Human Insulinomas Show Distinct Patterns of Insulin Secretion In Vitro

Insulinomas are uncommon, usually benign, tumors of pancreatic β-cells, which cause hyperinsulinemic hypoglycemia (1–4). Early morphological studies (5,6) have led to distinct classifications of insulinomas on the basis of their histological and structural organization. They also suggested correlations between the ultrastructural appearance of the tumors and either their insulin content (5) or the efficacy of diazoxide in inhibiting excessive insulin secretion in the patient (6). However, a study based on a large series of 76 case patients reported that the two main morphological groups of insulinomas, trabecular and solid (or medullary), are not homogeneous but display highly variable insulin immunostaining patterns (7). These findings have been confirmed (8), leading to the conclusion that the histological structure of insulinomas is not a satisfactory marker of distinct functional properties.

The characteristics of insulin secretion by human insulinoma cells are poorly known. Several in vitro studies have addressed the question, but no clear picture of the differences with normal β-cells has emerged, for a number of reasons. First, the techniques used were disparate. Minced pieces of the tumor were incubated either immediately (9,10) or after culture for several days or weeks (11–13). In other studies, pieces of the tumor were digested with collagenase before the measurement of insulin secretion either immediately (14,15) or after 1–8 weeks of culture (16–20). Second, only a few, dissimilar protocols of stimulation or inhibition were usually tested in static incubations. Third, with a few exceptions (12,13), these functional studies were limited to single case patients.

Over a period of 10 years, we obtained a fragment from 10 insulinomas, of which we then characterized the dynamics of insulin secretion in vitro, using the same methods as used in our recent studies (21,22) of the pancreas of infants suffering from congenital hyperinsulinism. The results show heterogeneity in the responses but sufficient similarities to permit a tentative subdivision into three functional groups. In one of these, abnormal insulin secretion is attributed to the undue expression of hexokinase-I (HK-I) in tumoral β-cells.

Between 2001 and 2010, we obtained a fragment of fresh insulinoma tissue from nine patients who had been operated on at the University Clinics Saint-Luc in Brussels, Belgium, and one patient who had been operated on at the Hôtel-Dieu in Paris, France. The clinical diagnosis of insulinoma had been established on the basis of classic medical, biological, and imaging criteria (3,4,23,24), which led to surgical ablation of a tumor, resulting in the cessation of hypoglycemic episodes in all patients. The study was conducted with the requested approvals and according to the regulations of the Commission d’Ethique Biomédicale of the Faculty of Medicine of the University of Louvain (Brussels, Belgium).

The fragment of insulinoma used for in vitro studies of insulin secretion was macroscopically sampled from the tumor, avoiding contamination with normal pancreatic tissue when present. Upon arrival in the laboratory, the tissue was minced and collagenase digested like pancreatic fragments from infants undergoing surgery for congenital hyperinsulinism (21). After ∼20 h of culture in RPMI medium containing 5 mmol/L glucose, portions of digested tissue were distributed into perifusion chambers for the measurement of insulin secretion (21) and, at the end of the experiments, of insulin content (25). Because the amount of tissue placed in each chamber was variable, the insulin secretion rate was expressed relative to the insulin content of the tissue (fractional insulin secretion as a percentage of insulin content per minute). Details on the system, techniques, and solutions (containing 1 μmol/L forskolin during perifusions) can be found elsewhere (26). The number of tests performed with each preparation was determined by the amount of available tissue. Two protocols evaluating acute and concentration-dependent insulin responses to glucose and testing the effects of diazoxide and tolbutamide could be performed in all insulinomas. Insulin secretion in response to high extracellular CaCl₂ concentration was tested in 9 of 10 insulinomas. Two other protocols were performed in a number of insulinomas, which are indicated in the text and legends. The initial insulin content of the tumor was not directly measured but was estimated by adding the amount of insulin secreted during culture and experiments to the insulin content measured at the end of perifusions (21). In two case patients, attempts to measure insulin secretion from pancreatic tissue outside the tumor were unsuccessful because insulin was not measurable in the perifusion medium of the small available fragments.

Fragments of insulinomas, sometimes with small adjacent regions of normal tissue, were fixed in formalin and/or Bouin solution for conventional microscopy and immunohistochemical identification of insulin, glucagon, and somatostatin (27) or of the proliferation marker Ki-67 (28). The volume density of β-cells in insulinomas was determined by point counting on slices immunostained for insulin (27). Staining with Congo Red was used to identify amyloid deposits. Immunodetection of HK-I was performed as previously described (22).

Subdivision of the studied insulinomas in three functional groups was performed retrospectively based on of the pattern of insulin secretion observed in vitro and on the presence or absence of HK-I in insulinoma β-cells (see below). The major characteristics of the 10 case patients, all of whom were negative for mutations in the MEN-1 gene, are given in Table 1. The first symptoms suggestive of hypoglycemia occurred 8–50 months before surgery. Hypoglycemic episodes occurring after meals were clearly documented in three patients from group A.

The diagnosis of insulinoma was histologically confirmed in all case patients. The insulinomas were classified as “solid” or “trabecular” according to established criteria (6,7) (Table 1). The typical appearances of the two forms can be seen in Fig. 1. As reported previously (6–8), insulin immunostaining was often polarized in β-cells of trabecular insulinomas (Fig. 1, case patient 8) and was more diffuse and heterogeneous in β-cells of solid insulinomas (Fig. 1, case patient 2). Notably, both histological types were observed in the three functional groups (Table 1). All were well differentiated, with a low proliferation index (2–10%) (Supplementary Fig. 1). The size of the tumor ranged between 0.4 and 1.9 mL in nine case patients and was very large (16 mL) in case patient 4. The volume density of β-cells in the tumors averaged 44%, but was highly variable (13–73%) owing to the sometimes high proportions of mesenchymal tissue, but not of non–β-endocrine cells (Table 1). After staining with Congo Red, amyloid deposits were detected in five of nine insulinomas, a similar proportion as in a larger study (29) (Supplementary Fig. 1). Amyloid deposits were abundant in only two insulinomas from group A and were rare or absent in insulinomas from groups B and C (Table 1).

The tumor insulin concentration was highly variable, ranging from 40 to 395 ng/mg (mean 132 ng/mg). This variability persisted after normalization for differences in β-cell proportions (Table 1). None of the three groups was characterized by consistently high or low insulin concentrations (Table 1).

In insulinomas from group A, the “basal” insulin secretion rate, measured in 1 mmol/L glucose, ranged from 0.02% to 0.12%/min. Increasing the glucose concentration to 15 mmol/L stimulated insulin secretion several-fold in the four case patients (Fig. 2A). This stimulation was partially (case patient 1) or completely inhibited by treatment with 100 μmol/L diazoxide and reversibly restored by 100 μmol/L tolbutamide. These responses are thus qualitatively similar to those observed in isolated islets from normal subjects (26) and fragments of normal pancreas from infants (21). The major difference is a sometimes more sluggish onset and a less pronounced biphasic pattern of the response to glucose.

In group B, insulinomas 5 and 6 were characterized by a very high insulin secretion rate in 1 mmol/L glucose (0.40% and 0.19%/min), and, except for a short-lived initial increase, no further stimulation by 15 mmol/L glucose (Fig. 2B). In case patient 7, a slower but more sustained response to 15 mmol/L glucose occurred above a less markedly elevated basal rate in 1 mmol/L glucose (0.09%/min) (Fig. 2B, bottom panel). The three insulinomas were sensitive to diazoxide, which inhibited insulin secretion well below the values measured in 1 mmol/L glucose, and to tolbutamide, which reversibly reversed this inhibition (Fig. 2B).

In insulinomas from group C, the basal rate of insulin secretion in 1 mmol/L glucose was low, ranging from 0.01% to 0.02%/min. Stimulation with 15 mmol/L glucose only transiently and minimally increased insulin secretion (case patient 8) or was ineffective (case patient 9); addition of diazoxide and tolbutamide also was without effect (Fig. 2C, top panel). It was only in case patient 10 that qualitatively normal but quantitatively small responses to glucose, diazoxide, and tolbutamide were observed (Fig. 2C, bottom panel).

In group A, stepwise stimulation with glucose augmented insulin secretion in a concentration-dependent manner (Fig. 3A). In case patients 1, 2, and 3, 1 mmol/L glucose approximately doubled (1.5-, 2.2-, and 2.4-fold) the insulin secretion rate, which then kept increasing up to 5 or 10 mmol/L glucose. In case patient 4, in which the experiment was started in 1 mmol/L glucose, stimulation of secretion occurred only at 5 mmol/L glucose and above (Fig. 3A, bottom panel).

In group B, the concentration dependency of the response to glucose was markedly altered (Fig. 3B). In case patients 5 and 6, the insulin secretion rate displayed a rapid and large peak in response to 1 mmol/L glucose (40-fold and 6-fold increases) and then regularly declined in the face of stepwise increases in glucose concentration. The response of case patient 7 was slightly different (Fig. 3B, bottom panel). Insulin secretion was also induced by 1 mmol/L glucose (4.5-fold increase), but the response was of slower onset and remained sustained when the glucose concentration was increased further. In no case patient was the switch from 10 to 1 mmol/L glucose followed by an obvious decrease in insulin secretion, whereas the return to a glucose-free medium was inhibitory in case patients 5 and 7.

In group C, no stimulation of insulin secretion occurred in response to stepwise increases in glucose (Fig. 3C). In case patients 8 and 9, a paradoxical increase in secretion was seen when the glucose concentration was lowered from 10 to 1 mmol/L (Fig. 3C, top panel).

In the four insulinomas from group A, the addition of leucine and glutamine (5 mmol/L each) to a medium containing 3 mmol/L glucose induced biphasic insulin secretion. This is illustrated for case patients 1 and 2 in Fig. 4A. The omission of extracellular CaCl₂ reversibly abolished secretion, which was also strongly inhibited by the activation of α₂-adrenoceptors with clonidine. In group B, only insulinoma 5 was tested. The two amino acids were apparently ineffective, but the high rate of insulin secretion induced by 3 mmol/L glucose was completely and reversibly abolished by the omission of CaCl₂ and was partially inhibited by clonidine (Fig. 4B). In group C, the combination of leucine and glutamine was ineffective, and the omission of CaCl₂ was followed by a paradoxical, reversible increase in secretion (Fig. 4C).

In normal β-cells, glucose controls insulin secretion through changes in the cytosolic Ca²⁺ concentration (triggering pathway) and amplification of Ca²⁺ effects on the exocytosis of insulin granules (metabolic amplifying pathway) (30). Metabolic amplification can be studied when all β-cell K_ATP channels are closed by a high concentration of sulfonylurea. In insulinomas 3 and 4 from group A, 500 μmol/L tolbutamide triggered a large peak of insulin secretion followed by stabilization of the secretion rate above prestimulatory values. Subsequent augmentation of the glucose concentration to 15 mmol/L, in the continuous presence of tolbutamide, increased insulin secretion in a reversible manner (Fig. 4D). In group B, tolbutamide was hardly effective (case patient 6) or transiently (case patient 7) effective in 1 mmol/L glucose, and the subsequent addition of 15 mmol/L glucose had no obvious effect (Fig. 4E). In group C, the high tolbutamide concentration was ineffective in 1 mmol/L glucose, and a subsequent increase in the glucose concentration to 15 mmol/L was slightly but reversibly inhibitory (Fig. 4F).

A sudden increase in the concentration of extracellular CaCl₂ from 1.25 to 10 mmol/L in a medium containing 5 mmol/L glucose evoked a large reversible peak of insulin secretion in all tested insulinomas from groups A and B (Fig. 5A and B). In group A, the relative stimulation ranged from 9.7-fold to 14.7-fold. It was smaller in group B, ranging from 4.9-fold to 7.6-fold. In contrast, insulinomas from group C did not increase insulin secretion in response to high extracellular CaCl₂ concentrations (Fig. 5C).

The stimulation of insulin secretion by as little as 1 mmol/L glucose in certain insulinomas led us to search for the presence of a low-K_m hexokinase in insulinoma cells. Immunohistochemistry for HK-I was negative in the three insulinomas from group A that could be tested (Table 1). As illustrated for case patient 2, β-cells within the tumor (and islets outside the tumor) were not labeled (Fig. 6A), in contrast to neural, centro-acinar, and vascular cells (Fig. 6A, inset). In the three case patients in group B, insulinoma cells were positive for HK-I (Fig. 6B), whereas β-cells in islets outside the tumor were negative (Fig. 6B, inset). In case patients 5 and 6, all insulinoma cells were labeled, though with a variable intensity, whereas areas of negative and positive cells coexisted in case patient 7. The three case patients in group C displayed a punctated positivity for HK-I in many insulinoma cells intermingled with negative cells (Fig. 6C), whereas β-cells of the islets were negative (Fig. 6C, inset).

Although our series of 10 insulinomas is the largest that has been functionally studied in vitro, its limited size, inherent to the rarity of the pathology, remains a handicap in our attempt to classify these endocrine tumors according to functional features. Our subdivision into three groups has limitations and would have been strengthened by genetic analysis of the tumors. It may well be an oversimplification of the heterogeneous behavior of these tumoral β-cells, but enough similarities appear to justify their grouping rather than a multiplication of smaller categories.

No consistent differences in symptom duration, tumor size, histological type, or insulin concentration were found among the three functional groups of insulinomas. For the 10 case patients, the average tumor insulin concentration (132 ng/mg) was similar to that measured in a normal autopsy pancreas (125 ng/mg) (31), but the interindividual variability was larger (10-fold vs. 4-fold). Higher average insulin concentrations and even greater variability (100- to 200-fold) were found in larger series (5,32). Insulinomas contain variable, sometimes high, proportions of mesenchymal tissue (vessels, amyloid deposits, and connective tissue) that decrease the apparent insulin concentration when expressed per weight. Nevertheless, because the β-cell fraction (13–73%) largely exceeded that in a normal pancreas (∼1.25%), it is clear that the concentration of insulin per β-cell is much lower in insulinomas than in normal islets, as has also been concluded by others (5,32). Finally, from tumor size and β-cell volume density, we can calculate that the mass of insulinoma β-cells ranged between 155 and 2,080 mg in our 10 case patients, compared with 300–1,500 mg (average ∼900 mg) β-cells in the pancreas of healthy subjects (31). It is therefore unlikely that this additional number of β-cells would cause profound hypoglycemia if these insulinoma cells were functionally normal.

In previous in vitro investigations of human insulinoma cells, insulin secretion was unaffected by high glucose concentrations in five case patients from three studies (11,12,16) and was variably augmented in seven single case patients (9,10,14,17–20). Other agents were rarely tested: insulin secretion was increased by a sulfonylurea in two insulinomas (18,19) and leucine was ineffective in one of one insulinoma (11). An inhibition of insulin secretion was observed upon CaCl₂ omission in one insulinoma (11), whereas diazoxide addition was ineffective in two of two insulinomas (9,12). In our experience, basal insulin secretion by isolated human islets perifused with 1 mmol/L glucose was ∼0.01% of their insulin content per minute (26). Much higher (5–50×) basal rates of secretion were measured here in insulinomas from groups A (three of four case patients) and B (three of three case patients). Distinct procedures for tissue preparation may contribute to the difference but are not the sole explanation. Thus, compared with normal islets, the difference is only 1.5- to 2-fold in insulinomas from group C, and only 2- to 3-fold in collagenase-digested fragments of the normal pancreas of infants (21,22).

In group A, the four insulinomas shared several features that make them qualitatively similar to normal islets (26) or fragments of normal pancreata (21). These included sustained insulin secretion in response to 15 mmol/L glucose; normal inhibitory and stimulatory effects of diazoxide and tolbutamide, respectively; normal stimulation by amino acids; inhibition by the omission of CaCl₂ or the addition of clonidine; and functioning of the metabolic amplifying pathway. All of these characteristics suggest qualitatively normal stimulus-secretion coupling with operative K_ATP channels. However, insulin secretion was quantitatively excessive. Thus, case patients 1, 2, and 3 displayed abnormally high secretion rates in low- and high-glucose conditions and a left shift of the glucose dependency of the response. Whereas the normal threshold is at 3 mmol/L glucose (21,26), a doubling of insulin secretion was already observed at 1 mmol/L glucose, which could explain the occurrence of fasting hypoglycemic episodes in the patients. This threshold lowering was not correlated with HK-I detection in insulinoma cells, as in group B (see below), perhaps because expression of the enzyme was weaker, also explaining the smaller response to 1 mmol/L glucose than in group B. The insulin secretion rate induced by high glucose concentrations (0.3–0.4%/min) was also more than twofold higher than in normal islets or fragments of normal pancreata tested under similar conditions (∼0.15%/min) (21,26). We have no mechanistic explanation for this excessive insulin response, which probably explains why hypoglycemic episodes also occurred after meals, as documented in case patients 2 and 3. Unlike the other three case patients of group A, case patient 4 did not show an increase in glucose sensitivity with elevation of the secretion rate at a low glucose level, but the response to high glucose concentrations was also exceedingly large, which might explain why hypoglycemic episodes occurred only after meals. Notably, this case patient was also exceptional due to the large size of the tumor that was filled with amyloid deposits.

In group B, the main characteristics of insulinomas 5 and 6 were a large increase in insulin secretion in response to 1 mmol/L glucose, followed by a progressive decline in the face of subsequent increases in glucose concentration. Strong stimulation already by 1 mmol/L glucose may explain why stepping to 15 mmol/L glucose was poorly efficient and why diazoxide inhibited insulin secretion well below the rates measured at 1 mmol/L glucose. The abnormal sensitivity of insulinomas to low glucose concentrations with poor further response to higher glucose concentrations can result in hypoglycemic episodes in the fasting state and not after meals. Two insulinomas resembling our case patients 5 and 6 have previously been studied in vitro (18,20). Compared with a glucose-free medium, 1 mmol/L glucose induced an ∼10-fold increase in insulin secretion, which was not larger at higher glucose concentrations. This excessive sensitivity to a low glucose concentration remained unexplained (18) or was tentatively ascribed to the strong expression of the high-affinity GLUT-1 in the tumor cells (20). However, it is now established that GLUT-1 and GLUT-3 are the major glucose transporters in normal human β-cells (33,34). Using immunohistochemistry, we detected the presence of the low-K_m HK-I in tumor β-cells, whereas normal islet β-cells, which do not express HK-I but use the high-K_m glucokinase to phosphorylate glucose (35,36), were negative as expected. We therefore suggest that, in the insulinoma cells of group B, glucose metabolism is already markedly accelerated when the glucose concentration is low. The expected consequences are a virtually complete closure of K_ATP channels and full activation of the metabolic amplifying pathway (30), which would account for the poor effect of tolbutamide, except when K_ATP channels have been opened by diazoxide, and for the virtual lack of further stimulation by the combination of leucine and glutamine, which mainly acts through the acceleration of metabolism (37,38). In support of our proposal that the activity of HK-I in insulinoma cells is responsible for abnormal insulin secretion at low glucose concentrations in the case patients in group B, we wish to underline the similarities with our previous findings in a group of infants with focal hyperinsulinism (22). The in vitro abnormalities of insulin secretion associated with the presence of HK-I in a subset of their β-cells are quasi-superimposable on those of insulinomas from group B. A dominant form of congenital hyperinsulinism has also tentatively been attributed to abnormal HK-I expression in β-cells (39). Case patient 7 behaved slightly differently from case patients 5 and 6. Insulin secretion was also induced by 1 mmol/L glucose, but higher concentrations retained some effect, as did tolbutamide even in the absence of diazoxide. A plausible explanation is the weaker and heterogeneous presence of HK-I in the tumor, with some β-cells presumably maintaining a normal control of glucose phosphorylation by glucokinase.

In group C, the major characteristics of the three insulinomas were very low rates of insulin secretion despite substantial insulin stores in the tumor and a virtually complete absence of responses to stimuli and inhibitors active in the other two groups. Poor stimulation by glucose and the ineffectiveness of diazoxide and tolbutamide are reminiscent of the defects observed in focal lesions of the pancreas causing congenital hyperinsulinism owing to a lack of functional K_ATP channels in β-cells (21). However, major differences distinguish the two pathological entities, as follows: insulin secretion rates were 5- to 10-fold lower in these insulinomas than in focal lesions; the decrease in secretion produced by high glucose levels in the presence of high tolbutamide levels in insulinomas was never seen in focal lesions; the omission of CaCl₂ paradoxically increased insulin secretion in insulinomas, whereas it consistently suppressed it in focal lesions; clonidine was ineffective in insulinomas; and a high extracellular CaCl₂ concentration was ineffective in insulinomas but strongly stimulated insulin secretion in focal lesions (21). Several morphological features also distinguish focal lesions of infants and insulinomas (27). The functional significance of HK-I detected in insulinomas from group C is unclear in view of the virtually complete ineffectiveness of glucose on secretion. Intriguingly, HK-I immunostaining was not present in all cells, suggesting heterogeneity within the tumor. A poor viability of these three insulinomas cannot be formally excluded but would be expected to cause insulin leakage rather than low baseline secretion and is not supported by their preserved morphology and similar pattern of HK-I staining. Distal defects in the exocytotic machinery are possible, but how they caused hypoglycemic episodes (in the fasting state only) is unclear.

Selective arterial calcium stimulation of the pancreas with hepatic venous sampling is used for preoperative localization of insulinomas (23,40–42). We observed strong stimulation of insulin secretion by 10 mmol/L CaCl₂ in all insulinomas from groups A and B, and no response in group C. In vitro stimulation by high CaCl₂ concentration has previously been reported in three other case patients and was attributed to the activation of an extracellular Ca²⁺-sensing receptor (13,43), but the augmentation of Ca²⁺ influx through the plasma membrane may also contribute. In fragments of the normal pancreas of infants, a high CaCl₂ concentration caused an insulin response that was qualitatively similar to but quantitatively smaller (threefold increase) than that in insulinomas from groups A and B (21). The discrepancy with the lack of effect on normal pancreas in vivo (23,40–42) is probably due to higher concentrations of CaCl₂ and glucose during in vitro experiments than during in vivo tests. Notwithstanding, our results show that false-negative results in vivo may be due to the unresponsiveness of some insulinomas (group C).

In conclusion, human insulinomas showed distinct, though not completely heterogeneous, defects in insulin secretion in vitro. We subdivided our 10 case patients in three groups on the following basis. Group A was characterized by a qualitatively normal stimulus-secretion coupling with effective control by glucose. However, increased sensitivity to the sugar caused the elevation of baseline secretion at low glucose levels, and the response to high glucose levels was quantitatively excessive, which may explain postprandial episodes of hypoglycemia. No HK-I was detected, but its presence at low levels cannot be excluded. Group B was characterized mainly by undue expression of HK-I in insulinoma β-cells, which resulted in the strong stimulation of insulin secretion by as little as 1 mmol/L glucose and the loss of regulation by higher glucose concentrations, two features that may explain why hypoglycemic episodes only occurred during fasting. Group C showed the heterogeneous presence of HK-I, very low rates of insulin secretion, and a virtually complete absence of responses to stimuli and inhibitors, features that do not readily explain fasting hypoglycemic episodes in these patients. Without investigation of the mechanisms of tumorigenesis and without genetic studies, we cannot determine whether these three groups have distinct origins. However, we can speculate that transitions might exist between groups, as illustrated by case patient 7, which expressed HK-I in some parts of the tumor only and was difficult to assign to functional group B rather than group A.

Acknowledgments. The authors thank Sebastien Godecharles and Fabien Knockaert for technical assistance at the medical faculty of the University of Louvain (Brussels, Belgium). The authors also thank Professors Christian Boitard and Etienne Larger for giving access to the insulinoma of one of their patients from the Hôtel-Dieu, Paris, France. In addition, the authors thank Professor Dominique Maiter and Dr. Raluca Furnica from the University Clinics Saint-Luc (Brussels, Belgium) for their help in tracing clinical information on the patients.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.-C.H. designed the research, researched and analyzed the data, and wrote the article. M.N., Y.G., J.R., and C.S. researched and analyzed the data and reviewed the article. J.-C.H. 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.