Preclinical In Vitro and In Vivo Evaluation of [18F]FE@SUPPY for Cancer PET Imaging: Limitations of a Xenograft Model for Colorectal Cancer

Molecular imaging probes such as PET-tracers have the potential to improve the accuracy of tumor characterization by directly visualizing the biochemical situation. Thus, molecular changes can be detected early before morphological manifestation. The A3 adenosine receptor (A3AR) is described to be highly expressed in colon cancer cell lines and human colorectal cancer (CRC), suggesting this receptor as a tumor marker. The aim of this preclinical study was the evaluation of [18F]FE@SUPPY as a PET-tracer for CRC using in vitro imaging and in vivo PET imaging. First, affinity and selectivity of FE@SUPPY and its metabolites were determined, proving the favorable binding profile of FE@SUPPY. The human adenocarcinoma cell line HT-29 was characterized regarding its hA3AR expression and was subsequently chosen as tumor graft. Promising results regarding the potential of [18F]FE@SUPPY as a PET-tracer for CRC imaging were obtained by autoradiography as ≥2.3-fold higher accumulation of [18F]FE@SUPPY was found in CRC tissue compared to adjacent healthy colon tissue from the same patient. Nevertheless, first in vivo studies using HT-29 xenografts showed insufficient tumor uptake due to (1) poor conservation of target expression in xenografts and (2) unfavorable pharmacokinetics of [18F]FE@SUPPY in mice. We therefore conclude that HT-29 xenografts are not adequate to visualize hA3ARs using [18F]FE@SUPPY.


Introduction
Colorectal cancer (CRC) is the fourth leading cause of cancer-related deaths in men and women worldwide [1]. The primary diagnosis is usually made by colonoscopy and biopsy, which often does not reflect the full extent of the disease due to tumor heterogeneity and disregard of potential metastases. Positron Emission Tomography (PET) provides a noninvasive imaging technique, which is valuable for tumor staging and clinical decision making and to estimate the patient's prognosis [2]. Besides the routinely used PET-tracer [ 18 F]FDG, the availability of specific tumor tracers would enhance the characterization of colorectal tumors and help in CRC staging and with the choice of treatment.
An essential characteristic of most solid tumors is hypoxia, which inevitably leads to accumulation of adenosine within the tumor microenvironment as a result of the breakdown of adenine nucleotides, which has been recognized in the 1990s [3,4]. Since then, many efforts have been made to clarify the role of adenosine and its receptors in cancer [5][6][7]. The expression of the A 3 adenosine receptor (A 3 AR), which is one of four subtypes of the adenosine receptor family, has been reported in several human tumor cell lines including leukemia (Jurkat T, HL-60), melanoma (A375), and astrocytoma (ADF) [8][9][10][11][12]. In particular, there is a rising interest in the involvement of A 3 ARs in CRC as A 3 AR protein expression has been reported for various colon cancer cell lines, including Caco-2, HCT-116, CCL-228, DLD-1, and HT-29 [13][14][15]. Merighi et al. have shown that caffeine leads to hypoxia-inducible factor-1 (HIF-1) protein accumulation and increased vascular endothelial growth factor (VEGF) expression through A 3 AR stimulation in HT-29 cells under hypoxic conditions [16]. According to Sakowicz-Burkiewicz et al., treatment with the A 3 AR agonist IB-MECA (1 M) results in an A 3 AR-dependent growth promoting effect in HT-29 cells. In contrast, IB-MECA causes cell apoptosis in HCT-116 cells, similarly in an A 3 AR dependent manner [13].
High expression of A 3 AR mRNA and protein has been reported in colon and breast carcinoma compared to adjacent nonneoplastic tissue by Madi et al. Remarkably, even higher levels of A 3 AR mRNA have been found in lymph node metastases than in primary tumor tissue, suggesting A 3 AR-overexpression as a marker for tumor progression [17]. Additionally, Gessi et al. studied A 3 AR expression in colorectal cancer tissue samples of 73 patients and provided evidence that the A 3 AR has the potential to be used as a diagnostic marker for colon cancer. The authors have shown ≥2-fold increased A 3 AR protein expression in primary colon carcinomas compared to normal mucosa and describe a tendency towards higher A 3 AR expression in large adenomas compared to small adenomas. Therefore, the authors proposed a major role of the A 3 AR in cancer aggressiveness [18]. Moreover, radioligand binding experiments using the A 3 AR antagonist [ 3 H]MRE 3008F20 and western blot analysis indicated that the A 3 AR is the most abundant of all four adenosine receptor subtypes in colorectal cancer tissues as well as in colon cancer cell lines (Caco-2, DLD-1 and HT-29). On the contrary, RT-PCR experiments showed relatively low levels of A 3 AR mRNA in the mentioned colon cancer cell lines compared to mRNA levels of the other adenosine receptor subtypes [15]. As mRNA levels do not necessarily correlate with protein levels [19] and protein transcription is a prerequisite for targeted receptor imaging approaches such as PET imaging, protein expression data is the most relevant for this study.
The A 3 AR antagonist [ 18 F]FE@SUPPY has been presented as the first PET-tracer for hA 3 AR imaging in 2008 by Wadsak et al. [20,21]. First preclinical PET imaging using CHO-K1-hA 3 AR xenografts has shown promising results leading to further evaluation of this PET-tracer in oncology [22] [23][24][25]. To our knowledge, no preclinical in vivo PET imaging has been reported for these A 3 AR PET-ligands so far. In our preclinical study, we aimed to evaluate [ 18 F]FE@SUPPY as a PET-tracer for human cancer using in vitro imaging and in vivo PET imaging in a CRC tumor model.

Animals.
Six-week-old male BALB/c mice (BALB/ cAnNRj, Division of Laboratory Animal Science and Genetics, Himberg, Austria) were kept under conventional housing conditions, with food and water supply ad libitum and a 12 h day/night cycle. Male, immunodeficient CB17-SCID mice (CB-17/Icr-scid /Rj, Janvier Labs, France) of the same age were kept under specific pathogen-free conditions in individually ventilated cages. All animals were treated according to the European Union rules on animal care.  were monitored every second day by caliper measurement. The respective tumor volume was calculated according to the following equation: tumor volume (mm 3 ) = 2 × D/2 (where d is the shortest diameter and D the longest diameter). Animals were subjected to PET imaging 10 days after inoculation, when tumors reached a volume of at least 300 mm 3 . Tumor volume never exceeded 1 cm 3 .

Human Tissues.
Colorectal carcinoma tissue and adjacent healthy colon tissue were obtained directly after tumorectomy from two patients after full informed consent and quick-frozen in 2-methylbutane (−40 ∘ C). Tissue was sliced into 16 m slices using a microcryotome (Thermo Scientific Microm HM 560) and stored at −80 ∘ C until usage. Depending on the sample size, 3 to 4 different regions were defined and analyzed by means of autoradiography and immunohistochemistry. Increasing concentrations of test compounds were added, whereby the concentration of dimethyl sulfoxide (DMSO) in final assay volume remained ≤10% (hA 1 AR and hA 2A AR assay) and ≤1% in the hA 3 AR assay. Nonspecific binding was determined using 1 M DPCPX (hA 1 R assay), 1 M SCH-442,416 (hA 2A R assay), or 10 M I-AB-MECA (hA 3 AR assay). Filtration through GF/B filters (Whatman5, presoaked in 0.1% PEI or 0.5% BSA) was performed using a cell harvester (Bran-del5), and receptor-bound radioactivity was determined via gamma counting (2480 Wizard 2 , PerkinElmer) or liquid scintillation counting (Hidex 300 SL). IC 50 fitted binding curves were generated using the GraphPad Software 5.0, and values were calculated using the Cheng-Prusoff equation.

Flow Cytometry.
For the flow cytometric evaluation of hA 3 AR expression, single-cell suspensions of HT-29 cells (2 × 10 5 per tube) were fixed and permeabilized using Cytofix/Cytoperm6 kit (BD Biosciences). Cells were incubated with mouse monoclonal anti-human A 3 AR (100 L of 4 g/mL in PBS + 2% FCS, Abnova H00000140-M01) or mouse IgG2b kappa isotype control (100 L of 4 g/mL in PBS + 2% FCS, eBioscience6 14-4732-85) for 1 h at 4 ∘ C. Following a washing step, bound primary antibodies were detected with rabbit anti-mouse IgG FITC (100 L of 40 g/mL in PBS + 2% FCS, Dako F0261) for 30 min at 4 ∘ C in the dark. Samples were analyzed on a FACSCalibur6 flow cytometer (BD Bioscience), whereby 10,000 single cells were recorded.

Western Blot.
Cell lysates were prepared from 75 cm 2 cell culture flasks when cells reached 80% confluency using radioimmunoprecipitation assay (RIPA) buffer and protease inhibitor cocktail according to the manufacturer's instructions. Tissue lysates from HT-29 xenografts were prepared according to a standard protocol using RIPA buffer (according to sample size approx. 4 times of lysis buffer), protease inhibitor, and Ultra-Turrax5 for homogenization. The protein concentration of cell lysates was determined using Pierce6 BCA Protein Assay Kit (Thermo Scientific), and 20 g protein per well was loaded onto TGX6 precast gels (Bio-Rad). After gel electrophoresis (200 V, 30 min), proteins were transferred to nitrocellulose membranes (Amersham6 Protran6 Premium 0.2 m NC, GE Healthcare Life Sciences) via semidry blotting (80 mA per gel). Membranes were incubated with rabbit polyclonal anti-A 3 AR (Santa Cruz Biotechnology, Inc. sc-13938) (1 : 750, 2 h, RT) and further incubated with goat anti-rabbit IgG HRP conjugate (1 : 5000, 1 h, RT). Detection was performed using the dedicated kit (SuperSignal West Pico Chemiluminescent Substrate detection kit, Thermo Scientific), and chemiluminescence imaging was conducted (Bio-Rad VersaDoc6 Imaging System).

Autoradiography.
Tissue slices were thawed and reconstituted in assay buffer (50 mM Tris-HCl pH 7.4, 100 mM NaCl, 1 mM EDTA, 1% BSA, 1 unit adenosine deaminase/100 mL) for 30 min at RT. Radiosynthesis of [ 18 F]FE@SUPPY was performed as previously described and the product was physiologically formulated (EtOH/0.9% saline 10/90) [20]. Tissue slices were incubated with 50 kBq [ 18 F]FE@SUPPY (40-200 GBq/ mol) in 100 L assay buffer for 1 h at RT. Slices were thoroughly washed with ice-cold wash buffer (50 mM Tris-HCl pH 7.4), dried, and exposed to a phosphor screen overnight. The readout of the phosphor storage screen was performed on a Cyclone Phosphor Imager (Perkin Elmer), and data analysis was performed using OptiQuant5 Software as previously described [26]. Statistical testing was performed using GraphPad Prism 5.0 Software. Differences among groups (colorectal cancer versus healthy colon) were analyzed using a two-tailed, unpaired Student's -test with Welch's correction.

Immunohistochemistry.
Vicinal cryosections of colorectal carcinoma and healthy colon tissue were stained to identify regions with hA 3 AR expression following a standard protocol. In brief, cryosections were fixed (96% ethanol, 10 min), permeabilized (0.2% Triton X-100 in PBS, 5 min), and blocked using Bloxall6 Blocking Solution and a dedicated avidin/biotin blocking kit (Invitrogen, Thermo Fisher Scientific). Additionally, sections were incubated with goat serum (1 : 10 in PBS) to reduce nonspecific binding. Rabbit polyclonal anti-A 3 AR (1 : 100, ab203298; Abcam) was used 1 : 100 in PBS + 0.1% BSA for 1 h in a humid, dark chamber. Purified rabbit IgG (Life technologies) was used as an isotype control likewise. Cryosections were washed 3 times for 5 min (PBS + 0.1% Tween-20) and incubated with biotinylated anti-rabbit IgG (1 : 200, PBS + 5% goat serum) for 30 min. After washing, further detection was performed with the Vectastain5 ABC kit (Vector Laboratories) according to the manufacturer's instructions. DAB substrate kit (Abcam) was used as a chromogen to detect peroxidase, and haematoxylin was used for counterstaining of cell nuclei. Immunohistochemically stained slides were acquired on an automated TissueFAXS microscope system (TissueGnostics, Vienna, Austria) at a 5-fold and 20-fold magnification.

Tracer Stability in Mice
2.4.1. In Vitro Stability Tests. Stability of [ 18 F]FE@SUPPY was tested against mouse liver microsomes, mouse S9 fraction, and mouse plasma (BD Sciences). Amount of intact tracer (%) was determined after 5, 10, 20, 40, 60, and 120 min using an Agilent series 1100/1200 HPLC system connected to a radioactivity detector (Raytest, Ramona Star) ( = 2 in triplicate). The assay was conducted as previously described for respective rat and human enzymes [27].

2.5.
Biodistribution. Ex vivo biodistribution of [ 18 F]FE@SUPPY was assessed 70 min after tracer application in BALB/c mice. Radioactivity was determined using a gamma counter (2480 Wizard 2 , PerkinElmer), organs were wet-weighted, and percentage of injected dose per gram of organ was calculated (%ID/g).

In Vivo Imaging.
Xenograft-bearing CB17-SCID mice were anesthetized using isoflurane (2.5%) mixed with oxygen (1.5 L/min) to avoid movement during the imaging. Blocking agents (2 mg/kg BW FE@SUPPY or MRS1523) or the respective vehicle control (Tween-20/EtOH/0.9% physiological saline 1/9/90) was administered retroorbitally 2 min prior to the radiotracer administration ( = 3 per group). Subsequently, the animals received another retroorbital injection of 17.42 ± 4.5 MBq [ 18 F]FE@SUPPY into the venous plexus of the opposite eye. With a minor delay after the application of the radiotracer (2-3 min), mice were placed into the field of view of the scanner ( PET/CT Inveon, Siemens Medical Solution, Knoxville, USA), and dynamic imaging was performed for 60 min to follow tracer distribution. Vital parameters (respiration, body temperature) were continuously monitored using a dedicated monitoring unit (bioVet; m2m imaging, Cleveland, OH, USA) to ensure the depth of anesthesia and wellbeing of the animals. Retroorbital application volumes did not exceed 100 L per application.

Characterization of Binding and Target Expression.
Affinity and selectivity of FE@SUPPY and its potential metabolites upon cleavage by carboxylesterases, DFE@SUPPY, and FE@SUPPY:11 [28] were determined in competitive binding assays. FE@SUPPY has been first described by Li et al., who reported a value of 4.22 ± 0.7 nM for human A 3 AR. However, this study only provided the selectivity ratio towards rat A 1 AR (rA 1 AR/hA 3 AR = 7400) [29]. Here, we confirmed the affinity of FE@SUPPY towards the human A 3 AR ( = 6.02 ± 0.4 nM, = 3) and demonstrated its selective hA 3 AR binding compared to the other human adenosine receptors. Moreover, the respective theoretical metabolites show little affinity for the hA 3 AR, supporting the potential of FE@SUPPY as a ligand for human in vivo application ( Table 1).
The human colorectal adenocarcinoma cell line (HT-29) was characterized regarding its hA 3 AR protein expression using flow cytometry and western blot. Flow cytometric analysis resulted in mean fluorescence intensity (ΔMFI) of 53.6 ± 22 in three independent experiments (Figure 1). Additionally, A 3 AR protein expression in HT-29 cells was determined by western blot (Figure 8) (western blot results are discussed separately below). This is in line with previous studies, which reported A 3 AR expression for this cell line as well [13,15]. Thus, HT-29 cells were subsequently chosen for tumor graft experiments.

In Vitro Imaging.
Fluorescence microscopy of HT-29 cells showed cell membrane-specific staining, pointing at the expression of hA 3 AR on the cell surface, which is typical for GPCRs ( Figure 2).
In all investigated regions of the two CRC patients, [ 18 F]FE@SUPPY accumulation was higher in colorectal carcinoma tissue slices than in healthy colon tissue slices of the same individual (for detailed analysis see supplementary (available here)). In 5 of 7 regions, a ≥2.3-fold higher binding Table 1: Affinity and selectivity data of FE@SUPPY and metabolites towards adenosine receptor subtypes ( = 3-5 in triplicate; amount of DMSO never exceeded 1% of total assay volume in hA 3 AR assay; DMSO was added up to 10% in hA 1 AR and hA 2A AR assay; * = 2 in triplicate; exact value could not be determined due to limited solubility).  of [ 18 F]FE@SUPPY was found ( < 0.05). This finding is in accordance with Gessi et al., who reported similar ratios by means of [ 3 H]MRE 3008F20 binding [18]. Regions with high accumulation of [ 18 F]FE@SUPPY corresponded to regions with high hA 3 AR expression identified by immunohistochemistry ( Figure 3).  Figure 4). This data indicates higher metabolism in mice compared to rats described in previously conducted studies, where 25.8 ± 5.3% intact tracer was found in plasma after 60 min [22]. However, these data could also be mimicked by the fact that intact [ 18 F]FE@SUPPY is rapidly cleared from blood hepatobiliary (into the bile fluid, compare Figures 5 and 7), and the equilibrium in blood is therefore shifted to the metabolites.

Biodistribution.
Biodistribution was assessed 70 min after tracer application in healthy BALB/c mice and revealed a high accumulation of radioactivity in fat-rich regions (brown adipose tissue, BAT) likely due to the tracer's lipophilicity [30]. Regarding the emunctory organs, liver showed the highest accumulation (14.57 ± 0.20% ID/g), followed by the kidneys (2.67 ± 0.24% ID/g). The additional analysis of body liquids pointed at a mainly hepatobiliary excretion of [ 18 F]FE@SUPPY, as the highest amount was found in bile fluid (162.78 ± 37.51% ID/g). The amount of radioactivity in the kidneys and urine (43.33 ± 9.23% ID/g) suggests the excretion of the hydrophilic radioactive metabolite, 2- [31], which was already proposed by Haeusler et al. [28]. The circulating radioactivity in blood was low after 70 min (1.6 ± 0.1% ID/g). This finding is in accordance with the results obtained by the ex vivo blood analysis. Moreover, pronounced accumulation of [ 18 F]FE@SUPPY was found in A 3 AR rich tissues such as the heart (1.13±0.04% ID/g) and lung (1.50 ± 0.23% ID/g). A similar biodistribution pattern was observed for rats in a previously conducted study [20]. [ 18 F]FE@SUPPY accumulation in the brain was low after 70 min (0.23 ± 0.03% ID/g) ( Figure 5).

3.5.
In Vivo Imaging. PET imaging of the mouse xenograft model revealed high uptake of [ 18 F]FE@SUPPY in the emunctory organs, which was again most pronounced in the liver (SUV = 6.68 ± 0.80). Low standardized uptake values were observed in tumor masses of both HT-29 and CHO-K1 xenograft tumors (SUV = 0.23 ± 0.06 and 0.25 ± 0.33), respectively. There was no difference between CHO-K1 xenografts, which served as a negative control (human A 3 AR negative), and HT-29 xenografts. Moreover, significant blocking could not be achieved ( Figure 6). The affinity of FE@SUPPY for the mouse A 3 AR is uncertain but is expected to be lower than that for the human A 3 AR due to the known species differences. The lack of adequate rodent models, mainly due to the low affinity of most hA 3 AR ligands to the rodent A 3 AR, was already recognized by Yamano et al. who proposed a humanized mouse model [32]. Specific uptake was therefore not expected in mouse tissues. Interestingly, a significant influence of the blocking was observed in BAT (decrease in uptake) and lung (increase in uptake). However, the data is based on a set of three individuals in each group, and displacement was not performed in the same individuals. Since tumor uptake in the chosen model was insufficient and not blockable, this phenomenon was not investigated any further.
For a detailed analysis of the pharmacokinetics, volumes of interest were also generated for mouse body liquids including blood, urine, and bile fluid (  Adenosine concentrations of ∼0,5 M have been proposed in HT-29 tumors grown as xenografts [4]. Even though adenosine displays only intermediate affinity for the A 3 AR (∼1 M at the rat A 3 AR [33]), the PET-tracer would have to compete with the endogenous ligand for A 3 AR occupancy. This may decrease accumulation of [ 18 F]FE@SUPPY in the xenografts. However, more importantly, despite the fact that western blot analysis demonstrated hA 3 AR expression in HT-29 cells, hA 3 AR protein could not be detected in tissue lysates derived from HT-29 xenografts. This indicates that the human receptor is poorly conserved in mice upon tumor graft (Figure 8). To our knowledge, this phenomenon has not been described in literature so far but has tremendous impact on in vivo imaging. PET imaging is only feasible if an abundant amount of the target is available, as only nanomolar or even lower concentrations of PET-tracers are applied.

Conclusion
We found a favorable binding profile of [ 18 F]FE@SUPPY displaying high affinity for the human A 3 AR besides low affinity for the other human adenosine receptor subtypes. Autoradiography showed ≥2.3-fold higher uptake in human CRC compared to adjacent healthy colon tissues. First in vivo studies using HT-29 xenografts showed insufficient tumor uptake. After initial high expression rates of the A 3 AR in the HT-29 cells, tumor masses, derived from HT-29 xenografts, revealed low target expression. The receptor was not conserved in the xenograft, which hampered the PET imaging strategy. An additional drawback of the used mouse model is the unfavorable pharmacokinetics of the PET-tracer visualization of the A 3 AR has not been successful to date and deeper understanding of A 3 AR function is still missing.

Conflicts of Interest
The authors declare that they have no conflicts of interest.