Hydrogen is increasingly viewed as a key fuel for low-carbon energy systems, but most industrial hydrogen is still produced from fossil resources. Anaerobic dark fermentation offers a biological alternative that can convert organic waste into hydrogen under relatively mild conditions. However, practical yields remain constrained by competing metabolic pathways, especially routes that divert carbon and reducing power into ethanol, acetate, biomass, or other byproducts. Existing studies have often optimized culture conditions or individual pathways without fully capturing whole-cell resource allocation. Due to these challenges, deeper research is needed into the system-level metabolic rules that govern microbial hydrogen production.

In a study in Environmental Science and Ecotechnology, researchers from the National Technology Innovation Center of Synthetic Biology, Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences; the Key Laboratory of Environmental Biotechnology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences; Tianjin Key Laboratory of Aquatic Science and Technology, School of Environmental and Municipal Engineering, Tianjin Chengjian University; and the State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology (Shenzhen), reported an enzyme-constrained genome-scale metabolic model (ecGEM) that quantitatively resolves the growth–hydrogen production trade-off in Ethanoligenens harbinense YUAN-3.

The team first constructed a genome-scale metabolic model (GEM), named ixeh674, for YUAN-3, covering 674 genes, 977 metabolites, and 1,063 reactions. To make the model more biologically realistic, they rebuilt the biomass equation using experimentally measured protein, deoxyribonucleic acid (DNA), ribonucleic acid (RNA), glycogen, and amino acid composition. After manual curation and carbon-source validation, the model’s prediction accuracy improved from 64.71% to 91.42%. The researchers then added enzyme turnover numbers, or kcat values, predicted using DLkcat, a deep-learning-based tool for enzyme kinetic prediction, and built the enzyme-constrained model ecixeh674 with ECMpy. Unlike the conventional GEM, ecixeh674 avoided unrealistic overestimation of growth and hydrogen yield by accounting for finite enzyme resources. It showed that rapid growth draws enzyme capacity toward precursor synthesis, leaving fewer resources for hydrogen-producing pathways. During the stationary phase, growth slowed while hydrogen yield increased, matching experimental observations. The model also revealed that diverting carbon and reduced nicotinamide adenine dinucleotide (NADH) flux toward glutamate and glutamine biosynthesis reduced ethanol formation and supported higher hydrogen production. In single-gene knockout simulations, deletion of Ethha_1547, encoding phosphoglycerate kinase, increased hydrogen flux by about 30% under low-carbon conditions.

The authors said the study shows that improving biohydrogen production is not simply a matter of pushing one pathway harder. "Microbial cells must divide limited enzyme and energy resources between growth, survival, and product formation," they said. "By making those hidden trade-offs visible, the model offers a practical route for selecting engineering targets. It helps explain when hydrogen production can be enhanced, when growth becomes a limiting cost, and why balanced strain design is essential for future fermentation systems."

The findings provide a model-guided path for engineering hydrogen-producing microbes beyond trial-and-error optimization. By identifying metabolic routes that redirect carbon, NADH, and adenosine triphosphate (ATP), the ecGEM framework can support rational strain design to improve hydrogen yield while maintaining viable growth. The approach may also be extended to mixed-substrate fermentation, microbial communities, and reactor-scale process design, where substrate competition and community stability remain major bottlenecks. As biological hydrogen production moves toward industrial use, enzyme-constrained modeling could become a useful decision-making platform for linking microbial metabolism with cleaner energy production.