Direct Effect of Cholesterol on Insulin Secretion: A Novel Mechanism for Pancreatic β-Cell Dysfunction

C57BL/6J mice (Harlan Laboratories, Indianapolis, IN) were used, unless otherwise noted. Apolipoprotein E (apoE)-deficient mice (ApoE^−/−) (B6.129P2-Apoe^tm1Unc/J), matching control mice (C57BL/6J), and nNOS knockout mice (B6.129S4-Nos1^tm1Plh/J) were purchased from The Jackson Laboratory (Bar Harbor, ME). ob/ob;apoE^−/− mice (on C57BL/6J background) were produced by intercrossing ob/+ and apoE^−/− mice (16). All mice were kept on a rodent chow (LabDiet 5001, 12% of calories from fat) and cared for according to the guidelines of the Vanderbilt Institutional Animal Care and Use Committee. 832/13 INS-1 cells were a kind gift from Dr. Newgard (Duke University, Durham, NC). βTC3 and 832/13 INS-1 insulinoma cells were cultured and prepared for experiments as previously described (17,18). Pancreatic islets were isolated and cultured using methods developed in our lab (19).

Cholesterol content was quantified in a 96-well plate by a fluorometric method using an enzyme-coupled reaction provided by the Amplex Red Cholesterol Assay kit (Molecular Probes). After isolation, islets were divided into two groups for cholesterol and insulin assays. For cholesterol measurement, material from each tube containing 10 islets was subjected to lipid extraction with chloroform/methanol (2:1; vol/vol), dried down to a thin film, and resuspended in 60-μl 1× working solution (Amplex Red Cholesterol Assay kit) supplemented with 0.1% Triton X-100. From that, 50 μl was used for cholesterol assay and 10 μl for protein assay (BCA Protein Assay; Pierce, Rockford, Il).

Static incubations of pancreatic islets for insulin secretion measurements were done as previously described (19). Insulin radioimmunoassays were performed by the Diabetes Research and Training Center Hormone Core Resource at Vanderbilt University. Mouse serum total cholesterol, triglycerides, and free fatty acids were measured by the Vanderbilt Mouse Metabolic Phenotyping Center Lipid Core.

For metabolic depletion, islets were grown for 3 days in metabolic depletion medium (RPMI-1640; similar to the growth medium but with 10% lipoprotein-deficient serum in place of FBS, supplemented with 200 μmol/l mevalonate [required to maintain cell viability] and 10 μmol/l mevastatin) to block cholesterol synthesis and deplete cholesterol stores (20). For depletion by methyl-β-cyclodextrin (MβCD) (Sigma), islets and β-cells were incubated with 10 mmol/l MβCD at 37°C for 1 h, which removed cholesterol from both plasma membrane and intracellular stores because intracellular cholesterol effluxes efficiently to the plasma membrane where it is quickly removed by MβCD as the acceptor (9). To cholesterol overload, cells were incubated with 10 mmol/l soluble cholesterol (Sigma; 1 g contains ∼40 mg cholesterol) at 37°C for 1 h.

Enzymatic activity of GK was measured by a continuous spectrophotometric rate determination method (21) using 60 mmol/l Tris, 20 mmol/l MgCl, 2.5 mmol/l dithiothreitol, 100 mmol/l KCl, 4 mmol/l adenosine 5′-triphosphate, 0.9 mmol/l β-nicotinamide adenine dinucleotide phosphate, and 10 units glucose 6-phosphate dehydrogenase. One unit of GK activity is defined as phosphorylation of 1 μmol d-glucose to d-glucose 6-phosphate per minute at pH 9.0 at room temperature. GK activity was measured by subtracting the activity measured at 0.5 mmol/l glucose from that at 100 mmol/l glucose to account for hexokinase activity in the cell lysates.

Cells were collected in cold PBS and resuspended in cold lysis buffer (20 mmol/l HEPES, 100 mmol/l NaCl, 1 mmol/l EDTA, 1% cholate, 1% Triton X-100, and 1× protease inhibitor cocktail for mammalian cells [Sigma]). Cold SDS loading buffer (0.05 mol/l Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 100 mmol/l dithiothreitol, and 0.1% bromphenol blue) was added to the cell lysate and incubated for 5 min at 0°C. Samples were loaded on 6% polyacrylamide gels, subjected to electrophoresis at 0°C, and visualized by immunoblotting with nNOS monoclonal antibody (BD Transduction Laboratories) and peroxidase-conjugated secondary antibody. To separate the cytoplasmic and membrane-bound fractions, digitonin permeabilization was performed as described previously (22), with the exception of using 10 μg/ml instead of 20 μg/ml digitonin.

Results are expressed as means ± SD. Unpaired two-tailed Student's t tests were performed for comparisons between treated and control groups. Differences were considered significant at P < 0.05. In cholesterol and insulin secretion assays, n denotes the number of mice used for each condition. Samples of 10 islets each were taken from each mouse. In all other experiments using cultured β-cells, n denotes the number of times each experiment was repeated with different plates of cells for each condition.

Traditionally, a high-fat diet is used to increase serum cholesterol levels in mice. However, this is also accompanied by elevations in FFA levels. To circumvent this issue, we used apoE-deficient (apoE^−/−) mice fed a standard diet. Compared with control C57BL/6J animals, apoE^−/− mice have elevated plasma cholesterol levels without changes in plasma FFA. Furthermore, when the apoE^−/− mice are made obese by placing them on a leptin-deficient ob/ob background, plasma cholesterol levels are further elevated, again without concomitant increases in FFA compared with ob/ob controls (Table 1). Because apoE is not present in the pancreatic islets of wild-type mice (23), its absence does not directly affect islet function when studied in vitro, and we observed similar islet morphology between apoE^−/− and control mice (apoE^−/− vs. C57BL/6J and ob/ob;apoE^−/− vs. ob/ob). Cholesterol measurements in pancreatic islets show that an increase in serum cholesterol (Table 1) leads to increased islet cholesterol (Fig. 1A). Although total insulin content does not vary significantly among different mouse models (Fig. 1B), the increase in islet cholesterol causes a dramatic reduction in the islets’ ability to secrete insulin in response to basal and high glucose concentrations (Fig. 1C). This result indicates that cholesterol directly impacts GSIS independent of the cellular FFA content.

Because plasma triglyceride levels are also elevated in our mouse models relative to their respective controls (Table 1), we studied isolated islets in which cholesterol can be specifically manipulated. To determine whether direct cholesterol depletion could enhance GSIS from isolated C57BL/6J islets, MβCD was used for 1 h to cause acute depletion of cellular cholesterol (9) and mevastatin (a cholesterol biosynthesis inhibitor) to cause metabolic depletion over several days in culture (20). With these treatments, islet cholesterol was reduced by 30–40% (Fig. 2A), whereas total insulin content remained unaffected (Fig. 2B). Direct cholesterol depletion from isolated islets significantly enhances insulin secretion under both basal and high glucose concentrations (Fig. 2C). Cholesterol depletion of the ob/ob;apoE^−/− islets using MβCD also partially restores GSIS (Fig. 1C). GSIS (20 mmol/l glucose) is shown as a function of islet cholesterol for the genetically modified mouse models (Fig. 1D) or isolated wild-type islets with different cholesterol levels (Fig. 2D). Although many factors affect insulin secretion, in all of these cases, the islet cholesterol level is inversely correlated to β-cell GSIS.

To investigate the molecular mechanisms underlying cholesterol-dependent GSIS, we turned to cultured pancreatic β-cells. Both 832/13 INS-1 and βTC3 cells secrete insulin in a regulated manner (18,24). INS-1 cells are used for the functional experiments because of their robust responsiveness to glucose (18), and βTC3 cells are used for the imaging experiments because of their low autofluorescence. The capability to manipulate cholesterol in these insulinoma cells is greatly improved relative to that in islets, so that cholesterol overloading using soluble cholesterol (cholesterol loaded on MβCD) is possible (Fig. 3A). In INS-1 cells, using MβCD to deplete cholesterol before glucose stimulation increases insulin secretion at all tested glucose concentrations (Fig. 3B). In contrast, cholesterol overload suppresses GSIS (Fig. 3B). Similar results are observed in βTC3 cells for changes in intracellular Ca²⁺, a critical signaling molecule in GSIS (Fig. 3C). The increase in intracellular Ca²⁺ on glucose stimulation is reduced in islets from ob/ob;apoE^−/− versus ob/ob mice (Fig. 3D). To determine whether cholesterol depletion alone is sufficient to stimulate insulin secretion, we used a fluorescent zinc indicator (FluoZin-3) to monitor insulin release from MβCD-treated βTC3 cells. FluoZin-3 has been successfully used to study insulin release in pancreatic β-cells because insulin and Zn²⁺ are co-released by exocytosis (25). When 10 mmol/l MβCD is added to βTC3 cells kept in 2 mmol/l glucose, a rise in FluoZin-3 fluorescence indicates the initiation of insulin secretion after cholesterol removal, even in the absence of other insulin secretagogues (Fig. 3E).

Because cholesterol directly affects GSIS, we examined the glucose-stimulated signal transduction pathway to see where cholesterol action occurs. To determine whether cholesterol is coupled to proximal glucose metabolism, we monitored the effect of MβCD treatment on intracellular Ca²⁺ in the presence or absence of a nonmetabolizable glucose analog, 2-deoxyglucose (2-DG). As shown in Fig. 3F, adding MβCD to βTC3 cells in 2 mmol/l glucose raises intracellular Ca²⁺. However, the increase in intracellular Ca²⁺ caused by cholesterol depletion is blunted when glucose is replaced by 2-DG, suggesting that glucose metabolism is required for initiating insulin secretory response in cholesterol-depleted cells. We also used NADPH autofluorescence to assay cellular redox state (26). Pretreating βTC3 cells with MβCD leads to a slightly higher NADPH production on 20 mmol/l glucose stimulation compared with the control cells (Fig. 4A). Overloading with cholesterol before glucose stimulation decreases NADPH production (Fig. 4A). Compared with ob/ob mice, NADPH production in islets from ob/ob;apoE^−/− mice is significantly reduced (Fig. 4B). As with Zn²⁺ secretion (Fig. 3E) and intracellular Ca²⁺ (Fig. 3F), cholesterol depletion with MβCD alone initiates NADPH production (Fig. 4C). NADPH fluorescence does not increase in response to MβCD when 2-DG is present or when soluble cholesterol is added to the cells (Fig. 4D). These results suggest that the effect of cholesterol depletion using MβCD on intracellular Ca²⁺ increase, NADPH production, and insulin secretion (indicated by Zn²⁺ release) involves proximal glucose metabolism.

GK critically regulates GSIS by exerting tight control over glucose metabolism. Its unique glucose-sensing property underlies the ability of pancreatic β-cells to respond to plasma glucose fluctuations. Because of this, even small changes in GK activity have an impact on the rate of GSIS (27). This is supported by many studies of inactivating mutations in the gene encoding GK in humans with maturity-onset diabetes of the young (28). It has been shown on glucose stimulation that GK dissociates from the insulin granules and is released to the cytoplasm, where it is activated (29), as depicted in Fig. 5A. In this study, pretreating INS-1 cells with MβCD enhances GK activity under both basal (3 mmol/l) and high glucose (20 mmol/l) concentrations (Fig. 5B). In contrast, cholesterol overloading decreases GK activity on glucose stimulation (Fig. 5B). We next examined GK translocation by immunoblotting of the pellet (membrane) and soluble (cytoplasmic) fractions of digitonin-permeabilized cells under different conditions. Translocation of GK from the membrane bound to the cytoplasmic fraction is observed on cholesterol depletion (Fig. 5C). Further, glucose stimulation results in an additional increase in GK activity and translocation in MβCD-treated cells (Fig. 5B and C), which may suggest that cholesterol depletion and glucose stimulation utilize separate mechanisms for GK regulation.

Previous evidence indicates that glucose-stimulated GK translocation involves S-nitrosylation of by nNOS (29). To determine whether nNOS activity impacts GK in cholesterol-depleted cells, we examined nNOS activity using a fluorescent NO indicator, DAF-FM (4-amino-5-methylamino-2′,7′-difluorofluorescein). Figure 6A shows that cholesterol depletion before insulin stimulation prevents NO production observed in control cells. Insulin was used here as the stimulant because it caused a more rapid change in nNOS activity than glucose (29). The inhibition in NO production is consistent with the observation that disruption of lipid microdomains by MβCD results in reduced NO release from insulin-secreting cells (15). Adding MβCD to βTC3 cells in 2 mmol/l glucose leads to a decrease in DAF-FM fluorescence, indicating that cellular NO content is reduced on cholesterol depletion (Fig. 6B). These results suggest that nNOS activity is inhibited by cholesterol depletion.

Since cholesterol modulation using MβCD occurs within 30 min (30), its effect on nNOS and GK activity is unlikely to work through changes in gene transcription or protein stability. To determine how cholesterol depletion might decrease nNOS activity and release GK from insulin granules, we looked at the role of cholesterol-rich membrane microdomains in determining nNOS properties. nNOS is localized to insulin granules through its interaction with the transmembrane protein insolinoma-associated protein 2 in pancreatic β-cells (31). Characterization of nNOS has shown that the native protein is a homodimer (32), and dimerization is absolutely required for nNOS activity (33). Thus, we hypothesized a model where cholesterol depletion disrupts nNOS dimers, which in turn weakens the association between nNOS and GK, thereby releasing GK to the cytoplasms, where it becomes activated. Excess cholesterol, on the other hand, stabilizes nNOS dimers and keeps GK bound to insulin granules, where it is less active (Fig. 6C).

To test this hypothesis, we immunoprecipitated nNOS from cell lysates and probed by immunoblot for GK. As shown in Fig. 7A, cholesterol depletion by MβCD treatment results in dissociation of GK from nNOS under both low and high glucose. In contrast, cholesterol overloading increases association of GK with nNOS on glucose stimulation. There is an additional weakening of the interaction between GK and nNOS when MβCD-treated cells are stimulated with high glucose (Fig. 7A), again possibly suggesting that nNOS regulation of GK by cholesterol depletion is different from that by glucose stimulation. Our hypothesis is further supported by the fact that GSIS is greatly enhanced in nNOS knockout mice (Fig. 7B), consistent with the idea that a weaker association of GK with insulin granules (due to the lack of granule-associated nNOS) leads to greater GK activity in β-cells. Using low-temperature SDS-PAGE (34), we show that there is less nNOS dimerization in cholesterol-depleted cells compared with untreated cells and that the opposite is observed with cholesterol overloading (Fig. 7C). We also measured the nNOS dimer and monomer distribution in the membrane-bound and cytoplasmic fractions using digitonin permeabilization. In both control and cholesterol-depleted cells, there are more nNOS dimers in the pellet fraction, suggesting that the dimer structure is enriched in the insulin granule membranes (Fig. 7D). Furthermore, the nNOS oligomeric state dramatically shifts from predominately dimers to mostly monomers in the membrane-bound fraction on cholesterol depletion. These results show that cholesterol depletion disrupts nNOS dimers associated with insulin granules and leads to decreased nNOS activity (Fig. 6A and B). This is consistent with the previous finding that disruption of the dimeric state of nNOS by its inhibitors affects the secretagogue-induced insulin response (35). Although the exact mechanism by which nNOS dimerization is affected by membrane structure remains unknown, it is conceivable that clustering of insolinoma-associated protein 2 in cholesterol-rich domains effectively increases the local concentration of nNOS, thus promoting dimer formation.

Our results indicate that excess cellular cholesterol plays a direct role in pancreatic islet dysfunction and may well be a key factor underlying the progression of type 2 diabetes. Using different animal models, we show that elevated serum cholesterol leads to increased cholesterol in pancreatic islets. More importantly, islet cholesterol levels directly and significantly impact the extent of GSIS, independent of FFA levels. This has great implications because it indicates the existence of a novel mechanism linking hyperlipidemia and the pathogenesis of type 2 diabetes that is independent of FFAs. It also suggests that the regulation of cellular cholesterol may be a potential target for therapeutic intervention aimed at preserving or improving GSIS function in pancreatic β-cells.

It has been shown that VLDL is markedly increased in apoE^−/− mice (36), and VLDL exposure leads to a decrease in insulin mRNA levels in βTC3 cells (37). However, we did not find a significant difference between control and apoE^−/− mice in either plasma insulin levels (Table 1) or total islet insulin content (Fig. 1B). The discrepancy might be due to the difference in the cells used (cultured insulinoma cells vs. islets), cholesterol-to-triglyceride ratio of the VLDL particles (triglyceride-rich human VLDL vs. triglyceride-poor apoE^−/− VLDL), or the length of exposure time (short term vs. chronic). ApoE deficiency is shown to induce greater insulin sensitivity (38). This may explain the relatively normal glucose tolerance in apoE^−/− mice despite the reduced GSIS observed in our study.

Because of the profound effects of cholesterol on lipid organization and cellular functions, cells have formulated comprehensive mechanisms to maintain membrane cholesterol levels within a narrow range (39). Among all the signal transduction pathways in which cholesterol may be involved in β-cells, we have focused on the role of cholesterol-rich membrane domains in GSIS and one molecular mechanism in particular involving nNOS regulation of GK. Our data show that cholesterol depletion results in a change of nNOS dimer-to-monomer ratio, which in turn causes translocation of GK from insulin granules to the cytoplasm, where it is activated. In contrast, excess cholesterol prevents GK activation by keeping GK bound to insulin granules. Further studies are underway to provide additional details for other GSIS pathways that may also be affected by cholesterol modulation.

Membrane microdomains have raised great interest because they are proposed to be involved in many cellular functions, including signal transduction, protein trafficking, and cell polarization, as well as pathogenesis of many diseases (40). Cholesterol is an essential component of many putative membrane domains, and removal of cholesterol disrupts the function of proteins residing in these domains. It has been shown that ΜβCD treatment results in enhanced GSIS from HIT-T15 hamster insulinoma cells, due to redistribution of voltage-gated K⁺ channel and soluble N-ethylmaleimide–sensitive factor attachment protein receptor proteins of the putative membrane microdomains (13). However, our results indicate that glucose metabolism is directly affected by cholesterol depletion in addition to any changes in plasma membrane protein distribution that may occur. Our data also suggest that the function of two proteins critically involved in GSIS, nNOS, and GK depends on the cholesterol-enriched lipid environment of insulin granules. nNOS is found to be associated with granule membranes in many cell types. In neurons, MβCD redistributes nNOS from the Triton X-100 insoluble to the soluble fraction, and cholesterol repletion reverses nNOS localization (41). Several lines of evidence have shown that some intracellular membranes not only contain high amounts of cholesterol but also form detergent insoluble domains (42). For example, cholesterol makes up ∼70% of the lipids in bovine pituitary secretory granule membranes (43). Cholesterol offers the rigidity necessary for membrane curvature during secretory granule biogenesis. Insulin is packaged into dense core granules, which bud from the trans-Golgi network containing cholesterol and sphingolipid-rich microdomains (44).

Although the relationship between elevated cholesterol and diabetes has not been extensively examined, there is some evidence supporting a potential link between the two. First, a locus on chromosome 9 is associated with both the control of cholesterol levels and diabetes (45). Second, blood lipoprotein levels are an indication of the plasma cholesterol content and have been confirmed as important predictors for the onset of type 2 diabetes (46). In a recent study (46) aimed at examining the link between the use of a statin, which lowers serum cholesterol concentration, and subsequent risk of developing diabetes, it was shown that preventative administration of pravastatin reduced the occurrence of type 2 diabetes by 30%. However, results from this type of study are thus far inconclusive. Even though obesity has been identified as an important risk factor for type 2 diabetes, only a small percentage of obese patients will eventually develop the disease. Research suggests that predisposition to the development of type 2 diabetes may be due to genetic differences, although the specific genes involved have not yet been identified. Type 2 diabetes has been associated with high synthesis and low absorption of cholesterol, indicating a link between glucose and cholesterol metabolism (47–50). Indeed, our study opens the possibility that genetic variations relating to cellular cholesterol management may play a role in this genetic predisposition for type 2 diabetes.

This work is supported by grants from the National Institutes of Health (DK67821 to S.C.G., DK53434 to D.W.P., U24-DK59637 to the Metabolic Phenotyping Center, and DK-20593 to the Vanderbilt Diabetes Training and Research Center), the American Heart Association (0330011N to A.H.H.), and the American Diabetes Association (1-04-JF-20 to A.H.H.).

We thank Dr. Kevin D. Niswender for helpful discussions and Drs. Dick Goodman, Britton Chance, Roland Stein, Owen McGuinness, Al Beth, and Alan Cherrington for critical reading of the manuscript.