Host preference and invasiveness of commensal bacteria in the Lotus and Arabidopsis root microbiota

Abstract

Roots of different plant species are colonized by bacterial communities, that are distinct even when hosts share the same habitat. It remains unclear to what extent the host actively selects these communities and whether commensals are adapted to a specific plant species. To address this question, we assembled a sequence-indexed bacterial culture collection from roots and nodules of Lotus japonicus that contains representatives of most species previously identified using metagenomics. We analysed taxonomically paired synthetic communities from L. japonicus and Arabidopsis thaliana in a multi-species gnotobiotic system and detected signatures of host preference among commensal bacteria in a community context, but not in mono-associations. Sequential inoculation experiments revealed priority effects during root microbiota assembly, where established communities are resilient to invasion by latecomers, and that host preference of commensal bacteria confers a competitive advantage in their cognate host. Our findings show that host preference in commensal bacteria from diverse taxonomic groups is associated with their invasiveness into standing root-associated communities.

Fig. 1. Lotus and Arabidopsis root-associated bacterial communities.

a, Alpha diversity analysis of soil- (n = 8), rhizosphere- (n = 13 for Gifu, n = 15 for Col-0) and root-associated bacterial communities (n = 13 for Gifu, n = 15 for Col-0) from Lj and At plants grown in natural soil (exp. A), assessed using the Shannon index. b, PCoA of Bray–Curtis dissimilarities of the same communities (n = 64). c,e, Rank abundance plots of OTUs found in the Lotus (c) and Arabidopsis (e) natural root communities. Community members captured in the corresponding culture collection are depicted as black while non-recovered OTUs are shown in white. The vertical axis on the right shows the accumulated relative abundance in natural communities of all recovered OTUs. d,f, Percentage of abundant OTUs (≥0.1% RA) associated with Lotus (d) or Arabidopsis (f) roots in nature (natural communities, NatComs) that are captured in the Lotus or the Arabidopsis IRLs (At- and Lj-IRL). Source data

Fig. 2. Whole-genome phylogeny of the Lotus and Arabidopsis core culture collections.

Maximum likelihood phylogeny, constructed from a concatenated alignment of 31 conserved, single-copy genes (AMPHORA) showing the taxonomic overlap of the Lj-SPHERE (n = 294, blue track) and At-SPHERE (n = 194, red track) core culture collections. Arrows in the outer rings indicate the strains selected for five mixed communities used in reconstitution experiments. Source data

Fig. 3. Reconstitution experiments recapitulate culture-independent patterns and show signatures of host preference by commensal communities.

a, Setup of the competition experiments. b–d, Constrained PCoA (CPCoA) of Bray–Curtis dissimilarities (constrained by all biological factors and conditioned by all technical variables) of soil, rhizosphere and root samples. b, Lj wild-type Gifu, nfr5 mutant and At wild-type Col-0 plants cocultivated with the mixed community LjAt-SC2 (exp. B, n = 155, variance explained 53.8%, P = 0.001). c, Gifu, Col-0, A. lyrata MN47 (Al) and L. corniculatus cocultivated with LjAt-SC3 (exp. F, n = 173, variance explained 65.1%, P = 0.001). d, Dead roots of Gifu and Col-0, and toothpick cocultivated with LjAt-SC3 (exp. J, n = 250, variance explained 43.9%, P = 0.001). e–g, Aggregated RA of the 16 Lj-derived and the 16 At-derived strains in the live (e,f) or dead roots (g) of Lotus and Arabidopsis plants inoculated with LjAt-SC2 (n = 66, e) or LjAt-SC3 (n = 72, f and n = 89, g). Source data

Fig. 4. SynCom-specific transcriptional outputs in Lotus and Arabidopsis roots.

a,b, Whole transcriptome-level principal component analysis of Arabidopsis (n = 12 biologically independent samples, a) and Lotus (n = 12, b) roots after coinoculation with host-specific SynComs (SC) (Lj- and At-SC3, exp. K). In the case of Lotus plants, a nodule isolate from the Lj-SPHERE collection was added to all treatments to prevent transcriptional outputs from being dominated by symbiosis or nitrogen starvation responses. c,d, Heatmaps showing scaled counts of genes arranged according to k-means clustering results (only differentially expressed genes shown) for Arabidopsis (c) and Lotus (d). e, Distribution of expression patterns for clusters of genes upregulated after coinoculation with native SynComs. f, Overlap in terms of homologues identified in the same clusters between the two host and a list of relevant transcription factors identified as potential key regulators of differential transcriptional responses. Source data

Fig. 5. Invasion and persistence of commensal bacteria.

a, Setup of the sequential inoculation experiment. Lj Gifu and At Col-0 plants were cocultivated with the mixed community LjAt-SC3, or individual SynComs Lj-SC3 and At-SC3, followed by inoculation with the contrasting SynCom (exp. L). b,c, Constrained PCoA (CPCoA) of Bray–Curtis dissimilarities (constrained by all biological factors and conditioned by all technical variables; n = 267; variance explained 14.7%, P = 0.001) of soil, rhizosphere and root samples (b), and PCoA of root samples only (n = 137, c). d,e, Aggregated RA of the 16 Lj-derived and the 16 At-derived strains in Lotus and Arabidopsis root (d) (n = 120) and rhizosphere (e) (n = 120) samples in the indicated treatments. Different letters above boxes indicate different significance groups according to a Kruscal–Wallis test, followed by a Dunn’s post hoc. Source data

Fig. 6. Host preference is linked to invasiveness.

a, Analysis of host preference of individual commensal strains across gnotobiotic experiments (n = 366). Each strain is represented by a dot, whose colour corresponds to its host preference index and whose size to its average relative abundance. A significant host preference (Mann–Whitney test, false-discovery-rate corrected) is depicted by a black circle around a dot. NS, not significant. b,c, Correlation between host preference and invasiveness index for each strain in root (n = 115) (b) and rhizosphere samples (n = 119) (c), respectively, obtained from the sequential inoculation experiment (exp. L). The colour of each point designates the host of origin of each strain and the size denotes its mean relative abundance (log₂ transformed). Each point is labelled with a numeric identifier that corresponds to the strains in a (LjAt-SC3). At, A. thaliana; Lj, L. japonicus; Al, A. lyrata; Lc, L. corniculatus. Source data

Extended Data Fig. 1. Flow chart of experimental procedures.

a, Overview of experiments and type of analyses performed on root, rhizosphere, and soil samples, or millifluidic droplets. b, Culture collection establishment of Lotus japonicus root and nodule bacterial isolates. Image created with BioRender.com.

Extended Data Fig. 2. Culture-independent diversity analysis of root-associated bacterial communities from Lotus and Arabidopsis.

Rank abundance plot of bacterial communities from Lotus or Arabidopsis roots, aggregated to the class level, including all families with a mean accumulated relative abundance of > 0.1% on either host. Statistical differences were assessed using a two-sided, non-parametric Mann–Whitney test. Asterisks represent significant values after multiple testing correction using the Benjamini–Hochberg method (P < 0.05).

Extended Data Fig. 3. Functional overlap of Lotus and Arabidopsis culture collection genomes.

PCoA of functional distances of genomes from bacterial isolates of the Lotus (Lj-SPHERE; n = 294) and the Arabidopsis (At-SPHERE; n = 194) culture collections.

Extended Data Fig. 4. Host-species specific bacterial root communities and commensal host preference is confirmed using independent mixed communities.

a and c, Constrained PCoA of Bray-Curtis dissimilarity (constrained by all biological factors and conditioned by all technical variables) of soil, rhizosphere, and root samples from L. japonicus wild-type Gifu, nfr5 mutant, and A. thaliana wild type Col-0 plants co-cultivated with the mixed community LjAt-SC1 (a, experiment C, n = 155, variance explained 53.8%, P = 0.001), from Gifu and Col-0 co-cultivated with LjAt-SC4 (b, experiment D, n = 87, variance explained 60%, P = 0.001), or from Gifu and Col-0 co-cultivated with LjAt-SC5 (b, experiment M, n = 100, variance explained 67%, P = 0.001). d, e and f, Aggregated RA of the 16 Lj-derived and the 16 At-derived strains in the roots of Lotus and Arabidopsis plants inoculated with LjAt-SC1 (d; n = 68), LjAt-SC4 (e; n = 34), or LjAt-SC5 (f; n = 40). n refers to biologically independent samples.

Extended Data Fig. 5. Host preference is retained after in silico removal of individual bacterial families.

Aggregated relative abundance of Lotus- (a, n = 20) and Arabidopsis-derived (b, n = 16) strains in roots from plants inoculated with the mixed community LjAt-SC3 (experiment L). Host preference was assessed using a Mann-Whitney non-parametric test after in silico removal of each family. The x-axis labels indicate each depleted family. n refers to biologically independent samples.

Extended Data Fig. 6. Bacterial abundance in mono-association with host plants.

Bacterial abundances of commensal bacteria (strain IDs indicated at x-axis) colonizing Col-0, Gifu, and nfr5 roots, assessed by counting of colony forming units (CFUs) after extraction from root tissue. Plants were grown for two weeks on agar plates in mono-association with the indicated Lj-SPHERE (a) and At-SPHERE (b) strains (exp. E). n = 6 (3 biologically independent samples × 2 technical replicates). Statistical differences were assessed using a two-sided, non-parametric Mann-Whitney test. Asterisks represent significant values after multiple testing correction using the Benjamini–Hochberg method (P < 0.05).

Extended Data Fig. 7. Plant performance in mono-associations.

Shoot fresh weight of Gifu (a) and Col-0 (b) plants grown for two weeks on agar plates in mono-association with the indicated Lj-SPHERE and At-SPHERE strains (exp. E). n = 30 (3 biologically independent samples × 10 technical replicates).

Extended Data Fig. 8. Sister species of L. japonicus and A. thaliana establish distinct bacterial root communities.

Constrained PCoA of Bray-Curtis dissimilarity (constrained by all biological factors and conditioned by all technical variables) of root samples from L. japonicus wild type Gifu and L. corniculatus (a; n = 87; variance explained 58.7%, P = 0.001), and or root samples from A. thaliana wild type Col-0 and A. lyrata MN47 (b; n = 86; variance explained 65%, P = 0.001), inoculated and grown with the mixed community LjAt-SC3, and of the corresponding rhizosphere and bulk soil communities (exp. F).

Extended Data Fig. 9. Tested plant immune receptors and signaling pathways do not affect host preference of commensals.

a, Constrained PCoA of Bray-Curtis dissimilarity (constrained by all biological factors and conditioned by all technical variables; n = 98; variance explained 24.6%, P = 0.001) of root samples from L. japonicus wild type Gifu, Ljfls2 mutant, A. thaliana wild type Col-0, Atfls2 mutant, and Atbbc mutant inoculated and grown with the mixed SynCom LjAt-SC1 (exp. G), and of the corresponding bacterial input communities. b, Aggregated relative abundance of the 16 Lj-derived and the 16 At-derived strains in the roots of Lotus and Arabidopsis plants. n = 21 for Gifu, n = 16 for Col-0, n = 20 for Ljfls2 and Atfls2, n = 18 for Atbbc. c, Constrained PCoA of Bray-Curtis dissimilarity (constrained by all biological factors and conditioned by all technical variables; n = 64; variance explained 42.2%, P = 0.001) of soil and root samples from Gifu, Col-0, and Atdeps mutant inoculated and grown with the mixed SynCom LjAt-SC3 (exp. H). d, Aggregated relative abundance of the 16 Lj-derived and the 16 At-derived strains in the roots of Lotus and Arabidopsis plants. n = 14 for Gifu, n = 20 for Col-0 and Atdeps. n refers to biologically independent samples.

Extended Data Fig. 10. Effect of secreted indole glucosinolates on host preference of commensals.

a, Constrained PCoA of Bray-Curtis dissimilarity (constrained by all biological factors and conditioned by all technical variables; n = 50; variance explained 47.2%, P = 0.001) of soil and root samples from L. japonicus wild type Gifu, A. thaliana wild type Col-0, and Arabidopsis cyp79b2 cyb79b3 mutant inoculated and grown with the mixed SynCom LjAt-SC3 (exp. H). b, Aggregated relative abundance of the 16 Lj-derived and the 16 At-derived strains in the roots of Lotus and Arabidopsis plants. n = 20 for Col-0, n = 14 for Gifu, n = 6 for Atcyp79b2b3. n refers to biologically independent samples.