Identification of gut microbial bile acid metabolic enzymes via an AI-assisted pipeline

Cell

August 07, 2025

Yong Ding, Xi Luo, Jiasheng Guo, Baiying Xing, Haoyu Lin, Haohan Ma, Yicun Wang, Meng Li, Chuan Ye, Sen Yan, Kangjie Lin, Jinxin Zhang, Yingying Zhuo, Qixing Nie, Donghui Yang, Zhipeng Zhang, Yanli Pang, Kai Wang, Ming Ma, Luhua Lai, Changtao Jiang

https://doi.org/10.1016/j.cell.2025.07.017

ABSTRACT: The modifications of bile acids (BAs) are fundamental to their role in host physiology and pathology. Identifying their synthetases is crucial for uncovering the diversity of BAs and developing targeted interventions, yet it remains a significant challenge. To address this hurdle, we developed an artificial intelligence (AI)-assisted workflow, bile acid enzyme announcer unit tool (BEAUT), which predicted over 600,000 candidate BA metabolic enzymes that we compiled into the human generalized microbial BA metabolic enzyme (HGBME) database (https://beaut.bjmu.edu.cn). We identified a series of uncharacterized BA enzymes, including monoacid acylated BA hydrolase (MABH) and 3-acetoDCA synthetase (ADS). Notably, ADS can produce an unreported skeleton BA, 3-acetoDCA, with a carbon-carbon bond extension. After determining its bacterial source and catalytic mechanism, we found that 3-acetoDCA is widely distributed among populations and regulates the microbial interactions in the gut. In conclusion, our work offers alternative insights into the relationship between microbial BAs and the host from an enzymatic perspective.

摘要：胆汁酸 (BAs) 的修饰对其在宿主生理和病理学中的作用至关重要。鉴定其合成酶对于揭示BAs的多样性和开发有针对性的干预措施至关重要，但这仍然是一项重大挑战。为了解决这一难题，我们开发了一种人工智能 (AI) 辅助工作流程——胆汁酸酶播报单元工具 (BEAUT)。该工具预测了超过600,000种候选BA代谢酶，并将其汇编到人类通用微生物BA代谢酶 (HGBME) 数据库 ( https://beaut.bjmu.edu.cn ) 中。我们鉴定了一系列未表征的BA酶，包括单酸酰化BA水解酶 (MABH) 和3-乙酰DCA合成酶(ADS)。值得注意的是，ADS可以生成一种未报道的骨架BA，即3-乙酰DCA，其具有碳碳键的延伸。在确定了其细菌来源和催化机制后，我们发现3-乙酰DCA在人群中分布广泛，并调节肠道微生物的相互作用。总之，我们的研究从酶学角度为微生物BA与宿主之间的关系提供了新的见解。

Keywords: ADS; 3-acetoDCA; MABH; gut microbiota; bile acid; biosynthesis enzyme; artificial intelligence

关键词：3-乙酰DCA合成酶；单酸酰化BA水解酶；肠道微生物群；胆汁酸；生物合成酶；人工智能

SUMMARY: Through functional validation, the study identified more than 40 previously unrecorded BA-metabolizing enzymes. Among these, monoacyl 3-acylated BA hydrolase (MABH)—derived from Hungatella hathewayi—is responsible for hydrolyzing monoacyl 3-acylcholic acids (such as 3-acetylcholic acid and 3-propionylcholic acid) but exhibits no activity toward 3-succinylcholic acid. Additionally, a novel class of thiamine pyrophosphate (ThDP)-dependent BA-metabolizing enzymes, namely 3-acetoxydeoxycholic acid synthases (ADS), was discovered. These enzymes can produce the previously unreported carbon-skeleton bile acid 3-acetoxydeoxycholic acid (3-acetoDCA) via an aldol condensation reaction. It was confirmed that Bacteroides ovatus is the primary producer of 3-acetoDCA, and its ADS gene knockout strain completely loses the ability to synthesize 3-acetoDCA. Structural biology studies revealed that BoADS (ADS derived from Bacteroides ovatus) belongs to the 1-deoxy-D-xylulose-5-phosphate synthase (DXPS) family. It binds to the substrate 3-oxodeoxycholic acid (3-oxoDCA) through a unique "sandwich"-style hydrophobic interaction and catalyzes the reaction between 3-oxoDCA and pyruvate to generate 3-acetoDCA. Analysis of clinical samples showed that 3-acetoDCA is widely present in human feces, and its concentration is positively correlated with the abundance of Bacteroides ovatus and the read count of the ADS gene. At physiological concentrations, 3-acetoDCA does not act on traditional BA receptors such as FXR, VDR, PXR, and TGR5. Instead, it regulates gut microbiota composition by promoting the growth of Lactobacilli (particularly Lactobacillus gasseri) in the presence of arginine/ornithine, thereby increasing the levels of tryptophan metabolites ((such as indole-3-lactic acid) in mouse feces and colons—a effect that may play a protective role in metabolic diseases. The application of BEAUT and the discovery of novel BA-metabolizing enzymes provide a "top-down" paradigm for research on microbial BA metabolism and expand the understanding of BA chemical diversity and microbe-host interactions.

总结：该研究开发了一种人工智能辅助流程BEAUT（胆汁酸酶播报单元工具），用于挖掘肠道微生物中的胆汁酸（BA）代谢酶。BEAUT 基于蛋白质语言模型ESM2生成蛋白质功能感知表征，并通过密集神经网络从宏基因组数据中筛选BA代谢酶，结合已知BA代谢酶序列及具有相似底物结合口袋的扩展序列进行训练，从人类肠道微生物基因组序列中预测出超60万个潜在BA代谢酶，并构建人类通用微生物BA代谢酶（HGBME）数据库（https://beaut.bjmu.edu.cn）供公开访问。通过功能验证，研究发现了40多种未被记录的BA代谢酶，其中单酸酰化BA水解酶（MABH）来自Hungatella hathewayi，负责单酸3-酰胆酸（如3-乙酰胆酸、3-丙酰胆酸）的水解，对3-琥珀酰胆酸无活性；还发现一类依赖硫胺素焦磷酸（ThDP）的新型BA代谢酶——3-乙酰DCA合成酶（ADS），其能通过羟醛缩合反应生成此前未被报道的碳骨架胆汁酸3-乙酰DCA（3-acetoDCA），明确多形拟杆菌（Bacteroides ovatus）是3-acetoDCA的主要产生菌，且其ADS基因敲除菌株完全丧失3-acetoDCA合成能力。结构生物学研究揭示，BoADS（多形拟杆菌来源的ADS）属于1-脱氧-D-木酮糖-5-磷酸合成酶（DXPS）家族，通过特殊的“三明治”式疏水相互作用结合底物3-氧代脱氧胆酸（3-oxoDCA），催化其与丙酮酸反应生成3-acetoDCA。临床样本分析显示，3-acetoDCA 在人类粪便中广泛存在，且其浓度与多形拟杆菌丰度及ADS基因正相关；生理浓度下的3-acetoDCA不作用于FXR、VDR、PXR、TGR5等传统BA受体，而是通过促进精氨酸/鸟氨酸存在时乳杆菌（尤其是加氏乳杆菌）的生长来调节肠道菌群组成，进而增加小鼠粪便和结肠中吲哚-3-乳酸等色氨酸代谢物水平，在代谢疾病中发挥保护作用。BEAUT的应用及新型BA代谢酶的发现，为微生物BA代谢研究提供了“自上而下”的新范式，拓展了对BA化学多样性及微生物-宿主相互作用的认知。