Metabolic dysfunction in obstructive sleep apnea: A critical examination of underlying mechanisms

Abstract

It has recently become clear that obstructive sleep apnea (OSA) is an independent risk factor for the development of metabolic syndrome, a disorder of defective energy storage and use. Several mechanisms have been proposed to explain this finding, drawing upon the characteristics that define OSA. In particular, intermittent hypoxia, sleep fragmentation, elevated sympathetic tone, and oxidative stress – all consequences of OSA – have been implicated in the progression of poor metabolic outcomes in OSA. In this review we examine the evidence to support each of these disease manifestations of OSA as a unique risk for metabolic dysfunction. Tissue hypoxia and sleep fragmentation are each directly connected to insulin resistance and hypertension, and each of these also may increase sympathetic tone, resulting in defective glucose homeostasis, excessive lipolysis, and elevated blood pressure. Oxidative stress further worsens insulin resistance and in turn, metabolic dysfunction also increases oxidative stress. However, despite many studies linking each of these individual components of OSA to the development of metabolic syndrome, there are very few reports that actually provide a coherent narrative about the mechanism underlying metabolic dysfunction in OSA.

INTRODUCTION

Obstructive sleep apnea (OSA) is a common syndrome, affecting approximately 20–30% of men, and 10–15% of women.^1–3 In the last 15 years, many studies have shown that OSA is a risk factor for the development of metabolic syndrome, independent of obesity.^4,5 Metabolic syndrome is marked by the presence of at least three of the following: abdominal obesity, hypertriglyceridemia, low plasma high-density lipoprotein (HDL) levels, hyperglycemia, and elevated blood pressure.⁶ Insulin resistance is a primary manifestation of metabolic syndrome,⁷ and metabolic syndrome is a major risk factor for cardiovascular morbidity and mortality.⁶ The most commonly cited pathways by which OSA is believed to affect metabolism include oxidative stress, sympathetic nervous system activation, tissue hypoxia, and sleep fragmentation (Fig. 1). The purpose of this review is to critically examine these theories and the literature supporting their validity in linking OSA to metabolic syndrome.

Figure 1.

Proposed mechanisms that may underlie metabolic dysfunction in obstructive sleep apnea. ROS, reactive oxygen species.

OXIDATIVE STRESS

Links between OSA and oxidative stress

Reactive oxygen species (ROS) are molecules possessing one or more unpaired electrons, such as superoxide (O₂·⁻), hydrogen peroxide (H₂O₂), or hydroxyl radicals (HO·). Reactive nitrogen species (RNS) describe nitric oxide synthase products, including nitric oxide (NO) or peroxynitrate (OONO⁻). The odd number of electrons in these molecules renders them highly reactive, which is why both ROS and RNS have important roles in both health and disease. ROS are formed from the reduction of oxygen during respiration, or may be generated by ROS-generating enzyme systems. ROS may also enter biological systems from the environment in the form of pollutants, UV radiation, carcinogens, smoking, and infection. They play critical physiological roles in signal transduction, anti-microbial defense, and cell proliferation. However, when present in high levels, ROS interfere with cell structure and/or function, leading to “oxidative stress.”

Some manifestations of oxidative stress include modification of nucleic acids, leading to genetic mutations; oxidation of lipids, which can foster atherosclerosis; or oxidation of proteins, which can alter enzyme function and signaling pathways. To limit excessive ROS, anti-oxidant systems have evolved, including superoxide dismutase and glutathione reductase. In 1956, Harman proposed the free radical theory of aging.⁸ Since then, oxidative stress has been implicated to varying degrees in the development and progression of atherosclerosis, neurodegenerative disorders, diabetes, and cancer.

Given the transient existence of most ROS/RNS, quantitative assessment of oxidative stress is difficult. It is often easier to measure the impact of ROS on their cellular environment. For instance, ROS interact with glutathione, a free radical scavenging protein, changing it from reduced (GSH) to oxidized (GSSG) gluthathione in the process. Inferences about the redox state of a biological system can also be drawn by measuring levels or types of oxidized cellular lipids, protein, or nucleic acids. In addition, investigators may analyze the activity of ROS-generating enzyme systems such as xanthine oxidase, lipooxgenase, or nicotinamide adenine dinucleotide phosphate-oxidase (NADPH) oxidase.

Several studies have demonstrated that OSA patients exhibit elevated markers of oxidative stress. These biomarkers include modified lipids,^9–13 protein,^14,15 nucleic acids,¹⁶ and reduction in antioxidant capacity.¹⁷ In addition, OSA is associated with increased ROS production from leukocytes,^18,19 with attenuation of this phenomenon after CPAP treatment. A thorough cataloguing of these studies appears in expert reviews.^20,21

The mechanism by which OSA induces oxidative stress is unknown, but is commonly ascribed to hypoxia. OSA is characterized by transient falls in hemoglobin saturation during obstructed or flow-limited breathing, followed by recovery of oxygen levels during resumption of effective breathing. This cyclic pattern of hypoxia and re-oxygenation is termed intermittent hypoxia (IH). Chronic IH in rats increases brain oxidative stress²² and NADPH oxidase activity.²³ Chronic IH in mice induces cardiac lipid peroxidation²⁴ in association with left ventricular dysfunction, as well as increased nuclear factor (NF-κB) expression in the aorta, heart and lungs.²⁴ However, other studies either show no significant oxidative stress from IH exposure²⁵ or selective changes in oxidative stress only in certain tissues such as the liver.²⁶ The leading theory of how IH causes oxidative stress contends that “the oscillation of O₂ concentrations during chronic IH remarkably mimics the processes of ischemia/re-oxygenation and could therefore increase cellular production of ROS”.²² Ischemia-reperfusion (I-R) injury is a devastating form of tissue damage often occurring in the setting of thrombosis or transplantation,²⁷ where ROS mediate reperfusion capillary leak and organ dysfunction.²⁸ During the ischemic phase of I-R injury, ATP depletion causes toxic accumulations of metabolites and buildup of intracellular calcium. The high calcium level may be a stimulus for protease activation of xanthine oxidase,²⁹ an enzyme that catabolizes purines and in so doing, generates ROS. During reperfusion, oxygen becomes available as a substrate for xanthine oxidase H₂O₂ production. Influx of activated neutrophils may cause further oxidative injury from NADPH oxidase.³⁰

It is not known whether the IH of OSA in patients with preserved blood flow is sufficient to induce the same pathophysiology as I-R injury. Inadequacy of oxygen to sustain usual metabolic activity (i.e. ischemia) is not evident even when breathing severely hypoxic gases, which has led to a critical view of the term “hypoxia”.²⁷ In addition, sustained hypoxia (i.e. from high altitude exposure) without reoxygenation induces oxidative stress.^31–34 This oxidative stress may be due to a paradoxical increase in the amount of superoxide normally generated by mitochrondria during respiration. This phenomenon has been termed “reductive stress”.³⁵ It might be argued that the distinctions between ROS generation from I-R theory and reductive stress are merely academic. However, there may be practical implications for how sleep-related hypoxemia is addressed, depending upon the nature and origin of ROS in OSA.

Obstructive sleep apnea may also induce oxidative stress by other pathways that have received comparatively little attention. First, OSA dynamically increases circulating substrate levels during sleep, including glucose³⁶ and free fatty acids (FFA).³⁷ These elevations may be related to hypoxia-induced stimulation of the autonomic nervous system and lead to ROS generation through the aforementioned pathways. In particular, elevated FFA induce endothelial dysfunction via vascular ROS³⁸ and induce skeletal muscle insulin resistance and inflammation.³⁹ Second, OSA^40,41 as well as experimental IH exposure in humans⁴² and rodents⁴³ activates the renin-angiotensin system. Angiotensin II has potent effects on blood pressure, inflammation and oxidative stress,⁴⁴ in part through its action on endothelial NADPH oxidase.⁴⁵ ACE inhibitors have been shown to improve insulin signaling in skeletal muscle.⁴⁶ In addition, the fluid retention from elevated renin-angiotensin could be a unifying factor behind the development of OSA and hypertension.⁴⁷ Third, OSA can lead to sleep loss. Sleep deprivation, through as yet poorly understood pathways, induces production of the pro-inflammatory cytokines tumor necrosis factor-α (TNF-α) and interleukin-6 (IL-6), and increases appetite, which can lead to systemic oxidative stress through excessive consumption of calorie-rich foods.⁴⁸ Fourth, obesity is a common risk factor for both OSA and oxidative stress⁴⁹ and may therefore drive the association between the two. In summary, OSA may induce oxidative stress through several possible pathways, although I-R injury is the most widely cited theory.

Links between oxidative stress and metabolism

Oxidative stress could play a role in the development or progression of diabetes or metabolic syndrome.⁵⁰ Hypertension, a key component of metabolic syndrome, may be mediated by suppressed levels of the arterial vasodilator nitric oxide. Superoxide, generated by the stimulation of NADPH oxidase by Angiotensin II,⁵¹ degrades endothelial nitric oxide.⁵² Diets rich in fruits and vegetables^53,54 and antioxidant vitamins⁵⁵ have been shown to reduce blood pressure and oxidative stress. Another key element of metabolic syndrome is obesity. While it is not clear whether oxidative stress induces obesity, it is abundantly clear that obesity is as a pro-inflammatory state. Obese mice exhibit high levels of adipose tissue NADPH oxidase activity, in association with reduced expression of adiponectin and PPAR-γ and elevated TNF-α, reflecting an inflammatory phenotype. Inhibition of NADPH oxidase decreased inflammation and attenuated hyperlipidemia and hyperglycemia as well as liver steatosis.⁵⁶ This study illustrates that ROS can directly alter metabolism through-cross talk with inflammatory pathways in adipose tissue.

One of the primary insults leading to diabetes mellitus is insulin resistance, which is also considered the core defect in metabolic syndrome.⁵⁷ Insulin stimulates glucose uptake in muscle and adipose tissues, inhibits hepatic glucose production, and suppresses lipolysis. Circulating insulin binds its receptor on the surface of insulin-responsive cells. Upon insulin binding, the receptor undergoes auto-phosphorylation of tyrosine residues, which in turn leads to phosphorylation of intracellular insulin receptor substrates (IRS). Tyrosine-phosphorylated IRS proteins bind to Src-homology-2 (SH2) domains including phosphatidylinositol 3′-kinase (PI3K). PI3K in turn phosphorylates downstream messengers, which leads to GLUT-4 translocation (glucose transport), GSK-3 activation (glycogen synthesis), and fatty acid synthesis. The insulin signaling cascade can be interrupted by inflammatory stimuli. For example, TNF-α or increased FFA levels⁵⁸ cause phosphorylation of IRS-1 on serine, instead of tyrosine residues, effectively preventing downstream insulin signal transduction, and IRS-1 binding to the insulin receptor.⁵⁹

Several clinical studies show an association between oxidative stress and insulin resistance.⁶⁰ Interestingly, physiologic levels of ROS (H₂O₂) are generated during insulin stimulation of target cells, and are actually necessary to propagate insulin signaling.^60,61 Paradoxically, high levels of exogenous H₂O₂ inhibit insulin signaling^62–64 and these levels of ROS may occur during in vivo insults such as hyperglycemia.⁶⁰ In fact, chronically elevated glucose and FFA have been proposed as pathways leading to excessive mitochondrial ROS generation and insulin resistance.⁶⁵ However, the more precise mechanisms by which ROS impair insulin signaling are unclear. Some of the proposed pathways are reviewed elsewhere⁶⁶ and include (a) activation of Ser/Thr “stress” kinases such as NF-κB or p38-MAPK, which cause inappropriate phosphorylation of the insulin receptor or IRS; (b) disruption of cellular compartmentalization of insulin signaling elements; (c) reduced GLUT-4 transcription; and (d) impaired mitochondrial oxidation of fatty acids.⁶⁷ A second “hit” in the pathogenesis of diabetes mellitus is pancreatic β-cell failure. ROS have been proposed as mediators of β-cell failure from excessive glucose and fatty acid loads.⁶⁸ The hyperglycemia of diabetes induces oxidative stress, which may mediate devastating neurovascular consequences.^69,70 Hence, a vicious cycle may be operant in which oxidative stress leads to diabetes, which in turn causes further oxidative stress.

A practical question is whether treating oxidative stress can improve glucose homeostasis. The data are limited, but small studies involving short-term administration of antioxidants have shown improved glucose disposal and/or insulin action.^71–74 However, no long-term, large-scale studies have been published that demonstrate the efficacy of antioxidants for metabolic syndrome. Nutrition guidelines by the American Diabetes Association do not recommend antioxidant supplements⁷⁵ and enthusiasm for the use of antioxidants has been tempered by disappointing trials for the primary and secondary prevention of cardiovascular diseases.⁷⁶ Whether this lack of evidence should be construed as proof against the role of ROS in diabetes is controversial,⁷⁷ reflecting the larger debate about the free radical theory of aging and disease.⁷⁸

Evidence that metabolic dysfunction is mediated by oxidative stress in OSA

Several studies demonstrate that elements of metabolic syndrome, such as hypertension and insulin resistance, improve with CPAP treatment⁷⁹ while other studies show that CPAP treatment attenuates oxidative stress.²⁰ One month after CPAP treatment, Murri et al. showed a decrease in blood pressure that correlated with improvements in oxidative stress.⁸⁰ However, these studies do not directly prove a functional role of oxidative stress in OSA.

One approach to demonstrate that oxidative stress is a mediator of OSA consequences is to examine the physiological or metabolic impact of antioxidant therapies in OSA. Intravenous administration of vitamin C to patients with OSA acutely normalized vascular responses to acetylcholine⁸¹ and improved flow-mediated dilation (FMD) of the brachial artery.⁸² Allopurinol, a xanthine oxidase inhibitor, also improved FMD compared to placebo in patients with OSA, in association with a lowering of plasma lipid peroxides.⁸³ A diet enriched in fruits and vegetables led to weight loss and reductions of blood pressure in patients with OSA, albeit without a detectable change in antioxidant levels measured using a ferric-reducing/antioxidant power (FRAP) assay.⁸⁴ Interestingly, the antioxidant N-acetylcysteine (NAC) improved lipid peroxidation, and improved OSA itself.⁸⁵

Overall, there is compelling evidence that ROS may play a role in OSA-mediated endothelial dysfunction, but the role of oxidative stress in mediating other OSA consequences is decidedly less clear.

SYMPATHETIC NERVOUS SYSTEM

Links between OSA and sympathetic activation

The sympathetic nervous system (SNS) is responsible for eliciting adaptive responses to stressful stimuli. Physiologic stress stimulates the locus coeruleus in the hypothalamus,^86,87 which instigates a variety of “fight-or-flight” reflexes, such as an increase in cardiac output, and redistribution of blood flow to skeletal muscles. In terms of metabolism, the SNS orchestrates a host of tissue signals that result in a net efflux of glucose and FFA into the bloodstream. Some of these signals arise from sympathetic efferent fibers acting upon adipose tissue, liver, and skeletal muscle. In addition, the adrenal medulla secretes catecholamines into the plasma, which are carried to target tissues. In contrast to these rapid effects of SNS stimulation, other responses to stress are activated more gradually, including the hypothalamic-pituitary-adrenal (HPA) axis, which mediates the secretion of corticotropin-releasing hormone, pituitary adrenocorticotropic hormone, and ultimately corticosteroids produced in the adrenal cortex.^88,89

OSA has been firmly linked to SNS activation, confirmed by measurements of muscle sympathetic nerve activity and elevated urine and plasma catecholamines^90–94 with normalization after CPAP.^90,95 One of the hallmarks of OSA that increases SNS activity is IH. IH stimulates chemoreflexes in the carotid body that in turn stimulate the SNS,^96,97 with evidence of increased efferent stimulation in cervical, renal, splanchnic, thoracic, and lumbar sympathetic nerves.^96,98–101 In rodents, IH activates the sympathetic nervous system and increases catecholamine efflux by the adrenal medulla.^{96,97,102–113} Besides IH, OSA also can cause arousals from sleep and hypercapnia, and these stimuli have interactive effects on sympathetic signaling. For example, hypoxia and hypercapnia augment SNS activity in a synergistic manner.^114,115 In experimentally-induced apneas in sleeping dogs, hypoxia and sleep arousals showed additive effects on SNS activity and systemic blood pressure.¹¹⁶

What are the functional consequences of IH-induced SNS activation? Healthy humans exposed to acute IH develop insulin resistance.¹¹⁷ Qualitatively similar findings occur in subjects exposed to acute sustained hypoxia.¹¹⁸ Moreover, acute hypoxia-induced insulin resistance is attenuated by the sympatholytic drug clonidine.¹¹⁹ It is not yet clear whether long-term IH also induces glucose intolerance or insulin resistance via SNS activity. In rats, IH causes a time-dependent increase in blood pressure. This elevation in blood pressure can be abolished by sympathetic denervation with 6-hydroxydopamine, denervation of the carotid bodies or sympathetic nerves, or adrenal medullectomy.^110,120 However, it remains unclear whether IH is the primary driver of SNS activation and hypertension in human OSA. Exposure of humans to 2 weeks of IH increases daytime blood pressure¹²¹ in association with increased SNS activity, but preventing hypoxemia in OSA with supplemental oxygen does not lower blood pressure, although CPAP does.¹²²

Links between the sympathetic nervous system and metabolism

Catecholamines are considered counter-regulatory hormones since they oppose the anabolic, energy-conserving, and glucose utilizing role of insulin. Catecholamines stimulate glucagon secretion, activate glycogenolysis and gluconeogenesis in the liver, and cause breakdown of muscle glycogen and adipose tissue triglycerides. They also inhibit insulin secretion and insulin-mediated glucose uptake by skeletal muscle.^123–125 The tight coupling of SNS and metabolism is particularly evident in the anatomy of liver tissue. In the liver, sympathetic nerve fibers travel with the hepatic artery and portal vein, branching into smooth muscle layers to reach hepatic lobules.^126–130

Sympathetic nervous system activation is known to rapidly alter glucose homeostasis. For decades it has been recognized that epinephrine, the prototypical stress hormone, causes insulin resistance in humans.¹³¹ Norepinephrine and epinephrine also inhibit insulin secretion from pancreatic islets^132–136 and blunt the usual stimulation of insulin secretion by glucose.^134–137 At least in some species, insulin secretion may be under tonic inhibition by this pathway: Mice lacking α_2A-adrenoreceptors have elevated plasma insulin and reduced blood glucose levels, as well as improved glucose tolerance.¹³⁶ SNS signaling, predominantly via α-adrenoreceptors, controls hepatic glucose output.^{128,138–140} SNS activation also has potent effects on lipid metabolism. Lipolysis is chiefly activated by stimulation of β-adrenoreceptors located on the surface of adipocytes. A signaling cascade leads to an elevation of intracellular cAMP, followed by hydrolysis of intracellular triglycerides and release of FFA, which serve as an important source of fuel for oxidation, particularly in the skeletal muscle and heart.

Although the SNS is critical for survival during life-threatening stress, excessive SNS activity may have deleterious consequences. For example, the SNS was implicated in the development of hypertension many decades ago, which led to the development of sympatholytic drugs for lowering blood pressure. Recently, catheter-based renal sympathetic denervation has been offered to treat poorly controlled hypertension.^141,142 Brotman and Girod envision a duel between insulin and counter-regulatory hormones as a “tug of war with no winner,” culminating in metabolic syndrome.¹⁴³ It is theoretically plausible that repetitive episodes of insulin resistance could ultimately progress towards pancreatic β-cell failure and type 2 diabetes. In particular, excessive lipolysis with an over-abundance of FFA can lead to systemic “lipotoxicity,” which refers to the ectopic accumulation of lipids in skeletal muscle and liver.^144–147 Norepinephrine infused into dogs caused acute insulin resistance, high plasma FFA levels, and acute fatty degeneration of the liver, which “was yellow and cut like butter” within 48 h.¹⁴⁸ More physiologic elevations of FFA have been implicated in skeletal muscle insulin resistance, where FFA and their metabolic intermediates inhibit tyrosine phosphorylation of insulin receptor substrate 1 (IRS-1), which in turn decreases insulin-stimulated glucose transport.^149,150 In the liver, increased FFA flux stimulates gluconeogenesis¹⁵¹ and the assembly of triglyceride rich very low density lipoproteins (VLDL) and VLDL secretion to the bloodstream, resulting in dyslipidemia.¹⁵²

In summary, acute stimulation of the SNS can induce multiple physiologic changes that resemble metabolic syndrome. Chronic stress and elevated SNS activity have been proposed as causal factors in the development of metabolic syndrome.^89,153,154

Evidence that metabolic dysfunction is mediated by SNS activity in OSA

Evidence for the role of the SNS in mediating OSA consequences is mostly circumstantial. For example, CPAP decreases both sympathetic activity and blood pressure.^{90,95,122,155} Exposure of healthy humans to IH causes insulin resistance with increased sympathetic activity.¹¹⁷ Patients with OSA have elevated plasma FFA levels, perhaps via SNS-mediated lipolysis.^37,156

TISSUE HYPOXIA

Links between OSA and tissue hypoxia

Intermittent hypoxia is a defining characteristic of OSA.¹⁵⁷ The nadir arterial oxyhemoglobin saturations seen in OSA may be quite low, and are typically lower than in other pulmonary illnesses. In some cases, the hypoxemia of OSA can lead to cor pulmonale. As such, considerable importance has also been attached to tissue hypoxia as a mechanism that may underlie metabolic dysfunction in OSA.

No studies have directly measured tissue oxygen tension in patients with OSA. Rather, inferences about inadequate tissue oxygenation have been made using indirect evidence. For example, two case reports describe acute elevation of liver enzymes in OSA; in these cases, severe nocturnal desaturations in OSA were thought to have caused ischemic liver damage.^158,159 Another study reported higher morning creatine phosphokinase (CPK) levels in OSA patients, which were then lowered with CPAP treatment.¹⁶⁰ These results suggest that patients with OSA may experience liver and muscle damage during sleep, but do not specifically prove that tissue ischemia is the underlying mechanism. Morning lactate levels are modestly elevated in OSA.^161,162 During exercise, subjects with OSA show impaired aerobic and glycolytic capacity, a finding similar to that seen at high altitude.¹⁶³ In addition, OSA severity correlates modestly with hematocrit level,¹⁶⁴ although the disease rarely leads to polycythemia.^165,166 Taken together, these findings suggest that there may be patients with OSA who have inadequate tissue oxygenation but more invasive assessments would be needed to confirm this hypothesis.

Translational models of OSA using IH exposure in rodents have provided some insight about the potential burden of tissue hypoxia in OSA. Reinke et al. measured tissue oxygen levels during exposure of mice to IH. As arterial oxygen levels fluctuated between baseline and a nadir saturation of 60%, oxygen tension in fat and liver also varied with the same periodicity.¹⁶⁷ Interestingly, obese mice showed an even greater degree of tissue hypoxia, suggesting possible interactions of hypoxia and obesity. Similar reductions in oxygen tension were reported during extrinsic airway obstruction in rats.¹⁶⁸ From these animal studies, it is apparent that IH can rapidly decrease oxygen tension in tissues, and a similar phenomenon might occur in human OSA. Whether these transient reductions in tissue oxygen are sufficient to cause anoxic or ischemic injury is unclear.

Tissue hypoxia can also be assessed by examining pathways of cellular adaptation to hypoxia. One might surmise that tissue hypoxia could be accompanied by activation of mediators of a global cellular hypoxic response, such as hypoxia inducible factor-1 (HIF-1). HIF-1 is a heterodimer that consists of a constitutively expressed β subunit and an O₂-regulated α subunit.¹⁶⁹ HIF-1α activation by sustained hypoxia occurs due to the inhibition of O₂-dependent prolyl hydroxylation,¹⁷⁰ thereby preventing ubiquitination and proteasomal degradation. An alternative pathway of ROS-mediated accumulation of HIF-1α has been demonstrated in adrenal cells during in vitro IH.¹⁷¹ HIF-1 regulates a multitude of cellular functions, including metabolism, angiogenesis, and cellular growth.¹⁷² Few studies have actually examined HIF-1 activation in OSA. One such study examined gene expression profiles of skin biopsy samples from OSA patients, mouse aortas from animals exposed to 4 weeks of IH, and human dermal microvascular and coronary artery endothelial cells cultured in IH.¹⁷³ Severely hypoxic OSA patients had increased skin HIF-1α expression relative to less hypoxic patients, but there was no increase in mouse aorta HIF-1α expression in IH, and of the two cell types examined, only coronary artery endothelial cells showed an increase in HIF-1α. Another study reported no increase in HIF-1 activity in OSA patients relative to a control population, as measured by DNA binding assay.¹⁷⁴ However, HIF-1α expression increased in the liver and lung tissues of mice after chronic IH,¹⁷⁵ and in vitro exposure of cells to IH has mirrored this finding.¹⁷⁶ Whether these experimental paradigms are realistic simulations of cellular hypoxia in OSA remains to be shown. Still, these results do suggest that HIF-1α activation depends upon factors including tissue, species, as well as the nature and severity of hypoxia.

Links between tissue hypoxia and metabolism

Hypoxia has complex effects on metabolism, many of which are mediated by the central nervous system. Hypobaric hypoxia encountered at high altitude leads to anorexia, weight loss, and SNS activation with associated insulin resistance.¹⁷⁷ In many mammals, hypoxia causes a decrease in the thermoregulatory set point, leading to a fall in body temperature.¹⁷⁸ Therefore, hypoxia at relatively cool ambient temperatures can lead to a dramatic fall in metabolic rate, deceleration of tissue lipid uptake, and the development of hyperlipidemia¹⁷⁹ – a finding not seen at warmer ambient temperature.¹⁸⁰ Besides eliciting these systemic responses, hypoxia also can directly affect cellular metabolism at the tissue level. For the purposes of this review, we will focus on tissue-level hypoxia, since this is a commonly cited hypothesis connecting OSA to metabolic dysfunction.

There is mounting evidence that oxygen tension in adipose tissue may have important metabolic consequences. It has been proposed that HIF-1 plays a role in the development and progression of metabolic syndrome, since the oxygen tension of adipose tissue in obese humans^181,182 and mice¹⁸³ is decreased. This reduced oxygen tension may be sufficient to induce HIF-1 expression. However, HIF-1 may also be upregulated in obesity by high levels of insulin.¹⁸⁴ Regardless of whether adipose HIF-1 is stimulated by hypoxia or insulin, the consequences of HIF-1 activation have been demonstrated by both gain-of-function and loss-of-function studies: Overexpression of HIF-1α in adipose tissues causes weight gain,¹⁸⁵ while inhibition of HIF-1 leads to weight loss in mice fed a high-fat diet.¹⁸⁶ Adipose tissue hypoxia may also have other consequences. In vitro exposure of cultured adipocytes to IH significantly reduced secretion of adiponectin,¹⁸⁷ a hormone that fosters insulin sensitivity and weight loss. Additionally, cultured adipocytes increase expression of the glucose transporter GLUT-1 in hypoxia¹⁸⁸ and decrease expression of the insulin-sensitive transporter GLUT-4,¹⁸⁹ suggesting that hypoxia, via HIF-1 activation, may cause impaired glucose handling in adipocytes.

Hypoxia in the liver may play a role in the progression of non-alcoholic fatty liver disease (NAFLD), the hepatic manifestation of metabolic syndrome.¹⁹⁰ Baze et al. examined hepatic gene expression in mice by microarray, in response to chronic continuous hypoxia of increasing severity. Their team found a difference in the expression profile in hypoxia in genes associated with immune response, angiogenesis, oxygen transport, glycolysis, and carbohydrate, fatty acid and lipid metabolism.¹⁹¹ Hypoxia led to impaired glucose metabolism and induced liver injury in PTEN knockout mice, a transgenic animal that develops steatohepatitis.¹⁹² Hypoxia was also found to increase the expression of several lipogenic genes, such as PPAR-γ (peroxisome proliferator activated receptor-gamma), SREBP-1c (sterol regulatory element binding protein-1c), and ACC1/2 (acetyl CoA carboxylase 1 and 2); and was associated with decreased expression of mitochondrial beta oxidation genes. Hepatic HIF-1 activation is necessary for the development of liver fibrosis in a mouse model of NAFLD using bile duct ligation,¹⁹³ and is involved in the activation of hepatic Kupffer cells¹⁹⁴ and stellate cells¹⁹⁵ in the development of hepatic fibrosis.

Hypoxia may be involved in the development of atherosclerosis. Atherosclerosis was accelerated in rabbits exposed to sustained chronic hypoxia.¹⁹⁶ In fact, arterial wall hypoxia has been hypothesized as an inciting event in early atheroma formation.¹⁹⁷ In humans, OSA is associated with early signs of atherosclerosis, and OSA severity correlated with the severity of vascular abnormalities.¹⁹⁸ Moreover, CPAP reverses early signs of atherosclerosis in patients with OSA.¹⁹⁹ Chronic IH also accelerates atherosclerosis in mice. This finding was evident at early stages of plaque formation²⁰⁰ and at advanced stages, assessed in mice deficient in ApoE (a lipoprotein necessary for catabolism of cholesterol).²⁰¹ The increased atherosclerosis was associated with hypertension, vascular stiffness, and dyslipidemia, and could be attenuated by correcting the lipid transport defect caused by IH.²⁰² It remains to be seen whether human OSA exacerbates atherosclerosis by these pathways examined in translational IH models.

However, not all studies suggest that hypoxia is harmful for metabolic health. In fact, hypoxic exposures have been used to treat obesity and to improve body composition.²⁰³ Hypoxia was found to improve glucose metabolism in some studies,^204,205 suggesting that tissue-level hypoxia may actually have insulin-sensitizing effects that prevail over SNS-mediated insulin resistance. Mice exposed to hypoxia for 4 weeks displayed marked improvement in insulin sensitivity, contrasting with the acute effects of hypoxia.²⁰⁶ Authors speculated that improvements in skeletal muscle blood flow, perhaps induced by endothelial nitric oxide, mediated the observed improvements in glucose disposal. Acute hypoxia was also shown to improve, rather than worsen, the plasma lipid profile in mice housed at thermoneutrality.¹⁸⁰ This reduction in plasma lipids was ascribed to hypoxia lowering hepatic VLDL secretion, and increasing fatty acid uptake in the heart.

Evidence that metabolic dysfunction is mediated by tissue hypoxia in OSA

Several studies have shown an association between metrics of hypoxia during sleep and metabolism. For example, in the Sleep Heart Health Study, sleep related hypoxemia was associated with glucose intolerance after controlling for multiple confounding variables.²⁰⁷ However, it is difficult to study the effects of OSA-related hypoxia in complete isolation from other aspects of the disorder. Some of this information can be gleaned from studies where OSA was treated with supplemental oxygen rather than CPAP. Several studies have shown that supplemental oxygen alone is effective in reducing the magnitude of nocturnal desaturations in OSA, although it does not correct the syndrome itself (reviewed in Mehta et al.²⁰⁸). A few studies have examined the effect of oxygen therapy on blood pressure;^209–211 two of these studies showed no effect,^210,211 and the third showed a modest reduction in diastolic blood pressure relative to patients placed on supplemental air.²⁰⁹ One larger recent study has examined the effect of CPAP or supplemental oxygen on blood pressure in patients with moderate to severe OSA (AHI 15–50 events/h).¹²² Supplemental oxygen did not alter mean arterial pressure. CPAP therapy, however, caused a 2.4 mm Hg decrease in mean arterial pressure relative to untreated patients – a finding replicated in several other studies.²¹² These findings call into question whether hypoxia is the major factor that leads to hypertension in OSA. Other features of OSA, such as repetitive arousals, hypercapnia, and shifts in intrathoracic pressure, may be more important. Currently, there are no studies that have examined any other metabolic outcomes in OSA when comparing supplemental oxygen to CPAP.

SLEEP FRAGMENTATION

Links between sleep fragmentation and metabolism

Fragmented sleep is another of the cardinal manifestations – indeed, one of the defining characteristics – of sleep apnea.¹⁵⁷ There is ample evidence, both in animal models and in human subjects, that sleep fragmentation results in insulin resistance and hypertension.

A few studies have examined the role of acute sleep fragmentation in the development of metabolic dysfunction in humans. One study used mechanical or auditory stimuli to fragment sleep over two nights in healthy, nonobese volunteers.²¹³ After sleep fragmentation, the volunteers showed worsened glucose handling and insulin sensitivity, as measured by intravenous glucose tolerance test. Similar results were seen in patients after slow wave sleep deprivation.²¹⁴ Auditory sleep fragmentation also blunted the physiologic decrease in blood pressure during sleep.²¹⁵ Mechanisms by which sleep fragmentation induces these effects are unknown but are likely related to physiologic changes in autonomic tone that occur between sleep and wake states.²¹⁶

The role of chronic sleep fragmentation in the development of metabolic syndrome is unclear. One cross-sectional study of an elderly population found that each standard deviation increase in fragmented sleep, assessed with actigraphic recording, was associated with an increase in body mass index of 0.59 kg/m².²¹⁷ Frequent arousals from sleep in snorers without OSA was also found to be associated with hypertension.²¹⁸

Animal models of sleep fragmentation have been developed, often using mechanical disruption to prevent REM sleep^219–221 or manipulation of the airway or environment to simulate obstructive apneas.^222–224 In one study, rats were subjected to either restricted sleep,” or “disturbed sleep,” imposed by timed revolutions of a wheel in which the animals were housed. Both protocols caused hyperglycemia and decreased insulin levels during an IV glucose tolerance test.²²⁵ Longer durations of sleep fragmentation (up to 20 weeks) have also been shown to induce insulin resistance, in association with activation adipose tissue inflammation and oxidative stress.²²⁶ Sleep fragmentation in mice also caused glucose intolerance and increased food intake after 2 weeks,²²⁷ and obesity after 8 weeks.²²⁸

Evidence that metabolic dysfunction is mediated by sleep fragmentation in OSA

There is little evidence to support a direct link between sleep fragmentation and the development of specific manifestations of metabolic syndrome in patients with OSA. As noted above, CPAP therapy appears to improve some metabolic outcomes in OSA,⁷⁹ and CPAP clearly decreases fragmented sleep, but the effects of CPAP are myriad, so ascribing the metabolic improvements to less fragmented sleep specifically is impossible.

CONCLUSION

In conclusion, OSA is believed to affect metabolism by several pathways including oxidative stress, SNS activation, tissue hypoxia, and sleep fragmentation. However, the relative contribution of each of these mechanisms to metabolic dysfunction in OSA is unclear and probably varies considerably from patient to patient depending on a disease phenotype. Future investigation may focus on specific disease phenotypes, which lead to the development of poor metabolic outcomes, and on utilizing these phenotypic markers to optimize future therapeutic strategies and outcomes.