Effects of different acute hypoxic regimens on tissue oxygen profiles and metabolic outcomes

Abstract

Obstructive sleep apnea (OSA) causes intermittent hypoxia (IH) during sleep. Both obesity and OSA are associated with insulin resistance and systemic inflammation, which may be attributable to tissue hypoxia. We hypothesized that a pattern of hypoxic exposure determines both oxygen profiles in peripheral tissues and systemic metabolic outcomes, and that obesity has a modifying effect. Lean and obese C57BL6 mice were exposed to 12 h of intermittent hypoxia 60 times/h (IH60) [inspired O₂ fraction (Fi_O2) 21–5%, 60/h], IH 12 times/h (Fi_O2 5% for 15 s, 12/h), sustained hypoxia (SH; Fi_O2 10%), or normoxia while fasting. Tissue oxygen partial pressure (Pti_O2) in liver, skeletal muscle and epididymal fat, plasma leptin, adiponectin, insulin, blood glucose, and adipose tumor necrosis factor-α (TNF-α) were measured. In lean mice, IH60 caused oxygen swings in the liver, whereas fluctuations of Pti_O2 were attenuated in muscle and abolished in fat. In obese mice, baseline liver Pti_O2 was lower than in lean mice, whereas muscle and fat Pti_O2 did not differ. During IH, Pti_O2 was similar in obese and lean mice. All hypoxic regimens caused insulin resistance. In lean mice, hypoxia significantly increased leptin, especially during SH (44-fold); IH60, but not SH, induced a 2.5- to 3-fold increase in TNF-α secretion by fat. Obesity was associated with striking increases in leptin and TNF-α, which overwhelmed effects of hypoxia. In conclusion, IH60 led to oxygen fluctuations in liver and muscle and steady hypoxia in fat. IH and SH induced insulin resistance, but inflammation was increased only by IH60 in lean mice. Obesity caused severe inflammation, which was not augmented by acute hypoxic regimens.

intermittent hypoxia, a hallmark feature of obstructive sleep apnea (OSA), is believed to be a major factor in the development and progression of metabolic dysfunction (11, 23, 26, 28, 37, 46). However, despite a copious body of literature on metabolic and cardiovascular outcomes, almost nothing is known about changes in tissue oxygen partial pressure (Pti_O2) during intermittent hypoxia (3).

The majority of patients with OSA are obese (38, 55, 56). Obesity per se causes insulin resistance, Type 2 diabetes, systemic inflammation, oxidative stress, and adverse cardiovascular outcomes (10). Adipose tissue hypoxia is implicated in the development of insulin resistance and systemic inflammation of obesity (12, 53). Obese individuals show significantly lower oxygen levels in adipose tissue than normal-weight subjects (9, 19, 32). Oxygen levels in other insulin-sensitive tissue, such as liver and skeletal muscle, in obese and lean subjects have not been compared.

The authors of the present study have developed a mouse model of intermittent hypoxia, recapitulating the nocturnal oxygen profile in patients with OSA, and have shown that chronic intermittent hypoxia induces insulin resistance, systemic inflammation, and oxidative stress in mice (14, 17, 35, 42). Nevertheless, multiple aspects of OSA and intermittent hypoxia remain poorly understood. Metabolic outcomes of OSA have been correlated with the severity of the disease, but this relationship has not been sufficiently addressed in the animal model. A more severe nadir of Sp_O2 (oxygen saturation measured by pulse oximetry) led to more severe hepatic oxidative stress and dyslipidemia at the same frequency of intermittent hypoxia (25), but the effects of frequency of intermittent hypoxia on tissue O₂ levels and metabolic outcomes have not been systematically investigated in animal models.

In the present study, lean and obese mice were exposed to two different paradigms of intermittent hypoxia: 12 times/h and 60 times/h, as well as sustained hypoxia and measured 1) oxygen levels in blood, liver, muscle, and epididymal fat in real time; 2) indexes of insulin resistance, including fasting blood glucose and plasma insulin levels, as well as levels of adipokines affecting insulin resistance, leptin and adiponectin; and 3) indexes of oxidative stress and inflammation. Commonly used by investigators regimens of intermittent and sustained hypoxia were chosen (34, 35, 47).

We hypothesized that a regimen of hypoxic exposure determines both oxygen profiles in peripheral tissues and systemic metabolic outcomes, and that obesity has a modifying effect.

METHODS

Ethical approval.

All protocols complied with and were approved by the Johns Hopkins University Animal Care and Use Committee and were in accordance with the current National Institutes of Health guidelines (Guide for the Care and Use of Laboratory Animals).

Experimental animals.

In total, 143 6- to 8-wk-old male lean C57BL/6J mice, purchased from Jackson Laboratory (Bar Harbor, ME), were used in the study. Seventy-two animals were used for all reported metabolic changes, oxidative stress, and inflammation. Forty-eight animals were used for immunoblotting and immunohistostaining for hypoxia probe. Twenty-three animals were used solely for assessing oxygen profiles in arterial blood and metabolically active tissues. Lean mice were fed a regular chow diet, whereas obese mice were fed a high-fat diet (TD 03584, Teklad WI, 5.4 kcal/g, 35.2% fat, 58.4% of kcal from fat). Mice were housed in a light-dark regulated environment with the 12-h light phase administered from 9 AM to 9 PM. All measurements were obtained at an age of 25–30 wk.

Experimental design.

The authors of the present study have developed a mouse model of intermittent hypoxia, as previously described (39). Briefly, a gas control delivery system was designed to regulate the flow of room air, nitrogen, and oxygen into cages. During intermittent hypoxia 60 times/h, the fractional inhaled O₂ was reduced from 20.9 to ∼5% over 30 s and then reoxygenated to room air levels in the subsequent 30 s, resulting in 60 hypoxic events per hour. During intermittent hypoxia 12 times/h, the fractional inhaled O₂ was reduced from 20.9 to ∼5% within 20 s, kept at ∼5% for the subsequent 15 s, and then rapidly reoxygenated to room air levels. This cycling was repeated every 5 min, resulting in 12 hypoxic events per hour (34). During sustained isobaric hypoxia, the fractional inhaled O₂ was reduced to a sustained level of 10%. Normoxic mice had continuous exposure to room air. For all measurements, except for assessing oxygen profiles, mice were exposed for one night (12 h) to normoxia, intermittent hypoxia 12 times/h, intermittent hypoxia 60 times/h, or sustained hypoxia while fasting. Thirty minutes before the end of the exposure, blood glucose was measured in unanesthetized mice by tail bleeding. Immediately after the exposure, mice were anesthetized with 1–2% isoflurane, and blood, liver, muscle, and epididymal fat were obtained for further analysis. In a second set of 11 lean and 12 obese mice without previous conditioning to hypoxia, Pti_O2 in liver, muscle and epididymal fat was measured during the different hypoxic regimens using a fast-response oxygen microelectrode (OX-50, Unisense A/S, Denmark; 50-μm diameter, 90% response time <5 s) simultaneously with noninvasive pulse oximetry by applying a Mouse CollarClip (Starr Life Sciences, Oakmont, PA). The oxygen microelectrode was calibrated before and after measurement in each animal, according to the manufacturer's protocol. The mice were anesthetized with 1.5 g/kg urethane intraperitoneally for this procedure. The snout was positioned through a hole into the same type of cage used for the first set of mice. Then the skin over the different tissues was cut open to allow measurement of tissue oxygenation with the probe. The oxygen microelectrode was mounted on a micromanipulator and inserted vertically 2 mm from the surface and then slightly retracted to a point where a stable Pti_O2 signal was obtained. All three hypoxic regimens, as well as normoxia, were delivered as described above and recorded in the same position before moving the oxygen microelectrode to a different position in the same tissue or to a different tissue. The duration of measurement was at least 2 min for both normoxia and sustained hypoxia and 5 min for the intermittent hypoxic regimens. Overall, each experimental animal was used for multiple measurements (four experimental regimens in three tissues).

Hypoxia probe.

Immunohistostaining for hypoxia probe was done using Hypoxyprobe-1 kit (HP2–100, Chemicon International), according to the manufacturer's protocol. Briefly, Hypoxyprobe-1 was injected intraperitoneally at 60 mg/kg body wt at 30 min before tissue collection, while exposure to hypoxia or normoxia continued. Fresh tissue was fixed in 10% buffered formalin, dehydrated in ethanol, and embedded in paraffin. Immunohistostaining was performed with the primary Hypoxyprobe-1 antibody conjugated with FITC and a secondary antibody of anti-FITC antibody conjugated to horseradish peroxidase. For Western blotting, fresh liver, muscle, and epididymal fat were frozen immediately in liquid nitrogen. SDS-PAGE and Western blot were performed using Bio-Rad precast gel system. Protein (20 μg) was applied per lane. We used Hypoxyprobe-1 antibody as the primary antibody.

Biochemistry.

Fasting plasma total cholesterol and triglycerides were measured with kits from Wako Diagnostics (Richmond, VA). Blood glucose was tested with Accu-Chek Comfort Curve kit from Roche Diagnostics (Indianapolis, IN). Plasma insulin and leptin were measured with ELISA kits from Millipore (Billerica, MA). Adiponectin was assessed with an ELISA kit from R&D Systems (Minneapolis, MN). Malondialdehyde was measured with an assay from Percipio Biosciences. To measure tumor necrosis factor-α (TNF-α) secretion ex vivo, epididymal fat tissue fragments were incubated for 3 h in M199 medium (Gibco, Langley, OK) containing 1% BSA. TNF-α levels in the medium were assessed with an ELISA kit from R&D Systems (Minneapolis, MN). For real-time polymerase chain reaction (PCR), total RNA was extracted from liver using TRIzol (Invitrogen, Carlsbad, CA), and complementary DNA was synthesized using Advantage RT for PCR kit from Clontech (Palo Alto, CA). Real-time reverse transcriptase PCR was performed with primers and probe specific for TNF-α, as previously described (42). The mRNA expression levels were normalized to 18S ribosomal RNA concentrations using the standard ΔΔCt approach (http://www.ambion.com/techlib/basics/rtpcr/index.html), as our laboratory has previously done (25), and then expressed as a ratio of mRNA in a sample of interest to the average mRNA level in normoxic mice.

Data analysis.

Analyses were performed using Stata version 9.0. Data are presented as means ± SE. Comparison of oxygen levels between lean and obese mice was performed using a two-sample Wilcoxon rank-sum (Mann-Whitney) test. Comparison of mean variance in oxygen levels was performed using a Wilcoxon signed-rank test. Comparison of metabolic and inflammatory indexes between hypoxic and normoxic conditions within lean or obese phenotype were performed using unpaired t-tests. Comparisons of metabolic and inflammatory indexes between lean and obese mice were performed using a general linear model ANOVA. A P value of <0.05 was considered significant.

RESULTS

Lean mice.

In lean mice, preexposure body weights in the normoxic group were lower than in all hypoxic groups (Table 1). The initial weight difference was intentional in anticipation of a greater weight loss in hypoxic groups (16). The body weight at the time of death was similar in all groups, except for mice exposed to sustained hypoxia, which were heavier than other groups. Weight loss during all hypoxic exposures was significantly greater compared with normoxia, with both intermittent hypoxic regimens showing the greatest weight loss.

Table 1.

Basic characteristics

	Lean Mice				Obese Mice
	Air	IH12	IH60	SH	Air	IH12	IH60	SH	P Value
N	18	18	18	18	12	12	12	12
Age, wk	25–30	25–30	25–30	25–30	25–30	25–30	25–30	25–30
Body weight before exposure, g	28.7 ± 0.4	30.5 ± 0.6*	31.0 ± 0.6*	30.2 ± 0.5*	42.4 ± 1.9	43.3 ± 2.1	43.0 ± 1.9	42.2 ± 1.7	<0.001
Body weight after exposure, g	26.9 ± 0.3	26.5 ± 0.4	26.8 ± 0.4	28.2 ± 0.5*	40.2 ± 1.8	39.7 ± 2.0	39.4 ± 1.9	39.1 ± 1.7	<0.001
Body weight loss, %	7.6 ± 0.6	12.7 ± 0.6*	12.8 ± 0.6*	9.3 ± 0.3	5.2 ± 0.4	8.4 ± 0.3†	8.8 ± 0.5†	7.5 ± 0.5†	<0.001
Plasma triglycerides, mg/dl	28.4 ± 1.9	17.4 ± 1.1*	21.7 ± 0.9*	32.0 ± 1.9	54.7 ± 2.3	47.0 ± 3.7	51.4 ± 2.3	33.1 ± 2.6†	<0.001
Plasma total cholesterol, mg/dl	97 ± 2	107 ± 3*	105 ± 3	106 ± 5	176 ± 10	158 ± 8	177 ± 11	181 ± 11	<0.001

Values are means ± SE; n, no. of mice. Values are shown for normoxia (air), intermittent hypoxia 12 times/h (IH12), intermittent hypoxia 60 times/h (IH60), and sustained hypoxia (SH) in both lean and obese mice.

^*P < 0.05 compared with lean air group.

^†P < 0.05 compared with obese air group. The P values in the right column compare lean mice vs. obese mice using a general linear model ANOVA.

The mean oxygen levels in blood and all organs were highest in normoxia and decreased progressively from normoxia to infrequent intermittent hypoxia (12 times/h), frequent intermittent hypoxia (60 times/h), and sustained hypoxia being the lowest (P < 0.001, Fig. 1A, Table 2). Comparison of Pti_O2 in different tissues showed that the mean Pti_O2 was significantly higher in the liver than in muscle and fat (P < 0.001). During frequent intermittent hypoxia, high-amplitude swings of Pti_O2 were observed in blood and liver and, to a lesser degree, in muscle. The variability of Pti_O2 in muscle and epididymal fat during frequent intermittent hypoxia was significantly less compared with Sp_O2. The lowest variability of Pti_O2 during frequent intermittent hypoxia was observed in epididymal fat (Fig. 1, Table 3).

Fig. 1.

A: average distribution of oxygen saturation measured by pulse oximetry (Sp_O2) and tissue oxygen partial pressure (Pti_O2) during normoxic and hypoxic conditions in lean and obese mice. The red curves show normoxic group, the purple curves show infrequent intermittent hypoxia (12 times/h; IH12), the blue curves show frequent intermittent hypoxia (60 times/h; IH60), and the green curves show sustained hypoxia (SH). Sp_O2 and Pti_O2 values were normalized to 100% (%time spent) for each regimen. B: example tracings during frequent intermittent hypoxia in a lean mouse. The top curves show the Sp_O2 in arterial blood, the bottom curves show the Pti_O2 in the different tissues at the same time.

Table 2.

Mean oxygen levels in arterial blood and tissues

	Regimen	Lean Mice	Obese Mice	P Value
SpO2, %	Air	98.3 ± 0.1	98.0 ± 0.1	0.090
	IH12	92.2 ± 2.1	84.7 ± 1.7	0.022
	IH60	79.2 ± 1.2	73.4 ± 1.5	0.045
	SH	53.2 ± 1.6	43.9 ± 3.3	0.055
Liver PtiO2, Torr	Air	71.1 ± 3.6	61.6 ± 1.4	0.028
	IH12	60.0 ± 7.2	52.2 ± 4.0	NS
	IH60	39.9 ± 3.1	37.8 ± 2.5	NS
	SH	25.0 ± 2.2	22.4 ± 2.4	NS
Muscle PtiO2, Torr	Air	51.8 ± 5.7	52.0 ± 2.0	NS
	IH12	37.9 ± 7.8	36.7 ± 4.3	NS
	IH60	26.0 ± 4.4	24.2 ± 2.9	NS
	SH	17.7 ± 4.4	13.7 ± 6.0	NS
Epididymal fat PtiO2, Torr	Air	56.3 ± 3.8	55.4 ± 3.4	NS
	IH12	47.1 ± 5.2	38.9 ± 3.6	NS
	IH60	26.7 ± 2.8	28.8 ± 2.0	NS
	SH	19.0 ± 2.8	18.9 ± 3.5	NS

Values are means ± SE; n = 11 lean mice, n = 12 obese mice. Oxygen saturation measured by pulse oximetry (Sp_O2) and tissue oxygen partial pressures (Pti_O2) are shown for normoxia (air), IH12, IH60, and SH. NS, not significant. Bolded P values are significant.

Table 3.

Variability in oxygen levels during intermittent hypoxia in arterial blood and tissues

		Nadir		Peak		Standard Deviation
Regimen	Tissue	Lean	Obese	Lean	Obese	Lean	Obese
IH12	SpO2, %	47.0 ± 7.5	32.5 ± 3.2	96.7 ± 0.9	94.1 ± 1.2	13.9 ± 2.3	16.8 ± 1.2
	Liver PtiO2, Torr	22.0 ± 6.3	15.8 ± 5.5	65.1 ± 7.0	59.4 ± 3.5	11.4 ± 1.6	11.3 ± 1.2*
	Muscle PtiO2, Torr	9.2 ± 5.1	8.0 ± 3.0	43.8 ± 9.2	43.0 ± 4.6	10.0 ± 1.9	9.7 ± 0.9*
	Epididymal fat PtiO2, Torr	21.7 ± 3.3	21.4 ± 2.7	55.8 ± 6.1	45.3 ± 3.5	10.0 ± 2.3	7.0 ± 1.5*
IH60	SpO2, %	49.2 ± 3.2	44.6 ± 3.1	96.6 ± 0.3	95.2 ± 1.0	16.7 ± 1.0	17.6 ± 1.0
	Liver PtiO2, Torr	16.8 ± 3.3	20.1 ± 2.7	59.9 ± 5.0	52.4 ± 4.0	13.1 ± 1.8	9.4 ± 1.2*
	Muscle PtiO2, Torr	11.2 ± 2.1	8.7 ± 3.2	37.7 ± 7.3	36.8 ± 3.9	8.9 ± 2.0*	9.0 ± 1.3*
	Epididymal fat PtiO2, Torr	21.1 ± 2.5	23.9 ± 2.0	31.7 ± 4.0	33.3 ± 2.7	3.7 ± 1.1*†	3.2 ± 0.8*†‡

Values are means ± SE; n = 11 lean mice, n = 12 obese mice. Sp_O2and Pti_O2 are shown for IH12 and IH60.

^*P < 0.05 compared with Sp_O2.

^†P < 0.05 compared with liver Pti_O2.

^‡P < 0.05 compared with muscle Pti_O2.

Hypoxyprobe-1 (pimonidazole hydrochloride) was used to assess hypoxia in the different tissues. Following injection, Hypoxyprobe-1 is distributed to all tissues and forms adducts with thiol containing proteins only in those cells that have a Pti_O2 < 10 Torr (29). Lean mice demonstrated no evidence of tissue hypoxia in the immunohistostaining (Fig. 2) or immunoblotting (not shown), regardless of hypoxic regimen or control condition.

Fig. 2.

Immunohistostaining of the liver with Hypoxyprobe (pimonidazole) counterstained with hematoxylin. Representative immunohistostaining is shown for both lean and obese mice in the normoxic group (air) and the IH60 group. Original magnification: ×200. Hypoxic areas are stained in brown (pimonidazole). Arrows mark hepatic vein. Scale bar is 50 μm.

Fasting blood glucose significantly increased by ∼20 mg/dl in all three hypoxic groups compared with normoxia, without changes in plasma insulin (Fig. 3, A and B). Hypoxia significantly increased plasma leptin levels, especially in the sustained hypoxia group (44-fold, Fig. 3C). There was no significant change in adiponectin at any hypoxic condition compared with normoxia (Fig. 3D). Plasma total cholesterol increased during all hypoxic regimens, but only intermittent hypoxia 12 times/h differed significantly from normoxia (Table 1). Plasma triglycerides were decreased in both intermittent hypoxia regimens. Lipid peroxidation was significantly increased by intermittent hypoxia 60 times/h in the liver, but not in muscle or fat (Fig. 4). There was a greater than twofold increase in TNF-α secretion by adipose tissue in the frequent intermittent hypoxia group compared with normoxia (Fig. 5). Hypoxia did not affect TNF-α expression in the liver of lean mice (Fig. 6).

Fig. 3.

Indexes of insulin resistance after different hypoxic regimens and 12-h fast. A: glucose. B: insulin. C: leptin. D: adiponectin. Values are means ± SE. Values are shown for normoxia (air), IH12, IH60, and SH. *P < 0.05 compared with lean air group. †P < 0.05 compared with obese air group.

Fig. 4.

Lipid peroxidation measured by malondialdehyde (MDA) levels after different hypoxic regimens. A: liver. B: muscle. C: epididymal fat. Values are means ± SE. Values are shown for normoxia (air), IH12, IH60, and SH. *P < 0.05 compared with lean air group. †P < 0.05 compared with obese air group.

Fig. 5.

Secretion of tumor necrosis factor-α (TNF-α) by adipose tissue ex vivo after different hypoxic regimens. Values are means ± SE. Values are shown for normoxia (air), IH12, IH60, and SH. *P < 0.05 compared with lean air group.

Fig. 6.

Results of the real-time PCR for TNF-α in liver tissue. Values are means ± SE. TNF-α levels are shown as mRNA expression levels normalized to 18S ribosomal RNA concentrations and then expressed as a ratio to the lean air group.

Obese mice.

Obese mice were intentionally grouped to yield a higher preexposure body weight in both intermittent hypoxia groups due to anticipated weight loss during hypoxic exposure (Table 1). After exposure, body weights were similar in all exposure groups. Hypoxic mice lost more weight than control animals, regardless of the hypoxic regimen. Absolute weight loss was similar to that of lean mice.

Obese mice showed a trend towards lower Sp_O2 than lean mice at normoxic conditions, although the difference was very small (Table 2). Compared with lean mice, obese animals had lower levels of mean Sp_O2 during all hypoxic exposures. Of interest, liver Pti_O2 at baseline conditions was lower in obese mice than in lean animals, but not during hypoxic exposures. Pti_O2 in muscle and fat was similar in both groups, regardless of experimental conditions. As expected, the mean oxygen levels in blood and all organs were highest in normoxia and declined progressively from normoxia to infrequent intermittent hypoxia, frequent intermittent hypoxia, and sustained hypoxia (P < 0.001). As in lean mice, comparison of Pti_O2 in different tissues showed that the mean Pti_O2 was significantly higher in the liver than muscle and fat (P < 0.001). Similar to lean mice, frequent intermittent hypoxia induced high-amplitude swings of oxygen levels in blood, liver, and muscle, but not in epididymal fat tissue (Table 3). Mean variability during frequent intermittent hypoxia was significantly less in all tissues compared with Sp_O2. In contrast to lean mice, these differences were also prominent during infrequent intermittent hypoxia. As in lean mice, the lowest variability was observed in epididymal fat during frequent intermittent hypoxia. In the livers of obese mice, Hypoxyprobe-1 staining revealed evidence of hypoxia, predominantly in the areas surrounding hepatic veins, which were present at all conditions, including normoxia (Fig. 2). Hypoxyprobe-1 staining in muscle and fat was negative in lean and obese mice, regardless of the condition (not shown).

As expected, fasting blood glucose was higher in obese than in lean mice at all experimental conditions. As in lean mice, fasting blood glucose was increased by acute hypoxia (Fig. 3A). Fasting insulin levels were increased by frequent intermittent hypoxia and sustained hypoxia, which did not occur in lean mice (Fig. 3B). Obese mice showed a greater than 100-fold increase in leptin levels compared with baseline in lean mice, which were not modified by acute hypoxia (Fig. 3C). Plasma adiponectin levels were decreased by obesity without significant effects of acute hypoxia (Fig. 3D). Plasma total cholesterol and triglyceride levels were higher in obese than lean mice across all experimental conditions. Hypoxic regimens did not affect cholesterol levels, and there was a reduction in triglyceride in the sustained hypoxia group (Table 1). Lipid peroxidation was similar to that in lean mice in all organs. There was a mild increase in malondialdehyde levels in fat tissue during frequent intermittent hypoxia compared with normoxia and infrequent intermittent hypoxia (Fig. 4). There was a 10-fold increase in TNF-α secretion rate by adipose tissue in obese mice compared with lean mice at baseline conditions (Fig. 5), but acute hypoxia had no further impact. There was a four- to sixfold increase in TNF-α liver mRNA levels in obese mice compared with lean mice at baseline conditions (Fig. 6), and acute hypoxia had no further impact.

DISCUSSION

To the best of our knowledge, this is the first study to determine the impact of different regimens of acute hypoxia on tissue oxygen levels in lean and obese mice. Several novel findings resulted from the study. First, frequent intermittent hypoxia caused large-amplitude oxygen swings in the liver, mirroring swings of Sp_O2, whereas fluctuations of oxygen levels were attenuated in muscle and nearly abolished in adipose tissue. Second, at baseline normoxic conditions, obesity led to lower oxygen levels in the liver, predominantly in the areas surrounding hepatic veins, whereas oxygen levels in muscle and fat did not differ between the lean and obese states. Third, during hypoxic exposures, oxygen levels in liver, muscle, and fat were similar in obese and lean mice. Several additional findings resulted from the study. In lean mice, all types of hypoxia induced insulin resistance and increased circulating leptin levels, but only frequent intermittent hypoxia (60 times/h) increased lipid peroxidation and TNF-α production by adipose tissue. In obese mice, frequent intermittent hypoxia and sustained hypoxia augmented insulin resistance, but obesity-induced increases in plasma leptin levels and TNF-α secretion overwhelmed any effect of acute hypoxia. In the discussion below, we will further elaborate on significance of these findings.

Patterns of hypoxia in different tissues during intermittent hypoxia.

Oxygen levels in various tissues at baseline normoxic conditions have been previously measured in humans and several animal species, including cats, guinea pigs, dogs, rabbits, and rats (49). Mean Po₂ values ranged from 1–3 Torr in rat pons to 20–30 Torr in rat liver, and 30–35 mmHg in dog liver to 60–70 mmHg in dog kidneys. We report relatively high baseline liver Pti_O2 in lean mice of 71 Torr (Table 2). Using the same type of oxygen electrode and the same technique, Almendros et al. (2) recently reported baseline oxygen levels of 40–45 Torr in rat skeletal muscle and ∼50 Torr in rat visceral fat, which was ∼5 Torr lower than in our study (Table 2). Thus discrepancies between presently reported and previously described baseline values could be attributed to a different type of electrode used in the past (49) or interspecies differences, since none of the previous measurements were done in mice.

An important implication of the present work is that it describes patterns of oxygen in different tissues during intermittent hypoxia. A growing body of clinical and experimental literature previously assumed a priori that fluctuations in blood oxygen observed in intermittent hypoxia and sleep apnea were translated into oxygen swings in different organs and tissues. There was a general assumption that the pathogenesis of intermittent hypoxia and sleep apnea is similar to ischemia-reperfusion injury (22, 23, 31). The effects of intermittent asphyxia and intermittent hypoxia on oxygen levels in cerebral cortex, skeletal muscle, and visceral fat of a rat have recently been examined (2, 3). The report by Almendros et al. (2) suggests that fluctuations of blood oxygen levels during intermittent asphyxia and intermittent hypoxia result in significant oxygen swings in the brain and skeletal muscle, whereas the amplitude of the swings is attenuated in visceral fat. The present study established that intermittent hypoxia indeed generates hypoxia-reoxygenation in liver and skeletal muscle, whereas white adipose tissue exhibits a sustained hypoxia pattern, especially during frequent intermittent hypoxia. The sustained pattern could be attributed to relatively poor perfusion of adipose tissue (32).

Over the last decade, a number of papers have been published attempting to address molecular mechanisms of intermittent hypoxia utilizing cell culture (27, 30, 36, 40, 57). However, exposure regimens were arbitrary in these systems, because Pti_O2 in vivo during intermittent hypoxia had been unknown. Based on the present data, a realistic modeling of in vivo conditions in intermittent hypoxia cell culture can be employed.

Effects of different regimens of hypoxic exposure.

Compared with infrequent intermittent hypoxia, frequent intermittent hypoxia and sustained hypoxia induced more severe hypoxia in insulin-sensitive tissues, liver, skeletal muscle, and white adipose tissue (Fig. 1A). Infrequent intermittent hypoxia induced similar Sp_O2 and Pti_O2 nadirs and peaks to frequent intermittent hypoxia, but the mean Sp_O2 and Pti_O2 levels were much higher during the former regimen compared with the latter (Tables 2 and 3). An overshoot in Pti_O2 values during intermittent hypoxia compared with normoxia (Fig. 1A) is likely explained by compensatory hyperventilation immediately after the hypoxic episode (33) and a compensatory increase in cardiac output (1, 43). In lean mice, all types of hypoxia caused identical increases in fasting blood glucose without changes in serum insulin levels, suggesting that peripheral tissues are resistant to insulin action and pancreatic β-cells are incapable of an increase in insulin secretion. The authors of the present study have previously performed a euglycemic hyperinsulinemic clamp in mice and have shown that acute intermittent hypoxia (60 times/h over 9 h) induces insulin resistance (14). Hypoxic injury of pancreatic β-cells during intermittent hypoxia 60 times/h has also been reported in mice (52, 54). Louis and Punjabi (26) have utilized an intravenous glucose tolerance test in healthy human volunteers and reported similar finding of decreased insulin sensitivity and the lack of adequate increase in pancreatic insulin secretion after intermittent hypoxia. The present study shows that even infrequent episodic hypoxia (12 times/h) is sufficient to disturb insulin and glucose regulations in lean mice.

Suppression of insulin-sensitizing adipokine adiponectin (20) has been invoked as a possible cause of insulin resistance in OSA and intermittent hypoxia (27). The present study found no evidence that adiponectin is involved in intermittent hypoxiainduced insulin resistance, but the levels of biologically active high-molecular-weight adiponectin were not measured. Mechanisms of pancreatic endocrine insufficiency in intermittent hypoxia are not clear, but one possibility is leptin, which suppresses insulin secretion (21, 44). In lean mice, all types of hypoxia caused striking increases in serum leptin, 17-fold in infrequent intermittent hypoxia, 9-fold in frequent intermittent hypoxia, and 44-fold in sustained hypoxia. The authors of the present study have previously shown that intermittent hypoxia increases leptin levels (24, 35). Notably, human OSA is associated with high levels of circulating leptin (15), which can be decreased by treatment with continuous positive airway pressure (7, 13). Overall, the present study shows that infrequent episodic hypoxia was sufficient to induce hyperleptinemia in lean mice, and that sustained hypoxia induced both more severe tissue hypoxia and higher leptin levels than intermittent hypoxia.

Frequent chronic intermittent hypoxia induces oxidative stress in different organs and systems. In rats, chronic intermittent hypoxia causes myocardial oxidative stress that is associated with cardiac dysfunction (6, 51). In contrast to rats, mice are relatively protected from myocardial oxidative stress during intermittent hypoxia (17), but develop oxidative stress in the liver. In the present study, frequent intermittent hypoxia was the only type of exposure that induced a small (20%) increase in lipid peroxidation in liver tissue of lean mice, whereas muscle and fat were not affected. The unique effect of frequent intermittent hypoxia on the mouse liver could be attributed to more vigorous production of reactive oxygen species during large and frequent swings of Pti_O2 levels, which were attenuated in muscle and fat. Liver Pti_O2 swings did not occur during sustained hypoxia and happen with a lower frequency during infrequent intermittent hypoxia. Of interest, an increase in liver lipid peroxidation did not occur in obese mice, which might be related to the lower amplitude of oxygen fluctuations. Thus the present findings suggest that an increase in hepatic lipid peroxidation in lean mice may be related to frequent high-amplitude swings of Pti_O2.

Another unique feature of frequent intermittent hypoxia was inflammatory response manifested by an increase in TNF-α secretion by adipose tissue. This increase did not occur during sustained hypoxia, despite mean fat Pti_O2 being lower than nadirs of Pti_O2 in both intermittent hypoxic groups (Tables 2 and 3). The inflammatory response cannot be attributed to hypoxia-reoxygenation, because there was virtually no fluctuation in Pti_O2 levels in fat. The present findings suggest that excessive TNF-α secretion during frequent intermittent hypoxia is likely attributable to systemic inflammatory response rather than adipose tissue hypoxia per se.

Effects of obesity.

Pti_O2 levels in the liver in lean and obese mice at baseline have not been compared previously. The present study showed that, at normoxic conditions, Pti_O2 levels in obese livers were, on average, 10 Torr lower than in lean livers (Table 2). Pimonidazole staining revealed that hypoxic areas were in the centers of hepatic lobules around hepatic veins (zone 3). Previous studies have shown that liver tissue is not homogeneous, and hepatocytes surrounding hepatic vein are markedly more hypoxic than periportal hepatocytes (4, 48). The novelty of the present findings is that zone 3 hypoxia was much more prominent in obese mice. Obese mice had higher baseline fasting glucose levels than lean mice, suggesting higher hepatic glucose output. Hypoxia upregulates a key enzyme of hepatic gluconeogenesis, phosphoenolpyruvate carboxykinase (8), but the role of tissue hypoxia in obesity-induced gluconeogenesis has not been explored. Thus obesity is associated with hypoxia in liver tissue.

Obesity did not induce hypoxia in epididymal fat and skeletal muscle at baseline conditions. While effects of obesity on Pti_O2 in skeletal muscle have not been previously studied, the present data in fat differ from previous reports in mice (53). However, Ye et al. (53) measured Pti_O2 in ob/ob mice, which are more obese than mice with diet-induced obesity. Lacking leptin, ob/ob mice have poor adipose tissue perfusion (50), because the lack of leptin induced angiogenesis (5, 45). Of note, human studies also consistently show that obesity induces adipose tissue hypoxia (9, 19, 32). It is conceivable that, if Pti_O2 would have been measured in the present study in other adipose tissue depots (omental, retroperitoneal, or subcutaneous), hypoxia might have been detected.

The effect of obesity on Pti_O2 during hypoxic exposure has been previously unknown. Obese mice showed more severe oxyhemoglobin desaturation than lean mice during intermittent hypoxia (Table 2). In context of this finding, it is surprising that obese and lean mice exhibited the same liver, muscle, and fat oxygen levels during all types of hypoxic exposure. It is conceivable that tissue oxygenation is preserved in obesity by a mechanism yet to be determined.

Another important finding of this study was that, despite a similar impact on tissue oxygen levels in obese and lean mice, metabolic effects of hypoxia in obese animals were modest. We propose that effects of acute hypoxia were likely overwhelmed by obesity. Indeed, obesity induced striking changes in adipokine production and increased systemic inflammation. Compared with lean mice, obese mice showed a 170-fold increase in plasma leptin, 11.5-fold increase in adipose TNF-α secretion, a 4- to 6-fold increase in TNF-α expression in the liver, and a 2-fold decrease in plasma adiponectin. Notably, the magnitude of the impact of obesity greatly exceeded effects of hypoxia. The only metabolic effect of acute hypoxia, which was present in obese mice, was insulin resistance. Unlike lean mice, obese mice nearly doubled serum insulin levels in response to frequent intermittent hypoxia and sustained hypoxia, suggesting that pancreatic β-cell secretion was preserved. Obese mice did not increase leptin levels during intermittent and sustained hypoxia, and the lack of inhibitory action of leptin could be accountable for preserved insulin secretion (21, 44). Overall, obesity did not augment levels of hypoxia in metabolically active tissues during intermittent and sustained hypoxia, and the effects of obesity overwhelmed and masked the majority of proinflammatory and metabolic effects of acute hypoxia, except insulin resistance.

Limitations.

The present work has a number of limitations. First, we measured Pti_O2 during acute hypoxic exposure. It is conceivable that chronic intermittent hypoxia will result in a different Pti_O2 tissue profile. Metabolic and proinflammatory responses to chronic intermittent hypoxia, a hallmark of OSA, may also differ from acute hypoxia. Second, the sustained hypoxia regimen was more severe than both intermittent hypoxia regimens. However, despite the lowest mean inspired O₂ fraction, Sp_O2, and tissue Pti_O2, sustained hypoxia induced similar increases in insulin resistance, whereas inflammatory and oxidative stress responses were less prominent, compared with intermittent hypoxia. Third, Pti_O2 measurements were performed under general anesthesia. Nevertheless, during both intermittent hypoxic regimens, Sp_O2 was rebounding to normoxic values between the hypoxic events, suggesting that the observations may be pertinent for unanesthetized animals. Fourth, the probe measured oxygen tension within a small random area, but Pti_O2 varies within each organ and tissue. For instance, in the liver, oxygen tension is lowest in the area surrounding the hepatic vein (18). Immunohistochemistry with Hypoxyprobe was used as a complementary approach. Fifth, immunostaining with Hypoxyprobe has a low sensitivity, detecting only severe hypoxia with Pti_O2 < 10 Torr. Finally, OSA leads to nocturnal hypercapnia, whereas, in the present model, mice were either hypocapnic or normocapnic (41).

Conclusions and implications.

Frequent acute intermittent hypoxia caused reciprocal oxygen swings in liver and skeletal muscle, whereas fluctuations of Pti_O2 were absent in adipose tissue. Intermittent and sustained hypoxia resulted in similar levels of tissue hypoxia and insulin resistance in lean and obese mice. In lean mice, all types of hypoxia raised serum leptin levels, whereas only frequent intermittent hypoxia increased oxidative stress and adipose inflammation. Obesity induced liver hypoxia at baseline and caused a profound impact on adipokine and TNF-α production, which overwhelmed effects of intermittent hypoxia. The present data suggest that infrequent and frequent intermittent hypoxia have similar effects on insulin resistance and adipokine secretion. Given that high-frequency fluctuations of O₂ in cell culture are not possible due to a limited rate of gas perfusion in the media, the present work demonstrates that low-frequency swings of O₂ achievable in cell culture have biological relevance for OSA.

GRANTS

This work was supported by Research Fellowship Grant RE 2842/1-1 of the German Research Foundation to C. Reinke; Conselho Nacional de Desenvolvimento Científico e Technológico Grant 200032/2009-7 and Fundação Zerbini, Brazil, to L. F. Drager; and National Institutes of Health (NIH) Grant HL-80105 and American Heart Association Grant 10GRNT3360001 to V. Y. Polotsky. This work was also partly supported by NIH Specialized Center of Clinically Oriented Research Grant 5P50-HL-084945 and by the Mid-Atlantic Nutrition Obesity Research Center (formerly Clinical Nutrition Research Unit of Maryland) NIH Grant P30 DK072488.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).