Influences of Hypoxia Exercise on Whole-Body Insulin Sensitivity and Oxidative Metabolism in Older Individuals

Abstract

Context

Aging is a primary risk factor for most chronic diseases, including type 2 diabetes. Both exercise and hypoxia regulate pathways that ameliorate age-associated metabolic muscle dysfunction.

Objective

We hypothesized that the combination of hypoxia and exercise would be more effective in improving glucose metabolism than normoxia exercise.

Design and Participants

We randomized 29 older sedentary individuals (62 ± 6 years; 14 women, 15 men) to bicycle exercise under normobaric hypoxia (fraction of inspired oxygen = 15%) or normoxia (fraction of inspired oxygen = 21%).

Intervention

Participants trained thrice weekly for 30 to 40 minutes over 8 weeks at a heart rate corresponding to 60% to 70% of peak oxygen update.

Main Outcome Measures

Insulin sensitivity measured by hyperinsulinemic-euglycemic glucose clamp and muscle protein expression before and after hyperinsulinemic-euglycemic glucose clamp.

Results

Heart rate and perceived exertion during training were similar between groups, with lower oxygen saturation when exercising under hypoxia (88.7 ± 1.5 vs 96.2 ± 1.2%, P < 0.01). Glucose infusion rate after 8 weeks increased in both the hypoxia (5.7 ± 1.1 to 6.7 ± 1.3 mg/min/kg; P < 0.01) and the normoxia group (6.2 ± 2.1 to 6.8 ± 2.1 mg/min/kg; P = 0.04), with a mean difference between groups of –0.44 mg/min/kg; 95% CI, –1.22 to 0.34; (P = 0.25). Markers of mitochondrial content and oxidative capacity in skeletal muscle were similar after training in both groups. Changes in Akt phosphorylation and glucose transporter 4 under fasting and insulin-stimulated conditions were not different between groups over time.

Conclusions

Eight weeks of hypoxia endurance training led to similar changes in insulin sensitivity and markers of oxidative metabolism compared with normoxia training. Normobaric hypoxia exercise did not enhance metabolic effects in sedentary older women and men beyond exercise alone.

Aging increases the risk for many chronic and cardiometabolic diseases. Sarcopenia, reduced skeletal muscle performance, and diminished oxidative phosphorylation capacity are hallmarks of aging (1–3). Consequently, the probability for impaired exercise tolerance and impaired insulin sensitivity increases (2, 4). Aerobic exercise improves cardiovascular and metabolic health and holds the potential to slow down several age-associated changes in skeletal muscle properties and function (5). Endurance exercise under hypoxia could be even more effective (6–8). Indeed, hypoxia and aerobic exercise regulate the activation of hypoxia inducible factor-1α (HIF-1α) and peroxisome proliferator-activated receptor γ coactivator 1α (9–11). Both pathways ameliorate metabolic muscle dysfunction observed with aging (5, 12–14). In patients with type 2 diabetes, a single exercise bout under hypoxia as well as intermittent hypoxia breathing improved glucose metabolism compared with breathing room air (8, 15). In healthy young and middle-aged subjects, regular hypoxia exercise improved markers of glucose tolerance and insulin resistance (16–18). These results, together with animal and molecular data, raised the hope that hypoxia exercise attenuates the risk for insulin resistance and type 2 diabetes (10, 12, 13, 19–21). Effective training at reduced workload is a potential advantage of hypoxia exercise (17, 22); thereby, less stress may be imposed on the locomotor system while realizing similar beneficial cardiovascular and metabolic effects compared with normoxia exercise. We tested the hypothesis that combining normobaric hypoxia and exercise is more effective in improving whole-body insulin sensitivity and metabolic skeletal muscle function in older sedentary volunteers than exercise alone.

Research Design and Methods

Experimental design

This prospective, randomized, parallel-group and single-blinded trial was conducted between June 2015 and September 2018 at Hannover Medical School. Women and men aged 55 to 75 years were eligible for participation if they were weight stable during the past 6 months (±2% body weight), had a body mass index between 20 and 35 kg/m², and a homeostasis model assessment insulin resistance index (HOMA-IR) at screening between 2.0 and 4.0 U. We screened for preexisting diseases by a detailed medical history and examination, blood pressure evaluation, 12-lead ECG, and blood sampling for routine laboratory tests. Exclusion criteria were more than 1 hour of scheduled exercise training per week; known diagnosis of type 2 diabetes or measured hemoglobin A1c >6.5% (>48 mmol/mol); known alcohol or drug abuse; acute or chronic infections; increased bleeding risk by history or laboratory testing; any contraindication (e.g., orthopedic, cardiopulmonary) to perform exercise training. The institutional review board of Hannover Medical School approved the study, and written informed consent was obtained from each participant before enrollment.

Study protocol

Participants were advised to maintain their physical activity level and dietary habits throughout the study to avoid bias introduced by lifestyle changes. After baseline evaluations, participants performed 8 weeks endurance training under either normobaric hypoxia (fraction of inspired oxygen, 103 mm Hg, corresponding to 2750 m altitude) or normoxia (fraction of inspired oxygen, 150 mm Hg). Participants were centrally randomized (by the Institute of Biometry) 1:1 using a computer-generated list of random numbers stratified for sex. For randomization, variable block length was used to avoid selection bias from predictability. The primary outcome was the glucose infusion rate during a 3-hour hyperinsulinemic-euglycemic glucose clamp (HEGC), an established measure of whole-body insulin sensitivity. Investigators assessing the primary end point were blinded for allocation of participants. After completion of the training program, we repeated all outcome measurements conducted at baseline.

Anthropometric, cardiopulmonary, and metabolic measurements

At screening, after an overnight fast, a general medical examination by a physician was performed, body weight, waist circumference, and height were measured in a standardized fashion. Sitting heart rate and blood pressure were assessed in triplicate after a 5-minute rest in a quiet room with an automated device (Dinamap, GE Healthcare, Madison, WI). A 12-lead ECG reading was obtained in the supine position (Cardiovit CS-200, Schiller, Geisenfeld, Germany). Venous blood samples were obtained for routine laboratory tests and to calculate HOMA-IR [insulin (µU/mL) × glucose (mmol/L)]/22.5 (23).

During study visit 1, body fat and fat-free mass were determined by air-displacement plethysmography (BodPod, Life Measurement, Inc., Concord, CA) and venous blood samples were obtained. All participants performed an incremental exercise test on a bicycle ergometer (Ergoline 900, Bitz, Germany) until voluntary exhaustion. Workload started at 30 W and increased every 1 minute by 10 W until the requested 60 rpm pedal frequency could not be maintained. Brachial blood pressure, ECG, and gas exchange breath-by-breath (MasterScreen CPX, VIASYS Healthcare, Conshohocken, PA) were monitored, and arterialized blood samples were collected from the ear lobe to determine lactate concentrations. Peak oxygen uptake (VO_2peak) was documented as the highest averaged 15-second oxygen uptake interval throughout the test.

On a second visit, scheduled at least 48 hours later, participants came to the laboratory in the morning after an overnight fast. They refrained from strenuous physical activity and alcohol and caffeine intake for the previous 24 hours. Muscle biopsies from the vastus lateralis were taken before the 3-hour HEGC using a spring needle technique. Samples were shock frozen in liquid nitrogen and stored at −80°C until analysis. After the biopsy, participants were instrumented for the 3-hour HEGC. The procedure consisted of a primed infusion of insulin followed by a continuous infusion and a variable glucose infusion to achieve euglycemic conditions as previously described in detail (24). Arterialized glucose values were obtained every 5 minutes, and the last 30 minutes of the clamp (minutes 150 to 180) were considered as the steady-state period. During this period, the mean glucose infusion rate (mg/min) and the insulin sensitivity index (calculated from steady-state glucose infusion rates and plasma insulin and glucose concentrations) (24) were calculated. A second muscle biopsy was performed directly at the end of the 3-hour HEGC under continued insulin perfusion to obtain muscle tissue samples under hyperinsulinemic conditions.

At least 48 hours after HEGC, participants performed a 30-minute constant workload test (visit 3) on a bicycle ergometer with the workload set at 60% of individual VO_2peak as assessed during the incremental exercise test at visit 1. We recorded heart rate (ECG) and gas exchange and collected arterialized blood samples from the ear lobe to determine lactate concentrations. From breath-by-breath gas exchange, we calculated whole-body fat and carbohydrate oxidation using stoichiometric equations (25). The test was repeated with the same absolute workload (corresponding to 60% of pretraining VO_2peak) after the 8-week training period.

Training program

Following all baseline assessments, participants started the supervised endurance exercise program. According to randomization, participants trained either under normobaric hypoxia or normoxia in a temperature-controlled hypoxia chamber (Höhenbalance GmbH, Cologne, Germany), with ambient conditions (hypoxia vs normoxia) unrevealed to the participants. Oxygen and carbon dioxide concentrations within the chamber were continuously monitored by sensor electrodes throughout all training sessions. Participants trained on a bicycle ergometer 3 days per week for 8 weeks (24 training sessions). To achieve similar relative exercise intensities in both intervention groups, participants trained at a heart rate corresponding to 60% of pretraining VO_2peak for 30 minutes for the first 4 weeks. After 4 weeks, exercise intensity was increased to 70% of pretraining VO_2peak and exercise duration to 40 minutes. Oxygen saturation and heart rate were recorded every 10 minutes to calculate mean training values for all participants during all exercise sessions. Perceived exertion was rated using the Borg scale after any exercise session.

Laboratory tests

Plasma glucose, serum insulin, and serum lipids were measured by standard methods in a certified clinical chemistry laboratory. Arterialized whole blood glucose during the clamp was measured every 5 minutes on-site by sampling whole blood in a glass capillary and photometric assessment according to the manufacturer’s instructions (Glucose Analyzer Super GL compact, Hitado, Möhnesee, Germany). A portion of frozen muscle (∼30 mg) was prepared for immunoblot analysis for glucose transporter 4 (GLUT4) phospho-Akt (Ser⁴⁷³) normalized to total Akt (p-Akt), and OxPhos protein expression. The frozen muscle samples were homogenized in 1.5-mL tubes using ReadyPrep mini grinders (Bio-Rad Laboratories, Hercules, CA) and 200 µL of ice-cold cell lysis buffer (50 mM Tris HCl pH 6.8, 1.6% SDS, 7% glycerol, 4% β-mercaptoethanol, and 0.016% bromophenol blue) including a protease inhibitor cocktail (Bimake, Houston, TX). For determination of GLUT4, an additional 8 M urea was added to the lysis buffer to avoid dimer structures. The homogenates were incubated 15 minutes on ice and then centrifuged and the supernatant removed. Protein content of the supernatants was determined by Bradford assay using Roti-Nanoquant (Carl Roth, Karlsruhe, Germany). Aliquots of supernatant were mixed with 4x Laemmli buffer (8% SDS, 40% glycerol, 4% β-mercaptoethanol, 0.05% bromophenol blue, 0.24 M Tris-HCl pH 6.8), and denatured by heating. Samples were separated on a 4% to 20% SDS-PAGE gel followed by transfer onto Protran nitrocellulose membranes (GE Healthcare, Amersham, GB). Membranes were blocked in 5% nonfat milk and incubated with the following primary antibodies: anti-GLUT4 (cat. no. 2213, Cell Signaling, Danvers, MA), anti-Akt (cat. no. 4691, Cell Signaling), anti-p-Akt (Ser⁴⁷³) (cat. no. 4060, Cell Signaling), anti-OXPHOS antibody cocktail (cat. no. MS601, Mitosciences, Eugene, OR), and anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH; cat. no. MAB374, EMD Millipore, Billerica, MA). Membranes were then incubated in appropriate species-specific secondary antibodies (antimouse IgG-horseradish peroxidase cat. no. 115-035-003, or antirabbit horseradish peroxidase cat. no. 111-035-003, Dianova, Hamburg, Germany). Protein bands were visualized using a LAS-3000 imaging system (Fujifilm, Tokyo, Japan) and analyzed with ImageJ software (Wayne Rasband, National Institutes of Health). Protein loading was controlled by normalizing bands of interest to GAPDH.

For measuring enzyme activity by spectral photometry, a postnuclear supernatant was prepared from s muscle biopsy specimen as described by our group before (26). Citrate synthase activity as a mitochondrial marker enzyme was measured according to Srere, mitochondrial complex IV activity (cytochrome c oxidase) and V (ATP synthase) activities were measured as described previously (26). All enzyme activity values were normalized to protein content of the lysates.

Statistical analysis

For descriptive analysis, absolute frequencies were calculated for categorical variables and mean with SD for continuous variables. Differences between study groups in baseline parameters were compared by Student t test for unpaired samples or the χ² test. For the primary outcome (glucose infusion rate), the changes between baseline and 8 weeks were calculated and considered as response variable in an analysis of covariance (ANCOVA). Explanatory covariables included insulin sensitivity at baseline (continuous), age (continuous), sex (binary), and the study group allocation (binary). The difference between groups over time was given as adjusted mean (for age, sex, baseline value) with 95% CIs. For secondary outcomes, the same analysis method was used. All analyses were carried out in the intention-to-treat population (all randomized participants who participated in at least one training session). Missing values were replaced with the baseline measurements (baseline observation carried forward). As sensitivity analysis, we also performed a per-protocol analysis (only participants with pre- and postintervention data). To test for within-group differences from baseline to end of intervention, a two-sided Student t test for paired samples was used. Univariate associations between parameters were tested using Pearson correlation coefficient. The type I error was set to 5% (two-sided). All statistical analyses were performed with IBM SPSS 25 Statistics (IBM Corporation, Armonk, NY).

Results

We screened 68 older individuals for eligibility. Of those, 29 were randomized and underwent pretraining measurements. Twenty-five (86%) subjects completed the entire intervention period and four dropped out (for details, see Fig. 1). Anthropometric and laboratory baseline characteristics for both training groups are given in Table 1. Groups were well matched for age, sex distribution, body mass index, fasting glucose concentrations, daily physical activity level, and cardiorespiratory fitness; all participants were nonsmokers.

Figure 1.

Participant flow throughout the study.

Table 1.

Subject Characteristics at Baseline

	Hypoxia Group	Normoxia Group
Subjects (women/ men), n	14 (7/7)	15 (7/8)
Age, y	60.4 ± 5.1	63.8 ± 5.8
Body weight, kg	86.7 ± 10.3	88.4 ± 14.5
Body mass index, kg/m2	28.6 ± 3.0	28.3 ± 1.9
Body fat mass, kg	31.6 ± 8.1	32.2 ± 7.0
Heart rate, beats/min	68 ± 7	67 ± 9
Fasting glucose, mmol/L	5.9 ± 0.5	5.8 ± 0.8
Triglycerides, mmol/L	1.47 ± 0.81	1.44 ± 0.86
Hypertension, n (%)	3 (21)	4 (27)
Dyslipidemia, n (%)	8 (57)	9 (60)
Total physical activity, MET-h/wk	67 ± 51	76 ± 35
VO2peak, mL/min/kg	24.1 ± 5.3	23.2 ± 5.0

Data are mean ± SD. Hypertension: systolic BP values ≥140 mm Hg and/or diastolic BP values ≥90 mm Hg. Dyslipidemia: low-density lipoprotein >3.0 mmol/L and/or triglycerides >1.7 mmol/L. No significant differences between groups as analyzed with Student paired t test for unpaired samples or the χ² test.

Abbreviations: BP, blood pressure; MET, metabolic equivalent of task; VO₂, oxygen uptake.

Adherence to the training schedule was 22.8 ± 2.6 days for participants in the hypoxia group (95%) and 23.0 ± 2.7 days in the normoxia group (96%). During the 8-week training sessions, heart rate (hypoxia, 114 ± 14 beats/min; normoxia, 115 ± 10 beats/min; P = 0.90), heart rate relative to maximum heart rate (Fig. 2), training workload (hypoxia, 0.82 ± 0.25 W/kg body weight; normoxia, 0.94 ± 0.24 W/kg body weight; P = 0.23), and perceived exertion during exercise were not significantly different between training groups, whereas arterial oxygen saturation was lower during hypoxia training sessions (Fig. 2).

Figure 2.

Relative heart rate (top), subjective rating of exertion using the Borg scale (center), and arterial oxygen saturation (bottom), averaged from all performed exercise sessions for each subject randomized to training in normoxia or in hypoxia. *P < 0.01. HRmax, maximum heart rate.

After the 8-week intervention, the glucose infusion rate had increased in both groups (hypoxia, 500 ± 125 to 574 ± 142 mg/min, P < 0.001; normoxia, 552 ± 210 to 606 ± 226 mg/min, P = 0.049) with no significant difference over time between groups (mean difference, –34.7 mg/min; 95% CI, –106.6 to 37.1; P = 0.33). Changes in glucose infusion rate normalized for body weight (M value, Fig. 3), HOMA-IR (Table 2), and the insulin sensitivity index (hypoxia, 1.28 ± 0.29 to 1.45 ± 0.30 μg/kg/min/(mM × pM), P < 0.05; normoxia 1.39 ± 0.46 to 1.55 ± 0.57 μg/kg/min/(mM × pM), P = 0.28, between-group P = 0.87) were also not different between study groups. A protocol analysis for the primary outcome (only participants with complete pre- and postintervention data, n = 25) revealed similar results for changes in glucose infusion rate (mean difference between groups over time, –43.8 mg/min; 95% CI, –124.8 to 37.1; P = 0.27).

Figure 3.

Individual changes of the glucose infusion rate normalized for body weight (M value) from baseline (set at zero) to 8 weeks of exercise training in subjects randomized to training in normoxia (left) or in hypoxia (right). Framed P values indicate differences from baseline to 8 weeks of training for each group as assessed with Student paired t tests. No important difference between groups over time was detected with an ANCOVA controlled for baseline values, age, and sex.

Table 2.

Cardiometabolic Risk Factors and Exercise Test Parameters Before and After 8 Weeks of Training

	Hypoxia Group		Normoxia Group		P Value
	Before	After	Before	After	(Group × Time)
Risk factors
Waist circumference, cm	107 ± 7	104 ± 6	107 ± 11	105 ± 9	0.98
Body fat, %	36.4 ± 8.4	36.2 ± 8.8	37.7 ± 7.2	36.9 ± 7.3	0.47
Total cholesterol, mmol/L	5.5 ± 1.2	5.5 ± 1.0	5.5 ± 0.7	5.4 ± 0.8	0.52
LDL cholesterol, mmol/L	3.8 ± 1.1	3.8 ± 0.9	3.7 ± 0.9	3.6 ± 0.8	0.98
HOMA-IR, U	3.5 ± 1.4	3.0 ± 1.0	2.9 ± 1.5	2.7 ± 1.6	0.42
Incremental exercise test
VO2peak, mL/kg/min	24.1 ± 5.3	26.2 ± 6.5	23.2 ± 5.0	25.0 ± 5.4a	0.60
Power output maximum, watt/kg	1.67 ± 0.22	1.86 ± 0.28a	1.80 ± 0.43	1.90 ± 0.46a	0.13
RQ max, units	1.20 ± 0.09	1.18 ± 0.06	1.19 ± 0.08	1.18 ± 0.10	0.65
Lactate maximum, mmol/L	6.7 ± 2.2	6.8 ± 2.3	6.4 ± 2.2	6.4 ± 2.1	0.20
Constant workload test
Work load, W	76 ± 24	76 ± 24	75 ± 34	75 ± 34	1.00
Lactate, mmol/L	2.5 ± 1.1	2.0 ± 1.2a	2.2 ± 0.8	1.7 ± 0.7a	0.82
RQ, U	0.97 ± 0.06	0.94 ± 0.06a	0.95 ± 0.04	0.92 ± 0.04a	0.88
CHO oxidation, g/min	14.9 ± 3.9	13.3 ± 4.3a	16.7 ± 6.3	12.8 ± 6.3a	0.26
Fat oxidation, g/min	1.2 ± 0.9	1.8 ± 1.2a	0.8 ± 1.4	1.6 ± 1.4a	0.80

Data are mean ± SD and are from an incremental exercise test until voluntary exhaustion and a constant work load test at 60% of baseline VO_2peak on a bicycle ergometer.

Abbreviations: CHO, carbohydrates; LDL, low-density lipoprotein; RQ, respiratory quotient; VO₂, oxygen uptake.

^a

P < 0.05 compared with before training as analyzed with Student paired t test for paired samples; differences between groups over time analyzed with an ANCOVA controlled for baseline values, age, and sex.

Body weight (hypoxia group, –0.8 ± 1.9 kg; normoxia group, –0.4 ± 1.3 kg; between-group P = 0.86) and body composition (Table 2) did not change significantly with training irrespective of oxygen conditions. For cardiorespiratory and metabolic parameters assessed during exhaustive incremental exercise tests, no differences between groups over time were detected (Table 2). During constant workload exercise (60% VO_2peak), fat oxidation was enhanced after 8 weeks training, whereas carbohydrate oxidation decreased (Table 2), but changes in both parameters were similar between study groups (Table 2).

Paired skeletal muscle samples in fasting and insulin-stimulated conditions were obtained from eight hypoxia-training and seven normoxia-training participants. The other participants refused muscle biopsies. At baseline, the M value was correlated to the activity of mitochondrial respiratory chain complex IV (r = 0.48, P < 0.05) and complex V (r = 0.60, P < 0.05), but not to citrate synthase activity (r = 0.36, P = 0.15). After 8 weeks training, citrate synthase and mitochondrial complexes IV and V activities did not change significantly in both study groups, and not between groups over time (Fig. 4). Immunoblotting of GLUT4 showed a significant increase after the intervention for the combined group under fasting conditions (fold change to baseline 0.25 ± 0.20, P < 0.01), but not during hyperinsulinemic conditions. For p-Akt, no changes were observed during fasting or during insulin-stimulated conditions [Fig. 5(a)]. Between-group differences were not observed for GLUT4 and p-Akt. Similar to enzyme activities, respiratory chain complexes I to V content showed no substantial changes after 8 weeks training in both groups, nor were differences detected between groups over time [Fig. 5(b)].

Figure 4.

Activity for citrate synthase and complex IV (cytochrome c oxidase) and V (ATP synthase) of the respiratory chain measured in skeletal muscle tissue lysates normalized by protein content of the lysates for subjects randomized to training in normoxia or in hypoxia. No significant differences were detected from pre- to posttraining as assessed with Student paired t test or between groups over time assessed with an ANCOVA controlled for baseline values, age, and sex. Data are mean ± SE.

Figure 5.

(a) Skeletal muscle phospho-Akt serine⁴⁷³ normalized to total Akt (top) and GLUT4 normalized to GAPDH (bottom) under fasting and insulin-stimulated conditions. Data represent fold changes to bsl values under fasting conditions. Representative Western blots of p-Akt and GLUT4 are shown. (b) OxPhos under fasting conditions from skeletal muscle. Data represent fold changes to baseline values. A representative Western blot is shown. For all parameters, no within-group differences from baseline to 8 wk were detected with Student paired t test for paired samples. No between-group differences over time were detected as analyzed with an ANCOVA controlled for baseline values, age, and sex. Data are mean ± SE. bsl, baseline; OxPhos, respiratory chain complexes I to V.

Discussion

The important finding of our study is that both normoxia and hypoxia endurance training over 8 weeks improved whole-body insulin sensitivity in older sedentary individuals. We also observed a modest increase in fat oxidation during exercise and an increase in skeletal muscle GLUT4 content; however, normobaric hypoxia exercise did not further improve any of these responses compared with exercise alone. The finding can neither be explained by group differences at baseline nor by differences in exercise workload or cardiopulmonary strain during the training sessions.

Aging is associated with worsened insulin sensitivity in skeletal muscle (4, 27). Insulin resistance not only affects glucose homeostasis. Insulin stimulates amino acid transport across the muscle membrane, stimulates protein synthesis, and inhibits protein breakdown (28). Thus, age-related muscle metabolic dysfunction and frailty may share common cellular mechanisms. Muscle contraction and hypoxia regulates pathways that are crucial to glucose and lipid metabolism. At the molecular level, both stimuli increase HIF-1 expression (9, 11). HIF-1α targets genes involved in glycolysis, glucose transport, and mitochondrial function (11, 12, 19, 29), which have all been implicated in insulin resistance (4, 30). Recently, a study provided further insight into the molecular regulation of glucose metabolism and insulin action during muscle contraction, implicating a key role for HIF-1α in this process by directly controlling the transcription of RAB20 and TXNIP (13).

Our study confirms physical exercise as an efficacious measure to enhance insulin sensitivity, in particular in the less-studied cohort of older individuals (31). Yet, our results also suggest that applying hypoxia during exercise does not lead to a major additional effect on whole-body insulin sensitivity. Previous data indicate greater improvements of insulin sensitivity or glycemic control after acute (8, 15) and regular exercise training under hypoxic conditions (16–18). Therefore, the negative outcome of our study is somewhat unexpected. Possibly, adaptations to the delivery, transport, and intramyocellular metabolism of glucose and the insulin signaling pathway (32, 33) might have been sufficiently large with exercise training alone. Pancreatic beta-cell function, inflammatory processes, or muscular mechanical stretch during exercise are further candidates that are not necessarily coupled to hypoxic signaling pathways, leading to improved insulin action and glucose handling after regular training (34–37). Of note, the positive outcomes on markers of glucose homeostasis after regular hypoxia exercise so far were observed in young or in well-trained individuals (10, 17, 18). Given that HIF-1α expression is reduced after regular training (38), additional hypoxia could have a stronger effect in trained individuals by restoring exercise-induced HIF-1α expression compared with untrained subjects with already high HIF-1α expression under normoxic conditions. Our subjects trained at moderate intensities as recommended by current guidelines. Yet, higher exercise intensity during hypoxia might be beneficial and possibly result in greater effects as suggested by some studies (6, 39). Finally, whereas previous studies used fasting blood samples or glucose tolerance tests to estimate glycemic control and insulin action (16–19), we applied HEGCs to more directly estimate whole-body´s insulin sensitivity.

With advancing age, impaired mitochondrial function is evident in skeletal muscle, which is thought to be one reason for age-related insulin resistance (4, 40). In this line, we observed that cytochrome c oxidase and ATP synthase activity, as markers of muscle oxidative phosphorylation capacity, were related to HEGC-estimated whole-body insulin sensitivity at baseline. After exercise training, some studies observed improved mitochondrial function or content for hypoxia compared with normoxia conditions (7, 10), whereas others did not (39, 41, 42). Yet, all of these studies were performed in young and apparently healthy subjects. In our group of older individuals, we did not observe substantial changes of mitochondrial respiratory complex activity or protein content, which might be, at least partly, from a diminished exercise-induced response on mitochondrial biogenesis with aging (43).

Akt and GLUT4 are components of insulin-stimulated glucose uptake and are, to a varying extent, dysregulated in insulin-resistant skeletal muscle (30, 44). As expected, Akt phosphorylation and GLUT4 increased after acute insulin infusion. However, the extent of the insulin-induced increase was not different following hypoxia or normoxia endurance training. The observation that changes in Akt and GLUT4 protein content were independent of inspired oxygen content during exercise parallels our clinical results for whole-body insulin sensitivity. Cross-sectional studies reported increases in glucose uptake or insulin sensitivity resulting from hypoxic treatment or combination of hypoxia with exercise (12, 13, 15), which led to the suggestion that hypoxia might amplify the response of exercise training on insulin action and glucose homeostasis. Our data do not confirm these suggestions, either on the whole-body or on the skeletal muscle level. However, our study does not rule out the possibility that other hypoxia interventions like breathing hypoxic air for longer periods might have positive effects on insulin action.

Our study has strengths and limitations. A limitation might have been the short exposure time to hypoxia or not high enough exercise intensity during training sessions. Because we planned to develop a training program suitable for older sedentary people, neither longer duration of hypoxic stimulation nor higher exercise intensities appeared to be feasible and meaningful in this context. Our cohort was relatively healthy, leaving open the possibility that older people with more pronounced metabolic or cardiovascular disturbances could benefit from hypoxia training. Finally, the sample size in our trial might have been too small to detect important differences between study groups. Strengths of our study include the use of HEGCs in conjunction with muscle biopsies to determine whole-body insulin sensitivity, markers of the insulin signal cascade and oxidative metabolism, as well as studying a group of subjects for whom effects of the intervention might have been of great clinical importance.

We conclude that training under normobaric hypoxia did not further enhance insulin sensitivity compared with training under normoxia in older women and men. Although we cannot exclude subtle additional metabolic benefits of hypoxia given the potential limitations of the intervention, the additional infrastructure and costs required for training under hypoxia appears not reasonable in the studied population.

Acknowledgments

We thank Ina Groen and Momme Kück for expert technical assistance.

Financial Support: This work was supported by a grant from the German Research Foundation (reference no. HA 7037/2-1) to S.H.

Clinical Trial Information: ClinicalTrials.gov no. NCT02196623 (registered 22 July 2014).

Disclosure Summary: The authors have nothing to disclose.

Abbreviations:

Abbreviations:
 
• ANCOVA

  analysis of covariance

 
• GAPDH

  glyceraldehyde 3-phosphate dehydrogenase

 
• GLUT4

  glucose transporter 4

 
• HEGC

  hyperinsulinemic-euglycemic glucose clamp

 
• HIF-1α

  hypoxia inducible factor-1α

 
• HOMA-IR

  homeostasis model assessment insulin resistance index

 
• p-Akt

  phospho-Akt (Ser⁴⁷³) normalized to total Akt

 
• VO_2peak

  peak oxygen uptake

Author notes

K.C.-J., R.J.S., J.J. and S.H. contributed equally to this study.