Photoacoustic discrimination of antibiotic resistant and sensitive Staph. aureus isolates

Abstract

Objectives:

Bacteremia is a serious and potentially lethal condition. Staph. aureus is a leading cause of bacteremia and methicillin resistant Staph. aureus (MRSA) accounts for more than a third of the cases. Compared to methicillin sensitive Staph. aureus, MRSA is more than twice as likely to be fatal. Furthermore, sub-populations of seemingly isogenic bacteria may exhibit a range of susceptibilities, often called heterogenous resistance. These heterogeneous antibiotic resistant infections are often misdiagnosed as hospital acquired secondary infections because there are no clinically used tests that can differentiate between homogeneous and heterogeneous antibiotic resistance. We describe the development and proof of concept of rapid bacterial identification using photoacoustic flow cytometry and labeled bacteriophages with the characterization and differentiation of heterogeneous antibiotic resistant bacterial infections.

Methods:

In photoacoustic flow cytometry, pulsed laser light is delivered to a sample flowing past a focused transducer and particles that absorb laser light create an acoustic response. Optically labeled bacteriophage are added to a bacterial mixture that flows through the photoacoustic chamber. The presence of target bacteria is determined by bound labeled phage which are detected photoacoustically. Incubation of bacterial samples in the presence and absence of the antibiotic daptomycin creates a difference in bacterial cell numbers that is quantified using photoacoustic flow cytometry.

Results:

Four clinical isolates were tested in the presence and absence of daptomycin. Photoacoustic events for each isolate were recorded and compared to growth curves. Samples treated with daptomycin fell into three categories: resistant, susceptible, and heterogeneous resistant.

Conclusions:

Here we show a method to determine the presence of bacteria as a marker for blood stream infection level and antibiotic sensitivity in less than 4 hours. Additionally, these results show an ability to identify heterogeneous resistant strains that are often misidentified.

INTRODUCTION

Antimicrobial resistance has become one of the most pressing concerns for global health and therefore has become an ever increasing focus of research and product development. Antibiotic resistance arises most often when bacteria gain and express gene cassettes that confer the ability to outmaneuver the action of an antibiotic. Bacteria employ two main strategies to gain resistance to antibiotics; pumps and enzymes[1]. Pumps work to evacuate the antibiotic from cell cytoplasm before they can reach a critical level. Enzymes work to degrade antibiotic molecules before they can have an effect. Non-genetic antibiotic resistance also occurs, though in a minority of cases, and is often mediated by small molecule communication between cell populations or general impermeability of the cell wall[2]. Resistance to antibiotics is an inevitable outcome of their use. The first cases of penicillin resistance was reported only 2 years after its widespread use[3]. There are multiple pathways of acquiring antibiotic resistance in each type of bacteria. Horizontal gene transfer is the dominant method of acquiring antibiotic resistance[4].

The most rapid test for antibiotic resistance in Staph. aureus is a polymerase chain reaction (PCR) based test for the mecA gene. PCR is much faster than the standard antibiotic disc method[5], but is used in only a minority of cases. The antibiotic disc method requires an antibiotic disc placed on a lawn of bacteria and measuring the zone of inhibition caused by the antibiotic. This method requires an additional 24 hours after initial bacterial identification. PCR methods can be completed in 3–4 hours but can suffer from failure due to unknown sample conditions and concentrations or amplification of sub-populations that mask clinically relevant genetic traits or markers. Most problematic for clinicians are the false negatives when a resistant strain is misidentified as a susceptible strain. This delay in appropriate antimicrobial treatment can be detrimental to patients. Occasionally, these tests give intermediary or conflicting results regarding a strain’s resistance.

Isolates that are heterogeneous in their expression of a resistance gene can lead to serious problems for patient treatment. A low expressing strain will appear to be susceptible when using the antibiotic disc method and when treated with first line antibiotic such as oxacillin. In some cases, the majority of bacteria are killed and the small remaining percentage are dealt with by the immune system. In other cases, the infection appears to be cleared only to reemerge a few days later. Despite the infection still being the original bacterial strain that the patient presented with, it will be treated by the clinician as an unresolved infection or a new infection, depending on the time taken for the infection to reemerge. The reemergence of infection will likely be classified as a hospital acquired infection. Hospital acquired infections are no longer reimbursed by the Centers for Medicare and Medicaid Services. Additionally, increased hospital acquired infections negatively impact each institutions Hospital-Acquired Condition score resulting in overall reduction of funding and reimbursement rates from Medicare and Medicaid Services[6]. For healthcare institutions this is a costly mistake since they can no longer charge for the initial community acquired infection.

In order to develop a system of determining antibiotic resistance, we used the antibiotic daptomycin in conjunction with photoacoustic flow cytomtery (PAFC) and bacteriophage as molecular tags. Daptomycin has shown a lack of cross resistance with other antibiotic classes[7] as well as being broadly active against MRSA isolates[8]. Flow cytometry has been effectively used to analyze large heterogeneous cell populations since Wolfgang Gohde first developed it in 1968. Fluorescent flow cytometry relies on the absorption of laser light by an object and the detection of the fluorescence from that object at an alternative wavelength. Fluorescence flow cytometry has a few disadvantages that can limit its usefulness for bacterial identification. Light, though very powerful, is easily quenched or blocked in a turbid environment such as blood. Additionally, small amounts of light from single cells can be hard to detect in dilute samples. Fluorescent flow cytometry works best with large numbers of cells and clear non turbid environments. As an alternative, PAFC relies on the absorption of laser light and the detection of ultrasound waves created by the photoacoustic effect[9].

The photoacoustic effect has been used in many ways. Depth profiling in human tissues for the treatment of port wine stains[10] has been done as well as photoacoustic imaging of blood vessels[11]. PAFC has successfully been used to enumerate circulating tumor cells and has been shown to be a robust predictor of metastasis in melanoma[12, 13]. Additionally, PAFC has been used successfully for the isolation of circulating tumor cells by several groups[14, 15]. PAFC was invented to find rare, individual particles in complex environments[16]. The ultrasonic waves created by the photoacoustic affect are robust and not quenched in turbid media, such as cell suspensions or blood samples. Recently, PAFC has been used on blood samples, in vitro, to detect bacteria using modified bacteriophage as optical tags[17].

Bacteriophage, viruses that infect bacteria in a specific manner, have been used for many years as a way of classifying bacterial strains[18]. A bacteriophage’s ability to discriminate and bind tightly to their host bacteria is vital to their fitness and evolutionary survival. Even in complex environments, bacteriophage are able to identify and bind target bacteria within seconds[19]. Bacteriophage-host attachment is achieved via protein-protein interactions with the long tail fibers or tail spike proteins[20].

These proteins have developed to be among the most stable protein complexes found in nature[21]. As bacterial probes, tail fiber proteins have many advantages over antibodies. Tail fibers are produced as part of bacteriophages that are self-replicating within a bacterial host, making them cheaper to produce than antibodies. An electron micrograph of bacteriophage tail fibers is shown in Figure 1. Additionally, tail fibers are more stable than antibodies[22, 23]. Bacteriophage tail fibers also have greater specificity than antibodies and have evolved to bind to essential surface antigens of the bacteria and are therefore hard for the bacteria to change.

Figure 1.

Bacteriophage Det7 virion particles with Tails sheath, Capsis, and Tail _bers labeled. Picture take by Robert H. Edgar at a magni_cation of 52,000x using a Leica TEM with uranyl acetate stain

Bacteriophage have evolved alongside bacteria as they have differentiated into new subspecies of bacteria, even those that have acquired antibiotic resistance. Bacteriophage attaching to bacterial surface antigens are displayed in Figure 2[24].

Figure 2.

Multiple bacteriopahge particles attached to a single E. coli cell imaged using Helium Ion microscopy by Leppnen et al.2017

Materials and Methods

Our photoacoustic setup is based on our system used to detect circulating melanoma cells in blood[13, 25]. PAFC has been described previously and was shown to be a robust method of detecting rare particles in dilute samples. The PAFC system was tested with phosphate buffered saline (PBS) to demonstrate a level of background noise or variability. As a positive control for system function we used 10μm polystyrene spheres obtained from Polybead (Warrington, PA) and titered them through the PAFC system. Staph. aureus strain SA113 (ATCC 35556, Manassas, Virginia) was obtained from American Type Culture Collection and we used E.coli K12 as a control bacterial cell type. Dyed SP1 bacteriophage were added to resuspended cultures at a ratio of 1000 phage per bacterial cell. Phage/bacteria mixtures were incubated on the bench top for 10 minutes to allow phage attachment, then processed thought PAFC system.

Sample preparation

Clinical isolates of Staph. aureus were obtained from the Urisch laboratory at the UPMC Department of Othopaedic Surgery. All isolates were sequenced and identified as mecA positive or negative. Isolates were de-identified according to IRB protocol and were stored in 50% glycerol solution at −80◦C. Isolates were streaked for single colonies on mannitol salt agar (MSA) plates and cultures were grown in mannitol salt broth, shaking at 36.5◦C. Overnight cultures were then diluted into fresh media and regrown for 2 hours to ensure that bacteria were entering exponential growth phase. Exponentially growing cultures were mixed with media or daptomycin in a 1:1 ratio. Daptomycin was added to 0.5 ml of culture to a final concentration of 0.25 μg/ ml. Identical cultures were used in all experiments using PAFC and BioTek H1 plate reader. For photoacoustic testing, bacteriophage SP1 was added to a concentration of 1000 phage per bacterial cell. Growth curves were performed for 16 hours with measurements taken once every minute. Growth curves for each strain and corresponding PAFC results are displayed in Figure 4.

Figure 4.

Untreated culture are shown in red and treated cultures shown in blue. Left hand panel shows 2 hours growth di_erentiation. Right hand panel shows full 16 hours of growth. Tables shows the number of cells detected in each sample tested at two hour time point. Homogeneous susceptible (1) strain shows a complete inhibition of growth and zero detection in treated samples. Heterogeneous resistant (3) strain A shows a delay of exponential growth. Homogeneous resistant (2) strains show no delay or inhibition of growth and both treated and untreated samples reach exponential growth simultaneously. Heterogeneous resistant (4) strain B shows no delay in initial growth but sever retardation of growth starting at 1.5 hours and continuing until the end of the test.

Treated and control cultures were placed in a round bottom 96 well culture plate (Falcon microtest 96 well plate 35077, ThermoFisher, Waltham, Massachusetts) and were placed into the BioTek H1 plate reader. Optical density (OD) measurements were taken every minute at 600nm wavelength. Between measurements, the plate was shaken at 100 rpm and maintained at 36.5◦C allowing for bacterial growth. The two hour time point was determined to be sufficient to differentiate the growth rates. Additionally, multiple replicates of each bacterial strain was grown at 36.5◦C in the BioTek H1 plate reader and 100 μL was removed and plated on MSA plates every 10 minutes. Growth curves were made directly from these titers for each strain to give a quantifiable number of bacterial cells for each OD. Each strain was found to consistently correlated between cell titer and OD.

Photoacoustic Flow Cytometery

The photoacoustic flow chamber is shown in Figure 3. A 1000μm optical fiber with numerical aperture of 0.39 (Thorlabs, Newton, New Jersey) was used to deliver laser light at 532 nm with a 5 ns pulse duration from a Nd:YAG laser(Litron Nano, Boseman, Montana). Laser beam energy was maintained and measured between 1.9 and 2.1/mJ for all PAFC experiments. The optical fiber was placed 5mm away from a quartz tube (Quartz 10 QZ, Charles Supper, Natick, Massachusetts) with 10μm thick walls. Samples were pumped through the quartz tube inside a 3D printed flow chamber. The laser beam was assumed Gaussian in shape and fluence was calculated to be 0.014 mJ/cm² and detection volume was determined to be 0.04 μL.

Figure 3.

Schematic of photoacoustic ow chamber with parts labeled for identification

Sonotech LithoClear acoustic gel (Next Medical Products Company, Branchburg, New Jersey) filled the internal space of the 3D printed flow chamber and a 2.25 MHz transducer was focused on the quartz sample tube. The acoustic gel provided acoustic coupling between the quartz tube and the transducer along the propagation of acoustic waves generated from thermoelastic expansion. A Tegam 4040B amplifier (Tegam, Inc., Geneva, Ohio) amplified signals with a gain of 50. A computer running a LabView program recorded signal waveforms[26].

Bacteriophage preparation

SP1 bacteriophage were grown using Staph. aureus strain SA113 (ATCC, Old Town Manassas, Virginia) and concentrated using methods described previously[17]. Purified phage of 1 × 1012 plaque forming units per milliliter (PFU/ml) were added to a saturated solution of Direct Red 81 dye (Sigma Aldrich, Saint Louis, Missouri). Virion particles were then pelleted and resuspended in buffer (10mM Tris, pH 7.5, 10mM MgCl2, 68mM NaCl). This process was repeated to ensure the removal of unbound dye. The absorbance spectrum of dyed phage was determined using the BioTek H1 and compared to that of undyed phage particles. Dyed phage were titered to ensure no detrimental effects were observed from the dying process. Dyed phage were retested for their ability to infect after 150 days and no difference in titer was observed.

All strains were tested in the presence and absence of daptomycin. Cultures were diluted into fresh media and regrown for 2 hours in the presence and absence of 25ug/ml of daptomycin. Each sample was then incubated at room temperature for 10 minutes with multiple dyed phage per bacterial cell. Incubated samples were then processed through the PAFC system and number of detected cells recorded. Isolates were tested in triplicate for both the plate reader and PAFC system.

Results

Differential growth rates were identifiable after two hours of the sixteen hour growth curve. Two hour differentiation was confirmed by Newton-Raphson root finding method.

Samples treated with daptomycin fell into three categories: resistant, susceptible, and heterogeneous resistant.

Resistant strains are those where no inhibition of growth was observed in the treated sample versus the untreated control. In resistant strains, the rate of exponential growth was identical between treated and untreated samples as well as nearly identical carrying capacity. Susceptible strains showed near complete inhibition of growth in treated sample versus untreated control.

Heterogeneous resistant strains were those where clonal isolates (genetically identical) growth curves where intermediary to susceptible and resistant growth curves. Heterogeneous samples displayed either a delay in reaching exponential growth phase or a complete retardation in achieving exponential growth. We tested all samples using the PAFC system in parallel to measuring growth rates. Growth curves for each of the bacterial strains were matched with their reported genotype. Strains that showed susceptibility at 0.25 μg/ml of daptomycin matched genotypically with MRSA strains where the mecA gene was not present in the genome.

Number of cells detected are displayed with corresponding growth curves in Figure 4.

Discussion

Bacterial resistance continues to be a growing problem worldwide. The 2019 antibiotic resistance threat report[?] from the CDC estimates over 2.3 million antibiotic resistant infections occur in the United States each year[27]. MRSA is one of the best known resistant bacteria and is also the most common. Using PAFC we have shown a way to rapidly identify antibiotic resistance and to quantify the level of resistance expression using a short treatment of antibiotics.

Growth curves for each bacterial strain were performed over 16 hours using a BioTek H1 plate reader. This technique allowed us to confirm reliable differentiation between strains growth at 2 hours. Daptomycin was used to test for antibiotic resistance. Daptomycin has been found to be more broadly active against MRSA isolates than the standard oxacillin[8]. Additionally, daptomycin has shown a lack of cross resistance with other antibiotic classes[7]. Sensitivity to daptomycin is dosage dependent for both MRSA (MIC 0.25–1 μg/ml) and VRSA (MIC < 4 μg/ml)[8]. Daptomycin was used at a concentration of 0.25 μg/ml as has been used by several other groups and been widely effective against antibiotic resistant Staph. aureus strains tested throughout Europe[28].

Growth curves for each of the bacterial strains were matched with their reported genotype. Strains that showed susceptibility at 0.25 μg/ml of daptomycin matched genotypically with MRSA strains where the mecA gene was not present in the genome. Susceptible strains were tested in the presence and absence of daptomycin. In the presence of daptomycin, little or no growth was observed at either the 2 hour or 16 hour time point as can be seen in Figure 4 panel 1.

To correlate the OD growth curves with photoacoustics, we tested multiple strains in our PAFC system in the presence and absence of daptomycin. Dyed bacteriophage SP1 was added to resuspended cultures to add optical absorption to the bacterial cells for photoacoustic detection. For homogeneous susceptible strains, no bacteria were detected in the treated bacterial cultures demonstrating a complete inhibition of growth as seen in Figure 4, panel 1. This inhibition is expected, due to the efficacy of antibiotic treatment suggested by the susceptible nature. Our PAFC system results are corroborated by the 16 hour growth curves demonstrating our ability to identify antibiotic sensitivity in under 4 hours. The photoacoustic method allows for more rapid determination of susceptibility, indicating clinical utility of the method.

MRSA strains were tested using PAFC and OD growth curves. Resistant strains fell into two distinct categories. What we defined as homogeneous resistance were strains in which the growth in the presence and absence of daptomycin was indistinguishable, examples are shown in Figure 4 panel 2. Homogeneous resistance could also be described as having complete penetrance of mecA gene expression. When tested with PAFC, we see identical numbers of cell detected between the treated and untreated cultures. In contrast, heterogeneous strains were ones where a slight retardation in growth was observed, Figure 4 panel 3 and 4. Many genotypically resistant strains displayed an intermediary growth in the presence of daptomycin. The level of growth inhibition is correlated to the amount of penetrance of mecA gene expression in that population. The ability of PAFC to provide bacterial numbers gives an indication of this heterogeneity and may be used for clinical management of such infections. In contrast, homogeneously resistant infections are more easily interpreted during the course of treatment and knowledge of heterogeneity may provide needed information when clinical decisions are made regarding modifying treatment.

Heterogenous resistant A and B (Figure 4, panel 3 and 4)are representative examples of strains where growth curves showed an intermediary inhibition in the presence of daptomycin. When these strains are tested in our PAFC system, we again see corroboration with the growth curves and the intermediary number of cells detected. Heterogeneous Resistant B in Figure 4, panel 4 demonstrates both a delay in growth and large, but not complete, inhibition of growth.

In our current study, production and purification of phage was the limiting factor for how many samples we could test. Though the production and purification is relatively cheap, even on a laboratory scale, it is time consuming and requires specialized skills. With advancements in phage therapy several companies such as Advanced Phage Therapeutics (Gaithersburg, MD) and ARMATA pharmaceuticals (Marina del Ray, CA) have started large scale production of FDA approved GMP phage. The availability of high titer purified phage greatly increases the practicality and lowers the expense of this system of bacterial detection. Additionally, this system of bacterial detection can also be used in phage therapy to rapidly test bacterial susceptibility to particular phage for treatment purposes.

Photoacoustic flow cytometers are relatively economical to build and are much simpler than many common laboratory equipment. Laser sources are the greatest cost and all parts are commercially available. Total cost of a laboratory setup is around $30,000 while commercial setups could be produced for much less. This low equipment cost suggests photoacoustic flow cytomtery could become common clinical tools, similar to x-ray or ultrasound machines.

Bacteriophage Attachment Variability. An additional advantage when used with PAFC, bacteriophage size plays an important role. Single free floating phage particles are below the detection threshold of our photoacoustic system. Bacteriophage attach to outer surface antigens on the surface of bacterial cells allowing for multiple bacteriophage to bind to a single cell. Regarding the number of bacteriophage attaching to an individual cell, Max Delbruck demonstrated in the 1930’s that bacteriophage binding to surface receptors followed a pattern where the vast majority of bacterial cells will have a full complement of bacteriophage bound and very few will have more or less bound.[29]. When labeled bacteriophage aggregate on the surface of bacterial cells, they form a large enough optical absorber to create a detectable acoustic response. Due to this phenomenon, acoustic signals are only detected from target bacteria of our bacteriophage of interest and not from unbound bacteriophage. A threshold was set for the acoustic signal amplitude at 2.5 times the noise floor, meaning that even bacterial cells with far fewer bacteriophage attached create a detectable and quantifiable signal.

Additionally, the bacterial cell suspension is diluted so that the expected value of cells in the detection volume is one, following a Poisson distribution. Assuming a uniform distribution of cells from a well mixed sample, the vast majority of detections will have a single bacterial cell, though a few might have two. A negligible number will have three or more. With our simple amplitude threshold detection, the method is not dependent on relative numbers of attached bacteriophage, as long as there are enough to reach threshold. We are considering automated classifiers for the photoacoustic signals, though our work is too preliminary to provide enough data for a robust classifier at this point.

Conclusions

Combining the growth curve data and PAFC results leads us to postulate that Heterogenous A and B strains are heterogeneous in their expression for the mecA gene. Our results suggest that strain Heterogeneous Resistant A in Figure 4, panel 3 expresses resistance derived from the mecA gene at a higher penetrance while Heterogeneous Resistant B in Figure 4, panel 4 expresses the mecA gene at a low penetrance. These results demonstrate our ability to identify antibiotic resistance in Staph. aureus in less than 4 hours. This 4 hour period includes 2 hours of incubation with and without antibiotic, followed by photoacoustic testing, which is less than 2 hours. Additionally, these results show an ability to identify heterogeneous resistant strains that are often misidentified. Correct identification of heterogeneous resistant Staph. aureus could potentially save hospitals money and resources.

Future Work

We will extend the ability to determine antibiotic sensitivity by testing a larger sample size as well as other types of bacteria and resistance. We are building collaborations with commercial entities to supply larger quantities of bacteriophage to expand our current studies. Partnering with a group that can supply larger quantities of bacteriophage will allow us to increase our sample size and develop strong, predictive models for bacterial identification and determination of antibiotic resistance. Until then, this preliminary work suggests that PAFC can provide this type of information for clinical management of bacterial infections.

Additionally, RNA-Seq will be performed to further correlate the level of gene expression penetrance with growth curves and cells detected from PAFC. Vancomycin resistant strains, though less common, are dramatically harder to treat and use up hospitals limited resources[30]. Vancomycin resistant Enterococcus, carbapenem resistant Enterobacteriaceae, and multi-drug resistant Pseudomonas aeruginosa are listed as serious threats by the CDC and would benefit from early detection and susceptibility determination. On both an individual and global scale, rapid identification and characterization is essential. The potential for worldwide crippling pandemics from bacterial pathogens is of central concern to the CDC and WHO[27]. Multi-drug resistant bacterial infections regularly have mortality rates closer to 50% and transmission rates similar to that of flu and SARS-COV-19 which have mortality rates closer to 1%[31]. There is a clear need for better and more advanced rapid diagnostics to protect individual patients. The disk diffusion method has been the gold standard since 1956 but requires 16–24 hours after cultures are grown[32]. PAFC has the potential to supplant the gold standard by directly counting the difference between cells in treated and untreated cultures in less than four hours.