Study participants

Study participants were recruited from the outpatient clinic at the Department of Surgery in the Radboud university medical center (Nijmegen, the Netherlands). All individuals had a minimum age of 18 years and were scheduled for (partial) pancreatic resection after being diagnosed with a pancreatic tumour (see Table 1). One of these individuals had dexamethasone-induced diabetes, others did not have diabetes. Recent blood samples that were taken for standard laboratory tests prior to surgery were used to assess kidney function and liver enzymes. All study procedures were performed in the Radboud university medical center. Participants were included from two clinical studies with similar inclusion and exclusion criteria (Clinicaltrials.gov ID NCT03889496 and NCT04733508). Although intended for more study participants, funding for the trial was terminated early as a consequence of the COVID 19 pandemic (NCT03889496). Another challenge was the inclusion of participants since we frequently encountered unexpected changes in surgery date affecting the planning of the study or changes in treatment plan resulting in no surgery at all (instead, most patients were included in clinical studies receiving chemotherapy), which have led to early termination of NCT04733508. Notwithstanding these unforeseen situations, the data acquired from both studies with the current number of inclusions are sufficient to answer our research question.

Table 1 Clinical characteristicsPreparation and labelling of 111In-labeled exendin-4

Preparation and labelling of [111In]In-[Lys40(Ahx-DTPA)]exendin-4 (where Ahx represents aminohexanoic acid) for clinical use was produced (piCHEM, Graz, Austria), dissolved and portioned according to GMP standards to a final concentration of 2 µg/ml in 1 mol/L HEPES (pH 5.5) containing 0.1% Tween-80. The vials were stored at − 20 °C. Radiolabelling was performed by adding exendin-DTPA to the vial containing [111In]InCl3 (Curium, Petten, the Netherlands) to a final concentration of 150 MBq 111InCl3 per µg [Lys40(Ahx-DTPA)]exendin-4 (molar activity of 72 GBq/µmol). After incubation at room temperature for 20 min 0.150 mL EDTA (3.75 mg/ml) was added. Due to regulation changes during the runtime of the project, additional purification steps were added to the procedure to remove the excess of HEPES. Therefore, the labelling procedure for the final patient included subsequent purification of [111In]In-[Lys40(Ahx-DTPA)]exendin-4 on an HLB cartridge (Waters, UK) and by gel filtration on a PD-10 column. The end product was sterilized by passing through a 0.2 μm filter (Millex GV) in a Hotcell Class A. Quality control was performed by reversed-phase high performance liquid chromatography (RP-HPLC) and instant thin layer chromatography (ITLC). Labelling efficiency was determined on Varian silicagel strips (ITLC-SG, Agilent Technologies, Amstelveen, the Netherlands) using 0.1 mM ammonium acetate buffer with 0.1 mM EDTA, pH 5.5 as the mobile phase. RP-HPLC was done using a C18 reversed-phase column (HiChrom alltima C18, Breda, the Netherlands). For elution, a linear gradient of 0.1% TFA in acetonitrile (3–100% over 10 min with a flow rate of 1 ml/min) was used. The labelling efficiency was 99.9% and the purification with HLB and PD-10 was necessary for the reduction of HEPES concentration.

Administration of 111In-exendin and image acquisition

After a 4-hour fasting period, participants received an intravenous injection with 151 ± 5 MBq [111In]In-Lys40(Ahx-DTPA)]exendin-4 (further referred to as 111In-exendin) (peptide dose of ~ 1 µg) as a slow bolus over 1 min. In the first seven participants, SPECT/CT images of the abdomen were acquired on a Siemens Symbia T16 SPECT/CT system, using the low-dose CT (without contrast) for anatomical correlation (total scanning time of approximately 45 min). For the SPECT acquisition, 128 views were obtained with an acquisition time of 40s per view. The acquired SPECT/CT images were reconstructed as previously described [6]. The other participants underwent surgery and the resected pancreatic tissue was used for microSPECT/CT imaging (MILabs, Utrecht, The Netherlands). A 1 h microSPECT/CT scan of the resected pancreatic tissue was performed using a 1.5-mm diameter pinhole rat collimator tube followed by a CT scan for anatomical reference. The microSPECT/CT was reconstructed according to the following parameters: 0.8 voxel size, 4 iterations and 128 subsets using pixel-based similarity-regulated ordered subsets expectation maximization (SROSEM) algorithm. A gaussian filter of 2.5 mm was applied to the microSPECT images.

Quantification of SPECT/CT images

Reconstructed SPECT/CT images were analysed using Inveon Research Workplace 4.1 software (Siemens Healthcare). Three spherical volumes of interest (VOIs) of 9.6 mm were manually placed in the pancreatic head, corpus, and between head and corpus. Positions for each VOI were localized on the CT scan. To determine the pancreatic uptake of 111In-exendin, the counts measured in the VOIs were corrected for the administered activity (MBq), the time after injection (tscan-tinjection), and the half-life of indium-111 (t1/2=2.81 days).

$$Exendin\,uptake\, = \, \over )}^} - })} \over }}}}}}}}$$

(1)

Autoradiography and immunohistochemistry of pancreatic tissue

Surgical resection of pancreatic tissue was done one day after injection of the radiotracer. On the day of surgery, frozen sections of 4 μm were obtained from the resected pancreatic tissue and fixated with ice cold acetone for 10 min before autoradiography. The dried sections were exposed to a phosphor imager screen (Fuji Film BAS-SR 2025; super resolution) for a period of 2 weeks.

The slides used for autoradiography (frozen tissue sections) were incubated with a guinea pig-anti-insulin antibody (1:750 in phosphate buffered saline (PBS) + 1% bovine serum albumin (BSA) ab195956, Abcam, Cambridge, UK) for 60 min at room temperature (RT) in a humidified chamber. Slides were then washed 3 times with 10 mM PBS and incubated with goat-anti-guinea pig-HRP (1:1000 in PBS + 1% BSA, A18775 Thermo Fisher) for 30 min at RT. After washing 3 times with 10 mM PBS, the bound antibodies were visualized using diaminobenzene (DAB, Bright DAB, BS04 Immunologic, VWR, Dublin, Ireland). Slides were counterstained for 5 s with hematoxylin (Klinipath, Olen, Belgium). Tissue was mounted with a cover slip using Permount (Fisher Scientific, Waltam, MA, USA).

Formalin fixed and paraffin embedded pancreas tissue was sectioned at 4 μm thickness. Sections were deparaffinized with xylene and rehydrated in ethanol. Antigen retrieval was performed with TRIS- buffered EDTA (TBE) + 0,05% Tween-20 in a PT Module (Thermo Fisher Scientific, Waltam, MA, USA) (10 min at 96 °C). Slides were washed 3 times with tris buffered saline (TBS) + 0.05% Tween-20 between all the incubation steps. Endogenous peroxidase activity was quenched with 3% H2O2 for 10 min, followed by an incubation of 30 min at RT with 20% normal goat serum + 0.1 M glycine in PBS in a humidified chamber. Sections were stained with Mouse-anti-GLP-1 receptor (1:10, DSHB Mab 3F52, Iowa City, IA, USA) for 2 h at RT. After washing, the slides were incubated with Bright Vision poly HRP-GAM/Rb IgG (Immunologic, VWR, Dublin, Ireland). The bound antibodies were visualized using diaminobenzene (DAB, Bright DAB, BS04 Immunologic, VWR, Dublin, Ireland). Slides were counterstained with hematoxylin (Klinipath, Olen, Belgium) and mounted with a cover slip using Permount (Fisher Scientific, Waltam, MA, USA).

Adjacent sections were stained for insulin and underwent antigen retrieval with 10 mM citrate pH 6.0 in a PT Module (Thermo Fisher Scientific, Waltam, MA, USA) (10 min at 96 °C). Slides were washed 3 times with 10 mM PBS between all the incubation steps. Endogenous peroxidase activity was quenched with 3% H2O2 for 10 min, followed by an incubation of 30 min at RT in 20% normal goat serum. The slices were incubated with guinea pig-anti-insulin (1:500, ab195956, Abcam, Cambridge, UK) for 60 min at RT. After washing, the slides were incubated with goat anti guinea pig-HRP (1:1000, A18775 Thermo Fisher) for 30 min at RT. The bound antibodies were visualized using diaminobenzene, counterstained and mounted as described before.

Image acquisition of tissue sections

Sections that were stained for insulin and the GLP-1 receptor were scanned with an automated microscope (Pannoramic 1000 (P1000)) using an objective magnification of 20x, resulting in pixel size of 0.24 μm (in x- and y-direction). The obtained digital images and the previously mentioned autoradiography images allowed to check for overlap between insulin-positive areas, GLP-1 receptor positive areas, and radiotracer uptake.

Endocrine-to-exocrine ratio

Ratios between endocrine and exocrine tissue were determined using the digital autoradiographic images analysed with MATLAB (R2018a; MathWorks, Natick, MA, USA). First, the complement of the autoradiographic image was computed so that high pixel values correspond with high uptake of 111In-exendin in the image. The average pixel value of the background was determined by manually drawing a region of interest (ROI) in the image. Exocrine and endocrine tissue were excluded from the ROI to exclude signal from exendin uptake in the tissue. Subsequently, six ROIs were drawn in different locations in the exocrine tissue to acquire the average pixel value from these exocrine regions. These ROIs were drawn in regions with low fat content. In total, 10 ROIs were positioned in endocrine regions (Islets of Langerhans) to again obtain the average pixel value from these ROIs combined. Locations of all ROIs were carefully placed in the image and were always based on the section stained for insulin after it had been used for autoradiography, to localize the desired regions (background, exocrine or endocrine tissue; see Supplementary Fig. 1 for graphical representation of the analysis). To correct for the background activity, the average pixel value from the background ROI were subtracted from the exocrine and endocrine ROIs. This enabled to calculate the endocrine to exocrine ratio corrected for the background signal.

Statistical analysis

Data were expressed as mean ± SD, median (interquartile range (IQR)), or number (%). Relationships between parameters were checked for linearity using the Pearson correlation coefficient (r), with a two-tailed analysis of variance. The level of significance was set at p < 0.05. GraphPad Prism software was used for all analyses (GraphPad Prism 10 for Windows, San Diego, California USA).