Directed evolution to re-adapt a co-evolved network within an enzyme

Two-substrate enzyme engineering using small libraries that combine the substrate preferences from two different variant lineages

Improving the range of substrates accepted by enzymes with high catalytic activity remains an important goal for the industrialisation of biocatalysis. Many enzymes catalyse two-substrate reactions which increases the complexity in engineering them for the synthesis of alternative products. Often mutations are found independently that can improve the acceptance of alternatives to each of the two substrates. Ideally, we would be able to combine mutations identified for each of the two alternative substrates, and so reprogramme new enzyme variants that synthesise specific products from their respective two-substrate combinations. However, as we have previously observed for E. coli transketolase, the mutations that improved activity towards aromatic acceptor aldehydes, did not successfully recombine with mutations that switched the donor substrate to pyruvate. This likely results from several active site residues having multiple roles that can affect both of the substrates, as well as structural interactions between the mutations themselves. Here, we have designed small libraries, including both natural and non-natural amino acids, based on the previous mutational sites that impact on acceptance of the two substrates, to achieve up to 630× increases in kcat for the reaction with 3-formylbenzoic acid (3-FBA) and pyruvate. Computational docking was able to determine how the mutations shaped the active site to improve the proximity of the 3-FBA substrate relative to the enamine-TPP intermediate, formed after the initial reaction with pyruvate. This work opens the way for small libraries to rapidly reprogramme enzyme active sites in a plug and play approach to catalyse new combinations of two-substrate reactions.

Directed evolution for soluble and active periplasmic expression of bovine enterokinase in Escherichia coli

Bovine enterokinase light chain (EK_L) is an industrially useful protease for accurate removal of affinity-purification tags from high-value biopharmaceuticals. However, recombinant expression in Escherichia coli produces insoluble inclusion bodies, requiring solubilisation, refolding, and autocatalytic activation to recover functional enzyme. Error-prone PCR and DNA shuffling of the EK_L gene, T7 promoter, lac operon, ribosome binding site, and pelB leader sequence, yielded 321 unique variants after screening ~ 6500 colonies. The best variants had > 11,000-fold increased total activity in lysates, producing soluble enzyme that no longer needed refolding. Further characterisation identified the factors that improved total activity from an inactive and insoluble starting point. Stability was a major factor, whereby melting temperatures > 48.4 °C enabled good expression at 37 °C. Variants generally did not alter catalytic efficiency as measured by k_cat/K_m, which improved for only one variant. Codon optimisation improved the total activity in lysates produced at 37 °C. However, non-optimised codons and expression at 30 °C gave the highest activity through improved protein quality, with increased k_cat and T_m values. The 321 variants were statistically analysed and mapped to protein structure. Mutations detrimental to total activity and stability clustered around the active site. By contrast, variants with increased total activity tended to combine stabilising mutations that did not disrupt the active site.

Therapeutic enzyme engineering using a generative neural network

Enhancing the potency of mRNA therapeutics is an important objective for treating rare diseases, since it may enable lower and less-frequent dosing. Enzyme engineering can increase potency of mRNA therapeutics by improving the expression, half-life, and catalytic efficiency of the mRNA-encoded enzymes. However, sequence space is incomprehensibly vast, and methods to map sequence to function (computationally or experimentally) are inaccurate or time-/labor-intensive. Here, we present a novel, broadly applicable engineering method that combines deep latent variable modelling of sequence co-evolution with automated protein library design and construction to rapidly identify metabolic enzyme variants that are both more thermally stable and more catalytically active. We apply this approach to improve the potency of ornithine transcarbamylase (OTC), a urea cycle enzyme for which loss of catalytic activity causes a rare but serious metabolic disease.

Biophysical characterization of the inactivation of E. coli transketolase by aqueous co-solvents

Transketolase (TK) has been previously engineered, using semi-rational directed evolution and substrate walking, to accept increasingly aliphatic, cyclic, and then aromatic substrates. This has ultimately led to the poor water solubility of new substrates, as a potential bottleneck to further exploitation of this enzyme in biocatalysis. Here we used a range of biophysical studies to characterise the response of both E. coli apo- and holo-TK activity and structure to a range of polar organic co-solvents: acetonitrile (AcCN), n-butanol (nBuOH), ethyl acetate (EtOAc), isopropanol (iPrOH), and tetrahydrofuran (THF). The mechanism of enzyme deactivation was found to be predominantly via solvent-induced local unfolding. Holo-TK is thermodynamically more stable than apo-TK and yet for four of the five co-solvents it retained less activity than apo-TK after exposure to organic solvents, indicating that solvent tolerance was not simply correlated to global conformational stability. The co-solvent concentrations required for complete enzyme inactivation was inversely proportional to co-solvent log(P), while the unfolding rate was directly proportional, indicating that the solvents interact with and partially unfold the enzyme through hydrophobic contacts. Small amounts of aggregate formed in some cases, but this was not sufficient to explain the enzyme inactivation. TK was found to be tolerant to 15% (v/v) iPrOH, 10% (v/v) AcCN, or 6% (v/v) nBuOH over 3 h. This work indicates that future attempts to engineer the enzyme to better tolerate co-solvents should focus on increasing the stability of the protein to local unfolding, particularly in and around the cofactor-binding loops.

Fine-tuning the activity and stability of an evolved enzyme active-site through noncanonical amino-acids

Site-specific saturation mutagenesis within enzyme active sites can radically alter reaction specificity, though often with a trade-off in stability. Extending saturation mutagenesis with a range of noncanonical amino acids (ncAA) potentially increases the ability to improve activity and stability simultaneously. Previously, an Escherichia coli transketolase variant (S385Y/D469T/R520Q) was evolved to accept aromatic aldehydes not converted by wild-type. The aromatic residue Y385 was critical to the new acceptor substrate binding, and so was explored here beyond the natural aromatic residues, to probe side chain structure and electronics effects on enzyme function and stability. A series of five variants introduced decreasing aromatic ring electron density at position 385 in the order para-aminophenylalanine (pAMF), tyrosine (Y), phenylalanine (F), para-cyanophenylalanine (pCNF) and para-nitrophenylalanine (pNTF), and simultaneously modified the hydrogen-bonding potential of the aromatic substituent from accepting to donating. The fine-tuning of residue 385 yielded variants with a 43-fold increase in specific activity for 50 mm 3-HBA and 100% increased k_cat (pCNF), 290% improvement in K_m (pNTF), 240% improvement in k_cat /K_m (pAMF) and decreased substrate inhibition relative to Y. Structural modelling suggested switching of the ring-substituted functional group, from donating to accepting, stabilised a helix-turn (D259-H261) through an intersubunit H-bond with G262, to give a 7.8 °C increase in the thermal transition mid-point, T_m , and improved packing of pAMF. This is one of the first examples in which both catalytic activity and stability are simultaneously improved via site-specific ncAA incorporation into an enzyme active site, and further demonstrates the benefits of expanding designer libraries to include ncAAs.

Exploiting correlated molecular-dynamics networks to counteract enzyme activity-stability trade-off

Rigidifying flexible sites is a powerful method to improve enzyme stability. However, if the highly flexible regions form the active site, modifying them risks losing activity due to the activity–stability trade-off. We hypothesized here that regions outside the active site whose dynamics were highly correlated to flexible active sites, would provide good targets for stabilizing mutations. To test this hypothesis, six variants were constructed in the 3M variant of Escherichia coli transketolase. The best variant had a 10.8-fold improved half-life at 55 °C, and increased the T_m and T_agg by 3 °C and 4.3 °C, respectively. The variants even increased the activity, by up to threefold. This study highlights how protein engineering strategies could be potentially improved by considering long-range dynamics.

The directed evolution of enzymes for improved activity or substrate specificity commonly leads to a trade-off in stability. We have identified an activity–stability trade-off and a loss in unfolding cooperativity for a variant (3M) of Escherichia coli transketolase (TK) engineered to accept aromatic substrates. Molecular dynamics simulations of 3M revealed increased flexibility in several interconnected active-site regions that also form part of the dimer interface. Mutating the newly flexible active-site residues to regain stability risked losing the new activity. We hypothesized that stabilizing mutations could be targeted to residues outside of the active site, whose dynamics were correlated with the newly flexible active-site residues. We previously stabilized WT TK by targeting mutations to highly flexible regions. These regions were much less flexible in 3M and would not have been selected a priori as targets using the same strategy based on flexibility alone. However, their dynamics were highly correlated with the newly flexible active-site regions of 3M. Introducing the previous mutations into 3M reestablished the WT level of stability and unfolding cooperativity, giving a 10.8-fold improved half-life at 55 °C, and increased midpoint and aggregation onset temperatures by 3 °C and 4.3 °C, respectively. Even the activity toward aromatic aldehydes increased up to threefold. Molecular dynamics simulations confirmed that the mutations rigidified the active-site via the correlated network. This work provides insights into the impact of rigidifying mutations within highly correlated dynamic networks that could also be useful for developing improved computational protein engineering strategies.

Enzyme engineering: reaching the maximal catalytic efficiency peak

The practical need for highly efficient enzymes presents new challenges in enzyme engineering, in particular, the need to improve catalytic turnover (k_cat) or efficiency (k_cat/K_M) by several orders of magnitude. However, optimizing catalysis demands navigation through complex and rugged fitness landscapes, with optimization trajectories often leading to strong diminishing returns and dead-ends. When no further improvements are observed in library screens or selections, it remains unclear whether the maximal catalytic efficiency of the enzyme (the catalytic 'fitness peak') has been reached; or perhaps, an alternative combination of mutations exists that could yield additional improvements. Here, we discuss fundamental aspects of the process of catalytic optimization, and offer practical solutions with respect to overcoming optimization plateaus.

Structural Analysis of an Evolved Transketolase Reveals Divergent Binding Modes

The S385Y/D469T/R520Q variant of E. coli transketolase was evolved previously with three successive smart libraries, each guided by different structural, bioinformatical or computational methods. Substrate-walking progressively shifted the target acceptor substrate from phosphorylated aldehydes, towards a non-phosphorylated polar aldehyde, a non-polar aliphatic aldehyde, and finally a non-polar aromatic aldehyde. Kinetic evaluations on three benzaldehyde derivatives, suggested that their active-site binding was differentially sensitive to the S385Y mutation. Docking into mutants generated in silico from the wild-type crystal structure was not wholly satisfactory, as errors accumulated with successive mutations, and hampered further smart-library designs. Here we report the crystal structure of the S385Y/D469T/R520Q variant, and molecular docking of three substrates. This now supports our original hypothesis that directed-evolution had generated an evolutionary intermediate with divergent binding modes for the three aromatic aldehydes tested. The new active site contained two binding pockets supporting π-π stacking interactions, sterically separated by the D469T mutation. While 3-formylbenzoic acid (3-FBA) preferred one pocket, and 4-FBA the other, the less well-accepted substrate 3-hydroxybenzaldehyde (3-HBA) was caught in limbo with equal preference for the two pockets. This work highlights the value of obtaining crystal structures of evolved enzyme variants, for continued and reliable use of smart library strategies.

Correlated positions in protein evolution and engineering

Statistical analysis of a protein multiple sequence alignment can reveal groups of positions that undergo interdependent mutations throughout evolution. At these so-called correlated positions, only certain combinations of amino acids appear to be viable for maintaining proper folding, stability, catalytic activity or specificity. Therefore, it is often speculated that they could be interesting guides for semi-rational protein engineering purposes. Because they are a fingerprint from protein evolution, their analysis may provide valuable insight into a protein's structure or function and furthermore, they may also be suitable target positions for mutagenesis. Unfortunately, little is currently known about the properties of these correlation networks and how they should be used in practice. This review summarises the recent findings, opportunities and pitfalls of the concept.

(Dis)similarity Index To Compare Correlated Motions in Molecular Simulations

Molecular dynamics (MD) simulations are widely used to complement or guide experimental studies in the characterization of protein dynamics, thanks to improvements in force-field accuracy, along with in the software and hardware to sample the conformational landscape of proteins. Among the different applications of MD simulations, the study of correlated motions is largely employed for different purposes. Several metrics have been developed to describe correlated motions in the MD ensemble, such as methods based on Pearson Correlation or Mutual Information. Cross-correlation analysis of MD trajectories is indeed appealing not only to identify residues characterized by coupled fluctuations in protein structures but also since it can be used to extrapolate motions along directions in which major conformational changes should occur, for example on longer time scales than the ones that are actually simulated. Nevertheless, most of the MD studies employ average correlation maps and mostly in a qualitative way, even when different systems or different replicates of the same system are compared. The broad application of correlation metrics in the analysis of MD simulations, especially for comparative purposes, requires a step forward toward more quantitative and accurate comparisons. We thus here employed a simple but effective index, which is based on a normalized Frobenius norm of the differences between protein correlation maps, to compare correlated motions. We applied this index for a quantitative comparison of correlated motions from MD simulations of seven proteins of different size and fold. We also employed the index to assess the robustness of correlation description when multi-replicate MD simulations of a same system are used, and we compared our index to metrics for comparison of structural ensembles such as Root Mean Square Inner Product and the Bhattacharyya Coefficient.

Synthetic biology for the directed evolution of protein biocatalysts: navigating sequence space intelligently

The amino acid sequence of a protein affects both its structure and its function. Thus, the ability to modify the sequence, and hence the structure and activity, of individual proteins in a systematic way, opens up many opportunities, both scientifically and (as we focus on here) for exploitation in biocatalysis. Modern methods of synthetic biology, whereby increasingly large sequences of DNA can be synthesised de novo, allow an unprecedented ability to engineer proteins with novel functions. However, the number of possible proteins is far too large to test individually, so we need means for navigating the 'search space' of possible protein sequences efficiently and reliably in order to find desirable activities and other properties. Enzymologists distinguish binding (Kd) and catalytic (kcat) steps. In a similar way, judicious strategies have blended design (for binding, specificity and active site modelling) with the more empirical methods of classical directed evolution (DE) for improving kcat (where natural evolution rarely seeks the highest values), especially with regard to residues distant from the active site and where the functional linkages underpinning enzyme dynamics are both unknown and hard to predict. Epistasis (where the 'best' amino acid at one site depends on that or those at others) is a notable feature of directed evolution. The aim of this review is to highlight some of the approaches that are being developed to allow us to use directed evolution to improve enzyme properties, often dramatically. We note that directed evolution differs in a number of ways from natural evolution, including in particular the available mechanisms and the likely selection pressures. Thus, we stress the opportunities afforded by techniques that enable one to map sequence to (structure and) activity in silico, as an effective means of modelling and exploring protein landscapes. Because known landscapes may be assessed and reasoned about as a whole, simultaneously, this offers opportunities for protein improvement not readily available to natural evolution on rapid timescales. Intelligent landscape navigation, informed by sequence-activity relationships and coupled to the emerging methods of synthetic biology, offers scope for the development of novel biocatalysts that are both highly active and robust.

Patch cloning method for multiple site-directed and saturation mutagenesis

BACKGROUND

Various DNA manipulation methods have been developed to prepare mutant genes for protein engineering. However, development of more efficient and convenient method is still demanded. Homologous DNA assembly methods, which do not depend on restriction enzymes, have been used as convenient tools for cloning and have been applied to site-directed mutagenesis recently. This study describes an optimized homologous DNA assembly method, termed as multiple patch cloning (MUPAC), for multiple site-directed and saturation mutagenesis.

RESULTS

To demonstrate MUPAC, we introduced five back mutations to a mutant green fluorescent protein (GFPuv) with five deleterious mutations at specific sites and transformed Escherichia coli (E. coli) with the plasmids obtained. We observed that the over 90% of resulting colonies possessed the plasmids containing the reverted GFPuv gene and exhibited fluorescence. We extended the test to introduce up to nine mutations in Moloney Murine Leukemia Virus reverse transcriptase (M-MLV RT) by assembling 11 DNA fragments using MUPAC. Analysis of the cloned plasmid by electrophoresis and DNA sequencing revealed that approximately 30% of colonies had the objective mutant M-MLV RT gene. Furthermore, we also utilized this method to prepare a library of mutant GFPuv genes containing saturation mutations at five specific sites, and we found that MUPAC successfully introduced NNK codons at all five sites, whereas other site remained intact.

CONCLUSIONS

MUPAC could efficiently introduce various mutations at multiple specific sites within a gene. Furthermore, it could facilitate the preparation of experimental gene materials important to molecular and synthetic biology research.

CC bond formation using ThDP-dependent lyases

The present review summarizes recent achievements in enzymatic thiamine catalysis during the past three years. With well-established enzymes such as BAL, PDC and TK new reactions have been identified and respective variants were prepared, which enable access to stereoisomeric products. Further we highlight recent progress with 'new' ThDP-dependent enzymes like MenD and PigD, which catalyze the Stetter-like 1,4 addition of aldehydes and YerE, which is the first known ThDP-dependent enzyme accepting ketones as acceptors.

Using analyses of amino Acid coevolution to understand protein structure and function

Determining which residues of a protein contribute to a specific function is a difficult problem. Analyses of amino acid covariation within a protein family can serve as a useful guide by identifying residues that are functionally coupled. Covariation analyses have been successfully used on several different protein families to identify residues that work together to promote folding, enable protein-protein interactions, or contribute to an enzymatic activity. Covariation is a statistical signal that can be measured in a multiple sequence alignment of homologous proteins. As sequence databases have expanded dramatically, covariation analyses have become easier and more powerful. In this chapter, we describe how functional covariation arises during the evolution of proteins and how this signal can be distinguished from various background signals. We discuss the basic methodology for performing amino acid covariation analysis, using bacterial two-component signal transduction proteins as an example. We provide practical suggestions for each step of the process including assembly of protein sequences, construction of a multiple sequence alignment, measurement of covariation, and analysis of results.