A protocol for the generation of dynamic chemical landscapes by photolysis within microfluidic and millifluidic setups is presented. This methodology is suitable to study diverse biological processes, including the motile behavior, nutrient uptake, or adaptation to chemicals of microorganisms, both at the single cell and population level.
We demonstrate a method for the generation of controlled, dynamic chemical pulses―where localized chemoattractant becomes suddenly available at the microscale―to create micro-environments for microbial chemotaxis experiments. To create chemical pulses, we developed a system to introduce amino acid sources near-instantaneously by photolysis of caged amino acids within a polydimethylsiloxane (PDMS) microfluidic chamber containing a bacterial suspension. We applied this method to the chemotactic bacterium, Vibrio ordalii, which can actively climb these dynamic chemical gradients while being tracked by video microscopy. Amino acids, rendered biologically inert ('caged') by chemical modification with a photoremovable protecting group, are uniformly present in the suspension but not available for consumption until their sudden release, which occurs at user-defined points in time and space by means of a near-UV-A focused LED beam. The number of molecules released in the pulse can be determined by a calibration relationship between exposure time and uncaging fraction, where the absorption spectrum after photolysis is characterized by using UV-Vis spectroscopy. A nanoporous polycarbonate (PCTE) membrane can be integrated into the microfluidic device to allow the continuous removal by flow of the uncaged compounds and the spent media. A strong, irreversible bond between the PCTE membrane and the PDMS microfluidic structure is achieved by coating the membrane with a solution of 3-aminopropyltriethoxysilane (APTES) followed by plasma activation of the surfaces to be bonded. A computer-controlled system can generate user-defined sequences of pulses at different locations and with different intensities, so as to create resource landscapes with prescribed spatial and temporal variability. In each chemical landscape, the dynamics of bacterial movement at the individual scale and their accumulation at the population level can be obtained, thereby allowing the quantification of chemotactic performance and its effects on bacterial aggregations in ecologically relevant environments.
Microbes rely on chemotaxis, the process of detecting chemical gradients and modifying motility in response1, to navigate chemical landscapes, approach nutrient sources and hosts, and escape noxious substances. These microscale processes determine the macroscale kinetics of interactions between microbes and their environment2,3. Recent advances in microfluidics and microfabrication technologies, including soft lithography4, have revolutionized our ability to create controlled microenvironments in which to study the interactions of microbes. For example, past experiments have studied bacterial chemotaxis by generating highly controlled, stable gradients of intermediate to high nutrient concentrations5,6. However, in natural environments, microscale chemical gradients can be short-lived―dissipated by molecular diffusion―and background conditions are often highly dilute7. To directly measure the chemotactic response of microbial populations first exposed to unsteady chemical environments, we devised and here describe methods to combine microfluidic technology with photolysis, thereby mimicking gradients that wild bacteria encounter in nature.
Uncaging technology employs light sensitive probes that functionally encapsulate biomolecules in an inactive form. Irradiation releases the caged molecule, allowing the targeted perturbation of a biological process8. Due to the rapid and precise control of cellular chemistry that the uncaging affords9, photolysis of caged compounds has traditionally been employed by biologists, physiologists and neuroscientists to study the activation of genes10, ion channels11, and neurons12. More recently, scientists have leveraged the significant advantages of photolysis to study chemotaxis13, to determine the flagella switching dynamics of individual bacterial cells exposed to a stepwise chemoattractant stimulus14,15, and to investigate motility patterns of single sperm cells in three-dimensional (3D) gradients16.
In our approach, we implement photolysis of caged amino acids within microfluidic devices to study the behavioral response of a bacterial population to controlled chemical pulses, which become near-instantaneously available through photorelease. The use of a low-magnification (4x) objective (NA = 0.13, depth of focus approximately 40 μm) allows both the observation of the population-level aggregative response of thousands of bacteria over a large field of view (3.2 mm x 3.2 mm), and the measurement of motion at the single-cell level. We present two applications of this method: 1) the release of a single chemical pulse to study bacterial accumulation−dissipation dynamics starting from uniform conditions, and 2i) the release of multiple pulses to characterize the bacterial accumulation dynamics under time-varying, spatially heterogeneous chemoattractant conditions. This method has been tested on the marine bacteria Vibrio ordalii performing chemotaxis toward the amino acid glutamate17, but the method is broadly applicable to different combinations of species and chemoattractants, as well as to biological processes beyond chemotaxis (e.g., nutrient uptake, antibiotic exposure, quorum sensing). This approach promises to help elucidate the ecology and behavior of microorganisms in realistic environments and to uncover the hidden trade-offs that individual bacteria face when navigating ephemeral dynamic gradients.
1. Fabrication of the Microfluidic Device for the Single Chemical-pulse Experiment
2. Fabrication of the 3D-printed Millifluidic Device for the Experiment with Multiple Pulses
3. Cell Culture
4. Calibration of the Uncaging Protocol
5. Single Chemical-pulse Experiment
6. Multiple Chemical-pulse Experiment
7. Image Analysis and Data Analysis
We used the microfluidic and millifluidic devices (Figure 1) to study bacterial accumulation profiles under dynamic nutrient conditions. Bacterial trajectories were extracted from recorded videos acquired by phase contrast microscopy of the accumulation-dissipation dynamics of a bacterial population following a chemical pulse released by photolysis (Figure 2 and Figure 3). By averaging millions of trajectories, the spatiotemporal dynamics of the radial drift velocity and bacterial concentration were obtained. Statistics describing the swimming in the absence of a chemical gradient were obtained in separate experiments with higher spatial and temporal resolution (Figure 4).
Figure 1: Microfluidic and millifluidic devices for bacterial experiments under dynamic nutrient conditions. (A) Photomask of the microfluidic channel used to observe the bacterial accumulation to a single chemical pulse. (B) Photomask of the microfluidics channel used to wash the millifluidic bacterial chamber for the multi pulse experiments. The small dots are micropillars (200 µm in size) that help the bonding of the membrane to the PDMS, and at the same time help prevent the membrane from buckling or collapsing in the center. (C) Design of the 3D printed master used to create the bacterial chamber. (D) The PDMS layer with the patterning of the bacterial chamber. (E) The complete millifluidic device (top view, PCTE membrane in white). (F) Schematic of the millifluidic device for the generation of dynamic nutrient conditions (side view). Optics ray diagram is shown on the bottom of the device, where a violet beam (395 nm) performs the uncaging of the chemoattractant, whereas an (independent) blue beam (470 nm) is used for the observation of fluorescent bacteria (harboring pGFP) through a sCMOS camera, which captures the bacterial density (bottom centre). The device is placed on a motorized stage (bottom right), which can be moved in the x-y plane to release chemical pulses at user-defined positions (controlled via NIS software, bottom left). Please click here to view a larger version of this figure.
Figure 2: Representative bacterial responses to a chemical pulse. (A,B) Maximum intensity projection showing bacterial position (white) over a 0.5 s interval, shown (A) immediately following, and (B) 40 s after the pulse release (objective: 4x, NA = 0.13, Ph 2; video recording at 12 fps). (C) Bacterial trajectories (black) shown 60 s after the pulse release. The shaded region represents the chemical pulse released by photolysis at t = 0 s in the middle of the field of view (black cross), which subsequently diffuses. Trajectories were reconstructed using custom in-house software. (D,E) Temporal dynamics of the radial drift velocity (D) and of the bacterial concentration (E) following a pulse release in the center of the field of view. In panel D, negative values of the drift velocity (in blue color) correspond to directed chemotactic motion towards the center of the pulse. This figure has been modified after Brumley et al.17. Please click here to view a larger version of this figure.
Figure 3: Spatio-temporal profiles of the radial drift velocity and bacterial concentration following a chemical pulse release. (A,B) The radial drift velocity (A) and the bacterial concentration (B) as a function of time and distance from the center of a 35 µM glutamate pulse. In panel A, negative values of the drift velocity (in blue color) correspond to directed chemotactic motion towards the center of the pulse. Values were calculated by averaging over all trajectories. The black dashed line at t = 250 s roughly delimits the period of active chemotaxis (blue shading in panel A, region I), after which the cells diffuse isotropically outward (green shading in panel A, region II). (C) Bacterial concentration (rescaled over the background B0) as a function of distance at t = 10, 30, 60 s after the creation of a diffusing chemical pulse. Bacteria aggregate close to the site of the pulse due to chemotaxis. (D) Bacterial concentration takes ~20 min to relax to the background B0 by bacterial diffusion. The inset shows the fit of the diffusive spreading with a diffusion coefficient of DB = 165 µm2 s-1. Panels A, C, and D have been modified from Brumley et al.17. Please click here to view a larger version of this figure.
Figure 4: Swimming statistics for a bacterial population in the absence of chemical gradients. (A) Trajectories represent the two-dimensional projections of the three-dimensional bacterial motion in the microchannel. Bacterial trajectories are extracted using a custom in-house tracking script. Here the blue cross indicates the start of the track, and red and green symbols indicate reorientation events. Data were recorded at 50 fps with a 20x objective, NA 0.45. (B) Probability distribution of measured bacterial swimming speed (black) together with a gamma distribution fit (blue). Because the depth of focus when imaging with a 20x objective (NA 0.45) is only a few microns24, the recorded trajectories are essentially planar and measurements of the swimming speed are not biased by the projection. (C) Probability distribution of the time between successive reorientations. These data were required to effectively calibrate an individual based model that takes into account the reorientation pattern, the distribution of swimming speed, and the reorientation statistics of the organisms. This figure has been modified from Brumley et al.17. Please click here to view a larger version of this figure.
This method allows researchers to study bacterial chemotaxis under controlled, dynamic gradients in micro- and millifluidic devices, enabling reproducible data acquisition. The near-instantaneous creation of chemical pulses at the microscale by photolysis aims to reproduce the types of nutrient pulses that bacteria encounter in the wild from a range of sources, for example, the diffusive spreading of plumes behind sinking marine particles25, or the nutrient spreading from lysed phytoplankton cells26.
We presented two applications of this method: 1) the release of a single chemical pulse to study the bacterial accumulation−dissipation dynamics starting from uniform conditions, and 2) the release of multiple pulses to characterize the bacterial accumulation profiles under non-equilibrium nutrient conditions. The first application is particularly suited to characterize the behavioral responses of microbes when first encountering a nutrient source. Under such conditions, in which the concentration of chemoattractant molecules is extremely low, the early phase of chemotaxis is dominated by the stochastic binding−unbinding events of chemoattractant molecules to the chemoreceptors17. Our method, by rapidly and precisely releasing a known mass of chemoattractant in a zero-nutrient background, offers significant advantages over previous approaches to characterize the bacterial response under dynamic conditions27,28,29. Advantages include knowing the full distribution of chemoattractant at all locations and at all times (since its diffusivity is known), and completely avoiding the generation of fluid flow that is inherently associated with other devices, such as the microinjector27 or three-inlet geometries30.
In the second application, the chemical pulses occur in a large, quasi-2D domain according to a user-defined sequence in space and time that is fully customizable via software, which can be used to impose random sequences or particular patterns. Importantly, this method provides a powerful link between the high-resolution behavioral dynamics of bacterial chemotaxis and nutrient uptake over timescales of seconds and long-term dynamics, such as growth and potentially evolution. The bacterial arena is considerably larger (2 cm x 2 cm) than the spatial range of chemical interaction of the bacteria with a single pulse (from hundreds of micrometers for the smallest pulses to a few millimeters for the largest pulses). Key to the maintenance of a low chemical background (much lower than the concentration of the chemical pulses at their release) is the inclusion of the nanoporous PCTE membrane sandwiched between the two PDMS layers18. By applying a fluid flow in the microfluidic channel placed at the top of the device, a continuous wash-out of the uncaged compounds and spent medium in the bacterial arena is realized by means of molecular diffusion through the nanoporous membrane, without creating flow in the test section of the device where bacteria are located (Figure 1).
By modulating the focused LED beam in time, amplitude, size, and geometry, photorelease technology endows the experimenter with great flexibility to generate different types of chemical environments. At the same time, while the tests presented here were performed under quiescent conditions, our method can be further expanded to test bacterial chemotaxis under different flow configurations. By faithfully reconstructing the bacterial accumulation dynamics through video microscopy, our method generates large quantities of high quality data that can be used to estimate the statistics of bacterial behavior and the potential nutrient uptake by bacterial cells. Our experimental microfluidic approach, mimicking nutrient landscapes that bacteria might face under natural environmental conditions, allows the systematic study of the foraging behavior of microbial species that are essential in the cycling of nutrients at the macroscopic scale2,3. As such, the type of data generated through this methodology is useful to effectively calibrate population uptake rates and better derive nutrient kinetics in mesoscale ecosystem models.
The fabrication of the PDMS molds to make the large bacterial arena was performed using a commercial 3D printing service. However, similar results can be achieved using a high-end 3D printer in house, with a resolution of 50−100 µm required to resolve the smallest features of the microchannel designs. A smooth surface of the 3D-printed material for the PDMS mold is required to achieve a good bonding between the cast PDMS and the other surfaces of the device (i.e., glass, polystyrene, PDMS). For our application, the use of a polystyrene Petri dish (90 mm x 15 mm) as the lower surface of the 3D arena presents two advantages over the use of a glass slide as commonly employed in microfluidics studies: first, it considerably reduces the attachment of bacterial cells compared to a glass surface (although attachment might depend on the particular surface properties of the microbe under consideration); second, it provides secondary containment, which can prevent leakage of media over microscopy equipment in the case of spills. The PDMS mold curing process typically occurs at a high temperature (70−80 °C), but in this application the experimenter must bake the PDMS mold at a considerably lower temperature (45 °C in the current case, see Table of Materials), below the heat deflection and the melting temperature of the material used for the 3D printing of the master. The lower baking temperature considerably lengthens the curing process (overnight), but does not change the mechanical and chemical properties of the PDMS.
Although our method has been applied to one particular combination of bacteria and chemoattractant, the methodology is suitable to test diverse biological processes, including nutrient uptake or antibiotic exposure, and can be applied to model systems of different species and chemoattractants, given that a myriad of molecules have been (or can be) caged8. One potential limitation arises from the costs of commercially available caged compounds, but these costs are comparable to those incurred when using the molecular probes typically employed in the life sciences for cell viability, counting, or intracellular staining. Notwithstanding this potential limitation, the proposed methodology may find broad applications across biophysical and biomedical sciences, to characterize early responses and adaptation dynamics of microbial populations at single cell resolution to dynamic chemical gradients.
The authors have nothing to disclose.
The authors thank the FIRST microfabrication facility at ETH Zurich. This work was supported by an Australian Research Council Discovery Early Career Researcher Award DE180100911 (to D.R.B.), a Gordon and Betty Moore Marine Microbial Initiative Investigator Award GBMF3783 (to R.S.), and a Swiss National Science Foundation grant 1-002745-000 (to R.S.).
(3-Aminopropyl) triethoxysilane (APTES) | Sigma-Aldrich | A3648 | >98% purity, highly toxic |
CELLSTAR tube | Greiner Bio-One | 210261 | 50 ml |
Centrifuge | Eppendorf | 5424R | to eliminate spent media from the bacterial culture |
Digital Incubators Incu-Line | VWR-CH | 390-0384 | to bake 3D master |
Duster | VWR-CH | 16650-22 | to clean the wafer and microchannels |
Hot plate | VWR-CH | 444-0601 | to bond the microchannels |
Isopropanol | Sigma-Aldrich | W292907 | |
LightSafe micro centrifuge tubes | Sigma-Aldrich | Z688312 | 1.5 ml |
MATLAB | Mathworks | for image analysis and bacterial tracking | |
Microcentrifuge tube | Eppendorf | 30120086 | 1.5 ml |
Microscope glass slide | VWR-CH | 631-1552 | |
Microscope Nikon Eclipse TiE | Nikon Instruments | MEA53100 | with motorized stage |
MNI-Glutamate | Tocris Bioscience | 1490 | >98 % purity, photosensitive |
Mold printing equipment | Stratasys | Objet30 3D printer | |
Mold printing service | 3D Printing Studios | Custom | https://www.3dprintingstudios.com/ |
Nanodrop One UV-Vis Spectrophotometer | Thermo Fisher Scientific | ND-ONE-W | to calibrate the uncaging |
NIS Elements | Nikon Instruments | Microscope Imaging Software | |
Oven Venti-Line | VWR-CH | 466-3516 | to bake PDMS (with forced convection) |
Photoresist SU-8-3050 | MicroChem Corp. | SU8-3050 | |
Plasma chamber Zepto | Diener Electronic | ZEPTO-1 | to functionalize the surfaces before bonding |
Polycarbonate membrane | Sterlitech | PCT0447100 | 0.4 µm pore size, 19 % open area, 24 µm thickness |
Polyethylene microtubing | Scientific Commodities | BB31695-PE/2 | I.D. x O.D.: 0.015" x 0.043" / 0.38mm x 1.09mm |
Polystyrene Petri dish | VWR-CH | 25373-100 | bottom surface (90 mm x 15 mm) to bond the millifluidic device |
Scale | VWR-CH | 611-2605 | to weight PDMS mixture |
sCMOS camera Andor Zyla | Oxford Instruments | for phase contrast and fluorescence microscopy (max 100 fps) | |
Sea salt | Instant Ocean | Product No. SS1-160p | |
SolidWorks 2015 | Dassault Systemes SolidWorks | Used to design the mold | |
Spectra X light engine | Lumencolor | for LED 395 nm | |
Sylgard 184 | Dow Corning | 110-41-155 | PDMS Si Elastomer Kit; curing agent |
Syringe (Luer-Lok) | B Braun Omnifix | 4616308F | |
Syringe Needle | Agani | A228 | from 10 to 30 ml |
Syringe Pump 11 Pico Plus Elite | Harvard Apparatus | 70-4506 | Terumo Agani 23 gauge 5/8 inch (16mm) |
VeroGrey | Stratasys | Dual Syringe Pump | |
Vortex-Genie | Scientific Industries | SI-0236 | Mold Material |