5.3 Channel-type chips
5.3.3 In vivo studies
For the study of bacterial communication in these chips, we used two genetic circuits that were engineered in Chapter 4 and thoroughly characterised via a range of tech- niques in discrete and continuous time series experiments. These corresponded to the optimised sender and receiver devices presented in Figure 4.27.
5.3.3.1 Cell size control
Our study relied on trapping cells in specific microfluidics chambers. However, with the design features of channel-type microfluidics chips, observation channels/chambers were inaccessible to E. coli under normal size conditions. The standard protocol to load E. coli into microfluidics chips involves growing bacteria from a fresh colony, refreshing
5.3. Channel-type chips 145
FIGURE5.11: Fluorescence images of channel-type microfluidics features
immersed in a red dye. Features visible over a black background corre- spond to cellular traps and diffusion channels. (A), (B), (C) and (D) re- spectively represent bottom traps widths of 10µm, 20µm, 30µm and 40µm. Top cellular channels have a width of 1µm and diffusion channels 0.7µm
on average.
saturated cultures and providing exponential phase cells to the input system. Given the facilities provided by the Centre for Bacterial Cell Biology, we may have been able to load cells via an optical tweezer[182], but this method would have implied to move the microfluidics apparatus after cellular loading to a fluorescence microscope for live imaging. Since E. coli cells undergo physiological changes as they grow in different phases, we rather tested inputting cells extracted from higher cellular density condi- tions - thus smaller cells - into microfluidics chips[126,136]. Although this step did not help the efficient loading of cells into biochips, we overcame this issue by exploring poorer growth media alternatives prior to live cell imaging. We tested a range of min- imal media, derivative of M9 salts, and managed to reduce E. coli size by an average of 3-4x fold compared to cultures that were grown in richer media (LB-based). This poor growth medium consisted of 1x M9 supplemented with 0.02% casamino acids and 0.5% glycerol. By growing cells to exponential phase in this poorer medium, we significantly improved cellular loading in microfluidics chips, from ≤ 1 cell per trap to several cells per observation feature (cf. Figure 5.12A).
Microfeatures aimed at maintaining cells in observation chambers were initially designed to accommodate standard E. coli cells grown in rich medium. Therefore, we could not keep mini-cells obtained from starvation in poor medium in bacterial traps. These were too tiny and could move in small compartments supposed to hinder cells. We thus needed to swap medium after loading of smaller cells for a richer, standard
FIGURE 5.12: E. coli cells in channel diffusion microfluidics chips. (A) and (B) show the same location within the biochip at two different time points: after cell starvation (small cells, A) and after reverted physiological change in LB (bigger cells, B). (C) displays the fluorescence levels observed in two bacterial communities respectively sending (bottom) and sensing
5.3. Channel-type chips 147 LB growth medium supplemented with eventual inducer. Cells started recovering nor- mal size within half an hour after providing richer medium. In general, we allowed a 2-3h recovery time for cells to grow back to normal size, and afterwards performed a final medium change to induce cultures. The physiological change from small cells (just loaded) to recovering cells after 1h culture at 37◦C is shown in Figure 5.12A and
B. After the obtention of normal size E. coli that would fill bacterial traps, cultures were induced and followed by fluorescence over time. Figure 5.12C displays a fluo- rescence snapshot of two sender and receiver communicating colonies in channel-type microfluidics chips. Here, sender devices were located at the bottom of the image in the large cellular compartment, while sensor cells were growing above these. Over ≥ 10 experiments, we consistantly started to detect AHL sensing by the activation of sfGFP signal in receiver colonies after 2h of sender cultures induction. Overall, we managed to successfully load bioengineered cells into channel-type biochips, and we observed significant signal for intercellular communication between microcolonies of bacteria separated by 50µm within 120 minutes.
5.3.3.2 Limiting design factors
We have previously demonstrated how channel-type microfluidics chips may be engi- neered and adapted for in vivo studies of E. coli. However, after fabricating channel- type PDMS chips and visualising their features, it became quickly obvious that trap- ping cells in 1µm wide aperture channels would become challenging (cf. Figure 5.12). Although the process of cellular loading could be optimised for our purpose as de- tailed in the previous paragraph, we ran into major challenges when reaching high cellular densities in bacterial traps. The design of 0.7µm diffusion channels appeared to be fine to block the passage of cells over the first couple of hours, but subsequent cel- lular divisions ended up squeezing cells within diffusion planes. Individual bacterial colonies would then join and contaminate each other, rendering the screening of bac- terial behaviour impossible in extended time-course assays. Figure 5.13A and B show the same bacterial trap at 6h of interval during which cells gradually overfilled the ob- servation chamber. At the top of the circular trap, we observe in panel A a single cell starting to enter the diffusion channel getting progressively filled over a few hours, as shown in panel B. Besides the stacking of cells within observation chambers rendering fluorescence measurements less precise (e.g. cells growing in 3D), we also noticed a big deterioration in E. coli physiology. As shown in Figure 5.13C, cells under observa- tion eventually acquired mutations and developed significant phenotypical changes. The cell shown here and other occurences displayed atypical lengths of ≥ 50µm. The
FIGURE 5.13: E. coli cells in channel diffusion microfluidics chips after
extended culture. (A) and (B) show the same location within the biochip at two different time points: at cell induction (A) and after 6h of culture (B). (C) shows a bacterial trap where a mutant cell is hanging in deeper
flow channels after an overnight culture.
development of such abnormalities could be due to the trigger of a SOS response that ends up affecting the cellular division machinery in subsequent biological pathways. It decreases cellular fitness and is proof that cellular observation chambers were not op- timal for bacterial growth. This could be reasonably well explained by an aperture for input/output of fresh medium/cells too small to accommodate larger colonies. Over- all, channel-type biochips were poorly adapted for bacterial growth. This stresses out the need for a correct balance between ideal design and compromises for bacterial fit- ness to achieve a certain function when using microfluidics chips. Therefore, based on the knowledge gained from experimenting with these biochips, we engineered other types of microfluidics chips aimed at screening for intercellular communication, and present these in the next section.