size of our CLECs (vide infra).Therefore, a lower amount of CLECs [50 mg] was subsequently used and the reactor was additionally filled with glass beads exhibiting a bigger particle size [>106 µm] than our CLECs.With this optimization, HheG D114C CLECs in the azidolysis of cyclohexene oxide achieved in the fluidized-bed reactor (FBR) at different flow rates (0.25, 0.5, 1, 2, 4 mL min −1 ).The reaction flow contained different concentrations of cyclohexene oxide (20, 50, 75, 100, 150 mM) and azide in a 1:1.1 ratio in 50 mM Tris•SO 4 , pH 7.0.The flow was achieved using a KDS100 Legacy syringe pump.The FBR contained 4 mg CLECs, and the continuous reaction was performed at room temperature.The continuous-flow reaction rate was determined by achiral GC via the attained conversion after 10 min of each run.The highest achieved continuous-flow reaction rate of 18.9 g product h −1 g CLEC −1 at a flow rate of 4 mL min −1 corresponds to a conversion of 2.3% of the initially applied 100 mM cyclohexene oxide.
clogging could be prevented.The resulting PBR was operated using an FPLC system with two different inlets for cyclohexene oxide and azide, which enabled the use of higher azide concentrations than in the FBR.Again, productivity of the PBR (in terms of continuous-flow reaction rate) was investigated at different flow rates using a low (20 mM cyclohexene oxide) as well as a high epoxide concentration (150 mM cyclohexene oxide) and always 2 eq of azide.
Despite the fact that the continuous-flow reaction rate again increased with increasing substrate concentration, this time a 1.6fold higher reaction rate of 30.4 g product h −1 g CLEC −1 was achieved with the PBR (Figure 4) compared with the FBR when using 150 mM cyclohexene oxide and a flow rate of 4 mL min −1 .The latter might be influenced by the applied higher azide concentration in the PBR.
However, a similar increase in continuous-flow reaction rate was also ] and product enantiomeric excess (ee P ) of the formed 2-azido-cyclohexan-1-ol in the packed-bed reactor (PBR) over different flow rates (0.5, 1, 2, 4 mL min −1 ).The reaction flow contained 20 or 150 mM cyclohexene oxide and 40 or 300 mM azide, respectively, in 50 mM Tris•SO 4 , pH 7.0 with epoxide and nucleophile fed separately from two different inlets.The flow was controlled using an FPLC.The PBR contained 50 mg CLECs and 500 mg glass beads [>106 µm], the reaction was performed at room temperature.The continuous-flow reaction rate was determined by achiral GC via the attained conversion after 1 h of each run; product enantiomeric excess was analyzed by chiral GC.Reactions were performed in duplicate and resulting standard deviations are indicated.The highest achieved continuous-flow reaction rate of 30.4 g product h −1 g CLEC −1 at a flow rate of 4 mL min −1 corresponds to a conversion of 30% of the initially applied 150 mM cyclohexene oxide.
evident in the experiment using only 20 mM epoxide and 40 mM azide in the PBR at a flow rate of 4 mLmin −1 , when compared with the corresponding reaction in the FBR using the same epoxide to azide ratio (see Figure 4 and Supporting Information: Figure S10).As the reaction rate of immobilized enzymes does not necessarily increase linearly with increasing enzyme amount or catalyst loading (Ganguly & Nandi, 2015;Ryu et al., 2003), the higher CLEC amount applied in our PBR compared with our FBR might also play a role here.Further, the productivity of the PBR was overall higher than the productivity of the FBR, which is in agreement with other literature precedents (Bódalo-Santoyo et al., 1999;Lorenzoni et al., 2015).
Moreover, the product enantiomeric excess of the formed azido alcohol was again independent of the applied flow rate, but slightly higher at increased substrate concentration (Figure 4).
To further compare all reactor systems with each other, the residence time of the FBR and the PBR at a flow rate of 4 mLmin −1 was determined to enable the calculation of STYs.Thus, residence times of 23.1 s and 19.1 s for the FBR and PBR, respectively, were obtained (Supporting Information: Figure S11).As the residence time is defined as the time the fluid needs to pass the effective reactor volume, a smaller residence time for the PBR is expected, as the effective reactor volume of the PBR is smaller compared with the FBR due to the higher catalyst loading and the use of glass beads (Rodrigues, 2021).].Moreover, the PBR approach was even more productive reaching a STY of 23.9 kg product g CLEC −1 h −1 L reactor volume −1 (Table 2).This result is in line with other literature reports that compare immobilized enzyme preparations in batch and PBR applications (Planchestainer et al., 2017;Tomin et al., 2010;Zhang et al., 2018).
The fact that the PBR is twice as productive as the FBR is likely explained by the larger amount of CLECs used in the PBR as well as the higher azide concentration in that reaction.Despite this high nucleophile concentration, the resulting product enantiomeric excess of the formed azido alcohol was also highest [ee P = 64%] in the PBR reaction.This indicates that the actual impact of the nonenzymatic product formation (i.e., chemical background) was the lowest in this continuous approach.
Using the same high azide concentration in a batch reaction, a significantly lower ee P of only 41% was obtained.In contrast, the FBR yielded the worst ee P of only 33% (when using 150 mM epoxide) among all tested reactions, despite the applied lower azide concentration.The latter is actually caused by premixing of cyclohexene oxide and azide in the FBR approach.At high azide concentration, this causes an unfavorably high chemical background reaction resulting in racemic product.The use of two separate inlets for cyclohexene oxide and azide, as in the PBR approach, can significantly reduce this undesired nonenzymatic reaction.Altogether, the continuous-flow application in a PBR with two individual inlets for epoxide and nucleophile seems to be the ideal reactor concept for the biocatalytic application of our HheG D114C CLECs in epoxide ring opening reactions.This setup yielded the highest productivity and selectivity among all tested reactor systems in our study.Moreover, the achieved STY in the PBR might be further optimized by fine-tuning the CLEC amount in the reactor.
Even though a long-term operation of our CLEC-containing PBR has not been studied herein, the operational stability of the PBR is expected to be comparable with the operational stability of the FBR mentioned above.Thus, industrially relevant TON values of >1 • 10 6 mol product per mol of enzyme or ≥5000 g product per gram of enzyme (Hagen, 2015;Tufvesson et al., 2011) should be attainable with the PBR (a calculated TON of 3.2 • 10 6 mol product per mol of enzyme is obtained assuming a stable continuous-flow reaction rate of 30.4 g product h −1 g CLEC −1 over 21 days of operation).
This emphasizes once more the outstanding potential of our HheG CLECs for application in industrial processes.

| CLEC formulation
For a possible industrial application of our HheG D114C CLECs in the future, their storage in aqueous solution after crystallization and crosslinking is still impractical.Lyophilization is often used to transfer proteins into a more suitable state regarding handling, storage, and transportation (Roy & Gupta, 2004), and has been reported for other CLECs before (Abraham et al., 2004;Fernández-Penas et al., 2021;Hetrick et al., 2014;Jegan Rajan & Emilia Abraham, 2008).Therefore, we also studied a ] and product enantiomeric excess (ee P ) of the formed 2-azido-cyclohexan-1-ol achieved with the different reactor systems under optimal reaction conditions using always 150 mM cyclohexene oxide.(Supporting Information: Figure S12).Also, corresponding median values of particle size are almost identical [2.24 µm before and 2.18 µm after lyophilization].Thus, in agreement with previous reports (St. Clair & Navia, 1992), lyophilization did not affect CLEC size, and the CLECs were also not damaged during the formulation process.
To investigate a possible activity loss upon lyophilization, one aliquot of lyophilized CLECs was resuspended in 50 mM Tris•SO 4 , pH 7.0, and applied in azidolysis reactions of cyclohexene oxide.The other aliquots of lyophilized CLECs from the same batch were stored at room temperature over a period of 6 months and every month residual activity was determined as described above.This revealed a good storage stability of the lyophilized HheG D114C CLECs at room temperature (activity recovery of 67% over 6 months of storage), even though an initial activity loss of 25% upon lyophilization was observed compared with nonlyophilized CLECs (Supporting Information: Figure S13).The latter is in agreement with other literature examples (Roy & Abraham, 2006) (St. Clair & Navia, 1992).Thus, lyophilization is a convenient means of preparing dry and storagestable preparations of our HheG D114C CLECs.

| CONCLUSION
Overall, we have demonstrated that the crystallization and cross-linking of HheG D114C for CLEC generation can be successfully transferred to a stirred approach, which enables further upscaling of the process beyond the herein-used 50 mL scale in the future.Moreover, the application of those CLECs in a PBR facilitated the production of 2-azidocyclohexan-1-ol with an impressive space-time-yield of 23.9 kg product g CLEC −1 h −1 L reactor volume −1 , which would correspond to 28.7 kg product L −1 day −1 for our PBR setup containing 50 mg HheG D114C CLECs.Additionally, high operational stability of our CLECs in repetitive batch reactions over 3 weeks has been demonstrated before (Staar, Henke, et al., 2022) and is reinforced herein using continuous flow, which highlights the great potential of our HheG CLECs for industrial application.Although only cyclohexene oxide in combination with azide has been used as model substrate in our study, we anticipate that our CLEC-containing PBR can be applied as well in other epoxide ring opening reactions that have previously been reported for HheG (Calderini et al., 2019;Solarczek et al., 2022;Wan et al., 2019;Wang et al., 2022).To further facilitate their implementation and handling, lyophilization of our HheG D114C CLECs has been demonstrated to yield a storage-stable preparation retaining more than 50% of initial CLEC activity (compared with nonlyophilized CLECs) after a storage period of 6 months at room temperature.In summary, our results highlight the efficacy of HheG CLECs for continuous-flow applications of this enzyme and set the stage for their potential application in the industry.Moreover, since many HHDHs exhibit high crystallizability, we expect that CLEC generation will be an effective means for immobilization of other HHDHs as well.
2.11 | PBRA PBR was constructed using a PureCube Cartridge 1 mL FPLC column filled with 50 mg CLECs and 500 mg glass beads (>106 µm diameter, Merck) and connected to an Äkta prime FPLC (Cytiva).A schematic representation of the PBR setup is given in Supporting Information: FigureS5.Cyclohexene oxide and azide (both in 50 mM Tris•SO 4 , pH 7.0) were not premixed this time, but taken from separate inlets.Different flow rates (0.5, 1, 2, 4 mL min −1 ) and substrate concentrations (20 or 150 mM cyclohexene oxide and each 2 eq of azide) were analyzed.The continuous-flow reaction rate [g product h −1 g CLEC −1 ] and product enantiomeric excess were determined each after 1 h operation time using achiral and chiral GC, respectively.The PBR column was washed with 50 mL 50 mM Tris•SO 4 , pH 7.0 between two runs.Reactions were performed in duplicate.Space-time yield (STY) [kg product g CLEC −1 h −1 L reactor vol- ume −1 ] was calculated according to the following equation: buthylmethylether and centrifuged for 1 min at 17,000g.The upper organic phase was dried over anhydrous MgSO 4 and then used for GC analysis.Achiral separation was carried out using an OPTIMA 5 MS column (30 m length, 0.25 mm inner diameter, 0.25 µm film thickness, Macherey Nagel), whereas chiral analysis was performed on a HYDRODEX γ-DIMOM column (25 m length, 0.25 mm inner diameter, 0.25 µm film thickness, Macherey Nagel) with 40 cm s −1 hydrogen carrier gas flow.Temperature programs and respective retention times for the different compounds are listed in Supporting Information: Table

For a comparison
of the CLEC performance in different reactor systems, the STR was investigated first.The epoxide ring opening of cyclohexene oxide with azide was used as model reaction (Supporting Information: Scheme S1).Initially, the CLECs were applied in an EasyMax 102 Thermostat system with 500 rpm stirring speed over 24 h to evaluate their mechanical stability in a stirred system.Online measurement during the stirring process was performed using a focused beam reflectance measurement (FBRM) to determine the chord length distribution of the CLECs.Percentiles d 10 , d 50 , and d 90 did not decrease over 24 h of mechanical stress time (Supporting Information: Figure S6), indicating that the CLECs did not change in size and form (e.g., through breakage) during the stirring process.To confirm this initial result, additional offline particle size distribution by F I G U R E 1 Avrami-based crystallization kinetics of HheG D114C in different scales.Crystallization in 2 and 20 mL scale was performed with gentle shaking at 8°C.Crystallization in 50 mL scale was performed with stirring at 200 rpm.A final protein concentration of 12 mg mL −1 was used in each case, corresponding to 24, 240, and 600 mg of total protein in 2, 20, and 50 mL scale, respectively.The crystallization data for HheG D114C in 2 mL scale are taken from Staar, Staar, et al. (2022).
well as microscopic analysis of the CLECs before and after mechanical stress was performed.Percentiles (d 10 , d 50 , d 90 )

F
I G U R E 3 Continuous-flow reaction rate [g product h −1 g CLEC −1 ] and cumulative turnover number (TON) [mol product mol CLEC −1 ] over 21 days in the fluidized-bed reactor (FBR) achieved at a flow rate of 0.25 mL min −1 .The reaction flow contained 20 mM cyclohexene oxide and 40 mM azide in 50 mM Tris•SO 4 , pH 7.0.The flow was achieved using a KDS100 Legacy syringe pump.The FBR contained 4 mg CLECs, and the continuous reaction was performed at room temperature.Sampling was carried out every 24 h and the corresponding continuous-flow reaction rate was determined by achiral GC via the attained conversion.The mean continuous-flow reaction rate of 1 g product h −1 g CLEC −1 corresponds to a conversion of 10% of the initially applied 20 mM cyclohexene oxide.F I G U R E 4 Continuous-flow reaction rate [g product h −1 g CLEC −1 our D114C CLECs.To this end, 353 mg of CLECs from our 50 mL CLEC production were lyophilized at −35°C for 12 h, resulting in a total of 332 mg lyophilized CLECs with a high formulation yield of 95.1 mg lyophilized CLECs per liter enzyme production volume.Subsequent particle size distribution analysis of the resuspended lyophilized CLECs resulted highly similar density distribution curves for the CLECs before and after lyophilization Note: STY was calculated via the attained conversion as determined by achiral GC; ee P was determined by chiral GC.Except for the FBR reaction, which has been performed only once, reactions were performed in duplicate with standard deviations of space-time yield below ±0.03 kg product g CLEC −1 h −1 L reaction volume −1 and ee P below ±0.9%.