Spatiotemporal Dynamics of Synthetic Microbial Consortia in Microfluidic Devices

Abstract

Synthetic microbial consortia consist of two or more engineered strains that grow together and share the same resources. When intercellular signaling pathways are included in the engineered strains, close proximity of the microbes can generate complex dynamic behaviors that are difficult to obtain using a single strain. However, when a consortium is not cultured in a well-mixed environment the constituent strains passively compete for space as they grow and divide, complicating cell-cell signaling. Here, we explore the temporal dynamics of the spatial distribution of consortia cocultured in microfluidic devices. To do this, we grew two different strains of Escherichia coli in microfluidic devices with cell-trapping regions (traps) of several different designs. We found that the size of the traps is a critical determinant of spatiotemporal dynamics. In small traps, cells can easily signal one another, but the relative proportion of each strain within the trap can fluctuate wildly. In large traps, the relative ratio of strains is stabilized, but intercellular signaling can be hindered by distances between cells. This presents a trade-off between the trap size and the effectiveness of intercellular signaling, which can be mitigated by increasing the initial seeding of cells in larger traps. We also built a mathematical model, which suggests that increasing the number of seed cells can also increase the strain ratio variability due to an increased number of strain interfaces in the trap. These results help elucidate the complex behaviors of synthetic microbial consortia in microfluidic traps and provide a means of analysis to help remedy the spatial heterogeneity inherent to different trap types.

Keywords: microbial consortia; microfluidics; spatiotemporal patterning.

Figure 1:. Designs of the microfluidic devices.

(a) Schematic of the “hallway” device. The blue region is the tall flow channel and the red regions are the shorter cell-trapping regions. (b) Cross section of the hallway trap device shown with cells in the trapping region. (cd) Same as (ab), but for the “open” device. See text for more details.

Figure 2:. Strain fraction time series.

(a) Images of the two control strains grown in a hallway trap over time. (b) The fraction of the yellow strain in all experiments performed in the hallway trap with the experiment from (a) in blue. Red dotted lines signify traps that lose one strain completely. (c) Images of the two control strains grown in the open trap over time. (d) The fraction of the yellow strain in all experiments performed in the open trap with the experiment from (c) in blue. Time 0 was defined as the first image at which the entire trap was filled with cells.

Figure 3:. Strain ratio comparisons.

(a) The range of the yellow strain ratio fluctuations over time is significantly different in hallway traps (Fig. 2b) than in open traps (Fig. 2d, p < 0:01, student’s t-test). (b) When pooling data from all 20 hallway traps to mimic the population size in an open trap, the pooled yellow strain fraction is more stable. The range of this data is shown in panel (a) as “pooled hallways”. (c) Segmenting the data from the open trap experiment shown in Fig. 2c into hallway-sized segments and plotting the yellow strain fraction within each segment over time results in higher variability, compared to the overall ratio (blue line). Red dotted lines are lines that go to 0 or 1. The range of yellow strain fluctuationss within each segment are shown in panel (a) as “segmented open”. These fluctuations do not differ from those in hallway traps (p > 0:01), but do differ from fluctuations in the entire open trap ( p < 0:01).

Figure 4:. Strain fraction stability for traps of varying size.

(a) Open traps with decreasing lengths were manufactured to determine how population strain ratio dynamics depend on the size of the population in each trap. (b) Filled circles show the empirical mean of the yellow strain fraction ranges ( max – min) for each trap size investigated; black circles: open-walled traps, red circle: hallway trap. The hallway trap and largest area open trap data points use the same data as in Fig. 2a,c. The other data points are averages of 6 experiments (Fig. S8). Starting with the smallest trap size, both the mean of the yellow strain fraction ranges and their standard deviations (error bars) generally decrease with increasing trap area for open traps. However, the largest open trap appears to level off, which we attribute to the larger average number of cells seeded for this data point (see Fig. 5 and Fig. S11, Supplementary Information).

Figure 5:. Emergent single-strain banding vs. number of cells seeded into trap.

(a) Images from five representative experiments with two strains grown in the largest open trap at the time the trap fills completely with cells. The number of cells seeded into each trap is given on the left, and the number of bands formed when the trap fills is given on the right. The greater the number of cells seeded into the trap, the more heterogeneous the population. (b) The number of bands (single-strain regions) increases approximately linearly with the number of seeded colonies. The points are from the same experiments shown in Fig. 2. Open dots correspond to the 5 examples shown in panel (a). The predicted range of bands (mean ± 3 s.d.) as a function of seeded cells is shaded in grey (see Supplementary Information).

Figure 6:. Sender-receiver consortium and signaling depth measurement.

(a) Schematic of the sender and receiver strain gene circuits. Genes shown are on plasmids. (b) In the top pair of images the larger number of thinner, interspersed bands results in sufficient intermingling of strains so that all receiver cells respond to QS molecules. The smaller number of thicker single-strain bands in the lower image pair leaves some receiver cells at a large distance from the sender strain. As a result, CFP signal decreases to background levels as the distance from a sender band increases. Top images of each pair are merged phase contrast, and yellow, red, and cyan fluorescent images from the microscope. Bottom images of each pair are only yellow and cyan fluorescent images merged to more clearly visualize CFP decay. (c) We selected a subset of the cells in the experimental image in (b) to determine the decay of CFP signal with distance from the sender strain. We averaged the signal in the red square over columns, and in a series of ten frames forward from the shown image, to generate the experimental CFP signal. Transition boundary from left (sender cells) to right (receiver cells) was automatically detected and set to the x-axis (distance) origin. (d) After threshold detection for receiver cells, response data was normalized and fit to an exponential decay with spatial decay parameter ξ. Experimental data shown as blue stars, fit curve in orange. Resulting fit: ξ =20 µm.