Integration of metabolome and microbiota datasets
To test our hypothesis in which stress-specific root exudates can influence the relative abundance of microbial taxa, we built a framework that integrates microbiota and metabolome data. This investigation targets at elucidating the influence of each treatment on both exudates and microbiota, and to identify microbial taxa and compounds that may coordinately interact in response to stress(es). Excitingly, the variance decomposition step revealed three factors explaining a high proportion of the variance, jointly for metabolome and microbiota in the three different treatments (Supplementary material 1). For each treatment, we then selected the top 0.1% of bacterial taxa and the top 10% of metabolites that mostly contribute to explain the variation within each factor, and then we correlated the abundance of the bacterial taxa and metabolites, then extracting significative correlations.
The analysis of drought-stresses plants suggests that two compounds (p-coumaric acid ethyl ester and L-serine) negatively correlate with several bacterial taxa. Both compounds were negatively correlated with the same pool of taxa (Mucilaginibacter , Sphingomonas ,Acidibacter, Paenibacillus, Filimonas, Labrys, Mesorhizobium, Heliimonas, Rudaea ) with the only difference that L-serine negatively correlated with a single unidentified bacterium, while p-coumaric acid ethyl ester negatively correlated with 2 unidentified bacterial OTUs (Supplementary material 1).
Plants exposed to heat treatment showed five compounds [aminocyclopropane-1-carboxilic, rhamnetin, L-histidine, Quercetin 3-O -(6-acetyl-galactoside) 7-O -rhamnoside and 2-methoxy-5-prop-1-enylphenol/eugenol] correlating negatively with a total of 23 bacterial taxa (Supplementary material 2). Interestingly, the first four molecules negatively correlated with a common pool of 22 taxa (Supplementary Material 1) including the genera Bacillus ,Bacteriovorax , Caulobacter , Chitinophaga ,Dyadobacter , Heliimonas , Mesorhizobium ,Micromonospora , Niastella , Pseudonocardia ,Rhizobacter together with 6 OTUs of unidentified bacteria. 2-Methoxy-5-prop-1-enylphenol/Eugenol negatively correlated with 4 OTUs: one belonging to the genus Alsobacter and three toCaulobacter (Supplementary material 1).
When plants were exposed to combined stress, the analysis suggests that two compounds (L-homocysteine and cinnamoyl glucose) negatively correlate with bacterial taxa, while a positive correlation was observed only for carnosic acid. L-homocysteine influenced the abundance of 25 taxa (Supplementary material 1) including Bacillus ,Clavibacter , Devosia , Dietzia ,Microbacterium , Paenibacillus , Paracoccus ,Rhizobium , Rhodococcus , Stenotrophomonas . Cinnamoyl glucose and carnosic acid affected, in opposite direction, a single OTU of Paenibacillus (Supplementary material 1).