Roles of ceramide and sphingolipids in pancreatic β-cell function and dysfunction

β-cell cytokine signaling

Early investigations of ceramide generation in β-cells focused in the area of Type 1 diabetes (T1D), prompted by studies of inflammatory signaling cascades in other cell systems. It was first shown that exogenous ceramide provided either directly or generated following addition of SMase was sufficient to reproduce some of the effects of IL-1β in repressing insulin production and promoting β-cell death in fetal islets and β-cell lines.^17-19 These treatments did not activate the NFκB pathway that mediates induction of iNOS and generation of nitric oxide.^19,20 Direct evidence for activation of SMase by cytokines, however, has been harder to demonstrate using β-cells than in other systems. IL-1β reportedly promoted a transient (5min) decrease in SM in clonal RINm5F cells,¹⁹ but this was not consistent with other studies using rat islets²⁰ or β-TC3 cells¹⁷ possibly explained by technical limitations of the tracer methodologies that had been employed in all three reports. Moreover, increases in ceramide mass under these conditions were only observed in RINm5F cells,¹⁹ but not β-TC3 cells.¹⁷ These modest or non-significant effects are perhaps consistent with observations that cytokines can activate both neutral and acidic ceramidases in β-cells,^21,22 which would tend to counteract any ceramide accumulation. Pharmacological inhibitors of ceramide synthesis or SMase activity also failed to alter IL-1β-stimulated responses in a rat β-cell line,²³ although the potentiation of IL-1β effects on cell death by TNFα might involve an enhanced sensitivity to basal acid SMase.²⁴ Finally, Il-1β and TNFα augment S1P in β-cells,²⁵ which might represent a protective response against cell death via S1P receptor mediated pathways.²⁴ The overall evidence would thus tend to argue against alterations in sphingolipid metabolism playing major roles in cytokine-mediated β-cell destruction in the context of T1D. However, this conclusion is perhaps not definitive and this topic might benefit from further investigation focusing on longer-term effects of cytokines, employing more sensitive assay methods, and extending measurements to metabolites in addition to ceramide and SM.

Immune cell infiltration in Type 1 diabetes

The in vitro studies discussed above focused on the downstream responses of β-cells to the extracellular milieu of cytotoxic cytokines encountered during autoimmune cellular destruction of the islets in T1D. These cytokines come from the many different immune cell types that migrate to the islet in response to islet self-antigens. Fewer studies have investigated the effect of altering sphingolipid metabolism from the reverse perspective; whether alterations in the sphingolipid content of β-cells, immune cells or serum might modulate immune cell infiltration, or otherwise act upstream of cytokine release. This seems feasible given known effects of sphingolipid modulators in other inflammatory diseases,²⁶ and some recent indications that downregulation of certain SM species in the circulation is associated with decreased progression to T1D in both humans and non-obese diabetic (NOD) mice.²⁷ However, SM was only one of several lipids that were altered and this observation was not pursued mechanistically. Other studies have focused on glycosphingolipids that can serve as ligands for cell surface receptors. One such is the macrophage scavenger receptor, CD14, which is primarily expressed on monocytes, macrophages and other immune cell types with lipopolysaccharide (LPS) as its natural ligand. CD14 is also expressed on a subset of β-cells from human islets.²⁸ In addition to LPS, GalCer, acting via the CD14 receptor and presumably in an auto/paracrine manner, was also found to also stimulate β-cells to secrete TNF-α, IL-1β and IL-8. Conversely sulfatide, another endogenous glycolipid found in insulin secretory granules, inhibited LPS effects mediated via the CD14 receptor.²⁸ However, the real physiological role of CD14 on the β-cell surface and whether it contributes to β-cell death seen in diabetes is as yet unclear.

There is some evidence that S1P may be critical in the protection of β-cell destruction by cytokines. Recent studies of islet allografts in diabetic mice demonstrated that the S1P receptor agonist, FTY720, enabled long-term survival of these grafts.^29,30 This S1P analog was found to prevent lymphoangiogenesis that was key to subsequent T cell infiltration, inflammation and β-cell destruction of these allografts.³⁰ Long-term FTY720 treatment also prevented the onset of diabetes in rodent models of T1D by halting islet immune cell infiltration and pro-inflammatory cytokine-mediated destruction.^31,32 Modulation of S1P signaling may therefore be an attractive candidate for future studies focused on the preservation of islet architecture for the prevention of T1D.

There is a longer-standing literature suggesting that various sphingolipids in β-cells act as auto-antigens. Dotta and colleagues identified GM2–1 as a pancreatic islet specific species of ganglioside³³ that was a target of IgG islet cell autoantibodies. The presence of these antibodies not only correlated strongly with progession to diabetes in relatives of subjects with T1D, but the content of this glycolipid was hyperexpressed in the NOD mouse.³³ Furthermore, this ganglioside is present in β-cell secretory granules, similar to other autoantigens in Type 1 diabetes such as insulin and carboxypeptidase H. More recently, cell surface patches of SM were shown to be the epitope for IC2 a monoclonal antibody that specifically recognizes the surface of β-cells.³⁴ Since this antibody was developed from the diabetic BB rat, which possesses circulating auto-antigens to β-cells, these findings at least raise the possibility that cell surface SM might be involved in β-cell autoimmunity.

Ceramide and β-cell apoptosis

Although ceramide had long been implicated in apoptosis, its involvement in lipotoxicity was first suggested by studies in hemopoietic cells showing that saturated but not unsaturated FAs induced cell death, which was associated with enhanced de novo synthesis of ceramide.³⁵ This concept was extended to β-cells in pioneering work from the Unger laboratory using the Zucker diabetic fatty (ZDF) rat model of T2D.^36,37 Key evidence included observations that ceramide content was augmented in islets of the diabetic animals and that the accompanying enhancement of apoptosis was reduced by fumonisin B1, an inhibitor of SPT.³⁷ Tracer analyses implicated an enhancement of de novo ceramide synthesis, which was consistent with observed increases in SPT expression in islets of ZDF rats, and islets of control rats treated with elevated FAs.³⁶ In vivo treatment of diabetic animals with an SPT inhibitor also reduced apoptosis of islets assayed ex vivo.³⁶ This work proved seminal for the future development of the lipotoxicity field, not only in β-cells, but also in the broader context of the whole metabolic syndrome. In retrospect, however, it could be argued that the ZDF model represented an extreme case in that these islets accumulate massive amounts of lipid. Indeed, the partitioning of palmitate tracer into ceramide in islets of ZDF rats greatly outweighs the amount that is oxidized.³⁸ More recent work using another animal model of the metabolic syndrome has shown only modest increases in ceramide late in the progression of β-cell failure,³⁹ and in fact β-cells appear relatively resistant to lipid overload as compared with other tissues in this model.⁴⁰

A milder and much more commonly employed model of lipotoxicity comprises the chronic exposure of β-cells to elevated concentrations of FAs in vitro. A caveat of this approach is that the actual concentration of free FA depends on the ratio of total FA to albumin in the medium, and so can vary markedly between protocols and in a manner that is not always easy to evaluate.⁴¹ The discussion below is focused on reports that are less likely to be confounded by the use of very high levels of free FAs. The general conclusion from these studies is that whereas both saturated and unsaturated FAs promote defects in glucose-stimulated insulin secretion,^42,43 β-cell apoptosis is much more restricted to saturated FAs, and indeed unsaturated FAs appear to be protective in this case.^44-46 Islets from ZDF rats appear to be an exception as these were sensitive to mixtures (2:1 unsaturated to saturated) of FA³⁶ but this might reflect the high overall concentrations employed or be a special feature of the ZDF model as discussed above. Although a variety of mechanisms have been proposed to account for β-cell lipo-apoptosis one of the best substantiated relates to ER stress.^47,48 According to this hypothesis, apoptosis ensues from a failure of the adaptive mechanisms serving to protect cells from ER stress due to an accumulation of misfolded proteins in the lumen of ER. Secretory cells such as β-cells are particularly susceptible to ER stress due to the high rate of protein traffic through the ER.^49,50 Evidence in support of this hypothesis includes the observation that ER stress is preferentially triggered by saturated FAs,^47,48,51 and that (at least mild) lipo-apotosis is reduced by inhibition of ER stress signaling.^47,48,52

A separate and longer standing body of work has pointed to involvement of ceramide synthesis in β-cell apoptosis due to saturated FAs, consistent with the substrate preference of SPT for palmitate.⁶ This has been supported by recent observations showing that ³H-serine incorporation into ceramide (also catalyzed by SPT) was augmented by pre-exposure to palmitate.⁵¹ Enzymatic measurements of ceramide mass, however, have generally reported only modest increases with FA pretreatment unless glucose is concomitantly elevated.^53-55 And yet inhibitors of de novo ceramide synthesis, acting at the level of either SPT (cycloserine or myriocin) or CerS (fumonisin B1), generally reduce lipo-apoptosis in β-cells^{44-46,54,56-58} although this is not universally observed, particularly with higher concentrations of FAs or when combined with elevated glucose.^59,60 A possible explanation for these somewhat disparate observations might be that apoptosis is triggered by increases in ceramide that are restricted to particular species or cellular locations, and which are less apparent when the total cellular content is assayed. Alternatively, ceramide generated via different pathways (de novo synthesis vs. salvage pathway or SMase activity) might differ in its apoptotic potential. Finally, the toxic intermediate might not be ceramide itself, but a metabolite of it.

To some extent these possibilities have been resolved by two recent studies using β-cell lines that combine genetic modulation of sphingolipid metabolism with accurate quantification of ceramide mass using mass spectroscopy.^56,58 Our work, aimed at identifying the lipid metabolite responsible for lipotoxic ER stress, comprised a comprehensive and unbiased lipidomic screen and largely excluded involvement of phospholipids or neutral lipids.⁵⁶ Following pretreatment of the MIN6 β-cells with palmitate for 48h we were unable to detect increases in major ceramide species at the whole cell level, but hexosylceramide (predominately GluCer) was augmented, as was flux through the de novo synthetic pathway as determined by incorporation of ³H-serine. However, overexpression of GluCer synthase to enhance conversion of ceramide to GluCer did not exacerbate lipotoxicity, but partially protected against apoptosis and induction of ER stress markers. This intervention also overcame the defects in ER-to-golgi protein trafficking due to palmitate pretreatment⁵⁶ that we had previously postulated contributes to lipotoxic ER stress through protein overload.⁶¹ These results suggested that chronic palmitate induces ER stress via increases in a subpool of ceramide, without altering total ceramide mass. Broadly consistent with our findings Veret et al. were able to show increases in ceramide using INS-1 cells treated with higher concentrations of palmitate, but that this effect was maximal at 12 h and was no longer apparent at 48 h, except in the presence of elevated glucose concentrations.⁵⁸ The increased ceramide was observed with most side-chain species, not just C16, and appeared to arise from the de novo synthetic pathway because there were accompanying increases in di-hydroceramide and sphinganine, but not sphingosine. In this model, however, there was minimal lipo-apoptosis unless glucose was also augmented and this correlated best with accumulation of C18:0, C22:0 and C24:1 ceramide species, also consistent with increased expression of CerS4 under these conditions. Finally, up- or downregulation of CerS4 expression enhanced or reduced glucolipotoxicity respectively.⁵⁸ The conclusion that C18:0 and C22:0 Cer are the toxic metabolites would be consistent with unpublished data that these species are also selectively increased in our model of lipotoxicity. One caveat, however, is that C18:0 and C22:0 ceramide are very minor metabolites in MIN6 cells⁵⁶ and mouse islets (EB, PJM, TJB unpublished). Indeed there may be pronounced species differences since C16:0 greatly abounds over C24:0 ceramide in mouse β-cells⁵⁶ (EB, PJM, TJB unpublished), but the opposite is true in rat INS-1 cells.^58,62 It will thus be important to investigate key aspects of sphingolipid metabolism in rat, and ultimately human islets, to confirm findings obtained in the cell lines, and to determine which of the rodent species is the better model for human islets.

Although de novo synthesis appears to be the major mechanism for lipotoxic ceramide generation there are other possibilities. Ramanadham and co-authors have demonstrated that strong ER stress, triggered either pharmacologically in vitro^62-64 or using a genetic model of pro-insulin misfolding⁶⁴ itself promotes ceramide accumulation secondary to the activation of neutral SMase 2. In this instance ceramide accumulation in the mitochondria would be one mechanism whereby ER stress might trigger apoptosis.⁶² Whether ceramide generated via de novo synthesis during lipotoxicity could also act in this manner (perhaps independently of ER stress) is yet to be determined. In contrast, using our milder lipotoxic model, ER stress was neither necessary nor sufficient for any of the observed alterations in lipid metabolites, including ceramide, SM and GluCer.⁵¹ Nevertheless it cannot be excluded that small, perhaps localized, increases in ceramide and decreases in SM might contribute to lipo-apoptosis, either up or downstream of ER stress. This will require detailed analyses of sphingolipid metabolites in different subcellular compartments, and testing the functional effects of manipulating those metabolites in selected compartments.

Mechanisms of ceramide-induced apoptosis in β-cells

Our studies based on the protection afforded by GluCer synthase overexpression clearly implicate ceramide as acting upstream of defective protein trafficking and the activation of lipotoxic ER stress.^56,61 This mechanism does not appear to apply in liver,^65,66 but ceramide can promote ER stress in some other cell types.^67,68 Ceramide has also been reported to disrupt ER-to-golgi protein trafficking, although this seemed to involve the salvage rather than de novo synthetic pathway.⁶⁹ In principle, ceramide might also promote ER stress independently of effects on trafficking, by depleting Ca²⁺ from the ER and thereby causing protein misfolding.⁷⁰ This has not been assessed in β-cells.

As intimated above, ceramide could promote apoptosis by mechanisms other than ER stress, especially during glucolipotoxicity or very strong lipotoxicity. One such possibility would be disruption of mitochondrial function as mentioned previously⁶² and which has been implicated as causative in some aspects of defective β-cell function using SMS1 knockout mice.⁷⁰ This is discussed further below. Although exogenous ceramide is sufficient to disrupt mitochondrial function⁷¹ there is currently no evidence linking ceramide accumulation in mitochondria to lipo-apoptosis in β-cells. Ceramide also has the ability to form large stable channels in membranes which directly affects membrane conductance and hence may have biological relevance at the mitochondria or plasma membrane, particularly in regard to cytochrome c release and apoptotic signaling.⁷ Another potential apoptotic mechanism would be alterations in plasma membrane DRMs. Indirect evidence for this comes from studies of β-cell apoptosis due to islet amyloid polypeptide (IAPP),⁷² accumulation of which is linked to the pathogenesis of T2D. In this instance IAPP appeared to augment ceramide, and apoptosis was not observed in islets of mice deficient in acid SMase.⁷³ The latter enzyme was also required for IAPP-mediated inhibition of a K⁺ current that had previously been implicated in apoptosis and whose function might depend on clustering in DRMs.⁷³ If confirmed such a mechanism would potentially apply to many signaling pathways in addition to the gating of K⁺ channels. A further means whereby ceramide might mediate β-cell lipotoxicity is via inhibition of the anti-apoptotic kinase, Akt, as best described in muscle.^74,75 In β-cells, however, Akt is inhibited by both oleate and palmitate,^76,77 which might argue against involvement of ceramide because of the substrate specificity of SPT. Moreover, several labs have shown that chronic exposure to moderately elevated palmitate activates rather than inhibits Akt^55,78,79 and high FA concentrations stimulate Akt phophorylation at early time points, and only inhibit at 24h or more.⁷⁶ In muscle activation of the protein phosphatase PP2A is implicated in mediating the palmitate/ceramide effects on Akt.⁸⁰ It is therefore conceivable that PP2A might contribute to β-cell lipotoxicity via substrates other than Akt,⁸¹ although this remains to be investigated.

Ceramide and insulin gene expression

Co-treatment with palmitate inhibits the augmentation of pro-insulin mRNA levels due to chronically elevated glucose.^80,81 The fact that this did not seem to occur with oleate⁸² and that ceramide was sufficient to repress insulin content¹⁸ potentially implicated de novo ceramide synthesis in this context. This was directly confirmed by Kelpe et al., who demonstrated that inhibitors of the de novo synthetic pathway overcame both the increases in ceramide due to glucolipotoxicity, as well as the accompanying reduction in pro-insulin gene expression.⁵⁵ Mechanistically there appears to be a key role for the transcription factor C/EBPβ, whose expresssion is enhanced by palmitate^82-84 but not oleate⁸³ and which competes with Pdx1 and MafA on the proinsulin gene promoter.^85,86 Administration of exogenous ceramide or blockade of further ceramide metabolism in the presence of palmitate is sufficient to activate ERK⁸⁷ or modulate proinsulin and C/EBPβ gene expression respectively.^57,84 Whether ERK activation occurs upstream⁸² or independently⁸⁷ of C/EBPβ, remains to be resolved. To date, however, there is no direct evidence that ceramide accumulation is actually necessary for the enhancement of C/EBPβ function due to palmitate, although this would seem to be likely and could easily be tested by genetic or pharmacological intervention. With the use of lipidomic tools such as mass spectrometry, the contribution of endogenous ceramides to transcriptional regulation will be easier to elucidate, obviating the reliance on short-chain length exogenous ceramides (such as C2-ceramide), whose physiological relevance within the cell has often been questioned.

Ceramide and insulin secretory defects

In contrast to the strong evidence favoring ceramide as a mediator for lipo-apoptosis and reduced insulin gene expression, a corresponding role in defective insulin secretion appears less probable. Ceramide analogs have only marginal effects on glucose-stimulated insulin secretion,¹⁷ and inhibitors of de novo ceramide synthesis do not overcome the secretory defects due to lipotoxicity.^60,79 This is probably not surprising since unsaturated FAs (as well as saturated FAs) inhibit glucose-stimulated insulin secretion⁴³ and these are not substrates for SPT. It is possible that chronic exposure to oleate might augment ceramide through other mechanisms, perhaps via side-chain incorporation. This seems unlikely, however, since we observed no evidence for accumulation of sphingolipid metabolites in a palmitate-resistant subline of MIN6 cells,⁵⁶ which display an augmented ratio of unsaturated to saturated FAs due to enhanced expression of stearoyl CoA desaturase.⁸⁸ Even so, it is conceivable that localized, or more subtle, alterations in sphingolipid metabolism might still impact on stimulus-secretion coupling. Indeed a whole body knockout of SMS1 resulted in the expected decreases in SM and increases in ceramide in β-cells and was accompanied by defects in insulin secretion and mitochondrial function and increased ROS production.⁷⁰ However, because these mice also display lipodystrophy it is possible that there is disruptive lipid accumulation in β-cells secondary to increases in circulating FAs. Confirming whether this interesting secretory phenotype is truly intrinsic to the β-cell would require further investigation using tissue-specific knockout mice. However, knockdown or pharmacological inhibition both SMS1 and SMS2 in INS-1 has also been recently shown to inhibit insulin secretion in vitro.⁸⁹ The mechanism in this instance appears to involve defective export of secretory proteins from the golgi to the plasma membrane. Whether this is explained by increased ceramide, or decreases in either SM or DAG, which are both products of SMS activity, was not addressed. Although taken together these studies suggest that SMS function is essential for the maintenance insulin secretion, it remains to be determined whether loss of SMS activity contributes to the secretory defects observed in the setting of diabetes.

The sequestration of SM and/or ceramide within DRMs post-stimulus, such as in the case of IAPP accumulation described above,⁷³ promotes clustering of receptors and other plasma membrane proteins into these microdomains to augment downstream signaling. Recent work characterizing the expression, gating and association of potassium and calcium channels at the plasma membrane of β-cells has revealed a regulated compartmentalization of these ion channels into DRMs during the time-frame of insulin release.⁹⁰ Voltage-gated Ca²⁺ channels (Ca_v1.2) and K⁺ channels (K_v2.1), two critical determinants of β-cell electrical excitability, are directly associated within these DRMs with exocytotic proteins such syntaxin 1A, SNAP-25 and VAMP2. Dissolution or disordering of those lipid microdomains with β-cyclodextrin (which removes cholesterol from the membrane), inhibited Kv2.1 activity, but not Cav1.2 channel activity and this actually enhanced single-cell exocytotic events and insulin secretion.⁹⁰ It is therefore possible that alterations in sphingolipid metabolism that impact on DRMs could influence insulin secretion, but that this has yet to be tested directly.

Although it is unclear whether ceramide itself mediates defective insulin secretion, the minor yet highly physiologically active ceramide metabolite, S1P, has recently been shown to accumulate in response to glucose in both MIN6 and pancreatic islets.⁹¹ Furthermore, the activity of SK2 was responsible for this S1P accrual and its inhibition reduced GSIS in vitro and impaired insulin availability and glucose tolerance in vivo. The downstream signaling of S1P from its five G-protein coupled receptors may therefore play an essential and as yet to be defined role in GSIS.

Sulfatide

A glycosphingolipid derivative of GalCer, sulfatide, appears to be particularly important to secretory cells such as the pancreatic β-cell and neuronal cells. This lipid, GalCer-3-O-sulfate, is sulfated by sulfate transferase and present in β-cells but not exocrine tissue of the islets in humans and other species including rat, mouse, pig and monkey.⁹⁸ Sulfatide is synthesized in the golgi and is packaged into insulin secretory granules with insulin in the trans-golgi network.^99,100 It binds to insulin crystals to preserve the crystal structure at pH 5.5 as well as aiding the conversion of insulin hexamers to monomers at pH 7.4 at the cell surface.¹⁰¹ It also promotes proinsulin folding and oxidation within the secretory pathway.¹⁰¹ Patch clamp studies show that sulfatide negatively regulates glucose stimulated insulin secretion potentially via its action on K⁺_ATP channels.^101,102 Furthermore, loss of the C16:0 isoform of sulfatide in β-cells has been implicated in the pathogenesis of T2D. This specific isoform is lacking in islets from ob/ob and db/db mouse models of T2D as compared with normal human pancreatic tissue, BALB/c mice and the non-diabetic Lewis rat.⁹⁹ Moreover, C16:0 sulfatide significantly improves insulin crystal preservation.⁹⁹ As discussed earlier, the co-secretion of sulfatide with insulin also appears to negatively regulate CD14 signaling to prevent excessive secretion of pro-inflammatory cytokines from the β-cell that may precipitate β-cell destruction.²⁸ Collectively these studies make a strong case for sulfatide playing an important role in β-cell biology, and this is another topic for reinvestigation using newer genetic and analytical tools.