Glucose-Sensing in Glucagon-Like Peptide-1-Secreting Cells Free

GLUTag cells (28) were cultured in Dulbecco’s modified Eagle’s medium (5.6 mmol/l glucose), supplemented with 10% (vol/vol) FCS, penicillin, and streptomycin. Medium was exchanged every 3 days, and cells were trypsinized and reseeded at a 1:4 dilution when ∼70% confluence was reached (approximately every 5 days).

For secretion experiments, cells were plated in 24-well culture plates and allowed to reach 60–80% confluence. On the day of the experiment, cells were washed twice with 500 μl of glucose-free Krebs-Ringer medium containing (in mmol/l) 120 NaCl, 5 KCl, 2 CaCl₂, 1 MgCl₂, 22 NaHCO₃, and 0.1 DiprotinA, bubbled with 95% O₂/5% CO₂ for 10 min and supplemented with 0.5% (wt/vol) BSA. Experiments were performed by incubating the cells with test reagents in 500 μl of Krebs-Ringer for 2 h at 37°C, 5% CO₂. Tolbutamide, forskolin, and isobutylmethylxanthine (IBMX) were prepared as 1,000× stocks in DMSO, and the final DMSO concentration was adjusted to 0.2% (or 0.3% for the experiment with tolbutamide, forskolin, and IBMX simultaneously). At the end of this incubation period, medium was collected and centrifuged to remove any floating cells. GLP-1 was assayed in the supernatants using an enzyme-linked immunosorbent assay (ELISA) that is specific for GLP-1 (7-36) amide and GLP-1 (7-37) (GLP-1 active ELISA-kit; Linco). Each sample was analyzed in duplicate.

Cells were plated onto 35-mm dishes 1–3 days before use. Experiments were performed on single cells and well-defined cells in small clusters. Microelectrodes were pulled from borosilicate glass (GC150T; Harvard Apparatus), and the tips were coated with melted beeswax. Electrodes were fire-polished using a microforge (Narishige) and had resistances of 2.5–3 M.ohms when filled with pipette solution. Membrane potential and currents were recorded using an Axopatch 200B (Axon Instruments) linked through a Digidata 1320A interface and analyzed using pCLAMP software (Axon Instruments). Experiments were performed at 22–24°C. Current clamp recordings were filtered at 5 kHz and digitized at 10 kHz. Membrane conductance was measured in perforated patch voltage clamp experiments by applying a series of 1.2-s voltage ramps from −100 to −50 mV (holding potential −70 mV). The current voltage relation was linear in this voltage range, and the slope conductance was calculated from the average of 10 voltage ramps.

The bath solution contained (in mmol/l) 5.6 KCl, 138 NaCl, 4.2 NaHCO₃, 1.2 NaH₂PO₄, 2.6 CaCl₂, 1.2 MgCl₂, and 10 HEPES (pH 7.4). Glucose and drugs were added to the bath solution as indicated. The perforated patch pipette solution contained (in mmol/l) 76 K₂SO₄, 10 KCl, 10 NaCl, 55 sucrose, 10 HEPES, and 1 MgCl₂, (pH 7.2), to which amphotericin was added to a final concentration of 200 μg/ml. The pipette solution for recording washout currents contained (in mmol/l) 107 KCl, 1 MgCl₂, 1 CaCl₂, 10 EGTA, 10 HEPES, and 0.3 ATP (pH 7.2). Tolbutamide was prepared as a 100-mmol/l stock solution, and diazoxide was prepared as a 68-mmol/l stock solution, in KOH.

Results are presented as mean ± SE. Statistical significance was tested using Student’s t test, using a threshold for significance of P < 0.05.

Whole RNA was prepared from ∼2.5 × 10⁷ GLUTag cells using Tri-reagent (Sigma). First-strand cDNA synthesis was performed for 1 h at 42°C in a total reaction volume of 25 μl containing 1 μg of RNA, 500 ng of random hexamer primers (Promega), 25 nmol of each dNTP (Promega), 40 units of RNAsin (Promega), 25 mmol/l Tris-HCl (pH 8.3), 50 mmol/l KCl, 2 mmol/l dithiothreitol, 5 mmol/l MgCl₂, and 21 units of RT (SuperRT; HT Biotechnology). One microliter of this reaction was used for subsequent PCR amplification using the following specific primers, which span introns wherever possible (from 5′ to 3′): Kir6.1 (accession no. D88159) sense GTCTTCTAGGAGGACGCGTG (position 126), antisense TATCATACAGGGGGCTACGC (position 786); Kir6.2 (accession no. D50581) sense AAGAAAGGCAACTGCAACGT (position 243), antisense CCCCATAGAATCTCGTCAGC (position 1066); SUR1 (accession no. L40624 [ rat]) sense GTGAGTTCCTGTCCAGTGCA (position 2091), antisense GTGATGTTCTCCTCCACCGT (position 2622); glyceraldehyde-3-phosphate dehydrogenase (GAPDH; accession no. M17701 [rat]) sense CGATCCCGCTAACATCAAAT (position 264), antisense GGATGCAGGGATGATGTTCT (position 654); glucokinase (accession nos. L41629 and L41631) sense GCTGAGGCGTGAGGGACAG (position 361 [exon-1 β-cell]), sense AGCCAGACAGTCCTCACCC (position 9210 [exon-1 liver]), antisense TCTGAGCCTTCTGGGGTGG (position 13863). The reaction was performed in a volume of 50 μl containing 50 pmol of each primer of one sense/antisense pair, 0.2 mmol/l of each dNTP (Promega), 1.8 mmol/l MgCl₂, 50 mmol/l KCl, 10 mmol/l Tris-HCl (pH 9.0), 0.1% Triton X-100, and 0.75 units of Super-Taq-polymerase (HT Biotechnology) in a PCR-express thermocycler (Hybaid) using the following cycling protocol: initial 3-min denaturation at 95°C (in which time the polymerase was added [hot start]) was followed by 35 cycles of 94°C for 60 s, 58°C for 60 s, and 72°C for 60 s and a final elongation at 72°C for 7 min. Twenty microliters of the reaction was used for subsequent analysis by agarose gel electrophoresis (2% gel), and products were visualized by ethidium bromide fluorescence. The predicted sizes (bp) of fragments were as follows: Kir6.1, 661; Kir6.2, 824; SUR1, 532; GAPDH, 391; glucokinase, β-cell promoter, 605; glucokinase, liver promoter, 331. The primer pairs for SUR1, Kir6.1, and Kir6.2 have been described previously (30). No bands were detected in control experiments using water in place of the RT product.

To investigate whether the membrane potential of GLUTag cells is modulated by glucose, we studied intact cells by current clamp using the perforated-patch method. GLUTag cells perfused at low glucose concentrations (0–2 mmol/l) were electrically silent or fired infrequent action potentials. The mean resting potential in the presence of 2 mmol/l glucose was −53 ± 1.3 mV (n = 13), measured during action potential-free intervals. Increasing the glucose concentration to 20 mmol/l resulted in membrane depolarization and the generation of spontaneous action potentials (Fig. 1A and B). The mean action potential frequency in 20 mmol/l glucose was 1.3 ± 0.3 Hz (n = 14; Fig. 1C). We also observed a few cells that fired rapidly in 2 mmol/l glucose and were not affected by perfusion with low or high glucose. As this behavior was frequently associated with subsequent loss of the patch, such cells were not included in the analysis.

Metabolic inhibition with 3 mmol/l Na azide abolished action potentials in the presence of 20 mmol/l glucose (Fig. 2A) and resulted in membrane hyperpolarization to a mean voltage of −61 ± 4 mV (n = 3). Similar results from pancreatic β-cells have been attributed to the opening of K_ATP channels in response to the falling concentration of ATP and rising concentration of ADP (21).

As K_ATP channel closure would result in a reduced membrane conductance, we investigated the effect of glucose on the conductance measured in perforated-patch voltage-clamp experiments. Increasing the glucose concentration progressively reduced the slope conductance (Fig. 3), indicating that it causes net channel closure.

We investigated the amplitude of washout currents in standard whole-cell voltage-clamp experiments. Current amplitudes were small immediately after establishment of the whole-cell configuration but increased after 2–5 min (Fig. 4A). The reversal potential of the washout current was −75 ± 1 mV (n = 5), consistent with the opening of K⁺ selective ion channels. The currents were blocked by 500 μmol/l tolbutamide (Fig. 4B), indicating that they flowed through open K_ATP channels.

Tolbutamide (500 μmol/l) triggered the firing of action potentials in 2 mmol/l glucose (Fig. 2B). The mean action potential frequency in the presence of tolbutamide was similar to that in 20 mmol/l glucose alone (1.6 ± 0.6 Hz [n = 6] in tolbutamide vs 1.3 Hz in 20 mmol/l glucose). The K_ATP channel opener diazoxide (340 μmol/l) abolished action potentials in the presence of 20 mmol/l glucose and caused membrane hyperpolarization (Fig. 2C).

We used RT-PCR to investigate which K_ATP channel subunits are present in GLUTag cells and whether they also express glucokinase. Glucokinase primers were designed to distinguish between the upstream (β-cell) and downstream (liver) promoters. Figure 5 shows that GLUTag cells express Kir6.2 and SUR1 and glucokinase under the control of the upstream promoter. Similar results were obtained using RNA extracted from the insulinoma cell line min6 (data not shown).

We next investigated whether glucose and tolbutamide stimulate GLP-1 secretion from GLUTag cells. Experiments were performed in Krebs Ringer solution to prevent interference by other possible nutrient secretagogues, such as fatty acids and amino acids. Under these conditions, 0.5, 5, and 25 mmol/l glucose enhanced GLP-1 secretion 2.4 ± 0.1-, 3.1 ± 0.3-, and 3.4 ± 0.3-fold, respectively, relative to the secretion observed in the absence of glucose (Fig. 6). As described previously, strong enhancement of GLP-1 secretion was observed when cytoplasmic cAMP levels were elevated by the addition of forskolin and IBMX (10 μmol/l each); however, under these conditions, secretion was still modulated by the glucose concentration. Tolbutamide (500 μmol/l) also stimulated secretion in the absence of glucose (2.1 ± 0.02-fold in the absence of forskolin/IBMX and 3.1 ± 0.05-fold in the presence forskolin/IBMX, relative to secretion in 0 mmol/l glucose under the same conditions). By contrast, we detected no further stimulation of GLP-1 secretion by tolbutamide in the presence of 0.5 mmol/l glucose.

Despite the interest in GLP-1 as a novel treatment for type 2 diabetes, the electrophysiological events underlying GLP-1 secretion have not previously been addressed. Using the GLUTag cell line, we show that GLP-1 secretion and electrical activity are triggered by an increase in glucose over the concentration range 0–25 mmol/l.

Our results support the idea that regulation of K_ATP channel activity plays a critical role in glucose sensing by GLUTag cells. Thus, cells were hyperpolarized (to ∼−55 mV) at low glucose concentrations, consistent with a potassium conductance playing a major role in determining the resting membrane potential. Increasing the glucose concentration caused net channel closure, as shown by the reduced membrane conductance. Sulfonylurea-sensitive whole-cell washout currents confirmed the presence of K_ATP channels. Tolbutamide triggered action potentials in the presence of 2 mmol/l glucose and stimulated GLP-1 secretion in zero glucose, indicating that a significant proportion of the K_ATP channels are open at low glucose concentrations. The identification of Kir6.2 and SUR1 by RT-PCR is consistent with the electrophysiological finding that the channels were sensitive to tolbutamide and diazoxide.

K_ATP channels have been detected previously in the poorly differentiated and multipotential CCK-secreting cell line STC-1, which also releases GIP and GLP-1 (31–33). CCK secretion from STC-1 cells was stimulated by 5 and 25 mmol/l glucose and by the nonspecific channel blocker disopyramide (31).

In pancreatic β-cells, glucokinase plays a major role in controlling flux through glycolysis at physiological glucose concentrations (19). Using RT-PCR, we showed that GLUTag cells also express glucokinase and that, as in the pancreatic β-cell, this is under the control of the upstream promoter. It has been shown previously that the β-cell glucokinase promoter can direct expression in neuroendocrine cell types, including some containing GLP-1 immunoreactivity (34). The finding that glucokinase is expressed in GLUTag cells further supports the idea that glycolytic flux, ATP generation, and K_ATP channel activity may vary at physiological glucose concentrations.

The idea that low concentrations of glucose trigger GLP-1 release through metabolism and K_ATP channel closure seems contrary to results obtained using perfused rat ileum, which showed that some nonmetabolizable glucose analogues also stimulate GLP-1 release (35). The ileal perfusion experiments, however, tested substrates at much higher concentrations (∼250 mmol/l) than those used in the current study, and might stimulate secretion by an alternative mechanism.

The glucose dependence of the membrane conductance in GLUTag cells is similar to that reported previously for pancreatic β-cells (36,37). The rate of GLP-1 secretion closely mirrored the change in membrane conductance, rather than reflecting the action potential frequency. Thus, the greatest fall in membrane conductance and the greatest increase in GLP-1 secretion occurred when the glucose concentration was increased from 0 to 1 mmol/l or 0.5 mmol/l, respectively. Action potential frequency, by contrast, continued to increase significantly when the glucose concentration was raised from 2 to 20 mmol/l. The results suggest that GLP-1 release may be elicited by relatively infrequent action potentials but shows little augmentation when the action potential frequency is further enhanced. This idea is supported by the finding that tolbutamide increased the action potential frequency in 2 mmol/l glucose but only stimulated secretion in the absence of glucose. The finding that GLUTag cells respond to glucose at very low concentrations (0–5 mmol/l) but are not sensitive to additional increments (5–25 mmol/l) is consistent with previous reports that L-cells in primary culture do not respond to changes in the glucose concentration from 5 to 25 mmol/l (16).

Fetal rat intestinal cells have been shown to secrete GLP-1 in response to glucose, but only in the additional presence of insulin (24). Here we provide—to our knowledge, for the first time—evidence for direct sensing of glucose in an l-cell in vitro system. The high glucose sensitivity (0–5 mmol/l) of GLUTag cells raises the question of whether glucose concentrations in the gut lumen or plasma are ever low enough to switch off GLP-1 secretion in vivo. Although glucose concentrations as low as 0.5 mmol/l are not normally reached in plasma, measurements from the distal small intestine of rats during normal feeding behavior revealed glucose concentrations in the range of 0–3 mmol/l (38). As L-cells in vivo are strongly polarized (1,10), they may detect glucose differentially at apical and basolateral surfaces. Low expression of glucose transporters on the basolateral membrane, for example, could render the cells largely blind to glucose in the plasma. It is also possible, however, that immortalization of L-cells to form the GLUTag cell line may be associated with altered glucose sensitivity, as has been reported in some insulinoma cell lines (39), or that cultured cells may lack a tonic K_ATP channel opener. There is evidence, for example, that the neurotransmitter galanin may activate K_ATP channels in the pluripotent intestinal neuroendocrine cell line STC-1 (33) and could thereby influence the glucose sensitivity in vivo.

The distribution of L-cells along the gut shows some species variation, although the cells are universally present in ileum and colon. In humans, L-cells are largely absent from the duodenum but are numerous in the jejunum, ileum, and colon (10). The high frequency of L-cells in the distal gut, taken together with the lack of glucose sensitivity of isolated cells when measured between 5 and 25 mmol/l, has led to the conclusion that the rapid rise in GLP-1 that follows meal ingestion is not due to direct glucose sensing by L-cells. If, however, the L-cells are capable of responding at low glucose concentrations, as our data suggest, then cells in the jejunum may experience stimulatory conditions very soon after a meal. Indeed, studies on human subjects showed that the rate of glucose entry into the duodenum exceeded the maximal duodenal absorptive capacity within 5 min after a glucose meal (40), which would result in the early delivery of small amounts of glucose to the jejunum.

There is considerable evidence supporting the hypothesis that luminal glucose is the trigger for the late phase of GLP-1 release. Infusion of glucose in rat, pig, and dog (13–15) or carbohydrate in human (41) directly into the ileal lumen triggers GLP-1 secretion. Retarding sugar absorption in humans using the α-glucosidase inhibitor acarbose results in earlier and larger increments in plasma GLP-1 levels, which has been attributed to the greater supply of sugar reaching distal populations of L-cells (42,43). The reactive hypoglycemia observed in dumping syndrome, which is associated with the arrival of high concentrations of sugar in the distal small intestine and colon, has also been attributed to increased GLP-1 release (44, 45).

Forskolin and IBMX enhanced GLP-1 secretion approximately fourfold in the presence of different glucose concentrations or tolbutamide. This is a similar magnitude of stimulation to that reported previously in the GLUTag cell line (28). Activation of PKA and PKC has also been shown to enhance GLP-1 release from cultured L-cells. One mechanism underlying the action of PKA may be increased transcription of the proglucagon gene (28,29, 46). In addition, protein kinases and/or cAMP may directly stimulate exocytosis as in the pancreatic β-cell (4). Modulation of protein kinase activity may provide an important mechanism for the neural and hormonal regulation of GLP-1 release, e.g., by GRP and GIP (16–18).

The finding that glucose increments from 5 to 25 mmol/l or tolbutamide added in the presence of 0.5 mmol/l glucose failed to stimulate GLP-1 release contrasts with the effect of these agents on insulin release from pancreatic β-cells. Despite the similarity in the glucose sensitivity of the membrane conductances in GLUTag cells and β-cells, insulin release is largely stimulated at glucose concentrations between 3 and 20 mmol/l. In β-cells, there is evidence that both glucose and sulfonylureas have K_ATP channel-independent effects on exocytosis, although the underlying mechanisms and significance remain controversial (23,47). One explanation for our findings, therefore, is that these actions are absent or not significant in GLUTag cells.

Our results indicate that GLP-1 secretion from GLUTag cells is triggered by glucose through a mechanism involving K_ATP channel closure. The glucose-sensing machinery in GLUTag cells is similar to that found in pancreatic β-cells, although the glucose set point for GLP-1 release is set to the left of that for insulin release (37). Although it is possible that this is a feature of the GLUTag cell line and not of L-cells in vivo, it is also possible that the difference may reflect the normal physiology of the cells, as insulin release is regulated by the plasma glucose concentration, whereas L-cells may detect glucose in the gut lumen. Indeed, although intravenous glucose infusion potently stimulates insulin release in humans, it does not markedly affect GLP-1 release (3). The effect of sulfonylureas on GLP-1 secretion in vivo is not clear. Total GLP-1 levels, measured using nonspecific antibodies, were similar in sulfonylurea-treated and diet-treated subjects with type 2 diabetes (48). However, more recent studies using antibodies specific for active GLP-1 and GLP-1 have shown that meal-related GLP-1 secretion is impaired in type 1 and type 2 diabetes (49,50). An answer to whether sulfonylureas stimulate GLP-1 secretion in normal subjects is not apparent in the literature. Additional investigations into the mechanisms underlying GLP-1 secretion may provide novel drug targets for treating type 2 diabetes.

We thank Prof. Daniel Drucker (Toronto) for kind permission to use the GLUTag cell line. F.G. is a Wellcome Trust Clinician Scientist Fellow, and F.R. is a Diabetes UK R.D. Lawrence Fellow.