Glucose Transporter Proteins in Human Insulinoma

Methods

Patients and Tumors

Clinical characteristics of five patients with the clinical insulinoma syndrome and their tumors are shown in Table 1. In four patients in the General Clinical Research Center of Temple University Hospital (patients 1, 2, 3, and 5) inappropriate hyperinsulinemia (insulin concentration, ≥ 30 pmol/L [5.0 µU/mL]) occurred during hypoglycemia (glucose concentration, ≤ 2.2 mmol/L [40 mg/dL]) that was induced by fasting. In patient 4, who was on a surgical ward of the hospital, hypoglycemia was documented during an episode of hypoglycemia that occurred after an overnight fast before breakfast. As determined in our laboratory, none of the five patients had insulin-binding antibodies, and none had detectable levels of blood sulfonylurea (determined by Smith-Kline Bioscience Laboratories, King of Prussia, Pennsylvania). Pancreatic masses were identified in all patients by ultrasound or computed tomographic scans or during exploratory laparotomy. The surgically removed tumors were diagnosed as insulinomas by light and electron microscopy. After resection, the tumors were immediately rinsed with ice-cold saline, weighed, measured, and frozen at −80°C.

Acid Ethanol Extraction

Pieces of the frozen tumors were finely minced with scissors and extracted overnight at 4 °C with acid ethanol [11] for measurement of immunoreactive insulin and glucagon content.

Preparation of Solubilized Membranes

Aliquots of frozen tumor or normal human liver were homogenized with a polytron in a medium containing 25 mmol/L of HEPES, 4 mmol/L of ethylenediaminetetraacetic acid, 25 mmol/L of benzamidine, and 1 µM each of leupeptine, pepstatin, and aprotinin and 200 µM of phenazine methosulfate. Triton X-100 (final concentration of 1%) (Union Carbide, Danbury, Connecticut) was then added to make the homogenate soluble. The suspension was shaken for 60 minutes at 4 °C, and the solubilized membranes were centrifuged at 150 000 g for 35 minutes. The supernatant fluid was filtered through a 0.45-micrometer filter. Its protein concentration was determined [12], and samples were applied directly to nitrocellulose membranes for immune blotting (2 to 20 µg of protein per assay) [13].

Immune Blotting of Glucose Transporter Proteins

The nitrocellulose membranes into which test samples had been blotted were incubated for 2 hours with TRIS buffer containing 5% bovine serum albumin and specific antihuman Glut antiserum. The Glut 1 antiserum RAB 379 was directed against the amino acid sequence 480-492 of rat Glut 1, which is identical to the same sequence in human Glut 1 [14]. The Glut 2 antiserum was directed against the C-terminal 30 amino acids of human Glut 2. A mouse antirabbit IgG antibody (provided by Sigma Chemical Co., St. Louis, Missouri) and a Iodine-125-labelled sheep antimouse antibody were added as second and third antibodies. Membranes were then subjected to autoradiography and were counted in a γ counter [13].

Fasting

Patients were admitted to the General Clinical Research Center of Temple University Hospital. A flexible intravenous catheter with a heparin lock was inserted into the antecubital vein of one arm. Patients were permitted to move around freely and had access to water as desired, but they did not receive any calories. Routine chemistries, complete blood count plus differential, and serum insulin and sulfonylurea concentrations were measured in blood samples at the beginning of the fast. Thereafter, blood samples were drawn every 2 hours for glucose determinations. Aliquots of these samples were kept for later determination of insulin. When the fasting blood sugar concentration decreased to less than 2.8 mmol/L (50 mg/dL), samples were taken every 30 minutes; when it decreased to less than 2.2 mmol/L (40 mg/dL), three blood samples were taken (one every 5 minutes), and the fast was terminated by intravenous infusion of glucose.

Analytical Procedures

Plasma glucose was measured with a Beckman Instruments (Palo Alto, California) glucose analyzer. Serum insulin [15] and plasma glucagon [16] concentrations were measured by radioimmunoassay and the C-peptide concentration was measured with a double-antibody radioimmunoassay kit (INCSTAR, Stillwater, Minnesota).

Results

Insulin and C-peptide Concentrations during Hypoglycemia

Figure 1 and Table 2 show the insulin and C-peptide concentrations during hypoglycemia. Patients 1, 2, 3, and 5 underwent a fast in the General Clinical Research Center until hypoglycemia developed after 20, 10, 25, and 9 hours of fasting, respectively. In patient 4, hypoglycemia developed on a surgical ward after an overnight fast that lasted approximately 12 hours.

In patient 1, serum insulin concentrations increased paradoxically when blood sugar levels decreased during the fast. This patient was found to have a multinodular, malignant insulinoma and metastatic tumors in the liver. In patients 2 to 5, who had mononodular, nonmetastatic insulinomas, serum insulin and C-peptide concentrations decreased during the fast. However, in none of the five patients did serum insulin concentrations decrease to lower than 30 pmol/L, nor did C-peptide concentrations decrease to lower than 0.08 nmol/L when plasma glucose concentrations had decreased to 2.2 mmol/L or lower. Moreover, all five patients showed central nervous system signs of hypoglycemia, which disappeared promptly after intravenous administration of glucose.

Glut 1 and Glut 2

Figure 2 shows the Glut 1 and Glut 2 concentrations in tumor membranes of four patients. Total insulinoma membranes of patient 1 contained exclusively Glut 1 protein. Tumors of the other four patients contained what appeared to be small amounts of Glut 2 and substantial amounts of Glut 1. Only Glut 2 protein was detected in a sample of a normal human liver. Thus, patterns of distribution of Glut 1 and Glut 2 in insulinoma were completely different than those in tissues from a normal liver.

Discussion

We studied five patients with the clinical insulinoma syndrome. During hypoglycemia induced by fasting (blood glucose concentration, <2.2 mmol/L), endogenous insulin release in all five patients was not appropriately suppressed. The islet cell tumors removed from these patients contained what appeared to be small amounts of Glut 2 protein (four patients) or no Glut 2 (one patient) and an abundance of Glut 1 protein as determined with specific anti-human Glut 1 and Glut 2 antisera. It should be emphasized, however, that this technique does not permit quantitative comparison between Glut 1 and Glut 2. Because normal human pancreatic islets were unavailable, we used a tissue sample from a normal human liver as a control. Both islets and the liver have been reported to almost exclusively contain Glut 2 and little or no Glut 1 protein [3, 4, 6, 17]. In concordance with these reports, only Glut 2 was detected in the liver membranes. To our knowledge, no reports have been published on glucose transporter protein content in human insulinoma. However, our findings are in accordance with a recent report by Seino and colleagues [18], who found Glut 1 messenger RNA but no Glut 2 messenger RNA in two cases of human insulinomas [18]. Their failure to detect Glut 2 messenger RNA may have been caused by the relative insensitivity of the Northern blot technique. Current data therefore suggest that human insulinomas overexpress the high-affinity (low K_m) Glut 1 and may underexpress the low-affinity (high K_m) Glut 2 glucose transporters. The assumption that Glut 2 was underexpressed was based on the finding that Glut 2 was undetectable in patient 1 and was present in much smaller amounts in the other four patients than the amount found in the tissue from the normal liver.

The reason for these abnormalities in insulinoma Glut content is not known. Previous studies by others have shown that islet cell expression of Glut 1 protein was substantially increased after oncogenic transformation. For instance, RINm5F cells, which are derived from a radiation-induced rat insulinoma, expressed large amounts of Glut 1 and little or no Glut 2 [3]. Similarly, high expression of Glut 1 was found in HIT cells, which are derived from simian virus 40 (SV-40)-transformed hamster islet cells [19] and in MIN7 cells, an insulinoma cell line derived from transgenic mice [10]. This phenomenon is not restricted to insulinomas; it has been shown that expression of Glut 1 messenger RNA and protein was increased in rat fibroblasts transformed by viral oncogenes [20, 21]. Presence of glucagon in all five cases of insulinoma suggests that α-cells, which constitutively express Glut 1 but not Glut 2, may have contributed some Glut 1 [3, 6, 22]. The possible contribution of Glut 1 from α-cells, however, could have been only minor because these insulinomas contained little glucagon. The ratio of extractable glucagon to insulin ranged from 1:35 in patient 5 to 1:21 000 in patient 1. Moreover, the tumor of patient 5 contained a glucagon concentration approximately 8 times less than that of patient 4 but also contained a Glut 1 concentration more than 5 times higher.

The reason for a possible reduced expression of Glut 2 is even less clear. Down-regulation of Glut 2 expression by hyperinsulinemia has been shown [23] and may be a possible cause. Three of our five patients did have persistently elevated basal insulin levels, but the other two did not. Moreover, it is unknown whether the relatively modest degree of hyperinsulinemia present in these patients was sufficient to cause Glut 2 down-regulation.

It has been proposed that the high-K_m Glut 2, perhaps in conjunction with the high-K_m phosphorylating enzyme glucokinase, serves as a glucose sensor for the β-cells [3, 5, 6]. According to this concept, under normal conditions, a decrease in extracellular glucose concentrations to less than 4.0 to 5.0 mmol/L would result in a proportional decrease in glucose transport, phosphorylation, and glycolytic flux in β-cells and would result in decreased insulin release. In contrast, presence of substantial amounts of the low-K_m Glut 1 in insulinoma cells would maintain a high glucose transport rate into the islet cells, even under hypoglycemic conditions, and thus would enable continued secretion of insulin. Our findings, as well as those by Seino and colleagues [18], of large amounts of Glut 1 or Glut 1 messenger RNA in seven of seven cases of human insulinomas are compatible with this concept. Moreover, human insulinoma cells grown in culture have been shown to secrete insulin maximally when exposed to glucose concentrations (0.8 mmol/L) at which insulin release by normal islet cells is completely inhibited [24].