2.1.

Photoacoustic Flow Cytometry

A Nd:YAG laser (Litron Nano, Bozeman, Montana) coupled into a $1000\mu \mathrm{m}$, 0.39 numerical aperture, optical fiber (Thorlabs, Newton, New Jersey) was used to produce 532-nm laser light with 5-ns pulses. The laser beam energy coupled through the optical fiber was maintained and measured from 1.9 to 2.1 mJ for most detection experiments. Laser energy was increased to 4 mJ for single cell detection experiment. Laser light was directed to a quartz tube (Quartz 10 QZ, Charles Supper, Natick, Massachusetts) with $10\text{-}\mu \mathrm{m}$-thick walls passing through a 3D-printed flow chamber. The $10\text{-}\mu \mathrm{m}$-thick walls allow the propagation of ultrasonic waves, as well as providing an optically transparent pathway for the sample to flow through. Optical fiber was placed 5 mm from the quartz tube to create a detection volume of $0.04\mu \mathrm{l}$. Laser beam shape was Gaussian and fluence was calculated to be $0.014\mathrm{mJ}/{\mathrm{cm}}^2$.

A 2.25-MHz transducer focused on the quartz sample tube was fitted to the base of the 3D-printed flow chamber (Fig. 2). The internal volume of the chamber was filled with Sonotech LithoClear acoustic gel (NeXT Medical Products Company, Branchburg, New Jersey) to provide a medium for the propagation of acoustic waves. Syringe pumps were used to create an alternating flow of sample and mineral oil equal to 60/min flow rate. The introduction of sample and the immiscible mineral oil induced two-phase flow.³⁷ Two-phase flow was employed to allow for future collection of the samples for further analysis while eliminating the possibility of samples becoming stuck or delayed inside the tubing. Signals were amplified with a gain of 50 using a Tegam 4040B amplifier (Tegam, Inc., Geneva, Ohio) and sent to a desktop computer running a customized LabView program. This computer also served for system control and data collection. This flow chamber setup served as the excitation and acoustic wave collection device.

Fig. 2

Schematic of photoacoustic flow chamber with parts labeled for identification.

Detections were determined by two methods. Primary detection was through amplitude detection above a given threshold. Threshold was empirically set at 1.5 times the root-mean-square noise value. A scoring function derived from known positive detections was derived to score each waveform based on its key aspects. Each detection above threshold was scored using the scoring function and given a confidence value (Fig. 3).

Fig. 3

Schematic of photoacoustic flow setup with parts labeled for identification.

2.2.

Bacteriophage Preparation

Det7 bacteriophage lysates were concentrated using polyethylene glycol 8000 precipitation described by Castro-Mejía et al.³⁹ and initially described by Yamamoto et al.⁴⁰ Differential centrifugation and cesium chloride gradients were used to further purify and concentrate bacteriophage Det7 stocks.⁴¹ Stock concentrations of $5\times {10}^{11}$ plaque forming units per milliliter (PFU/ml) or greater were produced. Pure stocks of bacteriophage were then diluted into a saturated solution of Direct Red 81 dye (Sigma Aldrich, Saint Louis, Missouri). Bacteriophage Det7 virion particles were then pelleted and resuspended in 10 mM Tris, pH 7.5, 10 mM ${\mathrm{MgCl}}_2$, 68 mM NaCl (bacteriophage buffer) to remove any unbound dye. A NanoDrop 2300 spectrophotometer was used to record the absorbance spectrum and to verify an increase in dyed bacteriophage absorbance at 532-nm wavelength. An increase of absorbance at 532 nm was identified for dyed bacteriophage Det7 when compared to the absorbance of undyed Det7 bacteriophage. Dyed Det7 bacteriophages were retested for their ability to infect their target bacteria and stocks were titered to determine the total number of dyed infectious particles. Titers before and after dyeing protocol are expected to be within 3% to 5%, as this is the expected variation derived from multiple titers. Any decrease in titer would suggest that the dye modification is inactivating bacteriophage particles. No difference in titer was determined even when Det7 bacteriophages were kept in an excess of dye for 60 days, demonstrating that the bacteriophages are robust and unaffected by the attachment of the Direct Red 81 dye.

3.1.

Bacteriophage Detection

Bacteriophage buffer was run through the PAFC system to demonstrate a level of background detection and any variability with PBS. Purified bacteriophages were next titered through the PAFC system; 0.5 ml of each concentration ranging from $1\times {10}^1\mathrm{PFU}/\mathrm{ml}$ to $1\times {10}^{12}\mathrm{PFU}/\mathrm{ml}$ were tested. No detections were observed for either undyed bacteriophages or phage buffer, demonstrating their inability to absorb laser light and produce a photoacoustic response using 2-mJ laser energy.

Next, purified dyed bacteriophages were tested with a laser energy of 2 mJ; 0.5 ml of each concentration ranging from $1\times {10}^1\mathrm{PFU}/\mathrm{ml}$ to $1\times {10}^{12}\mathrm{PFU}/\mathrm{ml}$ were tested. As can be seen in Table 1, no detections were recorded until bacteriophages reached a concentration of $1\times {10}^{11}\mathrm{PFU}/\mathrm{ml}$. At a concentration of $1\times {10}^{11}\mathrm{PFU}/\mathrm{ml}$, there are $\sim 1\times {10}^5$ dyed bacteriophages in the detection volume of $0.04\mu \mathrm{l}$. Using 2 mJ of laser energy, $1\times {10}^5$ dyed bacteriophages per detection volume produce a signal that crosses our threshold of 1.5 times the root-mean-square noise value. All concentrations below $1\times {10}^{10}\mathrm{PFU}/\mathrm{ml}$ bacteriophage were assumed to be free-floating and evenly dispersed throughout the sample (Fig. 4).

Table 1

Detection of bacteria, bacteriophage, and dyed bacteriophage.

Fig. 4

(a) Signal from irradiating PBS, resulting in background noise. (b) Signal generated from irradiating bacteriophage bound to target bacteria. A detection is defined as any single waveform with 1.5 times or greater amplitude than the root-mean-square noise value.

3.2.

Bacterial Detection

The photoacoustic response of free-floating bacteria was determined next. Overnight cultures of Salmonella LT2 and E. coli K12 were grown and diluted into fresh LB media. Cultures were grown at 37°C for 3 h to ensure bacteria were in exponential growth phase. Dilutions of each exponential culture were made and concentrations from $1\times {10}^8\mathrm{CFU}/\mathrm{ml}$ through $1\times {10}^1\mathrm{CFU}/\mathrm{ml}$ were tested for their photoacoustic response. Neither Salmonella LT2 nor E. coli K12 produced a photoacoustic response and no detections were recorded.

After determining background and baseline detection thresholds, we turned our attention toward our goal of detecting bacteria. When bacteriophages bind to their target bacteria, they are localized on the cell surface. This localization of bacteriophages, even when total concentration is well below detectable concentrations, creates a local increase in concentration that is then above the detection threshold. It is this localization of bacteriophages that results in the production of signals above our detection threshold.

Bacteriophage Det7 dyed with Direct Red 81 was incubated with Salmonella LT2 or E. coli K12 and allowed to bind to the bacterial cell surface. The host range of Det7 has previously been tested and described in detail.²⁸ Det7 infects a wide variety of Salmonella serovars but does not infect any E. coli strains. Salmonella LT2 bacteria were incubated with dyed Det7 bacteriophage in increasing ratios from 1:1 (bacteria:bacteriophage) increasing by order of magnitude to 1:1000. Mixed cultures were held at room temperature for 10 min to allow the bacteriophage time to adsorb to the surface of the Salmonella cells. Tests were run with bacterial cell concentrations ranging from $1\times {10}^4\mathrm{CFU}/\mathrm{ml}$ to $1\times {10}^8\mathrm{CFU}/\mathrm{ml}$. Each test was repeated using target bacteria, Salmonella LT2, and nontarget bacteria, E. coli K12. Table 2 demonstrates that in the presence of target bacteria, Salmonella LT2, and below threshold concentrations of dyed bacteriophage, Det7, multiple detections were recorded. Detections were limited with a built-in delay between signals to allow recording of each waveform. This delay limited the total number of signals that could be detected to 660 signals per test. Both the $1\times {10}^4\mathrm{CFU}/\mathrm{ml}$ and $1\times {10}^8\mathrm{CFU}/\mathrm{ml}$ were near constant detections, and the $1\times {10}^5\mathrm{CFU}/\mathrm{ml}$ and $1\times {10}^6\mathrm{CFU}/\mathrm{ml}$ showed a much lower number of detections. Table 3 shows that for nontarget, E. coli K12, no detections were recorded when mixed with dyed Det7 bacteriophage, except at a concentration of dyed bacteriophage of $1\times {10}^{11}\mathrm{PFU}/\mathrm{ml}$. At a concentration of $1\times {10}^{11}\mathrm{PFU}/\mathrm{ml}$, there are $1\times {10}^6$ bacteriophages per detection volume of $0.04\mu \mathrm{l}$.

Table 2

Detection of target bacteria.

Table 3

Detection of nontarget bacteria.

We next directed our attention toward our goal of detecting single bacterial cells tagged with labeled bacteriophage. LT2 Salmonella was mixed with dyed Det7 bacteriophages in a 1:1000 ratio. The cell/bacteriophage mixtures were serially diluted to produce 100 cells per test volume of 0.5 ml and laser energy was increased to 4 mJ. The experiment was replicated 5 times with new serial dilutions of bacterial cells and bacteriophages to ensure the significance of detection numbers. As seen in Table 4, we detected an average of 43.4 out of every one hundred cells.

Table 4

Single cell detection.

4.1.

Bacteriophage Detection

Undyed bacteriophage showed no photoacoustic response when run through our detection system. When Direct Red 81 dye was added to the phage particles, detections were only observed when concentrations of phage reached $1\times {10}^{11}\mathrm{PFU}/\mathrm{ml}$. At a concentration of $1\times {10}^{11}\mathrm{PFU}/\mathrm{ml}$, there will be $\sim 4\times {10}^5$ dyed bacteriophages per detection volume of $0.04\mu \mathrm{l}$, resulting in a signal. At a concentration of $1\times {10}^{10}\mathrm{PFU}/\mathrm{ml}$, bacteriophages are present at about the level of one bacteriophage per $0.6\mu {\mathrm{m}}^3$, which is about the volume of a bacterial cell. We hypothesize that at $1\times {10}^{11}\mathrm{PFU}/\mathrm{ml}$ concentration, bacteriophages start to clump together and form multiphage complexes, as has previously been seen by electron microscopy. Multiphage complexes can form for a variety of reasons, chief among them would likely be entanglement of tail fibers or low pH as described by Goldwasser et al.⁴² All concentrations of bacteriophages below $1\times {10}^{11}\mathrm{PFU}/\mathrm{ml}$ are assumed to be free-floating and evenly distributed. Free-floating bacteriophages less than $\sim 4\times {10}^5$ per detection volume are below the detection threshold for our system. A signal is produced when target bacteria are present that allow bacteriophage binding. Binding of multiple bacteriophages to a bacterial cell surface will increase the local concentration of dyed bacteriophage. We hypothesize that this increase in local concentration of bacteriophages is what leads to a positive signal above our detection threshold.

4.2.

Bacterial Detection

Tables 2 and 3 demonstrate that our bacteriophages are specific to their target bacteria and that we do not get a signal from unattached bacteriophages except when in extremely high concentrations. Table 2 shows our ability to detect bacterial cells when tagged with dyed bacteriophages. When cells are tagged with 1, 10, or 100 bacteriophages, they are below our detection threshold. Some cells were missed due to our built-in delay for recording of signals while others were simply missed due to our testing of only a single laser energy. In Table 2, the concentrations of bacteria with 1000 bacteriophages per cell showed detections. Due to our built-in delay, $1\times {10}^5\mathrm{CFU}/\mathrm{ml}$ and $1\times {10}^8\mathrm{CFU}/\mathrm{ml}$ showed saturated detections. Laser energies of 4 mJ have previously been shown to increase detection sensitivity of the system. In future trials, increasing laser energies will be tested until our background noise increases or we reach 100% cell detections. The variation between the number of detections between the $1\times {10}^4\mathrm{CFU}/\mathrm{ml}$ and $1\times {10}^8\mathrm{CFU}/\mathrm{ml}$ and $1\times {10}^5\mathrm{CFU}/\mathrm{ml}$ and $1\times {10}^6\mathrm{CFU}/\mathrm{ml}$ could be due in part to nonhomogeneous mixing of our bacteria and phage. Additionally, there could have been imperfections or bubbles in our acoustic gel that led to decreased signal propagation. Additional work is being done to remove bubbles from acoustic gel and develop better and more permanent ways of producing flow chambers and ensuring acoustic coupling in our system. Currently, repeated measurements in alternating orders are used to rectify this inconsistency. This represents an area of refinement and future work in preparing this system for more diagnostic purposes. Table 4 demonstrates our system’s ability to detect individual cells when tagged with modified bacteriophage. Cells were serially diluted to produce roughly 100 cells per test volume. Chase and Hoel⁴³ first described and modeled the error associated with serial dilutions of bacteria and bacteriophage. We therefore expect some variation and loss of cells from manual pipetting and serial dilutions. Despite this loss, we detected nearly 50% of estimated cells. Future work to resolve this challenge and produce 100% detection rate will come from using higher concentrations of cells with less chance of loss as well as optimizing our flow system. Using higher concentrations of bacteria will reduce the error from pipetting and serial dilutions. In future trials, bacteria can be collected after exiting our flow system and plated to determine relative number of bacteria present and calculate loss. Additionally, larger sample sizes will provide more robust measurements and greater accuracy in number of bacteria present. Moreover, the ability to detect about half of all single cells is probably much more sensitive than needed clinically, as the concentration of bacteria in blood would need to be much higher to cause illness in a human being.

4.3.

Conclusion

Bacteriophages have evolved to identify and bind to their target bacteria with high specificity. Bacteriophage host attachment is mediated solely by tail fibers.²⁸ Tail fibers are differentiated into long tail fibers, such as bacteriophage T4, and tail spike proteins, such as P22 TSP. Bacteriophage host attachment has many advantages over antibodies. Antigens used by antibodies are often the most abundant surface molecules or those that cause the greatest immune response.⁴⁴ These surface molecules can often change to avoid antibody detection.⁴⁵ Conversely, bacteriophages have evolved to use surface epitopes that are essential and difficult to change.⁴⁶ Bacteriophages have even been shown to target cell surface pumps used in bacterial antibiotic resistance. Though bacteriophage resistance can evolve, it happens at a much lower rate than antibody avoidance and always has a negative fitness effect on the bacteria.⁴⁷ Bacteriophage attachment proteins are also among the most stable protein structures to be discovered and bind the phage irreversibly to the bacterial cell.⁴⁸ Antibodies are more expensive to produce,⁴⁹ are less stable,⁵⁰ and bind less strongly than bacteriophages. Antibodies have a binding constant, kD, in the range of 1 to 10 nM while bacteriophages have a binding constant closer to 10 to 50 nM.^51,52

PAFC presents a rapid way to detect microscopic particles under flow based on their ability to absorb laser light. These initial experiments demonstrate our ability to use readily available protein dyes on bacteriophages without affecting their ability to attach to target bacteria. This research presents an innovative way of identifying and differentiating bacterial strains. This method can be further developed for use with other bacterial pathogens in blood cultures, representing a major step forward in clinical practice. The time and money saving potential of rapid detection and identification of bacterial infection are overshadowed only by the number of potential lives saved. Often the limiting factors for treatment of patients is the time spent waiting for results. It is our hope that the work presented above can be a foundation for future work and an ability to detect bacterial pathogens in blood cultures. Bacterial plate cultures and Gram staining are 19th-century technologies that have been the gold standard for decades, but current trends in resistant bacteria have necessitated a move toward more rapid and quantifiable diagnostic tools.