Ozone Exposure Triggers Insulin Resistance Through Muscle c-Jun N-Terminal Kinase Activation

Type 2 diabetes (T2D) is one of the most prevalent metabolic diseases worldwide and is projected to increase dramatically over the next decades because of aging populations, chronic overnutrition, and sedentary lifestyle. Environmental and lifestyle-related factors can account for roughly 90% of adult-onset diabetes (1), and it has been suggested that T2D can be prevented by lifestyle and diet modifications (2). There is also growing evidence for an association between traffic-related air pollution and the incidence of T2D. For instance, the hazards for diabetes were increased with particulate matter (PM) exposure (3), whereas a statistically significant association of nitrogen dioxide (NO₂) was detected with confirmed cases of diabetes or with mortality from diabetes (4,5). In advanced polluted air environments, the action of ultraviolet (UV) rays on volatile organic compounds or NO₂ is responsible for the formation of ozone (O₃), another major pollutant in urban areas (6). O₃ pollution has become a major environmental challenge because of large exposed human populations both in the Western world and in developing countries (7). Children and the elderly are particularly sensitive to the pulmonary health effects of O₃, inducing or worsening asthma, chronic obstructive pulmonary diseases, and lung inflammation (8,9). In addition, many extrapulmonary effects of O₃ have been described: activation of stress-responsive regions, catecholamine biosynthesis, cell degeneration, neurochemical alterations in the central nervous system (10,11), increased production of nitric oxide, enhanced protein synthesis and inflammation in the liver (12), compensatory changes, edema, and increased oxidative stress in heart tissue (13,14). Taken together, these studies demonstrate that O₃ can exert many deleterious effects in distant tissues.

The role of oxidative stress in the etiology and pathogenesis of human diseases has been widely described in the past 2 decades (15). Many studies pinpoint oxidative stress as a potential cause to both the onset and progression of T2D (16–18) and its associated complications, such as endothelial dysfunction and cardiovascular diseases (19,20). O₃ is a strong oxidant, and, although not a radical species, most of its toxic effects are mediated through free radical reactions (21,22). Oxidative stress is indeed implicated in many deleterious effects of O₃, such as pulmonary inflammation and dysfunction (23), Alzheimer's disease (24), and cardiovascular diseases (25). In a recent study, Bass et al. (26) demonstrated that O₃ impaired glucose homeostasis in rats; however, very little is known about the effects of O₃ on metabolism and metabolic diseases.

Despite being a key issue for public health, the putative role of O₃ exposure in the onset and progression of diabetes remains poorly defined. Our study demonstrates for the first time that exposure to O₃, at a concentration realistically mimicking human exposure, induces insulin resistance (IR) in rodents, therefore suggesting a significant contribution of O₃ in the etiology of T2D.

All chemicals and culture media were purchased from Sigma-Aldrich (Saint Quentin Fallavier, France) when no other origin is specified. Recombinant human insulin (100 IU/mL; Actrapid) was from Novo Nordisk (La Défense, France). Anti-phospho-Akt 1/2/3 rabbit IgG (catalog #7985R) and anti-Akt 1/2/3 rabbit IgG (catalog #8312) antibodies were purchased from Santa Cruz Biotechnology (Heidelberg, Germany), anti-tubulin mouse IgG antibody was from Sigma-Aldrich, and anti-mouse IgG and anti-rabbit IgG antibodies were from Bio-Rad (Marnes-la-Coquette, France). Endoplasmic reticulum (ER) stress antibodies were all included in the ER Stress Antibody Sampler Kit from Cell Signaling Technology (Saint-Aubin, France). Super Signal West Pico Chemiluminescent Substrate and Restore Western Blot Stripping Buffer were obtained from Thermo Scientific (distributed by Perbio Sciences France, Brebières, France).

Animal experiments were performed under authorization 69–266–0501 (INSA-Lyon, Direction Départementale de la Protection des Populations-Services Vétérinaires du Rhône), according to the ethical guidelines laid down by the French Ministère de l’Agriculture (87–848) and the European Union Council Directive for the Care and Use of Laboratory Animals of November 24th, 1986 (86/609/EEC). Authors A.G. (license 69266332) and C.O.S. (license 69266257) hold special licenses to experiment on living vertebrates that were issued by the French Ministry of Agriculture and Veterinary Service Department. Wistar rats weighing 400–450 g (Janvier SA, Le Genest-Saint-Isle, France) were housed in an air-conditioned room at 24 ± 1°C with a 12-h light/dark cycle (light on at 6:30 a.m.) with free access to food (13.4 kJ/g, 65% carbohydrates, 11% fat, 24% proteins, weight for weight; diet A03; SAFE, Augy, France) and water. Great care was taken to minimize animal discomfort and suffering during the experimental protocol.

Rats were kept within a Plexiglas hermetic environmental chamber (width 0.35 m, length 0.70 m, height 0.40 m, surface 0.25 m², volume 0.1 m³) supplied with a constant airflow (6 m³/h) and subjected to 1,570 µg/m³ O₃, corresponding to 0.8 parts per million (ppm) for 16 h, as previously described (10). O₃ was generated by passing filtered air across a UV light, and concentration inside the cage was controlled by adjusting the inlet flow of air. The O₃ concentration was continuously measured using a calibrated UV photometric O₃ monitor (range ± 0.001 ppm; catalog #41M; Environement SA, Poissy, France) connected to the outlet line of the chamber. Control exposure was performed in a similar chamber provided with filtered room air at the same flow rate. All the exposures were performed during the dark phase (i.e., the active phase for the rats). Rats were fasted during the exposure but had free access to water. The temperature and relative humidity measured in the chambers during the exposure were 25.3 ± 0.4°C and 75%, respectively. Ammonia and CO₂ concentrations remained undetectable. The mean O₃ concentration inside the exposure chamber was 0.800 ± 0.05 ppm. Of note, the O₃ concentration in the control chamber remained undetectable (<0.001 ppm). Rats were exposed during nighttime, namely, during their active phase, because O₃ pollution peaks generally take place during the day (i.e., during human active phase).

Some rats were pretreated for 10 days with N-acetylcysteine (NAC) prior to O₃ exposure. Rats were given NAC orally in drinking water (10 mmol/L). The water intake was monitored daily to calculate the daily NAC intake. The mean daily intake of NAC was 225 ± 10 mg/kg/day. Another independent set of rats was gavaged daily with 4-phenylbutyric acid (PBA; 150 mg/kg) for 4 days prior to O₃ exposure.

Euglycemic-hyperinsulinemic (EH) clamps were performed essentially as previously described (27). In brief, Wistar rats were implanted with indwelling catheters in the left carotid artery and right jugular vein. The rats were allowed to recover from the surgery for 5 days. Prior to the EH clamp, the rats were fasted overnight and a standard 2-h EH clamp was conducted using a primed and continuous infusion of human insulin at a rate of 6 mU/kg/min coupled with a variable infusion of 25% (weight for volume) glucose to maintain blood glucose concentrations at ∼6 mmol/L. The glucose infusion rate (GIR) was calculated as the amount of glucose perfused to maintain euglycemia during the second hour of the clamp and was expressed as milligrams of glucose per kilogram per minute.

After an overnight fast, Wistar rats with indwelling catheters were injected intravenously with 1 g/kg l-arginine in saline solution. Blood was withdrawn from an arterial catheter and centrifuged (1 min, 3,500g), and plasma samples were snap frozen in liquid nitrogen and stored at −20°C until insulin assay.

After overnight fast, animals were injected intraperitoneally with 0.50 IU/kg body wt recombinant human insulin. Blood glucose was measured before, and 15, 30, 60, 90, and 120 min after insulin injection. Blood glucose values were determined from a drop of blood sampled from the terminal portion of the tail using a glucometer. As previously described (28), the glucose disappearance rate for an insulin tolerance test (ITT) (K_ITT; reported as the percentage per minute) was calculated as follows: K_ITT = 0.693 × 100/t_1/2, where t_1/2 was the half-life calculated from the slope of the plasma glucose concentration, considering an exponential decrement of glucose concentration during the 30 min after insulin administration.

Rats were anesthetized with sodium pentobarbital (35 mg/kg), and a PE-240 tracheal catheter was inserted into the trachea. Bronchoalveolar lavages were performed twice with 1 mL of sterilized Dulbecco’s PBS. An aliquot fraction of bronchoalveolar lavage fluid (BALF) was removed for cellular numeration and leukocyte count while the BALF was centrifuged (8,000g, 2 min, 4°C) to pellet cells. Supernatants were removed and kept at −80°C. Total proteins in the BALF were determined by the Bradford protein assay. Lipid peroxidation was measured spectrophotometrically by the thiobarbituric acid method (29) using 1.1.3.3-tetramethoxipropane as the standard. Cells were stained by the May-Grünwald Giemsa method for leukocyte counts, and 500 leukocytes were randomly counted in a Malassez cell under an optical microscope (original magnification ×100).

To study insulin signaling in skeletal muscle, liver, or adipose tissue, rats were injected with insulin (Actrapid 0.75 IU/kg i.p.) or saline solution at the end of O₃ exposure. Thirty minutes after insulin injection, rats were deeply anesthetized with sodium pentobarbital (60 mg/kg i.p.). Lungs, gastrocnemius or tibialis muscles, retroperitoneal white adipose tissue, and liver were rapidly dissected out, snap frozen in liquid nitrogen, and stored at −80°C.

Plasma total cholesterol, triacylglycerol, and nonesterified fatty acid (NEFA) levels were measured with commercial kits Cholesterol RTU (bioMérieux, Marcy-l’Etoile, France), Triglyceride PAP (bioMérieux), and NEFA-C (WAKO). Levels of plasma insulin (SPIBio, Montigny-le-Bretonneux, France), corticosterone (Cayman Chemicals), tumor necrosis factor-α, and interleukin-1β (eBioscience, Paris, France) were determined with enzyme immunoassays according to manufacturer’s recommendations. Plasma aldehydes (i.e., 4-hydroxy-2-nonenal [HNE] and 4-hydroxy-2-hexenal [HHE]) were measured using gas chromatography–mass spectrometry as previously described (30). Plasma malondialdehyde (MDA), reduced glutathione (GSH), and oxidized glutathione (GSSG) levels were measured by high-performance liquid chromatography coupled to fluorescence detection. Carbonyl groups on proteins were determined using 2.4-dinitrophenylhydrazine as described previously (31).

Total RNAs from gastrocnemius muscle samples (80–100 mg) were extracted using TRI Reagent (Sigma-Aldrich). Total RNA quantities and qualities were assessed using the Model 2100 Bioanalyzer and RNA 6000 LabChip Kit (Agilent Technologies, Massy, France). Reverse Transcriptions (RNAse H, Takara enzyme) were performed on 1 μg of total RNA. Real-time PCR assays for ER stress genes (Grp78/Bip, Atf4, Atf6, Ire1-α, Perk, total and unspliced Xbp1, and Eif2) were performed using a Rotor-GeneTM 6000 (QIAGEN). Values were normalized to TBP gene expression. The full list of genes and corresponding primer sequences is shown in Supplementary Table 1.

C2C12 myoblasts (Mus musculus) (catalog #CRL-1772; ATCC, LGC Standards, Molsheim, France) were grown and differentiated to myotubes as described previously (28). C2C12 myoblasts were cultured in DMEM supplemented with 10% heat-inactivated FBS (South America Origin; Biowest), 2 mmol/L l-glutamine, 100 units/mL penicillin, and 100 µg/mL streptomycin. When reaching 90% confluence, differentiation was induced by shifting to DMEM supplemented with 2% heat-inactivated horse serum (South America Origin; Biowest). Experiments were performed when >80% of cells had formed myotubes. Cells were serum starved for 2 h prior to the experiments. Cells were incubated for 30 min with 10% (volume for volume [v/v]) BALF diluted in cell culture media (either from O₃ or from control rats). The cell culture medium containing 10% (v/v) BALF was discarded, and cells were further incubated in a fresh medium containing 100 nmol/L insulin for 20 min. Then, cell culture medium was discarded and cells were frozen. LBA cytotoxicity was determined measuring cell viability with an MTT assay (Roche).

Pretreatment for 1 h with the c-Jun N-terminal kinase (JNK) inhibitor SP600125 (10 μmol/L) or PBA (10 mmol/L) was performed prior to BALF incubation. An UltraLinkHydrazide Resin (Thermo Scientific, Illkirch, France) was used as an affinity support for immobilizing reactive lipid aldehydes present in BALF. C2C12 myotubes were incubated with 10% (v/v) of aldehyde-depleted BALF, as described above.

C2C12 myotubes were incubated for 20 min with 100 nmol/L insulin. Glucose uptake was measured using 2-deoxy-d-[³H]-glucose (747 GBq/mmol; PerkinElmer, Courtaboeuf, France), as described previously (28,32).

Muscle tissue and C2C12 cells were lysed on ice in Standard Lysis Buffer (20 mmol/L Tris, 138 mmol/L NaCl, 2.7 mmol/L KCl, 1 mmol/L MgCl2, 5% glycerol, and 1% NP40, supplemented extemporaneously with 5 mmol/L EDTA, 1 mmol/L Na3VO4, 20 mmol/L NaF, 1 mmol/L dithiothreitol, and a protease inhibitor cocktail) and centrifuged (13,000g, 15 min, 4°C). Protein concentration was determined with a Bradford protein assay. Sixty micrograms of protein lysate from skeletal muscle or 30 μg protein lysate from C2C12 cells were separated on SDS-polyacrylamide gels (10%) after heat denaturation (95°C, 5 min) and then transferred onto a nitrocellulose membrane (Hybond-ECL; GE Healthcare, Meylan, France). After transfer, the membranes were blocked with 5% BSA in Tris-buffered saline with Tween for 2 h. Blots were probed overnight with specific primary antibodies at 4°C followed by 1-h incubation at room temperature with secondary antibodies. Protein bands were detected with an enhanced chemiluminescence substrate kit (SuperSignal West Pico; Perbio Sciences France) using an Image Master VDS-CL camera (Amersham Pharmacia, Orsay, France) and quantified by Quantity One (Bio-Rad) or open-source ImageJ (http://rsbweb.nih.gov/ij/index.html) software. Protein phosphorylation levels were normalized to the matching densitometric values of total proteins. Results were expressed as arbitrary units (a.u.).

Data are expressed as the mean ± SEM. All data were analyzed using GraphPad Prism version 5.0 software (GraphPad Software, La Jolla, CA). Multiple comparisons were performed using ANOVA followed, when appropriate, by post hoc Fisher protected least significant differences tests. The results of ITTs, arginine tests, and clamps were compared by two-way ANOVA (time and treatment). Simple comparisons were performed using the Student t test. When appropriate, Welch correction for inequality of variances was applied. Differences were considered significant at the P < 0.05 level.

O₃-exposed rats exhibited a significant rise in fasting blood glucose and insulin concentrations (+23% [P < 0.01] and +128% [P < 0.05], respectively), while plasma lipid concentrations remained unaltered. Consequently, the HOMA-IR was higher after O₃ exposure (+150%, P < 0.05), suggesting an impairment in insulin sensitivity (Table 1). GIR measured using EH clamps (Fig. 1A and B) was significantly decreased in O₃-exposed rats compared with control rats (9.9 ± 1.9 vs. 24.2 ± 2.1 mg ⋅ kg⁻¹ ⋅ min⁻¹, P < 0.01), indicating marked whole-body IR. This was not due to a decrease in insulin secretion capacity, as the acute insulin secretion induced by the nonglucose secretagogue arginine was similar in the two groups (Fig. 1C). In good agreement, the results of an intravenous glucose tolerance test (Supplementary Fig. 1) indicated that glucose-induced insulin secretion was normal after O₃ exposure. Taken together, these results demonstrate that O₃-induced glucose intolerance is due to peripheral IR rather than a deficit in insulin secretion.

To gain insights into the molecular determinants of O₃-induced IR, we assessed activation of the insulin signaling pathway in skeletal muscle. Insulin-induced protein kinase B (PKB)/Akt phosphorylation was significantly impaired in gastrocnemius (P < 0.01 vs. control) (Fig. 1D) and extensor digitorum longus (data not shown) muscles of O₃ rats, confirming that O₃-induced glucose intolerance is due to skeletal muscle IR. In contrast, insulin signaling was not impaired in liver and white adipose tissue (Supplementary Fig. 2A and B), suggesting that IR was restricted to skeletal muscle.

Since activation of stress protein kinases can interfere with insulin signaling pathways, we studied the activation of JNK, extracellular signal–related kinase (ERK) 1/2, and p38 mitogen-activated protein kinase (MAPK) in gastrocnemius muscle. O₃ exposure specifically increased the phosphorylation of JNK (5.6-fold, P < 0.01), while p38 and ERK remained unaffected, suggesting a particular role of p46/p54 JNK in O₃-induced IR (Fig. 1E–G). A dose-response experiment (with O₃ concentrations ranging from 0.1 to 0.8 ppm) gave evidence that a significant effect on insulin sensitivity was noticeable from a concentration of 0.3 ppm O₃ (Supplementary Fig. 3).

As O₃ is too reactive to penetrate deeply into the tissues and reach the bloodstream, we hypothesized that an oxidative and/or inflammatory insult in the lung might be responsible for the peripheral IR. Animals exposed to O₃ had increased infiltration of cells in the lung (15-fold, P < 0.001) (Fig. 2A) due to an increase in neutrophils, lymphocytes, and monocytes (Fig. 2B). Accordingly, alveolar capillary membrane permeability was increased, as revealed by a markedly increased protein concentration in the BALF of O₃-exposed rats (12-fold, P < 0.01) (Fig. 2C). Infiltration of immune cells and increased alveolar permeability confirmed that O₃ induced an important inflammatory response in the lung; however, no difference in the number of inflammatory cytokines was found in either BALF or plasma (Supplementary Table 2), suggesting that there was little contribution of inflammation to the remote metabolic effects of O₃ exposure.

O₃ is a strong oxidant, and most of its toxic effects are mediated through free radical reactions. Indeed, lipid peroxidation, measured through thiobarbituric acid–reactive substances (TBARS) concentration, was markedly increased in BALF from O₃-treated animals (threefold, P < 0.001) (Fig. 2D). HNE and HHE, two reactive lipid aldehydes (among the most cytotoxic products of lipid peroxidation) were increased in BALF from O₃-exposed rats (Fig. 2E). Protein carbonyls, end products of oxidative stress, were also significantly increased in the lung (ninefold, P < 0.001) (Fig. 2F), confirming the oxidative stress in lungs after O₃ exposure. Surprisingly, and unlike inflammation, oxidative stress was not limited to the lung. O₃ exposure induced an increase in many oxidative stress biomarkers in blood (Table 1), such as lipid peroxidation by-products (2.5-, 2.0-, and 1.6-fold increases in HHE, HNE, and MDA concentrations, respectively, P < 0.05), and protein carbonyls (by 63%, P < 0.001). In good agreement, the GSH-to-GSSG ratio in erythrocytes was decreased in O₃ exposed rats (by 66%, P < 0.05). Even more striking was the increase in protein carbonyls in gastrocnemius muscle (3.3-fold, P < 0.001) after O₃ exposure (Fig. 2F), suggesting that oxidative stress can spread into peripheral tissues and be responsible for the remote effects of O₃.

In order to confirm that toxic mediators produced in the lung could affect peripheral tissues, plasma and BALF samples collected from control and O₃-treated rats were used to treat C2C12 myotubes in vitro. No significant difference in cell viability was observed between control and BALF-treated cells (Supplementary Fig. 4), but a 30-min pretreatment with BALF (10% v/v) from O₃ rats completely abolished the insulin-stimulated [³H]-2-deoxy-d-glucose transport in myotubes (Fig. 3A). Accordingly, insulin-induced serine phosphorylation of PKB/Akt (Fig. 3B) and tyrosine phosphorylation of insulin receptor substrate-1 (IRS-1) (Fig. 3C) were totally inhibited after O₃ BALF exposure compared with clean-air BALF. Taken together, these results demonstrate that BALF from O₃-exposed rats impaired insulin signaling in muscle cells. In good agreement with in vivo experiments, C2C12 myotubes preincubated with BALF from O₃-exposed rats exhibited a 1.5-fold increase in JNK phosphorylation with or without insulin (P < 0.05) (Fig. 3D), while no change was observed for ERK1/2 (data not shown). Pretreatment of cells with the JNK inhibitor SP600125 (10 μmol/L, for 1 h) prior to BALF incubation totally inhibited JNK activation (Fig. 3E) and restored insulin-induced PKB/Akt phosphorylation (Fig. 3F), pointing out the specific role of JNK in mediating the effects of BALF in C2C12 myotubes. In contrast with BALF, rat plasma decreased C2C12 cell viability (Supplementary Table 3). Insulin-induced serine phosphorylation of PKB/Akt (Supplementary Fig. 5A) was, however, totally inhibited after incubation with plasma from O₃-exposed rats, while JNK phosphorylation was increased (Supplementary Fig. 5B). These data fully confirm the results observed in experiments with BALF. Lipid aldehyde levels, which were increased in both BALF and plasma from O₃-exposed rats, are potent electrophiles that could mediate the effects of O₃ beyond the air/tissue interface. To support this view, we used a hydrazide resin as an affinity support for immobilizing lipid aldehydes present in BALF and incubated C2C12 cells with “aldehyde-depleted BALF.” After treatment with hydrazide resin, the concentrations of HHE and HNE in BALF remained lower than the limit of detection. C2C12 muscle incubated with aldehyde-depleted BALF did not display any disruption of the insulin pathway, as evidenced by insulin-induced serine 473 phosphorylation of PKB/Akt (Fig. 3G), suggesting that lipid aldehydes, produced in the lung under the action of O₃, could contribute to IR.

As oxidative stress seemed to be a major determinant of the noxious effects of O₃, we studied the ability of the antioxidant NAC (225 mg ⋅ kg ⁻¹ ⋅ day⁻¹, for 10 days) to prevent O₃-induced IR in rat. NAC pretreatment significantly increased the GSH-to-GSSG ratio (1.5-fold, P = 0.03) in muscle (Supplementary Fig. 6) and prevented its decrease after O₃ exposure. NAC treatment prevented the decrease in insulin sensitivity (Fig. 4A) observed after overnight O₃ exposure, as well as the increase in protein carbonyl content in lung and muscle (Fig. 4B) and the TBARS increase in BALF (Fig. 4C). Interestingly, NAC pretreatment failed to prevent lung inflammation, as evidenced by increased protein concentrations (ninefold, P < 0.01) and cell counts (eightfold, P < 0.001) in BALF (Fig. 4D and E), suggesting that the inflammatory components of the noxious effects of O₃ were upstream of oxidative stress. Finally, treatment of C2C12 myotubes with BALF from rats pretreated with NAC prior to O₃ exposure neither impaired insulin-induced Akt phosphorylation (3.6-fold, P < 0.001) (Fig. 4F) nor increased JNK phosphorylation (Fig. 4G), demonstrating that oxidative stress is a major mediator of the peripheral effects of O₃ on muscle insulin sensitivity.

Under conditions of oxidative stress, the accumulation of oxidized lipids or advanced oxidation protein by-products can induce ER stress and JNK activation. O₃ exposure, however, triggered a limited effect on ER stress protein gene expression in rat gastrocnemius muscles (Fig. 5A) since changes were restricted to the overexpression of Atf6 (increase of 106%, P < 0.01) and Ire1a (increase of 36%, P = 0.06) genes, while the expression of other ER stress protein genes such as Bip (Gpr78), Perk, Eif2s1, or Atf4 remained unaltered. O₃-exposed rats exhibited a modest post-translational activation of inositol-requiring protein-1α (IRE-1α) (1.6-fold, P = 0.1) (Fig. 5B), but not PERK (data not shown), in their gastrocnemius muscles. Incubation of C2C12 myotubes with BALF from O₃ rats recapitulated the activation of IRE-1α (Fig. 5C) (1.6-fold, P < 0.05). To confirm the activation of JNK downstream of ER stress, we pretreated C2C12 myotubes with a chemical chaperone, 4-PBA (10 mmol/L, for 1 h), to prevent ER stress. Pretreatment with PBA prevented the O₃-induced activation of JNK (Fig. 5D) and restored the activation of PKB/Akt in response to insulin (Fig. 5E). These results were confirmed in vivo, as PBA pretreatment (150 mg ⋅ kg⁻¹ ⋅ day⁻¹, for 4 days) significantly prevented O₃-induced IR, as measured with an ITT (Fig. 5F). Taken together, these data demonstrate that the induction of selective ER stress pathways in skeletal muscle could contribute to O₃-induced IR.

O₃ pollution peaks are transient and rarely last for more than a few days. To decipher how long O₃-induced IR could persist after the end of the exposure, intraperitoneal ITTs were performed immediately after O₃ exposure, and 24, 72, and 96 h later. Strikingly, even 96 h after O₃ exposure, animals exhibited a significant impairment in insulin sensitivity (Fig. 6A). The K_ITT was lower in all O₃-exposed groups compared with control rats (n = 5, P < 0.05). In skeletal muscle, JNK phosphorylation also lasted for up to 96 h (Fig. 6B) after O₃ exposure, while ER stress (IRE-1) nearly returned to baseline after 24 h (Fig. 6C). In contrast, oxidative stress and inflammatory markers in BALF returned to baseline in 24–48 h (Fig. 6D–G). These results show that an overnight acute exposure to O₃ can induce defects in skeletal muscle and impair insulin sensitivity for up to 3 days.

Rats underwent a subchronic exposure to a lower dose of O₃ (0.25 ppm, 12 h/day) for 4 days. We deliberately chose an intermittent O₃ exposure to take into account the day-to-night variations that generally characterize O₃ pollution in urban areas. Of note, long-term exposure to O₃ (0.25 ppm), as observed with the acute exposure study, yielded pulmonary inflammation, oxidative stress, and IR (Fig. 7).

Diet and sedentary lifestyle are the most common risk factors for cardiometabolic disorders; however, recent observations have also shown a strong association between other environmental factors such as air pollution (especially from traffic-related sources) and T2D (5). More and more recent cohort studies indeed associate traffic-related air pollution with the incidence of T2D. For instance, exposure to NO₂ is associated with diabetes prevalence among women (33) and with the development of diabetes in healthy individuals (4). PM pollution, consisting of airborne particles in solid or liquid form of <10 µm (PM₁₀), is also associated with incident T2D among elderly women (3). Although most aspects of the association of metabolic diseases with air pollution remain to be elucidated, these observations are of importance, given the high levels of airborne pollutants in urbanized environments, which could represent an underappreciated but critical link between industrialization and the increased incidence of metabolic diseases (34). Since it is produced from other pollutants, O₃ is the reflection of advanced polluted air environments and only recently has attracted the attention of scientists. O₃ is a major photochemical pollutant in urban areas, produced from the action of UV rays on volatile organic compounds or NO_x (NO, NO₂) emitted during the combustion of fossil fuels by automobile exhausts and industrial activities (6). As with PM and NO₂, O₃ is also associated with increased fasting glucose and insulin levels, and HOMA indices in elderly (35). The association of O₃ with these parameters is stronger among participants with a history of diabetes or increased susceptibility to oxidative stress, suggesting that O₃ could boost the development of diabetes.

The major finding of the current study is that a realistic acute O₃ exposure induces peripheral IR in vivo. This result was confirmed using the following three different methods: HOMA-IR, insulin and glucose tolerance tests, and the gold standard EH clamp technique. Skeletal muscle is the major tissue responsible for insulin-stimulated glucose disposal in lean individuals and, as such, has been identified to be pivotal for the development of IR (36,37). The measurement of insulin-induced PKB/Akt phosphorylation in skeletal muscles (gastrocnemius and extensor digitorum longus) further confirmed that O₃ induced a selective defect in muscle insulin sensitivity compared with liver and adipose tissue (Fig. 1).

Ground-level O₃ air pollution can often reach 0.18 ppm (i.e., 360 µg/m³), corresponding to level 3 of the alert threshold established by European Directive 2008/50/CE. This concentration of O₃ can cause a number of respiratory health problems, and can worsen asthma, allergies, and chronic obstructive pulmonary diseases (38). O₃ was primarily studied in this context, and a rat model that mimics O₃ exposure was developed and extensively characterized. In particular, it was demonstrated that rats remove a smaller fraction of their inhaled amount of O₃ (40–47%) than humans (75% with large interindividual variations) (39,40). As a consequence, the toxicity of O₃ observed for a given concentration in rats strongly underestimates the effect observed for the same concentration in humans, and thus, the effects observed in humans are comparable to those produced in rats exposed to a fivefold higher O₃ concentration (41). The concentrations chosen in the current study (0.25–0.8 ppm) are therefore realistic to mimic the actual levels of O₃ commonly encountered in urban areas during pollution peaks. Of note, IR was noticeable from concentrations of 0.25–0.3 ppm O₃ (Supplementary Fig. 3).

O₃ is too reactive to penetrate deeply into the tissues and reach the bloodstream (42). Instead, it promptly reacts with molecules of lung-lining fluid and/or with plasma membranes of airway epithelial cells to produce lipid aldehydes and lipid ozonation products, which are highly susceptible to initiating pathophysiological cascades (42). Aldehydes and/or lipid ozonation products have a lower reactivity and longer t_1/2 than O₃ itself and could mediate the effects of O₃ beyond the air/tissue interface. To support the idea of O₃-induced lung mediators that trigger IR, we recapitulated the effects of O₃ in vitro by treating C2C12 myotubes with BALF. Muscle cells treated with BALF from O₃-exposed rats exhibited an inhibition of insulin-stimulated glucose uptake and Akt phosphorylation, similar to in vivo O₃ exposure (Fig. 3). Moreover, partial inhibition of insulin-induced PKB/Akt phosphorylation could also be achieved with BALF from control animals submitted to in vitro ozonation (Supplementary Fig. 7). As most of these factors are likely oxidation products, this is consistent with previous reports (43,44) showing that lipid ozonation products such as cholesterol secoaldehydes have deleterious effects, inducing apoptosis in cardiomyocytes and neurons. In addition, publications by us and others (32,45,46) showed that lipid aldehydes and carbonyl stress trigger IR in muscle cells and adipocytes. Our results support that it is the reaction of O₃ with components of lung-lining fluid and/or surfactant that produces toxic by-products/factors that are able to induce IR. We further demonstrated that lipid aldehydes (e.g., HNE and HHE) produced under O₃ exposure could participate in these extrapulmonary effects of O₃ (Fig. 3G).

The lung is obviously the first tissue affected by O₃, and we indeed observed a large increase in protein concentration and immune cell infiltration, which are hallmarks of lung inflammation (Fig. 2). Although inflammation seems to be an important component of the O₃ response, it is limited to the lung, as overnight exposure to O₃ yielded only minor changes in systemic inflammatory cytokine and prostaglandin levels. In addition, the treatment of rats with the antioxidant NAC does not prevent lung inflammation (Fig. 4) but does restore insulin sensitivity in the muscle. Furthermore, recovery experiments showed that infiltration of the lung by immune cells resolves after 24 h, whereas muscle IR lasts for up to 3 days (Fig. 6). Taken together, these results strongly suggest that inflammation is not directly responsible for the defects observed in muscle. This is consistent with a recent study (14) showing that lung inflammation was not a requirement for the production of circulating vasoactive factors after O₃ exposure. Lung inflammation is, however, an important source of radical species (47) and may contribute to the oxidative chain reactions induced by O₃. We indeed observed that rats exposed to O₃ exhibit a systemic increase in blood reactive aldehyde levels (Table 1), as well as an increase in muscle carbonyl levels. In addition, pretreatment of rats with the antioxidant NAC prevented oxidative stress, carbonylation of muscle proteins, and IR, strongly suggesting that the systemic effects of O₃ are due to oxidative stress (Fig. 4).

In response to several types of stress, the aggregation of misfolded or abnormal proteins leads to ER stress, consequently triggering the unfolded protein response (UPR) pathways in order to restore ER homeostasis. This response can be induced by pollutants and is involved in PM toxicity. In vitro studies (48) have demonstrated that PM_2.5 (PM <2.5 µm) induces ER stress and significantly increases the UPR-associated proteins activating transcription factor-4, HSP70 and HSP90, and BiP. ER stress has been proposed to play a central role in the development of IR in liver and adipose tissue (49), and, indeed, the UPR intersects with a variety of inflammatory and stress-signaling systems, including the nuclear factor-κB and JNK pathways, as well as oxidative stress responses, all of which may influence lipid and glucose metabolism (49). Our study demonstrates that exposure to O₃ in vivo or to BALF from O₃-exposed rats in vitro triggers a modest activation of ER stress pathways in muscle, mostly restricted to the IRE1-α pathway. Using the chemical chaperone 4-PBA, which prevents ER stress and ER stress–mediated activation of JNK (50), we demonstrated in vivo and in vitro that 1) activation of selective ER stress pathways could contribute to the O₃-induced IR in muscle cells and 2) JNK activation is downstream of the ER stress activation in this context (Fig. 5).

Whole-body IR can be a consequence of defects in the canonical insulin signaling pathway IRS-1–phosphatidylinositol 3-kinase–PKB/Akt. Increased serine phosphorylation of IRS-1 is indeed triggered by many diabetogenic factors, especially through the stress signaling pathways that have been shown to contribute to the development of IR (51). PM_2.5 exposure activates JNK and downregulates the IRS1-mediated signaling in the liver, leading to glucose intolerance and IR (52). In addition, specific overexpression of a constitutively active JNK in skeletal muscle has been shown to be sufficient to induce whole-body IR through impairment of insulin signaling (53). In our rat model, O₃ is responsible for a specific activation of JNK signaling in muscle, either after in vivo exposure or after treatment with BALF from O₃-exposed rats. The inhibition of JNK can fully restore the impaired insulin-induced PKB/Akt phosphorylation, demonstrating the important role of JNK in O₃-induced IR (Fig. 3).

In summary, we report here that O₃ plays a causative role in the development of IR and that a realistic, short-term exposure to O₃ causes whole-body IR that lasts for up to 3 days. This alarming phenotype is due to the production of lung mediators that induce oxidative stress and the subsequent activation of JNK in skeletal muscle, therefore disrupting insulin-induced signaling and glucose uptake. Air pollutants generally occur in complex mixtures that create the potential for synergistic effects among pollutants and may worsen their pathological properties. Our results suggest that nontraditional factors such as O₃ pollution could synergize with other dominant factors (obesity and sedentary lifestyle) to accelerate the propensity for T2D and that chaperone molecules or antioxidants could protect susceptible people from complications induced by O₃ pollution peaks.

Acknowledgments. The authors thank Dr. Kevin P. Foley (Cell Biology Program, The Hospital for Sick Children, Toronto, ON, Canada) for proofreading and corrections.

Funding. This work was supported by the Institut National de la Santé et de la Recherche Médicale and the Institut National des Sciences Appliquées de Lyon. R.E.V., N.J.P., B.Z., and M.L.C. were supported by grants from the French Ministère de la Recherche et de la Technologie. L.K. held a fellowship from Fondation pour la Recherche Médicale.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. R.E.V., N.J.P., and C.O.S. researched the data and wrote the manuscript. B.Z., M.L.C., M.G., and A.G. researched the data, and reviewed and edited the manuscript. L.K., S.P., M.-A.C., and J.R. researched the data. R.E.V., N.J.P., and C.O.S. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.