Longitudinal Assessment of ¹¹C-5-Hydroxytryptophan Uptake in Pancreas After Debut of Type 1 Diabetes

Adequate mass and function of pancreatic β-cells are crucial for maintaining normal glucose control, a fact that is painfully evident in type 1 diabetes (T1D). However, our current understanding of human β-cell mass relies mainly on nonrepeatable biopsy studies (e.g., from autopsy and pancreatectomy). Despite progress in surgical procedures, the risk of complication from biopsy of the pancreas is high (1,2). Hence, we have a very limited understanding of the longitudinal alterations and potential dynamic modulation of β-cell mass in normal physiology and T1D development and progression. The establishment of a noninvasive in vivo method for the quantification of β-cell mass would serve as an invaluable tool for increasing our understanding of diabetes pathophysiology.

The number of islets in the human pancreas is ∼1–2 million, with an absolute majority of islets with a diameter <50 µm. Notably, the islets with largest contribution to the total endocrine volume have a diameter of ∼120–150 µm. ¹¹C-5-hydroxytryptophan (¹¹C-5-HTP) positron emission tomography (PET) imaging cannot show individual islets due to the resolution of the PET scanner (a few millimeters) but provides an accurate and quantitative integrated signal of the entire islet volume in the pancreas (3–10). In previous studies, ¹¹C-5-HTP PET imaging has confirmed several important morphological findings, i.e., a large interindividual, but a small intraindividual, variation in ¹¹C-5-HTP uptake in the pancreas in subjects without T1D and a significant reduction of ¹¹C-5-HTP uptake in the pancreas in subjects with T1D without detectable C-peptide, examined >10 years after diagnosis (6).

The loss of β-cell function in T1D occurs over a period of several years both before and after clinical diagnosis. However, the major loss of β-cell mass is considered to occur in the peri-diagnosis period, and in most people with T1D the endogenous insulin production (measured as plasma C-peptide) has dramatically decreased within 5 years after diagnosis. The loss of insulin secretion capacity could tentatively be caused by 1) a specific immune-mediated loss of β-cells, the currently dominating view; 2) loss of function, i.e., degranulation and exhaustion of β-cells; 3) loss of β-cell phenotype due to dedifferentiation; or 4) a combination. β-Cells could potentially dedifferentiate into either other functional endocrine cells, i.e., α-cells, polyhormonal immature cells, or endocrine nonhormonal chromogranin A–positive cells. Dedifferentiation can be induced by β-cell exhaustion in in vivo experimental studies (11) and in vitro studies of isolated human islets (12) and is currently considered a detrimental process in type 2 diabetes (13). A specific immune-mediated destruction of the β-cells would cause a significant reduction of the total endocrine volume, since β-cells constitute ∼65% of the islet volume (14). A reduction of this magnitude would be possible to detect with ¹¹C-5-HTP PET imaging, as previously demonstrated (6) (Fig. 1). A dedifferentiation process would theoretically have a marginal effect on the total endocrine volume and, if so, most likely could not be detectable with ¹¹C-5-HTP PET imaging due to the lack of β-cell specificity (Fig. 1). ¹¹C-5-HTP is a serotonin precursor, which have a preferential uptake in β-cells, but is also retained within other endocrine cells in the pancreas, especially α-cells (5,9).

In the current study, we have performed repeated ¹¹C-5-HTP PET/MRI examinations in young adults with new-onset T1D over the first 2 years of disease with the aim to quantify changes from baseline in pancreas volume, perfusion, and uptake of ¹¹C-5-HTP (i.e., the endocrine pancreas). The observed imaging findings correlated with residual β-cell function as assessed by mixed-meal tolerance tests (MMTT) and long-term metabolic control as estimated by HbA_1c at the same time points.

This was an observational, longitudinal single-center study to investigate changes in pancreas endocrine volume, perfusion, and morphology during the progression of T1D during the first 2 years of disease. Individuals (n = 16 in total) were recruited within 6 months of T1D diagnosis until n = 10 participants completed the entire study. Study participants were examined four times by PET/MRI: at recruitment and then after 6, 12, and 24 months. Clinical examinations and assessment of β-cell function by MMTT and fasting blood samples were performed in connection with the imaging examination. The study was conducted at Uppsala University Hospital, Uppsala, Sweden. All study participants provided written informed consent prior to any study-related investigations. Study protocols were approved by the Uppsala County Ethical Review Board (2014/035), and the trial was performed in accordance with the guidelines established by the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice.

Sixteen individuals with newly diagnosed T1D (mean ± SEM duration 7.1 ± 0.7 weeks) were enrolled, with the aim of n = 10 completing all four PET/MRI examinations. The sample size for this trial was based on empirical considerations and the observation that the test-retest variability of ¹¹C-5-HTP uptake in the nondiabetic pancreas is 6.4% (3,6). Eligible participants were C-peptide–positive (>0.1 nmol/L) men and women age >18 years and diagnosed with T1D within the last 6 months. Apart from T1D, the patients were healthy with normal vital signs as assessed by the investigator. Pregnant or breast-feeding patients were excluded; a pregnancy test was administered prior to each PET/MRI scan. Further exclusion criteria were impaired renal function (plasma creatinine >110 µmol/L), fasting C-peptide <0.1 nmol/L at diagnosis, and the presence of any general contraindications to undergoing MRI (claustrophobia and metal implants including pacemaker).

β-Cell function was assessed with an MMTT, which was conducted after overnight fasting and at bed rest. Fasting plasma samples for HbA_1c, glucose, and C-peptide were taken prior to the test. Following oral administration of 6 mL/kg (maximum 360 mL) Resource Protein (Nestlé Health Science, Switzerland), plasma glucose and C-peptide were sampled after 15, 30, 60, 90, and 120 min.

Prior to the PET/MRI examinations, all participants were fasting for >4 h. Before each examination, all participants were asked questions for exclusion criteria and all female participants performed a pregnancy test. An intravenous catheter was placed in the median cubital vein for administration of radiopharmaceuticals. The participant was placed in supine position in an integrated SIGNA PET MRI (GE Healthcare) scanner allowing for simultaneous dynamic PET scanning and MRI sequences. The pancreas was positioned in the center of the PET axial field of view (25 cm) by a scout MRI scan.

First, a 10-min dynamic PET examination (time frames: 1 × 10, 8 × 5, 4 × 10, 2 × 15, 3 × 20, 2 × 30, and 6 × 60 s) was performed after intravenous administration of a target dose of 400 MBq ¹⁵O-H₂O (radiolabeled water, radioactive half-life 2 min) for assessment of the pancreatic blood perfusion at the basal state. Next, ∼5 MBq/kg ¹¹C-5-HTP was administered intravenously. Each individual was examined with ¹¹C-5-HTP for 60 min, and the PET list mode data were reconstructed into 30 frames (12 × 10, 6 × 30, 5 × 120, 5 × 300, and 2 × 600 s) with an iterative VPFX-S algorithm (three iterations, three subsets, matrix 256 × 256, Z-axis postfilter 3 mm) with all relevant corrections performed.

During the PET acquisitions, T1-weighted MRI sequences were performed to generate attenuation maps, axial T2 and T1 two-point Dixon images for pancreas segmentation and anatomical coregistration, and T1 two-point Dixon sequence for fat fraction maps. Pancreas fat content was measured with a dedicated Dixon scan using a three-dimensional spoiled six-point multi-echo gradient sequence. An in-house developed method was used to reconstruct water and fat data with correction for T2* decay. The anatomical MRI images were acquired during either breath hold or respiratory triggering in exhaled position; thus, they best matched the PET images (which were acquired during free breathing, without movement correction or respiratory gating).

Three of the participants were examined by Discovery ST PET/CT scanner (GE Healthcare) instead of PET/MRI for their initial visit (PETLo2) or their first two visits (PETLo1 and PETLo3). The PET examination was performed as described above while a low-dose abdominal computed tomography examination during exhalation (120 kV, Auto mA 10–80 mA) was performed to provide anatomical coregistration and attenuation correction for the PET images.

The PMOD software (PMOD Technologies, Zürich, Switzerland) was used for image analysis. The entire pancreas volume was segmented on trans-axial projections of T1 MRI images. A separate segmentation of a smaller region was drawn in the central part of the pancreas, which was applied to the fat fraction MRI images, for readout of fat content. The whole pancreas segmentation was applied to the ¹⁵O-H₂O and ¹¹C-5-HTP data sets. Summation images were generated to verify that the segmentation was correctly coregistered. If not, smaller manual corrections were made. In the case of ¹¹C-5-HTP, portions of the segmentation in the tail were in some cases omitted in case spill in from the renal excretion signal was evident.

The pancreatic perfusion (mL/min/g tissue) was calculated by application of the “Flow and Dispersion” one-tissue compartment model (1 TC) in the PKIN module of PMOD to the ¹⁵O-H₂O dynamic data of the pancreas. Segmentation of the descending aorta was used as input function (at least 10 voxels fully inside the aortic lumen). The total perfusion of the entire pancreas gland (mL/min) was obtained by multiplication of the perfusion with the pancreas volume as measured by MRI.

Pancreatic uptake of ¹¹C-5-HTP at each time point was normalized to the percentage of injected dose taken up per gram of pancreas (%ID/g). The total percentage of injected dose taken up by the pancreas (%ID) was calculated by multiplication of %ID/g by the pancreatic weight (g), which was assessed by MRI with the assumption that the relationship between pancreas mass and volume was 1. The 50- to 60-min time frame after administration of ¹¹C-5-HTP provides the optimal contrast for imaging of the islets of Langerhans in pancreas (4–8), and therefore the %ID_{50–60 min} was used as a marker for the volume of the endocrine pancreas (i.e., the islet volume).

Data were reported as ranges, averages, and SDs or SEMs where noted. Changes in end points over time (between time points) were assessed by one-way ANOVA (mixed effects analysis, α = 0.95) in GraphPad, version 8, for Mac (GraphPad Software, San Diego, CA).

Correlation between parameters was analyzed as a set of linear random-effects regression models, where, for each pair of variables, one serves as outcome and the other as independent variable. A random offset for patients was used. Thus, the model is

where x and y are the variables of interest, u is a patient-specific random effect (assumed to follow a centered normal distribution), and is the residual error. The results contain estimated values of β together with a 95% CI and a P value for the null hypothesis β = 0 (no association).

The data sets generated during and/or analyzed during the current study are not publicly available due to protection of the personal integrity of the study participants but are available from the corresponding author upon reasonable request.

The aim was to enroll n = 10 participants who would complete the entire study. Sixteen participants with new-onset T1D were enrolled in total, of whom n = 10 completed the full study (24 months), while n = 4 dropped out after the first PET/MRI examination, n = 1 after 6 months, and n = 1 after 12 months. Descriptive data are presented in Table 1.

All participants were C-peptide positive (>0.1 nmol/L) at the start of the study. B-cell function assessed by MMTT and expressed as the area under the curve (AUC) for C-peptide response decreased by 21.0 ± 47.5% with a large interindividual variability (Fig. 2A). Three participants (PETLo10, PETLo16, and PETLo17) exhibited substantial and sustained loss of β-cell function (−70.7, −82.1, and −56.3%, respectively) after 24 months.

Most participants exhibited close to normalized HbA_1c at the 6-month assessment, due to initiation of well-controlled insulin treatment (Fig. 2B). HbA_1c was then relatively stable from 12 months to 24 months (10.9 ± 11.2%), except for PETLo10, in whom HbA_1c persisted in the range of 100 mmol/mol due to disinclination to adhere to the insulin treatment regimen.

Mean ± SD pancreas volume was 50.2 ± 10.3 mL (range 33.0–76.8) at T1D debut and had a tendency to decrease after 24 months (−16.0% to 42.2 ± 14.6 mL [range 22.0–68.1]) (P = 0.098) (Fig. 3A). Most participants demonstrated a similar slow but persistent decrease in volume.

The mean ± SD pancreatic uptake of ¹¹C-5-HTP at diagnosis was 0.23 ± 0.10 %ID (range 0.071–0.49) (Fig. 2B). The interindividual range of ¹¹C-5-HTP uptake was approximately sevenfold for these n = 16 participants. The pancreatic uptake of ¹¹C-5-HTP after 24 months was 0.21 ± 0.14 %ID (range 0.069–0.51), i.e., a small decrease of −7.6 (P = 0.59). PETLo1 demonstrated high ¹¹C-5-HTP uptake (∼0.5 %ID) in all four PET examinations with very low variability over time (within 4%) (Fig. 3B and Fig. 4). On the opposite end of the spectrum, PETLo10 exhibited low pancreatic uptake of ¹¹C-5-HTP (∼0.07%ID) in all four examinations (Fig. 2B and Fig. 3B). Generally, all participants exhibited pancreatic uptake of ¹¹C-5-HTP with low intraindividual changes between examinations.

The mean ± SD for total pancreas perfusion was 83.9 ± 40.7 mL/min (range 37.7–197.1) at diagnosis, which decreased by −18.2% to 68.2 ± 37.2 mL/min (range 25.5–132.2) after 24 months (P = 0.34) (Fig. 3C). Pancreas fat fraction at diagnosis (2.2 ± 1.5% [range 0–5]) was similar to that at 24 months (2.5 ± 1.1% [range 1–5] (Fig. 3D).

Correlations were analyzed with use of all examination points at different time points and with correction for the fact that up to four data points originate from each individual. Pancreatic uptake of ¹¹C-5-HTP correlated with the long-term metabolic control as estimated by HbA_1c (P = 0.047) but not the β-cell function as assessed by MMTT (Fig. 5 and Table 2). Pancreas volume did not correlate with HbA_1c or β-cell function (Fig. 5). ¹¹C-5-HTP uptake in pancreas correlated with the pancreatic volume (P < 0.001) (Fig. 5). Total pancreatic perfusion correlated with both the pancreas volume (P < 0.001) and ¹¹C-5-HTP uptake in pancreas (P < 0.001).

Obtained results are intriguing and provide novel information important for our understanding of the endocrine pancreas in the crucial period after diagnosis of T1D. At the time of T1D diagnosis, we observed a considerable interindividual variation of ¹¹C-5-HTP uptake, here considered as a marker of the endocrine cells in the pancreas. The uptake in new-onset T1D patients spanned from the normal range to ∼20% of what we have observed in young healthy subjects (on average 0.4%ID) (6) (Fig. 6A). Notably, there was only a minor and nonsignificant decrease of pancreatic ¹¹C-5-HTP uptake over the following 2-year period after T1D onset—a finding that correlates well with the functional assessment of the β-cells. As previously demonstrated for group levels there is only a loss of fasting plasma C-peptide of ∼0.01 pmol/mL/month over a period of 2 years after diagnosis of T1D (15). Obtained results adhere nicely with these previous findings with an average 2-year reduction of stimulated C-peptide of 21%.

Several autopsy and pancreatectomy studies report a major reduction of insulin-positive cells already at onset of T1D in comparisons with healthy control subjects (16–24). Based on these results, a significant reduction in pancreatic uptake of ¹¹C-5-HTP in the subjects with even only marginally reduced stimulated C-peptide during the study period was expected. Notably, in three of the subjects examined we could detect a pronounced loss of stimulated C-peptide, ranging from 56 to 82%. Even so, the pancreatic ¹¹C-5-HTP uptake in all subjects remained virtually unchanged during the study period. Obtained findings argue against a major loss of β-cells during the important 2-year period after diagnosis of T1D and are in agreement with the recent findings, with use of optimized techniques, of a lack of circulating β-cell-specific DNA in subjects with recent-onset T1D (25). Moreover, the results presented are in line with most morphological studies published on the human pancreas in the immediate period after diagnosis of T1D reporting presence of a discrete number of immune cells mainly in the periphery of only a few islets and without infiltrative destructive insulitis as seen in the NOD mouse model (23,24). Also, no correlation has been found between remaining β-cell mass and the frequency of insulitis (24,26).

¹¹C-5-HTP is taken up by both exocrine and endocrine cells within the pancreas by large amine transporters but rapidly exits the cell unless further metabolized. Therefore, the retention of ¹¹C-5-HTP depends on ability to synthesize serotonin, which exocrine cells of the pancreas lack. In validation studies, in vitro studies using a mix of exocrine tissue and pancreatic islets show a clear correlation between ¹¹C-5-HTP retention and the number of pancreatic islets (6). Furthermore, pancreatic uptake of ¹¹C-5-HTP in subjects, without detectable stimulated C-peptide, examined >10 years after diagnosis of T1D show markedly reduced pancreatic uptake of ¹¹C-5-HTP with no overlap with individuals without diabetes (6) (Fig. 6). With results taken together, it is highly unlikely that the pancreatic uptake of ¹¹C-5-HTP would represent an unspecific retention within the exocrine pancreas. ¹¹C-5-HTP is thus a marker for the endocrine cells in the pancreas, of which the β-cells of the islets of Langerhans contribute the largest part. Other islet cell types, mainly the α-cells, as well as, for example, enterochromaffin and nerve cell, constitute the remainder of the pancreatic endocrine cells, which could exhibit uptake of ¹¹C-5-HTP. Other neuroendocrine cells in the gut and abdomen will similarly exhibit uptake of ¹¹C-5-HTP, but due to the distance to pancreas and the resolution of PET, such extrapancreatic binding will not spill over into the pancreas or confound the ¹¹C-5-HTP signal.

Preserved islet size distribution, but a marked reduction in islet density in subjects with recent-onset T1D, was recently reported in comparison with subjects without diabetes, and no further reduction in islet density occurred with increased disease duration (27). In fact, all studies indirectly reporting findings from which the number of islets per volume or area of pancreas can be extracted report a reduction in number of islets in the range of 30–70% in subjects affected by T1D in comparison with control subjects without diabetes (16,28–32). Notably, there is seemingly no difference when T1D patients with <3 months’ duration (n = 5) are compared with those with a disease duration of 1–3 years (n = 4) regarding the mean ± SEM area of insulin-positive cells (0.16 ± 0.06% vs. 0.19 ± 0.1%, respectively; P > 0.99) or number of islets (165 ± 59 vs. 140 ± 8; P = 0.72) (22). However, most islets in subjects with recent-onset T1D contain a significantly larger proportion of α-cells in comparison with that in subjects without diabetes, and in subjects with long-standing T1D almost all remaining cells within the islets stain positive for glucagon (16,27,33). Dedifferentiation of β-cells into α-cells in T1D is thus a potential, but not established, mechanism explaining the lack of longitudinal decline seen with ¹¹C-5-HTP in this study.

The presently available methods for measuring β-cell function all have their inherent limitations. The MMTT is the most commonly used assessment in clinical trials and triggers a vast insulin secretion via a combination of an initial small increase in blood glucose and the secretion of incretins from the gut (34). The MMTT is performed on a nonexhausted β-cell mass in subjects subjected to overnight fast and with a blood glucose level <10 mmol/L (mean ± SEM 7.1 ± 0.4 mmol/L)—if higher, the test should be rescheduled. In subjects without diabetes, insulin secretion during an MMTT is sufficient to stabilize blood glucose within a range of a few millimoles per liter and thereby to prevent postprandial hyperglycemia. In the current study, no correlation was found between stimulated plasma C-peptide during an MMTT and the ¹¹C-5-HTP uptake in pancreas. This observation adheres with the notion that the MMTT cannot discriminate between a large β-cell mass with suboptimal function and a small β-cell mass with preserved function. In contrast, a significant correlation was found between HbA_1c levels and the ¹¹C-5-HTP uptake. This may seem counterintuitive; however, in contrast to an MMTT, which shows stimulated insulin secretion capacity of a resting β-cell mass, HbA_1c is regarded as an estimate of average blood glucose control over a period of a few months depending on the turnover of erythrocytes. Therefore, it may be speculated that the total volume of the endocrine pancreas (i.e., the pancreatic uptake of ¹¹C-5-HTP) contributes to preservation of glucose metabolism by preventing periods of β-cell exhaustion.

The pancreas volume tended to decrease from T1D diagnosis to 24-month follow-up in this study. In a retrospective comparison with a previous PET study in subjects with >10 years’ T1D duration and age-matched control subjects without diabetes (6), we observe that the pancreatic volume in this study was already at the level of those with long-standing T1D (on average ∼50 mL) (Fig. 6B). Furthermore, the pancreas volume at diagnosis was almost one-half that of control subjects without diabetes, indicating that the pancreas volume was decreased already before diagnosis or was never as large as in subjects without diabetes to begin with. Similar results were recently obtained in a dedicated MRI study in a larger cohort of >50 individuals with T1D scanned up to 100 days after diagnosis in comparison with siblings without diabetes (35). The pancreatic volume did not correlate with either β-cell function (MMTT) or long-term glycemic control (HbA_1c), as in the current study. The pancreatic volume then exhibited a slow decline over the 1st year after diagnosis. Potential loss of the islets (∼1–2% of the pancreas) are not enough for explaining the reduction in pancreatic volume, which is usually attributed to atrophy and dysfunction of the exocrine pancreas likely related to one or more of, for example, dysregulated blood flow, widespread inflammation, and fibrosis (36).

The main limitation of the current study is the relatively low number of enrolled participants, n = 16, of which n = 10 underwent all imaging examinations. However, all participants demonstrated functional and imaging end point values within physiological range for all time points. Additionally, the longitudinal design means that all participants are their own control for assessment of change from baseline (T1D diagnosis). Nevertheless, these results must be verified in a larger study, potentially incorporating other PET techniques for β-cell or islet assessment or other peripheral markers of β-cell dysfunction.

Collectively, our findings with ¹¹C-5-HTP PET imaging and MMTT argue against major β-cell destruction in the 2-year period after diagnosis of T1D. Presented findings challenge the current view on disease progression in T1D, provide an explanation for the lack of therapeutic value in clinical trials targeting the immune system (37), and call for principally new approaches for prevention and cure of the disease. Furthermore, the current results point to the need of β-cell-specific PET tracers for enabling direct assessment of the pancreatic β-cells during the development of T1D. Such a technique could, in combination with a marker for the endocrine pancreas, such as ¹¹C-5-HTP, be used to test the hypothesis of dedifferentiation as the major fate of the β-cells in T1D.

Acknowledgments. The authors thank the National PET/3T MRI facility (Anders Lundberg, Marie Åhlman, Gunilla Arvidsson), Francisco Ortiz-Nieto, Lina Carlbom, and the Preclinical PET/MRI Platform for support during the conduct and analysis of this study. The expert technical assistance from research nurses Violeta Armijo Del Valle, Rebecka Hilmius, and Karin Kjellström (Uppsala University Hospital) is gratefully acknowledged. The authors thank Lars Lindhagen (biostatistician at Uppsala Clinical Research Center) and Johan Bring (biostatistician at Statisticon) for consultations regarding the statistical analysis.

Funding. The study was funded by Novo Nordisk Foundation, an EFSD/Novo Nordisk grant, the Ernfors Family Fund, Barndiabetesfonden, Diabetesfonden, the Sten A Olssons Foundation, Helmsley Charitable Trust, JDRF, Diabetesfonden, Diabetes Wellness Sweden, Swedish Research Council (2019-01415 and 2020-02312), Swedish Heart-Lung Foundation, and Science for Life Laboratory.

Duality of Interest. O.E. is an employee of Antaros Medical AB. H.A. is an employee and one of the founders of Antaros Medical AB. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.E., P.-O.C., O.K., and O.E. designed, analyzed, and interpreted the studies and wrote the manuscript. R.K.S., M.R., P.C., and H.A. analyzed the imaging study and contributed to writing the manuscript. O.E. and D.E. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.