Open Access

Chiral amines represent high-value fine chemicals serving as key intermediate products in pharmaceutical, chemical and agrochemical industries. In the past decades, application of amine transaminases (ATAs) for stereoselective amination of prochiral ketones emerged to an environmentally benign and economically attractive alternative to transition metal-catalyzed asymmetric synthesis to afford optically pure amines at industrial scale. However, the restricted substrate scope of wild-type transaminases prohibited the conversion of particularly sterically demanding substrates, making protein engineering indispensable. The following thesis covers elaboration of a novel assay for transaminases (Article I) and identification and development of transaminase variants in order to achieve biocatalytic preparation of a set of pharmaceutically relevant model amines, ideally in optically pure form for both stereoisomers, preferentially using asymmetric synthesis and most preferably using isopropylamine as cost-efficient amine donor co-substrate (Article II-IV). The aforementioned target amines and the corresponding precursor ketones (see Scheme 4.1) were conceived and provided by the company F. Hoffmann-La Roche to attain suitable biocatalysts for a variety of potential intermediates for active pharmaceutical ingredients. Protein engineering of the transaminase scaffolds investigated in this thesis comprised: Initial screening for suitable starting enzyme scaffolds, structure-guided rational design of these scaffolds to enable bulky planar substrate acceptance, elaboration of a sequence motif, verification of the motif and preparative-scale asymmetric synthesis reactions (Article II). For non-planar and structurally different target substrates, namely spatially bulky or bi-cyclic bridged substrates, the transaminase variants were specifically refined and a different evolutionary route had to be pursued (Article III and Article IV). These results (Article II) represent not only the first successful endeavor to engineer a PLP-fold type I amine transaminase (commonly denoted as (S)-selective) for the conversion of highly sterically demanding substrates, but also generally expanded the scope of available fold type I amine transaminases by enzymes having a novel and exceptionally broad substrate spectrum. Aside from structure-guided rational protein engineering, as well non-rational methods, such as site-specific saturation mutagenesis or directed evolution, were applied for protein-engineering. In order to do so for all of the target compounds, a novel high-throughput solid phase activity assay for transaminases that was actually developed during the master thesis, was refined and published (Article I). In the context of this thesis, the same assay principle was as well adapted for quantification of specific activities in liquid phase (Article III). A comparison of different methodologies for developing agar plate assays and a detailed step by step protocol of our transaminase assay are illustrated in a book chapter.

Chiral primary amines are valuable building blocks for many biologically active compounds. Environmentally friendlier alternatives to the classical methods for α-chiral primary amine synthesis are highly desired. A biocatalytic alternative that recently proved beneficial for industrial applications is asymmetric synthesis utilising (S)-selective amine transaminases (S-ATAs). These enzymes can be utilized to transaminate a prochiral ketone with an amino donor (e.g. isopropylamine), to achieve a chiral amine and a carbonyl product (e.g. acetone). However, for several potential applications protein engineering is required to fit (S)-ATAS to the demands of an industrial process. Since no (S)-ATA crystal structure required for understanding the substrate recognition and thus protein engineering was available, we first aimed at obtaining structural data. Instead of solving crystal structures ourselves, we took advantage of structural genomics projects and discovered, that the protein data bank (PDB) already contained crystal structures of four enzymes with unknown function that we hypothesised to possess (S)-ATA activity. After developing a screening method, the four enzymes could be characterized as ω-amino acid:pyruvate transaminases (ωAA:pyr TAs). (S)-amine conversion was suggested to be a ‘substrate-promiscuous’ activity of these enzymes, as it is pronounced differently in the four investigated ones. By comparing the active sites of the highly and poorly active (S)-ATAs, the residues that determine the ability of amine conversion in these enzymes were discovered. Furthermore, the mechanism for dual substrate recognition, the binding of both, carboxyl and bulky hydrophobic substrates in the same active site, could be elucidated with the crystal structures. A flexible arginine side chain is able to adopt various positions thus enabling carboxylate binding and by ‘flipping’ out of the active site, to create space for amine binding. Then, a limitation of these enzymes, the restricted substrate scope caused by a small binding pocket was addressed. First, a rational protein engineering approach was set up to create more space. The tested mutations, however, destroyed most of the activity for both regular and more bulky substrates. We thus learned that the structural requirements for (S)-ATA activity are more complex than initially anticipated and a semi-rational approach was applied to broaden the substrate scope. By systematic saturation of active site positions, substantially improved mutants for bulkier amine synthesis could be obtained. As this study highlighted a lack of understanding of (S)-ATA, the functional important residues in the enzymes belonging to the class III TA family were surveyed. This family is defined by common sequence and structure features and besides (S)-ATAs mainly comprises TAs of various substrate scopes but also a few phospholyases, racemases and decarboxylases. To enable the comparison of active site residues among them, a commercial bioinformatics tool was used to create a family wide structure-based alignment of around 13,000 sequences. Based on statistical analyses of this alignment, structural inspections and literature evaluation, active site residues crucial for certain specificities within this family have been identified. By investigating the ingenious active site designs that enable such a plethora of reactions, and by identifying sets of functional important residues termed ‘active site fingerprints’, the understanding of catalysis in this enzyme family could be broadened. Furthermore, these functional important residues can on the one hand be applied to predict the specificity of uncharacterised enzymes, if a fingerprint is matched. On the other hand, if no fingerprint is matched, they can help to discover yet unknown activities or mechanisms to achieve a known specificity. We exemplified the latter case by functionally characterising a Bacillus anthracis enzyme with the crystal structure 3N5M, whose substrate specificity was unknown and could not be predicted. The 3N5M enzyme was found to possess ωAA:pyr TA and (S)-ATA activity even though it lacks the above-mentioned ‘flipping’ arginine. Based on molecular dynamics simulations we were able to propose an alternative mechanism for dual substrate recognition in the B. anthracis ωAA:pyr TA. By these findings the understanding of the requirements for (S)-ATA activity could be further broadened and a functional knowledge gap within the class III TA family was closed. The active site residue composition in 3N5M is now connected to enzymatic function and may be applied for future specificity predictions.