INTRODUCTION

Section:

ChooseTop of pageABSTRACTINTRODUCTION <<RESULTS AND DISCUSSIONCONCLUSIONMETHODSSUPPLEMENTARY MATERIALREFERENCESA range of microfluidic devices and methods have been developed to cultivate and study bacteria1,21. F. Wu and C. Dekker, Chem. Soc. Rev. 45, 268 (2016). https://doi.org/10.1039/C5CS00514K2. A. K. Wessel, L. Hmelo, M. R. Parsek, and M. Whiteley, Nat. Rev. Microbiol. 11, 337 (2013). https://doi.org/10.1038/nrmicro3010 as micro- (and nano-) structures are well-suited to the tasks of capturing and manipulating small, highly motile, and rapidly multiplying bacteria. Some strategies include confining bacteria to growth in either one33. M. R. Bennett and J. Hasty, Nat. Rev. Genet. 10, 628 (2009). https://doi.org/10.1038/nrg2625,44. P. Wang, L. Robert, J. Pelletier, W. L. Dang, F. Taddei, A. Wright, and S. Jun, Curr. Biol. 20, 1099 (2010). https://doi.org/10.1016/j.cub.2010.04.045 or two dimensions.55. A. Grünberger, N. Paczia, C. Probst, G. Schendzielorz, L. Eggeling, S. Noack, W. Wiechert, and D. Kohlheyer, Lab Chip 12, 2060 (2012). https://doi.org/10.1039/c2lc40156h So-called “mother machine” devices for one-dimensional growth have been particularly useful for studying cell dynamics and tracking lineages44. P. Wang, L. Robert, J. Pelletier, W. L. Dang, F. Taddei, A. Wright, and S. Jun, Curr. Biol. 20, 1099 (2010). https://doi.org/10.1016/j.cub.2010.04.045 and can even be used for single-cell phenotypic and genotypic analysis of heterogeneous populations.66. S. Luro, L. Potvin-Trottier, B. Okumus, and J. Paulsson, Nat. Methods 17, 93 (2020). https://doi.org/10.1038/s41592-019-0620-7 Bacteria can also be embedded within a hydrogel matrix,77. J. Heo, K. J. Thomas, G. H. Seong, and R. M. Crooks, Anal. Chem. 75, 22 (2003). https://doi.org/10.1021/ac0259717 which restricts motility and helps to contain progeny while still allowing for the delivery of nutrients through diffusion. Another approach has been to encapsulate bacteria within droplets,88. T. S. Kaminski, O. Scheler, and P. Garstecki, Lab Chip 16, 2168 (2016). https://doi.org/10.1039/C6LC00367B which has the advantage of generating thousands of individual cultivation chambers with general ease. However, it can be difficult to alter the contents of the droplets, thus making it difficult to perform experiments that investigate transient dosing of compounds. One method to overcome this is the use of bacteria encapsulated in hydrogel droplets. After gelification, the hydrogel droplets (or beads) can be transferred in and out of different solutions and the beads are also compatible with flow cytometry.99. Y.-J. Eun, A. S. Utada, M. F. Copeland, S. Takeuchi, and D. B. Weibel, ACS Chem. Biol. 6, 260 (2011). https://doi.org/10.1021/cb100336p While flow cytometry offers incredibly high-throughput analysis, it only offers a single snapshot in time and does not easily lend itself to tracking individual cells or colonies over time. Despite the many proposed microfluidic devices, those that are simple and easy to use tend to not be suitable for multiplexing. On the other hand, those that are powerful at multiplexing across different conditions and in a time-dependent manner tend to be too complex. Therefore, a middle ground with modest multiplexing capabilities arising from minimal complexity would be useful.Lately, much attention has been given to the development of miniaturized platforms for rapid testing of antibiotic resistances and susceptibility.10–1310. J. Choi, Y. Y. G. Jung, J. Kim, S. Kim, Y. Y. G. Jung, H. Na, and S. Kwon, Lab Chip 13, 280 (2013). https://doi.org/10.1039/C2LC41055A11. J. Choi, J. Yoo, M. Lee, E.-G. Kim, J. S. Lee, S. Lee, S. Joo, S. H. Song, E.-C. Kim, J. C. Lee, H. C. Kim, Y.-G. Jung, and S. Kwon, Sci. Transl. Med. 6, 267ra174 (2014). https://doi.org/10.1126/scitranslmed.300965012. N. Qin, P. Zhao, E. A. Ho, G. Xin, and C. L. Ren, ACS Sens. 6, 3 (2021). https://doi.org/10.1021/acssensors.0c0217513. P. Jusková, S. Schmitt, A. Kling, D. G. Rackus, M. Held, A. Egli, and P. S. Dittrich, ACS Sens. 6, 2202 (2021). https://doi.org/10.1021/acssensors.1c00020 For instance, microfluidic tools for bacterial classification14,1514. D. Zhang, H. Bi, B. Liu, and L. Qiao, Anal. Chem. 90, 5512 (2018). https://doi.org/10.1021/acs.analchem.8b0039915. A. A. Sklavounos, C. R. Nemr, S. O. Kelley, and A. R. Wheeler, Lab Chip 21, 4208 (2021). https://doi.org/10.1039/D1LC00609F as well as studying bacteria under continuous concentration gradients1616. S. Kim, S. Lee, J.-K. Kim, H. J. Chung, and J. S. Jeon, Biomicrofluidics 13, 014108 (2019). https://doi.org/10.1063/1.5066558 and combinations of antibiotics1717. E. Sweet, B. Yang, J. Chen, R. Vickerman, Y. Lin, A. Long, E. Jacobs, T. Wu, C. Mercier, R. Jew, Y. Attal, S. Liu, A. Chang, and L. Lin, Microsyst. Nanoeng. 6, 92 (2020). https://doi.org/10.1038/s41378-020-00200-7,1818. M. Azizi, B. Davaji, A. V. Nguyen, S. Zhang, B. Dogan, K. W. Simpson, and A. Abbaspourrad, ACS Sens. 6, 1560 (2021). https://doi.org/10.1021/acssensors.0c02428 have been reported. These tools can be useful in minimizing antibiotic resistance by reducing the time that inappropriate or broad-spectrum antibiotic therapies are administered. However, these tools study bacterial death under steady-state dosing, which does not necessarily represent in vivo conditions where antibiotic concentrations may fluctuate.1919. M. Müller, A. Dela Peña, and H. Derendorf, Antimicrob. Agents Chemother. 48, 1441–1453 (2004). https://doi.org/10.1128/AAC.48.5.1441-1453.2004 Most importantly, the persistence of bacteria against antibiotics has been linked with transient exposure to antibiotics.2020. N. Q. Balaban, J. Merrin, R. Chait, L. Kowalik, and S. Leibler, Science (80-.) 305, 1622 (2004). https://doi.org/10.1126/science.1099390 To account for the concentration profiles of antibiotics in vivo, pharmacokinetic/pharmacodynamic models have been developed.2121. J. Gloede, C. Scheerans, H. Derendorf, and C. Kloft, J. Antimicrob. Chemother. 65, 186 (2010). https://doi.org/10.1093/jac/dkp434 Among them, so-called hollow fiber infection models enable long-term monitoring of drug absorption and elimination in vitro. Cells are retained in a compartment of 10–20 ml and medium, and the drug is supplied via multiple porous fibers. The compartments are, however, large scale, i.e., 10–20 ml in volume, and pumping rates of 60–120 ml are applied.2222. T. Velkov, P. J. Bergen, J. Lora-Tamayo, C. B. Landersdorfer, and J. Li, Curr. Opin. Microbiol. 16, 573 (2013). https://doi.org/10.1016/j.mib.2013.06.010,2323. J. J. S. Cadwell, Adv. Pharmacoepidem. Drug Saf. S1, 007 (2012). https://doi.org/10.4172/2167-1052.S1-007 In addition, the hollow fiber cartridges are usually not suitable for observing cells under a microscope.Here, we present a microfluidic device for bacteria cultivation, capable of generating short-pulse dosing profiles of antibiotics. Ring-shaped pneumatic isolation valves define individual culture chambers of approximately 400 pl in volume and control exposure of the chamber's contents to the surrounding media. Similar isolation valve devices have been previously used where hydrodynamic traps have captured mammalian cells reliably.17–1917. E. Sweet, B. Yang, J. Chen, R. Vickerman, Y. Lin, A. Long, E. Jacobs, T. Wu, C. Mercier, R. Jew, Y. Attal, S. Liu, A. Chang, and L. Lin, Microsyst. Nanoeng. 6, 92 (2020). https://doi.org/10.1038/s41378-020-00200-719. M. Müller, A. Dela Peña, and H. Derendorf, Antimicrob. Agents Chemother. 48, 1441–1453 (2004). https://doi.org/10.1128/AAC.48.5.1441-1453.2004 However, hydrodynamic traps are less suitable for trapping the much smaller bacterial cells and challenging to fabricate with sub-micrometer gap sizes as required for bacteria.2424. P. Jusková, Y. R. F. Schmid, A. Stucki, S. Schmitt, M. Held, and P. S. Dittrich, ACS Appl. Mater. Interfaces 11, 34698 (2019). https://doi.org/10.1021/acsami.9b12169

Therefore, we use here plugs of agarose hydrogel formed within the isolation valves to immobilize bacteria within the culture chamber and prevent their escape during media exchange. These plugs entrap the bacteria but enable exchange of nutrients, antibiotics, stains, and potentially other reagents. We demonstrate the ability to cultivate Escherichia coli in hydrogel plugs in the microfluidic device and measured the steady-state metric of minimum inhibitory concentration (MIC) for ampicillin with E. coli K12 MG1655 (pSEVA271-sfgfp). However, the true value of the presented microfluidic device is its ability to administer transient and time-dependent dosing profiles, which we demonstrate by supplying short-term pulses at concentrations above the MIC. Although ampicillin's bactericidal activity is characterized as being time-dependent in the steady state when dosed at concentrations greater than the MIC, we observed that, for short pulses, its activity appears to be dependent on duration and concentration.

RESULTS AND DISCUSSION

Section:

ChooseTop of pageABSTRACTINTRODUCTIONRESULTS AND DISCUSSION <<CONCLUSIONMETHODSSUPPLEMENTARY MATERIALREFERENCESThe administration of time-dependent dosing profiles of antibiotics to populations of bacteria requires the ability to exchange media without dislodging the bacteria. Our method, outlined in Fig. 1(a), uses a microfluidic device to cultivate bacteria in addressable isolation valves and control the exposure of the bacteria to the antibiotic. The device comprises 64 individual isolation valves that each confine a volume of approximately 400 pl and that are addressable in groups of eight, through connections to eight pressure lines. Bacteria were immobilized in hydrogel plugs formed within the valves to prevent them from dislodging during experiments. We used ultra-low gelling temperature agarose because it allowed us to load the bacteria pre-mixed with the hydrogel into the microfluidic device; the risk of cell damage associated with UV-crosslinking or the added complexity of additional crosslinking reagents was avoided. Individual hydrogel plugs were formed within the isolation valves as illustrated in Fig. 1(b). A mixture of bacteria and liquid agarose is loaded into the microfluidic device. The isolation valves are then closed while excess agarose is cleared out with wash buffer before placing the microfluidic device on ice. After gelation, the valves can be opened, revealing a plug of agarose with immobilized bacteria. The valve can be closed forming a contact seal with the glass as shown in Fig. 1(c) to isolate the cells during experiments.The agarose gel plugs do not interfere with bacteria cultivation within the isolation valve and they keep the bacteria in place when the valves are opened for reagent exchange with the surrounding channel. Growth of sfgfp-expressing E. coli within agarose plugs was monitored by fluorescence microscopy [Fig. 2(a)]. E. coli appear to remain in the position where they were initially seeded during gelation, and colonies grow from these individual immobilized bacteria. Because of the agarose gel's porosity, it is possible to introduce reagents from the surrounding microchannel by opening the isolation valves. We monitored bacterial growth using the alamarBlue viability reagent (resazurin) and sfGFP fluorescence, which served as an arbiter for biomass13–2513. P. Jusková, S. Schmitt, A. Kling, D. G. Rackus, M. Held, A. Egli, and P. S. Dittrich, ACS Sens. 6, 2202 (2021). https://doi.org/10.1021/acssensors.1c0002024. P. Jusková, Y. R. F. Schmid, A. Stucki, S. Schmitt, M. Held, and P. S. Dittrich, ACS Appl. Mater. Interfaces 11, 34698 (2019). https://doi.org/10.1021/acsami.9b1216925. T. E. Gorochowski, E. van den Berg, R. Kerkman, J. A. Roubos, and R. A. L. Bovenberg, ACS Synth. Biol. 3, 129 (2014). https://doi.org/10.1021/sb4001245 [Fig. 2(b)]. Both alamarBlue and sfGFP fluorescence track together demonstrate the viability of the growing E. coli population. For subsequent experiments, sfGFP fluorescence was chosen as an indicator of E. coli growth.The ability to generate time-dependent dosing profiles can be useful in elucidating mechanisms of antibiotic persistence. We sought to create such profiles in our microfluidic device by using isolation valves to control the exposure of hydrogel plugs to the surrounding medium. Antibiotic exposure and clearance are dependent on diffusion in and out of the hydrogel. Therefore, we characterized diffusion in and out of the agarose plugs by monitoring the increase or decrease in fluorescence intensity of a fluorophore into or out of the plug, respectively. Sulforhodamine B (SRB), 580.65 Da, is of comparable size to the antibiotic ampicillin, 349.41 Da, and was chosen to give an indication of antibiotic transport into and out of hydrogel plugs. Figure 3 reports the fluorescence intensity of SRB within the isolation valve as it either diffuses into or out of the hydrogel. Maximum and minimum fluorescence intensities were observed after approximately 20 s. For later experiments, we used 1 min as the minimum time for keeping the isolation valve open when adding or removing antibiotics.Antibiotic susceptibility is typically determined by measuring MIC of an antibiotic on a bacteria culture and is used to identify resistance to different antibiotics and doses are often given in reference to the MIC.2626. G. Kahlmeter, D. F. J. Brown, F. W. Goldstein, A. P. MacGowan, J. W. Mouton, A. Österlund, A. Rodloff, M. Steinbakk, P. Urbaskova, and A. Vatopoulos, J. Antimicrob. Chemother. 52, 145 (2003). https://doi.org/10.1093/jac/dkg312 Though there are various methods and tools for measuring MIC,2727. B. Behera, G. K. Anil Vishnu, S. Chatterjee, V. S. N. Sitaramgupta V, N. Sreekumar, A. Nagabhushan, N. Rajendran, B. H. Prathik, and H. J. Pandya, Biosens. Bioelectron. 142, 111552 (2019). https://doi.org/10.1016/j.bios.2019.111552 it is typically performed by cultivating bacteria overnight with a serial dilution of antibiotic in a microtiter well plate. This is a test that is very well-suited to our platform as the valves enable different hydrogel plugs to be incubated with different concentrations of antibiotics on the same chip. Furthermore, we sought to measure the on-chip MIC to identify if small volumes and volume excluded by the hydrogel affected the MIC. E. coli were cultivated on-chip in agarose plugs. Ampicillin in varying concentrations (0–16 μg/ml) was delivered to the plugs, which were then isolated by the valves for 8 h. sfGFP fluorescence was monitored over that same period and used to determine bacterial growth, as shown in Fig. 4. As ampicillin is bacteriolytic,2828. P. I. Rafailidis, E. N. Ioannidou, and M. E. Falagas, Drugs 67, 1829 (2007). https://doi.org/10.2165/00003495-200767130-00003 sfGFP fluorescence only provides information on colony growth; we expect sfGFP from lysed cells to remain in the isolation valves until cleared. We found, after 8 h, the MIC of ampicillin to be 4 μg/ml. This is consistent with our own microtiter plate-based MIC measurements (Fig. S1 in the supplementary material) as well as literature. E. coli MG1655 has a reported MIC for ampicillin of 8 μg/mL using a conventional microtiter plate method,2929. Y. Tashiro, H. Eida, S. Ishii, H. Futamata, and S. Okabe, Microbes Environ. 32, 40 (2017). https://doi.org/10.1264/jsme2.ME16121 and microfluidic methods have reported an MIC of 4–8 μg/ml.3030. G. Pitruzzello, S. Thorpe, S. Johnson, A. Evans, H. Gadêlha, and T. F. Krauss, Lab Chip 19, 1417 (2019). https://doi.org/10.1039/C8LC01397GOur main goal was to develop a means of observing the effects of transient dosing of antibiotics on bacteria. The antibacterial effects of antibiotics can be categorized by pharmacokinetic and pharmacodynamic parameters (i.e., serum concentration and minimum bactericidal concentration, respectively) into two broad categories: time dependent and concentration dependent.3131. M. Mueller, A. de la Peña, and H. Derendorf, Antimicrob. Agents Chemother. 48, 369 (2004). https://doi.org/10.1128/AAC.48.2.369-377.2004 In the former, the duration that the dose remains above the MIC is the critical parameter; in the latter, it is the area under curve (AUC) where the antibiotic remains above the MIC. Ampicillin belongs to the time-dependent class of antibiotics where time-above-MIC is the critical parameter. Here, we begin to explore whether this time-dependency holds true at very short durations (10–20 mins) of high concentration doses (approx. 6–25 × MIC) for which the hydrogel-isolation valve platform is well suited to address.We compared exposing E. coli in agarose hydrogel plugs to 100, 50, 25, and 0 μg/ml of ampicillin for either 10 or 20 min and then monitoring growth over the subsequent 3 h. The maximum killing rate for time-dependent antibiotics has been found to occur at four to five times the MIC,3232. C. Nightingale, J. Int. Med. Res. 8(Suppl 1), 2 (1980).,3333. E. Yourassowsky, M. P. Van der Linden, and F. Crokaert, Chemotherapy 35, 423 (1989). https://doi.org/10.1159/000238706 which is comparable to the lowest concentration of ampicillin we chose. Thus, it would be expected that there should be no difference in the bactericidal effect between 10 and 20 min doses. However, our results, shown in Fig. 5, suggest that at such short durations, the bactericidal response depends on both concentration and duration (i.e., AUC). For a 10 min dose, only 100 μg/ml of ampicillin seemed to show a bactericidal effect (p μg/ml appeared to have the bactericidal effect as one-way ANOVA and Tukey's post-hoc test did not reveal a significant difference (p = 0.9995). From an AUC perspective, 100 μg/ml ampicillin for 10 min and 50 μg/ml ampicillin for 20 min are equivalent, as both dose profiles resulted in inhibition of bacterial growth. There was also an increase in the bactericidal activity for a concentration of 25 μg/ml ampicillin when exposed for 20 min but not to the same degree as the two higher concentrations. Indeed, it seems to be a critical exposure concentration and duration where cells eventually still grow.The combination of hydrogels for cell culture and isolation valves to control reagent exposure has allowed us to dose antibiotics in short pulses. This is a meaningful advantage over other microfluidic techniques for studying antibiotic susceptibility and persistence. To our knowledge, the only previous report to study transient dosing of antibiotics exposed bacteria to the antibiotic for periods on the order of hours;2020. N. Q. Balaban, J. Merrin, R. Chait, L. Kowalik, and S. Leibler, Science (80-.) 305, 1622 (2004). https://doi.org/10.1126/science.1099390 here, we present dosing on the order of tens of minutes. For instance, while droplet-based techniques can reveal heterogeneity among populations of bacteria and do so with high throughput,3434. O. Scheler, K. Makuch, P. R. Debski, M. Horka, A. Ruszczak, N. Pacocha, K. Sozański, O.-P. Smolander, W. Postek, and P. Garstecki, Sci. Rep. 10, 3282 (2020). https://doi.org/10.1038/s41598-020-60381-z adding and then removing antibiotics are difficult. Encapsulating bacteria within hydrogel beads or spheres3535. G. Leonaviciene, K. Leonavicius, R. Meskys, and L. Mazutis, Lab Chip 20, 4052 (2020). https://doi.org/10.1039/D0LC00660B provides a possible solution, as gel beads can readily be exchanged from one solution to another. Once the immiscible phase used to generate the gel beads has been removed and the beads have been washed, such gel beads could conceivably be loaded into a microfluidic device for reagent exchange. However, our proposed solution offers a single microfluidic device in which immobilization and antibiotic dosing are executed in a straightforward manner. With the device and method presented here, it is possible to provide short doses of antibiotics, track the cells over time, and compare multiple conditions on the same microfluidic device. While we relied on E. coli constitutively expressing sfGFP for quantitative measurements, it is possible to adapt this platform to non-fluorescent strains of bacteria. We demonstrated compatibility with fluorescent dyes for viability (i.e., alamarBlue), but it might also be possible to visually estimate bacteria growth with brightfield microscopy.3636. K. Kim, S. Kim, and J. S. Jeon, Sensors 18, 447 (2018). We also explored whether it was possible to recapitulate our microscale method on the macroscale by immobilizing E. coli in a thin layer of agarose on the bottom of a 96-well plate. While a 20 min exposure could be executed and could achieve similar results (Fig. S2 in the supplementary material), it was difficult to perform all pipetting steps within a shorter time frame (e.g., 10 min). Furthermore, controlling the thickness of the agarose layer in the bottom of a microtiter plate proved difficult, whereas the microfluidic isolation valves provided a consistent geometry of hydrogel.

METHODS

Section:

ChooseTop of pageABSTRACTINTRODUCTIONRESULTS AND DISCUSSIONCONCLUSIONMETHODS <<SUPPLEMENTARY MATERIALREFERENCES

Unless otherwise stated, all reagents were purchased from Sigma Aldrich (Saint Louis, MO, USA).

Microfluidic device fabrication

A microfluidic device with ring-shaped isolation valves was used to form hydrogel plugs, cultivate bacteria, and deliver antibiotics with a time-dependent dosing profile. The device comprised two layers, a bottom fluidic layer and a top pneumatic layer. In the bottom layer, a single inlet was divided into eight parallel channels, which merged together into a single outlet. The pneumatic layer contained eight sets of eight ring-shaped isolation valves to create a total of 64 isolation chambers. The inner diameter of the isolation valves was 175 μm, confining a volume of approximately 400 pl. The device was made from poly(dimethylsiloxane) (PDMS) (Dow Chemical Company, Midland, MI, USA) by mixing the elastomer with a curing agent in a 10:1 ratio (Sylgard 184, Dow Corning, Midland, MI, USA). Both bottom and top layers were formed over SU-8 (MicroChem, Westborough, MA, USA) moulds on silicon wafers with a feature height of 17 μm. For the bottom layer, approximately 2–3 g of the PDMS mixture was spin-coated on an SU-8 mould resulting in an ∼30 μm thick layer. The top layer was formed by pouring ∼40 g of PDMS onto an SU-8 mould. Both wafers were baked at 80 °C for 2 h. Using a razor blade, the top layer was cut into individual devices and eight 1 mm holes were punched for the pneumatic channels. The top and bottom layers were then aligned and bonded together with an intermediate layer of the curing agent; the bonded PDMS layers were then baked at 80 °C for 2 h. Individual devices were cut out and 1.5 mm holes were punched for inlets and outlets to the fluidic layer. The PDMS devices were bonded to No. 2 thickness (170–250 μm) cover glass (Menzel, Fisher Scientific, Hampton, NH, USA) by plasma bonding.

Microfluidic device operation

Prior to use, the fluidic channels were filled with 4% (w/v) bovine serum albumin (BSA) in phosphate-buffered saline (PBS; pH 7.4) and the pneumatic channels were filled with de-ionized H2O by centrifugation. The microfluidic devices were mounted onto the stage of an automated inverted microscope (Nikon Ti2, Nikon, Tokyo, Japan) enclosed in an environmental chamber. Temperature was set to 37 °C at the start of all experiments. A rack of 8 three-way stopcock luerlock valves (Cole-Parmer, Vernon Hills, IL, USA) connected in series was used to connect a single pressurized air supply (2 bar) to the eight pneumatic channels. This allowed each channel to be operated independently. Polytetrafluorethyl (PTFE) tubing (inner diameter 0.25 mm, outer diameter 1.59 mm) was connected directly to the inlet of the microfluidic chip and to a 1 ml plastic syringe. A programmable syringe pump (neMESYS from Cetoni GmbH, Korbussen, Germany, or Aladdin AL-1000 from World Precision Instruments, Friedberg, Germany) was used to control the syringe.

Bacterial culture

E. coli strain K12 MG1655 transformed with pSEVA271-sfgfp2424. P. Jusková, Y. R. F. Schmid, A. Stucki, S. Schmitt, M. Held, and P. S. Dittrich, ACS Appl. Mater. Interfaces 11, 34698 (2019). https://doi.org/10.1021/acsami.9b12169 was used for the characterization of the microfluidic device. The strain overexpresses super folding green fluorescent protein (sfGFP), enabling quantitation of bacteria. Bacteria were streaked onto lysogeny broth (LB) agar plates (BD Difco LB agar, Miller, Thermo Fisher Scientific, Waltham, MA, USA), and single colonies were cultivated in 2 ml LB culture media supplemented with 50 μg ml−1 kanamycin sulfate. Liquid cultures were grown at 37 °C and with a shaking speed of 200 rpm in a shaking incubator overnight. The following day, bacteria were diluted 1:10 in fresh media and grown until optical density at 600 nm (OD600) > 0.5.

Hydrogel plug formation

Agarose hydrogel plugs were formed within the microfluidic device as summarized in Fig. 1(b) and described below. A 3% (w/v) agarose solution was prepared by dissolving ultra-low gelling temperature agarose in sterile-filtered PBS and heating at 80 °C with 1000 rpm shaking in a ThermoMixer (Eppendorf, Hamburg, Germany) until dissolved. The hydrogel was then cooled to 40 °C and diluted to 2.85% (w/v) by the addition of a suspension of E. coli, buffer, or solution of fluorophore. Agarose was loaded into the microfluidic device from a reservoir (200 μl pipet tip) by withdrawing the syringe at a rate of 2 μl min−1. The valves were closed and wash buffer (100 mM NaCl, 10 mM Tris base, 1 mM EDTA, and 0.1% (v/v) Tween-20) was injected into the chip at 2 μl min−1 for 20 min. Temperature control was turned off and the microfluidic device was placed on an aluminum slide (25 × 75 × 2 mm3) on ice with a continuous flow (2 μl min−1) of either wash buffer or, for experiments with bacteria, LB containing 0.1% (v/v) Tween-20. After 30 min, the microfluidic device was removed and the temperature of the environmental enclosure set to 37 °C.

Diffusion experiments

A 2.85% (w/v) agarose gel was prepared as above. For experiments monitoring the diffusion of SRB out of the hydrogel plug, the agarose solution also contained 10 μM SRB. Hydrogel plugs were then prepared in the isolation valves on the microfluidic device, as described above. For monitoring diffusion into the hydrogel plug, a 10 μM solution of SRB in PBS was injected for 5 min at a flow rate of 2 μL min−1; for monitoring diffusion out of the hydrogel plug, the channel contained PBS without any SRB. After resting for 1 min, the valves were opened. Fluorescence images were taken every 1.5 s for 37.5 s.

On-chip cell cultivation and imaging

All on-chip cell cultivation experiments began by opening the valves for 1 min and exposing the E. coli to a fresh LB medium. Bacteria were then cultivated with closed valves and the LB medium containing 0.1% (v/v) Tween-20 was perfused through the chip at a rate of 1 μl min−1. Brightfield and fluorescence images were captured using a Nikon Ti2 automated inverted microscope equipped with a Nikon DS-Qi2 CMOS camera and a SOLA II LED light source for fluorescence excitation. GFP fluorescence was measured with a 470/40 nm excitation filter, a 500 nm dichroic mirror, a 535/50 nm barrier filter, and an exposure time of 50 ms at 10% LED intensity. alamarBlue fluorescence was measured with a 560/40 nm excitation filter, a 595 nm dichroic mirror, a 630/60 nm barrier filter, and an exposure time of 100 ms at 10% LED intensity. SRB fluorescence was measured with a 540/25 nm excitation filter, a 565 nm dichroic mirror, a 605/55 barrier filter, and an exposure time of 50 ms at 10% LED intensity. A 10×  objective was used.

MIC determination

A simple off-chip MIC assay was performed in a 96-well plate. A stock solution of ampicillin (512 μg ml−1) was serially diluted 1:2 in LB containing 50 μg ml−1 kanamycin and 0.1% Tween-20. An overnight pre-culture of E. coli was diluted 1:100 to OD600:0.1 and added in equal volume to the ampicillin-containing wells. Positive (only E. coli) and negative (only media) controls were also prepared. The 96-well plate was incubated overnight at 37 °C without shaking. alamarBlue (Thermo Fisher) was prepared as a 1:10 dilution in LB containing 50 μg ml−1 kanamycin and 0.1% Tween-20 from the stock and 30 μl was added to each well. The 96-well plate was incubated for a further 2 h at 37 °C without shaking before measuring GFP and alamarBlue fluorescence on a Cytation 5 plate reader (BioTek Instruments Inc., Sursee, Switzerland).

An on-chip MIC assay was performed by injecting LB with 0.1% Tween-20 and 0, 2, 4, 8, and 16 μg ml−1 ampicillin using a syringe pump operating at 2 μl min−1 into the microfluidic device prepared with E.coli-laden hydrogel plugs, as described above. The isolation valves were opened to expose the selected hydrogel plugs to the desired concentration of ampicillin. Valves were opened for 1 min in the presence of the desired concentration of ampicillin and then closed for the remainder of the experiment. The LB medium with 0.1% (v/v) Tween-20 was perfused through the chip at a rate of 1 μl min−1. Each concentration of antibiotic was injected for 5 min before opening the selected isolation valves. The MIC was defined as the concentration that resulted in 99% inhibition of bacteria growth.

Pulsed dosing experiments

E.coli-laden hydrogel plugs were prepared as described above. Prior to pulsed dosing experiments, E. coli were incubated on-chip at 37 °C for 4 h after exposing to fresh LB media. Solutions of 25, 50, or 100 μg ml−1 ampicillin were prepared in LB supplemented with 50 μg ml−1 kanamycin and 0.1% Tween-20. For 10 min dosing times, a single chip was used for each concentration of ampicillin and included a control of no ampicillin administered to some of the hydrogel plugs. Ampicillin was administered by injecting the solution at 2 μl min−1 for 2 min, waiting for 1 min, opening the valve for 1 min, and closing the valve for 10 min. During this time, LB supplemented with 50 μg ml−1 kanamycin and 0.1% Tween-20 was then injected into the device at 2 μl min−1 for 5 min. Isolation valves for hydrogel plugs that were to be exposed to 0 μg ml−1 ampicillin were opened at this time for 1 min and closed for 10 min. After respective 10 min incubations, valves were opened for 1 min to expose the hydrogel plugs to fresh LB media and then closed again. The LB medium containing 0.1% (v/v) Tween-20 was perfused through the chip at a rate of 1 μl min−1. For 20 min incubation with ampicillin, the method was repeated much as described above, except all four concentrations (0, 25, 50, and 100 μg ml−1) of ampicillin were dosed on a single chip. In this case, imaging was done with a 4× objective and stitched with a 5% overlap. Ampicillin was added to the microfluidic chip in increasing concentration such that each concentration was added approximately 5 min after the previous. Likewise, the valves were subsequently washed with fresh LB media in 5 min intervals such that the whole procedure took approximately 35 min to deliver and clear all four concentrations of ampicillin. After the hydrogel plugs exposed to 100 μg ml−1 ampicillin had been washed with LB, valves were then closed and then brightfield and fluorescence images were taken every 30 min for 3 h. The LB medium containing 0.1% (v/v) Tween-20 was perfused through the chip at a rate of 1 μl min−1.

Data analysis

Fluorescence intensities were measured using Fiji image analysis software.4040. J. Schindelin, I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J.-Y. Tinevez, D. J. White, V. Hartenstein, K. Eliceiri, P. Tomancak, and A. Cardona, Nat. Methods 9, 676 (2012). https://doi.org/10.1038/nmeth.2019 SRB, FITC-Dextran, and alamarBlue fluorescence intensities were measured using mean fluorescence over the inner circle of the isolation valve. Bacteria growth via GFP was calculated by comparing the percentage change in integrated fluorescence intensity to t = 0 for each individual hydrogel plug measured. Microsoft Excel and GraphPad Prism 8 were used to process the data and generate plots. For comparing 10 min doses, multiple t-tests with the Holm-Šídák method was used to compare the antibiotic with the control, performed on the same chip. For 20 min doses, a one-way ANOVA with Tukey's multiple comparisons was used.