Two experts and a newbie: [18F]PARPi vs [18F]FTT vs [18F]FPyPARP—a comparison of PARP imaging agents

Purpose Imaging of PARP expression has emerged as valuable strategy for prediction of tumor malignancy. While [18F]PARPi and [18F]FTT are already in clinical translation, both suffer from mainly hepatobiliary clearance hampering their use for detection of abdominal lesions, e.g., liver metastases. Our novel radiotracer [18F]FPyPARP aims to bridge this gap with a higher renal clearance and an easily translatable synthesis route for potential clinical application. Methods We developed a less lipophilic variant of [18F]PARPi by exchange of the fluorobenzoyl residue with a fluoronicotinoyl group and automated the radiosyntheses of the three radiotracers. We then conducted a comparative side-by-side study of [18F]PARPi, [18F]FPyPARP, and [18F]FTT in NOD.CB17-Prkdcscid/J mice bearing HCC1937 xenografts to assess xenograft uptake and pharmacokinetics focusing on excretion pathways. Results Together with decent uptake of all three radiotracers in the xenografts (tumor-to-blood ratios 3.41 ± 0.83, 3.99 ± 0.99, and 2.46 ± 0.35, respectively, for [18F]PARPi, [18F]FPyPARP, and [18F]FTT), a partial shift from hepatobiliary to renal clearance of [18F]FPyPARP was observed, whereas [18F]PARPi and [18F]FTT show almost exclusive hepatobiliary clearance. Conclusion These findings imply that [18F]FPyPARP is an alternative to [18F]PARPi and [18F]FTT for PET imaging of PARP enzymes. Supplementary Information The online version contains supplementary material available at 10.1007/s00259-021-05436-7.


Introduction
Critical for DNA repair [1], inhibition of poly(ADP ribose) polymerase (PARP) enzymes can be lethal for homologous recombination deficient (e.g., BRCA1/2 mutated) tumors due to the resulting inability to repair DNA single-strand breaks [2,3]. The clinical potential of such PARP inhibitors is reflected by the efforts taken to develop different drugs targeting the PARP enzymes: since the first milestone, the clinical approval of olaparib [4], other inhibitors followed this path with the prominent candidates niraparib, veliparib, and rucaparib [5], and optimized next generation inhibitors (e.g., talazoparib/BMN 673) demonstrate even higher efficacy [6]. Competing with the cofactor NAD + to interact with the enzyme's binding site [7], PARP inhibitors both inhibit catalytical poly-ADP-ribosylation (PARylation) of target proteins but at the same time can hamper the function of the replication fork by trapping the enzyme on the DNA [8] in a not yet fully understood process, ultimately leading to enhanced susceptibility of the tumor for DNA-damaging agents [9][10][11]. Mainly applied as maintenance therapy after regular chemotherapy for recurrent cisplatin-sensitive BRCA-mutated ovarian cancer types [12,13], the potential for the use of PARP inhibitors in non-BRCA-deficient tumors is evident as well [14].
Besides the therapeutic relevance of PARP inhibitors, non-invasive imaging of PARP with the aid of positron emission tomography (PET) possesses the potential to predict tumor malignancy since elevated PARP1 expression is associated with poor prognosis and lower survival rates in breast cancer and leukemia [15,16]. Enhanced PARP expression levels are found in different tumor entities pointing towards a broad application spectrum of PARP imaging not only limited to BRCA-deficient tumors [17].
Within the last decade, several inhibitors of the enzyme were radiolabeled and preclinically evaluated for nuclear imaging [18][19][20][21] but two outstanding PARP imaging agents evolved to gold standards of PARP radiotracer development: [22] and [ 18 F]FTT [23,24]. Both radiotracers successfully reflect PARP1/2 expression levels and detect target engagement of PARP inhibitors [25][26][27]. The fluorescent variant PARPi-FL shows promising clinical data for its use to detect early oral squamous cell carcinomas [28,29] and was already radiolabeled for use as dual-modality PET/ optical imaging probe [30]. While the benefit of the radiotracers was assessed independently from each other, there is no direct comparison of the pharmacological properties in the same animal model available. Both radiotracers entered ([ 18 F]FTT) [31] or successfully completed ([ 18 F]PARPi) [32] clinical phase I trials; however, they suffer from mainly hepatobiliary clearance hampering their use for abdominal lesions such as liver metastases.
To expand the clinical scope, we present the alternative PARP imaging agent [ 18 F]FPyPARP with a reduced logP value that is expected to shift the excretion route towards renal clearance. This was based on the empirical observations that the clearance pathway of a drug is often rather well predicted by only four physicochemical parameters including lipophilicity [33]. The new imaging agent [ 18 F] FPyPARP indicates promise for improved non-invasive imaging of abdominal lesions. Furthermore, we compared the state-of-the-art [ 18 F]PARPi and [ 18 F]FTT to our novel radiotracer in the same animal model side-by-side, shedding light on clearance pathways and tumor uptake differences.

Organic chemistry
All reagents and solvents were purchased from commercial suppliers and used without further purification if not stated otherwise. High-performance liquid chromatography (HPLC) columns were purchased from Phenomenex (Aschaffenburg, Germany). Electrospray ionization mass spectrometry (ESI-MS) was performed on a 1200 series HPLC system (Agilent, Waldbronn, Germany) equipped with a 6120 quadrupole mass spectrometer. A gradient of acetonitrile (MeCN) in 0.1% aqueous formic acid on a Luna C18(2) column (50 mm × 2 mm, 100 Å, 5 μm) was used for separation. Nuclear magnetic resonance (NMR) spectra were acquired using a 600-MHz Avance III spectrometer (Bruker Biospin, Ettlingen, Germany).

Synthesis of nonradioactive FTT
4-(2-Fluoroethoxy)benzoyl chloride was produced by activating 0.25 g of 4-(2-fluoroethoxy)benzoic acid (1 eq, 1.36 mmol) with 1 ml thionyl chloride (10 eq, 13.7 mmol) for 3 h at 85 °C. The vacuum-dried compound was added without further purification to 9-amino-1,2,3,4,-tetrahydro-5H-1,4,-benzodiazepin-5-one (1 eq, 0.057 g) dissolved in 5 ml DCM and 5 ml pyridine and stirred overnight. The solvents were evaporated and the residue was dissolved in 50 ml methanol. One milliliter methanesulfonic acid was used to induce cyclization for 2 h under reflux at 75 °C. The solvent was again evaporated, dissolved in 75 ml ethyl acetate, and washed with 50 ml of each saturated Na 2 CO 3 , water, and brine. After drying with MgSO 4 , the solvent was evaporated and a portion of the product was purified by flash column chromatography using 15% methanol in ethyl acetate (yield: 15%). 1   were added to the reaction vial. After cooling to 50 °C and addition of 1 ml 5% acetic acid to prevent hydrolysis, the reaction was diluted in 25 ml H 2 O and trapped on a Sep-Pak Plus Light C18 cartridge (Waters) preconditioned with 10 ml ethanol and 10 ml H 2 O. The synthon [ 18 F]SFB was eluted back in the reactor vial with 0.5 ml DMF containing 10 mg 4-(4-fluoro-3-(piperazine-1-carbonyl)benzyl) phthalazin-1(2H)-one (AB478852, abcr) and 40 µl DIPEA and heated for 10 min to 120 °C. Two milliliter 0.1% TFA containing 30% MeCN was added and the reaction mixture was purified on a Luna C18(2) column (250 mm × 10 mm, 100 Å, 10 µm) with 30% aqueous MeCN containing 0.1% TFA and a flow of 7 ml/min. The product peak (retention time 15-20 min, detected using an online radioactivity detector) was collected, diluted in 50 ml H 2 O, and trapped on a Sep-Pak Plus Light C18 cartridge (Waters) preconditioned with 10 ml ethanol and 10 ml H 2 O. The product was eluted into the product vial with 0.5 ml ethanol, followed by 4.5 ml phosphate-buffered saline (PBS). Quality control was performed on a Luna C18(2) column (250 mm × 4.6 mm, 100 Å, 5 µm) with 55% 0.1% aqueous TFA and 45% MeCN with a flow rate of 1 ml/min on a 1260 Infinity II HPLC system (Agilent) with radioactivity detector. Data on starting activity, radiochemical yield, and molar activity of the individual syntheses are provided in Supplementary  FPyPARP precursor in 500 µl tert-butanol (t-BuOH) and MeCN (8:2) was added to the reactor for 10 min at 40 °C, followed by the addition of 4-(4-fluoro-3-(piperazine-1carbonyl)benzyl)phthalazin-1(2H)-one (AB478852, abcr) in 500 µl MeCN containing 40 µl DIPEA and incubation for 10 min at 120 °C. Three milliliter water were added and the reaction mixture was purified on a Luna C18(2) column (250 mm × 10 mm, 100 Å, 10 µm) with 25% aqueous MeCN containing 0.1% TFA and a flow rate of 7 ml/min. The product peak (retention time 15-20 min, detected using an online radioactivity detector) was collected, diluted in 50 ml H 2 O, and trapped on a Sep-Pak Plus Light C18 cartridge (Waters) preconditioned with 10 ml ethanol and 10 ml H 2 O. The product was eluted into the product vial with 0.5 ml ethanol, followed by 4.5 ml PBS. Quality control was performed on a Luna C18(2) column (250 mm × 4.6 mm, 100 Å, 5 µm) with 65% 0.1% aqueous TFA and 35% MeCN with a flow rate of 1 ml/min on a 1260 Infinity II HPLC system (Agilent) with radioactivity detector. Data on starting activity, radiochemical yield, and molar activity of the individual syntheses are provided in Table S1.

[ 18 F]FTT
[ 18 F]Fluoride was trapped on an ion exchange cartridge (Sep-Pak Plus Light QMA Carb, Waters) preconditioned with 10 ml 1 M NaHCO 3 and 10 ml H 2 O and eluted with 2 ml MeCN containing 4% H 2 O, 9.5 mg Kryptofix 2.2.2, and 1.7 mg K 2 CO 3 which was subsequently dried azeotropically under a helium stream. One milligram [ 18 F]FTT precursor in 750 µl DMF was added to the reactor and the reaction mixture was stirred for 10 min at 105 °C. Three milliliter 17% MeCN in 0.1% aqueous TFA was added and the reaction mixture was purified on a Luna C18(2) column (250 mm × 10 mm, 100 Å, 10 µm) with 17% aqueous MeCN containing 0.1% TFA and a flow rate of 5 ml/min. The product peak (retention time 9-10 min, detected using an online radioactivity detector) was collected, diluted in 50 ml H 2 O, and trapped on a Sep-Pak Plus Light C18 cartridge (Waters) preconditioned with 10 ml ethanol and 10 ml H 2 O. The product was eluted into the product vial with 0.5 ml ethanol, followed by 4.5 ml PBS. Quality control was performed on a Luna C18(2) column (250 mm × 4.6 mm, 100 Å, 5 µm) with 0.1% aqueous TFA containing 32% MeCN with a flow rate of 1 ml/min on a 1260 Infinity II HPLC system (Agilent) with radioactivity detector. Data on starting activity, radiochemical yield, and molar activity of the individual syntheses are provided in Table S1.

Serum stability analysis
Serum stability was assessed by mixing [ 18 F]FPyPARP solution 1:1 with C57BL/6 J mouse or human (blood type AB + , Sigma-Aldrich, Steinheim, Germany) serum. After 0-, 30-, 60-, 120-, and 240-min incubation at 37 °C, samples were drawn and the proteins were precipitated by adding ice-cold MeCN to a final concentration of 50%. The supernatant after centrifugation (12,100 × g, 90 s) was analyzed by HPLC as described for radiotracer quality control.

Experimental logP and logD determination
Water (logP), PBS (logD), and 1-octanol were saturated with the respective other phase for 24 h before the experiment. One microliter radiotracer solution (0.4 MBq) was added to a mixture of 500 µl water or PBS and 500 µl 1-octanol. After thorough shaking, the suspension was shortly centrifuged for phase separation, samples were drawn from each phase, and radioactivity concentration was measured in a gamma counter (WIZARD2, PerkinElmer, Waltham, MA, USA).

Cell culture
Human breast carcinoma cells (HCC1937, ACC513) were purchased from the German Collection of Microorganisms and Cell Cultures (DSMZ GmbH, Braunschweig, Germany) and cultured in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 16% fetal calf serum (FCS), 100 U/ml penicillin, and 100 µg/ml streptomycin at 37 °C under 5% CO 2 atmosphere. The absence of mycoplasma infection was confirmed by PCR analysis in monthly intervals.

Western blot
HCC1937 cells were lysed using radioimmunoprecipitation assay (RIPA) buffer (Thermo Scientific, Waltham, MA, USA) containing protease inhibitor (cOmplete Mini, EDTAfree, Roche, Basel, Switzerland) and protein concentration was determined using a commercial bicinchoninic acid (BCA) assay kit according to the manufacturer's instruction (Thermo Scientific). Samples containing 40 µg of protein were boiled in reducing loading buffer and discontinuous Laemmli SDS-PAGE was performed with gels containing 12% polyacrylamide using the Mini-PROTEAN Tetra system (Bio-Rad, Hercules, CA, USA). The proteins were transferred on a polyvinylidene fluoride membrane using the same system and blocked for 1 h with Odyssey blocking buffer (Li-Cor, Lincoln, NE, USA) at room temperature after which the blot was incubated at 4 °C overnight in PBS with primary antibodies (1: mouse anti-PARP-1 IgG, clone C-2-10, BML-SA250-0050, Enzo life sciences, New York, NY, 1:3000; 2: rabbit anti-β-actin IgG, clone 13E5, Cell Signaling Technology, Danvers, MA, USA, 1:3000). The next day, the blot was washed twice with PBS-T, incubated with secondary antibodies (1: goat anti-mouse IgG, IRDye 680 RD, 1:7,000; 2: goat anti-rabbit IgG, IRDye 800 CW, 1:7,000, Li-cor) for 1 h at room temperature, and imaged after washing twice with PBS-T on an Odyssey Sa Infrared Imaging System (Li-Cor). For the acid wash, cells were distributed in 96-well filter plates as described before and incubated with the radiotracer solution for 30 min at 37 °C. After an initial wash with 100 µl medium, cells were either washed twice (1: 100 µl, 2: 200 µl) with medium or with glycine-HCl in PBS (50 mM, pH 2.8) followed by a final wash with 200 µl medium and measured in a gamma counter.

PET/MR imaging and ex vivo biodistribution
All animal experiments were performed according to the German animal welfare act and approved by the local authorities (Regierungspräsidium Tübingen, R3/18). Animals were housed in individually ventilated cages (IVCs, 5 mice per cage) with bedding and enrichment, and food and water was provided ad libitum. Animals were kept under isoflurane anesthetic (1.5% in pure oxygen, 1.5 l/ min) during the experiments. In 1:1 ice-cold Matrigel (Thermo Scientific)/PBS, 1 × 10 7 cells were injected subcutaneously in the right shoulder area of 7-week-old female NOD.CB17-Prkdc scid /J mice (n = 10 per tracer). After the xenografts reached a suitable size (302 ± 152 mm 3  anatomical scans using a 7-T Biospec 70/30 USR (Clinscan, Bruker Biospin MRI GmbH) and a T2-weighted spin echo sequence. Five mice of each dynamically acquired group underwent a second, 10-min static PET scan 2-h post-injection (p.i.). Mice were sacrificed by cervical dislocation, the collected organs were weighed, and the tissue uptake was determined by gamma-counting (WIZARD2). Liver and kidney tissue was frozen in Tissue-Tek (Labtech, East Sussex, Britain) and 20-µm slices were prepared for autoradiography. A storage phosphor screen (Molecular Dynamics, Caesarea, Israel) was exposed to the sections for 18 h and scanned at a resolution of 100 µm/px with a STORM phosphor imager (Molecular Dynamics). PET image reconstruction and correlation with MR images was performed with Inveon Acquisition Workplace and Inveon Research Workplace, respectively, using a user-defined dynamic framing and an ordered subset expectation maximization (OSEM3D) algorithm. Regions of interest (ROIs) were drawn according to the acquired MR images and co-registered with the PET data to obtain time-activity curves (TACs).

Ex vivo immunofluorescence microscopy
Immunofluorescence microscopy was performed by the Department of Dermatology at the University of Tuebingen, Germany. Sections of paraffin-embedded xenografts were blocked with donkey serum for 30 min and incubated with primary antibody overnight (rabbit polyclonal anti-human PARP1 ab74290, Abcam, Cambridge, UK, 1:50). After washing, the sections were incubated for 1 h at room temperature with secondary antibody (Cy3-conjugated donkey anti-rabbit IgG 711-166-152, Dianova, Hamburg, Germany, 1:250). Nuclei were stained with YO-PRO-1 iodide (Thermo Fisher Scientific) for 5 min according to the manufacturer's instructions; the samples were subsequently mounted with Mowiol (Sigma-Aldrich) and imaged on a LSM 800 microscope (Carl Zeiss, Oberkochen, Germany).

Statistical analyses
Statistical analyses are represented as mean values ± standard deviation. Analyses were performed using GraphPad Prism 9 and non-parametric t tests (comparison of two groups) and one-way ANOVA (comparison of more than two groups). Blood half-life was calculated using a twophase decay fit in GraphPad Prism 8. p values < 0.05 were considered statistically significant according to the software (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Vector graphics
All vector graphics in this work were created with Inkscape 0.92.
Serum stability analysis of [ 18 F]FPyPARP was conducted in C57BL/6 J mouse serum and human serum showing no significant radiometabolites and good radiotracer stability over 240 min (Fig. 1D, Supplementary Fig. 4)

In vivo radiotracer comparison
The chosen HCC1937 cell model was already successfully applied in [ 18 F]FTT radiotracer evaluation and thus used in this work as a standard xenograft model. To verify PARP expression of the cell line, HCC1937 cells were analyzed by Western blotting showing a prominent band at the expected size of PARP1 ( Fig. 2A). In vitro tracer uptake experiments with using olaparib as blocking controls were conducted to ensure specificity of the radiotracer uptake in the cell model displaying significant (p < 0.001) and quantitative blocking of the radiotracer uptake (Fig. 2B). No reduction in the radiotracer signal was observed when the cells were washed with an acidic buffer instead of only medium ( Supplementary  Fig. 5).
A standardized study protocol was applied for all in vivo experiments where immunodeficient mice were injected with HCC1937 cells subcutaneously. After the xenografts reached a sufficient size (302 ± 152 mm 3 ), the respective tracer was injected i.v. and dynamic as well as static PET data and anatomical MR images were acquired and biodistribution was analyzed ex vivo using gamma-counting.
Since specificity of the uptake was already demonstrated in the original publications of the gold standard tracers [22,24] and (in this study) in vitro, in vivo blocking controls were not considered necessary. PET images showed high excretion-related abdominal signal throughout all groups but interestingly, for [ 18 F]FTT, no noticeable renal clearance was observed as indicated by the absence of bladder uptake in contrast to [ 18 F]PARPi, where some of the animals showed renal excretion and [ 18 F]FPyPARP with all animals exhibiting strong radioactive signal in the bladder (Fig. 2C, Supplementary Figs. 6 and 7). Radiotracer uptake was found to be heterogeneously distributed within the xenografts which is in line with the general heterogeneity of PARP1 expression in xenografts [35,36] and confirmed by ex vivo immunofluorescence microscopy (Fig. 3C). The ex vivo biodistribution analyses at 1.5 h and 2.5 h were comparable in their absolute organ uptake values within the cohorts and confirmed high abdominal uptake especially in the liver, intestine, and kidney (Fig. 2D, E). The TACs from the dynamically acquired PET data revealed a higher overall xenograft uptake for [ 18 F]FTT whereas [ 18 F]FPyPARP and [ 18 F]PARPi were within the same range. As already indicated by the bladder uptake pointing towards partial renal clearance, the liver TACs revealed lower uptake values in the [ 18 F]FPyPARP cohort in direct comparison to [ 18 F]PARPi at early time points (Fig. 3A, Supplementary  Fig. 8 Fig. 3B). To determine the potential for imaging of liver metastasis, the tumor-to-liver ratios (TLR) were compared. Although the TLR of [ 18 F]FTT was significantly higher at both time points, all TLRs exhibit a value lower than 1 (Supplementary Fig. 9). For comparison of the excretion routes, the liver-to-kidney ratio (LKR) was calculated and found to be significantly lower for the [ 18 Fig. 6B); however since this only represents one time point and we observed inconsistent spacing to the radiography screen in our experimental setup and thus suboptimal resolution, these results should be taken with caution.

Discussion
The current PARP radiotracers are rapidly taken up into the liver and cleared via bile in rodents and are thus suboptimal for imaging of abdominal lesions. In order to identify the PARP radiotracer most suitable for this purpose, we aimed to compare the most relevant tracers in the same animal model. In addition, we synthesized a variant with reduced lipophilicity with the intention to shift the excretion route towards renal clearance. This was based on the observation that the physicochemical parameter "lipophilicity" is apparently a good parameter for many drugs to predict their excretion route [33].
We automated radiosynthesis procedures for the two literature-known probes and our novel PARP tracer in order to compare them side-by-side in regard to clearance route and tumor uptake. While [ 18 F]FTT radiosynthesis only requires one step, [ 18 F]PARPi synthesis developed by Carney et al. [22] involves many additions of low reagent volumes and has not been automated in literature. We instead successfully used our established automated [ 18 F]SFB labeling protocol for synthesis of this compound although during the preparation of the manuscript, a one-pot synthesis of [ 18 F]PARPi has been published by Wilson et al. [37]. To synthesize the novel variant [ 18 F]FPyPARP, we designed a one-pot reaction utilizing the synthon [ 18 F]FPyTFP. This synthesis can be further simplified in future by late-stage labeling of an already conjugated trimethylammonium nicotinate precursor. The radiotracer is highly stable in mouse as well as human serum; however, first-in-human clinical studies of [ 18  Of note, PARP inhibitors need to enter the cells to reach their intranuclear targets. The radiotracer uptake experiments clearly demonstrated high cell-associated uptake in the chosen HCC1937 cells that was blockable to baseline by olaparib, indicating specific interactions of all three tracers with their intranuclear targets. This was further confirmed by the comparison of HCC1937 washed with either medium or an acidic glycine buffer that did not show any reduction in the radiotracer signal that would hint to ionic binding to the cell surface. The mechanisms of cellular uptake of PARP inhibitor across the membrane are currently unknown. Because of the well-known role of membrane transporters in drug uptake [38][39][40], we speculate that the three radiotracers are also substrates of solute carrier (SLC) uptake transporters. An in silico analysis of transporter gene expression in HCC1937 cells identified a number of expressed SLC transporters (Supplementary Table 2). Within the SLC families implicated in drug transport, several SLC transporters might be candidates of PARP inhibitor uptake (e.g., SLC22A5, SLC22A18, SLC29A1, SLC29A2, SLCO3A1, SLCO4A1). In-depth functional characterization of these transporter candidates warrants further investigation, which is beyond the scope of this article. Of interest, ABCB1 (encoding MDR1 P-glycoprotein), which had been identified as efflux transporter of olaparib [41], is not expressed in HCC1937 cells.
The in vivo analyses employed a standardized protocol to ensure comparability within the different animal cohorts and indeed revealed decent xenograft uptake. Despite the lower TMRs of the [ 18 F]FPyPARP cohort caused by higher muscle uptake of this group, the TBRs indicate more effective blood clearance in comparison to [ (Fig. 3A). The preference of either radiotracer is thus dependent on the choice of reference tissue.
All radiotracers showed high abdominal uptake particularly in the liver, spleen, kidneys, and intestines. These findings point towards mainly hepatobiliary clearance of all three radiotracers; however, only for [ 18  A close-up on the xenografts is provided below. As time point, the last 10 min of the dynamic PET scans was chosen (minutes 50-60). Size bars represent 50 mm (whole body) or 10 mm (xenografts) and color-coded intensity bars range from 0 to 7.8 × 10 5 Bq/ml (whole body) or 0 to 4 × 10 5 Bq/ml (xenografts). E Ex vivo biodistribution analysis of the three radiotracers 1.5 h p.i. F Ex vivo biodistribution analysis of the three radiotracers 2.5 h p.i ◂ this group was caused by a lower retention of the signal in the liver compared to kidney tissue. Thus, the reduced logD of our novel tracer resulted in higher renal than hepatobiliary excretion, although only to a minor degree. The recently published [ 18 F]olaparib only shows hepatobiliary clearance although it has an even lower calculated logP value of 2.02 [42] than [ 18 F]FPyPARP or [ 18 F]PARPi. The reported logD value of olaparib, however, is higher than the experimentally determined logD of [ 18 F]FPyPARP (Lynparza monograph, AstraZeneca: 1.49, vs 1.16, respectively), suggesting that the logD is a more accurate parameter than logP for the prediction of clearance routes. This underlines that the physicochemical parameter "lipophilicity" may be a good predictor for the clearance pathways of many drugs, yet it does not account for all the different processes involved in hepatobiliary and renal clearance. Mechanisms such as transporter-mediated cellular uptake and efflux are important determinants of drug distribution and excretion [38][39][40]. The involvement of membrane transporters in the cellular uptake of PARPi radiotracers, beyond the role of ABCB1 in cellular efflux of olaparib [41], is currently unknown and will be studied in further investigations.

Conclusion
We here present the alternative PARP imaging agent [ 18  Side-by-side comparison of the three radiotracers revealed tumor-to-tissue ratios in the same range although minor differences were observed depending on the reference tissue. However, [ 18 F]FPyPARP was the only radiotracer showing a significant decrease in the LKR when comparing early to late time points, which hints towards a lower retention of this molecule in liver tissue compared to the benchmark radiotracers. This is of particular interest since low retention in non-target tissue is beneficial for targeted radiotherapy of PARP overexpressing lesions that bears high potential of a targeted therapy with promising first outcomes [43][44][45].
According to the obtained PET/MR images and the resulting TACs, [ 18 F]FPyPARP excretion is partially renal, demonstrating that small changes in the molecule can have a beneficial influence on the pharmacokinetics without affecting the uptake performance. Our data highlight the advantages of the three different radiotracers: [ 18 F]PARPi exhibits the highest initial TMR, [ 18 F]FPyPARP demonstrated improved clearance from liver tissue and sufficient tumor uptake, and [ 18 F]FTT showed continuously increasing tumor uptake due to the long blood retention time. This is in line with recent findings on the different modes of action of various PARP inhibitors, warranting the exploration of different pharmacophores for imaging to address unmet needs [46].
Since clinical data obtained by the group of Thomas Reiner indicate that [ 18 F]PARPi already has a 30% renal clearance in humans [32], it can be concluded that our radiotracer might have even lower excretion-related abdominal background signal in humans. With these data, [ 18 F]FPy-PARP has emerged as decent radiotracer for PARP expression with the benefit of improved renal clearance.
Funding Open Access funding enabled and organized by Projekt DEAL. This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy-EXC 2180, 390900677 (Gefördert durch die Deutsche Forschungsgemeinschaft (DFG) im Rahmen der Exzellenzstrategie des Bundes und der Länder-EXC 2180, 390900677). MS and ATN are in parts supported by the Robert Bosch Stiftung, Stuttgart, Germany.
Data availability Primary data are available upon reasonable request.

Declarations
Ethics approval All animal studies were approved by local authorities (Regierungspraesidium Tuebingen, R3/18) and performed in accordance with the German Animal Welfare Act.

Conflicts of interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long Fig. 3 Time-activity curves and tumor-to-muscle, tumor-to-blood, and liver-to-kidney ratios in comparison. A Mean TACs of the liver, kidney (separated in medulla and cortex), tumor, and heart of [ 18 F] PARPi (green, n = 7), [ 18 F]FPyPARP (red, n = 7), and [ 18 F]FTT (blue, n = 6) in comparison. B Tumor-to-muscle, tumor-to-blood, and liverto-kidney ratios 1.5 h (lighter colors) and 2.5 h p.i. (darker colors). C PARP1 immunofluorescence microscopy images. PARP1 is displayed in red and nuclei in green and the scale bar represents 20 µm ◂ as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.