Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Feb 3.
Published in final edited form as: Environ Int. 2020 Dec 21;147:106323. doi: 10.1016/j.envint.2020.106323

Metabolome-wide association study of flavorant vanillin exposure in bronchial epithelial cells reveals disease-related perturbations in metabolism

Matthew Ryan Smith 1,*, Zachery R Jarrell 1,*, Michael Orr 1, Ken H Liu 1, Young-Mi Go 1,†, Dean P Jones 1,†
PMCID: PMC7856097  NIHMSID: NIHMS1653511  PMID: 33360165

Abstract

Electronic cigarettes (e-cig) are an increasingly popular alternative to traditional smoking but have been in use for too short of a period of time to fully understand health risks. Furthermore, associated health risks are difficult to evaluate because of a large range of flavoring agents and their combinations for use with e-cig. Many flavoring agents are generally regarded as safe but have limited studies for effects on lung. Vanillin, an aromatic aldehyde, is one of the most commonly used flavoring agents in e-cig. Vanillin is electrophilic and can be redox active, with chemical properties expected to interact within biologic systems. Because accumulating lung metabolomics studies have identified metabolic disruptions associated with idiopathic pulmonary fibrosis, asthma and acute respiratory distress syndrome, we used human bronchial epithelial cells (BEAS-2B) with high-resolution metabolomics analysis to determine whether these disease-associated pathways are impacted by vanillin over the range used in e-cig. A metabolome-wide association study showed that vanillin perturbed specific energy, amino acid, antioxidant and sphingolipid pathways previously associated with human disease. Analysis of a small publicly available human dataset showed associations with several of the same pathways. Because vanillin is a common and high-abundance flavorant in e-cig, these results show that vanillin has potential to be mechanistically important in lung diseases and warrants in vivo toxicity testing in the context of e-cig use.

Keywords: electronic cigarettes, flavoring agents, high-resolution metabolomics, metabolic disruption, oxidative stress

Graphical Abstract

graphic file with name nihms-1653511-f0001.jpg

1. Introduction

Electronic nicotine delivery system (ENDS) usage has been rising rapidly over the last decade while traditional smoking habits have been decreasing (Jankowski et al. 2019). A critical health concern exists in teens and young adults who use ENDS with flavorings, e.g., inflammation of the lung as well as other pathologies (Chun et al. 2017; Lim and Kim 2014; Wu et al. 2014). Until recently, thousands of flavorings were available for ENDS users; however, recent government regulations have severely limited the flavoring chemicals that can be commercially sold (Products 2020). While the harmful effects of nicotine, one of the major ENDS components, are well-documented, the effects of the flavoring agents in lung are not well understood. In human studies, understanding specific contributions are confounded by the large number of exposures in the human exposome (Vermeulen et al. 2020).

Flavoring agents compose between 1–4% of e-liquid composition, and one of the most frequent flavoring agents used in commercial ENDS is vanillin, which is found in approximately 56% of all e-liquids tested, (Behar et al. 2018a; Tierney et al. 2016). Additionally, recent study by Behar et al showed that vanillin, at high concentration (203 mM), exhibits cytotoxic effects on lung cells (Behar et al. 2018b). Vanillin is a chemically synthesized version of vanilla, and it is commonly used as a flavoring agent to imitate the flavor of vanilla extract in many products such as food, beverages, and electronic cigarette (e-cig) cartridges (Hua et al. 2019). Direct biotransformation products of vanillin metabolism have been well characterized, and studies show that glucuronidation, a phase II conjugation reaction, is the predominant elimination route for vanillin from the body (Yu et al. 2013). Vanillin exerts many pharmacological effects, including chemoprevention and increased antioxidant activities (Bezerra et al. 2016; Lirdprapamongkol et al. 2009; Panoutsopoulos and Beedham 2005); limited information is available, however, on the effects of vanillin and its metabolites on lung metabolism and potential for contribution to disease mechanisms.

In recent years, we and others have advanced the sensitivity and coverage of high-throughput metabolic profiling (Go et al. 2019; Go et al. 2015b; Liu et al. 2020; Liu et al. 2016; Smith et al. 2019a; Smith et al. 2019b; Smith et al. 2019c; Uppal et al. 2016; Walker et al. 2016a; Walker et al. 2016b). Importantly, application of metabolomics to study of human lung diseases are beginning to clarify the metabolic associations of disease. For instance, changes in serum sphingolipid, energy metabolism (citric acid cycle, hexose), and amino acid metabolism (arginine, glutamate, aspartate) are associated with idiopathic pulmonary fibrosis (Zhao et al. 2017); sphingolipid, energy (citric acid cycle, hexose, pentose, butanoate), amino acid (tryptophan, branched chain amino acids, arginine), and micronutrient (pyridoxine, biopterin) metabolism are associated with asthma (Jeong et al. 2018); and in bronchoalveolar lavage fluid, lipid (phospholipid, arachidonate, linoleate), energy (citric acid cycle, hexose, fatty acid), amino acid (tyrosine, methionine, glutamate, branched chain amino acids), and pyrimidine metabolism are associated with acute respiratory distress syndrome (Evans et al. 2014). Importantly, metabolic reprogramming is increasingly recognized to predispose lung to disease, even without overt cell death (Hu et al. 2020). Little information is available, however, concerning metabolic changes caused by ENDS components.

Untargeted high-resolution metabolomics (HRM) with pathway enrichment analysis provides an approach to understand potential contributing factors to disease by testing for effects of exogenous xenobiotic chemicals, such as vanillin, on metabolism in model systems. Understanding effects of specific exposures is especially important for lung disease due to the complexity of human lung exposures. In the present study, we sought to determine whether vanillin, a common flavoring chemical used in ENDS, could be mechanistically linked to lung disease through impact on metabolic pathways previously associated with human lung disease. We used BEAS-2B as a model system for analysis of effects occurring from vanillin in e-cig vapors. BEAS-2B is a human bronchial epithelial cell line similar to airway lung epithelial cells, which are exposed to e-cig vapors. To avoid complications due to cell death, we used vanillin concentrations which do not cause overt cell toxicity, within a range expected from e-cig use, and also at lower concentrations, as occur from second-hand e-cig exposure.

2. Materials and Methods

2.1. Chemicals

Acetonitrile (HPLC grade), formic acid (HPLC grade), propylene glycol (PG), vegetable glycerin (VG), vanillin (HPLC grade), and water (HPLC grade) were from Sigma–Aldrich (St. Louis, MO). Vanillin (Van) was prepared as a 100 mM stock solution in ethanol and added to a mixture of PG and VG (3:7) commonly used as electronic liquid (e-liq) base to provide final concentrations of 50, 250 and 1000 μM.

2.2. Cell Culture, Vanillin treatment and Metabolite Extraction

BEAS-2B human bronchial epithelial cells obtained from ATCC (Manassas, VA) were cultured in a humidified incubator (5% CO2, 37°C) in complete media [10% fetal bovine serum (FBS) in RPMI 1640 (RPMI)] with penicillin-streptomycin until fully confluent. Cells were treated with various concentrations of Van with PG:VG or PG:VG alone for 18 h in low serum containing medium (0.5% FBS in RPMI). Culture media was removed, and cells were washed three times with ice-cold phosphate buffered saline (PBS). To extract metabolites, cells were lysed and extracted by addition of 300 μL of lysis solution [2:1, acetonitrile (ACN): water] containing a mixture of isotopic standards (Go et al. 2015a). Samples were then centrifuged at 14,000 × rpm for 15 min to remove protein, and 100 μL of the supernatants were transferred to autosampler vials for untargeted high-resolution metabolomics (HRM).

2.3. High Resolution Metabolomics (HRM)

Samples were randomized and maintained at 4°C in an autosampler prior to analysis. All samples were measured in triplicate using 10 μL injections on two platforms, hydrophilic interaction liquid chromatography (HILIC; ThermoFisher Scientific, Accucore, 50×2.1mm, 2.6μm) with positive electrospray ionization (ESI) and reversed phase liquid chromatography (RP; C18 column; Higgins Analytical, 50×2.1mm, 2.6 μm) with negative ESI, operated in a dual chromatography mode as previously described (Liu et al. 2016). Data were collected on a mobile phase gradient over a 5-minute period on a Thermo Orbitrap Fusion mass spectrometer (Thermo Fisher, San Diego, CA) set to collect data from mass-to-charge ratio (m/z) of 85 to 1,275 at a resolution of 120,000. Analyte separation for HILIC was performed with a Waters XBridge BEH Amide XP HILIC column (2.1 mm × 50 mm, 2.6 μm particle size) and gradient elution with mobile phases A: LCMS grade water, B: LCMS grade acetonitrile, C: 2% formic acid. The initial 1.5 min period consisted of 22.5% A, 75% B, and 2.5% C, followed by a linear increase to 75% A, 22.5% B, and 2.5% C at 4 min and a final hold of 1 min. C18 chromatography was performed on an end-capped C18 column (Higgins Targa C18 2.1 mm × 50 mm, 3 μm particle size) with mobile phases A: water, B: acetonitrile, C: 10 mM ammonium acetate. The initial 1 min period consisted of 60% A, 35% B, and 5% C followed by a linear increase to 0% A, 95% B, and 5% C at 3 min and held for the remaining 2 min. For both methods, the mobile phase flow rate was 0.35 mL/min for the first min, and increased to 0.4 mL/min for the final 4 min. Tune parameters for sheath gas were 45 for ESI+ and 30 for ESI−. Auxiliary gas was set at 25 for ESI+ and 5 for ESI−. Spray voltage was set at 3.5 kV for ESI+ and −3.0 kV for ESI−.

2.4. Data Preprocessing and Metabolic Feature Selection

Mass spectral files in .raw format were converted to .cdf files using XCalibur file converter software (Thermo Fisher, Waltham, MA) and were extracted using apLCMS (Yu et al. 2009) and xMSanalyzer (Uppal et al. 2013) to generate feature tables which contain mass spectral features defined by mass-to-charge ratio (m/z), retention time, and ion abundance. For feature selection, triplicate technical replicates were median summarized, and m/z features with at least 80% non-missing values in any of the groups and more than 10% non-missing values across all samples were retained (n=6 for all groups excluding the group treated with 50 μM vanillin, for which n=5 due to contamination of one sample). Data were then log2 transformed and quantile normalized after filtering. Selection of differentially expressed m/z features (hereafter termed metabolites) was performed based on one-way ANOVA and linear regression using the limma() and lmreg() packages in R respectively. Benjamini-Hochberg false discovery method (Benjamini and Hochberg 1995) was used for multiple hypothesis testing correction at FDR<0.2 threshold. The hclust() function in R was used to determine the pattern of selected metabolites and similarity of patterns in samples, visualized in an unsupervised two-way hierarchal clustering analysis (HCA) plot. Principal component analysis (PCA) was performed using the pca() function implemented in R package pcaMethods. Post hoc comparisons of ANOVA results were performed using Tukey’s honest significant difference in R.

2.5. Pathway Enrichment Analysis

Pathway enrichment analysis was performed using mummichog1.10 (Li et al. 2013) using the significantly associated metabolic features using the filtering criteria previously described above. Mummichog uses a probability-based approach to predict pathway matches without metabolite identification. Filtering thresholds for pathway annotation included 5 ppm m/z tolerance and 1000 iterations of permutation testing. Included pathways had a minimum of 4 annotated metabolites to be considered a valid match. In previous applications to more than 100 research papers, only one pathway identification error was found when apply this minimum criterion (DP Jones, unpublished results). In the present study, pathways which are reported had at least one metabolite with confirmed identification.

2.6. Metabolite annotation and identification

Metabolites were annotated using xMSannotator (Uppal et al. 2017), and all confidence scores provided by xMSannotator are derived from a multistage clustering algorithm. Identities of selected metabolites (Liu et al. 2020) were confirmed by co-elution relative to authentic standards and ion dissociation mass spectrometry (Level 1 identification by criteria of Schymanski et al. (Schymanski et al. 2014)). Annotations with high or medium confidence in xMSannotator (≥ 4) had M−H/M+H adducts detected in the negative/positive mode respectively. Annotations were made using KEGG, (Kyoto Encyclopedia of Genes and Genomes); HMDB (Human Metabolome Database) (Wishart et al. 2007); T3DB (Toxin and Toxin Target Database) (Wishart et al. 2015), and Lipid Maps (Fahy et al. 2007) databases at 5 ppm tolerance. Commercial sources of authentic standards for vanillin metabolites were not found so these were generated enzymatically using pooled human lung S9 fractions (mixed gender, H0610.PS9), NADPH regenerating system (K5000–10) from Sekisui Xenotech (Kansas City, KS) and 3 μl/ml addition of phase 2 xenobiotic metabolism co-substrates (10 mM UDPGA, 2 mM GSH, 2 mg/ml PAPS, 0.1 mM acetyl-CoA). Metabolites showed time-dependent formation from vanillin, requirement for respective co-substrates and accurate mass matches to predicted products. Vanillin metabolites in BEAS-2B cell extracts and human plasma were identified by coelution and accurate mass match to these confirmed vanillin metabolites.

2.7. xMWAS

Vanillin-related metabolites were integrated with the remaining HRM data using xMWAS (Uppal et al. 2018) based on the partial least-squares (PLS) regression method for data integration. Community detection was obtained with xMWAS through the multilevel community detection algorithm which identifies groups of nodes that are heavily connected with other nodes in the same community but have sparse connections with the rest of the network. The input for xMWAS included the vanillin-related annotations (23 samples × 50 putative vanillin metabolites) and the metabolome (23 samples × 10,144 metabolic features). The threshold criteria were set to |r| > 0.7 and p < 0.05 as determined by Student’s t-test.

2.8. Bubble Plots

Bubble plots of associated metabolic pathways were generated using corrplot() in R. Using the metabolic features selected by the parameters from xMWAS, mummichog v1.0.10 was used to identify relevant pathways associated with top vanillin-related metabolites independently, and significant pathways were selected using the criteria described above within the pathway enrichment methods. Both the size as well as the color of the bubble represent the pathway significance level based on the −Log10P value.

2.9. Cell Viability Assay

BEAS-2B cells were plated at a density of 5.5 × 104 per well in a 96 well plate. After confluency, cells were exposed to PG:VG mixture alone, ethanol vehicle, vanillin alone (50–1000 μM) or vanillin with PG:VG (50–1000 μM) for 18 h, or hydrogen peroxide as a positive control for cell death (10 mM) for 1 h. Viability was then assessed using the Alamar Blue reagent. The 96-well plate was incubated for 1 h at 37 °C and absorbance was measured (Spectramax M2) at 570 nm, with 600 nm used as a reference wavelength according manufacturer’s specifications. The data reported included 8 technical replicates, presented as mean ± SEM. Statistical significance was determined using a Student’s t-test, with p<0.05.

2.10. Human metabolomics and pathway analysis

This data is available at the NIH Common Fund’s National Metabolomics Data Repository (NMDR) website, the Metabolomics Workbench, https://www.metabolomicsworkbench.org, where it has been assigned Project ID PR000722. The data can be accessed directly via it’s Project DOI: 10.21228/M8BT3D. The MWAS was performed to determine metabolic features varying in association with vanillin at p <0.05. Filtering thresholds for pathway annotation included 5 ppm m/z tolerance and 1000 iterations of permutation testing. Included enriched pathways had a minimum of 4 annotated metabolites at to be considered a valid match at P< 0.05.

2.11. Data Sharing

Datasets corresponding to figures and metabolomics datasets used for xMWAS analysis, and full xMWAS network membership, are available upon reasonable request.

3. Results

3.1. Vanillin exposure perturbed metabolism in the BEAS-2B lung epithelial cells

To examine metabolic responses of a bronchial epithelial cell model to the flavorant vanillin and determine dose response, studies were conducted using BEAS-2B cells which were exposed to vanillin concentration found in e-cig (1000 μM) and lower concentrations (0, 50, and 250 μM) with e-liq base PG/VG (3:7) for 18 h. To evaluate flavorant effect only, nicotine in e-liq was excluded. Duration of vanillin exposure was chosen based on prior studies showing that vanillin exposure can cause cell death after longer periods of time (Sassano et al. 2018). Consistently, we observed no cytotoxicity at the above condition with vanillin doses examined by cell viability using the Alamar Blue assay while we observed significant cell death by H2O2 treatment used for positive cell death control (88 ± 2% death compared to vehicle control). Results were confirmed with viability assayed by WST-1. These results are consistent with previous studies showing vanillin to have relatively low toxicity, with LD50 in the range of 162 – 387 mM depending on the animal model (NTIS; Taylor et al. 1964). Thus, under the conditions selected for metabolomics analysis (removal of media and 3X wash with saline prior to lysis), effects of vanillin reflect intracellular effects of the 18 h incubation rather than changes due to cell death.

3.1.1. Metabolic associations with vanillin using the HILIC column

After data extraction and quality filtering for HILIC with positive ESI (HILIC+), 10,144 metabolic features remained for statistical analyses. Using linear regression, 1053 features were altered with increased vanillin dose with somewhat more features decreased with increasing vanillin concentration (Fig 1A, FDR<0.2). The heatmap result shows a clear separation with 2 groups between 1mM vanillin and doses lower than 1 mM vanillin. Of 1053 metabolic features, 599 were decreased while 454 were increased in their abundances by 1 mM vanillin compared to lower vanillin doses. We found that some metabolites, such as the amino acids tyrosine and methionine were increased after exposure to 1mM vanillin, while leucine/isoleucine and threonine were decreased. These features (599 positively and 454 negatively associated with increasing vanillin dose) as well as the features determined at p<0.05 (1721) are also shown by Manhattan plot (Fig 1B). Detailed characteristic information on these metabolic features are provided in Supplemental Table 1. To determine metabolic pathways associated with these metabolites, we performed pathway enrichment analysis using mummichog (Li et al. 2013). The result showed that 15 metabolic pathways were associated with vanillin exposure (Fig 1C). Most strikingly, many of these were amino acid metabolism pathways, including glutamate, tyrosine, methionine and cysteine, arginine and proline, aspartate and asparagine, glycine, serine, alanine and threonine, branched chain amino acids (valine, leucine, isoleucine) and lysine. Identity of amino acids in these pathways, along with selected metabolites, have been previously confirmed (Schymanski Level 1) by the analytical methods used (see (Liu et al. 2020)). Consistent with widespread effects on amino acid metabolism, the urea cycle was also significantly affected. Other metabolic pathways included the aminosugars pathway, which has been linked to the maintenance of extracellular matrix and biomechanical stress. Metabolic pathways for antioxidant regulation including methionine and cysteine, glutathione, ascorbate and biopterin were also perturbed with vanillin exposure. In addition, the pentose phosphate pathway which has been linked to oxidative stress (Perl et al. 2011), was also found to be changed with vanillin, consistent with previously observed effects on oxidative stress pathways (Bezerra et al. 2016; Lirdprapamongkol et al. 2009; Panoutsopoulos and Beedham 2005). A complete list of metabolite annotations within each pathway is provided in Supplemental Table 2, and responses of selected metabolites to vanillin concentrations are shown below.

Figure 1.

Figure 1.

Open in a new tab

Metabolome-wide association study (MWAS) of vanillin in BEAS-2B cells. A) Unsupervised hierarchal cluster analysis (HCA) heatmap using the HILIC column coupled with positive ionization showed that 1053 features were associated with vanillin. FDR cutoff indicated by the dashed dark blue line, p<0.05 cutoff indicated by dashed grey line. B) Type I Manhattan plot showed 599 features negatively associated (blue) and 454 features positively associated (red) with vanillin. C) Pathway enrichment analysis showed 15 (out of 119) pathways associated with vanillin. (Filled gray bars indicate significance and the cutoff (p<0.05) is indicated by the dotted line). D) HCA for C18- show 235 features associated with vanillin. FDR cutoff indicated by the dashed dark blue line, p<0.05 cutoff indicated by dashed grey line. E) Type I Manhattan plot shows 156 features negatively (blue) and 79 features positively associated (red) with vanillin. F) Pathway enrichment analysis showed 16 pathways associated with vanillin. (Filled gray bars indicate significance and the cutoff (p<0.05) is indicated by the dotted line). n=6 for all groups except 50 μM vanillin, for which n=5.

3.1.2. Metabolic associations with vanillin using the C18 column

Analysis with C18 chromatography with negative ESI (C18-) complements the HILIC+ analysis to enhance metabolic coverage. After extraction and quality filtering, we used 9,027 metabolic features for statistical analyses. Using linear regression, 235 metabolic features were associated with vanillin (Fig 1D, FDR<0.2, see Supplemental Table 3 for detailed characteristics of metabolites). As with the result of HILIC+ analysis, the heatmap shows a clear separation with two groups between 1 mM vanillin and the lower than 1 mM vanillin concentrations in terms of metabolites decreased with increasing vanillin dose. In addition, of these vanillin-associated metabolites, more metabolites were decreased with increasing vanillin. In these data, we found that some of amino acids including proline as well as the fatty acid stearic acid were increased with vanillin, while some of TCA cycle metabolites and the nucleotides were decreased. These metabolites (79 positively and 156 negatively associated with increasing vanillin dose) and metabolites determined at p<0.05 (1014) are also shown by Manhattan plot (Fig 1E). Subsequent pathway enrichment analysis showed 16 metabolic pathways (Fig 1F), which overlapped with HILIC+ data (designated by *) for several amino acid pathways, including tyrosine, arginine and proline, glutamate, and branched chain amino acid metabolism. Other canoate and the anti-oxidant pathways, ascorbate and aldarate. Ten additional pathways were associated with vanillin (Fig 1F), including the top pathway, pentose and glucuronate interconversions, linked to detoxification. Other pathways included the energy pathways related to hexose metabolism (galactose, starch and sucrose, glycolysis, pyruvate), and mitochondrial function (citric acid cycle), suggestive of energetic disruption and mitochondrial dysfunction. Two lipid pathways associated with vanillin involved glycosphingolipids biosynthesis and breakdown. N-glycan pathway, related to the aminosugar pathway above, has also been linked to extracellular matrix maintenance. Changes were also observed in the pyrimidine pathway, which functions in nucleic acid biosynthesis and degradation and is dependent upon mitochondrial function. See Supplemental Table 4 for a complete list of metabolite annotations within each pathway. Together with the HILIC+ data, the C18- results show widespread dose-dependent effects of vanillin on amino acid, energy, amino sugars, antioxidants and sphingolipid metabolism.

3.2. Alterations in key selected metabolites with vanillin exposure

Dose-response characteristics of selected metabolites with confirmed identification (Level 1 (Schymanski et al. 2014)) from the significant pathway in Fig 1C and 1F) are shown in Figures 2–4. Some of the non-essential (glutamine, proline) and essential (leucine + isoleucine) amino acids were positively associated with increased vanillin dose (Fig 2A–C). A corresponding increase was observed for hydroxyproline (Fig 2D), a degradation product of collagen and procollagen, suggesting increased protein degradation relative to protein synthesis. In contrast, cysteine, a precursor for glutathione and protection against oxidative stress and alkylating agents, was negatively associated with vanillin exposure (Fig 3A). UDP and UDP-glucuronic acid, essential for elimination of vanillin as well as other detoxification reactions, were also negatively associated with vanillin (Fig 3B, 3C). Disruption in phospholipid metabolism was evident from decline in phosphatidylglycerol, phosphatidylinositol and phosphatidylserine (Fig 4A–C) and increase in the free fatty acid, stearic acid, in association with increased vanillin (Fig 4D).

Figure 2.

Figure 2.

Open in a new tab

Alterations in selected amino acids with vanillin in BEAS-2B cells. Abundance of selected amino acids including glutamine (A), proline (B), (iso)leucine (C) and hydroxyproline (D) are shown in response to vanillin exposure. Data is represented as quantile normalized abundance, with the 95% confidence interval indicated in the grey shaded area, and the regression curve indicated by the blue line.

Figure 4.

Figure 4.

Open in a new tab

Alterations in phospholipid metabolites with vanillin exposure. Abundance of selected phospholipid metabolites including phosphatidylglycerol (A), phosphatidylinositol (B), phosphatidylserine (C) and stearate (D) associated with vanillin exposure are shown. Data is represented as quantile normalized abundance, with the 95% confidence interval indicated in the grey shaded area, and the regression curve indicated by the blue line.

Figure 3.

Figure 3

Open in a new tab

Alterations in selected detoxication pathway metabolites with vanillin exposure. Abundance of selected metabolites associated with detoxication pathway including cysteine (A), UDP (B) and UDP-Glucuronate (C) are shown in response to vanillin exposure. Data is represented as quantile normalized abundance, with the 95% confidence interval indicated in the grey shaded area, and the regression curve indicated by the blue line.

3.3. Pathway correlations of vanillin metabolites

To determine metabolic pathways altered by vanillin, we used xMSannotator (Uppal et al. 2017) to select 51 mass spectral features which were annotated as putative vanillin metabolites. Using enzymatic generation from vanillin using pooled human liver S9 fractions with co-substrates, we subsequently detected vanillic acid-4-O-glucuronide, among several others, providing level 2 confidence, while other vanillin metabolites detected provided confidence levels 3, 4, and 5 respectively (see Methods 2.6; Supplemental Table 5). To determine pathway associations with vanillin metabolites, we performed a metabolome-wide association study against the metabolome, with the putative vanillin metabolites using xMWAS (Uppal et al. 2018). The results showed that the four vanillin metabolites were correlated in three metabolic communities (Fig 5A). The largest, Community 1 (orange), was associated with the major elimination product, vanillic acid-4-O-glucuronide (vanillate glucuronide). Subsequent pathway analysis on the metabolites in this community showed 16 pathway associations (Fig 5B, see Supplemental Table 6 for annotated metabolites associated with communities). The extensive overlap with pathways associated with vanillin (Fig 1C and F) suggest that the effects of vanillin are closely related to vanillin metabolism. An additional pathway, purine metabolism, shows that secondary metabolic effects can occur through bioconversion to vanillate glucuronide. Community 2 (blue) was associated with another major elimination product, vanillin-4-sulfate. Pathway analysis of this cluster revealed that all 6 of the pathways associated with vanillin sulfate overlapped with vanillate glucuronide. Lastly, Community 3 (green) was associated with vanillate and vanillin acetate, although vanillin acetate had no pathway associations that met the requisite criteria. Collectively, the results show that targeted analysis of metabolic pathways linked to vanillin metabolites extensively overlapped with the pathways from the untargeted analysis linked to vanillin dose (Fig 1C and 1F).

Figure 5.

Figure 5.

Open in a new tab

Association of vanillin-related metabolites with the metabolome. A) xMWAS network of vanillin treated cells revealed 3 metabolic communities. Community 1 (orange) had 603 metabolic features (squares) which were associated with vanillic acid 4-O-glucuronide (circle), while Community 2 (blue) had 276 metabolic features associated with vanillin 4-sulfate. Community 3 (green) had 125 features which were associated with both vanillate and vanillin acetate. (|r| > 0.7 at p < 0.05). Red lines indicate positive associations, and blue lines indicate negative associations. B) Bubble plot showing metabolic pathways associated with each community using mummichog. The size and color of the bubbles represents the pathway significance level based on −log10P. Metabolic pathway analysis was performed using metabolic features associated with vanillin metabolites at |r|>0.7 and p<0.05 using mummichog.

3.4. Pathway correlations of vanillin in human plasma

To determine whether similar pathways associated with vanillin in humans, we examined a dataset available from the Metabolomics Workbench (see Methods 2.10) for children with second-hand ENDS exposures (n = 78). We performed an MWAS for vanillin using HILIC+ data (881 metabolites at |r| ≥0.2, P<0.05), and the resulting pathway enrichment analysis with mummichog showed many of the same pathways as identified in the analysis with BEAS-2B cells (Fig 6). For instance, amino acid pathways included several of those associated with vanillin dose in the cell studies, i.e., Asp and Asn, Arg and Pro, Val, Leu and Ile, Tyr, Lys and Glu. The urea cycle was also associated with vanillin, as was the butanoate pathway. A complete list of metabolite annotations within each pathway is provided in Supplemental Table 7. Thus, the results show that the dose response studies in the cell model described above recapitulate associations seen in human plasma samples.

Figure 6.

Figure 6.

Open in a new tab

Vanillin-associated metabolic pathways in human plasma. The 881 metabolites correlated with vanillin in plasma (|r| ≥ 0.2, P < 0.05) were used for pathway enrichment analysis using mummichog. Eighteen metabolic pathways were determined to be associated with vanillin. Filled gray bars indicate significance and the cutoff (p<0.05) is indicated by the dotted line. Asterisks indicate overlap with vanillin-associated pathways determined in BEAS-2B cells.

4. Discussion

Electronic cigarettes (e-cig) use is an emerging trend that has quickly supplanted traditional cigarette use as the primary mode of nicotine consumption (Jankowski et al. 2019). While the effects of nicotine are very well characterized in the lung, the effects of the flavorants, commonly utilized in e-cig, are less understood. Many of these flavoring chemicals can form reactive aldehydes upon heating, and these aldehydes are associated with pathological effects in the lung (Gillman et al. 2016). Among over 7000 e-cig flavorants, vanillin is a very commonly used flavoring agent and can be found in over 50% of the formulations of e-cig liquids with concentrations ranging up to 200 mM (Behar et al. 2018a; Behar et al. 2018b) (Hua et al. 2019). However, limited studies are available to determine whether flavorants such as vanillin perturb metabolic homeostasis. In the present study, we found that vanillin had a broad impact on lung metabolism, which encompassed many amino acids, fatty acids, lipids, as well as mitochondrial function (Fig 1C and 1F). In an initial analysis of associations with vanillin in human samples, several of these pathways were also identified (Fig 6). The amino acid proline, as well as it’s hydroxylated form, 4-hydroxyproline were positively associated with vanillin exposure (Fig 2B and 2D). Several studies have suggested that proline metabolism is a biomarker of microenvironmental stress (Phang et al. 2008). The 4-hydroxyproline is synthesized from proline through the enzyme prolyl-hydroxylase, and requires ascorbic acid as a cofactor, and has a critical role in stabilizing collagen and preventing apoptosis (Nemethy and Scheraga 1986). Thus, the increase in hydroxyproline suggests that the rate of protein synthesis was impaired relative to the rate of degradation, perhaps through effects on procollagen stabilization. Such an imbalance could also contribute to increased levels of other amino acids, such as glutamine and branched chain amino acids (Fig 2). Vanillin exposure also was associated with antioxidant metabolism, with affected pathways such as glutathione and methionine (Met) and cysteine (Cys) metabolism (Fig 1). We found that glutathione disulfide, glutamyl-glutamine, as well the precursor Cys were negatively associated with vanillin exposure. Several phospholipids were decreased with vanillin, and the pro-inflammatory fatty acid, stearate, was increased with vanillin exposure.

Many of the metabolic pathways identified to be perturbed by high level vanillin have been previously associated with lung disease, including asthma, idiopathic pulmonary fibrosis, and acute respiratory distress syndrome (ARDS) (Fig 7). This data suggests that there are common mechanisms and metabolic pathways perturbed with vanillin exposure and lung disease. Importantly, future surveillance studies with humans exposed to ENDS will be needed to confirm the current findings from cell and human data. Interestingly, in a recent study, Yang et al. systematically reviewed human data from 99 published studies and discussed association of e-cigarette exposure with oral disease. In this article, they also discussed how the flavorant cinnamaldehyde contained in e-cigarettes has been associated with mouth irritation (Yang et al. 2020), supporting the need for human surveillance studies.

Figure 7.

Figure 7.

Open in a new tab

Vanillin-related metabolic disruptions in BEAS-2B cells include pathways associated with human lung diseases. Pathways found to be altered in association with asthma (Jeong et al. 2018), idiopathic pulmonary fibrosis (Zhao et al. 2017)and acute respiratory distress syndrome (Evans et al. 2014) were selected which overlap with pathways altered in association with vanillin dose (see Fig 1). Similar and overlapping pathways were grouped together to simplify comparisons. For instance, common pathways impacting energy metabolism, glycolysis, pyruvate metabolism and hexose metabolism, were grouped together. Some pathways, such as glucuronate metabolism, are not well identified in this simplification, and other metabolomics resources should be consulted for details.

Earlier research has indicated that lower levels of vanillin are beneficial for antioxidant responses. However, the present results raise concern that higher concentrations of vanillin, commonly used commercially in the use of electronic cigarettes cartridges, could also drive the lung metabolic microenvironment to a more pathogenic state. Specifically, vanillin caused changes in hexose, pentose and citric acid cycle pathways functioning in energy metabolism of all of these disease processes (Evans et al. 2014; Jeong et al. 2018; Zhao et al. 2017) and could, in principle, impact onset or progression. Changes in branched chain amino acid and pyrimidine metabolism are common indicators of mitochondrial dysfunction and previously associated asthma and ARDS (Fig 7) (Evans et al. 2014; Jeong et al. 2018). Biopterin, essential for blood flow regulation by nitric oxide synthase and also for tyrosine metabolism, is altered in asthma (Jeong et al. 2018) and ARDS (Evans et al. 2014). Arginine is also needed for blood flow regulation, and both arginine and tyrosine are altered in inflammation (Ckless et al. 2008; Reynaert et al. 2005). Sphingolipids are commonly altered in lung disease (Kowal et al. 2019; Mizumura et al. 2018) and also perturbed by vanillin (Fig 7). Changes in antioxidant metabolism, including Met, Cys and ascorbate pathways, are caused by vanillin and also occur in ARDS (Evans et al. 2014) (Fig 7). Finally, changes in aminosugars, which function in N-glycans to provide a structural lattice for cells (Lau et al. 2007), are altered by vanillin (Fig 1C, 1F) and could contribute to both fibrotic disease as well as sensitivity to inflammation and ARDS.

The concentration of vanillin used in e-cig products varies over a very wide range. In the present study, we used a reference concentration of 1 mM as a common expected dose (Behar et al. 2018a), and performed a series of lower concentrations to evaluate metabolic responses associated with vanillin dose. Results for individual metabolites show dose response over the 20-fold concentration range (Fig 2–4), but not all metabolites were significantly altered at the lowest concentration, 50 μM, relative to the vehicle control. Furthermore, the analyses were performed at a time point when no cell death had occurred. Based upon previous research (Sassano et al. 2018), some cell death would be expected at 1 mM vanillin. Because cell death is a natural part of cell turnover, however, cell death per se is not a good disease marker. Similarly, metabolic effects can be compensatory and beneficial. Thus, studies will be needed in model systems with specific disease biomarkers to estimate lowest effect level for vanillin. Similarly, longitudinal studies will be needed to evaluate long-term effects on metabolic reprogramming.

Whether metabolic effects are a direct consequence of vanillin or a result of vanillin metabolism is unclear. Vanillin elimination can undergo both phase I and phase II reactions in vivo, with glucuronidation being the primary route, followed by sulfation (Bingham 2001). Both vanillin, as well as the oxidized product, vanillate, can undergo glucuronidation and sulfation (Coughtrie et al. 1994; Warner et al. 2017; Yu et al. 2013). The pathways associated with vanillin dose in an untargeted MWAS of vanillin were largely the same as those obtained using a targeted association study of vanillin metabolites. For the latter, putative vanillin metabolites were obtained by using xMSannotator to select all metabolites annotated as possible vanillin metabolites. We only verified identities of a subset of metabolites, and future studies of other vanillin related-elimination products are needed. Interestingly, studies have also shown that vanillin can also inhibit glucuronidation (Salleh et al. 2017). This coupled with our pathway results suggest that the levels of vanillin used in ENDS can inhibit antioxidant activity and enhance lung inflammation.

Clustering analysis showed significant associations for vanillate, vanillate glucuronide, vanillin sulfate and vanillin acetate. Use of xMWAS with these four metabolites and all of the untargeted data showed three central communities, two of which almost completely overlapped with those obtained by the untargeted MWAS of vanillin dose. Thus, the concurrence of completely untargeted results based upon statistical selection followed by pathway enrichment analysis (Fig 1) with the annotation-based selection followed by data-driven community detection (Fig 5), provides confidence in the overall conclusion that vanillin metabolism causes major disruptions which could contribute in an adverse way to a range of lung diseases.

5. Conclusion

Vanillin exposure in a human epithelial cell model causes substantial metabolic perturbations which have previously been associated with human lung diseases. These changes occur over a range of vanillin concentration and impact multiple central pathways function in energy metabolism, lung structure and physiology and antioxidant defense. The results are mirrored by an initial study of vanillin in human samples and warrant detailed studies to evaluate contribution of specific e-cig flavorants to risks of acute and chronic lung diseases.

Supplementary Material

Supplementary Table 1
Supplementary Table 2
Supplementary Table 3
Supplementary Table 4
Supplementary Table 6
Supplementary Table 7
Supplemental Table 5

Highlights.

  • Vanillin in e-cig causes widespread metabolic disruption in airway epithelial cells

  • Metabolic disruption by vanillin is associated with lung disease

  • High-resolution metabolomics supports mechanistic toxicology of vanillin in e-cig

Acknowledgement

Dr. Young-Mi Go and Dr. Dean P. Jones share equal senior authorship in this collaborative research. The authors would like to acknowledge Yating Wang for her technical expertise with the mass spectrometer.

Funding: This study was supported by National Institute of Environmental Health Sciences grants R01 ES023485 (DPJ and YMG), R21 ES031824 (DPJ and YMG), P30 ES019776 (DPJ), U2C ES030163 (DPJ) and T32 ES012870 (ZJ), and NIH grant S10 OD018006 (DPJ).

Abbreviations

E-cig

electronic cigarette(s)

ENDS

electronic nicotine delivery system

FDR

false discovery rate

HMDB

Human Metabolome Database

HRM

high-resolution metabolomics

KEGG

Kyoto Encyclopedia of Genes and Genomes

PG

propylene glycol

VG

vegetable glycerin

EtOH

ethyl alcohol

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Declaration of Competing Financial Interests

The authors declare they have no actual or potential competing financial interests.

References

  1. Behar RZ, Luo W, McWhirter KJ, Pankow JF, Talbot P. 2018a. Analytical and toxicological evaluation of flavor chemicals in electronic cigarette refill fluids. Sci Rep. 8(1):8288. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Behar RZ, Wang Y, Talbot P. 2018b. Comparing the cytotoxicity of electronic cigarette fluids, aerosols and solvents. Tob Control. 27(3):325–333. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B (Methodological). 57(1):289–300. [Google Scholar]
  4. Bezerra DP, Soares AK, de Sousa DP. 2016. Overview of the role of vanillin on redox status and cancer development. Oxid Med Cell Longev. 2016:9734816. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bingham E, Cohrssen B 2001. Patty’s toxicology. New York, NY: John Wiley & Sons. [Google Scholar]
  6. Chun LF, Moazed F, Calfee CS, Matthay MA, Gotts JE. 2017. Pulmonary toxicity of e-cigarettes. Am J Physiol Lung Cell Mol Physiol. 313(2):L193–L206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Ckless K, Lampert A, Reiss J, Kasahara D, Poynter ME, Irvin CG, Lundblad LK, Norton R, van der Vliet A, Janssen-Heininger YM. 2008. Inhibition of arginase activity enhances inflammation in mice with allergic airway disease, in association with increases in protein s-nitrosylation and tyrosine nitration. J Immunol. 181(6):4255–4264. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Coughtrie MW, Bamforth KJ, Sharp S, Jones AL, Borthwick EB, Barker EV, Roberts RC, Hume R, Burchell A. 1994. Sulfation of endogenous compounds and xenobiotics--interactions and function in health and disease. Chem Biol Interact. 92(1–3):247–256. [DOI] [PubMed] [Google Scholar]
  9. Evans CR, Karnovsky A, Kovach MA, Standiford TJ, Burant CF, Stringer KA. 2014. Untargeted lc-ms metabolomics of bronchoalveolar lavage fluid differentiates acute respiratory distress syndrome from health. J Proteome Res. 13(2):640–649. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Fahy E, Sud M, Cotter D, Subramaniam S. 2007. Lipid maps online tools for lipid research. Nucleic Acids Res. 35(Web Server issue):W606–612. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Gillman IG, Kistler KA, Stewart EW, Paolantonio AR. 2016. Effect of variable power levels on the yield of total aerosol mass and formation of aldehydes in e-cigarette aerosols. Regul Toxicol Pharmacol. 75:58–65. [DOI] [PubMed] [Google Scholar]
  12. Go YM, Kim CW, Walker DI, Kang DW, Kumar S, Orr M, Uppal K, Quyyumi AA, Jo H, Jones DP. 2015a. Disturbed flow induces systemic changes in metabolites in mouse plasma: A metabolomics study using apoe(−)/(−) mice with partial carotid ligation. AmJPhysiolRegulIntegrCompPhysiol. 308(1):R62–R72. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Go YM, Smith MR, Walker DI, Uppal K, Rohrbeck P, Krahl PL, Hopke PK, Utell MJ, Mallon TM, Jones DP. 2019. Metabolome-wide association study of deployment to balad, iraq or bagram, afghanistan. J Occup Environ Med. 61 Suppl 12:S25–S34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Go YM, Walker DI, Liang Y, Uppal K, Soltow QA, Tran V, Strobel F, Quyyumi AA, Ziegler TR, Pennell KD et al. 2015b. Reference standardization for mass spectrometry and high-resolution metabolomics applications to exposome research. Toxicol Sci. 148(2):531–543. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Hu X, Go YM, Jones DP. 2020. Omics integration for mitochondria systems biology. Antioxid Redox Signal. 32(12):853–872. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Hua M, Omaiye EE, Luo W, McWhirter KJ, Pankow JF, Talbot P. 2019. Identification of cytotoxic flavor chemicals in top-selling electronic cigarette refill fluids. Sci Rep. 9(1):2782. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Jankowski M, Krzystanek M, Zejda JE, Majek P, Lubanski J, Lawson JA, Brozek G. 2019. E-cigarettes are more addictive than traditional cigarettes-a study in highly educated young people. Int J Environ Res Public Health. 16(13). [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Jeong A, Fiorito G, Keski-Rahkonen P, Imboden M, Kiss A, Robinot N, Gmuender H, Vlaanderen J, Vermeulen R, Kyrtopoulos S et al. 2018. Perturbation of metabolic pathways mediates the association of air pollutants with asthma and cardiovascular diseases. Environ Int. 119:334–345. [DOI] [PubMed] [Google Scholar]
  19. Kowal K, Zebrowska E, Chabowski A. 2019. Altered sphingolipid metabolism is associated with asthma phenotype in house dust mite-allergic patients. Allergy Asthma Immunol Res. 11(3):330–342. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lau KS, Partridge EA, Grigorian A, Silvescu CI, Reinhold VN, Demetriou M, Dennis JW. 2007. Complex n-glycan number and degree of branching cooperate to regulate cell proliferation and differentiation. Cell. 129(1):123–134. [DOI] [PubMed] [Google Scholar]
  21. Li S, Park Y, Duraisingham S, Strobel FH, Khan N, Soltow QA, Jones DP, Pulendran B. 2013. Predicting network activity from high throughput metabolomics. PLoS ComputBiol. 9(7):e1003123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Lim HB, Kim SH. 2014. Inhallation of e-cigarette cartridge solution aggravates allergen-induced airway inflammation and hyper-responsiveness in mice. Toxicol Res. 30(1):13–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Lirdprapamongkol K, Kramb JP, Suthiphongchai T, Surarit R, Srisomsap C, Dannhardt G, Svasti J. 2009. Vanillin suppresses metastatic potential of human cancer cells through pi3k inhibition and decreases angiogenesis in vivo. J Agric Food Chem. 57(8):3055–3063. [DOI] [PubMed] [Google Scholar]
  24. Liu KH, Nellis M, Uppal K, Ma C, Tran V, Liang Y, Walker DI, Jones DP. 2020. Reference standardization for quantification and harmonization of large-scale metabolomics. Anal Chem. 92(13):8836–8844. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Liu KW, Walker DI, Uppal K, Tran V, Rohrbeck P, Mallon TM, Jones DP. 2016. High-resolution metabolomics assessment of military personnel: Evaluating analytical strategies for chemical detection. JOccupEnvironMed. 58(8 Suppl 1):S53–S61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Mizumura K, Justice MJ, Schweitzer KS, Krishnan S, Bronova I, Berdyshev EV, Hubbard WC, Pewzner-Jung Y, Futerman AH, Choi AMK et al. 2018. Sphingolipid regulation of lung epithelial cell mitophagy and necroptosis during cigarette smoke exposure. FASEB J. 32(4):1880–1890. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Nemethy G, Scheraga HA. 1986. Stabilization of collagen fibrils by hydroxyproline. Biochemistry. 25(11):3184–3188. [DOI] [PubMed] [Google Scholar]
  28. Ots0536271. Springfield, VA 22161: National Technical Information Service; [accessed].
  29. Panoutsopoulos GI, Beedham C. 2005. Enzymatic oxidation of vanillin, isovanillin and protocatechuic aldehyde with freshly prepared guinea pig liver slices. Cell Physiol Biochem. 15(1–4):89–98. [DOI] [PubMed] [Google Scholar]
  30. Perl A, Hanczko R, Telarico T, Oaks Z, Landas S. 2011. Oxidative stress, inflammation and carcinogenesis are controlled through the pentose phosphate pathway by transaldolase. Trends Mol Med. 17(7):395–403. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Phang JM, Pandhare J, Liu Y. 2008. The metabolism of proline as microenvironmental stress substrate. J Nutr. 138(10):2008S–2015S. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Products CfT. 2020. Enforcement priorities for electronic nicotine delivery system (ends) and other deemed products on the market without premarket authorization. Rockville, MD: Food and Drug Administration; p. 1–52. [Google Scholar]
  33. Reynaert NL, Ckless K, Wouters EF, van der Vliet A, Janssen-Heininger YM. 2005. Nitric oxide and redox signaling in allergic airway inflammation. Antioxid Redox Signal. 7(1–2):129–143. [DOI] [PubMed] [Google Scholar]
  34. Salleh NM, Ismail S, Ibrahim MNM. 2017. The inhibition of hepatic and renal glucuronidation of p-nitrophenol and 4-methylumbelliferone by oil palm empty fruit bunch lignin and its main oxidation compounds. Pharmacogn Mag. 13(Suppl 1):S102–S114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sassano MF, Davis ES, Keating JE, Zorn BT, Kochar TK, Wolfgang MC, Glish GL, Tarran R. 2018. Evaluation of e-liquid toxicity using an open-source high-throughput screening assay. PLoS Biol. 16(3):e2003904. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Schymanski EL, Jeon J, Gulde R, Fenner K, Ruff M, Singer HP, Hollender J. 2014. Identifying small molecules via high resolution mass spectrometry: Communicating confidence. Environ Sci Technol. 48(4):2097–2098. [DOI] [PubMed] [Google Scholar]
  37. Smith MR, Uppal K, Walker DI, Utell MJ, Hopke PK, Mallon TM, Krahl PL, Rohrbeck P, Go YM, Jones DP. 2019a. Environmental chemicals altered in association with deployment for high risk areas. J Occup Environ Med. 61 Suppl 12:S15–S24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Smith MR, Walker DI, Uppal K, Utell MJ, Hopke PK, Mallon TM, Krahl PL, Rohrbeck P, Go YM, Jones DP. 2019b. Benzo[a]pyrene perturbs mitochondrial and amino acid metabolism in lung epithelial cells and has similar correlations with metabolic changes in human serum. J Occup Environ Med. 61 Suppl 12:S73–S81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Smith MR, Woeller CF, Uppal K, Thatcher TH, Walker DI, Hopke PK, Rohrbeck P, Mallon TM, Krahl PL, Utell MJ et al. 2019c. Associations of benzo(ghi)perylene and heptachlorodibenzo-p-dioxin in serum of service personnel deployed to balad, iraq, and bagram, afghanistan correlates with perturbed amino acid metabolism in human lung fibroblasts. J Occup Environ Med. 61 Suppl 12:S35–S44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Taylor JM, Jenner PM, Jones WI. 1964. A comparison of the toxicity of some allyl, propenyl, and propyl compounds in the rat. Toxicol Appl Pharmacol. 6:378–387. [DOI] [PubMed] [Google Scholar]
  41. Tierney PA, Karpinski CD, Brown JE, Luo W, Pankow JF. 2016. Flavour chemicals in electronic cigarette fluids. Tob Control. 25(e1):e10–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Uppal K, Ma C, Go YM, Jones DP, Wren J. 2018. Xmwas: A data-driven integration and differential network analysis tool. Bioinformatics. 34(4):701–702. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Uppal K, Soltow QA, Strobel FH, Pittard WS, Gernert KM, Yu T, Jones DP. 2013. Xmsanalyzer: Automated pipeline for improved feature detection and downstream analysis of large-scale, non-targeted metabolomics data. BMC Bioinformatics. 14:15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Uppal K, Walker DI, Jones DP. 2017. Xmsannotator: An r package for network-based annotation of high-resolution metabolomics data. Anal Chem. 89(2):1063–1067. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Uppal K, Walker DI, Liu K, Li S, Go YM, Jones DP. 2016. Computational metabolomics: A framework for the million metabolome. ChemResToxicol. 29(12):1956–1975. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Vermeulen R, Schymanski EL, Barabasi AL, Miller GW. 2020. The exposome and health: Where chemistry meets biology. Science. 367(6476):392–396. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Walker DI, Go YM, Liu K, Pennell KD, Jones DP. 2016a. Population screening for biological and environmental properties of the human metabolic phenotype: Implications for personalized medicine In: Holmes E, Nicholson JK, Darzi AW, Lindon JC, editors. Metabolic phenotyping in personalized and public healthcare. Academic Press; p. 167–211. [Google Scholar]
  48. Walker DI, Mallon CT, Hopke PK, Uppal K, Go YM, Rohrbeck P, Pennell KD, Jones DP. 2016b. Deployment-associated exposure surveillance with high-resolution metabolomics. JOccupEnvironMed. 58(8 Suppl 1):S12–S21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Warner EF, Smith MJ, Zhang Q, Raheem KS, O’Hagan D, O’Connell MA, Kay CD. 2017. Signatures of anthocyanin metabolites identified in humans inhibit biomarkers of vascular inflammation in human endothelial cells. Mol Nutr Food Res. 61(9). [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Wishart D, Arndt D, Pon A, Sajed T, Guo AC, Djoumbou Y, Knox C, Wilson M, Liang Y, Grant J et al. 2015. T3db: The toxic exposome database. Nucleic Acids Res. 43(Database issue):D928–934. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Wishart DS, Tzur D, Knox C, Eisner R, Guo AC, Young N, Cheng D, Jewell K, Arndt D, Sawhney S et al. 2007. Hmdb: The human metabolome database. Nucleic Acids Res. 35(Database issue):D521–526. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Wu Q, Jiang D, Minor M, Chu HW. 2014. Electronic cigarette liquid increases inflammation and virus infection in primary human airway epithelial cells. PLoS One. 9(9):e108342. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Yang I, Sandeep S, Rodriguez J. 2020. The oral health impact of electronic cigarette use: A systematic review. Crit Rev Toxicol. 50(2):97–127. [DOI] [PubMed] [Google Scholar]
  54. Yu J, Han JC, Hua LM, Gao YJ. 2013. In vitro characterization of glucuronidation of vanillin: Identification of human udp-glucuronosyltransferases and species differences. Phytother Res. 27(9):1392–1397. [DOI] [PubMed] [Google Scholar]
  55. Yu T, Park Y, Johnson JM, Jones DP. 2009. Aplcms--adaptive processing of high-resolution lc/ms data. Bioinformatics. 25(15):1930–1936. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Zhao YD, Yin L, Archer S, Lu C, Zhao G, Yao Y, Wu L, Hsin M, Waddell TK, Keshavjee S et al. 2017. Metabolic heterogeneity of idiopathic pulmonary fibrosis: A metabolomic study. BMJ Open Respir Res. 4(1):e000183. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Table 1
NIHMS1653511-supplement-Supplementary_Table_1.xlsx (139.2KB, xlsx)
Supplementary Table 2
NIHMS1653511-supplement-Supplementary_Table_2.xlsx (21.3KB, xlsx)
Supplementary Table 3
NIHMS1653511-supplement-Supplementary_Table_3.xlsx (156.6KB, xlsx)
Supplementary Table 4
NIHMS1653511-supplement-Supplementary_Table_4.xlsx (14.8KB, xlsx)
Supplementary Table 6
NIHMS1653511-supplement-Supplementary_Table_6.xlsx (19.6KB, xlsx)
Supplementary Table 7
NIHMS1653511-supplement-Supplementary_Table_7.xlsx (20KB, xlsx)
Supplemental Table 5
NIHMS1653511-supplement-Supplemental_Table_5.xlsx (26.4KB, xlsx)

RESOURCES