Distinctive detection of insulinoma using [18F]FB(ePEG12)12-exendin-4 PET/CT

Specifying the exact localization of insulinoma remains challenging due to the lack of insulinoma-specific imaging methods. Recently, glucagon-like peptide-1 receptor (GLP-1R)-targeted imaging, especially positron emission tomography (PET), has emerged. Although various radiolabeled GLP-1R agonist exendin-4-based probes with chemical modifications for PET imaging have been investigated, an optimal candidate probe and its scanning protocol remain a necessity. Thus, we investigated the utility of a novel exendin-4-based probe conjugated with polyethylene glycol (PEG) for [18F]FB(ePEG12)12-exendin-4 PET imaging for insulinoma detection. We utilized [18F]FB(ePEG12)12-exendin-4 PET/CT to visualize mouse tumor models, which were generated using rat insulinoma cell xenografts. The probe demonstrated high uptake value on the tumor as 37.1 ± 0.4%ID/g, with rapid kidney clearance. Additionally, we used Pdx1-Cre;Trp53R172H;Rbf/f mice, which developed endogenous insulinoma and glucagonoma, since they enabled differential imaging evaluation of our probe in functional pancreatic neuroendocrine neoplasms. In this model, our [18F]FB(ePEG12)12-exendin-4 PET/CT yielded favorable sensitivity and specificity for insulinoma detection. Sensitivity: 30-min post-injection 66.7%, 60-min post-injection 83.3%, combined 100% and specificity: 30-min post-injection 100%, 60-min post-injection 100%, combined 100%, which was corroborated by the results of in vitro time-based analysis of internalized probe accumulation. Accordingly, [18F]FB(ePEG12)12-exendin-4 is a promising PET imaging probe for visualizing insulinoma.

Insulinoma is a rare pancreatic neuroendocrine neoplasm (PanNEN) and causes hypoglycemia owing to insulin oversecretion 1 . It is the most common cause of endogenous hyperinsulinemic hypoglycemia in adult patients without diabetes and might play a part in the multiple endocrine neoplasia type 1 (MEN1) syndrome 2 . Because surgical resection is the only curative treatment for insulinoma, accurate preoperative tumor localization is essential for the determination of the precise area of resection in the pancreas 3,4 . However, determination of localization remains challenging because of the small sizes of these tumors 5,6 . Whereas the conventional imaging methods such as ultrasonography, computed tomography (CT), and magnetic resonance imaging have proven to display low sensitivity, selective arterial calcium stimulation test and endoscopic ultrasound are invasive and operator-dependent 6,7 . Additionally, somatostatin receptor (SSTR)-targeted imaging approaches have demonstrated improved sensitivities, but are highlighted by limitations because insulinomas, especially benign ones, lack reasonable expression levels of SSTR subtype 2 and 5 8,9 . Moreover, they do not reflect endocrine functions of the tumors, thereby making it impossible to differentiate insulinoma from other panNENs in cases with multiple panNENs.
GLP-1R and SSTR2 mRNA expressions in pancreatic tumors of Pdx1-Cre;Trp53 R172H ;Rb f/f mice. For the resected pancreatic tumors of Pdx1-Cre;Trp53 R172H ;Rb f/f mice with a non-fasting blood glucose level of < 80 mg/dL (n = 10), the relative GLP-1R and SSTR2 mRNA expression levels as well as anti-insulin and -glucagon antibody staining of each tumor are as follows: The immunohistochemical analysis of the insulinpositive tumors (n = 6) demonstrated glucagon-negative staining, whereas that of the insulin-negative tumors (n = 6) showed glucagon-positive staining ( Fig. 2A), consistent with a previous report 22 . The expression levels of GLP-1R mRNA in insulin-positive tumors were significantly higher than those in insulin-negative tumors and the whole pancreas (Fig. 2B). In contrast, SSTR2 mRNA expression levels of insulin-positive tumors tended to be lower than those of insulin-negative tumors and whole pancreas (Fig. 2C), which was comparable with the clinical characteristics of insulinoma [8][9][10] . As for [ 18 Fig. 4. In this mouse, the two insulin-positive and the other two insulin-negative tumors were identified by histological examination (Fig. 4A Table 2. A total of nine pancreatic tumor sections were histologically examined; six were insulin-positive tumors and three were insulinnegative tumors, with mean tumor sizes of 1.  PET and SPECT imaging techniques with various radiolabeled exendin-4-based probes have been investigated in an attempt to localize insulinoma [14][15][16][17][23][24][25] . Among several radiolabel types, fluorine-18 is one of the most favorable radioisotopes for nuclear medicine imaging owing to its low positron emission energy and potential for high resolution and wide use in today's clinical settings 16,18 . However, most previous reports on 18 F-labeled exendin analogs demonstrated relatively high non-specific uptakes in preclinical studies, mostly due to the labeling methodologies 16,26,27 . Notably, PEGylation became an alternative owing to its ability to increase www.nature.com/scientificreports/ the molecular weight of the probe and its stability in circulation as well as improvement of the probe uptake by the tumor cells and leakage prevention from the tumor 20 1). In comparison with [ 18 F]FB-exendin-4, our PEGylated probe showed improved tumor uptake and higher contrast with surrounding organs including kidneys, although the available data obtained from the previous report regarding [ 18   www.nature.com/scientificreports/ of tumor per kidney that were comparable to our PEGylated probe 31,32 . In view of insulinoma visualization, the advantages and disadvantages of [ 18 F]FB(ePEG12)12-exendin-4 were summarized in Table 3.
Identifying a suitable mouse model for the preclinical evaluations of imaging probes in the localization of insulinoma remains an issue. INS-1 xenograft mouse models have been conventionally used for this purpose. Indeed, in most preclinical studies, using radiolabeled exendin-4-based probes for subcutaneous and/or intramuscular transplantation of INS-1 cells have been accomplished 18 . However, the visualization of INS-1 xenografts and tumor probe uptake were affected by their transplantation site and circumstances, which might be largely different in the case of endogenous insulinomas. As a mouse model of endogenous insulinoma, a mouse strain with rat insulin gene-2 promoter used to drive transgenic expression of simian virus 40 large T antigen (Rip-Tag2) was applied in a limited number of the previous studies 11, 33 . Although Rip-Tag2 mice are well-studied   22 . We thought that this mouse strain could be the useful model for differential imaging evaluation of our novel probe in functional panNENs. In the immunohistochemical study, insulin-positive pancreatic tumors exhibited positivity on [ 18 F]FB(ePEG12)12-exendin-4 PET/CT imaging and showed significant uptake (Fig. 4, Table 2); however, immunohistochemically, insulin-negative and glucagon-positive pancreatic tumors were negative on [ 18 F]FB(ePEG12)12-exendin-4 PET/CT images but positive on [ 68 Ga]DOTATOC PET/ CT images, suggesting the selectivity of our probe. These results are consistent with those of GLP-1R and SSTR subtype 2 (SSTR2) mRNA expressions, in which insulin-positive tumors showed significantly high GLP-1R and relatively low SSTR2 mRNA expression levels (Fig. 2). Therefore, Pdx1-Cre;TRp53 R172H ;Rb f/f is a suitable model in preclinical investigations of exendin-4-based probes for the specific detection of insulinoma.  www.nature.com/scientificreports/ In our investigation of [ 18 F]FB(ePEG12)12-exendin-4 PET/CT scans of Pdx1-Cre;Trp53 R172H ;Rb f/f mice, the scans at 30-and 120-min post-injection demonstrated different visualization states and probe uptake, which led to the best detection ratios in combined analyses of both scans (Fig. 4, Table 2). The scan at 120-min postinjection showed higher detection ratios among insulin-positive tumors than at 120-min post-injection. As previously reported, 18 F-labeled exendin-4-based probes have features of more rapid clearance from the kidney than other radiometal labeled probes 18,35,36 . In addition, our in vitro analyses showed the internalization and stable retention of [ 18 F]FB(ePEG12)12-exendin-4 probe in the cells; whole-cell and cell-internalized probe accumulated even at 120 min of incubation following the rapid increase within 30 min (Fig. 5B,C). These characteristics might contribute to a better contrast between the tumors and the surrounding organs in delayed-phase scans. Michalski et al. reported that a single 60-min early-phase [Nle14,Lys40(Ahx-DOTA-Ga-68)NH2]exendin-4 ([ 68 Ga]DOTA-exendin-4) PET/CT scan is sufficient for localizing insulinoma, and dual-time-point scans at 60 and 120 min after injection of [ 68 Ga]DOTA-exendin-4 had no additional diagnostic value 39 . However, their scan at 60-min post-injection might not be an early-phase one. Moreover, a longer half-life than gallium-68 (109.8 min vs. 67.7 min) and different chemical modifications could affect the optimized time of scans. Although our results suggested that the scan at 120-min post-injection and/or repeated scans might have potential benefit on insulinoma detecting in Pdx1-Cre;Trp53 R172H ;Rb f/f mice, further clinical investigations are warranted to confirm the optimal scan protocol for detecting insulinoma using our probe.
Finally, our study has limitations; the relatively small number of mice used in the imaging analysis might impede more definitive conclusions about the protocol being optimal for our analysis of the [ 18 F]FB(ePEG12)12exendin-4 PET/CT scans for the detection of insulinoma. However, since most of the Pdx1-Cre;Trp53 R172H ;Rb f/f mice develop insulinoma only at the age of over 180 days and those mice die within 2 weeks after presenting with hypoglycemia as previously reported 22 , it is challenging to employ a larger number of the mice to complete our comparative protocol between [ 18 F]FB(ePEG12)12-exendin-4 and [ 68 Ga]DOTATOC PET/CT scans. In addition, this is a preclinical study and further clinical investigations are needed in order to confirm the utility and the optimal scan protocol of [ 18 F]FB(ePEG12)12-exendin-4 PET/CT. Moreover, a dynamic study for 120-180 min might provide additional information on the whole body pharmacokinetics and reveal the tumor uptake kinetics and kidney clearance that would allow optimization of the scanning time point in terms of the image contrast. Further investigations including comparative studies with other exendin-4 based probes are warranted to confirm any benefits over other chemical modifications of exendin-4.
In conclusion, employing [ 18 F]FB(ePEG12)12-exendin-4 PET/CT for the detection of insulinoma yielded favorable sensitivity and specificity in the INS-1 xenograft mouse models and the Pdx1-Cre;Trp53 R172H ;Rb f/f mice with endogenous insulinoma and glucagonoma.

Materials and methods
Animals. Male ddY and BALB/c slc-nu/nu mice were purchased from SLC Japan (Hamamatsu, Japan). Male Pdx1-Cre;TRp53 R172H ;Rb f/f mice were obtained from the established colonies 22 . The mice enabled us to examine the PET probe's specificity to both insulin-positive and -negative neuroendocrine neoplasms in the intact pancreas at one scan. We measured non-fasting blood glucose levels of Pdx1-Cre;Trp53 R172H ;Rb f/f mice weekly using the glucose oxidase method (GT-1670; Arkray, Kyoto, Japan) and only the mice with blood glucose levels of < 80 mg/dL were enrolled in this study because such mice were expected to develop insulinoma 22  Ribonucleic acid isolation and real-time polymerase chain reaction. Total ribonucleic acid (RNA) was extracted from each of three samples, cultured cells, snap-frozen pancreas, and enucleated pancreatic tumors of Pdx1-Cre;Trp53 R172H ;Rb f/f mice (n = 10) using an RNA isolation kit (Qiagen, Valencia, CA, USA) 38 , following the manufacturer's instructions. For complementary deoxyribonucleic acid synthesis, 1 μg total RNA from each sample was reverse transcribed using a High Capacity RNA-to-cDNA Kit (Applied Biosystems, Alameda, CA, USA), following the manufacturer's protocol. SYBR Green polymerase chain reaction (PCR) Master Mix (Applied Biosystems) was applied for quantitative real-time PCR using an ABI StepOne-Plus Real-Time PCR System (Applied Biosystems, CA, USA). The signals of the products were standardized against glyceralde- isoflurane anesthesia using a Triumph LabPET12/SPECT4/CT (TriFoil Imaging Inc., Chatsworth, CA, USA), as previously described 12 . The image analysis including maximum intensity projection (MIP) reconstructions was performed using the Amira software version 5.6.0 (FEI Visualization Sciences Group, Düsseldorf, Germany). Subsequently, the mice were sacrificed by cervical dislocation 60 min after injection, followed by immediate resection of the INS-1 tumors, and radioactivity measurements of the resected tumors were performed using an auto-well γ-counter (COBRA II, Perkin Elmer, Waltham, MA, USA). The radioactivity measurements were decay-corrected to the time of injection. Then, the tissue and tumor radioactivity levels were expressed as their uptake values per injected dose of the probe (%ID/g) 12 www.nature.com/scientificreports/ mice completing the protocol were analyzed since the most of the Pdx1-Cre;Trp53 R172H ;Rb f/f mice developing insulinoma die within 2 weeks after presenting with hypoglycemia as previously reported 22 .
Histological analysis of Pdx1-Cre;Trp53 R172H ;Rb f/f mice. After completion of the PET/CT procedures, mice were sacrificed by cervical dislocation, followed by immediate resection of the pancreas. Tissues were spread on filter paper and immediately fixed in 10% formalin at 4 °C. For the analysis of pancreatic tumors, serial formalin-fixed paraffin-embedded sections (4-μm thickness) were stained with anti-insulin and anti-glucagon antibody, as previously reported 39 . Briefly, primary rabbit polyclonal antibody was used for insulin detection at an antibody dilution of 1:100 (Cat# sc-9168; Santa Cruz Biotechnology, USA) and a mouse monoclonal antibody glucagon detection at an antibody dilution of 1:100 (Cat# ab10988; Abcam, USA). The secondary antibodies utilized for the assay were Alexa Fluor 488 goat anti-rabbit antibody at 1:200 dilution (Cat# A-11008; Thermo Fisher Scientific, USA) and Alexa Fluor 546 goat anti-mouse antibody at 1:200 dilution (Cat# A-11030; Thermo Fisher Scientific, USA), respectively. The slides were prepared for analyses under a fluorescence microscope (BZ-X700; Keyence, Osaka, Japan). The radioactivity measurements were decay-corrected to the time of incubation. For the measurement of the cell-internalized activity, the surface-bound radioactivity was removed by incubation with acid stripping buffer (50 mmol/L glycine-HCl/100 mmol/L NaCl, pH 2.8) at room temperature for 20 min. Then, the internalized activity was measured relative to the total radioactivity added.

In vitro internalization study of [
Statistical analysis. All data were expressed as the mean ± SEM. Statistical analyses were performed using a one-way analysis of variance with the Tukey-Kramer post hoc test and Student's or Welch's t-test. P-values of < 0.05 were considered statistically significant. The statistical analysis was carried out using SPSS Statistics 24 software (IBM Corp., Armonk, NY, USA) and JMP Pro ® , version 15.1 (SAS Institute Inc., Cary, NC, USA).