Introduction

Cyclooxygenase (COX) or prostaglandin endoperoxidase synthase, is an enzyme involved in the biosynthesis of prostaglandins, prostacyclins and thromboxanes from arachidonic acid [1,2,3] Among the three known isoforms of COX (COX-1, COX-2 and COX-3), COX-1 has predominantly constitutive activity and is involved in many normal physiological functions [1,2,3]. In contrast, COX-2 is inducible, with relatively low constitutive activity and mostly found in kidney, brain, and heart [1,2,3,4,5,6]. The third isoform, COX-3, may be responsible for febrile response and mediates the antipyretic effects of aspirin and acetaminophen. COX-2 inhibition mediates the analgesic activities of non-steroidal anti-inflammatory medications (NSAIDs) [7,8,9]. COX-2, induced by inflammatory stimuli, catalyzes prostanoid formation associated with inflammation and proliferative diseases.1–3 In the central nervous system (CNS). Neuroinflammation and COX-2 induction are implicated in the pathogenesis of neurodegenerative diseases such as Alzheimer’s disease (AD), Parkinson’s disease (PD), amyotrophic lateral sclerosis (ALS), as well as psychiatric disorders, smoking, seizure disorders, and traumatic brain injury (TBI) [10,11,12,13,14,15]. COX-2 induction is also involved in pain, arthritis, cancers, myocardial infarction, and acute allograft rejection [5, 16,17,18,19,20,21]. COX-2 inhibitors (COXIBs) have anti-inflammatory effects [22,23,24].

Monitoring in vivo changes in COX-2 expression allows quantification of inflammation, tracking of disease course, and assessing target occupancy of newly developed NSAIDs and or monitoring clinical use of FDA-approved COXIB medications. At present, COX-2 quantification is accomplished by ex vivo assays of tissue samples, invasively obtained from biopsies. PET imaging would allow non-invasive and in vivo visualization of COX-2 throughout the body. Existing COX-2 PET ligands are not successful for in vivo quantification of COX-2 due to limitations such as suboptimal COX-2 affinity, high nonspecific binding, de18F-fluorination, and skeletal uptake, poor brain or organ uptake, inability to detect basal or low level of COX-2, and poor signal to noise ratio due to lipophilicity [25,26,27,28,29,30,31,32,33,34,35].[11C]TMI, a highly potent COX-2 inhibitor, developed by our group is the only radiotracer exhibiting partial blocking of constitutive COX-2 level in baboon brain, however, its ability to image inflammation in animal disease models is yet to be proven [36]. A recent report demonstrates that [18F]pyricoxib, a COX-2 inhibitor belonging to the class of 2-(4-methylsulfonylphenyl)lpyrimidines, and triazole analog [18F]triacoxib showed a higher binding in COX-2 positive cells (HCA-7) compared with COX-2 negative cells (HCT-116). However, both tracers demonstrated only modest binding in xenografts, in vivo and did not show significant uptake in the brain (Fig. 1) [37,38,39]. Similarly, [11C]MC1, another member of the class of arylpyrimidines, showed increased binding to COX-2 in animal models of neuroinflammation (Fig. 1) [40]. Despite promising uptake in neuroinflammation, [11C]MC1 failed to detect constitutive or low level of COX-2 in the brain under conditions without inflammation and cold MC1 confirmed no specific binding in CNS and periphery organs [41]. We posit that COX-2 selective inhibitors with sub-nanomolar affinity and radiolabeled with long-lived isotopes such as 18-F (decay half-life 110 min) may help to overcome some of the limitations of the known tracers. Hence, herein, we report the radiosynthesis and evaluation of [18F]6-fluoro-2-(4-(methylsulfonyl)phenyl)-N-(thioph-en-2-ylmethyl) pyrimidin-4-amine ([18F]FMTP; COX-2 IC50 = 2.5 nM), as potential PET ligand for COX-2 imaging in brain and periphery (Fig. 1) [42, 43]. The synthesis of unlabeled FMTP was reported by reacting thiophen-2-ylmethanamine with 4,6-difluoro-2-(4-(methylsulfonylphenyl) pyrimidine [44]. We choose [18F]FMTP, belonging to the pyrimidine class of COX-2 inhibitors, as a potential PET tracer for COX-2 imaging owing to the presence of a chemically and metabolically stable radiolabeling position on the aromatic ring, level of lipophilicity for passive brain entry, and high COX-2 affinity and high selectivity index of COX-1/COX-2 of the pyrimidine-based class of COX-2 inhibitors having the FMTP structure [44, 45].

Fig. 1
figure 1

Chemical structures of [11C]MCI, [18F]Pyricoxib, [18F]Triacoxib and [18F]FMTP

Full size image

Materials and methods

All commercial chemicals and solvents used in the synthesis were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO), or Fisher Scientific Inc. (Springfield, NJ) and were used without further purification. 18F-fluoride was produced using RDS112 cyclotron (Siemens, Knoxville, TN). Gamma-ray detector (Bioscan Flow-Count fitted with a NaI detector) coupled in series with the UV detector (Waters Model 996 set at λ 254 nm) was used for detection of radiolabeled products. Data acquisition for both the analytical and preparative systems was accomplished using a Waters Empower Chromatography System. BxPC3 and PANC1 pancreas carcinoma human cell lines were obtained from ATCC (Manassas, VA). Dynamic microPET image acquisitions were performed with Trifoil microPET/CT scanner.

Synthesis of Cl-MTP

Thiophen-2-ylmethanamine (68 mg, 0.6 mmol) was added to a solution of 4,6-dichloro-2-(4-(methylsulfonyl)phenyl) pyrimidine (0.3 mmol, 90 mg) in 4 mL dry dichloromethane and 0.1 mL triethylamine. The reaction mixture was stirred at room temperature for 1 h and at this time, HPLC analyses showed > 95% consumption of the chloro-substrate. The reaction mixture was evaporated and chromatographed over silica gel using 40:60 ethylacetate-hexane to afford 75 mg (65%) of Cl-MTP as pale yellow solid.

Cl-MTP: mp: 169–171 °C [42, 43, 45]; 1H NMR (400 MHz, CDCl3): 3.0 (3H, s, CH3), 4.85 (2H, bs, CH2), 5.3 (1H, bs, NH), 6.3 (1H,s), 6.9 (1H, m), 7.0 (1H,d, J = 2.81 Hz), 7.2 (2H, d, overlapped with CDCl3), 7.9 (2H, d, J = 8.53 Hz),8.5 (2H,d, J = 8.56 Hz).

Synthesis of fluoro-MTP

Tetra-n-butyl ammonium fluoride (TBAF) (0.1 mL, 1 M solution in. THF) was added to a solution of Cl-MTP (20 mg, 0.05 mmol) in DMF (2 mL). The resulting solution was heated at 140 °C for 1 h at which time point, HPLC analyses showed > 95% consumption of Cl-MTP. The reaction mixture was then allowed to cool, diluted with water (20 mL), and extracted with ethyl acetate (3 × 10 mL). The combined ethyl acetate fractions were further washed with saturated brine, dried over anhydrous MgSO4 and chromatographed over silica gel (20:80 ethyl acetate-hexane) to afford 15 mg (85%) of FMTP as pale yellow solid. The analytical data of F-FMTP is in agreement with the reported data [44].

FMTP: m.p: 131.5 °C; 1H NMR (400 MHz, CDCl3): δ 3.0 (3H, s, CH3), 4.8 (2H, bs, CH2), 5.9 (1H, bs, NH), 6.4 (1H, s), 6.9–7 (1H, m), 7.05 (1H, d, J = 2.82 Hz), 7.2 (2H, d, overlapped with CDCl3), 8 (2H, d, J = 8.51), 8.6 (2H,d, J = 8.54); HRMS (MH+) calculated for: C16H15FN3O2S2: 364.0512; Found: 364.0533.

Radiosynthesis of [ 18 F]FMTP

18F-fluoride (Eclipse cyclotron, Siemens) trapped from QMA was eluted with 1 mL of 10:1 acetonitrile: water, containing kryptofix K222 (36 mg) and potassium carbonate (2 mg). The reaction mixture was azeotropically heated and dried at 98 ºC under a stream of argon by the repeated addition of acetonitrile (4 × 0.5 mL). A solution of ~ 2 mg of Cl-MTP in 500 μL of DMSO was then added to the reaction vial, sealed, and heated for 20 min at 140 ºC. The reaction mixture was allowed to cool to room temperature and diluted with 20 mL water and passed through a classic C-18 Sep-Pak cartridge and eluted with 1 mL acetonitrile. The crude product in acetonitrile was injected onto a semi-preparative HPLC column (Phenomenex, Prodigy ODS-Prep 10 × 250 mm, 10 μ; and eluted with 50:50 acetonitrile-0.1 M ammonium formate with a flow rate of 8 mL/min).[18F]FMTP eluted at 9–10 min was collected based on the γ-detector reading, diluted with 50 mL of deionized water, and passed through a classic C-18 Sep-Pak cartridge, washed with 10 mL of deionized water and eluted with 1 mL of ethanol. Reconstruction of the product in 1 mL of absolute ethanol afforded [18F]FMTP in 35 ± 5% yield (EOS). A portion of the ethanol solution was analyzed by analytical HPLC (Phenomenex, Prodigy ODS(3) 4.6 × 250 mm, 5 μ; mobile phase; 60:40 acetonitrile-0.1 M AMF, flow rate 2 mL/min, tR =  ~ 5 min) to determine the specific activity, chemical and radiochemical purities. The ethanol solution was then diluted to a volume of 10 ml with saline and filtered through a sterile environment, and a portion of this solution was formulated for injection.

Cell uptake of [18F]FMTP.

BxPC3 and Panc1 pancreas carcinoma human cell lines were plated on a 24-well plate at 2 × 105 cells/well. After 48 h, [18F]FMTP was added (2.0 µCi/mL) to the cell medium for 30 min. For blocking experiments, non-labelled FMTP or celecoxib (5 M) was added to cells 30 min before [18F]FMTP. After incubation with [18F]FMTP cells were washed 4 times with cold PBS, lysed with 0.1 N NaOH, and counted in a gamma counter (Hidex AMG, LabLogic, Tampa, FL).

Micro PET imaging

The animals (male white mice) were anesthetized with isoflurane (1–2% isoflurane in 100% oxygen) using a nose cone, and a 29-gauge needle connected to a catheter was placed into the lateral tail vein (n = 3). For neuroinflammation, mice were stereotactically injected with 5 µg of LPS or PBS (controls) in the brain, adopting a known procedure [46, 47], and waited 24 h for imaging experiments (n = 4). The animal was placed in a prone position on the platform of the scanner and moved into the center of the field of view guided by laser beam calibration. Immediately after scan start, [18F]FMTP (~ 2.5 MBq 25 mL in 20% ethanol-saline solution) was injected through the tail-vein catheter manually. The error caused by the injector and catheter was corrected by subtracting the remaining dose. 40 min dynamic imaging was acquired on a microPET scanner (Siemens Inveon). The acquired list-mode data were reconstructed with 3-D ordered subset expectation maximization (OSEM) algorithm [48, 49] using the software of Siemens Inveon Acquisition Workplace with a framing protocol of 2 × 30 s, 4 × 60 s, 3 × 120 s, 3 × 180 s, and 4 × 300 s.

PET data analysis

Using PMOD software (version 4.0, Switzerland), the three-dimensional ellipsoid volume of interests (VOIs) ranging from 2-to-6 mm were placed manually at the center of the brain, heart (blood-pool), liver, proximal humerus (bone) and posterior cervical muscle. The standardized uptake values (SUVs) were estimated using a calibration factor calculated from the phantom study and time-activity curves (TACs) were derived from VOIs in the series of reconstructed images.

Results and discussion

Synthesis of FMTP and radiosynthesis of [ 18 F]FMTP

Synthesis of Cl-MTP, the radiolabeling precursor molecule, was achieved in four steps from commercially available 4-(methylthio)benzimidamide in 65% yield by adopting a previously reported procedure (Scheme 1) [45]. The non-radioactive standard FMTP was obtained in 85% yield by reacting Cl-MTP with TBAF in dimethyl sulfoxide (DMSO) as shown in Scheme 1. Our approach for the synthesis of FMTP was different than previously reported [43]. The radiosynthesis of [18F]FMTP was optimized by reacting 18F-fluoride/K222/K2CO3 using chlorine to 18F-fluorine displacement reaction with the precursor molecule (Scheme 1). To our knowledge, this is the first successful radiolabeling of a chlorine to-18F-fluorine displacement at the 6th position of aryl[1,3]pyrimidines.[18F]FMTP was obtained in 35 ± 5% yield at EOS with > 99% radiochemical purity and a molar activity of 92.5 ± 18.5 GBq/μ mol (n = 10). The total synthesis time of [18F]FMTP was 1 h at EOS and the radioligand was found to be stable in 10%-ethanol-saline formulation.

Scheme 1
scheme 1

Synthesis of FMTP and radiosynthesis of [18F]FMTP

Full size image

Cell uptake of [ 18 F]FMTP

After the successful radiosynthesis and formulation, proof of concept and selectivity of [18F]FMTP binding to COX-2 was examined by evaluating the tracer uptake in COX-2 positive BxPC3 and a control COX-2 negative PANC1 human pancreatic carcinoma cell lines. Figure 2 illustrates a > twofold uptake of [18F]FMTP in BxPC3 cells compared with PANC1 cells. This binding was substantially blocked with unlabeled FMTP. Blocking with celecoxib, an FDA approved COX-2 inhibitor, produced less proportional blocking compared with cold FMTP. As anticipated [18F]FMTP showed negligible binding to PANC1 cells and no specific binding with FMTP or celecoxib block.

Fig. 2
figure 2

Binding of [18F]FMTP in BxPC3 and PANC-1 cells. Values are reported as the mean ± SD from four independent experiments

Full size image

MicroPET evaluation of [ 18 F]FMTP

MicroPET evaluation of [18F]FMTP was initially performed in nude white mice (n = 3). Figure 3 shows the summed early frames (0–10 min) and total frames (0–40 min) of microPET images. As evident from the images, the tracer penetrates the BBB and subsequently shows a fast washout of activity from the brain. Time activity curves (TACs) also indicate an initial rapid influx of radioactivity followed by rapid washout (Fig. 4). The uptake of [18F]FMTP peaked around one minute in heart and brain and then decreases rapidly. The absence of retention of [18F]FMTP activity in these organs is predicted due to low COX-2 expression in the normal mouse brain. The highest radioactivity outside the heart was found in the liver, and the delay to peak at 5–10 min is probably due to the accumulation of radioactive metabolite(s). The tracer did not show uptake in the spine and skeleton which indicates a lack of de18F-fluorination, an advantage of [18F]FMTP for use in in vivo PET imaging. We studied the binding of [18F]FMTP in LPS-induced neuroinflammation in mice compared with vehicle (PBS) treated-mice [46]. Intracranial injection of LPS in mice is known to generate COX-2 induction and neuroinflammation after around 24 h [47]. Therefore, we performed PET imaging of LPS treated mice after 24 h with [18F]FMTP and found an approximately twofold increase of tracer binding in the brain compared with PBS-treated mice (n = 4) (Fig. 5). TACs in whole-brain confirmed higher binding of [18F]FMTP in LPS treated mice compared with PBS treated controls (Fig. 6).

Fig. 3
figure 3

microPET images of [18F]FMTP in mice, A: 0–10 min summed transaxial; B: 0–10 min summed sagittal; C: 0–40 min summed transaxial; D: 0–40 min summed sagittal. Crosses represent the region with the brain

Full size image
Fig. 4
figure 4

Representative TACs of [18F]FMTP in mice

Full size image
Fig. 5
figure 5

Representative microPET images of [18F]FMTP in PBS and LPS treated mice

Full size image
Fig. 6
figure 6

TACs of [18F]FMTP in PBS and LPS treated mice. Values are reported as the mean ± SD from three independent experiments

Full size image

Conclusions

In summary, we report a new method for the radiosynthesis of [18F]FMTP via a chlorine-to-18F-fluorine displacement reaction of aryl[1,3]pyrimidine core molecule. Proof of concept of the use of [18F]FMTP for quantifying COX-2 was established first, in vitro, in COX-2 positive BxPC3 cells. MicroPET imaging of normal mice demonstrated BBB penetration and a fast washout of radioactivity from the brain, likely due to the low concentration of COX-2 in the normal brain.[18F]FMTP showed a higher binding in LPS-induced neuroinflammation compared to binding in the brain of control mice. Specific binding to COX-2 in cell lines, lack of in vivo de-18F-fluorination and skeletal uptake, and BBB permeability, and higher brain binding in neuroinflammation qualifies [18F]FMTP as a potential PET tracer for imaging inflammation where COX-2 overexpression is reported.