Physiological Reviews

Positive and Negative Regulation of Insulin Signaling by Reactive Oxygen and Nitrogen Species

Nava Bashan, Julia Kovsan, Ilana Kachko, Hilla Ovadia, Assaf Rudich


Regulated production of reactive oxygen species (ROS)/reactive nitrogen species (RNS) adequately balanced by antioxidant systems is a prerequisite for the participation of these active substances in physiological processes, including insulin action. Yet, increasing evidence implicates ROS and RNS as negative regulators of insulin signaling, rendering them putative mediators in the development of insulin resistance, a common endocrine abnormality that accompanies obesity and is a risk factor of type 2 diabetes. This review deals with this dual, seemingly contradictory, function of ROS and RNS in regulating insulin action: the major processes for ROS and RNS generation and detoxification are presented, and a critical review of the evidence that they participate in the positive and negative regulation of insulin action is provided. The cellular and molecular mechanisms by which ROS and RNS are thought to participate in normal insulin action and in the induction of insulin resistance are then described. Finally, we explore the potential usefulness and the challenges in modulating the oxidant-antioxidant balance as a potentially promising, but currently disappointing, means of improving insulin action in insulin resistance-associated conditions, leading causes of human morbidity and mortality of our era.


A. What are ROS and RNS and How Do They Function in Biological Systems?

All biological systems functioning in aerobic conditions are exposed to oxidants, either generated intentionally or as byproducts. Multiple different species of oxidants are being generated, and as the first letter of the terms reactive oxygen species (ROS) and reactive nitrogen species (RNS) implies, they are generally reactive substances. In some ways, one can compare the function of oxidants in biological systems to that of acids, which we are more used to think about in quantitative terms. Reactive species (RS) are characterized by the following.

They are a very diverse group of substances: there are multiple types of ROS and RNS. The most widely known ROS are superoxide ion (O2•−), hydrogen peroxide (H2O2), and peroxyl radical (OH•), and the common RNS are nitric oxide (NO) and peroxynitrite (ONOO).

ROS and RNS chemically interact, as do acids: a very intricate chemistry of ROS and RNS generates multiple species from those originally generated, occasionally resulting in a “hybrid RS” (like the generation of peroxynitrite from the rapid chemical interaction between NO and O2•−). Furthermore, such interaction may convert a non-RS to a RS, generating chain reactions. In that regard, an “antioxidant” (or reductant) can be converted into an oxidant, reminiscent of an acid's conversion into its cognate base. Whereas pKa defines the thermodynamic balance of this conversion for a specific acid, the redox potential of a specific RS is more difficult to measure. Additionally, although acids react spontaneously, biological systems also include enzymes to accelerate some reactions (like carbonic anhydrase for carbonic acid). Similarly, enzymes are also involved in some reactions of RS, like the conversion of O2•− to H2O2 by superoxide dismutase (SOD). A detailed description of ROS and RNS chemistry and biochemistry is well beyond the scope of this review, and the interested reader is referred to other reviews on the topic (27, 155, 341, 345).

Different ROS and RNS vary in their chemical reactivity, i.e., their tendency to react chemically with neighboring molecules: in principle, the more reactive a RS is, the shorter its half-life, since it more rapidly interacts with other molecules. In a biological system, this of course results in different distances a particular RS could travel from its site of generation: a less reactive ROS like H2O2 has a half-life of seconds and can therefore theoretically travel relatively large distances, even crossing cells and tissues, whereas the diffusion distance of OH• is in the range of a small protein due to its high reactivity.

Chemical characteristics determine the biological milieu of activity: RS can be electrically charged or electroneutral, hydrophobic or hydrophilic, governing their ability to cross membranes and/or to partition between aqueous and lipophilic (membranes) environments.

Oxidants are balanced by reductants (antioxidants): for normal function (physiological conditions), homeostasis needs to be maintained, reminiscent of the balance between acids and bases in biological systems. In the case of acid-base balance, it is measured easily by the pH. For ROS and RNS, redox balance needs to be maintained, but is poorly defined and has no agreed-upon measures.

ROS and RNS have critical biological functions essential for normal physiology, just as acids do. Yet, overproduction, deficiency in the balancing forces [bases (in the case of acid-base), reductants (in the case of redox balance)], and/or generation of abnormal species (certain organic acids; specific ROS and RNS) result in impaired homeostasis and is associated with pathology. This point related to RS is further discussed in the following section.

B. ROS and RNS in Normal Physiology and in Pathophysiology

As outlined above, life in an aerobic environment is constantly accompanied by generation of ROS and RNS. In fact, multiple functions vital for normal physiology make use of regulated ROS/RNS generation and subsequent alterations in redox state, one example of which is the insulin signaling cascade. Yet, when normal regulation of ROS/RNS generating processes and/or mechanisms to shut-off ROS/RNS-initiated signals are disrupted, RS play an active role in pathophysiological processes, including in obesity and diabetes in which insulin signaling is impaired. A discussion of ROS/RNS generation and antioxidant mechanisms and the evidence for their dysregulation in obesity and diabetes are presented in sections iC and ii. Below is an overview of ROS/RNS participation in normal physiology and pathophysiology (Figs. 1 and 2).

FIG. 1.

Major triggers, generating mechanisms, types, and targets of reactive oxygen species (ROS) and reactive nitrogen species (RNS) in biological systems.

FIG. 2.

Examples for ROS and RNS participation in physiological and pathophysiological processes.

1. Triggers for ROS/RNS generation

Production of ROS and RNS can occur in response to diverse stimuli (Fig. 1A) . These include extracellular factors signaling through plasma membrane receptors, like hormones and growth factors, including platelet-derived growth factor (216, 413), epithelial growth factor (12), insulin (as detailed in sect. iii), proinflammatory cytokines like tumor necrosis factor (TNF)-α, (69), and physical-environmental factors (like ultraviolet irradiation) (167, 371). In addition, intracellular factors or processes that trigger ROS generation include nutrient metabolism (detailed in sect. iiA1b), endoplasmic reticulum stress (143, 166; detailed in sect. iiA1c), and the detoxification of various xenobiotics (135, 399).

2. Generation mechanisms

Generation mechanisms (Fig. 1B, and further detailed in sect. iiA) include designated systems for intentional ROS/RNS generation like the NADPH oxidase (22) and nitric oxide synthase (4, 316), mitochondrial electron transfer (474), as well as enzymatic systems in which ROS are generated as a byproduct. These include various oxidases (lipoxygenase, xanthine oxidase) and detoxification mechanisms like the P-450 system (471).

The biological outcome of ROS/RNS generation depends on multiple factors partly presented in section iA. Particularly important is which are the specific reactive oxygen and/or nitrogen species that are being generated, what is their cellular location and duration of production (temporal-spatial coordinates), how much is generated, and whether it overwhelms antioxidant defense systems. Together, the biological consequence of ROS and RNS generation is very much dependent on the “biological context,” representing a highly combinatorial signal with certain specificity (302).

3. “Cellular/molecular targets” of ROS/RNS generation

The “cellular/molecular targets” of ROS/RNS generation (Fig. 1C) in response to a specific stimulus are the consequence of the biological context in which they are generated and their specificity. Oxidants can react with multiple cellular components (proteins, lipids, nucleic acids), generating reversible or irreversible oxidative modifications. They also activate various signaling cascades, some of which are designated for sensing and responding to “stress,” like the “stress-activated protein kinases” of the mitogen-activated protein (MAP) kinase family, JNK, and p38 MAP kinase (109, 353). Such signaling pathways alter proteins' functions, half-life, and/or gene regulation. As a principle, RS involvement in normal physiological processes involves carefully regulated production in a tight spatial-temporal manner, leading to reversible oxidative modifications. Pathophysiological processes mediated by RS are more likely to involve irreversible modifications of cellular components, be it proteins, lipids, or DNA. Accumulation of irreversibly modified biomolecules is a reasonable definition of the ill-defined term oxidative stress (discussed further below, and presented in Table 1).

View this table:

Different approaches to assess “oxidative stress” in biological systems

4. Functional outcome of RS generation

Functional outcome of RS generation (Fig. 2) ranges from the physiological to the pathophysiological range. Vascular tone (149), cell adhesion (59, 416), immune responses (145), and growth factors and hormone action (96, 179) are examples of RS participation in normal physiology. Conversely, a causative role of RS has been implicated in ageing-related diseases (251), malignant transformation (405), atherosclerosis (268, 403), neurodegenerative diseases (236), obesity, and diabetes (detailed below).

The diversity of RS-mediated processes and their broad range of involvement in physiology and pathophysiology make the definition of “pathological RS-mediated processes” particularly challenging. To date, there is no agreed-upon direct measurement of oxidative stress in biological systems. Defining oxidative stress as the imbalance between pro- and anti-oxidant mechanisms has led to the use of indicators of each side of the balance (Table 1). However, this view is somewhat limited. First, an increase in antioxidants might be interpreted as “lower oxidative stress.” Yet, it may in fact indicate a biological adaptive response to an increase in pro-oxidant processes. This would imply that an elevation, rather than decrease, in antioxidant defense, indicates increased oxidative stress. Second, if an increase in pro-oxidant processes is fully balanced by an upregulation of antioxidant mechanism, a new steady-state is reached without adverse effects to the biological system. So, should oxidative stress be defined as an increase in the amount of oxidation products of biological macromolecules (lipids, proteins, and/or DNA, Table 1)? Finally, is an increase in “oxidative stress” the equivalent of “oxidative damage,” or does the latter also indicate a functional consequence to the oxidative insult? These different ways of defining oxidative stress, along with the different, mostly indirect ways of measuring it (237), provide the fundamental basis for confusion, controversies, and contradictory conclusions in the field (11). Keeping these limitations in mind, the following section summarizes current evidence that insulin resistance-related conditions, obesity and type 2 diabetes, are associated with increased oxidative stress.

C. Increased Oxidative Stress in Insulin Resistance States: Obesity and Diabetes

A wealth of evidence largely from the last 15 years suggests increased oxidative stress in obesity (10, 13, 72, 122, 228, 410) and diabetes (20, 109, 122, 132, 137, 343, 462). Decreased antioxidant capacity, increased production of ROS, and elevated oxidation products of lipids, DNA, and proteins have been reported in plasma, urine, and various tissues, suggestive of systemic and organ-specific oxidative stress. The more recent evidence for systemic oxidative stress includes detection of increased circulating and urinary levels of the lipid peroxidation product F2-isoprostane (8-epi-prostaglandin F) in both type I and type II diabetic patients (81, 82, 289), and in obesity (35, 83). Remarkably, this marker correlated with blood glucose levels and responded to antidiabetic intervention (82). In nearly 3,000 obese persons from the Framingham Heart Study, 8-isoprostane level strongly correlated with body mass index (228). Increase in plasma protein carbonylation, a group of different oxidative protein modifications characterized by various carbonyl moieties, was also reported in both type 1 and 2 diabetes (47, 426; reviewed in Ref. 78).

The evidence for oxidative stress in key target tissues for insulin action has been somewhat less consistent. Although this may largely reflect the inconsistencies (and limitations) of the various approaches to measure oxidative stress, it may point to true physiological differences between these tissues.

1. Adipose tissue

Supportive evidence for increased oxidative stress in adipose tissue in obesity (and diabetes) includes the documentation of elevated oxidation products of both lipids and proteins. Obese mice before developing diabetes exhibited increased H2O2 generation by adipose tissue, accompanied by decreased mRNA levels of SOD, catalase, and glutathione peroxidase (122). The apparent perturbation of the pro-oxidant/antioxidant balance was manifested by elevated levels of thiobarbituric acid reactive substances (TBARS)(122) or malondialdehyde (MDA)(126), crude measures of lipid peroxidation. Developing diabetes in these mice mildly exaggerated these alterations, which remained unobservable in liver, muscle, or aorta (122). In addition, total protein carbonylation was reported to be two- to threefold higher in adipose tissue of high-fat fed (HFF) mice (144). Concomitantly, mRNA levels of glutathione-S-transferase 4, a key enzyme responsible for reversing lipid peroxidation adducts (4-hydroxynonenal, 4-HNE) on proteins, was decreased three- to fourfold, providing a potential mechanism for increased protein carbonylation in obesity.

2. Liver

Increase in markers of lipid peroxidation has been reported in the livers of animal models of diabetes and obesity (91, 113, 284, 415). However, these were not matched by a decrease in the levels of dietary antioxidants like vitamin E, nor by a protective effect of antioxidant supplementation (113, 415). Obesity and the related insulin resistance are frequently associated with increased accumulation of lipids (triglycerides) in the liver. Hence, it remains unclear if the elevation in lipid peroxidation in liver really reflects increased oxidative stress, or simply the increase in available substrates for lipid peroxidation (415). Furthermore, a decrease in the ratio of reduced to oxidized GSH (GSH/GSSG) in the liver was reported in a model of type 1 (91), but not in type 2/obesity model (123). Despite this seemingly inconsistent evidence for oxidative stress in obesity and diabetes (possibly attributed to the different models and parameters used), oxidative stress is accepted as a major driving force in the pathophysiology of nonalcoholic steato-hepatitis (NASH, “fatty liver”) (reviewed in Refs. 112, 161). Fatty liver is closely associated with insulin resistance, yet oxidative stress may be particularly involved in progressive stages of NASH, like the conversion from steatosis to cirrhosis. It is therefore difficult to conclude if NASH-associated liver oxidative stress is representative of the liver in obesity and diabetes.

3. Muscle

Evidence for oxidative stress in muscle in human obesity and diabetes is also somewhat limited. A recent study reported no increase in 4-HNE in muscle of obese diabetics compared with obese nondiabetic persons (287). In experimental type 1 diabetes, decreased high- or low-molecular-weight thiols have been reported in skeletal or cardiac muscles, as well as decrease in antioxidant enzymes (91, 148, 231). Furthermore, increased S-nitrosylation of total muscle proteins as well as of key proteins in the insulin signaling cascade has been reported in nutritional and genetic obesity in rodents (49, 470). Despite these, in the KKAy mouse model of obesity and diabetes skeletal muscle H2O2 production, TBARS and the expression of antioxidant enzymes were unaltered, although they were affected in adipose tissue (122).

D. Key Message From Section I

RS, a diverse group of oxidant molecules, are involved in a broad range of biological processes ranging from physiology to pathophysiology. In the latter, an imbalance between pro- and anti-oxidative mechanisms occurs, resulting in “oxidative stress,” an ill-defined term likely representing an array of different conditions, underlying mechanisms, and accumulation of different oxidatively modified biomolecules. Consequently, there is currently no unifying measure to assess oxidative stress. Moreover, gaps exist in our understanding of organ-specific oxidative stress in conditions associated with impaired insulin action, like obesity and diabetes. These may reflect differences in models and parameters utilized to measure oxidative stress. Conversely, they may represent true organ-dependent differences, and/or the existence of different “types” of oxidative stress in each tissue. Regardless, the current common notion is that oxidative stress is an accompanying feature of the obese/diabetic state (109). The next section provides further detail on major oxidant-generating and antioxidant systems and their status in conditions of impaired insulin action.


A. Systems for Oxidants Generation

In the context of insulin resistance, excessive ROS/RNS generation occurs as part of several, partially interrelated, pathophysiological mechanisms, which include metabolic overload–the overabundance of glucose and free fatty acids (FFA), inflammation, endoplasmic reticulum (ER) stress and the unfolded protein response (UPR), and dysregulated hormonal and growth factors regulation (Fig. 3). These rely on the main cellular sources for ROS or RNS generation in physiological conditions, which include the NOX family of NADPH oxidases, mitochondrial respiration, the ER, and nitric oxide synthases (Fig. 4), each of which is discussed below.

FIG. 3.

Major sources of ROS and RNS in pahtophysiology related to insulin resistance and obesity. The main proposed cellular sources for increased ROS/RNS generation in obesity and diabetes are shown. The processes responsible include the following: i) High metabolic load: dysregulated carbohydrate and lipid metabolism in obesity and diabetes exposes cells to an overload of nutrients. Excessive mitochondrial oxidation may result in enhanced ROS generation. In addition, high levels of free fatty acids (FFA) can activate toll-like receptors, constituting an overlap between ROS generation in response to metabolic and inflammatory cues. Finally, ROS can affect mitochondrial function, potentially resulting in further ROS generation. ii) Inflammation: obesity has been demonstrated to be associated with a chronic inflammatory state characterized by both increased circulating levels of proinflammatory cytokines (TNF, IL-6) and macrophage infiltration of adipose tissue. Macrophages can be a direct source of ROS generation through their high-efficiency NADPH oxidase system (NOX2). Pro-inflammatory cytokines act via membrane receptors to activate NADPH oxidases (NOX4), to accelerate mitochondrial ROS generation, and to induce endoplasmic reticulum (ER) stress. iNOS is an NOS-generating system activated by inflammatory cues. iii) ER stress: has been initially proposed to occur in beta cells, but recently proposed to occur in adipose tissue in obesity. Like mitochondria, dysfunctional ER can be both a source and a consequence of increased ROS generation. iv) Endocrine dysregulation: impaired endocrine regulation can result in altered circulating levels of hormones (in type 2 diabetes, elevated circulating insulin levels are a common compensatory response to peripheral insulin resistance). Many hormones activate NADPH oxidases and rely on ROS generation as part of their signaling cascades. AGE, advanced glycation end products; G6PD, glucose-6-phosphate dehydrogenase; PKC, protein kinase C; TNF-R, TNF receptor; TLR, toll-like receptor; UPR, unfolded protein response.

FIG. 4.

Major ROS and RNS generation systems. Generation mechanisms of ROS or RNS are depicted in blue boxes, while reactive species are shown in stars. AGE's, advanced glycation end products; ER, endoplasmic reticulum; Mito, mitochondria; NO, nitric oxide; ONOO, peroxynitrite; NOS, nitric oxide synthase; O2•−, superoxide; OH•, hydroxyl radical; H2O2, hydrogen peroxide; XaOX, xanthine oxidase.

1. ROS generation


NADPH oxidases are membrane-bound enzymatic complexes comprising an electron transport chain that transfers electrons from the donor, cytosolic NADPH, to the acceptor, O2. Consequently, O2•− (and H2O2) are generated, initially extracellularly or within organelles (recently reviewed in Ref. 22). This highly specialized system for intentional ROS generation was thought for many years to be uniquely utilized by phagocytic cells for oxidative burst, a process required for pathogen killing. Yet, in recent years, homologs of the cytochrome subunit of phagocytic NADPH oxidase were discovered (NOX1-5 and DUOX1-2) (22). Phagocytes' capacity to generate ROS is indeed the highest due to the high level of expression and the specific kinetic properties of the phagocytic NADPH oxidase NOX2 versus other NOXs. Nevertheless, the discovery of multiple members of the NOX family made it clear that most cells have the capacity to generate ROS through an NADPH oxidase system. Such ROS generation is increasingly recognized to be required for diverse physiological functions, including the response to various hormones and growth factors. Insulin-stimulated ROS production relies on NOX4 in adipocytes (272) and possibly also on NOX2 expressed in skeletal and cardiac muscle (173, 209). NOX4 shares only a mild (39%) sequence identity to NOX2 and is distinct from it in several aspects worth mentioning here. NOX4 was proposed to be localized also within internal membranes (like the ER) and, hence, to generate ROS into the lumen of such organelles (279, 440). Superoxide may require a transporter to get into the cytosol, but H2O2, likely the consequence of O2•− dismutation within the lumen by SOD, can cross membranes directly. Second, different from NOX2 that depends on the phosphorylation and assembly of several cytosolic components, functional NOX4 is independent of those for activity (130, 279, 389). Nevertheless, a requirement for the membranal p22Phox has been proposed (227, 279), and for the small GTPase Rac demonstrated in cells endogenously expressing NOX4 (138, 195). Further discussion of the requirement of ROS production by NOX4 in normal insulin signaling is presented in section iii.

What is the evidence that NADPH oxidases contribute to enhanced ROS production in conditions related to insulin resistance? In adipose tissue of obese mice, NOX4 was increased, as were p22Phox, NOX2, and other components of its multiprotein complex (p67Phox, p47Phox, and p40Phox) (122). NOX2 in adipose tissue of obese mice may arise either from obesity-associated upregulation of NOX2 in the adipose cells, or from phagocytic cells infiltrating this tissue in obesity (454, 466). Regardless of the exact NOX, increased ROS production in 3T3-L1 adipocytes incubated with FFA (122) or in hyperglycemic conditions (463) was inhibited by chemical inhibitors thought to inhibit mainly NADPH oxidases. In HFF mice, increased ROS production by adipocytes was inhibited by both inhibitors of NADPH oxidases, as well as by inhibiting protein kinase C (PKC)-δ (420), indicating PKC-δ as an upstream regulator of NADPH-mediated adipocyte ROS production in vivo. Similarly, impaired cardiac and vascular function (including insulin-induced vasodilation) in diabetes, HFF, and in response to fatty acids or to TNF were all reported to be mediated by NADPH oxidase-dependent ROS overproduction in the endothelium (60, 224, 257, 335, 395). In humans, expression of p47phox in endothelial cells correlated with body mass index and was significantly elevated in obese and overweight compared with lean controls (390). Studies on human leukocyte function in diabetes are somewhat contradictory. Some studies report higher levels of nonstimulated or stimulated superoxide generation and/or p47phox in neutrophils' membranes from diabetics (92, 176, 221, 312), consistent with a primed state of the NADPH oxidase complex. Other reports suggest decreased neutrophil ROS generation particularly in the stimulated state (86, 197, 369, 425). This was proposed to underlie, along with deficiencies in other neutrophil functions [like phagocytic capacity (6)], the elevated risk for bacterial infections in some diabetic patients. These discrepancies either reflect different patient populations and disease severity, methodologies, and/or the possibility that in vivo basal (unstimulated) ROS generation is elevated, but activation of the NADPH oxidases in response to infectious stimuli is impaired.

The exact mechanism for increased NOX-mediated ROS production in adipocytes in obesity is unclear. With the use of other cell systems, it was demonstrated that hypoxia (412, 439), inflammatory cytokines like TNF (286) and angiotensin II (175, 461), and ER stress (327) all induce upregulation of NOX4 mRNA and protein levels. All of these conditions or factors have been implicated as causative factors in insulin resistance of adipose tissue and/or muscle (143, 184, 185, 193). Furthermore, peroxisome proliferation activator receptor (PPAR)-γ ligands used as insulin sensitizing agents were shown to downregulate protein and mRNA expression of NOX4 (and NOX2) in endothelial cells (192). In adipose tissue, obesity was shown to increase the expression of glucose-6-phosphate dehydrogenase (G6PD) (324). This rate-limiting enzyme of the hexose monophosphate shunt is a major source of NADPH. Thus both elevated NOX4 expression and increased production of its substrate NADPH potentially contribute to enhanced O2•− production in adipose tissue in obesity. Interestingly, vascular cells were shown to have decreased activity of G6PD in response to hyperglycemia (475), suggesting cell-type specific adaptations to metabolic stress.

Collectively, the NADPH oxidases are a major source of ROS production in normal physiology and of overproduction in pathophysiological settings related to insulin action.


The mitochondrial electron transport chain is a major source of ROS production. Its primary function is in respiration, delivering electrons from NADH and FADH2 to molecular oxygen while generating an electrochemical gradient across the inner mitochondrial membrane. Yet, even under normal physiological conditions, it is estimated that incomplete electron transfer to oxygen resulting in ROS production (mainly O2•−) occurs with at least 0.2–2% of O2 molecules, generating a large oxidative burden (51, 398). In this respect, mitochondrial ROS production is a byproduct of aerobic ATP production. As such, it has been implicated in various pathophysiological conditions, including degenerative disease, ischemia-reperfusion injury, and aging (8, 267, 386, 404). Yet, it is increasingly recognized that mitochondrial ROS generation is regulated and may be important for various cellular functions, suggesting that its view exclusively as a toxic byproduct is an oversimplification.

Electrons can leak from the electron transfer chain generating O2•− mainly from complexes I and III, as a result of partial reduction of molecular oxygen. Superoxide ion is released into the mitochondrial matrix (by complexes I and III-Qi), or outwards to the intermembrane space (by complex III-Qo) (404, 436, 474) (Fig. 5). Given their negative charge, O2•− formed in the inner matrix cannot readily diffuse out of the mitochondrial matrix, though a passage through the voltage-dependent mitochondrial anion channel of the mitochondrial permeability transition pore has been described (156). More commonly, O2•− in the matrix is rapidly dismutated to the readily permeable H2O2 by Mn-SOD (SOD2), a process that can also occur in the intermembrane space or the cytosol by CuZn-SOD (SOD1) (404). Final detoxification of mitochondrial superoxides can occur by converting H2O2 to H2O by catalase (at high H2O2 concentrations) or glutathione peroxidase (at lower concentrations), as per the relative Km of the two enzymes (see sect. iiB). These systems are essential for limiting oxidative damage of mitochondrial structures, particularly in the inner mitochondrial matrix, as exemplified by the lethal phenotype of SOD2 knockout mice (441). Furthermore, even in the presence of SOD, O2•− may react with NO in a chemical reaction far faster than the dismutation to H2O2, generating peroxynitrite. Likewise, H2O2, through interaction with transition metals, may form OH• (the Haber-Weiss or Fenton reactions, Fig. 4). Both OH• and peroxynitrite are powerful oxidants that propagate oxidative chain of reactions. Hence, the biological outcome of mitochondrial ROS production and its potential involvement in physiological versus pathological processes depends on the extent and site of superoxide generation, other RS formed simultaneously, and the balance with detoxifying mechanisms. The reader is referred to several excellent reviews on ROS and RNS (bio)chemistry in the mitochondria (404, 436, 474).

FIG. 5.

ROS generation by the ER and its interrelations to mitochondrial ROS generation. ROS can be generated in the ER as part of oxidative protein folding. To generate correctly folded protein with disulfide bonds formed between the correct cysteines (in green), electrons are donated to oxidoreductases (PDI, Ero1) and ultimately to molecular oxygen, yielding H2O or ROS. When incorrect disulfide bonds form (in orange), they need to be reduced by GSH, resulting in a further decrease of GSH/GSSG ratio, altering the redox state within the ER. Alternatively, misfolded proteins can be directed to degradation through ER-associated degradation machinery (ERAD). Accumulation of misfolded proteins in the ER initiates the unfolded protein response (UPR), part of which is programmed to enhance antioxidant defense. Calcium ions released from the ER can augment mitochondrial ROS generation. This occurs through the elevation in electron donors to the electron transport (respiratory) chain through the stimulation of the tricarboxylic acid cycle. In addition, calcium ions increase cytochrome c release, impairing electron transfer and increasing ROS generation. This in turn can impair respiratory complexes, further augmenting ROS generation. ATF6, activating transcription factor 6; ATF4, activating transcription factor 4; eIF2α, eukaryotic translational initiation factor 2α; ERAD, ER-associated degradation; IRE, inositol-requiring enzyme-1; Nrf2, nuclear respiratory factor 2; ox/red-Ero1, ER oxidoreductin 1; ox/red-PDI, oxidized/reduced protein disulfide isomerase; PERK, PKR-like eukaryotic initiation factor 2α kinase; XBP1, X-box protein 1.

Increased mitochondrial ROS generation has been strongly implicated as a mediator between hyperglycemia and its pathological consequences in the vasculature (306), kidney (233), neurons (445, 446), retina (260), and pancreatic beta cells (364). As presented above, superoxide is generated in the mitochondria from complexes of the respiratory chain in their reduced (electron-bound) state. Respiratory chain complexes remain electron-bound when they cannot transfer electrons downstream. Most frequently, such conditions arise when the proton gradient between the mitochondrial matrix and the intermembrane space is high (high membrane potential, Δψ), preventing further outward pumping of H+. In such state, chemical uncouplers or uncoupling proteins (UCPs) could relieve ROS production, as they allow proton reentry into the mitochondrial matrix and electron transfer to oxygen. Indeed, this scenario has been suggested in response to high glucose. The increased glucose flux through glycolysis and the tricarboxylic acid (TCA) cycle generates excess electron donors (NADH), initially resulting in high Δψ and increased mitochondrial ROS production. In principle, such a mechanism should also operate in response to fatty acids (an additional source of NADH), as was shown in endothelial cells (306). Increasing antioxidant capacity or decreasing mitochondrial superoxide production with uncouplers prevented high-glucose-induced activation of PKCs, NFκB, generation of advanced glycation end products (AGEs) and sorbitol accumulation in endothelial cells (42, 306), all pathogenic mechanisms for diabetic vascular complications (reviewed in Ref. 348).

In adipose tissue and in skeletal muscle, the primary targets for the metabolic actions of insulin, evidence supporting these mechanisms is scarce. In 3T3-L1 adipocytes, high glucose was shown to induce mitochondrial ROS, which mediated insulin resistance and activation of inflammatory pathways (261). This scenario is essentially the one proposed for endothelial cells presented above (306). In contrast, in response to FFA, ROS production in 3T3-L1 adipocytes (122), or in isolated adipocytes from HFF mice (420), was proposed to be mediated through an NADPH oxidase, not mitochondria, based on chemical inhibitors. However, what exactly is the mechanism for altered mitochondrial function in adipocytes and skeletal muscle?

In both tissues, evidence exists for decreased respiratory capacity associated with insulin resistance and/or diabetes (287, 334, 460). Theoretically, a decrease in mitochondrial respiration could represent decreased number (mass) of mitochondria, impaired function of the respiratory chain, and/or a decrease in mitochondrial uncoupling (decrease in UCPs). The latter two can be associated with elevated ROS production due to higher likelihood of respiratory complexes remaining in an electron-bound state. A dysfunctional downstream complex creates a “bottle neck” for electron transfer, thus retaining upstream complexes in a reduced state. Such dysfunction of a respiratory complex could be genetically determined, and/or could represent a consequence of oxidative damage to the mitochondria, suggesting a pathological futile cycle. Decreased UCPs expression/function is associated with high Δψ, which as explained above, is associated with enhanced ROS generation. These different mechanistic possibilities are difficult to differentiate in vivo. In adipocytes, decreased mitochondrial gene expression attributed to reduced mitochondrial mass was documented with the initiation of obesity in the leptin-deficient ob/ob mouse (460). Remarkably, this could be prevented by a thiazolidinedione (TZD) PPAR-γ agonists, potentially by activating transcription factors involved in mitochondrial biogenesis, like PGC1 (460). Intriguingly, PGC1 can enhance mitochondrial antioxidant defense (upregulating Mn-SOD), possibly contributing to the capacity of TZDs to prevent ROS production and protein carbonylation in adipocytes exposed to dexamethasone or TNF (188). Counterintuitively, UCP2 expression was shown to be markedly increased in white adipose tissue of mouse obese models compared with lean controls (346, 414). Thus, in adipose tissue, only minimal evidence exists for enhanced mitochondrial ROS production and is associated with decreased mitochondrial mass.

In skeletal muscle, evidence for mitochondrial ROS overproduction in both human and rodent models is scarce, and in some studies was not documented even when such evidence was documented in adipose tissue (122, 287). Yet, in skeletal muscle of insulin-resistant or diabetic people, it appears that decreased respiratory function is a consequence of reduced mitochondrial mass and/or increased type II (glycolytic) muscle fibers (287, 370), and could underlie the development of insulin resistance without enhanced ROS production (334). Whether skeletal muscle UCP expression is altered in diabetes also remains controversial (17, 379). A recent study proposed that in muscle of diet-induced insulin resistant mice, mitochondrial dysfunction is in fact a consequence, not the cause, of increased ROS production (38).

Thus clear evidence linking mitochondrial ROS production and impaired insulin action in muscle and fat is limited. Nevertheless, a mitochondrial source of ROS/RNS production in these tissues, possibly from endothelial cells, is possible. Its relative contribution to overall oxidative burden and its physiological/ pathophysiological significance in vivo awaits further studies.


The ER is an additional organelle that recently attracted attention as a source of ROS generation (73, 275). The ER is where an estimated one third of translated proteins undergo full assembly and posttranslational protein modifications, which are required for their normal subcellular targeting and biological functions. Among other ER-based posttranslational modifications like glycosylation, the local environment within the ER supports the generation of disulfide bonds, which are critical for proteins' correct tertiary and quaternary structure stabilization. This is achieved by the presence of oxidoreductases and a relatively oxidizing environment (lower GSH-to-GSSG ratio) compared with the cytoplasm. It is estimated that oxidative protein folding may account for as much as 25% of total cellular ROS generation (433). When the requirement for ER-supported protein folding increases, whether as a consequence of increased demand for newly synthesized proteins or secondary to accumulation of misfolded proteins that initiate the unfolded protein response (UPR), ER-derived ROS generation may increase (275). Furthermore, interorganellar communication between mitochondria and ER may underlie increased mitochondrial ROS generation originating from ER-derived signals. Intriguingly, excessive activation of the UPR, or “ER stress,” has been implicated in the pathogenesis of insulin resistance associated with obesity (and with pancreatic beta cell failure related to type 2 diabetes). Gaps still exist in the basic understanding of the intricate interrelations between oxidative and ER stresses, but the two conditions seem to share common downstream signaling pathways that can impinge on insulin signaling. Thus, as several excellent recent reviews are available describing ER stress and UPR in molecular details (73, 143, 249, 275, 372), this section highlights the proposed mechanisms for ER-based ROS generation, and the possibility that within the context of ER stress, ROS generation contributes to the pathogenesis of insulin resistance.

Oxidative protein folding in the ER requires an electron transfer chain, which is less characterized than that in mitochondrial respiration, but shares some similarities with it (Fig. 5). Electrons freed from the two cysteines that form a disulfide bond are donated to ER oxidoreductases like protein disulfide isomerase (PDI), then to ER oxidoreductin 1 (Ero1), a flavin adenine dinucleotide (FAD)-containing protein, and finally, to molecular oxygen (433). As in mitochondrial ROS generation, which is augmented when metabolic load is increased, the ER electron transfer to oxygen is a likely source of ROS generation particularly when the demand for it is increased. In such conditions, a more severe redox imbalance may result from the excessive utilization of GSH to reduce inappropriate disulfide bonds formed between mispaired cysteine residues. Although the mechanisms regulating the redox balance within the ER are not well characterized, it is highly conceivable that the combination of increased ROS generation and GSH consumption by the ER would on its own induce oxidative stress. Yet, ER-derived signals can also augment mitochondrial ROS generation (Fig. 5). As a major Ca2+ storage site, perturbation of the ER, like in response to oxidative or ER stress, can readily result in elevated cytosolic and mitochondrial Ca2+ levels (139). This can stimulate mitochondrial ROS production by enhancing generation of reducing equivalents (mitochondrial Ca2+ would enhance the TCA cycle) while interfering with mitochondrial electron transfer (due to Ca2+-induced release of cytochrome c through the permeability transition pore, inhibition of complex IV secondary to Ca2+-stimulated generation of NO by nitric oxide synthase) (28, 204). Moreover, Ca2+-induced opening of the mitochondrial permeability transition pore could also result in depletion of intramitochondrial reducing equivalents. Thus the ER, particularly when the UPR is activated, may be a significant source of ROS generation and of oxidative stress.

Although the section above describes how ER stress/UPR is a likely source of oxidative stress, the two phenomena are more intricately interrelated. Oxidative stress can also be a cause, rather than a consequence, of ER stress, as it can be the primary reason for protein misfolding. In addition, the UPR constitutes a complex response (reviewed in Ref. 249) that can also induce antioxidative capacity as one of its arms (Fig. 5). Specifically, one of the primary effectors of the UPR is the activation of PKR-like ER kinase (PERK). The best characterized functional consequence of PERK activation is the inhibition of protein translation through Ser51 phosphorylation of the translation initiation factor eIF2α. This arm of the UPR serves to decrease the load of “client proteins” on the ER machinery. However, another substrate of PERK is Nrf2 (74), a member of ubiquitously expressed transcription factors that are best known to be activated in response to oxidative stress, and to upregulate detoxifying and antioxidant enzymes. This function is mediated by Nrf2 binding to the antioxidant response element (ARE) in the promoters of the genes encoding for these enzymes (361). Nrf2 target genes include heme oxygenase 1, glutathione-S-transferase, and the rate-limiting enzyme in glutathione biosynthesis γ-glutamylcysteine synthetase (457). Hence, in the context of the UPR, PERK-mediated Nrf2 activation serves to enhance the increased demand for reduced glutathione (GSH), which as mentioned above, frequently accompanies ER stress. An additional “antioxidant” mode of the UPR is the upregulation of the ER-associated degradation (ERAD) machinery through which misfolded proteins are targeted to degradation (208). This response, mediated by IRE through activation of Xbp1, decreases the load of proteins that consume GSH for reducing inappropriate disulfide bonds, hence preventing deterioration of the GSH/GSSG ratio and oxidative stress (166). Collectively, ER stress and oxidative stress are intimately interrelated. Each can cause the other, they share common signaling pathways (the activation of stress-responsive kinases like JNK and other effectors like CHOP), and also, they share the attempt to enhance antioxidant defense through Nrf2.

In the context of obesity and type 2 diabetes, ER stress has been mostly implicated as a potential mechanism for pancreatic beta-cell failure (recently reviewed in Ref. 372). The response to obesity-related insulin resistance is an attempt to increase circulating insulin levels to maintain glucose homeostasis, thereby increasing the synthetic load on the beta cells. This challenge on the ER folding machinery predisposes the beta cells to ER stress, and renders them particularly sensitive to defective UPR response. Indeed, PERK knockout mice and humans with mutant PERK (Wolcott-Rallison syndrome) develop diabetes due to beta-cell apoptosis (87, 160), demonstrating the requirement of normal UPR response for beta-cell survival and function. More recently, ER stress was proposed as a putative mechanism also for insulin resistance, particularly in adipose tissue and the liver (143, 300, 314, 315). Adipose tissue from both nutritional and genetic models of obesity exhibit markers of increased ER stress (314), including phosphorylation of eIF2α, PERK, and elevated expression of the ER chaperone GRP78. Mice with heterozygous deletion of XBP1, a genetic means of exaggerating ER stress, faired worse metabolically compared with wild-type controls when placed on HFF (314). Similarly, genetic loss of a protective protein induced by the UPR, ORP150, caused impaired glucose tolerance, even when this loss was restricted to the liver (300, 313). Furthermore, mice with a mutant eIF2α also have more severe metabolic response to HFF (373). Although this may be attributed also to beta-cell dysfunction, the possibility that defective eIF2α signaling impaired the antioxidant response to ER stress makes it particularly noteworthy herein. The cause-effect relationship between ER stress and insulin resistance was further offered by the demonstration that impaired insulin response can be relieved in mice by chemical chaperones that reduce ER stress (315), or by overexpression of ORP150.

The mechanism for the contribution of ER stress to insulin resistance, and particularly, the role ROS/RNS exert in this process, is still largely unknown. Strongest evidence exists for IRE-mediated phosphorylation of the stress-activated MAP kinase JNK as a molecular mechanism connecting ER stress and impaired insulin signaling (314). This kinase has been implicated in phosphorylating insulin receptor substrate protein 1 (IRS1) on serine residues that are inhibitory for insulin-initiated signal transduction (see also sect. ivA). Indeed, cells deficient in IRE exhibit lower activation of JNK and lower Ser phosphorylation of IRS1 in response to tunicamycin (314), a commonly used inducer of experimental ER stress. Yet, JNK can be activated by inflammatory cytokines and directly by oxidants, both of which can either be or not be related to ER stress, as described above for ER-related ROS generation. Therefore, it would seem fair to conclude that current understanding of the role of ROS generation in the context of ER stress as a mechanism for insulin resistance still requires further studies. These may be particularly challenging given the complex interrelations between oxidative and ER stresses and would still require proof for playing a significant role in the pathogenesis of human insulin resistance.

2. RNS generation: NO synthases

NO synthases (NOS) are the major source of NO production, an RNS that perhaps best reflects the complex and ambivalent roles of RS in physiology and pathophysiology. Its major physiological action is in mediating endothelium-derived vasodilatation. Moreover, given its high reactivity with O2•− (to form peroxynitrite, ONOO), it is even considered to exert “antioxidant effect” and anti-inflammatory effects by limiting O2•−-induced NFκB activation and the ensuing state of chronic inflammation (207). In contrast, NO is also an important mediator of inflammation, and protein nitrosative modifications can alter protein function, in some instances leading to irreversible protein damage. Moreover, peroxynitrite and other nitrogen-containing RS are potent oxidants and are major factors in mediating oxidative stress. As will be discussed in following sections, all these translate into a dual role in insulin action, with NO exerting a physiological as well as a pathophysiological role (i.e., in insulin resistance). How is this duality in function that is so typical of the entire oxidant field possible? In the case of NO biology, much of this can be attributed to the family of NOS, their expression profile, regulation, and enzyme kinetics properties, as detailed next.

NOSs catalyze the generation of NO from l-arginine, relying on the cofactors flavin mononucleotide (FMN), flavin adenine dinucleotide (FAD), heme, reduced nicotine adenine disphosphononucleotide (NADPH), and (6R)-5,6,7,8-tetrahydrobiopterin (BH4) (4). Compared with other RS, NO has a rather high diffusion distance in biological systems given its lipophylic nature, neutral charge, and relative low reactivity (316). Its major route of elimination is the conversion to nitrate through the reaction with oxyhemoglobin, resulting in a half-life of 1 s. The best-characterized physiological functions of NO, and those which participate in the physiological response to insulin, are attributed to the constitutively expressed endothelial NOS (eNOS, NOS3) and the neuronal NOS (nNOS, NOS1). nNOS is in fact expressed in all major insulin target tissues (except for adipose tissue), including liver (105) and skeletal muscle (220, 298). eNOS, though, may be found in some nonendothelial cell types, is probably mainly expressed in the endothelial cell component of various insulin target tissues including liver (383), skeletal (220), and cardiac muscles (15), and in adipose tissue (220). Its prime physiological function is the generation of NO to promote vasodilatation. In accordance, and as will be further detailed in section iii, the vasodilatatory action of insulin is mediated through the activation of NO production by eNOS. Both isoforms of NOS are regulated by intracellular Ca2+ levels and by calmodulin, generating NO in relatively short bouts to levels reaching nanomolar concentration range (316).

In contrast to the constitutive NOS isoforms and their role in normal physiology, the inducible NOS (iNOS, NOS2) is responsible for NO generation reaching micromolar range. iNOS is expressed in inflammatory cells and is robustly induced in most cell types including adipocytes and muscle cells by inflammatory signals (303, 304). This response is mediated through the activation of NFκB, and via various stress and inflammation-responsive kinases like p38 MAPK and ERK (244). Differently from eNOS and nNOS, its dependence on calmodulin is unaffected by Ca2+ (464), suggesting that iNOS is regulated differently from the two other NOSs by both gene expression and allosteric regulation. Furthermore, under inflammatory states, O2•− generation may also be enhanced. This likely reflects enhanced production through NADPH oxidases (NOXs; see sect. iiA1a), but may also be the consequence of increased mitochondrial ROS generation (see sect. iiA1b). Alternatively, O2•− generation may be the results of “uncoupled” NOS activity itself due to reduced availability of tetrahydrobiopterin (BH4), a NOS cofactor essential for the coupling of l-arginine to NADPH to generate NO (70, 442). Regardless of the exact O2•− source, increased levels of NO (generated by iNOS) vis-à -vis O2•− results in the generation of the highly reactive oxidant peroxynitrite (ONOO) through a rapid direct chemical reaction (Fig. 4). Peroxynitrite, either directly or through complex chemical reactions with CO2, protons, and other RS intermediate products, modifies thiol groups (to promote S-nitrosylation and disulfides), side chains of tyrosine or other amino acids (nitrotyrosine, histidinyl), and can interact with transition metals centers. The consequential alteration in proteins' function renders peroxynitrite a central mediator of NO-derived oxidative stress and damage. The reader is referred to an excellent recent review on the topic for further details (316).

Implicating NOSs as a significant source of RS contributing to the pathophysiology of diabetes and obesity is complex. First, because direct measurements of NO products (nitrate and nitrite) are not clearly increased in obesity and diabetes (270, 366), possibly since they represent a complex measure of NO generation and elimination. Nevertheless, increased NO2 and NO3 have been reported, though mainly in persons with late diabetes complications (58, 200, 270). Second, since NO has vital physiological functions (vasodilatation), decreased NO generation and/or bioavailability may in fact be the pathological link between NO and obesity/diabetes, rather than excessive production (146, 459). This would potentially contribute to endothelial dysfunction and hypertension, which frequently associate with both conditions. The main putative mechanisms for lower NO bioavailability in obesity and/or diabetes-related conditions include 1) decreased NOS expression (396); 2) increased levels of endogenous inhibitors (99, 277); 3) increased cogeneration of O2•−, either by uncoupled NOS activity in diabetes (222, 223, 388) or through non-NOS-dependent mechanisms for O2•− generation (104); and 4) a manifestation of insulin resistance itself, as insulin increases eNOS-mediated NO generation (232; further discussed in sect. iii).

Complementing the above is the evidence of dysregulated expression of NOSs in various tissues of human and model diabetes and obesity. Genetic and nutritional animal models of obesity demonstrate increased iNOS expression in skeletal and cardiac muscle, in adipose tissue (329, 476), and in ob/ob mice also in liver (120). Yet, in Zucker rats, total NOS activity and eNOS immunoreactivity in skeletal muscle were decreased compared with lean controls (396). In humans, eNOS and iNOS, but not nNOS, were shown to be more highly expressed at mRNA levels in adipose tissue and isolated adipocytes of obese persons (102, 103), although Western blot analysis confirmed increased protein levels only for eNOS (102). Consistently, in diabetic kidneys, eNOS was increased due to elevated expression in endothelium, whereas iNOS was increased due to higher presence of inflammatory cells (178).

Jointly, it appears that dysregulated NOS activity accompanies diabetes and obesity, contributing to altered ROS/RNS generation, though the proposed mechanisms involved are variable. The increased abundance of nitrosative protein modification in diabetes and obesity (discussed in sects. iC and ivD) underscores the notion that NOS contribute to dysregulated ROS/RNS generation in insulin resistance-associated disorders. Consistent with this view are studies with iNOS knockout mice. These mice are protected against insulin resistance, attributed to improved skeletal muscle, but not adipose tissue, insulin action, and glucose disposal (329). iNOS was also proposed to play a role in hepatic insulin resistance in ob/ob mice (120). Since iNOS is induced by inflammatory signals, frequently in conjunction with increased superoxide generation, it is conceivable that in these tissues peroxynitrite is elevated, contributing insulin resistance potentially through nitrative protein modification.

3. Other pathways and mechanisms contributing to ROS production in insulin resistance states

Several additional pathways have been proposed to contribute to ROS generation in insulin resistance-related conditions (Fig. 4). Of particular interest in the context of hyperglycemia is the generation of ROS related to auto-oxidation of glucose (190) and to the generation of nonenzymatically glycated proteins (Amadori products and AGEs)(189). Receptors for AGE (RAGE) have been identified and shown to mediate ROS generation through the interaction with extracellular AGE proteins (376, 468). Interestingly, it was suggested that ROS generation in response to activated RAGE is mediated via the NADPH oxidase complex (452), and inhibition of NADPH oxidases prevented AGE-mediated damage in diabetes nephropathy (427).

Xanthine oxidase (XO) catalyzes final steps in purine nucleotide degradation, with generation of superoxide. Hence, theoretically, elevated activity of this enzyme could be a source of increased ROS generation. Indeed, some evidence links XO as a source of increased ROS generation in diabetes and obesity. High-fat feeding was shown to increase muscle arterioles XO activity and protein level, and allopurinol (an inhibitor of XO) improved endothelial function (104). Increased XO activity was also reported in livers of a rodent model of type 1, and allopurinol treatment diminished the increase in lipid hydroperoxide levels in plasma, liver, and heart of these mice (91).

Finally, in addition to the relative contribution each mechanism for ROS/RNS generation exerts, in the living cell in which all or some mechanisms may operate simultaneously they frequently interact, raising various feed-forward loops culminating in further oxidant production. For example, S-nitrosylation of proteins of the mitochondrial electron transport chain (particularly complex I and IV) may exaggerate mitochondrial ROS generation (77). Evidence for mitochondrial NOS activity exists (85), which may switch from NO to superoxide generation in response to hyperglycemia (41). ROS and lipid peroxides may also stimulate mitochondrial ROS (56, 296), constituting an additional amplification loop of ROS generation.

B. The Antioxidant Systems

Oxidative stress is frequently defined as the consequence of an imbalance between pro-oxidant processes (oxidants generation, described in sect. iiA) and antioxidant defense mechanisms. The latter likely evolved to counterbalance and limit the availability and potential hazardous effects of excessive ROS or RNS generation, and is thus an important component of the redox balance. Yet, as mentioned above, when assessed in isolation, the antioxidant side of the redox balance somewhat indirectly reflects on the role of oxidants, the topic of this review. This is because low levels of antioxidants may be the primary cause of oxidative stress, or rather represent its consequence (reflecting increased consumption). Likewise, upregulation of antioxidant enzymes may reflect increased antioxidant capacity, but may in fact represent the biological response to increased RS generation. Nonetheless, antioxidant supplementation is a common potential therapeutic strategy applied in conditions associated with “increased oxidative stress.” The shortcomings inherent to this approach lead to frequently disappointing results in clinical trials related to insulin resistance, and will be described in more detail in section v. Here, for the sake of completeness, we briefly describe the antioxidant defense systems, with emphasis on its components currently available for intervention trials described later. The reader is referred to other reviews on the subject for more detail (64, 153, 154, 250, 382).

The antioxidant defense capacity largely depends on small molecules with antioxidant activity and on enzymes catalyzing oxidant-modifying reactions (Fig. 6). Since antioxidant activity is broadly defined as the capacity to antagonize oxidants, both “designated” antioxidants and other endogenous metabolites (like uric acid and bilirubin) contribute to the low-molecular-weight-derived antioxidant capacity. Hence, antioxidants are diverse compounds classically divided into preventive and chain-breaking antioxidants based on their mode of biochemical action (154). The preventive antioxidants prevent the initiation of radical chain reactions through direct scavenging of ROS, or through the binding of transition metals. The chain-breaking antioxidants are capable of interrupting the propagation of radical chain reactions by forming a stable product. This is based on the potential for conversion reaction between the reduced and oxidized forms. The hydrophilicity versus hydrophobicity of the compound determines the major milieu in which a particular antioxidant will be more efficient. A classical example for this concept is the ability of the lipid-soluble vitamin E to primarily prevent lipid peroxidation. Yet, the oxidation-reduction potential of each antioxidant determines its ability to recycle other antioxidants, thereby contributing to “total antioxidant capacity.” Described below are the major antioxidants and their state in conditions of insulin resistance.

FIG. 6.

Major ROS and RNS detoxifying systems. The major enzymes of ROS detoxification are shown in blue, and the major low-molecular-weight antioxidant molecules are listed according to their lipophilicity or hydrophilicity. GPx, glutathione peroxidase; GR, glutathione reductase; GSH, reduced glutathione; GSSG, oxidized glutathione; SOD, superoxide dismutase.

1. Dietary (exogenous) small molecule antioxidants

In mammals, dietary consumption of both the water-soluble vitamin C and the lipid-soluble vitamin E contributes to the antioxidant potential (328). α-Lipoic acid, a cofactor of the α-keto-acid dehydrogenases, is an important antioxidant that participates in redox cycling of vitamins C and E as well as glutathione by virtue of its hydrophilic and lipophilic properties and its redox potential (317319). The trace element selenium is also a diet-derived factor associated with antioxidant defense, as the major selenoproteins are key antioxidant enzymes (glutathione peroxidase, thioredoxin reductase), or exert antioxidant properties (selenoprotein P, W) (21, 44).

The circulating level of dietary antioxidants is the most frequently used measure of the antioxidant status in clinical trials, although how well it represents the concentration of these factors in tissues relevant for insulin action is questionable. Furthermore, using this approach has yielded inconsistent and occasionally conflicting results when comparing diabetic patients with controls (48, 443, 458; reviewed in Ref. 33). This possibly reflects differences in the populations studied (type of diabetes, severity and duration of the diseases, late diabetic complications, alterations in nutritional intake of vitamins), as well as methodology used for the various measurements. Intracellular levels have mostly been assessed in red blood cells, platelets, and leukocytes (33). These studies somewhat more consistently demonstrated decreased levels of dietary antioxidants, but again, it is unknown how these results reflect the intracellular concentrations in muscle, liver, fat, and the vasculature. Selenium levels have not been consistently shown to be decreased in diabetics and may even be somewhat elevated (31, 32).

An additional approach to look for an association between decreased circulating levels of dietary antioxidants (as a potential cause or reflection of increased oxidative stress) and impaired insulin action leading to type 2 diabetes has been using prospective-observational or intervention trials. Low plasma levels of vitamin E were suggested to be associated with increased risk of incident type 2 diabetes (282, 291, 344, 365). Inconsistent results were reported for vitamin C (114, 291). Intriguingly, vitamin E supplementation was not protective (263, 282). This could potentially suggest either that only dietary source of vitamin E is effective, or that a related food-derived factor other than vitamin E (114, 117) is in fact responsible for the protective effect seen in observational trials. Selenium supplementation has been recently shown to have no protective effect (and even potentially increase the risk) for type 2 diabetes (31, 406). These studies are further discussed in section v. Collectively, there is only circumstantial and indirect evidence linking dietary antioxidants with the regulation of insulin action in healthy humans.

2. Endogenous low-molecular-weight antioxidants: glutathione (GSH)

GSH is a tripeptide (γ-Glu-Cys-Gly) present in tissues in the millimolar range. It is a potent reductant by virtue of its thiol moiety, acting as the major intracellular water-soluble antioxidant (165). Through the action of enzymes that catalyze various oxidation-reduction processes, GSH participates in the detoxification of toxic peroxides (like H2O2 by the activity of glutathione peroxidases, Fig. 6), in the maintenance of protein SH groups (Fig. 5), and in conjugation with xenobiotics to enable their elimination (through the action of glutathione-S-transferase). After participation in redox reactions, GSH is regenerated from GSSH (its oxidized form) by the enzyme GSSG reductase, using reduced NADPH as a cofactor. It has been demonstrated in various systems that depletion of GSH renders cells or tissues more susceptible to oxidative injury while loading cells and particularly the mitochondria with GSH or the thiol antioxidant N-acetylcysteine can limit such damage (115, 297, 338, 347).

In diabetes, most human studies do demonstrate decreased blood (plasma, blood cells) GSH levels, and animal studies also reveal decreased levels of reduced glutathione in heart, liver, and muscle (reviewed in Ref. 351). Some studies suggest that decreased GSH levels precede the occurrence of diabetes (308), but GSH as a biomarker of diabetes risk has not been established by large-scale clinical trials. Studies in animal models demonstrate that N-acetylcysteine, a cell-permeable thiol antioxidant reminiscent of GSH, has been shown to protect against the development of diabetes (218). Yet, this effect likely represented the prevention of pancreatic beta-cell dysfunction rather than maintenance of insulin action. In contrast, GSH depletion in rats by pharmacological inhibition of γ-glutamylcysteine synthetase induced impaired glucose tolerance (229) and defective insulin action on its metabolic target tissues (311). This is further discussed in section iv.

3. Small protein antioxidants: thioredoxins and glutaredoxins

Thioredoxins (Trx1-2; cytosolic and mitochondrial, respectively) and glutaredoxins (Grx1,2,5; cytosolic and two mitochondrial, respectively) have an increasingly appreciated array of functions (182, 259). Best characterized is their role in preserving the reduced state of SH groups in proteins that undergo sulfhydryl oxidation to form intramolecular disulfide bonds, or undergo glutathionylation (reaction of protein-SH with GSH) or S-nitrosylation (reaction of protein-SH with NO). Grx can also reduce dehydroascorbate, thereby participating in small-molecule external antioxidant regeneration (455). This reducing capacity is facilitated by a typical dicysteine motif (Cys-X-X-Cys) in the active site of all the members of this family (take out Grx5, which has only one) and can be regulated by the oxidation-reduction of additional, nonactive site cysteines (259). The regeneration of the reduced form of Trxs is typically catalyzed by thioredoxin reductases (TrxR1-3, cytosolic, mitochondrial, and testis-specific, respectively), whereas Grxs are reduced mostly by GSH.

The importance of the Trx-Grx system as a major component of the antioxidant defense system, and its involvement in diverse pathologies, has been demonstrated using a variety of cellular and animal experimental models. Surprisingly, despite their involvement in multiple biological processes including signal transduction, metabolism, and gene expression, a direct connection with insulin action is not very strong. Worth mentioning are early studies that demonstrated the ability of Trx to reduce the disulfide bonds of the insulin molecule itself, thereby having the potential capacity to affect its bioactivity (183). A more recently discovered potential connection may be through the role of Trx and Grx in regulating the activity and expression of apoptosis signal-regulated kinase (ASK1) (363). This MAP kinase kinase kinase (MAP3K) is activated by oxidants through the oxidation of Trx and/or Grx and the subsequent dissociation of the ASK1-Trx complex. By this means, ASK1 mediates the oxidation-induced activation of the MAP kinase pathways JNK and p38MAPK, two MAP kinases proposed to be involved in the induction of insulin resistance, particularly of adipose tissue (419). Activation of ASK1 was indeed proposed to mediate TNF-induced insulin resistance (194), suggesting a putative link between Trx/Grx and insulin action.

4. Antioxidant enzymes

Maintaining endogenous antioxidants relies on enzymes which catalyze their synthesis and regeneration (from the oxidized to the reduced form). For example, these include γ-glutamylcysteine synthetase and glutathione reductase, respectively, in the case of glutathione, and the protein synthesis machinery and thioredoxin reductases for thioredoxin. Yet, the enzymes classically referred to as “antioxidant enzymes” are those that directly act on oxidants, converting them into complete nonoxidant moieties (like catalase), or at least to less reactive species (like SOD, Fig. 6) (451). Catalase converts H2O2, mostly when in high concentrations based on its high Km, to H2O2, whereas the three isoforms of SOD (cytosolic, mitochondrial, and extracellular: SOD1-3, respectively) all convert O2•− to H2O2 and molecular oxygen. The selenoproteins glutathione peroxidases (GPx) also “neutralize” H2O2 using GSH, and GPx-1 may be an important front-line defense against the effects of this oxidant in the extracellular environment.

Data implicating changes in antioxidant enzymes in diabetes are conflicting (33, 351). Moreover, it is difficult to interpret, since, as mentioned above, both increase and decrease in their expression or activity can be rationalized to indicate oxidative stress. In long-term diabetes, the activity and expression of catalase, GSH reductase, GSH peroxidase, and SOD is decreased in various tissues (2, 140, 310, 381, 392), but whether this indicates a primary or secondary phenomenon is largely unclear. Given these limitations, and the fact that different from dietary and low-molecular antioxidants, manipulating antioxidant levels is not a readily available tool to modulate the oxidant-antioxidant state, further discussion of antioxidant enzymes is beyond the scope of the present review.

C. Key Message From Section II

Complex pro-oxidant and antioxidant systems are operational in biological systems. These include designated systems for RS generation and detoxification (enzymes, external and endogenous small molecules), as well as processes and molecules whose contributions to the oxidant-antioxidant balance are indirect or are not their primary function. A delicate and highly dynamic balance governs the availability of ROS and RNS in normal physiology, a balance which is likely perturbed at multiple levels in diabetes and obesity. This underlies the common notion that these conditions are associated with “increased oxidative stress.” Yet, when specific components of the oxidant-antioxidant balance are assessed, the emerging result is highly complex and variable, with differences between various models, tissues, cell types, and severity of the pathophysiological states. It seems fair to speculate that there are numerous “types of oxidative stress,” each representing different combinatorial perturbation of various component(s) of the complex oxidant-antioxidant balance. Even if such perturbation is accepted to occur, assessing whether it plays a functional role in physiology and/or pathophysiology, rather than representing their consequence, is a separate question. Section iii describes how ROS and RNS participate in normal insulin action. Sections iv and v deal with mechanisms and evidence for perturbed oxidant-antioxidant balance playing a functional role in the pathophysiology of insulin resistance-related conditions. The major concepts used to support a role for ROS/RNS participation as positive versus negative regulators of insulin action, assigning them a physiological versus pathophysiological roles, respectively, are summarized in Figure 7.

FIG. 7.

Major lines of evidence for ROS/RNS as positive and negative regulators of insulin signaling.


A. Oxidants Have Insulin-Mimicking Effects

One of the earliest connections between ROS and insulin has been the documentation that high (millimolar) concentrations of H2O2 activate insulin signaling and/or induce typical metabolic actions of insulin (75, 164, 169, 238, 280). The activation of the insulin signaling cascade was shown to arise from the insulin-independent tyrosine phosphorylation of the insulin receptor β-chain (164), suggesting that H2O2employs the same pathway as insulin. Propagation of the signal downstream yielded typical metabolic actions of insulin. In adipocytes and muscle, H2O2 increased glucose uptake (174, 238, 431), and in adipocytes also stimulated GLUT4 translocation (273) and lipid synthesis (280).

The initial steps in insulin signal transduction include phosphorylation on tyrosine residues of the insulin receptor and its direct substrates. The phosphorylation state of a protein reflects the balance between kinases and phosphatases, such that increased phosphorylation can result from an increase in the activity of a kinase, and/or inhibition of a respective phosphatase. Thus, theoretically, inhibiting phosphatase activity may have a similar effect as enhancing activity of a related kinase. Millimolar concentrations of H2O2 modulate the tyrosine kinase-phosphatase balance. The insulin receptor kinase activity itself is susceptible for oxidative regulation, as binding of ATP, which is required for the autophosphorylation process of the receptor, is modulated by hydrogen peroxide (374, 377). Yet, the main underlying cellular mechanism for the insulinomimetic effect of millimolar H2O2 concentrations is the inhibition of the catalytic activity of various protein and lipid phosphatases (Fig. 8A). For the insulin receptor and for the insulin receptor substrate 1 (IRS1) protein, the intracellular single-domain enzyme PTP1B, a member of an extensive family of protein tyrosine phosphatases (PTPs), is the major tyrosine phosphatase. Hence, it is a physiological negative regulator of the insulin signaling cascade, a notion supported by the finding that mice with PTP1B gene deletion exhibit increased insulin sensitivity (88, 101, 234). Numerous studies have characterized the biochemical regulation of PTPs by oxidation of their catalytic cysteine thiol moiety (90, 252, 274). The activity of PTP1B is dependent on the oxidation state of its Cys215 residue: reduced Cys215 is required for catalytic activity via the formation of a phosphoenzyme intermediate, whereas oxidized Cys215 renders the enzyme inactive and has been recognized as a critically important mode of PTP regulation in vivo (19, 283).

FIG. 8.

Physiological role of ROS in insulin signaling cascades. A: ROS are second messengers in insulin singaling, as they are reversible inhibitors of protein and lipid phosphatases (in blue) that function as negative regulators in the insulin signaling cascades. This can affect both the “metabolic arm” and the “mitogenic arm” of the insulin signaling cascade (in green or gray symbols, respectively). B: ROS generation in response to insulin via Nox4. In addition to generation of extracellular superoxide by Nox complex at the plasma membrane, Nox4 may be also localized on internal membrane pools, generating superoxide intraluminally. To reach the cytoplasm, superoxide anion would require a transporter, whereas H2O2, the product of superoxide dismutation by SOD, can cross the membrane. IR, insulin receptor; IRS, insulin receptor substrates; PI3K, phosphatidylinositol 3-kinase; PIP3, phosphatidylinositol 3,4,5-trisphosphate; PIP2, phosphatidylinositol 4,5-bisphosphate; PDK, phosphoinositide-dependent kinase; PKB, protein kinase B (also Akt); PTEN, phosphatase and tensin homolog; PP2A, protein phosphatase 2A; MAPKAP-1, mitogen-activated protein kinase-associated protein 1; PTP1B, protein tyrosine phosphatase 1B; ERK, extracellular signal-regulated kinase; NOX4, NADPH oxidase 4.

Downstream of the tyrosine phosphorylation of IR and IRSs, additional phosphorylation steps of proteins (on Ser/Thr residues) and of phospholipids participate in propagating the insulin signaling cascade. Reminiscent of the effects of oxidants on PTP1B, a number of additional cellular enzymes were suggested as potential targets for oxidative inhibition by oxidants. A serine-threonine phosphatase, protein phosphatases 2A (PP2A), is a major negative regulator of PKB (by dephosphorylating its Ser473), p70 S6-kinase, and ERK (438). PP2A has a redox-sensitive cysteine residue that is potentially susceptible to inhibition by H2O2 (150). Likewise, the dual-specificity phosphatase MAP kinase phosphatase-1, which attenuates insulin-stimulated MAP kinase activation, also depends on a reduced thiol for activity (247). The lipid phosphatase PTEN is another important negative regulator of insulin signaling, as it catalyzes the dephosphorylation of 3′-phosphoinositides, lipid second messengers essential for insulin signal transduction (299). Similar to the aforementioned protein phosphatases, it is also inactivated by oxidation of an essential cysteine residue in its active site (253, 254).

Collectively, external oxidants applied onto cells may activate the insulin signaling pathways at several points through the inactivation of protein and lipid phosphatases, all of which are negative regulators and off-mechanisms of insulin signaling.

B. Insulin Induces ROS and RNS Production

Already in the late 1970s, insulin itself was shown to elicit the generation of H2O2 in adipocytes (281). An early characterization of the enzymology of this process revealed that insulin activated a plasma membrane-associated enzyme system with the properties of a Nox, resulting in the downstream production of H2O2 (242, 243, 293). Differentiated 3T3-L1 adipocytes that were loaded with a redox indicator dye generated cellular oxidants (likely H2O2) within 1 min of stimulation by 0.1–10 nM insulin (273). The oxidant signal peaked at 5 min and began to dissipate by 10 min. To verify the participation of Nox4 in insulin-induced generation of H2O2, its presence was first demonstrated at both the mRNA and protein levels in mature primary adipocytes and differentiated 3T3-L1 cells (272). Furthermore, adenovirus-mediated expression of Nox4 deletion constructs acting in a dominant-negative fashion attenuated insulin-stimulated generation of oxidants. Similar results were obtained using siRNA gene silencing approach (272). Complementarily, when overexpressed, enhanced Nox4-mediated oxidant generation was associated with enhanced insulin receptor autophosphorylation and with the inhibition of PTP1B catalytic activity. Collectively, these findings suggest that Nox4 provides a link between the insulin receptor and ROS generation by a mechanism that enhances insulin signal transduction, at least in part via the oxidative inhibition of cellular PTPs, specifically PTP1B (Fig. 8A).

How does insulin stimulate Nox4-mediated ROS generation? One intriguing observation has been that insulin-mediated stimulation of NADPH oxidase activity may not require the kinase activity of the insulin receptor (241), yet this possibility was not reported by other studies. Otherwise, the regulation of Nox4 activity by insulin remains largely unknown. It was suggested that Rac and the small G protein Gαi2 may play a role in the regulation of Nox4 in insulin-sensitive cells (Fig. 8B) (134, 243), possibly mediating the critical role proposed for these G proteins in insulin action (55, 210, 292).

In addition to ROS, RNS production is also stimulated in response to insulin. In vascular endothelium, insulin-stimulated NO production is mediated by increasing expression and/or activation of eNOS (see detailed description of NOS in sect. iiA2). This RNS-generating effect of insulin underlies the vasodilatatory response to insulin. Insulin-stimulated activation of eNOS is mediated via the PI 3-kinase branch of the signaling cascade (294). PKB phosphorylates eNOS on Ser1179 (or Ser1177 in humans), leading to enhanced production of NO, an effect that was diminished in cells expressing a mutant eNOS with a disrupted PKB phosphorylation site (93, 290). Consistently, overexpression of a dominant inhibitory mutant of PKB in human umbilical vein endothelial cells (HUVEC) nearly completely inhibited production of NO in response to insulin (472). Knocking out the predominant PKB isoform in the vasculature, PKBα (Akt1) resulted in significantly lower levels of active eNOS in endothelial cells (54). On the basis of these data, it is likely that PKBα (Akt1) isoform mediates insulin-induced activation of eNOS. However, although necessary, PKB-dependent phosphorylation of eNOS is not sufficient for full eNOS activation, which can be achieved only when eNOS binds calmodulin (294). Intriguingly, insulin stimulates the protein complex formation consisting of eNOS, calmodulin and the heat shock protein (Hsp)90 (417, 473). This complements the activation step facilitated by PKB-mediated phosphorylation of eNOS on Ser1177 (290).

Similar to its effect on the endothelium, insulin stimulates NO production in vascular smooth muscle cells (VSMCs) in a PI 3-kinase-dependent manner (23, 432). However, since in VSMC iNOS and nNOS are expressed in addition to eNOS (39, 168), the relative contribution of each NOS isoform to insulin-induced NO production is unknown (23, 43). In addition to the short-term effect of insulin on NO production, long-term (>4 h) insulin stimulation also increases eNOS expression via mammalian target of rapamycin (mTOR)-dependent activation of AP-1 and Sp-1 transcription factors (141, 245).

C. RNS and ROS Are Second Messengers in Insulin Signal Transduction

Are insulin-stimulated RS required for insulin action, or in other words, do RS act as secondary messengers in the insulin signaling cascade? Theoretically, RS are ideal second messengers: they are short-lived, hence having an “intrinsic shut-off mechanism” and functionality that is limited in time and space, they can initiate chain reactions, and they can be generated upon request through regulated mechanisms, as detailed above for insulin. Different RS may have different properties and chemical reactivity, providing an additional aspect of specificity which characterizes the potential second messenger's function of ROS and RNS (302). To prove a second messenger function for RS, it would be required to show that in response to insulin-stimulated RS generation rises prior to measured biological effects, that RS can potentiate insulin action when administered at submaximal concentrations, and even better, that insulin action is diminished if ROS or RNS generation is inhibited.

Studies in adipocytes and in muscle cells have shown the ability of ROS to enhance insulin signal. Physiologically relevant concentrations (<0.1 mM) of H2O2 enhance the response to 100 nM insulin, indicating a coregulatory function of ROS in insulin receptor activation (375). In addition, various antioxidants, such as NAC, were found to inhibit the induction of insulin receptor kinase activity, potentially by antagonizing the effects of insulin-induced oxidants (374). Blocking the generation of cellular H2O2 with catalase or diphenyleneiodonium (DPI), a commonly used inhibitor of cellular Nox activity, reduced insulin-stimulated autophosphorylation of the insulin receptor and tyrosine phosphorylation of IRS proteins by up to 48% (273, 274). Moreover, expression of deletion constructs of Nox4 led to an inhibition of insulin receptor and IRS-1 tyrosine phosphorylation (272). Similar results were obtained using siRNA approach. Decreasing Nox4 was also associated with decreased insulin-stimulated PKB phosphorylation, suggesting that the diminution of insulin receptor and IRS tyrosine phosphorylation caused by inhibiting ROS generation may be of physiological significance. Intriguingly, patients with Leprechaunism (severe insulin resistance caused by rare mutations in the insulin receptor) exhibit diminished Nox4-mediated ROS generation (323). While this may explain more generalized diminution of the biological response of various growth factors, it also supports the role of Nox4-mediated ROS generation in insulin action.

As discussed in section iiiB, ROS generation by Nox4 in adipocytes is associated with oxidative inhibition of cellular PTP1B activity (274). Overexpression of recombinant PTP1B was shown to inhibit insulin-stimulated tyrosine phosphorylation of the insulin receptor, and this effect was reversed when the cells also overexpressed active Nox4. The proof that insulin-induced ROS production precedes biological outcomes is somewhat less established, but in one study ROS generation was shown to peak at 1 min after insulin stimulation and gradually return to prestimulus levels within 10 min (273). Collectively, these studies provide significant evidence in support of assigning a second messenger, or at least a positive modulator role for ROS generated by Nox4 in the insulin signaling cascade (134).

The role of NO as a second messenger or a positive modulator of insulin signaling was demonstrated in various systems. In isolated rat heart muscles and in cultured human vascular smooth muscle cells, insulin-stimulated glucose transport and GLUT4 recruitment were blocked by an inhibitor of NO synthesis (26, 256). Moreover, using in vivo models, intravenous administration of NG-monomethyl-l-arginine (l-NMMA), a competitive inhibitor of all NOS isoforms, acutely induced insulin resistance in rats (18, 362). In addition, eNOS knockout mice were shown to be insulin resistant at the level of the liver and peripheral tissues, whereas the nNOS knockout mice displayed insulin resistance in peripheral tissues, but excluding the liver (384). Taken together, these results indicate that insulin-induced NO production is necessary for successful propagation of the metabolic actions of insulin.

D. Insulin-Like Effects of ROS and RNS Through Alternative Pathways

In addition to this direct activation of the insulin signaling cascade leading to typical metabolic effects of the hormone, RS may participate in the activation of “insulin-like” metabolic effects by activating other, non-insulin-initiated signaling pathways. One of the best physiologically relevant examples is the stimulation of glucose transport in skeletal muscle during exercise (67, 266). Like insulin, skeletal muscle contraction stimulates glucose transport by up to 50-fold during maximal exercise in humans (226). Adding exogenous ROS to skeletal muscle in vitro stimulates glucose transport (174), whereas the thiol antioxidant N-acetylcysteine (NAC) was shown to diminish contraction-mediated glucose uptake by ∼50% (367). This effect of NAC was associated with a similar degree of inhibition of contraction-induced activation of AMP-activated protein kinase (AMPK). AMPK is a central signaling kinase mediating increased glucose uptake in muscle under conditions of high AMP/ATP ratio, like hypoxia and muscle contraction, constituting a non-insulin-dependent pathway to increase muscle glucose utilization (119, 159, 162, 181, 213, 246). Yet, ROS did not seem to mediate AMP-mediated activation of AMPK, as glucose uptake induced by AICAR, which mimics AMP-mediated activation of AMPK, was not affected by NAC (367). Thus the proposed role of ROS in mediating the stimulation of glucose transport in response to skeletal muscle contraction is as follows: skeletal muscle contraction increases O2•− production via mitochondrial respiration. O2•− is then converted to H2O2 by SOD, resulting in direct activation of AMPK, Glut4 translocation to the plasma membrane, and a subsequent increase in glucose transport (225). Interestingly, in skeletal muscle, but not in adipocytes, NAC did not affect insulin-mediated glucose uptake (367). Thus, in muscle, NAC specifically antagonized the second messenger role of ROS in mediating the increase in glucose uptake in response to contraction, but not to insulin. This differential effect of NAC unravels a largely poorly understood specificity in the second messenger function of ROS.

H2O2 functioning upstream of AMPK seems to represent only one role of RS in exercise-induced glucose uptake. A role for RNS was suggested in murine muscle cell line, in which NOS was activated after stimulation of AMPK by AICAR (118). Cell treatment with NOS inhibitors completely blocked the increase in glucose transport after activation of AMPK. Furthermore, as assessed in human subjects in vivo, NOS inhibition reduces leg glucose uptake during cycling exercise (40), and similar observations were made with electrical muscle stimulation in rats (350). Hence, activation of NOS and subsequent NO production is required for contraction-mediated glucose transport downstream of AMPK activation.

E. Key Message From Section III

In normal physiology, ROS and RNS play a coregulatory second messenger function necessary for achieving the full extent of biological effect of insulin. In addition, they mediate other stimuli that culminate in “insulin-like” metabolic effects, such as muscle contraction. Clearly, poorly understood spatial and temporal coordinates, as well as other “biological context” factors, ensure that such intentional ROS and RNS generation yields physiological effects rather than harmful oxidative stress. As will be discussed in subsequent sections, this physiological role of ROS and RNS poses a major challenge when seeking to alleviate the pathological effects of oxidants using antioxidant agents.


As presented above, the insulin signaling cascade constitutes a complex signaling network, culminating, when adequately activated, in the induction of diverse biological functions. Insulin resistance is the reduced capacity of insulin to induce its biological actions in its target organs. Identifying which signaling step(s) mediate insulin resistance induced in response to exposure to oxidants is a daunting task, mainly because of the following two reasons.

  1. Even when a signaling step is found to be defective, it may be difficult to prove that it functionally mediates the impairment in insulin-stimulated action. This is because each signaling step functions within a complex network rather than in a linear pathway. Moreover, even assuming a linear pathway model, a certain degree of decreased activity may still encompass residual function that is sufficient to propagate the signal to the next step.

  2. As discussed earlier, oxidants are commonly generated by various potential inducers of insulin resistance (Fig. 3). Yet, each inducer may produce a unique “oxidant stimulus” characterized by the specific oxidant species generated, their temporal and spatial determinants, and the specific biological context. These would result in a highly diverse array of biological effects. Moreover, the production of oxidants is frequently only one part of several effects various stimuli induce, compromising the ability to dissect out their respective role in a complex biological setting. Thus it is very difficult to make generalizations about the effects of oxidants on the insulin signaling cascade.

The reductionist's approach to circumvent these difficulties is to investigate how exposure of cells or tissues to oxidants affects insulin action, then to identify the impaired signaling steps induced, and ultimately, to establish a cause-and-effect relationship between the oxidant-induced alteration and the impaired insulin action. Most frequently, oxidants, oxidant donors, or enzymatic oxidant-generating systems are directly applied to generate an exogenous oxidizing environment. To modulate endogenous oxidants production, strategies to manipulate the pro-oxidant-antioxidant balance need to be applied. These most frequently include the use of antioxidants, or regulating the expression and activity of antioxidant enzymes pharmacologically or genetically. Clearly, these approaches should be interpreted cautiously, since they generate an artificial environment: exposing cells to exogenously applied oxidants suffers from taking the oxidants out of their true biological context when produced in vivo in response to an oxidant-generating signal. Moreover, altering a single component of the oxidant-antioxidant balance may create paradoxical effects, since most antioxidants have pro-oxidant activities in their oxidized form, depending on the biological context. Nevertheless, these approaches can assist in identifying specific mechanism(s) that then should be confirmed in the more complex and pathophysiological in vivo setting.

Baring in mind the limitations described above, the following sections describe mechanisms offered to explain how oxidants impinge on the insulin signal transduction cascades, thereby inducing insulin resistance.

A. IR and IRS1 Serine Phosphorylation and Enhanced Degradation

The IRS proteins are immediate substrates for the tyrosine kinase activity of the IR upon insulin binding, which constitutes a critical step in the insulin signal transduction cascades. Various knock-down approaches have demonstrated that the members of this group of proteins, particularly IRS1 and IRS2, are critical for normal insulin action, although the relative expression and contribution of each isoform varies among different cells and tissues, as is the degree of functional overlap between them (301, 428, 456). In addition to multiple tyrosine residues accessible to the IR upon insulin stimulation, IRS molecules undergo phosphorylation on multiple Ser residues. Multiple kinases have been demonstrated to exert IRS Ser/Thr kinase activity. Some are “intrinsic” to the insulin signaling cascade, i.e., are activated by insulin as part of its known signaling pathways. These include PI 3-kinase, PKB, PKC-ξ, GSK3, and p70 S6K (147, 258, 264, 326, 385, 423). Indeed, insulin stimulation results also in Ser phosphorylation of IRS1, exerting both positive-feedback loops as well as negative termination signals (477). In addition, kinases “extrinsic” to the classical insulin signaling cascades are bona fide IRS Ser/Thr kinases, including the stress-activated kinase JNK and IKKβ (3, 125).

The most frequently described consequences of IRS Ser phosphorylation are a decrease in its capacity to undergo tyrosine phosphorylation, and enhanced targeting for degradation (477). Phospho-Ser residues can be easily envisioned to prevent phosphorylation on an adjacent tyrosine. In addition, conformational changes in the IRS molecule rendering it a poorer substrate for a tyrosine kinase including the IR can be induced also by Ser residues distant from a specific critical phosphotyrosine site. Ser307, Ser636, and others have been described to mediate reduced tyrosine phosphorylation of IRS1 (3, 34, 360, 424). In addition, enhanced IRS1 Ser phosphorylation was proposed to release IRS1 from internal membrane pools and to result in its increased proteasomal degradation (66, 339, 422).

Both “intrinsic” (p70 S6-kinase, ERK) and “extrinsic” (JNK, IKK) IRS Ser/Thr kinases may be activated by oxidants (reviewed in Refs. 108, 109). Moreover, as discussed in section iii, various phosphatases are a target for inhibition by oxidants, further contributing to enhanced phosphorylation. What is the functional consequence of oxidant-mediated IRS Ser/Thr phosphorylation on insulin action? Both enhanced degradation and impaired insulin-induced tyrosine phosphorylation have been implicated.

1. Oxidant-induced IRS protein degradation (Fig. 9C)

FIG. 9.

Cellular mechanisms for ROS-induced insulin resistance. A: in normal, unstimulated state, insulin signaling molecules are distributed between the cytosol (PI 3-kinase, PKB, and PDK) and internal membrane pools (like the majority of IRS molecules). The actin cytoskeleton is mainly organized in muscle cells in stress fibers. B: upon insulin stimulation, tyrosine residues on the IR and IRS are phosphorylated through the actived insulin receptor kinase, recruiting PI 3-kinase to the plasma membrane and in internal membrane pools. Subsequent generation of PIP3recruits PDK and PKB to the membrane fractions and activates the small GTPase Rac, which induces cytoskeletal reorganization. These propagate the insulin signals, finally culminating in typical metabolic effects of insulin, like increased glucose uptake. IRS proteins not phosphporylated on tyrosine residues are released into the cytosol. C: under conditions of increased oxidative stress (reactive species generation), stress-responsive signaling cascades like the MAP kinase cascades are being activated, leading to increased Ser/Thr phosphorylation of IRS molecules. In addition, RNS induce additional protein modifications of IR, IRS, PI3K, PKB, and actin (red triangles). Modified IRS molecules are released from internal membrane pools and are subjected to increased protein degradation. D: under these conditions, insulin fails to normally elicit metabolic effects. This is because IRS molecules are decreased in content and cannot be normally tyrosine phosphorylated when hyperphosphorylated on certain Ser/Thr residues. Alternatively, PI3K fails to be activated in internal membrane pools which are required for subsequent propagation of the signal to PKB and Rac. IR, insulin receptor; IRS, insulin receptor substrates; PI3K, phosphatidylinositol 3-kinase; PDK, phosphoinositide-dependent kinase; PKB, protein kinase B (also Akt).

When 3T3-L1 adipocytes and Fao hepatoma cells were exposed to micromolar concentrations of H2O2 for 3 h or longer, IRS1 protein content was decreased (339). Pharmacological inhibition of PI 3-kinase and mTOR indicated that this effect was largely mediated by “intrinsic” kinases. However, unlike the effect of chronic exposure to insulin (which theoretically could also be mediated by insulin-induced oxidants generation, see sect. iiiB), the decrease in IRS1 protein content could not be prevented by inhibitors of the proteasome degradation machinery. Furthermore, the degree of IRS1 decrease did not correspond to metabolic insulin resistance (339): when IRS1 decrease was prevented with rapamycin, insulin-induced glycogenesis was not protected. Conversely, the antioxidant lipoic acid protected against the decrease in insulin induction of PKB phosphorylation and of glucose transport (339, 358) without inhibiting oxidation-induced decrease in IRS1 (339). These data suggest that H2O2-induced decrease in IRS1 protein content may not constitute a functional limiting step in insulin action. A different notion was proposed in response to other oxidants: 3T3-L1 adipocytes exposed to the lipid peroxidation product 4-HNE exhibited enhanced IRS1 and IRS2 degradation attributed to increased phosphorylation of IRS1 on Ser307 and elevated levels of HNE-modified IRS molecules (89). As with H2O2, this was not mediated by enhanced proteasomal degradation. Yet, overexpression of fatty aldehyde dehydrogenase, which reverses HNE modification of proteins partially protected IRS from degradation, along with protection against insulin resistance (89). In muscle cells, an NO donor (GSNO) induced enhanced proteasomal degradation of IRS1 (411). In vivo relevance was proposed by demonstrating that whereas in ob/ob mice IRS levels were diminished in skeletal muscle, this was not observed when iNOS was knocked down on the ob/ob background, and these mice exhibited improved insulin sensitivity (411). Enhanced proteasomal degradation of IRS1 was also demonstrated in vascular smooth muscle cells in response to angiotensin II (422). Oxidants were implicated by demonstrating that this could be blocked by overexpressing catalase or by pharmacological antioxidants. Thus enhanced IRS protein degradation remains a potential mechanism for insulin resistance in response to some oxidants. The protein degradation machinery (proteasomal or lysosomal) utilized may be cell-type or tissue-specific.

2. Role of oxidants-induced serine phosphorylation in decreasing IRS tyrosine phosphorylation (Fig. 9, C and D)

Decreased insulin-stimulated tyrosine phosphorylation of IRS molecules also related to enhanced Ser/Thr phosphorylation appears to involved kinases both “intrinsic” and “extrinsic” to the insulin signaling cascade. In Fao cells exposed to H2O2, pSer307 and pSer636 appeared to identify minimally overlapping pools of IRS1 molecules (34): Ser636-phosphorylated IRS1 exhibited peripheral distribution and exhibited intact tyrosine phosphorylation, interaction capacity with PI 3-kinase, and cellular redistribution upon insulin stimulation. In contrast, Ser307-phosphorylated IRS1 molecules were retained in the cellular membrane fraction and exhibited diminished capacity to undergo tyrosine phosphorylation and to interact with PI 3-kinase following insulin stimulation. The combinatorial nature of the oxidant-mediated regulation of IRS molecules by phosphorylation was exemplified by the fact that partial prevention of insulin resistance was achieved only using a combination of JNK and IKK inhibitors (34). Furthermore, even this combination did not fully restore IRS1 Ser/Thr phosphorylation state in control cells, suggesting additional determinants linking oxidants to impaired insulin action through IRS1 phosphorylation.

B. Impaired Signal Transmission From PI 3-kinase to PKB

Section ivA discussed the evidence that oxidants impaired insulin signal transduction by affecting IRS proteins. In fact, H2O2 were shown to impair insulin signal transduction at steps as proximal as the IR: insulin-stimulated IR activation was decreased by H2O2 in HEK293 cells and NIH3T3 overexpressing the insulin receptor and IRS1 (157) and in vascular smooth muscle cells and in whole rat muscles exposed to micromolar H2O2 (94, 127). Yet, other experimental models implicated more distal oxidation-sensitive steps in the insulin signaling cascade. In those, when assessed in total cell lysates (i.e., when subcellular compartments are disregarded; these will be discussed in sect. ivC), intact IR and IRS proteins tyrosine phosphorylation were observed, as well as an intact insulin-stimulated interaction between IRS proteins and PI 3-kinase (211, 357, 429). In fact, these steps following insulin stimulation were even reported to be exaggerated in response to preexposure to oxidants, possibly reflecting the oxidant-induced inactivation of cellular phosphatases, particularly PTP1B (see sect. iiiA for further details), or impaired feedback inhibition by downstream signaling components of the system. In contrast to this effect on proximal insulin signaling events, both 3T3-L1 adipocytes and L6 muscle cells exposed to submillimolar stable concentrations of H2O2 resulted in impaired PKB activation (211, 429). Metabolic end points downstream of PKB were all shown to be affected, including impairment in insulin-stimulated glucose uptake and GLUT4 translocation, insulin-induced protein and glycogen synthesis, all consistent with the induction of metabolic insulin resistance (211, 356, 430). Interestingly, this effect of oxidants was stimulus specific, since under the same conditions the metabolic effects of platelet-derived growth factor were preserved (430). The oxidative mechanism of this inhibition was further supported by demonstrating that enhancing the antioxidant capacity of the cells using the antioxidant lipoic acid or N-acetylcysteine protected against this signaling effect (211, 358). Finally, it is noteworthy that a signaling defect between PI 3-kinase and PKB is not a mechanism for insulin resistance unique to oxidative stress, as it was observed in cultured cells in response to FFA (378), ceramide (449), growth hormone (418), and others (305). Both pharmacological antioxidants and overexpression of antioxidant enzymes provided partial protection against TNF- and dexamethasone-induced defects in PKB phosphorylation, suggesting that oxidants mediated their effect (188). Thus insulin signal transmission from PI 3-kinase to PKB may constitute an oxidation-sensitive step in the insulin signaling cascade, which is impaired by various inducers of insulin resistance through oxidant generation.

Closer analysis of the steps through which the insulin receptor-initiated signal is transmitted from PI 3-kinase to PKB reveals a complex process: phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P3] needs to be generated, “sensed” by the PH domains of PKB and one of its upstream kinases PDK1, thereby promoting their translocation from the cytosol to the membranes where PDK1 can phosphorylate PKB on Thr308 (Fig. 9B). An additional phosphorylation step on Ser473 of PKB is required for full activation, but although this step is also PI 3-kinase dependent, the exact kinase responsible has not been unequivocally revealed (95, 368). Which of these steps is oxidation sensitive is largely unknown. The human immunodeficiency virus (HIV) protease inhibitor nelfinavir induces insulin resistance characterized by normal insulin-induced PI 3-kinase activity but impaired PKB activation (25, 359), and a recent study suggests that production of reactive oxygen species mediate this effect (24, 248, 448), which can be relieved by cotreating the cells with an SOD/catalase mimetic agent or ascorbic acid (24, 448). Insulin-induced PI(3,4,5)P3 production was intact in nelfinavir-treated cells, and a constitutively active PI 3-kinase could not rescue impaired insulin action, suggesting that the oxidation-induced defect was downstream of these steps (unpublished results). In contrast, membrane-targeted PKB was normally phosphorylated, suggesting intact PKB-kinase activity. Furthermore, membrane-targeted PKB promoted GLUT4 translocation, confirming that nelfinavir-induced oxidative stress impaired PKB sensing of PI(3,4,5)P3, leading to defective insulin-stimulated GLUT4 recruitment to the plasma membrane. Whether this is the result of oxidation-induced posttranslational modification of PKB, as was suggested to occur in response to ceramide (340), or due to its binding to a protein that prevents its translocation, as was demonstrated with TRB3 (97), is unknown. The direct oxidative or nitrative modification of PKB, as a cellular mechanism for oxidation-induced insulin resistance, will be detailed in section ivD.

C. Disruption of spatial organization of the insulin signal: potential role for the actin cytoskeleton

Insulin signaling components undergo spatial redistribution upon insulin stimulation (Fig. 9, C and D). Best described is the recruitment of PH-containing proteins, like PKB and its upstream kinase PDK1 from the cytosol to the sites of PI(3,4,5)P3 production by PI 3-kinase, a site most frequently identified as the plasma membrane (187, 330). In addition, upon acute insulin stimulation, cellular redistribution of IRS1 and PI 3-kinase occurs in adipocytes: a significant proportion of IRS1 is associated in the unstimulated basal state with internal structures sedimenting at high velocity, consisting of internal membranes (“low density microsomes,” LDM) and/or cytoskeletal elements (65, 196, 429). Insulin induces the release of IRS1 from this fraction to the cytosol, potentially with preference to IRS1 molecules phosphorylated on Ser residues, which are then available to the proteasome for protein degradation as mentioned in section ivA (Fig. 9B). In contrast, tyrosine-phosphorylated IRS1 accumulates in the LDM, promoting the recruitment of PI 3-kinase from the cytosol through the interaction between the SH2 domains of PI 3-kinase with phosphorylated tyrosine residues of IRS1. These effects of insulin, initially supported by biochemical analyses (subcellular fractionation studies)(65, 196, 429), were later supported by cellular imaging approaches of L6 myoblasts (325). In these cells, PI(3,4,5)P3 is produced in the plasma membrane in response to insulin stimulation, but also in intracellular locale, most likely internal membranes captured within actin cytoskeleton structures beneath the plasma membrane, reminiscent of lamellopodia (325, 354). These insulin-induced actin structures gather IRS1, the p110α subunit of PI 3-kinase (but not the β-isoform), and PKB (but not PKC-ξ, another PI 3-kinase-dependent enzyme). Thus insulin-induced actin remodeling appears as a means of generating “signaling microdomains” spatially segregating some signaling components from others.

How does insulin promote actin remodeling, and is it important for insulin action? Actin dynamics are regulated by small GTPases, and a few have been demonstrated to be activated (GTP-loaded) in response to insulin, and to mediate actin dynamics. The small GTPase TC10 was suggested when overexpressed in adipocytes to be activated in a class I PI 3-kinase-independent pathway and to promote actin dynamics in adipocytes (219). This signaling pathway was suggested to be required for insulin-induced GLUT4 translocation to the plasma membrane, potentially through the activation of PI3P generation (271). In contrast, in muscle cells, although overexpressed TC10 could be activated by insulin, it was not required for the induction of actin structures (210). Rather, another small GTPase, Rac1, was activated by insulin in a PI 3-kinase-dependent (but PKB-independent) pathway and was required for insulin-induced actin structures (210, 211). Furthermore, Rac activation was required for GLUT4 translocation to the plasma membrane in response to insulin.

Several studies suggest that actin-facilitated redistribution of insulin signaling components is an oxidation-sensitive event in insulin signaling. 3T3-L1 adipocytes exposed to micromolar H2O2 concentrations exhibited normal PI 3-kinase activation when assessed in total cell lysates, but markedly decreased activation of PI 3-kinase in the intracellular high-speed pellet/LDM fraction (357, 429). Furthermore, insulin-induced recruitment of PI 3-kinase into this fraction and IRS1 release from it were diminished (429). This failed ability to promote the cellular redistribution of these two key proteins was associated with diminished PKB activation and impaired insulin-induced metabolic actions (stimulation of glucose uptake, glycogenesis, and lipogenesis). In Fao cells, Ser636-phosphorylated IRS1 was still released from the membrane fraction upon insulin stimulation, but not IRS1 molecules phosphorylated on Ser307, which demonstrated decreased tyrosine phosphorylation and interaction with PI 3-kinase (34). Thus oxidation-induced disruption of cellular redistribution of signaling molecules in response to insulin stimulation was associated with impaired insulin action. An in vivo model of oxidative stress provided support for this notion. In this model, oxidative stress was induced in rats using buthionine sulfoximine (BSO), an inhibitor of γ-glutamylcysteine synthetase, the rate-limiting enzyme in the glutathione biosynthesis. The consequential drop in tissue levels of glutathione, a major cellular antioxidant, was shown to result in increased markers of oxidative stress in both cells and animals, and in vivo induced impaired glucose homeostasis. Whereas one study failed to demonstrate impaired insulin responsiveness in adipose tissue and cultured adipocytes (229), another study using higher doses (and different administration route) of the inhibitor demonstrated impaired insulin signaling and metabolic actions in both adipose tissue and skeletal muscle (311). Remarkably, the underlying mechanism proposed was the impaired ability of insulin to induce the translocation to and activation of PI 3-kinase in intracellular compartments, fully consistent with the observations in the adipocyte cell line (429).

Recent studies suggest that insulin-induced actin remodeling is an oxidative stress-sensitive step underlying the abnormalities in the cellular distribution of its signaling components and in signal transmission (211) (Fig. 9D). Actin has been shown to be oxidatively modified, as will be discussed in section ivD. In addition, in L6 muscle cells exposed to H2O2, insulin-induced actin remodeling was diminished, as was the GTP loading of Rac1 (211). Dose-response analysis of the effects of H2O2 demonstrated a closer association between the impairment in GLUT4 translocation and the diminution of insulin-induced Rac1 activation, than the inhibition of PKB activation. Since Rac1 and PKB are parallel PI 3-kinase downstream substrates, these results suggest that in muscle cells, insulin-induced Rac activation (and the resulting actin remodeling) constitute an oxidation-sensitive point between insulin stimulation and GLUT4 translocation.

D. Direct Modification of Insulin Signaling Components by Oxidants

ROS/RNS can react and chemically modify various cellular components, including proteins. Protein oxidative/nitrative modifications are largely classified into two main categories (reviewed in Refs. 96, 131 and summarized in Table 2). 1) Readily reversible modifications can be used also as a strategy to regulate protein function in normal physiology, somewhat reminiscent of the well-characterized regulation by reversible phosphorylation/dephosphorylation. In this group, oxidation-reduction processes (redox regulation) are the prototypes, involving protein moieties such as vicinal sulfhydryls (133, 205). Only excessive accumulation of this type of modification and an ensuing “pathological” functional consequence signifies “oxidative/nitrative stress”. 2) Irreversible modifications are those for which cellular reversal mechanisms are limited or absent. Accumulation of this group of modifications is considered a marker of oxidative/nitrative damage as it usually results in protein dysfunction, with affected proteins frequently destined to degradation and/or to form dysfunctional protein aggregates. Of these, best known is protein carbonylation, a large group of oxidative/nitrative modifications resulting in the introduction of various carbonyls (aldehydes and ketones) into the protein (57, 78, 309). Carbonyls can form by direct attack of the protein backbone by various reactive species resulting in protein fragmentation. Alternatively, they occur by interaction of various amino acid side chains (Arg, Lys, Pro, Thr) with either ROS/RNS or secondarily with other active carbonyls. The latter include AGEs or lipid peroxidation products like 4-HNE. One of the most studied irreversible modification is tyrosine nitration, the result of peroxynitrite reaction with the OH moiety in the tyrosine side chain (198, 342, 434).

View this table:

Examples of oxidative and nitrative protein modifications

Along the discussion in section iC, increased total serum protein carbonylation and nitrotyrosine levels have been reported in diabetes (37, 50, 426), potentially implicating both ROS, RNS, and active carbonyls (like 4-HNE) in systemic protein modification associated with this disorder. In cells, altered protein function occurred in response to in vitro oxidation/nitration. This was demonstrated on several specific proteins with potential relevance for insulin action [PKC-ξ (61), GAPDH (46, 158, 288), and actin (5, 79, 450)], though a connection to impaired insulin action has not been reported for those. Nevertheless, different types of modifications affecting multiple proteins could impinge in a combinatorial fashion on the insulin signaling cascade, thereby contributing to insulin resistance. To this end, several studies provide direct evidence tying oxidative/nitrative modifications of insulin signaling components to impaired insulin signaling and actions.

The reversible S-nitrosylation of IR, IRS1, and PKB have been shown to occur in response to NO donors in cells in culture and in isolated muscle, as well as in muscle of genetic and nutritional mouse models (217). S-nitrosylation of Cys224 (470) or of Cys296 (265) of PKB results in impaired activity, the latter because it prevents an essential disulfide bridge with Cys310. Intriguingly, nitrosylation of PKB may uncouple the enzyme's phosphorylation state from its activity, suggesting that modified PKB may be dysfunctional even if normally phosphorylated (470). Support for a causative role of S-nitrosylation of insulin signaling components in insulin resistance was suggested by the finding that the insulin sensitizer rosiglitazone (49) and exercise (84) decreased iNOS expression and S-nitrosylation of IR, IRS1, and PKB in muscle of rodent models of obesity. Furthermore, to direct S-nitrosylation of insulin signaling proteins as a cause of insulin resistance, nitrosative stress may mediate inflammatory responses and ER stress, which in turn can cause impaired insulin action (217).

The irreversible modification of nitrosylation of tyrosine residues was also shown to impair the function of insulin signaling components at different levels. 3T3-L1 adipocytes exposed to peroxynitrite exhibited impaired insulin action (glucose uptake), along with normal insulin-stimulated tyrosine phosphorylation of IR, but decreased IRS1 tyrosine phosphorylation, interaction with PI 3-kinase, and activation of PKB (307). Mass spectrometry analysis of IRS1 revealed tyrosine nitration of at least five Tyr residues, one of which, Tyr939, is known to be critical for the interaction with PI 3-kinase (307). Tyrosine nitration of the regulatory p85 subunit of PI 3-kinase was shown to cause dissociation from the catalytic p110 subunit of the enzyme (100), providing an additional molecular mechanism for nitration-mediated impairment in insulin signal propagation.

Lipid peroxidation adducts on proteins is potentially the major carbonyl oxidative modification (447). 3T3-L1 adipocytes exposed to the lipid peroxidation product 4-HNE exhibit increased protein HNE modification and impaired insulin actions (89). These were attributed to enhanced degradation of IRS proteins secondary to their HNE modification, as overexpressing fatty acid aldehyde dehydrogenase, which reverse protein HNE modification, prevented the decrease in IRS expression and in insulin responses (89). A recent proteomic approach to identify carbonylated proteins in adipose tissue of HFF mice identified the fatty acid binding protein as being one of the proteins modified by 4-HNE (144). Fatty acid binding protein was linked to the development of insulin resistance (knockout mice were protected from diet-induced insulin resistance; Refs. 171, 186) and was shown to activate hormone-sensitive lipase when in the fatty acid-bound state (212). Thus, if HNE modified fatty acid binding protein mimics the fatty acid-bound state, then these data provide an additional putative mechanism linking protein oxidative modification to insulin resistance (144).

E. Alterations in Gene Regulation

Reactive oxidants and radicals are prominent regulators of gene expression, as they can directly (through oxidative modifications) and/or indirectly (by activating stress-sensing kinases) regulate the activity of many transcription factors families (382). Direct modification, mostly oxidation of cysteines critical for DNA binding capacity, usually lead to decreased transcriptional activity, similar to the effect oxidants exert on protein phosphatases (discussed in sect. iiiA). In contrast, phosphorylation by kinases regulates the transcriptional capacity in a more diverse manner, as reviewed in References 80, 382, and 435. Oxidants therefore induce alterations in the expression of multiple genes, including gene products that are related to insulin action, thereby providing an additional mechanism for oxidant-induced insulin resistance.

1. Oxidant regulation of GLUT4 gene expression

GLUT4 is expressed in cell types in which insulin induces an increase in glucose uptake, mainly skeletal and cardiac muscle and fat cells. GLUT isoform-selective inhibitors demonstrated that at least 60–80% of glucose uptake into these cells and tissues is facilitated by GLUT4 (295, 355). In both human and experimental insulin resistance conditions, GLUT4 protein and mRNA expression is largely unaltered in muscle, suggesting that the decrease in insulin-stimulated glucose uptake is largely mediated by impaired signaling promoting GLUT4 mobilization from internal membrane pools to the plasma membrane (124, 354). In contrast, decreased expression of GLUT4 in adipocytes is tightly associated with total body insulin resistance. Furthermore, fat-specific GLUT4 knockout mice (1), and heterozygote total-body GLUT4 null mice that have decreased expression of GLUT4 in fat (400), exhibit an insulin-resistant and type 2 diabetes-like phenotype. These suggest that adipose tissue gene regulation of GLUT4 is an important determinant of total body glucose homeostasis. Recently, retinol binding protein 4 (RBP4) was suggested to mediate liver and muscle insulin resistance in response to diminished levels of GLUT4 in adipose tissue (142, 469).

In vitro, adipocytes exposed to oxidants exhibited a decrease in GLUT4 protein and mRNA content (332, 357). The GLUT4 promoter includes putative binding sites for C/EBPs and for NF1, among other transcription factors, and is also thought to be regulated by PPARγ, although the exact binding site for this factor in the GLUT4 promoter has not been clearly identified. Upon exposure of adipocytes to micromolar H2O2 concentrations, EMSA studies suggested that oxidants decreased DNA binding of C/EBPα-containing dimers while increasing that of C/EBPδ dimers (331). Intriguingly, DNA binding capacity of the C/EBPs transcription factor family is not known to be particularly sensitive to redox regulation (62) [though the ξ isoform (also known as GADD153 or CHOP10) is known to be upregulated by oxidative stress (151)]. This suggested that expression level, rather than direct alteration in DNA binding capacity, altered C/EBP's regulation of GLUT4 expression. Consistent with this proposition, C/EBPα protein and mRNA levels were downregulated, whereas those of C/EBPδ upregulated in response to oxidative stress (331). Moreover, the antioxidant lipoic acid could prevent the decrease in C/EBPα mRNA and in GLUT4 mRNA, consistent with an oxidation-mediated mechanism. Intriguingly, similar observations were observed in response to TNF (206, 401, 402), which was proposed to exert part of its actions by inducing ROS generation (276, 467).

Despite the above, reporter gene assays suggest that the C/EBP binding sites in the GLUT4 promoter cannot fully mediate oxidation-induced GLUT4 gene repression (331). An additional region proposed to mediate this effect was located ∼700 bp upstream of the transcription initiation site of the GLUT4 gene, an area that includes the insulin responsive element (IRE)(68, 332). Although this sequence was shown to mediate GLUT4 repression also in response to chronic insulin (68), the mechanism exerted by H2O2 differed, being associated with decreased DNA binding of NF1 to IRE. Different from the C/EBPs, DNA binding capacity of this transcription factor is known to be redox sensitive, and indeed, partial recovery of its DNA binding to the IRE sequence from the GLUT4 promoter could be demonstrated in nuclear extracts from oxidized cells treated with DTT (332).

Thus, collectively, oxidants appear to repress GLUT4 gene expression by direct oxidation of NF1 as well as via suppression of C/EBPα expression.

2. Oxidant regulation of adiponectin gene expression

Adiponectin is currently the adipokine (fat-derived hormone) most clearly related to human health and disease. Secreted almost exclusively from adipocytes, it is strongly inversely correlated with fat mass in obesity and with its associated cardiovascular risk. Consistently, its physiological effects all seem to be beneficial to health and include anti-inflammatory actions, insulin-sensitizing actions at the level of the liver and muscle, direct antiatherosclerotic effects on the vasculature, and possibly anorexigenic signals in the hypothalamus, limiting excessive food intake. Obesity-associated decrease in adiponectin gene expression is thus a potentially strong mechanism linking obesity to its common comorbidities.

Plasma and urinary lipid peroxidation markers indicative of systemic oxidative stress (TBARS and 8-epi-PGF, respectively) correlated with lower circulating adiponectin levels (122). Establishing a causative role for oxidants as an isolated factor in the repression of the adiponectin gene was performed in cells in culture: 3T3-L1 adipocytes expressed lower adiponectin mRNA levels when exposed to oxidants generated by either glucose oxidase (214, 393) or hyoxanthine and xanthine (122) added to the medium, when H2O2 was added directly to the medium (52, 214), or when cells were exposed to the lipid peroxidation derivative 4-hydroxynonenal (4-HNE) (393). The sulfhydryl antioxidant N-acetylcysteine (52, 122, 214) or catalase (52, 393) prevented the drop in adiponectin. Remarkably, 10-min exposure to 500 μM H2O2 induced this effect 16 h later without compromising cell viability (214), suggesting cellular “memory” and a delayed biological response to a severe short-term oxidative insult. A transcriptional repression mechanism utilized by H2O2 was shown by a reporter gene assay using the adiponectin promoter (122). In vivo, hypoadiponectinemia and decreased adiponectin mRNA in adipose tissue were observed in BSO-treated rats (199), a model for oxidative stress, which also exhibits insulin resistance (311). Collectively, the studies discussed in this section suggest that oxidative stress in adipose tissue mediates obesity-induced hypoadiponectinemia by regulating adiponectin gene expression, thereby contributing to insulin resistance. Finally, a recent study demonstrated that adipose tissue hypoxia through ER stress and activation of C/EBPξ/CHOP10/GADD153 directly repressed the transcription of adiponectin (184). The responses to both hypoxia (391) and ER stress (143, 166) are associated with enhanced production of radicals, and C/EBPξ is a transcription factor known to be upregulated by both ER and oxidative stress (151, 349). Thus, although not supported by some reports (52), it is plausible that reactive species also mediate hypoadiponectinemia induced by hypoxia and ER stress. Similarly, hypoadiponectinemia induced by fatty acids was proposed to be mediated by increased production of reactive species secondary to decreased FOXO1 (409).

Thus oxidants may directly mediate the suppression of the adiponectin gene in response to various adipose tissue alterations in obesity, thereby contributing to whole-body insulin resistance.

F. Key Message From Section IV

Multiple molecular mechanisms have been proposed to contribute to the induction of cellular and whole body insulin resistance by reactive species. These include oxidant-induced alterations in phosphorylation state and protein modifications, altered cellular localization and half-life, and changes in gene regulation. Affected proteins are either direct insulin signaling molecules, oxidant-sensitive signaling pathways which impinge on the insulin signaling cascade, transcription factors, or hormones and cytokines indirectly affecting whole body insulin sensitivity.


As detailed in the above sections, ROS and RNS generation affects insulin action in seemingly opposing modes: on one hand, their production is required for insulin action, as RS play a second messenger role in insulin signaling (see sect. iii). On the other hand, they are likely mediators in the development of impaired insulin action (see sect. iv). This dual role creates a paradoxical situation when attempting to manipulate ROS/RNS generation for therapeutic purposes; decreasing oxidant load would be accepted to prevent or improve insulin resistance, but theoretically could induce insulin resistance by decreasing the action of required second messengers. This paradox likely contributes to the overall disappointing results of antioxidant intervention trials so far (discussed further in sect. vC).

This section describes the available information on the possibility to prevent or improve insulin resistance by manipulating the oxidant-antioxidant balance. Studies using in vitro, cell line systems are described first (sect. vA), followed by studies in animal models and in humans (sect. v, B and C, respectively). In cellular and animal studies, attempts were made to inhibit ROS/RNS generation, as well as to facilitate the antioxidant defense. In humans only the latter is currently available.

A. Cellular Systems

1. Preventing ROS/RNS generation

Cellular evidence for the ability to prevent insulin resistance by decreasing ROS/RNS generation is available for each of the three major ROS/RNS generating systems: mitochondria, NADPH oxidases, and NOS (described in sect. iiA). Inhibiting mitochondrial ROS generation by overexpression of uncoupling proteins 1 (UCP1) in hepatoma cells prevented TNF-induced increase in the serine phosphorylation of IRS1 and the attenuation of insulin-stimulated tyrosine phosphorylation of this protein (194). Consistently, preventing mitochondrial ROS generation in 3T3-L1 adipocytes partially prevented insulin resistance induced by TNF, dexamethasone, and high glucose and increased adiponectin levels (188, 261).

ROS production by NADPH oxidases was implicated in the induction of insulin resistance (impaired insulin-stimulated IRS1 tyrosine phosphorylation, activation of PKB and GLUT4 translocation) by angiotensin II in L6 muscle cells (193). This was supported by demonstrating that both pharmacological inhibitors of NADPH oxidases, as well as siRNA-based gene silencing of p47Phox prevented these effects. Yet, a recent paper questioned the possibility that NADPH oxidase is significantly expressed in this cell line (191).

Inhibiting iNOS-induced NO generation in muscle cells was proposed as a mode of action of the insulin-sensitizing drugs metformin and troglitazone. Muscle cells exposed to an inflammatory stimulus (TNF, interferon-γ, and lipopolysaccharide) increased iNOS expression and NO generation, which were associated with impaired insulin-stimulated activation of PI 3-kinase. Metformin and troglitazone prevented these effects through their capacity to activate AMPK (both an activator of AMPK and a pharmacological inhibitor of iNOS prevented inflammation-induced insulin resistance) (337). In contrast, in cultured human adipocytes, iNOS mRNA was activated only in a transient manner in response to inflammatory cytokines, but the protein was undetectable (although readily observable in 3T3-L1 adipocytes) (262). Consistently, iNOS inhibitors could not prevent the decrease in insulin-stimulated glucose uptake that was induced by exposure to inflammatory agents (262). Regardless of this discrepancy between the two cell types, it appears based on in vitro experimental systems that inhibiting either mitochondrial or NADPH-derived ROS generation, and at least in muscle cells also iNOS-induced NO generation, can relieve insulin resistance.

2. Enhancing antioxidant defense

The most frequent strategy to facilitate antioxidant defense is to supplement cells with antioxidant agents. This is then followed by assessment of the ability to induce insulin resistance in these cells, either in response to direct exposure to oxidants or exposure to inducers of insulin resistance thought to induce oxidants as part of their mode of action. This line of research has yielded numerous reports with positive results (109, 111 and references within).

The antioxidant lipoic acid partially prevented in both 3T3-L1 adipocytes and in L6 muscle cells the induction of insulin resistance by exposure to a H2O2 enzymatic generating system (used to generate stable, micromolar concentrations of the oxidant) (269, 358). This thiol antioxidant has several possible modes of action: it directly reacts with radicals, is involved in the recycling of other antioxidants including vitamin C and E, enhances cellular GSH by maintaining it in a reduced form and by increasing its synthesis, and chelates transition metals (317, 319). In addition, lipoic acid is a cofactor of the α-keto-acid dehydrogenases like pyruvate dehydrogenase, and thus may directly affect the oxidative metabolism of glucose through nonantioxidant mode of action. Thus comparison of the ability of lipoic acid to prevent insulin resistance to other antioxidants can be used to assess the likelihood of its antioxidant mode of action. Supplementing 3T3-L1 adipocytes with vitamin C did not confer the protective effect observed with lipoic acid upon exposure to H2O2-generating system (358), whereas vitamin E supplementation protected L6 muscle cells (444). Yet, the thiol antioxidant NAC was shown to protect 3T3-L1 adipocytes from insulin resistance induced by exposure to H2O2 (387). Moreover, this antioxidant provided partial protection against insulin resistance induced by angiotensin II in vascular smooth muscle cells (422), or by TNF and dexamethasone in 3T3-L1 adipocytes (188). In the latter cells, NAC prevented the H2O2-induced decrease in adiponectin gene and protein expression (122). Collectively, these studies suggest the potential beneficial effect of thiol antioxidants in protecting against insulin resistance generated either in response to direct exposure to oxidants, or to factors thought to induce insulin resistance at least partly by provoking ROS generation.

An additional approach to enhance antioxidant defense has been to increase the activity of antioxidant enzymes. This is most frequently achieved either by overexpression gene delivery approaches, or by chemical agents mimicking the activity of antioxidant enzymes. The SOD/catalase mimetic agent manganese (III) tetrakis(4-benzoic acid) porphyrin (MnTBAP) provided 50–60% recovery of insulin-stimulated glucose uptake activity that was decreased in 3T3-L1 adipocytes by TNF or dexamethasone (188). Insulin-stimulated phosphorylation of PKB and of p70S6 kinase were similarly recovered by MnTBAP. A similar partial protection by MnTBAP against decreased insulin-stimulated PKB phosphorylation was observed in 3T3-L1 adipocytes exposed to the HIV protease inhibitor nelfinavir (24), which are thought to involve ROS generation in the induction of insulin resistance (24, 248, 352). Increasing intracellular catalase activity, either by stable overexpression or by adding the enzyme to the cells, prevented H2O2, TNF, dexamethasone, or angiotensin-induced insulin resistance (157, 188, 422).

B. Experimental Animal Studies

1. Preventing ROS/RNS generation in animals

Interfering with mitochondrial ROS production in vivo to test its contribution in the induction of insulin resistance has been difficult to accomplish. Although overexpressing uncoupling proteins is indeed expected to decrease mitochondrial ROS generation as shown in cell culture systems (discussed in sect. vA1), it also augments mitochondrial respiration and, hence, energy expenditure. Hence, transgenic mice overexpressing UCP1 or UCP3 in adipose tissue or muscle were found to be protected from obesity and its metabolic consequences (63, 240, 255). Yet, these studies cannot prove an isolated role of mitochondrial ROS production in insulin resistance.

Interfering with ROS generation by the NADPH oxidase system frequently involves pharmacological inhibitors like DPI and apocynin. In vivo administration of DPI decreased glucose levels (163, 180), but this was likely a result of this inhibitors capacity to inhibit mitochondrial NADH: coenzyme Q reductase, hence representing the classical Pasteur effect (increased glucose utilization and lactate generation in response to interference with mitochondria energy production). In vivo administration of apocynin decreased adipose tissue ROS generation and lipid peroxidation products in a genetic mouse model of obesity and diabetes, decreased glucose and insulin levels, and elevated adiponectin (122). Although to the best of our knowledge NOX4 knockout mice have not been tested in the context of insulin resistance, these data provide supporting evidence for an in vivo role of adipocyte NADPH oxidase generation of ROS in whole body insulin resistance.

2. Enhancing antioxidant defense in animals

Different antioxidants, including vitamins C and E, lipoic acid, NAC, and flavanoids, have been shown to attenuate different markers of systemic oxidative stress in a variety of experimental animal models for obesity, type 1 or type 2 diabetes. Yet, whether such antioxidants can prevent or improve insulin action by interfering with oxidation processes is still an open question. This is partially because at the whole body level, an intricate relationship between metabolic and endocrine control and the generation of oxidants complicates the interpretation of the results and limits determining directionality in cause-effect relationships: Does the decrease in oxidative stress cause the improved insulin action, or does it only reflect (i.e., is the consequence of) lower glucose and insulin levels? Nevertheless, several studies demonstrated that lipoic acid supplementation positively affects glucose homeostasis and insulin action. With the use of the fatty Zucker rat, a model of insulin resistance and obesity, lipoic acid treatment was shown to improve markers of insulin resistance including reduction in circulating insulin and FFA concentrations (203). At the skeletal muscle level, isolated epitrochlearis muscles from rats treated with lipoic acid exhibited improved insulin stimulated glucose transport, glycogenesis, and CO2 production (170, 408). In general accordance with these findings, improved insulin sensitivity both systemically as well as at the skeletal muscle level was observed in the hypoinsulinemic streptozotocin (STZ)-induced diabetic rat model (230). This was attributed to enhanced protein levels of GLUT4 in muscle and was associated with improvement in a variety of oxidative stress parameters (231). Further supporting a thiol-antioxidant mode of action of lipoic acid, NAC was also shown to protect against insulin resistance in ZDF rats or db/db mice (218, 421), or in normal mice rendered insulin resistant and hyperglycemic in response to a 6-h infusion of high glucose (152). In accordance, in high-sucrose- or fructose-fed rats, increasing dietary cysteine content either by high cysteine proteins or by NAC markedly reduced diet-induced oxidative stress and insulin resistance (36, 394). Thus most available information presents a striking agreement between cell culture work (see sect. vA) and experimental animal models in the potential therapeutic potential of thiol-antioxidants to improve insulin action.

In addition to thiol antioxidants, enhancing antioxidant enzyme activity in vivo using the SOD/catalase mimetic antioxidant MnTBAP improved insulin sensitivity in ob/ob mice (188). Tempol and other SOD mimetic antioxidants improved insulin action in various rats with insulin resistance, including Zucker rats (16), aged rats (9), and insulin resistance induced by angiotensin II (30, 453, 465).

C. Human Studies

Observational studies (116, 282, 365) have provided a rationale for conducting intervention trials to assess if directly manipulating the oxidant-antioxidant balance can improve insulin action in humans, thereby preventing the occurrence of and/or improving metabolic control in diabetes. Such human intervention trials are restricted to attempts to increase antioxidant defense by antioxidant supplementation. (Theoretically, attenuating ROS/RNS generation may occur secondary to other interventions, like those that would improve metabolic control or inflammatory state, but approaches to directly inhibit the major ROS/RNS generators are largely unavailable for clinical applications.) Improved insulin sensitivity and/or improved markers of glucose tolerance have been demonstrated with lipoic acid (110, 201, 202, 239), NAC (121), vitamin E (45, 320322), and vitamin C (177) (and recently reviewed in Refs. 106, 107). Yet, this was mainly observed in small-sized, short-term trials, and frequently, reports exist demonstrating no measurable effects of the same agents trialed on similar study populations (53, 76, 136, 263). Late diabetes complications known to associate with metabolic control (407) also do not seem to be positively affected by antioxidant therapy (reviewed in Ref. 380). Consistently, despite some supporting evidence for the ability to improve insulin action with antioxidants, current clinical guidelines do not recommend antioxidant supplementation for the general population of persons with impaired insulin action, like type 2 diabetes (7). This is not only because evidence for a beneficial effect is anecdotal or too weak, but also due to evidence suggesting potential harm, including increase in all-cause mortality, with vitamin E, carotene, selenium, and other antioxidant supplementation (29, 32, 285). The recent demonstration that selenium supplementation may increase the incidence of type 2 diabetes (406) potentially reflects negative effects of selenium on insulin sensitivity. Yet, whether this is mediated through an antioxidant effect is unknown.

The careful assessment of evidence-based literature conducted by clinical guideline committees reflects the overall disappointing results in the ability to positively affect insulin sensitivity-related health outcomes with currently used antioxidants in placebo controlled intervention trials. Yet, it is difficult to determine whether this current assessment rules out a basic mechanistic role for ROS/RNS in the induction of insulin resistance in humans. More likely, the extremely complex oxidant-antioxidant systems and their intricate, occasionally opposing roles in physiology and pathophysiology render currently existing modes aimed at manipulating the oxidant-antioxidant balance clinically ineffective. This could be either because supplemented antioxidant is efficiently balanced (buffered) by other components of the system that are in equilibrium with it, because excessive antioxidants can be converted to pro-oxidants, and/or because a specific antioxidant simply does not affect the specific mechanism(s) responsible for excessive oxidant generation. Reflecting this notion is the recommendation endorsed by clinical guidelines (7), to consume sufficient amounts of naturally occurring dietary antioxidants by balanced eating habits. Whether consuming diets proposed to be high in antioxidants exerts its beneficial effects by providing carefully balanced, bioavailable antioxidant mix, and/or includes novel, yet uncharacterized antioxidants, or simply includes yet unidentified dietary components unrelated to an antioxidant mode of action altogether, is largely unknown. Clearly, addressing these possibilities experimentally will enhance current understanding of the role of ROS/RNS in insulin action and resistance, the relevant mechanisms by which these effects occur, and hopefully, would eventually lead to interventions to prevent insulin resistance and its enormous toll on human health.

D. Key Message From Section V

Studies in cell culture systems and in animal models of obesity and diabetes have demonstrated the ability to improve insulin action by modulating the pro-oxidant/antioxidant balance. These provide the most compelling evidence available implicating ROS/RNS in negatively affecting insulin signaling. However, human studies have generally failed to yield evidence sufficiently powerful to include currently available means of pharmacologically manipulating the antioxidant system for improving health, and have even raised safety concerns. Yet, consumption of an antioxidant-rich diet is an accepted approach to promote health. The complexity of the oxidant-antioxidant systems and their positive and negative influences on the insulin signal transduction cascades likely underlie the inconclusive state of this field, despite the enormous investigative effort invested.


ROS/RNS act as a double-edged sword in modulating insulin signaling. On one hand, they are generated in response to insulin and are required for insulin to exert its full physiological actions. On the other hand, ROS/RNS can impair insulin action, and since increased ROS/RNS generation has been proposed to occur in obesity and diabetes, they have been implicated in the pathogenesis of insulin resistance. This seemingly paradoxical function of bioactive agents is actually more the rule than it is the exception in biology. Insulin itself, like most hormones, would impair its own bioactions when present in excess. However, in “free radical biology,” understanding physiological versus pathophysiological functions appears more perplexing. The reasons for this are likely multiple: 1) ROS/RNS are a vast group of agents, each with unique chemical-physical characteristics, interacting with each other to generate yet additional ROS/RNS. Thus, unlike insulin, this is not a single agent whose activity is governed by timing and concentration alone. 2) ROS/RNS are short lived and are difficult to measure. Most existing means of assessing their generation are indirect, and their accuracy is questionable. 3) Biological systems developed complex, interconnected antioxidant machineries that exist in normal physiology in a carefully maintained homeostatic balance. We can assume that physiological function of ROS/RNS is possible as long as homeostasis is maintained, whereas pathophysiology occurs when perturbed, but we currently have at hand limited tools to quantitatively measure the state of the oxidant-antioxidant balance. Hence, ROS/RNS biofunction is influenced by a multidimensional matrix of factors constituting a “biological context.” Its individual components can be partially defined (in which organ, subcellular location, timing and duration, etc.), but its overall output is difficult to assess.

Despite this, it is possible to conclude that physiological function of ROS/RNS in insulin signaling includes the generation of superoxide and/or H2O2 by NADPH oxidase systems, and NO by eNOS. Likely, this regulated RS generation is confined in time and space. Molecular targets for such ROS generation are protein and lipid phosphatases acting within the insulin signaling cascades as shut-off mechanisms. Hence, their temporary, reversible inactivation propagates the insulin signal transduction cascades. Pathological ROS/RNS generation involves multiple ROS/RNS generation systems, RS types, organs and organelles, and is compounded by alterations in the antioxidant defense systems. Molecular targets of such ROS/RNS generation are multiple and include modifications of insulin signaling components (oxidative modifications, phosphorylation as a consequence of the activation of ROS/RNS-sensitive kinases), transcription factors modulating the gene regulation of signaling components and/or regulatory proteins (adipokines), and more generic components of cellular function (cytoskeleton, protein degradation machineries).

Translational research in this field, i.e., the ability to improve insulin action by modulating the oxidant/antioxidant balance, has been disappointing so far. The fundamental features of the field outlined above, particularly, the dual role of ROS/RNS in insulin action, the complexity of the system, and our limited ability to assess it are part of the reason. In addition, high availability of agents perceived as “nontoxic natural antioxidant” at low cost has led to numerous studies, and on the other hand, to lack of financial incentive to support large-scale, high-quality research. But has this field exhausted itself? Our hope is that this review, by discussing both the achievements of research in the field as well as the shortcomings in our current understanding of this complex system, will aid in directing research effort to prove otherwise. One can envision (and hope) that better understanding of components of the biological context of RS function will assist in developing “context-specific” modulators of RS, that would help alleviate the health burden of insulin resistance-associated conditions.


Our own work cited herein was supported by grants from the Israel Science Foundation, the Israeli Ministry of Health, D-Cure Diabetes Care in Israel, and the Leslie and Susan Gonda (Goldschmied) Center for Diabetes Research and Education. N. Bashan is Chair of the Fraida Foundation in Diabetes Research.


We are grateful to Dr. Daniel Konrad for critical reading of this manuscript and for helpful discussions. We apologize to colleagues whose work has not been sufficiently discussed given the vast magnitude of the field.

Address for reprint requests and other correspondence: N. Bashan, Dept. of Clinical Biochemistry, Faculty of Health Sciences, Ben-Gurion University of the Negev, Beer-Sheva 84103, Israel (e-mail: nava{at}