|Home | About | Journals | Submit | Contact Us | Français|
Pulmonary responses to the air pollutant, ozone, are increased in obesity. Both obesity and ozone cause changes in systemic metabolism. Consequently, we examined the impact of ozone on the lung metabolomes of obese and lean mice. Lean wildtype and obese db/db mice were exposed to acute ozone (2 ppm for 3 h) or air. 24 hours later, the lungs were excised, flushed with PBS to remove blood and analyzed via liquid-chromatography or gas-chromatography coupled to mass spectrometry for metabolites. Both obesity and ozone caused changes in the lung metabolome. Of 321 compounds identified, 101 were significantly impacted by obesity in air-exposed mice. These included biochemicals related to carbohydrate and lipid metabolism, which were each increased in lungs of obese versus lean mice. These metabolite changes may be of functional importance given the signaling capacity of these moieties. Ozone differentially affected the lung metabolome in obese versus lean mice. For example, almost all phosphocholine-containing lysolipids were significantly reduced in lean mice, but this effect was attenuated in obese mice. Glutathione metabolism was also differentially affected by ozone in obese and lean mice. Finally, the lung metabolome indicated a role for the microbiome in the effects of both obesity and ozone: all measured bacterial/mammalian co-metabolites were significantly affected by obesity and/or ozone. Thus, metabolic derangements in obesity appear to impact the response to ozone.
Ozone (O3) is a common air pollutant produced by exposure of automobile exhaust to sunlight. Exposure to O3 causes respiratory symptoms, increases susceptibility to pulmonary infections, and even increases the risk of mortality in those with underlying cardiorespiratory conditions [1–5]. O3 is a particular problem for asthmatics. Even O3 concentrations near the EPA standard are sufficient to reduce lung function in asthmatic children . Hospital admissions and emergency room visits for asthma increase after days of high ambient O3 concentrations [6–8] and O3 causes airway hyperresponsiveness (AHR) , a canonical feature of asthma. O3 injures lung epithelial cells, and the ensuing inflammatory response, which includes production of numerous cytokines and chemokines and recruitment of neutrophils , likely also contributes to the capacity of O3 to trigger asthma.
Obesity amplifies the impact of O3 on the lungs. O3-induced increases in asthma symptoms are greater in obese than lean children . Similarly, O3-induced reductions in lung function are greater in obese than lean adult human subjects, especially if those subjects also exhibit AHR [12,13]. Similar results are obtained in mice. Obese mice exhibit innate AHR . O3-induced increases in airway responsiveness and in pulmonary neutrophil recruitment are also greater in obese than in lean mice [15–17]. The mechanistic basis for these augmented responses to O3 remains incompletely understood.
Metabolomic profiling offers a means of discovering metabolic pathways that underlie disease. For example, the importance of trimethylamine N-oxide (TMAO) for cardiovascular disease was first predicted from metabolomic profiling studies . Obesity is a metabolic disease and there are marked effects of obesity on the serum and urinary metabolomes in humans, rats, and mice including changes in carbohydrate, lipid, and branched chain amino acid (BCAA) metabolism [19–21]. Lungs of naive obese mice also exhibit metabolic changes, including changes in lipid, phospholipid, and cholesterol metabolism . In rats, acute exposure to O3 causes profound changes in the serum metabolome including increases in sugars, free fatty acids, BCAAs (valine, leucine, and isoleucine, and urea, indicating impaired glycemic control, lipolysis, and proteolysis . Similar results are obtained in human subjects . The lipid mobilization and increased glucose induced by O3 are consistent with effects of O3 on adipose tissue and liver, tissues that are substantially altered in obese mice. Indeed, transcriptomic profiling of livers from O3 exposed mice confirmed increased expression of genes involved in gluconeogenesis and decreased expression of genes involved in triglyceride biosynthesis . Importantly, these systemic effects of O3 appear to contribute to O3-induced injury and inflammation within the lungs . Nevertheless, effects of O3 on the lung metabolome have not been described in either lean or obese mice, though changes in the lung metabolome do accompany pulmonary exposures to other inhaled irritants and infectious agents [26–30].
The purpose of this study was to perform global metabolomic profiling on lungs of obese db/db mice and their lean wildtype (C57BL/6J) controls exposed to air or O3 in order to identify metabolites that could be contributing to the augmented responses to O3 observed in obese mice [15–17]. Mice were exposed to room air or to O3 (2 ppm) for 3 hours. Twenty-four hours later, the lungs were harvested, flushed with PBS to remove blood, flash frozen in liquid nitrogen, and analyzed via liquid-chromatography or gas-chromatography coupled to mass spectrometry for metabolites. The dose and timing of exposure were chosen to correspond with exposure conditions for which functional responses to O3 have already been established . Our results indicate profound differences in the lung metabolomes of unexposed obese and lean mice that include elevations in lipids and carbohydrates. These changes may be due to elevations in these moieties in the blood  and subsequent diffusion into the lung extracellular fluid. O3 also affected the lung metabolome. Importantly, there were differential effects of O3 in obese and lean mice, including effects on BCAA metabolites, lysolipids, and glutathione. The lung metabolome also indicated a role for the microbiome in the effects of obesity on pulmonary responses to O3.
These studies were approved by the Harvard Medical Area Standing Committee on animals. Female db/db mice on a C57BL/6 background and age- and sex-matched wildtype (WT) C57BL/6J mice were purchased from The Jackson Labs at age 6 weeks, allowed to acclimatize within the mouse vivarium at the Harvard T.H. Chan School of Public Health for 4 weeks, and studied at 10 weeks of age. A separate cohort of female WT and db/db mice that were bred and raised in house were used for measurements of serum insulin. Db/db mice lack the longform of the receptor for the satiety hormone, leptin.
For metabolomics, db/db and WT mice were exposed for 3 hours to room air or to O3 (2 ppm) in stainless steel and plexiglass exposure chambers . During exposure, mice were placed within individual wire mesh cages and food and water were withdrawn. Mice were returned to their home cages immediately after exposure at which time food and water were restored. Mice were euthanized with an overdose of sodium pentobarbital 24 hours after cessation of exposure.
For mice in the metabolomics study, after euthanasia, blood was obtained by cardiac puncture for the preparation of serum. The chest wall was then opened and a small incision was made in the left ventricle. 10 ml of ice cold PBS was gradually injected into the right ventricle in order to flush blood from the lungs. The lungs were then flash frozen in liquid nitrogen and stored at -80°C until shipped on dry ice to Metabolon Inc. (Durham, NC). Upon receipt, the lungs were again frozen at -80°C until analysis.
In another cohort, db/db and WT mice were exposed to room air or O3 as described above. Blood was obtained and the lungs flushed of blood and frozen in liquid nitrogen as described above. In this cohort, lung tissue was used to prepare RNA for microarray and qRT-PCR analysis. Microarray data have been deposited at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE81800).
The automated MicroLab STAR® system from Hamilton Company was used to prepare samples for metabolomics profiling. Equal weight of lungs was used from each mouse. Prior to the first step in the extraction process, recovery standards were added for quality control (QC) purposes. A series of organic and aqueous extractions proprietary to Metabolon Inc. was then used to remove protein while allowing maximum recovery of small molecules. Extracts were divided into two fractions. One was used for analysis by liquid chromatography (LC) and the other was used for analysis by gas chromatography (GC). Samples were placed briefly on a TurboVap® (Zymark) to remove the organic solvent, frozen, and dried under vacuum. Samples were then prepared for the appropriate instrument, either LC/MS (mass spectrometry) or GC/MS, as described by Evans et al . For a brief description of the methods used for LC/MS and GC/MS, see the S1 Method. After analysis, raw data were extracted, peak-identified and QC processed using Metabolon's proprietary hardware and software. At the time of this analysis, identification of known chemical entities was based on comparison to metabolomic libraries of more than 1000 commercially available purified standard compounds. The combination of chromatographic properties and mass spectra gave an indication of a match to the specific compound or an isobaric entity. This data is available at the NIH Common Fund's Metabolomics Data Repository and Coordinating Center (supported by NIH grant, U01-DK097430) website, the Metabolomics Workbench, http://www.metabolomicsworkbench.org ID:934.
After excision, the right lung was immersed in RNAlater (Qiagen) for subsequent preparation of RNA . A small volume spectrophotometer (Nanodrop, Thermo Scientific) was used to assess RNA concentration and purity and a commercial kit (SuperScript III for qRT-PCR, Invitrogen) was used to convert RNA into cDNA. Gclc mRNA abundance was quantified using real time PCR (7300 Real-Time PCR Systems, Applied Biosystems) with SYBR-green detection and normalized to 36B4 ribosomal RNA (Rplp0). Primers for Gclc were forward–TGTGGTATTCGTGGTACTGCT and CTGGGCCACTTTCATGTTCTC. Primers for Gsta1 were forward: ACCTGATGCACTCCATTCTG and reverse: GCTGGACTGTGAGCTGAGTG. Primers for Rplp0 were as described . The ΔΔCt method was used to assess changes in mRNA abundances.
Serum was analyzed for insulin using an ELISA assays (EMD Millipore) according to the manufacturers’ recommendations.
For mRNA abundances and serum analytes: factorial ANOVA using genotype and expression as main effects and Fisher’s LSD test as follow up, was used to assess the significance of differences. For these outcomes, p<0.05 was considered statistically significant.
For metabolomics data analysis: missing values when present due to being under the limit of detection of the instruments were imputed with the minimum value on a per metabolite basis. For each metabolite, raw peak area counts were rescaled to set the median across all samples for that metabolite to 1 and the data log transformed. Then, a two way ANOVA consisting of the factors genotype and ozone treatment. Follow-up pairwise contrasts were also conducted to compare individual group means using F-tests. Storey’s q-values were calculated to estimate the proportion of false positives (see S1 Table for p and q values for each metabolite). PCA analysis was performed using the prcomp function in R 3.4  (https://www.R-project.org/) and plotted with gglot2 package .
For metabolic pathway enrichment analysis, enrichment factors (EF) were calculated as follows where the significance of individual metabolites was assessed as p<0.05:
Fisher’s exact test was used to assess the significance of pathway or superpathway enrichment. A false discovery rate (FDR) was computed to account for multiple comparisons.
A p-value or q value <0.05 was considered statistically significant depending on whether individual metabolites (q value) or metabolites within a significantly affected pathway (per FDR analysis) (p value) were being assessed.
Among the 321 biochemicals identified in lung tissue, two-way ANOVA identified 171 that were significantly affected by db genotype, 71 that were affected by O3 exposure, and 14 for which there was an interaction between genotype and O3. Differences between individual experimental groups are shown in Table 1. Both in air and O3 exposed mice, approximately 1/3 of identified biochemicals were different in db/db versus WT mice, with the majority being increased in the db/db mice. Principal component analysis (PCA) indicated a clear separation in the lung metabolomes of the db/db versus WT mice, both when all 4 groups were considered together (Fig 1), and when either the air-exposed or O3-exposed mice were considered separately (S1 Fig). O3 exposure also caused significant changes in lung biochemicals in both WT and db/db mice, with the majority being reduced in the O3- versus air-exposed mice (Table 1).
Below, we first describe the metabolic pathways affected by db genotype and by O3 exposure using metabolic pathway enrichment analysis of the significantly affected metabolites from the two-way ANOVA (Table 2). We then describe the results of a similar analysis of metabolic pathway enrichment using the results of the analysis comparing the 4 individual groups (Table 3). Next we discuss the lung metabolomes of air-exposed db/db versus WT mice, the lung metabolomes of WT mice exposed to air versus O3, and the differential impact of O3 exposure on the lung metabolomes of db/db versus WT mice, in each case focusing on those metabolites in pathways identified in Tables Tables22 and and3.3. Lastly, we describe the impact of obesity and O3 on metabolic substrates used for energy production, metabolites related to oxidative stress, and on bacterial mammalian co-metabolites, since these categories include metabolites from numerous different metabolic pathways.
Lung metabolites significantly affected by db genotype or by O3 exposure in the two-way ANOVA are highlighted in blue in S1 Table. The metabolic superpathways and pathways to which these metabolites belong are also indicated in S1 Table. In order to determine whether there were metabolic pathways that were enriched among those metabolites affected by db genotype or by O3, we calculated an enrichment factor (Table 2) for each superpathway, and for affected pathways that were part of any significantly affected superpathway or for which there were pre-existing data to suggest an effect of O3 or obesity (e.g. metabolism of BCAAs). Enrichment factors were calculated as described in the methods. Pathways with a significant enrichment of significantly altered metabolites (as indicated by an FDR < 0.1) are indicated in bold text in Table 2. This analysis indicated a significant effect of db genotype on carbohydrate metabolism, particularly metabolites involved in fructose, mannose, galactose, starch, and sucrose metabolism and metabolites involved in glycolysis, gluconeogenesis, and pyruvate metabolism. Metabolites involved in BCAA metabolism were also affected by db genotype. Enrichment factor analysis also indicated a significant effect of O3 exposure on lipid metabolites, particularly lysolipids and metabolites involved in carnitine metabolism.
We performed a similar analysis on lung metabolites identified as being significantly affected by db genotype in either air or O3 exposed mice or significantly affected by O3 exposure in either WT or db/db mice (Table 3). These metabolites are highlighted in red (increased) and green (decreased) in S1 Table. Enrichment factor analysis indicated that the effect of db genotype on carbohydrate metabolism was observed in both air and O3 exposed mice and that there was also an effect of db genotype on fatty acids in the air but not O3 exposed mice. The data also indicated that the effect of O3 exposure on lysolipids was limited to the WT mice.
As described above, pathway enrichment analysis indicated that metabolites related to carbohydrate and lipid metabolism were different in lungs of db/db and WT mice exposed to room air (Table 3). This metabolomic profile is similar to that reported in the blood of db/db versus WT mice . The carbohydrates involved were primarily those related to glucose metabolism and the metabolism of other simple sugars (Table 3 and S1 Table). As shown in Fig 2, most of these metabolites were increased in the lungs of db/db versus WT mice, consistent with the systemic hyperglycemia and insulin resistance of db/db mice. Only 1,5-anhydroglucitol (1,5-AG) was reduced in the db/db mice (Fig 2), likely as a result of increased loss in the urine because of the high filtered glucose load in the kidney.
Lipids, particularly fatty acids, were also altered in the lungs of db/db versus WT mice exposed to air (Table 3). Most fatty acids were increased in db/db versus WT mice (Fig 3), as were glycerol, and the ketone BHBA (Fig 3), consistent with the marked systemic insulin resistance and consequent lipolysis characteristic of db/db mice. There were also changes in other lipids in the lungs of db/db mice (S1 Table). For example, although cholesterol itself was unchanged, there were elevations in several cholesterol metabolites, including dihydrocholesterol, 7-α-hydroxcholesterol, 7-β-hydroxycholesterol, and 7-ketocholesterol (S1 Table). Milner et al  also observed significant changes in many lipid moieties in lungs of obese versus lean mice, although their analytic methods differed from ours.
Other metabolites that differed in lungs of air exposed db/db versus WT mice are found in S1 Table.
In WT mice, examination of metabolic pathways affected by O3 exposure indicated an effect on lysophospholipids (Table 3). In particular, O3 exposure caused substantial decreases in lysolipids in WT mice (Table 4). Of the 38 measured lysolipids, 17 were significantly lower in O3- than air-exposed WT mice, Notably, of the 17 lysolipids reduced by O3 in lean mice, most were choline-containing lysolipids; the phospholipids that make up the majority of the phospholipids in lung surfactant . There was also a significant effect of O3 on the monoglycerides 1-linoleoylglycerol and 2-linoleoylglycerol (S1 Table).
Other biochemicals were also significantly affected by O3 in WT mice (see S1 Table). For example, O3 exposure caused an approximate 2-fold increase in lung citrulline in WT mice, and a similar effect in db/db mice. Arginine is converted to citrulline by nitric oxide synthase (NOS). Hence, the increase in citrulline following O3 exposure is consistent reports of increased NOS expression following O3 exposure in mice . O3 also caused a marked decrease in the lung heme in WT mice. The decrease in lung heme was likely the result of increased expression of heme-oxygenase, which catalyzes heme degradation, and which is known to increase following O3 exposure .
There were several notable differences in the metabolites affected by O3 in db/db versus WT mice (Tables (Tables22 and and3).3). For example, whereas there was a significant enrichment of lysolipids among the metabolites affected by O3 exposure in WT mice, this was not the case in db/db mice (Table 3), although 8 of 38 lysolipids were significantly lower in O3- than air-exposed db/db mice (Table 4). However, even for these latter lysolipids, the magnitude of the reduction was not as great in db/db as in WT mice. It is noteworthy that of the 8 lysolipids reduced by O3 in obese mice, all were choline-containing lysolipids. Similarly, whereas monoglycerides were elevated by O3 in WT mice no such effect was observed in db/db mice (S1 Table).
Hypotaurine, an osmolyte, was significantly increased after O3 in db/db but not WT mice (S1 Table). Others have reported reductions in hypotaurine in blood of db/db versus WT mice , indicating that the increases in lung hypotaurine were unlikely to derive from systemic sources. Hypotaurine acts as an antioxidant within the mammalian reproductive tract  and it is conceivable that elevated levels in db/db mice exposed to O3 are acting to protect the lungs from O3-induced oxidative damage.
Several other metabolites were different in db/db versus WT mice exposed to O3 even though no similar difference or trend was observed in db/db versus WT mice exposed to air. For example, the prostanoids PGE1, PGE2, PGI2, and 6-keto prostaglandin F1α were each significantly greater in db/db than WT mice exposed to O3 but not air (S1 Table). Many of the other metabolites in this category were lysolipids (see above) or were bacterial mammalian co-metabolites. The latter are discussed in more detail below.
Pathway enrichment analysis indicated a significant effect of db genotype on BCAA metabolism (Table 2). Two-way ANOVA indicated a significant effect of genotype on all 3 BCAAs (isoleucine, leucine, and valine) (Fig 4A, 4B and 4C), and the BCAA metabolites 3-hydroxyisobutyrate, alpha-hydroxyisovalerate, isobutyrylcarnitine, 2-methylbutyrylcarnitine, isovalerylcarnitine and hydoxyisovaleroylcarnitine (Fig 4D, 4E, 4F, 4G, 4H and 4I). Increases in BCAAs and their short chain acylcarnitine metabolites are typically observed in serum and tissues from obese mice and obese humans [19,40–42], and are thought to reflect reduced catabolism of BCAAs, a process that yields carbon sources that enter into energy production pathways. Indeed, BCAA metabolites were each higher in lungs of db/db versus WT mice (Fig 4). However, the effect was only observed after O3 exposure, in part because there were significant reductions in isobutyrylcarnitine, 2-methylbutyrylcarnitine, isovalerylcarnitine, and hydroxyisovaleroyl carnitine after O3 exposure in WT but not db/db mice (Fig 4F, 4G, 4H and 4I). Reductions in these BCAA metabolites are observed during catabolism of BCAAs . Hence, these data suggest increased reliance of lungs of WT but not db/db mice on BCAA catabolism for energy production after O3 exposure.
Pathway enrichment analysis also indicated a significant effect of O3 exposure on carnitine metabolism (Table 2). Conjugation of fatty acids with carnitine facilitates transport of these molecules across mitochondrial membranes where they can undergo subsequent β-oxidation. In particular, two-way ANOVA indicated that O3 caused a significant reduction in 3 of 3 measured long-chain acylcarnitines: oleoylcarnitine, stearoylcarnitine, and palmitoylcarnitine (Fig 5). The magnitude of the effect of O3 on these acylcarnitines was generally greater in the db/db than the WT mice: all 3 measured long-chain acylcarnitines were significantly reduced in obese mice after O3. However, a significant reduction in stearoylcarnitine was also reduced after O3 in lean mice (Fig 5A, 5B and 5C). Elevations in plasma acylcarnitines are typically observed when specific enzyme deficiencies prevent their metabolism via β-oxidation . Consequently, reductions in oleoylcarnitine, stearoylcarnitine, and palmitoylcarnitine observed after O3 exposure likely reflect increased β-oxidation in lung cells.
Metabolism regulating hormones: Since corticosteroids promote β-oxidation  and also attenuate BCAA catabolism , similar to the result obtained in obese O3 exposed mice (Figs (Figs44 and and5),5), we hypothesized that greater O3-induced increases in corticosterone in db/db than WT mice might account for the different effects of O3 on β-oxidation (Fig 5) and BCAA metabolism (Fig 4) observed in db/db versus WT mice. Corticosterone was among the biochemicals identified in our metabolomic analysis and indeed two-way ANOVA did indicate greater lung corticosterone in O3- than air-exposed mice (Fig 6A and S1 Table), presumably as a result of increases in serum corticosterone. Notably, the effect of O3 only reached significance in the db/db mice.
Differences in insulin (Fig 6B) could also account for obesity-related differences in the effects of O3 on lung BCAA metabolism (Fig 4). Db/db mice are markedly resistant to insulin . In human subjects, the development of insulin resistance is associated with increases in circulating BCAAs and reductions in their catabolism [19,41]. Thus, db/db mice should be limited in their ability to catabolize BCAAs, as observed (Fig 4). Consequently, we also measured serum insulin. Serum insulin was higher in db/db than WT, as expected, but there was no effect of O3 exposure on serum insulin in either group of mice (Fig 6B).
O3 causes oxidative stress within the lungs, and an associated induction of anti-oxidant enzymes and systems . There is also evidence of increased airway oxidative stress in obese asthmatics . Consequently, we examined markers of oxidative stress among the lung metabolites identified. The tripeptide, glutathione (GSH), is a key component of the systems that maintain cell redox status and exposure to O3 caused changes in glutathione metabolism in lean and obese mice (Fig 7A and 7B). Compared to air, O3 caused a significant increase in GSH in lung tissue from both lean and obese mice (Fig 7A). GSH is produced by ligation of glutamate and cysteine by the catalytic subunit of glutamate—cysteine ligase (Gclc)(the rate limiting step) followed by addition of glycine by the enzyme glutathione synthetase (Gss). Survey of a microarray analysis we performed assessing lung gene expression in db/db mice exposed to air or O3 indicated no change in Gss mRNA expression in db/db mice after O3 (GSE81800). To determine whether increases in GSH were consequent to increased expression of Gclc, a redox-regulated gene , we measured pulmonary Gclc mRNA abundance by qRT-PCR. O3 did increase Gclc expression, but the effect was only observed in obese mice (Fig 7C). As the lung is one of the organs with the highest uptake of circulating GSH , it is possible that the increase in GSH after O3 (Fig 7A), especially in lean mice, is derived from non-pulmonary production of GSH rather than an increase in lung-derived GSH.
Given the observed increases in GSH with O3 (Fig 7A), we also examined O3-induced changes in pulmonary expression of glutathione-S-transferases (Gst), to determine whether there were likely to be obesity-related differences in the glutathionylation of targets such as lipid peroxides that are generated by O3 exposure. Using our microarray data, we identified two Gst genes that were highly expressed in the lungs and also significantly changed by O3: Gsta1 and Gsta2. RT-PCR confirmed increased expression of Gsta1 after O3, but there was no difference in the impact of O3 on Gsta1 in lean versus obese mice (Fig 7D).
Under conditions of oxidative stress, GSH is converted to oxidized GSH (GSSG). Although significant O3-induced increases in GSSG were not observed in either lean or obese mice, levels of GSSG were significantly higher in obese mice versus lean mice exposed to O3 (Fig 7B).
Others have proposed that airway oxidative stress in obese asthmatics may be the result of NOS uncoupling, which results in the production of superoxide anion instead of nitric oxide . In particular, NOS uncoupling occurs when L-arginine is reduced or when asymmetric dimethyl arginine (ADMA) is increased. However, we did not observe any significant effect of either obesity or O3 on lung arginine or ADMA levels (S1 Table), though we cannot rule out the possibility that there were changes in these moieties in the blood.
There were no changes in other measured biochemical markers of oxidative stress (e.g. methionine sulfoxide, dimethylarginine (SDMA + ADMA), cysteine-glutathione disulfide, and 13-HODE + 9-HODE) with either O3 or obesity (see S1 Table). However, the antioxidants, ascorbate (vitamin C) (Fig 8A) and alpha-tocopherol (vitamin E) (Fig 8B) were changed. In particular, two-way ANOVA indicated an effect of obesity on both ascorbate and alpha-tocopherol: in obese mice, lung ascorbate was increased and alpha-tocopherol was decreased regardless of exposure status.
There are differences in the metabolomic profile of tissues harvested from germ free versus conventionally housed mice and from antibiotic-treated versus control mice [51–54]. These data indicate that bacteria are required for the generation of certain metabolites present in mammalian tissues. Such metabolites are termed bacterial-mammalian co-metabolites. We have observed reductions in the pulmonary response to O3 in mice after administration of antibiotics, suggesting a role for the microbiome in responses to O3 . There are also differences in the gut microbiomes of obese and lean mice (reviewed in ). Consequently, we examined bacterial-mammalian co-metabolites in our data set. Twelve such metabolites [51,57–60] were identified (Table 5). Note that propionylcarnitine and butyrlcarnitine are the carnitine derivatives of propionate and butyrate, short chain fatty acids that are products of bacterial metabolism of dietary fiber. Remarkably, every one of these metabolites was significantly altered by obesity, by O3, or the combination of obesity and O3 (Table 5). These data support the hypothesis that some of the observed differences in the lung metabolomes of obese and lean mice and their changes upon exposure to O3 may derive from differences in their microbiomes.
Both obesity and O3 had substantial effects on the lung metabolome (Tables (Tables11–3). There were also differences in the impact of O3 in obese and lean mice. For example, O3 had differential effects on lipid and BCAA metabolism (Figs (Figs44 and and55 and Tables Tables22–4), and on hormones that impact these metabolic pathways (Fig 6). There were also differential effects of O3 on glutathione metabolism and other markers of oxidative stress in obese versus lean mice (Figs (Figs77 and and8).8). Finally, both obesity and O3 had substantial effects on microbial-mammalian co-metabolites (Table 5), suggesting that obesity-related changes in the gut microbiome may be impacting responses to O3.
Before discussing these data, it is important to note several caveats related to the results presented here. First, we examined metabolites in the lung. We did so because we were interested in the possibility that the effects of certain metabolites might be altering lung responses to O3 in obesity. Thus, it was important to know whether these metabolites changed within the lung. However, there are marked systemic effects of both obesity [19–21] and O3 [23,24] on metabolism, and we cannot necessarily conclude that observed changes in various lung metabolites were the result of altered metabolism within the lung. Blood was flushed from the lungs prior to tissue harvest, so the observed changes do not simply reflect obesity-related differences in the blood within the lungs. However, moieties in the blood can diffuse into the extracellular fluid (ECF) in the lung, and this process may be enhanced following O3 exposure, which increases the permeability of the alveolar/capillary barrier . Consequently, the observed increases in total lung carbohydrate and lipid metabolites in db/db mice (Figs (Figs22 and and3)3) may be the result of increases in these biochemicals in the lung ECF stemming from systemic alterations in metabolism that affect these metabolites.
Second, we studied only female mice. We did so because increases in asthma prevalence with BMI are stronger in women than men , and because obesity-related effects on the response to O3 are greater in female than male human subjects . However, sex differences in the serum and urine metabolomic profiles of male and female obese mice have been reported . Thus, it is possible that obesity would also have different effects on the metabolomic response to O3 in male and female mice. Indeed, others have reported that 17β-estradiol increases the production of GSSG in cultured alveolar epithelial cells exposed to O3, indicating effects of sex hormones on oxidative stress .
Third, our study was primarily descriptive in nature. It was performed to identify metabolites that might be contributing to the innate AHR or the augmented O3-induced AHR observed in obese mice. Nevertheless, there may be important mechanistic implications of our results: many of the metabolites we identified as being altered by obesity and/or O3 have effects that may contribute to AHR. For example, several cholesterol metabolites, including 7-α- and 7- β-hydroxycholesterol were increased in lungs of obese mice (S1 Table). These moieties are precursors for the production of 7-α 27-hydroxycholesterol and 7-β-27-hydroxycholesterol, ligands for a transcription factor RORγt that controls the production of innate lymphoid cells type 3 (ILC3) . IL-17A+ ILC3s are increased in lungs of obese versus lean mice and contribute to the innate AHR of these mice . Elevations in fatty acids in the lungs of db/db mice (Fig 3) would also be expected to increase the ability of ILC2s to produce type 2 cytokines , and we have reported that IL-33 causes ILC2 activation and type 2 cytokine release are important for obesity-related increases in the response to O3[16,17]. Additionally, the ketone, BHBA, was higher in the lungs of obese versus lean mice (Fig 3). Milner et al  reported that lung BHBA correlated positively with the number of T regulatory cells (Tregs) in the lungs during influenza virus infection in lean mice but inversely in obese mice. Whether Tregs impact pulmonary responses to acute O3 has not yet been established. Finally receptors for lactate (HCA1/GPR81), for many fatty acids (GPR40, GPR41, GPR43, GPR84, and GPR120), for BHBA (HCA2/GPR109A), and for other lipid moieties [67,68] exist. Consequently, the observation that these moieties are increased not only in the blood , but also in the lung (Figs (Figs22 and and3),3), suggests that obesity-related changes in these biochemicals could impact airway responsiveness via direct activation of these receptors. Saturated fatty acids like lauric acid and palmitic acid, which were increased in lungs of obese mice (Fig 3), also have the capacity to stimulate the innate immune system by activating pattern recognition molecules and the NLRP3 inflammasome.
O3 caused substantial reductions in choline-containing lysophospholipids, especially in lean mice (Tables (Tables22–4). Changes in phospholipids and/or lysolipids are also observed in the lungs after PM2.5 exposure , and after allergen sensitization and challenge  suggesting that these changes may represent a common response to lung injury and/or inflammation. Reductions in lysolipids after O3 in WT mice could be the result of reduced production from membrane phospholipids resulting from decreases in phospholipase activity. However, examination of published microarray data from C57BL/6 mice exposed in the same manner as we did (GSE38014) indicates that pulmonary mRNA abundance of most phospholipases is either increased or unchanged after O3. Choline containing phosphoplipids make up the majority of the phospholipids in surfactant  and O3 causes oxidation of surfactant phospholipids with consequent loss of their surface active properties . Hence, it is also possible that reductions in lysolipids after O3 reflect their increased incorporation into surfactant phospholipids in order to replete these lipids after O3-induced degradation. If so, greater reductions in lysolipids in lean than obese mice after O3 (Tables (Tables33 and and4),4), would be expected to result in less loss of surfactant function in lean than obese mice. Consistent with this prediction, we have previously reported changes in the pressure-volume curve of the lung consistent with loss of surfactant function in obese but not lean mice after O3 exposure . LysoPC acyltransferases (LPCATs), enzymes involved in conversion of choline containing lyso lipids into phospholipids, are increased in a mouse model of sepsis in conjunction with reductions in lysophospholipids . Similarly, LPCAT3 is increased after O3 exposure in WT mice (GSE38014) but is unchanged in db/db mice after O3 exposure, and might account for the reductions in lysolipids observed in the WT mice.
We also observed differential effects of O3 on the substrates used for energy production in the lungs of db/db and WT mice (Figs (Figs44 and and5).5). In particular, O3 caused reductions in most BCAA metabolites in lean mice but had no effect in obese mice (Fig 4), suggesting increased reliance upon BCAA catabolism for energy in lean but not obese mice exposed to O3. Instead, the obese mice demonstrated increased reliance upon β-oxidation for energy after O3 exposure: long chain acylcarnitines were reduced after O3 exposure, particularly in obese mice, even though these mice had elevations in the lung fatty acids from which these acylcarnitines are derived (Fig 5). Changes in lung β-oxidation also occur following exposure to another inhaled pollutant, acrolein . In particular, mice that are resistant to the effects of acrolein have evidence of increased β-oxidation, whereas sensitive mice have impaired β-oxidation.
As discussed above, greater O3-induced increases in corticosterone in obese than lean mice (Fig 6A) may account for the different effects of O3 on β-oxidation (Fig 5) and BCAA metabolism (Fig 4). Consistent with this hypothesis, O3 exposure also increases serum corticosterone in rats, many of the metabolomic changes induced by O3 in rats are attenuated in adrenalectomized rats [23,25]. Differences in insulin (Fig 6B) could also account for obesity-related differences in the effects of O3 on BCAA metabolism (Fig 4).
Our results support the hypothesis that lung oxidative stress was greater in obese than lean mice, especially after O3 exposure (Figs (Figs77 and and8).8). First, the redox-regulated gene, Gclc, increased to a greater extent in obese than lean mice after O3 exposure. Second, lung GSSG was increased in obese versus lean mice after O3 exposure. Reductions in the antioxidant, alpha tocopherol, in lungs of obese versus lean mice may have contributed to the differences in oxidative stress. These observations are consistent with previous observations in lungs of obese human subjects .
To our knowledge, this is the first report of the impact of acute O3 exposure on the lung metabolome. However, as discussed above, others have reported the effect of O3 on the serum metabolome of lean rats . One of the key observations in that study was that acute O3 exposure causes lipolysis within adipose tissue leading to increases in circulating fatty acids. In contrast, we did not observe substantial changes in lung fatty acids in either obese or lean mice after O3 exposure (S1 Table), though increases in some monoglycerides were observed in lean mice. Miller et al  also reported increases in BCAAs and their metabolites in lean rats exposed to O3, whereas we saw reductions (Fig 4). It is possible that the differences lie in the tissue examined—lung in our study versus serum in that of Miller et al. However, another key difference between the two studies is the time point at which these changes were noted (24 hours post O3 in our study versus immediately post O3 in the study of Miller et al). In this context, it is important to note that in the study of Miller et al  many of the metabolomic changes noted had substantially resolved by 18 hours after cessation of O3 exposure.
Of the 12 bacterial-mammalian co-metabolites identified in lungs of the mice in this study, each one was affected by obesity, by O3 exposure, or by the combination of these two factors (Table 5), suggesting a possible role for the microbiome in obesity-related differences in the response to O3. That obesity might impact metabolites of bacterial origin is not unexpected. Changes in the gut microbiome are observed both in genetically obese mice and in mice with diet-induced obesity, and obesity also impacts the human gut microbiome (see  for review). Gut bacteria-derived metabolites can enter the circulation and diffuse into the lungs. That O3 also affected these metabolites is somewhat more surprising. However, generation of many of the bacterial-mammalian co-metabolites identified (Table 5) requires a metabolic step that occurs in the liver. In rodents, acute O3 exposure has substantial effects on gene expression within the liver  and could thus impact the generation of these metabolites. Hepatic steatosis is common in obesity  and could alter the effect of O3 on the liver.
Whether or not these bacterial-mammalian co-metabolites contribute to the functional and inflammatory changes observed in the lungs after O3 exposure and/or the effect of obesity on these responses to O3 [16,17] remains to be established. However, it is increasingly appreciated that other bacterially derived metabolites contribute to pathological processes not only within the gut but also within the heart, the immune system, and the nervous system [18,74,75]. Indeed, the microbiome has already been shown to play a role in pulmonary responses to allergen . Importantly, a preliminary report from our lab indicated that O3-induced airway hyperresponsiveness was reduced in mice treated with a cocktail of antibiotics, indicating a role for the microbiome in responses to O3 .
The metabolomic profile of the lung was fundamentally altered in obesity and with O3 exposure. Obesity caused changes in carbohydrates and lipids in the lungs. O3 caused differential effects on lung lysolipids and also induced an increased reliance upon BCAA for energy production in lungs of lean mice and an increased reliance upon fatty acids for energy in obese mice, possibly as a result of greater O3-induced increases in corticosterone in the obese mice. Together, these metabolomic changes may have the capacity to promote the asthma-like phenotype observed in obese mice. We have previously reported that both IL-33  and TNF  play a role in the effects of obesity on O3-induced AHR. Consequently, it is conceivable that these cytokines contribute to the metabolomic changes observed here. Finally, the marked effects of both obesity and O3 on bacterial mammalian co-metabolites also suggest a role for the microbiome in the effects of obesity on the lung.
The authors would like to thank Lorena Pantano and John Hutchinson of the Harvard Chan Bioinformatics Core, Harvard. T.H. Chan School of Public Health, for assistance with the PCA and pathway enrichment analyses.
This study was supported by National Institutes of Health grants ES013307, F32ES02256, HL007118, ES024032, and ES000002. Metabolon provided support in the form of salaries for authors (LB, PRG and EK), but did not have any additional role in the study design, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.
All relevant data are within the paper and its supporting information files.