University of Groningen
Characterization and computation-supported engineering of an ω-transaminase Meng, Qinglong
DOI:
10.33612/diss.172243517
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2021
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Meng, Q. (2021). Characterization and computation-supported engineering of an ω-transaminase. University of Groningen. https://doi.org/10.33612/diss.172243517
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
157
SUMMARY AND PERSPECTIVES
Summary
ω-Transaminases (ω-TAs) are PLP-dependent enzymes with growing importance for the conversion of ketones to amines. The latter are used in a diversity of applications, such as the synthesis of fine chemicals. The high enantioselectivity of ω-TAs that is often observed in asymmetric amination reactions makes them especially attractive for the preparation of chiral building blocks for pharmaceutical synthesis. However, the use of ω-TAs in industrial biocatalysis is often hampered by enzyme instability and a rather limited substrate scope. The work described in this thesis is aimed at improving relevant properties of the fold-type-I homodimeric ω-TA from Pseudomonas jessenii (PjTA) via protein engineering, using computational design to find the necessary mutations (Scheme 1).
158
The transaminase PjTA was previously discovered and crystallized by our group. It naturally converts 6-aminohexanoic acid to 6-oxohexanoic acid in the caprolactam biodegradation pathway. In Chapter 2 we investigated the substrate scope of PjTA with 34 different amino donors, including aliphatic amines, aromatic amines, and proteinogenic as well as
non-proteinogenic amino acids. For this we used a coupled enzyme assay that detects formation of alanine by linking it to NADH-dependent regeneration to pyruvate by alanine dehydrogenase. PjTA displayed decent activities towards aromatic amines such as (S)-1-phenylethylamine, benzylamine, and 4-phenylbutylamine, and also with aliphatic amines like aminoheptane and 1-aminohexane. Non-proteinogenic amino acids that were accepted included 6-aminohexanoic acid, 5-aminopentanoic acid and 4-aminobutanoic acid. The activity differences with various substrates were explored by docking simulations. In comparison the well-studied ω-TAs from
Chromobacterium violaceum (CvTA) and Vibrio fluvialis (VfTA), PjTA displayed a preference for 6-aminohexanoic acid and 4-aminobutanoic acid. The modeled structures of the external aldimines of these two amino acids showed that phenylalanine side chains in CvTA (Phe89) and VfTA (Phe85) could restrict space at the active site entrance by pointing towards a conserved arginine (Arg416 in CvTA, Arg415 in VfTA). At the corresponding position of this phenylalanine, PjTA possesses the smaller Ser87 which may provide a more spacious active site entrance allowing Arg417 to form a salt bridge with the carboxylate group of an ω-amino acid. This slight difference in active site geometry allows a dual interaction of Arg417 with the substrate and the presence of Ser87 may play a role in the role of this specific enzyme in the caprolactam biodegradation pathway.
In view of the potential applications of PjTA as a biocatalyst for the synthesis of valuable amines and the very modest stability of the wild-type enzyme, PjTA was stabilized by
computational protein engineering (FRESCO) as described in Chapter 3. According to previous research, the poor stability of dimeric ω-TAs may be due to loss of the aminated cofactor PMP, which was proposed to diffuse out of the active site after the first half-reaction, which is then followed by irreversible denaturation. Subunit dissociation may facilitate this cofactor release. After computational prediction and experimental verification of mutations that enhance stability, the spatial distribution of the best stabilizing mutations and the extent of stabilization indicated that the subunit interface was critical for stability. After a rational combination of confirmed stabilizing mutations, two robust variants called PjTA-R4 (∆Tmapp = +18 °C) and PjTA-R6 (∆Tmapp = +23 °C) were obtained. These variants were more active at their respective higher optimum temperatures, more tolerant to cosolvents (DMSO and methanol) present in reaction mixtures, and better accepted high concentrations of the amine donor isopropylamine than the wild-type PjTA. With PjTA-R6, the yield of (S)-1-phenylethylamine in reaction mixtures increased to 92% (ee > 99%) under harsh reaction conditions (1 M isopropylamine as the amino donor, 100 mM acetophenone with 20% DMSO at 56 °C). The crystal structures of the PjTA-R4 and PjTA-R6 variants were solved and mostly confirmed the expected structural changes. A rarely described stabilization
159
mechanism, i.e. removal of steric strain, was identified as the effect of the most stabilizing mutation I154V. In short, this enzyme stability engineering study indicates that computational interface redesign can be a rapid and powerful strategy for the stabilization of an ω-TA.
Based on the activity of PjTA, which naturally acts on aliphatic substrates as amine donor (Chapter 2), the catalytic activity of the robust variant PjTA-R6 in the synthesis of optically pure aliphatic amines from ketones was explored using the cheap amino donor isopropylamine
(Chapter 4). The results showed that PjTA-R6 displayed better performance in terms of product yield and enantioselectivity (ee > 95%) in the synthesis of ten relevant aliphatic amines than the homologous enzymes CvTA and VfTA. Since a salt bridge between the conserved Arg417 and the carboxylate group of a keto acid acting as amino acceptor is an important and conserved feature of the active site geometry of these transaminases, it was considered that the iminium function of Arg417 could provide repulsive interactions with aliphatic substrates lacking a carboxylate group. Thus, mutant R417L was constructed since it might better accept aliphatic amines. The results showed that the PjTA-R6-R417L mutant displayed a similar performance as PjTA-R6, indicating the switching arginine Arg417 is indeed dispensable when a carboxylate functionality in the substrate is absent. For a set of aliphatic amines, the Rosetta Interface Energies of PjTA-R6 and VfTA by Rosetta docking simulations of the external aldimines exhibited a clear correlation with the yield of amine that was obtained. The docked structures revealed differences between PjTA-R6 and VfTA in switch-on/off positions for the conserved arginine, which explained some of the differences between these enzymes when producing aliphatic amines.
After obtaining two highly stable variants of PjTA (Chapter 3), the substrate scope of the most robust variant PjTA-R6 was expanded (Chapter 5). Previous studies described that the substrate scope of ω-TAs in asymmetric amination reactions is limited due to steric hindrance in the active site binding pockets that should accommodate the rest groups connected to the carbonyl carbon of the ketone. A large and a small binding pocket can be distinguished, where in case of PjTA the large binding pocket overlaps with the active site tunnel that is occupied by the
carboxyalkyl group in the crystal structure. Many bulky amines cannot be produced by PjTA and related ω-TAs. In Chapter 5, six bulky amines were selected for which PjTA-R6 has no detectable activity. A computational protein engineering strategy was explored to reshape the binding pockets of PjTA-R6. Docking simulations were used to construct models of PjTA-R6 with the external aldimine form of the target substrates bound, and the binding energies were minimized using the Rosetta search algorithm. This algorithm searches for variants with low energy by varying the identities and rotamer conformations of a chosen set of residues surrounding the bound substrate. In total seven residues (Met54, Leu57, Trp58, Tyr151, Ala230, Ile261 and Arg417) in the large binding pocket and one residue in the small binding pocket (Phe86) were included in the search space for all six substrates. The potentially improved variants with low Rosetta interface energy were further filtered through visual inspection and for each target substrate a small library was created for experimental verification. These six small libraries contained 40 unique designs in total, and when tested in the laboratory 38 of these 40 mutants indeed produced the targeted enantiopure
160
amine (ee > 99%). Moreover, the stability of most mutants was not compromised by the introduced mutations. It appeared that the yields of six selected bulky amines were strongly correlated to Rosetta interface energies, thus illustrating the predictive power of Rosetta. The W58G mutant displayed the best performance in the synthesis of five structurally similar bulky amines, while the best mutant for the synthesis of the remaining bulky amine was W58M+F86L+R417L. The crystal structures of these two best variants were solved and mostly confirmed the modeled structures. Replacing Trp58 in the large binding pocket was identified as a key step in the redesign of PjTA-R6 to accept bulky substrates. Overall, the use of an efficient and accurate computational
methodology was demonstrated for expanding the substrate scope of PjTA-R6. We expect it to be a similarly useful strategy for the engineering of other ω-TAs.
Perspectives
In the literature (reviewed in Chapter 1) and in this work, several ω-TAs variants with an increased stability and expanded substrate scope have been obtained by protein engineering. The work on PjTA has produced engineered variants that can be used in the asymmetric synthesis of chiral amines, including aliphatic amines (Chapter 4), aromatic amines (Chapter 3) and bulky amines (Chapter 5). However, the yields of some bulky amines in laboratory-scale
biotransformations were limited. For example, the best variant for the synthesis of (S)-6-methoxytetralin-1-amine (PjTA-R6-W58G) only gave a 29% product yield (Chapter 5) when tested with a high concentration of the cheap amino donor isopropylamine.
In many cases the limited yields that are observed are probably due to the reactions reaching thermodynamic equilibrium. This obviously cannot be solved by further enzyme engineering but will require bioprocess engineering, where product removal is an evident option. Thus, the equilibrium of transaminase reactions with isopropylamine as the amino donor can possibly be shifted to the direction of amine accumulation by removing the volatile coproduct acetone with the desired amine staying in the reaction mixture. In Chapter 5, a low air pressure (40 kPa) was tested for its effect on amine synthesis, but the yield was not improved since the substrate ketone was also lost under low air pressure. Ways to remove acetone while retaining ketone in the reaction mixture have been studied. For instance, hydrophobic pervaporation using the ability of polydimethylsiloxane membranes to separate an organic component from an aqueous stream was used to improve acetophenone amination1. This method reduced, but did not eliminate, ketone loss and improved product yield by 13%. To prevent loss of the ketone substrate,
continuous in situ acetone removal using a selective membrane was also studied2. This gave a substantially improved conversion of acetophenone (from 50% to 98%). Coupled reactions for improving product yields also deserve attention. At small scale, an enzymatic method with in situ use of a specific dehydrogenase was developed to remove acetone3. The system consisted of a yeast alcohol dehydrogenase from S. cerevisiae (acetone reduction) and a formate dehydrogenase from Candida boidinii (NADH regeneration) and could drive the equilibrium to the direction of
161
amine synthesis. With four selected amines, production increased to 99%. These studies on acetone removal deserve further investigation, taking into account the large scale on which ketone to chiral amine conversions will need to be performed if applied for the production of pharmaceutical intermediates. The availability of highly robust enzyme variants with tailored substrate scope will facilitate such studies on acetone removal and reduce process costs when scaling up.
The other prevailing amino donor in asymmetric amine synthesis is alanine, which is converted to the coproduct pyruvate. Pyruvate can be removed by oxidation to lactate, to
acetaldehyde by decarboxylation, or to alanine by regeneration using alanine dehydrogenase. The latter reaction goes at the expense of NADH, which in turn can be regenerated by glucose
dehydrogenase or formate dehydrogenase. Such a cascade has been shown to improve production of aliphatic amines and aromatic amines like butylamine, pentylamine,
(S)-2-octylamine, and (S)-1-phenylethylamine using the (S)-selective ω-TA termed ATA-1134. With alanine as the amino donor, product removal requires cascades that employ two or more enzymes which complicates the process of amine synthesis. Overall process economics and sustainability issues will determine which amine donor is most attractive in practice.
Besides transaminases, amine dehydrogenases have been also explored to produce chiral amines from prochiral ketones. Amine dehydrogenases are NADH-dependent enzymes that catalyze the reductive amination of ketones, and glucose (or formate) dehydrogenase is used to regenerate NADH5. Amine dehydrogenases are generally obtained by protein engineering of amino acid dehydrogenases such as leucine dehydrogenase from Bacillus stereothermophilus5, phenylalanine dehydrogenase from Bacillus badius6, L-lysine dehydrogenase from Geobacillus stearothermophilus7, and 4-oxopentanoic acid dehydrogenase from Clostridium sticklandii8. Several chiral amines including aliphatic amines like (R)-2-pentylamine, (R)-2-hexylamine and aromatic amines like (R)-1-phenylethylamine, (R)-1-phenylbutylamine, (R)-(+)-1-aminotetralin and (S)-pentan-2-amine have been produced by engineered amine dehydrogenases. Several (R)-amines can also be synthesized by ω-TAs. A prominent example is the biocatalytic sitagliptin production process that is based on a highly engineered variant of the (R)-selective ω-TA from Arthrobacter sp. (ATA-117)9. The choice between amine dehydrogenase- and transaminase-based biocatalytic routes will require a thorough comparison of bioprocess engineering aspects, but obviously also the possibility to acquire enzymes with the required activity will be of key
importance. We expect that the work in this thesis and studies by other groups will contribute to the predictability of protein engineering projects aimed at tailoring transaminases for specific
conversions, which is important when selecting a certain approach for developing a biocatalytic process.
The results presented in the thesis also suggest steps to further improve the protein engineering protocols that were used. In Chapter 3, the procedure of mutant library design by FRESCO contains different in silico screening steps before experimental verification: folding energy calculations, high-throughput MD simulations to predict local flexibility, and visual inspection. Each step can eliminate a large amount of potentially negative mutations. Stabilizing
162
mutations missed in the protocol can possibly be discovered by including additional principles of protein stabilization, such as the introduction of disulfide bond10, which may include inter-subunit disulfide bonds. Furthermore, stabilization of flexible surface loops or replacing such loops by rigid ones may increase the stability of a protein11. Whereas computationally designed surface mutations reported in Chapter 3 had little effect on the stability of wild-type PjTA, it is well possible that they would further enhance the stability of variants that have already been improved by other methods, because the weakest spots in the structure may have shifted. In the current work we did not use biophysical methods to establish the regions of PjTA where unfolding may start. Finding these so-called early unfolding regions and focusing mutations there may further improve the stability of the robust variants.
In Chapter 5, PjTA-R6 was used as the template for engineering transaminases that can produce bulky amines. Whereas nice designs were obtained, further studies aimed at a larger group of bulky amines should be done to fully explore the possibilities of PjTA redesign. An interesting compound would be (R)-1,2-diphenylethylamine, which has the amino group located between two phenyl groups. The reaction has been shown with an engineered CvTA12. It would be important to develop in-silico strategies for selecting the best template enzyme for a certain target reaction. Engineering PjTA for production of such very bulky amines would certainly offer a challenging protein engineering project. Furthermore, only ketones are tested in this thesis as substrates for amination reactions. Acceptance of aldehydes as substrates by ω-TAs allows synthesis of primary amines, and it would be important to test the robust variant PjTA-R6 in the acceptance of a series of aldehydes such as benzaldehyde and 4-phenylbutyraldehyde. The wild-type PjTA gave a good activity in the deamination of the corresponding amines (benzylamine and 4-phenylbutylamine) (Chapter 2). Even though the complex kinetics of transaminases does not imply that good activity with a compound as an amino donor implies good activity in the production of that compound from ketone, some correlation is to be expected for mechanistic and thermodynamic reasons.
An interesting primary amine is cinnamylamine, which is the precursor of the antifungal agent naftifine. That compound has been produced from cinnamyl alcohol in a cascade involving five enzymes, including galactose oxidase (oxidation of cinnamyl alcohol), horseradish peroxidase (regeneration of the copper ion cofactor of galactose oxidase), transaminase (amination of
cinnamaldehyde), alanine dehydrogenase (regeneration of the amino donor alanine), and glucose or formate dehydrogenase (regeneration of the cofactor NADH)13. Although this might be a feasible strategy for producing cinnamylamine, a simpler cascade which contains three enzymes, i.e. alcohol dehydrogenase (oxidation of cinnamyl alcohol and regeneration of the cofactor NADH), PjTA-R6 or a similar transaminase (amination of cinnamaldehyde), and alanine dehydrogenase (regeneration of the amino donor alanine at the expense of NADH) can be considered.
Furthermore, the synthesis of cinnamylamine from cinnamyl aldehyde using isopropylamine as the amino donor may also be feasible.
Besides amines, non-proteinogenic amino acids are also interesting targets for further investigation because the wild-type PjTA naturally deaminates aminohexanoic acid to
6-163
oxohexanoic acid and also displays decent activities towards 5-aminopentanoic acid and 4-aminobutanoic acid (Chapter 2). Different non-proteinogenic amino acids such as 3-amino-3-phenylpropionic acid and L-norvaline have been studied with other ω-TAs14,15. Related high-value chemicals including amino alcohols such as (R)-phenylglycinol and (R)-3-amino-1-butanol have also been synthesized by engineered ω-TAs16,17. The amines which are used as pharmaceuticals such as sitagliptin9 and imagabalin18 have been produced by engineered ω-TAs. Production of some other pharmaceutical chemicals carrying an amino group like fluvoxamine19 by ω-TAs has not been reported. These examples illustrate that the diversity of products that deserves exploration with engineered variants of PjTA-R6 is broader than the range of compounds discussed in this thesis.
In our study, almost all the produced chiral amines were (S)-enantiomers with >99% e.e. However, the (R)-amines such as (R)-1-phenylethylamine have not been synthesized by PjTA. A protein engineering study aimed at switching the enantioselectivity from S to R at first sight seems challenging, but probably not competitive to an approach that takes an R-selective fold-type IV transaminase as the template. The general idea would be to create more space for the small binding pocket of the active site to fit the large group of the substrate, and to simultaneously reduce the large binding pocket to fit the small group of the substrate. This strategy has been applied to switch the enantioselectivity of CvTA in the synthesis of 2-aminotetraline20.
Computational methods are at the heart of the protein engineering studies reported in this thesis. Computational docking simulations were used to explain activity differences between enzymes and substrates (Chapters 2 and 4). Computational design was used in Chapters 3 and 5 to discover variants of PjTA with improved stability and tailored substrate scope, respectively. In each case small libraries were created by computational prediction of variants that are likely to display the desired properties, which was confirmed by experimental verification. However, often the success rate of computational design is still modest, making it too challenging to design the single best variant for a certain performance in a one-step protocol. The current protocols might be optimized via a deep understanding of an enzyme catalytic mechanism before computational design, and different algorithms could be used in parallel to increase the success rate of
computational predictions. In Chapter 3, the success rate for mutations designed to improve the stability of PjTA was only 13%. The high success rate (56%) for the predicted interface mutations suggests that selection of search space based on insight in underlying mechanisms (in this case subunit dissociation) can contribute enormously to the predictability of computational design. Another way to improve computation-supported design would be to use bioinformatics. As
described in the Introduction of this thesis (Chapter 1), towards-consensus mutations can improve stability, and methods that combine computational design and the consensus approach have been reported. Sequence alignments in combination with functional predictions by bioinformatics may also become useful for selecting templates and mutations that contribute to protein engineering aimed at a desired activity or specificity.
164
References
1. Satyawali, Y., del Pozo, D. F., Vandezande, P., Nopens, I. & Dejonghe, W. Investigating
pervaporation for in situ acetone removal as process intensification tool in ω-transaminase catalyzed chiral amine synthesis. Biotechnol. Prog. 35, e2731 (2019).
2. Rehn, G., Adlercreutz, P. & Grey, C. Supported liquid membrane as a novel tool for driving the equilibrium of ω-transaminase catalyzed asymmetric synthesis. J. Biotechnol. 179, 50–55 (2014). 3. Cassimjee, K. E., Branneby, C., Abedi, V., Wells, A. & Berglund, P. Transaminations with isopropyl
amine: equilibrium displacement with yeast alcohol dehydrogenase coupled to in situ cofactor regeneration. Chem. Commun. 46, 5569–5571 (2010).
4. Koszelewski, D. et al. Formal asymmetric biocatalytic reductive amination. Angew. Chem. Int. Ed.
47, 9337–9340 (2008).
5. Abrahamson, M. J., Vázquez-Figueroa, E., Woodall, N. B., Moore, J. C. & Bommarius, A. S. Development of an amine dehydrogenase for synthesis of chiral amines. Angew. Chem. Int. Ed. 51, 3969–3972 (2012).
6. Abrahamson, M. J., Wong, J. W. & Bommarius, A. S. The evolution of an amine dehydrogenase biocatalyst for the asymmetric production of chiral amines. Adv. Synth. Catal. 355, 1780–1786 (2013).
7. Tseliou, V., Knaus, T., Masman, M. F., Corrado, M. L. & Mutti, F. G. Generation of amine dehydrogenases with increased catalytic performance and substrate scope from ε-deaminating L-Lysine dehydrogenase. Nat. Commun. 10, 3717 (2019).
8. Mayol, O. et al. A family of native amine dehydrogenases for the asymmetric reductive amination of ketones. Nat. Catal. 2, 324–333 (2019).
9. Savile, C. K. et al. Biocatalytic asymmetric synthesis of chiral amines from ketones applied to sitagliptin manufacture. Science. 329, 305–309 (2010).
10. Xie, D. F. et al. Improving thermostability of (R)-selective amine transaminase from Aspergillus terreus through introduction of disulfide bonds. Biotechnol. Appl. Biochem. 65, 255–262 (2018). 11. Nestl, B. M. & Hauer, B. Engineering of flexible loops in enzymes. ACS Catal. 4, 3201–3211
(2014).
12. Land, H., Ruggieri, F., Szekrenyi, A., Fessner, W. & Berglund, P. Engineering the active site of an (S)-selective amine transaminase for acceptance of doubly bulky primary amines. Adv. Synth. Catal.
362, 812–821 (2020).
13. Fuchs, M. et al. Amination of benzylic and cinnamic alcohols via a biocatalytic, aerobic, oxidation-transamination cascade. RSC Adv. 2, 6262–6265 (2012).
14. Hwang, B. Y. & Kim, B. G. High-throughput screening method for the identification of active and enantioselective ω-transaminases. Enzyme Microb. Technol. 34, 429–436 (2004).
15. Han, S. W., Park, E. S., Dong, J. Y. & Shin, J. S. Active-site engineering of ω-transaminase for production of unnatural amino acids carrying a side chain bulkier than an ethyl substituent. Appl. Environ. Microbiol. 81, 6994–7002 (2015).
16. Nobili, A. et al. Engineering the active site of the amine transaminase from Vibrio fluvialis for the asymmetric synthesis of aryl-alkyl amines and amino alcohols. ChemCatChem. 7, 757–760 (2015). 17. Gao, X. et al. Reshaping the substrate binding region of (R)-selective ω-transaminase for asymmetric
synthesis of (R)-3-amino-1-butanol. Appl. Microbiol. Biotechnol. 104, 3959–3969 (2020).
18. Midelfort, K. S. et al. Redesigning and characterizing the substrate specificity and activity of Vibrio fluvialis aminotransferase for the synthesis of imagabalin. Protein Eng. Des. Sel. 26, 25–33 (2013). 19. Figgitt, D. P. & McClellan, K. J. Fluvoxamine: An updated review of its use in the management of
adults with anxiety disorders. Drugs. 60, 925–954 (2000).
20. Humble, M. S., Cassimjee, K. E., Abedi, V., Federsel, H.-J. & Berglund, P. Key amino acid residues for reversed or improved enantiospecificity of an ω-transaminase. ChemCatChem. 4, 1167–1172 (2012).
