Biocatalysis@TUDelft

Machine learning holds great promise for accelerating enzyme optimization, but its power is fundamentally constrained by the limited availability of sequence-fitness data. Here, we introduce MillionFull, a low-cost method that enables high-throughput full-length sequence- fitness mapping for enzymes of arbitrary length. Each run yields on the order of 10-10 data points, capturing sequence-function relationships at unprecedented scale. By overcoming the data bottleneck, MillionFull provides a foundation for dramatically advancing AI-driven enzyme engineering.

Vanadium-dependent haloperoxidases (VHPOs) catalyze the halogenation of organic molecules under mild aqueous conditions. Selective bacterial VHPOs exhibit exquisite regio- and enantiocontrol, however the precise mechanisms dictating selectivity have remained elusive. We have solved the single-particle cryo-electron microscopy (cryo-EM) structure of a selective bromoperoxidase from Enhygromyxa salina (esVHPO). Mutagenesis demonstrates that halide oxidation and substrate halogenation occur in two distinct pockets, with halide transfer mediated by critical lysine residue K329. Isolation of a stable intermediate following bromide oxidation (BrOx) enables single turnover catalysis in the presence of organic substrate; subsequent application of a chemoselective fluorescent probe provides support for an intermediate bromamine involved in selectivity. Cryo-EM of the BrOx state reveals a camera shutter mechanism that compacts the halide entry tunnel and vanadate pocket, minimizing the premature dissociation of hypohalous acid. These findings collectively unveil a multilayered halogen trapping and transfer mechanism and provide a rationale for selective VHPO catalysis. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=114 SRC="FIGDIR/small/684477v1_ufig1.gif" ALT="Figure 1"> View larger version (25K): org.highwire.dtl.DTLVardef@17c427forg.highwire.dtl.DTLVardef@f1a7caorg.highwire.dtl.DTLVardef@13ed2bforg.highwire.dtl.DTLVardef@17e1c16_HPS_FORMAT_FIGEXP M_FIG C_FIG

Enzyme mediated bioconjugation provides a method for easy and rapid formation of protein-protein and protein-small molecule conjugates under mild conditions. Promiscuous enzymes are of particular interest because they can catalyze conjugation reactions on a broad set of substrates. However, this promiscuity carries the risk of undesirable off-target modifications. To mitigate this effect, we used computational design to install a substrate recruitment domain (SRD) onto the promiscuous enzyme, tyrosinase. The redesigned tyrosinase, called D42, preferentially modifies tyrosine residues within the peptide core (core) linked to a 6-amino acid recognition motif/sequence (RS) specific for the SRD. Incorporation of the recognition sequence along with a neighboring tyrosine in peptides or proteins allows for rapid D42-mediated conversion of the tyrosine to an orthoquinone, which can be selectively modified with a variety of nucleophiles. We demonstrate the utility of our design system by rapidly installing cytotoxic molecules on a monoclonal antibody. For Table of Contents Only O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=72 SRC="FIGDIR/small/684804v1_ufig1.gif" ALT="Figure 1"> View larger version (20K): org.highwire.dtl.DTLVardef@d78510org.highwire.dtl.DTLVardef@160e406org.highwire.dtl.DTLVardef@1a52eeorg.highwire.dtl.DTLVardef@3a9bd3_HPS_FORMAT_FIGEXP M_FIG C_FIG

Although enzymatic retrosynthesis planning is widely used to accelerate drug synthesis and discovery, it still faces key challenges, including limited knowledge transfer from general chemical space and incomplete enzyme assignment. Here, we introduce Enzyformer, an innovative model that integrates the pretraining of molecules and reactions to capture both the syntax of SMILES and the transformation rules of organic reactions. With this two-stage pretraining strategy, Enzyformer improves top-1 and top-10 accuracies in retrosynthesis prediction by 7.5% and 11.7%, respectively, compared with the baseline R-SMILES. In addition, the incorporation of a contrastive learning model enables the highest reported F1-score of 0.922 for first-level EC number assignment. In conclusion, Enzyformer offers a promising solution for improving the accuracy and interpretability of enzymatic retrosynthesis prediction, which may help to reduce researchers’ workload and cost while accelerating drug design and mitigating the environmental impact of drug synthesis.

Modified (non-natural) prenoids are useful as chemical biology tools, designer fragrances and building blocks of bioactive compounds. However, their laborious synthesis generally limits their widespread application. Although prenylelongases offer potential biocatalytic access to modified long-chain prenoids, these enzymes have remained underexplored since i) they are difficult to assay in high throughput and ii) their pyrophosphate substrates are challenging and expensive to synthesize. To address these gaps, we herein report how prenylelongases can be interrogated with continuous high-throughput assays using readily accessible monophosphates of their native and modified substrates, bypassing the need for laborious pyrophosphate synthesis. We demonstrate the utility of this assay platform for the biochemical characterization of a panel of diverse prenylelongases, their profiling with non-native substrates, and two exploratory engineering campaigns. Our results illustrate that prenylelongases are biochemically diverse, promiscuous, engineerable, and synthetically useful enzymes. As such, our work provides the methodological repertoire and enzymological understanding required for biocatalytic access to modified prenoids.

Enzymes are a key enabling technology for sustainable chemical manufacturing due to their exquisite selectivity under mild conditions. However, their broader application is bottlenecked by the resource-intensive process of enzyme engineering. Here, we introduce a multi-modal model designed to predict the fitness of galactose oxidase and accelerate design of high performing mutants. The model predicted enzyme activity with 75% accuracy and achieved 67% precision for mutants with ≥0.8-fold activity against 1-phenyl-1-butanol. Retrospective sequence analysis reduced training data requirements by 41% without significant precision loss. Applied to 14 additional substrates, the model achieved 73% precision for ≥0.8-fold activity mutants. Finally, iterative model application for in-silico directed evolution yielded double mutants with 70% precision for ≥0.8-fold activity mutants. This work contributes toward understanding of galactose oxidase functionality, speeds up the enzyme optimization process by minimizing experimental demands, and paves the way to more efficient, eco-friendly, and cost-effective industrial enzyme applications.

The effect of substitutions at the 2′ position of the nucleotides at the ligation site on phosphoramidate-type chemical ligation reaction efficiency is examined. The efficiency varies depending on the substituents, with the combination of a fluorine group on 5′NRNA and an O-methyl group on RNA3′P showing the highest efficiency.

Chemical ligation of RNA fragments is an effective method for synthesizing long RNAs. It is particularly useful for incorporating site-specific modifications into long RNAs; however, its low reaction efficiency remains a major challenge. Herein, 5′-amino-2′-substituted uridine or cytidine nucleotides are synthesized, which are attached to the 5′ end of a synthetic RNA and conjugated with a phosphate group at the 3′ end of an alternative RNA fragment, to perform a head-to-tail-type RNA ligation. It is examined how the 2′ position of nucleotides with 5′-amino or 3′-phosphate groups in the ligated site affects the ligation reaction. The 2′-fluoro-5′-amino-nucleotide shows enhanced reactivity compared with the 2′-ribo- or 2′-deoxy-5′-amino-nucleotide. In contrast, the 2′-O-methyl modification demonstrates optimal efficacy with the 3'-phosphate nucleotide. Interestingly, the 2′-fluoro-5′-amino nucleotide maintains remarkable reactivity, even under acidic conditions. A pre-miRNA comprising 84 nucleotides is synthesized through ligation, and the intracellular functionality of the ligated RNA is confirmed. This study elucidates the effect of 2′ substituents on chemical ligation reactions, providing insights into the chemical synthesis of long RNAs.

An enzyme-catalyzed strategy for the stereoselective synthesis of chiral 3-n-Butylphthalide (NBP), an effective commercial drug for treating acute ischemic stroke, is developed. Two carbonyl reductases, SmCR_K6 and SsCR_K1, are used to produce (S)- and (R)-NBP continuously from aromatic ketones with 94% ee (S) and 99% ee (R), achieving space-time yields that are 9-fold and 30-fold higher, respectively, than those obtained in batch reactions.

3-n-Butylphthalide (NBP) is an effective commercial drug for the treatment of acute ischemic stroke, with its S-enantiomer, (S)-NBP, demonstrating clinical superiority over (R)-NBP. However, the stereoselective synthesis of both enantiomers with high enantiomeric excess (ee) presents significant challenges. Herein, a novel enzymatic strategy for the efficient and highly stereoselective synthesis of (S)- and (R)-NBPs under mild reaction conditions is presented. Specifically, two carbonyl reductases, SmCR_K6 and SsCR_K1, are engineered to facilitate the asymmetric reduction of the prochiral aromatic ketone, 2-pentanoyl benzonitrile (1), followed by intramolecular cyclization to produce chiral NBPs. SmCR_K6 exhibits a catalytic activity of 0.46 Umg^{− 1} protein, which is 23-fold greater than that of its parent enzyme, SmCR_V4, and the ee value of (S)-NBP increases from 2% to 94% (S). The catalytic activity of SsCR_K1 (9.46 U mg⁻¹ protein) is fourfold higher than that of its parent, SsCR, and the ee of (R)-NBP reachs 99%. In the preparative synthesis, (S)-NBP and (R)-NBP are generated continuously in a 3D microfluidic reactor, achieving space-time yields that are 9-fold and 30-fold higher, respectively, than those obtained in batch reactions. This continuous-flow enzymatic process has the potential for future scale-up for the industrial production of these important chiral drugs and similar derivatives.

Light-triggered bond cleavage is an essential biotechnology in chemical biology research. This concept highlights three photocleavable modalities—proteins, small molecules, and metal complexes—distilling design principles for wavelength tuning, cleavage efficiency, and biocompatibility. By comparing strengths/limits, we discuss the next generation of photo-responsive tools for biological research.

The spatiotemporal control of biomolecular functions via light-triggered bond cleavage has emerged as a powerful approach in chemical biology and cell biology. In this concept review, three major modalities of photo-cleavable systems—proteins, small molecules, and metal complexes—are classified and discussed, highlighting their design principles, biological applicability, and remaining challenges. Emphasis is placed on recent efforts to address key design challenges—such as balancing functional performance, biological compatibility, and optical responsiveness—across different molecular modalities, offering perspectives for the next generation of photo-responsive tools for biological research.

Nature Synthesis, Published online: 24 October 2025; doi:10.1038/s44160-025-00898-1

Inspired by nature’s spatially organized catalytic systems, multiscale confinement emerges as a powerful strategy for programmable biocatalysis. This Review highlights recent advances in designing artificial enzymes, enzyme immobilization methods and substrate channelling techniques, discussing current challenges and future directions towards efficient, sustainable enzyme catalysis for organic synthesis.

Polystyrene (PS), long considered non-degradable by biocatalytic pathways, can now be broken down under aqueous conditions using atmospheric air and a laccase–mediator system composed of a commercially available fungal enzyme and 1-hydroxybenzotriazole as small organic mediator. By formulating PS into stable colloidally nanoparticles, we unlock access for biocatalytic attack, leading to a drastic molar mass reduction. This mild, effective method opens new paths for sustainable plastic depolymerization without harsh chemicals or energy-intensive processes.

Abstract

Polystyrene (PS) is one of the most widely used synthetic polymers, with annual global production of around 20 million tons. However, its robust C─C backbone renders it highly recalcitrant to (bio)chemical depolymerization, and no sustainable re-/up-cycling method has yet been developed. Here, we establish a proof-of-concept for the efficient depolymerization of PS under mild aqueous conditions, using a laccase–mediator system (LMS) composed of Trametes versicolor laccase, 1-hydroxybenzotriazole (HBT), and ambient oxygen. To overcome substrate accessibility issues, PS is formulated into colloidally stable nanoparticles, promoting interfacial remote biocatalysis. Under such conditions, up to 99.9% decrease in molar mass is achieved from an initial PS of over 2 million g mol⁻¹, synthesized by ab initio free-radical emulsion polymerization. This colloidal dispersion strategy is also effective for commercial PS and expanded PS waste processed by post-dispersion in surfactant-containing aqueous media. Mechanistic studies suggest that LMS-mediated depolymerization proceeds via HBT radical diffusion into PS nanoparticles, triggering hydrogen atom transfer (HAT)-based oxidation and β-scissions of PS chains. This approach provides an efficient method for PS depolymerization using aqueous conditions, ambient O₂ and a native enzyme without harsh solvents or experimental conditions.

A phomactatriene synthase (SiPS) and two verticillene synthases (LxVS and AxVS) from bacteria were identified through genome mining. DFT calculations show that phomactatriene and verticillene compounds share a common rearrangement pathway. Comparative analyses of the synthase structural models, together with molecular modelling and site-directed mutagenesis, enabled their functional interconversion.

Abstract

Terpenoids represent the most structurally diverse class of natural products on Earth. Terpene synthases are key enzymes for constructing the complex and varied terpene skeletons by catalyzing the formation of multiple carbon–carbon bonds. Phomactatriene and verticillene family natural products are both classified as bicyclic diterpenoids, sharing a unique bicyclo[9.3.1]pentadecane skeleton. In this study, we used genome mining to identify the phomactatriene synthase SiPS from bacteria, together with two verticillene synthases, LxVS and AxVS. Our DFT calculations revealed that the rearrangement pathways for compounds in the phomactatriene and verticillene families follow a shared biosynthetic route. Furthermore, through comparative structural model analyses of the phomactatriene and verticillene synthases, we employed molecular modelling and site-directed mutagenesis to facilitate functional interconversion between these distinct terpene synthases. This work enhances our understanding of terpene biosynthesis and the potential for engineering terpene synthases for biotechnological applications.

We have converted the only known true bacterial monoterpene synthase, cineole synthase from Streptomyces clavuligerus (bCinS, C₁₀ substrate), to a highly competent sesquiterpene synthase (C₁₅) with a minimum number of rational mutations. By comparison with diterpene synthases (C₂₀), we were then able to bestow diterpene synthase activity on bCinS. This is the first time the doubling of the substrate length has been reported for a monoterpene synthase.

1,8-cineole synthase from Streptomyces clavuligerus (bCinS) is the only known bacterial terpene synthase that shows exclusive activity towards the monoterpene substrate geranyl diphosphate (GPP; C₁₀). Unlike most plant terpene synthases, bCinS is a high-fidelity enzyme producing 1,8-cineole as the predominant product (> 95%). A large number of bulky aromatic residues in the active site steer the carbocationic intermediates down a single path and restrict the conversion of larger prenyl-diphosphate substrates. Previously, we have shown that a single Phe-to-Ala mutation (F74A or F179A) allows bCinS to convert farnesyl diphosphate (FPP; C₁₅) into sesquiterpenoid products, including sesquicineole and germacrene A. Here, we made combinatorial mutations of aromatic active site residues to further expand the substrate scope of bCinS. The F74A-F179A double variant was not only more active than the wild type but showed increased activity towards FPP over GPP, with sesquicineole and cineole as the main products from these substrates, respectively. Computational active site volume analysis identified an additional residue, W58A, that unlocked activity towards the diterpene substrate geranylgeranyl diphosphate (GGPP; C₂₀), with the W58A-F74A-F179A triple variant showing the highest activity on this substrate. Remarkably, these key variants all appear to use the same 1,6 cyclisation cascade to form their main products from GPP, FPP, and GGPP. These results show that even high-fidelity terpene synthases such as bCinS can be engineered to accept different prenyl-pyrophosphate substrates without affecting the fundamental reaction cascade.

Science Advances, Volume 11, Issue 43, October 2025.

Science Advances, Volume 11, Issue 43, October 2025.

Nature, Published online: 28 October 2025; doi:10.1038/d41586-025-03546-y

The developers of OpenFold3 have released an early version of the tool, which they hope will one day perform on par with DeepMind’s protein-structure model.

A novel non-productive substrate binding conformation was identified in the PcHNL5_L331A mutant by crystallography and computation. Destabilizing this state through mutagenesis led to a triple mutant with enhanced hydrocyanation activity. These findings highlight a strategy to improve enzyme function by targeting unfavorable states associated with substrate binding.

Abstract

Understanding enzyme–substrate conformational transformations is crucial to the design and engineering of biocatalysts. However, the mechanisms by which substrates undergo dynamic transformations that regulate the function of an enzyme remain poorly understood. Hydroxynitrile lyase from Prunus communis (PcHNL5) catalyzes the cleavage of cyanohydrins. Its reverse reaction holds significant synthetic potential for the preparation of pharmaceutical precursors. Using a combination of crystallography and computational experiments, a novel flipped substrate binding state is identified within the substrate tunnel of the PcHNL5_L331A mutant. This binding state is non-productive and undergoes a conformational change before the catalytic cycle can proceed. Site-saturation mutagenesis led to the discovery of a triple mutant, PcHNL5_{L331A/S333V/P340L}, that destabilizes the non-productive substrate binding state thereby facilitating its transition to the catalytically productive conformation and significantly enhancing catalytic efficiency. Crystallographic studies provide a structural description of the factors that stabilize versus destabilize the different binding conformers in the different enzyme variants and thus the differing catalytic efficiencies. These findings demonstrate that destabilizing unfavorable substrate binding conformations within an enzyme active site can improve functionality and provide a promising strategy for designing efficient biocatalysts.

An unexpected heme-dependent enzyme, Alp1J, was identified as the catalyst that installs N─N bond in kinamycin biosynthesis. Alp1J forms a stable complex with ferredoxin Alp1I, which is essential for activity. Together they convert l-aspartate and nitrite into a hydrazine intermediate through a four-electron reductive pathway. The discovery establishes a platform for genome mining and synthetic biology aimed at novel N─N-containing therapeutics.

Abstract

A nitrogen–nitrogen (N─N) bond is a core feature of diverse natural products with interesting structural and biological properties. Kinamycin and lomaiviticin, featuring a diazobenzo[b]fluorene core, exhibit exceptional potency as chemotherapeutic agents. However, the N─N bond forming step in their biosynthesis has remained elusive. Through extensive mutagenesis and biochemical studies, we herein report that Alp1J, belonging to a new family of heme-dependent enzymes, catalyzes the N─N bond formation in kinamycin biosynthesis. Interestingly, Alp1J forms a stable complex with its partner ferredoxin Alp1I, which can protect the cofactors and is critical for the N─N bond formation activity. With its partner ferredoxin, Alp1J catalyzes formation of the hydrazine intermediate directly from l-aspartate and nitrite by a pathway involving four-electron reduction. Our findings expand the knowledge of enzymatic N─N bond formation and show the potential for the discovery and development of novel N─N bond containing natural products through genome mining and synthetic biology.