Chronic Effects of Fatty Acids on Pancreatic β-Cell Function: New Insights From Functional Genomics

Since the initial observations of Zhou and Grill (18), many studies have now confirmed that isolated islets, or β-cell lines, cultured with elevated fatty acids (FAs) for longer than 24 h display an increased basal secretion and, generally, a blunted response to high glucose (19–24). Additionally, there is direct evidence that even moderately elevated FAs, especially palmitate, are toxic to β-cells (15,16). On the other hand, chronic fat-feeding only results in diabetes in animals with certain genetic backgrounds (25–27). This is true even in the db/db mouse, which is massively obese and insulin resistant because of the mutated leptin receptor; these animals are only diabetic on a C57BL/KsJ, but not C57BL/6, background (25). The rat equivalent of this model, the ZDF (Zucker) rat, has been actively exploited by Unger and colleagues (8,15,28) to advance the so-called lipotoxicity hypothesis. In addition to displaying an enhanced basal secretion, islets derived from these animals had markedly increased triglyceride content, and this was true even in vitro when cultured with identical FA concentrations. Moreover, interventions (whether in vivo or in vitro) that ablated islet-lipid deposition also reversed basal hypersecretion (28).

There are three major levels at which lipid oversupply might contribute to β-cell decompensation. The first level is metabolic in that FA oxidation might disrupt glucose metabolism via Randle cycle effects and thereby inhibit glucose-stimulated insulin secretion (GSIS). The partial reversal of secretory defects by FA oxidation inhibitors has been taken as evidence of such a mechanism (10,18). However, more comprehensive metabolic analyses have shown that secretory defects can occur in the absence of a functional Randle Cycle, at least in clonal β-cells (20,22). On the other hand, metabolic effects do probably contribute to β-cell cytotoxicity, because de novo synthesis of ceramide from palmitate appears important in this context. The second level is that of signal transduction. Unlike lipid-induced insulin resistance, in which signaling interfaces such as protein kinase C (PKC) and stress-activated kinases have been implicated (29,30), there has been little investigation of these aspects in terms of β-cell dysfunction. The time of onset of the defect in GSIS (>24 h) has contributed to the neglect of this level, as well as arguing against the relevance of the metabolic level.

The third level is that of transcription. This is consistent with the knowledge that FAs and their acyl CoA derivatives are important ligands for several classes of nuclear receptors (31). Furthermore, the recent demonstrations by transcript profiling that FAs regulate the expression of hundreds of β-cell genes (24,32) has extended earlier work on the transcriptional regulation of a few important candidate genes (33,34). Indeed over- or underexpression of lipid-binding transcription factors, such as sterol response element-binding proteins and peroxisome proliferator-activated receptors, has led not only to coregulation of many of these candidate genes, but has also reproduced some of the hallmark features of secretory dysfunction (17,35,36). Finally, overexpression of some of the candidate genes themselves, such as carnitine palmitoyl transferase 1 and uncoupling protein 2, is also sufficient to promote β-cell dysfunction (37,38). For these reasons, we and others (33–36) contend that most, if not all, effects of FAs to cause β-cell decompensation are mediated by altered gene expression.

We have comprehensively analyzed both the functional (phenotypic) and transcriptional adaptations that occur in β-cells chronically exposed to FAs (24). Our chosen model is the glucose-responsive insulin-secreting cell line MIN6 (39). These cells possess the advantage over islets of cellular homogeneity, and their glucose-responsive genes were already defined before our studies (40). Although displaying secretory alterations similar to those observed in islets and other β-cell lines (18–22), we obtained several novel findings and potential insights. First, our transcript profiling provided the first comprehensive analysis of changes in global β-cell gene expression due to chronic exposure to the major circulating saturated and unsaturated FAs palmitate and oleate, respectively (24). Most importantly, this work conclusively demonstrated that the two FAs were not equivalent in their effects. Thus, of ∼200 genes whose expression was altered >1.8-fold, palmitate altered twice as many genes as oleate, and only a quarter of the total were coregulated by both FAs. Second, our work demonstrated that exposure to palmitate and oleate exerted differing effects on secretory function, in addition to regulating the expression of both consistent and exclusive sets of genes. Third, the effects of FA pretreatment varied depending on the nature of subsequent secretory stimulus.

Our detailed analyses established that there were at least four definable effects of FA pretreatment on subsequent secretory responses, and only two of which were widely documented beforehand. The first effect was the well-described enhancement of basal secretion, although our results now suggest that this is more pronounced in cells pretreated with palmitate rather than oleate (24). The second alteration was a selective inhibition of GSIS. This is again widely observed, but our findings suggest that when allowance is made for loss of insulin content, this is only genuinely observed with oleate pretreatment (24). In contrast, chronic exposure to palmitate slightly increases GSIS. However, this increase is relatively modest compared with a third (and more novel) alteration: a robust upregulation of distal steps in the secretory process, beyond the generation of coupling factors derived from metabolism of nutrient secretagogues (24). Again, this was more pronounced with palmitate than oleate. These three alterations have now been confirmed and extended in a new series of analyses using the nonmetabolizable FA bromopalmitate (Fig. 1). In these experiments, MIN6 cells were pretreated for 48 h at 5.6 mmol/l glucose with either BSA control or BSA coupled with 0.4 mmol/l bromopalmitate. Cells from each group were then stimulated for 1 h in Krebs-Ringer bicarbonate buffer at either 2.8 or 16.7 mmol/l glucose or 2.8 mmol/l plus other additions as described. Pretreatment with bromopalmitate resulted in alterations in the secretory phenotype that are similar to those due to chronic exposure to palmitate (24): enhancement of basal secretion and upregulation of the responses to either glucose, KCl, or the phorbol ester tetradecanoylphorbolacetate (TPA). However, forskolin-stimulated insulin secretion was virtually unaffected by pretreatment with bromopalmitate (Fig. 1) as opposed to the upregulation previously observed in cells chronically exposed to palmitate (24). Another dissociation was seen with KCl-stimulated secretion that was unaffected by prior treatment with oleate but enhanced after either palmitate (24) or bromopalmitate (Fig. 1). These results suggest that distal upregulation is not explained by a single common mechanism, but rather separate components interacting specifically with each of the cAMP, Ca²⁺, or diacylglycerol (DAG)/PKC signaling pathways. Although not previously analyzed in such detail, these findings are consistent with earlier work with fat-fed (or diabetic animals) that showed upregulated secretory responses to nonnutrients (such as the neurotransmitter acetylcholine) and that were attributed, at least in part, to a sensitization of PKC signaling pathways (41,42).

The fourth secretory alteration to palmitate pretreatment was a pronounced enhancement of responsiveness to subsequent addition of palmitate. This effect had not been previously described and is extended here to cells chronically exposed to oleate (Fig. 2). In these experiments, MIN6 cells were pretreated for 48 h at 5.6 mmol/l glucose with either BSA control or 0.4 mmol/l oleate or palmitate. Each of these groups were then stimulated for 1 h at either 2.8 or 16.7 mmol/l glucose, with either no further addition or in the presence of oleate or palmitate. Consistent with several earlier studies (43–45), palmitate is more effective than oleate in potentiating GSIS in control-pretreated cells, whereas both FAs exert only minimal effects at 2.8 mmol/l glucose. Oleate-pretreated cells showed an enhanced basal secretion and inhibition of GSIS as highlighted previously (24). The additional data (Fig. 2) now reveal that oleate pretreatment enables palmitate to act as a potent acute stimulus, even at 2.8 mmol/l glucose (equivalent in size to GSIS in control cells). Furthermore, potentiation of GSIS seen acutely with oleate in control cells was not further enhanced by oleate pretreatment, in contrast to the already larger potentiation due to acute palmitate addition. Compared with control cells, palmitate pretreatment resulted in an approximately threefold enhancement of basal secretion at 2.8 mmol/l glucose, but most notably now allows a large response to the acute addition of palmitate or even oleate under these conditions. Correspondingly, the palmitate-pretreated cells show a modest enhancement of GSIS, but a much more pronounced increase in the capacity of both oleate and palmitate to potentiate this response acutely.

To interpret these results, it should be stressed that they have been expressed as percent of cell insulin content. When not normalized in this manner, it appears as if FA pretreatment potently inhibits GSIS and does not affect basal secretion (24). The reason for this discrepancy is the depletion of cellular insulin content that occurs in vitro using cells pretreated with oleate and (especially) palmitate. This depletion limits the response to a subsequent secretory challenge and, in our view, confounds the other effects of FA pretreatment described above. This view, and the validity of correcting for loss of insulin content, was confirmed in our earlier study in which we included diazoxide (which inhibits secretion and hence insulin depletion) during pretreatment. Under these conditions, net secretory responses of palmitate-pretreated cells were essentially similar to those already described: there was an enhanced basal response, a slight increase in GSIS, and a marked increase in the acute effects of palmitate whether at 2.8 or 16.7 mmol/l glucose (24). It is very important to realize that this depletion has also been described in islet studies (18,19), and had this been fully taken into account at the time, the subsequent emphasis on inhibition of GSIS, as opposed to other effects of FA pretreatment, might not have been so pronounced.

The minimal consensus view for the effects of FAs on β-cells is based on a time-dependent duality. At early times, an elevation in circulating FAs is universally regarded as beneficial in that they potentiate GSIS and probably thus play a role in β-cell compensation. Over longer periods, FAs are generally regarded as deleterious, although various possibilities have been preferred as to why this would be the case. One explanation centers on downregulated GSIS and another on β-cell exhaustion due to chronically elevated basal secretion and/or depletion of insulin reserves. Above and beyond these secretory defects are the cytotoxic effects. However, a potential drawback of this model is that it does not fully explain the genetic component that underlies β-cell decompensation: the described effects of FAs can be generated in vitro, even in normal β-cells, not just those genetically programmed to become diabetic. Of course, genetic susceptibility to diabetes might be manifest as enhanced delivery of FAs to the β-cell, increased FA uptake into it, or altered partitioning of FA through various metabolic pathways within it. Such mechanisms appear to underlie β-cell failure in the ZDF rat, but it is less likely (to us) that this could fully explain the human disease, which is both polygenic and presents as a graded continuum ranging from slightly impaired glucose tolerance to full-fledged diabetes.

Our analysis of the secretory changes (summarized in Table 1) suggests that a reappraisal is warranted. It is thus apparent that acute responses to all other secretory stimuli are upregulated by palmitate pretreatment to a much greater extent than is GSIS (Table 1). Secondly, the glucose response in cells chronically exposed to oleate is the only condition that is inhibited relative to control. This suggests that both FAs indeed negatively regulate some proximal aspect of the stimulus-secretion coupling that is triggered specifically by glucose. This is manifest as an inhibition with oleate, as opposed to a weak increase with palmitate, because the latter is able to partially reverse the proximal defect by upregulating more distal components of the secretory process. The distinction between oleate versus palmitate has been rarely addressed, but it is noteworthy that this aspect of our study has recently been confirmed in vivo (46). Rats maintained for several months on a diet rich in saturated fat (e.g., palmitate) showed a slight potentiation of GSIS, in contrast to those fed mainly unsaturated fat (e.g., oleate), in which there was an inhibition of the glucose response. These differences appeared to be directly attributable to effects at the level of the β-cell, because they were maintained in ex vivo pancreas perfusions (46). This very much underscores the physiological relevance of our admittedly simple experimental system. However, the important point is that this inhibition of glucose coupling is relatively weak. We believe that the relevance of this inhibition has been previously exaggerated. First, most studies have not taken into account the depletion of insulin content that occurs in vitro, but that does not explain the early and selective loss of glucose responsiveness that characterizes type 2 diabetes (1–3). Second, the earlier investigations were usually based on a rather artificial experimental design, in which FAs were present during the pretreatment period but then removed for the assessment of GSIS. However, in the continuous presence of FAs (the physiological situation), the loss in glucose responsiveness is compensated for by enhanced sensitivity to FAs; insulin release in the presence of high glucose plus FA is actually greater in pretreated cells (especially cells pretreated with palmitate) (Fig. 2). Although not fully appreciated at the time, these results are consistent with an earlier study using isolated islets (19). Third, the effects on GSIS of chronic lipid-elevating protocols in humans have been inconsistent and not nearly as pronounced as with the in vitro experiments (11). This might be due to differing effects of saturated versus unsaturated FAs as shown here, continuing presence (or not) of FAs during glucose infusions, and the probable nondepletion of islet insulin content under these conditions.

In contrast to the modest effects on GSIS, the sensitization of FA-stimulated secretion by prior FA exposure is pronounced, with the sensitized cells responding robustly to FAs even at basal glucose. From this perspective, the slight inhibition of GSIS caused by chronic exposure to FAs would appear less as an inhibitory response per se, but as a switch in β-cell responsiveness away from glucose and toward FAs (and potentially nonnutrients). It is our working hypothesis that broad alterations in gene expression underlie this switch and that, in vivo, it would represent a normal and beneficial adaptation to elevations in circulating FAs (i.e., a compensatory response to insulin resistance). Many genes might be crucial to this reprogramming, including transcription factors and genes encoding metabolic enzymes, signaling proteins, and components of the exocytotic machinery. Allelic variations in any of these genes might compromise β-cell compensation and to varying extents. For example, loss or gain of function of a transcription factor, regulating expression of multiple genes, would be expected to produce a more profound phenotypic alteration than allelic variation in the gene encoding a component in a signaling pathway for which other proteins (pathways) might compensate. Variability could be accommodated even at the gross phenotypic level. For example too strong a switch to FAs might provoke hypersecretion and hence β-cell exhaustion, whereas too weak a switch might expose and exacerbate a preexisting and separate weakness in GSIS.

Obviously, much further work is required to test this hypothesis. It would first need to be determined in humans whether FA-induced insulin secretion is indeed sensitized by chronic prior exposure to FAs. In favor of this hypothesis is the observation that secretory responses to FAs (at basal glucose) are diminished in diabetic subjects (47). Additional investigations would focus on identifying all of those genes whose altered expression underlie each of the secretory phenotypes described. Whereas transcript profiling using microarrays makes this feasible in principle, secondary strategies would also be required to condense the large number of possible contributing genes to a workable number of candidates that can be then assessed more directly. Our future approaches will entail further analyses of the different secretory phenotypes to determine treatment conditions that either reproduce or dissociate them (for example, with bromopalmitate as above). Transcript profiling of the most informative of these treatment groups should then identify those genes that always cosegregate with a given phenotypic alteration. Proof of principle that this approach will identify key genes is provided with short-chain acyl CoA dehydrogenase, a gene that was downregulated in FA-pretreated cells (24) and whose loss of function in human subjects was independently shown to cause fasting hyperinsulinemia (48).

At a more basic level, however, this characterization will probably also increase knowledge of the mechanisms of stimulus-secretion coupling in the β-cell. This is especially true of the genes contributing to the enhancement of basal secretion and/or upregulation of distal secretory pathways (which may be interrelated). Our understanding of these distal pathways is particularly poor. Interestingly, our transcript profiling experiments identified a number of genes encoding proteins with known signaling function but that had not been previously implicated in the control of insulin secretion. An example of this is the gene for phospholipase D (PLD)-1 encoding an enzyme that catalyzes the hydrolysis of phosphatidylcholine to form phosphatidic acid (49). This gene was upregulated by both palmitate and oleate (24). Expression of a second transcript, phosphatidic acid phosphohydrolase 2, which converts phosphatidic acid to DAG, was also increased by palmitate. Because DAG is a well-known activator of PKC, these results suggest that palmitate-pretreated cells would display an enhancement of the PKC signaling pathway through the two-step conversion of phosphatidylcholine to DAG. It is noteworthy that PLD1 has also been implicated recently in the control of exocytosis in chromaffin and mast cells (50,51) but has only been studied to a limited extent in pancreatic β-cells (52–54). However, in light of our transcript profiling results, we have reexamined this issue using a more sensitive assay for PLD activation (49,51). Our data indicate that a combined secretory stimulus of glucose plus the acetylcholine analog carbachol activates PLD in MIN6 cells (Fig. 3A). Moreover, consistent with the gene-chip data (24), this activation is more pronounced in FA-pretreated cells. Note that isotopic dilution between the two saturated FAs, palmitate and myristate, prevented use of this technique with palmitate pretreatment. However, both oleate and palmitate did increase expression of the PLD1 protein (Fig. 3B). These results, together with our recent findings that PLD activation appears necessary for at least some aspects of stimulated insulin secretion (T.J.B., W.E.H., unpublished data), suggest that increased expression of PLD1 (and perhaps also of PAP2) contributes to the upregulation of distal secretory processes in FA-pretreated cells.

Completion of the genome projects for humans, mice, and other species opens up new avenues for investigation of complex diseases such as type 2 diabetes. One obvious example is the ability to use microarrays containing virtually the full genome to map alterations in gene expression on a comprehensive scale (55). The power of such studies is that they serve not only to test preexisting hypotheses, but that they also highlight novel and often totally unexpected areas for future investigation. We have applied the approach of transcript profiling to a much simpler system than the whole animal because we were prepared to compromise immediate (patho)physiological relevance for ease of manipulating inputs and clarity of data interpretation. Moreover, as described above, we have identified a strategy for attributing precise phenotypic alterations to changes in expression of discrete genes. The strength of this approach rests on the capacity to accurately define these phenotypic alterations, as witnessed by our discovery that FAs sensitize the β-cell to a subsequent FA challenge. Although a very simple model, it is reassuring that fully 25% of the genes recently shown to be differentially expressed in β-cells of diabetic mice (55) were also detected as changed in FA-pretreated MIN6 cells (24; A.K.B., T.J.B., unpublished data). This perhaps reinforces the point that a complete understanding of type 2 diabetes will only be achieved by integrating results from human, animal, and cell-based investigations, rather than focusing on any one approach in isolation.

This work was funded by grants from the National Health and Medical Research Council of Australia and the Diabetes Australia Research Trust.

We thank Professor Don Chisholm and Dr. Ross Laybutt for critical comments on the manuscript.