THOR's Hammer: the Antibiotic Koreenceine Drives Gene Expression in a Model Microbial Community

Abstract

Microbial interactions dictate the structure and function of microbiomes, but the complexity of natural communities can obscure the individual interactions. Model microbial communities constructed with genetically tractable strains known to interact in natural settings can untangle these networks and reveal underpinning mechanisms. Our model system, The Hitchhikers of the Rhizosphere (THOR), is composed of three species-Bacillus cereus, Flavobacterium johnsoniae, and Pseudomonas koreensis-that co-isolate from field-grown soybean roots. Comparative metatranscriptomics on THOR revealed global patterns of interspecies transcriptional regulation. When grown in pairs, each member of THOR exhibits unique signaling behavior. In the community setting, gene expression is dominated by pairwise interactions with Pseudomonas koreensis mediated either directly or indirectly by its production of the antibiotic koreenceine-the apparent "hammer" of THOR. In pairwise interactions, the koreenceine biosynthetic cluster is responsible for 85 and 22% of differentially regulated genes in F. johnsoniae and B. cereus, respectively. Although both deletion of the koreenceine locus and reduction of P. koreensis inoculum size increase F. johnsoniae populations, the transcriptional response of P. koreensis is only activated when it is a relative minority member at the beginning of coculture. The largest group of upregulated P. koreensis genes in response to F. johnsoniae are those without functional annotation, indicating that focusing on genes important for community interactions may offer a path toward functional assignments for unannotated genes. This study illustrates the power of comparative metatranscriptomics of microorganisms encountering increasing microbial complexity for understanding community signal integration, antibiotic responses, and interspecies communication. IMPORTANCE The diversity, ubiquity, and significance of microbial communities is clear. However, the predictable and reliable manipulation of microbiomes to impact human, environmental, and agricultural health remains a challenge. Effective remodeling of microbiomes will be enabled by understanding the interspecies interactions that govern community processes. The extreme complexity of most microbiomes has impeded characterization of the relevant interactions. Investigating the genetics and biochemistry of simplified, model microbiomes could unearth specific interactions and generate predictions about community-governing principles. Here, we use one such model community to quantify changes in gene expression of individual species as they encounter stimuli from one or more species, directly mapping combinatorial interspecies interactions. A surprising amount of gene expression is regulated by a single molecule, the antibiotic koreenceine, which appears to impact gene regulation across community networks.

Keywords: antibiotic signaling; microbial communities; rhizosphere microbes; transcriptional regulation.

FIG 1

THOR member with coculture fitness defect exhibits greatest transcriptional response. Each THOR member was inoculated at 1 × 10⁶ CFU/mL alone, in pairwise coculture and full community. Every 24 h, cultures were quantified by dilution plating on species-specific antibiotics to determine CFU/mL levels of F. johnsoniae (A), P. koreensis (B), and B. cereus (C) under the different coculture conditions over 6 days. Data are shown as biological and technical duplicates. Dotted horizontal lines indicate the limit of detection. (D) Schematic of comparisons to uncover differential expression changes in response to pairwise coculture (category I), the addition of the third THOR member (category II), and the full community (category III). A comparison of the B. cereus conditions is shown as an example. (E) Global pairwise expression changes (>2-fold) are shown as a percentage of the total number of genes within each species.

FIG 2

Single-species responses dominate THOR transcriptional regulation. (A) For each gene in F. johnsoniae with statistically significant category I regulation, the log₂-fold change in coculture with P. koreensis was plotted on the x axis, with the log₂-fold change in coculture with B. cereus plotted on the y axis. There was no fold change cutoff implemented; all statistically significant differentially expressed genes were included in the analysis. Points along the axes indicate genes that were only regulated in the presence of a single partner. F. johnsoniae differentially expressed many genes in response to either THOR member in both the same (blue quadrants) and opposite (white quadrants) regulation patterns. Pairwise coculture data were similarly displayed for B. cereus (B) and P. koreensis (C). The total number of genes regulated and their regulatory pattern breakdowns are shown in the table columns for F. johnsoniae (D), B. cereus (E), and P. koreensis (F).

FIG 3

Dual-species responses in F. johnsoniae are nonadditive with P. koreensis dominating opposite responses in the community setting. F = F. johnsoniae, B = B. cereus, and P = P. koreensis. DEG normalized transcript levels for the replicates from edgeR were averaged in each condition. Then, pairwise and community conditions were each divided by expression levels of the species alone. The data are displayed as populations on a log₁₀ scale. (A and B) Positively (A) and negatively (B) regulated DEGs in pairwise coculture >2-fold with B. cereus and P. koreensis (Fig. 2A, blue quadrants) were also regulated in the community. (C and D) F. johnsoniae expression of genes regulated in opposite directions by B. cereus and P. koreensis (Fig. 2A, white quadrants) in the community exhibited levels similar to coculture with P. koreensis. The measured BFP values differ from the predicted “Sum” (gray), which was calculated for each DEG to determine whether community expression level was a result of the additive effect of both pairwise interactions. All significance was determined by a paired, nonparametric Wilcoxon test with a Bonferroni correction. ***, P < 3.03 × 10⁻⁵.

FIG 4

Single-species responses to P. koreensis and B. cereus differ in the community setting. See the Fig. 3 legend for a DEG normalization description and Sum calculation. (A) F. johnsoniae upregulated genes >2-fold in response to P. koreensis (Fig. 2A and >1-log2 x axis) were also upregulated in the community. (B) B. cereus upregulated genes >2-fold in response to P. koreensis (Fig. 2B and >1-log2 x axis) were also upregulated in the community. (C) P. koreensis upregulated genes >2-fold in response to B. cereus (Fig. 2C and >1-log2 x axis) were also upregulated in the community. (D) F. johnsoniae upregulated genes >2-fold in response to B. cereus (Fig. 2A and >1-log2 y axis) demonstrated reduced regulation in the community. Representative genes (red) in each panel highlight overall trends, specifically, the expression levels in BFP did not match the dominant pairwise or the calculated pairwise sum. (A) FJOH_RS08820 clpP, (B) A9L49_RS16845 kinB, (C) BOW65_RS13585 putative oxidoreductase, (D) FJOH_RS04840 hypothetical protein. All significance was determined by a paired, nonparametric Wilcoxon test with a Bonferroni correction. *, P = 9.69 × 10⁻⁴; **, P = 1.54 × 10⁻⁴; ***, P < 2.26 × 10⁻¹⁰.

FIG 5

Loss of koreenceine reverses direction of gene regulation in pairwise coculture with P. koreensis. (A) F. johnsoniae DEGs in pairwise with P. koreensis, both single- and dual-species categories, aligned with increasing fold change (black) on the x axis. The comparable fold change (FC) for that gene in the wild-type FP pairwise compared to the FΔkec pairwise shown in red. (B) Linear regression and Spearman correlation between log₂ FC of F versus FP pairwise DEGs against FP versus FΔkec. (C) B. cereus genes differentially expressed in pairwise with P. koreensis aligned with increasing FC (black) on the x axis. The FC for that gene in the wild-type BP pairwise compared to the BΔkec pairwise shown in red. (D) Linear regression and Spearman correlation between log₂ FC of B versus BP pairwise DEGs against BP versus BΔkec.

FIG 6

Koreenceine elicits both conserved and species-specific responses. F. johnsoniae and B. cereus genomes were functionally annotated using eggNOG-mapper and the baseline prevalence of each COG category was quantified. Koreenceine-dependent genes sets were identified as >2-fold P. koreensis-regulated genes with opposing direction of regulation in the Δkec coculture (see Fig. 5). COG functional enrichment index was quantified as the percentage of COG category present in koreenceine-dependent DEGs divided by the percentage of that particular COG category in the whole genome. To be enriched, the functional annotation must have an index of >1 (dotted lines), meaning the frequency of that function in the data set is greater than found in the genome. Functional enrichment of both upregulated (yellow) and downregulated (blue) DEGs were analyzed for F. johnsoniae (A) and B. cereus (B). Only COG categories with at least one significant enrichment between the two species are shown. COG categories shaded gray are enriched in the same direction in both species. Significance was determined in Rstudio by Fisher exact test (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

FIG 7

Relative abundance of P. koreensis, not presence of koreenceine, drives changes in P. koreensis gene expression. (A) Heat map showing number of P. koreensis DEGs (>2-fold) in response to category I and category II comparisons for 1 × 10⁶ wild-type, 1 × 10⁶ Δkec mutant, or 2 × 10⁴ wild-type (P^LI) CFU/mL P. koreensis inoculations. (B) Cell density (CFU/mL) of indicated conditions at the time of RNA harvest with F. johnsoniae levels shown in orange and P. koreensis levels shown in green.

FIG 8

The major category of P. koreensis genes upregulated in response to F. johnsoniae have no functional annotation. (A) P. koreensis genome was functionally annotated using eggNOG and the baseline prevalence of each COG category was quantified. P. koreensis low inoculum genes sets were identified as >2-fold change in coculture with F. johnsoniae compared to P. koreensis alone. COG functional enrichment index was quantified as the percentage of COG category present in the FP^LI data set divided by the percentage of that particular COG category in the whole genome. To be enriched, the functional annotation must have an index greater than 1 (dotted lines), meaning the frequency of that function in the data set is greater than found in the genome. Upregulated genes shown in yellow and downregulated genes shown in blue. Only COG categories with significant enrichment are shown. Significance was determined in Rstudio by Fisher exact test (*, P < 0.05; **, P < 0.01; ***, P < 0.001). (B) Upregulated DEGs without COG annotation were searched for conserved domains (CD) and the frequency of CD hits, prophage genes, and remaining unannotated genes are shown.