One-Step 18F-Labeling and Preclinical Evaluation of Prostate-Specific Membrane Antigen Trifluoroborate Probes for Cancer Imaging

After the identification of the high-affinity glutamate-ureido scaffold, the design of several potent 18F- and 68Ga-labeled tracers has allowed spectacular progress in imaging recurrent prostate cancer by targeting the prostate-specific membrane antigen (PSMA). We evaluated a series of PSMA-targeting probes that are 18F-labeled in a single step for PET imaging of prostate cancer. Methods: We prepared 8 trifluoroborate constructs for prostate cancer imaging, to study the influence of the linker and the trifluoroborate prosthetic on pharmacokinetics and image quality. After 1-step labeling by 19F–18F isotopic exchange, the radiotracers were injected in mice bearing LNCaP xenografts, with or without blocking controls, to assess specific uptake. PET/CT images and biodistribution data were acquired at 1 h after injection and compared with 18F-DCFPyL on the same mouse strain and tumor model. Results: All tracers exhibited nanomolar affinities, were labeled in good radiochemical yields at high molar activities, and exhibited high tumor uptake in LNCaP xenografts with clearance from nontarget organs. Most derivatives with a naphthylalanine linker showed significant gastrointestinal excretion. A radiotracer incorporating this linker with a dual trifluoroborate-glutamate labeling moiety showed high tumor uptake, low background activity, and no liver or gastrointestinal track accumulation. Conclusion: PSMA-targeting probes with trifluoroborate prosthetic groups represent promising candidates for prostate cancer imaging because of facile labeling while affording high tumor uptake values and contrast ratios that are similar to those obtained with 18F-DCFPyL.

The prostate-specific membrane antigen (PSMA), a transmembrane metalloenzyme (1), is highly overexpressed in prostate cancer and tumor-associated neovasculature (2). PSMA-targeting constructs have been designed and evaluated as imaging agents for visualizing prostate cancer, most notably by PET (3)(4)(5)(6). The diamino acid glutamate-ureido is commonly used for PSMA targeting because of synthetic ease, rapid pharmacokinetics, and high contrast ratios (7). 68 Ga-PSMA-11 is currently the most commonly used radioligand for prostate cancer imaging (8,9). The short half-life of 68 Ga (68 min) generally restricts distribution to clinics that are close to a 68 Ge-68 Ga generator, which itself limits daily production to 2-4 clinical doses unless direct production using a more complex solid-target apparatus is implemented (10). In contrast, 18 F has several advantages, including a longer half-life (109.8 min); higher spatial resolution than 68 Ga due to its short positron range; and on-demand, scalable production of 18 F-fluoride ions up to a few hundred gigabecquerels (11).
We report the synthesis, radiolabeling, and PET imaging of radiotracers based on the glutamate-ureido-lysine scaffold bearing RBF 3 2 radioprosthetic groups (compounds 1-8, Fig. 1). We measured their binding affinity toward PSMA and LogD 7.4 values and then acquired PET images and ex vivo biodistribution data in mice bearing PSMA-expressing LNCaP prostate cancer xenografts. These results were compared with those of 18 F-DCFPyL, a clinically emergent 18 F-labeled tracer for prostate cancer imaging.

Synthesis of Trifluoroborate Probes and Radiosynthesis
18 F-DCFPyL was prepared following literature procedures (24). Precursors for tracers 1-8 were synthesized as described in the supplemental data section (available at http://jnm.snmjournals.org) to give azide-bearing precursors (6,18,21,(24)(25)(26)(27)(28), which were conjugated to previously reported alkyne-bearing RBF 3 2 (19). After conjugation, the final trifluoroborate conjugate was purified by high-performance liquid chromatography (HPLC), and purity was confirmed by electrospray ionization-mass spectrometry. Representative crude and quality control HPLC traces are provided in the supplemental data section. 18 F-1-8 were labeled via previously reported procedures (15,29). Briefly, 30-40 GBq of nocarrier-added 18 F-fluoride were trapped on a QMA light cartridge and eluted with 0.9% saline or phosphate-buffered saline (typically 100 mL) directly into a septum-sealed falcon tube containing 80-100 nmol of precursors 1-8 dissolved in 50:50 dimethylformamide:water containing 1 M pyridazinium-HCl buffer (pH 2.5). The reaction was heated to 80°C, and a vacuum was applied to reduce the reaction volume. After 15-20 min, the reaction was quenched by addition of 2 mL of 40 mM ammonium formate or phosphate-buffered saline, and the contents were purified by semipreparative HPLC. Radiochemical purity was confirmed by HPLC analysis using an analytic RP-C18 column with gradients of acetonitrile and water (both containing 0.1% trifluoroacetic acid). Measurements of molar activity values were based on standard curve analysis.

In Vitro Competition Binding Assay
Inhibition constants (K i ) of 1-8 and DCFPyL to PSMA were measured by in vitro competition binding assays using 18 F-DCFPyL as the radioligand. LNCaP cells were plated onto a 24-well poly-D-lysine coated plate for 48 h (400,000/well). Growth medium was removed and replaced with 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffered saline (50 mM HEPES, pH 7.5, 0.9% sodium chloride) After 1 h, 18 F-DCFPyL (0.1 nM) was added to each well (in triplicate) containing varied concentrations (0.5 mM-0.05 nM) of tested compounds (DCFPyL, 1-8). Nonspecific binding was determined in the presence of 10 mM unlabeled DCFPyL. The assay mixtures were incubated for 1 h at 37°C with gentle agitation followed by 2 washes with cold HEPES buffered saline. A trypsin solution (0.25%, 400 mL) was then added to each well to harvest the cells. Radioactivity was measured by g-counting, and K i values were calculated using the ''1 site-fit K i '' built-in model in Prism 7 (GraphPad). The dissociation constant value for 18 F-DCFPyL, used for K i determination, was 0.49 nM, as previously measured by saturation assays using LNCaP cells (30).

Distribution Coefficient (LogD 7.4 ) Measurements
LogD 7.4 values were measured using the shake flask method. Briefly, an aliquot of 18 F-labeled tracer was added to a vial containing 2.5 mL of n-octanol and 2.5 mL of phosphate buffer (0.1 M, pH 7.4). The mixture was vortexed for 2 min and then centrifuged at 3,000g for 10 min. A sample of the n-octanol (0.1 mL) and buffer (0.1 mL) layers was counted using a g-counter. Values of LogD 7.4 were calculated using the following equation: LogD 7.4 5 log 10 [(counts in n-octanol phase)/(counts in buffer phase)].

PET/CT Imaging and Biodistribution Studies
Imaging and biodistribution experiments were performed using NODSCID IL2RgKO male mice. All experiments were conducted according to the guidelines established by the Canadian Council on Animal Care and approved by the Animal Ethics Committee of the University of British Columbia. Mice were anesthetized by inhalation with 2% isoflurane in oxygen and implanted subcutaneously with 1 · 10 7 LNCaP cells behind the left shoulder. The mice were imaged or used in biodistribution studies once the tumor reached 5-8 mm in diameter (5-6 wk).
PET imaging experiments were conducted using an Inveon preclinical PET/CT scanner (Siemens). Compounds 18 F-1,2,3,5,7, and 8 were formulated in 10% ethanol/normal saline, whereas 18 F-4 and 6 were formulated in 10% ethanol/phosphate-buffered saline. Each tumor-bearing mouse was injected with 6-8 MBq of 18 F-1-8 or 18 F-DCFPyL through the tail vein under sedation (2% isoflurane in oxygen). For blocking controls, the mice were coinjected with DCFPyL (0.5 mg). After injection, the mice were allowed to recover and roam freely in their cage. After 50 min, the mice were sedated by 2% isoflurane inhalation and positioned in the scanner. A CT scan was performed first for localization and attenuation correction. This was followed by a 10-min static PET scan. The mice were kept warm by a heating pad during image acquisition. PET images were reconstructed using the IAW software (Siemens), using 2 iterations of the orderedsubset expectation maximization algorithm followed by 18 iterations of the maximum a posteriori algorithm.
For biodistribution and blocking studies, the mice were injected with 1-3 MBq of radiotracer. At 60 min, the mice were anesthetized with 2% isoflurane inhalation and euthanized by CO 2 inhalation. Blood was withdrawn by cardiac puncture, and the organs and tissues of interest were collected, weighed, and counted using an automatic g-counter (PerkinElmer). Uptake values were expressed as the percentage of the injected dose per gram of tissue (%ID/g).

Statistical Analysis
A standard 1-way ANOVA was performed to determine whether statistically significant differences in tumor uptake occurred between radiotracers. Each radiotracer was compared with 18 F-DCFPyL using the Dunnett test (a many-to-one t test comparison). This analysis was also performed for kidney and blood activity and for tumor-to-blood and tumor-to-muscle ratios. Reported P values were adjusted for multiple comparisons. The analysis was performed using Prism 8 (GraphPad).

Binding Assays
We determined K i via in vitro competition binding assays using LNCaP cells and 18 F-DCFPyL as the radioligand (Fig. 2A). The K i value for DCFPyL was 2.0 6 0.8 nM, consistent with the value previously reported by Chen et al. (1.1 6 0.1 nM) (31). Probes 1-4 and 7 had K i values in the 10-30 nM range, whereas 5 and 6 had up to 10-fold better affinities, comparable to that measured for DCFPyL. Probe 8 showed excellent binding affinity to PSMA, with a K i value of 0.22 6 0.01 nM (Table 1).

LogD 7.4
All compounds but 6 had LogD 7.4 values similar to 18 F-DCFPyL ( Figure 2B and Table 1). Using pyrBF 3 instead of AMBF 3 as the prosthetic group decreased hydrophilicity in 4 and 6 compared with 3 and 5, respectively. Compound 6 was the most lipophilic compound of the series.

PET/CT Imaging and Biodistribution
Imaging 18 F-DCFPyL confirmed good tumor uptake and fast clearance (31). Similarly, 18 F-1-8 showed significant tumor uptake in LNCaP xenografts, which was blocked by coinjection of unlabeled DCFPyL (Fig. 3), thus confirming the specificity of tumor uptake for PSMA. All images also showed high, specific kidney uptake along with urinary excretion. Bone accumulation was negligible for all radiotracers. The blocking agent caused significantly lower tumor and kidney uptake values for all compounds.
Since the 3 carboxylates of Glu-ureido-Lys are needed for binding to PSMA, we introduced modifications at the lysine side chain (31,33), off of which we introduced several well-established linkers along with a suitable RBF 3 2 . Binding assays confirmed low-nanomolar affinities for compounds 5 and 6, whereas compound 8 had subnanomolar binding affinity. Compounds 1-4 exhibited 10-fold higher affinities than DCFPyL, suggesting that the trifluoroborate prosthetic group may not interact well with the S1 binding pocket in PSMA, which exhibits pronounced affinity for hydrophobic groups (3). Compounds incorporating a naphthylalanine-tranexamic acid motif (5 and 6) exhibited improved binding affinities (K i 5 1.14 nM and 1.90 nM,  respectively) similar to those of DCFPyL (K i 5 2.0 nM) and Ga-PSMA-617 (K i 5 2.3 nM) (25). Interestingly, the tranexamic acid linker appears to contribute significantly to affinity, as its replacement by a polyethylene glycol 2 spacer (compound 7) resulted in a higher K i . The dual glutamate-BF 3 motif, introduced to improve the hydrophilicity of the BF 3 derivatives with a naphthylalaninetranexamic acid linker, unexpectedly improved the binding affinity of compound 8, with a K i approximately an order of magnitude better than DCFPyL.
All the RBF 3 2 -bioconjugates were radiolabeled at activity yields greater than 1.85 GBq at molar activity values of at least 56 GBq/mmol. The pyrBF 3 -modified conjugates showed higher activity yields than those modified with the AMBF 3 , along with higher molar activities, consistent with a report that compared  both prosthetic groups in the context of LLP2A-RBF 3 2 bioconjugates (19), as well as with the established stabilities of various trifluoroborates, as previously reviewed (34). High molar activities were also achieved with compound 8, with a dual glutamate-BF 3 motif. Although imaging and biodistribution studies were performed with HPLC-purified tracers to ensure the highest level of purity, a simple Sep-Pak purification of 18 F-6 (,5 min) afforded good radiochemical purity (95%) (supplemental data). This demonstrates potential for HPLC-free labeling where speed is preferred (overall synthesis time , 30 min).
Although radiochemical yields were lower for certain compounds, these syntheses have not been optimized. Notably, yields were dramatically improved by increasing the amount of precursor: the lowest yield (for tracer 18 F-2) was increased more than 8-fold to 34% when using 10 times more precursor. Consequently, the average molar activity of 18 F-2 decreased by a similar factor from 89 to 13.3 GBq/mmol. This demonstrates that yields dramatically increase when high molar activity is not critically needed.
To evaluate 18 F-1-8 for PSMA imaging, PET/CT imaging and biodistribution studies were conducted in mice bearing LNCaP tumor xenografts. Previously, Chen et al. and Harada et al. imaged 18 F-DCFPyL in different strains of mice bearing different tumor models (31,33), thus complicating a comparison between this work and prior work. Given these discrepancies, we directly compared 18 F-1-8 with 18 F-DCFPyL using a single mouse strain and the LNCaP xenograft tumor model, because it expresses PSMA endogenously and is commonly used to evaluate PSMA-targeting radiotracers (25,33).
Imaging and biodistribution studies showed that 18 F-1-8 and 18 F-DCFPyL were all retained in tumors and cleared from nontarget tissues and organs, mainly through the renal pathway for compounds 18 F-1-4 and 8, and a combination of renal and hepatobiliary clearance for compounds 18 F-5-7 (Fig. 4). Tumor uptake was higher with 18 F-8 than with 18 F-DCFPyL, a result that might be explained by improved affinity. All compounds showed significant renal uptake, which was blocked by DCFPyL, consistent with the well-documented, high PSMA expression in mouse kidneys (25,31,33,(35)(36)(37)(38). As with 18 F-DCFPyL, images acquired with 18 F-1-4 and 8 showed low uptake in nontarget organs, whereas those acquired with 18 F-5-7 showed high accumulation in the gallbladder and intestines. Blocking controls showed that this intestinal uptake was not receptor-mediated. Although it is likely that intestinal uptake is due to the hydrophobic naphthylalanine moiety, this was not noted with 68 Ga-or 177 Lu-labeled PSMA-617 tracers (32). We presume that the DOTA chelator promotes renal clearance.
Because many radiotracers were compared with 18 F-DCFPyL, this study did not have statistical power to evaluate small differences between radiotracers. The results confirmed the versatility of RBF 3 2 prosthetic groups for 18 F radiolabeling, and potential strategies to direct radiotracers to favor hepatobiliary or renal clearance.
Renal clearance can be a drawback for prostate cancer imaging, as focal retention in the ureters may be confused with small nodal metastases, and because high bladder activity may obscure the detection of primary prostate tumors or recurrences. Conversely, excessive bowel activity may also be detrimental for detection of small lesions in the pelvis and abdomen. High liver activity, as observed in clinical studies with 18 F-DCFPyL (12) and 18 F-PSMA-1007 (13), might impair detection of liver tumors, notably for detection of hepatocellular carcinomas, for which PSMA imaging may be of value (39).
Other 18 F-labeled PSMA binding radiotracers have recently been reported, notably 18 F-PSMA-1007 (13,40), among others (41)(42)(43)(44). The RBF 3 2 radiotracers presented in this article were not directly compared with these compounds. With an excellent binding affinity, high tumor accumulation, and no liver or gastrointestinal excretion, 18 F-8 represents an attractive radiopharmaceutical for clinical translation. CONCLUSION We report promising alternatives to current 18 F-and 68 Ga-labeled PSMA-targeting agents, as the RBF 3 2 prosthetic groups enable a facile, 1-step 18 F-labeling in aqueous medium. Labeling times could be further reduced to 30 min with a simple Sep-Pak purification. The 1-step labeling by isotope exchange provided for the simple production of a precursor that is chemically identical to the radiolabeled product, simplifying aspects of both production and labeling. These radiotracers were designed to explore the influence of both the spacer and the trifluoroborate prosthetic group. Compound 8, with a naphthylalanine-tranexamic acid linker and a dual glutamate-BF 3 moiety designed to enhance hydrophilicity, showed excellent binding affinity and high tumor uptake without liver accumulation or hepatobiliary clearance.

ACKNOWLEDGMENTS
We thank Nadine Colpo and Navjit Hundal-Jabal for their help with the animal studies.