In Vivo Fluorescence Imaging of Bacteriogenic Cyanide in the Lungs of Live Mice Infected with Cystic Fibrosis Pathogens

Background Pseudomonas aeruginosa (PA) and Burkholderia cepacia complex (Bcc), commonly found in the lungs of cystic fibrosis (CF) patients, often produce cyanide (CN), which inhibits cellular respiration. CN in sputa is a potential biomarker for lung infection by CF pathogens. However, its actual concentration in the infected lungs is unknown. Methods and Findings This work reports observation of CN in the lungs of mice infected with cyanogenic PA or Bcc strains using a CN fluorescent chemosensor (4′,5′-fluorescein dicarboxaldehyde) with a whole animal imaging system. When the CN chemosensor was injected into the lungs of mice intratracheally infected with either PA or B. cepacia strains embedded in agar beads, CN was detected in the millimolar range (1.8 to 4 mM) in the infected lungs. CN concentration in PA-infected lungs rapidly increased within 24 hours but gradually decreased over the following days, while CN concentration in B. cepacia-infected lungs slowly increased, reaching a maximum at 5 days. CN concentrations correlated with the bacterial loads in the lungs. In vivo efficacy of antimicrobial treatments was tested in live mice by monitoring bacteriogenic CN in the lungs. Conclusions The in vivo imaging method was also found suitable for minimally invasive testing the efficacy of antibiotic compounds as well as for aiding the understanding of bacterial cyanogenesis in CF lungs.


Introduction
Cystic fibrosis (CF) is the most common genetic disease in Caucasians [1]. Its main clinical features are excessive mucus production, airway obstruction, chronic airway infection and chronic inflammation, leading to severe bronchiectasis, irreversible lung damage and respiratory failure. Pseudomonas aeruginosa (PA) is the most common pathogen and is detected in 60-80% of patients' cultures according to age [2]. PA releases virulence factors that induce pulmonary deterioration in the lungs of CF patients [3], including exotoxin A, elastase, LasA protease and pyocyanin [4]. Burkholderia cepacia complex (Bcc), a group of at least nine species including B. cepacia, is more dangerous to CF patients than PA [5]. Bcc infection leads to the rapid deterioration of a patient's condition, termed 'cepacia syndrome', and approximately 20% of these patients die from high fevers, bacteremia and severe necrotizing pneumonia [6]. Bcc's innate multi-drug resistance to antibiotics is often associated with the high mortality rates [3]. Bcc produces a variety of virulence factors, such as proteases, lipases, hemolysins and exopolysacchride (EPS) [3].
Most PA and Bcc strains are capable of producing cyanide (CN) [7,8], which binds ferric iron and inhibits the function of cytochrome c oxidase in mitochondria. The lungs of CF patients are typically in a microaerobic condition, where cyanogenic pathogens survive through anaerobic respiration that can increase CN production [7,9]. Although studies have not yet determined whether CN alone aggravates the conditions of CF patients, growing evidence suggests that it may be a predictor of PA infection in the CF lung [10,11]. CN has been detected up to 130 mM in the sputa of CF and non-CF bronchiectasis patients infected with PA [10] and has not been detected in the sputa of patients without PA infection [10,11]. However, monitoring the amount of CN produced by microorganisms in the lung is not easy because sputum samples are not easily obtained from either pediatric patients or small animals, such as mice.
While conventional imaging methods, such as X-ray and MRI (magnetic resonance imaging) show the pathological state of the infected organs, in vivo molecular imaging [12] offers noninvasive insight into living organisms and provides spatial and temporal information of disease-related changes in the body. Optical molecular imaging [13,14] can detect light emitted from chemiluminescent or fluorescent probes and is relatively safe, cheap and easy to handle compared with MRI or nuclear imaging. Fluorescent chemosensors selectively bind to ions [15] or metals [16], changing their emission or excitation/emission profiles. This property has been exploited to study cell physiologies, such as heavy metal transitions [16]. CN specific chemosensors were developed [17][18][19][20], but their application was limited to detecting CN in environmental samples, such as water. Recently, we have successfully visualized CN in the nematode Caenorhabditis elegans using chemosensors [21,22].
In this study, CN production was monitored and quantified in the lungs of mice infected with PA or B. cepacia strains. For chronic infections, mice were intratracheally infected with these strains embedded in agar beads [23,24]. An ''off-on'' type CN chemosensor was directly injected into the lungs before the whole animal imaging. The effects of antimicrobial compounds on CN production were visualized as well.

Synthesis and fluorescent assays of the CN sensor
The sensor was synthesized by a previously reported method [25]. Fluorescein (4 g, 12 mmol), 10 mL CHCl 3 , 6 mL CH 3 OH and 0.06 g 15-crown-5 were placed in a 100 mL flask to which 20 g 50% NaOH solution was carefully added at 55uC. The mixture was stirred for 5 h. After cooling, it was acidified with 10 M H 2 SO 4 . The precipitates were collected and dried in vacuum. Finally, the white solid product (142 mg, yield 3.0%) was obtained by chromatography on a silica gel column using CH 3 COOCH 2 CH 3 /CH 2 Cl 2 (5:95, v/v) as eluent. Fluorescent assays used test solutions of the sensor (3 mM) and a various concentration of the sodium salts of different anions (CN 2 , H 2 PO 4 2 , HSO 4 2 , NO 3 2 , CH 3 CO 2 2 , F 2 , Cl 2 , Br 2 and I 2 ) in HEPES buffer (0.01 M, pH 7.4, 33% CH 3 CN). Fluorescence emission spectra were obtained using a RF-5301/PC spectrofluorophotometer (Shimadzu Co.).

Bacterial strains and CN determination in the bacterial culture
All PA strains were grown overnight in LB medium (BD) at 37uC and 220 rpm. The mutant required gentamicin (Sigma Chemical Co.). Supernatants were harvested at the optical density, OD 600 = 0.9 and CN was assayed using SpectroquantH CN detection kits (Merck).
B. cepacia (KCTC 2966) was obtained from the Biological Resource Center (Daejeon, Korea). It was grown in 470 bacteria culture medium (Sigma-Aldrich) at 30uC and 220 rpm. The supernatant was harvested at OD 600 = 2.6 and CN was assayed with the kit. To determine the amount of CN produced from B. cepacia biofilm, 12 mL of a 1:10 dilution of the culture (OD 600 = 0.5) was inoculated in a Petri dish containing 48 g of glass beads and CN was trapped in a separate dish containing 4 M NaOH as described previously [8]. CN trapped in NaOH was assayed using the CN detection kits (Merck).
Introduction of NaCN to the lungs 6-week-old female BALB/c nude mice (17 to 19 g) were purchased from Orientbio Inc. (Seongnam, Korea). They were treated in accordance with the university's guidelines. The mice were anesthetized by intraperitoneal (i.p.) injection with 10 mL (each 0.25 mg tiletamine, zolazepam) zoletilH50 (Virbac, Carros, France) using an insulin syringe (Shinchang medical co., Gumi, Korea). Then, 40 mL of various concentrations of NaCN (0.1 mM-1 M) were injected into their lungs using a syringe.
Infection with PA or B. cepacia 50 mL of bacterial suspension in agar beads containing approximately 4.5610 6 CFU was inoculated into the anesthetized mice via intratracheal instillation [23,24] using a 24-gauge catheter (BD).

In vivo imaging
A multimodal-imaging system IS4000MM (Kodak) was used at 465 nm and an emission at 535 nm. After 40 mL of the CN sensor (1 mM) was injected into the lungs of the anesthetized mice, the animals were exposed for 10 sec. Two rectangular regions of interest (ROI) with dimensions of 280 pixels (width)6222 pixels (height) were manually drawn on each animal image with a dimension of 800 pixels (width)61,400 pixels (height) using the program Image J (NIH, USA). One of the regions was the infectious or affected area, covering the lobes, for the signal intensity; the other ROI, outside the infectious area, was for background intensity. The value of the background intensity was subtracted from the signal intensity in order to obtain the normalized signal intensity from each image.
Ex vivo imaging of CN and biofilm in the lungs 20 mm thick lung samples on glass slides [26] were incubated with the sensor (2 mM) and TRITC (tetramethyl-rhodamineisothiocyanate)-labeled concanavalin A (50 mg/mL) (Sigma-Aldrich) for 5 min. The images were obtained using a confocal laser scanning microscope (CLSM) (LSM510, Carl Zeiss) with 488 nm excitation and 505-530 nm emission filters for the sensor and 543 nm excitation and 560-615 nm emission filters for concanavalin A-TRITC at 4006magnification.

Bacterial load determinations
The lungs that were obtained from the sacrificed mice were homogenized in 1 ml of sterile PBS, plated in duplicates on BBL TM eosin methylene blue agar (BD) and incubated at 37uC for 18 h prior to CFU determinations.
Cell death in the infected lung tissue 20 mm thick fresh frozen sections from the mice that were previously either injected NaCN or infected with PA14 were incubated on a 45uC hot plate for 2 h and then washed sequentially with 100%, 95%, 70% ethanol and PBS. Samples were fixed with 3.7% buffered formaldehyde for 10 min. and washed with PBS. A TACSH 2 TdT-Fluor in situ apoptosis detection kit (Trevigen, Gaithersburg, MD, USA) was used to examine whether bacteriogenic CN induced apoptosis in the infected lungs. As a positive control, TACS-nuclease was added to tissue sections, which induced apoptosis. To examine necrotic cells, tissue sections were counterstained with 1.5 mM propidium iodide (Invitrogen).

Statistics
The data are reported as dot blots including every data points of n independent mice and independent experiments. Significant difference was calculated with one-way analysis of variance (ANOVA) was with Bonferroni post-tests and defined as p,0.05 using an unpaired two-tailed t test [29] (Graphpad, La Jolla, CA, USA).

Sensitivity and selectivity of CN sensor in solution
The title sensor 49,59-fluorescein dicarboxaldehyde (Fig. 1a), reported as an intermediate in previous work [30], was synthesized and used as a CN sensor for the first time in this study to fluorescently observe CN in lungs. The sensing mechanism involved CN acting as a selective nucleophilic agent for ohydroxy-substituted aromatic aldehyde [31] and could thus attack the carbonyl group, which was activated by phenol protons (Fig. 1a). Protons were then transferred from phenol hydrogen to the developing alkoxide anion, resulting in the fluorescence enhancement (Fig. 1a). In HEPES buffer (0.01 M, pH 7.4, 33% CH 3 CN), the enhancement of fluorescence depended on CN concentration (Fig. 1b). In 100% water, there was no significant change in fluorescence intensities when pH was adjusted from 6.0 to 8.2 (Fig. 1c). The CN sensor selectively reacted with CN 2 ions among the various other anions in HEPES buffer, such as F 2 , Cl 2 , Br 2 , I 2 , H 2 PO 4 2 , HSO 4 2 , ClO 4 2 , NO 3 2 and CH 3 COO 2 (Fig. 1d). The sensor did not react with PA14 supernatant or pyocyanin, a toxin produced by PA (Fig. 1e). Its reactivity with CN interfered with B12a [28], known to bind strongly with CN (Fig. 1f). Taken together, these results indicate that the sensor specifically binds to CN, enhancing its fluorescence in aqueous solution.

In vivo fluorescence imaging of exogenous CN in mice
The sensor was prepared in dimethyl sulfoxide (DMSO), known to have low toxicity [32]. Before applying the CN sensor into mouse, its toxicity was tested by daily injecting 40 mL of the sensor (1 mM) into the lungs of three mice for 10 days. The direct injection of the sensor into the lobes did not cause vomiting or death. During the period, the injected mice did not show any differences in behavior relative to controls that had not been injected with the sensor (data not shown).
In the following experiments, 40 mL of various concentrations of NaCN (0.1 mM-1 M) and 40 mL of the CN sensor (1 mM) were injected into anesthetized mice (Fig. 2a). Whole animal imaging system was used to determine whether the sensor was capable of visualizing the exogenous CN in the lungs. Strong fluorescence was observed in lungs injected with NaCN (0.1 mM or higher), whereas the fluorescence was not observed in the lungs without NaCN (Fig. 2c and Figure S1), indicating that the sensor selectively reacted with CN in the lungs. Region of interest (ROI) analyses showed that the sensor's fluorescence intensity the lungs exhibited a linear response with respect to the concentration of injected NaCN (Fig. 2b). The correlation coefficient of the linear regression was 0.992 for 4 to 7 independent measurements, and the detection limit was 0.1 mM NaCN (p = 0.0026). The reaction of the sensor with CN in mice was inhibited by B12a ( Figure S2).
Short-term in vivo imaging of biogenic CN in the lungs of mice infected with PA or B. cepacia To chronically infect the mouse lungs, mice were intratracheally infected with PA or B. cepacia strains embedded in agar beads [23,24] (Fig. 3a). As shown in Fig. 3b  40 mL of the CN sensor (1 mM) was directly injected into the lungs at 18 h. after the infection. The dose-response curve (Fig. 2b) of the fluorescence signals in the lungs was used to estimate CN concentration in the lungs. Surprisingly, both PA strains produced millimolar concentrations (1.8 to 2.9 mM) of CN in the lung (Fig. 3c). Interestingly, the wild type PA14 produced stronger signals than wild type PAO1. These in vivo data were verified by ex vivo imaging results (Fig. 3d), which exhibited a similar tendency in fluorescence intensity. The reaction of the sensor with CN in PA14-infected lungs was inhibited by B12a ( Figure S2). Although intranasal application of the sensor appeared to be least invasive, it was not used in this study as it caused vomiting, and thus required a high volume (ca. 200 mL) of the sensor in the infected mice ( Figure S3). In contrast, direct injection of the sensor in the lung did not cause vomiting and 40 mL of the sensor (1 mM) generated the highest signal to noise ratio to bacteriogenic CN ( Figure S4).

Long-term in vivo imaging of biogenic CN in lungs of mice infected with PA or B. cepacia
In vivo imaging was used to monitor CN production in the lungs of live mice infected with either wild type PA14 or B. cepacia strains for up to 9 days. In PA14-infected lungs, CN concentration rapidly increased within 24 hours but gradually decreased over the following days (Fig. 4a,c). In contrast, in B. cepacia-infected lungs, CN concentration slowly increased, reaching a maximum at 5 days, after which it remained constant (Fig. 4b,d). Irrespective of this difference in CN production pattern between PA14 and B. cepacia strains, both strains were able to continuously produce millimolar concentrations of CN in the lung even 7 days after infection (Fig. 4c,d), indicating that the lungs were chronically infected with the strains.
Tunnel assay showed that the apoptotic signal was stronger than the necrotic signal when lungs were injected with 1 mM NaCN, while the reverse occurred with injection of 10 mM NaCN (Fig. 4e), indicating that the mode of cell death depended on CN concentration. Lung sections of mice infected with PA14 for 18 h. also revealed stronger necrotic signals (Fig. 4e), implying that the CN concentration in the PA14-infected lungs was higher than 1 mM. CN concentration-dependent death mode has been reported in cultured primary rat cortical cells [36]. The ex vivo imaging results show that the bacteria in the lungs produced a large amount of CN as well as EPS ( Figure S5), a major constituent of bacterial biofilm.
In vivo efficacy of antimicrobial compounds against PA and B. cepacia infections When mice infected with wild type PA14 were treated with either ceftazidime (200 mg/kg) or ciprofloxacin (30 mg/kg) [26] at 18 h. after infection, both CN production (Fig. 5a,b) and bacterial load (Fig. 5c) in the lungs significantly decreased (p,0.0001), indicating that the antibiotic treatments were effective against the lung infection. Both CN production and bacterial loads significantly decreased (p,0.0001) when mice infected with PA14 for 18 h were treated with patulin (50 mg) [27], a fungal toxin that inhibits bacterial communication, for 3 days intraperitoneally, indicating that the toxin was capable of inhibiting the infection in vivo.
In contrast, the individual antibiotic treatments were not effective against B. cepacia infection. Neither CN production nor   . 50 mg patulin was treated intraperitoneally (i.p.) daily for 3 days. The sensor was then directly injected into the lungs at 6 h. after the antibiotic treatment. (b) Quantification of the fluorescence intensity in the PA14-infected lungs treated with antibiotics. Each dot represents a CN concentration from a mouse in each treatment group, and every three mice were given each treatment (total n = 15). ANOVA with Bonferroni post-tests (p,0.0001). The efficacy of each antibiotics against PA14 infection was determined using Student's unpaired t-test: ceftazidime, p,0.0001: ciprofloxacin, p = 0.0029: patulin for 3 days, p,0.0001. (c) Reduced bacterial loads in the PA14-infected lungs after each antimicrobial treatment. The PA14-infected lungs were excised from the mice at 6 h. after the antibiotic treatment and CFUs in the excised lungs were determined. Each dot represents the mean CFU of four different lung samples from independent mice and three independent experiments were performed per treatment. ANOVA with Bonferroni post-tests (p,0.0001). (d) In vivo images of CN in B. cepacia-infected lungs treated with antibiotics. 200 mg/kg ceftazidime was injected into the tail vein, 30 mg/kg ciprofloxacin was orally administered, and 50 mg patulin was bacterial loads significantly decreased in the infected lungs ( Fig. 5d-f), indicating that the B. cepacia strain was resistant to all the antibiotic treatments.

Discussion
The CN sensor was highly specific for CN both in vitro and in vivo. Only CN induced fluorescence enhancement of fluorescein dicarboxyaldehyde among various anions in aqueous solution (Fig. 1d). This selectivity was attributable to the specific reaction between the aldehyde group of the sensor and CN (Fig. 1a). The aldehyde group allows intramolecular hydrogen bonding, thereby facilitating nucleophilic attack from CN. The specificity of the sensor to CN in aqueous solution was further verified by results showing that its fluorescence response to CN was not affected by either PA14 supernatant or pyocyanin (Fig. 1e), while it was reduced by a CN antidote, B12a [28] (Fig. 1f). Similarly, hcnC mutant strain did not induce any fluorescent signal in the infected lung (Fig. 3b,c) and B12a was able to interfere with the reaction of the sensor with CN in the lung ( Figure S3), demonstrating the in vivo specificity of the sensor to CN. It is expected that physiological pH (6.9) in the lung [37] would not affect the fluorescence response of the sensor, as there was no significant difference in the fluorescence intensities of the sensor over the pH range 6.0-8.2 (Fig. 1c). Using the sensor and whole animal imaging, concentrations as low as 0.1 mM NaCN were observed in the lung (Fig. 2).
The presence of CN at high concentrations (1.8 to 4 mM) allowed visualization of bacteriogenic CN in the lungs of live mice infected with PA or B. cepacia strains. These strains in the lung produced more CN than they did in the planktonic culture (Table S1). CN production by PA or B. cepacia strains might be enhanced by a microaerophilic condition in the infected lungs, where HCN synthases are highly activated [7,9]. The microaerophilic condition in the lung might have been contributed by biofilm formation ( Figure S5). Especially, B. cepacia is known to produce CN only when it establishes mature biofilm [8]. This may explain why it did not produce detectable amounts of CN in the lung until 2-3 days after infection (Fig. 4b). Once it started to produce CN, its CN production was continued even at 9 day after infection (Fig. 4b,d). These results suggest that CF lung, which is microaerophilic, is suited for the cyanogenic bacteria to produce CN.
Cell death by necrosis (Fig. 4e) indicates the presence of millimolar CN in the infected lungs. It was shown that CN at 1 mM or higher concentrations caused necrosis in epithelial lung cells [38]. Other major virulence factors including pyocyanin in PA are known to apoptosis in the lung [39,40]. The CN concentration (2.9 mM) in the murine lungs infected with PA14 can be converted to be 1.7 mg/kg, a fifth of the lethal dose (8.4 mg/kg) in mice [41]. At this concentration, CN was capable of inhibiting cytochrome oxidase activity by up to 15%, causing unconsciousness in the mice for 10 to 60 min [42]. In fact, many of the infected mice vomited and did not eat for two days, some died (data not shown). Necrosis in the lung has been reported to enhance inflammation at the site of infection [43]. These data suggest that, along with other virulence factors, biogenic CN in the lungs may have detrimental effects in CF patients.
CN concentration in the lungs was clearly related to the respective bacterial loads (Fig. 5), suggesting that bacteriogenic CN acted as a biomarker of the cyanogenic bacteria in the lungs. Information about the effective route of antibiotic administration, as well as the in vivo efficacy of the antibiotics, was obtained from this correlation (Fig. 5). PA14 was susceptible to b-lactam (e.g., ceftazidime) or fluoroquinolone (e.g., ciprofloxacin) antibiotics as well as patulin (Fig. 5a-c). Patulin inhibits quorum sensing [44], which allows bacteria to produce responses such as virulence factors or biofilm at a critical cell density. PA has been reported to be effectively cleared from mice by treatment with patulin [27]. In contrast, B. cepacia was resistant to all the antimicrobial treatments ( Fig. 5d-f). This is possible because it has b-lactamase as well as antibiotic efflux system [45]. These results suggest that the current CN imaging method is suitable for testing the in vivo efficacy of antibacterial compounds and CN inhibitors. The method has several advantages in testing antibiotic susceptibility of PA in vivo over the in vivo imaging method using luciferase-tagged isolates [46,47]. Unlike the latter, the former does not require a construction of a strain carrying the luciferase gene and its fluorescence signal is not affected by exhaustion of flavin mononucleotide, which is a substrate of luciferase [46,47].
Collectively, these results suggest that the chemosensor-based in vivo imaging method is useful for fast and convenient animal testing of drug efficacy against the PA and B. cepacia infections as well as understanding cyanogenesis in CF lungs. In future, a CFTRdeficient murine model [48] could be used to elucidate the pathological role of biogenic CN in CF patients. CLSM images of the 20 mm thick cryo-sections incubated with 2 mM CN sensor for 5 min and 50 mg/mL concanavalin A-TRITC for 5 min. The surface plot image of merged images was obtained using Imaris TM program. Scale bar 220 mm.

(DOCX)
Table S1 CN production in liquid and solid cultures and murine lungs by PA and B. cepacia strains. CN in the liquid cultures was determined using the SpectroquantH method* in triplicate (at OD 600 = 0.9 for PA, 2.6 for B. cepacia). CN in the solid cultures containing glass beads were trapped in 4 M NaOH and measured using the method described above. Means 6 standard deviations of three independent experiments are shown (n. d. = not detected). Biogenic CN in the lungs previously infected with PA and B. cepacia was determined from the images using a multimodalimaging system (IS4000MM, Kodak). The standard curve (Fig. 2b) was used to determine the concentration of biogenic CN in the lungs infected with either PA or B. cepacia. *Detailed description in the methods. (DOCX)