Deuteration versus ethylation – strategies to improve the metabolic fate of an 18F-labeled celecoxib derivative

The inducible isoenzyme cyclooxygenase-2 (COX-2) is closely associated with chemo-/radioresistance and poor prognosis of solid tumors. Therefore, COX-2 represents an attractive target for functional characterization of tumors by positron emission tomography (PET). In this study, the celecoxib derivative 3-([18F]fluoromethyl)-1-[4-(methylsulfonyl)phenyl]-5-(p-tolyl)-1H-pyrazole ([18F]5a) was chosen as a lead compound having a reported high COX-2 inhibitory potency and a potentially low carbonic anhydrase binding tendency. The respective deuterated analog [D2,18F]5a and the fluoroethyl-substituted derivative [18F]5b were selected to study the influence of these modifications with respect to COX inhibition potency in vitro and metabolic stability of the radiolabeled tracers in vivo. COX-2 inhibitory potency was found to be influenced by elongation of the side chain but, as expected, not by deuteration. An automated radiosynthesis comprising 18F-fluorination and purification under comparable conditions provided the radiotracers [18F]5a,b and [D2,18F]5a in good radiochemical yields (RCY) and high radiochemical purity (RCP). Biodistribution and PET studies comparing all three compounds revealed bone accumulation of 18F-activity to be lowest for the ethyl derivative [18F]5b. However, the deuterated analog [D2,18F]5a turned out to be the most stable compound of the three derivatives studied here. Time-dependent degradation of [18F]5a,b and [D2,18F]5a after incubation in murine liver microsomes was in accordance with the data on metabolism in vivo. Furthermore, metabolites were identified based on UPLC-MS/MS.


Introduction
Celecoxib (tradename Celebrex®) is an FDA-approved anti-inammatory, analgesic, and antipyretic drug which belongs to the class of selective cyclooxygenase-2 (COX-2) inhibitors and can be prescribed for the treatment of rheumatoid arthritis, osteoarthritis, and acute pain. The development of this drug in the late 1990s represented a breakthrough in the development of anti-inammatory drugs taking into account that the longterm gastrointestinal toxicity of common non-steroidal anti-inammatory drugs (NSAIDs) like ibuprofen could be circumvented with this new class of COX-2-selective inhibitors (Coxibs). Today however, only a few Coxibs such as celecoxib, etoricoxib, and the prodrug parecoxib are in clinical use because of general concerns about the cardiovascular safety in long-term applications of this inhibitor class. 1,2 The mode of action is based on the inhibition pattern for cyclooxygenase (COX), an enzyme which converts arachidonic acid to prostaglandin H 2 . Thereby, COX catalyzes a key step in the synthesis of autocrine and paracrine acting prostaglandins and thromboxane A 2 that are involved in the regulation of a multitude of physiological and pathophysiological processes. Two isoforms of cyclooxygenases are known: the constitutively and throughout the whole body expressed isoform COX-1 which regulates mainly homeostatic processes and its counterpart COX-2 that is nearly absent in the body with exception of some organs, such as kidney and brain. However, COX-2 is induced in inammatory conditions and neoplasia. The COX-2 pathway hence provides prostaglandins at inamed sites and, furthermore, in tumors which makes it a potential target for diagnosis and treatment of a variety of disorders. COX-2 is overexpressed in chronic inammatory diseases such as rheumatoid arthritis or in neurodegenerative diseases such as Parkinson's and Alzheimer's. In addition, as a key player in tumorassociated inammation, COX-2 modulates radiation sensitivity, progression, and metastasis in a variety of cancers and leads to a reduction in overall survival. [1][2][3][4][5][6][7][8] A high 'molecular contrast' between healthy tissue and pathological lesions also forms the basis for successful radionuclide-based approaches. From the perspective of the oncologically oriented radiochemist/radiopharmacologist, COX-2 is a highly promising drugable target because (i) under physiological conditions it is nearly absent in tissues except kidneys, heart, and brain, (ii) its expression is inducible and COX-2 is overexpressed at inammatory sites and in (inamed) tumor tissue, and (iii) the availability of clinically approved drugs would allow the clinician to intervene in a personalized manner in, e.g., radioresistant tumors with high COX-2 expression. Among a multitude of labeled COX-2 inhibitors, [9][10][11] a wide variety of celecoxib-based radiotracer approaches have been described in the literature ( Fig. 1) for studies on absorption, distribution, metabolism, and excretion (carbon-14 (ref. 12)) as well as for positron emission tomography (PET; carbon-11, 13-18 uorine-18 (ref. [19][20][21][22][23][24]) or single photon emission computed tomography (SPECT; iodine-123/125, [25][26][27][28][29] , technetium-99m 30,31 ). However, up to now no radiotracer could be transferred to clinical application because of high non-specic binding, metabolic instability, inability to demonstrate COX-2-specic binding in vivo, and/or lack of data from primates or rst-in-patient studies. As one example, the [ 18 F]uoromethylsubstituted celecoxib derivative was synthesized by Uddin et al. 21 and evaluated in an inammation model as well as in tumor xenogras of COX-2-negative human colorectal carcinoma (HCT116) and COX-2-expressing human head and neck squamous cell carcinoma (1483 HNSCC) providing evidence for COX-2-specic uptake of this radiotracer in vivo. Unfortunately, no further studies on this radiotracer have been presented up to now. One reason might be, as shown by Uddin et al., 21 that the [ 18 F]uoromethyl-substituted radiotracer is prone to 18 F-deuorination in vivo and, presumably, binds to carbonic anhydrase in the erythrocytes as it is known for celecoxib and other sulfonamide-substituted Coxibs. 28,32,33 Based on our interest in the development of radiolabeled COX-2 inhibitors 34-37 and to minimize non-specic binding, we aimed to synthesize the respective methylsulfonyl-substituted celecoxib derivatives and thus we herein describe two attempts to increase their metabolic stability. In general, possibilities to increase the metabolic stability of a radiotracer containing 18 F bonded to an sp 3 -hybridized carbon atom include (i) the deuteration vicinal to the uorine substituent [38][39][40][41] or at other parts of the molecule, [41][42][43] (ii) the replacement of an alkyl chain by a cycloalkyl ring, 44 (iii) the use of CF 2 18 F groups, 20 and (iv) the application of uorophenyl-instead of uoroalkylsubstituted analogs wherever possible. [45][46][47] As an early example, Zhang et al. demonstrated that utilization of a [ 18 F] uoromethoxy-d 2 group in a peripheral benzodiazepine receptor ligand decreased its metabolization and, hence, 18 F-deuorination in the brain of rats but not in primates. 39 Accordingly, the use of [ 18 F]uoroethoxy-d 4 group was found to increase the metabolic stability of verapamil analogs. 41 In the present study, we focus on the development of a deuterated and a uoroethyl-substituted celecoxib analog ( Fig. 1). We present the synthesis, COX inhibition studies, radiolabeling with uorine-18 as well as radiopharmacological evaluation in vitro and in vivo for three methylsulfonyl-substituted celecoxib derivatives to identify potent and metabolically stable radiotracers.
2 Results and discussion

Synthesis
The synthetic route for the labeling precursors 4a,b and [D 2 ]4a as well as the uorine-19-substituted reference compounds 5ad and [D 2 ]5a comprises a three-step synthesis starting from the diketoacid esters 1a-c (Scheme 1). At rst, Knorr pyrazole synthesis by cyclocondensation of the b-dicarbonyl compounds 1a-c with methylsulfonylphenylhydrazine hydrochloride in methanol furnished the acid estersubstituted pyrazoles 2a-e in 15-96% yield. As commonly utilized for the synthesis of celecoxib, the reaction of the a,gdiketoacid ester 1a with the respective phenylhydrazine leads predetermined by the ÀI and ÀM effect of the neighboring ester group to the initial condensation at the a-keto group and, hence, regioselectively to the 1,5-diphenyl-substituted pyrazole 2a. In comparison, in the b,d-diketoacid ester 1b the corresponding keto group is adjacent to a methylene group exerting a +I-effect instead which disfavors the initial condensation at this site and therefore results in the non-regioselective reaction towards the 1,5-diphenyl-(2b,c: 47-69% yield) and 1,3diphenyl-substituted (2d,e: 15-17% yield) products. The structural identity of the regioisomers was conrmed by NMR spectroscopy, and X-ray crystal structure analysis in case of 2e (Fig. 2, le). Exemplarily, 2e showed coupling in NOESY and ROESY spectra for the signal of the methylene group (d ¼ 3.91 parts per million (ppm)) to the pyrazole proton (6.85 ppm) and to the aromatic protons of the para-methylsulfonyl-substituted phenyl ring (7.79 ppm). In contrast, in NOESY and ROESY spectra of 2c only the coupling of the methylene group (3.76 ppm) to the pyrazole proton (6.57 ppm) was observed (for selected expansions of NOESY spectra of 2c and 2e see ESI, Fig. S1 and S2 †). The asymmetric unit of crystals of 2e contains only the regioisomer 2e, not 2c (see ESI, Fig. S3 †). Although the pyrazoles 2d,e do not exhibit the vicinal diaryl heterocyclic structure of Coxibs we selected 2d and further processed this compound to generate 5d as a negative control for our COX inhibition assay. In the next step, the four pyrazole derivatives 2a-d were reduced using lithium aluminum hydride in THF to form the alcohols 3a-d, and in case of 2a also with lithium aluminum deuteride in THF to form the deuterated alcohol [D 2 ] 3a, in 61-96% yield. The radiolabeling precursors 4a,b and [D 2 ] 4a, bearing a tosyl leaving group, were synthesized in 48-76% yield by conversion of the respective alcohols with tosyl anhydride in DCM utilizing 4-(dimethylamino)pyridine (DMAP) and pyridine as bases. The uoro-substituted reference compounds 5a-d and [D 2 ]5a were obtained by uorination of the alcohols 3a-d and [D 2 ]3a, respectively, with (diethylamino)sulfur tri-uoride (DAST) in DCM in 47-90% yield. The X-ray crystal structure obtained for the 1,5-diphenyl-3-uoroethylsubstituted reference compound 5c (Fig. 2, right) conrmed the structural identity of this regioisomer. Crystals of 5c contained two symmetry-independent molecules in the asymmetric unit (ESI, Fig. S4 †). Both have the same group connectivity, such that they are the same regioisomers. They only differ in dihedral angles around single bonds (i.e. C1-C4, C3-C13, C9-S1) as a result of crystal packing effects.

COX inhibition
COX inhibition potency was determined in vitro for the reference compounds 5a-d and [D 2 ]5a (Table 1) using puried ovine COX-1 and recombinant human COX-2 enzyme in the commercially available 'COX Fluorescent Inhibitor Screening Assay KIT' (Cayman Chemical, Ann Arbor, MI). The data obtained for the COX-2-selective inhibitor 5a are in accordance with the inhibition pattern previously determined by Uddin et al. with a 14 C-based COX assay. 21 As expected, deuteration did not markedly change COX-2 inhibitory potency of [D 2 ]5a, while both prolongation to the uoroethyl chain resulting in the tolyl derivative 5b and the lack of a methyl substituent in derivative 5c weakened the inhibitory potency and selectivity for COX-2. The regioisomer 5d, which does not contain the typical vicinal diarylheterocyclic structure of Coxibs, showed neither COX-2 nor COX-1 inhibition as anticipated.

Radiosynthesis
Radiolabeling with [ 18 F]uoride was achieved starting from the tosylated precursors 4a,b and [D 2 ]4a under standard conditions (Scheme 2 and Table 2) in a synthesis sequence using an automated radiosynthesizer (TRACERlab FX-N). Aer azeotropic drying of [ 18 F]uoride, labeling was performed at 80 C in acetonitrile for 15 min followed by dilution of the mixture with eluent, ltration, and purication by semi-preparative HPLC. In order to obtain the radiotracer in high chemical purity, we adjusted the eluent to 40% MeCN/60% water with 0.1% tri-uoroacetic acid (TFA) and allowed the tracers to be eluted from the HPLC column aer a separation time of approximately 50 min. Aer a C 18 -based solid phase extraction to remove the eluent, the nal tracers [ 18 F]5a,b and [D 2 , 18 F]5a were isolated in radiochemical yields (RCY) between 33 and 38% and high radiochemical purity (RCP > 99%) as well as chemical purity (CP Fig. 2 Molecular structures of compounds 2e (left) and 5c (right) in the crystal (ORTEP plot with atom labeling scheme, displacement thermal ellipsoids are drawn at 50% probability level. Only one of the two independent molecules of the asymmetric unit of 5c is shown).

Cellular uptake
For evaluation of COX-2 selectivity in vitro, cell uptake was studied in two melanoma cell lines with known COX expression pattern: 35 the COX-2-negative Mel-Juso and COX-2-positive A2058 cell line, both expressing COX-1 at low but constitutive levels. Cellular binding and uptake of [ 18 F]5a,b and [D 2 , 18 F]5a was found to be independent of the COX-2 expression since cellular uptake by COX-2-negative Mel-Juso cells was slightly higher compared to uptake by COX-2-positive A2058 cells (Fig. 3). Furthermore, inhibition with the known COX-2selective inhibitor celecoxib resulted in a reduction of cell uptake, however, in both cell lines in a comparable manner pointing to a saturation of other binding sites instead of COX-2.
For celecoxib it is known that it can also bind and interact with other targets which contributes to its anticarcinogenic effects. 49 group which is a major determinant of substrate and CA inhibitor specicity 57 and hence are not likely potent CA inhibitors. However, it is possible that other interactions play a role in cellular binding effects observed in this investigation because in both cell lines a COX-independent blocking effect was observed at a concentration of 100 mM celecoxib which is higher than the IC 50 value of the interactions with moderate affinity described above. Because of that and insufficient COX inhibition potencies, compounds 5a,b and [D 2 ]5a did not meet the requirements for in vivo studies in tumor xenogra-bearing mice. Instead, we focused on biodistribution studies in healthy rats to evaluate the inuence of deuteration or elongation on metabolic stability for the celecoxib-based radiotracers.

Metabolic stability in vivo
To assess the in vivo behavior of [ 18 F]5a,b and [D 2 , 18 F]5a in healthy rats with focus on metabolic stability, (i) ex vivo biodistribution at 5 and 60 min post injection (p.i.) was analyzed, (ii) small animal PET imaging was performed up to 90 min p.i., and (iii) blood samples were analyzed for metabolites at 20 and 60 min p.i. Biodistribution experiments (Fig. 4, see ESI, Tables S5 and S6 †) showed a similar pharmacokinetic pattern for [ 18 F]5a,b and [D 2 , 18 F]5a. At 5 min p.i., the highest initial radioactivity concentration was found in liver, adrenals, and brown as well as white adipose tissue (BAT/WAT). Whereas radioactivity concentration considerably decreased in almost all organs from 5 to 60 min p.i., radioactivity concentration in intestine, urine, and femur increased over time. At 60 min p.i., the hepatobiliary and renal excretion was found to increase within the row [ 18   F]5b, metabolites of the three radiotracers were analyzed in samples of blood as well as of liver, urine, and intestinal content by radio-HPLC (Fig. 6,  injection in healthy rats. Analysis of blood samples (Fig. 6A) conrmed the rapid blood clearance of [ 18 F]5a, [D 2 , 18 F]5a, and [ 18 F]5b resulting from very fast tracer distribution (t 1/2 ¼ 0.6, 1.3, and 1.0 min, respectively) followed by a considerably slower tracer elimination (t 1/2 ¼ 19.5, 17.0, and 20.7 min). At 60 min p.i., for the three radiotracers 16-23% of activity in blood was found in erythrocytes, whereas about 70% of activity in blood was retrieved in blood plasma (Fig. 6B). Precipitation with trichloroacetic acid (TCA) revealed that about 25-29% of activity in blood was bound to plasma proteins. Low binding of [ 18 F] 5a,b and [D 2 , 18 F]5a to erythrocytes is in line with the absence of a sulfonamide group typically interacting with carbonic anhydrase. With regard to metabolization, the formation of free [ 18 F] uoride and more hydrophilic metabolites was observed for all three radiotracers (Fig. 6C). However, radiotracers showed a different number of metabolites and different amounts of intact compound. For all three radiotracers, no intact parent compound was excreted into the urinary bladder or the intestine so that only metabolites and [ 18 F]uoride could be observed there.
For the [ 18 F]uoroethyl derivative [ 18 F]5b, 30% of the parent compound was intact at 60 min p.i (Fig. 6D). showing the presence of three metabolites in blood (S ¼ 43%). For this tracer, excretion into the intestine via four metabolites and into the urinary bladder via two metabolites was observed (ESI,   The lead celecoxib is metabolized by CYP2C9 and to lower extent (<25%) by CYP3A4 to the hydroxyl form, oxidized to the carboxyl derivative by cytosolic alcohol dehydrogenases ADH1 and ADH2, and nally converted to a minor extent to the O-glucuronide by UDP glucuronosyltransferases. 12,51 Also Takashima-Hirano et al. 13,14 showed that isotopically labeled [ 11 C]celecoxib is oxidized via the hydroxymethyl form to the carboxylic acid derivative. While in blood and liver [ 11 C]  celecoxib was found to be mostly intact, the carboxylic acid was the major radioactive component in the bile suggesting dominant excretion of this radiometabolite into the intestine. Similarly, for [ 18 F]5a,b and [D 2 , 18 F]5a only more polar radiometabolites were found to be excreted into the urinary bladder and the intestine while both the intact compound and radiometabolites were present in blood plasma. The deuterated analog [D 2 , 18 F]5a showed faster blood clearance compared to [ 18 F]5a (Fig. 6A) presumably caused by both tissue tracer distribution of intact compound, e.g., to white adipose tissue (Fig. 4) as well as a metabolic shi towards the more extensive oxidation to the carboxylic acid and excretion in this form. Of note, a fast metabolization of [ 11 C]celecoxib was also observed in male baboon, i.e. only 17% of intact radiotracer was found at 90 min p.i. in blood. 18

Metabolic stability in vitro
Murine liver microsome assay was performed to investigate the time-dependent formation of 18 F-bearing metabolites and their structure in more detail. Different liver subcellular fractions such as liver microsomes, liver S9 fractions, and liver cytosol are principally available differing in their capability to catalyze phase I and phase II reactions. We decided to use liver microsomes to follow cytochrome P450 (CYP)-mediated phase I oxidation to the hydroxylated metabolites as rate limiting steps in radiotracer metabolization. Moreover, liver microsomes contain avin monooxygenases and uridine glucuronide transferase, which allow further phase I and II transformations. Enzymes like aldehyde oxidase, glutathione transferase, sulfotransferase, or other cytosolic cofactors are not present in this subcellular fraction. 60 For the experiment, [ 18 F]5a, [D 2 , 18 F]5a, and [ 18 F]5b were prepared at the same day to ensure comparable activity of the liver microsomes. All three radiotracers were subjected to oxidative conditions in the murine liver microsome assay for 10, 30, 60, and 120 min and samples were analyzed by radio-HPLC and radio-TLC (Fig. 7). Furthermore, carrier-added samples were subjected to the same conditions and analyzed aer radioactive decay by UPLC-MS/MS to identify the structure of the metabolites. While for [ 18 F]5a and [D 2 , 18 F]5a the formation of one major metabolite was observed, two metabolites were formed from [ 18 F]5b ( Fig. 7A and ESI, Fig. S13-S24 †). Of note, glucuronidation was exemplarily investigated for [ 18 F]5b by applying oxidative conditions as described in the experimental section with addition of MgCl 2 (5 mM), alamethicin (50 mg mL À1 ), and uridine diphosphate glucuronic acid (UDPGA, 5 mM) but glucuronidation was not observed in this experimental set-up (data not shown). UPLC-MS/MS experiments veried the formation of the respective hydroxylated 18 Fig. 7B and ESI, Fig. S10-S12 †). Hence, deuteration effectively enhanced the metabolic stability as observed in vitro and in vivo.
Our results showed that oxidative metabolism represents a dominant factor determining the metabolic fate of [ 18 F]5a,b and [D 2 , 18 F]5a which is in accordance with previous reports. For example, aer administration of a non-radioactive uoromethyl-substituted celecoxib derivative Uddin et al. 21 also identied the oxidatively deuorinated metabolite bearing a hydroxymethyl group in the inamed footpad. In accordance, [ 11 C]celecoxib is oxidized to the carboxylic acid derivative, 13,14 a fact that also causes fast metabolization of [ 11 C]celecoxib in male baboon 18 which shows the need for metabolic stabilization also in primates. Deuteration as shown for [D 2 , 18 F]5a in this study effectively suppressed metabolization in vitro and in vivo suggesting that this might be a useful strategy for further developments. Interestingly, other studies indicated that the metabolic degradation at the 5-phenyl ring can be lowered as well. For example, a 18 F-labeled celecoxib derivative having a 4-[ 18 F]uoro-5-phenyl-pyrazole 22 as structural key feature but no oxidizable methyl group has been reported to be metabolically stable and was found to be intact in brain, liver, intestine, and blood plasma of healthy mice. Also, only a small amount of metabolites has been observed for the ortho-18 F-celecoxib derivative 23 which has a uorine in ortho-position to the methyl group indicating that steric hindrance or substitution nearby can also lower the oxidative metabolism at this site. For other celecoxib derivatives, only non-specic 17,19,24,26,28 or no 16,20,25,27,[29][30][31] information about the detailed metabolism is currently available.

Conclusion
In this study, alterations in the substitution pattern of a celecoxib derivative were studied with respect to deuteration and elongation of the side chain in position 3 of the pyrazole ring. The radiotracers [ 18 F]5a,b and [D 2 , 18 F]5a were examined within detailed in vitro and in vivo studies. COX inhibitory potency of [D 2 ]5a was found to be comparable to the potent and selective COX-2 inhibitor 5a, while the ethyl derivative 5b turned out to be slightly less potent. However, all three radiotracers [ 18 F]5a,b and [D 2 , 18 F]5a did not show COX-2 specic binding in vitro so that radiopharmacological evaluation in vivo focused on metabolic stability in healthy rats. Our results demonstrate marked effects of both deuteration and elongation on 18 F-deuorination and formation of 18 F-bearing metabolites in vitro and in vivo (Fig. 8).
In principle, oxidative metabolism and excretion to urinary bladder and intestine were found to be dominant factors determining the metabolic fate of all three investigated radiotracers which is in accordance with other results from the literature. As a main result, deuteration was shown to be most benecial as it effectively decelerates metabolic transformation without compromising the biological activity of the molecule. In this sense, it is part of ongoing research to envisage the additional exchange of the CH 3 with a CD 3 group at the tolyl ring to result in a radiotracer candidate with highest possible metabolic stability for this kind of substance class. Other COX-2targeting radiotracers, e.g. the [ 18 F]uoromethyl-substituted valdecoxib derivative 61 for which the evaluation in monkey was substantially hampered by 18 F-deuorination, might accordingly benet from similar modications.

Conflicts of interest
There are no conicts to declare.