Introduction
Prostate cancer is the second most prevalent cancer among men, with more than 1.1 million diagnoses in 2012 [
1] and more than 292,000 deaths due to prostate cancer reported worldwide in 2013 [
2]. The disease burden continues to grow—157,000 deaths were reported in 1990 [
2]—and it is estimated that more than 180,000 men will be newly diagnosed with prostate cancer in the United States in 2016, and that more than 26,000 deaths due to the disease will be registered [
3]. When detected early and the disease is confined to the prostate gland and regional lymph nodes, the 5-year survival rate is nearly 100 %, but the survival rate drops below 30 % when the disease is metastatic [
4]. Early diagnosis can significantly improve patient prognosis, while sensitive and specific localization of the disease is an important feature in the diagnosis and staging of the disease. Accurate staging is critical for appropriate patient management [
5].
Prostate-specific membrane antigen (PSMA; also known as glutamate carboxypeptidase II) is significantly overexpressed in prostate cancer primary tumors and many metastatic lesions, while expression in healthy prostate and other tissue is limited [
6]. Several other characteristics combine to make PSMA an ideal target for molecular diagnostics and therapeutics for prostate cancer: (1) it is overexpressed at all stages of the disease; (2) expression typically correlates with tumor grade, disease aggressiveness, metastasis and biochemical recurrence; (3) it is a transmembrane protein with an extracellular ligand-binding domain; and (4) the bound ligand-protein complex is internalized via receptor-mediated clathrin-dependent endocytosis [
7,
8]. The potential utility of PSMA as a target for diagnostic imaging and therapy was demonstrated with the radiolabeled monoclonal antibody J591 [
9], using In-111 or Zr-89 for imaging and Y-90 or Lu-177 for therapy, however the pharmacokinetics of the antibody make it unsuitable for diagnostic imaging with short-lived radionuclides [
10].
A number of low molecular weight, urea-based small molecules have been described as single photon emission computed tomography (SPECT) or positron emission tomography (PET) imaging agents for prostate cancer, and several of them are undergoing clinical investigation in humans (structures can be found in Fig.
S1 of the Supplementary Materials). Currently, seven such molecules are in Phase I/II clinical trials in the United States and/or Europe: (i) radioiodinated MIP-1095 (I-123 for SPECT/CT and I-124 for PET/CT) and MIP-1072 (I-123 for SPECT/CT), developed by Molecular Insight Pharmaceuticals, Inc., (ii)
99mTc-MIP-1404 and
99mTc-MIP-1405, two further SPECT imaging agents emerging from the Molecular Insight Pharmaceuticals platform, (iii) [
68Ga]Ga-PSMA-HBED-CC (also known as [
68Ga]PSMA-11 and [
68Ga]DKFZ-PSMA-11) for PET/CT, and (iv) [
18F]DCFBC and its next-generation derivative [
18F]DCFPyL for PET/CT [
11,
12]. Newly introduced compounds to undergo first-in-human evaluation include
68Ga-DKFZ-617, developed to be a theranostic ligand and evaluated in a therapeutic context as
177Lu-DKFZ-617 [
13], and the structurally related
68Ga-CHX-A”-DTPA [
14].
The greater sensitivity and higher spatial resolution of PET relative to SPECT has made this technique the preferred imaging platform in a number of clinical environments [
15,
16]. Among the positron-emitting isotopes currently incorporated into PSMA-targeting ligands, fluorine-18 and gallium-68 are preferred to iodine-124 because of their higher efficiency of positron emission (97 % and 89 % vs. 23 %, respectively) and shorter half-lives. Furthermore, iodine-124 scans require complex reconstruction algorithms to minimize the signal-to-noise ratio, which, in combination with the long half-life of iodine-124 (t
1/2 = 4.18 d) and the undesired emission of beta particles, is often a poor match for the pharmacokinetics of small molecules [
17]. In addition, gallium-68 is currently produced from a
68Ge/
68Ga generator, enabling its use in single-batch syntheses in radiopharmacies independent of access to a cyclotron, and chelation of gallium-68 is both clean and rapid under conditions that are compatible with most small molecules and peptides. These considerations have contributed to the emergence of [
68Ga]Ga-PSMA-HBED-CC as the most widely used radiotracer currently in clinical development [
18].
Fluorine-18 presents a number of practical advantages compared to gallium-68, including: (i) a longer half-life [t
1/2(
18F) = 109.8 min vs. t
1/2(
68Ga) = 67.7 min], which permits multiple step radiosyntheses and allows a longer time for background signals to clear before imaging is performed, (ii) large-scale cyclotron production that allows multiple patient doses to be produced from a single synthesis, and (iii) chemical characteristics, such as a similar atomic radius to hydrogen, that allow diverse types of ligands to be prepared. On this basis, the development of PSMA-targeting ligands labeled with fluorine-18 has emerged recently as a goal of great interest. [
18F]DCFBC is based on a Glu-urea-Cys pharmacophore modified with a 4-[
18F]fluorobenzyl group, and was reported to show good uptake in PSMA
+ xenograft tumors [
19]. In humans, however, it has shown limited clearance from soft tissue, resulting in a decreased tumor-to-background ratio and poor visualization of small lesions [
20,
21]. To address the slow clearance, the second-generation [
18F]fluoropyridine-modified Glu-urea-Lys analogue [
18F]DCFPyL was developed [
22]. Despite more promising pharmacokinetics, fast tumor washout is evident as early as 1 h post-injection (p.i.) and accumulation of radioactivity in evacuated bladders is considerable [
22]. In addition, [
18F]DCFPyL suffers from poor radiochemical yields (in the range 5–12 % decay-corrected [
23,
24], although recently an improved synthesis by direct fluorination has been reported to increase yield to greater than 20 % [
25].
In an effort to address the aforementioned challenges in the development of fluorine-18-PSMA ligands of high specificity and affinity, we describe the synthesis and preliminary structure-activity relationship (SAR) studies of two new classes of [18F]fluoroethyltriazolylphenyl urea-based PSMA ligands, afforded by click chemistry in high radiochemical yield (20–40 %), excellent radiochemical purity (>99 %) and high specific activity (182–391 GBq/μmol) from starting activities of less than 7.4 GBq (200 mCi). Each of these ligands shows substantial tumor uptake in nude mice bearing LNCaP xenografts using μPET/CT. The two most promising ligands, RPS-040 and RPS-041, show excellent PSMA imaging characteristics based on their high specificity for PSMA, high tumor uptake and prolonged tumor retention, rapid clearance from non-target tissues and resulting high tumor-to-background ratios. They also show superior pharmacokinetics when compared to [68Ga]Ga-PSMA-HBED-CC in mice, thereby warranting development as clinical PET imaging agents for prostate cancer.
Cell culture
The human prostate cancer cell line, LNCaP, was obtained from the American Type Culture Collection. Cell culture supplies were from Invitrogen unless otherwise noted. LNCaP cells were maintained in RPMI-1640 medium supplemented with 10 % fetal bovine serum (Hyclone), 4 mM L-glutamine, 1 mM sodium pyruvate, 10 mM N-2-hydroxyethylpiperazine-N-2-ethanesulfonic acid (HEPES), 2.5 mg/mL D-glucose, and 50 μg/mL gentamicin in a humidified incubator at 37 °C/5 % CO2. Cells were removed from flasks for passage or for transfer to 12-well assay plates by incubating them with 0.25 % trypsin/ethylenediaminetetraacetic acid (EDTA).
In vitro determination of IC50
The 50 % inhibition concentrations (IC
50 ) values of the non-radioactive fluorine-containing ligands were determined by screening in a multi-concentration competitive binding assay against
99mTc- ((7S,12S,16S)-1-(1-(carboxymethyl)-1H-imidazol-2-yl)-2-((1-(carboxymethyl)-1H-imidazol-2-yl)methyl)-9,14-dioxo-2,8,13,15-tetraazaoctadecane-7,12,16,18-tetracarboxylic acid technetium tricarbonyl complex;
99mTc-MIP-1427) for binding to PSMA on LNCaP cells, according to methods previously described [
30]. The LNCaP cells were plated 48 hours prior to the experiment to achieve a density of approximately 5 x 10
5 cells/well (in triplicate) in RPMI-1640 medium supplemented with 0.25 % bovine serum albumin prior to performing the assay. LNCaP cells were incubated for 1 hour with 1 nM
99mTc-MIP-1427 in serum-free RPMI-1640 medium in the presence of 1–10,000 nM test compounds. Radioactive incubation media was then removed by pipette and the cells were washed twice using 1 mL of ice-cold HEPES buffer. Cells were harvested from the plates and transferred to tubes for radioactive counting using a Packard Cobra II gamma counter. IC
50 values were determined by non-linear regression using GraphPad Prism software.
Inoculation of mice with xenografts
All animal studies were approved by the Institutional Animal Care and Use Committee of Weill Cornell Medicine and were undertaken in accordance with the guidelines set forth by the USPHS Policy on Humane Care and Use of Laboratory Animals. Animals were housed under standard conditions in approved facilities with 12-h light/dark cycles. Food and water was provided ad libitum throughout the course of the studies. Male inbred athymic nu/nu mice were purchased from The Jackson Laboratory. For inoculation in mice, LNCaP cells were suspended at 4 x 107 cells/mL in a 1:1 mixture of PBS:Matrigel (BD Biosciences). Each mouse was injected in the left flank with 0.25 mL of the cell suspension. The mice were imaged when the tumors reached approximately 200–400 mm3, while biodistributions were conducted when tumors were in the range 100–400 mm3.
Imaging
LNCaP xenograft tumor-bearing mice (two per compound) were injected intravenously via the tail vein as a bolus injection of 7.03–7.77 MBq (190–210 μCi) of the tracer ([18F]RPS series), 5.5–6.5 MBq (150–175 μCi) [18F]DCFPyL or 5.5 MBq (150 μCi) [68Ga]Ga-PSMA-HBED-CC. Specific activity was greater than 190 GBq/μmol. The mice were imaged by μPET/CT (Inveon™; Siemens Medical Solutions, Inc.) at 1 h, 2 h, 4 h and 6 h p.i. ([18F]fluorinated compounds) or 1 h and 3 h p.i. ([68Ga]Ga-PSMA-HBED-CC). Total acquisition time was thirty minutes, and a CT scan was obtained either immediately before or immediately after the acquisition for both anatomical co-registration and attenuation correction. The data were reconstructed using the commercial Inveon™ software supplied by the vendor. Tumor uptake was estimated by drawing a region of interest (ROI).
Biodistribution
LNCaP xenograft tumor-bearing mice (n = 5 per time point) were injected via the tail vein with a bolus injection of 370 kBq (10 μCi) of either [18F]RPS-040 or [18F]RPS-041. The specific activity of the compounds was 341 GBq/μmol and 391 GBq/μmol, respectively. The mice were euthanized by asphyxiation under isofluorane at 1 h, 2 h and 4 h p.i. An additional set of mice (n = 5) was co-administered [18F]RPS-040 (370 kBq; 10 μCi) and 2-PMPA (approx. 250 μg; 10 mg/kg) and sacrificed at 1 h p.i. to determine the uptake specificity. A full biodistribution study was conducted on all mice, and tissues were excised, weighed and counted in an automated γ-counter. Tissue time-activity values were expressed as percentage injected dose per gram of tissue (%ID/g). Statistical comparisons were performed using the standard Student’s t test for a 95 % confidence interval.
Discussion
A prosthetic group strategy was envisioned for the synthesis and radiosynthesis of the six PSMA inhibitors. By this approach, it was possible to overcome the interaction between fluoride and the urea protons [
35] which contributes to low yield and high variability in the direct fluorination of urea-based imaging probes. Copper(I)-catalyzed click reactions have proven to be useful in radiochemistry due to their rapid kinetics and broad functional group tolerance, which allow labeling under mild conditions and in the absence of protecting groups as well as reduced reaction times [
36‐
38]. 2-[
18F]Fluoroethylazide has been reported to be synthesized in good radiochemical yields and to react with small molecules and peptides [
27,
39‐
41]. It has three principal advantages in the context of radiochemical synthesis: 1) it is the smallest azide that can be radiofluorinated, and the size of the resulting [
18F]fluoroethyltriazole permits it to be incorporated into small molecules without necessarily disrupting activity [
39]; 2) it can be purified by distillation, leading to high specific activity productions [
27,
39]; and 3) when Cu(I) is in stoichiometric excess, as it typically is for reactions on a radiochemical scale, the click reaction is second-order with respect to the alkyne [
39,
42,
43].
The application of the 2-[18F]fluoroethylazide/Cu(I)-catalyzed click chemistry methodology to the radiosynthesis of small PSMA-binding molecules was demonstrated to be a straightforward and reproducible route to high-affinity ligands synthesized in good radiochemical yield. The synthesis of 2-[18F]fluoroethylazide was found to be highly efficient and reproducible, though substantial losses were observed during distillation. Although addition of small aliquots of MeCN to the reaction vial increased the recovery of 2-[18F]fluoroethylazide after distillation, the additional volume of MeCN was found to suppress the yield of the click reaction.
Optimization of the click reaction is ongoing, but early experiences have highlighted the sensitivity of the reaction to total reaction volume and the composition of reaction solvents. Conversion to the [
18F]fluorinated triazoles was better in smaller reaction volumes; this is likely to be the consequence of higher reagent concentration. However, in reactions with similar volumes, those that had a higher MeCN content gave poorer triazole yields. This is consistent with reports of the percentage of DMF in the reaction mixture playing an important role in maintaining high levels of Cu(I) [
39]. The similar boiling points of MeCN and 2-[
18F]fluoroethylazide prevent concentration of the distillate, limiting the minimum volume of MeCN that can be used. Therefore, improvements in the yield of the click reaction will likely arise from increasing the concentration of alkyne and/or the CuSO
4/sodium ascorbate mixture.
Recently, a potential [
18F]fluorinated PSMA ligand synthesized by a click approach using 2-([
18F]fluorophenyl)acetylene as a synthon has been reported [
41]. The biological characteristics and pharmacokinetics of the ligand have yet to be established. Although the decay-corrected radiochemical yield was reported to be 30 %, the synthon is synthesized in 3 steps, and this limits even further the potential for automation of the process. Therefore, the use of 2-[
18F]fluoroethylazide as the prosthetic group for radiofluorination appears to be a more promising route to a [
18F]fluorinated PSMA ligand produced in quantities appropriate for clinical use.
The potent PSMA inhibitor MIP-1095 was used as a structural lead due to its high affinity for PSMA and high tumor uptake in both LNCaP xenograft tumor-bearing mice [
44] and in humans [
45]. The rigidity of the phenylurea is credited with improved potency relative to the amide analogue MIP-1072 [
46] and was, therefore, a key feature retained in the structure of the newly described ligands. The ethynyl moiety has been described as a potential bioisosphere of iodine [
47], and the subsequent 1,2,3-triazole has been proposed to be a bioisosphere of amide bonds [
48], suggesting that the structural modifications might not have a severely adverse effect on affinity for PSMA. Furthermore, the fluoroethyl moiety might project more deeply into the S1 accessory hydrophobic pocket identified in crystal structures of glutamate carboxypeptidase II with bound ligands [
49]. While the six triazolyl urea compounds ultimately displayed lower affinities for PSMA than MIP-1095, the imaging and biodistribution studies demonstrated that of the promising pharmacokinetics of the structural lead were retained.
The SAR studies highlight two clear trends in tumor uptake. Compounds in series 2 had greater uptake than their counterparts in series 1 (Table
1), indicating a preference for direct conjugation of the triazole ring to the phenylurea moiety. Derivatization of the ε-amine of Glu-urea-Lys with a rigid arm has previously been shown to improve affinity for PSMA relative to flexible linkers in a series of amide-based PSMA-targeting fluorescent probes [
50] . It is likely that the lack of conformational flexibility in members of series 2 relative to members of series 1 contributes to the improvements observed in binding affinity and tumor uptake.
The second trend emerges from a comparison of the substitution position on the phenyl ring. Tumor uptake is in the order 3 > 4 > 2 for both series of compounds (Table
1). This order of preference was unexpected given previous SAR studies with a halogenated small molecule PSMA inhibitor, which indicated a strong preference for substitution at the 4-position [
26], and with (alkoxyphenyl)urea derivatives of Glu-urea-Lys, for which substitution at the 2-position led to higher-affinity compounds [
51]. It also did not correspond to the rank order of IC
50 values determined in LNCaP cells. Furthermore, the two compounds substituted at the 2-position, RPS-039 and RPS-042, showed clearance via the hepatobiliary pathway in addition to renal excretion, contributing to a lower contrast tumor image. These trends suggest that substitution at the 2-position of the phenylurea does not appear to be detrimental to ligand potency as determined by in vitro assay, but it does negatively influence the in vivo imaging characteristics of the tracer.
In spite of the increased tumor uptake of the 3-substituted [
18F]RPS-040 relative to 4-substituted [
18F]RPS-041, the clearance of [
18F]RPS-041 is more rapid, leading to greater image contrast and a higher tumor-to-background ratio at 2 h p.i. (Figs.
4 and
7). The clearance of [
18F]RPS-038, the phenyl ether analogue of [
18F]RPS-041, is similarly more rapid than [
18F]RPS-043 (Fig.
4). It is not apparent based on the μPET/CT images alone whether the same trend is true for compounds [
18F]RPS-043 and [
18F]RPS-038, so the significance of the finding in the context of SAR studies requires further investigation.
The imaging characteristics of each of the six [
18F]fluorinated triazole PSMA ligands compare favorably to [
68Ga]Ga-PSMA-HBED-CC, the most widely used diagnostic PET imaging agent for prostate cancer. In addition to the greater sensitivity and higher spatial resolution that fluorine-18 offers over gallium-68 [
52], the two- to three-fold higher tumor uptake is evident when images are compared in the same intensity scale (Fig.
8). The improved image quality is reinforced by the biodistribution studies with ligands [
18F]RPS-040 and [
18F]RPS-041, which showed significantly greater tumor-to-background and tumor-to-kidney ratios than [
68Ga]Ga-PSMA-HBED-CC (Fig.
9).
[
18F]DCFPyL, a second-generation [
18F]fluorinated PSMA ligand currently undergoing clinical evaluation, has recently been studied in LNCaP tumor xenografts, and the maximum standardized uptake value (SUV
max ) was reported to be 1.1 ± 0.1 at 1 h p.i. [
25]. This is lower uptake than observed for each of the six RPS ligands, for which SUV at 1 h p.i. is estimated to range from 1.5 to 2.5 and SUV
max in the tumor is calculated to range from approximately 1.5 to 2.9. Moreover, [
18F]DCFPyL is reported to have a tumor-to-blood ratio of 8.3 at 1 h p.i. in LNCaP xenograft tumor-bearing mice [
25]. At this same time point, the tumor-to-blood ratios for [
18F]RPS-040 and [
18F]RPS-041 are 28.8 ± 8.06 and 20.78 ± 7.87, respectively. By 4 h p.i., the prolonged tumor retention and rapid blood clearance drives the ratios to 92.93 ± 75.67 and 118.4 ± 69.4, respectively.
[
18F]DCFPyL was compared by μPET/CT imaging to the RPS ligands in the same LNCaP xenograft tumor-bearing mice. Clearance from the kidneys was rapid, but rapid tumor washout was also evident (Fig.
4). These pharmacokinetics contribute to a reduction in signal in the kidneys, but also to poorer tumor delineation than can be achieved with the RPS series. The in vitro binding of PSMA-targeting ligands to mouse kidney cells was reported to be at least two-fold greater than binding to human kidney cells [
53], suggesting that rapid kidney clearance in pre-clinical mouse models of prostate cancer is not an essential requirement. In this light, the greater tumor uptake, longer tumor retention and greater tumor-to-blood ratios of [
18F]RPS-040 and [
18F]RPS-041 compared to [
18F]DCFPyL are favorable characteristics of these potential PET imaging agents.
Our approach to the development of PET-based PSMA tracers focused on fluorine-18 due its near optimal PET imaging characteristics [
54] and the application of the radiosynthon, [
18F]fluoroethylazide, allowed us to achieve our radiosynthetic goals simply and efficiently. The use of [
18F]fluoroethylazide may add further complexity to automation when using single-purpose or cassette-based synthesis units, in use today by many PET centers, due to the need to label the synthon, purify it by distillation, and carry out the subsequent click reaction in a single reaction pot. Initial work has begun on the adaptation of commercially available cassettes to facilitate the distillation step. In parallel, an affordable, fit-for-purpose custom radiosynthesis box is being developed. The robust chemical and mechanical reproducibility of a synthesis box of this type has already been demonstrated by our group [
55].