1. Introduction
The transmembrane G-protein coupled receptor C-X-C chemokine receptor type 4 (CXCR4), which specifically binds to the chemokine CXCL12, is expressed on many cell types, including various types of cancer cells. Activation of CXCR4 by its natural ligand stromal cell-derived factor 1 (SDF-1) is known to be involved in a number of vital physiological processes such as inflammation [
1], cell migration [
2], and tumorigenesis [
3]. CXCR4 plays a key role in organ-specific metastasis, directing the migration of malignant cells expressing this receptor toward microenvironments where the cognate ligand is secreted [
4,
5,
6,
7,
8,
9,
10].
To date, there is no clinically approved imaging agent to assess CXCR4 expression in patients. The radiotracer, [
68Ga]Pentixafor, a cyclic pentapeptide incorporating the chelate, DOTA, is under clinical investigation for imaging CXCR4 expression in Europe [
11,
12,
13,
14,
15,
16,
17,
18]. Although [
68Ga]Pentixafor accumulates in CXCR4-positive tumors with acceptable tumor avidity and contrast, the short half-life of
68Ga (t
1/2 = 68 min) limits its utility. In addition, as
68Ga is obtained from the
68Ge/
68Ga generator the maximum produced activity of
68Ga is typically low (1.85 GBq) [
19] precluding large-scale production for regional distribution.
The DOTA moiety in Pentixafor was added to conserve the presence of the high affinity cyclam ring [
20] found in the small molecule CXCR4 antagonists AMD-3100 and AMD-3465. However, the CXCR4 affinity of the
68Ga complex was much greater than the complexes of other PET or SPECT radiometals such as
111In, precluding the substitution of longer-lived isotopes for
68Ga. Bicyclam-containing [
64Cu]AMD-3100 was taken up in tumors in a CXCR4-dependent manner, but high signal in liver and bone resulted in modest tumor-to-background ratios [
21]. The higher affinity radioligand, [
64Cu]AMD-3465, accumulated in tumors to a greater extent than [
64Cu]AMD-3100, although liver and kidney uptake remained significant and persistent [
22]. Nevertheless, this early study supported the monocyclam structure of AMD-3465 as a scaffold for the design of novel CXCR4-targeting radioligands.
The positron emitting radionuclide,
18F (t
1/2 = 109 min), can be produced in large quantity (>185 GBq) and at high molar activities, making it a more practical choice for CXCR4 probes if higher volume clinical use is to be achieved [
23]. Recently we introduced, [
18F]RPS-544, a derivative of the monocyclam AMD-3465, which demonstrated CXCR-dependent tumor uptake. However, accumulation of [
18F]RPS-544 in the liver and intestines was substantial [
24]. We hypothesized that further modification of the monocyclam structure may lead to a tracer with improved tumor uptake and/or normal tissue kinetics compared with [
18F]RPS-544.
Our strategy was to create a library of fluorine-containing derivatives of AMD-3465, and select leading candidates for synthesis and in vitro and in vivo evaluation on the basis of an in silico screen. To generate structural diversity, we considered modifications and substitutions of the phenyl and pyridyl rings of the parent compound. We present herein a novel series of [18F]fluoroethyltriazolyl-containing monocyclam derivatives and highlight in particular [18F]RPS-534 and [18F]RPS-547, which demonstrate improved CXCR4-dependent tumor uptake and tumor-to-background ratios over [18F]RPS-544.
3. Discussion
The novel fluoroethyltriazole-containing derivatives of AMD-3465, [
18F]RPS-534 and [
18F]RPS-547, show high and specific uptake in CXCR4-positive tumors, which might be predicted based on their binding to PC3-CXCR4 cells. While we acknowledge the compounds prepared as part of this work represent a narrow chemical space, we observed a surprising lack of correlation between the docking score, CXCR4 affinity, and CXCR4 binding. On the basis of the docking score alone, [
18F]RPS-544 (−8.18 kcal/mol) and [
18F]RPS-534 (−8.17 kcal/mol) would be expected to exhibit similar CXCR4 affinities. However, the apparent affinity of [
18F]RPS-534 is 30-fold lower. The Hill plots of both of these ligands support a one site, independent binding model, though we cannot rule out the possibility that multiple molecules of [
18F]RPS-534 can bind CXCR4 independently. This might explain why, notwithstanding its weaker affinity, uptake of [
18F]RPS-534 in PC3-CXCR4 xenograft tumors is 2-fold higher than [
18F]RPS-544. Similarly, although both ligands can be partially displaced from PC3-CXCR4 tumors by co-injection of AMD-3100, the mass of AMD-3100 required to displace [
18F]RPS-534 is greater than the mass required to displace [
18F]RPS-544. An alternative explanation is that [
18F]RPS-534 may also interact with related chemokine receptors such as CXCR7 to a greater extent than [
18F]RPS-544. The affinity or binding interaction of [
18F]RPS-534 for CXCR7 might differ to AMD-3100, which is an allosteric agonist [
25], leading to incomplete blocking of tumor uptake. We are currently developing methods for determining the affinity of [
18F]RPS-534, [
18F]RPS-547, and [
18F]RPS-552 for other chemokine receptors.
When we fitted a curve of the docking score versus IC
50, we derived linear plot with R
2 = 0.76. Our choice of AMD-3465 as a parent structure restricted our ligands to a relatively narrow band of docking scores. Within this narrow band, the docking score proved to have little predictive value. However, inclusion of RPS-510 (docking score = −2.13 kcal/mol; IC
50 >10,000 nM in our assay) [
24] as a negative control in the plot improved the fitting (R
2 = 0.89;
Figure S10). In this context, perhaps the greatest value of the in silico screen is to exclude structures with extremely poor affinity. Our experience suggests that structures whose docking score against CXCR4 exceeds −7.0 kcal/mol should be synthesized and comprehensively evaluated. We acknowledge a potential bias in favor of triazole-containing structures in our choice of candidate CXCR4 ligands due to our familiarity with synthetic and radiosynthetic strategies to producing these compounds. It is possible that some of the other monocyclam-containing analogues that we screened in silico might prove to be superior ligands in vivo. It is apparent that there is a lack of correlation between affinity and tumor uptake which cannot be attributed solely to compound pharmacokinetics. Given the PC3-CXCR4 tumor uptake of the radioligands in our bilateral tumor model, we found that the most relevant metric of evaluating these CXCR4 ligands in vitro is to measure the total cell associated activity (%IA) over 1 h.
The first reported
18F-labeled analogue of [
68Ga]Pentixafor, [
18F]AlF-NOTA-pentixather, was limited by relatively high plasma protein binding and in vivo defluorination [
26]. The effect of these properties was to reduce the tumor-to-background contrast. To our knowledge, no
18F-labeled analogue of pentapeptide 1c [
20] has been developed, perhaps due to the challenges of synthesizing the 4-[
18F]fluorobenzoic acid synthon and selectively coupling it to the ornithine side chain. Therefore, small molecule CXCR4 ligands based around
para-xylyl groups, such as AMD-3100, AMD-3465, and WZ811 [
27], have served as the primary platforms for the derivatives of new fluorine-containing antagonists. [
18F]-3 [
28], alternatively reported as [
18F]RPS-510 [
24], a pyrimidine-pyridine amine compound, had low, but CXCR4-specific tumor uptake. [
18F]RPS-544 showed improved tumor uptake compared to [
18F]-3, but failed to address its pharmacokinetic deficiencies.
The tumor-to-background ratios of [
18F]RPS-534 and [
18F]RPS-547 are noticeably higher than those of [
18F]RPS-544, but generally lower than those of [
68Ga]Pentixafor. One exception is the tumor-to-blood ratio of [
18F]RPS-534, 27.5 ± 0.4 at 1 h p.i., which is more than double the ratio of [
68Ga]Pentixafor. Rapid clearance of [
18F]RPS-534 from the blood is matched by its rapid clearance from the heart, which is captured by a tumor-to-heart ratio of 28.8 ± 0.3. These properties may be suitable for PET/CT imaging of myocardial infarction, an emerging application of CXCR4 molecular imaging [
29]. Uptake in spleen and bone was evident for [
18F]RPS-534 but not observed to the same extent with [
68Ga]Pentixafor, [
18F]RPS-547, or [
18F]RPS-552. A blocking dose of 5 mg/kg AMD-3100 nearly eliminated radioactivity in these tissues, but only reduced activity in the tumor by approximately 70%. This may indicate that [
18F]RPS-534 has a higher affinity than the other radioligands for native murine CXCR4, which is highly expressed in the bone and spleen [
19]. These observations lead us to suggest that, at a minimum, the tumor-to-spleen and tumor-to-bone ratios of [
18F]RPS-534 could improve upon translation to humans.
4. Materials and Methods
4.1. Schrodinger Molecular Modeling
All molecular modeling was performed using Schrodinger v. 2014-3 (Schrodinger, New York, NY, USA). Structures were docked against the available protein structure of the human CXCR4 chemokine receptor in complex with small molecule antagonist (3ODU). The small molecule antagonist was used to define the binding pocket and generate the docking grid for the modeling experiments. An Extra Precision (XP) docking study was performed on >200 fluorine-containing compounds structurally derived from AMD-3465. To calibrate the resulting docking scores, known CXCR4 ligands including AMD-3100, AMD-3465, cyclic pentapeptide 1c [
20], and RPS-544 [
24] were included in the modeling. The compounds were ranked by the docking score, reported in kcal/mol.
4.2. Synthesis of Compounds and Precursors
All precursors and non-radioactive standards were synthesized from commercially available building blocks purchased from Sigma Aldrich (St. Louis, MO, USA), unless otherwise indicated. Compounds were characterized by mass spectrometry and nuclear magnetic resonance (NMR) and were greater than 98% pure. A full description of synthetic procedures and compound characterizations is available in the
Supplemental Data (S002).
4.3. Radiochemistry
4.3.1. General Methods
Reactions were monitored by HPLC and radioHPLC using a PRP polymeric column (50 mm, Hamilton Company, Reno, NV, USA) and a Symmetry C18, 4.6 × 50 mm column (Waters Corporation, Milford, MA, USA). Chemical and radiochemical purity of the final compounds were determined by analytical (radio)HPLC. Compound identity was confirmed by co-injection with non-radioactive standards. Analytical and semi-preparative HPLC were performed on a dual pump Varian Dynamax HPLC (Agilent Technologies, Santa Clara, CA, USA) fitted with an Agilent ProStar 325 Dual Wavelength UV-Vis Detector, and radiochemical purity was determined using a NaI(Tl) flow count detector (Bioscan, Poway, CA, USA). UV absorption was monitored at 220 nm and 280 nm. A binary solvent system was used, with solvent A comprising H2O + 0.1% TFA and solvent B consisting of 90% v/v MeCN/H2O + 0.1% TFA. For analytical HPLC the following gradient was used at a flow of 2 mL/min: 0–2 min: 0% B; 2–8 min: 0–100% B; 8–9 min: 100% B; 9–10 min: 100–0% B. Semi-prep HPLC was performed using a μBondapak® C18, 10 μm, 125 Å, 7.8 × 300 mm column (Waters, Milford, MA, USA). The following gradient was used at a flow of 10 mL/min: 0–1 min: 0% B; 1–16 min: 0–30% B; 16–18 min: 30–100% B; 18–19 min: 100% B; 19–20 min: 100–0% B.
4.3.2. Radiosynthesis of [68Ga]Pentixafor
An ITG 68Ge/68Ga Generator was eluted using 4 mL 0.05M HCl and collected in a sterile 15 mL centrifuge tube. To this eluate, containing 555–1110 MBq (15–30 mCi) of 68GaCl3, was added 25 µL of a 1 mg/ml solution of pentixafor in DMSO. The pH was adjusted to pH 4–5 by addition of 80 µL of a 3N NaOAc solution. The reaction was then placed in an Eppendorf ThermoMixer® (ThermoFisher Scientific, Waltham, MA, USA) at 95 °C and 600 rpm shaking for 15 min. A sample of the reaction was taken and analyzed by radio HPLC to confirm the labeling. Immediately after labeling was confirmed, the reaction was diluted with 9 mL H2O and the product was trapped on a Sep-Pak C18 Plus Lite (Waters, Milford, MA, USA) cartridge. The cartridge was washed with 5 mL normal saline solution (VWR International, Radner, PA, USA) and the product was eluted with 200 µL of a 50 % EtOH/saline solution, followed by 900 µL saline to produce a final injectable solution in 9% EtOH. A final QC HPLC was performed to assess purity.
4.3.3. Radiosynthesis of 18F-Labeled Monocyclam Derivatives
No-carrier-added 18F− was obtained from a (18O(p,n)18F) reaction by irradiating a liquid H218O target in a 19 MeV EBCO Cyclotron (Vancouver, BC, Canada). H218O containing [18F]F− was passed through a Sep-Pak® Light Waters Accell™ Plus QMA cartridge (Waters, Milford, MA, USA). The [18F]F− was eluted using a solution of 10 mg kryptofix (K222) and 5 mg K2CO3 in 80% v/v MeCN/H2O. The eluate was dried at 100 °C under vacuum (25 mmHg) and a constant flow of N2 (99.999%, 30 mL/min flow). Drying was completed in approximately 20 min following two azeotropic cycles with MeCN (0.5 mL). To the dried residue was added 0.5 mL of a 10 mg/mL solution of 2-azidoethyltosylate in 20% v/v MeCN/DMSO, and the reaction was heated for 10 min at 80 °C. The reaction mixture was purified by distillation at 120 °C for 10 min. The distillate was collected in a clean V-shaped reaction vial containing 300 µL of DMF cooled to 0 °C. To this vial was added 100 µL of a 10 mg/mL solution of alkyne precursor in DMSO, followed by a pre-mixed solution of reduced Cu+ consisting of 50 µL of a 0.5 M aqueous CuSO4 solution, 50 µL of a 1.5 M aqueous sodium ascorbate solution, and 100 µL DMF. The reaction vial was heated for 10 min at 100 °C before it was diluted with H2O (10 mL) and passed through a pre-activated Sep-Pak® Plus C18 Lite cartridge (Waters, Milford, MA, USA). The cartridge was washed with H2O (10 mL) and eluted with TFA (1 mL) into a clean V-shaped reaction vial. Deprotection was completed in 10 min at 100 °C, and the remaining TFA was evaporated under N2 flow. The dried residue was re-dissolved in 200 µL of 1M NH4OAc solution for final purification via semi-prep HPLC. The peak corresponding to the product was collected and diluted with H2O (20 mL). The product was trapped on a pre-activated Sep-Pak® C18 Plus cartridge (Waters, Milford, MA, USA) and eluted with 0.5 mL EtOH (190 proof) followed by 5 mL of isotonic saline. The final product solution contained 9% v/v EtOH and a radioactivity concentration of 185–370 MBq/mL.
4.4. Cell Lines
The human prostate cancer cell line PC3-WT and the stably transduced human PC3-CXCR4 cell line were cultured as previously described [
24].
4.5. Cell-Based Assays
4.5.1. Determination of IC50 by Competition Binding Assay
PC3-CXCR4 cells were plated in 12-well plates at a density of 500,000 cells per well and incubated for 48 h at 37 °C/5% CO
2. Prior to the binding assay the cells were incubated in fresh media and 370–555 kBq of [
68Ga]Pentixafor or [
18F]RPS-544 was added to each well. The corresponding concentration of [
68Ga]Pentixafor in each well was 100 nM, and the concentration of [
18F]RPS-544 was approximately 5 pM. Then non-radioactive ligands were added to give a final ligand concentration in the range 0.001–10,000 nM. A total of eight concentrations were studied, and each concentration was tested in triplicate. The cells were incubated for 60 min at 37 °C/5% CO
2. The media was removed under suction, the cells were washed with cold PBS 1X and detached and transferred to tubes for counting in 1 M NaOH. The cells were counted in a Wizard
2 Gamma Counter (PerkinElmer, Waltham, MA, USA). Non-specific binding was subtracted, and the IC
50 values were determined by assuming a simple competitive binding interaction and fitting the data to a sigmoidal Hill plot in Origin 9.0 (OriginLab, Northampton, MA, USA). Potencies are expressed as IC
50 ± standard deviation (
Supplemental Data, Figure S7).
4.5.2. Determination of Kd by Saturation Binding Assay
To determine the dissociation constants, saturation binding assays were performed in the PC3-CXCR4 cells. The cells were plated as described above, and approximately 5 nM of the radioligand was added to each well. Then the corresponding non-radioactive ligand was added to give a final concentration in the range 0.001–10,000 nM. A total of eight concentrations were tested, and each concentration was tested in triplicate. The samples were washed and counted as described above. Non-specific binding was subtracted, and Kd values were determined by performing a nonlinear one site binding curve fit with equation y = (B
max · x)/(Kd + x). Values are expressed as Kd ± standard deviation (
Supplemental Data, Figure S8).
4.5.3. Determination of Cell Binding and Internalization
Internalization studies were performed for [18F]RPS-534 and [18F]RPS-547 in a time dependent manner. PC3-CXCR4 cells were plated as described above in 12-well plates and incubated for 48 h prior to the experiment. The media was exchanged and [18F]RPS-534 or [18F]RPS-547 (10−5 μM; approximately 400 kBq) was added to each well. All experiments were performed in triplicate. The plates were incubated at 4 °C or 37 °C for 15, 30, 60, and 120 min. Then the media was removed under suction at the corresponding temperature, and the cells washed with 1 mL PBS. Cell surface-bound activity was collected by two successive 5 min incubations of 0.5 mL of a 50 mM glycine buffer (pH = 2.8) performed at 4 °C or 37 °C. The cells were then washed with 1 mL ice-cold PBS and detached with 1 mL ice cold 1 M NaOH. Detached cells were collected and counted along with the washes. An aliquot of the administered radioactivity (10% IA) was also counted and used for quantification. The activity found in the NaOH fractions of the 4 °C plates was subtracted from the activities of the fractions at 37 °C to correct for non-specific binding. Internalized activity was expressed as the fraction of activity in the NaOH fraction relative to the total activity added.
4.5.4. Determination of Cell-Associated Activity
Maximum cell-associated activity was calculated by subtracting the cell uptake of the radioactive ligand at the highest non-radioactive ligand concentration of 10 µM (which we considered to be non-specific uptake/binding) from the cell uptake at the lowest non-radioactive ligand concentration of 10−5 µM.
4.6. Animal Studies
All animal studies were approved by the Institutional Animal Care and Use Committee of Weill Cornell Medicine (Protocol Number: 2015-0066) 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. Hairless 6-to-8-week old male BALB/c inbred athymic nude mice (Jackson Laboratory, Bar Harbor, ME, USA) were inoculated bilaterally with 0.2 mL of a 1:1 mixture of 107 PC3-CXCR4 cells in PBS and Matrigel (Corning, Tewksbury, MA, USA) subcutaneously in the left shoulder and 0.2 mL of a 1:1 mixture of 107 PC3-WT cells in PBS and Matrigel subcutaneously in the right shoulder. Imaging and biodistribution studies were performed three weeks post-implantation, when tumors were approximately 100–300 mm3.
For the imaging studies, the tumor-bearing mice (n = 2/compound) were injected intravenously with 100 μL of a 9% v/v EtOH/saline solution containing approximately 3.7 MBq of the radioligand. Blocking of CXCR4-dependent tissue uptake of [18F]RPS-534, [18F]RPS-547, [18F]RPS-544, and [68Ga]Pentixafor was evaluated by co-injection of 0.1, 0.5, 1, or 5 mg/kg of AMD-3100 (n = 2 per compound per mass of AMD-3100). The animals were imaged under isoflurane at from 60–90 min p.i. in an Inveon μPET/CT (Siemens Medical Solutions, Malvern, PA, USA). Total image acquisition time was 30 min, and a CT scan was performed directly before the PET imaging for anatomical co-registration and attenuation correction. Images were reconstructed using the software provided by the vendor and post-reconstruction analysis, including decay correction, was performed using AMIDE (Open Source Software) prior to comparison. The images are laterally inverted relative to the imaging position, such that the PC3-CXCR4 tumor appears on the right shoulder and the PC3-WT tumor on the left shoulder. Tumor uptake was quantified by the following procedure: CT-aided regions of interest were drawn, and tissue concentration was determined by comparison with a 3 %ID/cm3 standard included in the field of view. Uptake was expressed as percent injected dose per centimeter cubed (%ID/cm3) ± standard deviation.
For the biodistribution studies, tumor-bearing animals (n = 5/compound/time point) were injected intravenously with 50–100 μL of a 9% v/v EtOH/saline solution containing approximately 1.1 MBq of the radioligand. Co-administration of AMD-3100 (5 mg/kg) was used to confirm CXCR4-dependent uptake of [18F]RPS-534, [18F]RPS-547, and [18F]RPS-552. The mice were euthanized at 1 or 2 h p.i. and tissues including blood, heart, lungs, liver, small intestine, large intestine, stomach (with contents), spleen, pancreas, kidneys, muscle, bone, PC3-WT tumors, PC3-CXCR4 tumors, and tail were collected. All organs were weighed and placed in test tubes for counting in a Wizard2 Gamma Counter (PerkinElmer, Waltham, MA, USA). For comparison, 1% injected activity standards were prepared and counted along with the tissue samples. The tails were counted to correct for any dose extravasation. Activity in each tissue was corrected for decay and for injected activity and expressed as percent injected dose per gram of tissue (%ID/g) ± standard error.
5. Conclusions
As predicted by their binding to PC3-CXCR4 cells, the fluoroethyltriazole-containing derivatives of AMD-3465, [18F]RPS-534, and [18F]RPS-547 showed high and specific uptake in CXCR4-positive tumors. Uptake of [18F]RPS-547 was comparable to [18F]RPS-544, a first-generation AMD-3465 derivative, but the improved clearance profile of [18F]RPS-547 led to higher tumor-to-normal tissue ratios. The tumor-to-normal tissue ratios of [18F]RPS-534, particularly tumor-to-blood, tumor-to-muscle, and tumor-to-lung, were even greater and comparable to [68Ga]Pentixafor. The chemical and pharmacokinetic properties of [18F]RPS-547 and especially [18F]RPS-534, including high yielding radiosynthesis, high tumor uptake, and good contrast to background, render these radiotracers promising candidates for imaging CXCR4 expression by PET.