An 18F-Labeled Poly(ADP-ribose) Polymerase Positron Emission Tomography Imaging Agent

Poly(ADP-ribose) polymerase (PARP) is involved in repair of DNA breaks and is over-expressed in a wide variety of tumors, making PARP an attractive biomarker for positron emission tomography (PET) and single photon emission computed tomography imaging. Consequently, over the past decade, there has been a drive to develop nuclear imaging agents targeting PARP. Here, we report the discovery of a PET tracer that is based on the potent PARP inhibitor olaparib (1). Our lead PET tracer candidate, [18F]20, was synthesized and evaluated as a potential PARP PET radiotracer in mice bearing subcutaneous glioblastoma xenografts using ex vivo biodistribution and PET–magnetic resonance imaging techniques. Results showed that [18F]20 could be produced in a good radioactivity yield and exhibited specific PARP binding allowing visualization of tumors over-expressing PARP. [18F]20 is therefore a potential candidate radiotracer for in vivo PARP PET imaging.


■ INTRODUCTION
Poly(ADP-ribose) polymerase-1 (PARP-1) is a nuclear protein that exhibits a broad range of functions and is involved in transcription, mitosis, apoptosis, and DNA damage repair. 1,2 PARP inhibition has been investigated as a therapeutic approach to treat cancers by either synthetic lethality in which tumor cells deficient in a type of DNA repair termed homologous recombination are sensitized to PARP inhibition, or chemoradiosensitization, in which PARP inhibition sensitizes tumor cells to conventional chemo-or radiotherapy. To date, olaparib (Lynparza), niraparib (Zejula), and rucaparib (Rubraca) are the only PARP inhibitors to receive approval for clinical use in the United States or Europe. 3,4 Olaparib (1; Figure 1) was the first agent in its class to receive such approval. In the European Union, it is currently indicated for the treatment of BRCA-mutated (homologous recombination deficient) ovarian, fallopian-tube, and peritoneal cancers, 4 in which it has been shown to increase progression-free 5 and overall 6 survival. In the United States, 1 can also be used for treatment of BRCA-mutated metastatic breast cancer 7 and as a maintenance therapy for patients with platinum-sensitive recurrent epithelial ovarian, fallopian-tube, or primary peritoneal cancer irrespective of BRCA mutations. 8 In both cases, 1 was once again shown to increase progression-free survival. 9,10 Olaparib 1 is also being investigated as a radio-and chemosensitizer for the treatment of solid cancers, including gliomas. However, adding PARP inhibitors to cytotoxic chemotherapy agents has been shown to exacerbate bone marrow toxicity in humans, hindering the establishment of effective PARP inhibitor and chemotherapy dosage regimens with acceptable safety profiles. 11 In the case of brain tumors, matters are further complicated as 1 suffers from poor blood− brain barrier (BBB) permeability, and delivery of the drug to the tumor is reliant on BBB disruption. 12 The degree of BBB disruption in brain tumors is very variable; 13−15 this could affect tumor penetration by 1 and, hence, reduce the clinical effectiveness of PARP inhibitor therapy. Furthermore, in vivo animal studies have revealed that prolonged treatment with 1 can result in increased tumor P-glycoprotein efflux transporter expression and subsequent drug resistance. 16 The above-mentioned issues highlight the challenges that are associated with PARP inhibitor therapy in the context of synthetic lethality as well as chemo-and radiosensitization. Nuclear imaging of an appropriately radiolabeled PARP inhibitor could be used to overcome these challenges. In combination with a suitable blockade study protocol, nuclear imaging could indirectly establish the distribution and retention of PARP inhibitors in tumors and normal tissues and subsequently identify therapeutic dosage regimens for which the combination of PARP inhibitors and cytotoxic agents exerts maximal tumor and minimal bone marrow cytotoxicity. Furthermore, a radiolabeled PARP probe could be used to indirectly ascertain the occupancy, retention, and target engagement of 1 in brain tumor tissue. This type of approach has been successfully applied in preclinical models of small-cell lung 17 and epithelial ovarian cancer. 18 In a clinical setting, this could be used to identify patients that are unlikely to respond to PARP inhibitor therapy due to weak target engagement as a consequence of poor drug tumor uptake or resistance caused by efflux transporter over-expression.
Previously, we reported the synthesis and characterization of a [ 123 I]-labeled compound with potential for single-photon emission computed tomography (SPECT) imaging of PARP. 19 Despite the established nature of the SPECT imaging modality, the use of positron emission tomography (PET) is rapidly expanding and, in many cases, has become the preferred nuclear imaging modality in the clinic. This can be attributed to the superior spatial resolution, 20 quantification, and sensitivity 21 of PET compared to SPECT. It is, therefore, not surprising that many of the current nuclear imaging agents for PARP were designed with PET imaging in mind ( Figure 2). 22,23 Importantly, the rationale for developing PARP PET imaging agents has been recently solidified by [ 18 F]4, which was shown to be capable of noninvasively ascertaining PARP-1 expression in epithelial ovarian tumors in humans. 18,24 Here, we report the synthesis of a small library of fluorinated analogs of the clinical PARP inhibitor 1 with potential for PARP PET imaging and the in vitro characterization of these compounds. The lead analog 20 was synthesized in its radiofluorinated version and evaluated as a potential PARP PET radiotracer in mice bearing subcutaneous glioblastoma xenografts using ex vivo biodistribution and PET−magnetic resonance (MR) imaging techniques.

■ RESULTS AND DISCUSSION
Chemistry and in Vitro Characterization. Due to poor accessibility of the central ring fluorine atom of 1 for radiolabeling, 25 focus was directed at synthesizing analogs of 1 bearing distal fluorinated moieties that were more likely to be amenable to radiofluorination methods. Initially, six fluorinated PARP inhibitors (8−13) containing the characteristic phthalazinone scaffold were synthesized through amide or N-alkyl coupling of commercially available benzoic carboxylic acids or benzyl halides with piperazine 7, the synthesis of which we described previously 19 (Table 1). The structures of compounds 8−13 were confirmed in part by nuclear magnetic resonance (NMR) spectroscopic analysis, which showed that the majority of these exist as a mixture of amide rotamers. The design of these analogs was partly driven by previous reports, 25−27 which have shown that structural modifications can be performed in the cyclopropane bearing region of 1 without having a marked effect on PARP inhibition. To confirm this, cell-free PARP-1 IC 50 assays were performed on compounds 8−13, and the results of these experiments were compared against the cell-free IC 50 of 1 (Table 1). Compared to 1, all compounds showed improved PARP-1 potency except for compound 13, which had an overlapping cell-free IC 50 95% confidence interval. Compounds 8−13 were also evaluated for their lipophilic and plasma protein binding properties, defined by log P oct and percentage plasma protein binding (%PPB) parameters (Table  1). All six analogs exhibited greater log P oct and %PPB values in comparison to 1. This may be attributed to the addition of aromatic and methyl moieties, which have the potential to increase lipophilicity and can, in turn, result in an increased % PPB due to the hydrophobic nature of plasma protein interactions. 28 From the perspective of nuclear imaging, radiotracers with high log P oct (>3.0) 29 and %PPB (>95%) 30 values can be associated with poor passive diffusion across biological membranes and vascular retention, which can in turn result in a poor target to background signal ratio. The physiochemical parameters of 8−12 were all found to be within the optimal range for radiotracer development.

Journal of Medicinal Chemistry
Article However, following analysis of the established in vitro parameters and potential radiochemical accessibility, compound 8 was identified as an initial lead candidate for further advancement in this research program. To establish the potency of 8 against PARP in living cells, cellular IC 50 assays were performed using primary (G7) and secondary (T98G) human glioblastoma cell lines ( Table 2) and previously described methodology. 19 Compound 8 exhibited low nano-molar IC 50 values that were in line with those observed for 1 in both cell lines, suggesting that 8 was able to effectively penetrate cellular membranes and reach PARP localized within cellular nuclei.
To access the [ 18 F]-radiofluorine analog of 8, standard aromatic nucleophilic substitution chemistry was employed using p-nitrobenzamide precursor 14, which was generated through amide coupling of 7 with commercially available 4nitrobenzoic acid (Scheme 1). However, optimization of the radiofluorination step proved challenging (see the Supporting Information), and the maximum radiochemical yield (based on high-performance liquid chromatography [HPLC] analysis of crude product) achieved was only 19% (Scheme 1). It was proposed that the poor yield was a consequence of a lack of activation of the system for aromatic nucleophilic substitution due to the weak electron withdrawing properties of the amide located para-to the nitro leaving group. Since commencing this work, Carney et al. were able to synthesize [ 18 F]8 with an optimized radioactivity yield of 38 ± 2.5% (isolated product)

Journal of Medicinal Chemistry
Article using a multistep early-stage radiofluorination approach that circumvented the aforementioned issue of poor activation. 17 Therefore, our attention was redirected toward an alternative target compound, 20, bearing a p-(fluoromethyl)benzamide group. It was proposed that a precursor analog of 20 would be more amenable to radiofluorination than p-nitrobenzamide 14, thereby allowing for late-stage radiofluorination and radiosynthetic automation. Compound 20 was synthesized by first performing an amide coupling reaction between mono-Boc protected piperazine 15 and commercially available 4-(chloromethyl)benzoic acid, giving access to intermediate 16 in 40% yield (Scheme 2). p-Chlorobenzamide 16 was then subjected to nucleophilic fluorination with tetra-n-butylammo-nium fluoride (TBAF), and this was followed by acid-mediated cleavage of the Boc-protecting group to give 18 in 87% yield over two steps. Finally, reaction of carboxylic acid 19, previously synthesized within our research group, 19 and piperazine 18 under standard amide coupling conditions gave target 20. The log P oct , %PPB, and cell-free and cellular IC 50 values were established for 20 in the same manner as described for the other analogs (8−13) in this series ( Table 3). As expected, compound 20 exhibited higher log P oct and %PPB values than 1 (Table 3). However, these were still within the optimal range for a nuclear imaging agent. Interestingly, despite 20 having an identical log P oct value to 8, the former compound exhibited a higher degree of %PPB. Furthermore,

Scheme 1. Radiofluorination Approach Used to Generate [ 18 F]8 From the Nitro Precursor 14
a Radiochemical yield was determined by radio-HPLC analysis of the crude product.

Journal of Medicinal Chemistry
Article compound 20 had a cell-free IC 50 value that was 6-and 3-fold less than that acquired for 1 and 8, respectively. Conversely, the cellular PARP inhibitory properties of 20 were comparable to compound 1 and marginally weaker than ascertained for compound 8. Collectively, the optimal physiochemical properties and low nanomolar PARP cell-free and cellular IC 50 values supported further investigation of 20 as a potential radiotracer for PARP. In the body, radiotracers can be exposed to a number of metabolic pathways, including blood plasma hydrolysis and liver functionalization or conjugation reactions that can have a significant effect on the kinetic properties of the tracer and, subsequently its usefulness in nuclear imaging. With this in mind, the in vitro plasma and liver microsome stability of lead candidate 20 were established by incubating the compound in mouse plasma proteins and human liver microsome enzymes, respectively, using previously described methodologies. 19 Candidate 20 appeared stable in mouse plasma with negligible decomposition following a 20 h incubation (Table 3). However, the intrinsic clearance parameter (a predictor of phase I liver metabolism) was approximately 3-fold greater for compound 20 in comparison to 1. Despite this, it was proposed that compound 20 would exhibit sufficient tissue retention to allow for nuclear imaging of PARP. This was justified by previously acquired data using a radioiodinated p-iodobenzamide analog that exhibited similar in vitro intrinsic clearance properties to 20 but still displayed a degree of retention in PARP over-expressing tumor tissue. 19 Initial attempts to generate the radiofluorinated version of 20 involved performing nucleophilic substitution reactions between the 18 F − nucleophile and the chloromethyl group of precursor 21, which was obtained by amide coupling of 7 with commercially available 4-(chloromethyl)benzoic acid (Scheme 3). However, these attempts were not successful because competing oligomerization reactions between the chloromethyl group of 21 and the phthalazinone core prevented effective radiofluorination. To overcome this issue, the phthalazinone core of 21 was Boc-protected to give compound 22.
Precursor 22 was then subjected to a screen of radiofluorination conditions as outlined in Table 4. The radio-chemical yield (based on HPLC analysis of the crude product) was 30% when tetra-n-butylammonium hydrogen carbonate (TBAHCO 3 ) was used as a phase-transfer agent, which was markedly higher in comparison to that obtained for Kryptofix (K 222 ) (entries 1 and 2). It has been reported in the literature that the introduction of a sterically hindered protic alcohol can have a beneficial effect on aliphatic nucleophilic radiofluorination reactions. 31,32 With the use of a 2:1 mixture of t-BuOH and MeCN as the reaction solvent, the radiofluoride incorporation increased from 30% to 48% in comparison to the same volume of MeCN alone (entries 1 and 3). A 30 min reaction time was established to be optimal based on lower radiochemical yields observed after a shorter reaction time (entry 4) and the short-lived nature of the 18 F radioisotope (half-life of 109.8 min), which prevented longer reaction periods. Radiofluoride incorporation was further improved by increasing the reaction temperature from 100 to 110°C, which resulted in a 51% radiochemical yield (entry 5). Interestingly, doubling the reaction solvent volume had a negative impact on the degree of radiofluorination (entry 6). Based on these data, the reaction conditions described in entry 5 were deemed optimal.
To access [ 18 F]20, compound [ 18 F]23 was subjected to Boc deprotection, which was achieved in 5 min and with minimal defluorination taking place by using water as an acid−base catalyst (Scheme 4). The use of hydrochloric acid was also investigated as a deprotecting agent, but its use was associated with marked defluorination (see the Supporting Information). The optimized two-step one-pot radiochemical reaction allowed access to [ 18 F]20 in a radioactivity yield (isolated product) of 9 ± 2% (n = 7) and a molar activity of >4.32 ± 1.46 Ci μmol −1 (n = 3). Importantly, the one-pot nature of the reaction opens up the potential for radiosynthetic automation.
In Vivo Characterization. Following successful optimization of the radiochemistry, the behavior of [ 18 F]20 was investigated in vivo in mice bearing subcutaneous U87MG-Luc2 human glioblastoma tumor xenografts using ex vivo biodistribution and PET−MR imaging techniques. Ex vivo biodistribution of [ 18 F]20 was established at 30, 60, and 120 min after intravenous radiotracer administration, and PET data were acquired by performing a 45 min dynamic scan. These experiments showed that a large proportion of radioactivity was detectable in the liver and small bowel at 30−45 min post tracer administration (Figure 3a,b) and mostly concentrated in the cecum matter and solid feces after 120 min (Figure 3a). This is in line with our previous findings 19 and other literature reports, 25,33,34 which showed in vivo hepatobiliary clearance of a range of related radioiodinated and radiofluorinated compounds based on the structure of 1. Interestingly, the mean percentage of injected dose per gram (%ID/g) of femur tissue remained relatively high across all three biodistribution time points (i.e., >8.5%; Figure 3a), which was also confirmed by PET imaging, in which high skeletal uptake of radioactivity was visible (Figure 3b). This is in contrast to observations made by Carney et al., who reported <2%ID/g of bone of [ 18 F]8 at 120 min. 34 The high bone uptake seen with [ 18 F]20 could be explained by in vivo defluorination and subsequent radiofluoride accumulation in bone tissue. 35 Despite this, apparent radiotracer tumor uptake was identified in both biodistribution and PET−MR imaging experiments (Figures 3a,c), whereas in the case of the former, the mean ratio of %ID/g of glioblastoma tumor to muscle increased from 1.9 ± 0.5 (n = 4) to 3.6 ± 0.5 (n = 4) between the 30 and 120 min time points. This was Scheme 3. Synthetic Route Used to Generate p-Chloromethylbenzamide Precursor 22

Journal of Medicinal Chemistry
Article suggestive of [ 18 F]20 retention in U87Mg-Luc2 glioblastoma tissue, which we have shown to be highly proliferative (see the Supporting Information) and to over-express PARP relative to muscle tissue (Figure 3d).
To establish the specificity of [ 18 F]20 for PARP in bone and tumor tissue, further biodistribution studies were performed on subcutaneous U87MG-Luc2 glioblastoma bearing mice that had PARP binding sites blocked by pretreatment with nonradioactive compound 1 at a dose of 50 mg/kg 20 min prior to [ 18 F]20 administration. The uptake of [ 18 F]20 in bone tissue remained high 60 min after radioligand administration and was not influenced by the presence of excess nonradioactive 1 (Figure 3e). This supports the earlier proposed possibility of in vivo [ 18 F]20 defluorination, leading to free radiofluoride accumulation in the bone. Despite this, preblockade using 1 resulted in a statistically significant decrease in the ratio of %ID/g of tumor to muscle from 1.79 ± 0.39 (n = 4) to 1.07 ± 0.15 (n = 4) 60 min after [ 18 F]20 administration in comparison to vehicle-pretreated mice (unpaired t-test: P < 0.05). These findings suggested that the uptake of [ 18 F]20 in tumor was due to specific PARP binding, which was in line with our previous findings for a related radioiodinated analog of 1. 19

■ CONCLUSIONS
We described here the identification and one-pot manual radiosynthesis of a PARP PET imaging agent, [ 18 F]20, that was accessed in good radioactivity yield and showed desirable physiochemical properties but suffered from rapid hepatobiliary clearance and in vivo defluorination followed by nonspecific 18 F − bone tissue uptake in mice. Despite this, the radiotracer compound exhibited specific PARP binding and subcutaneous glioblastoma tumor retention properties, which allowed for preclinical PET visualization of the PARP over-expressing tumor. Therefore, we believe that [ 18 F]20 should remain a potential candidate radiotracer for in vivo PARP PET imaging and that future imaging studies using small-animal intracranial glioblastoma models are required to further assess the utility of this compound.

■ EXPERIMENTAL SECTION
All reagents and starting materials were obtained from commercial sources and used as received. Dry solvents were purified using a solvent purification system, and all reactions were performed under an atmosphere of argon unless stated otherwise. Macherey-Nagel aluminum-backed plates precoated with silica gel 60F 254 were used for thin layer chromatography and were visualized with a UV lamp. Flash column chromatography was performed using Fisher matrix silica gel 60 (35−70 μm). 1 H NMR spectra were recorded on a Bruker DPX 400 or Bruker 500 spectrometer, and data are reported as follows: chemical shift in parts per million (ppm) relative to Me 4 Si or the solvent (CDCl 3 , δ 7.26 ppm; CD 3 OD, δ 3.31 ppm; or dimethyl sulfoxide [DMSO-δ 6 ] δ, 2.50 ppm) as the internal standard, multiplicity (s, singlet; d, doublet; t, triplet; q, quartet; m, multiplet or overlap of nonequivalent resonances, integration). 13 C NMR a PTA: phase-transfer agent. b Radiochemical yields were determined by radio-HPLC analysis of the crude product from a single reaction.

Scheme 4. Optimized Radiofluorination Approach Used to Generate [ 18 F]20
a Radioactivity yield was determined using the measured radioactivity of the isolated product.

Journal of Medicinal Chemistry
Article spectra were recorded on a Bruker DPX NMR spectrometer at either 101 or 126 MHz, and data are reported as follows: chemical shift (ppm) relative to Me 4 Si or the solvent as internal standard (CDCl 3 , δ 77.2 ppm; CD 3 OD, δ 49.0 ppm; or DMSO-δ 6 , δ 39.5 ppm), multiplicity with respect to proton (deduced from DEPT experiments; C, CH, CH 2 , or CH 3 ). Infrared spectra were recorded using a Shimadzu IRPrestige-21 spectrometer; wavenumbers are indicated with units of cm −1 . Mass spectra were recorded using electron impact, chemical ionization, or electrospray techniques. High-resolution mass spectrometry (HRMS) results were recorded using a JEOL JMS-700 spectrometer. Melting points are uncorrected. All compounds used for biological testing exhibited >95% purity as per HPLC. Purity assessment of compounds 8−13 was performed by injecting 0.5−1.    (10). To a solution of 4-fluoro-2-methylbenzoic acid (0.042 g, 0.27 mmol) in dichloromethane (5 mL) was added EDCI (0.052 g, 0.27 mmol), and the reaction mixture was stirred at ambient temperature for 0.5 h. 4-Dimethylaminopyridine (DMAP; 0.017 g, 0.14 mmol) was then added and the mixture was stirred for a further 0.5 h. A solution of 7 (0.100 g, 0.273 mmol) in dichloromethane (5 mL) was then added drop-wise and the resultant reaction mixture stirred vigorously under reflux for 24 h. Upon cooling to ambient temperature, water (5 mL) was added and the organic layer washed with water (3 × 5 mL) and a NaHCO 3 solution (5 mL), dried with MgSO 4 , filtered, and concentrated in vacuo. Purification using flash column chromatography (MeOH/ CH 2 Cl 2 , 1:49) gave 10 (0.043 g, 31%) as a white solid. NMR spectra showed a 59:41 mixture of rotamer. Only data for the major rotamers  (14). To a solution of 4-nitrobenzoic acid (92.0 mg, 0.550 mmol) in DMF (10 mL), was added Et 3 N (119 μL, 0.880 mmol), followed by HBTU (231 mg, 0.610 mmol) and the mixture was stirred at room temperature for 1 h. Compound 7 (200 mg, 0.550 mmol) was added, and the mixture was stirred at room temperature for a further 48 h. Water (12 mL) was then added, followed by 1 h of stirring, after which the mixture was cooled to 0°C. The resulting precipitate was collected by vacuum filtration, washed with water (4 × 20 mL), and dried in vacuo to yield 14 (239 mg, 84%) as an orange foam. NMR spectra showed a 3:2 mixture of rotamers. Only data for the major rotamer were recorded; IR tert-Butyl Piperazine-4-[[(4′-chloromethyl)benzoyl]-1′carbonyl]-1-carboxylate (16). The reaction was carried out as described for 14 using a solution of 4-(chloromethyl)benzoic acid (183 mg, 1.07 mmol) in DMF (5 mL), Et 3 N (23.1 μL, 1.71 mmol), and HBTU (446 mg, 1.18 mmol). Compound 15 (200 mg, 1.07 mmol) was added, and the mixture was stirred for a further 112 h. After this, water (10 mL) was added and the crude product was extracted into dichloromethane (3 × 10 mL). The organic layers were combined and washed with water (6 × 20 mL), dried with MgSO 4 , filtered, and concentrated in vacuo to give a white solid. Purification by

Journal of Medicinal Chemistry
Article h until pH 7.7 was reached, followed by the removal of excess CO 2 by bubbling argon through the mixture until pH 9.0 was reached. The 0.750 M TBAHCO 3 solution (10 μL) was added to a 2 mL v-vial containing 469−572 MBq of 18 F − /H 2 18 O (0.16 mL). The mixture was vortexed, and the[ 18 F]fluoride was dried by passing a constant stream of argon over the solution at 100°C for approximately 20 min; anhydrous MeCN was added in three aliquots of 0.50 mL to facilitate azeotropic drying. A solution of 22 (4.9 mg, 0.0079 mmol) in anhydrous MeCN (0.1 mL) and tBuOH (0.2 mL) was then added to the v-vial, and the reaction was allowed to proceed for 30 min at 110°C . After this, distilled water was added (0.45 mL), and the reaction was heated at 120°C for a further 5 min. Next, the crude reaction mixture was cooled to room temperature, purified using HPLC, and concentrated in vacuo in an evaporator flask. The flask was rinsed with MeCN (3 × 0.3 mL) to extract the radiolabeled compound, and the solution was transferred to a 2 mL v-vial; the solvent was removed by passing a constant stream of argon over the solution at 100°C for approximately 15 min. The radiotracer was reconstituted in up to 0.65 mL of 5% v/v DMSO in 0.9% w/v saline. The total radiosynthetic time was 138 ± 16 min (n = 6).
HPLC Analysis and Purification of Radiotracers. All radiochemical yields were determined by analytical radio-HPLC of the crude product. Radioactivity yield was determined using the measured radioactivity of the isolated product. Analytical and preparatory Dionex UltiMate 300 series HPLC systems with Phenomenex Synergi 4 μm Hydro-RP 80 Å 150 mm × 4.60 mm and 150 mm × 10 mm columns, respectively, were used. Column temperatures were set to 25°C, and mobile-phase flow rates were 1 and 3 mL/min for the analytical and preparatory systems, respectively. Analytical UV detection was performed using a UltiMate diode array detector (190−800 nm), and radiodetection was achieved using a Berthold Technologies Flow Star LB513 detector. Preparatory UV detection was performed using a Knauer Advanced Scientific Instruments Smartline UV Detector 2500, and radiodetection was achieved using a pin-diode connected to a Lab Logic Flow-Count radiodetector. Analysis of [ 18 23 were compared with the retention times obtained for nonradioactive 8, 20, and 23 using the same chromatographic conditions. Because the amount of [ 18 F]20 that was produced fell below the sensitivity threshold of the UV detector (<1.68 × 10 −4 μmol), molar activity (Ci/μmol) was calculated using the lowest detectable amount of nonlabeled 20 established from a calibration plot of a range of concentrations (0.001−1.000 mg/mL). All HPLC data acquisition and analyses were carried out using the Chromeleon 6.8 Chromatography software.
Mouse Glioblastoma Model. U87MG-Luc2 glioblastoma cell culture and implantation procedures were performed as reported previously. 19 Briefly, U87MG-Luc2 cells were purchased commercially from ATCC and cultured in minimum essential media that has been supplemented with 10% w/v fetal calf serum and 2 mM L-glutamine at 37°C and 5% v/v CO 2 . Unconscious female CD1 nude mice (6−11 week old; purchased from Charles River Laboratories) had 5 × 10 6 U87MG-Luc2 cells injected subcutaneously into the right flank. The resulting tumor xenografts were measured and monitored visually every 3 days, and tumor-bearing animals were used for in vivo studies 28−30 days post-implantation.

Ex Vivo Biodistribution with and without Preblockade.
Subcutaneous tumor bearing mice were administered 1.8−2.6 MBq of [ 18 F]20 in 0.11−0.21 mL of 5% v/v DMSO in 0.9% saline via bolus tail-vein injections. The remainder of the experiment was conducted as described previously. 19 PET−MR Imaging. A single subcutaneous U87MG-Luc2 tumor bearing female CD1 nude mouse was anaesthetized using inhaled isoflurane (in medical air; induction 5% v/v; maintenance 2.0−2.5% v/ v) and placed in the nanoScan PET−MRI scanner (Mediso Medical Imaging Systems). The mouse received a bolus tail-vein injection of 2.2 MBq of [ 18 F]20 in 5% v/v DMSO in 0.9% saline, and a 45 min dynamic PET scan (continuous list mode) was initiated immediately after. This was followed by a whole-body 1T GRE 3D Cor/Sag MRI scan. The dynamic PET data were corrected for random coincidences, dead time, scatter, and decay and subsequently reconstructed using 3D Tera-Tomo (Mediso). The reconstructed PET data were automatically co-registered with the MRI data, and they were subsequently analyzed using the PMOD 3.504 software. The PET frames from the last 15 min of the scan were summed to allow for better visual representation of radiotracer biodistribution. Standardized uptake values (SUV) were determined by dividing the image radiotracer concentration by the injected dose divided by the animal weight.
Immunohistochemistry. Tumor and muscle tissue was prepared and stained using anti-PARP-1 antibody (mouse antihuman and mouse antibody; sc-8007; Santa Cruz), as reported previously. 19 Haematoxylin and eosin and Ki67 staining were performed using a Leica ST5020 multi-stainer. Histology images were acquired using a Zeiss AX10 brightfield microscope at a 5× magnification.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jmedchem.8b00138.
Molecular formula strings for compounds presented in the manuscript. (CSV) 1 H and 13 C NMR spectra for all new compounds, cellfree and cellular IC 50 curves, metabolic stability graphs, radiofluorination optimization, optimization of deprotection chemistry, immunohistochemical staining of the U87MG-Luc2 subcutaneous xenograft, radiochemistry HPLC chromatograms, and purity assessment HPLC chromatograms for key compounds. (PDF)