Development of Continuous Flow Systems to Access Secondary Amines Through Previously Incompatible Biocatalytic Cascades

Abstract A key aim of biocatalysis is to mimic the ability of eukaryotic cells to carry out multistep cascades in a controlled and selective way. As biocatalytic cascades get more complex, reactions become unattainable under typical batch conditions. Here a number of continuous flow systems were used to overcome batch incompatibility, thus allowing for successful biocatalytic cascades. As proof‐of‐principle, reactive carbonyl intermediates were generated in situ using alcohol oxidases, then passed directly to a series of packed‐bed modules containing different aminating biocatalysts which accordingly produced a range of structurally distinct amines. The method was expanded to employ a batch incompatible sequential amination cascade via an oxidase/transaminase/imine reductase sequence, introducing different amine reagents at each step without cross‐reactivity. The combined approaches allowed for the biocatalytic synthesis of the natural product 4O‐methylnorbelladine.


Introduction
Multi-step synthesis in biocatalysis has been made more efficient through the utilization of enzymatic cascades.N o longer are reactions limited by intermediate isolation, thus significantly reducing time and waste. [1,2] hed esign of multienzyme systems has provided convenient syntheses of an umber of high value compounds and are now seen as the method of choice in implementing biocatalytic reactions. [3,4] mputer-aided synthesis planning (CASP) is also poised to revolutionize biocatalytic synthesis planning,a llowing in silico prediction of novel biocatalytic cascades. [5]Nevertheless,while biocatalytic cascades are increasingly considered as the method of choice for more complex syntheses,t here are still limitations to one-pot systems due to incompatible enzyme/substrate combinations.
Flow chemistry has seen rapid development in recent years,w ith the potential to improve about 50 %o fc hemical processes. [6]Small scale flow systems can be translated into larger scale production with minimal optimization as the systems can be run for longer to increase productivity,aluxury not afforded in batch processes which need to be scaled up. [7,8] other advantage is that continuous flow systems can be composed of different modules (different reactor types), which can enable the combination of ab road range of chemistries that are incompatible under batch conditions (Figure 1). [9]Further to this,the utilization of continuous flow systems facilitates the integration of several reaction steps resulting in telescoped synthetic sequences. [10]ssues that arise with analytical biotransformations not being translatable to scale or different biocatalytic reaction types being incompatible,h as led to the emergence of continuous flow biocatalysis as afavorable option to alleviate these issues. [11]Operating under ac ontinuous regime allows for greater control of reaction conditions as well as providing opportunities for inline analysis and purification, as well as automation. [12,13] 16] Thec ombination of immobilization and continuous flow reactors means that enzymes can be compartmentalized and segregated, allowing access to previously incompatible reactions in sequential reactor modules.Herein, efforts to exploit the power of biosynthetic cascades with ac ontinuous,c ompartmentalization approach containing multiple reactor types and ac ombination of free and immobilized enzyme are described.Issues of cross reactivity and reaction incompatibility in linear enzymatic cascades are addressed whilst also showing compartmentalized flow systems can be used to overcome metabolic flux issues associated with in vitro cascades (Figure 1). [17]Thea im was to adopt am odular approach using standard laboratory equipment to generate versatile systems which enable access to ab road range of chemistries.

Results and Discussion
Ther eaction cascades described here started with the generation of aldehydes from stable and commercially available alcohols.A sa ldehydes are versatile yet unstable intermediates,i tw as thought ac ontinuous flow system that generated this group in situ at high concentrations could allow for ar ange of subsequent enzymatic modifications.B iocatalytic oxidations are fundamental in transitioning to ab iobased economy,inp articular the use of oxidases which carry out selective oxidation using molecular oxygen as the sole oxidant. [18,19] [22] An engineered choline oxidase (AcCO 6 )w as initially chosen as an ideal biocatalyst to test in the MPIR due to its broad substrate scope and as it has been applied in biocatalytic cascades. [23,24]sing the MPIR, AcCO6 productivity was vastly improved for the oxidation of an umber of aromatic and aliphatic primary alcohols with a5 8f old improvement in space time yield and 4f old improvement in productivity in the oxidation of phenylethanol (see the Supporting Information).Thebiooxidations were further optimized to ensure no over-oxidation to the corresponding acid was observed, an issue associated with batch reactions and ap revious flow report for this enzyme. [14]ollowing the successful implementation of the MPIR for the generation of aldehydes,t he system was expanded by carrying out subsequent continuous bio-reductive aminations.Amines are important molecules in synthetic chemistry,with as ignificant proportion of the reactions performed by medicinal chemists reported to be associated with aminogroup chemistry. [25]Biocatalytic amine synthesis has been advanced by transaminases and imine reductases/reductive aminases (IREDs/RedAms), with the use of aminating biocatalysts providing selective and sustainable options for the synthesis of chiral amines. [26,27] ne of the most frequently applied aminating enzymes in flow is transaminase. [28,29] erefore,t he transaminase from Bacillus megaterium (BmTA)w as immobilized and loaded it into ap acked-bed reactor. [23,30] his was connected to the output of the MPIR (with inline mixing via am icro static mixer), generating an MPIR-packed bed system (MPBS).Pleasingly,f ull conversion to the corresponding primary amine 2 was achieved at steady state and maintained for four hours (STY: 1.58 gL À1 h À1 ).
Fort he generation of secondary amines such as 3 from aldehyde 1,R edAm and glucose dehydrogenase (GDH) combinations can be used and have also been demonstrated in flow systems previously. [14,31,32] Hre,t he reductive aminase from Ajellomyces dermatitidis (AdRedAm, 200 mg, 10 wt %) and the GDH from Bacillus megaterium (BsGDH, 10 mg, 10 wt %) were immobilized and both loaded into ap acked bed reactor.Pleasingly,this also gave full conversion of the in situ generated hydrocinnamaldehyde 1 to the corresponding N-allyl amine 3 at steady state which was maintained for four hours (STY: 2.1 gL À1 h À1 )( Table 1).
Following the optimization of AcCO 6 -amination cascades in continuous flow,t he combination of previously batchincompatible enzyme systems was investigated.[35] Using GOase instead of choline oxidase would lead to ap anel of benzylic amines as shown in Scheme 1.However, due to the reactive copper center in GOase,a mines can inhibit the biocatalyst meaning one-pot batch reactions are not feasible with aminating enzymes.I ndeed, when the cascades were carried out in batch, performance as expected was poor,with no observable conversion to the corresponding amines (see Supporting Information).
a] [a] Key:*Steady state conversions determined by GC-FID and compared to chemical standards.

Forschungsartikel
To overcome these issues the use of compartmentalized MPBS was investigated.Pleasingly,e xtremely efficient cascades were observed with > 95 %conversion achieved in the reductive amination of five benzaldehydes 4 (generated in situ by GOase in the MPIR, Scheme 1), using the transaminase from Vibrio fluvialis (VfTA)a nd AdRedAm to generate the primary 5 and secondary amines 6,respectively.Fresh samples of immobilized enzyme preparations were used for each of the described systems to avoid the need to study long term stability between experiments.Importantly,due to immobilization of the enzymes,s everal substrates could be screened using the same columns.T his approach efficiently surveyed the activity of VfTA and AdRedAm in the reductive amination of the benzaldehyes 4a-e (20 mM:five or 10 equiv.amine).At otal system residence time of < 38 mins was enough to obtain > 90 %conversion in all instances.Apanel of 10 primary 5a-e/secondary amines 6a-e (10 mM) were generated merely by changing the flow path and the initial benzyl alcohol, an approach far less time consuming than the preparation of individual batch reactions and enzyme preparations.T ofurther increase efficiency and automation, three-way switching valves (including aretrofitted HPLC waste valve) were added to the system (Scheme 1);this enabled exquisite control of the flow path without the laborious need of changing lines or columns.U sing this approach, the system could be quickly switched between "RedAm" and "TA" flow paths without intermittent washing steps,t hus vastly improving efficiency.T he same approach was also used for substrate feeding to the MPIR again greatly improving the efficiencyofthe system.These cascades represent the first demonstration of continuous biocatalyst substrate screening.This streamlined flow system has the potential to complement current high throughput biocatalytic screening methods in future, [36] and ultimately facilitate automation for biocatalytic cascade discovery in ahighly efficient manner.
Following the successful implementation of the oxidase-bioamination cascades,the wide scope of the system was demonstrated with in situ amine donor generation in aT A-RedAm cascade (Table 2).Combinations of TA and RedAm/IRED in batch are not compatible because both enzymes use amine substrates with potential for cross reactivity.H ence,t he in situ generation of primary amines by TAs for use in subsequent reductive aminations by RedAm/IRED has yet to be explored.Fort he initial flow bioamination cascade, BmTA and AdRedAm (with BsGDH) were immobilized on EziG amber support (200 mg 10 wt %) and placed in separate packed bed reactors,t op revent the previously mentioned potential issues for cross amination.Initially,asolution containing butanal (40 mM), racemic alanine (400 mM) and PLP (1 mM) was passed through the TA module,a chieving full conversion to butylamine 7 at steady state with aresidence time of 12 minutes.T he effluent of this reaction was mixed via am icrostatic mixer (see Supporting Information) with another solution containing cyclohexanone (10 mM), NADP + (1 mM) and glucose (50 mM).Initial results with AdRedAm showed about 20 % conversion at steady state,w ith four equivalents of amine 7 necessary (20 mM butylamine 7,and 5mMketone in RedAm module).To further improve the productivity of this reaction am ore suitable RedAm was selected from am etagenomics panel. [37]TheR edAms were identified using ap reviously described screen containing 384 different enzymes,w hich gives ac olourimetric indication of activity using ac oupled diaphorase system (see supporting information and references). [37,38]Fort he generation of Nbutylcylohexlyamine 10,I R-79 was selected and immobilized on the EziG amber support (100 mg, 10 wt %), with 85 %c onversion to the secondary amine 10 observed and maintained for three hours at steady state (Table 2).This represented atotal system residence time of 21 mins (STY: 1.87 gL À1 h À1 ), and only required two equivalents of butylamine 7 (20 mM butylamine and 10 mM ketone in RedAm module).This approach was expanded to generate N-(2-phenyl)ethyl cyclohexylamine 11 using another metagenomic RedAm (IR-23), again only requiring two equivalents of amine with as ystem residence time of 24 mins (STY: 2.53 gL À1 h À1 ).At hird approach was then used to generate N-benzylcyclohexylamine 12 with the transaminase from Pseudomonas putida (PpTA) [39] mediated trans-amination of benzaldehyde followed by reductive amination with cyclohexanone using IR-79.Pleasingly 95 %c onversion to 12 was achieved and maintained for four hours with as ystem residence time of 24 mins (STY: 2.24 gL À1 h À1 ).Again, the implementation of several switching valves allowed rapid change of flow path and enabled the generation of three RedAm products (10, 11, 12)w ithout isolation or purification of any intermediates.
De novo designed biocatalytic cascades are often limited to 2-4 steps due to incompatibility issues as discussed above, which become more limiting as the number of steps increases.After demonstrating that MPBS could over-come these issues,acompartmentalized six-enzyme cascade was targeted, highlighting that multiple reactor technologies can be combined to enable previously inaccessible reaction sequences.A synthetic sequence that was able to convert primary alcohols into amines via an oxidase-TAsequence,which could then be used as the amine substrate for aR edAm reaction was envisaged (Table 3).Using the compartmentalized MPBS, AcCO6 efficiently oxidized phenylethanol (120 mM).The resulting aldehyde 14 was mixed with racemic alanine (400 mM) and PLP (1 mM) via am icrostatic mixer, which then passed through immobilized BmTA forming 2-phenylethylamine 8.T he effluent from this reaction was mixed with aR edAm line (glucose (50 mM), NADP + (1 mM), cyclohexanone (10 mM)) and passed through immobilized IR-79 to quantitatively generate 11 under continuous conditions.Pleasingly,a mine donors were generated in two biocatalytic steps from the alcohols,w hich then fed directly into the RedAm reactors that successfully utilized the in situ generated amine substrates.A ll three examples reached steady state conversions of > 90 %.
a] [a] Cascades were applied in amulti packed-bed reactor system with switching valves enabling efficient control of flow path.Conversions to primary and secondary amines were calculate by comparison of GC-FID spectra with product standards.K ey:* Reductive amination carried out by AdRedAm.
a] [a] Conversion was determined by GC-FID.

Angewandte Chemie
Forschungsartikel (Scheme 2). [5]4O-Methylnorbelladine is an important precursor in the synthesis of an umber of pharmacologically active alkaloid scaffolds,s uch as the dementia drug galantamine. [40,41] orthis cascade,itwas thought tyramine 16 could be generated via ac ontinuous oxidase-TAc ascade.C holine oxidase activity was tested against tyrinol in the MPIR.The resulting biogenic aldehyde is ak ey intermediate in the biosynthesis of benzylisoquinoline alkaloids and often requires multistep syntheses. [42,43] espite the biooxidation proving highly efficient (120 mM substrate loading; > 90 % conversion;S TY: 8.37 gL À1 h À1 ;p roductivity: 5.1 g product g enzyme À1 )t yramine was commercially available at al ower cost than the alcohol so this step was not necessary.
Fort he reductive amination of isovanillin and tyramine 16,t he previously described high throughput colorimetric assay was used to identify potential RedAms. [38]From this screen IR-80 was selected and the reaction validated with an analytical scale biotransformation.Theproduct 17 was found to be insoluble in aqueous buffer and proved difficult to analyze by HPLC.A sa na lternative,t he crude biotransformation was filtered, and the residue suspended in DMSO/ MeOH for MALDI analysis with an observed m/z of 274.2 corresponding to the protonated adduct of the product 17 (see Supporting Information).Encouraged by this result, IR-80 was immobilized on EziG amber (150 mg, 10 wt %) for translation into continuous flow (Scheme 2).To implement ac ascade for this synthesis,aGOase M 3-5 (5 mg mL À1 ) mediated biooxidation of isovanilyl alcohol 15 (30 mM) was carried out in the MPIR.Thee ffluent was mixed with the RedAm feed (tyramine 16 (50 mM), NADP + (0.1 mM) and glucose (50 mM)) which was flowed through an IR-80 module to quantitatively generate 17 for four hours with as ystem residence time of 36 min (STY: 2.26 gL À1 h À1 ).Filtration of the effluent allowed recovery of the excess tyramine and provided asimple purification method for the isolation of 17.

Conclusion
Fort he first time it has been shown that previously incompatible enzyme cascades can be run continuously, greatly improving the metrics (up to 58-fold for STY and 4fold for enzyme productivity) when compared to the equivalent batch reactions.T his system is extremely versatile with several enzyme combinations tested that enabled the gen-eration of arange of amine intermediates.T he reactors were also extended to demonstrate proof-of-concept, high throughput reaction screening in continuous flow.Acontinuous synthesis of the biologically relevant molecule 4O-methylnorbelladine was also successfully demonstrated.Using ac ontinuous approach allows many additional benefits to those discussed here to further improve biocatalytic reaction efficiency.I np articular,r ecycling of co-factors is made operationally more simple with flow systems which significantly reduces the economic impact. [29,44] urthermore,c ombination with automated high throughput biocatalyst screening, [36] and computer aided synthesis planning, [5] paves the way for fully autonomous biocatalytic synthesis.

Figure 1 .
Figure 1.Advanced compartmentalized continuousf low systemsc an be used to overcome incompatible enzyme combinations/chemistries can deliver the lab-based technological advancementnecessary to fully realize this transformation.

Scheme 1 .
Scheme 1. GOase M 3-5 and AdRedAm/VfTA were applied in the MPBS continuous flow system and used for cascade screening.Steady state conversions determined by GC-FID analysis.Triangle in circle:pump;c ross in circle:mixing valve.