ABSTRACT
Bacterial biofilms can form persistent infections on wounds and implanted medical devices and are associated with many chronic diseases, such as cystic fibrosis. These infections are medically difficult to treat, as biofilms are more resistant to antibiotic attack than their planktonic counterparts. An understanding of the spatial and temporal variation in the metabolism of biofilms is a critical component toward improved biofilm treatments. To this end, we developed oxygen-sensitive luminescent nanosensors to measure three-dimensional (3D) oxygen gradients, an application of which is demonstrated here with Pseudomonas aeruginosa biofilms. The method was applied here and improves on traditional one-dimensional (1D) methods of measuring oxygen profiles by investigating the spatial and temporal variation of oxygen concentration when biofilms are challenged with antibiotic attack. We observed an increased oxygenation of biofilms that was consistent with cell death from comparisons with antibiotic kill curves for PAO1. Due to the spatial and temporal nature of our approach, we also identified spatial and temporal inhomogeneities in the biofilm metabolism that are consistent with previous observations. Clinical strains of P. aeruginosa subjected to similar interrogation showed variations in resistance to colistin and tobramycin, which are two antibiotics commonly used to treat P. aeruginosa infections in cystic fibrosis patients.
IMPORTANCE Biofilm infections are more difficult to treat than planktonic infections for a variety of reasons, such as decreased antibiotic penetration. Their complex structure makes biofilms challenging to study without disruption. To address this limitation, we developed and demonstrated oxygen-sensitive luminescent nanosensors that can be incorporated into biofilms for studying oxygen penetration, distribution, and antibiotic efficacy—demonstrated here with our sensors monitoring antibiotic impacts on metabolism in biofilms formed from clinical isolates. The significance of our research is in demonstrating not only a nondisruptive method for imaging and measuring oxygen in biofilms but also that this nanoparticle-based sensing platform can be modified to measure many different ions and small molecule analytes.
INTRODUCTION
While the bulk of bacterial understanding has come from planktonic bacteria, where bacteria grow independently in solution, natural biofilm growth is “a major mode of microbial life” (1). The biofilm microenvironment contains three-dimensional (3D) chemical gradients, and this chemical heterogeneity is associated with complex physiological heterogeneity with the potential to influence various pathogenic and antimicrobial-resistant behaviors (2, 3).
Pseudomonas aeruginosa is the most common infectious pathogen in the lungs of patients with cystic fibrosis (4). Chronic infection typically results in biofilm formation, which advances lung damage and often leads to respiratory failure (5). Unfortunately, treatment of P. aeruginosa infections can be difficult due to natural antibiotic resistance (6). There is frequently a lack of correlation between planktonic antibiotic susceptibility tests and biofilm-based infections (7). Thus, the ability to observe the internal biochemical microenvironment as antibiotics alter the biofilms would provide useful information for researchers.
Spatial heterogeneity exists in other biological systems as well. Other biofilms, such as those found in pipes (8) and the natural environment (9), all exhibit spatial heterogeneity where the ability to measure metabolites, such as oxygen, would be beneficial. Sulfate-reducing bioreactors, a passive wastewater treatment approach, show spatial heterogeneity in metabolites and community distribution (10). Models of the human trabecular meshwork (11) and intestinal crypts (12) display similar levels of heterogeneity that would also benefit from a way to investigate and understand complex 3D environments.
While chemical sensors are an advancement in the investigation of biofilms, they have been limited to measuring pH gradients (13, 14), oxygen (13, 15–17), or specific metal ions (18, 19). Microelectrodes have also been used to measure spatial distributions of gradients and concentrations in biofilms for a wide range of analytes (3, 17, 20–24). However, the spatial resolution is limited by the physical tip size, and the insertion of the microelectrode directly into the biofilm intrinsically disturbs the biofilm physical structure (25). Furthermore, this approach becomes difficult for frequent sampling and prohibitive for measurements spanning a heterogeneous sample. The majority of microelectrodes are only capable of producing one-dimensional (1D) measurements in the z-dimension. Kenney et al. (26), Acosta et al. (27), and others (28–30) have detected and mapped gradients of oxygen and pH in two-dimensional (2D) and 3D cell culture scaffolds. In order to capture temporal dynamics in addition to 2D spatial information, planar oxygen optodes are also used, but the information they provide is limited to the optode-biofilm interface (31–33). However, 2D approaches cannot capture depth-wise changes in samples, and other 3D options require particles that are large and potentially invasive. Oxygen is an important analyte of interest because it can be used as a measure of metabolic activity (34). Acosta et al. demonstrated the use of silica microparticles for measuring oxygen in bacterial biofilms (35). Many attempts to image or measure other analytes in biofilms have resulted in two-dimensional maps, limiting spatial understanding (15, 31, 32, 36).
Nanosensors are a technology which can overcome the limitations of current measuring methods by enabling continuous spatiotemporal monitoring of analyte concentrations in growing biofilms. These nanosensors are a tunable family of sensors that are designed for uninterrupted monitoring of in vitro or in vivo physiological parameters. The nanosensors are highly plasticized hydrophobic polymer nanoparticles which respond to changes in analyte concentration by altering their optical properties (37). These nanosensors (∼200-nm diameter) are useful in locations like biofilms where other techniques, like microelectrodes, fail due to size. Like microelectrodes, nanosensors have been developed for a variety of analytes, including pH, ions like K+ and Li+, and small molecules and metabolites, like oxygen and histamine. These polymer nanosensors can be fabricated with a variety of fluorescent emitters, such as quantum dots (38), carbon dots (39), or organic dyes (40). In addition, sensor response properties, including dynamic range and response midpoint, can be controlled through sensor formulation (41).
In this study, we report the development of our oxygen-sensitive luminescent nanosensor (O2NS) technology for monitoring oxygen spatiotemporal gradients and demonstrate its utility by monitoring the metabolism of both laboratory and clinically derived P. aeruginosa biofilms and the biofilm response to antibiotic attack.
RESULTS
Method development: sensor design and characterization.Figure 1a illustrates the sensor mechanism, where at low O2 concentrations, both the platinum porphyrin [platinum (II) meso-tetra(pentafluorophenyl)porphine (PtTFPP)] and the {4-[4-(dihexadecylamino)styryl]-N-methylpyridinium iodide} (DiA) reference dye luminesce. At higher O2 concentrations, oxygen collides with the PtTFPP molecule and quenches the luminescence via nonradiative decay. The fluorescence intensity of the platinum porphyrin PtTFPP (650 nm) decreases as the sample is exposed to increasing oxygen up to 21% in a linear manner (Fig. 1b). The peak of the DiA reference dye (585 nm) remains constant with respect to oxygen concentrations. Physical properties of these nanosensors, as determined by dynamic light scattering (DLS) and phase analysis light scattering (PALS), were an effective diameter of 163.4 ± 1.5 nm, a polydispersity of 0.15 ± 0.02, a zeta potential of −11.5 ± 1.34 mV, and a mobility of −0.90 ± 0.10 μm-s−1/V-cm−1. Values are listed as the arithmetic means and standard deviations.
(a) Schematic of nanosensor components coloaded into the nanoparticle matrix and luminescence quenching mechanism in response to increasing oxygen concentrations. (b) Fluorescence spectra for ratiometric oxygen nanosensors. Excitation occurs at 450 nm and produces the spectra shown under different oxygen concentrations. As oxygen concentrations increase, the platinum porphyrin peak at 650 nm decreases and the reference peak at 585 nm is insensitive to changes. This ratiometric sensor response allows us to calculate oxygen concentrations without complications resulting from variable nanosensor concentration. Normalized ratiometric calibration curves for oxygen nanosensors demonstrate effective sensing of oxygen concentration changes using both a spectrometer (c) and a confocal microscope (d) (used in remaining figures).
A Stern-Volmer constant, KSV, that allows for back calculation of oxygen concentration was determined. KSV for the O2NS was found to be instrument-specific, and thus, the O2NS was calibrated for each instrument used. The intensities of the PtTFPP signal at 650 nm and the intensities of the reference dye (DiA) signal at 585 nm were divided to form a ratiometric signal. The ratiometric signal at 0 mg/liter dissolved oxygen (DO) was then divided by the ratiometric signal at each concentration to create a linear calibration curve. The linear Stern-Volmer relationship between If0/If and oxygen concentration was used to fit the calibration data (Fig. 1c), where If0 is the ratiometric fluorescence intensity in the absence of oxygen and If is the ratiometric fluorescence intensity at the given concentration of oxygen. Calibration was also performed on the confocal scanning laser microscope used to obtain results from all microscopy experiments. Measurements were taken under ambient conditions (21% O2 conditions/6.65 mg/liter DO) and 0 mg/liter DO to form a two-point calibration curve (Fig. 1d).
As an initial test to determine if sensors function in the biofilm matrix, dead biofilms were imaged before and after deoxygenation with glucose and glucose oxidase. As can be seen in Fig. 2, the ratiometric signal increases as the oxygen in the biofilm decreases. These results indicate that the sensors function similarly in the complex matrix of the biofilm in comparison with how they function in solution and in simpler matrices, such as alginate (see Fig. S1 in the supplemental material).
Biofilms killed with colistin before (panels a and c) and after (panels b and d) the addition of glucose and glucose oxidase. (a and b) Average ratiometric intensity z-projections of data used to calibrate O2NS in the confocal microscope. (c and d) The 3D images show that oxygen concentration is uniform through the biofilm. This finding demonstrates that the nanosensors respond to oxygen concentrations as expected even in the complex biofilm environment.
Example application: measuring antibiotic-induced metabolic changes.Ratiometric signal data can be used to render 3D images of biofilms so that their structure can be visualized along with information about oxygen concentration within the biofilm (Fig. 3). The 2D projections of the raw 3D images (before ratiometric division) can be seen in Fig. S2 in the supplemental material. In addition, quantitative information on the oxygen concentration at an arbitrary location within the biofilm can also be determined via Matlab processing and graphed in a manner similar to data obtained from oxygen microelectrodes. This can be seen in Fig. 4. At four locations within the biofilm, oxygen concentration values were extricated from the Matlab array and graphed versus depth. There is large variation in oxygen concentration between locations, such as location 3 (a large central area with high DO concentrations) and location 4 (low DO concentrations throughout).
The 2D orthogonal (a) and 3D (b) views of a live P. aeruginosa PAO1 biofilm created using ratiometric intensity data. Raw confocal images of the PtTFPP (O2 sensitive) and DiA (reference) signals were processed and divided in ImageJ to produce a stack of images of the ratiometric intensity that recapitulates oxygen concentration throughout the biofilm.
Location versus oxygen concentration plot (right) of PAO1 image (left). Four locations across the biofilm were chosen for graphing, and error bars represent standard error of n = 9 pixels surrounding the chosen location. The four locations were chosen arbitrarily and show large spatial variation. Location 3, whose representative orthogonal views are shown, has a large gap in the center that has DO concentrations close to those in the surrounding medium. This finding highlights the massive spatial differences in biofilm oxygen concentrations.
The location in the biofilm we chose for temporal analysis (location 4 in Fig. 4) has an oxygen concentration of 1 mg/liter before the addition of colistin, with concentrations at or approaching 6 mg/liter at the edges of the biofilm, which can be seen in Fig. 5. This finding is consistent with a diffusion limitation of oxygen into the center of the biofilm as the bacteria consume it. As colistin penetrates the biofilm, oxygen concentration also increases, indicating cell death. The center of the biofilm takes the longest amount of time to approach the surrounding oxygen concentration, which can be seen in Movie S1 in the supplemental material.
Time lapse of oxygen concentration within PAO1 biofilm after the addition of 512 μg/ml colistin. Oxygen concentration within the biofilm reaches ambient values by 60 minutes. Error bars represent a standard error of n = 9 pixels surrounding the chosen location.
In clinical strain 1 (CS1), there is an overall higher concentration of oxygen than in PAO1, indicating lower metabolic activity or improved oxygen transport (Fig. 6). After the addition of colistin sulfate, the majority of the biofilm reaches an internal oxygen concentration approaching ambient conditions (indicating no metabolic activity) after 30 min. However, there is a small region that maintains an oxygen concentration below 1 mg/liter (see Fig. S3 and S4 in the supplemental material), indicating that there is continued aerobic respiration in this region. This is potentially a spatially inhomogeneous area occupied by cells with resistance to colistin.
Local oxygen concentrations over time within biofilms of P. aeruginosa strain PAO1 and three clinical strains after the addition of 512 μg/ml colistin. Error bars represent a standard error of n = 9 pixels surrounding the chosen location.
CS2 shows similar oxygen concentrations near the beginning of the experiment but maintains concentrations of <5 mg/liter for the majority of the observations. The oxygen concentration does not appear to increase until the 90-min mark, which could indicate a delayed reaction to or resistance to colistin.
CS3 shows large differences in structure compared with that of PAO1 (Movie S2 in the supplemental material) and the other two clinical strains (Movies S3 and S4 in the supplemental material) tested. In addition, there is a large increase in metabolic activity (indicated by a decrease in oxygen concentration) a few minutes after the addition of the colistin sulfate. After this initial change, oxygen concentration is variable over 90 min but is never above 4 mg/liter DO, indicating sustained metabolic activity within the biofilm. For PAO1 and CS1 samples, plating after experiments showed minimal or no growth (see Fig. S5 in the supplemental material), which supports the assumption of cell death when observing increased oxygen concentration within the biofilms. In the case of CS2 and CS3, both of which demonstrate sustained oxygen consumption after antibiotic addition, plating after experiments shows growth similar to plating before experiments. This supports the hypothesis that low oxygen concentration (high oxygen consumption) is indicative of active cell metabolism.
In addition to the ability to observe differences in clinical samples, our approach also enables the measurement and observation of biofilm responses to different antibiotics. The effect of two different antibiotics (colistin and tobramycin) on oxygen concentration in PAO1 biofilms can be seen in Fig. 7.
Comparison of the effects of different antibiotics (4 mg/ml tobramycin and 512 μg/ml colistin) on oxygen concentration within PAO1 biofilms. Error bars represent a standard error of n = 9 pixels surrounding the chosen location.
DISCUSSION
Method development: sensor design and characterization.Our nanosensor-based approach is a method to spatiotemporally monitor analyte heterogeneity and gradients in 3D systems, such as biofilms. Our approach utilizes nanosensors which are significantly smaller than the bacteria rather than larger, potentially minimizing impacts on biofilm growth while still maintaining the ability to monitor extracellular analyte concentrations. This is possible as the size of the nanosensors is small enough to not measurably impact growth, but the approximately 160-nm diameter of the nanosensors ensures that the particles are entrapped in the extracellular polysaccharides (EPSs) of the biofilm and do not leak into the surrounding medium substantially. Along with the comparison of LIVE/DEAD viability staining of biofilms with and without oxygen nanosensors (see Fig. S6 in the supplemental material), this suggests that the addition of nanosensors to growth medium does not significantly impact biofilm structure or bacterial activity.
Fabrication is straightforward, which enables quick application for monitoring analytes in a variety of settings. Because these sensors can be calibrated in a variety of matrices and equipment, they are adaptable to many different applications, including biofilms of other species, cell culture scaffolds, hydrogels, and other systems where 3D monitoring is valuable. In addition, the polymeric nanoparticle platform described here is adaptable to a wide variety of analytes that might be of interest in biological systems, including oxygen, glucose, ions, and small molecules. The O2NS we developed responds linearly to oxygen concentration, which is consistent with previously reported observations (42). Lee and Okura note that the oxygen-sensitive metalloporphyrin used here (PtTFPP) demonstrates excellent linearity in the range of interest, but the response does become nonlinear at oxygen concentrations greater than 21 mol% in the gas phase (42). The DiA exhibits minimal change at different oxygen concentrations, although photobleaching begins to occur near the end of the 90-min observation periods used in the above-described experiments (Fig. S2).
In our current implementation, the biofilm is grown with nanosensors in the medium. This makes the technique directly applicable to in vitro interrogation of bacterial isolates as we showed here, which were grown in such a manner as to mimic the microaggregate structure (43, 44) of biofilm infections found in the lungs of cystic fibrosis (CF) patients. However, due to the negligible diffusion of the nanosensors into the biofilm, this approach is not currently suitable for ex vivo or in situ biofilm samples. In addition, imaging was performed with confocal microscopy. These points limit nanosensors as a tool for scientific research rather than a rapid-use clinical tool. However, with a consideration to such research applications, the fabrication and use of these nanosensors build upon equipment already present in many laboratories and core facilities. This technique is not necessarily faster than culture-based techniques for antibiotic screening applications since biofilm growth (up to 72 h for mature biofilms) is required. However, the use of nanosensors is well suited for research applications (i.e., measuring metabolism of subpopulations and biofilm homogeneity) that cannot be addressed with current tools.
Example application: measuring antibiotic-induced metabolic changes.The ability to differentiate between the three clinical strains demonstrates the value of this approach. Using these oxygen nanosensors, it is possible to obtain information about the entire biofilm and not just a single point or a few points within it. Biofilms formed by PAO1 have a microaggregate structure (biofilms ∼100 μm in size) similar to what is seen in the lungs of cystic fibrosis patients (43–45). In these biofilms, we see spatial variations (Fig. 4) in dissolved oxygen concentration and secondary structures that are reminiscent of liquid and nutrient transport channels like those observed by Wilking et al. (46) and others (47, 48). It is also possible to see parts of the biofilm that are either not surface attached or loosely attached to the surface, which would be lost with the physical disruption of monitoring with the standard microelectrode-based measurements. In the case of the first clinical strain, we show that this technique can elucidate areas of different metabolism (Fig. S4). If one were to interrogate this biofilm with a microelectrode, only measuring a small area (approximately 50 μm by 50 μm) would obscure the heterogeneity of the biofilm as a whole. This localization of persistent oxygen consumption could indicate the presence of antibiotic-resistant cells or persister cells within the biofilms. In the cases of clinical strains 2 and 3, this approach is able to distinguish responses to antibiotic attack as opposed to only demonstrating resistance. We note different rates of DO concentration variation between the two, with these variations occurring on measurable time scales (around 15 minutes for CS3 and 40 minutes for CS2). CS3 also had remarkably different architecture, producing a larger structure than those seen in PAO1 and the other clinical strains tested. This highlights our ability to individually profile these strains with improved measurements over traditional approaches. The ability to obtain information on spatial homogeneity of biofilm metabolism could be elevated with the addition of sensors for other analytes. Two-dimensional maps can provide x-y spatial information on biofilms but are unable to provide z axis information. Because biofilms have a complex, three-dimensional structure with channels for the movement of nutrients, water, and oxygen, the complete understanding of dynamics within the system requires three-dimensional information. Our approach allows us to rapidly gather all spatial information, and we are able to do so without disturbing the structure. Temporal information is limited only by the imaging equipment’s ability to capture data. Similar to any confocal microscopy experiment, we can increase time resolution at the expense of spatial resolution or with more advanced imaging hardware.
Both antibiotics used in this study were chosen for their relevance to clinical therapy for CF patients. Tobramycin belongs to the aminoglycoside class of antibiotics and shows the greatest antipseudomonal activity of the aminoglycosides (49). Because of this, it is a common choice for combination therapy in CF patients (50). Colistin is a polymyxin antibiotic that is often used to treat P. aeruginosa airway infections in CF patients. Polymyxins are popular, as they usually retain activity against multidrug-resistant strains of P. aeruginosa (51). Kill curves in P. aeruginosa PAO1 biofilms as well as clinical strains have been demonstrated by Hengzhuang et al. using the Calgary biofilm device method (52), and our sensors were able to recapitulate these data with a finer time resolution. The results obtained from O2NS measurements in Fig. 7 are consistent with the kill curves obtained by Hengzhuang et al. (52) using the modified Calgary biofilm device method, although our results also provide information on the structure of the biofilms and spatial distribution of high- and low-oxygen concentration regions within the biofilm. Calculations (53) and experimental data (54) from microelectrodes show that oxygen should penetrate P. aeruginosa biofilms attached to surfaces to a depth of 60 to 70 μm, but this assumes diffusion into a uniform biofilm. We note in Fig. 4 that there is considerable spatial variation within the biofilm. Therefore, we chose a point near the center of the larger aggregate for temporal analysis, which presumably has the largest diffusion barrier.
Conclusions.Here, we used polymeric oxygen-sensitive nanosensors to monitor the response of lab and clinical Pseudomonas aeruginosa biofilms to antibiotic attack with three-dimensional spatial data collection and 5-min temporal resolution. The nanosensors’ small size, rapid response, and ratiometric signal allow for detailed interrogation of clinical samples to obtain antibiotic susceptibility information as well as data from which pharmacodynamic parameters can be determined. Additionally, these oxygen nanosensors enable the elucidation of biofilm heterogeneities that microelectrodes and other techniques cannot capture, such as microaggregate size and shape, localization of oxygen consumption, and even biofilm features which are not surface attached. We are able to measure differences in the response of three clinical samples to antimicrobial attack and overall metabolism via changes in oxygen concentration. This approach for biofilm monitoring is straightforward to implement for in vitro spatial and temporal monitoring of biofilm metabolism, and future work will focus on the addition of the nanosensors to preexisting biofilms. The sensor platform’s adaptability and ease of manufacture enable the monitoring of multiple or different analytes within the biofilm structure in future work as well.
MATERIALS AND METHODS
High-molecular-weight poly(vinyl chloride) (PVC), bis(2-ethylhexyl) sebacate (BEHS), tetrahydrofuran (THF), dichloromethane (DCM), and phosphate-buffered saline (PBS) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Platinum (II) meso-tetra(pentafluorophenyl)porphine (PtTFPP) was purchased from Frontier Scientific (Logan, UT, USA). 1,2-Dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-750] ammonium salt in chloroform (PEG-lipid) was purchased from Avanti Polar Lipids (Alabaster, AL, USA). 4-Di-16-ASP {4-[4-(dihexadecylamino)styryl]-N-methylpyridinium iodide} (DiA) was purchased from Thermo Fisher Scientific (Waltham, MA, USA).
NS fabrication.A total of 30 mg of polyvinyl chloride (PVC) was weighed into a 2-ml vial and combined with 66 μl of BEHS. In a separate vial, 5 mg of Pt(II) meso-tetra(pentafluorophenyl)porphine (PtTFPP) and 0.2 mg 4-Di-16-ASP {4-[4-(dihexadecylamino)styryl]-N-methylpyridinium iodide} (DiA) were dissolved in 500 μl of tetrahydrofuran (THF). The THF and dissolved dyes were then added to the PVC/BEHS mixture and vortexed for 1 min until all solids were dissolved. A total of 750 μl of dichloromethane (DCM) was then added to the vial, and the mixture was vortexed thoroughly. The optode cocktail was then stored at 4°C until use.
To fabricate nanosensors from the optode cocktail, 2 mg PEG-lipid (80 μl of a 25-mg/ml solution in chloroform) was dried in a 4-dram scintillation vial and then resuspended in 5 ml of PBS (pH 7.4) with a probe tip sonicator (Branson Digital Sonifier 450; Branson Ultrasonics Corporation, Danbury, CT) for 30 s at 20% intensity. A total of 125 μl of the optode cocktail mixture was injected into the PBS/PEG-lipid solution under probe tip sonication (3 min, 20% intensity). Following sonication, excess polymer was removed by filtration via a 0.22-μm syringe filter (Pall Corporation, Port Washington, NY).
NS characterization.Oxygen-sensitive nanosensors were calibrated on an Avantes AvaSpec-ULS2048L StarLine versatile fiber-optic spectrometer (Apeldoorn, the Netherlands) with a 100-μm slit width. O2NSs were sealed in an airtight screw top quartz cuvette (Starna Cells, Atascadero, CA), and the oxygen content of the sample was altered by bubbling a compressed air/N2 mixture through the cuvette for 30 min. The O2NS sample was then excited with a 450-nm laser diode, and spectra were collected at various concentrations of oxygen. A linear regression of ratiometric signals was performed in GraphPad Prism 7 software (La Jolla, CA). Values were fit to the Stern-Volmer relationship describing collisional quenching of an excited species where the quencher Q is molecular oxygen. In the equation below, If0 is the ratiometric fluorescence intensity in the absence of oxygen, If is the ratiometric fluorescence intensity at the given concentration of oxygen [Q], and KSV is the Stern-Volmer constant.
The ratiometric mode is a method of internal self-calibration that is made possible by the comparison of emission at two different wavelengths (55). In this case, If refers to the ratio of the oxygen-sensitive fluorescence intensity of PtTFPP at 650 nm and the insensitive fluorescence intensity of DiA at 585 nm (I650/I585). This allows the sensors to overcome limitations, such as concentration variation or heterogeneity of the local microenvironment, which will impact both fluorescent peaks similarly, while oxygen will only impact one of the peaks (56, 57).
O2NSs were calibrated in alginate hydrogels on a Zeiss LSM780 confocal microscope. Hydrogels with embedded O2NS were fabricated according to Fletcher et al. (58). Briefly, 0.5 ml of 3 wt% alginate was mixed with 200 μl of O2NS and 40 μl of 0.1 M CaSO4. The mixture was cast between two glass plates to produce hydrogels of 380-μm thickness, which was measured with a micrometer. The resulting hydrogels were placed into chamber slide wells and covered with 500 μl of PBS. Images were taken under ambient conditions (21% O2 or 6.65 mg/liter dissolved O2 at the lab elevation of 5,750 ft above sea level) and under oxygen deficiency (0% O2 or 0 mg/liter dissolved O2). Chamber slide wells were deoxygenated with 10 mM glucose and 2 IU/ml glucose oxidase, as described by Baumann et al. (59). The average intensities of the biofilms in the images at 0% O2 and 21% O2 were then used in a linear regression to determine the Stern-Volmer constant KSV. In both calibration scenarios, the intensities of the PtTFPP signal at 650 nm and the intensities of the reference dye (DiA) signal at 585 nm were divided to form a ratiometric signal. The ratiometric signal at 0 mg/liter dissolved oxygen (DO) was then divided by the ratiometric signal at each concentration to create a linear calibration curve. The linear Stern-Volmer relationship between If0/If and oxygen concentration was used to fit the calibration data (60).
Image analysis was performed in Matlab using the Batch Image Processor. The image stack was transformed into a 3D array in Matlab with dimensions of 1,024 by 1,024 by the number of slices in the image stack containing 8-bit brightness values (intensity values, 0 to 255). These slices were converted via the Stern-Volmer calibration curve function into DO values in milligrams per liter, which could be extricated and graphed.
Dynamic light scattering (DLS), zeta potential via phase analysis light scattering (PALS), and mobility measurements were performed on a Brookhaven zetaPALS instrument with Particle Solutions software v 2.2 (Brookhaven Instruments Corporation, Holtsville, NY).
Biofilm construction and growth.(i) P. aeruginosa PAO1 (ATCC 15692). Biofilm construction and growth procedures were adapted from methods by Kirchner et al. (61). Briefly, P. aeruginosa strains were plated onto Luria Bertani (LB) agar (Sigma-Aldrich, St. Louis, MO, USA) from frozen stocks and incubated at 37°C for 24 h. One colony was pulled and dispersed in 1 ml of LB broth and incubated for 24 h at 37°C. Liquid culture was diluted to an optical density at wavelength of 600 nm (OD600) of 0.05. Biofilms of PAO1 were grown statically in a 16.7% vol/vol solution of oxygen nanosensors in PBS mixed with LB broth (Sigma-Aldrich). A total of 600 μl of O2NS + LB was added to each well of a 4-well LabTek chamber slide and then inoculated with 5 μl of liquid P. aeruginosa culture at an OD600 of 0.05. Chamber slides were placed in a humidity chamber and incubated for 72 h at 37°C.
(ii) P. aeruginosa clinical strains. Respiratory samples collected for clinical care were processed by the clinical microbiology laboratory at Children’s Hospital Colorado using standard CF Foundation guidelines for culture. P. aeruginosa isolates were frozen and shipped to study investigators at the Colorado School of Mines. All three clinical strains are deidentified and in the manuscript are labeled as clinical strain 1 (CS1), CS2, and CS3. With all three of the provided isolates, we expanded and grew the biofilms as detailed above with the PAO1 samples.
(iii) Biofilm characterization. Three biofilms of PAO1 were grown as described above, namely, one containing no polymer nanoparticles, one containing polymer nanoparticles with no fluorescent components, and one containing O2NS. The FilmTracer LIVE/DEAD biofilm viability kit (Thermo Fisher Scientific) was used for biofilm viability. A total of 12 μl of SYTO9 green fluorescent nucleic acid stain (3.34 mM in dimethyl sulfoxide [DMSO]) and 12 μl of propidium iodide (20 mM in DMSO) were added to 4 ml of PBS, and 400 μl of the resulting solution was added to each well of the chamber slide and incubated at room temperature for 15 min in the dark. The solution was then aspirated, and each well was washed 2 times with sterile PBS. A total of 400 μl of 10% formalin was added to each well and incubated for 30 min in the dark. Formalin was then aspirated, and each well was washed 2 times with PBS. A final volume of 200 μl PBS was added to each well before imaging. Three regions of interest containing the entirety of the biofilm were chosen for each sample containing no nanoparticles (n = 4 images) or containing blank polymer nanoparticles (n = 3) or O2NS (n = 3).
(iv) Antibiotic testing. All antibiotic testing was performed on live biofilm samples that had not been stained or fixed prior to imaging. A total of 10 μl of 10 mg/ml tobramycin in PBS was added to the 490 μl of PBS already present in the chamber slide well, for a final tobramycin concentration of 4 mg/ml (62). This was administered to each biofilm sample, and images were taken every 5 min for 2 h.
For comparison, a final colistin concentration of 512 μg/ml was administered to each biofilm sample in a similar manner, and images were taken every 5 min for 2 h. Here, colistin was administered as colistin sulfate as opposed to the prodrug colistimethate sodium.
ACKNOWLEDGMENTS
This work was supported by Children’s Hospital of Colorado Research Institute, the Office of Research and Technology Transfer at Colorado School of Mines, and start up funds from Colorado School of Mines. Imaging experiments were performed in the University of Colorado Anschutz Medical Campus Advance Light Microscopy Core supported, in part, by NIH/NCATS Colorado CTSI grant number UL1 TR001082. The contents of the paper are the authors’ sole responsibility and do not necessarily represent official NIH views.
We thank the Children’s Hospital of Colorado (CHCO) sample bank for providing the clinical isolates.
FOOTNOTES
- Received 17 May 2019.
- Accepted 9 August 2019.
- Accepted manuscript posted online 16 August 2019.
Supplemental material for this article may be found at https://doi.org/10.1128/AEM.01116-19.
- Copyright © 2019 American Society for Microbiology.
