Host genotype and age shape the leaf and root microbiomes of a wild perennial plant

Abstract

Bacteria living on and in leaves and roots influence many aspects of plant health, so the extent of a plant's genetic control over its microbiota is of great interest to crop breeders and evolutionary biologists. Laboratory-based studies, because they poorly simulate true environmental heterogeneity, may misestimate or totally miss the influence of certain host genes on the microbiome. Here we report a large-scale field experiment to disentangle the effects of genotype, environment, age and year of harvest on bacterial communities associated with leaves and roots of Boechera stricta (Brassicaceae), a perennial wild mustard. Host genetic control of the microbiome is evident in leaves but not roots, and varies substantially among sites. Microbiome composition also shifts as plants age. Furthermore, a large proportion of leaf bacterial groups are shared with roots, suggesting inoculation from soil. Our results demonstrate how genotype-by-environment interactions contribute to the complexity of microbiome assembly in natural environments.

Figure 1. Summary of experimental design and analysis.

(a) Map of the study region in central Idaho, USA (map data from the R package ‘maps'70). The five genotypes used in this experiment were collected from the five B. stricta populations shown. We collected seeds from 8–10 individual B. stricta plants from each population, for a total of 48 accessions or genetic ‘lines'. For our analyses these lines were grouped back into five ‘genotypes' corresponding to the populations from which their ancestors were collected. The populations marked with triangles correspond to the ‘sites' of the three common gardens where the experiment took place. Scale bar, 50 km. (b) Schematic representation of common garden layout. Each garden contained six replicated, randomized blocks per planting cohort (2008 and 2009). Each block contained one replicate of each ‘line', for a total of 8–10 replicates per ‘genotype'. In both 2011 and 2012, one individual of each genotype was haphazardly chosen for destructive sampling in each block. (c) A temporally staggered planting/harvesting design disentangled the effects of plant age and year of observation. (d) Abbreviations and colour codes are shown for the five genotypes and three sites featured in this study. (e) The relative abundances of major phyla are shown for each leaf or root sample.

Figure 2. Habitats differ strongly in richness and composition of leaf- and root-associated bacterial communities.

Sample sizes are N=306 for leaves and N=310 for roots. (a) Mean Chao1 richness and effective Shannon diversity (e^{Shannon entropy}) differed among field sites for both roots and leaves (ANOVA, P<1.5e−11; detailed statistics are found in Table 1). The bottom and top edges of the boxes mark the 25th and 75th percentiles (that is, first and third quartiles). The horizontal line within the box denotes the median. Whiskers mark the range of the data excluding outliers that fell more than 1.5 times the interquartile range below the first quartile or above the third quartile (dots). (b) PCoA of weighted UniFrac distances reveals that field site is a major source of bacterial community variation in both leaves and roots (ANOVA, P<9e−7; detailed statistics in Table 2). Similar patterns result from PCoA of unweighted UniFrac distances, which only considers presence/absence of OTUs (Supplementary Fig. 4). (c) α-Diversity increased between 2011 and 2012 at most sites (ANOVA, P<0.05; detailed statistics in Table 1). Least-squares mean Chao1 richness is plotted for each year and each site; error bars represent 1 s.e.m. (d) Microbiome composition changed moderately between 2011 and 2012 (ANOVA, P=0.055 in leaves, P<0.01 in roots; detailed statistics in Table 2). Least-squares mean PCo1 and PCo2 are plotted with s.e.m. to show effects of year while controlling for other sources of variation using LMMs.

Figure 3. Individual bacterial taxa at multiple taxonomic levels are sensitive to several interacting factors.

Abundances of bacterial OTUs, families, orders, classes and phyla were individually modelled using NBMs. Sample sizes are N=306 for leaves and N=310 for roots. Significance was assessed using a Likelihood ratio test at P<0.05 after correction for multiple comparisons using Benjamini–Hochberg false discovery rate. ‘Geno'=genotype; ‘GxS'=genotype-by-site interactions; ‘AxS'=age-by-site interactions; ‘YxS'=year-by-site interactions. (a) Bar plots show the total relative abundance of bacterial taxa predicted by each source of variation (top horizontal axis) at multiple taxonomic levels (bottom horizontal axis), in variance-stabilized NBMs. (b) Effect sizes (fold changes between experimental groups for each term on the top horizontal axis) are plotted for all statistically significant pairwise contrasts predicted by variance-stabilized NBMs (Wald test, P<0.05). The bottom and top edges of the boxes mark the 25th and 75th percentiles (that is, first and third quartiles). The horizontal line within the box denotes the median. Whiskers mark the range of the data excluding outliers (green or grey dots) that fell more than 1.5 times the interquartile range below the first quartile or above the third quartile. To improve readability of the plot, the vertical axis was truncated at 2¹¹, concealing seven extreme outliers: all were root-associated OTUs with extreme changes in abundance due to age-by-site interactions.

Figure 4. Leaf- and root- associated bacterial communities change as plants age.

Detailed statistics for all tests are found in Tables 1 and 2. Sample sizes are N=306 for leaves and N=310 for roots. (a) Within-sample diversity declines with age in roots (ANOVA, F_2,57=6.08; P=0.0081) but not in leaves (F_2,64=1.11; P=0.67). Least-squares mean diversity estimates are plotted to show the effect of age after controlling for other sources of variation. Bars depict 1 s.e.m. (b) PCoA of weighted UniFrac distances between samples reveals that root bacterial community composition shifts over the lifetime of the plant. In roots, communities of experimental plants become more similar to those of endogenous plants, suggesting a role of succession after transplant (see Supplementary Note 2 for a detailed treatment of this hypothesis). Detailed statistics for the top three PCoA axes are found in Table 2. (c) Least-squares mean estimates of Chao1 richness are plotted for each age group in each site, illustrating how the distinct plant-associated bacterial communities in these habitats respond differently to host age. Leaves: F_4,63=1.48; P=0.36. Roots: F_4,56=7.85; P=8.6e−5. Error bars depict 1 s.e.m. (d) In leaves, the effect of plant age on the abundance of several phyla differs among sites (Likelihood ratio test, P<0.05 after Benjamini–Hochberg correction for multiple comparisons). Estimated mean abundances from NBMs are plotted for each age group at each site. ‘Actino.'=Actinobacteria; ‘Armatim.'=Armatimonadetes; ‘Verruco.'=Verrucomicrobia.

Figure 5. Host genotype shapes α and β diversities of the leaf microbiome but has weaker effects on root communities.

The results shown here represent only the constitutive or environment-independent component of genetic variation, that is, genotype differences averaged across all sites. Sample sizes are N=306 for leaves and N=310 for roots. (a) Least-squares mean estimates of within-sample diversity are plotted for each plant genotype, after controlling for other sources of variation using linear mixed effects models. Leaves: F_4,265=4.68, P=0.0023; Roots: F_4,261=0.344, P=1. Bars depict 1 s.e.m. (b) Each point is an estimated differential abundance of one OTU between a pair of plant genotypes, plotted as a fold change (note log scale on vertical axis). Leaf OTUs are shown in green, root OTUs in grey. Differential abundance estimates and statistical significance were both generated using variance-stabilized NBMs that also controlled for site, year of observation, plant age, genotype-by-site interaction, age-by-site interaction and year-by-site interaction. Only statistically significant contrasts are shown (Wald test, P<0.05 after Benjamini–Hochberg correction for multiple comparisons). (c) Bacterial community composition separates by host plant genotype in weighted UniFrac ordination for leaves, but not roots. Least-squares mean PCoA coordinates are plotted to highlight the influence of host genotype after controlling for other sources of variation using LMMs. Detailed statistics for each PCoA axis are found in Table 2. Bars depict one standard error of the mean.

Figure 6. Host genetic control of the bacterial microbiome differs among habitats.

Sample sizes are N=306 for leaves and N=310 for roots. (a) Estimates of broad-sense heritability (H²) of individual OTUs are plotted for leaves (top) and roots (bottom). The bottom and top edges of the boxes mark the 25th and 75th percentiles (that is, first and third quartiles). The horizontal line within the box denotes the median. Whiskers mark the range of the data excluding outliers that fell more than 1.5 times the interquartile range below the first quartile or above the third quartile (dots). (b) Between-sample diversity of the leaf microbiome is plotted as least-squares mean PCo1 of the weighted UniFrac distance for each plant genotype in each site, showing the genotype-by-site interaction after controlling for other sources of variation in a LMMs, including the constitutive effect of plant genotype and average site effects; F_8,257=4.53, P=0.00011. Bars depict one standard error of the mean. (c) Least-squares mean leaf Shannon diversity is plotted for each genotype and each site, revealing site-dependent differences in the relative leaf community richness among host genotypes; F_8,261=2.33, P=0.04. Bars show 1 s.e.m. (d) The total relative abundance of OTUs that were predicted by site-specific genotype effects in NBMs is shown for leaves and roots in each site (Wald test, P<0.05 after Benjamini–Hochberg correction for multiple comparisons).

Figure 7. Multiple non-mutually exclusive hypothetical mechanisms could cause genotype–environment interactions for microbiome composition.

Each panel shows the ambient microbial community for two environments (abstracted as coloured circles); the expressed phenotypes of two different host genotypes (abstracted as funnels that are labelled with the plant genotype); and the resulting microbiome of the genetically distinct plants in each environment. In a the genetic variant is expressed in both environments (that is, there is no phenotypic plasticity) but it affects only certain microbes, which are absent from Environment 2. As a result, genetic diversity for the trait is expressed in both environments, but genetic diversity for the microbiome is only observed in Environment 1. In b, environmental variation alters the phenotype expressed by plant genotype B, but not plant genotype A (that is, there is genetic variation for phenotypic plasticity, or a genotype–environment interaction for the functional trait). As a result, host genotype affects microbiome composition only in Environment 2, even though the ambient communities are identical in both environments.

Figure 8. A large fraction of leaf OTUs were also observed within roots of the same plant.

(a) Each bar shows the total fraction of leaf (green) and root (grey) OTUs (binned at 99% sequence similarity) belonging to the ‘shared set' of OTUs that were observed in both organs within one plant (N=237 plants). (b) Root OTUs from across the relative abundance spectrum contributed to the bacterial leaf microbiome. All root OTUs are plotted in increasing order of their relative abundance in roots (black dots). For each OTU that was also observed in leaves, its relative abundance in the leaf microbiome is plotted above or below it (green dots).