GLP-1 and Glucagon Secretion from a Pancreatic Neuroendocrine Tumor Causing Diabetes and Hyperinsulinemic Hypoglycemia

Context:

Glucagon-like peptide-1 (GLP-1) is a gut peptide that promotes insulin release from pancreatic β-cells and stimulates β-cell hyperplasia. GLP-1 secretion causing hypoglycemia has been described once from an ovarian neuroendocrine tumor (NET) but has not been reported from a pancreatic NET (pNET).

Objective:

A 56-yr-old male with a previous diagnosis of diabetes presented with fasting hypoglycemia and was found to have a metastatic pNET secreting glucagon. Neither the primary tumor nor metastases stained for insulin, whereas the resected normal pancreas showed histological evidence of islet cell hyperplasia. We provide evidence that GLP-1 secretion from the tumor was the cause of hyperinsulinemic hypoglycemia.

Methods:

GLP-1 levels were determined in the patient, and immunohistochemistry for GLP-1 was performed on the tumor metastases. Ex vivo tissue culture and a bioassay constructed by transplantation of tumor into nude mice were performed to examine the tumor secretory products and their effects on islet cell function.

Results:

The patient had high levels of glucagon and GLP-1 with an exaggerated GLP-1 response to oral glucose. Immunohistochemistry and primary tissue culture demonstrated secretion of glucagon and GLP-1 from the tumor metastases, whereas insulin secretion was almost undetectable. Ex vivo coculture of the tumor with normal human islets resulted in inhibition of insulin release, and transplanted mice developed impaired glucose tolerance.

Conclusions:

This is the first description of glucagon and GLP-1 secretion from a metastatic pNET causing sequential diabetes and hypoglycemia. Hypoglycemia was caused by insulin secretion from hyperplastic β-cells stimulated by tumor-derived GLP-1.

Insulin hypersecretion from pancreatic neuroendocrine tumors (pNET) is the principle cause of spontaneous hypoglycemia in adults after excluding iatrogenic causes. In occasional cases, patients with type 2 diabetes (T2DM) can subsequently develop fasting hypoglycemia due to the development of a pNET (insulinoma), and there is one case reported of an insulinoma in a patient with type 1 diabetes (1).

Hypoglycemia in association with hyperinsulinemia, but in the absence of insulinoma, is termed noninsulinoma pancreatogenous hypoglycemia. Pancreatic histological features include β-cell hyperplasia and increased periductular islets, morphological findings similar to those seen in persistent hyperinsulinemic hypoglycemia of infancy or childhood. Hypoglycemia due to islet cell hyperplasia occurring in adults against a background of diabetes has been described, but is extremely rare with only four cases reported (2–5). More recently, noninsulinoma pancreatogenous hypoglycemia has been identified in patients after Roux-en-Y gastric bypass surgery associated with islet cell hyperplasia (6). In these Roux-en-Y gastric bypass subjects, excess insulin has been associated with increased glucagon-like peptide-1 (GLP-1); hypoglycemic symptoms typically emerge years after their surgery, suggesting an islet cell adaptation to prolonged stimulus.

GLP-1 is released from the ileum in response to glucose ingestion and is one of the incretin hormones, promoting the release of insulin from pancreatic islet cells. GLP-1 is derived from the same proglucagon gene product as glucagon but is expressed at very low levels in the glucagon-producing α-cells of the islets due to a lack of the enzyme proconvertase 1/3 (7). Neoplastic secretion of GLP-1 causing reactive hypoglycemia has been reported on one occasion from an ovarian neuroendocrine tumor (NET) (8) and has not been previously described in patients with pNET.

We describe a male patient with previously diagnosed diabetes who was subsequently found to have a glucagon-secreting pNET, which was the likely cause of hyperglycemia. He subsequently developed hyperinsulinemic hypoglycemia due to GLP-1 secretion from metastases. Insulin staining of the primary and metastatic tissue was negative, but we demonstrated high levels of GLP-1 secretion in the patient. We infer that the hyperinsulinemic hypoglycemia was the result of excessive stimulation of β-cells by GLP-1, which had consequently become autonomous. We further identified from the literature a number of cases where previously diabetic patients developed hypoglycemia due to metastatic pNET, in whom it is likely that the true etiology was glucagon and GLP-1 secretion from the tumor.

Case Report

A 56-yr-old Caucasian man with a body mass index of 26 kg/m² and no family history of note was diagnosed with diabetes in 2003 with two elevated fasting blood glucose levels (14.0 and 9.4 mmol/liter) and a glycosylated hemoglobin of 10.0%. Glycemic control was achieved with metformin and gliclazide.

Three years later, the patient developed diarrhea, dyspepsia, and weight loss. Initial investigation by endoscopy demonstrated gastritis, and the diarrhea improved after initiation of a proton pump inhibitor. Subsequent imaging revealed multiple somatostatin receptor-positive liver lesions, left paraaortic lymphadenopathy, and a 3-cm pancreatic lesion. Fasting gut hormone profile, taken while on a proton pump inhibitor, demonstrated increased secretion of glucagon [170 pmol/liter; normal range (NR), 1–50 pmol/liter], gastrin (>400 pmol/liter; NR, <40 pmol/liter), and pancreatic polypeptide (473 pmol/liter; NR, <301 pmol/liter) with raised chromogranin A (307 pmol/liter; NR, 0–41 pmol/liter). Somatostatin was normal (9 pmol/liter; NR, <150 pmol/liter). During this period, the patient suffered severe symptomatic hypoglycemia that persisted despite cessation of all diabetes treatment, with his capillary blood glucose measured as low as 1.8 mmol/liter.

A supervised fast was conducted and terminated at 18 h after the patient developed symptomatic hypoglycemia (glucose, 1.6 mmol/liter) with inappropriately high insulin (675.7 pmol/liter; NR, <180 pmol/liter) and C-peptide (1800 pmol/liter; NR, <5 pmol/liter) concentrations confirming hyperinsulinemic hypoglycemia.

Cytoreductive surgery was planned in two stages, commencing with a distal pancreatectomy, splenectomy, and right hepatectomy. Tumor material from the pancreas and liver demonstrated a well-differentiated neuroendocrine carcinoma with positive immunostaining for glucagon and somatostatin and weakly positive staining for gastrin. Neither the primary nor the metastatic tumor cells stained for insulin, but pancreatic islet cell hyperplasia of insulin-producing β-cells was noted with focal “peliosis”-like vascular ectasia and numerous ductuloinsular complexes (Fig. 1, B and C). After the resection of pancreatic tissue and despite the persistence of hepatic metastases, the patient had no further episodes of hypoglycemia, and his blood glucose levels returned to the diabetic range.

The demonstration of islet cell hyperplasia rather than insulin secretion by the NET was unexpected and prompted further investigation. Measurement of GLP-1 during a 75-g oral glucose tolerance test (GTT) revealed a basal level of 32.1 pmol/liter (NR, 30.7–86.6 pmol/liter) with a very high peak at 30 min (321.37 pmol/liter; NR, 54.6–112.1 pmol/liter). Further surgery was performed to resect the remaining liver metastases. The metastatic tumor cells expressed predominantly GLP-1 on immunohistochemistry (Fig. 1A), with low levels of glucagon, somatostatin, and gastrin. The patient is currently being treated with octreotide and radionuclide therapy for residual liver metastases.

Methods

After surgery to the hepatic metastases, the expression of pancreatic hormone mRNA by RT-PCR and immunohistochemistry for GLP-1 were assessed in the fresh tissue. The tumor was dispersed and secretory products were analyzed by determination of the hormones in tissue culture media during ex vivo cell culture. The effect of tumor secretion on islet cell function was further assessed both by coculture of tumor cells with fresh islets and by transplantation of tumor cells into nude mice as a form of bioassay, followed by assessment of the glucose tolerance of the mice. All detailed methodology for the in vitro experiments and in vivo animal work is available as Supplemental Data (published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org).

Hormone assays

All samples were assayed in duplicate. GLP-1 was measured by an established in-house RIA with a detection limit of 7.5 pmol/liter and intraassay coefficient of variation of 6.1%, as previously described (9). There is no cross-reactivity between the GLP-1 assay and other proglucagon products. Active GLP-1 was not specifically measured. Specific and sensitive RIA were used to measure gastrin, vasoactive intestinal polypeptide, pancreatic polypeptide, somatostatin, neurotensin (10), and enteroglucagon (11). Insulin was measured by a commercial assay (Siemens Healthcare, Surrey, UK) with inter- and intraassay coefficients of variation of 4.8 and 3.3%, respectively. The level of human-specific C-peptide in the culture medium was determined by immunoradiometric assay in triplicate (Diagnostic Systems Laboratories UK Limited, Kent UK).

Literature search

A literature search was undertaken in Medline using the following terms in different combinations: type 1 diabetes, type 2 diabetes, neuroendocrine tumor, insulinoma, glucagonoma, hypoglycemia, hyperinsulinemia, and nesidioblastosis. After evaluation of the abstracts, all relevant full papers were retrieved, and subsequent further examination of the references was undertaken.

Results

Ex vivo coculture studies

Tumor cells were cultured either alone or in the presence of human islet cells. Tumor cells expressed insulin at very low levels just above the threshold of detection. Pancreatic polypeptide expression from the tumor was equivalent to native islets, whereas somatostatin and proglucagon were expressed at levels reduced in relation to islets alone (Fig. 2A). There was a reduction in the amount of insulin expressed in cocultured tumor and islet cells compared with islet cells alone, indicating a paracrine effect of the tumor inhibiting insulin release. This inhibitory effect of the tumor cells on islet cell insulin secretion is supported by C-peptide assessment, which showed a 38% decrease in C-peptide in the culture medium of tumor and islet cells compared with human islets alone (P < 0.001; Fig. 2B).

In vivo animal studies

Intraperitoneal GTT demonstrated that nude mice transplanted with tumor cells alone had impaired glucose tolerance compared with mice transplanted with islet cells alone, with highly significant increases in glucose concentration at 30 min (P < 0.01), 60 min (P < 0.001), and 90 min (P < 0.001) providing evidence for an endocrine diabetogenic effect of the tumor cells on islet cell function (Fig. 3A). This diabetogenic effect was partially overcome by the cotransplantation of tumor and islet cells.

On postmortem analysis of the mice, there was reduced insulin staining in the transplanted islet cells in the mice transplanted with tumor and islet cells compared with mice transplanted with islet cells alone, again demonstrating the inhibitory effect of tumor cells on insulin production (Fig. 3, B and C). Mice transplanted with tumor cells alone or with tumor and islet cells both demonstrated GLP-1 and glucagon expression, indicating that GLP-1 synthesis occurred in the transplanted tumor cells. Pancreatic tissue from the mice did not show evidence of a change in morphology or insulin synthesis (data not shown).

Literature search

A total of 71 titles and abstracts were identified by the initial search criteria, from which 37 cases describing hypoglycemia or insulinoma or nesidioblastosis in patients previously known to have diabetes were identified in 35 publications. Full documents were retrieved for 18. We were able to divide these into the following groups, as shown in Table 1:

1. 1. 

   Clear evidence of insulinoma occurring in patients with diabetes with a single pNET and either immunohistochemical evidence of insulinoma or surgical cure (12 patients).

2. 2. 

   Hypoglycemia in patients with previous diabetes with histological evidence of islet cell hyperplasia but no other cause (4 patients).

3. 3. 

   Metastatic NET with initial diagnosis of diabetes and subsequent hypoglycemia due to endogenous hyperinsulinemia (7 patients including this case).

4. 4. 

   Others (2 patients), including one patient with ovarian NET, and one patient with MEN-1 and two distinct NET secreting glucagon and insulin.

In addition, there were 13 patients where data were insufficient or unobtainable, predominantly from the older literature.

A clear pattern emerged from our analysis. Of the 12 patients with a solitary insulinoma on the background of T2DM, the majority had surgical cure and, where reported, positive immunohistochemistry for insulin. In contrast, among the seven metastatic NET that were associated with hypoglycemia, there was no direct evidence of insulin secretion from the tumor because either immunostaining was negative or data were not supplied. Three of seven patients with metastatic NET had confirmed glucagonoma.

Discussion

We describe a patient with an initial presentation of diabetes who later developed fasting hypoglycemia. Secondary diabetes was not suspected at diagnosis, and he was considered to have T2DM. In hindsight, however, there were no predisposing risk factors for T2DM other than age of 56 yr; he was Caucasian, active and otherwise healthy, had a body mass index of 26 kg/m², and had no family history of T2DM. With the later identification of a pNET, there is substantial evidence that his initial diabetes was caused by glucagon secretion. First, initial investigations of the patient revealed elevated fasting serum glucagon levels. Second, immunohistochemistry demonstrated positive glucagon staining of the tumor cells. Third, ex vivo cell culture demonstrated glucagon and somatostatin expression, albeit at levels lower than in normal human islets. Fourth, we observed a paracrine inhibitory action of the tumor cells on normal human islet cell insulin expression. Fifth, coculture of tumor with human islet cells led to reduced C-peptide secretion into the cell culture media. Finally, transplantation of tumor cells into severe combined immunodeficiency mice led to impaired glucose tolerance. The pNET primary tumor was 3 cm and unlikely to have caused significant islet cell disruption. Although somatostatin was present in vitro, the circulating levels were normal. Taken together, these results make it likely that the patient's initial diabetes was caused by the pNET, primarily due to glucagon secretion.

There is unequivocal evidence for endogenous hyperinsulinemic hypoglycemia in the patient, both clinically and from laboratory glucose, insulin, and C-peptide levels taken during a supervised fast. However, our studies exclude significant insulin secretion by the tumor. Neither the resected primary tumor nor the metastases showed immunopositivity for insulin. Ex vivo studies demonstrated almost undetectable levels of insulin mRNA expression by the cultured tumor cells, low levels of C-peptide secretion, and negative insulin immunohistochemistry. Furthermore, the histological diagnosis of insulin-positive islet cell hyperplasia indicates that the pancreatic islets provided the source for insulin secretion. Hypoglycemia resolved after subtotal pancreatectomy, despite the presence of substantial residual liver metastases, suggesting that the original excess insulin secretion was derived from the pancreatic islets rather than the tumor tissue itself. Our data show that the tumor production of GLP-1 persisted, although the pancreatic islet cell population had been depleted and therefore hypoglycemia was overcome.

We contend that the islet cell hyperplasia was a consequence of very high postprandial levels of GLP-1, greater than those we have previously demonstrated in normal individuals (29) or after Roux-en-Y gastric bypass surgery (30). GLP-1 staining was confirmed in the fresh tumor tissue from the metastasis but could not be determined in the primary tumor for technical reasons. GLP-1 mRNA expression cannot easily be determined specifically because it is encoded by a single exon and the mature peptide is derived by differential splicing from the same gene as glucagon.

GLP-1 has trophic effects on pancreatic β-cells, stimulating β-cell proliferation and neogenesis (31). Although we cannot define the duration of GLP-1 excess, it is likely that for several years the patient had increased GLP-1 levels that potentially exerted a chronic trophic stimulus. Increased GLP-1 secretion is thought to be the mechanism for the development of islet cell hyperplasia and nesidioblastosis years after gastric bypass (6, 32, 33). GLP-1 acts as an incretin-stimulating glucose-dependent insulin secretion underlying the use of GLP-1 agonists (e.g. exenatide) in the treatment of T2DM (34). GLP-1 analogs (e.g. exenatide) are not known to cause hypoglycemia. Nevertheless, it remains likely that the fasting hypoglycemia in our case was caused by the long-term trophic effects leading initially to islet cell hyperplasia and subsequently to the development of autonomous function. The mechanism for this remains unclear, although prolonged hormone stimulation resulting in autonomous behavior can occur in endocrine tissues—for example, in tertiary hyperparathyroidism.

Todd et al. (8) described an ectopic GLP-1-secreting ovarian NET giving rise to symptomatic postprandial hypoglycemia in a patient with preexisting diabetes. It is unknown whether there was hyperplasia of the pancreatic islet cells. There are two previous reports of NET producing GLP-1 and GLP-2 (35, 36). Both patients had severe constipation due to reduced gut motility, attributed to GLP-1, and one patient had gut hypertrophy related to GLP-2. Neither had reports of severe hypoglycemia, although the authors noted low/normal glucose.

We identified 12 cases of benign solitary insulinoma occurring in T2DM. However, among the seven metastatic pNET where hypoglycemia developed on the background of previous diabetes (1, 23, 25–27) or other features of glucagonoma (24), there was no clear evidence that the tumors secreted insulin. Insulin staining was either negative or not reported. In three of seven, including our case, the NET was described and initially diagnosed as a glucagonoma (24, 26). Because GLP-1 and glucagon are derived by differential processing of the preproglucagon gene product by proconvertase, it is highly plausible that these two hormones may be produced by the same tumor. pNETs that present with hypoglycemia as large mass lesions or with metastases may well represent malignant insulinomas. In contrast, where hypoglycemia emerges in the context of previous diabetes with a metastatic NET, our analysis would suggest that this scenario might well be due to the release of GLP-1, rather than insulin, directly from the tumor. Neoplastic GLP-1 secretion should form part of the differential diagnosis in metastatic NET.

Conclusion

We provide the first description of combined glucagon and GLP-1 secretion from a metastatic pNET causing diabetes and hypoglycemia. The diabetes was likely related to pNET glucagon secretion, and the hypoglycemia was caused by insulin secretion from hyperplastic β-cells stimulated by metastases-derived GLP-1. This appears to represent a novel paraneoplastic endocrine syndrome.

Acknowledgments

We thank Professor Alan McGregor (King's College London) and Dr. John Miell (University Hospital Lewisham) for reviewing the manuscript, and the surgeon Prof. Mohamed Rela (King's College London) for facilitating access to surgical material.

This work was partially funded by a Summer Scholarship advance from Ipsen Limited, Slough, UK.

Disclosure Summary: The authors have nothing to declare.

Abbreviations

 
• GLP-1

  Glucagon-like peptide-1

 
• GTT

  glucose tolerance test

 
• NET

  neuroendocrine tumor

 
• NR

  normal range

 
• pNET

  pancreatic NET

 
• T2DM

  type 2 diabetes.

References