tions
Besides differences in bacterial colonization, it is also possible that there are quantitative differences in bacterial relative abundance under different nutritional conditions. We used a Zero-Inflated Negative Binomial (ZINB) model to identify such differences (Lebeis et al., 2015) (section 3.7.9). We found 25 instances where there are significant relative abundances differences that can be attributed to an interaction between sample fraction and media (Table 3.1). These differences involved 11 strains and all media types, but strain CL21 and CL14 were the most sensitive to this parameter combination with seven and four instances, respectively, in which they were affected (Fig. 3.8). These two strains were relatively rare in bulk soil samples but are highly enriched in both neighboring soil and root samples across all media; but we were able to detect quantitative variation between the media types in specific fractions (Table 3.1). This is in stark contrast to what we observed in terms of presence/absence variation (section 3.1), where most of the differences between media types occurred among low prevalence isolates (Fig. 3.5e-f), with the notable exception of strain CL21 which showed variation in both prevalence (presence/absence; Fig. 3.5d) and relative abundance (Fig. 3.8; Table 3.1).
Figure 3.7: Presence/Absence variation of isolates between hosts. a Strain 8 is mostly excluded from roots, but it is present more often than expected in neighboring soils and roots of B. distachyon Bd21. b Strain 496 is present at a lower rate in the roots than in the soils, but it is absent from the roots of Col-0 and Oy-0. c Strain 29 is equally prevalent between neighboring soils and roots, but it is absent from both fractions of Sha-0 samples. d
Bacterial richness (number of different strains detected in each sample) of samples of different fractions and genotypes. The only significant difference (after controlling for batch effects) is an increased richness in Oy-0 samples of both fractions with respect to the other genotypes (q-value <0.05).
Taxon Variable Estimate q-value CL21 LowN.Root -6.408931 8.454272E-16 CL21 LowS.Root -7.067537 1.329227E-09 CL14 LowN.Root -5.774673 4.23655E-09 50 1/25 MS.Root -8.837079 3.348696E-08 CL14 LowN.N -5.289415 1.655905E-05 161 LowN.N -3.251344 3.726209E-05 CL41 LowP.Root -3.537116 0.000119682 CL14 LowS.Root -4.907614 0.000119682 161 Johnson.Root 3.43132 0.000119682 345A LowN.N -5.570872 0.0003216634 267 LowS.Root 6.860572 0.0003895009 CL21 1/4 MS.N 3.142551 0.001023724 CL14 LowS.N -4.760557 0.001350852 CL21 LowP.Root -3.057111 0.001716946 CL9 LowN.Root 2.934064 0.002077016 161 1/4 MS.N -2.409676 0.002077016 40 1/25 MS.Root -5.556394 0.002293188 279 LowN.Root -3.517204 0.002687698 CL21 Johnson.Root -3.417778 0.003287743 CL21 Johnson LowP.Root -3.296168 0.003745 CL21 LowN.N -2.961913 0.003745 345A LowS.N -6.04034 0.006375684 371 LowS.N -3.841529 0.006447609 345A LowN.Root -3.693198 0.007363134 40 LowN.Root -4.643262 0.009988183
Figure 3.8: Isolates sensitive to media and sample fraction. Isolate CL21 (a) and CL14 (b) are both more abundant in neighboring soils and roots than in bulk soil samples, and show quantitative variation in relative abundances in different media. Results of statistical tests are in Table 3.1
Together, our results suggest that changes the nutritional conditions in the soil, lead to changes in the root environment and microbial community, that allow sporadic access to strains that would be normally excluded in the root (section 3.1; Fig. 3.5). Our observations can be explained by the fact that nutritionally challenged plants down-regulate defense (Castrillo et al., 2017; Yamada et al., 2016), and the observation that hypo-immune plants can be colonized by bacteria that are normally not able to do so (Lebeis et al., 2015). On the other hand, isolates that were highly abundant in the root microbiome under full nutritional conditions, continued to successfully colonize plant roots in other conditions, but at variable relative abundances (section 3.3; Fig. 3.5). This might be a result of a combination of bacteria-bacteria competition caused by the changes in colonization of low-prevalence members, and physiological changes in the plant host.
3.4 Specific changes in the root microbiome under different host genotypes
We also found a number of significant relative abundance differences due to plant host genotype. Interestingly, we found 13 instances in which the difference was consistently present in the neighboring soil and root samples, suggesting that the host either alters the surrounding environment strongly enough for it to mirror the root, or that bacterial colonization of the root provides a competitive advantage to those isolates, by maintaining a population that can expand into the neighboring soil. Consistent with host phylogeny, the largest number of differences from Col-0 were found with B. distachyon Bd21, with 5 instance; intriguingly the same number of differences from Col-0 were found with A. thaliana accession Oy-0, while another accession (Cvi-0) and a related Brassicaceae(C. rubella) showed only two and one differences, respectively. Figure 3.9a-b shows examples of two strains that with consistent differential abundances between a pair of hosts in both neighboring soils and root samples.
We have shown that the effect of host genotype is stronger in the root than in the neighboring soil (Fig. 3.12; section 3.2). Thus, we asked whether there are bacterial relative abundance differences that are specific to the root. We found nine such instances involving eight strains. We found the relative effect of Oy-0 to be even stronger than when looking for
Taxon Variable Estimate q-value 371 Oy-0.Root 3.953721 4.214462E-06 109 Bd21.Root 1.367001 0.0001777078 27 Oy-0.Root 2.053737 0.0003340636 371 C. Rubella.Root 3.507121 0.0003340636 41 Oy-0.Root 3.340118 0.0003340636 CL52 Oy-0.Root 2.377527 0.001058683 CL69 C. Rubella.Root 3.508078 0.004553106 217 Oy-0.Root 1.936684 0.004871109 40 Oy-0.Root 2.167675 0.004871109
Table 3.2: Significant relative abundance differences for specific bacterial between hosts. differences that were consistent across fractions. Six of the nine genotype by fraction specific differences involved Oy-0, two C. rubella and one B. distachyon Bd21 (Table 3.2). Two examples are showin in Fig. 3.9c-d. Overall, we observed that Oy-0 is the most distinct of the fourA. thaliana accessions tested, while the variation of related Brassicaceae C. rubella, clearly falls within the A. thaliana variation, consistent with a previous report on natural soils (Schlaeppi et al., 2014). Moreover, the root microbiome of the highly divergent monocot grass B. distachyon also falls within this range of variation.
A common feature among the genotype specific differences in the root is that they involved low abundance strains (7/8 are sporadic colonizers; Fig. 3.6), were depleted in the root with respect to the neighboring soils, but were robustly enriched in the roots of a specific host (Fig. 3.9c-d). Genotype-dependent enrichment of low abundance isolates was also observed in at least some cases of isolates enriched in both fractions from a particular host (Fig. 3.9a-c). The predominance of enrichments observed in Oy-0 roots (Table 3.2) explain the increased bacterial richness that we measured in that accession (Fig. 3.7d). In summary, the enrichment of normally depleted or low abundance strains and the qualitative-like enrichment in specific hosts, points to a parallelism withgene for gene interactions which are responsible for disease resistance in plants (Flor, 1971), and suggest a simple underlying genetic architecture that could be amenable to genetic mapping.
Figure 3.9: Isolates enriched in specific hosts. a) Strain 8 is more abundant in the neighboring soil than in the plant root, and it is more abundant in the neighboring soil and root of B. distachyon Bd21 than in A. thaliana Col-0. b) Strain 363 is equally abundant between fractions, but it is more abundant than expected in the neighboring soil and root of Oy-0 than in Col-0. c) Strain 371 is less abundant in roots than in neighboring soil, but it is enriched in the roots of C. rubella and Oy-0. d) Strain 27 is less abundant in roots than in neighboring soil, but it is enriched in the roots of Oy-0.
Figure 3.10: CAP analysis of bacterial composition of root and neighboring soil
samples from 18 Arabidopsis accessions. Plant accession explainedv17.6% of the
variance in community composition among root samples (p-value = 0.01499; permutation), while it explained only 7.76% of the variance among neighboring soil samples (p-value = 0.8541).