Efficient and selective energy transfer photoenzymes powered by visible light

nature chemistry Article https://doi.org/10.1038/s41557-025-01820-0 Efficient and selective energy transfer photoenzymes powered by visible light Received: 25 October 2024 Accepted: 1 April 2025 Published online: xx xx xxxx Rebecca Crawshaw 1,4, Ross Smithson 1,4, Johannes Hofer2, Florence J. Hardy 1, George W. Roberts1, Jonathan S. Trimble1, Anna R. Kohn1, Colin W. Levy1, Deborah A. Drost 3, Christian Merten 3, Derren J. Heyes 1, Richard Obexer 1, Thorsten Bach 2 & Anthony P. Green 1 Check for updates The development of [2 + 2] cyclases containing benzophenone triplet sensitizers highlights the potential of engineered enzymes as a platform for stereocontrolled energy transfer photocatalysis. However, the suboptimal photophysical features of benzophenone necessitates the use of ultraviolet light, limits photochemical efficiency and restricts the range of chemistries accessible. Here we engineer an orthogonal Methanococcus jannaschii tyrosyl-tRNA synthetase/tRNA pair for encoding thioxanthone triplet sensitizers into proteins, which can efficiently harness visible light to drive photochemical conversions. Initially, we developed an enantioselective [2 + 2] cyclase that is orders of magnitude more efficient than our previously developed photoenzymes (kcat = 13 s−1, >1,300 turnovers). To demonstrate that thioxanthone-containing enzymes can enable more challenging photochemical conversions, we developed a second oxygen-tolerant enzyme that can steer selective C–H insertions of excited quinolone substrates to afford spirocyclic β-lactams with high selectivity (99% e.e., 22:1 d.r.). This photoenzyme also suppresses a competing substrate decomposition pathway observed with small-molecule sensitizers, underscoring the ability of engineered enzymes to control the fate of excited-state intermediates. Biological photocatalysis has emerged as a powerful strategy to unlock new chemical reactivity within protein active sites. In addition to a handful of natural photoenzymes1–4, a variety of cofactor-dependent enzymes have been repurposed as biocatalysts for stereocontrolled photoredox processes5–11. Recently our laboratory has introduced a powerful mode of photochemistry into proteins, namely triplet energy transfer (EnT) photocatalysis12–24, by developing an enantioselective [2 + 2] cyclase that relies on a genetically programmed benzophenone as a triplet sensitizer (Fig. 1)25. In a simultaneous report from Sun et al., a similar approach was used to develop selective photobiocatalysts for [2 + 2] cycloadditions of indole derivatives26. These studies suggest that designed photoenzymes could offer a versatile platform for mediating a wide variety of stereoselective EnT processes, including those that are beyond the scope of existing small chiral catalysts. Crucially, as benzophenone sensitizers can be genetically encoded27, they can be quickly and accurately positioned within a wide variety of protein scaffolds, in principle allowing the generation of photocatalytic sites with diverse sizes, geometries and arrangements of functional residues. In addition to this unrivalled flexibility, the efficiency and selectivity of designed photoenzymes can be readily optimized using directed-evolution workflows adapted to an expanded amino acid alphabet25,26. Despite their potential, the capabilities of EnT photoenzymes is currently limited by a reliance on benzophenone derivatives as triplet sensitizers, which can be encoded into proteins using a pre-existing Methanococcus jannaschii 1 Department of Chemistry & Manchester Institute of Biotechnology, The University of Manchester, Manchester, UK. 2Department of Chemistry and Catalysis Research Center, School of Natural Sciences, Technische Universität München, Garching, Germany. 3Ruhr-Universität Bochum, Faculty for Chemistry and Biochemistry, Bochum, Germany. 4These authors contributed equally: Rebecca Crawshaw, Ross Smithson. e-mail: ; Nature Chemistry Article https://doi.org/10.1038/s41557-025-01820-0 O – Weak absorbance features – Photocrosslinking and side reactions O UV C 250 nm – UV excitation S Thioxanthone 300 R196 Q195 BpA173 O NH2 W244 Acetophenone H287 350 Excitation O O H N H O O O In vivo translation O aa-tRNA 405 nm Visible O EnT1.3 OH Excitation H 365 nm, 1 h O S 365 nm 400 O S O UV A Y121 NH2 Engineered translation components Acetone HN – Requiring new translation components This work UV B Previous work N H + Strong absorbance features + Fewer side reactions + Existing translation components Benzophenone + Visible light excitation 450 Visible light S O New chemistries High efficiency Visible light energy transfer photoenzyme Fig. 1 | Comparison of benzophenone and thioxanthone triplet sensitizers. Left: previously we developed a photoenzyme for [2 + 2] cycloadditions (EnT1.3) by genetically encoding a non-canonical amino acid containing a benzophenone side chain (BpA, violet)25. Structural analysis of product-bound EnT1.3 (PDB: 7ZP7, product depicted in cyan) showed a sophisticated active site network, including π-stacking and hydrogen-bonding interactions. This photosensitizer requires irradiation with UV light (365–395 nm) for excitation. Chemical scheme showing the targeted intramolecular [2 + 2] photocycloaddition of an oxygen-linked quinoline substrate. Centre: Absorbance spectrum of smallmolecule benzophenone (violet) and thioxanthone (blue) recorded in PBS buffer with 10% DMSO. Right: here we have engineered translation components to allow genetic encoding of a non-canonical amino acid containing a thioxanthone side chain, which has strong absorbance features extending into the visible range. This system enhances the efficiencies and expands the range of chemistries accessible with energy-transfer photoenzymes. tyrosyl-tRNA synthetase/tRNA pair (MjTyrRS/tRNA)27. Benzophenone has weak absorbance features in the ultraviolet (UV) region, which overlap with many target substrates, leading to competing direct excitation processes that preclude selective catalysis28–30. Furthermore, excited benzophenones can undergo a variety of off-target processes, including electron and hydrogen-atom transfers, as evidenced by its common use as a photocrosslinking group27,31,32. Taken together, these limitations ultimately compromise photochemical efficiency and greatly restrict the range of chemistries accessible. Although new photocatalytic groups can be introduced post-translationally through covalent labelling, this approach requires a unique and accessible reactive handle within the target protein33–37. These methods also introduce long and flexible linkers, making accurate sensitizer positioning challenging, which compromises enzyme efficiency and engineerability. As a result, if we are to unlock a wider spectrum of EnT catalysis within proteins, new genetically programmable triplet sensitizers are required that surpass the photophysical properties of benzophenone. I (...truncated)