Low kidney uptake of GLP-1R-targeting, beta cell-specific PET tracer, 18F-labeled [Nle14,Lys40]exendin-4 analog, shows promise for clinical imaging

Background Several radiometal-labeled, exendin-based tracers that target glucagon-like peptide-1 receptors (GLP-1R) have been intensively explored for β cell imaging. The main obstacle has been the high uptake of tracer in the kidneys. This study aimed to develop a novel GLP1-R-specific tracer, with fluorine-18 attached to exendin-4, to label β cells for clinical imaging with PET (positron emission tomography). We hypothesized that this tracer would undergo reduced kidney uptake. 18F-labeled [Nle14,Lys40]exendin-4 analog ([18F]exendin-4) was produced via Cu-catalyzed click chemistry. The biodistribution of [18F]exendin-4 was assessed with ex vivo organ γ-counting and in vivo PET imaging. We also tested the in vivo stability of the radiotracer. The localization of 18F radioactivity in rat and human pancreatic tissue sections was investigated with autoradiography. Receptor specificity was assessed with unlabeled exendin-3. Islet labeling was confirmed with immunohistochemistry. The doses of radiation in humans were estimated based on biodistribution results in rats. Results [18F]exendin-4 was synthesized with high yield and high specific activity. Results showed specific, sustained [18F]exendin-4 uptake in pancreatic islets. In contrast to previous studies that tested radiometal-labeled exendin-based tracers, we observed rapid renal clearance of [18F]exendin-4. Conclusions [18F]exendin-4 showed promise as a tracer for clinical imaging of pancreatic β cells, due to its high specific uptake in native β cells and its concomitant low kidney radioactivity uptake.


Background
The determination of β cell mass is a challenge, because the target is a diffuse collection of cell clusters, highly dispersed throughout the pancreas. Indeed, the β cell mass constitutes less than 2% of the total pancreas mass, and this mass is likely to decrease during the course of diabetes. The ongoing development of therapeutic approaches to diabetes, whether based on pharmacology or islet transplantation, also calls for the development of methods for longitudinal in vivo assessments of β cell mass and function. There is an unmet need for an imaging agent that offers potential for clinical trial applications.
Glucagon-like peptide-1 (GLP-1) is secreted from the intestine in response to elevated blood glucose levels. GLP-1 binds to specific receptors on pancreatic islets to stimulate insulin secretion. Exendin-4 is a long-acting agonist of the glucagon-like peptide-1 receptor (GLP-1R); it mimics the actions of short-lived GLP-1 (plasma halflife <2 min). Several efforts have been made to label pancreatic β cells with exendin-based tracers that carry radiometals, such as 99m Tc and 111 In for SPECT imaging [1,2] and 64 Cu and 68 Ga for PET imaging [3]. However, high radioactivity concentrations were observed in the kidney. During imaging, this high radioactivity compromises tissue visualization in the kidney region, limits quantitative imaging, and reduces diagnostic accuracy. Because the kidney is a radiosensitive organ, the radiation burden might rapidly become unacceptably high with long-lived radioisotopes [4]. Some success has been achieved in efforts to reduce the renal radioactivity of various radiolabeled peptides. For example, some studies co-administered amifostine, albumin derivatives, polyglutamic acid, or Gelofusine [5][6][7] to limit renal uptake; and others used cleavable linkers, to allow excretion of radioactive metabolites into the urine [8,9].
Most peptides are excreted via the kidneys, but studies have shown that persistent retention is characteristic for radiometal-labeled exendin derivatives [10]. As an alternative, preclinical studies were conducted with 18 Flabeled exendin-4; they showed high, sustained tracer uptake in INS-1 tumor cells and xenograft models [11]. However, importantly, after initially high radioactivity levels in the kidneys, clearance from the kidneys was rapid, compared to radiometal-labeled analogs [12,13]. Wu et al. [14] observed specific uptake of [ 18  The vesicular monoamine transporter type II (VMAT2) has been suggested as a target to calculate β cell mass, but so far issues with this approach remain controversial [15][16][17].
Aqueous [ 18 F]fluoride was produced by proton bombardment of 18 O-enriched water as reported previously [18]. Synthesis started by allowing tosylate 1 to react with a [ 18 F]fluoride-Kryptofix complex in anhydrous DMSO (Fig. 1). After purification by HPLC, the collected fraction was concentrated on an Oasis HLB cartridge (Waters, Milford, MA, USA). After elution of the cartridge with THF, the obtained [ 18 F]fluoro-propargyltriethylene glycol was allowed to react with peptidyl azide 2 in the presence of pre-activated copper sulfate/sodium ascorbate for 10 min at room temperature. Subsequently, the reaction mixture was diluted in water and purified by HPLC. The collected fraction was retained on a Sep-Pak C8 cartridge (Waters) and formulated in ethanol/saline. Radiochemical analysis of 18 F-labeled [Nle 14 ,Lys 40 ]exendin-4 ([ 18 F]exendin-4) was performed with HPLC and thin layer chromatography (TLC). The reaction details and a full characterization of the tracer will be reported elsewhere.

Ex vivo autoradiography
The kidney, pancreas, and brain were sectioned and exposed to imaging plates (Fuji BAS-TR2025; Fuji Photo Film Co., Tokyo, Japan) for approximately 4 h. Imaging plates were scanned with a BAS-5000 scanner (Fuji Photo Film Co.) and the images were analyzed with AIDA 4.5 software (Raytest, Isotopenmessgeräte, Straubenhardt, Germany). Three pancreatic sections per rat were analyzed (N = 2-8 rats/time point). We determined the mean density values (photostimulated luminescence per square millimeter, PSL/mm 2 ) of ten prominently labeled islets and of the exocrine tissue and corrected for background noise. Finally, the islet-to-exocrine tissue ratio was calculated. This was done in order to be able to demonstrate the accumulation of radioactivity into the islets as a function of time and to be able to compare the islet-toexocrine tissue ratios between animals of different time points. The pancreatic sections used for autoradiography were then stained with insulin antibody (1' mouse-antiinsulin, Nordic Biosite, Täby, Sweden, 2' MAHC1 mouse probe, Biocare Medical, Concord, CA, USA), counterstained with hematoxylin, and imaged with a Slide Scanner (Pannoramic 250 F, 3DHistech, Budapest, Hungary).

Metabolite assay
The metabolic stability of [ 18 F]exendin-4 was evaluated by radio-HPLC, as reported previously [3]. Briefly, samples were collected at 5, 15, 60, and 120 min post injection (p.i.; N = 1-3 samples/time point). Plasma and urine proteins were precipitated with equivalent volumes of acetonitrile and removed by centrifugation. Samples of the pancreas, kidney, and duodenum (pyloric sphincter) were homogenized in acetonitrile, purified by centrifugation, and the supernatants were analyzed with radio-HPLC.

In vitro [ 18 F]exendin-4 binding in human and rat pancreas and rat brain
We acquired a piece of human pancreas that was frozen immediately after surgery. We dissected a rat pancreas and brain and froze them immediately after dissection. Prior to incubations with ligand, the tissue sections of human and rat pancreas and GLP-1R-rich areas from rat brain [19] were incubated for 30 min with 1% BSA in PBS (pH 7.4). Then, sections were incubated for 60 min with various concentrations of [ 18 F]exendin-4 (range 2.5-20 nM) in vitro. To investigate tracer specificity, sections were incubated for 30 min with an excess of cold exendin-3 prior to adding [ 18 F]exendin-4. Sections were then washed and exposed to imaging plates, as described above (ex vivo autoradiography). Next, sections were stained and processed, as described above.

Estimation of radiation doses for humans
Absorbed doses were calculated with OLINDA/EXM1.0 software [20,21], which includes radionuclide information and a selection of human body phantoms. Rat ex vivo biodistributions were estimated by integrating the area under the time-activity curves. The obtained residence times were converted into corresponding human values by multiplying with organ-specific factors that scaled organs to body weights, as follows: (W TB,rat /W Organ,rat ) × (W Organ,human /W TB,human ), where W TB,rat and W TB,human are the total body weights of rats and humans (70-kg male), respectively; and W Organ,rat and W Organ,human are the organ weights of rats and humans (70-kg male), respectively.

Statistical analyses
Results are reported as the mean ± SD. The two-tailed, unpaired Student's t test and one-way ANOVA were used for determining statistical differences between groups (GraphPad Prism 5.01, GraphPad Software, San Diego, CA, USA). Linear regression analysis was used to assess the relationship between the islet-to-exocrine tissue ratio and time. Values of P < 0.05 were considered statistically significant.

Radiochemistry
The synthesis of [ 18 F]exendin-4 is depicted in Fig. 1. The copper(I)-catalyzed reaction between exendin-4-azide 2 and [ 18 F]fluoro-propargyl-triethylene glycol proceeded rapidly and efficiently under mild reaction conditions. Technical details of the synthesis optimization studies will be reported elsewhere. The absence of major radioactive by-products ensured neat purification using HPLC.

Ex vivo biodistribution and PET imaging
Measurements of [ 18 F]exendin-4 in tissue samples indicated low but sustained uptake in the whole pancreas over the course of the study (Fig. 2a, Table 1). [ 18 F]exendin-4 accumulation in the kidneys was high at first, but with time, its clearance was rapid and relatively little was retained (Fig. 3). A co-injection of excess exendin antagonist did not significantly reduce the renal uptake of [ 18 F]exendin-4 (Table 1); this finding demonstrated that kidney tracer uptake was not mediated by the GLP-1R. In contrast, tracer uptake rates in the pancreas, lung, and stomach wall were significantly reduced after the exendin-3 injection, consistent with the expression profile of GLP-1R in rat [22].
Dynamic imaging revealed increasing accumulation of [ 18 F]exendin-4 in the kidneys within minutes p.i., and clearance had already begun in the late phase of the dynamic 60-min PET scan (Fig. 4a-b). Late static scans showed rapid elimination via the kidneys. The lung and stomach wall showed modest retention. Uptake in liver was low, which indicated that the tracer was not metabolized via the hepatic pathway ( Fig. 4c-d). Both ex vivo biodistribution measurements and PET imaging showed very low uptake in the brain. F]exendin-4 uptake (highlighted with a blue box) in the pancreas was significantly higher (P = 0.0001) than uptake of the 64 Cu-and 68 Ga-labeled PET tracer counterparts (one-way ANOVA) [3]. Conversely, the uptake of the [ 111 In]exendin-4 SPECT tracer (grey) was higher than uptake of the PET tracers [2] Analysis of tracer metabolites In rat plasma, we detected unchanged [ 18 F]exendin-4 and a more polar radioactive metabolite(s). The plasma radioactivity associated with intact [ 18 F]exendin-4 at 60 min p.i. represented 28.3 ± 3.5% of the total radioactivity, and at 120 min p.i., this level decreased to 16%. At 60 min p.i., [ 18 F]exendin-4 represented 20 ± 7.2% of the radioactivity in the pancreas, compared to 38 ± 3.5% of the radioactivity in the duodenum, and negligible levels of intact [ 18 F]exendin-4 in the kidney and urine. Additionally, in pigs, the plasma radioactivity associated with [ 18 F]exendin-4 at 5 min p.i. represented 87 ± 6% of the radioactivity, and at 60 min p.i., this level decreased to 77 ± 3%.

F]exendin-4 in rat
Differences between values at 60-min and 60-min blocking were determined with the Student t test without correction for multiple comparisons. One-way ANOVA and Bonferroni's multiple comparison test was used to determine the differences at 60, 120, 240, and 360 min. */ † P < 0.05, **/ † † P < 0.01, and ***/ † † † P < 0.001 statistically significant compared to [ 18 F]exendin-4 at 60 min post injection. Results are expressed as the mean ± SD percentage of the injected dose per gram of tissue Fig. 3 Ex vivo measurements of accumulation of exendin-based PET and SPECT tracers in rat kidney. The renal clearance of [ 18 F]exendin-4 (blue) was rapid, and the radioactivity in kidney was significantly lower at all investigated time points, compared to other, metal-labeled, exendin-based tracers designed for PET and SPECT analyses [2,3,41]. Ex vivo autoradiography images (blue inset) show the decreasing [ 18 F]exendin-4 accumulation in the kidney cortex at 60, 120, and 360 min p.i. body temperature from 37 to 41°C. The blood pressure decreased, and arrhythmia was observed. Similar symptoms and effects of exendin on heart rate and blood pressure have been reported previously [23].

Intrapancreatic distribution of [ 18 F]exendin-4 in rats
After [ 18 F]exendin-4 injections, tracer uptake was observed in the islets of rat pancreas (Fig. 5a). Islet labeling was not visible when rats were co-injected with excess cold exendin-3 and [ 18 F]exendin-4 ( Fig. 5b-c). These results indicated that [ 18 F]exendin-4 had bound specifically to GLP-1R in the islets. The autoradiography results demonstrated that [ 18 F]exendin-4 was cleared from the exocrine pancreas, but uptake was sustained in the islets. These findings were reflected in the islet-to-exocrine tissue ratio, which increased over time (Fig. 5d). These findings were consistent with previous reports [3,8]. In contrast, our autoradiography results of rat brain did not show any tracer uptake in the investigated GLP1-R-rich areas after injections of either high (50 MBq) or low (5 MBq) doses of radioactivity.

Tracer binding in tissue sections of human pancreas
We investigated tracer binding in human pancreas by incubating pancreatic tissue sections with various concentrations of [ 18 F]exendin-4. We found that tracer binding was clearly visible in human pancreatic islets (Fig. 8a-b)  was observed in human pancreatic islet tissue sections ( Fig. 8c-d).

Dosimetry
By extrapolating from our rat ex vivo results, we calculated that the mean [ 18 F]exendin-4 effective dose for a 70-kg human adult was 0.021 mSv/MBq ( Table 2). The kidney was the dose-limiting organ. When we based calculations on the ex vivo results, the absorbed kidney dose for [ 18 F]exendin-4 was 0.3 mSv/MBq.

Discussion
This report described the synthesis of [ 18 F]exendin-4 and its evaluation in rat and human pancreas. In this study, ex vivo γ-counting of rat whole pancreas and autoradiography analyses of pancreatic sections showed specific and sustained uptake of [ 18 F]exendin-4 in the islets. In vitro labeling of human pancreas revealed specific targeting of the islets and co-localization with insulin, detected with immunohistochemistry. Renal clearance of [ 18 F]exendin-4 was rapid compared to radiometal-labeled exendin derivatives [2,3,24,25].   5 nM) b co-localization with the islets was observed by staining the same sections with insulin antibody. c-d Islet labeling was not observed, when an excess of cold exendin-3 was added. Scale bar 1000 μm 18 F-labeled target-specific peptides are ideal in vivo imaging agents, because 18 F is readily available from most small medical cyclotrons, and it has a half-life (110 min) that matches the pharmacokinetics of most peptides. Moreover, 18 F decays with a 96% positron branching ratio, and it has a short β + trajectory (E βmax = 635 keV), which enables the acquisition of highresolution PET images. Previous reports have described various approaches for the radiofluorination of exendin-4 analogs, which utilize maleimide-or aldehyde-containing prosthetic reagents, and chelating agents [13,[26][27][28]. However, the main challenge is designing an efficient strategy for preparing radiotracers with very high specific activity, which would render in vivo imaging of native β cells. Click chemistry has generated increasing attention due to its rapid and high-yielding transformation, and tolerance to a broad range of functional groups. These properties make click chemistry one of the preferred strategies in developing 18 F-labeled radiopharmaceuticals [14,29,30]. The original copper-catalyzed variant of the Huisgen 1,3-dipolar cycloaddition of terminal alkynes and azides features very high reaction rates [31,32] and has prompted us to advance this reaction into the design of a potential clinical grade tracer for native β cell imaging. It was shown previously that the binding affinity (IC 50 ) of native exendin-4 for GLP-1 receptor was not  significantly altered by modification/addition of a fluorinated linker [14,33]. Noteworthy, the different fluorination strategies did not affect the binding affinity of the cold ligand towards the target receptor, compared to native exendin-4.
The GLP-1R is expressed specifically in pancreatic β cells [34,35]; little to no expression was found on other pancreatic islet cell types [36]. Because GLP-1R expression levels are low, a sensitive radiotracer with high specific activity is required to avoid receptor saturation with unlabeled peptide [37]. The number of GLP-1Rs in native β cells and transplanted islets is restricted compared to insulinomas, which overexpress the receptor. In this study, [ 18 F]exendin-4 was produced with a specific activity as high as 323 MBq/nmol and an injected mass of 2 ± 1 μg/kg. In humans, the pharmacological dose of exenatide (Byetta®) is low (10 μg BID), and in animal studies, exendin-4 was a potent glucose-lowering agent (dose 0.001 to 10 μg) [38]. These properties of high specific activity and low injected radiotracer mass must be considered carefully. With doses of 7-25 μg of [ 68 Ga]NOTA-exendin-4, subjects experienced palpitation, vomiting, and nausea after tracer injection [39].
In previous studies, the levels of pancreatic uptake of other labeled exendin-based tracers were similar at 60 min p.i. (Fig. 2b). Nevertheless, [ 18 F]exendin-4 accumulation in the pancreas was significantly higher than that of the 64 Cu-and 68 Ga-labeled counterparts [3]. Brom et al. [2] showed that the [ 111 In]DTPA-exendin-4 SPECT tracer was taken up by the pancreas at a higher rate than the PET ligands.
Kastin et al. [40] reported that exendin-4 crossed the blood-brain barrier in rat. In our studies, GLP-1R-rich areas were observed in in vitro studies of rat brain sections (Fig. 7). However, ex vivo autoradiography, tissue γ-counting, and PET imaging did not reveal any specific binding in rat brain. These results indicated that [ 18 F]exendin-4 did not cross the blood-brain barrier.
Our stability study results demonstrated that the native (unmetabolized) tracer was recovered in significantly higher amounts in pig plasma compared to rat plasma. However, in rats [ 18 F]exendin-4 was rapidly cleared from the blood, and low amounts were taken up in the liver. In contrast, when 18 F-labeled GLP-1R targeting radiotracers were investigated in xenograft models, substantial liver uptake was observed [11,13,14,28]. On the other hand, Yue et al. [12] reported moderate liver retention of [ 18 F]FNEM-[Cys 40 ]-exendin-4 in a xenograft model. High liver uptake might restrict the clinical use of these tracers, because the liver is the site of insulinoma metastasis, and it is also the primary site for islet transplantation. Low tracer accumulation in the liver enables the high target-to-background contrast required for high-quality PET imaging.
In this study, PET imaging and organ radioactivity measurements demonstrated high [ 18 F]exendin-4 uptake in the kidneys soon after injection, but thereafter, clearance was rapid and retention decreased (Fig. 3). In contrast, the renal excretion route and proximal tubular reabsorption cause high retention of metal-labeled exendin tracers in the kidney cortex. At the 60-min time point, kidney uptake of [ 18 F]exendin-4 was significantly lower than the uptake of the metal-labeled counterparts. Furthermore, at several hours p.i., high levels of both [ 111 In]DTPA-exendin-4 [2,41] and [ 64 Cu]NODAGAexendin-4 [3] had accumulated at in the kidneys, and [ 111 In]DTPA-exendin-4 showed only minor clearance over time. Consequently, repeated scanning in patients could cause radiation doses that exceed the tolerance of this delicate organ [4,42]. Melis et al. [43] showed that high radiation doses of [ 111 In]DTPA-Lys 40 -exendin-4 caused nephrotoxicity in mice. Several approaches have been tested to overcome high uptake of exendin-based tracers in the kidneys, including co-administration of competitive inhibitors of reabsorption [5][6][7], addition of metabolizable linkers [8,9], or modifications in the size, charge, or structure of the tracer [44,45]. Those attempts have been crucial in finding ways to improve the imaging accuracy, decrease the risk of nephrotoxicity, and improve the efficacy of peptide receptor radionuclide therapy. In this study, the absorbed kidney dose of [ 18 F]exendin-4 was similar to that of [ 68 Ga]Ga-DO3A-VS-Cys 40 -exendin-4 (Table 3) [46]. The absorbed kidney doses of 64 Cu-and 111 In-labeled tracers were 5 and 15 times higher, respectively, than that of [ 18

Conclusions
We found that in vivo injections of the β cell-specific PET tracer, [ 18 F]exendin-4, resulted in specific, sustained tracer uptake in rat pancreatic islets. Additionally, we demonstrated specific uptake of [ 18 F]exendin-4 in human pancreatic tissues in vitro. Unlike its metal-labeled counterparts, [ 18 F]exendin-4 renal clearance was rapid. These properties of [ 18 F]exendin-4 showed promise for its future development for use in clinical imaging of β cells.