Design and Operation of the Microfluidic Platform
We first tested a simple microfluidic chip containing a straight 100 µm wide channel out of polydimethylsiloxane (PDMS) bonded on glass. Unsurprisingly, inhomogeneous bacterial adhesion and formation of large bacterial agglomerates along the chip sidewalls were observed (Figure S1, Additional File 1). In general, a high density of bacteria was found close to the access holes as well as to the sidewalls of the channels due to the lower flow shear at these positions, similar to what has been reported before (11). Furthermore, formation of so-called streamers was also observed after 3 hours of continuous inoculum injection into the microchannel (Figure S2, Additional File 1 and Video, Additional File 2).
The investigated Escherichia coli not only adhered to the glass surface but also to each other, leading to the formation of large clumps of cells in the area of low flow rates. This phenomenon has been reported previously (15,16). During our observations, the bacterial clumps were eventually washed away and thereby not only removed bacteria from the channel floor but also clogged the microchannel as has been reported by others (17). This observation can be explained by the narrowing of the flow channel due to biofilm formation resulting in a local increase of the flow speed, which in turn increases the shear force. Along with the change of the flow profile the chances of removal of adhered cells is significantly increased.
In order to overcome these limitations, we designed a tailor-made microfluidic µFC. The design includes three inlet channels that merge into a single chamber followed by an outlet channel (Figure 1A). This arrangement enabled control of the flow rate from three different liquid reservoirs and thereby allowed spatially separated flows of different media in the same µFC chamber because of the laminar flow regime (Figure S3, Additional File 1).
The flow-focusing principle was previously shown to be a reliable way to steer bacterial adhesion to the center of a flow chamber, effectively preventing bacterial adhesion on sidewalls (13,18). However, the design of the flow chamber restricted microscopic observation to low magnification. Indeed, the architecture of the flow chamber prevented the microscope objective from coming into close proximity with the specimen therefore only allowing the use of low magnification objectives with long focal distances. With our novel microfluidic platform, we quantified the influence of growth medium on the kinetics of bacterial adhesion and biofilm formation at single-cell resolution under flow, which would have not been possible with the so-far reported systems.
The newly designed µFC was mounted on the stage of an inverted microscope so that the adherent bacteria on the bottom glass surface of the µFC could be imaged by wide-field microscopy. A bacterial suspension was injected in the central channel while medium was flown from the two outer channels, thus restricting the bacteria to the center of the µFC (Figure 1B: adhesion phase). Since non-adhered bacteria cells were removed by the fluid motion, only adhered bacteria were visible. After a given period of time, the flow of inoculum in the central channel was stopped while perfusion of sterile medium from both outer channels was continued for up to 65 hours in order to allow biofilm growth (Figure 1B: proliferation phase). During this phase, images were recorded at several defined locations in the µFC. The recorded images were then processed by automated single-cell tracking analysis in order to measure surface coverage and the behavior of adherent cells.
Finally, the platform was used to investigate the antimicrobial effect of colistin by injecting the antibiotic into one of the inlet channels (Figure 1B: Antibiotic treatment).
Automated and Spatially Controlled Inoculation in the Microflow Cell
Utilizing the thus improved novel platform, we investigated the effects of medium composition on the initial attachment of E. coli and subsequent proliferation on the glass surface. Cells either cultured in rich medium tryptic soy broth (TSB) or modified minimal medium M9 (later simply referred as M9) were injected into the microfluidic chip. The cells that were in their exponential growth phase in TSB and M9 had an OD600nm of 0.30 and 0.26, respectively, measured just before inoculation. Bacteria on the glass surface were exposed to a maximal hydrodynamic flow shear rate of 412 s-1 with a Reynolds number of 4.7, which lies well within the laminar flow regime. M9 medium led to much faster cell adhesion and higher numbers of adhered bacteria than TSB medium within the same time period (Figure 2), despite the fact that the M9 suspension contained a lower number of bacteria than the TSB suspension. After 0.5 hours, cells suspended in M9 led to 5% surface coverage, whereas those in TSB barely reached 0.1%. Prolonged perfusion of 4 hours resulted in the median surface coverage of 8 ± 0.7 % for bacteria suspended in M9, and below 1% for those in TSB medium. The adhesion profile of bacteria in TSB was linear with a constant adhesion rate over the 4 hour inoculation phase, whereas the adhesion rate of bacteria in M9 quickly decreased to reach a plateau after about 2.5 hours of injection (Figure 2). This saturation phenomenon could be due to the steric influence of adherent bacteria on the flow profile near the surface by preventing flowing bacteria from interacting with the glass surface, and thus hindering further colonization of the surface. On the other hand, since E. coli has been reported to exhibit a negative zeta potential (19), electrostatic repulsion from the adherent bacteria could prevent as well the new bacteria from interacting with the surface.
It has been reported that nutrient limitation and thus medium composition can play an important role in bacterial cell membrane compositions and extracellular polymeric substances (EPS) production (20,21) where starving bacterial cells were shown to have altered levels of sugars and proteins on their surface. However, bacterial adhesion and biofilm formation are dependent on experimental conditions (e.g. media and surface characteristics) and studied strains (22,23). Moreover, high level of adhesion of a given organism in a given medium does not imply high biofilm formation (24). Nevertheless, one can hypothesize that the reduced amount of nitrogen present in M9 medium compared to TSB could lead to different cell membrane composition and EPS production level and be the reason behind the different bacterial adhesion behavior that we observed.
Furthermore, the two different media used are likely to result in different conditioning films on the glass surface, which in turn can result in different physicochemical affinity between bacteria and glass surfaces (25), offering another explanation for the differences in adhesion. In order to test the latter hypothesis, water contact angle measurements were performed on glass slide pretreated with TSB, M9 medium or phosphate buffered saline (PBS) (Figure S4, Additional File 1). TSB significantly increased the hydrophobicity of the glass surface compared to glass pre-treated with PBS and M9 (43 ± 11°, 20 ± 3°, and 17 ± 3°, respectively). This increase could be explained by the higher concentration of amino acid in TSB compared to that of M9 that can readily adsorb on the glass surface. Such higher surface hydrophobicity could result in lower bacterial adhesion in TSB due to less favorable physicochemical affinity between bacteria and glass.
Broadly speaking, these results highlight the importance of medium selection when conducting adhesion assays and when comparing results obtained with different media. Moreover, biofilms occurring on medical devices such as catheters are formed in a different nutrient environment than standard media used in the laboratory (26). It is thus necessary to employ media that mimic the actual in vivo environment in order to aim at predictive conditions for in vitro studies.
Automated Single-Cell Tracking to Monitor Surface Colonization by Bacteria
The independent flow control of the three channels and the laminar flow regime enabled us to use two different media (TSB and M9, respectively) synchronously in one experiment. Hence, the influence of two different media on biofilm formation could be investigated within a single µFC with otherwise identical conditions. Moreover, the flow-focusing design conveniently allowed us to use dedicated channels to inject an E. coli inoculum and to deliver sterile medium for extended periods of time to the adhered bacteria in the µFC, thus eliminating the undesirable effects of bacterial growth in the feeding channels such as clogging and medium alteration.
TSB was injected into the left outer channel and M9 into the right outer channel so that the two sides of the µFC were in contact with either of the media (Figure. S5, Additional File 1). Bacteria were suspended in PBS and injected through the central channel of the observation chamber for one hour. Thereafter, microscopic images of the growing biofilm were taken at defined locations in the observation chamber either in the M9 perfused area, the TSB perfused area or at the interface region of the two media (Figure 3A and Videos, Additional Files 3-5). Bacterial growth is clearly visible during bright field time-lapse microscopy, namely bacterial elongation and binary fission, demonstrating the advantage of the established microfluidics platform in allowing analysis at single cell resolution. The biofilm that has developed after 66 h of incubation is shown in Figure 3B. The two different perfusion regions can be easily discriminated. The biofilm formed in TSB became much more opaque than the one that was incubated in M9 medium, indicating that TSB promoted the formation of a thicker biofilm.
The surface coverage was used to quantify the biofilm formation as it reflected the increase of biomass on the glass surface. This analysis showed a fast colonization of the surface by bacteria grown in TSB (Figure 3C); after 40 h the observed region were almost completely covered. By contrast, bacteria grown in M9 medium colonized the surface more slowly and some regions remained uncovered even after 66 hours of incubation. At first sight, this data is in agreement with results obtained from the semi-static assay, where more biofilm was formed in TSB than in M9 (Figure S6, Additional File 1). In addition, planktonic E. coli were also found to grow more slowly and reached a lower cell density in M9 than in TSB (Figure S7, Additional File 1). This can be explained by the readily available nutrients such as amino acids in TSB medium. However, unlike growth in static assay where the amount of nutrient is limited by the volume of liquid, bacteria grown in the µFC are constantly supplied with fresh medium. One should therefore expect that the nutrient concentration cannot be the limiting factor for the growth of E. coli cells in M9 at least during early phase of biofilm formation.
To gain better understanding on the observed colonization profile, we performed single-cell tracking analysis during early stage of biofilm formation in order to monitor the behavior of cells on the surface. This enabled us to precisely follow the dynamics of the bacterial colonization on the surface by quantifying the bacterial proliferation and the bacteria being released in the flowing liquid (Figure 4A).
The evolution of cell numbers with time in both media can be seen in Figure 4B. More cells were originally present on the glass surface in contact with M9 than with TSB and their rate of growth was faster for the first 2.5 hours of incubation. It then stagnated, while the growth rate of bacteria in TSB gradually increased. Bacteria growing in TSB eventually overtook those growing in M9 after 12 hours of incubation.
Single-cell tracking enabled us to not only record the amount of bacteria on the surface at a given time but also to follow the generation on new bacteria (through cellular division) and the release of bacteria from the surface (Figure 4B). We observed that bacteria actively divided at the beginning of the incubation in both media. Surprisingly, bacteria proliferated faster in M9 than in TSB for more than 10 hours. However, more bacteria in M9 were released from the surface than in TSB. An equilibrium between generation and release was observed between 2.5 hours and around 10 hours explaining the lag in surface colonization in M9. This phenomenon was not observed for cells exposed to TBS: after 5 hours of incubation, progressively more bacteria were generated than left the surface resulting in an exponential colonization phase.
To summarize, the lag in population growth on the glass surface in M9 is not due to delay or slower bacterial growth but rather caused by the release of bacteria into the liquid. These differences of cell growth dynamics between the two media were visible on the pictures as bacteria growing in TSB formed clusters early on while bacteria growing in M9 stayed isolated on the surface for longer periods of time (Figure 4A).
These discoveries highlight the importance of a suitable tool like the established microfluidic platform here to allow detailed investigations at single-cell level, and the medium selection for the design of relevant biofilm models. A major advantage of our platform is the fact that any planktonic bacteria released from the growing biofilm are readily cleared from the µFC thanks to constant medium perfusion. By contrast, planktonic cells have an unavoidable influence in biofilm studies performed under static condition, in which careful rinsing steps have to be involved for the removal of planktonic cells before quantification or further processing and analysis of the biofilm. Moreover, planktonic bacteria can compete with sessile cells for nutrients during growth of the biofilm. By ensuring perfusion of sterile medium inside the µFC and clearing away planktonic cells, the platform overcomes this issue and permits the study of sessile bacteria alone in a controlled environment.
Furthermore, the homogenous biofilm growth along the entire length of the µFC (Figure 3B) indicated that the medium flow rate was sufficient to provide enough nutrient to the biofilm formed on the whole length of the µFC. The used flow rate of 400 µl/min resulted in a mean flow velocity of 12 mm/s in the µFC and thus a residence time of approximately 2 seconds for the medium within the µFC (dilution rate of 30 min-1). It is thus highly likely that sufficient nutrient in the flow is provided to obtain homogenous biofilm, even at the end of the µFC. In addition, it can be speculated that the flow of oxygen was sufficient for aerobic growth condition.
Visualization and Quantification of Antibiotic Activity
To demonstrate the versatility of the developed platform, E. coli biofilms were treated with colistin, an effective antibiotic against Gram-negative bacteria, and the killing action was followed in real-time by using a method described by Avalos Vizcarra et al. (27). Briefly, propidium iodide (PI) was added to the antibiotic solution at a non-cytotoxic concentration. PI can penetrate membrane-deficient cells and confer to these (dead) cells a bright red fluorescent signal. By following the emergence of the red fluorescent signal, colistin was found to rapidly kill bacteria and after 80 minutes of injection, only a small fraction of the cells were still alive (Figure 5 and Video, Additional File 6).
The platform established here can be utilized for the screening and assessment of novel antimicrobial agents against surface-associated bacteria and biofilms. Unlike traditional assays that rely on indirect measurements of cell viability either by optical density or colony forming unit counting, our assay allows direct visualization of single live/dead cells and reveals in-depth information about the performance of an antimicrobial agent. Moreover, the platform is fully compatible with confocal microscopy and high magnification with oil immersion objectives making the recording of spatial-resolved antibiotic activity on biofilm possible.
Stability and Versatility of the Platform
The platform presented here is extremely versatile and can be customized to suit a broad range of applications such as investigating the biofilm formation ability of different bacteria including mutants under defined and controlled conditions or the effect of a molecule of interest on adherent bacteria to name a few. Besides, different flow shear stresses can be easily applied by adjusting the flow rate and the dimensions of the microchannels and different media can be perfused synchronously.
Gas bubbles are a common burden in most microfluidic applications, as they can severely alter the flow characteristics. The bubbles can spontaneously form inside a microfluidic system by degassing out of liquid phase. The likelihood of this phenomenon increases with prolonged operation time and long-term incubation is therefore more prone to bubble formation than short-term experiments. By using a pressure driven flow and working with a positive pressure in the range of 500 mbar, the spontaneous occurrence of bubbles was avoided without the use of a bubble trapping system, thus keeping the complexity of the platform low. The pressure-driven flow also enabled the use of conventional glass bottles as medium reservoir hence allowing large volumes to be perfused, which would have been impossible with a traditional syringe pump (28). Moreover, using a pressure controller drastically reduced the risk of leakage in case of clogging of the system by preventing the rise of pressure that can occur with traditional pumps. By using flow sensors, pressure applied to the reservoir could be continuously adjusted in order to maintain constant flow rate throughout the whole experiment.
Additionally, the relatively large dimension of the microfluidic system (channel width ranging from 1 to 3 mm) and the absence of intricate channel geometry made it possible to manufacture the mold used for chip fabrication by computer numerical control (CNC) milling instead of photolithography, which greatly reduces the cost of production. Moreover, stereolithography is another cost-effective manufacturing method also suitable for mold fabrication and gaining popularity (29).
Here we employed our platform to investigate the influence of medium composition on bacterial adhesion and biofilm formation under constant flow rate. This study demonstrated that the platform is suitable for studying the dependence of environmental parameters (nutrition, shear stress) on the described phenomena as well as for testing novel antimicrobial agents. Firstly, their performance can be assessed on adherent bacteria in a more clinically relevant setting compared to traditional assays. Secondly, real-time monitoring of the activity of an antimicrobial agent at the single cell level will allow a deeper understanding of its mode of action.