Glucagon-Like Peptide 1 Stimulates Lipolysis in Clonal Pancreatic β-Cells (HIT)

Growth and incubation of cells. Clonal pancreatic β-cells (HIT-T15) were cultured in a RPMI-1640 medium supplemented with 50 U/ml penicillin, 50 μg/ml streptomycin, and 10% fetal calf serum (used between passages 64 and 80), harvested with phosphate-buffered saline containing 0.02% EDTA or with 0.25% trypsin and 0.06% EDTA diluted in phosphate-buffered saline, and washed in a buffer (pH 7.4) (18).

The effect of GLP-1 on FFA oxidation was measured as the ¹⁴CO₂ released from cells that were labeled overnight in RPMI-1640 with [U-¹⁴C]oleic acid (1 μmol/l), washed free of remaining labeled FFAs, and incubated in the absence of added labeled or unlabeled FFAs (19).

For insulin secretion measurements, HIT-T15 cells were grown under standard conditions in 24-well plates and then used 2-4 days after passaging for secretion assays. Cells were washed twice in Krebs-Ringer bicarbonate (containing 2 mmol/l CaCl₂ and 0.25% bovine serum albumin) and buffered with 10 mmol/l HEPES at pH 7.4. The cells were preincubated with the above assay buffer for 30 min at 37°C. The buffer was replaced and the cells were then incubated for 30 min. Secretion was stopped by cooling the plates on ice. The sample solution was removed and centrifuged to remove any loose cells, then the insulin level was measured by radioimmunoassay using the assay protocol for rat insulin distributed by Linco Research.

Fluorescence measurements of pH_i, redox state, intracellular Ca²⁺, and extracellular FFA. Cells were loaded with 1 μmol/12′,7′-bis(2-carboxyethyl)-5(6)-carboxyfluorescein (BCECF), acetoxymethyl ester (AM) for 30 min at 37°C with continuous shaking (20). For fluorescence recordings, cells (0.2 mg protein/ml) were maintained in suspension by continuous stirring in a modified Krebs buffer at 30°C in a computer-controlled Hitachi F-2000 spectrofluorometer. BCECF fluorescence was monitored at excitation wavelengths of 440 and 495 nm and an emission wavelength of 535 nm. The average basal pH_i (mean ± SD) was 7.0 ± 0.1 (n = 14). The pH_i was calibrated by measuring the fluorescence of the dye at different pH values between 6 and 8 as determined with a pH electrode after the cells were permeabilized to equilibrate pH_i with extracellular pH (16). Traces shown are those obtained at the excitation wavelength of 495 nm. NH₄Cl (10 mmol/l final concentration) was added at the end of each experiment to confirm cell viability (16); the increase in pH_i followed by the restoration of the basal pH_i indicated viable cells and an intact plasma membrane. The engineered intestinal fatty acid—binding protein ADIFAB (0.2 μmol/l) (dissociation constant [K_D] for oleic acid = 390 nmol/l [21]), was monitored at an excitation wavelength of 390 nm and emission wavelengths of 432 and 505 nm (21,22) to measure the release of FFAs into the media. The redox state of the cells was determined by measuring reduced pyridine and oxidized flavin nucleotide fluorescence at the excitation and emission wavelengths of 340 and 460 nm and 460 and 540 nm, respectively, as described previously (20). Intracellular Ca²⁺ changes were measured in cells loaded with the Ca²⁺ indicator fura 2, as described previously (20). When pH_i and extracellular FFAs were measured simultaneously, only single emission and excitation wavelengths were used. Baseline drift was subtracted from traces when single wavelength measurements were used.

HSL. The total RNA from pancreatic islets was isolated using the guanidinium isothiocyanate-phenol method and was reverse transcribed using random hexamers. Polymerase chain reaction (PCR) was performed using the rat HSL sense and antisense primers 5′-gaacactacaaacgcaacgag-3′ and 5′-caagggggtgagatggtaac-3′. The PCR cycles included a hot start at 97°C, followed by 30 cycles of 94°C for 30 s, 62°C for 20 s, and 72°C for 1 min. Experimental samples were also processed in the absence of reverse transcriptase to rule out amplification from genomic DNA. The PCR products were displayed on a 1.2% agarose gel, stained with ethidium bromide, and photographed with UV light. Negative control reactions lacking cDNA were amplified simultaneously with the experimental samples.

Materials. BCECF-AM, fura 2-AM, and ADIFAB were purchased from Molecular Probes (Eugene, OR). Radioactive tracers were from New England Nuclear (Boston, MA), and triacsin C was from Biomol Research Laboratories (Plymouth Meeting, PA). Other reagents were from Sigma (St. Louis, MO). Orlistat (Xenical, Roche) capsule material was extracted with ethanol to make a nominally 500 mmol/l stock solution. Forskolin (Research Biochemicals International, Natick, MA) was dissolved in DMSO. Cells used for secretion studies contained 0.025% ethanol and 0.1% DMSO in Krebs-Ringer bicarbonate solution containing 0.25% FFA-free bovine serum albumin.

GLP-1 decreases pH_i. An acidification in clonal β-cells was observed when GLP-1 was added (Fig. 1A). The decrease in pH_i began immediately after the addition and continued over a period of minutes. A similar but larger decrease in pH_i was caused by the addition of forskolin (Fig. 1B), which indicated that this change was probably caused by an increase in cAMP. The extent of the decrease in pH_i was similar to that observed in adipocytes and attributed to the generation of FFAs from triglycerides after the stimulation of lipolysis (17). The data in Fig. 1A and B were obtained in the absence of glucose. Glucose has been shown to alkalinize pH_i (23), and this effect appears to dominate pH_i in the presence of glucose, although a transient acidification occurred when GLP-1 was added (Fig. 1C).

Elevation of Ca²⁺ does not influence pH_i. Agents that elevate cAMP may also elevate Ca²⁺ (24). Thus, if acidification results from the release of intracellular FFAs, the FFAs could originate from either triglyceride stores (via lipolysis) or from phospholipids (via Ca²⁺-activated phospholipase activity). Figure 2 shows that the elevation of cytosolic free Ca²⁺ by bombesin is not accompanied by a change in pH_i (top trace). As noted above, previous studies from our laboratories have shown that glucose stimulation of β-cells, which also elevates cytosolic free Ca²⁺, causes alkalinization (23). Other agonists (e.g., carbachol and bradykinin) that increase Ca²⁺ do not also decrease pH_i (data not shown). These findings make it very unlikely that Ca²⁺-stimulation of a phospholipase is responsible for the cAMP-induced decrease in pH_i.

GLP-1 stimulates lipolysis. If GLP-1 stimulates lipolysis, there should be evidence for an increase in FFA production. Evidence supporting an action of GLP-1 to stimulate lipolysis is presented in Fig. 3. Cells were incubated overnight with trace amounts of ¹⁴C-oleic acid to label the internal triglyceride pool, then the production of ¹⁴CO₂ was measured in cells that had been thoroughly washed to remove free oleate. We expected the stimulation of triglyceride breakdown to lead to increased oxidation of FFAs released from the labeled triglyceride pool. Indeed, cells acutely treated with GLP-1 exhibited more than a twofold increase in FFA oxidation, whereas the addition of glucose inhibited FFA oxidation in both control and GLP-1—treated cells. Inhibition of FFA oxidation by glucose (in the absence of GLP-1) was previously shown (19).

To determine whether FFAs were released from the β-cell into the medium, we used ADIFAB, a fluorescent fatty acid—binding protein that reports the binding of FFAs to the protein (21). Figure 4 directly demonstrates the basal FFA efflux from clonal β-cells by showing the increase in the ADIFAB fluorescence ratio. Glucose prevented FFA efflux (Fig. 4A), presumably by stimulating re-esterification. Lipolysis yields FFAs, which are converted to their active metabolite LC-CoA by acyl CoA synthetase (ACS). Triacsin C, an inhibitor of some isoforms of ACS (25), increased the release of FFAs when added under basal conditions (Fig. 4B) or after glucose addition (Fig. 4A). Control experiments showed that triacsin C itself neither altered the redox state nor qualitatively prevented the redox changes induced by glucose (Fig. 5). Also, it did not alter the Ca²⁺ changes induced by a variety of agonists (data not shown).

To determine whether the pHi effects of GLP-1 (which we attributed to the stimulation of lipolysis) corresponded to FFA exit from the cells, we monitored the release of FFAs to the external buffer using ADIFAB while simultaneously measuring the pH_i. Figure 6 directly demonstrates that GLP-1 increased FFA release into the medium; as in Fig. 1, GLP-1 caused acidification of the cytosol (Fig. 6A, bottom trace) and a small release of FFAs into the medium (Fig. 6A, top trace). FFA release in the presence of glucose plus GLP-1 was quite low, presumably because of glucose-promoted re-esterification similar to that seen in Fig. 4; however, the FFA production was unmasked by the addition of triacsin C, which sharply increased FFA release and intracellular acidification (Fig. 6B). The increases in ADIFAB fluorescence (top traces) showed a time dependence for the release of FFAs into the medium that corresponded to the time course of the decreases in pH_i (bottom traces). This implies that the FFAs can leave cells almost as quickly as they are generated.

HSL in the β-cell. To explain the cAMP-mediated stimulation of lipolysis, the presence of an HSL-like activity must be postulated. To confirm that the RNA for adipocyte HSL was expressed in islets, rat islet cDNA was used in a PCR with HSL-specific sense and antisense primers. The PCR products were separated by agarose gel electrophoresis and the gel was stained with ethidium bromide. An amplification product of the expected size (280 bp) was present in the rat islet cell lane (Fig. 7). DNA sequencing of the isolated band confirmed its identity to HSL. There was no DNA amplification in the negative control lane, which was lacking a cDNA template. There was also no amplification product in RNA samples processed without reverse transcriptase enzyme (data not shown), ruling out amplification from genomic DNA.

Inhibition of the incretin effect by a lipase inhibitor. Forskolin increased glucose-stimulated insulin secretion, and this incretin effect was progressively and largely diminished by 30 min preincubation of the cells with increasing concentrations of the lipase inhibitor orlistat (tetrahydrolipstatin) (26,27,28) (Fig. 8). Similar results were seen in a second experiment with a different passage of cells. Glucose-stimulated secretion itself was less potently affected by orlistat in the experiment illustrated by Fig. 8, and it was not affected at all in the second experiment. There was no effect of orlistat on basal secretion (data not shown). It should be noted that though orlistat acts in the intestinal lumen in vivo and is not taken up systemically (29), orlistat can nevertheless cross biological model membranes (30). Furthermore, orlistat inhibits HSL (28) in addition to pancreatic lipase, but it does not inhibit phospholipase A₂ (31).

The data presented here are consistent with a model (Fig. 9) in which GLP-1 activates adenyl cyclase, generating cAMP, which consequently stimulates HSL via PKA, thereby causing the breakdown of triglycerides and the generation of FFAs. The FFAs are converted to LC-CoA, either for oxidation in the mitochondria or for the synthesis of complex lipids (e.g., DAG or PA). Triacsin C interferes with this process by partially blocking the generation of LC-CoA and favoring the accumulation and efflux of FFAs. Support for this model is indicated by the actions of GLP-1 to 1) decrease pH_i, 2) increase the release of FFAs from the cell, and 3) stimulate FFA oxidation after overnight labeling of the triglyceride pool. Further support comes from the recent finding that high-fat feeding, which increases islet triglyceride content, leads to an exaggerated response to GLP-1 (32).

The suggestion that lipolysis (with the resulting production of FFAs and LC-CoA) may be involved in the incretin effect of GLP-1 is amenable with our hypothesis that both FFAs and cytosolic LC-CoA and other complex lipids are major effector signals in insulin secretion (12,13). The data also suggest that an important source of these lipid molecules is internal triglyceride stores, the regulation of which has not been previously investigated. The current studies indicate that cAMP may regulate lipolysis, and ultimately FFA and LC-CoA production, via HSL. This may explain why the pancreatic islet contains large triglyceride stores (9,10) and relies on basal FFA oxidation to provide most of its fuel needs (9). These triglyceride stores and their modulation may also play an important role in glucose homeostasis in vivo (14,15,33,34). During the course of this work, similar and more extensive evidence of HSL in islets was published by Mulder et al. (11).

It should be noted that PKA phosphorylation of either purified HSL or HSL in fat cell homogenates causes only a 2- to 3-fold increase in activity, far less than the 20- to 50-fold change in lipolysis in intact cells (35,36). This is attributed to the much greater importance of translocation of HSL to the substrate and changes in the accessibility of the substrate, probably also involving PKA-mediated phosphorylation of perilipin (37). Interestingly, PKA phosphorylation of purified lipase increases its activity only toward certain substrates (e.g., triglyceride and cholesteryl oleate), but not toward the diglyceride substrates that are commonly used in lipase assays and yield the highest rates of hydrolysis by HSL (35,38). Diglyceride substrates were used in the studies of HSL-like activity in β-cells by Mulder et al. (11). Although we have confirmed lipase activity in HIT cell homogenates using a diglyceride-based commercial kit, for the above reasons we have not pursued studies of assayed activity in homogenates. Monitoring the FFA release from intact cells, as we have done here, appears to be the more physiologically relevant means of measurement. Admittedly, although the HSL-like enzyme shown to be present in β-cells is the most likely candidate for mediating GLP-1—induced FFA release, it is possible that another lipase (e.g., a Ca²⁺-independent phospholipase) might be involved. Regardless of the identity of the lipase, the important finding for the proposed lipidmediated pathway of signal transduction in the β-cell is that GLP-1 causes FFA release.

GLP-1 acidified the β-cell under the conditions of our experiments; however, the acidification is unlikely to be important for its incretin action because secretion was not stimulated by GLP-1 in the absence of glucose where the decrease in pH_i was greatest, whereas in the presence of glucose the ability of GLP-1 to acidify was minimal. In contrast, glucose-stimulated secretion is accompanied by alkalinization of the β-cell (20,23). This alkalinization results from an early stimulation of pyruvate transport into the mitochondria (cotransport with a proton), as suggested by observations that blockage of the pyruvate transport with 3-hydroxycyanocinnamate prevented the pH change but not the subsequent glucose-induced rise in Ca²⁺ (20,23). Thus, the changes in pH_i may serve as useful indicators of metabolic events without necessarily playing an important role in stimulus-secretion coupling. In addition, our data indicate that the pH_i change did not act indirectly by causing changes in Ca²⁺ and that agonists that elevated cytosolic free Ca²⁺ did not change the pH_i.

The rise in FFA levels caused by GLP-1, as indicated by acidification or FFA efflux from the cell, is insufficient by itself to cause insulin secretion in the absence of glucose. Furthermore, glucose reduced the rate of release of FFAs, and hence probably the cellular level of FFAs, as well as the rate of FFA oxidation; these effects are presumably due to the ability of glucose to promote complex lipid formation and re-esterification. This postulation may suggest that a product formed from FFA but requiring glucose is involved in stimulus-secretion coupling.

Triacsin C was used here to help unmask FFA production by inhibiting both the activation of FFAs to LC-CoA and their further removal by either mitochondrial oxidation or complex lipid formation in the presence of glucose. In theory, triacsin C could be used to distinguish between effects mediated directly by FFAs or by LC-CoA or complex lipids. It has been reported that acutely administered triacsin C largely inhibited oxidation of exogenous FFAs yet had no effect on glucose-stimulated secretion (39); we have confirmed this and, likewise, we have found no consistent effect on the enhancement of secretion by GLP-1 or FFAs (data not shown). On the surface, this might seem to indicate that the FFAs, but not LC-CoA, were involved in the stimulation of secretion. However, the situation is complicated by the existence of pools of LC-CoA and various isoforms of ACS with differing sensitivity to triacsin C (25,40). Indeed, although FFA oxidation is strongly inhibited, the incorporation of FFAs into phospholipids is much less affected by triacsin C (25,39,40). Because glucose itself reduces FFA oxidation, the more relevant path of FFA metabolism for signal transduction is likely to be the formation of complex lipids. We propose that GLP-1 causes the release of FFAs and that this effect can provide more substrate for both FFA oxidation and complex lipid formation, as well as the observed enhanced FFA efflux from the cell; as noted above, high glucose levels tend to direct FFA carbon into the complex lipids, at the expense of the other two possible outcomes.

The potent inhibition of the incretin effect by the lipase inhibitor orlistat (Fig. 8) provides strong evidence for the involvement of PKA-stimulated lipolysis, as indicated in Fig. 9. Orlistat had less effect on glucose-stimulated secretion alone. This does not imply that glucose signaling does not involve a lipid-mediated component, merely that glucose modulation of lipid signals may not arise primarily from direct effects on lipolysis and thus FFA provision for LC-CoA synthesis. Instead, as discussed previously (12), glucose increases the malonyl CoA level, which in turn inhibits carnitine palmitoyl transferase-1 and reduces mitochondrial uptake and oxidation of LC-CoA. This leads to a rise in cytosolic LC-CoA and, together with the increased production of α-glycerophosphate from glucose, to increases in complex lipid formation.

These experiments were supported by National Institutes of Health Grants DK50662 (G.C.Y.), DK35914 (B.E.C.), DK46200 (B.E.C.), HL26335 (J.A.H.), and DK53064 (K.T.).