Metabolic Consequences of Sleep-Disordered Breathing

Abstract

There is increasing evidence of a causal relationship between sleep-disordered breathing and metabolic dysfunction. Metabolic syndrome (MetS), a cluster of risk factors that promote atherosclerotic cardiovascular disease, comprises central obesity, insulin resistance, glucose intolerance, dyslipidemia, and hypertension, manifestations of altered total body energy regulation. Excess caloric intake is indisputably the key driver of MetS, but other environmental and genetic factors likely play a role; in particular, obstructive sleep apnea (OSA), characterized by intermittent hypoxia (IH), may induce or exacerbate various aspects of MetS. Clinical studies show that OSA can affect glucose metabolism, cholesterol, inflammatory markers, and nonalcoholic fatty liver disease. Animal models of OSA enable scientists to circumvent confounders such as obesity in clinical studies. In the most widely used model, which involves exposing rodents to IH during their sleep phase, the IH alters circadian glucose homeostasis, impairs muscle carbohydrate uptake, induces hyperlipidemia, and upregulates cholesterol synthesis enzymes. Complicating factors such as obesity or a high-fat diet lead to progressive insulin resistance and liver inflammation, respectively. Mechanisms for these effects are not yet fully understood, but are likely related to energy-conserving adaptations to hypoxia, which is a strong catabolic stressor. Finally, IH may contribute to the morbidity of MetS by inducing inflammation and oxidative stress. Identification of OSA as a potential causative factor in MetS would have immense clinical impact and could improve the management and understanding of both disorders.

Introduction

Metabolism and Metabolic Syndrome

Metabolism refers to the whole range of biochemical processes in a living organism that produce energy and basic materials needed for important life processes, including cell growth, reproduction, response to environment, and maintenance of cell structure and integrity. The energy that all obligate aerobes require for survival is derived from the reduction of molecular oxygen (O₂) by nutrients to yield carbon dioxide (CO₂) and water:

$$\text{nutrients}+{\mathrm{O}}_2\to {\text{CO}}_2+{\mathrm{H}}_2\mathrm{O}$$

The modern Western diet has brought the cell into contact with an abundance and composition of nutrients not encountered by early humans—and with them diseases such as obesity, diabetes mellitus, atherosclerosis, and steatohepatitis. As the prevalence of these diseases has grown exponentially in the last century, scientists have identified several interrelated risk factors that appear to promote atherosclerotic cardiovascular disease: abdominal obesity, dyslipidemia, hypertension, and elevated plasma glucose (Grundy et al. 2006). Thirty years ago Hermann Haller (1977) coined the term “metabolic syndrome” (MetS¹) to collectively describe these risk factors, which had hitherto gone by several other names. A decade later, Gerald Reaven (1988) proposed insulin resistance as the cornerstone of MetS, which is now widely recognized as a leading risk factor for cardiovascular morbidity and mortality.

Diagnosis of MetS requires the presence of three or more of the following characteristics (NIH 2002):

• Abdominal obesity (waist circumference >40" in men and 35" in women)

• Serum triglycerides >150 mg/dL

• Cholesterol (HDL-C) <40 mg/dL in men and <50 mg/dL in women

• Blood pressure (BP) >130/85 mmHg

• Fasting glucose >110 mg/dL

Obstructive Sleep Apnea and Metabolic Syndrome

Many studies show that untreated obstructive sleep apnea (OSA¹) increases the risk of cardiovascular disease and mortality (Marin et al. 2005; Marshall et al. 2008; Mooe et al. 2001; Shahar et al. 2001; Valham et al. 2008; Yaggi et al. 2005; Young et al. 2008). Given the association between OSA and cardiovascular disease, it is reasonable to speculate that OSA may promote MetS, and one study has indeed proven that it is the cause of a key aspect of MetS, systemic hypertension (Peppard et al. 2000). An analogous link may exist between OSA and other features of MetS, but clustering of OSA with metabolic risk factors makes establishing such a connection problematic. For example, estimates indicate that 70% of patients with OSA are obese. Conversely, 40–90% of obese patients have OSA (Daltro et al. 2007; Vgontzas et al. 1994; Young et al. 1993, 2002).

Because the precise mechanisms regulating the metabolic consequences of OSA remain unclear, numerous studies have used animal models of OSA to further scientists’ understanding of OSA-MetS relationships.² In this review, we look at the role of OSA in diabetes mellitus (DM¹), dyslipidemia, obesity, and nonalcoholic fatty liver disease (NAFLD¹), all of which are associated with metabolic syndrome (Browning and Horton 2004; Marchesini et al. 2001). We then discuss animal models of OSA and present findings from these models related to components of MetS.

Possible Effects of Obstructive Sleep Apnea: Clinical Evidence

Diabetes

In a number of cross-sectional, prospective, and interventional trials over the past 15 years, scientists have attempted to determine whether OSA is an independent risk factor for diabetes mellitus. Several cross-sectional studies provide evidence for an association of OSA with glucose intolerance (Table 1). Nigel McArdle and colleagues (2007) studied 42 male sleep clinic patients matched for age, body mass index (BMI,¹ a measure of body fat), smoking, and OSA status; they calculated insulin resistance according to the homeostasis model assessment (HOMA) with the formula HOMA = fasting serum insulin (μU/ml) × fasting blood glucose (mmol/L)/22.5. The men with OSA had worse insulin resistance and higher plasma epinephrine levels, as well as lower insulin-like growth factor-1, which is predictive of diabetes risk (Sandhu et al. 2002).

Table 1.

Selected clinical studies relevant to OSA and diabetes^a

Type	Methods	Findings	Author
Case-control	Nondiabetic men and women screened with PSG for OSA; HOMA-IR	AHI correlates with degree of IR	Ip et al. 2002
	Males with OSA vs. sleep clinic controls matched for BMI, age, smoking; HOMA-IR	OSA associated with higher IR, plasma epinephrine, lower IGF-1	McArdle et al. 2007
	Obese male nondiabetics from community; OGTT	AHI and oxygen saturation associated with IR	Punjabi et al. 2002
	Nondiabetic subjects; IVGTT	OSA associated with impairments in insulin sensitivity and pancreatic β cell function	Punjabi and Beamer 2008
	Male/female overweight OSA subjects; HEC	BMI rather than AHI primary driver of IR	Stoohs et al. 1996
	Male/female snoring children vs. nonsnoring controls; HOMA-IR	BMI rather than AHI primary driver of IR	Tauman et al. 2005
Prospective	Snoring vs. nonsnoring nurses followed 10 years for development of DM	Adjusted OR for DM in regular snorers: 2.03	Al Delaimy et al. 2002
	Snoring vs. nonsnoring men followed 10 years for development of DM	OR for DM in obese snorers: 7.0; in obese nonsnorers: 5.1	Elmasry et al. 2000
	Baseline prevalence and 4-yr incidence of DM in OSA	OR for baseline DM: 2.3 (AHI>15); no increased 4-year risk	Reichmuth et al. 2005
Intervention	Diabetics with OSA before/after CPAP; 72-hour glucose; HgA1c	CPAP improved HgA1c and postprandial glucose	Babu et al. 2005
	OSA vs. matched controls; serum glucose and insulin	OSA with excessive daytime sleepiness had elevated fasting glucose and insulin	Barceló et al. 2008
	Obese diabetics with OSA before/after CPAP 4 months; HEC	CPAP improved insulin responsiveness	Brooks et al. 1994
	OSA and snorers vs. controls before/after CPAP for OSA group; fasting insulin	OSA subjects did not have elevated fasting insulin	Davies et al. 1994
	Children with OSA before/after tonsillectomy; fasting insulin and glucose	IR only in obese subjects; improved after tonsillectomy	Gozal et al. 2008
	Diabetics with OSA before/after CPAP 2 days/3 months; HEC	CPAP improved insulin responsiveness at 2 days and 3 months	Harsch et al. 2004
	Obese nondiabetics before/during CPAP; continuous nocturnal glucose and insulin	CPAP did not affect glucose or insulin profiles	Saini et al. 1993
	Sleep fragmentation (SWS deprivation) induced in healthy young adults	SWS deprivation caused IR and inappropriately low insulin release	Tasali et al. 2008

^aAHI, apnea-hypopnea index; BMI, body mass index; CPAP, continuous positive airway pressure; DM, diabetes mellitus; HEC, hyperinsulinemic euglycemic clamp; HOMA-IR, homeostasis model assessment of insulin resistance; IGF-1, insulin-like growth factor-1; IR, insulin resistance; IVGTT, intravenous glucose tolerance test; OGTT, oral glucose tolerance test; OR, odds ratio; OSA, obstructive sleep apnea; PSG, polysomnography; SWS, slow-wave sleep

Investigators in Hong Kong (Ip et al. 2002) compared the degree of insulin resistance with BMI and severity of OSA among 185 patients. Using multiple linear regression, they found that obesity was the primary determinant of insulin resistance, but the patients’ apnea-hypopnea index (AHI¹) and minimal arterial O₂ saturation were also significant contributors. Another study reported similar findings in mildly obese men; for every 4% reduction in saturation of peripheral oxygen (SpO₂¹), the odds ratio for glucose intolerance doubled (Punjabi et al. 2002). In the Sleep Heart Health Study (Punjabi et al. 2004), among 2656 nondiabetic patients AHI and average O₂ saturation were associated with elevated fasting glucose and impaired oral glucose tolerance. More recently, Naresh Punjabi and colleagues used an intravenous glucose tolerance test (IVGTT) in 118 nondiabetic subjects and found that AHI and severity of nocturnal oxyhemoglobin desaturation were associated with both impaired insulin sensitivity and pancreatic β cell dysfunction (Punjabi and Beamer 2009). In severely obese patients, insulin resistance was associated not only with the obesity but also with the severity of nocturnal hypoxemia (Polotsky et al. 2009). But not all studies are in agreement on this subject. Riccardo Stoohs and colleagues (1996) screened 50 healthy middle-aged adults for OSA and found that those with an AHI greater than 10 had significantly more insulin resistance. However, when corrected for multiple variables, BMI remained the sole determinant of insulin resistance. Another study reported similar dependence of insulin resistance on adiposity rather than OSA severity in children (Tauman et al. 2005).

OSA may also be a risk factor for the development of DM, according to two frequently cited studies of snoring. In one study, snoring Swedish men were more likely (5.4% vs. 2.4%) to develop diabetes over 10 years than their nonsnoring counterparts (Elmasry et al. 2000). In the US Nurses’ Health Study of more than 69,000 women, regular snorers were twice as likely to develop DM independent of BMI, smoking, or family history (Al Delaimy et al. 2002). As snoring is not diagnostic for OSA, Kevin Reichmuth and colleagues (2005) followed the incidence of DM in patients who underwent a full sleep study. Although there was a higher prevalence of DM in the OSA group at baseline (14.8% vs. 2.8%), its incidence over the next 4 years was no greater than in nonapneics.

Researchers have conducted therapeutic trials to consider whether treatment of OSA can improve insulin resistance; an extensive review is available (Punjabi and Polotsky 2005), so we present a few illustrative examples. Using a continuous 72-hour glucose monitoring system, Ambika Babu and colleagues (2005) showed that continuous positive airway pressure (CPAP¹) treatment lowered postprandial glucose levels and caused a 0.6% fall in hemoglobin A1c in diabetic OSA subjects, and greater decreases occurred with the use of CPAP for more than 4 hours per night. The gold standard to quantify insulin resistance is the hyperinsulinemic euglycemic clamp, which delivers simultaneous infusions of insulin and glucose; the insulin infusion is at a fixed rate, while glucose is titrated to keep serum glucose in a narrow range (DeFronzo et al. 1979). In obese diabetic patients with moderate to severe OSA, CPAP for 4 months improved insulin responsiveness measured by the clamp technique (Brooks et al. 1994), and another study found that only 2 days of CPAP were enough to enhance insulin sensitivity by clamp measurement (Harsch et al. 2004). Interestingly, nonobese subjects benefited the most from treatment, whereas in a pediatric population only obese subjects with impaired glucose tolerance improved with tonsillectomy (Gozal et al. 2008). In a few negative studies, CPAP for 1 night (Saini et al. 1993) or for 3 to 6 months did not improve fasting glucose or insulin (Davies et al. 1994; Ip et al. 2000). A different perspective was provided by Antonia Barceló and colleagues (2008), who showed in a study of 44 patients with OSA that only in those suffering from excessive daytime sleepiness did insulin resistance improve with CPAP.

Other sleep research may provide additional evidence of a connection between OSA and insulin resistance. First, OSA involves sleep fragmentation, which can independently affect glucose metabolism. Esra Tasali and colleagues (2008) subjected healthy young volunteers to 3 nights of sleep fragmented by auditory stimuli and evaluated the effects on glucose tolerance and insulin secretion. They found marked insulin resistance and increased sympathetic activity. Second, sleep fragmentation interrupts the normal secretion of growth hormone, which occurs predominantly during slow-wave sleep (Van et al. 2004). Growth hormone has antagonistic effects on insulin activity, so its suppression might actually improve glucose profiles; but it also stimulates the hepatic production of insulin-like growth factor-1 (IGF-1), which in turn promotes insulin sensitization and pancreatic β cell development and function (Yuen and Dunger 2007). Indeed, both growth hormone and IGF-1 levels are low in OSA (Aydogan et al. 2007; Meston et al. 2003). Third, the hypothalamic-pituitary axis may be hyperactive in OSA, producing elevated cortisol states (as exemplified by Cushing’s syndrome), which share many features of MetS (Iwasaki et al. 2008; Weaver 2008). OSA subjects showed increased 24-hour cortisol (Vgontzas et al. 2007) and impaired dexamethasone suppression of salivary cortisol (Carneiro et al. 2008), effects that CPAP reversed. These observations are consistent with the fact that cortisol secretion is pulsatile with each sleep arousal (Buckley and Schatzberg 2005; Follenius et al. 1992). Fourth, multiple studies have shown that sleep deprivation induces insulin resistance, pancreatic endocrine dysfunction, and hyperphagia (Knutson and Van 2008).

Thus although there is no conclusive evidence that OSA is a direct cause of DM, there is clinical evidence to support an adverse effect of OSA on insulin sensitivity, especially in patients with DM.

Dyslipidemia

Numerous large studies have shown both incremental cardiovascular risk with increasing levels of serum cholesterol (Multiple Risk Factor Intervention Trial Research Group 1982; NIH 2002; Stamler et al. 2000) and the benefits of lipid-lowering treatment (Heart Protection Study Collaborative Group 2002; LaRosa 2003; Law et al. 2003; NIH 2002) (see Table 2 for an overview). Given the association between OSA and atherosclerosis (Dematteis et al. 2009; Drager et al. 2005, 2007; Minoguchi et al. 2005), demonstration of a link between OSA and the development of dyslipidemia could have major therapeutic implications.

Table 2.

Selected clinical studies relevant to OSA and dyslipidemia^a

Methods	Findings	Author
Newly diagnosed OSA before/after 6 mo. CPAP; measured cholesterol, CRP, homocysteine	CPAP lowered total cholesterol, CRP, homocysteine	Steiropoulos et al. 2007
OSA subjects before/after CPAP, sham CPAP control; measured cholesterol and activated coagulation factors	CPAP lowered total cholesterol	Robinson et al. 2004
Children with OSA before/after tonsillectomy; measured cholesterol	CPAP lowered LDL-C, apolipoprotein B, and CRP, and raised HDL-C	Gozal et al. 2008
Males with OSA before/after CPAP; measured cholesterol and CRP	Only obese subjects had elevated CRP; no OSA or BMI effect on cholesterol	Barceló et al. 2004
Multivariate analysis of OSA subjects at differing levels of obesity and AHI; subset underwent CPAP treatment; measured homocysteine and CRP	CRP levels associated with obesity only	Ryan et al. 2007
Severely obese individuals undergoing bariatric surgery	VLDL triglyceride and LDL-C levels associated with severity of nocturnal oxyhemoglobin desaturation	Savransky et al. 2008
Severely obese individuals undergoing bariatric surgery	CRP levels associated with obesity only	Polotsky et al. 2009

^aBMI, body mass index; CPAP, continuous positive airway pressure; CRP, C-reactive protein; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; OSA, obstructive sleep apnea; VLDL, very low density lipoprotein

Most studies compare pre- and post-CPAP levels of serum cholesterol and inflammatory markers such as C-reactive protein (CRP), which indicates poor cardiovascular outcomes (Danesh et al. 2004; Ridker et al. 2005; Rutter et al. 2004). In one trial from Greece, researchers measured the cholesterol of newly diagnosed OSA patients both at baseline and after 6 months of CPAP and found that the intervention resulted in a 7% decrease in total cholesterol (TC; 240.2 to 223.7 mg/dL), mostly due to a fall in low-density lipoprotein cholesterol (LDL-C); they also noted concurrent reductions in high-sensitivity CRP and homo-cysteine (Steiropoulos et al. 2007). A study of 220 patients with OSA included a subtherapeutic CPAP control group and showed that therapeutic CPAP resulted in a decrease in TC from 220.4 to 209.6 mg/dL, which trended toward significance in comparison with the subtherapeutic group (p = 0.06) (Robinson et al. 2004). But the lack of information about what lipoprotein fractions were changed limits the usefulness of this information.

In children with OSA, tonsillectomy improved several parameters of the lipid profile as, 6 to 12 months after surgery, their LDL-C decreased from 92.0 to 66.0 mg/dL and their high-density lipoprotein cholesterol (HDL-C) increased from 44.6 to 64.2 mg/dL (Gozal et al. 2008). Apolipoprotein B, which more accurately reflects the atherogenicity of LDL-C (Lamarche et al. 1996, 1998), decreased from 102.2 to 56.3 mg/dL, and CRP levels also fell dramatically. The magnitude of these effects was similar in both obese and nonobese children (Gozal et al. 2008).

Several negative studies, however, highlight the confounding effects of obesity. In a Spanish study of 43 men with OSA, cholesterol levels and CRP were neither elevated at baseline compared to controls nor lowered by CPAP (Barcelo et al. 2004); BMI rather than AHI was the driver of elevated cholesterol, CRP, and insulin resistance. A few other studies confirm that BMI is a strong predictor of serum CRP levels, as OSA with or without CPAP did not affect CRP (Akashiba et al. 2005; Ryan et al. 2007).

In summary, there is limited evidence from adult clinical studies to support an independent effect of OSA on dyslipidemia. Even in studies where CPAP modestly improved lipid profiles, many subjects did not show comparatively elevated baseline cholesterol in OSA. Larger prospective studies may provide useful information in this area.

Obesity

Researchers have attempted in a number of studies to discern the nature of the relationship between OSA and obesity (Table 3). An increase of one standard deviation in any measure of the body habitus is associated with a threefold increase in the risk of OSA (Young et al. 1993). There is also speculation that OSA may reciprocally predispose to weight gain; retrospectively, patients with newly diagnosed OSA reported significant weight gain in the year prior (Phillips et al. 1999; Traviss et al. 2002). The possible indicators of a causal relationship would be more convincing if it could be shown that treatment of OSA can stimulate weight loss, but the data are scarce. Defining 4.5 kg as the threshold for significant weight loss, one small study showed that obese OSA patients compliant with CPAP for 6 months were more likely to lose weight than noncompliant patients (Loube et al. 1997). This finding contrasts with a similar study that found no weight change after 1 year of CPAP use (Redenius et al. 2008). Using a different metric of adiposity, Chin and colleagues (1999) showed that 6 months of CPAP led to visceral fat loss even if subjects did not lose weight.

Table 3.

Selected clinical studies relevant to OSA and obesity^a

Type	Methods	Findings	Author
Weight history	1-year weight histories in newly diagnosed OSA patients vs. screened controls	OSA patients exhibited significant antecedent weight gain	Phillips et al. 1999
	Obese newly diagnosed OSA subjects maintained lifestyle and dietary logs	Subjects reported rapid weight gain and adverse food intake profiles	Traviss et al. 2002
CPAP effects	OSA subjects with BMI>25 treated with CPAP for 6 months; compliance and body weight measured	CPAP-compliant subset more likely to lose >4.5 kg	Loube et al. 1997
	Retrospective comparison of BMI in compliant vs. noncompliant CPAP users for 1 year	CPAP did not induce weight loss	Redenius et al. 2008
	OSA subjects had visceral and subcutaneous fat, BMI, and leptin assessed (by CT scan) before/after CPAP for 3 days or 6 months	CPAP decreased leptin and visceral fat in all subjects; only those that lost weight also showed less subcutaneous fat	Chin et al. 1999
Metabolic rate	Sleeping metabolic rate in OSA subjects before/after uvuloplasty	Moderate-severe OSA increased metabolic rate, decreased by uvuloplasty	Lin et al. 2002
	24-hr metabolic rate in OSA subjects vs. snorers, 3 mo. before/after CPAP	OSA increased 24-hr metabolic rate, decreased by CPAP	Stenlof et al. 1996
	24-hour metabolic rate in male OSA subjects divided into quartiles of hypoxemia	More severe hypoxemia predicted lower metabolic rate	Major et al. 2007
Leptin	Leptin levels of male OSA subjects vs. BMI-matched controls	OSA associated with increased leptin	Phillips et al. 2000
	Leptin levels of male OSA subjects vs. BMI-matched controls	OSA associated with increased leptin	Vgontzas et al. 2000
	Leptin levels of male OSA subjects related to AHI and anthropometric data	No association between AHI and leptin	Schafer et al. 2002

^aAHI, apnea-hypopnea index; BMI, body mass index; CPAP, continuous positive airway pressure; CT, computed tomography; OSA, obstructive sleep apnea

From an equilibrium standpoint, obesity is a positive energy state that leads to storage in the form of fat. In normal sleep, there is an overall decrease in energy expenditure as compared to wakefulness, with relatively minor variations in energy expenditure as a function of sleep stage (Brebbia and Altshuler 1965; Fontvieille et al. 1994; Fraser et al. 1989; Webb and Hiestand 1975). A lower sleep metabolic rate might tip the scales toward anabolism and promote obesity, but data on the impact of OSA on sleep metabolic rate are highly inconsistent, as studies show increased (Lin et al. 2002; Stenlof et al. 1996), unaffected (Ryan et al. 1995), or decreased (Major et al. 2007) sleep metabolic rates between OSA and control subjects. The differing methods of these studies make it difficult to draw clear conclusions, but they reflect the complex determinants of metabolic rate during OSA-affected sleep; for example, fragmented slow-wave sleep should increase the metabolic rate, whereas severe nocturnal hypoxemia may reduce it.

Short sleep duration is associated with decreased levels of leptin, a hormone that lowers food intake, increases energy expenditure (Friedman and Halaas 1998; Halaas et al. 1995), and is secreted in proportion to body fat stores (Considine et al. 1996). Scientists hypothesized that leptin would be similarly lower in OSA patients, but two separate studies (Phillips et al. 2000; Vgontzas et al. 2000) found higher levels, consistent with a leptin-resistant state. CPAP treatment led to an early decrease in leptin, followed by a later loss of visceral fat (Chin et al. 1999).

Collectively these studies suggest that CPAP improves OSA-induced leptin resistance and thus facilitates visceral fat loss. However, not all studies support a relationship between OSA severity and leptin levels (Schafer et al. 2002), and the role of leptin resistance in obesity (Caro et al. 1996; Kalra 2001; Schwartz et al. 1996), let alone in OSA, is still incompletely understood. Therefore, while there is no question that obesity is a cause of OSA, evidence that OSA can in turn affect obesity is circumstantial at best.

Liver Disease

Although not technically part of MetS, nonalcoholic fatty liver disease (NAFLD) is becoming the most common liver disease of the developed world and shares many features of MetS (Bellentani et al. 2000; Browning et al. 2004; Browning and Horton 2004; Clark 2006; Hilden et al. 1977; McCullough 2006; Nomura et al. 1988). It comprises a spectrum of diseases ranging from simple fat accumulation to end-stage cirrhosis. Intermediate disease is characterized by hepatocytic fat accumulation (steatosis), which can progress to nonalcoholic steatohepatitis (NASH¹) when inflammatory changes become evident (Browning and Horton 2004; Day and James 1998; Diehl 2005). Most (60–90%) morbidly obese patients presenting for bariatric surgery have NAFLD (Angulo 2002; Clark 2006; Campos et al. 2008), and 53% of those with hepatic steatosis and 88% with NASH have MetS (Marchesini et al. 2003). NAFLD is a predictor of cardiovascular mortality (Hamaguchi et al. 2007) and increases the risk of liver-related death (Adams et al. 2005).

A number of case reports (Mathurin et al. 1995; Saibara et al. 2002; Trakada et al. 2004) offer proof of principle that severe OSA can lead to liver injury, supposedly due to tissue ischemia; direct liver ischemia is not necessarily related to NAFLD, but it may be one of several insults thought to be necessary for the disease process to occur (Day and James 1998; Dixon et al. 2001; Wanless and Shiota 2004). A handful of observational studies provide additional evidence of a possible relationship between OSA and NAFLD (Table 4). In 109 patients with OSA, serum aminotransferase levels correlated with low nocturnal O₂ saturation (Norman et al. 2008). In a study of 40 obese men with OSA, 14 had modestly increased morning aminotransferase levels that were blunted by a single night of CPAP (Chin et al. 2003). Obese snoring children commonly had elevated liver enzymes that normalized after tonsillectomy with adenoidectomy (Kheirandish-Gozal et al. 2008). However, elevated liver enzymes may simply reflect systemic inflammation (Yamada et al. 2006) and are not diagnostic of NAFLD. In fact, NASH can occur even in the absence of transaminitis (Garcia-Monzon et al. 2000). Koichiro Tatsumi and Toshiji Saibara (2005) used liver/spleen ratios and serum type III procollagen levels (Guechot et al. 1996) to characterize hepatic steatosis and latent steatohepatitis; by multiple regression analysis including AHI, BMI, and VFA as variables, only the average O₂ saturation inversely correlated with type III procollagen levels.

Table 4.

Selected clinical studies relevant to OSA and nonalcoholic fatty liver disease^a

Type	Methods	Findings	Author
Liver function tests	Retrospective regression analysis of OSA subjects	AHI correlated with elevated aminotransferases	Norman et al. 2008
	Obese males with OSA before/after CPAP aminotransferase levels	CPAP decreased aminotransferases	Chin et al. 2003
	Snoring children with suspected OSA evaluated for elevated aminotransferases before/after treatment	Aminotransferases elevated frequently in obese snorers, improved with OSA treatment	Kheirandish-Gozal et al. 2008
Liver biopsy	Liver biopsies taken at time of bariatric surgery	OSA associated with increased aminotransferases, trend toward progressive liver disease	Kallwitz et al. 2007
	Subjects with suspected OSA had polysomnography and a subset (18) had liver biopsy	Severe OSA associated with higher percentage of steatosis, necrosis, and fibrosis	Tanné et al. 2005

^aAHI, apnea-hypopnea index; CPAP, continuous positive airway pressure; OSA, obstructive sleep apnea

The most reliable method for diagnosing and staging NAFLD from steatosis to cirrhosis is liver biopsy (Clark and Diehl 2003), which was the assessment method in several bariatric surgery studies. Compared with their non-apneic counterparts, liver biopsies from OSA subjects trended (p = 0.06) toward more advanced inflammation and fibrosis (Kallwitz et al. 2007). In a similar bariatric surgery population (Jouet et al. 2007), OSA was a strong risk factor for elevated liver enzymes but not for the prevalence of NASH. In a larger study, Florence Tanné and colleagues (2005) found that increased BMI and AHI levels (>50) were predictive of elevated liver enzymes; in their study, half (n=18) of those with elevated liver enzymes volunteered to undergo liver biopsy, which revealed that severe OSA was associated with more advanced necrosis and fibrosis. In addition, one of the authors recently reported the results of a study that excluded patients with known liver disease from a bariatric surgery population and found that liver biopsy revealed NASH only in those with severe nocturnal hypoxia (n=20) (Polotsky et al. 2009). Taken together, these relatively small studies indicate that OSA is associated with liver enzyme elevations and possibly the progression of NAFLD. However, neither large prospective studies nor randomized CPAP trials have assessed the effects of OSA on the liver.

Clinical studies offer only limited mechanistic insights, and their data thus far have repeatedly been confounded by the prevalence of obesity in OSA populations. A translational approach in animals may therefore offer advantages. In the next two sections we describe and present data from animal models of OSA.

Animal Models of OSA

There are four general categories of OSA animal models: (1) animals with spontaneous sleep apnea, (2) mechanical or surgical airway obstruction, (3) sleep-dependent or -independent intermittent hypoxic gas delivery, and (4) induction of sleep fragmentation. We discuss studies and data based on the latter two types of model.³

Intermittent Hypoxia

In studies of metabolism in OSA, the mechanics of upper airway collapsibility are of less interest than the consequent exposure of the body to intermittent hypoxia (IH¹).⁴ Inducing IH precisely with each sleep onset requires sophisticated monitoring methods that make high-throughput data collection difficult. Instead, many researchers simply induce IH when rodents are most likely to be asleep. In rodents, 70% of sleep occurs in the light phase and 30% in the dark phase (Franken et al. 1999), so to mimic human OSA, IH exposure usually takes place during the light phase and ceases during the dark phase (Polotsky et al. 2006). Typically, a gas control system regulates the flow of air, nitrogen, and O₂ into customized cages to achieve cyclic O₂ desaturations similar to those experienced by subjects with OSA (Dematteis et al. 2009). Most investigators expose mice or rats to an oxygen nadir of 5–10%, cycling 9 to 60 times per hour for 6 to 12 hours per day (Chen et al. 2005; Xu et al. 2004; Zhan et al. 2005). During validation of this model in our laboratory, electroencephalogram tracings showed that O₂ desaturations accompanied arousal from sleep in the mice, capturing the sleep fragmentation aspect of OSA (Polotsky et al. 2006; more on sleep fragmentation models below).

Advantages

The main advantage of IH animal models lies in the direct control of all aspects of simulated OSA, from the genotype and phenotype of the rodent to all parameters of gas exchange. Perhaps the best-known work validating the use of IH models in OSA research is that of Eugene Fletcher, who showed that IH can induce systemic hypertension in rats (Fletcher et al. 1992b; Fletcher and Bao 1996b). He was able to leverage the animal model to great advantage by manipulating sympathetic signaling (Fletcher et al. 1992a; Lesske et al. 1997), inhaled gas mixtures (Fletcher et al. 1995), and rat strain (Fletcher and Bao 1996a).

Caveats

A number of concerns regarding animal models of IH require attention.

• The variety of protocols in the literature limits the generalizability of their findings.

• IH causes hypoxemia with hyperventilation and hypocapnia, rather than the hypercapnia characteristic of airway obstruction (Lee et al. 2008; Perry et al. 2006; Savransky et al. 2007a; Tagaito et al. 2001). In defense of this difference, Fletcher found that IH with supplemental CO₂ did not alter the hypertensive effect of IH, suggesting that, at least in this context, PaCO₂ did not significantly alter physiologic responses (Fletcher et al. 1995). On the other hand, it is likely that hypercapnic IH would induce greater sleep fragmentation (Ayas et al. 2000; Berthon-Jones and Sullivan 1984).

• Weight loss invariably occurs in rodents during early IH exposure and may interfere with the study of MetS- related phenotypes (Polotsky et al. 2003), although animal adaptation is evident after several weeks, with weight recovery (Li et al. 2007b). Such adaptation, geared toward survival but likely detrimental to long-term cardiovascular health, may actually be the basis for OSA-MetS connections.

• Although sleep fragmentation for up to 1 week has been reported in one IH model (Polotsky et al. 2006), the long-term effects on sleep architecture are unknown.

• A closed system with a fixed ambient O₂ is necessary for accurate measurements of metabolic parameters such as O₂ consumption and respiratory quotient, but the complex nature of intermittent oxygen delivery makes such measurements problematic.

• IH does not incorporate effects of upper airway occlusion such as thoracic negative pressure changes or upper airway inflammation.

• IH systems alter airflow and may thus disturb the animal environment; for this reason, most IH protocols expose control animals to intermittent air bursts without inducing hypoxia.

Although this list of caveats is long, IH models have successfully induced hypertension, hyperlipidemia, insulin resistance, and liver disease. Table 5 summarizes the strengths and weaknesses of IH models.

Table 5.

Strengths and weaknesses of intermittent hypoxia (IH) animal models in obstructive sleep apnea (OSA) metabolic research

Strengths	Weaknesses
Direct control over oxygen profile, diet, obesity, genotypes Induces sleep fragmentation (SF) Validated in studies to induce hypertension, hyperlipidemia, insulin resistance High throughput	Does not reproduce hypercapnia, thoracic strain, airway edema Long-term SF not characterized; difficult to separate hypoxia and SF effects on metabolism Heterogeneity of published protocols Problematic assessment of oxygen consumption, respiratory quotient Early weight lossa

^aWeight loss is followed by recovery, a potentially maladaptive long-term metabolic change.

Sustained Versus Intermittent Hypoxia

There are important differences between IH and sustained hypoxia research. Although some outcomes are likely to be similar, inasmuch as the severity of the hypoxia is the common pathway to injury, in cases such as sympathetic activation and oxidative stress IH may exert metabolic effects distinct from those of sustained hypoxia.

Oxygen Delivery

At the tissue level, sustained and intermittent hypoxia may have roughly equivalent impacts. During IH, lung O₂ stores, peripheral vasodilation, erythropoiesis, and O₂ diffusion properties should reduce the amplitude of O₂ fluctuations, but, in the absence of measures of O₂ tension in different organs, it is not possible to draw definitive conclusions. In the circulation, however, researchers in Christopher O’Donnell’s lab demonstrated that hemoglobin oxygen saturation varies widely throughout the IH cycle in mice (Lee et al. 2008). In addition, in our lab we have shown by pulse oximetry that relatively large amplitude swings in SpO₂ correlate closely with ambient cage O₂ levels during cycling IH (SpO₂ = 57–93%; Polotsky, unpublished data). Furthermore, in OSA and in most IH models, hypoxia is intermittent in two dimensions: first, there are rapid falls in oxyhemoglobin saturation with each apneic event; and second, hypoxemia is limited to the hours of sleep, which we might define as “circadian intermittency.” It is not clear whether rapid cycling IH, circadian IH, or simply hypoxia itself is the key stimulus, so it has not been possible to fully characterize oxygen delivery during IH—the metabolic effects may be related to the hypoxia, the intermittent nature of the stimulus, or both.

Sympathetic Activation

Scientists have long understood that sustained hypoxia activates the sympathetic nervous system (Korner and White 1966), and the same is proven true of IH (Baker-Herman and Mitchell 2008; Dimsdale et al. 1997; Fletcher 2003; Fletcher et al. 1992a,b; Gilmartin et al. 2008; Peng et al. 2001, 2003; Somers et al. 1988, 1995). At the chemoreceptor level, data from David Gozal’s and Nanduri Prabakhar’s laboratories suggest that IH differs markedly from sustained hypoxia as a stimulus for hypoxic ventilatory responses (Peng et al. 2004; Reeves et al. 2003): it induces efferent respiratory nerve activity persisting beyond the duration of exposure (Dick et al. 2007; Peng et al. 2003; Reeves et al. 2006). Correspondingly, a number of transcriptional changes occur in IH but not in sustained hypoxia (Nanduri et al. 2008). Harly Greenberg and colleagues (1999) provided evidence of the effect of IH on the sympathetic nervous system, showing that 30 days of IH augmented cervical sympathetic activity in response to hypoxia, hypercapnia, or both. Thus there are unique sympathomimetic effects attributable to IH that are likely to have metabolic consequences.

Inflammation

Researchers are aware of several mechanistic pathways for hypoxia- and IH-induced inflammation, although the actual contribution of such inflammation to MetS is unclear. One pathway is via the activity of hypoxia-inducible factor 1 (HIF-1¹), a DNA-binding protein that controls the expression of genes related to angiogenesis, erythropoiesis, cell survival, and iron homeostasis under hypoxic conditions (Wang and Semenza 1995; Wang et al. 1995). The transcription factor comprises two subunits: HIF-1 β, which is constitutively expressed, and HIF-1α, which is intact only during hypoxia due to rapid ubiquitination and proteasomal degradation under normal O₂ tension (Semenza 2001). HIF-1α promotes the expression of interleukin (IL) 8 and subsequent neutrophil chemotaxis in human endothelial cells (Kim et al. 2006), and upregulates inflammatory adenosine receptor signaling (Kong et al. 2006).

Hypoxia also induces nuclear factor-kappa B (NF-kB), a transcription factor that regulates inflammatory and anti-apoptotic responses (Cummins et al. 2007; Hoffmann and Baltimore 2006; Sen and Baltimore 1986). Animal IH models and OSA studies alike have shown elevations in NF-kB (Greenberg et al. 2006; Ryan et al. 2005, 2006; Savransky et al. 2007c), inflammatory cytokines (Li et al. 2006; Ryan et al. 2005, 2006; Savransky et al. 2007a,b; Vgontzas et al. 2000), and adhesion molecules (Chin et al. 2000; Ohga et al. 1999, 2003; Robinson et al. 2004). HIF-1 acts via NF-kB to promote neutrophil survival (Walmsley et al. 2005), and constitutive HIF-1α expression in mouse epithelial cells activates NF-kB and downstream cytokines (Scortegagna et al. 2008).

Hypoxia may also suppress adiponectin bioactivity (adiponectin is secreted by adipocytes and has an array of anti-inflammatory, antiatherogenic, insulin-sensitizing properties; Maeda et al. 1996; Matsuzawa et al. 2004). In vitro studies suggest that both hypoxia (Chen et al. 2006) and IH (Magalang et al. 2009) result in lower adiponectin release from adipocytes.

Oxidative Stress

Both intermittent and sustained hypoxia cause oxidative stress via mitochondrial dysfunction (Guzy and Schumacker 2006; Guzy et al. 2005), but IH may be more detrimental because it involves reoxygenation, analogous to ischemia/reperfusion injury, known to be mediated by reactive oxygen species (Jun et al. 2008; Prabhakar et al. 2007; Selmi et al. 2007; Weinberg et al. 2000; Yamauchi and Kimura 2008). Indeed, OSA has been shown to induce systemic oxidative stress, which responds favorably to CPAP (Dorkova et al. 2008; Lavie 2003; Lavie et al. 2004; Schulz et al. 2000). In rats, reactive oxygen species mediate sympathetic responses to IH, by both carotid body sensitization (Prabhakar and Kumar 2004) and adrenal catecholamine efflux (Kumar et al. 2006). In addition, IH induces oxidative stress in regions of the rat brain (Ramanathan et al. 2005), with detrimental consequences on neuropeptide signaling (Sharma et al. 2009). Ling Chen and colleagues (2005) showed that IH exposure leads to progressive increases in cardiac oxidative stress with concurrent left ventricular dysfunction. More recently investigators showed that the antioxidant effects of L-carnitine may attenuate muscle fatigue during IH (Dutta et al. 2008). But oxidative stress in OSA and IH has not been ubiquitously described (Ozturk et al. 2003; Svatikova et al. 2004, 2005) and the extent to which it induces metabolic dysfunction is under investigation.

Sleep Fragmentation

Another key aspect of OSA is sleep fragmentation, for which there are several animal models. One method, known as the platform or “flower-pot” technique (Grahnstedt and Ursin 1985; Hilakivi et al. 1984; Mendelson et al. 1974), involves placing an animal on a platform above water. When the animal enters rapid eye movement (REM¹) sleep, muscle atonia causes it to fall into the water and awaken. This method deprives the subject of REM sleep and, to a lesser extent, slow-wave sleep.

Dina Brooks and colleagues (1997a) developed a model of mechanical airway obstruction timed to sleep onset in dogs and modified it to induce sleep fragmentation by an acoustic alarm without interfering with the airway (Brooks et al. 1997b); and a similar, more recent study compared the effects of hypoxic arousals in an IH model with mice subjected to high-flow air blasts during the light phase (Polotsky et al. 2006). Both models induced sleep fragmentation over a span of several days without altering total sleep time. Polotsky’s IH protocol decreased Δ power and REM sleep during the light phase (an exposure period), but these alterations in sleep architecture did not occur during nonhypoxic sleep fragmentation. It is nearly impossible to identify the effects of IH as resulting from either hypoxia or sleep fragmentation because nonhypoxic sleep fragmentation is a weaker stimulus to disrupt sleep than IH. A limitation of most nonhypoxic sleep fragmentation methods is that the homeostatic drive to sleep increases over time, raising the arousal threshold and engendering microsleep between stimuli during the exposure period. Thus animals eventually adapt and sleep through any but the most noxious stimuli.

To summarize, an animal model of long-term sleep fragmentation without sleep deprivation does not yet exist, and animal data pertinent to metabolism in OSA are chiefly from IH models.

Effects of Intermittent Hypoxia: Evidence from Animal Models

Diabetes

Studies indicate that short-term exposure to intermittent hypoxia during the light (i.e., sleep) phase leads to insulin resistance in lean mice. IH induces a state of elevated sympathetic outflow that appears to be HIF-1 dependent (Peng et al. 2006), and this sympathetic activation leads to acute elevations in serum glucose, stimulates hepatic glucose secretion, decreases peripheral glucose use, and suppresses pancreatic insulin secretion (Deibert and DeFronzo 1980; Porte 1967; Rizza et al. 1980). Furthermore, catecholamine-stimulated lipolysis releases free fatty acids into the circulation that can contribute to insulin resistance and elevated triglycerides (Delarue and Magnan 2007; Kim et al. 2001). Even in the absence of sympathetic input acute IH in mice alters glucose metabolism (Iiyori et al. 2007): C57BL/6J mice exposed to 9 hours of IH exhibited insulin resistance and decreased glucose use in oxidative muscles, a finding that was not affected by autonomic blockade with hexa-methonium. These results suggest that impaired oxidative phosphorylation may explain insulin resistance (Iiyori et al. 2007), similar to the effect in humans at high altitude (Larsen et al. 1997).

Does the IH-induced insulin resistance carry over from the light phase, when the stimulus is administered, to the dark phase, when the animals are allowed to recuperate? One of the authors participated in a study that exposed lean C57BL/6J and leptin-deficient obese (C57BL/6J-Lep^ob) mice to IH during the light phase for 5 days. The researchers measured fasting glucose and insulin levels 6 hours after cessation of exposure (i.e., during the dark phase) and used a homeostasis model assessment (HOMA) to calculate insulin resistance. Surprisingly, the mice had lower fasting blood glucose and higher insulin sensitivity (Polotsky et al. 2003), with concurrent increases in serum leptin and adipose tissue leptin transcription. The C57BL/6J-Lep^ob mice, on the other hand, showed a sixfold increase in insulin levels and only modestly decreased fasting blood glucose, indicative of significant insulin resistance. Administration of leptin normalized the insulin level. When the research group then exposed the leptin-deficient animals to a 12-week course of IH, the animals showed stepwise, time-dependent increases in fasting insulin levels (Polotsky et al. 2003). Thus it appears that, at least in leptin-deficient obese animals, chronic IH can induce progressive and substantial dark-phase insulin resistance. The same cannot be concluded from lean animals.

A subsequent experiment illustrates the importance of considering glucose uptake during both the light and dark phase. Takuya Yokoe and colleagues (2008) subjected mice to 3 days of IH, with serial measurements of glucose, insulin, and functional insulin resistance at 8:00 a.m. (the end of the awake phase) and at 6:00 p.m. (the end of the sleep phase). Blood glucose in the intermittent air control group decreased during the light phase and increased during the dark phase. The opposite profile occurred in the IH group, but insulin levels in both groups were unchanged. Corticosterone levels, the principal stress glucocorticoid of rodents, were higher during both the light and dark phase. Furthermore, when mice were exposed to 4 weeks of IH, there was evidence of pancreatic β cell replication, resulting in elevated β cell mass. Without a change in average glucose levels over the experimental period and the lack of augmented β cell replication in another group receiving glucose infusions, the authors showed that a mechanism other than hyperglycemia caused the pancreatic changes. Acute IH thus reverses glucose diurnal patterns and stimulates the hypothalamic- pituitary axis. Collectively, these experiments demonstrate that IH can alter patterns of glucose clearance in lean mice. The results also suggest that IH increases insulin resistance in lean mice during the light phase, when the stimulus was administered, but that the resistance did not carry over to the dark phase. The effects of chronic IH on insulin resistance in lean mice have not been studied.

Additional animal studies relevant to hypoxia and insulin resistance concern growth hormone. The effects of cyclic IH on growth hormone are unknown, but it is possible to draw conclusions from experiments under other hypoxic conditions. In rats, after a 2-hour exposure to hypoxia, pituitary growth hormone levels were high and plasma levels low, suggesting impaired secretion. After a 25-day exposure, both pituitary and plasma growth hormone levels were low, consistent with impaired growth hormone synthesis (Zhang and Du 2000). Similarly, exposure to 12% O₂ led to reduced animal growth and decreased IGF-1 bioactivity. Exogenous growth hormone improved growth in the hypoxic group but not in pair-fed controls (Moromisato et al. 1999).

In summary, studies of IH in animal models have shown that even in the absence of obesity, IH can profoundly alter both acute and chronic glucose homeostasis. Sympathetic activation, decreased oxidative phosphorylation, leptin signaling, and growth hormone axis suppression are possible mediators of these observations. However, much remains unexplained. For example, are lean mice also predisposed to progressive insulin resistance with chronic IH exposure? What is the functional significance of IH-induced pancreatic β cell replication? Perhaps further studies will reveal the answers to these important questions.

Dyslipidemia

Researchers have used IH models to study the relationship between dyslipidemia and OSA. Jianguo Li and colleagues (2005b) exposed C57BL/6J and C57BL/6J-Lep^ob mice to 5 days of IH and then measured serum and liver lipids as well as transcriptional levels of hepatic cholesterol synthesis enzymes. They found that in lean mice IH increased serum total cholesterol, mostly due to a 20–30% increase in triglycerides (with a twofold increase in the liver) and HDL cholesterol, whereas obese mice had higher baseline levels of all lipids and IH did not alter these levels. The authors showed a number of alterations in the hepatic lipid biosynthesis pathways; in particular, IH increased expression of (1) sterol regulatory element binding protein 1 (SREBP-1¹), a key regulator of synthesis for several lipids, and (2) a downstream enzyme of lipoprotein secretion, stearoyl-CoA desaturase 1 (SCD-1¹). Accordingly, the ratio of monounsaturated to polyunsaturated fatty acids was higher, reflecting SCD-1 enzyme activity. In addition, IH downregulated scavenger receptor B-1, which is involved in the uptake of HDL cholesterol and phospholipids from the circulation. These results show that IH can alter lipid profiles and their associated synthesis pathways. However, the clinical significance of these findings is unclear because the animals lost weight and had increased HDL cholesterol, clear departures from MetS.

In the C57BL/6J-Lep^ob mice, short-term IH did not seem to affect the lipid metabolism but after 12 weeks of chronic IH the animals showed increased liver triglycerides and phospholipids, with concurrent transcriptional upregulation of enzymes such as SREBP-1 and SCD-1, as seen in the lean IH-exposed animals. Further experiments showed that severity of hypoxia is an important variable: when the hypoxic nadir was raised from 5% to 10%, IH had no effect on cholesterol, weight, or hepatic oxidative stress (Li et al. 2007b).

With only the 5% nadir of IH altering lipid metabolism in lean mice, Li and colleagues (2006) hypothesized that HIF-1 might be involved. They performed a 5-day IH experiment using either C57BL/6J (WT) or heterozygous Hif1α^+/− mice (with partial HIF-1α deficiency). During IH, Hif1α^+/− mice experienced blunted rises in serum triglycerides, liver triglycerides, light-phase fasting insulin, and glucose, and attenuated transcription or translation of several liver lipid biosynthesis enzymes. However, the genotype did not affect the rise in total cholesterol, HDL, phospholipids, or free fatty acids. HIF-1α deficiency diminished the rise of SREBP-1 and SCD-1 protein levels during IH without affecting serum cholesterol (Li et al. 2006). This could indicate that these enzyme pathways are only associations and not causal pathways of IH-induced hypercholesterolemia. However, the more than twofold increase in SCD-1 protein level (and associated increase in both serum and liver triglycerides that was attenuated by HIF-1α deficiency) appears to indicate a logical pathway for therapeutic investigation.

To more directly determine the role of SREBP-1 in IH-induced dyslipidemia, a conditional knockout of SREBP cleavage-activating protein (SCAP) was introduced onto a B6;129 background, generating L-Scap⁻ mice. After 5 days of IH, the L-Scap⁻ animals did not exhibit any changes in lipid profile or expression of previously upregulated lipid biosynthesis genes (Li et al. 2007a). However, SCAP deficiency lowered lipid levels as well as both baseline and IH levels of gene expression in comparison with those of wild-type controls. Thus, while this study shows the cholesterol-suppressing benefit of inhibiting SREBP-1, it does not prove definitively that this pathway is the mechanism of IH-induced hyperlipidemia. In a subsequent study (Savransky et al. 2008), injections of antisense oligonucleotides abolished the rise of hepatic SCD-1 and hyperlipidemia and downregulated SCD-1 expression by 90% in C57BL/6J mice exposed to IH for 10 weeks. However, low serum lipid levels were observed in both control and IH animals treated with SCD-1 antisense oligonucleotides. Hence, these results serve more to confirm that there are potential therapeutic effects of SCD-1 blockage (Cohen et al. 2002; Gutierrez-Juarez et al. 2006; Jiang et al. 2005; MacDonald et al. 2008; Sampath and Ntambi 2006) than to prove that IH mediates its lipid effects via SCD-1.

In one of the only studies to date to compare the independent effects of IH, REM sleep deprivation, and sleep restriction, Juliana Perry and colleagues (2006) subjected rats to 4 days of (1) IH, (2) REM sleep deprivation (using the platform technique), (3) IH and REM sleep deprivation, or (4) sleep restriction (similar to REM sleep deprivation but the rats were allowed to sleep up to 6 hours/day). A separate group of rats experienced chronic IH or chronic IH and sleep restriction for 3 weeks. The researchers found that REM sleep deprivation or the combination of IH and REM sleep deprivation increased TC and HDL-C but lowered triglycerides and very low density lipoproteins (VLDL). IH alone did not have any effect, in contrast to the findings from the study by Li and colleagues (2007b). (On the other hand, Perry’s IH protocol used a nadir of 10% O₂, so their findings were consistent with Li’s work in the equivalent model.) Chronic IH led to increases in triglycerides and VLDL, whereas sleep restriction or chronic IH in combination with sleep restriction caused lower VLDL and triglycerides. These findings show that alterations in sleep architecture alone can affect lipid metabolism, and that sleep restriction may have an antagonistic, lipid-lowering effect as compared to IH. Together with the studies by Li and colleagues, the findings appear to indicate that IH of sufficient duration and severity can raise VLDL and triglycerides (Perry et al. 2006), although there is as yet no adequate explanation for these observations.

Further examination of IH-induced changes to hepatic lipid synthesis, lipolysis, and fatty acid metabolism is necessary to advance the field.

Obesity

The metabolic responses to hypoxia are complex and depend on an animal’s species, developmental stage, environmental temperature, nutritional state, and activity. Sustained hypo-baric hypoxia induces anorexia (Hamad and Travis 2006; Tschop and Morrison 2001), catabolism (Consolazio et al. 1972; Larsen et al. 1997; Whitten et al. 1970), and, at ambient temperatures lower than an organism’s thermoregulatory set point, a fall in metabolic rate (Gautier 1996; Kellogg 1978), adaptations that lead to lower body mass (Zhou et al. 2007) and energy conservation. Another adaptive response to chronic hypoxia is an overall shift toward carbohydrate oxidation, supposedly because more ATP is produced from glucose than from lipid per O₂ molecule consumed (Braun 2008). HIF-1 appears to be instrumental in orchestrating these metabolic changes (Iyer et al. 1998). Consistent with sustained hypoxia, IH induces lower food intake and weight loss in mice (Li et al. 2005b).

How then is it possible that anorexia and catabolism during IH have anything to do with MetS, a disease of obesity? Thus far, IH-induced weight loss has been either regarded as evidence of an overly hypoxic or stressful model, or touted as an ancillary finding that strengthens a finding such as dyslipidemia, which ought to improve with weight loss. Instead, IH-mediated weight loss should receive attention for its possible connections to MetS. First, weight loss under hypoxic conditions may have detrimental consequences. High-altitude dwellers, chronic commuters between high and low altitudes (Siques et al. 2007), and rats exposed to hypobaric hypoxia (Louhija 1969) have elevated triglyceride levels. This observation has not been fully explained, but would be a logical consequence of elevated circulating and hepatic free fatty acids liberated from fat stores during hypoxia. Elevated free fatty acids in the circulation can induce insulin resistance (Delarue and Magnan 2007; Kim et al. 2001) thereby affecting other features of MetS. Second, adaptations to long-term hypoxia lead to attenuated weight loss (Boyer and Blume 1984; Tschop and Morrison 2001) and likely foster a state of “energy thrift,” ultimately promoting obesity when hypoxia is no longer present. By similar reasoning, hypoxia during early development may predispose to MetS in adulthood (Waters and Gozal 2004).

Nonalcoholic Fatty Liver Disease (NAFLD)

We showed above that liver triglycerides increased after IH, whether after a 5-day exposure for lean animals (Li et al. 2005b) or a 12-week exposure for obese animals (Li et al. 2005a). Accumulation of cholesterol in the liver is a hallmark of NAFLD, but the liver was not the focus of these studies. In a study of the effects of IH on the liver, Vladimir Savransky and colleagues exposed pair-fed C57BL/6J mice to 12 weeks of chronic IH and evaluated liver histopathology. Chronic IH led to increased liver enzymes, hyperglycemia, oxidative stress, and increased hepatic glycogen stores but did not induce hepatic steatosis or inflammation (Savransky et al. 2007c). However, intraperitoneal injection of acetaminophen (600 mg/kg) led to severe liver injury in the IH group (with a negligible effect in controls). The authors proposed that the synergistic glutathione-depleting effects of acetaminophen were compounded by IH-induced oxidative stress. Furthermore, with this two-hit model of liver injury in mind they hypothesized that diet-induced hepatic steatosis may predispose the liver to progress to NASH under IH conditions. Indeed, in a subsequent experiment, the combination of a high-fat, high-cholesterol diet and chronic IH led to hepatic inflammation and fibrosis as well as elevations in serum transaminases, hepatic inflammatory cytokines, collagen, and oxidative stress (the livers of control mice showed simple steatosis without inflammation; Savransky et al. 2007a).

These studies support the clinical observation that NAFLD is a multifactorial disease. Unfortunately, the mechanism by which IH induces injury in these models is unclear. Reactive oxygen species from liver nicotinamide adenine dinucleotide phosphate (NADPH) oxidases during IH may be a potential pathway (Jun et al. 2008). But further interventional experiments are necessary to establish causal relationships between IH, oxidative stress, and NAFLD.

Conclusions

Given the close association between MetS and cardiovascular mortality, there is immense clinical interest in reducing metabolic risk factors. Therefore, identifying OSA as a novel risk factor for MetS, in part or in whole, has great therapeutic significance. For example, clinical practice has changed with the identification of OSA as a cause of hypertension. In the same way, OSA may affect other components of MetS, and identifying the relevant physiologic and metabolic pathways may produce therapies that could support OSA treatments. Such therapies would be especially important to those who are intolerant of or noncompliant with CPAP. In addition, OSA may contribute to the inflammatory state of MetS without directly leading to a particular MetS phenotype, in which case an anti-inflammatory OSA treatment would be appropriate. IH research may also provide valuable insights into the metabolic consequences of other conditions that involve repetitive exposure to hypoxia, whether intermittent or sustained.