A Reactivity-Based 18F-Labeled Probe for PET Imaging of Oxidative Stress in Chemotherapy-Induced Cardiotoxicity

Oxidative stress underlies the pathology of many human diseases, including the doxorubicin-induced off-target cardiotoxicity in cancer chemotherapies. Since current diagnostic procedures are only capable of monitoring cardiac function, a noninvasive means of detecting biochemical changes in redox status prior to irreversible functional changes is highly desirable for both early diagnosis and prognosis. We designed a novel 18F-labeled molecular probe, 18F-FPBT, for the direct detection of superoxide in vivo using positron emission tomography (PET). 18F-FPBT was radiosynthesized in one step by nucleophilic radiofluorination. In vitro, 18F-FPBT showed rapid and selective oxidation by superoxide (around 60% in 5 min) compared to other physiological ROS. In healthy mice and rats, 18F-FBPT is distributed to all major organs in the first few minutes post injection and is rapidly cleared via both renal and hepatobiliary routes with minimal background retention in the heart. In a rat model of doxorubicin-induced cardiotoxicity, 18F-FBPT showed significantly higher (P < 0.05) uptake in the hearts of treated animals compared to healthy controls. These results warrant further optimization of 18F-FBPT for clinical translation.


■ INTRODUCTION
Reactive oxygen species (ROS) are generated as normal byproducts of metabolism in the electron transport chain and play an integral role in the regulation of cell growth, neurotransmission, and the immune response. 1 However, uncontrolled ROS production leads to the oxidation of DNA, proteins, and lipids, underlying the pathogenesis of many cardiovascular and neurodegenerative diseases, as well as cancers and inflammatory conditions. 2−7 In the cardiovascular system, elevated ROS are responsible for tissue injury during ischemia/reperfusion and have been linked to the progression from cardiac hypertrophy to heart failure, the evolution of atherosclerotic plaques, and the cardiac and microvascular dysfunction associated with diabetes. ROS production has also been linked to the cardiotoxicity of cancer chemotherapeutic agents, which severely limits their dosimetry and effectiveness. 8−11 Most notably, the cardiotoxicity induced by doxorubicin, a widely used cancer chemotherapy agent, has been linked to ROS generation. 12,13 A clinically translatable means of noninvasively identifying and quantifying elevated ROS production in vivo would be highly desirable for both diagnostic and prognostic purposes and the development and evaluation of emerging cardioprotective approaches. 14 Several radiolabeled small molecules have been reported to indirectly detect oxidative stress by positron emission tomography. 15,16 Radiotracers based on the fluorescent probe dihydroethidium have been tested in rodent models of superoxide-associated cardiotoxicity and neuroinflammation. 17,18 While some of these designs are promising, none have yet progressed to clinical evaluation, and the search for optimal ROS-sensing PET radiotracers continues.
Here, we describe a novel chemical scaffold with potential utility as 18 F-labeled PET tracers for the direct detection of ROS in vivo. Phenyl benzothiazoles were identified as promising candidates due to their low molecular weight, well-established chemistry, susceptibility to oxidation by superoxide, and physicochemical properties that are favorable to the cell membrane and blood−brain barrier permeability. 19,20 Our tracer, [ 18 F]3-(3-fluoropropyl)-2-phenyl-2,3dihydrobenzo[d]thiazole, 18 F-FPBT was designed to be sufficiently lipophilic in its reduced form to facilitate cell membrane and blood−brain barrier penetration, while sufficiently hydrophilic when oxidized to become transiently retained in cells under oxidative stress to provide ROS-sensitive PET contrast ( Figure 1). Herein, we report the synthesis, 18 F-labeling, in vitro, and preclinical in vivo evaluation of 18 F-FPBT in healthy mice and rats, as well as in a rat model of doxorubicin-induced cardiotoxicity.

■ MATERIALS AND METHODS
General Information. HPLC analysis was performed with an Agilent 1200 HPLC system equipped with a 1200 series diode array detector. Radio-HPLC analysis was performed with an Agilent 1200 HPLC system equipped with a series diode array detector and a Raytest GABI Star radioactivity detector. 18 F-Fluoride was purchased from either the PET Center at St. Thomas' Hospital or Alliance Medical UK. All reagents were purchased from Sigma-Aldrich and were used without further purification. The radiochemical yield was calculated as a percentage of purified tracer to starting activity. The radiochemical conversion was calculated based on the HPLC peak AUCs. Preclinical PET/CT images were acquired using a NanoScan PET/CT (Mediso, Budapest, Hungary) scanner. Data are represented as mean ± SD, unless stated otherwise. Chromatographs and graphs were plotted on GraphPad Prism 8.
Chemical Synthesis. The synthesis and characterization of the nonradioactive reference compound 1, intermediates, 2, 3, 4, and radiolabeling precursor 5 is described in the Supporting Information.
Radiosynthesis. Radiosynthesis of 18  F] 1) was purified with a ZORBAX column (300SB-C18, semipreparative 9.4 mm × 250 mm, 5 μm) using acetonitrile and water as the mobile phase, at a flow rate of 3.0 mL/min. The following gradient was used: from 50% to 90% acetonitrile in 15 min; kept at 90% acetonitrile for 10 min; from 90 to 50% acetonitrile in 5 min. [ 18 F] 1 has an HPLC retention time of 12.5 min. It was collected from the HPLC column and diluted to 10% acetonitrile in water. It was trapped onto a Sep-Pak C-18 light cartridge (preactivated with 5 mL of methanol followed by 5 mL of water). The cartridge was washed with 2 mL of water, and [ 18 F] 1 was then eluted with 1 mL of ethanol to give [ 18 F] 1 in 60 ± 20% radiochemical conversion, decay corrected isolated RCYs around 33 ± 8%, and >98% radiochemical purity (n = 10).
Radiosynthesis of 18 Lipophilicity. The lipophilicity of 18 F-FPBT and 18 F-FPBT-Ox was determined by a conventional partition method between 1-octanol and phosphate buffered saline (PBS), pH 7.4. 1-Octanol was presaturated with PBS before use. The radiotracer (∼2 KBq) was added to a mixture of PBS (200 μL) and 1-octanol (200 μL) in a 1.5 mL Eppendorf vial (n = 6). The mixture was vigorously agitated at rt for 5 min and then centrifuged at 3000 g for 10 min. A 100 μL aliquot from each layer was drawn for measurement in a gamma counter. The LogD oct/PBS was calculated as follows: log [(cpm in the 1octanol layer − cpm 1-octanol blank)/(cpm in the PBS layer − cpm PBS blank)].
Animal Studies. All experiments were performed in accordance with the Animals (Scientific Procedures) Act 1986 under project license nos. PPL P96678ED7 and PPL 70/ 8482. Male Wistar rats were obtained from Envigo Ltd. and C57BL/6 mice from Charles Rivers. All animals went through a 7 day acclimatization period at the Biological Services Unit at St. Thomas' Hospital prior to any experiments. 18 F-FPBT was formulated in ∼5% ethanol in PBS, containing 0.01 mg/mL sodium ascorbate for injection. 18 F-FPBT-Ox was formulated in ∼5% ethanol in PBS. All PET imaging studies were performed using a Nano-Scan PET/CT (Mediso, Budapest, Hungary) scanner. All PET/CT data were reconstructed with the Monte Carlo-based full-3D iterative algorithm Tera-Tomo (Mediso Medical Imaging Systems, Budapest, Hungary). All procedures were performed under 2% isoflurane in oxygen anesthesia.
PET/CT Imaging of Healthy Mice. C57BL/6 mice (male, 25 ± 2 g, n = 3) were placed on the PET scanner bed and injected with 18 F-FPBT (1.7 ± 0.4 MBq) through the tail vein at the start time of PET acquisition. A dynamic PET scan was performed for 60 min, followed by a CT scan and ex vivo biodistribution. The PET data was reconstructed into 5 frames of 60 s, 3 frames of 300 s, and 4 frames of 600 s. Regions of interest (ROIs) were drawn and quantified using VivoQuant software (v3.0, inviCRO, LLC, Boston, USA). Organ uptake was calculated as standard uptake values (SUV), and data are reported as mean ± SD.
Post-Mortem Biodistributions in Healthy Rats. Wistar rats (male, 319 ± 20 g) were injected intravenously via the tail vein with either 18 F-FPBT (1.0 ± 0.4 MBq) or 18 F-FPBT-Ox (1.1 ± 0.5 MBq). The animals were sacrificed by cervical dislocation at 1, 5, or 30 min postinjection (n = 3/group). Organs and tissues of interest were harvested and weighed, and the radioactivity was measured in a gamma counter. Organ uptake was calculated as the percentage injected dose per gram of tissue mass (% ID/g). Data are reported as mean ± SD.
Echocardiography. In order to assess cardiac function, all rats were subjected to a cardiac ultrasound (Vevo 770TM, VisualSonics) 1 day prior to and on day 6 of osmotic pump implantation. Rats were anesthetized with 2% isoflurane in 100% oxygen and maintained at 37°C via a homeothermic platform and rectal thermometer. High-resolution parasternal left ventricle (LV) long axis M-mode and B-mode images were obtained using an RMV710B transducer. Images were analyzed offline using Vevo Software to determine LV function. Data are reported as mean ± SD. Statistical analysis was performed using one-way ANOVA with Tukey's multiple comparisons test.
PET/CT Imaging in a Rat Model of Doxorubicin-Induced Cardiotoxicity. Wistar rats that had received either doxorubicin treatment (male, 267 ± 19 g, n = 6) or saline (male, 303 ± 9 g, n = 4) via osmotic pumps for 7 days were injected with 18 F-FPBT (2.6 ± 1.5 MBq) via the tail vein on the PET/CT scanner bed. The time of injection coincided with the beginning of PET acquisition in order to obtain dynamic tracer uptake information. PET scans with the chest area in the field of vision were acquired for 30 min, followed by a CT scan, after which the animals were sacrificed and organs harvested, weighed, and gamma-counted as for the biodistribution protocol. The hearts were frozen in liquid nitrogen for gamma counting and oxidative stress biomarker assays. The PET data was reconstructed into three frames: 0−3 min, 3−10 min, and 10−30 min. For the time−activity curve, the PET data was reconstructed into 1 min per frame in the first 5 min and then 5 min per frame between 5 and 30 min. Using VivoQuant software (v3.0, inviCRO, LLC, Boston, USA), regions of interest (ROIs) were drawn in the left ventricle and blood inside the myocardium. Radioactivity uptake in the heart was calculated as a standard uptake value ratio (SUVR) between left ventricle and blood. Data are reported as mean ± SD. Statistical analysis was performed using one-way ANOVA with Tukey's multiple comparisons test.
Ex Vivo Heart Biomarkers of Oxidative Stress. Total glutathione concentration was determined by an enzyme recycling method. 21 Lipid peroxidation was evaluated using the lipid peroxidation (MDA) assay kit from Sigma-Aldrich, which measures the colorimetric product that results from the reaction of malondialdehyde (MDA) present in the tissue with thiobarbituric (TBA) acid. Iron content was measured in isolated mitochondria and cytosolic fractions using the iron colorimetric assay kit from BioVision following the manufacture's protocol.

Molecular Pharmaceutics
pubs.acs.org/molecularpharmaceutics Article ■ RESULTS Synthesis. The nonradioactive reference compound of 18 F-FPBT and its iodinated precursor for nucleophilic radiofluorination were prepared in three steps from commercially available starting materials ( Figure 2). 2-Phenylbenzo[d]thiazole 2 was synthesized by reacting 2-aminobenzenethiol with benzoic acid in the presence of polyphosphoric acid. 22 N-Alkylation of 2-phenylbenzo[d]thiazole 2 took place with 3iodopropyl trifluoromethanesulfonate or 3-fluoropropyl trifluoromethanesulfonate to yield the triflate salts 3 and 4, which were subsequently reduced with sodium borohydride to obtain the nonradioactive reference compound 1 and precursor 5 for 18 F-labeling. Radiosynthesis. 18 F-FPBT was radiosynthesized by a onestep 18 F-labeling of alkyl iodide precursor 5 with high radiochemical conversion (60 ± 20%, n = 10) as indicated by radioHPLC chromatogram of the crude reaction mixture. ( Figure 3A) The decay corrected isolated radiochemical yields (RCYs) were consistently around 33 ± 8% (n = 10) with >98% radiochemical purity. The manual radiosynthesis, purification, and formulation of 18 F-FPBT took around 2.5 h from the end of bombardment until formulation. The oxidized analogue, 18 F-FPBT-Ox, was obtained by reacting 18 F-FPBT with an excess of potassium superoxide in PBS in 15 min and isolated with >90% radiochemical purity. The identity of both 18 F-labeled compounds was confirmed by HPLC coelution with the corresponding nonradioactive reference compounds ( Figures 3B and Supplementary Figure 1). The molar activity of 18 F-FPBT was measured as 168 ± 39 GBq/μmol (n = 3) when starting with around 1.2 GBq of 18 F-fluoride.
In Vitro Stability of 18 F-FPBT and 18 F-FPBT-Ox. 18 F-FPBT is stable in both pure ethanol and 5% ethanol in PBS in the presence of ascorbic acid (0.01 mg/mL) for 4 h (not tested for longer). When incubated in rat serum at 37°C, about 90% of 18 F-FPBT was intact in 1 h ( Figure 3C). The 18 F-FPBT-Ox also exhibits excellent serum stability ( Figure 3D).
In Vitro Reactivity of 18 F-FPBT toward Various ROS. 18 F-FPBT showed rapid (in 5 min) and selective oxidation 58 ± 8% (n = 4) by superoxide, which can be partially inhibited to 31 ± 5% (n = 3) in the presence of ascorbate (1 mg/mL) ( Figure 4). In contrast, 18 F-FPBT had little or no reactivity to other biologically relevant ROS. 18 F-FPBT In Vivo PET/CT Imaging of Healthy Mice. 18 F-FPBT had fast blood clearance. It was rapidly taken up and cleared by the heart, lungs, spleen, and brain with low nonspecific background retention in these organs. The radiotracer is excreted by both renal and hepatobiliary routes in less than 1 h (Supplementary Figure 3).
Post-Mortem Biodistributions in Healthy Rats. Ex vivo biodistribution studies in healthy Wistar rats also showed rapid uptake of 18 F-FPBT by major organs, including the brain, heart, lung, and spleen, and fast clearance from the blood pool at 1 min postinjection ( Figure 5A). 18 F-FPBT was excreted by both renal and hepatobiliary routes. Conversely, 18 F-FPBT-Ox shows significantly reduced uptake in major organs and was quickly cleared by the kidneys and excreted through the urine ( Figure 5B).
Rat Model of Doxorubicin-Induced Cardiotoxicity. Either doxorubicin or saline was delivered to two groups of rats via osmotic pumps to develop the animal model of doxorubicin-induced cardiotoxicity and the control group. From the day of minipump implantation to the day of imaging (7 days), animals in the saline control group maintained their weight with small variations [1.3 ± 2.5% (n = 4) increase], while doxorubicin-treated animals consistently lost weight [16.2 ± 8.9% (n = 6) decrease]. Rats treated with doxorubicin showed a statistically significant reduced (15%) left ventricular ejection fraction at day 6 compared to their baseline values at day 0. However, echocardiography cannot detect the statistical difference of the left ventricular ejection fraction between doxorubicin-treated animals and their time-matched vehicle controls at day 6 ( Figure 6B).
PET/CT Imaging in Rat Model of Doxorubicin-Induced Cardiotoxicity. A dynamic PET/CT scan (30 min) was performed on the rats that received either doxorubicin or saline. The PET data of each animal was reconstructed into three frames: 0−3, 3−10, and 10−30 min. The standard uptake value ratio (SUVR) between the left ventricle and blood was calculated to determine the radioactivity retention in the heart of each animal. The doxorubicin-

Molecular Pharmaceutics
pubs.acs.org/molecularpharmaceutics Article treated rats showed higher myocardial radioactivity retention than the vehicle control group in the PET imaging ( Figure  6C). The SUVR LV/blood of the doxorubicin-treated rats was statistically significantly (p < 0.05) higher than the vehicle control group ( Figure 6D). In the time−activity curve extracted from PET data (Supplementary Figure 4a,b), the median myocardial SUV values of the doxorubicin-treated animals are higher than their median blood SUV values in the 30 min PET scan, while the median myocardial and blood SUV values of the saline-treated control group are very similar at all of the time points. In addition, the median SUVR LV/blood of doxorubicin-treated animals is also higher than the saline control group in the 30 min PET scan (Supplementary Figure  4c).
In Vivo Blood Stability of 18 F-FBPT-Ox. Subsequently, we examined the in vivo blood stability of 18 F-FPBT-Ox at 30 min post IV injection. Plasma proteins were precipitated, and the supernatant was analyzed by radio-HPLC. The HPLC eluent was collected every minute in 30 vials separately and submitted to gamma counting. The counts per minute in each vial were plotted against the corresponding time point. Increased gamma counts (Supplementary Figure 5) were only observed very close to the HPLC retention time of the 18 F-FPBT-Ox ( Figure 3D).
Ex Vivo Heart Biomarkers of Oxidative Stress. Cardiac oxidative stress biomarkers including glutathione, maldonaldehyde, mitochondrial iron, and cytosolic iron were measured following the established literature methods. The levels of these biomarkers were marginally higher in the doxorubicintreated group comparing to the saline control group (Supplementary Figure 6).

■ DISCUSSION
The radiosynthesis of 18 F-FPBT was achieved in a late-stage one-step radiolabeling with sufficiently high yield to make it suitable for translation into a fully automated radiosynthesizer such as GE FASTlab. We systematically investigated the reactivity and chemoselectivity of 18 F-FPBT with various biorelevant ROS in vitro and observed rapid and selective oxidation by superoxide, which can be significantly inhibited by a large excess of ascorbic acid. We suspect the redox potential, half-life, and the steric hindrance of the ROS played a combination role in the chemoselectivity of 18 F-FPBT toward various ROS oxidation. The redox potential of hydroxyl radical (2.33 E°/V) is significantly higher than superoxide (0.94 E°/ V). 23 However, the hydroxyl radical has a much shorter half-  life (10 −10 s) than that of the superoxide (10 −6 s). 24 Therefore, 18 F-FPBT appears to be inert to hydroxyl radical but reactive toward superoxide. Such chemoselectivity of 2-benzothiazolines toward superoxide rather than hydroxyl radical has also been observed by others. For example, Zhang has reported that the 2-(2-pyridil)-benzothiazoline was oxidized by superoxide to form 2-(2-pyridil)-benzothiazole but not by the hydroxyl radical. 20 This might also be the reason why 18 F-FPBT has shown some reactivity toward other longer half-life ROS such as nitric oxide (a few seconds), peroxynitrite (10 −3 s), and hydrogen peroxide (stable) despite the fact that they have a much lower redox potential than the hydroxyl radical. 24 On the other hand, an alkylperoxyl radical (∼1.0 E°/V) such as a tertbutoxy radical has a similar redox potential as the superoxide (0.94 E°/V) but has a significantly higher steric hindrance. 23 This could explain that no oxidation was observed when 18 F-FPBT was treated with a tert-butoxy radical. The X-ray structure of 2-phenylbenzothiazoline indicates that the phenyl ring is perpendicular to the benzothiazoline, which increases the steric hindrance of C2 proton from oxidant attack. 19 Next, the LogD values of 18 F-FPBT and 18 F-FPBT-Ox were determined to be around 1 and −1, respectively. This raises the possibility that the lipophilic 18 F-FPBT can penetrate the cell membrane and react with intracellular superoxide. After oxidation, the hydrophilic 18 F-FPBT-Ox would be trapped inside the cells and generate contrast for PET imaging of oxidative stress. The biodistribution of both compounds in healthy rats demonstrated that 18 F-FPBT has rapid uptake by all major organs including the brain. Conversely, 18 F-FPBT-Ox showed significantly reduced uptake in these organs and was quickly cleared by the kidneys. The PET/CT imaging of 18 F-FPBT in healthy mice provided further evidence that 18 F-  FPBT has fast uptake and clearance by the heart, lung, and brain, which generates low background for imaging oxidative stress in these organs. Additionally, there was little radioactivity accumulation in the bone by both rats and mice indicating that 18

F-FPBT is inert toward defluorination in vivo.
To assess whether 18 F-FPBT can detect oxidative stress in vivo, PET imaging was performed in a rat model of doxorubicin-induced cardiotoxicity, which is characterized by mitochondrial dysfunction leading to excess intracellular superoxide generation. 11−13 Increased myocardial radioactivity uptake was observed in the PET images of doxorubicin-treated rats comparing to the saline control group. The doxorubicintreated animals also had a significantly higher LV to blood SUV ratio than the saline control group in the same period. Post-mortem biodistribution studies at 30 min post IV injection of 18 F-FPBT did not show an increased uptake in the heart of doxorubicin-treated animals. However, this represents a single late time point and does not account for radiotracer bioavailability. PET/CT imaging overcomes these limitations of sampling bias, and these results give us the confidence that 18 F-FPBT can be used to detect oxidative stress in cardiac pathologies. Moreover, blood metabolite analysis has shown that 18 F-FPBT-Ox is stable in blood at least 30 min post IV injection, which further supports the trapping mechanism of 18 F-FPBT post ROS oxidation.
Ex vivo measurements of cardiac biomarkers of oxidative stress (glutathione, malondialdehyde, and iron accumulation) were marginally elevated by doxorubicin treatment, but did not reach statistical significance (Supplementary Figure 6), largely due to the interindividual variability in response, which is a well-characterized aspect of doxorubicin cardiotoxicity, both experimentally and in patients; this phenomenon is, in fact, one of the prime motivators for developing imaging agents that would enable personalized medicine approaches for titrating chemotherapeutic regimes to maximize therapeutic effectiveness while minimizing cardiac risk. 18 F-FPBT is a novel small molecule probe that can be readily prepared by nucleophilic 18 F-labeling in one step with good radiochemical conversion. 18 F-FPBT exhibits chemoselectivity toward superoxide oxidation. The radiotracer was taken up by major organs and excreted through both renal and hepatobiliary routes rapidly. These characteristics would, in principle, make 18 F-FPBT a good candidate for clinical translation. Although further characterization in additional models of oxidative stress is required, we observed an increased uptake in the hearts of doxorubicin-treated rats when compared to controls, highlighting the potential of 18 F-FPBT as a PET tracer to noninvasively detect oxidative stress in vivo.
Synthesis and characterization of the nonradioactive reference compound 1, intermediates, 2, 3, 4, and radiolabeling precursor 5; HPLC chromatograms of 18 Flabeled compounds coelution with the corresponding nonradioactive reference compounds and quality control; 18 F-FPBT in vivo PET/CT imaging of healthy mice; time−activity curves of 18 F-FPBT in rats; blood stability of 18 F-FPBT-Ox 15 min post intravenous injection; ex vivo measurements of cardiac biomarkers of oxidative stress; 1 H, 13 C NMR, and 19 F spectra for all compounds prepared (PDF) not necessarily those of the NHS, the NIHR, or the Department of Health. The Comprehensive Cancer Imaging Centre of King's College London and UCL was funded by the CRUK and EPSRC in association with the MRC and DoH (England). This work was supported by a Wellcome Trust Multi-User Equipment grant [212885/Z/18Z]. This work was supported by EPSRC Program grants [EP/S019901/1 and EP/S032789/1]. The authors would like to thank the radiochemists at St. Thomas Hospital PET Centre for proving 18 F-fluoride and Kavitha Sunnassee and Jayanta K. Bordoloi for their assistance with the in vivo protocols and operation of the PET scanners.