Disentangling the syntrophic electron transfer mechanisms of Candidatus geobacter eutrophica through electrochemical stimulation and machine learning

• Published in: Scientific reports Vol. 11; no. 1; pp. 15140 - 14
• Main Authors: Yuan, Heyang, Wang, Xuehao, Lin, Tzu-Yu, Kim, Jinha, Liu, Wen-Tso
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 23.07.2021; Nature Publishing Group; Nature Portfolio
• Subjects: 631/326/171/1878; 631/61/168; Bayesian analysis; Diet; Electrochemistry; Electron transfer; Enzymes; Genomics; Geobacter; Humanities and Social Sciences; Hydrogen; Learning algorithms; Machine learning; Mathematical models; Metabolic pathways; Metabolism; Methanogenesis; multidisciplinary; Phylogeny; Science; Science (multidisciplinary); Silver chloride; Transfer learning
• DOI: 10.1038/s41598-021-94628-0
• ISSN: 2045-2322

Interspecies hydrogen transfer (IHT) and direct interspecies electron transfer (DIET) are two syntrophy models for methanogenesis. Their relative importance in methanogenic environments is still unclear. Our recent discovery of a novel species Candidatus Geobacter eutrophica with the genetic potential of IHT and DIET may serve as a model species to address this knowledge gap. To experimentally demonstrate its DIET ability, we performed electrochemical enrichment of Ca. G. eutrophica-dominating communities under 0 and 0.4 V vs. Ag/AgCl based on the presumption that DIET and extracellular electron transfer (EET) share similar metabolic pathways. After three batches of enrichment, Geobacter OTU650, which was phylogenetically close to Ca. G. eutrophica, was outcompeted in the control but remained abundant and active under electrochemical stimulation, indicating Ca. G. eutrophica’s EET ability. The high-quality draft genome further showed high phylogenomic similarity with Ca. G. eutrophica, and the genes encoding outer membrane cytochromes and enzymes for hydrogen metabolism were actively expressed. A Bayesian network was trained with the genes encoding enzymes for alcohol metabolism, hydrogen metabolism, EET, and methanogenesis from dominant fermentative bacteria, Geobacter , and Methanobacterium . Methane production could not be accurately predicted when the genes for IHT were in silico knocked out, inferring its more important role in methanogenesis. The genomics-enabled machine learning modeling approach can provide predictive insights into the importance of IHT and DIET.