Mild intermittent hypoxia exposure induces metabolic and molecular adaptations in men with obesity

Objective Recent studies suggest that hypoxia exposure may improve glucose homeostasis, but well-controlled human studies are lacking. We hypothesized that mild intermittent hypoxia (MIH) exposure decreases tissue oxygen partial pressure (pO2) and induces metabolic improvements in people who are overweight/obese. Methods In a randomized, controlled, single-blind crossover study, 12 men who were overweight/obese were exposed to MIH (15 % O2, 3 × 2 h/day) or normoxia (21 % O2) for 7 consecutive days. Adipose tissue (AT) and skeletal muscle (SM) pO2, fasting/postprandial substrate metabolism, tissue-specific insulin sensitivity, SM oxidative capacity, and AT and SM gene/protein expression were determined. Furthermore, primary human myotubes and adipocytes were exposed to oxygen levels mimicking the hypoxic and normoxic AT and SM microenvironments. Results MIH decreased systemic oxygen saturation (92.0 ± 0.5 % vs 97.1 ± 0.3, p < 0.001, respectively), AT pO2 (21.0 ± 2.3 vs 36.5 ± 1.5 mmHg, p < 0.001, respectively), and SM pO2 (9.5 ± 2.2 vs 15.4 ± 2.4 mmHg, p = 0.002, respectively) compared to normoxia. In addition, MIH increased glycolytic metabolism compared to normoxia, reflected by enhanced fasting and postprandial carbohydrate oxidation (pAUC = 0.002) and elevated plasma lactate concentrations (pAUC = 0.005). Mechanistically, hypoxia exposure increased insulin-independent glucose uptake compared to standard laboratory conditions (~50 %, p < 0.001) and physiological normoxia (~25 %, p = 0.019) through AMP-activated protein kinase in primary human myotubes but not in primary human adipocytes. MIH upregulated inflammatory/metabolic pathways and downregulated extracellular matrix-related pathways in AT but did not alter systemic inflammatory markers and SM oxidative capacity. MIH exposure did not induce significant alterations in AT (p = 0.120), hepatic (p = 0.132) and SM (p = 0.722) insulin sensitivity. Conclusions Our findings demonstrate for the first time that 7-day MIH reduces AT and SM pO2, evokes a shift toward glycolytic metabolism, and induces adaptations in AT and SM but does not induce alterations in tissue-specific insulin sensitivity in men who are overweight/obese. Future studies are needed to investigate further whether oxygen signaling is a promising target to mitigate metabolic complications in obesity. Clinical trial registration This study is registered at the Netherlands Trial Register (NL7120/NTR7325).


INTRODUCTION
Oxygen homeostasis, or the balance between oxygen supply and demand, is critical in cellular survival. Indeed, alterations in tissue oxygen partial pressure (pO 2 ) impact a variety of physiological responses. Compelling evidence points toward the important role of altered oxygen availability (i.e., 'hypoxia') in cardiometabolic perturbations [1,2].
Intriguingly, although metabolic adaptations to hypoxia are not fully understood, oxygen availability in crucial metabolic organs such as adipose tissue (AT) and skeletal muscle (SM) may play a significant role in the pathophysiology of obesity-related cardiometabolic complications [2]. We have recently demonstrated that AT pO 2 was increased in obese insulin resistant individuals compared to lean and obese insulin-sensitive individuals [3,4]. Moreover, AT pO 2 was positively associated with insulin resistance, independently of age, sex, and adiposity [3]. In line with these findings, diet-induced weight loss markedly decreased AT pO 2 in humans, accompanied by improved insulin sensitivity [5]. These findings suggest that targeting tissue oxygenation may lower the risk for cardiometabolic complications in people with obesity [2,6]. Conflicting findings on the effects of pO 2 on inflammation and insulin sensitivity have also been reported, which may be related to the severity, pattern, and duration of hypoxia exposure [2]. Indeed, acute exposure to severe hypoxia impaired insulin signaling in both murine and human adipocytes [7], whereas the prolonged, repeated exposure of differentiating adipocytes to transient hypoxia was able to reprogram these cells for enhanced insulin sensitivity [8]. Furthermore, hypoxia exposure increased glucose uptake in murine and human myotubes and human SM explants [9,10]. Interestingly, prolonged mild intermittent hypoxia (MIH) exposure, characterized by multiple hypoxic episodes per day, has been shown to improve glucose homeostasis in rodents [11e14]. Moreover, an uncontrolled trial in obese men indicated that mild hypoxia exposure (15 % O 2 ) improved whole-body insulin sensitivity [15]. However, placebo-controlled trials investigating the effects of prolonged exposure to MIH on tissue-specific insulin sensitivity and cardiometabolic risk factors in humans are lacking. In addition, a more detailed mechanistic understanding of the putative beneficial effects of MIH on glucose homeostasis is warranted. Therefore, we hypothesized that MIH improves insulin sensitivity and induces a shift toward increased glucose utilization in individuals that are overweight and obese. Here, we report the first randomized, controlled, single-blind crossover trial designed to investigate the effects of MIH exposure (FiO 2 15 % O 2 , 3 times 2 h/day vs. 21 % O 2 ) for 7 consecutive days on tissue-specific insulin sensitivity (primary study outcome), AT and SM pO 2, and fasting and postprandial energy/substrate metabolism in overweight/obese men. The reason we applied this MIH protocol was due to exposure to 15 % O 2 improving insulin sensitivity in a previous uncontrolled human study [15], and an intermittent pattern of mild hypoxia exposure (3e15 hypoxic episodes/ day) might elicit more beneficial effects on the cardiovascular risk profile [16] without inducing adverse effects. Furthermore, we determined the impact of MIH on the gene/protein expression in AT and SM and performed functional mechanistic experiments using human primary myotubes and adipocytes cultured under oxygen levels mimicking the AT and SM microenvironments in humans.

Human intervention study
2.1.1. Study design Twelve overweight/obese men (BMI !28 kg/m 2 , age 30e65 years) with HOMA IR !2.2 participated in the present study. Exclusion criteria were smoking, cardiovascular disease, type 2 diabetes mellitus, liver or kidney malfunction, use of medication known to affect body weight, glucose and/or lipid metabolism, and marked alcohol consumption (>14 alcoholic units/wk). Furthermore, subjects had to be weight stable (weight change <3.0 kg) for at least three months prior to the start of the study. Study participants were exposed to normobaric mild intermittent hypoxia (FiO 2 15 %; equivalent to~2800 m above sea level) and normobaric normoxia (FiO 2 21 %) for 7 consecutive days (3 cycles of 2 h exposure/day, with 1 h of normoxia exposure between hypoxic cycles) in a randomized fashion and separated by a 3-to 6-week wash-out period (Supplementary Figure 1). A random allocation sequence was performed by an independent researcher using a computer-generated randomization plan (block size n ¼ 4). Hypoxia exposure was performed in an in-house manufactured airtight clinical room with the capability to accurately adjust oxygen availability at the Metabolic Research Unit Maastricht (Maastricht University, The Netherlands). The oxygen level was set and maintained at 15.0 AE 0.2 % for the hypoxia exposure regimen. Participants were blinded for the exposure regimen (hypoxia or normoxia). Individuals were asked to refrain from drinking alcohol and perform no exercise 48 h before the start of and during exposure regimens. Furthermore, participants performed a standardized light-stepping activity (3 Â 5 min, p. day, 15 steps/min) during days 1e7. Measurements performed on days 6 (AT and SM pO 2 ), day 7 (high-fat mixed-meal [HFMM] test), and day 8 (two-step hyperinsulinemiceuglycemic clamp), as explained below, were performed after an overnight fast of at least 10 h. Participants were kept under energybalanced conditions. The diet was adjusted individually to match the energy requirements and maintain an energy balance throughout the study. Based on the estimated daily energy expenditure (basal metabolic rate [BMR; Ventilated Hood, Omnical, Maastricht University] multiplied by activity score of 1.55), subjects received a standardized diet consisting of 50 % carbohydrate, 35 % fat, and 15 % protein to maintain a stable body weight throughout the study. An assessment of hunger and satiety was performed on days 1, 3, and 7 by visual analog scale (VAS) questionnaires. Adverse events of MIH were monitored by the Lake Louise questionnaire for Acute Mountain Sickness (AMS). The study was approved by the Medical-Ethical Committee of Maastricht University and performed according to the Declaration of Helsinki. All subjects gave their written informed consent before participation in the study. This study is registered at the Netherlands Trial Register (NL7120/NTR7325).

Anthropometric measurements
Body weight was measured to the nearest 0.1 kg (model 220, Seca). Height was measured using a wall-mounted stadiometer (Seca). Blood pressure and heart rate were assessed (UA-789XL digital blood pressure monitor, A&D medical) on multiple occasions during the exposure protocol (days 1e5).
2.1.3. Systemic oxygen saturation, adipose tissue, and skeletal muscle partial oxygen pressure Systemic oxygen saturation levels were continuously monitored throughout the exposure regimens by pulse oximetry (Nellcore N-595 Pulse oximeter, Nellcor). On day 6 of each exposure regimen, AT and SM pO 2 were measured by a highly accurate optochemical measurement system to continuously monitor tissue pO 2 in vivo in humans (Figure 2A), as described previously [4]. Briefly, microdialysis catheters (CMA60, CMA microdialysis AB) were inserted into abdominal subcutaneous AT 6e8 cm lateral from the umbilicus (skin anesthetized using EMLA cream) and SM (m. gastrocnemius; local anesthesia using 2 % lidocaine after perfusion [Ringer solution]) After the insertion, both microdialysis catheters were perfused with Ringer solution (Baxter BV) at a flow rate of 2 ml/min (CMA400 microinfusion pump, CMA Microdialysis AB). The interstitial fluid was then directed toward a flow-through cell containing an optochemical O 2 -sensor [17]. Within 2e3 h after the insertion of the microdialysis probes in AT and SM, pO 2 had reached stable values (change in pO 2 < 2.0 mmHg over a 20-min period). Fasting AT and SM pO 2 values were calculated by averaging the 20-min periods with stable pO 2 readings.
Original Article 2.1.4. High-fat mixed-meal test On day 7 of each exposure regimen, a high-fat mixed-meal challenge test was performed (Figure 2A), fasting and postprandial (0e240 min) blood samples were collected, and substrate oxidation was determined using indirect calorimetry. Following a 30-min baseline period under fasting conditions, individuals were asked to ingest a liquid test meal, providing 2.6 MJ (consisting of 61 E% fat [35.5 E% saturated fat, 18.8 E% monounsaturated fat, and 1.7 E% polyunsaturated fat], 33 E% carbohydrates, and 6.3 E% protein), at t ¼ 0 within 5 min. Blood samples were collected from a superficial dorsal hand vein, which was arterialized by placing the hand into a hotbox (~55 C). Blood samples were taken under fasting (t ¼ 0 min) and postprandial conditions (t ¼ 30, 60, 90, 120, 180, and 240 min). Energy expenditure and substrate oxidation were assessed using indirect calorimetry (opencircuit ventilated hood system, Omnical, Maastricht University) under fasting conditions (t ¼ À30 -0 min) and for 4 h after the ingestion of the high-fat mixed-meal [18]. Calculations of energy expenditure and substrate oxidation were performed according to the formulas of Weir [19] and Frayn [20]. Nitrogen excretion was based on the assumption that protein oxidation represents~15 % of the total energy expenditure [21].
2.1.5. Two-step hyperinsulinemic-euglycemic clamp A two-step hyperinsulinemiceeuglycemic clamp combined with a D-[6,6-2 H 2 ]-glucose tracer infusion (Cambridge Isotope Laboratories, no. DLM-349) was performed on day 8, under normoxic conditions ( Figure 5A) to determine adipose tissue, hepatic, and peripheral insulin sensitivity [22]. First, a primed (bolus-injection of 2.4 mg•kg À1 ) continuous infusion of D-[6,6-2 H 2 ]-glucose was started and continued throughout the measurement at 0.04 mg•kg À1 •min À1 to determine the baseline endogenous glucose production (EGP), glucose rate of appearance (Ra), and glucose disposal (Rd). After 2 h, insulin was infused at a primed continuous low rate of 10 mU•m À2 •min À1 for 3 h to assess hepatic insulin sensitivity (% suppression of EGP) and adipose tissue insulin sensitivity (% suppression of free fatty acids [FFA]), followed by insulin infusion at a high rate of 40 mU•m À2 •min À1 for 2.5 h to determine peripheral insulin sensitivity (Rd). Arterialized blood samples (hotbox at~55 C) were frequently taken from a superficial dorsal hand vein, and blood glucose concentrations were directly determined. Euglycemia was maintained by a variable co-infusion of a 17.5 % glucose solution enriched by 1.1 % tracer (5.0 mmol•l À1 ). During the last 30-min of each step (0, 10, and 40 mU•m À2 •min À1 insulin), substrate oxidation was measured using indirect calorimetry, and blood was sampled every 15 min to determine steady-state kinetics.
2.1.6. Skeletal muscle and adipose tissue biopsies Before the start of the hyperinsulinemic-euglycemic clamp, an SM (80e100 mg; m. vastus lateralis) and abdominal subcutaneous AT biopsy (~1 g; 6e8 cm lateral from the umbilicus) were collected under local anesthesia (2 % lidocaine without epinephrine). Part of the fasting muscle biopsy was immediately placed in an ice-cold preservation medium (BIOPS, OROBOROS Instruments) to prepare intact permeabilized muscle fibers. Permeabilized muscle fibers were used to determine mitochondrial oxidative capacity using an oxygraph (ORO-BOROS Instruments), as described previously [23]. The remaining part of the SM biopsy was snap-frozen in liquid nitrogen and stored at À80 C until the analysis. The AT specimen was washed using sterile saline, visible blood vessels were removed using sterile tweezers, and the AT biopsy was snap-frozen in liquid nitrogen and stored at À80 C. Another muscle biopsy was collected at the end of the steady state of the high-insulin step of the clamp.

Plasma biochemistry
Blood was collected into pre-chilled tubes, centrifuged at 1,000 g, and plasma was snap-frozen and stored at À80 C until the analysis. Plasma glucose (ABX Pentra Glucose HK CP, Horiba ABX Diagnostics), FFA (WAKO NEFA-HR [2] ACS-ACOD method, WAKO Chemicals GmbH), TAG (ABX Pentra Triglycerides CP, Horiba ABX Diagnostics), and glycerol (Glycerol-kit UV-test, r-Biopharm) were determined. Plasma insulin was measured using a double-antibody radioimmunoassay (Millipore). Lactate was measured in plasma using standard enzymatic techniques automated on a Cobas Fara centrifugal spectrophotometer (Roche Diagnostics). Plasma TNFa, IFNg, IL-6, and IL-8 concentrations were determined in fasting plasma samples collected at day 8, subsequent to 7-d MIH exposure, using the V-PLEX Pro-inflammatory Panel I Human Kit (Mesoscale, 4-plex, no. K15052D), according to manufacturer's guidelines.
2.1.8. Adipose tissue and skeletal muscle gene expression Microarray analysis was performed on SM and AT samples, as described previously [24]. Total RNA was extracted from adipose tissue and skeletal muscle biopsies using the Trizol method (Qiagen, Venlo, Netherlands). Subsequently, 100 ng of intact total RNA was processed applying the GeneChipÒ WT PLUS Reagent Kit (Affymetrix) and Human Transcriptome Array (HTA) 2.0 GeneChip (Affymetrix) according to the manufacturer's manual. Gene set enrichment analysis was based upon FDR q-value <0.20 on the filtered data set (IQR > 0.2 [log2], intensity >20, >5 arrays, >5 probes per gene), using the databases Kyoto Encyclopedia of Genes and Genomes, Reactome, Wikipathways, and Biocarta. The present data have been deposited in NCBI's Gene Expression Omnibus and are accessible through the GEO Series accession number (accession number: to be added after manuscript acceptance) 2.1.9. Skeletal muscle protein expression To evaluate the effect of MIH on (p)-Akt, (p)-AMPK, and OXPHOS complex protein expression in skeletal muscle, we performed a Western Blot analysis on protein lysates derived from the biopsies. Proteins were lysed, as described previously [25], and were subsequently homogenized. Next, a BCA assay was performed to determine protein lysate concentrations. Equal amounts of protein were loaded ([p]AMPK and [p]Akt 25 mg, OXPHOS 7.5 mg) and separated on the different gels. Primary antibodies included mouse monoclonal antibodies cocktail directed against human OXPHOS (dilution: 1:10.000; no. ab110411/no. MS601, Abcam/MitoSciences), rabbit anti-human AMPKa (1:1000, no. 2603, Cell Signaling), rabbit anti-human p-AMPKa Thr172 (1:1000, no. 2535, Cell Signaling), rabbit anti-mouse Akt (1:1000, no. 9272, Cell signaling), and rabbit-anti mouse p-Akt Ser473 (1:1000, no. 9271, Cell Signaling). Secondary antibodies were donkey anti-mouse conjugated with IRDYe800 (OXPHOS), swine anti-rabbit HRP (AMPKa, p-Akt, and Akt), and goat anti-rabbit HRP (p-AMP-Ka Thr172 ). Visualization and analysis were performed using a CLx Odyssey Near Infrared Imager (OXPHOS) or ChemiDoc XRS system (AMPK, Akt) and Quantity One software. Individual intensities were quantified and are expressed in arbitrary units or derived ratios. PHG0311, Gibco). At 70e80 % confluence, a differentiation medium was added containing MEM-a-Glutamax (no. 32561-029, Gibco) supplemented with 2 % FBS (Bodinco BV), 0.5 mg/ml bovine fetuin (Invitrogen, Life Technologies), and 1x Antibiotic-Antimycotic (100 units•ml Àl penicillin, 100 mg•ml Àl of streptomycin, and 0.25 mg•ml Àl Amphotericin B [no. 15240-62, Gibco]) until the myotube formation was completed. After the differentiation, myotubes were exposed to 1, 3, and 21 % O 2 for 24 h, as described earlier [26]. Notably, more prolonged exposure to MIH was not feasible due to the relatively short lifespan of differentiated primary human myotubes. Gas mixtures were refreshed every 8 h. Subsequently, functional experiments were performed, and the cell culture medium, protein, and RNA were collected and stored at À80 C until the analysis.

Glucose uptake in primary human myotubes and adipocytes
Basal and insulin-stimulated glucose uptake were determined in both primary human myotubes and differentiated hMADS following serumstarvation for 2 h, as described previously [28]. Briefly, hMADS and myotubes were serum-starved 2 h prior to the glucose uptake experiment. After two washes in modified Krebs Ringer buffer (1. To investigate the involvement of adenosine monophosphate-activated kinase (AMPK) in the effects of hypoxia on glucose uptake in human primary myotubes, we co-incubated cells with and without 10 mM Compound C (no. P5499, Sigma) during hypoxia exposure and 1 mM AICAR (no. A9978, Sigma) as the positive control. Next, basal glucose uptake was measured.

Myotube protein expression
After exposure to 1, 3, and 21 % O 2 for 24 h, protein lysates were collected. Subsequently, protein concentrations were identified, and p-AMPKaThr 172 , AMPKa, and OXPHOS complex protein expression were determined by Western Blotting. Briefly, 1x RIPA buffer supplemented with a protease/phosphatase inhibitor cocktail (Cell Signaling Technology Europe, Leiden, The Netherlands) was added to lyse the myotubes. After scraping the cells, the protein lysate was collected, and the protein concentration was determined (BCA Protein Assay kit; Santa-Cruz Biotechnology Inc.). 15 mg protein was separated on BioRad Criterion TGX gels (Bio-Rad Laboratories; 12 % for OXPHOS, Any-kD for other targets) and subsequently blotted using Trans-Blot Turbo Transfer System (Bio-Rad; for AMPKa, OXPHOS, and GAPDH) and Bio-Rad Criterion Blotter (Bio-Rad; for p-AMPKaThr 172 , GAPDH), respectively. Membranes were incubated overnight at 4 C with primary antibodies in corresponding blocking buffers according to the manufacturer's protocols. Subsequently, blots were incubated with secondary antibodies for 1 h at room temperature. The primary antibodies used were AMPKa (no. 2603, Cell Signaling), p-AMPKaThr 172 (no. 2535, Cell Signaling), total OXPHOS antibody cocktail (no. MS601, Mitosciences), and GAPDH (no. 2118, Cell Signaling). Secondary antibodies used were swine-antirabbit HRP (no. P0339, DAKO), rabbit-anti-mouse HRP (no. P0161, DAKO), and goat-anti-rabbit HRP (no. PI-1000, Vector Laboratories). Antigen-antibody complexes were visualized by chemiluminescence using SuperSignalÔ West Femto and Dura extended Duration Substrates (Life Technologies). Visualization and analysis were performed using a ChemiDoc XRS system (Bio-Rad) and Quantity One software.

Adipokine secretion
At the end of each exposure regimen, the differentiated adipocytes and myotubes medium was collected to determine adipokine medium concentrations of IL-6, MCP-1, leptin, adiponectin, and VEGF using ELISA, as described previously [27].

Adipocyte and myotube gene expression
To obtain RNA for gene expression analysis, we added TRIzol Reagent (Invitrogen) to the cells at the end of exposure, and RT-PCR was performed, as described previously [27]. Firstly, the extracted RNA was precipitated and purified according to the manufacturer's protocol. Subsequently, cDNA was synthesized out of purified RNA using RT- Original Article

Subject characteristics
Twelve overweight and obese men with homeostasis model assessment for insulin resistance (HOMA IR ) !2.2 and without any chronic disease or endocrine disorder participated in the present randomized, single-blind crossover study. Subjects were recruited between April 2016 and December 2019. One participant did not undergo the twostep hyperinsulinemic-euglycemic clamp procedure. Participants' characteristics are shown in Table 1.   Figure 1). These findings provide the first proof-of-concept in humans that MIH exposure not only reduces systemic oxygen saturation but also consistently decreases AT and SM pO 2 by~40 %. In addition, SM pO 2 was significantly lower than AT pO 2 (p < 0.001), likely reflecting the higher metabolic rate in SM. No adverse events were reported.

Mild intermittent hypoxia exposure induces a shift toward glycolytic metabolism
Fasting and postprandial energy expenditure ( Figure 2B) were not significantly altered by MIH, though fasting energy expenditure tended to be reduced (p ¼ 0.053). MIH induced a pronounced increase in the fasting respiratory exchange ratio (RER, p ¼ 0.001) and carbohydrate oxidation (CHO, p ¼ 0.002), whereas fat oxidation was reduced (FAO, p ¼ 0.013) compared to normoxia exposure. Moreover, following meal intake, RER remained elevated (p AUC ¼ 0.003, Figure 2C) during MIH compared to normoxia exposure due to a marked shift toward increased postprandial CHO (p AUC ¼ 0.003, Figure 2D) and suppression of FAO (p AUC ¼ 0.018, Figure 2E). In addition, MIH exposure did not alter fasting and postprandial plasma glucose, insulin, free fatty acid, triacylglycerol, or glycerol concentrations compared to normoxia exposure (Figure 3BeF). Nevertheless, plasma lactate levels were increased by MIH compared to normoxia (p AUC ¼ 0.005, Figure 3G). Notably, the MIH-induced effects on fasting substrate utilization and plasma lactate concentrations were maintained but not more pronounced during postprandial conditions, as MIH did not alter the incremental areas under the curves (iAUC•min À1 ) for these parameters (Figure 3).

Mild intermittent hypoxia exposure does not affect tissuespecific insulin sensitivity but induces a metabolic shift toward increased carbohydrate oxidation
The two-step hyperinsulinemic-euglycemic clamp with a D-[6,6-2 H 2 ]glucose tracer infusion ( Figure 5A) revealed that MIH did not significantly alter basal endogenous glucose production ( Figure 5B), hepatic insulin sensitivity (suppression of endogenous glucose production [EGP] during low-insulin infusion, 10 mU•m À2 •min À1 ; Figure 5C), or AT insulin sensitivity (insulin-mediated suppression of free fatty acids during low-insulin infusion, 10 mU•m À2 •min À1 ; Figure 5D) compared to normoxia. Moreover, peripheral insulin sensitivity (insulin-stimulated glucose rate of disappearance, Rd) was not significantly affected by MIH ( Figure 5E). In addition, non-oxidative glucose disposal (NOGD) remained unaltered following MIH compared to normoxia exposure.
In agreement with these findings, we found that MIH did not induce changes in SM insulin signaling, determined using fasting (baseline) and insulin-stimulated (40 mU•m À2 •min À1 ) SM biopsies collected before and during the hyperinsulinemic-euglycemic clamp, respectively. More specifically, the insulin-induced increase in the p-Akt Ser473 /total Akt protein expression ratio in skeletal muscle was comparable following MIH and normoxia exposure ( Figure 5F,G). In addition, MIH did not change the muscle p-AMPK Thr172 /AMPK ratio compared to normoxia (p ¼ 0.346; Figure 5H,I). Intriguingly, the fasting RER (p ¼ 0.050) and CHO (p ¼ 0.050) remained increased at day 8, whereas FAO remained reduced (p ¼ 0.049), indicating that MIH impacts substrate utilization for at least 16 h after the cessation of the last bout of hypoxia (Supplementary Figure 3). In line with these findings, at day 8 (normoxia), the RER (p ¼ 0.070) and CHO (p ¼ 0.095) continued to be increased, whereas FAO tended to be reduced (p ¼ 0.091) during high-insulin infusion. However, insulin-induced changes in substrate oxidation rates were not altered, suggesting that metabolic flexibility was not affected by MIH. 3.6. Mild intermittent hypoxia exposure does not alter skeletal muscle oxidative capacity MIH did not induce alterations in state 3 and maximally uncoupled mitochondrial respiration using carbohydrate-and lipid-leads (Figure 6AeC). In agreement with these findings, the protein expression of OXPHOS complex components was unchanged following MIH ( Figure 6D,E). In line with these findings, western blot analysis revealed that mild hypoxia exposure did not affect OXPHOS complex protein expression in human primary myotubes ( Figure 6F,G).
3.7. Mild intermittent hypoxia exposure impacts inflammatory, metabolic, and ECM-related pathways in adipose tissue but not skeletal muscle In total, 171 gene sets were positively (and 285 negatively) enriched in AT following 7-day MIH compared to normoxia. In contrast, no significantly differentially expressed pathways could be identified in SM after gene set enrichment analysis (GSEA) at FDR q-value <0.2, p < 0.05 (all affected pathways in AT and individual genes in SM are in Supplementary Tables 1 and 2, respectively). In AT, pathways related to inflammation and carbohydrate and lipid metabolism were significantly upregulated, reflected by positive normalized enrichment scores. The expression of gene sets related to the cell cycle, mitochondrial translation, and extracellular matrix components was downregulated, reflected by negative normalized enrichment scores ( Figure 7). Interestingly, we found that MIH downregulated the collagen assembly pathway, which may reflect reduced ECM stress, possibly reflective of the initiation of AT remodeling. Furthermore, we found that MIH upregulated pathways related to inflammation transcription factor nuclear factor-kB (NF-kB) and various downstream cytokine signaling pathways involving cytokines, such as interleukin-1 (IL-1) and IL-2, and complement factors. However, systemic levels of inflammatory cytokines tumor necrosis factor a (TNFa), interferon-g (IFN-g), IL-6, and IL-8 were not affected by MIH exposure (Supplementary Figure 4).
3.8. Mild hypoxia exposure affects adipokine expression and secretion but not glucose uptake in primary human adipocytes MIH exposure reduced VEGF (p ¼ 0.005) and leptin (p ¼ 0.034) secretion by primary human adipocytes compared to 10 % and 21 % O 2 , respectively, but did not significantly affect IL-6, MCP-1, or adiponectin secretion (Supplementary Figure 5AeE). In line with reduced leptin secretion, leptin mRNA expression was significantly reduced by MIH exposure compared to 21 % O 2 (p ¼ 0.005; Supplementary Figure 5F). In contrast, MIH increased VEGFA mRNA expression compared to 10 % O 2 (p ¼ 0.011). Moreover, we determined the impact of MIH exposure on glucose uptake in primary human adipocytes. MIH did not significantly alter insulin-(in)dependent glucose uptake compared to 10 % O 2 (Supplementary Figure 5G).
3.9. Mild intermittent hypoxia induces slight hemodynamic adaptations but does not alter feelings of hunger and satiety MIH exposure did not affect SBP and DBP, despite a slight but significant increase in mean heart rate (p ¼ 0.017, Supplementary  Figure 6). Finally, we found that feelings of hunger, satiety, and thirst were not significantly affected by MIH exposure, determined by Visual Analogue Scales (VAS) completed by the participants before and after breakfast, lunch, and dinner (Supplementary Figure 7).

DISCUSSION
In the present randomized, controlled, single-blind crossover study, we demonstrated for the first time that MIH exposure for 7 consecutive days reduces systemic oxygen saturation, decreases AT and SM pO 2 , evokes a shift toward glycolytic metabolism, and induces adaptations in AT and SM but does not impact AT, hepatic, or SM insulin sensitivity in men with obesity. We found that MIH exposure markedly increased whole-body carbohydrate oxidation, an effect that persisted after the cessation of the exposure regimen until the following day at the least. This shift in substrate oxidation was also reflected by increased plasma lactate concentrations during the high-fat mixed-meal test under MIH, indicating an increased glycolytic rate. Importantly, MIH exposure did not induce hyperventilation, suggesting that the increased RER found reflects increased carbohydrate oxidation and does not seem to be explained by respiratory alkalosis. In accordance with our findings, it has been previously demonstrated that short-term mild (hypobaric) hypoxia exposure increased reliance on carbohydrate oxidation [29e 31] but did not affect postprandial glucose and insulin levels [32]. Mechanistically, the present data demonstrate that hypoxia exposure increased glucose uptake via an AMPK-dependent mechanism in primary human myotubes but not adipocytes, in line with previous in vitro studies [15,33e35]. In contrast to our in vitro findings in primary human myotubes and previous observations in rodent skeletal muscle [36,37], we did not find altered p-AMPK Thr172 /AMPK protein expression in human SM following MIH exposure. As the muscle biopsies were collected~16 h after the last hypoxic stimulus, we cannot exclude that AMPK was already dephosphorylated at the time of sample collection. Interestingly, the present randomized, controlled, single-blind crossover study demonstrates that for 7 days, MIH did not alter AT, hepatic, or SM insulin sensitivity. In contrast, findings from previous studies in rodents [11e14] and humans [15] have shown that prolonged exposure to mild hypoxia improves insulin sensitivity and/or glucose homeostasis. However, it is important to emphasize that the latter finding was based on a within-group comparison, as a control group was not included in the study design [15]. Moreover, our data show that MIH increased the AT gene expression of inflammatory/metabolic pathways and decreased extracellular matrixrelated pathways. However, MIH effects on adipokine expression/ secretion did not translate into changes in low-grade systemic inflammation. Conflicting findings have been reported on the effects of (severe) hypoxia on inflammatory processes, which seems to be related to the severity, pattern, and duration of hypoxia exposure [2].
We have recently shown that prolonged exposure to mild hypoxia reduces the expression of pro-inflammatory genes in cultured human adipocytes [27]. An alternative explanation for the upregulation of inflammation-related pathways in AT in the present study may include hypoxia-induced lactate production, contributing to increased NF-kBrelated gene expression, as observed in L6 myocytes and macrophages [38,39]. Conversely, the hypoxiaelactate axis might mitigate inflammation by suppressing macrophage activation and polarization toward anti-inflammatory M2-macrophages [40,41].
Furthermore, ex vivo SM mitochondrial respiratory capacity and the OXPHOS protein expression in SM were unaltered following MIH exposure. The hypoxia-induced activation of hypoxia-inducible factor (HIF)-1a activation might inhibit decarboxylation of pyruvate into acetyl-CoA by pyruvate dehydrogenase [42]. This effect may then shunt pyruvate away from the mitochondria, resulting in reduced mitochondrial respiration and enhancing glycolysis to ensure sufficient ATP production [42]. Notably, in the present study, the SM biopsies used to isolate the SM fibers used to determine oxygen consumption rates were collected~16 h after the last hypoxic stimulus of the exposure regimen and, therefore, reflect the chronic rather than the acute response of MIH on the oxidative machinery in human SM.
The strengths of the present study are that this study is the first randomized crossover trial investigating the effects of MIH exposure on many cardiometabolic risk factors in humans to our knowledge. Importantly, measurements were performed under well-controlled conditions, and participants were kept in energy balance throughout the study. Furthermore, we studied the isolated effects of MIH under normobaric conditions, thereby excluding the possible effects of a different ambient pressure present at high altitudes (i.e., hypobaric hypoxia). Finally, we performed extensive (functional) measurements in tissue biopsies and human primary myotubes and adipocytes, cultured under oxygen levels that mimic the adipose tissue and skeletal muscle microenvironments in humans to explore underlying mechanisms. Noteworthy, several nuances have to be made regarding the conclusions of the present study. First, as we studied men with obesity characterized by modest impairments in glucose homeostasis, we cannot exclude that MIH exposure may have other effects in individuals with a different metabolic phenotype (i.e., normal-weight individuals, patients with type 2 diabetes) or women. Secondly, the duration of the exposure to MIH was relatively short (42 h in total) due to the invasive nature of the study and the burden on the participants. Therefore, we cannot exclude that more prolonged exposure to mild (intermittent) hypoxia or exposure to more severe hypoxia is required to induce beneficial effects on glucose homeostasis, and this possibility warrants further investigation. Thirdly, as the primary human myotubes were continuously exposed to different oxygen levels for 24 h, we cannot exclude that exposure to an MIH regimen mimicking the human in vivo study (i.e., 3 Â 2 h per day) would have yielded less pronounced increases in AMPK activation and glucose uptake.

CONCLUSIONS
In summary, the findings of the present randomized, single-blind crossover trial demonstrate that 7 consecutive days of MIH exposure decreases AT and SM pO 2 , induces a metabolic switch toward increased reliance on glycolysis to ensure ATP production and evokes adaptations in AT and SM in humans with obesity. However, these findings do not translate into alterations in AT, hepatic, or SM insulin sensitivity. Further studies are warranted to investigate whether other hypoxia exposure regimens (longer intervention period and/or more severe hypoxic episodes) and/or pharmacological strategies targeting oxygen signaling may mitigate obesity-related metabolic perturbations. For example, the stabilization of HIF by prolyl-hydroxylase (PHD) inhibitors improved glucose tolerance, serum cholesterol, and inflammation in rodents [43,44]. Noteworthy, PHD inhibitors have recently been used in patients with chronic kidney disease [45], but the putative effects of these agents on metabolic homeostasis have not been examined yet to our knowledge.

DATA AND MATERIALS AVAILABILITY
The datasets generated by microarray analysis are deposited at the Gene Expression Omnibus (GEO) database (accession number < to be added upon acceptance>). Furthermore, primer sequences used in this study are included in Supplementary Table 3.