|Home | About | Journals | Submit | Contact Us | Français|
Insulin resistance is a sine qua non of Type 2 diabetes, and is associated with many other clinical conditions. Decades of research into mechanisms underlying insulin resistance have mostly focused on problems in insulin signal transduction and other mitochondrial and cytosolic pathways. In contrast, relatively little attention has been paid to transcriptional and epigenetic contributors to insulin resistance, despite strong evidence that such nuclear mechanisms play a major role in the etiopathogenesis of this condition. In this review we summarize the evidence for nuclear mechanisms of insulin resistance, focusing on three transcription factors with a major impact on insulin action in liver, muscle, and fat.
Insulin resistance is the condition in which a cell, tissue, or organism fails to respond appropriately to a given dose of insulin. Insulin performs a wide variety of functions, and not all of these activities need to be dampened in order to make a diagnosis of insulin resistance; typically insulin resistance refers to the metabolic actions of insulin, and specifically to the ability of insulin to promote glucose uptake into tissues like muscle and adipose and to repress glucose production in the liver. Insulin resistance is classically associated with Type 2 diabetes (T2D), where it is a driver of hyperglycemia. Currently, Type 2 diabetes affects 12–14% of US adults, and over half of all adults have pre-diabetes, a condition accompanied by significant insulin resistance (Menke et al., 2015).
Insulin resistance accompanies a wide range of pathological conditions, including obesity, lipodystrophy, sepsis, steroid use, growth hormone excess, polycystic ovarian disease, cancer, neurodegenerative disease, and even some physiological conditions, such as pregnancy. This can be modeled experimentally: investigators have developed numerous models of insulin resistance using a variety of chemical, drug, inflammatory, and nutritional challenges (Boucher et al., 2014). Thus arises a fundamental question: are there many independent paths to insulin resistance? Or do different perturbagens and clinical conditions converge on one or a few key pathways that are integral to any form of insulin resistance?
Despite years of study, there is still great uncertainty concerning how cells and organisms become insulin resistant. Enormous effort has been expended delineating the signal transduction pathways activated by insulin, resulting in a fairly detailed map of the intermediates involved (Boucher et al., 2014). As predicted, mice lacking many of these signaling intermediates are profoundly insulin resistant, as are humans with similar loss-of-function mutations (Biddinger and Kahn, 2006, Huang-Doran and Savage, 2011). However, most insulin resistant people do not harbor such mutations, so attention has focused on other causes of insulin resistance, including endoplasmic reticulum stress, inflammation, accumulation of toxic lipid intermediates (such as diacylglycerol and acylcarnitines), and reactive oxygen species (Hotamisligil, 2010, Houstis et al., 2006, James et al., 2012, Olefsky and Glass, 2010). Data supporting the contribution of these non-mutually exclusive pathways to insulin resistance are strong, but mechanisms by which they ultimately affect insulin action are still unclear. Many investigators ultimately conclude that altered insulin signaling must underlie the final manifestation of the disease.
There are, however, suggestions that dysregulated insulin signaling per se does not lie at the heart of insulin resistance. First, experimental inducers of insulin resistance tend to provoke their effects over a time course inconsistent with alterations in signaling. For example, TNF reduces insulin sensitivity in cultured adipocytes only after several days (Kang et al., 2015). The received wisdom on TNF action, however, is that it induces inhibitory phosphorylation events on key components of the insulin signaling cascade, effects that can be demonstrated in minutes to hours. Second, insulin resistance can develop in cells or animals without discernable or reproducible changes in insulin signaling, and conversely, animals with an engineered insulin signaling deficiency do not uniformly develop insulin resistance under normal conditions (Cleasby et al., 2007, Hoehn et al., 2008, Kang et al., 2015, Kim et al., 1999, Nadler et al., 2001).
There are compelling reasons to suspect that nuclear events, defined here as processes such as transcriptional and epigenetic regulation taking place in the nucleus, play a role in the development of insulin resistance. First, insulin sensitivity can be enhanced with drugs that primarily act through transcription factor targets, such as thiazolidinediones (TZDs), an activator of PPARγ (Ahmadian et al., 2013, Soccio et al., 2014) and glucocorticoids, which activate the glucocorticoid receptor. Second, drugs that affect chromatin remodeling, such as certain HDAC inhibitors, are known to affect insulin sensitivity in cells, animal models, and human subjects (Masuccio et al., 2010). Third, mice with genetic alterations in chromatin modifying enzymes, such as Jhdm2a and Ehmt1, develop obesity and insulin resistance (Inagaki et al., 2009, Ohno et al., 2013, Tateishi et al., 2009). Finally, and perhaps most compelling, there is a very large body of literature that indicates that the risk of developing insulin resistance in later life is strongly affected by nutritional conditions experienced in utero. For example, pregnant rodents that undergo caloric restriction give birth to offspring that have a significantly greater chance of developing insulin resistance as adults (Rando and Simmons, 2015). The same phenomenon has been reported in human populations, as with offspring of Dutch women who were pregnant during the ‘hunger winter’ of 1944–45 (Kyle and Pichard, 2006). Such examples of ‘metabolic memory’ are predicted to have an epigenetic basis, and there are data that support this mechanism directly, such as altered histone modification at the Slc2a4 (Glut4) locus in the offspring of calorically-restricted rats (Raychaudhuri et al., 2008).
Systemic insulin sensitivity is determined by the interactions of several different tissues and cell types (Odegaard and Chawla, 2013, Tomas et al., 2004). Classically, most people think of insulin action at the liver, muscle and adipose tissue, as the organs most responsible for insulin-dependent glucose production and disposal. Recent data, however, suggest a significant role for other tissues as well, notably the brain, which does not take up glucose in response to insulin but which can regulate the actions of insulin in classical target tissues indirectly (Parlevliet et al., 2014). Cells of the immune system are another example; although not insulin sensitive themselves, their number and activation state has a major effect on insulin action locally (in adipose tissue, for example) and systemically (Mathis, 2013, Odegaard and Chawla, 2013). Thus, a transcription factor can cause insulin resistance indirectly via effects on macrophage polarization, for example (Eguchi et al., 2013).
Finally, we should keep in mind that several (non-mutually exclusive) nuclear events may be involved, including changes in transcription factor expression, binding, post-translational modifications, and protein-protein interactions with co-factors/chromatin modifiers. In all of these scenarios, the relevant output is altered target gene expression. These target genes could be obvious, such as those encoding classic insulin signaling proteins, but may also be obscure. For example, several genes were recently identified in adipocytes that are dysregulated in insulin resistant states and which cause insulin resistance when overexpressed; none of these genes participates in a pathway known to be involved in insulin action (Kang et al., 2015). In the next section, we highlight three separate transcription factors and co-factors with significant actions on insulin sensitivity, and we discuss possible mechanisms by which they exert these effects. Although we focus on a few notable examples, many others exist, including transcription factors, nuclear co-factors, and chromatin modifying enzymes; some of these are indicated in Table 1.
Among transcription factors, the nuclear receptor (NR) superfamily plays a special role in regulating insulin sensitivity, given that they are activated by small ligands that are often fatty acid derivatives or other nutritional by-products. NRs thus provide a direct link between environmental conditions and the genome. Accordingly, several NRs have been implicated in the regulation of insulin action, most notably PPARγ and the glucocorticoid receptor (GR). Other examples of NRs that have been implicated in insulin resistance include VDR, LXR, FXR, and LRH-1.
The nuclear receptor best associated with insulin action is PPARγ, most notably via its position as the target of TZDs, which are used clinically as insulin sensitizers. Although originally identified as a dominant regulator of adipogenesis, PPARγ is now known to be expressed in many different tissues and cell types (albeit to a lower degree than adipose tissue) and to be involved in processes as distinct as lipid accumulation, glucose homeostasis, skeletal homeostasis, and inflammation (Ahmadian et al., 2013, Soccio et al., 2014). This broad tissue distribution has provoked numerous studies attempting to determine which site of action is most relevant for the insulin-sensitizing properties of PPARγ (Fig. 1). Somewhat amazingly, specific deletion of PPARγ in many different cell types, including adipose tissue, muscle, macrophages, and brain alters glucose homeostasis, and reduces the full activity of TZDs. The preponderance of evidence, however, suggests that adipose tissue is the major site of action for the insulin-sensitizing actions of PPARγ (Chao et al., 2000, He et al., 2003, Sugii et al., 2009). Mice lacking PPARγ in liver respond to TZDs normally unless adipose tissue is also defective (Chao et al., 2000), and liver- and muscle-specific PPARγ knockout mice were shown to display insulin resistance, though to a much lesser degree than adipose-specific knockout animals, and show variable response to TZDs (Hevener et al., 2003, Norris et al., 2003). A role for PPARγ has been speculated to be important in pancreatic β-cells, but while TZDs enhance insulin secretion from isolated islets in a PPARγ-dependent manner, mice lacking islet PPARγ have intact glucose homeostasis and respond normally to TZDs (Rosen et al., 2003).
Immune cells are another likely site of PPARγ action. PPARγ is highly expressed in macrophages, where it promotes alternative M2 polarization of macrophages (Bouhlel et al., 2007, Odegaard et al., 2007). Macrophage-specific depletion of PPARγ conferred protection from diet-induced insulin resistance, though this is still improved by rosiglitazone (Hevener et al., 2007, Pascual et al., 2007). A more recent study has provided evidence that PPARγ in a subset of regulatory T (Treg) cells plays a critical insulin-sensitizing role, as animals with PPARγ deletion in this cell type show no response to pioglitazone (Cipolletta et al., 2012) in glucose metabolism and insulin sensitivity. Lastly, there are indications that the weight gain associated with PPARγ activation is mediated partially through actions in the brain (Lu et al., 2011, Ryan et al., 2011). However, rosiglitazone still improved whole-body (but not hepatic) insulin sensitivity in mice lacking PPARγ in the brain (Lu et al., 2011), indicating that CNS effects may account for some but not all the metabolic effects of TZDs.
Several mechanisms have been proposed to explain the basis of the PPARγ-dependent insulin-sensitizing effect. In adipose tissue, the lipogenic and anti-lipolytic actions of PPARγ exert a so-called ‘lipid steal’ effect, in which fatty acids and other lipids are sequestered safely in adipose tissue. This counters the damaging actions of potentially toxic lipids in other tissues, which has been postulated to account for insulin resistance (Ye et al., 2004). TZDs also exert a direct glucose-lowering effect associated with improved insulin signal transduction in muscle and adipose (Iwata et al., 2001, Jiang et al., 2002). Yet another mechanism, at least in rodents, involves the ability of TZDs to promote ‘browning’ of white fat (Ohno et al., 2012, Qiang et al., 2012, Vernochet et al., 2009), which leads to increased energy expenditure and improved whole body metabolism. PPARγ also promotes both the expression and secretion of adiponectin, a potent insulin-sensitizing hormone (Riera-Guardia and Rothenbacher, 2008). Finally, the adipogenic actions of PPARγ may also mediate some of the insulin-sensitizing effect; the reason for this may involve a variation of the ‘lipid steal’ hypothesis (with new adipocytes contributing to the safe storage of toxic lipid species) and may also reflect a healthier adipokine and anti-inflammatory profile in newly formed, smaller adipocytes than in large, hypoxic adipocytes.
PPARγ is a transcription factor, of course, so its proximal actions are believed to involve alterations in gene expression, both positive and negative (Fig. 2). Positive targets include adiponectin, as well as a variety of lipogenic enzymes, transporters, and signal transduction intermediates. Negatively regulated genes in adipose tissue with relevance to insulin sensitivity include RBP4, resistin, and a variety of cytokines (Ahmadian et al., 2013). To date, several studies have been performed to comprehensively identify PPARγ-dependent gene expression and binding events with and without TZD treatment (reviewed in (Lefterova et al., 2014). However, our knowledge of the core transcriptional and epigenetic events through which PPARγ regulates insulin sensitivity remains incomplete.
Despite the well-known anti-diabetic effects of TZDs, their clinical use has been restricted due to serious adverse effects such as weight gain, plasma volume expansion, bladder cancer, and increased risk of cardiac heart failure, although some of these fears may have been overblown (Soccio et al., 2014). Nonetheless, the negative reports of TZD adverse effects led to a pronounced chilling of the PPARγ targeting programs of most major pharmaceutical companies. It remains to be seen whether the more reassuring data that has emerged recently will reinvigorate these activities. One promising area for future drug discovery relates to altering post-translational modifications of PPARγ to help treat metabolic disease. For example, preventing PPARγ phosphorylation at S112 preserved insulin sensitivity in response to a high fat diet, associated with reduced adipose hypertrophy, increased adiponectin, and reduced free fatty acid levels (Rangwala et al., 2003). Recent studies have also demonstrated that Erk/CDK5-dependent PPARγ phosphorylation at S273 regulates a selective set of PPARγ targets that are highly relevant to metabolic regulation (Choi et al., 2010, Choi et al., 2011). Furthermore, PPARγ agonists and MEK inhibitors that block this phosphorylation event exert anti-diabetic effects without causing weight gain or hemodilution (Banks et al., 2015). Other potentially interesting PTMs of PPARγ that may be amenable to therapeutic targeting include SUMOylation at Lys365 and Lys107, which mediates transrepression of inflammatory response genes, especially in macrophages (Pascual et al., 2007). SUMOylation at Lys107 can also lead to degradation of PPARγ, and this can be blocked by FGF21; this may mediate the insulin-sensitizing actions of that hormone (Dutchak et al., 2012). Sirt1-mediated deacetylation of PPARγ (Qiang et al., 2012) also leads to beneficial metabolic effects in metabolic syndrome.
Another nuclear receptor transcription factor with pronounced effects on insulin action is the glucocorticoid receptor (GR, encoded by NR3C1), which mediates the metabolic effects of endogenous and synthetic glucocorticoids (Oakley and Cidlowski, 2011, Patel et al., 2014). Activation of hypothalamic-pituitary-adrenal (HPA) axis by stress, including fasting, induces glucocorticoid synthesis and secretion from the adrenal cortex (Patel et al., 2014). Glucocorticoids exert a wide range of actions on metabolic tissues, the net effect of which is to raise blood sugar as required by the brain. Prolonged GR activation is associated with metabolic dysregulation, including insulin resistance, as occurs in Cushing’s syndrome (Plotz et al., 1952) or after long-term exposure to pharmacological doses of glucocorticoids (Rizza et al., 1982).
Glucocorticoids bind to the ligand binding domain of the GR in the cytoplasm, where it is sequestered by binding to Hsp90, Hsp70, and other chaperone proteins, and cause it to translocate to the nucleus (Fig. 2). Once in the nucleus, GR can alter gene expression in one of two ways. First, it can bind directly to glucocorticoid response elements (GREs) in the enhancers and promoters of various genes as a homodimer. Alternatively, GR can bind to other transcription factors as part of a ‘tethered’ complex that does not contact DNA directly. Both mechanisms can be used to promote gene activation or repression (Patel et al., 2014). There has been a longstanding belief that the beneficial actions of glucocorticoids, such as suppression of inflammation, primarily utilize the tethering model, and that adverse consequences of steroid action, including insulin resistance, require homodimerization and direct DNA binding. Recent studies call this into question, as mice that contain a mutation in the homodimerization domain of GR that does not affect tethering, display reduced insulin sensitivity at baseline, and still develop full insulin resistance upon treatment with dexamethasone (Roohk et al., 2013). It is true, however, that GR does directly bind and regulate a number of genes involved in gluconeogenesis, lipogenesis, and insulin signal transduction (Patel et al., 2014, Wang et al., 2004). In addition to these genomic actions of GR, nongenomic effects have also been proposed to contribute to glucocorticoid-induced insulin resistance; dexamethasone can inhibit insulin signaling very rapidly in cultured adipocytes in a manner that is not reversed by inhibiting transcription, for example (Lowenberg et al., 2006).
Most cells express the GR, including adipose tissue, muscle, and liver. In liver, GR activation promotes hepatic glucose output, an effect considered to be due to direct regulation of key gluconeogenic genes like Pck1 and G6pc. Thus, mice with liver-specific GR knockout or knock-down have reduced hepatic glucose output (Opherk et al., 2004) (Fig. 1). It is worth noting, however, that GR may also control gluconeogenesis via extra-hepatic sites like the hypothalamus, as direct injection of dexamethasone into the arcuate nucleus causes hepatic insulin resistance (Yi et al., 2012). These actions are independent of, but enhanced by, the increased food intake and weight gain associated with chronic glucocorticoid administration (Castonguay, 1991).
In skeletal muscle, chronic GR activation contributes to insulin resistance by inhibiting protein synthesis and promoting proteolysis, thus releasing amino acids which are used as substrate for glucose production. Glucocorticoids also inhibit insulin-stimulated glucose uptake. The underlying mechanism for this is not well understood but a recent ChIP-Seq study has shown that GR directly targets genes involved in insulin signaling (Kuo et al., 2012). GR mRNA levels in the skeletal muscle of diabetic patients correlates with the degree of insulin resistance, and expression normalizes following the administration of insulin sensitizers (Bodine et al., 2001).
Although the precise role of adipose tissue GR in the glucose metabolism and insulin resistance remains to be addressed in vivo, many studies have shown that activation of GR suppresses insulin-stimulated glucose uptake assay in cultured adipocytes (Houstis et al., 2006, Kang et al., 2015). GR causes this, in part, by directly regulating the expression of several downstream effector genes (e.g. Vdr, Tmem176a, Serpina3n, Lcn2), which reduce insulin-stimulated glucose uptake in as yet unclear ways. Interestingly, adipocyte GR can be activated by TNF-α in a partially ligand-independent way, representing an extraordinary example of an anti-inflammatory transcription factor mediating the metabolic effect of a pro-inflammatory cytokine (Kang et al., 2015).
FOXO transcription factors are key regulators of metabolism and are canonical mediators of insulin-dependent changes in gene expression (reviewed in (Kitamura, 2013). There are four mammalian FoxO genes (FoxO1, FoxO3a, FoxO4, and FoxO6) which have overlapping functions (Arden, Oncogene 2009). Although no common polymorphisms in the FOXO1 locus have been directly linked to insulin resistance or diabetes by large scale genome-wide association studies, a functional network analysis of the genes nearest GWAS signals for glycemic traits implicate FOXO1 as a shared interactor with multiple GWAS candidate proteins (Morris et al., 2012).
In the liver, insulin promotes the accumulation of lipid and represses gluconeogenesis; the latter is the major determinant of systemic glucose homeostasis in the fasted state. Among the many factors that regulate gluconeogenesis, the transcription factor FOXO1 and the co-factor PGC-1α have been identified as having a major role (Fig. 1). Liver-specific triple knockout of FOXO1, FOXO3a, and FOXO4 show increased fasting hypoglycemia, increased glucose tolerance, and enhanced insulin sensitivity with decreased plasma insulin levels; comparisons with single knockouts suggest all three work synergistically to regulate hepatic insulin sensitivity (Haeusler et al., 2010). Whole body FOXO6 knockouts also show decreased hepatic glucose production and enhanced insulin sensitivity (Calabuig-Navarro et al., 2015).
FOXO1 is thought to primarily effect insulin sensitivity in adipose tissue via inhibition of adipocyte differentiation, though transgenic expression of a dominant-negative FOXO1 in mature adipocytes improves glucose and insulin tolerance and increases energy expenditure in mice on high-fat diet (Nakae et al., 2008). In the pancreas, FOXO1 is involved in β-cell dysfunction via multiple mechanisms including suppression of β-cell proliferation, mediating oxidative, ER, and hypoxic stress, and increasing apoptosis (reviewed in (Kitamura, 2013). FOXO1 has also been shown to mediate insulin-regulated activity of hypothalamic neurons, with constitutively active FOXO1 causing hyperphagia, increased body weight, and inhibition of leptin action. Mice with Agrp neuron-specific deletion of FOXO1 are lean with reduced food intake and suppressed hepatic glucose production (Ren et al., 2012).
FOXO factors generally, and FOXO1 in particular, are regulated by a variety of post-translational modifications, including phosphorylation and acetylation (reviewed in (Calnan and Brunet, 2008). In its nonphosphorylated state, FOXO1 localizes to the nucleus, where it binds its cognate motif and drives expression of key insulin-regulated genes, such as G6pc (encoding glucose 6-phosphatase) (Fig 3). Insulin causes Akt-mediated phosphorylation of FOXO1 at residues T24, S256, and S319, resulting in its exclusion from the nucleus, where it is subsequently ubiquitinated and degraded. FOXO proteins are also a key element of the oxidative stress response pathway, which causes insulin resistance in multiple tissues. Reactive oxygen species (ROS) induce JNK-mediated phosphorylation of FOXO1, driving increased transcriptional activation of target genes, which include antioxidant scavengers (Kawamori et al., 2006). Transcriptional profiling of cells expressing a non-phosphorylatable mutant FOXO1 display up-regulated ROS response genes (Greer et al., 2007), while the vast majority of other FOXO1 targets are unchanged. FOXO1 is also activated by AMP-activated Protein Kinase (AMPK) phosphorylation, and its activation inhibits gluconeogenesis in the liver and glucose oxidation in muscle. In response to oxidative stress, β-catenins also directly interact with FOXO1 to increase its transcriptional activity (Almeida et al., 2007, Essers et al., 2005). Such β-catenin/FOXO1 interactions in liver have been shown to contribute to regulation of gluconeogenic gene expression and glucose homeostasis (Ip et al., 2015).
FOXO1 activity is also modulated by acetylation, serving as a direct target of the histone acetyltransferases CBP and p300, and the histone deacetylases SIRT1 and SIRT2. The effect of acetylation on FOXO1 activity appears to be context-dependent, with data showing positive, negative, and dual effects on transactivation (reviewed in (Daitoku et al., 2011). FOXO1 acetylation status is determined by the relative balance of protein acetylases (CBP, p300, and PCAF) and deacetylases (Sir2/Sirt family). FOXO1 acetylation promotes phosphorylation by Akt and inhibits DNA binding (Matsuzaki et al., 2005). Meanwhile, SIRT1-mediated deacetylation of FOXO family proteins, as occurs in response to ROS, promotes nuclear translocation and gluconeogenic gene expression, even in the face of Akt activation (Frescas et al., 2005). FOXO factors can also be activated by O-GlcNAcylation following increased oxidative stress; this modification is increased in diabetic livers and is associated with increased FOXO-driven expression of gluconeogenic and ROS detoxifying genes (Housley et al., 2008).
In the liver, the co-activator PGC-1α is a key insulin-responsive transcriptional regulator of gluconeogenesis. PGC-1α is induced in liver on fasting, is elevated in models of insulin resistance, and can activate the entire transcriptional profile of gluconeogenesis (Herzig et al., 2001, Yoon et al., 2001). The induction of gluconeogenesis by PGC-1α is mediated via direct interaction with FOXO1; Akt-mediated phosphorylation and expulsion of FOXO1 from the nucleus disrupts the FOXO1–PGC-1α interaction, thereby suppressing gluconeogenesis (Puigserver et al., 2003).
Insulin sensitivity is a carefully regulated process that goes awry in many different pathophysiological states. Its role as a major driver of Type 2 diabetes means that we must understand its antecedents and its consequences, in order to develop better and more rational targets for therapeutic intervention (see Outstanding Questions). Numerous theories have emerged to explain how insulin resistance develops and progresses, most of which involve changes in insulin signal transduction, or other processes that center on the cytosol, or organelles like mitochondria and the endoplasmic reticulum. Here we have endeavored to draw attention to transcriptional and epigenetic events that play an equally important role. These pathways involve a wide variety of organs and cell types and numerous molecular actors, including transcription factors, co-factors, and chromatin-modifying enzymes. They utilize a widely varying repertoire of mechanisms that include direct effects on the expression of insulin signaling components, repression or enhancement of inflammation, effects on cellular differentiation, and actions on targets with as yet unspecified roles in insulin sensitivity. Some of these pathways, such as those involving nuclear hormone receptors or chromatin-modifying enzymes, are inherently ‘druggable’, while other factors may prove to be more resistant to pharmacological intervention. Regardless, full explication of nuclear mechanisms and their downstream effectors will broaden our understanding of insulin resistance and should enable the identification of novel drug targets.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.