Investigating the dynamics of microbial consortia in spatially structured environments

Abstract

The spatial organization of microbial communities arises from a complex interplay of biotic and abiotic interactions, and is a major determinant of ecosystem functions. Here we design a microfluidic platform to investigate how the spatial arrangement of microbes impacts gene expression and growth. We elucidate key biochemical parameters that dictate the mapping between spatial positioning and gene expression patterns. We show that distance can establish a low-pass filter to periodic inputs and can enhance the fidelity of information processing. Positive and negative feedback can play disparate roles in the synchronization and robustness of a genetic oscillator distributed between two strains to spatial separation. Quantification of growth and metabolite release in an amino-acid auxotroph community demonstrates that the interaction network and stability of the community are highly sensitive to temporal perturbations and spatial arrangements. In sum, our microfluidic platform can quantify spatiotemporal parameters influencing diffusion-mediated interactions in microbial consortia.

Fig. 1. Design of a microfluidic platform to investigate the role of spatiotemporal parameters in microbial consortia.

a Schematic of the microfluidic device. The inlets I₁₁, I₁₂ or I_21, I₂₂ connect to the outlets O₁ or O₂, respectively, and allow temporal control of the environmental conditions. Cells are initially seeded into growth chambers and the continuous flow of media through the main channels removes excess cells. Pairs of growth chambers are separated by a lattice of pillars, defined as the interaction channel, which allows diffusion of biomolecules and prevents cells from entering the interaction channel. The device has ten pairs of growth chambers for each separation distance. b Top: schematic of the genetic circuit in the E. coli sender and receiver strains. In the sender strain, the operon containing the synthetase LuxI and GFP is induced in response to arabinose. The synthetase LuxI produces the acyl-homoserine lactone (3-oxo-C6-HSL or AHL), which diffuses through the interaction channel into the receiver strain growth chamber. In the receiver strain, AHL binds to LuxR to form an activated LuxR–AHL complex, which in turn activates expression of RFP driven by a LuxR-regulated promoter. Bottom: overlaid representative fluorescence and phase-contrast microscope images of the sender and receiver strains in the device for each interaction channel length. The scale bar represents 25 μm. c GFP fluorescence in sender growth chambers as a function of time. The vertical line indicates the time at which arabinose was introduced. Shaded regions represent one standard deviation from the mean. The dashed line denotes the model fit. d RFP fluorescence over time in the receiver growth chambers. The vertical line indicates the time at which arabinose was introduced. Shaded regions represent one standard deviation from the mean. The dashed line denotes the model fit. Source data are provided as a Source Data file.

Fig. 2. Computational model of inter-strain communication in defined spatial environments.

a Model schematic depicting the physical and biological processes represented by the model equations (Supplementary Methods). b The model RFP_p steady-states decrease with the distance from the sender strain. The blue dashed line represents the maximum RFP_p steady-state concentration for a 2 μm interaction channel. The gray dashed line denotes 1% of the maximum steady-state RFP_p concentration. Data points (squares) represent experimental measurements and error bars denote 1 SD from the mean. c Heat map of the simulated RFP_p time delays for diffusible signal molecules spanning a broad range of diffusion rates (D₁) and binding affinities of LuxRAHL to the RFP_p promoter (K_RFP). The circle represents the estimated parameters based on experimental data. d RFP_p steady-state dose response as a function of AHL for different K_RFP and K_LuxR values. The dashed curve represents the dose response of the parameterized model based on the experimental data. Inset denotes representative simulations of RFP_p for different interaction channel lengths. The RFP_p steady states for each interaction channel length (colored marker styles) were computed for ara = 10. e Steady-state RFP_p dose response as a function of AHL for two different n_RFP values in the model. The dashed curve represents the dose response for the parameterized model based on experimental data. The RFP_p steady-states for each interaction channel length (colored marker styles) were computed for ara = 10. Inset: relationship between distance and steady-state RFP_p concentration for a range of n_RFP values. f Heatmap of distance sensitivity across a broad range of total LuxR concentrations (LuxR_tot) and binding affinities of the activated LuxR complex to the RFP_p promoter (K_RFP). Distance sensitivity is defined as the ratio of steady-state RFP_p concentration in a 25–250 μm condition. Source data are provided as a Source Data file.

Fig. 3. The frequency of a periodic input determines the effect of distance on information transmission.

a Model simulations of RFP_p in response to a periodic arabinose input (2 h period). Gray shaded regions denote the presence of arabinose. b GFP as a function of time in response to an oscillatory arabinose input (2 h period). Gray shaded regions denote the presence of 0.1% arabinose. Colored shaded regions represent 1 SD from the mean. c RFP as a function of time in response to a periodic arabinose input (2 h period). d Model simulations of RFP_p in response to a periodic arabinose input (1 h period). e GFP over time in response to an oscillatory arabinose input (1 h period). f RFP in response to an oscillating arabinose input (1 h period). g RFP amplitude as a function of the input period at different distances. Square, plus, and circular data points represent the amplitudes in the step response, 2 h and 1 h forced oscillator experiments, respectively. Error bars represent 1 SD from the mean amplitude. h Top: mean RFP power spectra in response to an input period of 2 h (red dashed line) for 395–1536 min. The gray shaded region denotes a bandwidth of ±10 min around the expected input frequency. Bottom: mean RFP power spectrum in response to an input period of 1 h (blue dashed line) for 500–1249 min. i Signal-to-noise (SNR) ratios computed using bootstrapped power spectra for GFP or RFP for each distance (Methods) in the 1 h forced oscillator experiment. The horizontal lines within boxes denote the median and upper and lower edges represent the upper and lower quartiles, respectively. Upper and lower whiskers represent the 95th and 5th confidence intervals, respectively. Horizontal lines with stars denote statistically significant differences (P < 0.05) based on a one-tailed bootstrapped hypothesis test (Methods). *P < 0.05, **P < 0.01, and ***P < 0.001 (Supplementary Fig. 16). Source data are provided as a Source Data file.

Fig. 4. Feedback loops dictate the role of distance on a distributed gene circuit oscillator.

a Network schematic of activator and repressor strains in the MISTiC device. The activator and repressor exhibit bidirectional communication and dual-feedback loops mediated by the signaling molecules C4-HSL and 3OH-C14-HSL. The activator and repressor exhibit positive and negative feedback loops, respectively. b Representative normalized CFP and YFP fluorescence intensities as a function of time in the 25 μm (top) and 250 μm (bottom) conditions. The red line indicates the time of induction with 1 mM IPTG. c Amplitude of peaks in the mean-subtracted CFP (top) and YFP (bottom) fluorescence intensities across distances. Error bars represent 1 SD from the mean. Horizontal lines with stars denote a statistically significant difference (P < 0.05) based on a two-sided t-test. *P < 0.05, **P < 0.01, ***P < 0.001 (Supplementary Fig. 16). d The number of peaks as a function of distance for activator and repressor strains. Horizontal lines denote a statistically significant difference (P < 0.05) based on a two-sided t-test. *P < 0.05, **P < 0.01, and ***P < 0.001 (Supplementary Fig. 16). e Maximum cross-correlation between paired CFP and YFP time-series measurements for each distance. The horizontal lines within boxes denote the median and upper and lower edges represent the upper and lower quartiles, respectively. Upper and lower whiskers represent the 95th and 5th confidence intervals, respectively. Horizontal lines with stars denote a statistically significant difference (P < 0.05) based on a two-sided t-test. *P < 0.05, **P < 0.01, and ***P < 0.001 (Supplementary Fig. 16). f The coefficient of variation of the inter-peak distances (phase drift) as a function of separation distance for activator and repressor strains. Lines represent linear regression fits to the data. Source data are provided as a Source Data file.

Fig. 5. Spatial and temporal modes of amino acid cross-feeding in a synthetic E. coli consortium.

a Schematic of the experimental design. ΔmetA and ΔpheA are labeled with IPTG-inducible fluorescent reporters. The strains are initially cultured in the presence IPTG. At the denoted time, IPTG is removed and the media condition is altered. The maximum growth rate is inferred based on the rate of decay of the fluorescence signal. Relationship between distance and population-level doubling times of (b) ΔmetA or (c) ΔpheA. The dashed and solid lines represent presence and absence of methionine (M) and phenylalanine (F), respectively. Diamonds represent the model fits. Inset in c: the time of maximum ΔpheA growth rate in media lacking M and F. Horizontal lines denote a statistically significant difference (P < 0.05) based on one-tailed bootstrap hypothesis testing (Methods). *P < 0.05, **P < 0.01, and ***P < 0.001 (Supplementary Fig. 17). d Relationship between time and the average single-cell growth rates of ΔmetA and ΔpheA in a mixed community (solid line) or monoculture (dashed line) in media lacking M and F. Inset: fraction of cells with non-zero growth rates as a function of time. e Spearman’s correlation coefficient as a function of time between the fraction of the growth chamber occupied by the partner strain and the growth rate for individual cells (inset). The X symbol denotes correlations corresponding to P > 0.05. f Relationship between distance and the minimum doubling time in the presence of M and absence of F. ΔpheA exhibited two growth phases following the media switch (GP1 and GP2). Diamonds indicate model fits to the maximum growth rates in GP1. g Concentration of M or F divided by absorbance at 600 nm (OD600) in ΔpheA or ΔmetA conditioned media. Stars indicate the limit of detection for each measurement. 1× Amino acid fraction refers to 0.2 mM M and 0.4 mM F, respectively. The horizontal lines denote a statistically significant difference (P < 0.05) based on a two-sided t-test. *P < 0.05, **P < 0.01, and ***P < 0.001 (Supplementary Fig. 17). h Inferred interaction networks based on population-level maximum growth rates in spatially separated MISTiC experiments. The size of the node represents the maximum growth rate in the 25 μm (D₂₅) condition and is computed as (10 ln D₂₅⁻¹)². The edge widths represent the interaction strength defined as $D_d=1-D_{25}{D}_{250}^{-1}$ where D₂₅₀ denotes the maximum growth rate in the 250 μm condition. Source data are provided as a Source Data file.