Los Alamos National Laboratory Los Alamos, United States
Microbiome optimization could improve the performance of biological systems, from humans to plants and ecosystems. Yet, owing to challenges in finding and cultivating microbiomes that maintain their function in field conditions, use of microbes to improve plant productivity and crop stress tolerance has not become widespread despite years of trials. Based on the strong interactions and interdependency of rhizosphere microbes and plants, directed plant-microbiome evolution has been suggested as a means for developing microbiomes for these purposes. To test the feasibility of this method in developing microbial communities that improve plant performance under reduced irrigation, we cultivated Zea mays from seed in an artificial soil inoculated with soil microbiomes originating from a pine forest or a historically-droughted maize field. In the initial generation, water use efficiency (WUE) and stomatal closure point (SCP) were measured once the plants grew 10 leaves. The microbiomes of three plants demonstrating the best or worst WUE or SCP values for each microbiome source were selected for propagation to the next generation, and the process was repeated for two additional generations. Microbiome composition was analyzed after each generation using Illumina MiSeq sequencing, and microbial groups affecting plant performance were identified using Latent Dirichlet Allocation.
Our results show that, in three generations, the microbiome originating from the forest soil was able to consistently influence plant SCP, while the microbiome from the agricultural field had no significant effect on either WUE or SCP. The forest and agricultural microbiomes remained distinct from one another throughout the evolutionary process, but the microbiomes adapted to the greenhouse experiment such that the parent and offspring microbiomes became progressively more similar in subsequent generations. Interestingly, we also found that the microbiome originating from forest soil consistently produced faster growing plants than the microbiome from the agricultural field. This microbiome contained more bacteria related to the nitrogen cycle than the agricultural microbiome, while the agricultural microbiome had a higher abundance of bacteria commonly found in dry soils. We were able to consistently identify consortia of bacteria related to different plant traits using a dimension reduction method called Latent Dirichlet Allocation (LDA). Our experiment demonstrates that, in only a few generations (3 as opposed to 6-10 in previous studies), directed evolution can produce soil microbiomes that influence important functional drought tolerance traits in maize, but not all soils may have the microbial diversity or species structure needed to optimize plant traits.