Discussion
There is growing appreciation for the important and often complex interactions that exist between plants and their associated microbial communities. Exploring the genetic architecture of plant trait-microbiome interactions is an important step in determining if these interactions play a role in local adaptation and evolution. Here, we conducted a QTL study with a P. hallii RIL mapping population in soils inoculated with microbiomes from native P. halliihabitats to observe the impact of microbiomes on plant traits and genetic architecture. We found that the microbiota in the natural habitat of the RIL parents are distinct and served as suitable experimental treatments to quantify the effects of microbiota on host-plant traits. In this study, soils inoculated with native microbiomes drive trait plasticity in both above and below ground traits, and these effects were both general and location specific with respect to the origin of the microbial inoculum. We found QTL that displayed GxE for ten of twelve measured traits, suggesting widespread genetic variation in trait responses to plant-microbiome interaction. We also identified epistatically interacting QTL for root diameter present only in microbiomes from native locations, indicating that hybridization of ecotypes may disrupt genes and their interaction with microbes through root characteristics. Overall, our study suggests that the genetic architecture of host functional traits is significantly impacted by microbial associations.
It is clear that host traits are impacted by microbial communities. Although soil microbes interact directly and indirectly with the root system, they can induce changes that affect the entire plant. The presence of microbiomes from native soil inocula induced trait plasticity in above and belowground traits for the parents that was general and location specific (Fig. 2). For example, traits linked to resource acquisition such as specific leaf area (SLA) and specific root length (SRL) were altered in response to the presence of microbiomes. High SLA correlates with high nitrogen content and low structural investment in leaves, which yields high rates of photosynthesis to promote rapid growth (Cornelissen et al. 2003; Reich et al. 1997), a trait necessary in xeric environments with short seasons terminated by drought. This is consistent with high SRL, where plants produce longer and thinner roots with less structural input to search for water (Balachowski et al. 2016). SLA showed a plastic response to location specific native microbiomes: SLA was increased for plants with the AI microbiome and decreased for plants with the CI microbiome. This observed pattern in SLA is consistent with the directionality of ecotypic divergence. Moreover, SRL showed GxE in response to microbiomes that was also concordant with the direction of parental trait divergence: with xeric adapted hallii showing higher SRL in the presence of native microbiomes while mesic adapted filipes showed lower SRL.
We detected two groups of QTL interacting with native microbiomes. The first responded to native soil inocula regardless of their origin and the second interacted with native soil inoculum from only one site. For example, QTL for root number 8@33.1 was present only in the MI treatment and not detected in native treatments, suggesting that native microbiomes reduce genetic divergence for this trait (Fig. 4a). This could be explained by microbial taxa which flourished under MI treatment given that the niche competition was relaxed. QTL for SRL (4@19.1) and RMR (3@74.9) showed location specific GxE (Fig. 4b, c); plants with thefilipes allele in the CI treatment resulted in a higher trait value. This is opposite to the direction of SRL trait divergence in parental ecotypes and to their response to the CI treatment. Eight QTLs showed location specific GxE to the AI treatment (Fig. 4d-k). Our previous study conducted at the panicle emergence stage suggested that xeric hallii  employs a fast-acquisitive strategy for drought escape by acquiring nutrients rapidly and flowering quickly to enter dormancy before the onset of summer drought (Khasanova et al. 2019). This is consistent with current study conducted at the tillering stage where plants with the hallii allele in interaction with the AI microbiome produced more root and shoot biomass. This is accomplished by the increased production of tillers with roots to support them. Root systems of plants with these hallii QTL hotspots produced longer and thinner roots, putatively allowing increased foraging and resource acquisition. Four of these QTL present in interaction with AI clustered in the genomic “hotspot” on chromosome seven and three QTL clustered on chromosome nine. This common genetic control of ecotype differentiating traits involving above and below ground traits suggests that these factors interact with the AI microbiome in tandem, potentially contributing to ecotype divergence and local adaptation.
We also identified epistatically interacting QTL for root diameter present only in treatments with microbiomes from native locations (Fig. 5). When lines are homozygous for either hallii or filipesalleles at both of the interacting QTL, individuals produce smaller diameter roots. In contrast, individuals with mismatched genotypes (HH/FF) at the pair of interacting loci develop larger diameter roots. However, the observed epistatic QTL effects did not translate to decreases in aboveground biomass. This hybrid mismatched that have unusual phenotypes could represent either hybrid vigor or breakdown – but selection studies looking at how these phenotypes impact performance will be needed to further evaluate links between epistasis-microbiome interactions and root developmental responses.
A strength of our inoculation approach was prioritizing community effects, as opposed to the effect of single bacterial inoculants. However, given the exciting experimental advances of isolated bacterial strains in synthetic communities, targeted communities using locally adapted bacterial strains or combinatorics (Paredes 2018) could be used to address how the presence / absence of particular microbes impact plant phenotypes. In addition, it is intriguing to speculate on the genes and molecular mechanisms underlying the host x microbiome QTL detected in our study. It could be that these QTL harbor genes that interact only indirectly with the host microbiome, perhaps through abundance of soil nutrients as modified by microbes. For example, certain soil microbes in our inoculates may alter the abundance or availability of soil nutrients with subsequent consequences for genetic variation in root or shoot growth. It may be that QTL are related to root exudates or metabolites released that may recruit or amplify key beneficial microbes with subsequent impacts on available nutrients. There are many examples of soil resource abundances of key nutrients impacting plant growth, including genes that demonstrated plastic responses to nutrient availability (Brumbarova & Ivanov 2019). Alternatively, it may be that the genes within QTL intervals are involved in more direct interactions with microbes. For example, recent studies show that phytohormones, microRNAs and secreted peptides are known to recruit and foster the establishment of symbiotic arbuscular mycorrhizal fungi (Müller & Harrison 2019). Moreover, Finkel et al. 2020 recently discovered an important role of the bacterial genus Variovorax in attenuating the negative effects on root growth imposed by other bacterial isolates via modification of auxin concentration gradients in the rhizosphere. Plants also deploy extensive immune repertoires to ward off pathogens and control access of microbes to endophytic compartments (Chen et al. 2020) and some of our interactions may be related to ecotypic specific resistance or susceptibility. Our observation of an epistatic interaction is especially interesting as they may represent sensing and signaling pathways that are triggered or directed by microbes. In our case, epistatic interactions may also represent hybrid incompatibilities between ecotypes that are driven by the microbial community. Given the broad confidence intervals of our genome wide scans, we resist the temptation to consider and discuss specific candidate genes. Nevertheless, we emphasize that our approach leads to a direct pathway of fine-mapping and the identification and cloning of new genes involved in plant-microbiome interactions.
Our results show that microbiomes impact the influence of genetic architecture on plant traits in two ecotypes of Panicum hallii . These effects were broadly divided into two categories, effects dependent upon the presence of inoculated microbiomes in general and effects dependent upon microbiomes originating from a specific location of origin. This pattern sheds light on the role biotic factors may play in ecotype divergence and raises questions about how microbes impact the genetic architecture of plant quantitative traits, potentially leading to local adaptation and ultimately speciation. Further work in this system has several pathways forward. Broad characterization of microbial communities can be used to determine how genetic variation shapes microbial communities as well as individual microbes. For example, this approach may allow for differentiating the effects that are mediated by plant-fungal interactions vs. plant-bacterial interactions. Once more is known about specific members of the microbial community that play large roles in impacting plant traits, reductionist approaches including targeted inoculations of bacterial / fungal strains and reverse genetic approaches could be used to identify specific mechanisms underlying plant-microbe interactions.