Effect of microbiome inoculation on parental traits
Trait differences among parents were driven by plant genotype, environment and genotype by environment interactions (GxE). Parents differed in shoot and root traits across all treatments. For example, FIL2 produced 1.64-fold more shoot biomass (P <0.0001), 1.98-fold more root biomass (P <0.0001), 1.80-fold lower specific root length (SRL) (P <0.0001) and 1.46-fold higher RTD (P <0.0001) relative to HAL2 (Fig. 2a, d, e; Table S2; Table S3). These results mirrored earlier descriptive studies of P. hallii ecotypes (Palacio-Meija et al. , 2021; Lowryet al. , 2014), including studies of the shoot and root traits studied in current work. (Khasanova et al. 2019). Treatment also had a significant effect on plant traits (Fig. 2a-c, e, f; Table S2, Table S3). For example, inoculated plants had greater shoot biomass (1.35-fold more biomass in CI and 1.17-fold more in AI treatments relative to the MI treatment (P =0.027)), lower lateral root length (1.2-fold less in CI and 1.53-fold less in AI relative to MI (P =0.046)), and showed changes in specific leaf area (SLA) dependent upon treatment (1.05-fold increase in AI and 1.06-decrease in CI relative to MI (P =0.039)). Importantly, we also identified several ecotype x microbiome interactions (Fig. 2d-f; Table S2, Table S3). For example, SRL of FIL2 decreased 1.17-fold under AI and 1.33-fold under CI relative to MI soil, while HAL2 showed 1.1-fold increase in SRL under AI and no change under CI relative to MI (P =0.039; Fig. 2d; Table S2, Table S3). Root tissue density (RTD) of FIL2 increased 1.1-fold under AI and 1.36-fold under CI relative to MI, while HAL2 showed 1.1-fold decrease under AI and 1.1-fold increase under CI relative to MI (P=0.046; Fig. 2e; Table S2, Table S3). In total, seven traits showed ecotype differences between parents, five traits were affected by microbial treatment and three traits had significant ecotype x microbiome interaction (Fig. 2; Table S2, Table S3).
The impact of the microbiome on the quantitative genetic architecture of our measured traits was evaluated by comparing “base” and “GxE” linear mixed models. In 11 out of 12 cases, the GxE models were favored by AIC and log likelihood ratio tests (Table S4). Broad-sense heritability was low for most traits (ranging from 0.01 to 0.18; Table S2). Overall, we document considerable evidence that the microbiome modifies the expression of quantitative genetic variation in P. hallii .