The composition and diversity of the root-associated microbial community
We analysed all the experimental factors and their interactions on microbial community composition. Permutational multivariate analyses of variance (PERMANOVA) showed that rhizocompartment and soil type explained the largest percentage of data variation, with 16% and 18%, respectively (P < 0.01, Table 1). Therefore, microbiome structure in the rhizosphere and endosphere were analysed individually.
For rhizosphere microbiome composition, soil type accounted for most of the variation (about 80%, P < 0.01 based on PERMANOVA, Table 1), although the effects of root segment, genotype and their interactions were all significant (P < 0.01, Table 1). The biplot of non-metric multidimensional scaling (NMDS) based on Bray-Curtis distance showed that the samples clustered by soil type, and there was a high overlap between Soil-A and Soil-C (Fig. 1a). By contrast, the effects of root segment and host genotype were greater than soil type in determining the endosphere microbiome (11%-24% vs 6% PERMANOVA test, Table 1). In the NMDS biplot exploring root endophyte profiles, communities from the root apical and basal zone were clearly separated based on NMDS1, independently of different soil types (Fig. 1b, and Fig S4). Interestingly, endophyte composition in apical and basal root of the genotype EF overlapped considerably in the biplot (Fig. 1b). Consequently, we calculated the Bray-Curtis distance of microbiota composition between apical and basal root within each treatment and confirmed that this distance for the EF genotype was only half of the LF genotypes (LF1 and LF2) in the endosphere (P < 0.01, Table 1 and Fig. 1c). In addition, the variability of endophytic colonization across different soils was more stable in apical roots of all the genotypes, and in the EF basal root than other niches, as indicated by the shorter Bray-Curtis distance (P < 0.05, Table 1 and Fig. 1d).
At the phylum level, we analysed the taxonomic composition of the microbiome in each of the studied niches (Fig. 2). Proteobacteria and Actinobacteria were the dominant phyla in all the samples, accounting for 47% - 93% of the microbial consortia. A compensatory trend was observed in which the relative abundance of Proteobacteria increased, but Actinobacteria decreased from rhizosphere to endosphere, as well as from apical root to basal root, except for the EF genotype’s basal root tissue, which harboured an unaltered phylum composition relative to its apical zone, with less Proteobacteria but more Actinobacteria than the other two LF genotypes (FDR-adjusted P < 0.01).
Shannon index was used to estimate alpha diversity of the microbial community in each of the sampled niches. The magnitude of rhizocompartment effects on Shannon index was the greatest, indicated by the largest F value (P < 0.01, Table 1), and, on average, Shannon index was about 10 times higher in rhizosphere than endosphere samples (Fig. 3a and b). In the rhizosphere, the genotypic variation in the Shannon index was strongly dependent on the soil type and the longitudinal root fraction (Fig. 3a). While in the endosphere microbiome, the diversity of the EF genotype was greater than the two LF genotypes, and this observation was consistent in apical roots and basal roots in all soil types (Fig. 3b).