We present a method for the microfluidic analysis of individual bacterial cell lineages using Bacillus subtilis as an example. The method overcomes shortcomings of traditional analytical methods in microbiology by allowing observation of hundreds of cell generations under tightly controllable and uniform growth conditions.
Microfluidic technology overcomes many of the limitations to traditional analytical methods in microbiology. Unlike bulk-culture methods, it offers single-cell resolution and long observation times spanning hundreds of generations; unlike agarose pad-based microscopy, it has uniform growth conditions that can be tightly controlled. Because the continuous flow of growth medium isolates the cells in a microfluidic device from unpredictable variations in the local chemical environment caused by cell growth and metabolism, authentic changes in gene expression and cell growth in response to specific stimuli can be more confidently observed. Bacillus subtilis is used here as a model bacterial species to demonstrate a "mother machine"-type method for cellular analysis. We show how to construct and plumb a microfluidic device, load it with cells, initiate microscopic imaging, and expose cells to a stimulus by switching from one growth medium to another. A stress-responsive reporter is used as an example to reveal the type of data that may be obtained by this method. We also briefly discuss further applications of this method for other types of experiments, such as analysis of bacterial sporulation.
One of the most striking features of life on Earth is its great resilience and variety. A central goal of molecular biology is to understand the logic by which cells use genes and proteins to maximize their growth and fitness under a wide variety of environmental conditions. To achieve this goal, scientists must be able to confidently observe how individual cells grow, divide, and express their genes under a given set of conditions, noting how cells respond to subsequent changes in their environment. However, traditional analytical methods in microbiology have technical limitations that affect the types of questions that can be addressed. For example, bulk culture-based analyses have been very useful over the years, yet they offer only population-level data that can mask meaningful cell-to-cell variations or the behaviors of smaller sub-populations of cells in the total population. Single-cell analyses of living bacteria based on light microscopy reveal single-cell behavior but are also technically limited. Bacteria are typically immobilized on agarose pads containing growth medium, but cell growth and division crowds the microscopic view and depletes the available nutrients after just a few cell cycles, substantially limiting the observation time1,2. Moreover, the local depletion of nutrients and the concomitant buildup of metabolic byproducts due to cell growth are constantly changing the local cell growth environment in ways that are difficult to measure or predict. Such environmental changes using agarose pads pose a challenge to studies of steady-state behaviors or of cellular responses to specific changes in growth conditions3.
Microfluidic technology, in which a liquid medium is continuously flowed through microfabricated devices, offers a solution to classic experimental limitations. A microfluidic device can keep individual cells in position for live-cell microscopy while the flow of growth medium constantly provides cells with fresh nutrients and washes away metabolic byproducts and excess cells, thereby creating a highly uniform growth environment. Under constant growth conditions, cell behaviors can be observed in isolation from the influence of environmental factors, permitting an unimpeded view of the internal logic of cells. As the fluid flow prevents the microfluidic device from becoming crowded with cells, observation of single cell lineages for tens or hundreds of generations becomes possible4,5. Such long observation times permit the detection of otherwise undetectable long-term or rare cell behaviors. Finally, the composition of the medium that flows through the device can be altered at will, allowing cells to be observed as they respond to the onset of a stress or to the introduction or removal of a particular compound.
Microfluidics has already enjoyed a number of important applications. For instance, it has been used in tissue-, organ-, or body-on-a-chip devices, in which multiple human cell types are co-cultured to simulate an in vivo condition6; for the study of nematode movement in microstructured environments7; to examine interactions among bacterial biofilms (e.g., 8); and for the encapsulation and manipulation of tiny volumes of cells or chemicals (e.g., 9). Microfluidic devices have also become increasingly popular in the field of microbiology (for excellent reviews, see 10 and 11), especially as their physical and flow properties are well-matched to natural microbial niches12. For instance, microfluidics has been recently employed by microbiologists for such purposes as precisely measuring cell growth and division13,14,15, analyzing pathogen movement16, monitoring quorum sensing17 and physiological transitions18, and for protein counting19, among many other examples. The method presented here is specifically designed for the analysis of single bacterial cell lineages rather than combinations of strains or species. The microfluidic device demonstrated here utilizes one variation of the "mother machine" design4, in which cells are grown single-file within a microfluidic trench with one closed end and one open end; cell growth and division pushes progeny cells up and out of the open end into the fluid flow. Our analyses typically focus only on the "mother" cell that is confined at the closed end of the trench. We consider this method as an advancement over previous light microscopy-based single-cell analytical techniques, such as cell immobilization on agarose pads. While B. subtilis is used as a model here, the method is also applicable to other bacterial species (Escherichia coli is another common model; some species with different cell sizes or morphologies may require the fabrication of new devices with different dimensions). The use of fluorescent reporters to mark cells and to visualize changes in gene expression requires the use of genetically tractable species; however, analyses of cell growth and morphology are possible even without fluorescent markers.
The present protocol excludes the process of fabricating the silicon master using photolithography, which has been extensively described elsewhere5; masters can also be easily outsourced from microfabrication facilities. It includes the molding of a PDMS device from a reinforced silicon master; bonding the device to a piece of cover glass; assembling the microfluidic inlet and outlet plumbing, including pinch valves to permit medium switching; passivating the device, preparing bacterial cells, and loading the device with bacterial cells; attaching the plumbing to the device and equilibrating the cells; and loading the device onto a fluorescence microscope for imaging. Because many different image acquisition and processing software tools can be used to visualize and analyze different data of interest4,5, example images are shown, but image-capture methods are not included in this protocol.
1. PDMS Device Casting
2. Device Punching and Bonding
3. Microfluidic Plumbing Preparation
4. Cell Culture Preparation
5. Device Loading
6. Device Assembly, Equilibration, and Mounting
7. Medium Switching
Successful initial cell loading, as assessed by phase-contrast microscopy before attaching the microfluidic plumbing to the device, would be considered as having all or nearly all of the microfluidic side channels containing one or more bacterial cells (Figure 3A). Optimal loading would show several cells in each channel, but channels will nonetheless fill with cells due to cell growth during the equilibration period (Figure 3B). Lanes with poor loading, in which relatively few (<50%) side channels are loaded with cells, may be used, but the resulting data density will be lower due to fewer cell lineages being imaged at each imaging position.
Once the device is initially loaded, it is equilibrated by attaching the microfluidic plumbing and flowing medium through the device at a relatively high flow rate of 35 µL/min. The equilibration step serves to remove air bubbles and excess cells in the feeding channel from the device, and to encourage the cells in the side channels to begin growing. The cells that are removed from the device upon medium flow can often be observed by the naked eye as a light-colored band moving through the outlet tubing towards the waste container; the movement of such bands serves as a visual indicator that fluid is flowing normally through the plumbing and the device. Microscopic inspection of an equilibrated device should reveal a few cells confined in single file in most of the side channels (Figure 3B, 0 min).
Once an equilibrated device is placed on the microscope under flow at 1.5 µL/min at 37 °C, the cells will resume uniform and constant exponential-phase growth over the course of approximately 2 h (Figure 3B). Constant exponential growth should be directly verified after image analysis by the appearance of a plateau in the generation time of the cells. At this point, the cells are ready for fluorescence and/or brightfield imaging as desired. A successfully loaded and equilibrated device can typically be reliably imaged for periods of at least 24 h (Figure 4A, C). Introducing a medium switch (Figure 4A–C) expands the range of experiments that can be conducted but also increases the frequency of catastrophic cell-death events (Figure 4D), presumably through the introduction of air bubbles that were not successfully purged from the tubing. Another event that may cause catastrophic cell death is that clusters of cells located at the inlet may become dislodged and pass through the feeding channel, as shown in Figure 4D. We do not at present understand why this causes catastrophic cell death in the device, and we have not yet devised a method to prevent such events from occurring.
Figure 1: Schematic of the microfluidic device. A device schematic is shown approximately to scale. The alphabetic designator is labeled along with the inlet and outlet zones that are punched to connect the blunt-ended needles. Typically, half of the patterned channels are used (shaded gray). The cell trenches are oriented towards the designator. The accompanying CAD file (see Supplementary File 1) shows all of the features on the device. Please click here to view a larger version of this figure.
Figure 2: Microfluidic device passivation, cell loading, and equilibration. (A) A bonded device after passivation with BSA-containing medium. Gel-loading tips are kept in the outlet holes to monitor the passage of medium and cells through the device. The medium meniscus is visible in the thin portion of the gel-loading tips (arrow). (B) Device after cell loading. The cells are visible as a translucent white layer in the gel-loading tips (arrow). (C) The gel-loading tips are removed before centrifugation in a custom-fabricated centrifuge adaptor. Medical tape (blue) is placed on the aluminum parts to cushion the device. (D) Following centrifugation and verification of loading, the device is taped to a microscope stage insert and attached to the inlet and outlet pins. In the image shown, medium switching is made possible by pinch valves and Y-connectors that are arranged in an apparatus made from a micropipette tip box. (E) Schematic view of an entire assembled device, from the syringe pumps to the waste container. The outlet tubing runs to a small waste beaker. Please click here to view a larger version of this figure.
Figure 3: Cell loading and growth in the microfluidic device. (A) From left to right, phase-contrast micrographs of a device containing medium only before cell loading, the same device after cells have been added via pipetting but before centrifugation, and the same device after centrifugal loading. (B) Time course of cell adaptation and growth in the device (LB, 37 °C, flow 1.5 µL/min). At the beginning of adaptation, the stationary-phase cells are relatively short and narrow. As the cells adapt and return to exponential-phase growth (60 – 120 min), they adopt a longer and wider morphology. After 120 min, all cells in the device are resuming uniform exponential-phase growth, and the device is ready for imaging. Please click here to view a larger version of this figure.
Figure 4: Cell growth and medium switching in the microfluidic device. (A) Kymographs showing 550 min of growth before and 1,000 min of growth after a medium switch (grey dashed line) into 2% ethanol as a stressor. Images were captured at 10-min intervals. The top panel shows a transcriptional mNeonGreen reporter (green channel) for a stress-induced gene. The bottom panel shows a constitutive mNeptune reporter (red channel) that is used in this case for automated cell segmentation. (B) Close-up view of the portions of the kymographs in the dashed box in panel A, showing a transient response in the green channel following the medium switch. Note that growth continues through the switch, although the presence of the ethanol causes the cells to become shorter and grow at a slightly slower pace. Time scale is shown by the horizontal bar, and distance is shown by the vertical bar. (C) Example traces of analyzed fluorescence data of a single cell lineage from the experiment shown in panel A. The vertical dashed lines indicate the medium switch into 2% ethanol as a stressor. (D) Example of a failed medium switch. When the second medium reaches the cells, the cells in the channels are lysed. The switch is accompanied by a large number of cells that pass by in the main channel, causing a bright fluorescent signal (a re-scaled image corresponding to the washed-out region is shown just above it). After the switch, no further cells are observed, although dimly fluorescent cell debris remains in the channels. Time scale is shown by the horizontal bar, and distance is shown by the vertical bar. Please click here to view a larger version of this figure.
This microfluidics protocol is flexible in that many of the steps may be modified to optimize its use with a particular species or strain or for a specific purpose. Indeed, in this protocol we have made modifications to the original "mother machine" concept4 to optimize its use with B. subtilis. Often, the trenches in which the cells are confined constitute a single-layer feature, whereas in this protocol we use two-layer cell trenches, with a shallow channel surrounding the cells. The two-layer design was introduced as a way to maximize the flux of nutrients to and metabolites from B. subtilis cells, especially in long (>75 µm) trenches5; however, single-layer features often suffice for optimal cell growth and are commonly used for E. coli, especially when they are shorter (~25 µm)4. We typically use longer trenches, as they reduce the frequency with which B. subtilis cell chains, which arise stochastically, are pulled out of their trenches by the medium flow in the main feeding channel5.
Further modifications, such as the use of devices with different feature dimensions (e.g., channel widths) or characteristics (e.g., shape, number of open ends) may be substituted as appropriate. For instance, the use of cell trenches that are open on both ends (rather than at only one end, as in the present protocol) has been used in other studies1,20,21. Trenches that are open on both ends avoid cell aging because they are constantly being filled with newborn cells (as opposed to mother machines, where the oldest cell is tracked and newer cells are pushed out of the channel), although the fact that older cells are pushed out both ends typically limits their observational window to fewer than 10 generations1,21. We typically seek the longer observational windows (hundreds of generations) offered by a mother machine, which make it possible to measure even memoryless processes for which waiting times are tens or hundreds of generations long5. We have also used the mother machine design to observe cells that are not exponentially growing, as in our analyses of sporulation22. In such cases, additional cells throughout the growth trenches can be meaningfully analyzed22.
Many of the steps in the protocol are relatively forgiving in that they may be modified without substantially affecting the protocol outcome. For instance, an earlier version of the protocol called for cleaning the cover glass by sequential 10-min bath sonication steps in 10 M KOH and then pure water, but we found that good device bonding and operation was achieved using the simpler isopropanol-based cleaning step described here. If a bonding issue arose that cast suspicion on the cleanliness of the cover glass, reverting to a more-stringent cleaning protocol is recommended. Similarly, while we use a centrifugation step to maximize cell loading into the cell trenches, reasonably full loading can be achieved even without this step, provided that very concentrated cells are used for the loading step. This alternative strategy could be especially useful for researchers who do not have a readily available centrifuge adapter to spin the device. Again, different concentrations of passivating/lubricating agent (BSA in this protocol) may be used; we routinely use concentrations as high as 1 mg/mL. In cases where a strain may be sensitive to BSA, acceptable results may be achieved even in the absence of a passivating agent. Indeed, when we were using a microfluidic device to examine sporulation, which requires cell starvation, we were concerned that BSA might act as a potential nutrient source for the cells, and so we omitted it from the medium flow without observing any substantial adverse effects22.
Notably, the continuous flow of medium through the device can elicit slightly different phenotypes compared with cell growth in a closed system, such as a flask. For example, nutrient or compound depletion can be challenging, as cells can often scavenge trace compounds from the medium flow. Indeed, we have observed that inducing cell sporulation via starvation requires a longer time in a microfluidic device than in a flask. Similarly, we have observed that induction of energy stress using the oxidative phosphorylation decoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP) requires several-fold higher concentrations in the microfluidic device than reported in flask culture23. Therefore, some optimization may be required to achieve satisfactory results with different medium additives.
The authors have nothing to disclose.
This project was funded by the National Institutes of Health under GM018568. This protocol was performed in part at the Center for Nanoscale Systems (CNS), a member of the National Nanotechnology Coordinated Infrastructure Network (NNCI), which is supported by the National Science Foundation under NSF award no. 1541959. CNS is part of Harvard University. Many thanks are due to Thomas Norman and Nathan Lord for their work in conceiving and fabricating the master template used for the devices shown here and in building the original version of the apparatus. We also thank Johan Paulsson for his valuable collaborative advice and thank members of his lab for their advice and continued improvements to bacterial microfluidic apparatuses.
Sylgard 184 Silicone Elastomer Kit | Dow Corning | (240)4019862 | This is referred to as PDMS in the protocol. There are 2 components that are mixed at a 10:1 ratio |
0.75-mm biopsy punch | World Precision Instruments | 504529 | |
22 x 40 mm No. 1.5 cover glass | VWR | 48393 172 | |
Plasma Etch PE-50 | Plasma Etch Inc. | PE-50 | Instrument used to bond PDMS to glass using oxygen plasma treatment |
Tygon flexible tubing ID 0.02", OD 0.06" | Saint-Gobain PPL Corp. | AAD04103 | For the main part of the tubing, e.g. attached to the needles |
Silicone Tubing, 0.04" ID, 0.085" OD | HelixMark Standard Silicone Tubing | 60-795-05 | For pinch valves and Y-junction connection |
Bovine Serum Albumin | Sigma-Aldrich | A7906-100G | Used as passivation agent |
ART Gel Pipet Tips (P200) | Molecular BioProducts | 2155 | For loading the device with medium and cells |
Acrodisc 32-mm syringe filter with 5-μm Supor membrane | Pall Life Sciences | 4560 | For filtering cultures before device loading |
21-ga blunt needles, 1" | McMaster-Carr | 75165A681 | |
Y connector with 200-series barbs, 1/16" ID tubing, polypropylene | Nordson Medical | Y210-6005 | The company is AKA Value Plastics |
Eclipse Ti-E inverted microscope | Nikon | This unit has been discontinued by the manufacturer and is replaced by the Ti 2 E. | |
LB Lennox | Sigma-Aldrich | L3022 | |
Microfuge 18 | Beckman Coulter | 367160 | |
Bacillus subtilis strain NCIB3610 | Bacillus Genetic Stock Center | 3A1 | wild-type parental strain modified for use in the experiments shown in this protocol |
Gravity convection oven | VWR | 414005-106 | for curing PDMS and baking assembled devices |
Scotch-Weld Epoxy Kit | 3M | 2216 B/A | may be used to bond a silicon wafer to an aluminum backing plate |
Scotch Magic tape, 3/4" width | 3M | to clean dust from PDMS devices and to place over features to increase visibility (denoted "office tape" or "adhesive tape" in protocol) | |
Stereomicroscope | Nikon | SMZ800N | listed here as an example; nearly any stereomicroscope will do |
Isopropyl alcohol | Sigma-Aldrich | W292907 | To clean cover glass |
Syring pump, 6 channel | New Era Pump Systems Inc. | NE-1600 | |
Kimwipes | Kimberly-Clark | 34120 | a brand of dust-free wipes |