Obesity, insulin resistance, type 2 diabetes mellitus and cardiovascular disease form a metabolic disease continuum that has seen a dramatic increase in prevalence in developed and developing countries over the past two decades. Dyslipidaemia resulting from hypercaloric diets is a major contributor to the pathogenesis of metabolic disease, and lipid-lowering therapies are the main therapeutic option for this group of disorders. However, the fact that dysfunctional lipid metabolism extends far beyond cholesterol and triglycerides is becoming increasingly clear. Lipidomic studies and mouse models are helping to explain the complex interactions between diet, lipid metabolism and metabolic disease. These studies are not only improving our understanding of this complex biology, but are also identifying potential therapeutic avenues to combat this growing epidemic. This Review examines what is currently known about phospholipid and sphingolipid metabolism in the setting of obesity and how metabolic pathways are being modulated for therapeutic effect.
Hypercaloric diets lead to the dysregulation of multiple lipid metabolic pathways, which contributes to the onset and progression of metabolic disease
Lipidomic studies are starting to decipher the dysregulation of lipid metabolism associated with metabolic disease
Lipid metabolic pathways represent potential therapeutic targets to prevent or delay the onset and progression of metabolic disease
Further animal studies and clinical trials are required to define the stage of disease at which modulation of lipid metabolism will have maximal efficacy
Whether intervention into a single metabolic pathway or multiple pathways will produce optimal results remains to be determined
Obesity and its metabolic sequelae (which are largely the result of dyslipidaemia associated with obesity) have reached epidemic proportions and the prevalence continues to grow. Globally, the number of people with diabetes mellitus has more than doubled over the past three decades1. In 2013, an estimated 382 million people worldwide had diabetes mellitus, with almost half of these (175 million) thought to be undiagnosed2. A further 316 million people have impaired glucose tolerance, which is a major risk factor for type 2 diabetes mellitus and is remarkably underdiagnosed. Globally, the number of people with diabetes mellitus is projected to rise to 592 million by 2035 (Ref. 2).
Although this dramatic increase in obesity and type 2 diabetes mellitus might be expected to translate into a commensurate elevation in cardiovascular-related morbidity and mortality, the incidence of these diseases has partially been offset by the improved management of dyslipidaemia through statin-based therapies. Statins inhibit 3-hydroxy-3-methylglutaryl–coenzyme A (CoA) reductase, a key enzyme involved in the biosynthesis of cholesterol and other sterols, and thus reduce serum levels of LDL cholesterol and resultant coronary artery disease3. Indeed, the introduction of statin-based therapies has been one of the single most important developments in the management of metabolic and cardiovascular disease: statins are the first and most effective of the lipid-modulating therapies developed so far4.
Compared with statins, drugs targeting triglycerides have generally shown low clinical efficacy4. Fibrate-based and niacin-based therapies that lower serum triglyceride levels and raise serum HDL cholesterol levels have been used to supplement statins, particularly in patients with other features of the metabolic syndrome4. Although these types of therapies can successfully improve lipid profiles in individuals with abnormal profiles (such as patients with obesity, metabolic disease or diabetes mellitus), their effect on disease outcomes is modest, particularly when used in addition to statin therapies4.
The effect of statin therapy itself, although dramatic, is variable and tightly coupled to the level of LDL-cholesterol-lowering achieved. This effect was demonstrated by the JUPITER trial, which determined that statins reduced the relative risk of first-ever cardiovascular events by 9% in patients achieving no reduction in LDL cholesterol levels (who represented 10.8% of individuals treated with rosuvastatin), by 39% in patients achieving <50% reduction in LDL cholesterol levels (42.8% of individuals treated with rosuvastatin) and by 57% in patients achieving >50% reduction in LDL cholesterol levels (46.3% of individuals treated with rosuvastatin)3. Statins have been largely ineffective in modulating risk of type 2 diabetes mellitus onset, with some research suggesting that they might even increase this risk5,6.
The dyslipidaemia associated with obesity that drives adverse clinical outcomes is complex, and involves a large number of lipid species; the homeostasis of these lipids is controlled by diverse and intersecting metabolic pathways. Many sphingolipid and phospholipid metabolites have been implicated as critical components linking obesity to insulin resistance, type 2 diabetes mellitus and cardiovascular disease (Table 1). The new field of lipidomics has helped to define these relationships and has uncovered new avenues of investigation and therapeutic opportunities. In this Review, we highlight these pathways and identify several lipid-synthesizing and lipid-degrading enzymes that could emerge as new therapeutic targets.
Lipid metabolism and metabolic disease
One of the primary drivers of metabolic disease is an overabundance of calories in the diet, either as fats or carbohydrates. Within the intestinal epithelium, dietary fats are converted into triglycerides that are then transferred into the circulation within chylomicron particles. The majority of the triglycerides in these particles are subsequently hydrolysed by tissue-bound lipoprotein lipase to release free fatty acids that can be absorbed by peripheral tissues. In adipose tissue, fatty acids are converted back into triglycerides for sequestration within large lipid droplets, where they provide the vast majority of the body's energy reserves. In the liver, these fatty acids are also repackaged into triglycerides, which are delivered as VLDL to other tissues.
Saturated fats potentiate glucose-stimulated release of insulin7,8, the key anabolic hormone that promotes glucose uptake and storage in skeletal muscle and adipose tissue. Insulin also stimulates de novo lipogenesis in liver and fat tissue, and suppresses the rate of adipose lipolysis. The increased delivery of fats to peripheral tissues impairs some of these insulin responses. When in a state of insulin resistance, the body loses its ability to maximally stimulate muscle glucose uptake and to inhibit adipose lipolysis. However, under these conditions, the stimulatory effect of insulin on hepatic lipogenesis is retained9. The consequence of this 'selective' insulin resistance and associated hyperinsulinaemia is a sustained increase in fatty-acid production, which leads to enhanced delivery of fats to peripheral tissues. Lipotoxicity then ensues as increased circulating levels of lipids impair the function of the liver, skeletal muscle, pancreatic islets, vascular endothelium, heart and other tissues.
Within the liver, these fatty acids can have three fates. Firstly, they can be oxidized within the mitochondria to produce acetyl-CoA, which is then used to generate ATP or sterols. Secondly, they can be converted to triacylglycerols and then either exported as VLDL particles or stored as lipid droplets. Thirdly, they can be metabolized into a range of other phospholipid and sphingolipid species. Triglycerides, the main lipid stored in lipid droplets, are themselves not thought to be harmful. Other lipid classes or the dysregulation of other lipid metabolic pathways are probably the primary drivers of lipotoxicity. In particular, sphingolipid and phospholipid metabolism are affected in obesity and, as such, studying these pathways might reveal viable therapeutic targets to combat the growing epidemic of obesity and metabolic disease.
Remarkable advances in lipidomics have greatly expanded our understanding of the extent and complexity of lipid dysregulation in obesity and metabolic disease. This technology enables the measurement of several hundred individual lipid species in hundreds to thousands of samples (current capabilities of instrumentation, analytical strategies and statistical approaches in lipidomics have been reviewed elsewhere10,11,12,13). Although early studies were limited in their size and coverage of the lipidome, often focusing on cell or animal models, contemporary population-based studies of serum or plasma samples have revealed potential roles for a wide variety of different lipid species in the pathogenesis of metabolic disease.
One early plasma lipidomic study of obesity analysed samples from 14 pairs of monozygotic twins discordant for obesity14. The researchers found that obesity, independent of genetic influences, was primarily related to increases in levels of lysophosphatidylcholines and decreases in levels of ether phospholipids, including plasmalogens, compared with non-obese co-twins14. Subsequently, a large population-based study of 1,076 individuals of Mexican–American descent revealed that levels of 200 of the 312 plasma lipid species measured (from across 23 lipid classes and subclasses) were associated with BMI after adjustment for age, sex, systolic blood pressure, 2-h postload glucose plasma levels and smoking status15. In contrast to the prior report, the researchers observed a strong negative association between levels of lysophosphatidylcholine species and many ether phospholipid species and BMI. Negative associations were also observed with lysophosphatidylethanolamine and several classes of glycosphingolipids. As expected, triacylglycerol, diacylglycerol and cholesteryl esters were positively associated with obesity, as were phosphatidylethanolamine, phosphatidylglycerol, phosphatidylinositol and some sphingolipid classes (such as dihydroceramide and sphingomyelin). Ceramide itself, the central player in sphingolipid metabolism, showed no association with obesity15.
Further analysis of the same large cohort, using prediabetes and type 2 diabetes mellitus as outcomes, combined with cross validation in a subset of the Australian Diabetes, Obesity and Lifestyle (AusDiab) study, consisting of individuals with normoglycaemia (n = 170), prediabetes (n = 64) and type 2 diabetics mellitus (n = 117), identified that many of the same associations observed in obesity were also evident in patients with prediabetes or type 2 diabetes mellitus16. However, in this late stage of metabolic disease, the associations of prediabetes and type 2 diabetes mellitus with levels of ceramides were apparent, and the associations with the glycosphingolipids and sphingomyelin were weaker than with BMI, with some species of sphingomyelin showing negative associations. This result suggests a shift in the balance of sphingolipid metabolism as disease progresses, which leads to greater ceramide production and/or accumulation.
These plasma lipidomic profiling studies have limitations, as they measure plasma levels of lipids rather than lipids in the relevant target tissues. In plasma, lipids exist predominantly within lipoprotein particles and so although these measurements provide insight into lipid metabolism (predominantly within the liver) they will also reflect lipoprotein levels and metabolism (including production, circulatory metabolism and turnover), which makes interpretation of plasma lipidomic profiles challenging. Adjustment for anthropometric covariates known to influence lipoprotein levels (such as age, sex or obesity) and use of one or more clinical lipid measurements (such as total cholesterol, HDL cholesterol and triglycerides) can address this issue to some extent, but requires a careful interpretation of the resulting associations17. Further to these limitations, plasma lipidomic profiles do not reflect the temporal changes that occur as a result of altered feeding status or circadian rhythms, although standardized collection of plasma early in the day, after fasting, can minimize these effects. Notwithstanding these limitations, plasma lipidomic profiles provide an important framework around which we can build on our understanding of the spectrum of lipid changes that drive the pathogenesis of metabolic disease. Lipidomic analysis of isolated lipoprotein fractions can also be performed, and such studies are now starting to define how metabolic disease influences lipoprotein composition18,19 and potentially lipoprotein function20,21,22.
In any given cell, the majority of incoming or newly synthesized fatty acids are coupled onto a glycerol backbone23,24 to generate diacylglycerol, the foundation of all glycerophospholipids. Addition of another acyl side-chain by diglyceride acyltransferases produces triglycerides, the primary energy depot in mammals. Intermediates in this biosynthetic pathway, such as diacylglycerol, have been implicated in insulin resistance25, but the role of these molecules in this process is controversial26. Here, we will focus on the action of complex glycerophospholipid species that are synthesized downstream of diacylglycerol, which are probable contributors to metabolic diseases.
Phosphatidylcholine and phosphatidylethanolamine. The most abundant phospholipids in mammals are phosphatidylcholine and phosphatidylethanolamine which provide the majority of membrane lipids within cells. In most cell types, synthesis of phosphatidylcholine occurs via the cytidine 5′-diphosphate (CDP)-choline pathway, which adds CDP-choline to the diacylglycerol scaffold. However, in hepatocytes, up to 30% of phosphatidylcholine is supplied from the conversion of phosphatidylethanolamine to phosphatidylcholine by the sequential methylation of the choline head group, a reaction that is catalysed by phosphatidylethanolamine N-methyltransferase (PEMT). This conversion occurs at the expense of three S-adenosylmethionine molecules, which serve as methyl-donors27,28 (Fig. 1).
Phosphatidylethanolamine is synthesized by an analogous (CDP-ethanolamine) pathway, which couples diacylglycerol to CDP-activated ethanolamine. Phosphatidylethanolamine can also be synthesized by the decarboxylation of phosphatidylserine by phosphatidylserine decarboxylase, an enzyme that is located primarily within the mitochondria and is thought to regulate phosphatidylethanolamine content within this organelle and to modulate mitochondrial function29. In mice, muscle-specific knock out of ethanolamine-phosphate cytidylyltransferase, the enzyme involved in phosphatidylethanolamine production, leads to the accumulation of diacylglycerol and triacylglycerol within the tissue. However, these mice retained insulin sensitivity and showed marked increases in mitochondrial biogenesis and muscle oxidative capacity compared with wild-type mice30. These data suggest that phospholipids, rather than diacylglycerol or triacylglycerol, are the probable modulators of muscle insulin resistance.
Maintenance of the balance of the phosphatidylcholine:phosphatidylethanolamine ratio seems to have important health implications; obesity and the concomitant oversupply of fatty acids skews this balance. Within the liver, the phosphatidylcholine: phosphatidylethanolamine ratio reflects PEMT activity and is associated with steatosis in humans31 and in obesity or gene deletion mouse models32,33,34. Although a deficiency of PEMT is not sufficient to induce steatosis in mice (possibly owing to a 60% increase in choline-phosphate cytidylyltransferase in PEMT knockout mice compared with wild-type mice, leading to sufficient phosphatidylcholine production through the CDP-choline pathway), Pemt knockout mice fed a choline-deficient diet rapidly developed severe liver failure33. By contrast, mice deficient in both PEMT and phosphatidylcholine translocator ABCB4 (also known as multiple drug resistance 2 protein), which is a mediator of phosphatidylcholine conversion into bile acid, were resistant to liver failure and steatohepatitis33. This resistance was associated with a marked difference between the drop of the phosphatidylcholine:phosphatidylethanolamine ratio in Pemt knockout mice (57%) compared with Abcb4–Pemt double knockout mice (17%).
Plasma lipidomic studies in humans have also shown a clear association of phosphatidylethanolamine (and consequently, a decreased phosphatidylcholine: phosphatidylethanolamine ratio) with obesity15, prediabetes16 and type 2 diabetes mellitus16. Similarly, low hepatic and erythrocyte phosphatidylcholine:phosphatidylethanolamine ratios have been associated with patients with nonalcoholic fatty liver disease (NAFLD), compared with healthy individuals31. Moreover, genetic studies have shown an association between a loss-of-function polymorphism in the PEMT gene (Val175Met) in individuals and susceptibility to NAFLD and nonalcoholic steatohepatitis35,36.
The mechanism or mechanisms linking phosphatidylcholine, phosphatidylethanolamine and their ratio with steatosis and subsequent pathologies are unclear. Phosphatidylcholine is required for VLDL assembly and secretion, and dysfunction of VLDL secretion is seen in mice fed a diet deficient in choline and methionine37, as well as in mice lacking choline-phosphate cytidylyltransferase A38. Phosphatidylcholine regulation of hepatosteatosis might also occur through its feedback control on sterol regulatory element-binding protein 1 (SREBP1), the key regulator of de novo lipogenesis. Low cellular levels of phosphatidylcholine were found to activate SREBP1 and promoted its localization to the nucleus to switch on lipogenic gene expression, thus leading to lipid droplet formation within the liver39.
Although both phosphatidylcholine and phosphatidylethanolamine are major constituents of the plasma membrane, they are asymmetrically distributed in outer and inner leaflets respectively, such that phosphatidylcholine has a higher level in the outer compared with the inner leaflet, whereas phosphatidylethanolamine has a higher level in the inner compared with the outer leaflet. A low phosphatidylcholine:phosphatidylethanolamine ratio perturbs membrane integrity by presenting more phosphatidylethanolamine on the membrane outer surface. This rearrangement can fundamentally change the membrane potential and permeability to proteins such as cytokines. The correction of the phosphatidylcholine:phosphatidylethanolamine ratio by ethanolamine-phosphate cytidylyltransferase knock down in hepatocytes of Pemt knockout mice resulted in a decreased exposure of phosphatidylethanolamine on the cell surface and an increase in membrane integrity compared with Pemt knockout hepatocytes transfected with a control knockdown vector33.
In contrast to the decreased phosphatidylcholine: phosphatidylethanolamine ratio described earlier in the article, lipidomic analysis of the endoplasmic reticulum of obese mice showed an increase in the phosphatidylcholine:phosphatidylethanolamine ratio associated with endoplasmic reticulum stress and steatosis, through the effect of the increased phosphatidylcholine:phosphatidylethanolamine ratio on calcium ion reuptake into this organelle. Correction of the phosphatidylcholine:phosphatidylethanolamine ratio reduced endoplasmic reticulum stress and improved glucose homeostasis32. Thus, imbalances in the phosphatidylcholine:phosphatidylethanolamine ratio in either direction can lead to NAFLD, albeit through different mechanisms40.
Although these studies have examined the effects of broad changes in phosphatidylcholine and phosphatidylethanolamine classes, little is known about how the fatty-acid composition of these lipid species alters their distribution, the organism's susceptibility to obesity or resulting metabolic comorbidities.
Plasmalogen. Plasmalogens are a naturally occurring ether phospholipid species that are primarily present as phosphatidylcholine and phosphatidylethanolamine species. They contain a vinyl-ether-linked fatty alcohol at the sn-1 position of the glycerol backbone and an acyl-linked fatty acid in the sn-2 position. Plasmalogens are often esterified with polyunsaturated fatty acids such as arachidonic acid (C20:4) and the n-3 fatty acid docosahexaenoic acid (C22:6, a major constituent of fish oil), whereas the vinyl-ether-linked residue is usually saturated or monounsaturated (Fig. 2). Up to 50% of phospholipids in membranes are plasmalogens and levels are particularly high in cardiac and neural tissues41.
Plasmalogen biosynthesis is a complex process involving multiple enzymes within the peroxisome and endoplasmic reticulum42 (Fig. 3). The rate-limiting step in this pathway is the formation of the long-chain fatty alcohol by fatty acyl-CoA reductase 1 and 2 (FAR1 and FAR2). A second control of plasmalogen levels occurs during turnover. Plasmalogens can be deacylated by the action of the 85/88 kDa calcium independent phospholipase A2 enzymes (encoded by PLA2G6) to produce lysoplasmalogen. The lysoplasmalogen can then be either reacylated by a CoA-independent transacylase or further degraded by lysoplasmalogenase (encoded by TMEM86B). Lysoplasmalogenase is differentially expressed in different tissues and shows an inverse association with the relative proportion of plasmalogens in specific tissues43. Lysoplasmalogenase seems to be the major mechanism by which plasmalogen levels are regulated.
Lipidomic studies reveal that plasmalogens are negatively associated with obesity15, prediabetes, type 2 diabetes mellitus16, cardiovascular disease44 and Alzheimer disease45,46. Each of these conditions shares the common trait of heightened oxidative stress. In the setting of metabolic disease, plasmalogens might modulate oxidative stress, inflammation, cholesterol efflux and cell signalling. The vinyl-ether linkage of the plasmalogens is particularly susceptible to oxidation by reactive oxygen species, and thus could have an important antioxidant function in membranes and lipoproteins41. Ethanolamine plasmalogens prevent the oxidation of cholesterol in phospholipid bilayers and also regulate cholesterol esterification and efflux47. Plasmalogen levels in human serum decrease with ageing, positively correlate with serum levels of HDL48 and negatively correlate with serum levels of atherogenic small, dense LDL49. Enrichment of lipoproteins with plasmalogens increases their resistance to oxidation50. Plasmalogens are also central to the differentiation of monocytes to macrophages51. A mouse study has demonstrated enhanced cellular immunity in BALB/c mice and antitumour activity in BALB/c mice transplanted with spontaneous mouse mammary tumour, in response to administration of shark liver oil (a traditional source of alkylglycerols, which are precursors for plasmalogens)51. As a consequence of these multiple actions of plasmalogens, our knowledge of the sites and specific actions of plasmalogens in different forms of metabolic disease is limited.
Compared with glycerophospholipids, sphingolipids (such as ceramides, sphingomyelins and gangliosides) are very low in abundance, typically being present in the body at levels of <20% of their glycerolipid counterparts17,52. Complex sphingolipids derive from the addition of various head-groups to a ceramide or dihydroceramide backbone (Fig. 4). The initial reaction in the de novo synthesis pathway that produces ceramides involves the decarboxylation of a serine residue and condensation with a fatty acyl-CoA catalysed by serine palmitoyltransferase (SPT)53. Subsequent enzymatic reactions, catalysed by 3-ketodihydrosphingosine reductase, (dihydro) ceramide synthases (CerS) and dihydroceramide desaturases (sphingolipid δ(4)-desaturase DES1 and sphingolipid δ(4)-desaturase/C4-monooxygenase DES2, produce ceramides, the precursors for the majority of complex sphingolipids. Pharmacological or genetic inhibition of any of the enzymes required for de novo ceramide synthesis (namely SPT, CerS6 and DES1) improve postprandial glucose tolerance and reverse insulin resistance resulting from saturated fatty acids, high-fat feeding or deficiency of leptin or the leptin receptor in mouse models54. Moreover, in other mouse models, such interventions ameliorate cardiovascular complications such as atherosclerosis55,56 and lipotoxic cardiomyopathy57. These data suggest that one or more sphingolipids are toxic intermediates that contribute to obesity or dyslipidaemia-driven pathologies.
In addition to serine, alanine and glycine can also be condensed with fatty acyl-CoA by SPT to produce deoxysphingolipids. These lipids were initially reported to be elevated in hereditary sensory and autonomic neuropathy type 1, in which mutations in SPT alter the substrate selectivity leading to increased plasma levels of deoxysphingolipids, the causal factor in this rare genetic disease58. Subsequently, deoxysphingolipds were shown to be elevated in type 2 diabetes mellitus59,60 and are cytotoxic for insulin-producing cells61. These studies indicated that the intracellular metabolites of deoxysphinganine, in particular 1-deoxydihydroceramide, are important mediators of the cytotoxic effect. The researchers suggest that targeting deoxysphingolipid synthesis could complement the currently available therapies for type 2 diabetes mellitus. Indeed fenofibrate treatment of dyslipidaemia resulted in a reduction of deoxysphingolipids and consequently has been proposed as a novel therapeutic approach for diabetic neuropathy62.
Ceramide. Ceramide is the precursor for all the complex sphingolipids mentioned earlier in the article, and numerous studies suggest that it is also a major contributing factor of insulin resistance and cardiovascular disease63. Ceramides antagonize insulin signalling at the level of RACα serine/threonine-protein kinase (also known as Akt or protein kinase B)64,65,66 and their actions can be resolved from those of glucosylceramides67. Inhibition or ablation of the enzymes involved in ceramide biosynthesis is universally insulin sensitizing, antiatherogenic and cardioprotective54,56,57,63,68,69,70,71,72,73,74,75. Moreover, adiponectin was found to elicit its broad spectrum of antidiabetic and cardioprotective actions by activating a ceramidase to degrade ceramides76,77.
Profiling studies often support relationships between ceramides, obesity and/or insulin resistance in the circulation78,79,80,81 or skeletal muscle82,83,84,85,86. However, disagreement between some studies, particularly in skeletal muscle87, has caused controversy88,89. Studies examining the importance of saturated fatty acids in the ceramide side-chain might provide some resolution. Manipulation of the ceramide synthase isoforms 2, 5 and 6 revealed that ceramides comprised of long side-chains, such as C16:0 or C18:0, induce insulin resistance and/or hepatic steatosis, whereas those with very long side-chains (for instance, C24:0 or C24:1) do not90,91,92,93. Profiling studies also support these findings; in human liver, positive associations are emerging between ceramides containing saturated side-chains and insulin resistance that is independent of triglycerides94. In muscle, which predominantly contains the CerS1 isoform that preferentially utilises C18:0 fatty acid in ceramide synthesis, C18 ceramides associate most tightly, compared with other ceramides, with insulin resistance95.
Sphingomyelin. The sphingolipid that is most similar to phosphatidylcholine is sphingomyelin, which is produced by the transfer of a phosphocholine moiety from phosphatidylcholine to the ceramide backbone. Sphingomyelin is the most abundant of the sphingolipids and is the most prevalent class found in circulating LDL17. Sphingomyelins containing saturated, but not unsaturated, acyl-chains are associated with obesity, insulin resistance and decreased liver function in young adults (aged 18–27 years) with obesity96. Additionally, weight loss led to reductions in serum levels of HDL sphingomyelins in patients with overweight or obesity97. Further evidence for the involvement of sphingomyelins in metabolic disorders comes from interventions targeting the sphingomyelin synthase genes. Genetic ablation of the Sgms2 gene in mice reduces plasma-membrane levels of sphingomyelin and weight gain, and increases glucose tolerance and insulin sensitivity in animals fed a high-fat diet compared with wild-type controls98,99. In apolipoprotein E (Apoe) knockout mice, Sgms2 knockout ameliorated features of atherosclerosis compared with the Apoe knockout alone100.
Glucosylceramide and GM3 ganglioside. The addition of glucose to the ceramide scaffold gives rise to glucosylceramide, the precursor for complex gangliosides. Inhibitors of ceramide glucosyltransferase increase insulin sensitivity and improve glucose tolerance in leptin-deficient mice (referred to as ob/ob mice) and diet-induced obese mice, as well as in rats lacking the leptin receptor (Zucker diabetic fatty rats)101,102. When given to ob/ob mice, inhibitors of ceramide glucosyltransferase prevent hepatic steatosis and decrease expression of various lipogenic genes compared with untreated mice103. Moreover, these drugs prevent development of atherosclerosis in Apoe knockout and Apoe3-Leiden (a mutant form of Apoe3) mouse models104,105, although the finding is controversial as a separate study found no effect on atherosclerosis in Apoe knockout mice106. Further addition of hexoses, sialic acid or hexosamines to glucosylceramide produces a large family of gangliosides. In particular, the GM3 ganglioside (monosialodihexosylganglioside) is a precursor for many of the more complex ganglioside species. Moreover, the GM3 ganglioside impairs insulin signalling and blocks glucose uptake in cell culture studies67,107,108. Consistent with this finding, mice lacking lactosylceramide α-2,3-sialyltransferase (better known as GM3-synthase, the enzyme that catalyses the formation of GM3 ganglioside) display decreased fasting levels of glucose and improved glucose tolerance compared with wild-type mice109. Additionally, when challenged with high-fat diets, the GM3-synthase null mice maintained superior glucose tolerance, improved insulin-stimulated glucose uptake and enhanced suppression of hepatic glucose output compared with wild-type mice109.
Consequences of lipid dysregulation
What then are the metabolic, pathophysiological and clinical consequences of dysregulation of lipid metabolism in obesity? As discussed earlier in the article, the increased fatty-acid production resulting from either diet, lipolysis of adipose stores or lipogenesis (the latter two of which are exacerbated by insulin resistance) feeds into multiple lipid metabolic pathways that lead to increased production of glycerolipids (diacylglycerol and triacylglycerol) and of some sphingolipids (Fig. 5). Furthermore, elevation of diacylglycerol levels feeds into glycerophospholipid metabolism leading to a decreased phosphatidylcholine:phosphatidylethanolamine ratio, which at least partially results from dysfunctional methylation of phosphatidylethanolamine to phosphatidylcholine, owing to changes in the availability of the ubiquitous methyl donor S-adenosylmethionine (see also Fig. 1).
These changes in lipid metabolism and homeostasis in obesity can contribute to mitochondrial dysfunction, as has been reported for both ceramide91,110 (see discussion later in the article) and phosphatidylethanolamine30. Mitochondrial dysfunction can in turn increase the production of reactive oxygen species111. At the same time, the decrease in fatty-acid oxidation in obesity further exacerbates the accumulation of fatty acids, which leads to a vicious cycle of fatty-acid accumulation, dysregulation of lipid metabolism, mitochondrial dysfunction and oxidative stress. Similarly, the heightened oxidative stress in this context can result in the oxidation and increased turnover of multiple lipids, particularly plasmalogens (which are also important in mitochondrial function) that contain both a vinyl-ether linkage and, typically, polyunsaturated fatty acids that are susceptible to oxidation, thus creating a second vicious cycle. Sphingolipids (ceramide and sphingomyelin) also activate inflammatory pathways112,113. Additionally, the low phosphatidylcholine:phosphatidylethanolamine ratio increases membrane permeability, and the low cellular level of S-adenosylmethionine sensitizes the liver to lipopolysaccharide-induced expression and release of proinflammatory cytokines32,114. The inflammatory response can also lead to an increase in oxidative stress, and potentially feeds into mitochondrial dysfunction through damage to mitochondrial lipids, including plasmalogens. As mentioned earlier in the article, many lipids also impair insulin signalling: diacylglycerol and glucosylceramide reportedly inhibit insulin signalling at the level of the receptor, whereas ceramide impairs signalling at the level of Akt. These multiple overlapping pathways reinforce each other, leading to insulin resistance that further exacerbates the situation through feedback loops, resulting in a vicious cycle of inflammation, oxidative stress and insulin resistance. As this process extends to other tissues and organs, a progression towards type 2 diabetes mellitus and increased risk of cardiovascular disease is observed.
Several questions then arise: can we influence metabolic disease onset, progression and outcomes by intervention at a single point in this complex metabolic network? Can a single cycle or feedback loop be broken and consequently arrest the entire pathological process? If so, at what stage of the disease process can this intervention be achieved? Alternatively, will effective prevention or therapy require multiple interventions at different metabolic sites to arrest the cascade of deleterious events? The answers to these questions have been partially addressed by the multiple strategies and models being applied to modulate lipid metabolism in metabolic disease.
Therapeutic modulation of lipids
Modulation of phospholipid metabolism
Although animal models have demonstrated the therapeutic potential of altering phosphatidylcholine and/or phosphatidylethanolamine metabolism, which leads to an altered ratio of these two phospholipids (Table 1), how these observations might be translated into a therapeutic strategy is unclear. This ratio represents an obvious therapeutic target; however, the observations that both an increase and decrease in the phosphatidylcholine:phosphatidylethanolamine ratio leads to steatosis32,33,40 makes the application of such a therapy challenging. One potential approach is to utilise S-adenosylmethionine supplementation to increase PEMT activity and the conversion of phosphatidylethanolamine to phosphatidylcholine. S-adenosylmethionine production is controlled by S-adenosylmethionine synthase (also known as methionine adenosyltransferase; MAT) whereas its degradation is carried out by glycine N-methyltransferase115. Mat1A knockout mice show decreased phosphatidylcholine biosynthesis via PEMT, a decreased phosphatidylcholine:phosphatidylethanolamine ratio (leading to impaired VLDL secretion) and hepatic triglyceride accumulation116. Although S-adenosylmethionine supplementation can potentially increase VLDL and HDL production and secretion, thereby alleviating the accumulation of fat within the liver, this strategy might also increase endoplasmic reticulum stress if the balance of phosphatidylcholine:phosphatidylethanolamine swings too far toward phosphatidylcholine32.
MAT1A is underexpressed in patients with advanced NAFLD, but not in those with mild NAFLD117, and S-adenosylmethionine depletion leads to accumulation of lipid droplets in human skin fibroblasts39. This process seems to involve a feedback mechanism in which decreased phosphatidylcholine levels leads to maturation of nuclear, transcriptionally active SREBP-1 and thereby to increased cellular levels of S-adenosylmethionine39. Collectively, these human studies suggest that, at least in a subset of patients, reduced S-adenosylmethionine levels might contribute to disease pathogenesis; therefore, recovery of S-adenosylmethionine levels could have therapeutic benefit, particularly in limiting the progression from NAFLD to nonalcoholic steatohepatitis. However, few large definitive clinical trials of S-adenosylmethionine supplementation therapy have been performed and the potential therapeutic efficacy remains uncertain.
Modulation of plasmalogen metabolism
One can bypass the rate-limiting step in plasmalogen synthesis by orally administering naturally occurring alkylglycerols (1-O-alkylglycerol or 1-O-alkyl-2,3-diacylglycerol; Fig. 2). These molecules can be incorporated directly into the phospholipid pathway, thus bypassing the peroxisome118 (Fig. 3). This process leads to an increase in circulating and tissue levels of plasmalogens118,119.
Alkylglycerol supplementation has been used to increase plasmalogen content in cultured cells120, animals118 (Table 1) and humans119. Alkylglycerol supplementation in cell culture reduces the permeability120 and proliferation121 of porcine and calf aortic endothelial cells respectively. Early studies, in children (1–3 years of age) with the peroxisomal disorder rhizomelic chondrodysplasia punctata who have decreased levels of plasmalogens, demonstrated the efficacy of batyl alcohol supplementation in increasing red cell plasmalogen levels, and also showed subjective improvement of nutritional status, liver function, retinal pigmentation and motor tone. However, the treatment was largely ineffective in reducing the neurological pathology119. Mouse studies have demonstrated enhanced cellular immunity and antitumour activity in BALB/c mice transplanted with spontaneous mouse mammary tumour and administered with shark liver oil (which is highly enriched in alkylglycerols), compared with untreated mice51. Additionally, alkylglycerol supplementation in high-fat fed mice reduced the severity of diet-induced obesity and insulin resistance compared with untreated high-fat fed mice122. The study involved treatment of mice with three alkylglycerols (selachyl alcohol (O-18:1), batyl alcohol (O-18:0) and chimyl alcohol (O-16:0)) at either 20 mg/kg or 200 mg/kg body weight. The researchers reported decreased body weight, serum triglyceride, cholesterol and plasma levels of fasting glucose, insulin and leptin in the mice fed a high-fat diet supplemented with high-dose selachyl alcohol. By contrast, high-dose batyl alcohol increased fasting levels of insulin in the mice fed a high-fat diet compared with control mice122. These differential responses were also observed in cultured mouse adipocytes, in which selachyl alcohol decreased lipopolysaccharide-mediated phosphorylation of mitogen-activated protein kinase 8 (also known as JNK1) and mitogen-activated protein kinase 3 (also known as ERK) compared with untreated cells, whereas batyl alcohol had the opposite effect122. These divergent responses to alkylglycerols with different alkyl chain lengths suggest that the composition of plasmalogens within a cell or tissue, in addition to the total amount, could be critical for their capacity to attenuate metabolic disease.
One study examined the effect of batyl alcohol in Apoe knockout and Apoe–glutathione-peroxidase-1 (GPx1) double-knockout mouse models of atherosclerosis treated with a high-fat diet with or without supplementation with 2% batyl alcohol123. Lipidomic analysis showed a dramatic increase in plasmalogen levels in the circulation and in the heart in mice treated with batyl alcohol compared with untreated mice of the same genotype, with the increase restricted to those species containing an O-18:0 alkyl chain. Atherosclerotic plaque in the aorta was reduced in area by 70% in the treated mice compared with untreated mice, and oxidation of plasmalogen in the treated mice was evident from increased plasma levels of the plasmalogen oxidative by-product, sn-2 lysophospholipids. These data clearly support the antioxidant function of plasmalogens in metabolic disease.
Although alkylglycerols are readily available from shark liver oil, they have different O-alkyl modifications from their human counterparts. The major O-alkyl chain in shark liver oil is O-18:1, whereas in humans O-16:0 and O-18:0 are more prevalent124. Furthermore, if alkylglycerols are to become a dietary supplement that can attenuate the ever-increasing morbidity and mortality resulting from the obesity epidemic, alternate and sustainable sources will be required. One advancement has been the development of the drug PPI-1011, which consists of an O-16:0 alkylglycerol with a 22:6 acyl chain in the sn-2 position and a lipoic acid in the sn-3 position125. The lipoic acid is thought to stabilize the drug; the moiety is eventually removed by gut lipases, with the resulting 1-O-alkyl-2-acyl-glycerol metabolized to produce O-16:0-22:6 plasmalogens. Remodelling of the 1-O-alkyl-2-acyl-glycerol or plasmalogen can also occur to yield a plasmalogen containing alternate acyl chains in the sn-2 position126. Although this drug has not yet been evaluated in a setting of insulin resistance, animal studies suggest that it could have some efficacy in the treatment of dyskinesia in Parkinson disease by restoring deficient plasmalogen levels127,128.
Despite the ready availability of a number of forms of alkylglycerol and the development of the PPI-1011 drug, a paucity of clinical studies exists in this area. This lack of studies could relate to the largely ineffective trials aiming to modulate plasmalogen metabolism in childhood peroxisomal disorders that occurred in the 1970s, or to the challenges posed by pharmacokinetic studies in such a complex system. Nonetheless, with the developments in mass spectrometry and our understanding of plasmalogen metabolism, we are now in an ideal position to pursue the potential of plasmalogen modulation in metabolic disease.
Modulation of sphingolipid metabolism
Sphingolipid biosynthesis is driven by the availability of palmitate and serine, and the infusion of lipid cocktails enriched in saturated fats increases levels of various sphingolipids in mice and rats54. However, few studies have sought an understanding of the relationship between dietary macronutrient consumption and sphingolipid production in humans. One crossover trial compared the effects of diets enriched in palmitate (which drives ceramide biosynthesis) with diets enriched in oleate (which inhibits ceramide synthesis)129. In this study, women had decreased serum and muscle levels of ceramides, coupled with increased insulin sensitivity following the consumption of diets enriched in oleate, compared with the same women following diets enriched with palmitate. Similarly, an unbiased lipidomic screen was conducted in patients with the metabolic syndrome fed either a control diet consisting of low-fibre cereal products, dairy-fat-based spreads, regular-fat milk products and a limited amount of fruits, vegetables and berries or a 'healthy' Nordic diet consisting of whole grains, fruits, vegetables, berries, vegetable oils and margarines, fish, low-fat milk products and low-fat meat130.
The Nordic diet led to alterations in the plasma lipidome, including increased concentrations of plasmalogens and decreased levels of ceramides compared with the control diet. Various other insulin-sensitizing regimens have also been shown to decrease muscle levels of ceramides, including calorie restriction, exercise and bariatric surgery82,83,131.
Many of the enzymes involved in ceramide biosynthesis have emerged as potential therapeutic targets. The most commonly used reagent for these studies is the serine palmitoyltransferase inhibitor thermozymocidin (also known as myriocin), which prevents insulin resistance, diabetes mellitus, atherosclerosis, cardiomyopathy and hypertension in mice and rats53,132,133 (Table 1). Myriocin was originally isolated from an extract of the fruiting bodies of the fungus Isaria sinclairii and its parasitic host larva, and has since been identified as a common component in a number of closely related fungal species (including Cordyceps sinensis, Cordyceps cicadea, Cordyceps militaris and Cordyceps sinclairii)134,135. Although myriocin is not an approved drug for humans, extracts from I. sinclairii and Cordyceps spp. are commonly consumed as part of traditional Chinese medicines for treatment of a plethora of conditions, including diabetes mellitus, as they are believed by practitioners of traditional Chinese medicine to elicit an 'eternal youth' nostrum. These extracts are sold as nutritional supplements in numerous countries and seem to be efficacious against diabetes mellitus in mice and rats136,137. In humans, a few tests have also hinted at the efficacy of these fungal extracts for chronic kidney disease138. To our knowledge, none of these groups have evaluated the relationship between the myriocin content or sphingolipid-lowering ability of these supplements and their therapeutic efficacy.
The fourth enzyme in the biosynthetic pathway that produces ceramides is sphingolipid δ(4)-desaturase (also known as dihydroceramide desaturase), and pharmacological inhibition or genetic inactivation of the most abundant isoform (DES1) are also insulin sensitizing69,139. Several of the compounds with known inhibitory activity against DES1 have been administered to humans140. The most notable of these compounds is fenretinide, which is a potent DES1 inhibitor in mice and rats69 and has been tested in clinical trials for treating cancer and glucose intolerance139. Results from these trials have been mixed, with no significant effect seen in a large trial of 3,000 women with early breast cancer, but a possible benefit in premenopausal women141,142. However, in a phase III trial, fenetrinide did not reduce the reoccurrence of bladder cancer in patients143.
Another therapeutic means of targeting ceramides is by promoting their deacylation. Specifically, the broad spectrum of antidiabetic and cardioprotective actions of adiponectin might result from its ability to activate ceramidase77. Adiponectin receptors have some homology with ceramidase enzymes, and activating or depleting adiponectin receptors markedly alters cellular ceramidase activity77. Consequently, activating adiponectin could serve as another means of depleting tissue levels of ceramides.
Downstream of ceramide, ceramide glucosyltransferase inhibitors ameliorate obesity-induced insulin resistance in mice and rats144. Ceramide glucosyltransferase inhibitors can be given as an alternative to enzyme replacement for individuals with Gaucher disease type 1, a lysosomal storage disease that results from deficiency in glucocerebrosidase. This substrate reduction therapy limits the burden of disease resulting from the accumulation of nonmetabolizable cerebrosides in the lysosomes of these patients. One such medication, eliglustat tartrate, is a high affinity ceramide glucosyltransferase inhibitor that shows good oral bioavailability and efficacy145,146,147. Such compounds could have additional potential for the treatment of metabolic disorders, including type 2 diabetes mellitus and polycystic kidney disease146,147.
The studies highlighted in this Review reveal the probable involvement of a large number of distinct lipid species and metabolic pathways in the pathological sequelae of obesity. Researchers have disputed the relative importance of particular metabolites, but a large number of distinct species probably control tissue function by additive and synergistic mechanisms. Lipidomics reveals strong associations of certain species with various disease end points. Moreover, interventional studies in mice and rats show remarkable potential for drugs that alter lipid metabolic fate in the treatment of a whole cadre of conditions. Thus, the enzymes involved in glycerophospholipid and sphingolipid synthesis could have enormous untapped therapeutic potential.
A challenge for researchers attempting to uncover the molecular mechanisms linking distinct species to biological actions relates to the complexity of the lipidome and the interrelated nature of the lipid metabolic pathways. For example, diverting fatty acyl-CoAs from one pathway with a specific inhibitor invariably influences their rate of entry into another. As a result, studies using gene knock out and pharmacological reagents to alter cellular lipid profiles have struggled to produce a strong understanding of the mechanisms connecting lipids with biological consequences. However, we note that the same could also be said of statins and the sterols that they regulate, and this incomplete mechanistic understanding did not prevent their widespread use. We predict, therefore, that future therapies targeting the production of different lipid classes will result in the development of a large number of useful therapies. Many such therapies will probably be used in conjunction with existing lipid-modifying strategies (such as statins), and combinations of treatments to target multiple pathways will further enhance the efficacy of these therapeutic strategies.
P.J.M. is supported by a Senior Research Fellowship from the National Health and Medical Research Council of Australia.
About this article
iNKT cells prevent obesity-induced hepatic steatosis in mice in a C-C chemokine receptor 7-dependent manner
International Journal of Obesity (2018)