MetaFlowTrain: a highly parallelized and modular fluidic system for studying exometabolite-mediated inter-organismal interactions

Abstract

Metabolic fluxes between cells, organisms, or communities drive ecosystem assembly and functioning and explain higher-level biological organization. Exometabolite-mediated inter-organismal interactions, however, remain poorly described due to technical challenges in measuring these interactions. Here, we present MetaFlowTrain, an easy-to-assemble, cheap, semi-high-throughput, and modular fluidic system in which multiple media can be flushed at adjustable flow rates into gnotobiotic microchambers accommodating diverse micro-organisms, ranging from bacteria to small eukaryotes. These microchambers can be used alone or connected in series to create microchamber trains within which metabolites, but not organisms, directionally travel between microchambers to modulate organismal growth. Using MetaFlowTrain, we uncover soil conditioning effects on synthetic community structure and plant growth, and reveal microbial antagonism mediated by exometabolite production. Our study highlights MetaFlowTrain as a versatile system for investigating plant-microbe-microbe metabolic interactions. We also discuss the system´s potential to discover metabolites that function as signaling molecules, drugs, or antimicrobials across various systems.

Fig. 1. MetaFlowTrain system overview.

a MetaFlowTrain supports diverse inputs, including artificial media (e.g., TSB, ARE) and natural extracts (e.g., soil extracts, host exudates). b The system operates up to 24 trains per peristaltic pump with adjustable flow rates. c A modular setup allows for up to six microchambers per train. d Outputs include microbial and exometabolite collection for OD measurements, qPCR, amplicon sequencing, metabolomics, and plant growth assays. A movie describing the system is also available (Supplementary Movie 1).

Fig. 2. Growth and exometabolite profiling of Arabidopsis root microbiota using MetaFlowTrain.

a Schematic of MetaFlowTrain setup: sterile peat extract as input, 51 parallel trains, one microchamber per train, and two outputs: microbial growth (OD₆₀₀) and exometabolite profiles (targeted/untargeted). b Neighbor-joining phylogenetic trees of BF SynComs members (n = 11 bacteria, n = 4 fungi). Boxplots display individual strain growth after 62 hours (OD₆₀₀). Boxplots are delimited by the first and third quartiles, with the central line representing the median value. Whiskers extend to show the range of the data within 1.5 * IQR from the quartiles. n = 3 independent microchambers per strain. Significant differences compared to inoculum OD as control was determined with two-sided Wilcoxon tests (* P < 0.05, ** P < 0.01, *** P < 0.001). c PCA plots of untargeted (left, n = 6370 features) and targeted (right, n = 35 metabolites) metabolic profiles for each microbial strain. n = 3 independent microchambers per strain, except Variovorax, n = 2 independent microchambers per strain, and Mock, n = 5 independent microchambers per strain. Statistical significance was determined by PERMANOVA-based comparison of metabolic profiles of individual strains (CanberraDistances~Strains, permutations = 999). Colors represent the different strains. Source data are provided as a Source Data file.

Fig. 3. Effects of soil conditioning on SynCom assembly, metabolic profiles, and plant phenotypes.

a Schematic of MetaFlowTrain setup with three different sterile extracts as inputs. 42 parallel trains, one microchamber per train, and three outputs: community profiles (MetaBarcoding), exometabolic profiles (targeted), and plant phenotyping. b PCA based on Bray–Curtis distances showing bacterial (left) and fungal (right) beta diversity. n = 6 independent microchambers, except for Control n = 7 independent microchambers. Statistical significance was determined by PERMANOVA-based comparison of community profiles (BrayDistances~Medium, permutations = 999). c Dot plot showing quantitative abundance (log reads) of each SynCom member. n = 6 independent microchambers per treatment, except for Control n = 7 independent microchambers per treatment. Letters indicate statistically significant differences between extracts (Kruskal–Wallis test (P < 0.05), followed by two-sided Dunn’s post hoc tests with Benjamini-Hochberg correction for multiple comparisons (P < 0.05). d PCA plots of targeted exometabolite profiles (n = 35 metabolites). n = 6 independent microchambers per treatment, except for Control n = 7 independent microchambers. Statistical significance was determined by PERMANOVA-based comparison of metabolic profiles (CanberraDistances~Medium/Time/BFSynCom, permutations = 999). e Boxplots showing Log₂ fold change (Log2FC) of exometabolites in the BF SynCom compared to a control. Boxplots are delimited by the first and third quartiles, with the central line representing the median value. Whiskers extend to show the range of the data within 1.5 * IQR from the quartiles, and all data points are displayed as individual points, including outliers. Only significant exometabolites are shown. n = 6 independent microchambers per treatment, except for Control n = 7 independent microchambers. Letters indicate statistically significant differences between extracts (Kruskal–Wallis test (P < 0.05), followed by two-sided Dunn’s post hoc tests with Benjamini-Hochberg correction for multiple comparisons (P < 0.05). f Bar chart showing germination rates (% germinated vs. non-germinated seeds) in exometabolite-derived media. n = 60 individual plants per treatment. Stars denote significance (two-sided Fisher’s Exact test (* P < 0.05, ns = non-significant). g Boxplots showing root lengths (mm) of germinated plants in exometabolite-derived media. Boxplots are delimited by the first and third quartiles, with the central line representing the median value. Whiskers extend to show the range of the data within 1.5 * IQR from the quartiles, and all data points are displayed as individual points, including outliers. n = number of independent germinated plants. Significance, compared to Mock, was determined with two-sided Wilcoxon tests (* P < 0.05, ns = non-significant). Colors represent the different peat extracts used as media. Source data are provided as a Source Data file.

Fig. 4. Impact of bacterial and fungal exometabolites on SynCom structure and metabolic profiles.

a Schematic of the MetaFlowTrain setup with sterile extract as input, 28 parallel trains, two microchambers per train, and two outputs: community profiles (MetaBarcoding) and exometabolites (targeted metabolomics. b Boxplots showing the abundances of microbial community members (log₁₀ reads) for different microchamber combinations. Boxplots are delimited by the first and third quartiles, with the central line representing the median value. Whiskers extend to show the range of the data within 1.5 * IQR from the quartiles, and all data points are displayed as individual points, including outliers. n = 7 independent microchamber trains * n = 11 bacteria (n = 77) or n = 4 fungi (n = 28) per treatment. Letters denote statistically significant differences between bacterial abundance or fungal abundance (Kruskal–Wallis test (P < 0.05), followed by two-sided Dunn’s post hoc tests with Benjamini-Hochberg correction for multiple comparisons (P < 0.05)). c Heatmap showing bacterial and fungal community composition across microchamber combinations. Red color intensity indicates log-transformed absolute abundances. Balloon plot shows the mean absolute abundance of microbial community members. n = 7 independent microchamber trains per treatment. Letters indicate statistically significant differences (Kruskal–Wallis test (P < 0.05), followed by two-sided Dunn’s post hoc tests with Benjamini-Hochberg correction for multiple comparisons (P < 0.05). d PCA plots displaying targeted metabolic profiles (n = 35 metabolites). n = 7 independent microchamber trains per treatment, except for B–F, n = 8 independent microchamber trains. Statistical significance was determined by PERMANOVA-based comparison of metabolic profiles (CanberraDistances~Time/SynComs, permutations = 999). The microchamber configuration is denoted as Microchamber1–Microchamber2 (e.g., B–F: Bacterial SynCom in microchamber 1 and Fungal SynCom in microchamber 2). Colors represent the different combinations of SynCom used in the microchambers. Source data are provided as a Source Data file.

Fig. 5. Impact of Pseudomonas brassicacearum exometabolites on SynCom growth.

a Schematic of the MetaFlowTrain setup: TSB 50% as input, 18 parallel trains, two microchambers per train, with one output: microbial load (qPCR). b Boxplots showing microbial load (Ct values) for different combinations of SynComs (Bacterial, Fungal, Mock) and R401 WT or mutant. Boxplots are delimited by the first and third quartiles, with the central line representing the median value. Whiskers extend to show the range of the data within 1.5 * IQR from the quartiles, and all data points are displayed as individual points, including outliers. n = 4 independent microchamber trains per treatment, except for R401–F and R401–B, n = 3 independent microchamber trains. Letters denote significant differences between microchambers (Kruskal–Wallis test (P < 0.05), followed by two-sided Dunn’s post hoc tests with Benjamini-Hochberg correction for multiple comparisons (P < 0.05)). The microchamber configuration is denoted as Microchamber1–Microchamber2 (e.g., Mock–B: Mock in microchamber 1 and Bacterial SynCom in microchamber 2). Colors represent the different Chambers based on their position. Source data are provided as a Source Data file.