Biochemical Mechanism of Lipid-Induced Impairment of Glucose-Stimulated Insulin Secretion and Reversal with a Malate Analogue

Hyperlipidemia appears to play an integral role in loss of glucose-stimulated insulin secretion (GSIS) in type 2 diabetes. This impairment can be simulated in vitro by chronic culture of 832/13 insulinoma cells with high concentrations of free fatty acids, or by study of lipid-laden islets from Zucker diabetic fatty rats. Herein we show that impaired GSIS is not a simple result of saturation of lipid storage pathways, as adenovirus-mediated overexpression of a cytosolically localized variant of malonyl CoA decarboxylase in either cellular model results in dramatic lowering of cellular triglyceride stores, but no improvement in GSIS. Instead, the glucose-induced increment in “pyruvate cycling” activity (pyruvate exchange with TCA cycle intermediates measured by 13 C NMR), previously shown to play an important role in GSIS, is completely ablated in concert with profound suppression of GSIS in lipid-cultured 832/13 cells, while glucose oxidation is unaffected. Moreover, GSIS is partially restored in both lipid-cultured 832/13 cells and islets from ZDF rats by addition of a membrane permeant ester of a pyruvate cycling intermediate (dimethylmalate). We conclude that chronic exposure of islet β -cells to fatty acids grossly alters a mitochondrial pathway of pyruvate metabolism that is important for normal GSIS.


4
A major contributing factor to the development of type 2 diabetes is inadequate insulin secretion to compensate for insulin resistance. A hallmark of this β-cell dysfunction is the impairment and eventual complete loss of glucose-stimulated insulin secretion (GSIS). Hyperlipidemia, and the consequent accumulation of triglycerides (TG) and other lipid-derived intermediates in β-cells, is now well recognized as a variable that correlates with development of impaired insulin secretion (1)(2)(3)(4)(5)(6).
Furthermore, culture of pancreatic islets (3,7,8) or insulinoma cell lines (9) with elevated levels of free fatty acids in vitro results in loss of GSIS, and glucose sensing is also dramatically impaired in fat-laden islets from Zucker diabetic fatty (ZDF) rats (2,3).
However, a biochemical mechanism linking chronic exposure of islet cells to high levels of free fatty acids and impairment of GSIS has not emerged.
To gain more insight into this important issue, two independent model systems were exploited. First, we have recently described stable subclones of the rat insulinoma INS-1 cell line with robust GSIS, such as cell line 832/13 (10). As shown herein, chronic culture of these cells in 1 mM oleate:palmitate (2:1) causes profound impairment of GSIS. Second, islets from ZDF rats are both lipid-laden and poorly glucose responsive (3). Using these model systems, two hypotheses about the mechanism of lipid-induced impairment of GSIS were tested. The first is that accumulation of lipidderived metabolites caused by chronic exposure of β-cells to fatty acids plays a direct role in the functional impairment. To test this idea, we have employed a recombinant adenovirus encoding a variant, cytosolically-localized form of malonyl CoA decarboxylase (AdCMV-MCD∆5) (11) to lower malonyl CoA levels in lipid-laden cells.
by guest on March 24, 2020 http://www.jbc.org/ Downloaded from Application of this method caused a dramatic lowering of TG levels in both lipid-cultured 832/13 cells and in islets from ZDF rats, but failed to improve GSIS. This led us to test a second hypothesis based on our recent discovery of a critical link between pyruvate carboxylase (PC)-mediated pyruvate exchange with TCA cycle intermediates ("pyruvate cycling") and GSIS (12). This link between pyruvate cycling and GSIS was uncovered by NMR-based analysis of [U- 13 C] glucose metabolism in a set of variously glucose responsive INS-1-derived cell lines. More precisely, pyruvate cycling refers either to the "pyruvate/malate cycle", involving PC-catalyzed conversion of pyruvate to oxaloacetate, reduction of oxaloacetate to malate, and decarboxylation of malate to pyruvate via malic enzyme, and/or to the "pyruvate/citrate cycle", wherein the first and last steps are the same as in the pyruvate/malate cycle, but oxaloacetate formed in the PC reaction is converted to citrate, after which malate is regenerated via citrate lyase and cytosolic malate dehydrogenase (12). The NMR methods that we employ are not capable of distinguishing between these cycles, but do discriminate total cycling activity relative to TCA cycle flux. In the current study, we demonstrate that the profound impairment of GSIS that occurs in response of chronic exposure of 832/13 cells to fatty acids is accompanied by complete ablation of the normal glucose-induced increment in pyruvate cycling, with no change in the rates of glucose oxidation at basal or stimulatory glucose. removal of its N-terminal mitochondrial localization sequence and its C-terminal peroxisomal targeting sequence (AdCMV-MCD∆5; (11)). As controls, other groups of cells were treated either with a virus encoding a catalytically inactive form of MCD (AdCMV-MCD mut ; (11,14)), or the bacterial β-galactosidase gene (AdCMV-βGAL; (15) Malonyl CoA decarboxylase activity assay. MCD activity was determined as the rate of decarboxylation of malonyl-CoA to acetyl-CoA as previously described (18). In brief, the rate of acetyl-CoA formation was monitored by cleavage of its thioester bond by acetylcarnitine transferase over 5-10 minutes, a time period during which the rate of product accumulation was linear.  For studies of insulin secretion from ZDF islets, cells were washed with PBS and preincubated in HBSS containing 3 mM glucose for 1 h. Insulin secretion was then measured with the same static incubation protocol as described for 832/13 cells, using 2 uniformly sized islets/condition in triplicate, exposed to 3 mM or 16.7 mM glucose. (Acorn NMR, Fremont, CA). These multiplet areas were used to perform a 13 Cisotopomer analysis with the previously described program tcaCALC (19). The program was applied using the same model parameters as recently reported (12) to determine a metabolic profile for metabolism of [U -13 C 6 ] glucose in the TCA cycle. The R i values for other common substrates have been tabulated elsewhere (22). Total O 2 consumption by tissue may be defined as Q t = Q 0 + Q glucose , where Q 0 refers to O 2 consumption from oxidation of endogenous triglycerides or fats and Q glucose to O 2 consumption from oxidation of glucose. Similarly, total TCA cycle flux is defined as C t = C 0 + C glucose where C 0 refers to TCA cycle flux due to oxidation of endogenous substrates and C glucose to oxidation of glucose. Since the R i factor for each substrate relates O 2 consumption to TCA cycle flux, it follows that Q t = C 0 R 0 + C glucose R glucose . Given that the F Ci variables are defined by the fraction any given substrate makes to total acetyl-CoA entering the TCA cycle, F C0 = C 0 /C t and F C3 = C 3 /C t (Where F C0 and F C3 is the fraction of acetyl-CoA derived from fats or glucose respectively). These relationships can be combined to yield Q t /C t = F C0 R 0 + F C3 R glucose . Hence, if O 2 consumption (Q t ) can be determined as an absolute flux (using an oxygen electrode for instance), and the F Ci values measured by 13 (22). Thus, after measuring oxygen consumption, the fractional contribution of each substrate to acetyl-CoA and anaplerosis, C t is easily determined.

Oil
Q t /C t = F C0 R 0 +F C3 R glucose + yR a [1] It follows then that absolute glucose oxidation is simply F C3 C t,, and endogenous lipid   respectively). Figure 4A shows that this increase in MCD activity resulted in a 51 ± 5% decrease in TG stores, although levels in MCD-overexpressing fa/fa islets remained significantly elevated compared to islets from lean control Zucker rats (p < 0.05). Figure 4B shows that glucose caused a potent stimulation of insulin secretion in islets from lean control ZDF rats (8.2 ± 2.6-fold as glucose was raised from 3 to 16.7 mM). As expected, GSIS was modest in ZDF (fa/fa) control islets (2.1 ± 0.5-fold), and was not significantly enhanced by AdCMV-MCD∆5 treatment (3.1 ± 0.6-fold; p = 0.17). Arginine-stimulated insulin secretion was the same in the untreated and AdCMV-MCD∆5-treated fa/fa islets, suggesting that total secretory capacity was not affected by MCD overexpression (data not shown). We conclude that neither lipid-induced impairment of GSIS induced in tissue culture, or as occurs in lipidladen islets from fa/fa rats, is significantly reversed by depletion of TG stores in response to MCD overexpression.

Lipid-Induced Impairment of GSIS is Reversible by Removal of Lipids.
These findings raise the issue of whether islets or β-cell lines that are chronically exposed to high levels of fatty acids have undergone a permanent impairment in GSIS, perhaps indicative of a "lipotoxic" condition that is a prelude to programmed cell death. As one test of this idea, we performed studies on withdrawal of fatty acids from the medium after induction of impaired function to determine whether the impairment is permanent or reversible. To this end, 832/13 cells were cultured in 1  To overcome this concern in the current study, oxygen consumption was measured in parallel with the NMR experiments. Two important findings emerged from these experiments ( Figure 5B). First, chronic exposure to fatty acids raised oxygen consumption by 36% (p = 0.03) at 3 mM glucose relative to cells grown in the absence of lipids, confirming a previous report (9). Second, cells exposed to lipids exhibited a 30% decrease in oxygen consumption as glucose concentration was raised from 3 to 12 mM (p = 0.01), whereas this did not occur in control cells. These rates of oxygen consumption were then used to convert the relative fluxes measured by NMR into absolute measures of glucose oxidation (from the fraction of acetyl-CoA derived from [U-13 C]glucose), endogenous substrate oxidation (from the fraction of unlabeled substrate contributing to acetyl-CoA), and pyruvate cycling flux (22). As shown in Figure 5C, pyruvate cycling activity rises from 1. We realize that it may be surprising to some readers that an increase in glucose level from 3 to 12 mM glucose did not cause a significant increase in oxygen consumption in control cells in Figure 5B, as other groups have shown that oxygen consumption increases in response to glucose stimulation in β-cell lines and isolated islets (9,24,25). We believe that this discrepancy is explained by the fact that our oxygen consumption experiments were performed under conditions designed to mimic those of the pyruvate cycling and insulin secretion measurements shown in the various panels of

Discussion
The correlation between exposure of islet β-cells to elevated lipid concentrations and development of impaired insulin secretion has been well established in several laboratories (1,3,4,6), leading to the idea that this pathway may be a major contributing factor to β-cell failure of type 2 diabetes (3). However, the mechanism of this effect has not been clarified. In the current study, we have investigated two hypotheses about the development of lipid-induced impairment in islet function. First, we have evaluated the possibility that saturation of lipid storage pathways plays a direct role in loss of GSIS.
Support for this idea comes from studies in which TG content has been inversely correlated with glucose responsiveness (7,26). More recent work has suggested that lipid-induced impairment of β-cell function and accumulation of TG requires coexposure to elevated glucose concentrations, although the main variable measured in these studies has been insulin gene expression and synthesis rather than GSIS (6,(27)(28)(29).
The same investigators also reported that adenovirus-mediated expression of diacylglycerol acyltransferase in rat islets led to accumulation of TG and coincident impairment of GSIS (30). However, the experiments reported here show that the correlation between TG content and loss of insulin secretion has no mechanistic significance, and that TG levels are more likely serving simply as a marker of another fatmediated pathophysiologic event. We base this conclusion on our finding that overexpression of MCD either in fat-cultured 832/13 cells or in fat-laden islets from ZDF rats effectively lowers cellular TG levels, with no significant restorative effect on GSIS.
We recognize that it remains possible that MCD expression failed to remove lipid-derived metabolites other than TG that are the real causal agent for β-cell impairment.
For example, the fatty acid-derived metabolite ceramide has been linked both to impairment of insulin gene expression (31) and increased rates of β-cell apoptosis (32) in lipid-cultured β-cells. This issue will require further investigation.
Given that TG overstorage is not the direct cause of lipid-induced impairment of insulin secretion, it is somewhat difficult to develop further models from a survey of the literature in the field due to significant disagreement. For example, one group has reported that islets exposed to fatty acids experience a reduction in pyruvate dehydrogenase (PDH) activity in concert with a fall in glucose oxidation, leading to the suggestion that a glucose-fatty acid (Randle) cycle is operative in such cells (7,26,33). However, studies from two other laboratories failed to demonstrated significant lipid-induced impairment of PDH activity in INS-1 cells (9) or rat islets (34). In the Randle hypothesis, a rise in citrate is suggested to slow glycolytic flux via inhibition of phosphofructokinase activity. However, citrate levels are reported to be either unchanged (9) or decreased (35), and phosphofructokinase activity to be increased (36) in various β-cell preparations following lipid exposure. More recently it has been reported that long-term exposure of MIN-6 mouse insulinoma cells to fatty acids results in a reduction of the levels of PC protein (37). These authors also suggested that the consequence of such a lowering of PC might be a reduction in "malate-pyruvate shuttle flux", but this conclusion appeared to be based solely on a decrease in NAD(P)H autofluoresence in fat-cultured cells, rather than any direct measurement of a metabolic pathway. They further proposed that a decrease in NAPDH content may be involved in the fat-induced impairment in GSIS, but this speculation was based on data obtained by a method that cannot discriminate by guest on March 24, 2020 http://www.jbc.org/ Downloaded from NADH from NADPH. In contrast, another group has reported no change in PC V max in fatcultured islets, and further speculated that malate-pyruvate shuttle flux would be increased rather than decreased due to a 60% rise in intracellular pyruvate concentrations (34). The same group has also reported an increase in PC Vmax in islets from nondiabetic Zucker fatty rats, and have suggested that this would lead to increased pyruvate cycling, thereby possibly explaining the enhanced insulin secretion of such islets that compensates for insulin resistance (38). However, this conclusion was based on static measurement of enzyme activities and concentrations of selected metabolic intermediates rather than any direct measurement of metabolic flux.
In light of this confusion, a more comprehensive method for metabolic analysis of lipid-exposed versus normal cells was required. We have recently used 13 C NMR to analyze pathways of pyruvate metabolism in mitochondria, leading to the discovery that the activity of PC-catalyzed pyruvate cycling pathways can be used to distinguish robustly glucose responsive from poorly glucose responsive INS-1-derived cell lines (12). These findings are in agreement with another study employing radioisotopic tracers that demonstrated pronounced differences in glucose-driven anaplerosis in purified βversus α-islet cell preparations (39). Development of the NMR-based methods has allowed us to test our second hypothesis that chronic exposure of β-cells to lipids results in an alteration in PC-catalyzed pyruvate cycling activity.
Our approach has uncovered several metabolic perturbations that occur in β-cells in response to chronic exposure to elevated lipid concentrations. First, we observe a rise in oxygen consumption at 3 mM glucose that occurs in concert with an increase in endogenous substrate oxidation. Second, we find a loss of the glucose-induced increment in pyruvate cycling activity due to a large increase in cycling at basal glucose levels. Interestingly, these changes occur in the absence of any significant change in the rate of 13 C glucose oxidation in lipid-cultured versus control cells, at either 3 or 12 mM glucose.
We interpret these data as follows. The coordinate increase in oxygen consumption ( Figure 5B) and endogenous fuel oxidation ( Figure 5D) that occurs at 3 mM glucose in response to chronic lipid exposure is most likely explained by an increase in fatty acid oxidation. This follows from the obvious increase in supply of this substrate in fat-cultured cells, and is consistent with recent reports of increased expression of enzymes of lipid oxidation in β-cells in response to chronic lipid exposure (40,41). In liver, an increase in fatty acid oxidation has been linked to suppression of PDH activity via accumulation of ATP, NADH, and acetyl-CoA, part of the Randle mechanism. As discussed earlier, two laboratories have reported that this pathway is not operative in β-cells (9,34), and our data on a lack of change in glucose oxidation in response to chronic lipid culture is consistent with these findings. However, acetyl-CoA generated from lipid oxidation is also known to influence pyruvate metabolism via its capacity to activate PC. It would appear that this mechanism is retained in β-cells, given the large increase in PC-catalyzed pyruvate cycling activity that occurs at basal glucose in response to chronic lipid culture ( Figure 5C). This rise in basal pyruvate cycling activity eliminates the normal glucose-induced increment in this parameter, such that yet to be identified byproducts of this pathway that signal for insulin secretion are never generated. In our earlier study of several variously glucose responsive INS-1-derived cell lines, we noted a strong correlation between by guest on March 24, 2020 http://www.jbc.org/ Downloaded from pyruvate cycling and glucose responsiveness, with no correlation between GSIS and the fractional contribution of glucose to acetyl-CoA production, a measure of TCA cycle activity (12). This original observation is fully confirmed in the current study, as chronic exposure to lipids caused a profound impairment of GSIS ( Figure 5A) in concert with the loss of a glucosemediated increment in pyruvate cycling ( Figure 5C), but with no effect on the rate of glucose oxidation ( Figure 5E).

One potential issue that arises with this model is that overexpression of MCD is
expected to increase fatty acid oxidation by lowering of malonyl CoA levels, yet this maneuver has no effect on GSIS in 832/13 or parental INS-1 cells (11,18 (47). We therefore investigated whether the membranepermeant malate ester DMM could overcome the lipid-induced impairment in insulin secretion.
Remarkably, inclusion of DMM during the secretion assay almost completely restored GSIS in lipid-cultured 832/13 cells, and also significantly improved glucose responsiveness in lipidladen ZDF islets. Taken together, these data provide strong support for the idea that lipidinduced impairment of GSIS is caused at least in part by alteration of the metabolic fate of pyruvate. The effects of lipid to cause this diversion may be dependent upon concomitant overexposure of cells to elevations in glucose, based on work summarized earlier (6,27,29).
All of the experiments summarized in the current study were performed in the presence of       and control cells exposed to 3 or 12 mM U-13 C glucose for 4 h. The symbol * indicates differences between lipid-cultured and control cells, at 3 and 12 mM glucose, respectively, with P < 0.01. Panel B. Oxygen consumption was measured as described in Material and Methods. The symbol * indicates that oxygen consumption was significantly higher in FFA cultured cells than in control cells at 3 mM glucose, with p = 0.03; the symbol ** indicates that oxygen consumption was significantly lower in FFA cultured cells at 12 mM glucose than at 3 mM glucose, with P < 0.01. Panel C.
Pyruvate cycling activity measured by 13   each performed in triplicate. The symbol * indicates a significant increase in insulin secretion at 12 mM glucose in cells exposed to DMM, relative to cells exposed to 12 mM glucose in the absence of DMM, with p < 0.05.