Unraveling Metabolic Interactions in a Rhizosphere Microbial Community Through Genome-Scale Modeling
Manish Kumar* (email@example.com), Maxwell Neal, Deepan Thiruppathy, Oriane Moyne, Gabriela Canto Encalada, Anurag Passi, Rodrigo Santibañez, Diego Tec-Campos, and Karsten Zengler
University of California–San Diego
This project couples novel lab and field studies to develop the first predictive model of grass-microbiomes based on new mechanistic insights into dynamic plant-microbe interactions in the grasses Sorghum bicolor and Brachypodium distachyon that improve plant N use efficiency (NUE). The results will be used to predict plant mutants and microbial amendments that improve low-input biomass production for validation in lab and field studies. To achieve this goal, researchers will determine the mechanistic basis of dynamic exudate exchange in the grass rhizosphere with a specific focus on the identification of plant transporters and proteins that regulate root exudate composition and how specific exudates select for beneficial microbes that increase plant biomass and NUE. The team will further develop a predictive plant-microbe model for advancing sustainable bioenergy crops and will predictively shift plant-microbe interactions to enhance plant biomass production and N acquisition from varied N forms.
The team presents manually curated genome-scale metabolic models for 17 bacteria commonly found in the rhizosphere. These bacteria isolated from the switchgrass rhizosphere represent dominant members of the rhizosphere of grasses belonging to various genera, such as Arthrobacter, Bacillus, Bosea, Bradyrhizobium, Brevibacillus, Burkholderia, Chitinophaga, Lysobacter, Methylobacterium, Mucilaginibacter, Mycobacterium, Niastella, Paenibacillus, Rhizobium, Rhodococcus, Sphingomonas, and Variovorax. The models were generated using standard pipelines and incorporated a total of 3,877 reactions and 2,663 metabolites. Each model contains between 790 to 1,788 genes, covering 15% to 30% of the respective microbial genomes. The models were curated using information from literature and public databases such as KEGG, UniProt, and MetaCyc. All models were simulated on 215 different carbon and nitrogen sources and results were compared with experimental measurements and to improve model predictability. The models accurately predicted the growth phenotypes of rhizosphere bacteria 90% of the time. Specifically, there were 1,475 true-positive predictions (correctly identified growth), 1,542 true-negative predictions (correctly identified no growth), 388 false-positive predictions (incorrectly identified growth), and 250 false-negative predictions (incorrectly identified no growth).
Next, the team deployed these highly curated metabolic models to study the interactions in a synthetic microbial community (SynCom) of the rhizosphere. The team developed a computational framework for community metabolic models. For this the team included members in the community reconstruction with relative abundance above a minimum cutoff (>0.01%) under any tested condition. Each member’s metabolic model was treated as a separate compartment linked to a shared metabolic pool, with connections refined by experimentally determined phenotypic data (i.e., Biolog plates). Flux balance analysis was used to simulate the growth of individual members as well as that of the entire community. The biomass of each member and the community was optimized during the simulation. To further improve the computational framework, the team added a module that constrains the activity of reactions using metatranscriptomics and metatranslatomics data, reflecting internal resource allocation for each bacteria. This analysis predicts metabolic exchanges between community members and uncovers the nature of interactions, such as competition and cooperation, between rhizosphere microbes.
This work is supported by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research under Award DESC0021234.