2-(3-{1-Carboxy-5-[(6-[¹⁸F]Fluoro-Pyridine-3-Carbonyl)-Amino]-Pentyl}-Ureido)-Pentanedioic Acid, [¹⁸F]DCFPyL, a PSMA-Based PET Imaging Agent for Prostate Cancer

Prostate cancer (PCa) is the second leading cause of death from cancer in men in the United States (1). The vast majority of men dying of PCa succumb to metastatic, castration-resistant disease. Among the reasons to image PCa, including initial staging, therapeutic monitoring, guiding focal therapies, and determining the location of recurrence after prostatectomy, one elusive but important goal is to image with a view to distinguishing indolent from aggressive disease. Although no single marker is capable of providing that distinction, the prostate-specific membrane antigen (PSMA), a type II integral membrane protein overexpressed on prostate tumors, provides a step in that direction. Both disease-free survival and time to prostate-specific antigen (PSA) progression are decreased in patients with elevated levels of PSMA within their tumors (2, 3). PSMA expression has long been associated with androgen-independent disease (4). Recently, Evans and colleagues showed that a positron-emitting version of the anti-PSMA monoclonal antibody (mAb) J591 (5) was able to leverage PSMA expression into a noninvasive biomarker of androgen receptor (AR) signaling (6).

Several modalities have been applied to imaging PCa, but to study PSMA at high sensitivity in vivo others and we have focused on the radionuclide and optical molecular imaging techniques (7–18). PCa has not succumbed as readily to molecular imaging as other solid tumors as it is not easily visualized on positron emission tomography (PET) with [¹⁸F]fluorodeoxyglucose (FDG), the clinical gold standard, because PCa tends to grow slowly and is less metabolically active with respect to glucose transport and consumption. Another difficulty in imaging PCa with FDG, or any other radiopharmaceutical that is excreted through the urine, is the proximity of the prostate to the urinary bladder, which can obscure specific binding to intraprostatic PCa. There are ways around that problem, including rapid scanning soon after voiding (before accumulation of radiotracer within the bladder), catheterization, and application of postprocessing techniques (19). Accordingly, a variety of radiopharmaceutical imaging agents have been developed for PCa, including radiolabeled versions of choline (20, 21), [¹¹C]acetate (22–24), 1-amino-3-[¹⁸F]fluorocyclobutane-1-carboxylic acid ([¹⁸F]FACBC; 25), as well as a variety of radiolabeled antibodies specific for PSMA (6, 26–29), with several beginning to be seen in clinical trials.

We have previously reported the development of N-[N-[(S)-1,3-dicarboxypropyl]carbamoyl]-4-[¹⁸F]fluorobenzyl-L-cysteine, [¹⁸F]DCFBC (9), with which we have initiated a phase 1 trial (30). [¹⁸F]DCFBC showed 8% injected dose per gram (%ID/g) within PSMA+ PIP tumor, achieved at 60 minutes after injection, which decreased to 4.7% at 2 hours postinjection. Nearly 2% ID/g in bone was present at most time points. To improve upon the pharmacokinetics of [¹⁸F]DCFBC, we have now synthesized 2-(3-{1-carboxy-5-[(6-[¹⁸F]fluoro-pyridine-3-carbonyl)-amino]-pentyl}-ureido)-pentanedioic acid, [¹⁸F]DCFPyL ([¹⁸F]3), which uses the lys-C(O)-glu motif and contains a [¹⁸F]fluoropyridyl substituent, in analogy to a radioiodinated PSMA-binding ligand that we previously reported, which showed high uptake within PSMA+ tumor and fast clearance from nontarget tissues (11).

The tosylate salt of 1, previously described by us (10), was reacted with F-Py-TFP (31) to generate fluoropyridyl urea 2. Deprotection afforded DCFPyL (3) in 81% yield (Fig. 1). The ¹⁸F-labeled prosthetic group [¹⁸F]F-Py-TFP was prepared by a literature procedure (31) and used to generate [¹⁸F]DCFPyL ([¹⁸F]3; Fig. 1). The decay-corrected radiochemical yields of [¹⁸F]3 ranged from 36%–53% based on starting [¹⁸F]F⁻ (n = 3) with absolute yields of 7.7–10.4 mCi (285–385 MBq) after HPLC purification. The mean synthesis time was 128 minutes from the time of addition of [¹⁸F]F⁻. Starting from 44–61 mCi (1,628–2,257 MBq) of [¹⁸F]F⁻, the specific radioactivity of [¹⁸F]3 ranged from 340–480 Ci/mmol (12.6–17.8 GBq/μmol).

The K_i value for 3 was determined using a modification of the Amplex Red glutamic acid assay (32). The K_i value for 3 was 1.1 ± 0.1 nmol/L, comparable with that of ZJ-43, which is 1.4 ± 0.2 nmol/L when used as an internal reference.

Compound [¹⁸F]3 was assessed for its ex vivo pharmacokinetics in nonobese diabetic severe-combined immunodeficient (NOD-SCID) mice bearing both PSMA+ PC3-PIP and PSMA− PC3-flu xenografts. Table 1 shows the %ID/g of radiochemical in selected organs. Compound [¹⁸F]3 showed clear PSMA-dependent uptake within PSMA+ PC3 PIP xenografts, reaching a value of 46.7 ± 5.8%ID/g at 30 minutes postinjection (p.i.), which decreased by only about 10% over the ensuing 4 hours. At 60 minutes p.i. the kidney, liver, and spleen displayed the highest uptake. By that time, the urinary bladder also showed relatively high uptake. However, that uptake includes excretion at all time points. Rapid clearance from the kidneys was shown, decreasing from 74.1 ± 6.6%ID/g at 30 minutes to 7.4 ± 0.9%ID/g at 4 hours. The relatively high values noted in kidney are partially due to high expression of PSMA within proximal renal tubules (33, 34). The ratio of uptake within PSMA+ PIP to PSMA− flu tumors ranged from 40:1 to more than 1,000:1 over the 4-hour time period of the study. A possible explanation for that increased tumor uptake of radiochemical over time could be due to ligand-mediated PSMA internalization within tumor cells (35, 36). Less retention in kidney relative to tumor over time could be due to a lower degree of internalization in this (normal) tissue and/or different metabolism of [¹⁸F]3, which does not promote retention of radiochemical in kidney. Relatively low bone uptake (<1% ID/g at all time points) suggests little metabolic defluorination of [¹⁸F]3.

Intense radiochemical uptake was seen only in the kidneys and PSMA+ PC3 PIP tumor after administration of [¹⁸F]3 (Fig. 2). As noted above for the ex vivo study, the intense renal uptake was partially due to specific binding of the radiotracer to proximal renal tubules (33, 34) and to excretion of this hydrophilic compound. By 3.5 hours after njection, only the PSMA+ tumor is visible with no radiochemical background in liver or the gastrointestinal tract to obscure potential metastases.

Table 2 lists source organ time-integrated activity coefficients for [¹⁸F]3. Table 3 lists target organ absorbed doses. The organ with the highest mean absorbed dose per unit administered activity was the urinary bladder wall, 0.15 mGy/MBq, followed by the kidneys at 0.05 mGy/MBq. The absorbed dose to tissues listed in Table 3 that were not assigned a time-integrated activity coefficient reflects cross-fire photon contribution from organs that were assigned a time-integrated activity coefficient and contribution from radioactivity assigned to the remainder of the body. The effective dose based on the ICRP 60 tissue weighting factors was 13.6 μSv/MBq. On the basis of the dosimetry results a maximum of 9 mCi (331 MBq) can be administered without exceeding the 50 mGy critical organ dose limit (urinary bladder wall in this case), for a single administration of radioactive material for research use as specified in Code of Federal Regulations 21, part 361.

A variety of positron-emitting agents–as well as other techniques and modalities–for imaging PCa have been developed and have recently been reviewed elsewhere (37–39). With respect to PSMA specifically, in addition to imaging per se, PSMA affinity agents, such as low molecular weight compounds (40), antibodies (41) and aptamers (42), have been conjugated to and used to target nanoparticles and to deliver shRNA (43) to PSMA+ cells and tissues. Because PSMA is expressed in most solid tumor neovasculature (44–47), it may be used in principle as a general tumor imaging target with one clinical trial showing imaging of nonprostate tumors (29). However, as stated at the outset, the main value of a PSMA-based imaging agent may be in providing an avenue through which to begin determining the aggressiveness of an individual prostate tumor. Furthermore, PSMA may also be used as an indicator of AR signaling within prostate tumors (6, 48, 49), which would provide a particularly important function for a PSMA-based imaging agent as new AR-targeted drugs emerge (50). In this regard, targeting PSMA may provide more information than targeting AR directly with an agent such as [¹⁸F]fluordihydrotestosterone ([¹⁸F]FDHT; ref. 51), as the AR will be occupied in patients treated with AR-based therapeutics. Positron-emitting progestins have been pursued as imaging agents for breast cancer for similar reasons (52), namely, that estrogen receptor, the target for [¹⁸F]fluoroestradiol, is largely occupied in patients undergoing antiestrogen therapy.

Among the positron emitting isotopes that are integrated into tumor-targeting agents of low molecular weight, including ¹¹C, ¹⁸F, ⁶⁴Cu, ⁸⁶Y, ⁸⁹Zr, and ¹²⁴I, ¹⁸F is considered a particularly convenient radionuclide because of its nearly pure positron emission, isosterism with hydrogen, relative ease of incorporation into relevant affinity reagents at high specific radioactivity through ¹⁸F⁻ or a variety of prosthetic groups (as described here), and its relatively long physical half-life (110 min) enabling shipment of ¹⁸F-labeled radiotracers to sites distant from their production. Carbon-11-labeled PET agents are becoming increasingly viewed as being for proof-of-principle, with commercial entities showing interest primarily in those labeled with ¹⁸F, particularly for indications outside of the central nervous system (CNS). The pharmacokinetics of most agents for use outside of the CNS, particularly those of low molecular weight, are more amenable to an ¹⁸F radiolabel than to ¹¹C, which may not have a sufficiently long physical half-life to enable washout from nontarget sites during the imaging study. For those reasons, we have chosen to focus on development of a new ¹⁸F-labeled PSMA imaging agent for PET.

Antibodies (6, 27, 53) and antibody fragments (54), aptamers (55), and low molecular weight PSMA-binding affinity agents (9, 13, 18, 56, 57) have recently been derivatized with positron-emitting isotopes. Those agents will have different indications, as they have widely varying pharmacokinetics, however, each class has shown PSMA-specific binding in preclinical studies. In terms of specific, in vivo target tumor to nontarget tumor uptake ratio, the radiolabeled mAbs ⁶⁴Cu-DOTA-3/A12 (53) and ⁸⁹Zr-DFO-J591 (27) both showed values of approximately 3:1 at 48 hours postinjection. But comparisons are difficult due to the differences in tumor models used and, importantly, the variable degree of PSMA expressed within them. Antibodies may have an advantage over agents of lower molecular weight due to their putative inaccessibility to apically positioned PSMA on nonmalignant cells (27), suggesting enhanced tumor specificity for mAbs. PSMA-directed mAb fragments have not yet shown selective uptake in PSMA-expressing tumors in vivo (54).

Among nonprotein-based PET imaging agents for PSMA, Rockey and colleagues have optimized conditions for conjugating ⁶⁴Cu to a PSMA-targeting aptamer, which has shown PSMA-mediated uptake in PSMA+ 22RV1 prostate tumor cells relative to PSMA- PC3 cells (55). So far low molecular weight imaging agents for PSMA fall into 2 classes, the ureas, such as [¹⁸F]3, and the phosphoramidates (18). Both have a terminal glutamate at the P1′ position, which enables productive binding to PSMA. Both are amenable to modification with bulky substituents that interact with the arginine patch or tunnel region on PSMA (58–60). In addition to ¹⁸F, the urea-based compounds have been functionalized with ¹¹C (56) and ⁶⁸Ga (13) for PET. A phosphoramidate has been radiolabeled with ¹⁸F and tested in vivo (18). As suggested above, it is challenging to compare the pharmacokinetics of these compounds because of the different models used between investigators and even within the same research group, due to the variable expression of PSMA between experiments within what is considered a PSMA+ cell line. For example, the PSMA+ PC3-PIP cell line expresses significantly lower PSMA than does PSMA+ LNCaP (61). However, we prefer the isogenic PSMA+ PIP versus PSMA− flu comparison as the 2 cell lines are phenotypically identical, differing only in PSMA expression. We have also found that the PSMA+ PC3 PIP cells can lose PSMA expression after several passages. With those caveats in mind, [¹⁸F]3 shows 39.4% ID/g in PSMA+ PIP tumor with a PIP:flu uptake ratio of 358 at 2-hour postinjection, while the ¹⁸F-labeled phosphoramidate showed 1.2% ID/g in LNCaP with an LNCaP:PC3 of 3.5 (18). However, renal and liver values for [¹⁸F]3 were higher than for the ¹⁸F-labeled phosphoramidate, at 15.7 versus 2.2% and 2.1 versus 0.2%, respectively. [¹⁸F]3 also compares favorably with the first generation ¹⁸F-labeled urea, [¹⁸F]DCFBC (9), showing an 8-fold higher tumor uptake at 2-hour postinjection. That is important because [¹⁸F]DCFBC has proved capable of delineating metastases from prostate cancer in human subjects in a recent first-in-human trial (30).

Patterned after our previously reported radioiodinated PSMA-binding radiotracer that showed high uptake within PSMA+ tumor and fast clearance from nontarget tissues (11), [¹⁸F]3 showed high tumor and low normal tissue uptake and retention in PSMA+ PC3 PIP prostate tumor xenografts. The pharmacokinetics of [¹⁸F]3 compare favorably with other low molecular weight agents that bind PSMA selectively. The preclinical results obtained with [¹⁸F]3 warrant its further pursuit in a variety of clinical scenarios to help localize PCa.

All reagents and solvents were purchased from either Sigma-Aldrich or Fisher Scientific. The tosylate salt of 1 was prepared according to a reported procedure (10). ¹H NMR spectra were obtained on a Bruker Avance 400 MHz Spectrometer. ESI mass spectra were obtained on a Bruker Esquire 3000 plus system. High-resolution mass spectra (HRMS) were done by the Mass Spectrometry Facility at the University of Notre Dame using ESI by direct infusion on a Bruker micrOTOF-II. High performance liquid chromatography (HPLC) purification of 3 was done on a Waters 625 LC system with a Waters 490E multiwavelength UV/Vis detector.

[¹⁸F]Fluoride was produced by 18 MeV proton bombardment of a high pressure [¹⁸O]H₂O target using a General Electric PETtrace biomedical cyclotron. Solid-phase extraction cartridges (C₁₈ plus, Sep-Pak) were purchased from Waters Associates. Reverse phase radio-HPLC purification of [¹⁸F]3 was done using a Varian Prostar System with a Bioscan Flow Count PMT radioactivity detector (Varian Medical Systems). Radioactivity was measured in a Capintec CRC-10R dose calibrator. The specific radioactivity was calculated as the radioactivity eluting at the retention time of product during the semipreparative HPLC purification divided by the mass corresponding to the area under the curve of the UV absorption.

To a solution of 1 (0.015 g, 0.018 mmol) in CH₂Cl₂ (1 mL) was added triethylamine (0.010 mL, 0.072 mmol), followed by 6-fluoro-nicotinic acid 2,3,5,6-tetrafluoro-phenyl ester (F-Py-TFP; ref. 31; 0.005 g, 0.017 mmol). After stirring for 2 hours at ambient temperature, the solvent was evaporated. The crude material was purified on a silica column using methanol/methylene chloride (5:95) to afford 0.009 g (65%) of compound 2. ¹H NMR (400 MHz, CDC1₃) δ 8.65 (s, 1H), 8.22 (m, 1H), 7.15–7.24 (m, 7H), 6.83–6.97 (m, 7H), 5.35–5.56 (m, 2H), 4.93–5.08 (m, 6H), 4.31–4.35 (m, 2H), 3.76 (m, 9H), 3.31–3.36 (m, 2H), 2.34 (m, 2H), 2.07 (m, 1H), 1.88 (m, 1H), 1.72 (m, 1H), 1.54 (m, 3H), 1.23 (m, 2H). ESI-Mass calcd for C₄₂H₄₈FN₄O₁₁ [M + H]⁺ 803.3 found 802.9.

A solution of TFA in CH₂Cl₂ (1:1, 2 mL) was added to 2 (0.009 g, 0.011 mmol). The mixture was stirred at ambient temperature for 2 hours, then concentrated on a rotary evaporator. The crude material was purified by HPLC [Econosphere C18 10 μ, 250 × 10 mm, H₂O/CH₃CN/TFA (92/8/0.1), 4 mL/min] to afford 0.004 g (0.009 mmol; 81%) of 3. ¹H NMR (400 MHz, D₂O) δ 8.56 (s, 1H), 8.29 (m, 1H), 7.20 (m, 1H), 4.18–4.24 (m, 2H), 3.42 (m, 2H), 2.49 (m, 2H), 2.15 (m, 1H), 1.87–2.00 (m, 2H), 1.64–1.80 (m, 3H), 1.47 (m, 2H). ESI-Mass calcd for C₁₈H₂₄FN₄O₈ [M + H]⁺ 443.2, found 442.9. ESI-HRMS calcd for C₁₈H₂₄FN₄O₈ [M + H]⁺ 443.1573, found 443.1556.

In a vial containing 2 mg (0.002 mmol) of 1 and 0.005 mL of triethylamine was added [¹⁸F]F-Py-TFP (31) in 2 mL CH₂Cl₂. The reaction was heated at 45°C for 20 minutes followed by removal of solvent under a stream of nitrogen, addition of 0.1 mL of 3% anisole/TFA and further heating at 45°C for 10 minutes. The final product was obtained after HPLC purification (Econosphere C18 10 μ, 250 × 10 mm, H₂O/CH₃CN/TFA [90/10/0.1], 4 mL/min) at a retention time of approximately 9.5 minutes, and was neutralized with 1 mol/L NaHCO₃, concentrated under vacuum to dryness, reconstituted in PBS (pH 7.4) and passed through a 0.22 μm syringe filter into an evacuated sterile vial.

Cell lysates of LNCaP cell extracts were incubated with DCFPyL (0.01 nmol/L–100 μmol/L) in the presence of 4 μmol/L NAAG at 37°C for 2 hours. The amount of released glutamate was measured by incubating with a working solution of the Amplex Red glutamic acid kit (Molecular Probes Inc.) at 37°C for 30 minutes. Fluorescence measurements were done with a VICTOR³V multilabel plate reader (Perkin Elmer Inc.), with excitation at 490 nm and emission at 642 nm. Inhibition curves were determined using semi-log plots, and IC₅₀ values were determined at the concentration at which enzyme activity was inhibited by 50%. Assays were done in triplicate with the entire inhibition study being repeated at least once to confirm affinity and mode of inhibition. Enzyme inhibitory constants (K_i values) were generated using the Cheng–Prusoff conversion (62, 63). Data analysis was done using GraphPad Prism version 4.00 for Windows (GraphPad Software).

LNCaP cells used in the PSMA inhibition assay were obtained from American Type Culture Collection (ATCC) and were maintained as per ATCC guidelines. PC3 PIP (PSMA+) and PC3 flu (PSMA−) cell lines were obtained from Dr. Warren Heston (Cleveland Clinic) and were maintained as previously described (9). Cells were grown to 80%–90% confluence in a single passage before trypsinization and formulation in Hank's balanced salt solution (HBSS, Sigma) for implantation into mice. Animal studies were in compliance with guidelines related to the conduct of animal experiments of the local Animal Care and Use Committee. For biodistribution and imaging studies of [¹⁸F]3, male NOD-SCID mice (Johns Hopkins University, in-house colony) were implanted subcutaneously with 1 × 10⁶ PSMA+ PC3 PIP and PSMA− PC3 flu cells behind either shoulder. Mice were imaged or used in biodistribution studies when the tumor xenografts reached 3 to 5 mm in diameter.

PSMA+ PC3 PIP and PSMA− PC3 flu xenograft-bearing SCID mice were injected via the tail vein with 100 μCi (3.7 MBq) of [¹⁸F]3. In each case 4 mice were sacrificed by cervical dislocation at 30, 60, 120, and 240 minutes p.i. The heart, lungs, liver, stomach, pancreas, spleen, kidney, muscle, bone, small and large intestines, urinary bladder, and PC3 PIP and flu tumors were quickly removed. Stomach and other gastrointestinal contents were removed and the urinary bladder was emptied. A 0.1 mL sample of blood was also collected. Each organ was weighed, and the tissue radioactivity was measured with an automated γ counter (1,282 Compugamma CS, Pharmacia/LKBNuclear, Inc.). The %ID/g was calculated by comparison with samples of a standard dilution of the initial dose. All measurements were corrected for decay.

A single NOD-SCID mouse implanted with PSMA+ PC3 PIP and PSMA− PC3 flu xenografts was used for imaging. The mouse was anesthetized with 3% isoflurane in oxygen for induction and maintained under 1.5% isoflurane in oxygen at a flow rate of 0.8 L/min. Then the mouse was placed in the prone position on the gantry of a GE eXplore VISTA small animal PET scanner (GE Healthcare) and injected intravenously with 0.38 mCi (14.1 MBq) [¹⁸F]3 in 200 μL of PBS. The images were acquired as a pseudodynamic scan, that is, a sequence of successive whole-body images were acquired in 2 bed positions. The dwell time at each position was 1 minute such that a given bed position (or mouse organ) was revisited every 3 minutes. An energy window of 250 to 700 keV was used. Images were reconstructed using the FORE/2D-OSEM method (1 iteration, 16 subsets) and included correction for radioactive decay, scanner dead time, and scattered radiation. After PET imaging the mobile mouse holder was placed on the gantry of an X-SPECT (Gamma Medica Ideas) small animal imaging device to acquire the corresponding CT. Animals were scanned over a 8.9-cm field-of-view using a 230 μA, 75 kVp beam. The PET and CT data were then coregistered using NIH AMIDE software (http://amide.sourceforge.net/).

The human dosimetry values were obtained using the mouse biodistribution data. The mouse organ activity concentrations in %ID/g were converted to the human %ID/organ by setting the ratio of organ %ID/g to whole-body %ID/g in the mouse equal to that in humans and then solving for the human %ID/organ; the adult male phantom organ masses listed in the OLINDA/EXM 1.0 were used for the conversion (64). The human source organ time-activity curves were fitted using a monoexponential function. Because the biodistribution data were radioactive decay corrected, only the biological removal constants were obtained from the curve fits, and the physical decay constant for ¹⁸F was added in obtaining the time-integrated activity coefficients (TIAC). The source organ TIACs in MBq-h/MBq were entered in the OLINDA/EXM 1.0 for the dose calculations. The dynamic voiding bladder model was used to obtain the TIAC for the urinary bladder contents. The whole-body clearance half-life (obtained as sum of sampled tissues, excluding the tumors) was used as half-life to describe urinary bladder filling. All radioactivity was assumed eliminated via the urine. A 1 hour voiding interval was assumed.

No potential conflicts of interest were disclosed.

We thank CA92871 and CA134675 for financial support. We also thank James Fox for performing the imaging studies and Gilbert Green and Jianhua Yu for excellent technical support.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.