Long-term monitoring of bacteria undergoing programmed population control in a microchemostat

(Jana Verbančič)

F. K. Balagaddé, L. You, C. L. Hansen, F. H. Arnold, and S. R. Quake, “Long-term monitoring of bacteria undergoing programmed population control in a microchemostat.”, Science, vol. 309, pp. 137–140, 2005.

In this scientific paper Frederick K. Balagaddé and his colleagues implemented a microfluidic bioreactor or »microchemostat« that enables long-term culture and monitoring of extremely small populations of bacteria. They used the bioreactor, that operates at a working volume of 16 nL, to observe two strains of E. coli cells which contained a synthetic population control circuit that regulates cell density through a mechanism based on quorum sensing. The microchemostat enabled monitoring of sustained oscillations in cell density over hundreds of hours [1].

Introduction

Quorum sensing

Quorum sensing (QS) is a communication system that regulates the expression of genes and is based on population density. It is used by most bacteria and also other species. Because of fluctuations in cell-population density they produce and release chemical signal molecules called autoinducers to coordinate the behaviour in the group [2]. This types of molecules are different in every microorganism species and are therefore specific. In general, there are some common classes of molecules, used for signaling. Gram-negative bacteria use acylated homoserine lactones (AHL), gram-positive bacteria use processed oligopeptides, and autoinducer-2 (AI-2) is used for interspecies communications. At low cell densities this substances are in low concentrations, while at high cell densities this substances accumulate to the critical concentration required for activation of several genes [3]. Some of the best-known examples of quorum sensing come from studies of bacteria. They use quorum sensing to coordinate certain behaviours such as antibiotic resistance/production, biofilm formation, virulence, symbiosis, competence, conjugation, motility and sporulation. But quorum sensing was discovered and first described over 35 years ago in two luminous marine bacterial species, Vibrio fischeri and Vibrio harveyi, that use QS for light emission. In both species the enzymes responsible for light production are encoded by the luciferase structural operon luxCDABE and light emission was determined to occur only at high cell-population density in response to the accumulation of secreted autoinducer signaling molecules. Individual species of bacteria use several signals to communicate and they can be used to differentiate between species in consortia. This communication both within and between species is very important for bacterial survival and interaction in natural habitats [2]. Most quorum-sensing-controlled processes are unproductive when undertaken by an individual bacterium acting alone but become beneficial when carried out simultaneously by a large number of cells. With this quality quorum sensing enables bacteria to act as multicellular organisms [4].

Gram-negative bacteria: the LuxI/LuxR system

For better understanding of the circuit used in the article it is very important to know, how the LuxI/LuxR-type quorum sensing of the bioluminiscent bacterium Vibrio fischeri works. This system is also considered as the paradigm for quorum sensing in most gram-negative bacteria [4]. Vibrio fischeri inhabits a special light organ in the squid's (Euprymna scolopes) mantle. The bacteria are in symbiosis with the squid and can grow to large populations, reaching 10˄11 cells per ml, since there is a solution, full of sugars, amino acids and other nutrients in the light organ. In turn, the squid gets protected from predators, due to the emitted bioluminiscence. The light hides it's silhouette when viewed from below by matching the amount of light hitting the top of the mantle. This phenomenon is called counter-illumination [2]. V. fischeri also lives in symbiotic association with other eukaryotic hosts to serve various purposes, but the system by which the bioluminiscence is achieved is always the same. The expression of the luciferase operon (luxICDABE), required for the light production, is controlled by the proteins LuxI and LuxR. LuxI produces the signalling molecule 3OC6-homoserine lactone (acyl-homoserine lactone; AHL) and is therefore the autoinducer synthase. LuxR is the cytoplasmic autoinducer receptor and simultaneously a DNA-binding transcriptional activator [4]. The diffusible AHL accumulates in the surrounding environment during growth [3]. It increases in concentration with increasing cell density and when the signal reaches a critical threshold concentration in the cytoplasm (about 10 nM), it is bound by LuxR . This LuxR-AHL complex then binds to the luciferase operon and activates transcription. Notably, the complex also induces expression of luxI, which is encoded in the operon and therefore more AHL is produced with this positive feedback loop. Because of the abundancy of signal the entire population of cells switches into the quorum-sensing mode and produces light [4]. The LuxR proteins possess specific acyl-binding pockets that allow each LuxR to bind and be activated only by its corresponding signal. Therefore, in a mixed-species environment, where multiple AHL signals are present, each species can distinguish and respond only to higher concentrations of its own signalling molecule. It is known that bacteria have more than one LuxIR quorum-sensing system, therefore they can respond to several signals present in the environment [4].

Biofilm formation

For a mechanic device that measures the density of cells it is critical to prevent biofilm formation. In this section I will shortly explain why. Biofilms are communities of microorganisms attached to a surface. The adherent cells are embedded within a self-produced matrix of extracellular polymeric substance (EPS), which is composed of extracellular DNA, proteins and polysaccharides. The problem is, when cells switch from planctonic (free-swimming) to the biofilm mode of growth (a surface-attached community), they undergo a phenotypic shift in behaviour where a lot of genes get differetially regulated. It comes to a change in genes and regulatory circuits important for initial cell-surface interactions, biofilm maturation and the return of biofilm microorganisms to a planctonic mode of growth. These changes occur in response to a variety of environmental signals [5]. Microbial biofilms interfere with continuous bioreactor operation, because the changed bacteria shed their progeny into bulk culture and create mixed cultures. The prevention of biofim formation is therefore essential, as it could come to changes and mutations in the synthetic circuit that Balagaddé and his coworkers included in E. coli cells and subsequently to the disruption of the cell density oscillation that they wanted to achieve with the circuit.

Chemostat

The term »chemostat« (from »Chemical environment is static.«) was first described by Novick and Szilard in the 1950 paper »Description of the Chemostat.« As described from Paul Waltman in »The theory of the Chemostat«, »The chemostat is the best laboratory idealization of nature for population studies. It is a dynamic system with continuous material imputs and outputs, thus modeling the open system character and temporal continuity of nature. The input and removal of nutrient analogs mimics the continuous turnover of nutrients in nature. The washout of organisms is equivalent to non-age specific death, predation or emigration which always occurs in nature.« Thus, a chemostat is a bioreactor, where fresh medium with nutrients is continuously added and »old« culture liquid is continuously removed from the reactor [6]. This way, the microorganisms can be grown in a steady state, where growth occurs at a constant rate and the culture parameters like pH, culture volume, nutrient and product concentrations, temperature, cell density and dissolved oxygen concentration also remain constant. The conditions can be easily controlled. The steady state of growth is maintained with the limiting factor – when there is a low amount of cells in the bioreactor, the cells can grow at high grow rates, but when there is a high amount of cells, the growth is limited by the limiting nutrient, which is essential for growth. At a high cell concentration, the amount of cells, that are removed from the bioreactor cannot be replenished by growth as the addition of the limiting nutrient is insufficient and this results in a steady state [7]. A very important factor of the chemostat is that it allows control of the growth rate of bacteria and that is achieved by controlling the dilution or washout rate (D). At steady state the specific growth rate of bacteria equals the dilution rate, which is defined as the volumetric flow rate of nutrient supplied to the reactor divided by the volume of the culture. Initially, the rate of cell production increases as dilution rate increases. When Dmax is reached, the rate of cell production is at a maximum. This is the point where cells will not grow any faster. D = μ (dilution rate = specific growth rate) is also established at this point, where the steady-state equilibrium is reached. The concentration of cells starts to decrease once the dilution rate exceeds the Dmax. The cell concentration will continue to decrease until it reaches a point where all cells are washed out. At this stage, there will be a steep increase in substrate concentration because fewer and fewer cells are present to consume the substrate [8]. Some problems that may occur are foaming (thus the volume is not constant), contamination of the sterile medium, fragile or vulnerable cells can be ruptured during mixing and aeration, cells may form biofilms by adhering to surfaces or mixing may not be uniform. The continuous bioreactor has a requirement for large quantities of growth media and reagents. This and other challenges have pushed the research toward miniaturization and chip-based control [1].

Long-term monitoring of bacteria undergoing programmed population control in a microchemostat

The microchemostat

While chemostats exist in volumes from a few mililitres to a few litres, microchemostats work in microliter or even in nanoliter volumes and have to face other complications in functioning. Balagaddé and his colleagues created a very small chip-based bioreactor that uses microfluidic plumbing networks to actively prevent biofilm formation. The bioreactor is composed of six separate 16 nL (nanoliter) reactors, called »microchemostats«, that enable semicontinuous and planktonic growth of bacteria with no observable wall growth which points to an effective biofilm prevention. The cultures can be monitored in the microchemostats by optical microscopy with single-cell resolution and such measurements of cell density and morphology are automated, in real-time and noninvasive [1]. Description of the appearance and functioning of the microchemostat used in the main article experiments: each of the six reactors on the chip is made of a growth chamber with an integrated peristaltic pump, supply channels, input/output ports and a series of micromechanical valves to add fresh medium, remove waste (the so called wash out) and recover cells. The growth loop is divided into 16 separate segments, which can be individually maintained (Figure 1A and 1B). The microchemostat can work in one of two alternating states – one state is continuous circulation and the other is cleaning and dilution. When the microchemostat is in the continuous circulation state, the peristaltic pump pushes the microculture around the growth loop. With the ajdustment of pressure the researchers could set the linear velocity of moving the microculture around the loop to approximatelly 250 µm/s (Figure 1C). During the ceaning and dilution phase in one of the 16 segments the mixing with medium is paused. The segment gets isolated from the rest of the reactor with micromechanical valves, so that other parts don't get involved. Then, a lysis buffer is flushed through the segment to expel the cells it contains and all the cells that may adhere to the segment's wall (Figure 1D). At last, the segment is rinsed out with sterile medium and then filled with it. The micromechanical valves open and the segment gets joined with the growth loop. At this point the continuous circulation resumes and this alternating process is repeated sequentially on different growth loop segments. In such way the formation of biofilm gets interrupted in its early phase and the microchemostat can work pseudocontinuous. The treatment of microfluidic surfaces just with nonadhesive surface coatings proved innefective and biofims populated the channels within 48 hours (Figure S2 in the online supplementary materials).

The synthetic population control circuit

First, to test the microchemostat Balagaddé and his coworkers performed many experiments with Escherichia coli MG1655 cells (genotype: F-, λ-, rph-1; phenotype: reduced growth rate in minimal media due to reduced levels of PyrE) using a variety of growth media at 21 °C and 32 °C. Next, to demonstrate the ability of the microchemostat to facilitate analysis of complex growth dynamics, they performed experiments with E. coli MC4100Z1 cells (a casette containing lacIq, tetR and spect(R) genes was inserted int the chromosome of the MC4100 strain, genotype: araD139 Δ(argF-lac)205 flb-5301 pstF25 rpsL150 deoC1 relA1) and Top10F' cells (genotype: F´{lacIq, Tn10(TetR)} mcrA Δ(mrr-hsdRMS-mcrBC) Φ80lacZΔM15 ΔlacX74 recA1 araD139 Δ(ara leu) 7697 galU galK rpsL (StrR) endA1 nupG) with the integrated population control circuit, which autonomously regulates the cell density through a negative feedback system based on quorum sensing [1], [2], [9]. The population control circuit (pPopCtrl1) is based on a plasmid, that contains a p15A origin, kanamycinR, pluxI-lacZα-ccdB (lacZα-ccdB is the killer gene; CcdB is a toxin targeting the DNA girase in E. coli), luxR, luxI and a synthetic promoter plac/ara-1, that is inducible with isopropyl-β-ᴅ-thiogalactopyranoside or IPTG (Figure S1 in the online supplementary materials). To activate the circuit, 1 mM IPTG was used, but under this condition, the circuit in MC4100Z1 is only partially induced due to the presence of the AraC repressor, which binds to the araO sites in the synthetic promoter. Further inducing of the promoter with arabinose would be toxic for the cells. Because TopF10' cells do not produce AraC, 1 mM IPTG could fully induce circuit function. The induction of the synthetic promoter (plac/ara-1) starts with addition of IPTG and the circuit is in its ON state. The constant synthesis and sensing of the signaling molecule or autoinducer acyl-homoserine lactone (AHL) gives the information about the cell density (an increased concentration of AHL means increased cell density) and also modulates the expression of the lacZα-ccdB gene. At an increased cell density a lot of AHL accumulates in the cells and in the experimental medium. At a high enough concentration AHL can bind and activate the transcriptional regulator LuxR (it's production was induced with IPTG) that dimerizes. The C-terminal domain of activated LuxR relieves the repression exerted by H-NS nucleoid proteins that bind to the promotor (pluxI) and the trancsription of the lacZα-ccdB gene begins (Figure 1a), [9]. This killer gene regulates cell density by controlling the cell death rate.

Results

Growth experiments with E. coli MH1655 cells

The testing of the microchemostat began with more than 40 growth experiments with E. coli MH1655 cells in five different chips. There were several different conditions: variation in growth media (MOPS EZ Rich and LB broth with different concentrations of glucose (0.11 M, 1.1 M) and bacto-tryptone (0.1 g/L, 0.5 g/L, 3 g/L)), temperatures (21 °C and 32 °C) and dilution rates (0,24/h, 0.30/h, 0.34/h, 0.37/h) . In most cases, on the begining there was a short lag phase, followed by a exponential growth (log) phase and at a certain cell density a steady state where the same amount of cells that grew also died. The noticeable difference was in the steady-state cell concentrations. They decreased with the increasing dilution rate (at high dilution rates wash-out was observed) and increased with the increasing nutrient concentration (Figure 2A and 2B).

Growth experiments with E. coli MC4100Z1 and Top10F' cells carrying a synthetic population control circuit

Balagaddé and coworkers first tested E. coli MC4100Z1 cells to establish the growth dynamics with and without the synthetic circuit. Six simoultaneus experiments could be performed on a single microchemostat chip. The dilution rate was kept constant at 0,16/h. In reactors 1, 2 and 3 were cultures with the circuit and they were induced with IPTG (»circuit ON«). Reactor 4 contained populations without the circuit (negative control) and in reactors 5 and 6 were populations with the circuit but were not induced (»circuit OFF«). At the cultures without the circuit and the negative control cells (4, 5 and 6) the growth was similar to the MH1655 cells in the previous expreriments – the exponential phase that gave way to a steady-state regime and reached density of approximately 3,5 cells/pL after 6 h. The circuit ON populations (1, 2 and 3) had oscillatory dynamics that went on for almost 125 h (Figure 3A). As seen in (Figure 3B) the cleaning and dilution phases had almost no effect compared to the overall population fluctuations. Interestingly, the oscillations in cell density were associated with specific cell morphologies as seen in Figure 3A, a-e. On the micrograph a of the culture in reactor 3 there were healthy, cylindrical cells, because the population was small and not a lot of the killer protein was produced. In pannel b the density of cells got higher because of the exponential growth. The cells were generally healthy, but then the increased AHL concentration led to increased expression of the killer protein as seen in pannel c. Some cells changed their morphology and became filamented because of the toxin and the cell density began to decrease. The degradation process increased, as seen in pannel d, and the cell death intensified. Cells became more filamented. Further decrease in cell density led to a decrease in the killer protein concentration. The step of cleaning and dilution washed out all dead cells (pannel e) and the cells were again at the beginning of the cycle. After 6 oscillation cycles (~186 h) culture 3 went into steady state. The authors suggested it spontaneously escaped circuit regulation. In the E. coli strain Top10F' another cell growth dynamic and stronger growth regulation was expected, considering the more complete induction with IPTG. Because of that the authors tried to induce and halt the circuit (switching between ON and OFF states) during the experiment (Figure 4). The dilution rate was also kept at 0.16/h. All six cultures were with the circuit. At start, the cultures 1, 2 and 3 were induced (circuit ON) and cultures 4, 5 and 6 were not (circuit OFF). After 44 hours (point A), cultures 2 and 3 were turned OFF, then grew exponentially and went to steady state. When switched ON again after a period (point B, at 96 hours), they only shortly demonstrated circuit activation and then again went to steady state. Cultures 4, 5 and 6 were switched ON after 96 hours (point B) and only culture 4 established oscillations that were similar to cultures 1, 2 and 3 before switching OFF, the other two went into steady state after sharp decreasing in cell density. It was unusual that during the oscillations the morphology stayed the same as in circuit OFF cells.

Discussion

In the article they compared the results of the microchemostat to results they got in macroscale batch cultures (values are listed in the Supporting Online Materials, Figures S3, S4 and S5). The circuit was more stable in the microchemostats, because cultures lost regulation much later. In the macroscale experiments the cultures with circuits ON lost regulation after 70 (MC4100Z1 cells) and 48 hours (Top10F' cells), compared to the chemostat where population control could be maintained over 200 or even 500 hours (Figure 5). The theory is that in smaller populations there is a smaller rate at which mutations occur and change cell regulation (~10˄2 to ~10˄4 cells compared to ~10˄9 cells in microscale cultures), so a prolonged monitoring of genetically homogenous populations could take place. But then the oscillations in the misrochemostat would last longer than in the article results, so the exact reason is unknown. Also, there were different amplitudes of oscillations in the two E. coli strains. The reason may be circuit regulation or the circuit interacting with the continuous culture mechanism. Nonetheless, the oscillations in cell densiti were controllable and they occured only by adding IPTG (»circuit ON« state). The active approach to prevent biofilms showed good results and enabled the use of very small microfluidic channels that were used in the microchemostat.

Concluding remarks

The authors were able to make a microbioreactor with the working volume of only 16 nL. It was more than 300 times smaller as the smallest previous microfermentor and that alone was a big success. Along with it they managed to perform several experiments with different E. coli strains, where they succesfully prevented biofilm formation and simoultaneously monitored individual cells to look for specific morphologies. The authors suggested, that the measurements could be extended to some dynamic properties, such as gene expression dynamics and distributions reported by luminiscence or flourescence and that this could enable analysis of the phenotypisal caracteristics of different cell strains and networks.

References

[1] F. K. Balagaddé, L. You, C. L. Hansen, F. H. Arnold, and S. R. Quake, “Long-term monitoring of bacteria undergoing programmed population control in a microchemostat.,” Science, vol. 309, pp. 137–140, 2005.

[2] M. B. Miller and B. L. Bassler, “Quorum sensing in bacteria.,” Annu. Rev. Microbiol., vol. 55, pp. 165–199, 2001.

[3] W. C. Fuqua, S. C. Winans, and E. P. Greenberg, “Quorum sensing in bacteria: The LuxR-LuxI family of cell density- responsive transcriptional regulators,” J. Bacteriol., vol. 176, no. 2, pp. 269–275, 1994.

[4] C. M. Waters and B. L. Bassler, “Quorum sensing: cell-to-cell communication in bacteria.,” Annu. Rev. Cell Dev. Biol., vol. 21, pp. 319–346, 2005.

[5] G. O’Toole, H. B. Kaplan, and R. Kolter, “Biofilm formation as microbial development,” Annu. Rev. Microbiol., vol. 54, pp. 49–79, 2000.

[6] S. L. Novick A., “Description of the chemostat,” Science (80-. )., vol. 112, no. 2920, pp. 715–716, 1950.

[7] T. R. Herbert D, Elsworth R, “‘The continuous culture of bacteria; a Theoretical and Experimental study,’” J. Gen. Microbiol, vol. 14, no. 3, pp. 601–622, 1956.

[8] H. Smith and P. Waltman, “The Theory of the Chemostat.,” Cambridge University, Cambridge, UK. 1995.

[9] L. You, R. S. Cox, R. Weiss, and F. H. Arnold, “Programmed population control by cell-cell communication and regulated killing.,” Nature, vol. 428, no. April, pp. 868–871, 2004.