The field of biocatalysis has advanced from harnessing natural enzymes to using directed evolution to obtain new biocatalysts with tailor-made functions. Several tools have recently been developed to expand the natural enzymatic repertoire with abiotic reactions. For example, artificial metalloenzymes, which combine the versatile reaction scope of transition metals with the beneficial catalytic features of enzymes, offer an attractive means to engineer new reactions. Three complementary strategies exist: repurposing natural metalloenzymes for abiotic transformations; in silico metalloenzyme (re-)design; and incorporation of abiotic cofactors into proteins. The third strategy offers the opportunity to design a wide variety of artificial metalloenzymes for non-natural reactions. However, many metal cofactors are inhibited by cellular components and therefore require purification of the scaffold protein. This limits the throughput of genetic optimization schemes applied to artificial metalloenzymes and their applicability in vivo to expand natural metabolism. Here we report the compartmentalization and in vivo evolution of an artificial metalloenzyme for olefin metathesis, which represents an archetypal organometallic reaction without equivalent in nature. Building on previous work on an artificial metallohydrolase, we exploit the periplasm of Escherichia coli as a reaction compartment for the 'metathase' because it offers an auspicious environment for artificial metalloenzymes, mainly owing to low concentrations of inhibitors such as glutathione, which has recently been identified as a major inhibitor. This strategy facilitated the assembly of a functional metathase in vivo and its directed evolution with substantially increased throughput compared to conventional approaches that rely on purified protein variants. The evolved metathase compares favourably with commercial catalysts, shows activity for different metathesis substrates and can be further evolved in different directions by adjusting the workflow. Our results represent the systematic implementation and evolution of an artificial metalloenzyme that catalyses an abiotic reaction in vivo, with potential applications in, for example, non-natural metabolism.

Rational flux design in metabolic engineering approaches remains difficult since important pathway information is frequently not available. Therefore empirical methods are applied that randomly change absolute and relative pathway enzyme levels and subsequently screen for variants with improved performance. However, screening is often limited on the analytical side, generating a strong incentive to construct small but smart libraries. Here we introduce RedLibs (Reduced Libraries), an algorithm that allows for the rational design of smart combinatorial libraries for pathway optimization thereby minimizing the use of experimental resources. We demonstrate the utility of RedLibs for the design of ribosome-binding site libraries by in silico and in vivo screening with fluorescent proteins and perform a simple two-step optimization of the product selectivity in the branched multistep pathway for violacein biosynthesis, indicating a general applicability for the algorithm and the proposed heuristics. We expect that RedLibs will substantially simplify the refactoring of synthetic metabolic pathways.

Artificial metalloenzymes (ArMs) catalyzing new-to-nature reactions could play an important role in transitioning toward a sustainable economy. While ArMs have been created for various transformations, attempts at their genetic optimization have been case specific and resulted mostly in modest improvements. To realize their full potential, methods to rapidly discover active ArM variants for ideally any reaction of interest are required. Here, we introduce a reaction-independent, automation-compatible platform, which relies on periplasmic compartmentalization in Escherichia coli to rapidly and reliably engineer ArMs based on the biotin-streptavidin technology. We systematically assess 400 ArM mutants for five bioorthogonal transformations involving different metals, reaction mechanisms, and reactants, which include novel ArMs for gold-catalyzed hydroamination and hydroarylation. Activity enhancements up to 15-fold highlight the potential of the systematic approach. Furthermore, we suggest smart screening strategies and build machine learning models that accurately predict ArM activity from sequence, which has crucial implications for future ArM development.

An artificial deallylase is constituted on the E. coli surface and genetically optimized for the deprotection of caged aminocoumarin.

Elimination of metabolic flux imbalances in microbial cell factories is an important part in the establishment of viable biotechnological production processes. However, due to the high complexity of cellular metabolism, the limited a priori knowledge about the majority of production pathways and a lack of forward design standards, metabolic engineers strongly rely on empirical screening methodologies to achieve the required improvement of cell behavior. Combinatorial pathway engineering provides an interesting tool to identify global solutions for intricate pathways, but methods for the reduction of combinatorial library size are inevitably required to restrict the experimental effort to an affordable size. Here we review recent advances from this field by scrutinizing commonly applied diversification methods and highlighting crucial strategies for the minimization of experimental effort.

Residing at the interface of chemistry and biotechnology, artificial metalloenzymes (ArMs) offer an attractive technology to combine the versatile reaction repertoire of transition metal catalysts with the exquisite catalytic features of enzymes. While earlier efforts in this field predominantly comprised studies in well-defined test-tube environments, a trend towards exploiting ArMs in more complex environments has recently emerged. Integration of these artificial biocatalysts in enzymatic cascades and using them in whole-cell biotransformations and in vivo opens up entirely novel prospects for both preparative chemistry and synthetic biology. We highlight selected recent developments with a particular focus on challenges and opportunities in the in vivo application of ArMs.

Engineered enzyme cascades offer powerful tools to convert renewable resources into valueadded products. Man-made catalysts give access to new-to-nature reactivities that may complement the enzyme's repertoire. Their mutual incompatibility, however, challenges their integration into concurrent chemo-enzymatic cascades. Herein we show that compartmentalization of complex enzyme cascades within E. coli whole cells enables the simultaneous use of a metathesis catalyst, thus allowing the sustainable one-pot production of cycloalkenes from oleic acid. Cycloheptene is produced from oleic acid via a concurrent enzymatic oxidative decarboxylation and ring-closing metathesis. Cyclohexene and cyclopentene are produced from oleic acid via either a six-or eight-step enzyme cascade involving hydration, oxidation, hydrolysis and decarboxylation, followed by ring-closing metathesis. Integration of an upstream hydrolase enables the usage of olive oil as the substrate for the production of cycloalkenes. This work highlights the potential of integrating organometallic catalysis with whole-cell enzyme cascades of high complexity to enable sustainable chemistry.

Predicting effects of gene regulatory elements (GREs) is a longstanding challenge in biology. Machine learning may address this, but requires large datasets linking GREs to their quantitative function. However, experimental methods to generate such datasets are either application-specific or technically complex and error-prone. Here, we introduce DNA-based phenotypic recording as a widely applicable, practicable approach to generate large-scale sequence-function datasets. We use a site-specific recombinase to directly record a GRE's effect in DNA, enabling readout of both sequence and quantitative function for extremely large GRE-sets via next-generation sequencing. We record translation kinetics of over 300,000 bacterial ribosome binding sites (RBSs) in >2.7 million sequence-function pairs in a single experiment. Further, we introduce a deep learning approach employing ensembling and uncertainty modelling that predicts RBS function with high accuracy, outperforming state-ofthe-art methods. DNA-based phenotypic recording combined with deep learning represents a major advance in our ability to predict function from genetic sequence.