Monitoring β-Cell Survival After Intrahepatic Islet Transplantation Using Dynamic Exendin PET Imaging: A Proof-of-Concept Study in Individuals With Type 1 Diabetes

Intrahepatic transplantation of islets of Langerhans (ITx) is used in individuals with type 1 diabetes who have severe hypoglycemic episodes along with hypoglycemia unawareness or poorly controlled diabetes (1). The introduction of the Edmonton protocol allowed for successful implementation of this therapy in clinical care (2). Islet transplantation can achieve glycemic control in the nondiabetes range in individuals with type 1 diabetes, with full insulin independence in some, although islets of more than one donor are usually required to achieve this. Nevertheless, graft function often deteriorates over time, and after 5 years, at least 50% of the patients are back on insulin treatment (1,3,4). Despite this reduction in graft function, glycemic control remains better than before ITx in a vast majority of patients (1). Therefore, ITx is a viable approach for selected patients (1,5), although the necessity of immunosuppressive treatment to avoid rejection of the islet transplants and the large number of islets required to treat a single patient, combined with a shortage of donor pancreata/islets, limit the broader application of ITx for the treatment of type 1 diabetes (1). Therefore, the search for improved transplantation strategies and ways to reduce the number of islets needed per transplantation continues (6–10).

To improve ITx strategies, technologies that allow us to follow the fate of transplanted islets are potentially of great value. Currently, islet graft function is assessed using C-peptide and HbA_1c (1,4), which may indirectly reflect but do not actually measure β-cell mass (11). Biomedical imaging techniques hold significant promise to quantify islet and, more specifically, β-cell mass after ITx. Such information would aid the advancement of ITx approaches because it would allow us to measure the effect of medical interventions or transplantation/encapsulation strategies on islet survival, in addition to loss of function. MRI and positron emission tomography (PET) using ¹⁸F-fluorodeoxyglucose have been proposed for this purpose (12–14), but prelabeling with contrast agents (i.e., superparamagnetic nanoparticles) for MRI might impair islet engraftment (15), and the use of PET tracers for prelabeling does not allow for long-term follow-up of islet grafts because of the radioactive decay of radiotracers.

We have developed and validated a technology that does not depend on prelabeling; it allows specific imaging of viable β-cells using radiolabeled exendin, a glucagon-like peptide 1 (GLP-1) analog (16,17). We previously showed that pancreatic exendin uptake linearly correlates with β-cell mass in rodent models (17–19), and a proof-of-concept study in individuals with type 1 diabetes and healthy volunteers demonstrated reduced pancreatic tracer uptake in those with type 1 diabetes (17,20).

In this study, we aimed to assess the feasibility of quantifying β-cell mass in intrahepatic islet transplants in humans using [⁶⁸Ga]Ga-NODAGA-exendin-4 (⁶⁸Ga-exendin) PET imaging.

Thirteen adults with type 1 diabetes were included in the study conducted at the Department of Medical Imaging of Radboud University Medical Center (Radboudumc; Nijmegen, the Netherlands). Twelve individuals were recruited from the diabetes outpatient clinic of Leiden University Medical Center (Leiden, the Netherlands), and one was recruited from Lille University Hospital (CHU de Lille, Lille, France).

Study participants in the ITx group (n = 9) received intrahepatic islet grafts at least 3 months before enrollment. Islet graft function was biochemically proven by stimulated C-peptide >0.8 nmol/L. Participants in the control group (n = 4) were on the waiting list for ITx and had no intrahepatic islet grafts. None of the individuals in either group were or had been treated with synthetic exendin or dipeptidyl-peptidase IV inhibitors. Liver AST and ALT values >135 units/L and chronic kidney disease as defined by an estimated glomerular filtration rate (GFR) <30 mL/min/1.73 m² were criteria for exclusion.

All study procedures were performed at Radboudumc. This study was approved by the institutional ethics review committee of Radboudumc, and study participants provided written informed consent before participation. Initially, our intention was to include 10 individuals in the ITx group and five in the control group. This could not be realized because of the small number of individuals who underwent ITx and had islet graft function >0.8 nmol/L or were on the ITx waiting list. Enrollment was also complicated by the closing of research facilities and a temporary stop to enrollment because of the COVID-19 pandemic.

All individuals underwent a mixed-meal tolerance test (MMTT) to assess β-cell function. The MMTT was performed in the morning, preceded by an overnight 12-h fast, during which only water was consumed. Participants were asked to skip bolus insulin for a period of 6 h before the test. In addition, they were asked to inject approximately two-thirds of the regular long-acting insulin dose or, in case of an insulin pump, to set the basal rate at approximately two-thirds of the default setting. Before the MMTT, blood was drawn to determine fasting glucose, C-peptide, insulin, HbA_1c, kidney function, and liver enzymes. Participants consumed 6 mL/kg liquid meal (Nutridrink; Nutricia, Hoofddorp, the Netherlands) to a maximum of 360 mL within 5 min. Blood samples were collected at 0, 15, 30, 60, 90, and 120 min after ingestion to determine stimulated glucose, C-peptide, and insulin.

Glucose levels and basal and stimulated C-peptide levels were used to calculate the area under the curve (AUC) for glucose and C-peptide using Prism 5.03 software (GraphPad Software, Inc., San Diego, CA). To provide a better estimate of β-cell function than AUC_C-peptide alone, the ratio between AUC_C-peptide and AUC_glucose was calculated. The ratio of AUC_C-peptide to AUC_glucose and peak C-peptide measurements were then correlated with the results of the image analysis for each participant.

Image acquisition was performed on a Siemens Biograph 40 mCT time-of-flight PET/computed tomography (CT) system (effective resolution ∼3–4 mm). PET/CT acquisition for one of the control patients was unsuccessful because of a system failure. Participants fasted for at least 4 h before the PET examination, and insulin use was temporarily adjusted similar to the preparation for the MMTT. Dynamic PET images were acquired for 60 min starting immediately after a 100-s intravenous infusion of 1.2 MBq/kg ⁶⁸Ga-exendin (peptide dose 4–7 μg). Radiochemical preparation was performed as previously described (16). Image data were obtained with one bed position that included the liver in the field of view. For anatomical information and attenuation correction, low-dose CT without contrast of the abdomen was acquired.

Reconstructed PET/CT data were analyzed using PMOD version 3.508 software (PMOD Technologies, Zurich, Switzerland). The low-dose CT scan was used for the complete manual delineation of the hepatic volume. A cylindrical volume of interest (VOI) was placed fully in the lumen of the abdominal descending aorta to obtain an image-derived input function for the Logan plot, a graphical analysis technique (21). A spherical VOI was positioned in the erector spinae muscle to measure the background level of the tracer. Because exendin accumulation in the kidneys is high (16), and the right kidney is in the proximity of the liver, the spillover in the peripheral part of the liver was excluded from the hepatic VOI.

The standardized uptake value (SUV) in the liver of control patients was determined and averaged, resulting in a group average (hepatic SUV_controls). This was assumed to correspond to background uptake of ⁶⁸Ga-exendin. A cutoff level was then determined consisting of the hepatic SUV_controls plus twice the SD, as earlier described (22). This cutoff level (hepatic SUV_controls + 2SD = 1.7) was applied to all patients to identify hotspots with hepatic tracer uptake above this level. Subsequently, these hotspots within the hepatic volume were automatically identified and segmented by isocontouring. The dynamic uptakes of ⁶⁸Ga-exendin during all frames of the PET scan were then extracted for each individual and used as input for the Logan plot in the kinetic modeling module in PMOD (PKIN; PMOD Technologies). The blood signal as determined from the descending aorta was used as input function. The Logan plot allows the estimation of the total distribution volume (V_T) of the tracer (21). The V_T is defined as the ratio between the radiotracer concentration in target tissue (kBq/cm³) to that in plasma (kBq/mL) at equilibrium and thus provides information on the rate of the radiotracer uptake in the target tissue, compared to the background circulating tracer in blood. The combination of hotspot localization and information on tracer kinetics in those areas provides insight into tracer accumulation in intrahepatic islet grafts. The analysis of the dynamic PET data was performed as described in a previous study (22).

One of the patients treated with transplantation (ITx9) underwent a surgical liver biopsy at the time of a laparoscopic cholecystectomy performed for a symptomatic gallstone, 24 months after the last islet infusion. After fixation in formalin (formaldehyde 4%; Sigma-Aldrich, Saint Quentin Fallavier, France), sections of the liver tissue were immunostained for GLP-1 receptor (Developmental Studies Hybridoma Bank ref. no. monoclonal antibody 3F52 4ea 12/5/19 anti-mouse) and insulin diluted at 1/1,000 (Abcam ref. no. ab181547 anti-rabbit). After deparaffinization and rehydration of the paraffin sections, pretreatment in microwave was performed as described by Waser et al. (23). The slides were incubated overnight at 4°C with the primary GLP-1 receptor antibody, and after revelation, the anti-insulin antibody was incubated for 2 h at room temperature. GLP-1 receptor was revealed with secondary Alexa Fluor 594 (anti-mouse AF-594; cat. no. 11032; Invitrogen) and insulin with Alexa Fluor 488 (anti-rabbit AF-488; cat. no. 21206; Invitrogen) diluted at 1/800 and applied for 1 h at room temperature.

A Dixon sequence was used to map fat deposition in the liver. Hotspots identified on PET were compared with the mapped fat deposition to look for possible matches. In addition, the Dixon scan was used to qualitatively assess the presence of hepatic steatosis.

Proton magnetic resonance spectroscopy (¹H-MRS) was used to quantify hepatic lipid content. Spectra were obtained from a voxel positioned in the liver avoiding large vessels during a single breath hold in expiration. The MRS spectrum was obtained without water suppression. All MR experiments were performed on a 3 Tesla Siemens MR system (3T Magnetom Skyra; Siemens, Erlangen, Germany) using an 18-channel body-phased array coil positioned on the abdomen in combination with a 32-channel spine coil.

Acquired data of the groups are expressed as mean ± SD or median (interquartile range [IQR]). The Mann-Whitney U test was used to assess group differences, and data are shown as median (IQR). Correlations between V_T and ratio of AUC_C-peptide to AUC_glucose and islet equivalents (IEQs) were analyzed by the Pearson correlation coefficient (r) with a two-tailed ANOVA. The level of significance was set at P < 0.05. GraphPad Prism software was used for all analyses (GraphPad Prism 5 for Windows).

The data sets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Patient age, BMI, diabetes duration, and renal function were comparable between the two groups (Table 1). Individuals treated with ITx underwent one or more islet transplantations within 5 years before enrollment, and the number of transplanted islets ranged from 0.5 to 1.3 million IEQs. Of these nine individuals treated with ITx, five were insulin independent at the time of ⁶⁸Ga-exendin PET imaging. Compared with patients not undergoing transplantation, those treated with ITx had significantly lower HbA_1c (P = 0.02), higher stimulated peak C-peptide values (P = 0.003), and a higher ratio of AUC_C-peptide to AUC_glucose (P = 0.003) determined in the MMTT (Table 1). One of the participants was scanned twice, once before ITx and once 9 months after (Table 1). In this participant, HbA_1c fell from 7.7 (61 mmol/mol) to 6.5% (47 mmol/mol), and stimulated peak C-peptide and ratio of AUC_C-peptide to AUC_glucose increased from 0.42 to 2.17 nmol/L and 0.02 to 0.20 nmol · min/L, respectively.

After injection of ⁶⁸Ga-exendin, two patients experienced nausea and vomiting. No other adverse events related to study procedures were observed. One participant was hospitalized because of pneumonia, but this was deemed unrelated to study procedures.

Figure 1A shows an example of a transaxial projection from a ⁶⁸Ga-exendin PET/CT scan displaying distinct regions with strong hepatic uptake. The hepatic SUV_controls was 1.3 ± 0.2, which was in line with the hepatic SUV determined in a ventral VOI in individuals with type 1 diabetes from a previous study (0.86 ± 0.25) (25). Using a hepatic SUV_controls + 2SD resulted in a cutoff value of 1.7. Regional uptake above this threshold value was defined as a hotspot. The V_T of ⁶⁸Ga-exendin within defined hotspots was significantly higher in patients treated with ITx than in control patients (median 0.55 [IQR 0.51–0.63] vs. 0.43 [0.42–0.45]; P = 0.009) (Fig. 1B). The individual who was scanned twice showed an increase in V_T from 0.45 to 0.53 (Fig. 1B). No correlation was found between the V_T of the tracer and the IEQs of the transplant (Pearson r = 0.18; P = 0.64) nor between the V_T and the ratio of AUC_C-peptide to AUC_glucose (Pearson r = 0.15; P = 0.70) (Fig. 2A and B).

In all patients, multiple hepatic hotspots were found on the PET images through qualitative assessment, except in one control patient showing no hotspots on the PET scan (participant C4). The ITx group had a numerically higher number of hotspots (5.9 ± 4.3; total 53) than the control group (2.0 ± 2.0; total 6), but this difference failed to reach statistical significance (P = 0.11).

In the patient who underwent liver biopsy 24 months after the last ITx, immunohistochemical analyses of the biopsy tissue indicated the presence of insulin-positive cell clusters (Fig. 3). Anti–GLP-1 receptor staining showed that these insulin-positive cells expressed the GLP-1 receptor, which is required for exendin uptake.

To study whether the presence of islet mass correlates with regions with enhanced lipid content in the liver, we sought to map hepatic liver fat distribution using a modified Dixon sequence and compare these MR scans with the locations of the identified hotspots on the PET images (Fig. 4). Based on the Dixon maps, regions of enhanced fat deposition were identified in two patients treated with ITx and one control patient. Each fat deposition in these patients matched with hotspots identified on the PET images. Furthermore, the Dixon scans were used to qualitatively assess hepatic steatosis, which was observed on both examinations in the individual who was scanned twice and, in another patient, treated with ITx. These patients were, however, different from the individuals who had fat depositions in the liver. Based on these findings, no clear relationship between the PET hotspots and hepatic fat depositions on Dixon scans could be observed.

Intrahepatic fat was also determined with another approach, ¹H-MRS, and was expressed as the relative lipid content (Eq. 1). An overview of the results can be found in Table 2. All individuals had a normal quantity of intrahepatic fat (≤5.6%) except one participant in the ITx group. This patient with high intrahepatic fat content also demonstrated hepatic steatosis on the Dixon scan. Figure 5 illustrates the ¹H-MR spectrum of normal and high quantities of intrahepatic fat. There was no difference in intrahepatic fat between the ITx and control groups (P = 0.86).

Here, we provide the first clinical evidence that ⁶⁸Ga-labeled exendin uptake can be used for quantification and monitoring of intraportally transplanted islets. Quantification of the identified hepatic hotspots demonstrated a higher uptake of ⁶⁸Ga-exendin in individuals with transplanted islets than in individuals not treated with ITx. Circulating C-peptide was higher in individuals treated with ITx, which proved the functionality of the islet grafts. In addition, in one patient treated with ITx, GLP-1 receptor expression was demonstrated in islet grafts in liver tissue by immunohistochemistry.

We observed no correlation between the V_T of ⁶⁸Ga-exendin in the identified hotspots and the IEQs of the transplant. Based on data from ¹⁸F-fluorodeoxyglucose–labeled islets, a majority of intrahepatically transplanted islets fail to engraft and die shortly after transplantation (14). Furthermore, a large individual variability in graft survival in the peritransplantation period has been demonstrated. Because imaging was performed 6 to 68 months after transplantation, the transplanted IEQs do not necessarily reflect the viable islet mass at the time point of scanning (17), explaining the lack of correlation of exendin uptake with administered IEQs. Also, the V_T of ⁶⁸Ga-exendin did not directly correlate with functional parameters, which is not surprising, considering that the presence of viable β-cells does not necessarily mean that these cells are functional (or function equally well) (20,26,27). Our observations highlight the importance of an imaging biomarker reflecting β-cell mass to complement functional measures of islet graft performance.

Among the clinically available tracers to date, radiolabeled exendin-4 offers the highest sensitivity and specificity for β-cells (17–19,28–30). It is one of the few islet cell tracers that have been evaluated in clinical trials, with [¹¹C]C-5-HTP being the only other method having been used to monitor intraportally transplanted islet grafts (22). However, [¹¹C]C-5-HTP is a biomarker for endocrine cells rather than specifically for β-cells because uptake and retention are mediated by the serotonin biosynthesis pathway present in most endocrine tissues, including all islet cell subtypes (31). The potential of exendin to image transplanted human β-cells has been previously suggested with planar scintigraphy after intramuscular islet autotransplantation (32). Exendin is β-cell specific because of the strong expression of the GLP-1 receptor on β-cells. Although some GLP-1 receptor expression has been reported in islet δ-cells (33), these cells account for maximally 5% of islet mass (34,35), and potential overestimation of β-cell mass as a result of exendin uptake in δ-cells is most likely not relevant. Ductal cells can express the GLP-1 receptor as well, but the expression level is even lower than that in δ-cells (33), and because these cells are spread out over the whole organ, they will only mildly contribute to background activity.

To reliably quantify the uptake of ⁶⁸Ga-exendin in intrahepatic islets grafts, we used a previously described method (22) that reduces the influence of background signal and improves the detection of focal accumulations of islets in the liver, indicated by the significant increase in V_T of the tracer in the hotspots of patients treated with ITx. In addition, because radiolabeled exendin is cleared via the kidneys (36), substantial differences in GFR might influence tracer distribution and uptake. To mitigate this potential impact, we excluded individuals with an estimated GFR <30 mL/min/1.73 m² and included control patients with comparable kidney function. The focal distribution pattern we observed in the liver was in line with previous studies demonstrating an uneven islet distribution rather than a homogeneous distribution across the liver (14,37). Besides the background uptake in the blood pool, some receptor-mediated tracer uptake might theoretically also occur because human hepatocytes can express low levels of the GLP-1 receptor (38). However, in a previous study with 62 participants (including those with type 1 and type 2 diabetes), no uptake of ⁶⁸Ga-labeled exendin was observed in the liver (39). Similarly, ⁶⁸Ga-DOTA–labeled exendin-4 demonstrated hepatic uptake lower than that of blood in individuals with type 2 diabetes (40). The low background activity in the liver renders our approach a sensitive and reliable technique for detecting GLP-1 receptor–rich islet grafts.

For an imaging biomarker to reflect true β-cell mass, expression of the target needs to be stable and should not be affected by physiological or metabolic changes. A concern when using exendin can be the stability of GLP-1 receptor expression and its effect on image quantification (41). However, our preclinical validation studies have demonstrated that exendin uptake in islets is reproducible (42); it linearly correlates with β-cell mass (17–19,28), and uptake is not affected by inflammatory processes (18,19). The latter is of particular relevance because type 1 diabetes is an autoimmune disease, and recurrence of autoimmunity after ITx is one of the factors associated with loss of graft function (4). Only high blood glucose levels have been demonstrated in humans to downregulate GLP-1 receptor expression (43,44) and thus exendin uptake in human islets, an effect that is restored after normalization of blood glucose levels (41). None of the individuals in the present study experienced severely elevated blood glucose levels in the week before imaging, and five of nine individuals treated with ITx were insulin independent.

Another important aspect that should be considered is radiation exposure, in particular, radiation to the islets. Dosimetry studies of ⁶⁸Ga-exendin have demonstrated that the pancreatic islet–absorbed dose clearly remained <2 mGy. Considering that the risk of diabetes increases in adults receiving more than ∼10 Gy to the islets, with higher sensitivity in children (45,46), the radiation exposure of ⁶⁸Ga-exendin PET remains more than 5,000 times lower, so even multiple PET/CT scans could be performed in adults annually without risk (16).

Insulin secreted by transplanted islets can promote local fat deposition by focal hyperinsulinism caused by hypersecreting islet grafts (1,47–49). Therefore, we hypothesized that accumulation of ⁶⁸Ga-exendin may coincide with local fat deposition in the liver. We indeed found that the number of hepatic hotspots was almost ninefold higher in the ITx group, but colocalizations between hotspots and focal fat deposition could be demonstrated only in three individuals, including one control patient. Therefore, liver fat depositions and hotspots did not colocalize as hypothesized, which may have been due to limited spatial resolution and sensitivity of the imaging techniques. Also, it is possible that fat depositions reside in areas where islets initially engrafted but were later lost, leaving just the fat deposition and no viable islets (and thus no radiotracer uptake). Finally, hepatic steatosis can be a reversible process, and only hotspots on PET would then be visible, without fat depositions (48,50,51). Hepatic steatosis was observed in the individual who was examined before and after ITx and in the patient treated with ITx with the highest percentage of intrahepatic fat on ¹H-MRS. There was, however, no apparent relation between ITx and hepatic steatosis. Intrahepatic fat was also measured with ¹H-MRS but was not different between the groups.

In conclusion, our study indicates that radiolabeled exendin imaging can be used to monitor viable islet mass after intrahepatic islet transplantation in humans. Ideally, ⁶⁸Ga-exendin PET imaging might be used alongside functional measures. Combined, these parameters could significantly aid the evaluation of strategies designed to improve islet engraftment, survival, and function.

Acknowledgments. The authors thank the technicians for their assistance and help with the PET/CT acquisitions, MRI, and ¹H-MRS. The authors also thank the laboratory technicians for their help with the radiolabeling and assistance in the laboratory. The authors express their greatest gratitude to the study participants for their valuable contribution.

Funding. This work is supported by BetaCure (FP7/2014–2018; grant 602812) and the Diabetes Foundation The Netherlands Fellowship (2015-81-1845). This work is also supported by IMI2-JU under grants 115797 (INNODIA) and 948268 (INNODIA HARVEST). This joint undertaking receives support from the Union’s Horizon 2020 Research and Innovation Program as well as the European Federation of Pharmaceutical Industries and Associations, JDRF, and the Leona M. and Harry B. Helmsley Charitable Trust. O.E.’s position is supported by the Science for Life Laboratory and the Swedish Research Council (2020-02312).

Duality of Interest. O.E. is an employee of Antaros Medical AB and a cofounder of Antaros Tracer AB. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. T.J.P.J. drafted the manuscript with the assistance of M.Bu. T.J.P.J., M.Bu., M.Bo., and M.v.d.G. helped with the conduct of the study. T.J.P.J., M.Bu., M.v.d.G., S.v.K., M.-C.V., A.B., F.P., I.V., O.E., and M.G. performed data analysis. M.Bu., M.Bo., M.Br., B.E.d.G., M.A.E., O.E., E.J.P.d.K., and M.G. contributed to conception and design. M.F.N. performed patient screening. All authors contributed to the interpretation of the acquired data. All authors were involved in the revision process and approved the final version. M.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this work were given as oral presentations at the Annual Meeting for North Europe Young Diabetologists, Lyngby, Denmark, 8–10 May 2019, and the Annual Congress of the European Association of Nuclear Medicine, 22–30 October 2020. Parts of this work were also presented as poster presentations at the European Molecular Imaging Meeting, Glasgow, Scotland, 19–22 March 2019; the Annual Meeting of the European Association for the Study of Diabetes (EASD), Barcelona, Spain, 16–20 September 2019; and the Annual Meeting of the EASD, 21–25 September 2020.