Metagenome-enabled models improve genomic predictive ability and identification of herbivory-limiting genes in sweetpotato.

Authors: CHAM, ALHAGIE - University Of Tennessee; ADAMS, ALISON - University Of Tennessee; Wadl, Phillip; OJEDA-ZACARIAS, MARIA - Autonomous University Of Nuevo León; Rutter, William; Jackson, D; SHOEMAKER, D - University Of Tennessee; BERNARD, EARNEST - University Of Tennessee; YENCHO, G - North Carolina State University; OLUKOLU, BODE - University Of Tennessee

Submitted to: Horticulture Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/27/2024
Publication Date: 5/10/2024
Citation: Cham, A.K., Adams, A.K., Wadl, P.A., Ojeda-Zacarias, M., Rutter, W.B., Jackson, D.M., Shoemaker, D.D., Bernard, E.C., Yencho, G.C., Olukolu, B.A. 2024. Metagenome-enabled models improve genomic predictive ability and identification of herbivory-limiting genes in sweetpotato.. Horticulture Research. Uhae135. https://doi.org/10.1093/hr/uhae135.
DOI: https://doi.org/10.1093/hr/uhae135

Interpretive Summary: Plant-insect interactions are often influenced by microorganisms within host- and insect-associated genomes. While the understanding of genetic and mechanistic basis for host resistance can facilitate solutions targeted towards crop protection, breeding programs often depend on predictive models for crop improvement. Recent advancements in data acquisition and sequencing technologies offer unprecedented opportunities for investigating complex traits and the genes that control broad-spectrum and durable resistance. In our study, we present the implementation of genome wide associations and genomic prediction in sweetpotato using genome-wide data and species-level microorganism genome profiles that were simultaneously derived from next generation sequencing technologies. Our findings validate the theory of host-microorganism genome co-evolution and reveal that the concept can be applied to significantly improve accuracy of genomic estimated breeding values in sweetpotato. If only variation in the host genome is used for genomic prediction, predictive accuracy is expected to be limited by missing heritability problems. Consequently, we propose that accounting for variation due to the host-association microorganism genome can effectively explain some of this missing heritability. While some traits may be perturbed to a lesser degree by microorganisms, it is expected that microorganism-enabled genomic prediction models will be important for plant breeders focused on improving agronomic traits such as disease resistance, stress tolerance, nutrient acquisition in plants, feed efficiency and obesity in animal, and methane production in ruminants.

Technical Abstract: Plant-insect interactions are often influenced by microbes within host- and insect-associated metagenomes. Here, we used sweetpotato (Ipomoea batatas) leaf-associated metagenomes of diversity and biparental populations to obtain relative abundance estimates of insects, and the microbes that modulate host-insect interactions. A quantitative reduced representation sequencing (qRRS) of the metagenomes underscored the importance of strain/species-level resolution for functional metagenomics by identifying multipartite and multitrophic interactions. Correlation network analysis confirmed positive correlations between whitefly (Bemisia tabaci) and its endosymbionts (Candidatus Hamiltonella defensa, Candidatus Portiera aleyrodidarum, and Rickettsia spp.), while negative correlations with nitrogen-fixing bacteria implicate nitrogen (nitric oxide) in sweetpotato-whitefly interaction. Genome-wide associations (GWA) with 252,977 dosage-based markers implicated ethylene (ACC oxidase) and cell wall modification (lignin-forming anionic peroxidase and beta-xylosidase) in sweetpotato-whitefly interaction. Predictive abilities (PA) for whitefly and Ocypus olens abundance were higher in a 767 accession diversity population (69% and 35.8%, respectively) than in the biparental population (58.3% and 16.2, respectively; 61.7% for Frankliniella occidentalis). The high PA validates the qRRS-based quantification accuracy and role of host genetics in observed variation. Modeling the metagenome as a covariate improved PA by 25% and reduced false positive rates during GWA. Allele dosage (6x) and dominance relationship matrix significantly improved PA by as much as 76% and 18%, respectively. Pseudo-diploidized (2x) genotype data consistently underperformed for digenic dominance models. PA increased with increasing marker densities but plateaued at 10,000 markers. Our findings validate the holobiont theory of host-metagenome co-evolution and underscore its potential for breeding within the context of GxGxE interactions.