Here, we describe the use of a higher-throughput microfluidic bioreactor coupled with a fluorescent microscope for the analysis of shear stress effects on Pseudomonas aeruginosa biofilms expressing green fluorescent proteins, including instrument set up, the determination of biofilm coverage, growth rate, and morphological properties.
A higher-throughput microfluidic in vitro bioreactor coupled with fluorescence microscopy has been used to study bacterial biofilm growth and morphology, including Pseudomonas aeruginosa (P. aeruginosa). Here, we will describe how the system can be used to study the growth kinetics and the morphological properties such as the surface roughness and textural entropy of P. aeruginosa strain PA01 that expresses an enhanced green fluorescent protein (PA01-EGFP). A detailed protocol will describe how to grow and seed PA01-EGFP cultures, how to set up the microscope and autorun, and conduct the image analysis to determine growth rate and morphological properties using a variety of shear forces that are controlled by the microfluidic device. This article will provide a detailed description of a technique to improve the study of PA01-EGFP biofilms which eventually can be applied towards other strains of bacteria, fungi, or algae biofilms using the microfluidic platform.
Here, we will demonstrate a method to measure the effect of shear stress on the formation of fluorescent Pseudomonas aeruginosa (P. aeruginosa) PA01 biofilms using an automated higher-throughput microfluidic system.
Biofilms are communities of microorganisms, such as bacteria, organized by an extracellular polymeric substance that are attached to a support, and are typically found at the interface between a liquid and a solid surface1. These biofilm communities can be beneficial to the environment, such as improving water quality in water supply lines and in bioremediation of recalcitrant compounds2,3. However, biofilms can also be highly harmful to human health with undesirable consequences. For example, medical devices, such as hip and knee implants, are one type of surface where biofilm accumulation has been a challenge and causes severe medical complications4,5. Biofilms can also enter natural water systems, such as rivers and lakes, and infiltrate water supply pipes leading to bacteria contamination in drinking water resulting in infections6,7,8. Biofilms formed in marine environments adhere to ships and other man-made substrates and present a major economic and environmental problem as increased friction leads to increase fuel consumption9,10. Antimicrobial coatings, such as Tributyltin, have been developed to prevent these problems but are toxic to marine life11.
P. aeruginosa is a Gram-negative bacterium with high thriving capabilities in a variety of environmental and nutrimental conditions12. P. aeruginosa is a common cause of community- and hospital-acquired infections and found to be closely associated to injuries, such as severe burns, and immunocompromised hosts, such as in cystic fibrosis (CF)5,12,13, AIDS, and cancer patients5,13. The formation of P. aeruginosa biofilms has been most seriously connected to CF, where chronic lung infections are the leading cause of death for this disease5.
A reference strain of P. aeruginosa, PA01, is used in this report and is genetically modified to express an enhanced green fluorescent protein (PA-EGFP). EGFP represents a mutant form of GFP with greater fluorescence properties which allows for in situ biofilm analysis using fluorescence microscopy14,15,16. This type of fluorescence analysis is advantageous for the study of biofilms because GFP does not interfere significantly with cell growth and function17. For example, Escherichia coli cells that were tagged with GFP grew well and continuously without having suffered any toxic effects in comparison to the control bacteria17. Other reports substantiate this claim18,19,20. Furthermore, the use of a fluorescent reporter such as EGFP is quick and simple, yet only live cells will be measured because dead cells quickly cease to fluoresce21.
Biofilms can grow under various environmental conditions including those with different flow rates. For example, films can grow in high shear stress, such as in rivers, where the high water flow conditions lead to a greater microbial diversity22. Contrarily, standing water in ponds or oral biofilms experience a much lower shear force23. In addition to flow rate, there are other factors that influence biofilm adhesion, including surface roughness and hydrophobicity, media composition, and even the bacterial cell surface1,4,7,24. Conditions can also cause variation in the spatial structure or morphology of a biofilm. This includes environmental conditions such as shear stress exerted by a moving fluid or gradients in nutrient availability and biological factors such as the species present in the system, motility of cells, and the specific proteins present in the extracellular polymeric substance25,26,27. Under some conditions, the biofilm will be lawn-like (smooth and flat), while under other conditions the biofilm will be rough, fluffy, or even mushroom-like28. While the qualitative difference between biofilm lawns and mushroom structures can be clearly seen in microscopic images, understanding the relationship between film structure and biological processes within the film requires systematic and quantitative methods of describing the morphology. Morphological properties suggested for study by researchers include porosity, fractal dimension, diffusional length, microcolony area at the substratum, microcolony volume, roughness coefficient, and textural entropy29,30.
Bioreactors are used in the study of biofilms to mimic real life conditions31. Drip flow reactors (DFR) represent a low-shear environment where nutrients in media flow slowly across cells that are attaching to a surface over time to form a biofilm with a high cell density32. CDC reactors are bioreactors that create a high shear stress fluid environment by control of a stirring rod that is continuously rotating within the media filled tank33. These types of bioreactors are simple to set up, but they are limited in scope because of the relatively low sample size, the high consumption of media, large amounts of biohazardous waste produced from media drip flows ranging from 125 µL/minute for drip flow reactors to more than 1 mL/min for CDC reactors, and the need to autoclave large amounts of glassware and waste media34. Biofilms do not grow evenly across the surface in a drip flow reactor because the low shear of media causes trailing along the larger conglomerates of P. aeruginosa bacteria therefore, the biofilm growth is not very smooth and the uneven samples cannot be easily analyzed with fluorescence microscopy35,36 .
Some common bioreactor limitations are overcome by using a medium-throughput microfluidic bioreactor, where only milliliters of media are required, and the reaction plates are small and readily disposable after autoclaving37. Furthermore, depending on the number of wells, many replications can be performed in just one reactor run, which provides sufficient amount of data to conduct meaningful statistical analysis. In Figure 1, the different components of the microfluidic-microscopy system that allow for controlled conditions, including temperature and flow rate38,39,40, are shown. The bioreactor is coupled with fluorescence microscopy to visualize the fluorescence of the EGFP tag in PA01 under applied low through high shear conditions that will mimic more realistic scenarios that are encountered in the environment or in the biomedical field.
Figure 1: Individual components of the Microfluidic System. The individual components are listed from left to right: 1. CCD Camera, 2. High Resolution Inverted Microscope with Automated Stage, Automated Fluorescence Module, and Autofocus Module, 3. Plate Stage, 4: Imaging system Interface, 5: Manual Microscope Stage Control, 6: Vapor Trap, 7: Imaging system Controller (including Temperature Controller), 8: Hardware Controllers, 9: Fluorescence Controller, 10: Uninterruptible Power Supply, 11: External Hard Drive for Image Storage, 12: PC Workstation. Please click here to view a larger version of this figure.
An excerpt of the microfluidic plate is shown in Figure 2. The most commonly used plates consist of 48 wells. One experiment requires one inlet and one outlet wells, a total of 2 wells. This allows for 24 simultaneous experiments that can be performed with various experimental conditions, such as bacterial strains, antimicrobial treatments, and media varied from channel to channel and controlled shear flow for each column of six channels. The experimental temperature is also controlled with one temperature setting throughout the plate. The microfluidic channels show that each channel has a serpentine region to provide sufficient back pressure and controlled shear.
Figure 2: Visualization of the microfluidic channels and viewing window. Two inlet and outlet wells with the microfluidic channels connecting them are highlighted with red and green dyes. The dye makes visible a serpentine region in each channel which creates sufficient back pressure and controlled shear during fluid flow. Each viewing channel (within the red circle) can be imaged with desired wavelengths. Shown are bright field (top) and fluorescent (bottom) microscopy images of a single channel with a PA01-EGFP biofilm using a 20X objective. Scale bar = 80 µm. Please click here to view a larger version of this figure.
A step by step guide is provided to allow users of microfluidic bioreactors that are coupled with fluorescence microscopy to conduct novel biofilm experiments using different shear environments. This method will allow for the expansion of experiments involving other microorganisms besides bacteria, such as fungi and algae, that have medical and environmental applications41,42,43. The detailed approach describes how to culture PA01-EGFP, inoculate a 48-well plate and set up the microfluidic device and software, set up the fluorescent microscope, and demonstrate the software analysis to obtain the biofilm coverage, growth rate, and morphological properties such as surface roughness.
1. Media Preparation
2. Preparation of an Overnight and Experimental Culture of PA01-EGFP.
3. Equipment Startup
4. Priming and Seeding the Microfluidic Plate
NOTE: The priming, seeding, bacterial attachment and growth are illustrated in Figure 3.
Figure 3: Experimental overview of priming, seeding, and attachment of PA01-EGFP in the microfluidic channels. The priming, seeding, and attachment is outlined. The first step of priming requires fresh media introduced to the output. The seeding entails equal volumes of media and bacterial culture in the input and output wells, respectively. The culture should not pass the viewing segment of the experimental channel (red line) to avoid clogging the serpentine channels. After the incubation period is over, fresh media continuously flows from the input, to the viewing chamber and into the outlet. This initiates the attachment and growth of the bacterial biofilm. Please click here to view a larger version of this figure.
5. Setting the Software
6. Timed Biofilm Growth Experiment Setup
7. Review and Analyze Image Sequences after the Timed Biofilm Growth Experiment
8. Other Analysis Including Morphology and Surface Coverage Measure Using Python scripts from the Biofilm Morphology Suite
9. Other Software Applications – Making Overlays and Movies
Figure 4a demonstrates the percent thresholded area over time from a 24 h run at shear flows of 0.2, 0.5, and 1.0 dyne/cm2. The biofilm coverage or percent threshold surface area (C [%]) was different for all three shear settings. The biofilm coverage was fastest at 1.0 dyne/cm2 shear, where the threshold area increased from 2-5% to 100% after 200 min of growth and reached a stationary phase after 400 min. At 0.5 dyne/cm2, the biofilm coverage was delayed and began to increase at 400 min reaching 100% coverage after 800 min. The lowest shear at 0.2 dyne/cm2 clearly demonstrated the slowest increase in biofilm coverage, where percent coverage started to increase at 500 min but never reached beyond the 65% threshold area. These results indicated shear had a direct influence in biofilm surface coverage. The higher shear seemed to be a more optimal condition for biofilm growth, possibly due to the fact that the media provided the bacteria with more nutrients so that the biofilm can proliferate quicker.
Figure 4: Percent threshold area, total biofilm accumulation, roughness coefficient, and textural entropy. a) Percent threshold area (C [%]) over time using 0.2, 0.5, and 1.0 dyne/cm2 over a 24 h time period. Data was obtained from one channel for each shear condition. b) Total biofilm accumulation (relative measure) as a function of time using 0.2, 0.5, and 1.0 dyne/cm2 over a 24 h time period. The black lines are least-squares fits to an exponential model. Data was obtained from one channel for each shear condition. c) Roughness coefficient of PA01-EGFP using shear stress values of 0.2, 0.5, and 1.0 dyne/cm2 monitored over 24 h. Data was obtained from one channel for each shear condition. d) Textural entropy of PA01-EGFP using shear stress values of 0.2, 0.5, and 1.0 dyne/cm2. Data was obtained from one channel for each shear condition (Valquier-Flynn, H., Sutlief, A.L., Wentworth, C.D., 2018). Please click here to view a larger version of this figure.
Figure 4b is consistent with the findings shown in Figure 4a. The biofilm accumulation (N[rel]) was measured based on the assumption that the GFP signal at a point in an image is proportional to the live cell density at that position. It demonstrates that total relative measurement biofilm accumulation increased as a function of time using 0.2, 0.5, and 1.0 dyne/cm2 over a 24 h time period, and the growth rate declined from highest shear stress to lowest shear stress. There is a clear time period giving exponential growth from which a quantitative growth rate can be calculated.
Figure 4c demonstrates the roughness coefficient (Ra) at shear flows of 0.2, 0.5, and 1.0 dyne/cm2. Surface roughness, quantified by the roughness coefficient, measures the variance in the thickness profile of the film. The formal definition is
where Ti is the i-th thickness measurement, Ta is the average thickness, and N is the number of thickness measurement30. The procedure described in this investigation measures the thickness associated with living cells. A flat arrangement of cells will yield a roughness coefficient of zero while significant variations of thickness from the average will yield a roughness coefficient greater than one. Similar to the shear influence on the growth rate and the percent threshold coverage of the film, the biofilms exhibited different topographies over time. Overall, Ra decreased over time for all shear conditions indicating that all surfaces became smoother. However, in comparison to the lowest shear of 0.2 dyne/cm, the higher shear settings of 0.5 and 1.0 dyne/cm2 resulted in a smoother surface over time indicating that a faster shear flow contributed to smoother and more even surface, and the highest shear of 1.0 dyne/cm2 reaching below 0.2 Ra.
The smoothness, regularity, or coarseness of a surface can also be expressed in textural entropy (TE). TE is a property used in image analysis to measure the degree of randomness in a two-dimensional image. Its calculation is based on the gray level co-occurrence matrix, defined by Haralick et al., which looks at whether pixel values at one location are correlated with pixel values at another location44. A high degree of correlation will lead to low entropy. Figure 4d depicts the TE at shear flows of 0.2, 0.5, and 1.0 dyne/cm2. The TE increased over time for all shear conditions but the highest shear stress at 1.0 dyne/cm reached the maximum TE (1.0) earlier than the lower shear stresses at 900 min. The lowest shear of 0.2 dyne/cm2 had the lowest TE (0.8) reaching its maximum after 1,000 min. However, the intermediate shear stress of 0.5 dyne/cm2 reached its maximum TE (1.2) much later than the high or low shear stress conditions.
The roughness coefficient and TE measure different features. While a flat film would have a low roughness coefficient and low entropy, a film with significant variation in thickness would have a high roughness coefficient but could still be low entropy if the variation is sinusoidal rather than random. In this case, Ra decreased with increased shear stress and time while the TE trends cannot be directly related to shear stress applied to biofilm formation over time.
Figure S1: Image capture of the control software windows. Software window with Multi Dimensional Acquisition menu open (Top). Control Module set for an AutoRun (Bottom). Please click here to view a larger version of this figure.
Figure S2Image capture of the software for reviewing data. Application window after selecting the Data Set of interest using the Review Multi Dimensional Data tool from the Analysis Tools menu. Please click here to view a larger version of this figure.
The microfluidic system and image analysis procedure discussed here focuses on the execution of microfluidic biofilm experiments to determine morphological properties that do not require the full three-dimensional information typically found from confocal microscope studies. These included microcolony substratum coverage (percent coverage), surface roughness measured by the roughness coefficient, and textural entropy. A method for estimating total relative biofilm cell accumulation is also presented from which growth rates in log phase can be calculated.
There are several small but significant steps that should be highlighted in this method. Wiping the interface with alcohol helps avoid contamination of other bacteria in going from experiment to experiment but also from well to well in one experiment. Priming and seeding are also very important because priming allows the user to determine which channels are allowing the media to flow through the channels without any disturbances or clogging. Channels should also not be disturbed (i.e. they should be constantly full of media) after priming to increase the chances of a successful experiment with no air bubbles or clogging. The seeding step can be varied according the type of bacteria and so should be optimized for cell attachment. For example, if cells do not seem to attach, surface modifications may have to occur on the microfluidic plate prior to seeding or longer incubations times may be needed. It is also critical to make sure that the microscope is correctly set up to obtain an in-focus image and should be monitored periodically throughout the experiment to assure that quality images are obtained. If the focus is off, the microscope can and should be adjusted as the experiment continues. During the image acquisition time, the last wavelength should be set to All Closed in order to avoid the exposure of filters and lighting to only one channel during the waiting time that occurs in between the image acquisitions. Also, the image analysis that determines the % surface coverage was designed in house because the Montage software manual did not explicitly describe the procedure. Furthermore, in order to expand on the image analysis and determine other characteristics such as surface roughness, etc., open source Python-based code45 was developed in house and shared on the gitHub repository. There are also limitations on how much data can be stored and managed on the local hard drive, so an external hard drive or online data sharing is needed such as CyVerse46.
Conventional bioreactors, such as the CDC reactor and the drip flow reactor34, require a lot media, provide fewer sample sizes, and demand a high amount of sterilization of equipment. In contrast, the advantages of this higher throughput platform include the ability to control shear, flow rates, and the assumption that the in vitro experiments closely resemble in vivo conditions. Disadvantages of the system include the multiple accessories and software that require a meticulous setup that must be performed in the correct order of events. Furthermore, the manual that is provided for the equipment does not fully explain each step of the experiments, and the software commands, and consequently, many mistakes occur during the experiments, including clogging of channels, lack of growth or attachment, or the lack high quality microscopy images or movies. The instrument itself and consumables, such as the microfluidic plates, are also relatively expensive with a price tag of over $200 per plate and are not reusable. Thus, while the technique lends powerful results, the technical expertise required for its use is relatively high and requires repeated training by experts in the field. This report attempts to resolve this issue by providing a guide to new users of these bioreactors to study biofilms characteristics.
The microfluidic system, which is able to perform cellular analysis, has gained considerable attention for various scientific modalities, such as in microbiology, immunology, hematology, oncology, and stem cell research. More specifically, the technology has resulted in many publications describing topics that are highly relevant to medical applications37,47, including microbial oral adhesion48, determining the effects of biosurfactants on Pseudomonas aeruginosa and Staphylococcus aureus49,50, host pathogen interactions in E. coli51, Streptococcus adherence52, and treatment of cystic fibrosis53. Given the fact that this microfluidic system is very versatile, it is anticipated that more and more systems will be distributed throughout the world.
Some specific protocol steps should be carefully considered. Media can be diluted to 50% with dH2O to help prevent bubbles and clogging but was not required in this case. The specific value of OD600 used for seeding must be determined using trial runs of a growth experiment to see what works best for the particular set of conditions used. Bubbles in the wells prior to sealing can lead to bubbles in the microfluidic channels and should be removed by either popped or sucked out with a pipette tip. It is important to keep bacteria out of the small serpentine channels. By having equal volumes of media in the input and output during the seeding process, flow due to pressure from the liquid volume will be controlled so flow is only due to the applied pressure from the system. The calibration distances should be set up during installation by the company representative. These settings are specific for each camera.
There are several challenges that occur when finding the most representative threshold for an image. Setting maximum threshold values can be difficult if the average pixel intensity across regions of the background are not consistent caused by either selecting a stage position that is not at the center of the channel or from debris on the plate. Under the MM Standard, click on Process, and then select Background and Shading Correction tool to correction for these inconsistencies. However, this tool is generally only helpful if the user has taken images of the channels before seeding that they can use as reference images. Or, if background/shading reference images are unavailable, the user will need to use their judgment to set a threshold value that covers the most cell area without including background for the entire image. Alternatively, select representative areas to measure that exclude regions of inconsistency by click Rectangular Region, Ellipse Region, or Trace Region to select a region and select Active Region rather than Entire Image on the Show Region Statistics window (under Analysis Tools). If a representative region is utilized for thresholding the bright field image, the same region should be used for measurement of the corresponding FITC image. It is helpful to record the specific spatial statistics (Left, Top, 幅, 高さ, Area, Perimeter) associated with that representative region so the same region will be found and measured on the corresponding FITC image.
To prevent a buildup of data on the hard drive that will cause the computer to slow down, an external hard drive can be purchased for data storage. Another option for data storage and facilitating data sharing is the CyVerse bioinformatics platform. Create an account on the CyVerse system by going to http://www.cyverse.org/. Once logged in, launch the Discovery Environment then select "Log in CyVerse". Select "Data" and navigate to your the folder. If the image stack is on the local computer then select "Upload" then "Simple Upload from Desktop". Find the image stack file and select for upload. The file or a folder can be shared with collaborators if they have a CyVerse account and are granted permission. Sharing of the data folder to the general public requires that metadata be added for each file using CyVerse approved standards. This procedure will not be discussed here because this is not within the scope of this work.
The authors have nothing to disclose.
This work was made possible by grants from the National Institute for General Medical Science (NIGMS) (5P20GM103427), a component of the National Institutes of Health (NIH)
Ammonium Chloride, ACS | VWR | BDH9208-500G | Part of the minimal media composition |
BioFlux 1000 48 Well Low Shear Plate | Fluxion Biosciences | 910-0047 | |
BioFlux 1000Z Microfluidic Imaging System | Fluxion Biosciences | BF 1000Z | |
Calcium Chloride Dihydride, ACD Grade | VWR | 97061-904 | Part of the minimal media composition |
Dextrose, Anhydrous, ACS | VWR | BDH9230-500G | Part of the minimal media composition |
Magnesium Sulfate ACS Grade | VWR | EM-MX0070-1 | Part of the minimal media composition |
Potassium Phosphate Monobasic, ACS Grade | VWR | BDH9268-500G | Part of the minimal media composition |
Pseudomonas Aeurginosa GFP | ATCC | 15692GFP | Pseudomonas aeurinosa bacterial strain PA01 with GFP modification used for this study. |
Sodium Chloride, ACS | VWR | BDH9286-500G | Part of the minimal media composition |
Sodium Phosphate, Monobasic, Anhydrous, Reagent Grade | VWR | 97061-942 | Part of the minimal media composition |