Introduction

Healthcare-associated bacterial infections, particularly those that arise in conjunction with medical device implants and surgery, are difficult to distinguish from sterile inflammation without the use of invasive tissue sampling procedures1. In recent years, these infections have contributed to extended post-operative hospitalizations and increased patient mortality following routine medical procedures2,3. In addition to methods to prevent these infections, there is also a critical demand for non-invasive diagnostic tools that can rapidly and accurately detect and differentiate bacterial infections from sterile inflammation.

Conventional methods used to diagnose bacterial infections include magnetic resonance imaging (MRI), X-ray, ultrasonography and computed tomography (CT)3. However, these techniques are not specific to infection but rather display anatomical indications of infection-related pathologies. This limits their utility in detecting early stages of infection when notable anatomical signs may not yet be observable and when interventions are most effective. Molecularly targeted photoacoustic imaging (PAI) has emerged as an attractive modality for detecting early-stage infections because it offers the opportunity to develop infection-specific contrast agents. PAI also has improved depth resolution, as compared to purely optical imaging techniques (e.g., fluorescence imaging (FI)) since the generated acoustic waves experience less scattering in tissue than the photons generated during FI4. Additionally, the complimentary prerequisite of light absorption for both PAI and FI, highlights the possible utility of traditional fluorophores for PAI.

Recent reports on optical and PA imaging of infection have successfully utilized fluorophore-conjugated antibiotics5,6,7,8, compounds metabolized by bacteria9,10,11, and monoclonal antibodies6,12. However, antimicrobial resistance may impede the localization of certain probes to the site of infection, and monoclonal antibodies are known to experience long circulation times and could have a limited ability to penetrate mature biofilms due to their size. Zinc(II) dipicolylamine (Zn-DPA)-based bioconjugates can potentially be used to address these limitations as they have been established as suitable bacteria-targeting molecules due to their affinity for the anionic outer surfaces of bacteria cells, as compared to healthy mammalian cells which typically exhibit a net neutral surface6,13,14. The fluorescent Zn-DPA-based probe, PSVue794, has been well characterized in the literature for bacterial detection and is commercially available as a reagent for fluorescence imaging of bacteria15. Moving beyond fluorescence imaging methods, we evaluated the use of PSVue794 as a PAI probe (PIP) for imaging bacteria. The PA signal properties in phantom studies and in vitro bacteria binding were evaluated before radiolabeling the PIP with 99mTc to assess the biodistribution profile in a mouse model of bacteria-induced myositis. Proof-of-concept in vivo PAI studies were pursued to verify that PSVue794 could generate detectable PA signal at sites of infection compared to sites of sterile inflammation.

Results

PA Signal analysis of 2

Compound 2 (Fig. 1) was found to generate a PA signal within the range of 680–970 nm and possessed a signal maximum at 800 nm (Fig. 2a,c). An overlay of the PA signatures from 2 and those of oxygenated and deoxygenated hemoglobin (Fig. 2a) shows that 2 generates a unique signal, distinguishable from both blood chromophores. The signal of 2 was quantified in the presence of whole mouse blood using multiwavelength unmixing scans and a significant increase in the signal of 2 detected in blood (0.129 ± 0.012 au) was achieved, compared with the background (0.045 ± 0.002 au, where P = 0.006) (Fig. 2b). Additionally, the PA maximum of 2 was identified in the merged PA spectrum of blood + 2 (indicated by the dashed line at 810 nm), while there was no peak observed at ~ 810 nm in the spectrum of blood alone (Fig. 2c).

Figure 1
figure 1

PSVue794 (2) reagent kit complexation.

Full size image
Figure 2
figure 2

Characterization of the PA signal emitted by PSVue794 (2): (a) Normalized PA signal of 2 (in Diluent X) overlaid with the PA spectra of oxygenated and deoxygenated hemoglobin. (b) Quantified PA average single of 2 after spectrally unmixing samples of blood with or without 2. (c) Multiwavelength unmixing scans of the PA signal for 2 (20 µM), blood alone, and 2 (64 µM) in blood, with the PA spectra below the image. Signal characterization was performed in PA phantom. The white dashed line on the spectrum of blood + 2 indicates the peak of 2 within the signal generated from blood. Green = 2, red = oxygenated hemoglobin, blue = deoxygenated hemoglobin, grey-scale = ultrasound B-mode.

Full size image

In vitro bacterial binding analysis of 2 using fluorescence and photoacoustic detection

After subtraction of the signal observed in a blank sample, the fluorescence intensity of bacterial cell suspension incubated with 2 (60,263 ± 6136 au) was significantly higher than the no-cell control (20,167 ± 701 au, where P = 0.007), and the signal-to-background (S/B) ratio was 2.98 ± 0.20 (Fig. 3a). The same samples were then analyzed on a PA phantom, where the bacterial cell suspension was found to have a significantly higher PA signal (0.075 ± 0.014 au) above the no-cell control (0.024 ± 0.002 au, where P ≤ 0.005) (Fig. 3b). No significant differences were present between the no-cell control and the blank controls. A visual representation of the PA signal present within each phantom sample is depicted in Fig. 3c, where a green signal, pertaining to 2, was only seen in the bacterial cell suspension that was incubated with 2. A S/B ratio of 3.06 ± 0.30 was calculated from the data displayed in Fig. 3b.

Figure 3
figure 3

Florescent and PA evaluation of in vitro bacteria binding of 2. (a) Fluorescence intensity measurements of 2 incubated with or without bacteria. Signal was normalized by subtracting the values from a blank with or without bacteria. (b) PA average signal of 2 in each sample. (c) Phantom PA images of the samples with the PA spectra of each sample overlaid. Where green = 2.

Full size image

Radiosynthesis of [99mTc]Tc-PSVue794 (6)

A previously reported method, where technetium tricarbonyl [99mTc][Tc(CO)3(H2O)3]+ (4) (Figure SI-1) was used to label a dipicolylamine (DPA) ligand16,17, was adapted to generate [99mTc]-TcPSVue794 (6). Several parameters were explored (Table SI-1) to produce 6 in high yield and purity and the optimal conditions involved stirring the Apo-PSS precursor (1, 0.34 µmol) with 4 at room temperature for 1 h (Fig. 4). Radio-TLC analysis of the crude material showed a 90% radiochemical conversion to [99mTc]-TcApo-PSS794 (5) and a radiochemical purity (RCP) of 99% after passing the reaction mixture through a C-18 cartridge (Figure SI-2a,b,d). Zinc nitrate was then added to 5 to form 6, with a RCP of 99%, assessed using radio-TLC and gamma-HPLC (Figure SI-2c,e). The total preparation time was 2 h and the desired product was obtained in a non-decay corrected activity yield (Equation SI-1) of 52 ± 8% (n = 5) (starting activity of 108 ± 9 MBq) and molar activity of 0.76 ± 0.07 MBq/µmol at end of synthesis. The stability of compound 6, incubated in the formulation buffer at room temperature and mouse serum at 37˚C, was assessed over 18 h. Similar to previous 99mTc-radiolabeled DPA compounds,17 6 was stable at all time points (Figure SI-3 and Table SI-2).

Figure 4
figure 4

Radiosynthesis of [99mTc]Tc-PSVue794 (6) from the Apo-PSS precursor (1) and [99mTc][Tc(CO)3(H2O)3]+ (4). Note that the reaction mixture would also contain 1 (not shown here) given the excess of ligand compared to 99mTc used in a typical radiolabeling experiment.

Full size image

In vitro bacteria binding and quantitative biodistribution analysis of 6

An initial in vitro test demonstrated that 6 (37 kBq) binds to S. aureus cells. 15.0 ± 2.1% of the total activity successfully bound to the bacteria, while the no-cell control percent binding was 1.3 ± 0.2% (P < 0.001) (Fig. 5a). Subsequently, the whole-body distribution of 6 (0.74 MBq/mouse) was assessed in healthy naïve mice and in a mouse model of S. aureus-induced myositis (n = 4) (Fig. 5b and Table SI-3). A significant increase in the percent injected dose/gram (%ID/g) was noted between the infected muscle tissue (1.68 ± 0.61% ID/g) and contralateral muscle tissue (0.17 ± 0.07% ID/g, where P = 0.01), generating a target-to-non-target (T/NT) ratio of 10.1 ± 1.1 (Fig. 5c). No significant differences were found in the % ID/g in the spleen in the healthy naïve vs. infected mice. Differences were noted between the gall bladder and large intestine/caecum of the healthy vs. infected mice.

Figure 5
figure 5

(a) In vitro S. aureus binding analysis of [99mTc]Tc-PSVue794 (6). Percentage of total activity (CPM) of 6 present after incubation with or without S. aureus bacteria cells. In vivo biodistribution analysis of [99mTc]Tc-PSVue794 (6) in healthy naïve vs. infected mice (n = 4 per group) (b and c). Infected mice were inoculated with 1 × 108 CFUs of S. aureus in the right hindlimb muscle and with an equal volume of HBSS into the left (contralateral) hindlimb muscle. (b) Percent injected dose/gram (%ID/g) of 6 in each tissue. (c) Comparison of %ID/g of 6 in the infected vs contralateral and naïve muscle tissue.

Full size image

Bacteria- and LPS-induced myositis models

A model of sterile inflammation (Lipopolysaccharide [LPS]-induced myositis) was developed, and tissues were analyzed and compared to the previously described bacterial infection model. The hematoxylin and eosin (H&E) stained tissues (Fig. 6) presented collections of neutrophils in the muscle and perimysium of both LPS- and bacteria-injected tissue, while the spleen of LPS-injected mice exhibited lymphocytic pyknosis and karyorrhexis. The LPS-injected mice were diagnosed with a localized acute inflammation effect, resulting from the injection of material with some immunogenicity, indicated by the observed acute and thrombosing myositis, peri-myositis, and splenic lymphocytolysis. The bacteria-injected mice had acute peri-myositis with intense neutrophilic infiltrations often encompassing bacterial cocci with associated cellular necrosis. Myeloid hyperplasia was present. The contralateral vehicle control-injected muscles had no pathology on histologic analysis.

Figure 6
figure 6

Representative images of H&E-stained tissues highlighting key histological features. a-c. Tissue fields from LPS-injected mice, including muscle (a, b), there are large fibrinous thrombi infiltrated by neutrophils in veins. In the spleen (c) a proportion of cells in the lymphoid perivascular cuffs have karyorrhectic or pyknotic nuclei. (d–f) Tissue fields from the muscle (d, e) of bacteria-injected mice, with patchy areas of interstitium markedly widened by severe neutrophilic infiltration (d). In some areas, a massive neutrophilic infiltration encompasses large areas of necrosis (e). There is myeloid hyperplasia in the bone marrow (f).

Full size image

Photoacoustic imaging of LPS-induced sterile inflammation using 2

Following the administration of 2 (75 µL, 1 mM) to the LPS model, multiwavelength unmixing scans (Fig. 7a) did not depict a notable PA signal (green) within the inflamed muscle 12 h post-injection compared to the pre-injection background or the vehicle control muscle (contralateral). Determination of the average PA signal (Fig. 7b) was consistent with this result as the signals within the inflamed muscle (0.062 ± 0.012 a.u.), the pre-injection background (0.058 ± 0.006 a.u.), and the vehicle control muscle (0.066 ± 0.008 a.u.) were not statistically different from one another. The T/NT (inflamed vs. vehicle control muscle) and the S/B (inflamed muscle pre- vs. 12 h post-injection) ratios were 1.04 ± 0.17 and 1.07 ± 0.11, respectively. The signals of oxygenated and deoxygenated hemoglobin were observed in all the tissues through spectral analysis (Fig. 7c) and no spectral characteristics were noted to indicate the presence of 2.

Figure 7
figure 7

PAI of LPS-induced sterile myositis using 2: (a) Representative in vivo PA multispectral unmixing images (pre-injection and 12 h post-injection) of mice injected I.V. with 2 (1 mM, 75 µL, n = 3). (b) Bar graph representing the quantified 3D PA average signal, derived from the scans in (a), within the inoculation sites of the mice throughout the study. (c) Spectral analysis of the inoculation sites pre-injection and 12 h post-injection. ROIs drawn over the inoculation sites (white) in the unmixed Spectro scans with the spectral signature with the ROI overlaid (White dashed line indicates the spectral signature of 2). For all multiwavelength unmixing scans, greyscale = Ultrasound (US) images overlaid with unmixed PA signal, where green = 2. The signals from oxygenated and deoxygenated hemoglobin have been subtracted from each image.

Full size image

Photoacoustic imaging of S. aureus-induced bacterial infection using 2

Following administration of 2 (75 µL, 1 mM), the multiwavelength unmixing scans depicted a notable PA signal (green) within S. aureus-infected muscle (12 h post-injection) compared to the background (pre-injection) and the vehicle control muscle (contralateral) (12 h post-injection) (Fig. 8a). This was verified by quantifying the PA average signal within the images (Fig. 8b), where, at 12 h post-injection, the infected muscle (0.176 ± 0.011 a.u.) was significantly higher than the pre-injection background (0.081 ± 0.004 a.u., where P ≤ 0.03) and the vehicle control muscle (0.076 ± 0.017 a.u., where P ≤ 0.03). The T/NT tissue (infected vs. vehicle control muscle) and S/B (infected muscle pre- vs. 12 h post-injection) ratios were 2.40 ± 0.49 and 2.17 ± 0.05, respectively. Spectral analysis was used to verify that the signal in the infection site matched the expected spectral signature of 2, with a signal maximum of 810 nm (Fig. 8c). The signal associated with deoxygenated hemoglobin (760 nm) was present in all the tissue.

Figure 8
figure 8

PAI of S. aureus-induced myositis using 2: (a) Representative in vivo PA multispectral unmixing images (pre-injection and 12 h post-injection) of mice injected I.V. with 2 (1 mM, 75 µL, n = 3). (b) Quantified 3D PA average signal, derived from the scans in (a), within the inoculation sites of the mice throughout the study. (c) Spectral analysis of the inoculation sites pre-injection and 12 h post-injection. ROIs drawn over the inoculation sites (white) in the unmixed Spectro scans with the spectral signature with the ROI overlaid (White dashed line indicates the spectral signature of 2). For all multiwavelength unmixing scans, greyscale = US images overlaid with unmixed PA signal, where green = 2. The signals from oxygenated and deoxygenated hemoglobin have been subtracted from each image.

Full size image

Discussion

PA phantom analysis of 2 shows that it possesses a unique PA spectrum as compared to endogenous chromophores, oxygenated and deoxygenated hemoglobin. The PA signal of the dye was assessed in whole blood, using a concentration of 64 µM, which represents injecting 75 µL of 1 mM of 2, (a dose recommended in the reagent kit), into the circulating blood volume of a 20 g mouse.18 The signal from 2 was observable in the spectrum of the sample and successfully quantified after unmixing it from the blood chromophores, demonstrating that the amount of dye used in this experiment should be sufficient to produce an observable PA signal in vivo.

Bacteria binding of 2, using both fluorescence and PAI, was assessed following previously published protocols.14 Using sealed phantom tubing, a PA S/B ratio of 3.06 ± 0.30 was obtained when comparing the bacterial cell suspensions incubated with 2 versus the vehicle control. This aligned with the ratio that was determined using fluorescence intensity measurements (S/B = 2.98 ± 0.20), indicating that PA detection is consistent with the fluorescence measurements routinely used for in vitro studies.

To facilitate future comparisons using novel optimized PIPs, quantitative biodistribution analysis of PSVue794 was pursued. Reliable quantitation using FI and PAI is highly dependent on the solubility of the constructs and the absence of aggregation. This is because cyanine dyes, such as PSVue794, tend to aggregate in aqueous media, which induces shifts in the photophysical properties of the compound, and could impact the signal intensities observed in vivo19,20. Building on previous reports that employed DPA as a chelator for the 99mTc-tricarbonyl core16,17, PSVue794 was successfully radiolabelled to generate a gamma-emitting version of PSVue794 (6). Radiolabeling of PSVue794 enabled quantitative whole-body distribution analysis of the compound in a mouse model of bacterial infection and will facilitate future quantitative comparisons using different PIPs and models of disease.

The biodistribution of 6 in healthy naïve and infected mice were evaluated. By 12 h post-injection, it was found that the radiolabeled PIP was predominantly processed by the hepatobiliary system, exhibited by the elevated concentrations observed in the gall bladder, large intestine, liver, and spleen. The biodistribution profile of 6 was comparable to previously reported data on a 99mTc-labeled ZnDPA compound, where localization to gastrointestinal organs was observed.17 No significant differences were found in the % ID/g of the spleen and liver in healthy naive vs. infected mice, which indicated that uptake in these organs was most likely driven by the lipophilic nature of the dye and it’s technetium (I) tricarbonyl core. Significantly higher %ID/g was observed in the gall bladder and large intestine/caecum of the infected mice compared to the naïve mice. While this result was not further explored, recent literature reports highlight the possibility of bacteria spreading from the bloodstream into the gall bladder, where it can then seed into the intestines21. Additionally, it is possible that the stress of the bacterial infection on the immune system of the mice induced a state of dysbiosis, a phenomenon where imbalances in gut microbiota are commonly observed22.

The T/NT ratio of 10.1 ± 1.1, achieved using 6, is somewhat higher than that of a previous ZnDPA derivative that had a T/NT ratio of 6.2, where 111In was chelated to an alternate site on the molecule, maintaining the bis-Zn2+ complex23. The radiolabeled PIP (6) retained the ability to target sites of bacterial infection with comparable distribution properties to those of similar published radiolabelled probes17.

Subsequent studies utilizing PAI demonstrated the ability of the PIP to localize to and generate detectable PA signal at sites of bacterial infection as compared to sites of sterile inflammation. Bacterial cells activate inflammatory pathways, typically upon cellular recognition of pathogen-associated molecular pattern molecules produced by the bacteria24. Consequently, a model of sterile inflammation was produced to verify that any observed localization of the PIP in the bacterial infection was not due to innate inflammatory processes or cells present. A model of sterile myositis using LPS was adapted from literature to induce muscular and systemic inflammation25,26,27. Although PSVue794 has affinity towards the anionic components of apoptotic cells28, which has been documented to localize to sites of cell death to a lesser extent than to sites of bacterial infection29, the models used here were designed to eliminate this variable in order to showcase the specific interaction of the PIP with bacterial cells over immune cells that are likely also present at the site of bacterial infection. The histological assessment of tissues treated with LPS or with S. aureus confirmed that the two conditions effectively induced states of sterile or bacteria-induced myositis, respectively. The T/NT and the S/B ratios derived from the PAI analysis in LPS-induced (T/NT = 1.04 ± 0.17, S/B = 1.07 ± 0.11) and S. aureus-induced (T/NT = 2.40 ± 0.49, S/B = 2.17 ± 0.05) myositis showed that the Zn-DPA targeting moiety on 2 facilitates the specific localization of the PIP at sites of bacterial infection.

In conclusion, we demonstrated that compounds 2 and 6 can be used to specifically detect sites of bacterial infection over sites of sterile inflammation and showcased the potential of using PAI as a modality to detect infection. While 2 is currently commercially available, further optimization of bacteria-targeting PIPs to achieve superior aqueous solubility and optical absorption could lead to further improvements in T/NT and S/B ratios by facilitating better clearance profiles in healthy tissues along with improved signal unmixing capabilities from blood. Additionally, efforts toward addressing the role of cell death in clinically relevant models of healthcare-associated bacterial infections will be critical in establishing the utility of ZnDPA-based diagnostic agents for imaging bacterial infection.

Methods and materials

General materials and instrumentation

Unless otherwise noted, all reagents and solvents were ACS grade, purchased from commercial suppliers, and used without further purification. Deionized water was used for all experiments requiring the use of water. PSVue794 Reagent Kit was purchased from Cedarlane Labs (25101–1) and was used to prepare PSVue794 (2) from the Apo-PSS precursor (1) and a solution of 2 mM Zinc nitrate, 30 min before use. Lipopolysaccharide from Escherichia coli O55:B5, ready-made solution (1 mg/mL), 0.2 µm filtered, was purchased from Sigma-Aldrich (Cat#L5418-2ML).

Absorbance and fluorescence scans were collected on a Tecan Infinite M1000 plate reader using Corning 96-well, clear bottom plates (non-treated surface) (Corning 3631), PerkinElmer film was used to seal any plates containing bacteria cells. Microwave reactions were performed using a Biotage Initiator 60 microwave reactor under standard settings. Radio-TLC was performed using a Bioscan AR-2000 imaging scanner on iTLC-SG glass microfiber chromatography paper (SG10001, Agilent Technologies) plates and Aluminum oxide plates using 1% HCl in methanol as the eluent. For each TLC performed, plates were spotted with ~ 3 µL (481 kBq). A background (represented as a blue underline) was integrated on each radio-TLC plate and subtracted from the integration of each peak. Sep-Pak C18 Plus Light Cartridge (55–105 µm) (WAT023501, Waters) was used for isolating radiolabeled products. High-performance liquid chromatography (HPLC) of the radiolabeled compounds was performed on a Waters 1525 Binary (Midford, MA), monitored simultaneously with 2998 photodiode array detector at 245/700 nm and in line radioactivity Bioscan gamma detector with NaI (TI) scintillator using the Empower software package. Analytical HPLC of each compound was performed on a Gemini 5 µm C18 110 Å (250 × 4.6 mm) Column (00G-443-E0, Phenomenex), operating at a flow rate of 1 mL/min. A PerkinElmer Wizard 1470 automatic gamma counter was used to measure the amount of radioactivity in samples from the bacteria binding assay and the biodistribution studies.

Pertechnetate [99mTc][TcO4]− was obtained in 0.9% w/v saline from a 99Mo/99mTc generator supplied by Lantheus Medical Imaging. Caution: these materials are radioactive and should only be used in a properly licensed and equipped facility. Potassium boranocarbonate was prepared according to a literature method30, as was [99mTc(CO)3(OH2)3]+16.

HPLC method

Solvent A: Water + 0.1% Trifluoroacetic acid (TFA), Solvent B: Acetonitrile + 0.1% TFA. Gradient: 0–2 min 90% A, 2–6 min 90–40% A, 6–8 min 40% A, 8–10 min 40–10% A, 10–11 min 10–90% A, 11–15 min 90% A.

Photoacoustic phantom imaging

PA phantom work was performed using a Vevo Phantom imaging chamber and the Vevo-3100/LAZR-X (FUJIFILM VisualSonics, Inc., Toronto, ON, Canada) imaging system equipped with a 680–970 nm laser. Samples were injected into Vevo Contrast Agent Phantom tubing (PN52807) that was threaded into the Vevo Phantom chamber. A 30 MHz, linear array US transducer equipped with integrated fibre optic cables (MX400 and LZ-400 [15–30 mJ(cm3)−1, 20 Hz repetition rate, 10 ns pulse width], FUJIFILM VisualSonics, Inc., Toronto, ON, Canada) was positioned overtop of the tubes.

Spectral analysis was performed using ‘Spectro’ scans, where the PA signal of a 2D, x,y-cross-section of a sample is quantified at each wavelength (690–970 nm, step size of 5 nm, ~ 1 min/spectro scan, axial resolution of 50 µm) and is used to plot the spectral curve or ‘spectral signature’ from a defined region of interest (ROI). From this, the signal intensity of the chromophore of interest can be quantified by the signal unmixing algorithm utilized in the VevoLAB software (VisualSonics, Inc., Toronto, ON, Canada). This scan was also be used to demonstrate the merged spectra of multiple chromophores mixed. The 2D multispectral unmixing analysis was then performed on the spectrum, where the pre-programmed PA spectra of oxygenated and deoxygenated hemoglobin were unmixed from the spectra of 2, and were taken for each material in triplicate. Signals generated at 680, 692, 802, 924, and 942 nm were utilized in the signal unmixing algorithm. Images from the multiwavelength unmixing scans have been displayed using the render mode with the PA signal overlaid on US B-mode (greyscale) signal of a cross-section of tubing. All image processing was carried out using the VevoLAB software. A ROI was drawn around the US image of the tube and, applying ROIs of the same size and shape to each image, a spectrum of the PA average signal was generated from the overlaid PA data of each image.

PA signal analysis of 2

Samples of 2 at 20 µM in Diluent X and 2 at 1 mM in Diluent X diluted to 64 µM in whole mouse blood (collected from cardiac puncture using EDTA coated syringes and immediately supplemented with 10% EDTA in 0.9% w/v saline [0.15 g/mL]) were prepared alongside appropriate blank controls. The samples were injected into Vevo contrast agent phantom tubing and Spectro scans were taken of the material in triplicate. The data was analyzed as described in ‘Photoacoustic Phantom Imaging’.

Preparation of bacterial overnight (O/N) culture

Staphylococcus aureus cells (ATCC; 25,923) were incubated in tryptic soy broth (TSB) liquid media at 37 °C (450 rpm) for 18 h. The cells were centrifuged (2 min at 10 000 xg) and the pellet was resuspended in Hank’s Balanced Salt Solution (HBSS) to an OD600 = 19.9 for in vitro assay, or to an OD600 = 8.43 for in vivo studies.

In vitro bacteria binding of 2 using fluorescence and photoacoustic detection

2 (10 µM final concentration) was incubated with 50 µL of S. aureus suspension (1 × 109 CFUs in 50 µL HBSS) or 50 µL of HBSS (no-cell control), for 10 min in triplicate. The samples were washed 3 × and the pellet was resuspended in 250 µL of HBSS. 200 µL sample and 200 µL stock solution were transferred to a black flat clear bottom 96-well plate. A fluorescence scan (Ex = 736 nm; Em = 746–850 nm) was performed. The fluorescence intensities (at 829 nm) of blank samples were subtracted from samples of 2 incubated with or without bacteria and the change in intensity was evaluated. Approximately 20 µL of each sample was then injected into Vevo contrast agent phantom tubing. The ends of each tube were sealed with heat and Spectro scans were acquired of each sample in triplicate. The data was analyzed as described in the ‘Photoacoustic Phantom Imaging’ section.

Radiosynthesis of [99mTc]Tc-PSVue794 (6)

400 µL of Apo-PSS794 (1) (0.5 mg in 400 µL of methanol [0.87 mM]), was combined with 400 µL of [99mTc(CO)3(OH2)3]+ (4) (Prepared using a procedure reported by Causey, et al.16) (370 MBq, dissolved in 0.9% w/v saline, pH = 6–6.5), and mixed at room temperature for 1 h. Radio-TLC (1% HCl in methanol eluant) and HPLC (method A, UV = 700 nm tr = 8 min; \(\gamma\) trace tr = 8.8 min) were used to determine the radiochemical conversion (~ 85–90%). The material was then diluted in 10 mL of water and loaded onto a C18 cartridge for purification. Water was used to elute any remaining 4 from the cartridge and ethanol was used to elute 5 in 99% purity based on Radio-TLC. The ethanol was evaporated, and the material was reconstituted in 171.5 µL mL of Diluent X. 171.5 µL of Zn(NO3)2 (0.015 M in Diluent X) was added to 5 and stirred at 37 °C for 30 min. The material was then adjusted to the correct concentration using 10% ethanol in 0.9% w/v saline for subsequent experiments. HPLC UV = 700 nm tr = 8 min; \(\gamma\) trace tr = 8.8 min.

Stability testing of 6

The stability of compound 6, incubated in formulation buffer (Diluent X) (room temperature) and in mouse serum (37 °C), was assessed at 0, 2 and 18 h. At each time point an aliquot of the material was analyzed by radio-TLC (eluting in 1% HCl in methanol). For serum samples, the aliquot was mixed with cold acetonitrile and pelleted by centrifugation to remove the serum proteins.

In Vitro Bacteria Binding of 6

In a clear, conical bottom 96-well plate, 200 µL of 6 (37 kBq) in 10% ethanol in 0.9% w/v saline was incubated with S. aureus suspension (1 × 109 CFUs in 50 µL of HBSS) or 50 µL of HBSS (no-cell control), for 10 min (n = 3 per condition). The sealed plate was centrifuged for 1 min at max speed. The supernatants were removed and 200 µL of HBSS was added to resuspend the pellet. This was repeated a total of 3 times to remove the unbound fraction of 6. The washed pellet was resuspended in 200 µL of HBSS, 150 µL was transferred into gamma counting tubes, and the amount of radioactivity in each tube was determined using a gamma counter. A 100 µL aliquot of 6 (from the 0.37 MBq stock) was measured and corrected to represent a 100% signal control for each concentration. The counts per minute (CPM) from each sample were corrected based on the dispensed volume measured and specific binding was calculated by subtracting the non-specific binding (CPM from the samples without bacteria cells) from the total binding (CPM from the samples with bacteria cells). The percent binding of each sample was calculated using Equation SI-2.

Animal studies

All procedures were conducted according to the guidelines of the Committee for Research and Ethics Issues of the International Association for the Study of Pain, and guidelines established by the Canadian Council on Animal Care and the McMaster University Animal Research Ethics Board. All experimental protocols were approved by the McMaster University Animal Research Ethics Board. All experimental protocols and results are reported in accordance with ARRIVE guidelines. In each animal study 3–4 female BALB/c mice (7–10 weeks old, Charles River Laboratories, Raleigh, NC) from the same litter, weighing around 20 g, were used. The mice were sterile housed and maintained at 21 °C with a 12-h light/dark cycle and were provided autoclaved food and water ad libitum.

Staphylococcus aureus-induced myositis model

An aliquot of S. aureus (1 × 108 CFUs in 50 µL of HBSS) was innoculated intramuscularly (I.M.) on the right hindlimb of mice (n = 4). 50 µL of the vehicle control, HBSS, was injected IM into the contralateral hindlimb of the same mice. The mice were monitored for 18 h post-inoculation of the bacteria, tissues were then collected and analyzed as described in the section titled ‘Histological Analysis’. Confounders were not controlled in the study design. No data points were excluded from the experimental analysis.

Biodistribution of 6

Four healthy control mice and 4 mice with established S. aureus-induced myositis were injected in the lateral tail vein with 200 µL (0.74 MBq) of 6 formulated in 10% ethanol in 0.9% w/v saline. 12 h post-injection, animals were anesthetized and euthanized by cardiac puncture followed by cervical dislocation. Fluids, bone (knee), and select tissues were collected, weighed, and quantified on a gamma counter. Decay correction was used to normalize organ activity measurements to the time of dose preparation for data calculations. Data are expressed as percent injected dose per gram of tissue/fluid (%ID/g). Confounders were not controlled in the study design. No data points were excluded from the experimental analysis.

Lipopolysaccharide-induced myositis model

Lipopolysaccharide (LPS) (1 mg/mL, 100 µL/mouse) was injected intramuscularly (IM) on the right hindlimb of mice (n = 4). 50 µL of the vehicle control, HBSS, was injected IM into the contralateral hindlimb of the same mice. The mice were monitored for 15 h post-injection of LPS, tissues were then collected and analyzed as described in the section titled ‘Histological Analysis’. Confounders were not controlled in the study design. No data points were excluded from the experimental analysis.

Histological analysis

Mice were euthanized with sodium pentobarbitol and perfused with lactate Ringer’s solution and tissues were fixed with 10% formalin. The right and left hind limb, and the spleen were collected and decalcified in a solution of 10% formalin + 8% EDTA. The trimmed tissues were embedded into paraffin, sliced, and stained with hematoxylin–eosin (H&E). A veterinary research pathologist, blinded to the treatments, examined the slides under a Nikon Eclipse 50i light microscope and morphological features of the tissues were described. One slice of each tissue (the treated hindlimb, the contralateral hindlimb, and the spleen) from each mouse in the LPS- and bacteria-induced myositis groups and 1 naïve mouse were analyzed and contributed to the main diagnoses of each condition. The images selected for the figure are representative for each condition and taken from one mouse per group.

The bone marrow was present within the slices of the hindlimbs due to the decalcification of bone. The microphotographs were only provided for the tissues where pathological morphology was observed. This is true for the spleen as well. While the spleen tissue was analyzed from each mouse within each group, the microphotographs were only provided from tissues that displayed morphological changes relative to healthy tissue.

In vivo photoacoustic imaging studies

PAI was performed using the Vevo-3100/LAZR-X imaging system. Mice were anesthetized, and the fur of their right hindlimb was thoroughly removed before being placed in a prone position on a platform that enabled the monitoring of their respiration and heart rate. A generous amount of US gel was applied before imaging with the transducer (MX400/LX400) positioned over the leg. At each imaging time point, Spectro scans and 3D multispectral unmixing scans were taken of each hindlimb. The transducer was positioned laterally overtop of the hindlimb, corresponding with the inoculation site. Scan distances averaged 20 mm and scans were taken every 0.33 mm (~ 7 min/3D scan).

3D multiwavelength unmixing scans were performed, where the pre-programmed PA spectra of oxygenated and deoxygenated hemoglobin are unmixed from the spectra of 2, across a defined z,x axis to construct images depicting dye-specific PA signal present within the hindlimb. Designated wavelengths (680, 692, 802, 924, and 942 nm), based on notable signals generated within individual chromophore spectra (oxygenated and deoxygenated hemoglobin, or 2), were used by the Vevo-3100/LAZR-X unmixing algorithm. The series of 2D cross-sectional scans collected were then reconstructed by the instrument into a 3D image, where the signals from each chromophore are assigned a colour and overlaid onto a b-mode ultrasound scan. The scans are displayed as either render mode or texture mapping images where the PA signal is overlaid on US B-mode (greyscale) signal for sections of the hindlimb. Signals from oxygenated and deoxygenated hemoglobin were subtracted from the images to reduce visual complexity. The spectrally unmixed signal of 2, in the 3D multiwavelength unmixing scans, was quantified using 3D ROI drawn around the inoculation sites. All signal unmixing algorithms and image processing were carried out using the VevoLAB software. Image processing consisted of signal optimization, with parameters set the same for each image within the study. 3D regions of interest were drawn around the inoculation site.

Spectral analysis was performed using ‘Spectro’ scans (as described in the “Photoacoustic Phantom Imaging’ section) and was used to verify whether the spectral signature of the in vivo signal matched that of the exogenous dye. The ROI used for each spectral curve was displayed in a multiwavelength unmixing scan (oxygenated and deoxygenated hemoglobin unmixed from 3) derived from the Spectro scan.

Photoacoustic imaging of LPS-induced sterile inflammation using 2

Three hours after treating mice with LPS (n = 3), as described previously, the mice were injected intravenously with 75 µL of 2 (1 mM) and were imaged, as described in the section titled ‘In vivo photoacoustic imaging studies’, on the Vevo LAZR X, pre-injection, and 12 h post-injection. Confounders were not controlled in the study design. No data points were excluded from the experimental analysis.

Photoacoustic imaging of S. aureus-induced bacterial infection using 2

Six hours after inoculating mice (n = 3), as described previously, the mice were injected intravenously with 75 µL of 2 (1 mM) and were imaged, as described in the section titled ‘In vivo Photoacoustic Imaging Studies’, on the Vevo LAZR X, pre-injection, and 12 h post-injection. Confounders were not controlled in the study design. No data points were excluded from the experimental analysis.

Statistical analyses

Data are reported as mean ± standard error of the mean (SEM). Concentration and bacteria-binding data were analyzed by one-way analysis of variance (ANOVA). Differences between test groups and control were assessed using Dunnett’s Multiple Comparison Test on GraphPad Prism version 9 (GraphPad Software, San Diego, CA). In vivo PAI data were analyzed using a two-way ANOVA. Differences between tested groups and control were assessed using the Tukey Test on GraphPad Prism (Threshold *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001).