In Vivo Visualization of β-Cells by Targeting of GPR44

The trans-membrane G-protein–coupled GPR44 has recently been reported as being highly restricted to the β-cells in human pancreas by proteomic and transcriptomic screening (1,2). GPR44 is also designated as the prostaglandin D₂ (PGD₂) receptor 2, chemoattractant receptor-homologous expressed on Th2 lymphocytes receptor, or CD294. GPR44 binds endogenous PGD₂, and, because PGD₂ has been associated to asthma, GPR44 has previously been identified as a potential pharmaceutical target; antagonists of GPR44 are in clinical development. The exact role of GPR44 in β-cell physiology is currently unknown.

Recently, tritiated AZD3825, which is a high-affinity GPR44 antagonist, was evaluated for its selectivity for β-cells in human pancreas (3). [³H]AZD3825 binds to human islet preparations and β-cells with nanomolar affinity and specificity, whereas the binding profile in the exocrine pancreas was virtually nonexistent. These results indicated that AZD3825 could act as lead compound for development of a positron emission tomography (PET) ligand against GPR44.

AZ12204657 (a structural analog of AZD3825) was thus labeled by ¹¹C (4). In vitro evaluation of binding of [¹¹C]AZ12204657 to human pancreatic homogenates from islets or exocrine tissue showed similar results as [³H]AZD3825 (4).

The aim of the current study was to evaluate if the GPR44-targeting PET ligand [¹¹C]AZ12204657 is β-cell selective in vitro and in vivo in order to determine its potential as a surrogate imaging biomarker of β-cell mass in vivo.

[¹¹C]AZ12204657 batches (n = 21) were produced at Karolinska Institutet, with >95% purity as described previously (Fig. 1) (4). For the pig PET/MRI studies and the in vitro binding experiments, the tracer batches were transported to Uppsala University by car. The transportation took ∼40 min door-to-door. The specific radioactivity ranged from 615 to 3,342 GBq/µmol at end of synthesis, corresponding to 85–392 GBq/µmol at Uppsala University.

Three different GPR44 antagonists were used in order to evaluate the GPR44 receptor specificity of [¹¹C]AZ12204657 in different in vitro and in vivo models: AZ12204657, AZD3825, and AZ8154. The rational for using multiple compounds was to provide evidence of GPR44 displacement with several structurally similar but not identical antagonists. AZD3825 is a structural analog of AZ12204657, which is the unlabeled version of [¹¹C]AZ12204657. AZ8154 is from a different chemical series.

The potency of the antagonists at human GPR44 has been determined in vitro by quantifying the ability of AZ12204657, AZD3825, and AZ8154 to displace binding of [³H]PGD₂ from membranes of HEK293 cells transfected with human recombinant GPR44. In this assay, the obtained pIC₅₀ (half maximal inhibitory concentration) values for AZ12204657, AZD3825, and AZ8154 were 8.6, 9.5, and 8.4, respectively (corresponding to IC₅₀ values of 2.5, 0.32, and 4.0 nmol/L, respectively) (data not shown).

Pancreatic biopsies were collected from organ donors without diabetes (n = 6), with type 2 diabetes (T2D; n = 6), and with type 1 diabetes (T1D; n = 6). Additionally, a splenic biopsy was obtained from an organ donor (n = 1). The use of human tissue was approved by the Uppsala Ethical Review Board (#2011/473, #Ups 02–577, #2015/401, and #2016/465) and tissues obtained from Uppsala Biobank.

Pancreatic biopsies were also collected ex vivo from nondiabetic pigs (n = 2), streptozotocin (150 mg/kg i.v.) diabetic pigs (n = 2) (see Nalin et al. [5] for details), and a nondiabetic rat. A spleen biopsy was obtained from a nondiabetic pig. The use of pig and rat tissues collected ex vivo was approved by the Animal Research Ethical Committee of the Uppsala Region and performed according to the Uppsala University guidelines on animal experimentation (UFV 2007/724).

All biopsies were stored at −80°C, embedded in Tissue-Tek (Sakura Finetek Europe B.V., Leiden, Netherlands), and processed into 20-µm slices. Sections were preincubated in 100 mL 50 mmol PBS (pH 7.4) for 10 min. Then, 0.1 MBq/mL corresponding to ∼1 nmol [¹¹C]AZ12204657 (n = 6 separate experiments) was added, and the sections were incubated for 30 min at room temperature. Nondisplaceable binding was assessed by coincubation with 20 μmol/L AZD3825.

Subsequently, tissue sections were washed three times for 2 min each in 150 mL 50 mmol PBS at 4°C before being dried at 37°C for 5 min. The slices were then exposed against a phosphor-imager screen for 40 min, scanned using a Cyclone Plus Phosphor imager (PerkinElmer), and analyzed using ImageJ (National Institutes of Health). Regions of interest (ROIs) were drawn over the entire pancreatic or splenic sections as well as over regions corresponding to islets of Langerhans and exocrine tissue. Tissue binding in becquerels per gram was normalized for the incubation activity as becquerels per cubic centimeter and expressed as a unitless measurement of enrichment (similar to standardized uptake values [SUVs] for in vivo PET examinations). Specific binding was defined by subtracting nondisplaceable binding from total binding.

Adjacent sections from human pancreases were stained for insulin using Insulin A SC-7839 (Santa Cruz Biotechnology, Dallas, TX; goat-polyclonal 1:1,000). The sections were then incubated with secondary antibody Alexa Fluor 488 (Invitrogen, Carlsbad, CA; donkey anti-goat; dilution 1:100). Tile scan images were acquired with a Zeiss LSM780 confocal microscope (Carl Zeiss, Oberkochen, Germany).

Adjacent sections from animal pancreases were stained for insulin using A0564 (Agilent Technologies, Santa Clara, CA) as primary antibody. The stains were developed by the Envision DAB system K4010 (Agilent Technologies) using an anti-rabbit secondary antibody. All slices were counterstained using Mayer’s hematoxylin. The slides were captured digitally using a camera mounted on a Leica DMLB2 microscope (Leica Microsystems, Wetzlar, Germany).

Immunodeficient nu/nu Balb/c mice (male, n = 11; wt 36.0 ± 0.4 g) were transplanted with 500 human islets of Langerhans to the right kidney capsule as described previously (6). The study in mice was approved by the Animal Research Ethical Committee of the Uppsala Region and performed according to the Uppsala University guidelines on animal experimentation (UFV 2007/724). Human islets were isolated as previously described (7). The use of isolated human islets was approved by the Uppsala Ethical Review Board (#2011/473 and #Ups 02–577) and obtained through the Nordic Network for Islet Transplantation.

For organ distribution studies, the mice were divided into two groups: one group received only [¹¹C]AZ12204657 tracer (n = 6), whereas the other group received a pretreatment of i.v. 4 mg/kg AZD3825 (in 100 μL PBS 10 min prior to tracer) (n = 5). Wake mice were administered 2.5 ± 0.4 MBq [¹¹C]AZ12204657 i.v. in the tail by a 100-μL bolus.

After 40 min, the mice were euthanized by excess CO₂. Organs and tissues were immediately harvested, weighed, and measured for radioactivity in a γ counter (Uppsala Imanet, Uppsala, Sweden). The radioactive uptake was decay corrected to the time of injection and expressed as percent of injected dose (%ID/g) by correcting for the weight of the animal.

Nonhuman primates (cynomolgus monkeys, females, n = 4; wt 5–8 kg) were used for PET studies at Karolinska Institutet. The study was approved by the Animal Research Ethical Committee of the Northern Stockholm Region and was performed according to the “Guidelines for planning, conducting and documenting experimental research” of Karolinska Institutet.

PET measurements were conducted using either a High Resolution Research Tomograph (HRRT; Siemens Molecular Imaging, Malvern, PA) (n = 2) or Siemens Biograph 64 PET/CT (Siemens Molecular Imaging) (n = 2).

Anesthesia was induced by intramuscular injection of ketamine hydrochloride (∼10 mg/kg) and maintained by the administration of a mixture of isoflurane (1.5–2.0%), oxygen, and medical air after endotracheal intubation (HRRT measurement) or by a continuous i.v. infusion (1 mL/kg/h) of ketamine (4 mg/mL) and xylazine (0.4 mg/mL; PET/computed tomography [CT] measurement). Body temperature was maintained by a Bair Hugger model 505 (Arizant Healthcare, Eden Prairie, MN) and monitored by an esophageal thermometer. Electrocardiogram, heart rate, noninvasive blood pressure, oxygen saturation, and respiratory rate (under gaseous anesthesia) were continuously monitored throughout the experiments. Fluid balance was maintained by a continuous infusion of saline.

Each nonhuman primate was examined two times in one day. One of the nonhuman primates was examined with tracer alone on an additional experimental day.

For all examinations, the animals were positioned on the PET bed with pancreas in the center of field of view (FOV). Attenuation correction was performed by a transmission scan of 6 min using a single ¹³⁷Cs or a low-dose CT scan for the HRRT and the PET/CT scans, respectively. Each PET examination was performed for 93 min after i.v. injection of 138 ± 64 MBq [¹¹C]AZ12204657 either as tracer alone or 20 min following i.v. pretreatment with GPR44 antagonist AZD3825 (1 mg/kg). After the dynamic PET/CT examinations (n = 2), a 25-min static multibed scan from head to feet (8 min × 3 bed positions) was performed to assess the whole-body distribution of [¹¹C]AZ12204657.

For HRRT, data were reconstructed using the ordinary Poisson three-dimensional ordered-subset expectation maximization (OSEM) algorithm, with 10 iterations and 16 subsets including modeling of the point spread function. For PET/CT, data were reconstructed using an OSEM algorithm, with four iterations and eight subsets and using a 5-mm Gaussian filter.

Venous blood samples (1 mL) were acquired 4, 10, 20, 30, 60, 90, and 120 min following [¹¹C]AZ12204657 administration. The samples were immediately measured for radioactivity content as well as for blood glucose and then centrifuged at 4°C (4,500 rpm, 5 min). The blood plasma was separated and measured for radioactivity.

High-health herd-certified pigs (Yorkshire × Swedish Landrace × Hampshire; male, n = 5; wt 30–41 kg) were transported by car from the university farm on the morning of each experiment. The study was approved by the Animal Research Ethical Committee of the Uppsala Region and performed according to the Uppsala University guidelines on animal experimentation (UFV 2007/724).

Anesthesia was induced by an intramuscular injection of 5 mg/kg tiletamine and zolazepam (Zoletil Forte vet. 250 mg/mL; Virbac Laboratories, Carros, France) combined with 0.05 mg/kg medetomidine (Domitor vet. 1 mg/mL; Orion Pharma Animal Health, Sollentuna, Sweden) and 0.1 mg/kg butorphanol (Dolorex vet. 10 mg/mL; Intervet AB, Stockholm, Sweden). Before intubation, a bolus dose of 2 µg/kg fentanyl (Fentanyl B. Braun 50 μg/mL; B. Braun Medical AB, Danderyd, Sweden) was injected i.v. The anesthesia was maintained with an i.v. infusion, starting at a dose of 3 mg/kg/h continuously adjusted to achieve adequate anesthesia. During the PET examination, pigs were connected to an MRI-safe ventilator and ventilated with 30% oxygen in nitrogen. Oxygen saturation, heart rate, end-tidal carbon dioxide, electrocardiogram, noninvasive blood pressure, and rectal body temperature were recorded with a B40 patient monitor (GE Healthcare, Uppsala, Sweden). Fluid homeostasis was maintained with a continuous infusion of Ringer acetate (Fresenius Kabi AB, Uppsala, Sweden).

An arterial catheter was placed in the femoral artery and a venous catheter was placed in the right and left auricular vein for anesthesia and tracer infusion, respectively.

The pigs were examined up to three times per day in a Signa PET/MR scanner (GE Healthcare). The instrument is equipped with a cryo-cooled 3T magnet and allows for simultaneous PET/MR measurements.

Each PET examination was performed as tracer dose baseline studies (n = 6 as defined as <1 µg administered mass of AZ12204657), as coinjection dose-escalation studies (n = 4; 3–70 µg administered tracer mass, to elucidate the mass effect), or following pretreatment of a high dose of GPR44 antagonist AZ8154 (n = 3) (Table 1). AZ8154 pretreatment was performed to verify the GPR44 specificity of [¹¹C]AZ12204657 in vivo with a separate compound for AZ12204657.

Each pig was positioned with pancreas in the center of FOV of the PET/MR scanner by assistance of scout MR anatomic scans. For the initial baseline examination in each pig, [¹¹C]AZ12204657 was injected i.v. (150–250 MBq, corresponding to 0.15–0.74 µg [or 0.004–0.024 µg/kg] tracer mass as described in Table 1 depending on the specific radioactivity of the individual batches).

A dynamic PET sequence over 90 min was started simultaneously with the activation of the infusion pump. Additionally, an arterial blood drawer (SwissTrace; swisstrace GmbH, Menzingen, Switzerland) connected to a coincidence counter (1 Hz) was started just before tracer administration for continuous monitoring of arterial whole blood concentration of radiotracer. The radioactivity in the arterial blood was logged in the PMOD SwissTrace module (PMOD Technologies Ltd., Zurich, Switzerland). The arterial blood sampling was stopped after 10 min.

Arterial blood samples (2 mL) were acquired during the examination (5, 10, 15, 20, 30, 45, 60, and 90 min following [¹¹C]AZ12204657 administration). The samples were immediately measured for radioactivity content as well as for blood glucose and then centrifuged at 4°C (4,500 rpm, 5 min). The blood plasma was separated and measured for radioactivity.

The data sets were reconstructed into a 33-frame data set according to the following frame sequence: 12 × 10 s, 6 × 30 s, 5 × 2 min, 5 × 5 min, and 5 × 10 min.

Images were reconstructed using an iterative OSEM VUEPOINT-FX (GE Healthcare) algorithm (2 iterations, 28 subsets, and a 5-mm Gaussian postfilter, incorporating time-of-flight information). An MR anatomical examination was used for attenuation correction.

After the end of the baseline examination, a second and, in some cases, a third PET examination was performed during the course of an experimental day (Table 1). Subsequent examinations were performed at least 2 h after the preceding [¹¹C]AZ12204657 administration, giving sufficient time for the ¹¹C radioactivity to decay. In the case of dose-escalation studies, the tracer dose of [¹¹C]AZ12204657 was spiked with 3–70 µg/kg unlabeled AZ12204657 in order to inject increasingly higher doses of AZ12204657 to reduce to possible impact of remaining functional AZ12204657 in the biodistribution. In the case of preadministration of 5 mg/kg AZ8154, this was performed 30 min prior to [¹¹C]AZ12204657 administration and PET examination. A total of 5 mg/kg AZ8154 dissolved in 30–35 mL vehicle was administered i.v. over 10 min either manually or by an infusion pump.

The pigs were terminated at the end of the experimental day after up to three PET examinations.

The specific radioactivity for the baseline tracer examination for pig 1 was not accurately determined and thus excluded from the mass-effect calculation. Furthermore, the baseline tracer examination for pig 3 exhibited unusual tracer kinetics in all tissues and vessels (extremely strong apparent accumulation with total radioactivity within the PET FOV exceeding total injected radioactivity [i.e., not realistic]), which was assumed to be due to errors in scanner performance or reconstruction and thus was also excluded.

A diffusion weighted image (DWI) MRI protocol was performed in parallel for each PET examination in order to estimate the perfusion of the pig pancreas during the different dose-escalation and pretreatment examinations, according to the method described by Le Bihan et al. (intravoxel incoherent motion imaging) (8). The DWI protocol acquired images using different B-values as follows: B₀, B₂₀, B₅₀, B₇₅, B₁₅₀, B₄₅₀, and B₉₀₀.

The same pancreatic and splenic volumes of interest as used for the PET analysis were transferred to the DWI images. Additionally, a volume of interest outside of the pig was used to generate the background signal.

The background values were subtracted from the pancreatic values for each B-value in each scan. Then, the output was plotted against the B-values and the resulting curve fitted to the biexponential equation below, where x = B-values, f = perfusion (volume) fraction (or fraction of fast diffusion), D = true diffusion (mm²/s), and P = fast diffusion (perfusion-related diffusion) (mm²/s) in the software ORIGIN v. 8.6 (OriginLab, Northampton, MA).

Reconstructed PET data from pig and nonhuman primate examinations was analyzed using MATLAB (MathWorks Inc., Natick, MA) and SPM8 software (Wellcome Department of Cognitive Neurology, London, U.K.). Pancreatic and splenic ROIs were delineated on sequential coregistered images (MRI T2* in the case of the PET/MRI examinations and CT images for the PET/CT examinations). In the case of the standalone PET HRRT examinations, the pancreas was identified on early summation images (0–2 min). Pancreatic voxels exhibiting strong PET signal spill-in effects in the first 20 min postinjection from neighboring organs (left kidney and small intestines) were excluded from the ROI. This approach typically left a smaller (nonhuman primates) or bigger (pigs) part of the tail of pancreas in the analyzable ROI.

Image-derived arterial input curves were obtained automatically by identifying the hottest voxels in the early frame with maximum radioactivity, corresponding to voxels within the descending aorta. The aortic voxels were then pruned using the late radioactivity summation image to exclude voxels close to organs with high levels of radioactivity at late times (e.g., near the liver). The aortic input was corrected for the plasma-to-whole blood ratio as derived from the discrete blood sampling. For the pig studies, the continuous arterial whole-blood draw was used instead of image-derived aortic data for the first 10 min.

The total distribution volume (V_T) of [¹¹C]AZ12204657 in tissue was evaluated by the multilinear version of Logan’s linear graphical analysis of regional or voxel-wise PET data using the radioactive signal in arterial plasma as input. Voxel-wise analysis was performed using plasma input–based wavelet-aided parametric imaging (9).

The mass effect of [¹¹C]AZ12204657 in pancreas was assessed by calculating the in vivo K_d, defined as 50% of specific binding competed away. The tracer microdosing concept is fulfilled at administered doses of radiopharmaceutical that cause <5% receptor occupancy. Therefore, we also determined the [¹¹C]AZ12204657 dose corresponding to 5% reduction in binding (mass-effect dose) compared with the pancreas baseline in the dose escalation. Any dosing below the mass-effect dose will exert negligible mass effect and adhere to the tracer microdosing concept.

All data were reported as means ± SD. The normal distribution of data sets was assessed by the Shapiro-Wilk normality test (GraphPad Prism v6.0 for Mac; GraphPad Software, La Jolla, CA). Differences between groups (assessed as following normal distribution) were assessed by either a two-tailed, unpaired t test with Welch correction (not assuming equal SDs) (used for all in vitro autoradiography except for those noted below, immunodeficient mice, and pigs) or a two-tailed, paired t test (used for nonhuman primates) using a confidence level of 0.95.

In case normality could not be ensured, an unpaired, nonparametric Mann–Whitney test using a confidence level of 0.95 was used instead (in vitro autoradiography total binding versus block for pancreas from humans without diabetes or with T2D and human spleen).

Pancreatic uptake of [¹¹C]AZ12204657 in human pancreas sections from individuals without diabetes was heterogeneous and concentrated to the islets of Langerhans (Fig. 2A and B). Similarly, focal uptake of [¹¹C]AZ12204657 in pancreas from individuals with T2D corresponded to insulin-positive islets (Fig. 2C and D). No [¹¹C]AZ12204657 uptake was seen in pancreatic sections from individuals with T1D in whom insulin staining confirmed β-cell deficiency (Fig. 2E and F). No binding was seen in rat pancreas (intrinsically devoid of GPR44 expression; data not shown).

The enrichment in entire pancreatic sections (which is analogous with what is measured in a PET scanner) was displaceable by coincubation with a GPR44 blocker in excess in human pancreatic sections from individuals without diabetes (P < 0.05) and with T2D (P < 0.05) but not individuals with T1D (Fig. 2G). The enrichment in the islets of Langerhans was pronounced and significantly higher than the enrichment in the exocrine background (P < 0.01) (Fig. 2H). The enrichment in islets from individuals without diabetes or with T2D was comparable (i.e., there was no change in binding in individual islets in T2D).

In pancreatic sections from nondiabetic pigs (Fig. 3A and B), but not streptozotocin diabetic pigs (Fig. 3C), there was a heterogeneous binding pattern of [¹¹C]AZ12204657. Insulin staining on consecutive sections showed that the focal uptake corresponded to islets in the nondiabetic pig pancreas (Fig. 3D and E). No insulin-positive areas were seen in the pancreas from streptozotocin diabetic pigs, consistent with the lack of a focal binding pattern of [¹¹C]AZ12204657 in the corresponding autoradiogram.

Figure 4 compares the biodistribution in immunodeficient mice transplanted with human islets under the kidney capsule, either [¹¹C]AZ12204657 alone or after coinjection with 1 mg/kg AZD3825, as assessed by ex vivo organ distribution. In vivo injection of [¹¹C]AZ12204657 alone resulted in accumulation of radioactivity in the islet graft under the kidney capsule, but not in the opposing control kidney (P < 0.01) (Fig. 4). Additionally, radioactivity uptake in the islet graft was competed away by coinjection of AZD3825 (P < 0.05), indicating specific binding to the GPR44 receptor in the human islet graft.

GPR44-mediated binding was also seen in lungs consistent with the known biodistribution of GPR44 in mice (10). No pancreatic compartment in mouse expressed GPR44, which was also reflected in the biodistribution, as no specific pancreatic binding of [¹¹C]AZ12204657 was recorded. The high uptake in liver and kidney was not influenced by pretreatment and likely represents excretion of [¹¹C]AZ12204657.

In nonhuman primates, [¹¹C]AZ12204657 was excreted mainly through the intestines at later time points, which caused spillover into the pancreas in time frames after 20 min following injection (Fig. 5A). This is a species-specific technical issue due to the small size of the nonhuman primate in relation to the PET/CT scanner gantry size and not transferrable to the human situation.

Therefore, only uptake in the pancreatic tail (most distant from intestine) was analyzed and then only during the initial 20 min after administration. The kinetics of total binding were sufficiently rapid to allow for quantification using only 20 min of time activity curve data. Accordingly, the t* on the Logan plot corresponded to the time frame at ∼10 min postinjection.

The uptake of [¹¹C]AZ12204657 in the pancreatic tail was consistently reduced by 32–77% by pretreatment with AZD3825 (P < 0.05) (Fig. 5B). The splenic signal could be inhibited in a similar manner and with a similar range. Representative parametric images (color coded for V_T) are displayed in Fig. 5C ([¹¹C]AZ12204657 baseline) and Fig. 5D (decrease in [¹¹C]AZ12204657 binding following GPR44 inhibition).

Representative parametric images (color coded for plasma input–based wavelet-aided parametric imaging V_T) following baseline tracer administration (<1 µg) of [¹¹C]AZ12204657 are shown in Fig. 6A. Pancreatic SUV was in the range of 0.5–1.5 at the end of each baseline examination. Dose escalation by spiking the tracer with unlabeled AZ12204657 compound decreased the V_T in pancreas as well as other abdominal tissues dose dependently. Representative parametric images are shown in Fig. 6B and C. Pretreatment by 5 mg/kg AZ8154 decreased the V_T in pancreas to background (Fig. 6D).

The in vivo K_d in pancreas (defined as 50% of specific binding competed away) was calculated to ∼2.9 µg of administered mass (Fig. 6E). Negligible mass effect (<5% occupancy) was seen at doses <0.4 µg, as calculated from the dose-escalation data (red line indicating V_T = 0.95 × 1.41 = 1.33 on the y-axis, corresponding to the mass-effect dose 0.4 µg on the x-axis) (Fig. 6E).

The pancreatic binding of [¹¹C]AZ12204657 could be almost completely inhibited by preadministration of 5 mg/kg AZ8154 (P < 0.01) (Fig. 6F).

Additionally, there was a strong binding to the pig spleen at the baseline examinations, which could be abolished by blocking GPR44 (P < 0.01) (Fig. 6F).

There was no apparent correlation between the AZ12204657 dose escalation and the pancreas perfusion as estimated by changes in the capillary volume fraction (linear correlation R² = 0.02; P = 0.66) (data not shown). There was, however, a tendency of increase in pancreatic perfusion compared with baseline conditions when pretreating with 5 mg/kg AZ8154 (P = 0.063) (data not shown).

The splenic binding of [¹¹C]AZ12204657 was examined in vitro in pig and human tissue in order to predict if the high in vivo spleen uptake in pig would translate to the human situation.

The pig spleen tissue exhibited strong in vitro binding when incubated with [¹¹C]AZ12204657 alone (Fig. 7A and E), which is congruent with the high splenic signal seen in vivo. The signal could be almost completely abolished by coincubation with 20 µmol AZD3825 (P < 0.05) (Fig. 7B). In human spleen, only a weak signal was seen (Fig. 7C–E).

This study shows the feasibility of quantitative pancreatic imaging of the β-cell restricted protein GPR44 in vivo by means of PET.

[¹¹C]AZ12204657 radioactivity in pig and nonhuman primate pancreas was relatively low, which is expected because only the β-cells express the GPR44 protein. Indeed, an SUV of 1 (which would indicate a low binding in a tissue where most cells express the target protein) would, in the setting of imaging a β-cell–selective target (1% of all cells), indicate highly significant binding (SUV >100 in the β-cells, disregarding nonspecific binding in exocrine pancreas for sake of simplicity). Hence, the modest pancreatic accumulation is in line with imaging a target expressed in only a subpopulation of pancreatic cells.

The spillover seen from intestines into pancreas in nonhuman primates (Fig. 5A) is an animal model–specific technical issue due to the small size (5–8 kg) of a nonhuman primate cynomolgus (and corresponding small distance between nonhuman primate intestine and pancreas) compared with the PET/CT scanner and its resolution (∼3 to 4 mm). In humans (15 times larger), with correspondingly larger distances between intestines and pancreas, the entire human pancreas will be clearly visualized, as spillover is not expected. For example, spillover from neighboring tissues into pancreas was negligible in pigs with sizes of 30–40 kg in the PET/MRI scanner with similar resolution.

Importantly, the pancreatic baseline could be dose dependently competed away by an excess of unlabeled AZ12204657 or by pretreating with selective GPR44 inhibitors, demonstrating the specific binding of [¹¹C]AZ12204657 to GPR44.

Similarly, the dose-escalation study in pigs confirms that [¹¹C]AZ12204657 can be used for assessment of GPR44. A mass effect of 50% receptor saturation was seen at ∼2.9 µg AZ12204657. Therefore, it is advisable to administer a maximal dose of 0.4 µg in future clinical studies.

The pretreatment and dose-escalation (target saturation) studies indicate that at least 80% of total binding of [¹¹C]AZ12204657 in pancreas is due to specific binding, and thus, the in vivo binding potential of the tracer is at least 4, which corresponds to a very high affinity of the radioligand with a K_d in the picomolar range.

Proof of principle of in vivo targeting of human islets was shown in a transplantation model in immunodeficient mice. There was a strong radioactive signal in the explanted human islet graft, which was not found in the nontransplanted kidney. Furthermore, pretreatment with AZ12204657 abolished the islet-specific uptake, demonstrating a GPR44-mediated uptake mechanism of [¹¹C]AZ12204657. Lung, a positive control tissue in mouse, similarly demonstrated GPR44-mediated [¹¹C]AZ12204657 binding.

We found high [¹¹C]AZ12204657 uptake in spleen in the pig and, to a lesser extent, in spleen in nonhuman primates during the baseline PET examinations. The uptake was receptor mediated, as confirmed by both in vivo and in vitro competition studies with GPR44 antagonists. The [¹¹C]AZ12204657 splenic binding was surprising because the Human Protein Atlas did not indicate any significant GPR44 in human spleen (11), and furthermore, no similar binding was seen in immunodeficient mouse. Thus, we performed in vitro autoradiography on human spleen sections, which showed negligible binding to GPR44 in human spleen. Thus, splenic binding of [¹¹C]AZ12204657 in future human PET examinations will likely be low.

The DWI-MRI perfusion measurements did not show any change in perfusion in either pancreas or spleen, following dose escalation of AZ12204657. Preadministration of AZ8154 showed a tendency for increase in pancreatic perfusion. Thus, it is possible that GPR44 antagonism in pharmacologically meaningful doses increases pancreas perfusion. However, a dose below or just above the microdosing range (relevant for PET studies of basal physiology) does not influence pancreatic perfusion.

The in vitro autoradiography study was performed to complement the in vivo PET examinations by yielding higher resolution data on regional tissue distribution as well as providing [¹¹C]AZ12204657 binding data in several species to allow for better understanding of the translational possibilities.

The binding of [¹¹C]AZ12204657 to human pancreases corresponded to insulin-positive islets of Langerhans, both in sections from subjects without diabetes and with T2D. Importantly there was no difference in GPR44 density in islets from subjects without diabetes and with T2D, as detected by [¹¹C]AZ12204657, which is a crucial feature of a surrogate imaging marker for β-cell mass. The contrast between islets and the exocrine background was in the range of 20-fold.

The summed pancreatic binding (both islets and exocrine regions—analogous to the measurements in a PET scanner, which cannot resolve between pancreatic compartments) in both subjects without diabetes and with T2D was displaceable (i.e., mediated by GPR44) in contrast to the background pancreatic binding in subjects with T1D. Autoradiography of pig pancreas sections exhibited islet-specific binding of [¹¹C]AZ12204657 corroborating the in vivo results and validating the pig as a suitable large animal model for further PET studies of GPR44 in diabetes. Importantly, the pancreatic binding of [¹¹C]AZ12204657 approached background levels in sections from streptozotocin diabetic pigs without β-cells. Notably, [⁶⁸Ga]Exendin4 revealed a strong glucagon-like peptide 1 receptor–mediated signal also in the streptozotocin diabetic pigs without β-cells (5). Thus, GPR44 provides a more accurate measurement of β-cell mass than [⁶⁸Ga]Exendin4 in pig models of diabetes.

Taken together, GPR44 expression on the β-cells is species dependent and of a similar magnitude in humans, nonhuman primates, and pigs, but is absent in rodents. Based on in vivo results in pigs and nonhuman primates as well as ex vivo studies of human biopsies, we conclude that [¹¹C]AZ12204657 specifically targets GPR44 in the pancreas and constitute a promising PET imaging surrogate tracer for assessment of pancreatic β-cell mass in humans.

Acknowledgments. The authors thank Anneli Rydén and Elin Manell at the Swedish University of Agricultural Sciences for pig anesthesia and members of the Karolinska Institutet’s PET group for support with experiments in nonhuman primates.

Funding. This study was supported by JDRF and the Helmsley Charitable Trust (3-SRA-2014-265-Q-R), Swedish Medical Research Council (K2015-54X-12219-19-4, K2013-64X-08268-26-3, K2013-55X-15043, and 921-2014-7054), Diabetes Wellness, the Ernfors Family Foundation, the Göran Gustafssons Foundation, Barndiabetesfonden, Diabetesfonden, and ExoDiab (Excellence of Diabetes Research in Sweden). AstraZeneca provided supplies for study.

Duality of Interest. O.E. is an employee of Antaros Medical AB. P.J., Z.C., M.S.W., and S.S. are employees of AstraZeneca. AstraZeneca provided financial support for this study. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. O.E. and O.K. designed, performed, analyzed, and interpreted the studies and wrote the manuscript. P.J., Z.C., and S.S. designed, analyzed, and interpreted the studies and contributed to writing the manuscript. M.J., R.K.S., M.J.-W., A.T., M.S.W., and C.H. performed the study and contributed to writing the manuscript. O.E. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.