Ultra-miniaturized bioreactors show promise for bacterial surveillance

The RV broth, at both temperatures under test (37 and 41.5°C), effectively eliminated or drastically suppressed growth of non-S. Typhimurium species (Table 1). In traditional microbiological culturing using a selective medium, RV broth at 41.5°C would be superior in selecting for S. Typhimurium among the panel tested. Even at the sub-optimal temperature of 37°C, S. Typhimurium likely would be able to out-compete the only other bacteria species identified in this study that grow at a suppressed rate in RV broth, E. aerogens and C. freundii. In contrast to traditional selective methods, the objective of this study was to assess the feasibility for RV broth use in a microfluidic droplet. Only the S. Typhimurium under test successfully grew in RV broth at 41.5°C, preventing growth of all non-S. Typhimurium under test. At first glance, 41.5°C would be the temperature of choice given complete suppression of all non-S. Typhimurium species. However, the growth rate of S. Typhimurium was substantially reduced at 41.5°C, when compared to S. Typhimurium grown in RV at 37°C (Table 1). Using S. Typhimurium with a generation time of 91.2 minutes at 41.5°C as an example (Table 1), and assuming each droplet is starting with one S. Typhimurium cell, it would take 9 h to reach approximately 60 S. Typhimurium cells per droplet and 12 h to reach approximately 250 S. Typhimurium cells per droplet. For overnight culture purposes, the reduced generation time of RV at 41.5°C is not an issue of concern, since 18-to-24 h is utilized with inoculation of more than one S. Typhimurium cell. For this microfluidic droplet application, RV broth at an incubation temperature of 37°C was selected as the objective is to detect S. Typhimurium growth in 4–8 h.

As demonstrated by the collected data, a more prolific bacterial growth was supported during incubation in RV broth at 37°C as compared to 41.5°C. However, RV at 37°C does not suppress growth of both C. freundii and E. aerogens. In order to maximize selective growth rates in RV at 37°C, a fluorescently-labeled antibody specific for S. Typhimurium to increase specificity was utilized in conjunction with RV broth to screen out C. freundii and E. aerogens. Given fluorescently-labeled antibodies can create a relatively high detectable signal in the presence of their target antigen, and the multitude of fluorophores available for researchers to conjugate to their antibody of choice, this method proved to be the ideal way moving forward in creating a double positive indicator for S. Typhimurium. Masking of fluorescence must be taken into account, as these data indicate fluorescent intensities exhibit a log₁₀ reduction in RFU when diluted in RV broth (Table 2; analysis reported in SI). The reduced fluorescence intensity observed in RV broth is likely due to a pH of 5.3. Published findings have shown FITC to be pH-sensitive, exhibiting reduced fluorescence signal at a low pH [44,45]. It may be possible to increase specificity with an alternate fluorophore-antibody conjugate.

There are other fluorescent options commercially available for S. Typhimurium detection, however, not all options are applicable inside a droplet. Two of the more applicable fluorescent methods would be either a fluorescent medium or fluorescently-labeled antibody for S. Typhimurium. There is one fluorescent substrate, 4-methylumbelliferyl caprylate (MUCAP), for S. Typhimurium [46]. However, MUCAP is non-soluble in water, and thus is not compatible for aqueous in-droplet applications. For the purposes of this research, a fluorescently-labeled antibody specific for S. Typhimurium was a sound course of action when used in tandem with RV broth in-droplet. A direct binding antibody, such as an antibody targeting S. Typhimurium’s common surface antigen, could prove effective for detection purposes. A fluorescent indicator for detection purposes would be an ideal indicator for S. Typhimurium detection in-droplet, and a FITC-Ab could be applied in future microfluidic applications. It would be necessary to evaluate medium contributions to fluorescence in-droplet to best assess potential uses in the future. This research evaluated the impact medium can have on overall fluorescence (Fig 2; Table 2).

An additional aspect to consider when developing a method to rapidly detect S. Typhimurium would be incubation in-droplet. Some incubation would be necessary in order to reach optically detectable levels of colorimetric or fluorescence indicators. Current enumerating technologies are incapable of reliably detecting a single S. Typhimurium cell encapsulated inside a droplet on an enumeration or sorting microfluidic chip. This may seem to be a limiting factor, and in some ways, it is not ideal. However, using the selective properties of the RV broth can further help specificity of differentiating S. Typhimurium from non-S. Typhimurium in mixed samples. Current capabilities of the in-droplet microfluidic system necessitate an incubation time to achieve a detectable in-droplet bacterial cell concentration for enumeration or potentially sorting of the target bacterial organism in the future. The incubation step also differentiates viable versus non-viable cells as there is potential non-viable bacterial cells would still be capable of binding an antibody. The incubation step ensures single bacterial cells do not yield a false positive. By determining how many h it takes to achieve a detectable number of S. Typhimurium, in comparison to the non-S. Typhimurium, increased specificity can be achieved. For example, according to these data, C. freundii doubling time in RV broth at 37°C is over two times that of the S. Typhimurium strain under examination in this study (Table 1). The specificity of isolation and detection of S. Typhimurium can potentially be achieved given the suppressed growth state of C. freundii and relatively high replication rate of S. Typhimurium at 37°C. The E. aerogens performed considerably well in the RV broth at 37°C, and only had doubling times approximately 10 minutes longer than the S. Typhimurium strain under test in this study (Table 1). Therefore, given the narrow temperature window between 37°C and 41.5°C, it would be worthwhile for future research to address varying concentrations of mixed cultures with S. Typhimurium and non-S. Typhimurium, such as C. freundii and E. aerogens.

Finally, the U.S. Food and Drug Administration Bacteriological Analytical Manual (FDA BAM) states RV broth is selective for S. Typhimurium, and the data provided in this study reaffirms FDA guidance. The results also show RV broth has a precise temperature window and can be less selective if used at the incorrect temperature, i.e. less than 41.5°C. When used as directed at 41.5°C for incubation, RV broth is effective at suppressing non-S. Typhimurium growth. However, generation times are increased greatly for S. Typhimurium. It has already been noted S. Typhimurium would likely outcompete another serotype of bacteria in a mixed culture–at least the bacterial serotypes researched in the study. A more accurate determination for the concentration of S. Typhimurium can be achieved by carrying out future experiments of mixed cultures, and more effective selective media for S. Typhimurium could potentially be formally tested, certified, and eventually used by the scientific community. It is important to note that tetrathionate (TT) broth, which has been reported as an additional optional selective broth for S. Typhimurium by the FDA BAM [47], was not examined in this study due to the presence of a calcium carbonate precipitate that does not fully dissolve when the medium is prepared resulting in subsequent clogging or fouling of the microfluidic channels on-chip.

Initial in-droplet encapsulation experiments focused on assessing specificity of the FITC-Ab for S. Typhimurium, and the cross-reactivity with non-S. Typhimurium species. Potential medium contribution to fluorescence were eliminated by utilizing PBS for these experiments. Moreover, by suspending both bacterial cells and FITC-Ab in PBS, a more precise ratio of bacterial fluorescence to background could be assessed. A relatively high concentration of bacterial cells (approximately 100 cells) was generated in-droplet to better visualize cross-reactivity of the FITC-Ab with negative controls, and to provide ample sites for relative fluorescence measurements. The anti-S. Typhimurium FITC-Ab exhibited good sensitivity by binding to all encapsulated S. Typhimurium via S. Typhimurium’s CSA-1, and displayed robust relative fluorescence values (Fig 3). Deficiencies in specificity, however, are exposed by evaluating calculated relative fluorescence values and fluorescence images for the negative control panel (Fig 3). Cross-reactivity with the FITC-Ab and members of the negative control panel was anticipated as E. coli, E. aerogens, and S. Typhimurium are similar in many ways. All three are gram-negative, from the E. aerogensiaceae family, and can be found in the gastrointestinal tract of an animal host [48]. For this reason, use of a S. Typhimurium selective broth, RV, was then utilized to suppress growth of non-S. Typhimurium species in-droplet.

Assessment and evaluation of media identified by the FDA BAM was carried out, and RV broth was identified as the best candidate for in-droplet use to detect S. Typhimurium. Two key aspects must be noted for the use of RV broth in this study. RV broth exhibits fluorescence masking of the FITC-Ab, and this masking reduces relative fluorescence units by approximately one log₁₀ (reported in SI). Therefore, FITC-Ab concentrations were increased from 0.83 μg/ml to 10 μg/ml for in-droplet RV broth experiments carried out in this study. Also, S. Typhimurium had a much higher generation rate in RV broth at 37°C instead of the manufacturer-directed incubation temperature of 41.5°C (RV R10 Broth, Becton Dickinson) (Table 1).

RV broth performed as expected in suppressing most non-S. Typhimurium species under test, based on experiments outside of droplets. While growth was not completely inhibited, all E. coli species were noticeably suppressed by the presence of RV broth (reported in SI). One E. coli strain evaluated, ATCC 700891, was particularly suppressed. For instance, relative measurements were unattainable at time points 0 through 2 h. The FITC-Ab cross-reacted minimally with C. freundii in both PBS and RV broth with sterile deionized water. This is a promising observation as C. freundii is one of two bacterial species evaluated capable of growth in RV broth at 37°C.

In experiments utilizing PBS and RV broth in sterile deionized water, the FITC-Ab had moderate-to-high cross reactivity with E. aerogens (reported in S5 Fig). Moreover, E. aerogens grows well in RV broth at 37°C (Table 1). For this reason, E. aerogens was selected to be tested as the negative control when spiking the real-world shredded lettuce wash water acquired from the Producer. Fluorescent images with embedded metadata were captured every h and analyzed for a 5-h incubation of both S. Typhimurium and E. aerogens (Figs 7 and 8; E. aerogens images reported in S14 and S15 Figs). The wash water conditions under test were intended to simulate a highly contaminated sample, and encapsulate one bacterial cell per two-to-ten droplets generated. This allowed for monitoring of growth kinetics in-droplet over the five-hour time course. Both S. Typhimurium and E. aerogens had reduced relative fluorescence values in the real-world wash water experiments (Fig 7) when compared to experiments utilizing RV broth and sterile deionized water (S5 Fig).

The manufacturer of the RV broth formulation utilized in this study states the pH is 5.1 at 25°C (Becton Dickinson). Moreover, it has been reported that FITC fluorescence intensity has a direct correlation with pH gradients. Ma et al. explored the relationship between decreasing FITC intensity with decreases in pH [45]. Utilizing the figures reported by Ma et al., there is nearly a log₁₀ reduction in fluorescence intensity of FITC between a 7.5 pH solution and a 5.0 pH solution. This fluorescence relationship with pH could help explain the masking effect observed from RV broth when compared to other media (Table 2). A pH-related FITC intensity relationship is also something to consider in relation to the shredded lettuce wash water. The pH of the shredded lettuce wash water was 4.37 upon collection, and measured to have a pH of 5.97 after thawing from stock. A 20 ml volume of 50/50 2x RV broth and shredded lettuce wash water mixture (simulating in-droplet concentrations) was measured and had a pH of 5.26. This low pH likely explains the reduction in fluorescence intensity observed in the wash water experiments.

Bacterial growth in-droplet was also evaluated and compared to growth outside of droplets. Both S. Typhimurium and E. aerogens exhibited faster generation times and reduced observed lag times in-droplet (Table 3). More rapid droplet-based incubation growth rates have been reported in earlier studies, and is thought to be due to rapid mixing, reduced contaminants, reproducibility of droplets, and small volumes [43]. Lag times were minimal in-droplet as well. S. Typhimurium and E. aerogens initiated log phase growth between 1 and 2 h of incubation. These observed increased growth rates, and decreased lag times, further showcase microfluidic droplets as a rapid application for viable bacterial detection. It should be noted growth rates outside of droplets were calculated using turbidity measurements for growth, and these samples under test did not contain FITC-Ab.

Of note, this rapid microfluidic detection capability will need to be paired with a bacterial concentration technique. Recent published research highlights a range of concentration techniques that are potentially compatible with the droplet microfluidic system, such as continuous-flow centrifugation and ultrafiltration [49–52]. Follow-on laboratory testing and evaluation to determine the most suitable sample concentration technique is necessary to maximize the potential for this surveillance technique.

Why advancing food surveillance matters

Foodborne illness is an ever-present international threat to public health [1]. In the U.S. alone, 1 in 6 people will acquire foodborne illness annually and this national burden is estimated to cost between $55.5 and $93.2 billion [53]. Fresh-cut produce and leafy greens are often implicated in foodborne outbreaks, and it has been estimated leafy greens are the most common cause of foodborne outbreaks in the U.S. [54]. Developing and improving upon methods for surveillance and detection of food contamination that can be implemented in food industry HACCP control strategies are necessary to ensure safer food supplies in the future [13,55]. Moreover, developing novel methodologies to isolate, detect, and prevent bacterial outbreaks are essential to reduce mortality, illness, hospitalization, associated cost to taxpayers and overall international burden [56]. When deployed systematically with emerging engineering technologies, applications using in-droplet microfluidics show great promise for rapid, same-day/same-shift, and viable bacterial detection at CCP’s, such as fresh-cut produce wash waters.

Bacterial strains used, preparation, and protocols

A total of six ATCC bacterial strains, one S. Typhimurium positive control and five negative controls, were utilized (Table 4). All strains used were Biosafety Level 1 due to the microfluidic platform location outside of a biological safety containment hood. Stocks were generated from overnight culture and stored at -20°C. Working stocks at 10⁹ colony forming units (CFU)/ml in phosphate buffered saline (PBS) were stored at 5°C. Streak plates for isolated colonies were performed for each experiment on tryptic soy agar plates (bioMerieux) and incubated overnight at 37°C. Isolated colonies from overnight incubation were then diluted into PBS until an absorbance of 0.05 ± 0.005 at 660 nm, equating to 10⁹ CFU/ml. Following an absorbance confirmation of 0.05 ± 0.005 at 660 nm, a 1:10 dilution series was carried out to reach 10⁸ and 10⁶ CFU/ml, the concentrations ideally suited for microfluidic flow rates of encapsulation for control panel and Rappaport-Vassiliadis (RV) growth experiments, respectively. Control panel experiments utilized a bacterial input of 10⁸ CFU/ml. This bacterial input concentration, paired with microfluidic flow rates used, yielded approximately 100 bacterial cells encapsulated per droplet. RV broth growth kinetic experiments utilized a bacterial sample input of 10⁶ CFU/ml. This bacterial input concentration, paired with microfluidic flow rates used, yielded approximately one bacterial cell encapsulated every two-to-ten droplets. Bacterial dilutions and spread plate counts on tryptic soy agar plates (bioMerioux) were performed in shredded lettuce wash water growth experiments with S. Typhimurium and E. aerogens to identify length of lag phase, initiation of log growth phase, and generations time in minutes. Droplets with encapsulated bacteria were incubated at 37°C for the desired time and then droplets were lysed using 1H, 1H, 2H, 2H-Perfluoro-1-Octanol (PFO) (Sigma). Approximately 10–20 μl of PFO was used per sample with brief centrifugation. The amount of PFO used varied slightly based on volume of droplets captured. Once popped, the bacterial cells in PFO form an immiscible layer on top of the medium. An initial 1:100 dilution was performed due to the small volume of cells in PFO, and subsequent 1:10 dilutions were performed to achieve the desired spread plating CFU/ml concentration. Samples from time points zero and one h were assayed by spread plate in duplicate, and time points two through five were assayed by spread plate in triplicate.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Bacterial strains utilized in this study.

https://doi.org/10.1371/journal.pone.0233239.t004

Growth conditions, equipment, and measurement

Bacteria were grown for 24 h at either 37°C or 41.5°C (RV only) and turbidity measurements were taken every 15 minutes using an automated absorbance device (BioScreen C, Automated Microbiology Growth Curve Analysis System, Growth Curves USA, New Jersey, United States). Given one objective of this research was to examine the limiting capability of RV medium for non-S. Typhimurium bacterial species, all samples were grown at 37°C–the optimum growth temperature for the S. Typhimurium utilized. The control panel was also grown at 41.5°C in the RV broth for 24 h, per medium manufacturer instructions. Of note, optimal growth temperature for the E. aerogens strain used is 30°C. The starting cell concentration for all bacterial species was 10³ colony-forming unit CFU/ml and all bacterial species/strains studied were performed in triplicate.

Each bacterial species/strain used in this study was cultivated from stocks by an overnight culture. Streak plates for isolation were then performed for each bacterial species and allowed to grow overnight. Isolated bacterial colonies were then added to PBS until an absorbance of 0.05 ± 0.005 at 660nm was reached. An additional absorbance, 0.1 ± 0.05 at 600nm, was measured to ensure equipment was working properly. Starting with a 10⁹ CFU/ml suspension in PBS a 1:10 dilution series was repeated until reaching a 10⁵ CFU/ml concentration in PBS. Spread plating for CFU/ml counts were performed on tryptic soy agar plates (bioMerieux) in duplicate during this experiment to ensure CFU/ml concentrations were accurate, and all concentrations were found to have a starting concentration of 10⁹ CFU/ml. A 1:10 dilution series was performed in order to achieve 10⁵ CFU/mL concentration in PBS. A 1:100 dilution was then performed by transferring 10 μl of 10⁵ CFU/ml into 990 microliters of medium. A 1:100 dilution was performed at this stage to minimize the input of PBS, and thus not diluting the medium under test. A 200 microliter aliquot of 10³ CFU/mL bacterial concentration in medium were then added per well of the proprietary Bioscreen Honeycomb 100-well plate. Three replicates were performed per bacterial species/strain in respective medium.

Determination of generation time

Growth curves were generated for each bacterial strain, and stratified by each of the four conditions under test. The slope of the bacterial log-phase growth was utilized for the calculation of generation time. Length and slope of the log-phase varied across all bacteria. To calculate the slope, a log-linear trend line was fit to the log-phase, and an r² and natural log equation for each condition under test was attained.

Medium and antibody preparation, and pH measurements

RV broth was prepared at 2x concentration by adding double the manufacturer-directed amount in grams per unit volume (Beckton-Dickinson). Once mixed in filtered deionized water, the 2x RV broth mixture was autoclaved at 121°C for 15 minutes. An affinity purified polyclonal fluorescein isothiocyanate- (FITC)-conjugated antibody (Ab) targeting the common surface antigen- (CSA)-1 of S. Typhimurium was utilized for activated fluorescence detection (BacTrace, Anti-S. Typhimurium, CSA-1 Antibody, FITC-Labeled, SeraCare). FITC-Ab from stocks (0.5 mg/ml in PBS, 10 μl aliquots) were added to 2x RV broth to prepare a 20 μg/ml FITC-Ab concentration in 2x RV broth. A 2x RV broth with 20 μg/ml FITC-Ab was used for input on-chip as the concentration per droplet would be reduced by half upon droplet formation and bacterial encapsulation using the co-flow on-chip device, i.e. 1x RV broth with 10 μg/ml FITC-Ab upon droplet formation and bacterial encapsulation. The pH of RV broth medium was measured using a pH meter (VWR Symphony SB70P Digital, Bench-model pH Meter).

Fluorescent antibody and fluorescent measurement procedures

A FITC-conjugated antibody (FITC-Ab), anti-S. Typhimurium, CSA-1 Antibody (BacTrace), was utilized to measure medium (TSB, BPW, and RV) contribution to overall fluorescence at concentrations of 0.1, 1, and 10 μg/ml. Measurements were taken by fluorimeter (Tecan) in a 96-well black bottom plate at 488 nm excitation and 525 nm emission. The FITC-Ab was also merged into droplets at a 1:300 concentration from FITC-Ab stocks (0.5 mg/ml). Once merged into the droplet, at a 50/50 ratio with the complimentary input sample, the ratio in-droplet was 1:600 FITC-Ab (0.83 μg/ml in-droplet FITC-Ab concentration).

Wash water collection, characterization, and preparation

Wash water from a shredded lettuce processing line was acquired from a major ready-to-eat vegetable processing facility on the East Coast. Initial samples were divided into 500 ml aliquots and stored at -80°C. Working stocks were prepared in aliquots of 5 ml and stored at -20°C. Collected wash waters for all vegetable samples were analyzed for free chlorine and organic chloramines by the N,N-diethyl-p-phenylene diamine colorimetric method [57]. Turbidity, conductivity, and pH were also measured by Hach 2100N Turbidimeter (Hach Company, CO), Hach Sension5 portable conductivity meter (Hach Company, CO) and accutupH⁺ probe (Fisher Scientific), respectively. Once thawed from stocks, pH of the shredded lettuce wash water was measured using a VWR pH meter (VWR Symphony SB70P Digital, Bench-model pH Meter). For microfluidic experiments, the shredded lettuce wash water was filtered through a 0.22 μm syringe filter (Durapore PVDF Membrane, MILLEX GV) and spiked with bacteria to an approximate concentration of 10⁶ CFU/ml. Residual chlorine was quenched with 1 mM sodium thiosulfate (Sigma).

Microfluidic consumables, equipment, and procedure

The co-flow microfluidic device utilized in this research was designed at Harvard University, and published in Nature Protocols [29]. Microfluidic chips used in this research were fabricated by the Research and Exploratory Development Department (REDD), Johns Hopkins University Applied Physics Laboratory (JHUAPL). Microfluidic equipment, consumables, and technical support were supplied by the Applied Biological Sciences Group, JHUAPL. The microfluidic droplet system utilizes a proprietary surfactant (Phasex Corp.) at a concentration of 646 mg per 40 ml 7500 Novec Oil (3M). This surfactant-oil mixture aided in oxygen transfer, served as the carrier medium and facilitated uniform droplet formation of the sample input(s) on-chip. Flow rates reported in the Nature Protocol were 180:90:90 μl/h (2:1:1 flow rate ratio) for the surfactant-oil mixture and the two sample inputs [29]. Flow rates utilized in this research were modified from 2:1:1 to 2:0.5:0.5 (surfactant-oil to sample(s) input flow rate). Slight modifications were made for uniformity of observed droplet diameter upon formation during experiments. The on-chip device used in this research had one surfactant-oil input and two sample inputs. If two sample inputs are utilized, the overall input flow rate of the samples must be split between the two inputs. For example, if flow rate of the surfactant-oil input is 2,000 μl/h, sample input A will be 500 μl/h and sample input B will be 500 μl/h. Flow rates for this research were typically 1,000 μl/h surfactant-oil input, and 250 μl/h for each sample input. Samples were loaded into a syringe using a blunt fill needle (BD Blunt Fill Needle). The fill needle was then removed and replaced with a precision needle (BD Precision Glide). Tubing was connected between the sample syringe and the microfluidic device on-chip (95 Durometer LDPE, Scientific Commodities Micro Medical Tubing, 0.015” I.D x 0.043” O.D.). Proprietary apparatus pumps designed for syringes (1 ml to 30 ml volume BD Luer Lock Tip) were used to regulate flow rates (Harvard Apparatus). Pump flow rate was managed via a software program compatible with Harvard Apparatus (LabVIEW 16, National Instruments). Droplet generation was monitored in real-time using Nikon objectives equipped with a high-speed camera and complimentary software (FASTEC IL5). Droplets were captured in 1.5 ml experimental tubes (Eppendorf).

Microscopy equipment and procedures

Formed droplets were drawn into 50 μm diameter glass capillary tubing for visualization and measurement (0.05 x 0.5 mm ID, VitroTubes). Filled capillaries were then fixed to a microscope slide (Fischer) by first sealing the loading end and elevating the slide at a 45-degree angle for 1–5 minutes. This allowed the droplets to spread out equally and stack for imaging. The non-loading end was then sealed before microscopy. The Eclipse Ni-E motorized microscope system and Ni-E Analysis Elements software (Nikon) equipped with a digital camera (ORCA-Flash 4.0 LT, Hamamatsu) were used for fluorescent images and data analysis. Permission to utilize the Eclipse Ni-E motorized microscope system and software was granted by REDD, JHUAPL. Z-stack data at five μm intervals was collected throughout entirety of the droplet using the ND Acquisition tool in Ni-E Analysis Elements. Droplet diameter ranged from approximately 50–70 μm. Z-stack data were then merged into flat, data-rich images using the EDF tool in General Analysis, Ni-E Analysis Elements. Merging Z-stack data allowed for maximum image capture of encapsulated bacteria by creating a two-dimension data image of bacteria at differing focal planes. Point estimates of fluorescence intensity were gathered and a complimentary fluorescence intensity of background was paired with each measurement. Fluorescence measurements were normalized by reporting relative fluorescence. Relative fluorescence was calculated by dividing bacterial fluorescence by background fluorescence (Fluorescence_bacteria / Fluorescence_background). A sample size of five (n = 5) relative fluorescence measurements were used for all mean, standard deviation, and 95% confidence interval calculations. Fluorescent data images were smoothed and normalized using Ni-E Elements software in order to report comparable, high resolution images.