Diabetes mellitus has evolved as one of the leading causes of death in industrialized countries and is therefore one of the most challenging health problems in the 21st century. The International Diabetes Foundation assumes that 366 million people are suffering from diabetes in 2011 and that this figure will be rising to 552 million in 2030. In every country, the number of patients with type 2 diabetes mellitus (T2DM) is perpetually increasing, accounting for at least 90% of diabetes cases.1 But conditions predisposing for T2DM, so-called states of prediabetes, comprising impaired fasting glucose (IFG) and impaired glucose tolerance (IGT), are also major health problems, because of their greater risk for cardiovascular diseases.2 These findings underline the important task of public health to develop strategies for the prevention and early treatment of T2DM and prediabetic states. In order to address this task adequately, an understanding of the pathogenesis of these metabolic disorders is a prerequisite. Over the last decades, novel technologies have helped to broaden our knowledge and to elucidate pathways involved in the development of T2DM.
Type 2 diabetes mellitus represents a combination of impaired insulin sensitivity of tissues such as skeletal muscle, liver, adipose tissue and brain, also termed insulin resistance (IR), and inadequate insulin secretion, frequently characterized by impaired early prandial insulin release from pancreatic β cells.3 Of note, current definitions of T2DM also require the exclusion of mostly rare, but specific, causes of these abnormalities, such as monogenic defects of insulin signalling or secretion. There is a still-ongoing debate on whether IR or β-cell dysfunction is the primary cause or only the consequence of metabolic adaptation to the respective other abnormality. However, the tight feedback regulation of insulin action and secretion in order to maintain normal blood glucose concentrations makes a decision very difficult, if not impossible. Nevertheless, there is growing evidence that IR could be the earliest detectable abnormality ultimately leading to T2DM.4
This review will therefore primarily address the current knowledge on IR in the main target tissues of insulin, but also summarizes recent findings on lipid-induced β-cell dysfunction with relevance for the pathogenesis of T2DM.
In humans, two measures are mainly used to define IR: reduced insulin-stimulated glucose disposal and impaired insulin-dependent suppression of endogenous glucose production (EGP). Both measures are obtained in vivo from the gold standard test, the hyperinsulinemic–euglycemic clamp (HEC), and contribute to the increases in plasma glucose and insulin levels that occur in prediabetes and T2DM. However, in addition, tissues such as adipose tissue and the brain can indirectly contribute to IR and abnormal glucose and insulin concentrations5 (Figure 1).
Although the primary factors causing T2DM are still incompletely understood, it is widely accepted that IR plays a major role in its development. Recent results from the longitudinal Whitehall II cohort of British civil servants showed that insulin sensitivity starts to decrease linearly many years before T2DM manifestation, with an even steeper decrease during the last five years prior to diagnosis, whereas insulin secretion remains stable until four years before diagnosis.4 The risk of developing T2DM increases with male sex and increasing age, overweight and obesity, but first-degree relatives (FDR) of T2DM patients, women with a history of gestational diabetes (post-gestational diabetes mellitus; pGDM) or polycystic ovary syndrome, and persons with the so-called metabolic syndrome are also at a substantial risk.6
In IR states, insulin action is impaired and higher circulating insulin concentrations are required to ensure adequate cellular insulin signalling in its target tissues, thereby maintaining glucose homeostasis. Key questions regarding IR in humans are:
Which tissue-specific alterations define the insulin-resistant state?
Which are the main factors contributing to development of IR?
Here, we will review recent findings on the tissue-specific IR and discuss the impact of acquired and inherited factors.
Insulin resistance in skeletal muscle
On ingestion of a meal, peripheral tissues – mainly skeletal muscle – take up more than two-thirds of excess glucose, while less than one-third of glucose is stored in the liver.7,8 The meal-induced release of insulin is responsible for the increase in skeletal muscle glucose disposal. The HEC creates steady-state conditions of plasma insulin within the range of post prandial concentrations. Under these conditions, skeletal muscle is responsible for almost all of whole-body insulin-stimulated glucose disposal, also termed peripheral insulin sensitivity.9 Lean and obese persons with T2DM generally exhibit an approximately 50% lower whole-body glucose disposal than those without T2DM.10 Normal glucose tolerant (NGT) lean FDR of patients with T2DM show a similar reduction in whole-body glucose disposal compared with their diabetic parents, confirming the accredited importance of inherited factors in development of IR and T2DM.11–13 In obese humans and in pGDM, peripheral insulin sensitivity is frequently but not generally reduced.14–16 This indicates that both inherited and acquired factors can impair muscle insulin-sensitive glucose disposal in humans. The cellular pathways responsible for impaired muscle insulin sensitivity can involve reductions in glycogen synthesis and glucose transport/phosphorylation, but also in mechanisms contributing to impaired glucose oxidation such as altered mitochondrial function.17
Impaired glucose transport/phosphorylation and glycogen synthesis
Employing noninvasive 13C/31P magnetic resonance spectroscopy (MRS) revealed that muscle glycogen synthesis accounts for ∼90% of insulin-stimulated disposal and for virtually all non-oxidative glucose disposal in healthy NGT persons.18 Under conditions of combined hyperglycaemia and hyperinsulinaemia mimicking post prandial conditions, increases in skeletal muscle glucose-6-phosphate (G6P) concentrations as well as glycogen synthesis rates are markedly reduced in T2DM patients. Both abnormalities can be substantially improved by further raising plasma insulin concentrations,19–21 indicating that muscle glucose transport via glucose transporter 4 (GLUT4) is rate-controlling under these conditions.20 Recent data support this concept in humans carrying a mutation, present in 1 out of 70 UK whites, in the PPP1R3A gene encoding RGL, which is a key regulator of muscle glycogen content.22 This mutation results in a truncated version of RGL which fails to bind glycogen and co-localize with glycogen synthase, resulting in reduced glycogen synthesis. Carriers of this truncated PPP1R3A variant show severely reduced post prandial muscle glycogen synthesis rates, but no changes in glucose tolerance or insulin sensitivity, suggesting that impaired glycogen synthesis per se without other alterations does not necessarily lead to IR and/or IGT. Interestingly, insulin-stimulated increases in G6P concentration and glycogen synthesis rate are also impaired in young lean NGT insulin-resistant FDR of T2DM patients.23 Thus, inherited pre-existing defects in muscle glucose transport might be of importance for the development of IR in some groups at risk of T2DM. Of note, exercise training markedly improves muscle insulin-sensitive glucose transport/phosphorylation in some but not all FDR,11,24,25 whereas muscle glycogen synthesis does not normalize, suggesting additional abnormalities in nonoxidative glucose storage.11
Glucose transport rates adapt to demand by insulin-mediated translocation of GLUT4 to the plasma membrane as well as endocytosis of GLUT4 molecules.26 Indeed, reduction of insulin-mediated translocation of GLUT4 to the sarcolemmal membrane is a consistent feature of insulin resistant individuals and T2DM patients,27–29 but the underlying abnormalities remain in part unknown. Although impaired insulin receptor substrate 1 (IRS1) tyrosine phosphorylation and phosphatidylinositol-3-kinase (PI3K) activity have been widely reported,30 pathways downstream of IRS1 can also be altered in T2DM. Lower insulin-stimulated phosphorylation of the 160-kDa Akt substrate (AS160) has been reported for skeletal muscle of T2DM patients.31 Furthermore, the positive relationship between muscle insulin-stimulated glucose transport and AS160 phosphorylation observed in insulin-sensitive humans is disrupted in lean insulin-resistant FDR of T2DM, proposing an inherited impairment of AS160 action contributing to abnormal glucose metabolism.32 Another, PI3K-independent, defect in insulin signalling involves β-arrestin-2, which scaffolds Akt and Src to the insulin receptor.33 Expression of β-arrestin-2 is clearly reduced in skeletal muscle of mouse models of diabetes. Furthermore, knockdown of β-arrestin-2 exacerbates IR, whereas administration of β-arrestin-2 alleviates IR in these mice. Future studies are necessary to clarify the possible role of β-arrestin-2 in the pathogenesis of T2DM in humans.
Structural alterations in muscle cells may also inhibit insulin-induced recruitment of GLUT4 to the plasma membrane. Growing evidence supports a role for the small Rho family GTPase Rac1 in insulin-mediated GLUT4 membrane trafficking by reorganization of actin filaments.34 In muscle cells, reductions in insulin-activated Rac1 action and actin remodelling correlated well with reduced GLUT4 translocation.35 Furthermore, soluble NSF attachment protein (SNAP) receptor (SNARE) complexes have been implicated in insulin-regulated GLUT4 trafficking. SNAP23, which is required for insulin-stimulated translocation of GLUT4 to the plasma membrane, redistributes from the plasma membrane to the microsomal/cytosolic compartment in patients with T2DM.36 Apart from defective insulin-stimulated plasma membrane translocation, GLUT4 may not be properly localized in insulin-resistant and T2DM states under basal insulin conditions.28,32,37 Clathrin heavy-chain isoform, CHC22, might play a role in defective formation of insulin-responsive GLUT4 compartments, as CHC22 also relates to expanded compartments in skeletal muscle from patients with T2DM.37
Lower mitochondrial function and/or content
Whole-body glucose utilization consists of non-oxidative pathways, i.e. glycogen synthesis and oxidative pathways, i.e. mitochondrial oxidation. Although glycogen synthesis is quantitatively more important, impaired mitochondrial function can play a role for impaired muscle glucose disposal in IR and T2DM, at least under certain conditions.
Over the last few years, new technologies using in vivo 31P/13C MRS and ex vivo high-resolution respirometry have made it possible to exactly characterize muscle mitochondrial function in T2DM in humans. Unidirectional flux rates through adenosine triphosphate (ATP) synthase (fATP) reflect mitochondrial activity in the basal, non-exercising state; phosphocreatinine kinetics (PCr half-time) during recovery from submaximal exercise describe maximal adenosine diphosphate (ADP)-dependent mitochondrial capacity. In addition, the activities of electron transport chain complexes and tricarboxylic acid cycle fluxes can be assessed.
Overweight, metabolically well-controlled T2DM patients have 27% lower basal fATP than lean young NGT controls, but similar fATP to age- and body mass-matched nondiabetic humans.21 During stimulation by insulin, fATP increased, a response termed mitochondrial plasticity,38 but only in the nondiabetic groups. Of note, even combined elevation of both plasma insulin and glucose to overcome reduced muscle glucose uptake did not restore mitochondrial plasticity in T2DM patients, suggesting an abnormal mitochondrial response to altered substrate availability. A comparable group of overweight T2DM individuals also exhibited a 45% longer PCr half-time of recovery than matched humans with NGT, indicating impaired maximal oxidative phosphorylation in muscle with T2DM.39 In another comparable group of T2DM, the 25% reduction in in vivo mitochondrial capacity was in agreement with a 35% lower ADP-stimulated and 31% lower fluoro-carbonyl cyanide phenylhydrazone-driven maximal mitochondrial respiratory capacity.40 These findings suggest intrinsic abnormalities of mitochondrial function in T2DM patients occurring independently of local perfusion or substrate availability.
Impairment of resting and insulin-stimulated muscle fATP can be also present in lean insulin-resistant FDR of T2DM patients.13,41 However, another group of overweight elderly FDR exhibited only a trend towards lower oxidative capacity,40 whereas other rather younger FDR showed no difference in basal fATP.24 Furthermore, insulin sensitivity does not necessarily associate with mitochondrial function, either at baseline or after short-term or prolonged exercise training in FDR.24,25 It may be suggested that other factors, particularly age,42 hyperlipidaemia43 and obesity,44 influence mitochondrial function. Taken together, abnormal mitochondrial function may be an early feature of some groups at risk of T2DM, but does not fully explain IR in the pathogenesis of T2DM.
As overall mitochondrial function depends on mitochondrial content, changes in the latter could underlie the observed abnormalities in T2DM.45 Exact quantification of mitochondrial content requires transmission electron microscopy.38 This technique revealed no alterations in overall mitochondrial size in insulin-resistant and T2DM individuals in one study.46 On the other hand, mitochondrial density was indeed reduced in certain insulin-resistant states. Specifically, intermyofibrillar mitochondrial content was ∼40% and ∼24% lower in insulin-resistant humans with and without T2DM, respectively, whereas subsarcolemmal mitochondrial content was comparable between all groups.46 Likewise, mitochondrial density was 38% lower in young lean insulin-resistant FDR than in matched controls, suggesting that mitochondrial biogenesis may be – at least in part –genetically determined.47
Peroxisome proliferator-activated receptor γ co-activator 1 α (PGC1α) is a key regulator of skeletal muscle oxidative metabolism and mitochondrial biogenesis.48 Expression of genes encoded by PGC1α was found to be reduced in skeletal muscle of T2DM individuals and overweight non-diabetic individuals with a family history of T2DM.49,50 Most, but not all, recent studies support a role for PGC1α in explaining muscle IR.32,47,48 Experiments in muscle cell lines showed that increased expression levels of PGC1α improve IR by up-regulation of selected genes involved in β-oxidation, glucose transport and oxidative phosphorylation.51,52 Surprisingly, PGC1α-overexpressing animal models revealed that excessive PGC1α production causes deleterious metabolic effects by up-regulation of fatty acid translocase (FAT)/CD36 leading to accumulation of intramyocellular lipid intermediates, which interfere with insulin signalling.48 In contrast, modest increases in PGC1α improved mitochondrial fatty acid oxidation and increased GLUT4 expression as well as insulin-stimulated glucose transport.53 Thus, further studies are needed to elucidate the association of PGC1α with mitochondrial biogenesis, oxidative metabolism and the development of IR and T2DM in humans.
Intramuscular lipid content
Lipid deposition in other organs than in adipose tissue as so-called ectopic fat is generally negligible, but can be detected, e.g. in muscle (intramyocellular lipids, IMCLs) under certain conditions. In sedentary insulin-resistant humans, IMCL reflects an imbalance between energy supply and demand,54 which is determined by aerobic capacity.55 IMCLs strongly correlate with IR in various sedentary populations, independent of whole-body fat mass and diabetes status.56–58 In addition, in lean FDR of T2DM individuals, IMCLs in soleus muscle was up to 80% higher than in the control population along with about 30% and 85% lower fasting and insulin-stimulated fATP, respectively, indicating that the inherited mitochondrial impairment could predispose those subjects to IMCL accumulation with subsequent development of IR and T2DM.13,41 Nevertheless, recent studies showed reduced mitochondrial function independent of IMCL content in T2DM patients and FDR compared with body mass index (BMI)-matched NGT controls,39,40,54 so mitochondrial function is not necessarily coupled to IMCL content.
Role of lipotoxicity, inflammation and oxidative/endoplasmic reticulum stress
Although IMCLs per se are unlikely to explain the development of IR,59 augmented lipid availability and intracellular lipid metabolites can inhibit insulin signalling. This is supported by the fact that intravenous lipid infusion, which increases plasma free fatty acids (FFAs), clearly induces muscle IR with reduced glucose transport/phosphorylation, prior to reduction of fATP and without any alterations of IMCLs over a period of five hours.43,60 Of note, states of IR and T2DM are associated with greater concentrations of circulating plasma FFAs and triglycerides as well as of intracellular lipid intermediates in skeletal muscle. Growing evidence from animal models and clinical studies proposed lipid metabolites like diacylglycerol (DAG) and ceramides to be, at least in part, underlying skeletal muscle IR.59,61
Diacylglycerols are thought to promote protein kinase C (PKC) activation and PKC-mediated inhibition of insulin signalling, which in turn decreases activation of Akt59 (Figure 2). Short-term intravenous lipid infusion specifically activated PKCθ in rat skeletal muscle,62 which enhanced serine phosphorylation of IRS1,63 resulting in reduced insulin-stimulated GLUT4 membrane trafficking. Furthermore, PKCθ knock-out mice and transgenic mice carrying muscle-specific IRS1 serine (Ser) to alanine (Ala) mutations were both protected against lipid-induced IR.64,65 Similarly, in humans, PKC activity was found to be up-regulated by lipid infusion as well as in T2DM states.66,67 However, the contribution of PKCs to IRS serine phosphorylation in humans remains unclear, as another recent study did not confirm the previous results.68
In contrast to DAGs, both exogenous and endogenous ceramides lead to direct inhibition of Akt by inhibition of Akt phosphorylation without affecting upstream insulin signalling.70,71 As shown for DAGs, ceramide-mediated Akt inhibition also results in decreased GLUT4 membrane trafficking in skeletal muscle.71 Of note, only saturated fatty acids, but not unsatured fatty acids seem to enhance ceramide accumulation in skeletal muscle.72,73
Furthermore, serine/threonine phosphatase glycogen synthase kinase 3 (GSK3), which is involved in glucose transport and glycogen synthesis, has been implicated in serine phosphorylation of IRS1.74 GSK3 overactivity has been observed in skeletal muscle of insulin resistant animal models and T2DM patients75,76 (Figures 2 and 3). In female lean Zucker rats, a high-fat diet caused overactivation of GSK3, again linking lipid oversupply to impaired insulin signalling.77
In insulin-resistant states, tyrosine phosphorylation of IRS is markedly reduced and IRS serine phosphorylation enhanced, resulting in impairment of downstream insulin signalling. In skeletal muscle, IRS1 serine residues 307 (312 in the human IRS1 homologue) and 1011 have been identified as target residues of lipid-induced PKCθ.78 In liver, fatty acids also promote Ser307 phosphorylation by activation of PKCδ, in adipose tissue by PKCθ.
Free fatty acids induce skeletal muscle IR via two pathways: activation of Toll-like receptors (TLR) on the cell membrane and lipid metabolite accumulation. Lipid metabolites are linked to activation of PKCs, which apart from directly interfering with insulin signalling enhance inflammatory signalling pathways involving inhibitor of nuclear factor kappa-B kinase subunit beta (IKKB) and c-JUN NH2-terminal kinase (JNK). Activation of both kinases can also lead to serine phosphorylation of IRS1.79–81 As TLR4 deficiency and TLR2 knockdown protect from lipid-induced IR, the TLR system could contribute to IR.82–85 Similarly to lipid intermediates, TLRs activate pro-inflammatory IKKB and JNK.79 Finally, pro-inflammatory cytokines released from adipocytes and resident macrophages in adipose tissue,86,87 and also in skeletal muscle,88 can promote inflammation and IR in skeletal muscle.
Excessive dietary FFAs supply emerges as a negative regulator of insulin action due to induction of oxidative stress, not only by uncoupling oxidative phosphorylation but also by decreasing antioxidant defence mechanisms.89,90 In rats, hyperinsulinaemia enhanced muscle reactive oxygen species (ROS) production and also activated the proinflammatory IKKB pathway.91
The endoplasmic reticulum (ER) is the major site for protein synthesis and newly synthesised, properly folded proteins are released to the Golgi complex. Enhanced misfolding or unfolding of peptides, also termed unfolded protein response (UPR), reflects ER stress.92 Prolonged ER stress can also inhibit IRS1 activation by Ser307 phosphorylation through the JNK pathway, whereas prevention of ER stress in turn has beneficial effects on glucose homeostasis and insulin sensitivity.93–96 Interestingly, palmitate-induced ER stress can have no consequences for insulin signalling.97 Thus, the interaction between lipids and ER stress and its role in the development of IR is still a field of investigation.
Insulin resistance in liver
The liver is mainly responsible for EGP, which determines blood glucose concentrations under fasting conditions. Endogenous glucose is derived either from glycogen stores via glycogenolysis or by de novo synthesis via gluconeogenesis. Both pathways contribute to EGP under fasting conditions, in which insulin can decrease EGP mainly by inhibiting glycogenolysis.98 After meal ingestion, insulin decreases EGP by inhibiting gluconeogenesis as well as glycogenolysis.99 Fasting EGP is modestly elevated (∼10–25%) or unchanged in overt T2DM, dependent on glycaemic control.100–102 Furthermore, suppression of EGP during meal ingestion or combined glucose-insulin infusion is lower and delayed in T2DM patients.103 Of note, the excessive post prandial hyperglycaemia observed even in well-controlled T2DM patients is not solely explained by impaired suppression of EGP but also by decreased hepatic glycogen accumulation.103
The cellular pathways responsible for impaired hepatic insulin sensitivity in IR and T2DM involve several mechanisms such as alterations in hepatocellular lipid (HCL) storage and mitochondrial function.17
Impaired glycogen fluxes and gluconeogenesis
In T2DM, elevation of fasting EGP results mostly from increased rates of gluconeogenesis and becomes more pronounced with increasing severity of diabetes and obesity.98,101,104,105 In contrast, glycogenolysis contributes less to fasting EGP106 and may be even comparable in absolute terms between T2DM and normoglycemic BMI-matched controls.107 Nevertheless, the glycogenolytic fluxes of the patients seem inappropriately high for the ambient fasting glucose and insulin concentrations. Under comparable hyperinsulinemic–euglycemic conditions, the impaired insulin-mediated suppression of EGP in obese individuals with or without T2DM results from lower suppression of both gluconeogenesis101,104 and glycogenolysis.101 In young, lean insulin-resistant FDR, fasting EGP and rates of gluconeogenesis did not differ from insulin-sensitive controls.108 In the face of greater insulin concentrations, the unchanged gluconeogenesis suggests hepatic IR of gluconeogenesis.
In the post prandial state, T2DM patients exhibit markedly lower rates of net hepatic glycogen synthesis, which persist even during maximal stimulation by combined hyperglycaemia and hyperinsulinaemia.103 Hepatic glycogen content inversely correlates with fasting plasma glucose levels, suggesting that impairment of glycogen synthesis may be a secondary phenomenon arising from chronic hyperglycaemia (glucotoxicity) associated with prolonged IR and β-cell dysfunction.109 Likewise, post prandial glycogen synthesis was not impaired in young, lean but insulin-resistant FDR in the absence of hepatic steatosis.110
As in skeletal muscle, Akt is the important node for insulin-mediated signalling pathways in liver. Akt recruits downstream effector molecules like GSK3 and forkhead box O1 (FOXO1) to adapt hepatic metabolism to current needs by promoting glycogen synthesis and inhibiting gluconeogenesis as well as by regulation of lipid metabolism (Figure 2).59,111,112 FOXO1 is active during fasting and inactivated by Akt-mediated phosphorylation after feeding. Antagonism of FOXO1 by Akt is the predominant mechanism by which insulin suppresses EGP after a meal.113 Uncontrolled FOXO1 activity caused profound metabolic abnormalities in gluconeogenesis, glycogen storage as well as in lipid metabolism in insulin resistant mice, whereas antagonization or reduction of hepatic FOXO1 activity greatly improved glucose tolerance and insulin responsiveness.113 Humans with steatosis have elevated expression of hepatic FOXO1 protein suggesting a link between hepatocellular lipids (HCLs), IR and FOXO1.114 The other insulin effector enzyme, GSK3, is involved in regulation of glycogen synthesis115 and EGP. The role of GSK3 in hepatic IR is less clear, because the effects seem to be isoform-dependent. Targeted deletion of GSK3α in mice resulted in enhanced hepatic insulin sensitivity and glucose homoeostasis. Furthermore, fasted and glucose-stimulated hepatic glycogen content was enhanced in GSK3α knock-out mice. In contrast, liver-specific GSK3β knock down remained without any measurable metabolic effect.116,117 Concordantly, administration of a specific GSK3 inhibitor, acting on both isoforms, markedly improved insulin-stimulated suppression of EGP rates in high-fat diet-induced insulin-resistant mice.118
Glycogen synthesis further depends on the rate of glucose phosphorylation catalysed by glucokinase (GK). In Zucker diabetic fatty (ZDF) rats, GK expression in liver is progressively reduced with the development of hyperglycaemia indicating that elevated plasma glucose levels might severely interfere with GK function and therefore glycogen synthesis.119 In humans, data on hepatic GK activity and expression are rather conflicting.120,121 One recent study reported a positive correlation of GK messenger ribonucleic acid (RNA) expression levels with HCLs and markers of de novo lipogenesis, both in turn associated with T2DM.122 The contribution of altered GK expression to accumulation of HCLs and IR in humans needs to be addressed in future studies.
As with skeletal muscle, β-arrestins might also be relevant for hepatic IR, as β-arrestin-1 and -2 are down-regulated in livers of T2DM patients, and β-arrestin-2 knock-out mice have reduced insulin-mediated suppression of EGP.33
Impaired mitochondrial function
Metabolically well-controlled but slightly overweight T2DM patients exhibit features of lower mitochondrial functions, such as lower intrahepatic concentrations of γATP and inorganic phosphate (Pi) as well as hepatic fATP, than nondiabetic humans.123,124 Although HCL content and hepatic IR correlate negatively with measures of mitochondrial function, abnormal mitochondrial function is specifically present in T2DM individuals, indicating steatosis-independent hepatic mitochondrial abnormalities in T2DM.123 Furthermore, obese insulin-resistant but NGT subjects had similar hepatic γATP and Pi to lean insulin-sensitive but otherwise matched controls, suggesting that moderate chronic hyperglycaemia, rather than lipotoxicity or other causes, is responsible for the abnormal energy metabolism in human liver.124 At present, no data exist on hepatic ATP turnover in (insulin-resistant) FDR subjects, which would help to assess the contribution of inherited factors to altered hepatic mitochondrial function.
Accumulation of hepatocellular lipids
Hepatic IR correlates with increased HCLs, suggesting that HCLs may be a major contributor to reduced hepatic insulin action.125,126 Lipid accumulation in the liver associates with different cohorts featuring IR, such as people with obesity or the so-called metabolic syndrome, women with a history of gestational diabetes, and people with T2DM.127,128 There is still debate over whether hepatic steatosis develops prior to, in the setting of, or secondary to hepatic IR.129,130 A recent study proposed that the lipogenic transcription factor carbohydrate response element binding protein (ChREBP) dissociates steatosis from IR by conversion of saturated FFA into monounsaturated FFA, preserving insulin-responsive Akt phosphorylation.131 In humans with severe IR, liver ChREBP expression levels were reduced, and protection against impairment of insulin signalling may be lost by this mechanism in pre-diabetic states.
Hepatic FFAs are derived from different sources: dietary fat, lipolysis of adipose tissue and hepatic de novo lipogenesis. Alterations in cellular FFA synthesis, delivery, uptake and/or oxidation could lead to increased HCLs. In the animal model, hepatic steatosis developed quickly when rodents were fed a high-fat diet; and also in humans, a 2- or 3-week high-fat diet with 55% or 56% energy as fat, respectively, markedly increased HCLs and fasting insulin levels, the latter indicating IR.132,133 However, in obese individuals with hepatic steatosis, FFAs derived from adipose tissue contribute up to 60% of triglycerides in the liver,134 identifying an important role for adipose tissue in development and maintenance of hepatic steatosis and IR.
Increased de novo lipogenesis
Skeletal muscle IR has been implicated in the pathogenesis of hepatic steatosis and IR by altering the distribution pattern of post prandial energy storage.110 Defects in post prandial muscle glucose disposal in insulin-resistant individuals could cause tissue redistribution of carbohydrates with increased glucose uptake and de novo lipogenesis by the liver followed by a subsequent increase in hepatic triglyceride synthesis and plasma triglycerides.
In healthy lean individuals, de novo lipogenesis is only a minor source for hepatic triglycerides under fasting conditions.134,135 On the other hand, obese hyperinsulinemic individuals have already maximally stimulated de novo lipogenesis in the fasted state, which is responsible for up to one-fourth of the FFA pool.134 Increased post prandial de novo lipogenesis and plasma triglycerides are also present in young lean insulin-resistant FDR110 and attributed to increased nutrient delivery to the liver due to reduced skeletal muscle glucose disposal. Post prandial hyperinsulinaemia in insulin-resistant individuals may promote the expression of the sterol regulatory element binding protein 1 (SREBP1), a key regulator for transcription of enzymes involved in hepatic lipogenesis.136 In mice, hepatic overexpression of SREBP1 resulted in increased lipogenesis and the development of hepatic steatosis,137 whereas inactivation of the SREBP1 gene in an obese mouse model with hepatic steatosis markedly lowered the HCLs.138 Thus, in lean insulin-resistant humans, de novo lipogenesis may be one of the first alterations in hepatic metabolism contributing to HCL accumulation and IR.59 Increased hepatic de novo lipogenesis seems to precede adipose tissue IR, as lean insulin-resistant FDR have similar rates of subcutaneous fat lipolysis and plasma adipokines to insulin-sensitive controls13,110.
Role of lipotoxicity, inflammation and oxidative/endoplasmic reticulum stress
In analogy to skeletal muscle, FFA oversupply and intrahepatocellular lipid intermediates may be also important for the development of hepatic IR and HCL accumulation.139 A high-fat diet rapidly induces hepatic steatosis and IR in rats.140 Defective hepatic insulin signalling is associated with increased hepatocellular DAG, activation of PKCε and pro-inflammatory JNK141 (Figure 3). Of note, treatment with a specific PKC antisense oligonucleotide improved hepatic insulin sensitivity and insulin signalling in rats on a high-fat diet.142 The specific role of PKCε in hepatic IR has been confirmed also in other rodent models.143,144 Most importantly, levels of PKCε – and to a lesser extent PKC alpha and zeta – in the membrane fraction were also increased in the livers of obese T2DM patients.145 Recently, PKCε has been shown to be also modulated by ER stress.146
Nevertheless, other novel PKC isoforms such as PKCδ can be up-regulated in rodent liver on high-fat diet.147 Moreover, obese humans have elevated transcription of the PRKCD gene, which encodes PKCδ and positively correlates with fasting glucose and triglycerides. Targeted deletion of the PRKCD in mice resulted in increased hepatic insulin signalling and reduced expression of gluconeogenic and lipogenic enzymes, whereas overexpression had the converse effect. PKCδ may act via a 70-kDa ribosomol S6 kinase (p70SK6), which causes phosphorylation of IRS1 on Ser307 and thereby affects hepatic glucose fluxes. In parallel, PKCδ can also promote SREBP1c signalling and hepatic de novo lipogenesis.
Furthermore, accumulation of saturated FFAs and/or FFA metabolites has been reported to directly induce abnormal mitochondrial function and oxidative stress in liver cells by lysosomal breakdown and subsequent cathepsin B activation.148
As UPR is involved in regulation of lipogenesis, ER stress can be of great importance for the pathogenesis of hepatic IR and steatosis. In wild-type mice, enhanced UPR can result in a transient rise of HCLs,149 whereas suppression of UPR reduces HCLs in obese mice due to decreased expression of SREBP1 and ChREBP.150 Furthermore, ER stress is associated with reduced insulin-stimulated Akt phosphorylation and glycogen synthesis in the liver.151 The ER stress response protein X-box-binding protein-1 (XBP1) has been linked to regulation of hepatic lipogenesis.152 XBP1-deficient mice showed increased insulin sensitivity and reduced hepatic DAG content despite JNK activation.146 Therefore, defective modulation of hepatic lipogenesis due to ER stress may precede ER stress-induced hepatic IR.
Insulin resistance in adipose tissue
After skeletal muscle and liver, adipose tissue is a third important organ with insulin-stimulated peripheral glucose uptake and regulator of glucose homeostasis, even if it only accounts for about 5% of post prandial glucose disposal.153 Adipocytes need glucose to synthesise glycerol-3-phosphate, the basis for triglyceride production. Insulin acts in different ways on adipose tissue; it promotes fasting and post prandial glucose uptake as well as inhibition of triglyceride hydrolysis with subsequent FFA release into the circulation (lipolysis) and re-esterification of FFAs into triglycerides. In insulin-resistant and T2DM individuals, adipose tissue glucose uptake and insulin action are diminished (adipose tissue IR). This leads to only partial suppression of FFA release from adipose tissue under fasting conditions followed by permanent elevation of circulating FFA levels, which in turn promotes muscle and hepatic IR.154
In addition to its role in energy metabolism, adipose tissue serves as an endocrine organ, secreting various bioactive peptides, so-called adipokines, which regulate inflammatory and metabolic processes.155 Obese individuals who progress to T2DM display features of inflammation years before the onset of the disease termed ‘low-grade’ inflammation.156,157 Several reports demonstrate the important role of macrophage accumulation in the adipose tissue of obese subjects as well as their involvement in inflammatory pathways within adipocytes.158 These macrophages per se are a source of proinflammatory factors, but they may also act on adipokine secretion from adipocytes.158,159 Pro-inflammatory cytokines like interleukin 6 (IL-6), IL-β and tumour necrosis factor α (TNFα) are frequently increased in the plasma of humans with obesity, IR and T2DM.160–162 However, novel adipokines like resistin, retinol binding protein-4 (RBP4), CXC5 and lipocalin 2 have also been implicated in inhibition of insulin signalling.155 Accordingly, anti-inflammatory treatment with acetylsalicylate improves glycaemic control and IR in obese rodents and in human T2DM,163 most probably via inhibition of IKKB.164 Furthermore, studies in humans with massive weight loss on bariatric surgery underline the role of adipose tissue-mediated inflammation for IR and obesity-related T2DM.165,166
In adipocytes, insulin signalling activates Akt to regulate GLUT4-mediated glucose uptake and antilipolysis.167 Furthermore, insulin stimulates mitogen-activated protein kinases (MAPK) Erk 1 and 2 as well as stress-activated protein (SAP) kinases JNK and p38MAPK, which mediate reduction in insulin signalling and promote IR.168 Regulation of both signalling pathways is regarded as a key controller of adipose tissue insulin sensitivity, and the balance of both pathways can be disturbed by several mechanisms related to IR.169 Thus early growth response protein 1 (EGR-1), a zinc finger transcription factor highly expressed in obese mouse models and in fat tissue of individuals with T2DM, caused persistent Erk/MAPK activity due to chronic exposure to hyperinsulinaemia in adipocytes.168,170 The sustained activation of Erk1/2 lead to phosphorylation of IRS1 at Ser612, which impaired insulin signalling and glucose uptake in adipocytes. Furthermore, also the adipokine TNFα has been reported to induce IR in visceral adipocytes via enhanced JNK1/2 activation and serine307/312 phosphorylation of IRS1.171
Role of lipotoxicity, inflammation and oxidative/endoplasmic reticulum stress
Excessive availability of FFAs, in part released from the adipocytes themselves, appears to play a role also in the development of adipose tissue IR.172 In the presence of high FFA concentrations, 3T3-L1 adipocytes develop impairment of insulin signal transduction, insulin-stimulated GLUT4 translocation and glucose transport. In 3T3-L1 adipocytes, lipid oversupply activated PKCθ, which in turn induced IKKB and JNK, indicating that PKCs are also involved in adipocyte IR. As in skeletal muscle, PKCθ decreased IRS1 activity by enhanced phosphorylation of Ser307173 (Figure 3).
Lipid oversupply by a high-fat diet has also been linked to up-regulation of pathways for ROS production and oxidative stress, which can be detected even before the onset of IR.174 In mice, a high-fat diet led to down-regulation of genes regulating fatty acid synthesis and up-regulation of nicotinamide adenine dinucleotide phosphate (NADPH) oxidase complex, indicating that ROS may be involved in early development of adipose tissue IR.
In adipose tissue, early lipid-mediated inflammatory processes may be initiated, at least in part, by ER stress and/or reduced ER function.93 Mice with a heterozygous deletion of the 78-kDa glucose-regulated protein (Grp78), a key co-ordinator of ER homeostasis, were protected against diet-induced obesity, hepatic steatosis, adipose tissue inflammation and IR, but also showed increased energy expenditure175 linking ER integrity to energy balance, glucose homeostasis, and adipocyte stress.176
Insulin resistance in the central nervous system
Insulin receptors are widely expressed throughout the brain with high concentrations in the hypothalamus and hippocampus as well as the cerebral cortex, and insulin action plays an important regulatory role in the central nervous system (CNS).177 Furthermore, the insulin-responsive GLUT4 has been found in specific brain regions such as the hypothalamus, hippocampal formation, cortex and cerebellum.178 Insulin action in the human brain has been studied in vivo by using electroencephalography (EEG), magnetoencephalography (MEG), positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). In particular, fMRI allows for detection of neuronal activity even in deep brain structures.177 In healthy humans, an oral glucose load acutely mitigates hypothalamic neuronal activity and downstream neuronal circuits subsequently adapt food intake and metabolism to maintain energy homeostasis.179–182 In contrast, oral glucose ingestion fails to inhibit hypothalamic neuronal activity in patients with T2DM.179 Evidence for reduced insulin-stimulated neuronal activity comes from lower neuronal firing in obese than in lean individuals during a hyperinsulinemic–euglycemic clamp.183 Of note, the effects of insulin on cortical activity were negatively related to body fat mass and severity of IR.
One possible explanation for impaired CNS insulin signalling is a disturbance of blood–brain barrier function. Insulin concentrations in the cerebrospinal fluid (CSF) reflect those in peripheral plasma only in lean, but not in overweight individuals.177 However, intranasal administration of insulin in order to bypass the blood–brain barrier revealed that insulin reaches the CSF in obese subjects also. Nevertheless, intranasal insulin application mediated regulatory effects on body weight in lean but not in obese humans, indicating further defects in CNS insulin signalling in obesity apart from impaired transport across the blood–brain barrier.184 Furthermore, changes in CSF-to-plasma ratios of leptin and visfatin may contribute to alterations in central insulin action.185,186
Central nervous consequences of impaired insulin signalling
Mice with neuron-specific inactivation of the insulin receptor (NIRKO mice) provided important insights into the physiological role of insulin in the CNS. Inactivation of the insulin receptor had no impact on brain development, morphology or neuronal survival as shown by unchanged brain weights, apparent structures and glial cell activation.187 However, cerebral glucose metabolism has not been investigated in this model.
Insulin action on cerebral glucose metabolism remains uncertain despite the identification of insulin receptors and insulin-sensitive GLUT4.188 Although many studies failed to demonstrate an effect of insulin on cerebral glucose metabolism in humans,188 insulin-evoked changes in cerebral metabolic rate for glucose as measured with PET using [18F]fluorodeoxyglucose might be impaired in IR,189 particularly in regions subserving appetite and reward.
Central nervous system insulin signalling should involve IRS1/2, PI3K and phosphatidylinositol-3-phosphate (PIP3) with subsequent activation of Akt, MAPK signalling pathways as well as KATP channels.190,191 Insulin-mediated activation of these pathways leads to short-term changes within 3–4 hours in neuronal activity or prolonged changes in gene transcription and neuronal plasticity.192,193
Patients with Alzheimer's disease (AD), who share some pathophysiological features with T2DM patients, exhibit cerebral IR associated with enhanced serine phosphorylation of IRS1 at Ser616 and 636/639 and reduced IRS1 activity.194 Furthermore, impaired IRS1 activity in subjects with IRS1 Glycine972Arginine polymorphism was also related to an imbalance of insulin action and T2DM risk.195,196 Carriers of this polymorphism showed no insulin effects using MEG during hyperinsulinemic–euglycemic clamp when compared with carriers of the wild-type IRS1.183 Taken together, cerebral IR results from defects in canonical insulin signalling rather than from impaired insulin transport across the blood–brain barrier.177
Impaired CNS insulin signalling not only affects central, but also peripheral metabolic processes. In humans, intranasal administration of insulin reduces circulating leptin concentrations, body fat and body weight in lean healthy men, but not in women.197 Even an 8-week treatment with insulin did not affect body weight, fat or circulating leptin in obese men. In accordance to effects in normal-weight men, declarative memory and mood were improved and hypothalamic–pituitary–adrenal axis activity was reduced as assessed by circulating adrenocorticotropic hormone (ACTH) and cortisol levels,184 indicating specific CNS IR to regulation of energy homeostasis in obesity. However, intranasal insulin application can acutely improve insulin sensitivity in T2DM, which was accompanied by activity in different brain regions: first the hypothalamus, and later the putamen, right insula and orbito-frontal cortex.198
In animal models, central insulin signalling contributes to the regulation of peripheral insulin action in all major insulin-sensitive tissues.199 EGP, lipid fluxes in adipocytes and muscle glycogen synthesis are under the control of central insulin.200–202 Infusion of insulin into the third cerebral ventricle of rats effectively suppressed EGP independent of circulating levels of insulin and other glucoregulatory hormones. Conversely, blocking central molecules of CNS insulin signalling impaired the ability of circulating insulin to inhibit EGP.200 Furthermore, hypothalamic insulin administration increased white adipose tissue lipogenic protein expression, inactivated hormone-sensitive lipase and suppressed lipolysis. NIRKO mice, in contrast, exhibited unrestrained lipolysis and decreased de novo lipogenesis in adipose tissue.203 Thus the brain and, in particular, hypothalamic IR may contribute to hyperglycaemia and elevated plasma FFA concentrations in T2DM.
Of note, despite their obesity, NIRKO mice have elevated circulating plasma leptin concentrations, suggesting that peripheral IR also associates with some degree of CNS resistance to leptin action, a mechanism that may be involved in development of obesity in insulin-resistant states. Data supporting this hypothesis come from animal and human studies: obese leptin-deficient mice show both hyperglycaemia and IR, whereas lipodystrophic mice benefit from leptin treatment in the absence of white adipose tissue with favourable effects on IR.139 Furthermore, long-term leptin replacement therapy in humans with severe, generalized lipodystrophy and leptin deficiency reversed both hepatic and skeletal muscle IR, and this was accompanied by a complete resolution of the severe hepatic steatosis and reduction in IMCL content (Figure 4).204
Role of lipotoxicity, inflammation and oxidative/endoplasmic reticulum stress
High-fat feeding disrupts insulin signalling pathways in the rodent brain which regulate glucose and energy homeostasis205,206 by affecting the central insulin-signalling pathways lowering food intake in response to nutrient oversupply. This may be associated with an impairment of Akt activation, partly caused by activation of hypothalamic PKCθ and S6K and inhibition of IRS1/2,206 but also by activation of pro-inflammatory IKKB (Figure 4).207
As in skeletal muscle, lipid metabolites, especially DAGs, may be involved in FFA-induced activation of PKCθ. Indeed, animals exposed to a high-fat diet with saturated FFA had significantly elevated DAG levels relative to oleic acid- and low-fat-fed animals. Also, in in vitro experiments, cells exposed to palmitic acid exhibited a significant increase in DAG levels, which did not occur in cells exposed to oleic acid, providing a mechanism by which PKCθ is activated.208 In the brain, PKC-θ has been assigned to neuropeptide Y and leptin-responsive neurons in the arcuate nucleus, rendering these sites of modulation of EGP and body weight-regulation responsive to FFA.
Recent studies also link increased pro-ceramide gene expression in liver and increased serum ceramide levels to CNS IR. In rats, ceramide administration led to alterations in gene expression of insulin, but in addition signalling through IRS1 and Akt was impaired in the CNS. Thus, peripheral sources of ceramides might have detrimental effects on the CNS, such as the cognitive impairment seen in T2DM.208 Central ceramides further seem to induce CNS ER stress.209
High-fat feeding induces IKKB activation in hypothalamic neurons, at least in part via ER stress.210 Forced activation of hypothalamic IKKB decreases insulin and leptin signalling and actions, whereas IKKB inactivation/antagonism protected mice from development of obesity and glucose intolerance,210 proposing IKKB activation as a lipid-mediated neuronal mechanism of energy imbalance.
Central insulin further promotes a subtle, transient increase in hypothalamic ROS levels, which is required for insulin-mediated regulation of food intake inhibition.211 Normally, excess ROS are inactivated by the antioxidant defence systems. A high-fat diet induced excessive hypothalamic ROS production, which has been proposed to lead to stimulation of the antioxidant defence system, overquenching insulin-dependent ROS production. This may result in CNS IR and impaired energy intake contributing to development of obesity and peripheral IR. These data are supported by other studies which detected CNS oxidative stress in high-fat diet-induced obesity and hypothalamic IR.207,212
There is an ongoing debate about whether β-cell dysfunction apart from IR is an early factor promoting development of T2DM.4,213 Indeed, many obese patients do not develop T2DM although being severely insulin resistant. Conversely, FDR of T2DM patients without impairment of glucose homeostasis show diminished β-cell function with reduction in insulin and amylin responses indicating that genetic predisposition for impaired β-cell function might play a role in the early pathogenesis of T2DM.214 The pathological process of β-cell dysfunction in the pathogenesis of T2DM seems to be of multifactorial origin with glucose, dyslipidaemia, inflammatory processes and ER stress playing a role in the maladaptation of β cells.215–217
Chronically elevated levels of FFA have been implied in the pathogenesis of both IR and β-cell dysfunction,218 and the detrimental effects of FFAs on β-cell function have been referred to as lipotoxicity.219 Consequences of excessive lipid availability include decreased insulin secretion and synthesis as well as increased β-cell apoptosis. Various mechanisms underlying lipotoxicity have been proposed, and are summarized below.
In ZDF rats, elevation of circulating FFAs and intracellular triglyceride accumulation in pancreatic β-cells are associated with a marked reduction of glucose-stimulated insulin secretion.220 Endocrine and exocrine pancreas show an almost identical pattern of lipid accumulation in these rats, suggesting that measurement of pancreatic lipids by 1H MRS of the human pancreas may predict β-cell dysfunction and T2DM risk.221 Indeed, pancreatic steatosis has been observed in obese individuals and is further increased in obesity combined with IGT and/or IFG.222 Several studies in different cohorts showed that pancreatic steatosis negatively relates to β-cell function and insulin secretion,223,224 but no notable correlation between pancreatic steatosis and T2DM has been reported so far.218,224,225 At present, it is not clear if pancreatic steatosis is a real primary factor in the development of T2DM or just secondary to obesity.218 Of note, two recent reviews on ectopic fat accumulation share the opinion that pancreatic steatosis is not a cause of lipotoxicity in pancreatic β cells.54,218
Short-term exposure to elevated FFA enhances insulin secretion in vitro and in vivo226 in cell culture experiments and animal models. In humans, a four-day lipid infusion with elevation of FFA plasma levels comparable with T2DM patients resulted in increased insulin secretion in NGT subjects without a family history of diabetes, whereas NGT FDR of T2DM patients presented a decline in first- and second-phase insulin secretion.227 Therefore, in individuals at risk for T2DM, β-cell lipotoxicity may contribute to manifestation of the disease.228
There is growing evidence that the interplay of glucotoxicity and lipotoxicity promotes major damage to pancreatic islets.229,230 Recently, glucolipotoxicity has been proposed to contribute to poor pancreatic β-cell proliferation and reduced β-cell mass in T2DM.231 A four-day lipid infusion in mice did not alter basal β-cell proliferation but blocked glucose-stimulated proliferation without induction of excess β-cell death in mice. Elevated FFA increased the expression of the inhibitor of cyclin-dependent kinase 4 (INK4) family cell cycle inhibitors p16 and p18, and inhibition of p16 or p18 was sufficient to abolish the antiproliferative effects of FFAs, but the relevance of these studies for human β-cell dysfunction remains to be clarified.232
There is evidence for further lipotoxic effects, including oxidative and ER stress as well as inflammatory processes.226,233,234 Lipid metabolites such as DAG and ceramides may induce different pathways resulting in β-cell failure. FFA-induced activation of PKCε has been proposed to be involved in β-cell dysfunction. PKCε-deficient mice on high-fat diet showed augmented insulin secretion and a PKCε-inhibitory peptide ameliorated insulin secretion in obese diabetic mice.233 Furthermore, ceramides have been linked to β-cell apoptosis in ZDF rats.235
β-Cell inflammatory processes comprise involvement of multiple factors: interleukines, TLRs and inflammatory kinases like IKKB and JNK are proposed to play a role, the latter activated mainly by ER and oxidative stress. ER/oxidative stress further activates PKC, interfering with insulin signalling and promoting β-cell failure.236 Antioxidant treatment in humans reduced plasma markers of oxidative stress and improved FFA-induced decline in β-cell function, in most, but not all studies, suggesting that oxidative stress is implicated in lipid-induced β-cell failure in humans.226
The development of T2DM is of multifactorial origin, comprising inherited and acquired factors leading to metabolic and cellular changes, finally resulting in IR and β-cell dysfunction. Growing evidence suggests that IR precedes β-cell dysfunction and that the earliest abnormalities are detectable in skeletal muscle. Subsequently, liver and adipose tissue are important for further deterioration of IR, whereas the impact of CNS in human IR remains uncertain.
The major cellular mechanism operating during the initiation of IR resides in the tissue-specific accumulation of lipid species, which results from an imbalance between lipid uptake and oxidation as well as triglyceride turnover in insulin-responsive tissues.
Nevertheless, β-cell function also plays an important role in the development of T2DM by maintaining normoglycaemia for long periods despite increasing IR. Inherited factors, but also rising glucose and fatty acid concentrations, drive progressive β-cell failure prior to manifestation of overt T2DM.