We have developed a microfluidic platform for rapid antibiotic susceptibility testing. Fluid is passed at high speeds over bacteria immobilized on the bottom of a microfluidic channel. In the presence of stress and antibiotic, susceptible strains of bacteria die rapidly. However, resistant bacteria can survive these stressful conditions.
We have developed a rapid microfluidic method for antibiotic susceptibility testing in a stress-based environment. Fluid is passed at high speeds over bacteria immobilized on the bottom of a microfluidic channel. In the presence of stress and antibiotic, susceptible strains of bacteria die rapidly. However, resistant bacteria survive these stressful conditions. The hypothesis behind this method is new: stress activation of biochemical pathways, which are targets of antibiotics, can accelerate antibiotic susceptibility testing. As compared to standard antibiotic susceptibility testing methods, the rate-limiting step - bacterial growth - is omitted during antibiotic application. The technical implementation of the method is in a combination of standard techniques and innovative approaches. The standard parts of the method include bacterial culture protocols, defining microfluidic channels in polydimethylsiloxane (PDMS), cell viability monitoring with fluorescence, and batch image processing for bacteria counting. Innovative parts of the method are in the use of culture media flow for mechanical stress application, use of enzymes to damage but not kill the bacteria, and use of microarray substrates for bacterial attachment. The developed platform can be used in antibiotic and nonantibiotic related drug development and testing. As compared to the standard bacterial suspension experiments, the effect of the drug can be turned on and off repeatedly over controlled time periods. Repetitive observation of the same bacterial population is possible over the course of the same experiment.
The rise of bacterial resistance intensifies the need for fast phenotype-based antibiotic susceptibility tests in order to safeguard our drugs of last resort. Standard susceptibility tests are based on bacterial growth inhibition in the presence of antibiotics that take multiple (8-24) hours to complete. We have developed a novel antibiotic susceptibility test on a microfluidic platform that relies on the stress-activation of biosynthetic pathways to accelerate the action of antibiotics.
Antibiotic susceptibility tests at the microfluidic scale carry the advantage of effective sample usage, since they require small numbers of bacteria. Additionally, microfluidic devices can be multiplexed in order to test multiple samples under multiple conditions1,2. Recently, a number of microfluidic methods for antibiotic susceptibility testing have been reported3-9. In these methods, bacteria are grown inside nano- and picoliter droplets3,7, in the full volume of the microfluidic channel4-6,8, or as single bacteria electrically localized to the bottom surface of the channel9. Although these tests are carried out in microfluidic channels, they all monitor microbial growth in the presence and absence of antibiotics similar to traditional methods. Growth measurements are taken via optical density, pH sensitive dyes, or bright field/phase contrast or fluorescence images. Although some of these tests are faster than traditional methods, they each passively detect antibiotic resistance. In other words, these methods still require the user to wait for bacterial growth as the final read-out.
In contrast, we have developed a method that uses a combination of shear and enzymatic stress to activate antibiotic-sensitive biochemical pathways10. Challenging the stressed bacteria with those antibiotics creates a more rapid susceptibility test. Bacteria that are resistant to the antibiotic are able to withstand the stressful conditions. Susceptible bacteria, on the other hand, are rapidly killed by the combined stresses. The percentage of cell death after one hour, measured by microscopy using a fluorescent dead cell stain, defines the phenotype of the bacteria (resistant vs. susceptible).
For successful implementation of our method, bacteria must be immobilized on the bottom surface of the microfluidic channel. In this way, bacteria can be subjected to various stresses and simultaneously imaged under a microscope in a single plane. A coated microscope glass slide is used for bacteria immobilization. The slide is precoated by the manufacturer with epoxide groups for nonspecific protein binding. The nonspecific binding of these epoxides to bacterial surface proteins covalently attaches the bacteria to the slide surface.
Strains are tested under identical conditions (shear + enzymatic stress) in the absence (control) and presence (experiment) of antibiotic. Phase contrast and fluorescence microscope pictures of each channel are taken automatically every two minutes for one hour. Resistance designations are then made by comparing the percent of dead bacteria in the experimental channel to those present in the control channel. After one hour, a sample with a cell death percentage greater than 1% is deemed susceptible, while less than 0.5% death is indicative of resistance. Percentages that fall between these two cut-offs are considered indeterminate and the sample must be tested again.
Microfluidic channels are defined in PDMS, which is a material of choice for microfluidic devices11. PDMS is optically transparent in a wide range of wavelengths, biocompatible, inert, permeable to gases and has low permeability to liquids; therefore it is well suited for these experiments.
Mechanical/shear stress is created by the flow of room temperature media over the immobilized bacteria. (Note: Warming the media to 37 °C has no significant effect on assay outcome.) Automated syringe pumps force media (containing dead cell stain +/- antibiotic, as well as optional enzymatic stressors) through the microfluidic channels (200 µm x 400 µm) at a flow rate of 1 ml/min to give 6.25 kPa of shear force or a shear rate of 6,000 sec-1. This rate equals or exceeds previously studied shear stresses on Staphylococci.
The enzyme, lysostaphin, was selected for preliminary experiments because it causes direct damage to the Staphylococcus cell wall. The concentration of lysostaphin (0.7 ng/ml) was sufficient to cause bacterial cell wall damage, but not sufficient to cause bacterial cell death without antibiotic in the time frame of the experiment. Lysostaphin is not required for the correct designation of bacterial susceptibility but it does augment the outcome, leading to increased cell death in susceptible strains. In contrast, shear stress is critical for assay function. When methicillin-sensitive Staphylococcus aureus strains are treated with lysostaphin and oxacillin in the absence of flow, no cell death is recorded over the course of the experiment.
Cell viability is monitored with a fluorescent dead cell stain12. The selection of the dye was based on its ability to selectively stain only damaged cells, its nontoxicity to live cells, and its low background fluorescence, which allowed for its direct addition to the cell media without additional steps. The selection of a fluorescent dye concentration of 0.25 µM was to achieve acceptable signal levels during a 1.6 sec exposure time to fluorescence excitation light.
The beta-lactam, oxacillin, was used in our preliminary studies. Methicillin-resistant S. aureus (MRSA) species are resistant to oxacillin and will not show any appreciable cell death in the time frame of the experiment. The concentration of 50 µg/ml was determined in the preliminary studies. Lower concentrations of antibiotic gave less separation between resistant and susceptible strains, while higher concentrations did not cause an appreciable difference in experimental outcomes.
We have previously reported on the successful development of a test that combines mechanical and enzymatic stresses that directly affect the bacterial cell wall13 with an antibiotic that inhibits cell wall biosynthesis14,15. These proof-of-principle experiments were carried out on a panel of MRSA and methicillin-sensitive S. aureus (MSSA). However, with the selection of proper experimental parameters, our method should be applicable to multiple species of bacteria and multiple classes of antibiotics.
1. Make the PDMS Layer (Figure 1)
2. Assemble the Flow Cell According to Figure 2
Standard assembly of PDMS with glass slides is done through oxygen plasma treatment of both surfaces, which ensures leak-free bonding between the PDMS and microscope glass slide. In the presented protocol, the plasma treatment would destroy the chemical coating on the glass slide. Therefore the slide is pressure-sealed rather than plasma treated.
3. Prepare Log Phase Bacteria
4. Warm the Experimental Solution Components at Least 10 Minutes Before They are Needed
5. Prepare and Load the Bacterial Suspension
6. Prepare and Load the Experimental Solutions
7. Set Up the Flow Cell Under the Microscope
8. Run the 60 Minute Experiment
9. Disinfect the Flow Cell
10. Analyze Images and Generate Data
The data presented in Figure 4 show the response of a susceptible Staphylococcus aureus strain over time in an antibiotic-containing microfluidic channel. Phase contrast images acquired at 1 min and at the end of the 1 hr experiment are shown in Figures 4A and B. The analyzed 1 hr data are shown in Figure 4C with the bacteria highlighted in red (5,828 total). Corresponding fluorescence images are shown in Figures 4D and E. The analyzed 1 hr data are shown in Figure 4F with the fluorescing bacteria highlighted in red (174 dead).
This set of images illustrates several points. Most importantly, a significant increase in the number of fluorescing bacteria is observed by the end of the experiment (compare the number of bright fluorescent spots at 1 min and 60 min in Figures 4D and E), which indicates cell death inside the channel. These results are indicative of a susceptible strain.
Another important point to note is the necessity of the normalizations used in the equation above. Due to the stochastic nonuniformity of individual epoxy slide coatings, there may be bacterial cell loss throughout the experiment under sustained shear pressure (compare Figures 4A and B). To account for this variation, each fluorescence image of a dead cell population is normalized with the coacquired phase contrast image giving the total cell population at the same time point (within 1 sec).
One more normalization is done by subtracting the normalized fluorescence at the start of the experiment (t = 1 min) from all other time points (t = T). The dead cell fluorescence at 1 min corresponds to cell death prior to the beginning of the experiment. High fluorescence observed at the beginning of the experiment for either MSSA or MRSA usually indicates a poor state of the bacterial culture. On rare occasions, it represents contamination of the PDMS slide.
The insets on the bottom corners of the images are enlarged versions of the boxed image areas from both the raw and analyzed images. These enlargements show that our counting algorithm is more successful at accurately counting individual bacteria in fluorescence images than in densely populated phase contrast images.
The dead cell stain gives a high fluorescence contrast for individual dead bacteria. Since the number of fluorescing bacteria rarely surpasses 5% of the total number, each bacterium is very bright as compared to the background. For this reason, the fluorescent bacteria can be easily counted with a high degree of accuracy. On the other hand, bacteria are densely packed on the phase contrast image. Moreover, the phase contrast within a bacterium is not always uniform. These two factors present a challenge for a counting algorithm as it relies on thresholding of different areas based on their approximate size ranges and intensities.
Therefore, in order to compare experimental results, it is necessary to use the same counting algorithm with the same parameters throughout different experiments. It is worth noting that due to the low number of fluorescence counts, there is a higher sensitivity to fluorescence miscounts; therefore it is more important to accurately count the fluorescence images than the phase contrast images.
Finally, MSSA and MRSA data normalized according to the equation above are shown in Figure 5 for three different experiments. As expected for resistant strains, the normalized cell death is low in magnitude (<0.5%) and does not change over the course of the experiment. Susceptible strains show a steady increase in cell death and a higher value by the end of the experiment (>1%). The initiation of cell death for susceptible strains varies slightly between experiments, but usually stays between 10-30 min.
Figure 1. A photograph showing the geometries and sizes of the microfluidic channels in the PDMS layer. The narrow sections of the channels are 3.7 cm long, 200 µm tall and 400 µm wide.
Figure 2. A CAD model illustrating each of the flow cell components and their relative positions.
Figure 3. A photograph showing the experimental set-up of the microscope and syringe pump.
Figure 4. Phase contrast (A–C) and fluorescence (D–F) images for a representative MSSA strain within an antibiotic-containing channel. Insets show enlarged images of the boxed areas. Microscope images were acquired at 60X magnification. Click here to view larger image.
Figure 5. A graph showing normalized cell death percentages versus time. Each of the MSSA and MRSA strains were tested in three separate experiments.
Table 1. CellProfiler routine components for automated bacteria counting.
The presented protocol was validated and optimized in a set of experiments with methicillin-sensitive and methicillin-resistant Staphylococcus aureus strains10. Therefore, this protocol without modification should be directly applicable to other strains of S. aureus and other antibiotics with mechanisms of action affecting bacterial cell wall biosynthesis. Bacteria types other than S. aureus may require variation in the stress parameters: soluble (enzymatic) and mechanical stresses. If soluble stressors are desired in the testing of bacterial species other than Staphylococcus, another stress agent will be required, since lysostaphin is a staphylococcal–specific enzyme.
The amount of shear stress is proportional to the flow rate of the liquid and inversely proportional to the size of the channel, therefore in principle it can be varied within a large range of values. The achieved values of stress are limited by the following experimental factors: ability of the immobilization substrate to hold bacteria in place under higher flow rates, the microfluidic assembly sustaining high pressures without leaking, the liquid flow resistance inside the system, which can make the flow rate unstable or even halt it at lower flow rates. The strength of the binding was sufficient for the S. aureus experiments, but an increase in the flow rate or change of bacterial species may require a different coating type. The epoxy-coated slides are conventionally used for protein microarray analysis and are not specifically designed to hold large cellular objects. Since the formulation and density of the coating is proprietary, development of alternative formulations with the controlled surface coating deposition is highly desirable.
We have recently found that the amount of time required for bacterial culturing and attachment can be greatly reduced. Colonies can be grown to log phase directly from an agar plate in a small volume of media within three hours, eliminating the overnight culture step. Additionally, attachment can be accelerated by centrifuging the bacteria inside the flow cell for one minute, removing the 45 min settling time. Experiments in our laboratory are currently being run with this abbreviated protocol.
Although the protocol has been specifically designed for antibiotics targeting bacterial cell wall biosynthesis, recent studies indicate that antibiotics with different primary targets in bacterial cells activate the same downstream bacterial response pathways17,18. Therefore, our method may be applicable for the rapid identification of susceptibility to antibiotics that target biosynthetic pathways unrelated to cell wall biosynthesis, and initial studies in our laboratory lend credence to this hypothesis.
The authors have nothing to disclose.
We thank the engineers and students at the Fraunhofer Center for Manufacturing Innovation. For helping in the design, machining, and automation of the experimental system, we thank Andreas Prinzen, Holger Wirz, Doug Foss, David Chargin, and Dr. Sudong Shu. We thank Julia Kuckartz, Melanie Zimmermann, Niko Kraetzmar, Tim Gumbel, Josh Villanueva, Minori Shimizu, and Katarzyna Kuliga for help with testing experimental protocols and data collection. We acknowledge Drs. Anne E. Carpenter and Mark-Anthony Bray of the Imaging Platform at the Broad Institute of Harvard and MIT for help with development of the image analysis routine in CellProfiler. The project described was supported in part by Awards R21AI079474 and 1R01AI101446 from the National Institute of Allergy and Infectious Diseases. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Allergy and Infectious Diseases or the National Institutes of Health. The project was also supported by Fraunhofer USA.
SYTOX Green | Invitrogen Corporation | S7020 | Dead cell fluorescence stain |
Bovine Serum Albumin (BSA) | Sigma Aldrich, Inc | A9418-5G | Used for lysostaphin storage |
Sodium Acetate | Sigma Aldrich, Inc | S8750-500G | Used for lysostaphin storage |
Lysostaphin | Cell Sciences | CRL309A | Arrives as 1 mg solid. For storage: Dissolve in 20 mM sodium acetate. Mix with BSA solution to final concentration of 1% BSA and 100 µg/ml lysostaphin for storage |
Oxacillin salt | Sigma Aldrich, Inc | 28221-1G | Antibiotic |
Mueller Hinton Broth | Fisher Scientific | DF0757-17-6 | |
Sodium chloride | Sigma Aldrixh | S3014-500G | 2% added to Mueller-Hilton broth prior to autoclaving |
1 ml, Luer-lock syringe | BD (Beckton, Dickinson and Comp.) | 14-823-2F | |
2 oz, Luer-lock syringe | BD (Beckton, Dickinson and Comp.) | 309653 – 60 mL | Overfill to ~65 ml |
Microscope | |||
Inverted Fluoresccence Microscope Olympus IX-70 | Cambridge Scientific | 9349 | |
60X, Fluorescence/Phase contrast objective | Olympus Corp. | LCPlan F1 60x/0.70 Ph2 | |
Retiga 12-bit monochrome CCD camera | QImaging | RET-4000R-F-M-12-C | |
Microscope automation | |||
Shutters phase contrast/fluorescence | PRIOR Scientific | H204/H202 | |
X/Y Stage | PRIOR Scientific | H107AENN | |
Focus motor | PRIOR Scientific | H122 | |
Joystick for XYZ control | PRIOR Scientific | CS152EF | |
Proscan Controller | PRIOR Scientific | H3-XY2 | |
Image Acquisition Software | Fraunhofer CMI | ||
Flow Cell Assembly and PDMS | |||
Flow Cell | BU Scientific Instruments Facility/Fraunhofer CMI | 3333-1044 | Engineering drawings were produced by Fraunhofer CMI |
Glass window | Fraunhofer CMI | 3333-1054 | Glass window was cut to the proper size at Fraunhofer CMI |
BOROFLOAT Window 50 mm x 50 mm | Edmund Optics Inc. | NT48-543 | |
Sealing plate | BU Scientific Instruments Facility | 3333-1045 | |
Epoxide glass slide | Arrayit Corporation | SuperEpoxy 2 | |
PDMS master | Fraunhofer CMI | 3333-1053 | Master machined in aluminum or brass with UPM-0005 (ultrapresicion fly-cutting machine) |
PDMS slide design | Fraunhofer CMI | 3333-1053 | |
Tubing | |||
Nut, Super flangeless Tinytight, headless, 1/16 in, PEEK, green | IDEX Health & Science | M-644-03 | Flow cell inputs/outputs are tapped for this ferrule |
Ferrule, Tinytight, 1/16 in, 6-40, .030 in TH, PEEK w/ SS lock ring, black | IDEX Health & Science | M-657 | |
Nut, Super flangeless Tinytight, headless, 1/16-1/32 in, 1/4-28, PEEK, natural | IDEX Health & Science | P-255 | |
Ferrule, Super Falngeless, 1/16 in, Tefzel (ETFE), yellow | IDEX Health & Science | P-259 | Fits Luer-lock adapter |
Tubing, Teflon FEP, .030 in x 1/16 in x 20 ft, green | IDEX Health & Science | 1520G | |
Adapter, quick connect female Luer to female 1/4-28, PEEK, red | IDEX Health & Science | P-658 |