AI(AlphaFold3)を用いた光誘導ゲノム合成ツールの最適化

2025-10-22

2025-10-21 東京大学

東京大学大学院総合文化研究科の太田邦史教授らは、AI「AlphaFold3」を用いて光誘導型ゲノム合成技術「MagTAQingシステム」に使用する人工DNA切断酵素を合理的に設計し、ゲノム再編成効率を従来比77%向上させた。構造未知だったMboI酵素をAlphaFold3で精密予測し、基質DNAとの結合部位を最適化。これにより光照射で効率的に再編成を誘発する新ツールを開発した。本成果は、AIを活用したタンパク質機能改良の新たな応用例であり、今後は他の遺伝子編集技術への展開も期待される。研究結果は『The Journal of Biological Chemistry』誌に掲載。

拡散モデルを用いた構造未知のゲノム再編成ツールの構造推定(左から右へ細精化が進む)

<関連情報>

トップダウンゲノム工学のための光活性化エンドヌクレアーゼのAlphaFold3誘導最適化
AlphaFold3-guided optimization of a photoactivatable endonuclease for top-down genome engineering

Hideyuki Yone, Hiromitsu Kono, Moritoshi Sato, Kunihiro Ohta
Journal of Biological Chemistry Available online: 24 September
DOI:https://doi.org/10.1016/j.jbc.2025.110762

Recent advances in protein structure prediction by artificial intelligence have enabled the rational design of engineered enzymes with enhanced activity and precise regulatory features. Here, we report the AlphaFold3-guided enhancement of MagMboI, a photoactivatable restriction enzyme designed for light-controlled top-down genome engineering. MagMboI is derived from the type II restriction enzyme MboI and functions through a split-protein strategy in which its N- and C-terminal fragments are fused to light-inducible dimerization modules. Upon exposure to blue light, these domains heterodimerize, restoring nuclease activity in a controlled manner. Using AlphaFold3, we modeled the structure of the MagMboI-DNA complex and gained structural insights into the interaction between MagMboI and its target DNA recognition sequence (5′-GATC-3′) required for Mg²⁺-dependent DNA cleavage. Comparing neighboring split-site variants, we identified an alternative split that increases the MagMboI–DNA interface area and enhances complex stability relative to the original construct. This redesigned variant (designated MagMboI-plus) preserves α-helical integrity while strengthening protein-DNA contacts. Although MagMboI-plus, when introduced in Saccharomyces cerevisiae cells, exhibited slightly increased DNA-cleavage activity in vivo upon blue light activation, it was found to induce more pronounced genomic rearrangements compared to the original MagMboI construct. These findings demonstrate that AlphaFold3-based prediction can accelerate functional improvements in engineered enzymes, providing a strategy for developing light-controlled genome engineering tools.