2-(4-Methylsulfonylphenyl)pyrimidines as Prospective Radioligands for Imaging Cyclooxygenase-2 with PET—Synthesis, Triage, and Radiolabeling

Cyclooxygenase 2 (COX-2) is an inducible enzyme responsible for the conversion of arachidonic acid into the prostaglandins, PGG2 and PGH2. Expression of this enzyme increases in inflammation. Therefore, the development of probes for imaging COX-2 with positron emission tomography (PET) has gained interest because they could be useful for the study of inflammation in vivo, and for aiding anti-inflammatory drug development targeting COX-2. Nonetheless, effective PET radioligands are still lacking. We synthesized eleven COX-2 inhibitors based on a 2(4-methylsulfonylphenyl)pyrimidine core from which we selected three as prospective PET radioligands based on desirable factors, such as high inhibitory potency for COX-2, very low inhibitory potency for COX-1, moderate lipophilicity, and amenability to labeling with a positron-emitter. These inhibitors, namely 6-methoxy-2-(4-(methylsulfonyl)phenyl-N-(thiophen-2ylmethyl)pyrimidin-4-amine (17), the 6-fluoromethyl analogue (20), and the 6-(2-fluoroethoxy) analogue (27), were labeled in useful yields and with high molar activities by treating the 6-hydroxy analogue (26) with [11C]iodomethane, [18F]2-fluorobromoethane, and [d2-18F]fluorobromomethane, respectively. [11C]17, [18F]20, and [d2-18F]27 were readily purified with HPLC and formulated for intravenous injection. These methods allow these radioligands to be produced for comparative evaluation as PET radioligands for measuring COX-2 in healthy rhesus monkey and for assessing their abilities to detect inflammation.


Target Inhibitor Selection for Radioligand Development
We identified five 2-(4-methylsulfonylphenyl)pyrimidines (17)(18)(19)22, and 25) from the report of Orjales et al. [35] that presented: (i) high potency for inhibition of human COX-2 from human whole blood (IC 50 ≤ 5.1 nM); (ii) very low potency for inhibition of the isoform COX-1 from human whole blood (IC 50 > 100 nM); (iii) computed lipophilicity in an acceptable range for a PET radioligand (clogD 2-3, ideally about 2.2); and (iv) at least one site that was expected to be amenable to labeling with either carbon-11 or fluorine-18. In addition to moderate lipophilicity, we sought other properties in these inhibitors that would be consistent with ability to traverse the blood-brain barrier, such as molecular weight of less than 500 Da, computed total polar surface area (tPSA) of less than 90 Å 2 , low hydrogen bond donor number (HBD), less than 9 heteroatoms, pKa less than 9.5, and absence of acidic ionizable groups such as sulfonamido and carboxyl groups [32] (Table 1). We designed another five previously unknown 2-(4-methylsulfonylphenyl)pyrimidines (20,21,27,29, and 30) as potentially new and potent COX-2 inhibitors with other ostensibly favorable properties for PET radioligand development ( Table 1). Each of these compounds carried a group that was expected to be readily amenable for labeling with either carbon-11 or fluorine-18 (i.e., in Table 1 the R 1 group with carbon-11 for 21 and with fluorine-18 for 20, 27, and 29; and the R 3 group with carbon-11 for 30). We also prepared two hydroxy compounds with low COX inhibitory potency [35] as possible labeling precursors, namely 24 and 26.
In two cases, the yields from the non-optimized microwave-method matched (20) or surpassed those from the non-microwave method (21). Yields of the five N-(thiophen-2-ylmethyl)pyrimidin-4-amines by either method ranged from moderately low (22%) to high (80%). Treatment of 12 with benzylamine instead of thiophen-2-ylmethanamine under either set of reaction conditions gave 33% yield of target inhibitor 22. Treatment of the dichloro compound 11 with benzylamine under the non-microwave conditions gave inhibitor 23 in high yield (76%). The microwave method was not tested for this compound. Likewise, treatment of 11 with thiophen-2ylmethanamine gave ligand 25 in high yield (74%) by either method.
Treatment of 11 with potassium fluoride [37] replaced both chloro groups to give the difluoro analogue 28 in 30% yield. Treatment of 28 with thiophen-2-ylmethanamine under microwave conditions then gave the fluoro compound 29 in 53% yield, a slightly higher yield than obtained under the slower non-microwave conditions (47%).
The 4-chloro compounds 23 and 25 were readily converted with cesium hydroxide under palladium(II)-mediated conditions [38] into the respective 4-hydroxy analogues, 24 and 26, in low but useful yields (20% and 22%, respectively). Treatment of 26 with fluoroiodomethane under basic conditions gave the 4-fluoromethyl compound 27 in 56% yield. Finally, treatment of 25 with iodomethane with sodium hydride in DMF gave the N-methyl derivative 30 in 46% yield.

COX-1 and COX-2 Inhibition Assays
Species differences in the structures of COXs are well documented, especially between rodent and human [39]. Therefore, the choice of pre-clinical species for evaluation of prospective PET radioligands must be made with care. We aimed to use rhesus monkey if inhibitors from the 2-(4methylsulfonylphenyl)pyrimidine class showed very similar COX inhibitory potency between rhesus monkey and human. The 2-(4-methylsulfonylphenyl)pyrimidines of interest were tested for COX-1 and COX-2 inhibitory potencies using fresh monkey and human blood ( Table 2). For compounds whose potencies for inhibiting human COX-1 and COX-2 had already been reported (17-19, 22, 24, 25) [35], our assay results were in fair agreement for rank order of inhibitory potency, except that we obtained somewhat higher IC50 values for human COX-2 inhibitory potency. None of the tested compounds showed monkey or human COX-1 inhibitory IC50's below 1 nM, in accord with the observations of Orjales et al. [35] for all compounds in this class. These compounds appear to be too bulky for strong binding to the COX-1 isoform.
The thioethyl compound 19 and the hydroxy compound 26 showed much lower inhibitory potency for monkey COX-2 than that reported for human COX-2 [35]. Nevertheless, in several other cases (17, 20-22, 24, 27) inhibitory potencies for the monkey COX enzymes were found to be very close to those for the human enzymes. Therefore, we concluded that monkey could indeed serve as a useful preclinical model for testing candidate PET radioligands for imaging COX-2.

COX-1 and COX-2 Inhibition Assays
Species differences in the structures of COXs are well documented, especially between rodent and human [39]. Therefore, the choice of pre-clinical species for evaluation of prospective PET radioligands must be made with care. We aimed to use rhesus monkey if inhibitors from the 2-(4-methylsulfonylphenyl)pyrimidine class showed very similar COX inhibitory potency between rhesus monkey and human. The 2-(4-methylsulfonylphenyl)pyrimidines of interest were tested for COX-1 and COX-2 inhibitory potencies using fresh monkey and human blood ( Table 2). For compounds whose potencies for inhibiting human COX-1 and COX-2 had already been reported (17-19, 22, 24, 25) [35], our assay results were in fair agreement for rank order of inhibitory potency, except that we obtained somewhat higher IC 50 values for human COX-2 inhibitory potency. None of the tested compounds showed monkey or human COX-1 inhibitory IC 50 's below 1 nM, in accord with the observations of Orjales et al. [35] for all compounds in this class. These compounds appear to be too bulky for strong binding to the COX-1 isoform.
The thioethyl compound 19 and the hydroxy compound 26 showed much lower inhibitory potency for monkey COX-2 than that reported for human COX-2 [35]. Nevertheless, in several other cases (17, 20-22, 24, 27) inhibitory potencies for the monkey COX enzymes were found to be very close to those for the human enzymes. Therefore, we concluded that monkey could indeed serve as a useful preclinical model for testing candidate PET radioligands for imaging COX-2. Among the N-(thiophen-2-ylmethyl)pyrimidin-4-amines, monkey and human COX-2 inhibitory potency depended on compound structure in the following manner. Compounds containing an electron-donating group in the R 1 position of the general structure shown in Table 2, such as OMe (17), OEt (18), SEt (19), OCH2CH2F (20), SMe (21), or OCH2F (27), showed high COX-2 inhibitory potencies with IC50's between 1 and 5 nM. Compounds containing relatively small electron withdrawing groups, such as Cl (25)  Among the N-(thiophen-2-ylmethyl)pyrimidin-4-amines, monkey and human COX-2 inhibitory potency depended on compound structure in the following manner. Compounds containing an electron-donating group in the R 1 position of the general structure shown in Table 2, such as OMe (17), OEt (18), SEt (19), OCH 2 CH 2 F (20), SMe (21), or OCH 2 F (27), showed high COX-2 inhibitory potencies with IC 50 's between 1 and 5 nM. Compounds containing relatively small electron withdrawing groups, such as Cl (25) or F (29), in the R 1 position also showed quite high COX-2 inhibitory potency (IC 50 , 3-4 nM). N-Methylation of 25 effectively abolished inhibitory potency at human COX-2. Replacement of the 2-thienyl group of 17 with a phenyl group as in 22 resulted in decreased COX-2 inhibitory potency. Compounds 17, 20, and 27 showed the highest inhibitory potencies for COX-2 in both species. The phenols, 24 and 26, which were potentially useful as precursors for radiolabeling, showed low inhibitory potency for COX-2 [35].

Pharmacological Screening
At 10 µM concentration, inhibitors 17, 18, 27, and 30 showed less than 50% inhibition of radioligand binding at all tested receptors, transporters, and binding sites, except 17 (50, 50.6 and 53.7% inhibition at 5-HT1 B , benzodiazepine (BzP), and DAT (dopamine transporter), respectively), and 27 (50.8% inhibition at BzP). Therefore, this class of COX-2 inhibitor was generally devoid of very high binding affinity at any of the tested off-target sites, as is desirable for an effective PET radioligand.

Selection of Candidates for Radiolabeling
Inhibitors 17, 20, and 27 were selected from the prepared set (Table 2) for radiolabeling because they presented the highest inhibitory potencies for human and monkey COX-2, very low inhibitory potency for human and monkey COX-1, computed lipophilicities in a range generally acceptable for PET radioligands, and readily accessible positions for labeling with 11 C (17) or 18 F (20 and 27).

Pharmacological Screening
At 10 μM concentration, inhibitors 17, 18, 27, and 30 showed less than 50% inhibition of radioligand binding at all tested receptors, transporters, and binding sites, except 17 (50, 50.6 and 53.7% inhibition at 5-HT1B, benzodiazepine (BzP), and DAT (dopamine transporter), respectively), and 27 (50.8% inhibition at BzP). Therefore, this class of COX-2 inhibitor was generally devoid of very high binding affinity at any of the tested off-target sites, as is desirable for an effective PET radioligand.

Selection of Candidates for Radiolabeling
Inhibitors 17, 20, and 27 were selected from the prepared set (Table 2) for radiolabeling because they presented the highest inhibitory potencies for human and monkey COX-2, very low inhibitory potency for human and monkey COX-1, computed lipophilicities in a range generally acceptable for PET radioligands, and readily accessible positions for labeling with 11 C (17) or 18 F (20 and 27). The radiochemical purities of all three radioligands exceeded 99% (Figure 1). Formulated radioligands were free of appreciable amounts of labeling precursor (26) or derived impurities ( Figure 1). Each formulated radioligand was radiochemically stable for more than 1 h as measured by HPLC.  Table 3). The measured value for [ 18 F]20 was almost the same as the computed value. All the measured values were within the logD range (2-4) observed for many effective PET radioligands for imaging proteins in brain [44], but rather higher than might be considered ideal. The radiochemical purities of all three radioligands exceeded 99% (Figure 1). Formulated radioligands were free of appreciable amounts of labeling precursor (26) or derived impurities ( Figure 1). Each formulated radioligand was radiochemically stable for more than 1 h as measured by HPLC.

Calculated and Measured LogD for Prepared Radioligands
The measured lipophilicity values (logD's) for [ 11 C]17 and [d2-18 F]27 were higher than the computed values by 0.94 and 0.98, respectively ( Table 3). The measured value for [ 18 F]20 was almost the same as the computed value. All the measured values were within the logD range (2-4) observed for many effective PET radioligands for imaging proteins in brain [44], but rather higher than might be considered ideal.

General
Solvents (analytical grade) and reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless stated otherwise. The products were isolated with flash chromatography on a

General
Solvents (analytical grade) and reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless stated otherwise. The products were isolated with flash chromatography on a CombiFlash Rf+ system (Teledyne ISCO; Lincoln NE) equipped with a silica gel cartridge (12 g, RediSep ® Rf; Isco: Lincoln, NE, USA) using hexane and EtOAc as mobile phase (general method A). If necessary to achieve purity exceeding 95%, COX-2 inhibitors were subjected to reversed phase HPLC (Beckman Coulter, Fullerton, CA, USA) on an XBridge C18 column (19 × 150 mm, 10 µm; Waters; Milford, MA, USA) eluted at 6 mL/min with a gradient mobile phase composed of aq. ammonium hydroxide (0.25 mM) (A) and acetonitrile (B) (general method B), with eluate monitored for absorbance at 254 nm. Melting points were determined on an SMP20 apparatus (Stuart; Staffordshire, UK). 1 H (400.13 MHz), 13

Radiochemistry Materials and Methods
Radioactivity from carbon-11 or fluorine-18 was measured with a calibrated dose calibrator (Atomlab 300, Biodex Medical Systems, Shirley, NY, USA) or an automatic γ-counter (Wizard 3", 1480; PerkinElmer; Waltham, MA, USA). Radioactivity measurements were corrected for physical decay. Radiochemistry was executed in lead-shielded hot-cells for personal protection from radiation. Radioligands were separated with reversed phase HPLC (Beckman Coulter, Fullerton, CA, USA) on a Luna (C-18/2) column (10 × 250 mm, 10 µm; Phenomenex; Torrance, CA, USA). Radioligands were identified and measured for radiochemical purity with reversed phase HPLC (Beckman Coulter, Fullerton, CA, USA) on a Luna C18(2) column (4.6 × 250 mm, 10 µm). All HPLC column eluates were monitored for radioactivity and absorbance at 254 nm. Absorbance responses for analytical HPLC systems were calibrated for mass of injectates to enable molar activities [47] [40]. Thus, at the end of proton irradiation, [ 11 C]carbon dioxide was delivered to a PETtrace MeI Process Module (GE Medical Systems; Severna Park, MD, USA) through stainless tubing (OD 1/8 in, ID 1/16 in) over 3 min. The [ 11 C]carbon dioxide was reduced to [ 11 C]methane with hydrogen over heated nickel (360 • C) and trapped on molecular sieves (13X). Finally, the [ 11 C]methane was recirculated over iodine at 720 • C to generate [ 11 C]iodomethane, which was trapped on Porapak Q held in the recirculation path.

Radiosynthesis of [ 11 C]17
The phenol precursor 26 (1.01 mg, 2 µmol) was dissolved in DMF (80 µL) at 7 min before the end of radionuclide production. The solution was sonicated and then TBAH in methanol (1 M, 3.0 µL) was added. This solution was loaded into the loop (constructed of stainless steel tubing) of an AutoLoop apparatus (Bioscan; Washington, DC, USA). [ 11 C]Iodomethane was then released into the loop and held for 5 min at RT.

Conclusions
Three COX-2 inhibitors, 17, 20, and 27, were identified from the 2-(4-methylsulfonylphenyl)pyrimidine class as having promising physiochemical and pharmacological properties for development as PET radioligands for imaging COX-2 in monkey and human in vivo. Methods were established for satisfactorily labeling these ligands with positron-emitters for comparative PET imaging in monkey in vivo, including a neuroinflammation model [50]. These results will be published in detail elsewhere.