Data-Driven Protein Engineering for Improving Catalytic Activity and Selectivity.

Protein engineering is essential for altering the substrate scope, catalytic activity and selectivity of enzymes for applications in biocatalysis. However, traditional approaches, such as directed evolution and rational design, encounter the challenge in dealing with the experimental screening process of a large protein mutation space. Machine learning methods allow the approximation of protein fitness landscapes and the identification of catalytic patterns using limited experimental data, thus providing a new avenue to guide protein engineering campaigns. In this concept article, we review machine learning models that have been developed to assess enzyme-substrate-catalysis performance relationships aiming to improve enzymes through data-driven protein engineering. Furthermore, we prospect the future development of this field to provide additional strategies and tools for achieving desired activities and selectivities.

Biosensor-Guided Engineering of a Baeyer-Villiger Monooxygenase for Aliphatic Ester Production.

Esters are valuable aroma compounds and can be produced enzymatically by Baeyer-Villiger monooxygenases (BVMOs) from (aliphatic) ketone precursors. However, a genetically encoded biosensor system for the assessment of BVMO activity and the detection of reaction products is missing. In this work, we assembled a synthetic enzyme cascade - featuring an esterase, an alcohol dehydrogenase, and LuxAB - in the heterologous host Escherichia coli. Target esters are produced by a BVMO, subsequently cleaved, and the corresponding alcohol oxidized through the artificial pathway. Ultimately, aldehyde products are detected in vivo by LuxAB, a luciferase from Photorhabdus luminescens that emits bioluminescence upon the oxidation of aldehydes to the corresponding carboxylates. This biosensor system greatly accelerated the screening and selection of active BVMO variants from a focused library, omitting commonly used low-throughput chromatographic analysis. Engineered enzymes accepted linear aliphatic ketones such as 2-undecanone and 2-dodecanone and exhibited improved ester formation.

Discovery and Characterization of a Baeyer-Villiger Monooxygenase Using Sequence Similarity Network Analysis.

Baeyer-Villiger monooxygenases (BVMOs) are important flavin-dependent enzymes which perform oxygen insertion reactions leading to valuable products. As reported in many studies, BVMOs are usually unstable during application, preventing a wider usage in biocatalysis. Here, we discovered a novel NADPH-dependent BVMO which originates from Halopolyspora algeriensis using sequence similarity networks (SSNs). The enzyme is stable at temperatures between 10 °C to 30 °C up to five days after the purification, and yields the normal ester product. In this study, the substrate scope was investigated for a broad range of aliphatic ketones and the enzyme was biochemically characterized to identify optimum reaction conditions. The best substrate (86 % conversion) was 2-dodecanone using purified enzyme. This novel BVMO could potentially be applied as part of an enzymatic cascade or in bioprocesses which utilize aliphatic alkanes as feedstock.

Structure- and Data-Driven Protein Engineering of Transaminases for Improving Activity and Stereoselectivity.

Amine transaminases (ATAs) are powerful biocatalysts for the stereoselective synthesis of chiral amines. Machine learning provides a promising approach for protein engineering, but activity prediction models for ATAs remain elusive due to the difficulty of obtaining high-quality training data. Thus, we first created variants of the ATA from Ruegeria sp. (3FCR) with improved catalytic activity (up to 2000-fold) as well as reversed stereoselectivity by a structure-dependent rational design and collected a high-quality dataset in this process. Subsequently, we designed a modified one-hot code to describe steric and electronic effects of substrates and residues within ATAs. Finally, we built a gradient boosting regression tree predictor for catalytic activity and stereoselectivity, and applied this for the data-driven design of optimized variants which then showed improved activity (up to 3-fold compared to the best variants previously identified). We also demonstrated that the model can predict the catalytic activity for ATA variants of another origin by retraining with a small set of additional data.

A Universal, Continuous Assay for SAM-dependent Methyltransferases.

Enzyme-catalyzed late-stage functionalization (LSF), such as methylation of drug molecules and lead structures, enables direct access to more potent active pharmaceutical ingredients (API). S-adenosyl-l-methionine-dependent methyltransferases (MTs) can play a key role in the development of new APIs, as they catalyze the chemo- and regioselective methylation of O-, N-, S- and C-atoms, being superior to traditional chemical routes. To identify suitable MTs, we developed a continuous fluorescence-based, high-throughput assay for SAM-dependent methyltransferases, which facilitates screening using E. coli cell lysates. This assay involves two enzymatic steps for the conversion of S-adenosyl-l-homocysteine into H2 S to result in a selective fluorescence readout via reduction of an azidocoumarin sulfide probe. Investigation of two O-MTs and an N-MT confirmed that this assay is suitable for the determination of methyltransferase activity in E. coli cell lysates.

Folding Assessment of Incorporation of Noncanonical Amino Acids Facilitates Expansion of Functional-Group Diversity for Enzyme Engineering.

Protein design is limited by the diversity of functional groups provided by the canonical protein "building blocks". Incorporating noncanonical amino acids (ncAAs) into enzymes enables a dramatic expansion of their catalytic features. For this, quick identification of fully translated and correctly folded variants is decisive. Herein, we report the engineering of the enantioselectivity of an esterase utilizing several ncAAs. Key for the identification of active and soluble protein variants was the use of the split-GFP method, which is crucial as it allows simple determination of the expression levels of enzyme variants with ncAA incorporations by fluorescence. Several identified variants led to improved enantioselectivity or even inverted enantiopreference in the kinetic resolution of ethyl 3-phenylbutyrate.

Alteration of Chain Length Selectivity of Candida antarctica Lipase A by Semi-Rational Design for the Enrichment of Erucic and Gondoic Fatty Acids.

Biotechnological strategies using renewable materials as starting substrates are a promising alternative to traditional oleochemical processes for the isolation of different fatty acids. Among them, long chain mono-unsaturated fatty acids are especially interesting in industrial lipid modification, since they are precursors of several economically relevant products, including detergents, plastics and lubricants. Therefore, the aim of this study was to develop an enzymatic method in order to increase the percentage of long chain mono-unsaturated fatty acids from Camelina and Crambe oil ethyl ester derivatives, by using selective lipases. Specifically, the focus was on the enrichment of gondoic (C20:1 cisΔ11) and erucic acid (C22:1 cisΔ13) from Camelina and Crambe oil derivatives, respectively. The pursuit of this goal entailed several steps, including: (i) the choice of a suitable lipase scaffold to serve as a protein engineering template (Candida antarctica lipase A); (ii) the identification of potential amino acid targets to disrupt the binding tunnel at the adequate location; (iii) the design, creation and high-throughput screening of lipase mutant libraries; (iv) the study of the selectivity towards different chain length p-nitrophenyl fatty acid esters of the best hits found, as well as the analysis of the contribution of each amino acid change and the outcome of combining several of the aforementioned residue alterations and, finally, (v) the selection and application of the most promising candidates for the fatty acid enrichment biocatalysis. As a result, enrichment of C22:1 from Crambe ethyl esters was achieved either, in the free fatty acid fraction (wt, 78%) or in the esterified fraction (variants V1, 77%; V9, 78% and V19, 74%). Concerning the enrichment of C20:1 when Camelina oil ethyl esters were used as substrate, the best variant was the single mutant V290W, which doubled its content in the esterified fraction from approximately 15% to 34%. A moderately lower increase was achieved by V9 and its two derived triple mutant variants V19 and V20 (27%).

Fast, Continuous, and High-Throughput (Bio)Chemical Activity Assay for N-Acyl-l-Homoserine Lactone Quorum-Quenching Enzymes.

UNLABELLED: Quorum sensing, the bacterial cell-cell communication by small molecules, controls important processes such as infection and biofilm formation. Therefore, it is a promising target with several therapeutic and technical applications besides its significant ecological relevance. Enzymes inactivating N-acyl-l-homoserine lactones, the most common class of communication molecules among Gram-negative proteobacteria, mainly belong to the groups of quorum-quenching lactonases or quorum-quenching acylases. However, identification, characterization, and optimization of these valuable biocatalysts are based on a very limited number of fundamentally different methods with their respective strengths and weaknesses. Here, a (bio)chemical activity assay is described, which perfectly complements the other methods in this field. It enables continuous and high-throughput activity measurements of purified and unpurified quorum-quenching enzymes within several minutes. For this, the reaction products released by quorum-quenching lactonases and quorum-quenching acylases are converted either by a secondary enzyme or by autohydrolysis to l-homoserine. In turn, l-homoserine is detected by the previously described calcein assay, which is sensitive to α-amino acids with free N and C termini. Besides its establishment, the method was applied to the characterization of three previously undescribed quorum-quenching lactonases and variants thereof and to the identification of quorum-quenching acylase-expressing Escherichia coli clones in an artificial library. Furthermore, this study indicates that porcine aminoacylase 1 is not active toward N-acyl-l-homoserine lactones as published previously but instead converts the autohydrolysis product N-acyl-l-homoserine. IMPORTANCE: In this study, a novel method is presented for the identification, characterization, and optimization of quorum-quenching enzymes that are active toward N-acyl-l-homoserine lactones. These are the most common communication molecules among Gram-negative proteobacteria. The activity assay is a highly valuable complement to the available analytical tools in this field. It will facilitate studies on the environmental impact of quorum-quenching enzymes and contribute to the development of therapeutic and technical applications of this promising enzyme class.

Fully automatized high-throughput enzyme library screening using a robotic platform.

A fully automatized robotic platform has been established to facilitate high-throughput screening for protein engineering purposes. This platform enables proper monitoring and control of growth conditions in the microtiter plate format to ensure precise enzyme production for the interrogation of enzyme mutant libraries, protein stability tests and multiple assay screenings. The performance of this system has been exemplified for four enzyme classes important for biocatalysis such as Baeyer-Villiger monooxygenase, transaminase, dehalogenase and acylase in the high-throughput screening of various mutant libraries. This allowed the identification of novel enzyme variants in a sophisticated and highly reliable manner. Furthermore, the detailed optimization protocols should enable other researchers to adapt and improve their methods. Biotechnol. Bioeng. 2016;113: 1421-1432. © 2016 Wiley Periodicals, Inc.

Alteration of the Donor/Acceptor Spectrum of the (S)-Amine Transaminase from Vibrio fluvialis.

To alter the amine donor/acceptor spectrum of an (S)-selective amine transaminase (ATA), a library based on the Vibrio fluvialis ATA targeting four residues close to the active site (L56, W57, R415 and L417) was created. A 3DM-derived alignment comprising fold class I pyridoxal-5'-phosphate (PLP)-dependent enzymes allowed identification of positions, which were assumed to determine substrate specificity. These positions were targeted for mutagenesis with a focused alphabet of hydrophobic amino acids to convert an amine:α-keto acid transferase into an amine:aldehyde transferase. Screening of 1200 variants revealed three hits, which showed a shifted amine donor/acceptor spectrum towards aliphatic aldehydes (mainly pentanal), as well as an altered pH profile. Interestingly, all three hits, although found independently, contained the same mutation R415L and additional W57F and L417V substitutions.

Product Distribution of Steady-State and Pulsed Electrochemical Regeneration of 1,4-NADH and Integration with Enzymatic Reaction.

The direct electrochemical reduction of nicotinamide adenine dinucleotide (NAD+) results in various products, complicating the regeneration of the crucial 1,4-NADH cofactor for enzymatic reactions. Previous research primarily focused on steady-state polarization to examine potential impacts on product selectivity. However, this study explores the influence of dynamic conditions on the selectivity of NAD+ reduction products by comparing two dynamic profiles with steady-state conditions. Our findings reveal that the main products, including 1,4-NADH, several dimers, and ADP-ribose, remained consistent across all conditions. A minor by-product, 1,6-NADH, was also identified. The product distribution varied depending on the experimental conditions (steady state vs. dynamic) and the concentration of NAD+, with higher concentrations and overpotentials promoting dimerization. The optimal yield of 1,4-NADH was achieved under steady-state conditions with low overpotential and NAD+ concentrations. While dynamic conditions enhanced the 1,4-NADH yield at shorter reaction times, they also resulted in a significant amount of unidentified products. Furthermore, this study assessed the potential of using pulsed electrochemical regeneration of 1,4-NADH with enoate reductase (XenB) for cyclohexenone reduction.

State-of-the-Art Light-Driven Hydrogen Generation from Formic Acid and Utilization in Enzymatic Hydrogenations.

A concept of combining photocatalytically generated hydrogen with green enzymatic reductions is demonstrated. The developed photocatalytic formic acid (FA) dehydrogenation setup based on Pt(x)@TiO2 shows unprecedented hydrogen generation activity, which is two orders of magnitude higher than reported values of state-of-the-art systems. Mechanistic studies confirm that hydrogen generation proceeds via a photocatalytic pathway, which is entirely different from purely thermal reaction mechanisms previously reported. The viability of the presented approach is demonstrated by the synthesis of value-added compounds 3-phenylpropanal and (2R, 5S)-dihydrocarvone at ambient pressure and room temperature, which should be applicable for many other hydrogenation processes, e.g., for the preparation of flavours and fragrance compounds, as well as pharmaceuticals.

Formation of RNA phosphodiester bond by histidine-containing dipeptides.

A new scenario for prebiotic formation of nucleic acid oligomers is presented. Peptide catalysis is applied to achieve condensation of activated RNA monomers into short RNA chains. As catalysts, L-dipeptides containing a histidine residue, primarily Ser-His, were used. Reactions were carried out in self-organised environment, a water-ice eutectic phase, with low concentrations of reactants. Incubation periods up to 30 days resulted in the formation of short oligomers of RNA. During the oligomerisation, an active intermediate (dipeptide-mononucleotide) is produced, which is the reactive species. Details of the mechanism and kinetics, which were elucidated with a set of control experiments, further establish that the imidazole side chain of a histidine at the carboxyl end of the dipeptide plays a crucial role in the catalysis. These results suggest that this oligomerisation catalysis occurs by a transamination mechanism. Because peptides are much more likely products of spontaneous condensation than nucleotide chains, their potential as catalysts for the formation of RNA is interesting from the origin-of-life perspective. Finally, the formation of the dipeptide-mononucleotide intermediate and its significance for catalysis might also be viewed as the tell-tale signs of a new example of organocatalysis.

Interactions between catalysts and amphiphilic structures and their implications for a protocell model.

One of the essential elements of any cell, including primitive ancestors, is a structural component that protects and confines the metabolism and genes while allowing access to essential nutrients. For the targeted protocell model, bilayers of decanoic acid, a single-chain fatty acid amphiphile, are used as the container. These bilayers interact with a ruthenium-nucleobase complex, the metabolic complex, to convert amphiphile precursors into more amphiphiles. These interactions are dependent on non-covalent bonding. The initial rate of conversion of an oily precursor molecule into fatty acid was examined as a function of these interactions. It is shown that the precursor molecule associates strongly with decanoic acid structures. This results in a high dependence of conversion rates on the interaction of the catalyst with the self-assembled structures. The observed rate logically increases when a tight interaction between catalyst complex and container exists. A strong association between the metabolic complex and the container was achieved by bonding a sufficiently long hydrocarbon tail to the complex. Surprisingly, the rate enhancement was nearly as strong when the ruthenium and nucleobase elements of the complex were each given their own hydrocarbon tail and existed as separate molecules, as when the two elements were covalently bonded to each other and the resulting molecule was given a hydrocarbon tail. These results provide insights into the possibilities and constraints of such a reaction system in relation to building the ultimate protocell.

Engineering the protein dynamics of an ancestral luciferase.

Protein dynamics are often invoked in explanations of enzyme catalysis, but their design has proven elusive. Here we track the role of dynamics in evolution, starting from the evolvable and thermostable ancestral protein AncHLD-RLuc which catalyses both dehalogenase and luciferase reactions. Insertion-deletion (InDel) backbone mutagenesis of AncHLD-RLuc challenged the scaffold dynamics. Screening for both activities reveals InDel mutations localized in three distinct regions that lead to altered protein dynamics (based on crystallographic B-factors, hydrogen exchange, and molecular dynamics simulations). An anisotropic network model highlights the importance of the conformational flexibility of a loop-helix fragment of Renilla luciferases for ligand binding. Transplantation of this dynamic fragment leads to lower product inhibition and highly stable glow-type bioluminescence. The success of our approach suggests that a strategy comprising (i) constructing a stable and evolvable template, (ii) mapping functional regions by backbone mutagenesis, and (iii) transplantation of dynamic features, can lead to functionally innovative proteins.

An Ultrasensitive Fluorescence Assay for the Detection of Halides and Enzymatic Dehalogenation.

Halide assays are important for the study of enzymatic dehalogenation, a topic of great industrial and scientific importance. Here we describe the development of a very sensitive halide assay that can detect less than a picomole of bromide ions, making it very useful for quantifying enzymatic dehalogenation products. Halides are oxidised under mild conditions using the vanadium-dependent chloroperoxidase from Curvularia inaequalis, forming hypohalous acids that are detected using aminophenyl fluorescein. The assay is up to three orders of magnitude more sensitive than currently available alternatives, with detection limits of 20 nM for bromide and 1 µM for chloride and iodide. We demonstrate that the assay can be used to determine specific activities of dehalogenases and validate this by comparison to a well-established GC-MS method. This new assay will facilitate the identification and characterisation of novel dehalogenases and may also be of interest to those studying other halide-producing enzymes.

Library Growth and Protein Expression: Optimal and Reproducible Microtiter Plate Expression of Recombinant Enzymes in E. coli Using MTP Shakers.

Escherichia coli (E. coli) as heterologous host enables the recombinant expression of the desired protein in high amounts. Nevertheless, the expression in such a host, especially by utilizing a strong induction system, can result in insoluble and/or inactive protein fractions (inclusion bodies). Furthermore, the expression of different enzyme variants often leads to a diverse growth behavior of the E. coli clones resulting in the identification of false-positives when screening a mutant library. Thus, we developed a protocol for an optimal and reproducible protein expression in microtiter plates showcased for the expression of the cyclohexanone monooxygenase (CHMO) from Acinetobacter sp. NCIMB 9871. By emerging this protocol, several parameters concerning the expression medium, the cultivation temperatures, shaking conditions as well as time and induction periods for CHMO were investigated. We employed a microtiter plate shaker with humidity and temperature control (Cytomat™) (integrated in a robotic platform) to obtain an even growth and expression over the plates. Our optimized protocol provides a comprehensive overview of the key factors influencing a reproducible protein expression and this should serve as basis for the adaptation to other enzyme classes.

Program-Guided Design of High-Throughput Enzyme Screening Experiments and Automated Data Analysis/Evaluation.

The open source LARA software suite is designed for guiding manual or automated high-throughput screening experiments. Process planning, data reading, analysis, and visualization are herein explained in a step-by-step guide using exemplary dataset.

Normalized Screening of Protein Engineering Libraries by Split-GFP Crude Cell Extract Quantification.

The different expression level and solubility showed by each protein variant represents an important challenge during screening campaigns: Usually, the total activity measurement constitutes the only criterion for identifying improved variants. This hampers the chances of finding interesting mutants, especially if the aim is to improve activity: On the one hand, interesting but poorly soluble variants will remain undetectable. On the other hand, a mutation might not increase activity, but improve expression level or solubility. The split-GFP technology offers an affordable and technically simple manner for overcoming that constraints, making protein library screening more efficient through the normalization of the detected enzymatic activities in relation to the quantified protein contents responsible for them.

In-Depth High-Throughput Screening of Protein Engineering Libraries by Split-GFP Direct Crude Cell Extract Data Normalization.

Here, we report a widely and generally applicable strategy to obtain reliable information in high-throughput protein screenings of enzyme mutant libraries. The method is based on the usage of the split-GFP technology for the normalization of the expression level of each individual protein variant combined with activity measurements, thus resolving the important problems associated with the different solubility of each mutant and allowing the detection of previously invisible variants. The small size of the employed protein tag (16 amino acids) required for the reconstitution of the GFP fluorescence reduces possible interferences such as enzyme activity variations or solubility disturbances to a minimum. Specific enzyme activity measurements without purification, in situ soluble protein expression monitoring, and data normalization are the powerful outputs of this methodology, thus enabling the accurate identification of improved protein variants during high-throughput screening by substantially reducing the occurrence of false negatives and false positives.