Metabolomics Provides Novel Insights into the Potential Toxicity Associated with Heated Tobacco Products, Electronic Cigarettes, and Tobacco Cigarettes on Human Bronchial Epithelial BEAS-2B Cells

Smoking is an established risk factor for various pathologies including lung cancer. Electronic cigarettes (e-cigs) and heated tobacco products (HTPs) have appeared on the market in recent years, but their safety or, conversely, their toxicity has not yet been demonstrated. This study aimed to compare the metabolome of human lung epithelial cells exposed to emissions of e-cigs, HTPs, or 3R4F cigarettes in order to highlight potential early markers of toxicity. BEAS-2B cells were cultured at the air–liquid interface and exposed to short-term emissions from e-cigs set up at low or medium power, HTPs, or 3R4F cigarettes. Untargeted metabolomic analyses were performed using liquid chromatography coupled with mass spectrometry. Compared to unexposed cells, both 3R4F cigarette and HTP emissions affected the profiles of exogenous compounds, one of which is carcinogenic, as well as those of endogenous metabolites from various pathways including oxidative stress, energy metabolism, and lipid metabolism. However, these effects were observed at lower doses for cigarettes (2 and 4 puffs) than for HTPs (60 and 120 puffs). No difference was observed after e-cig exposure, regardless of the power conditions. These results suggest a lower acute toxicity of e-cig emissions compared to cigarettes and HTPs in BEAS-2B cells. The pathways deregulated by HTP emissions are also described to be altered in respiratory diseases, emphasizing that the toxicity of HTPs should not be underestimated.


Introduction
The Global Burden of Disease Project [1] estimated that approximately 1.14 billion people worldwide smoked in 2019.The majority of smokers are addicted to nicotine delivered via cigarettes, rationalizing the high prevalence of tobacco-induced diseases decades later [2].Smoking accounted for 8 million deaths globally in 2023 according to the WHO [3].The major causes of tobacco death include lung cancer, emphysema, heart attack, stroke, cancer of the upper aerodigestive areas, and bladder cancer [4].Smoking tobacco is also responsible of chronic diseases such as eye diseases, periodontal diseases, cardiovascular diseases, chronic obstructive pulmonary diseases, diabetes mellitus, rheumatoid arthritis, and disorders affecting the immune system.People who smoke can expect to lose an average of at least a decade of life compared to equivalent non-smokers [5,6].Smoking cessation greatly reduces the risks of smoking-related diseases.Nicotine administered alone in various nicotine replacement formulations (such as patches or gums) is safe and effective as an evidence-based smoking cessation aid.Novel forms of nicotine delivery Toxics 2024, 12, 128 2 of 22 systems have also emerged, including electronic cigarettes (e-cigs) and, more recently, heated tobacco products (HTPs).These products are marketed as lower-risk products than conventional cigarettes due to the absence of tobacco for e-cigs or tobacco combustion for HTPs.However, the safety of e-cigs or HTPs has not yet been fully established due to a lack of exhaustive independent toxicological studies.
Our previous findings demonstrated that HTP emissions contained fewer toxic compounds (polycyclic aromatic hydrocarbons, carbonyl compounds, and metals) than conventional cigarette smoke, but more than e-cig aerosols [7,8].In a model of human bronchial epithelial cells (BEAS-2B cell line) cultured at the air-liquid interface (ALI), conventional cigarettes were more cytotoxic and induced more oxidative stress and genotoxicity than HTPs, unlike e-cigs, which had no effect on these parameters under the studied conditions [9].In addition to these conventional methods for studying cellular damage, the use of global approaches such as "omics" (metabolomics, transcriptomics, MiRnomics, etc.) can provide a better understanding of the molecular and cellular mechanisms of toxicity, and it helps to identify relevant markers of effects and/or exposure.
Metabolomics aims to comprehensively assess changes in the metabolome induced by endogenous and/or exogenous factors, to screen for significantly different metabolite profiles, and thus to identify potential biomarkers [10].To date, the metabolic effects of exposure to HTPs or e-cig aerosols have been poorly characterized.A review of the corresponding studies, the experimental conditions implemented, and the main results obtained is summarized in Table 1.Several studies on e-cigs (most of them independent from the tobacco industry) have been carried out in humans [11][12][13][14][15] or using in vivo [16][17][18] or in vitro [19][20][21] models.These studies demonstrated the deregulation of many metabolic pathways after e-cig exposure, including glycolysis, the tricarboxylic acid (TCA) cycle, amino-acid metabolism, beta-oxidation, phospholipid metabolism, sphingolipid metabolism, or antioxidant metabolism.Two studies on HTPs showed the benefits of tobacco cessation or switching to HTPs on the human lipidomic lung profile [22,23].In addition, exposure to HTPs had a lower impact than cigarette smoke on the pathophysiology of human gingival organotypic cultures [24].However, these few studies were all conducted by the tobacco industry itself.Moreover, it is noteworthy that there are currently no metabolomic studies comparing the toxicity of e-cigs and HTPs in a similar model, although such a comparison would be essential given that both products are increasingly recommended as smoking cessation aids.Therefore, the aim of the present study was to compare the metabolome of human lung epithelial cells (BEAS-2B cell line) exposed to e-cigs, HTPs, or cigarette emissions in order to highlight potential metabolic fingerprints and to identify early markers of toxicity.We applied liquid chromatography-high resolution mass spectrometry (LC-HRMS)-based untargeted metabolomics to analyze endogenous and exogenous compounds in the lysate of cell cultures after HTP, e-cig, or cigarette aerosol exposure.Serum: no difference.Urine: A specific metabolomic signature characterized the stop-session, including 3-hydroxyisovalerate (↘), pyruvate (↘), trimethylamine oxide (↗), hippurate (↗), and N-phenylacetyl-glycine (↗).
In regular e-cig users, short-term vaping cessation shifted baseline urine metabolome Yes [11] Switch from cigarette to e-cig 1st e-cig generation Short-term (5 days) Human Urine, plasma n = 75 ↘ xenobiotic exposure (nicotine and its metabolites, other cigarette smoke constituents).Improved vitamin metabolism and ↘ oxidative stress.
Less toxic environment for consumers of e-cigs and potential health benefits compared to people who smoke cigarettes No [12] Effects of chronic e-cig vaping and cigarette smoking No information on e-cig exposure device Long-term (>2 years) Human Plasma n = 24 E-cig vaping deregulated TCA cycle-related metabolites, while cigarette smoking altered sphingolipid metabolism.

Specific metabolic signatures could serve as potential systemic biomarkers for early pathogenesis of cardiopulmonary diseases
Yes [13] Long-term effects of e-cigs compared to tobacco

No information on e-cig exposure device
Long-term effects (>6 months) Metabolomic signature of 839 and 396 features for people who smoke and vape, respectively, including 12% of common metabolites.↗ acylcarnitines and acylglycines in vapers, suggesting higher lipid peroxidation.Trend of ↗ in cancer-related biomarkers (Me-Fapy) in people who vape.

Deregulation of markers of inflammatory status and
fatty acid oxidation in people who vape, as well as a trend of elevated cancer-related biomarkers.

Potential biomarkers of periodontal disease in vapors.
Yes [15] Acute exposure to e-cigs 4th e-cig generation Short-term ( Vanillin perturbed specific energy, amino acid, antioxidant, and sphingolipid pathways previously associated with human disease such as lung disease including asthma, idiopathic pulmonary fibrosis, and acute respiratory distress syndrome.
Vanillin could drive the lung metabolic microenvironment to a more pathogenic state.
Yes [20] Effects of e-cig maltol (flavorant) 3rd e-cig generation Short-term (1 h) Perturbation of oxidative stress with e-liquids with or without maltol.Deregulation of amino acid metabolism specifically with maltol.Many effects of firsthand exposure were also observed with secondhand exposure.
Flavorants in e-liquids impact lung metabolism after both firsthand and secondhand exposure.

Chemicals and Reagents
The BEAS-2B cell line was purchased from the American Type Culture Collection (ATCC ® CRL9609™ Manassas, VA, USA).LHC-9 medium, phosphate buffer solution (PBS), and type I collagen solution were purchased from Life Technologies (Courtaboeuf, France).CellBIND 75 cm 2 tissue culture flasks and transwell culture inserts (4.67 cm 2 ) with a 0.4 µm pore size were bought from Corning (Amsterdam, The Netherlands) and Sigma Aldrich (Saint-Quentin Fallavier, France), respectively.
3R4F reference cigarettes were purchased from the University of Kentucky (Lexington, KY, USA).The tested HTP was an IQOS 2.4 model manufactured by Philip Morris International (Neuchâtel, Switzerland), used with IQOS heatsticks (amber box from Philip Morris International).A third-generation "ModBox-TC" model manufactured by NHOSS ® brand (Bondues, France) was used with an "Air Tank" clearomiser equipped with a 0.5 Ω kanthal coil and containing a "blond tobacco" flavored e-liquid (NHOSS ® brand, Bondues, France) with 16 mg/mL of nicotine.

Cell Culture and Experimental Design for Cell Exposure
Human bronchial epithelial BEAS-2B cells were selected as the in vitro model.Their cell culture protocol has already been described [9].Briefly, the cells were seeded onto transwell culture inserts.The ALI was established by removing the LHC-9 medium from the apical surface, exposing only the basal surface to the medium.The BEAS-2B cells cultured at the ALI were transferred to an exposure module (Vitrocell 6/4 CF module) and exposed to different doses of emissions from the 3R4F cigarette, HTP, or e-cig set up at 18 W (Mb-18W) or 30 W (Mb-30W) using the Vitrocell ® VC1 Smoking machine (Vitrocell, Waldkirch, Germany).The cells were not exposed to equal quantities of nicotine for all the devices but to comparable cytotoxicity conditions (>80% cell viability) for all exposures based on preliminary data: 60 and 120 puffs for Mb-18W, Mb-30W, and HTP and 2 and 4 puffs for the 3R4F cigarette [9].The Health Canada intense (HCI) puff profile was used to test all the products.For HTP exposure, the HCI regime was modified without blocking the ventilation holes of the IQOS heatsticks to avoid overheating of the device.The control cells consisted of unexposed cells that were maintained within the incubator [25].After exposure, cells were incubated at 37 • C and 5% CO 2 for 24 h before further sample preparation.The supernatant media were removed and stored at −80 • C until further measurement of lactate dehydrogenase (LDH).Four independent cell cultures were used to replicate each experimental point.

Cytotoxicity Evaluation
Cytotoxicity assays were performed using the Cytotoxicity Detection KitPLUS LDH (Cytotoxicity Detection Kit PLUS, Roche Diagnostics GmbH, Mannheim, Germany) according to the manufacturer's instructions.The assay relies on the evaluation of LDH activity that is discharged from the cytosol of damaged cells and measured in the supernatant medium 24 h after exposure.A positive control was included: the maximal LDH activity was measured by lysing the cells using Triton-X100.Cytotoxicity was determined as percentages related to positive control cells arbitrarily set at a value of 100%.

Sample Preparation
For sample harvesting, the cells were washed twice with PBS.Their metabolisms were subsequently quenched by the addition of 800 µL of ice-cold methanol/water (80:20, v/v).Here, metabolism quenching was achieved by combining a low temperature which decreases enzymatic activities and prevents metabolite spontaneous degradation and through the addition of an organic solvent which inactivates enzymes and contributes to the disruption of cell membranes, thus enabling metabolite extraction [26].After 20 min of incubation at −20 • C, the cells were scraped off the culture vessel on ice using rubbertipped cell scrapers.The lysates were collected and subsequently centrifugated for 5 min at 14,000× g and +4 • C to eliminate cell debris.The supernatants were recovered in a vial, and the precipitates were rinsed in another 200 µL of ice-cold 80% methanol, vortexed, centrifugated for 5 min at 14,000× g at +4 • C, then the supernatant was also transferred to the vial.Quality control samples (QC samples) were prepared by pooling equal aliquots of all the samples.The supernatants were concentrated to dryness using a speedvac, and the dried samples were stored at −80 • C until further use.Just before their injection into the chromatographic system, the samples were reconstituted using 100 µL of a water/methanol (90:10, v/v) mixture containing internal standards (methyl-clonazepam at 0.125 mg/L, βhydroxy-ethyltheophyllin at 1.6 mg/L, and phenobarbital-D5 at 1 mg/L in methanol).The samples were then centrifugated for 15 min at 14,000× g and +4 • C. The injected volume was set at 10 µL of supernatant for each sample.

Extraction of Raw Data and Pre-Processing
The LC-MS data were analyzed using the Progenesis QI software v1.0.5162 (Nonlinear Dynamics, Newcastle upon Tyne, UK).The retention time alignment, peak picking, and adduct deconvolution were sequentially performed.The sensitivity of the peak-picking step was tuned to recover approximately 5000 features from each analysis.The intensity data for each detected feature were exported from Progenesis QI as CSV files for further data analysis.

Data Processing and Statistical Analysis
Data processing and statistical analyses were conducted in the R environment [28].When analyzing untargeted metabolomic data, missing values are often encountered, but most multivariate statistical methods cannot be applied when the data have missing values.To use these incomplete data sets, features with more than 50% missing values were removed, and the remaining missing values were replaced by 1/5 of the minimum positive value of each variable (LoDs).The data were then filtered, removing features with relative standard deviations higher than 20% in the QC samples.The data were transformed via log transformation and normalized via cyclic loess normalization to obtain a normally distributed population.The quality of the pre-treatment was assessed and confirmed via principal component analysis and intra-class correlation.Further statistical analyses were performed on the processed data with a significance threshold set at a p-value of 0.05 after adjustment for the false discovery rate (FDR) [29] to correct for multiple statistical testing.First, we performed a multivariate supervised partial least squares discriminant analysis (PLS-DA) to discriminate between the different exposures to the 3R4F cigarette, Toxics 2024, 12, 128 8 of 22 HTP, Mb-18W, and Mb-30W.The features with the highest variable importance in the projection (VIP > 1.5) were selected for a cluster analysis via a multivariate unsupervised analysis heatmap to reveal the relationships of features.Second, an ANOVA test was used to compare the variances of the samples belonging to the 4 different exposure groups.This analysis revealed the deregulated features within each type of exposure.For significant features with a fold-change (FC) > 1.5, a Student's t-test was performed between the control and exposed samples.The intensity of deregulation compared to the control samples (D0) was expressed as Log2(FC), i.e., Log2(FC D1 ) for deregulation between the control and exposure at the lowest dose (D1) and Log2(FC D2 ) for deregulation between the control and exposure at the highest dose (D2).Significantly deregulated features were considered to be a metabolomic signature resulting from a specific exposure and were submitted to an identification process.

Feature Annotation and Pathway Analysis
Feature identification was performed using the Progenesis QI software.The altered features were further queried to propose feature annotations against three databases, which were, according to their rank of annotating confidence, as follows: (1) a homemade database containing the spectral properties, retention time (Rt), and collision cross section (CCS) for peak annotation from authentic standard compounds [27]; (2) an in-house predicted database containing Rt and CCS predicted using machine learning for almost 114,000 metabolites, as previously described [27]; and (3) a commercial metabolomic profiling CCS library (Waters, Manchester, UK).In a few cases, the spectra were manually checked against the MassBank spectral database [30] to broaden the identification possibilities.
A pathway analysis was carried out using MetaboAnalyst 5.0.The Homo Sapiens (KEGG) pathway library was queried using a hypergeometric test for metabolite enrichment analysis and a relative betweenness centrality topology analysis.The corrected p-values (FDR) < 0.25 represented notable enrichments of certain metabolites in a pathway.

Evaluation of Cytotoxicity
Two exposure doses (D1 and D2) were chosen for each device based on comparable subtoxic doses (>80% cell viability, measured via an ATP test), as previously reported [9], in order to evaluate a potential dose-dependent effect: D1 = 60 puffs and D2 = 120 puffs for Mb-18W, Mb-30W, and HTP and D1 = 2 puffs and D2 = 4 puffs for 3R4F cigarettes.To ensure that the cells were exposed to comparable subtoxic doses in the present metabolomic study, an LDH assay was used to evaluate the cytotoxicity in the cells exposed to the different emissions (Figure 1).HTP emissions caused a higher cytotoxicity after 120 puffs compared to the controls (Kruskal-Wallis test: p-value (HTP) = 0.03; Wilcoxon test: p-value (D0 vs. D2) = 0.02), while no differences were observed in the other groups.Consistently, we observed that all the devices showed a cytotoxicity below 20%.Thus, under the subtoxic conditions selected for the present metabolomic analysis, the effects of exposure reflected intracellular effects rather than changes due to cell death.

Evaluation of Metabolomics Data Pre-Treatment
Extraction of raw data permitted us to generate a data matrix made up of 5130 and 5037 features in the ESI+ and ESI− modes, respectively.After pre-treatment, the final data matrix was made up of 46 samples (two outlier samples were removed, possibly due to an injection failure) and 3591 features (2398 compounds from ESI+ and 1193 compounds from ESI−).

Impact of the Type of Emission
A multivariate supervised PLS-DA was first performed to discriminate between the different emissions (3R4F cigarette, HTP, Mb-18W, and Mb-30W).A PLS-DA model was created using the processed data to evaluate group separations and to calculate the VIP scores for each feature.The PLS-DA performed on the exposed groups (D1 and D2 combined) showed strong model statistics for the differentiation of the groups (R2X = 0.659, R2Y = 0.961), with a good reproducibility (Q2Y = 0.738).The obtained loading scatter plot (Figure 2) facilitated a global view of the relationships between variables.This model allowed us to separate the 3R4F cigarette, HTP, and non-tobacco groups (Mb-18W and Mb-30W).

Evaluation of Metabolomics Data Pre-Treatment
Extraction of raw data permitted us to generate a data matrix made up of 5130 and 5037 features in the ESI+ and ESI− modes, respectively.After pre-treatment, the final data matrix was made up of 46 samples (two outlier samples were removed, possibly due to an injection failure) and 3591 features (2398 compounds from ESI+ and 1193 compounds from ESI−).

Impact of the Type of Emission
A multivariate supervised PLS-DA was first performed to discriminate between the different emissions (3R4F cigarette, HTP, Mb-18W, and Mb-30W).A PLS-DA model was created using the processed data to evaluate group separations and to calculate the VIP scores for each feature.The PLS-DA performed on the exposed groups (D1 and D2 combined) showed strong model statistics for the differentiation of the groups (R2X = 0.659, R2Y = 0.961), with a good reproducibility (Q2Y = 0.738).The obtained loading scatter plot (Figure 2) facilitated a global view of the relationships between variables.This model allowed us to separate the 3R4F cigarette, HTP, and non-tobacco groups (Mb-18W and Mb-30W).The VIP lists were established, pinpointing 180 features with a VIP > 1.5, which were considered to be the discriminant features in this model.These discriminant features were used to build a heatmap (Figure 3).The heatmap analysis revealed a classification of the samples according to two arms.The first arm comprised samples exposed to tobacco products.Among them, those exposed to HTP were separated from those exposed to 3R4F.The VIP lists were established, pinpointing 180 features with a VIP > 1.5, which were considered to be the discriminant features in this model.These discriminant features were used to build a heatmap (Figure 3).The heatmap analysis revealed a classification of the samples according to two arms.The first arm comprised samples exposed to tobacco products.Among them, those exposed to HTP were separated from those exposed to 3R4F.The second arm consisted of samples exposed to e-cigs or not exposed (controls).Among them, those exposed to e-cigs were fairly well-discriminated from those not exposed.However, the samples exposed to Mb-18W or Mb-30W were not separated.Overall, these data show that both tobacco products induced similar metabolic deregulations, while the metabolome of cells exposed to e-cigs was not very different from that of unexposed cells.

Impact of the Exposure Dose
An ANOVA statistical study was conducted on the pre-treated data with a 95% confidence level (corrected p-value < 0.05), demonstrating the presence of significant differences between 214 feature levels in the cells exposed to 3R4F cigarettes, e-cigs, or HTPs compared to unexposed cells.Among them, 95 features were classified at the top of the highest VIP score (VIP > 1.5).A Venn diagram helps to visually represent the number of deregulated features among the four groups (Figure 4).

Impact of the Exposure Dose
An ANOVA statistical study was conducted on the pre-treated data with a 95% confidence level (corrected p-value < 0.05), demonstrating the presence of significant differences between 214 feature levels in the cells exposed to 3R4F cigarettes, e-cigs, or HTPs compared to unexposed cells.Among them, 95 features were classified at the top of the highest VIP score (VIP > 1.5).A Venn diagram helps to visually represent the number of deregulated features among the four groups (Figure 4).
Both 3R4F and HTP emissions significantly affected the metabolome of the BEAS-2B cells, whereas no difference was observed after e-cig exposure (only one deregulated compound with Mb-18W) compared to the controls.Notably, 84% of the compounds deregulated after 3R4F exposure were also deregulated after HTP exposure.This corresponded to 43 common compounds.To further identify the impact of the exposure dose, a Student's t-test was performed (Table 2).A total of 198 and 204 features were deregulated after 60 puffs and 120 puffs of HTP exposure, respectively, out of which 197 were common.Fifty-four percent of the deregulated features were upregulated.Exposure to 2 and 4 puffs of 3R4F cigarette smoke induced a fluctuation of 46 and 51 features, respectively (46 in common).Fifty-one percent of the deregulated metabolites were upregulated.The unique deregulated feature after Mb-18W exposure was upregulated after both durations of exposure.

Impact of the Exposure Dose
An ANOVA statistical study was conducted on the pre-treated data with a 95% confidence level (corrected p-value < 0.05), demonstrating the presence of significant differences between 214 feature levels in the cells exposed to 3R4F cigarettes, e-cigs, or HTPs compared to unexposed cells.Among them, 95 features were classified at the top of the highest VIP score (VIP > 1.5).A Venn diagram helps to visually represent the number of deregulated features among the four groups (Figure 4).

Feature Identification
The 214 altered features were further queried for compound annotations against databases.The annotation process followed the standards defined by the Metabolomics Standards Initiative [31].The results are detailed in Table 3. Eleven compounds were identified vs. authentic standards (confidence level = 1), 61 were putatively annotated (confidence level = 2), and 12 were attributed to chemical classes (confidence level = 3).The 84 annotated compounds varied in structure and covered a wide field of endogenous and exogenous metabolites.

Endogenous Compounds: Pathway Analysis and Biological Interpretation
Of the 51 discriminant features after 3R4F cigarette exposure, 13 were identified as endogenous compounds, including 8 lipids belonging to the eicosanoid class.This small number of identified endogenous metabolites did not permit us to perform an overrepresentation analysis.

Discussion
Considering the effects of tobacco smoke on health and lifespan and the limited information on the biological/health effects of e-cigs and HTPs use by consumers, there is a need to evaluate the health risks of these new tobacco and vaping products and to identify markers that would help us to understand their underlying physiopathological processes.In this study, we used metabolomic profiling to examine and compare the metabolic responses of a bronchial epithelial cell model to short-term exposure to cigarette smoke (2 or 4 puffs), HTP emissions (60 or 120 puffs), or e-cig aerosols (60 or 120 puffs) generated by a device set up at two powers (18 W or 30 W).The doses of exposure were chosen based on prior studies showing that exposure can cause cell death after higher doses of exposure to HTPs and 3R4F cigarettes [9].Here, the cytotoxicity assessed by measuring LDH in the culture media of the samples included in this study was consistent with these preliminary data, with a cytotoxicity <20% compared with the control samples.Working at low and comparable cytotoxicity conditions ensured that the differences observed in metabolite abundance reflected biological variability due to exposure and not due to cell death.
We differentially analyzed the 10,167 features detected using LC-HRMS.We were able to filter 214 differences between the exposed and control samples.Based on the metabolites that were significantly deregulated (FC > 1.5 and adjusted p-value < 0.05), a robust metabolomic fingerprint composed of 51 or 205 features was shown to be linked to 3R4F or HTP exposure, respectively.The number of significantly deregulated compounds and the intensity of these modulations (FC) increased to a limited extent with the dose of exposure, suggesting that there was a small dose-dependent effect.Eighty-four percent of the discriminant compounds after 3R4F exposure were also discriminant after exposure to HTP aerosols, suggesting common markers of exposure or effects between both tobacco products.These effects are detectable after lower-exposure doses for cigarettes (<4 puffs) than for HTPs (60 and 120 puffs).The greater number of discriminant metabolites after HTP exposure can be likely explained by the difference in exposure doses between HTP (60 and 120 puffs) and the 3R4F cigarette (2 and 4 puffs).Exposure to HTPs and e-cigs was more comparable, as the same doses were used for both devices.Only 1 metabolite was significantly deregulated after exposure to 60 or 120 puffs from Mb-18W, while no features were significantly deregulated with Mb-30W.This unidentified compound was not significantly modulated in the other exposure conditions.This suggests that there were few measurable metabolic alterations in our experimental conditions after exposure to e-cigs.
Among the 214 differential features, we identified 84 exogenous or endogenous compounds: 11 identified compounds (level 1), 61 putatively annotated compounds (level 2), and 12 putatively characterized compound classes (level 3).One hundred and thirty features (60% of the deregulated features) remained as unknown signals (level 4), which is not surprising given that feature identification is one of the main limitations of untar-geted metabolomics.These annotations allowed for powerful deciphering of deregulated compounds to better understand the cellular effects of tobacco or HTP exposure.
Of the compounds identified, eight were exogenous compounds, seven of which varied in a similar way when exposed to 3R4F smoke or HTP emissions.Another compound, 3-methylindole, was only increased through exposure to HTP emissions.In addition to nicotine, we highlighted an increase in the exogenous compounds that were already described in the literature as originating from the tobacco plant itself or from tobacco smoke: 1-naphthylamine (aromatic amines, naphthalene class) [32,33], scopoletin (polyphenol, coumarins, and derivatives class) [34], norharman, and harman (β-carbolines) [35].The compound 3-methylindole, which was increased after exposure to HTP aerosols but not after exposure to 3R4F smoke (potentially due to the low exposure dose for 3R4F), belongs to the indole and derivatives class.It is formed through the pyrolysis of tryptophan during tobacco combustion.This result could be in favor of a pyrolysis phenomenon occurring in response to the HTP device.The results published by Vivarell et al. indicate that HTP emissions contain pyrolysis and thermal degradation by-products identical to conventional cigarette smoke, and that they cause serious lung damage and increase the risk of cancer in animal models [36].The compound 3-methylindole has been described as cytotoxic in BEAS-2B cell lines after bio-activation [37] and as mutagenic in human lung microsome models, supporting the hypothesis of a probable pulmonary carcinogenic effect in humans [38].In normal human bronchial epithelial (NHBE) cells, exposure to 3-methylindole also caused significant DNA damage and mutations without triggering apoptotic defenses, reinforcing the hypothesis that this compound inhaled from cigarette smoke could be a selective lung carcinogen [39].In conclusion, these exogenous compounds can be considered as markers of exposure to tobacco products, both from cigarette and HTP emissions.While the tobacco industry described HTP as riskless to users' health, our results suggest that they could be nonetheless toxic, as one carcinogenic compound was identified.The toxicity of HTP should therefore not be underestimated.
Exposure to tobacco products was associated with a pulmonary cellular stress response, notably an oxidative stress response directly caused by exposure to chemical compounds or induced by the generation of reactive oxygen species (ROS).First, we observed disturbances in glutathione metabolism, notably with an increase in oxidized glutathione after exposure to the higher dose, both for HTP and 3R4F emissions.These results support the generation of oxidative stress following exposure to tobacco products.These findings were consistent with our previously published results, which showed an increase in the oxidized to reduced glutathione ratio following exposure to cigarette and HTP emissions [7].We also showed that HTPs and 3R4F (but not e-cigs) induced activation of the transcription factor Nrf2 and expression of its target genes, heme oxygenase 1 and NAD(P)H-quinone dehydrogenase 1, demonstrating an antioxidant response after exposure to tobacco products [9].In our metabolomic study, an increase in biliverdin was observed after exposure to 3R4F emissions (two and four puffs).Biliverdin is a heme metabolite produced under the action of a cryoprotective enzyme, heme oxygenase (HO) [40,41].Biliverdin is then rapidly converted to bilirubin by biliverdin reductase.Biliverdin can be regenerated from bilirubin through reactions with ROS.Various properties have been attributed to biliverdin and bilirubin, including antioxidant properties in response to oxidative stress.A study by Titz et al. [23] analyzing the metabolome of lung tissue after exposure of mice to conventional cigarettes or HTPs also revealed deregulation of bilirubin metabolism after exposure to cigarettes but not to HTP.Second, our study also revealed modulations in nucleotide metabolism (purines and pyrimidines) after exposure to HTPs (but not after 3R4F cigarettes).Purine metabolism has been reported to produce ROS via the xanthine oxidase pathway [42].In addition, a deregulation in nucleotide metabolism has already been described in cigarette smokers, with a role in cancer development [43,44].Third, we observed a decrease in methionine after exposure to HTP.Oxidative stress can induce methionine oxidation [45], leading to a decrease in methionine.An increase in the products of this oxidation (methionine sulphoxide and N-acetyl-methionine sulphoxide) has been described following exposure of human gingival epithelial cells for 28 min over 3 days to conventional cigarette smoke, but not to HTP emissions [24].A decrease in plasmatic methionine has also been observed in cigarette users compared with people who vaper or unexposed individuals [13].
The deregulation of several metabolites indicated an alteration in energy metabolism following exposure to tobacco products.Increased energy demand is an expected response for all cells under stress.Thus, ADP levels were increased after exposure to HTP aerosols.An intermediate of the TCA cycle, isocitrate, was decreased in cells exposed to HTP compared with controls.It is well known that cigarette smoke inhibits mitochondrial respiratory function and deregulates the TCA cycle [46], a central pathway for cellular energy metabolism.Isocitrate had also been previously described as decreased after repeated exposure of 3D bronchial tissue culture to cigarette smoke [47].Our analysis also showed a significant decrease in isovalerylcarnitine, a short-chain carnitine, and a decrease in an intermediate-chain acylcarnitine, hexanoylcarnitine, after exposure to HTP emissions.Acylcarnitines are organic compounds containing a fatty acid.They play a central role in the transport of fatty acids across the inner mitochondrial membrane during beta-oxidation, the metabolic pathway through which fatty acids are broken down to produce acetyl-CoA, which feeds the TCA cycle.Their disruption confirms mitochondrial dysfunction and disruption of beta-oxidation of fatty acids after exposure to tobacco products, which may reflect a high intracellular energy demand.
Lipids other than carnitines were also affected by exposure to tobacco products (Figure 6), with significant variations in several classes of lipids such as glycerophospholipids (glycerophosphocholines (LysoPC) or glyceroethanolamines (LysoPE)) or fatty acids (carnitines, eicosanoids).Most of these lipids increased in a dose-dependent manner in favor of lipid accumulation.Eicosanoids, derived from arachidonic acid through the action of the lipooxygenase and cyclooxygenase enzymes, are signaling mediators of the inflammatory response.An increase in the activity of these enzymes had already been described after HTP exposure [36], but also in smokers [48], and was associated with inflammation and pulmonary pathologies [49] and cancers [50].Other lipid classes were also increased after exposure to HTP emissions.These included lysoglycerophospholipids of the glycerophosphocholine subclass (lysophosphatidylcholines, LysoPC) and glycerophosphoethanolamines (lysophosphatidylethanolamines, LysoPE).LysoPCs are produced through the cleavage of phosphatidylcholine by phospholipase A2 (PLA2), a reaction that also forms free fatty acids such as arachidonic acid, which is the precursor of eicosanoids [51].This reaction is also possible under the effects of ROS.LysoPEs are also produced through the cleavage of phosphatidylethanolamine by PLA2.Altogether, exposure to cigarette and HTP emissions could therefore affect the regulation of the inflammatory response in bronchial cells.Studies conducted by the tobacco industry comparing the murine bronchial tissue lipidome after exposure to tobacco and HTP have shown lipid deregulation, but much more was marked for cigarettes than for HTP (very limited effects for HTP), allowing them to conclude that HTP had a lower toxic effect than conventional cigarettes [22,23].Given our results, the impact of HTP on those metabolic pathways should not be underestimated.
Some limitations of the present study must be mentioned.A unique regime (HCI) was used for all the tested products to compare their toxicity under the same laboratory conditions.However, other standardized smoking or vaping regimes have been created based on users' puffing behavior (ISO 20768:2018 [52] for e-cigs, HCI and ISO3308:2012 [53] for tobacco products).In this study, the unique regime choice could be a limitation to an accurate assessment of potential health implications in the context of different product categories.In addition, the nontargeted metabolomic analysis is an initial screening.Level 2 annotations should be interpreted with caution, as some of them may derive from a false positive annotation.In this context, further target verification is essential in the future.In particular, it should be noted that the majority of studies analyzing eicosanoids use analytical techniques with a negative ionization mode, whereas here, the identifications were made for compounds detected using a positive ionization mode.In this context, and given the risk of erroneous identifications, the confirmation of these identifications will be essential to confirm these conclusions.By using our in vitro model, we were able to evaluate the acute effects of exposure exclusively.While no significant deregulation of metabolites was found after e-cig acute exposure in the present study, other researchers have demonstrated modifications in the metabolomes in the urine, plasma, or saliva of people who vape [11,[13][14][15], or in mice chronically exposed to e-cigs [16][17][18].Therefore, this comparative study of the acute toxicity of e-cigs and HTPs using a metabolomic approach will need to be completed by animal or population studies in order to compare the long-term toxicity of these new tobacco and vaping devices.Some limitations of the present study must be mentioned.A unique regime (HCI) was used for all the tested products to compare their toxicity under the same laboratory conditions.However, other standardized smoking or vaping regimes have been created based on users' puffing behavior (ISO 20768:2018 [52] for e-cigs, HCI and ISO3308:2012 [53] for tobacco products).In this study, the unique regime choice could be a limitation to an accurate assessment of potential health implications in the context of different product categories.In addition, the nontargeted metabolomic analysis is an initial screening.Level 2 annotations should be interpreted with caution, as some of them may derive from a false positive annotation.In this context, further target verification is essential in the future.In particular, it should be noted that the majority of studies analyzing eicosanoids use analytical techniques with a negative ionization mode, whereas here, the identifications were made for compounds detected using a positive ionization mode.In this context, and given the risk of erroneous identifications, the confirmation of these identifications will be essential to confirm these conclusions.By using our in vitro model, we were able to evaluate the acute effects of exposure exclusively.While no significant deregulation of metabolites was found after e-cig acute exposure in the present study, other researchers have demonstrated modifications in the metabolomes in the urine, plasma, or saliva of people who vape [11,[13][14][15], or in mice chronically exposed to e-cigs [16][17][18].Therefore, this comparative study of the acute toxicity of e-cigs and HTPs using a metabolomic approach will need to be completed by animal or population studies in order to compare the long-term toxicity of these new tobacco and vaping devices.

Conclusions
This study provides innovative data to compare the acute toxicity of alternative tobacco products (e-cigs and HTPs) and to better understand their cellular and molecular mechanisms of toxicity.The metabolomic data observed strongly suggest a lower acute toxicity of e-cig aerosols compared to cigarette and HTP emissions in the BEAS-2B cell line.The metabolomic fingerprint identified for both tobacco products (HTP and 3R4F) consisted of exogenous compounds, one of which is carcinogenic, as well as endogenous metabolites, which can be considered as markers of effects.Their deregulations, which are only observed after more intensive exposure to HTPs (60 or 120 puffs) than to cigarettes (2 or 4 puffs), indicate alterations in various metabolic pathways, including oxidative stress and mitochondrial and lipid metabolisms.The metabolites deregulated by HTPs are involved in metabolic pathways that are also altered in respiratory diseases, confirming that the toxicity of HTPs should not be underestimated.Further long-term studies in animal models should be conducted to allow for the assessment of chronic exposures to HTPs.This work provides health agencies and authorities with additional information for the regulation of these products as well as for the development of public health policies to reduce smoking and tobacco product consumption.

22 Figure 1 .
Figure 1.In vitro cytotoxicity of HTP, e-cig (Mb-18W and Mb-30W), and 3R4F cigarette emissions on BEAS-2B cells.Cell viability was evaluated by measuring the LDH released 24 h after exposure.The results are expressed as percentages relative to the LDH released in the cells treated with the positive control (Triton-X100), arbitrarily set at a value of 100%.Types of exposure are: e-cigs [Mb-18W] (green), e-cigs [Mb-30W] (cyan), 3R4F cigarettes (blue) and HTPs (red).The data represent the median and interquartile range from four independent culture replicates.* p < 0.05 compared with control cells (Wilcoxon test).

Figure 1 . 22 Figure 2 .
Figure 1.In vitro cytotoxicity of HTP, e-cig (Mb-18W and Mb-30W), and 3R4F cigarette emissions on BEAS-2B cells.Cell viability was evaluated by measuring the LDH released 24 h after exposure.The results are expressed as percentages relative to the LDH released in the cells treated with the positive control (Triton-X100), arbitrarily set at a value of 100%.Types of exposure are: e-cigs [Mb-18W] (green), e-cigs [Mb-30W] (cyan), 3R4F cigarettes (blue) and HTPs (red).The data represent the median and interquartile range from four independent culture replicates.* p < 0.05 compared with control cells (Wilcoxon test).Toxics 2024, 12, x FOR PEER REVIEW 9 of 22

Toxics 2024 , 22 Figure 3 .
Figure 3. Heatmap analysis showing normalized abundance of discriminant features (VIP > 1.5) during 3R4F cigarette, HTP, Mb-30W, or Mb-18W exposure and in controls (unexposed cells).Redcolored tiles indicate a high abundance and blue indicates a low abundance of the compounds.

Figure 3 .
Figure 3. Heatmap analysis showing normalized abundance of discriminant features (VIP > 1.5) during 3R4F cigarette, HTP, Mb-30W, or Mb-18W exposure and in controls (unexposed cells).Redcolored tiles indicate a high abundance and blue indicates a low abundance of the compounds.

Figure 3 .
Figure 3. Heatmap analysis showing normalized abundance of discriminant features (VIP > 1.5) during 3R4F cigarette, HTP, Mb-30W, or Mb-18W exposure and in controls (unexposed cells).Redcolored tiles indicate a high abundance and blue indicates a low abundance of the compounds.

Figure 5 .
Figure 5.Chemical taxonomy of the identified metabolites deregulated after HTP exposure and the proportion of each super class and class.

Figure 5 .
Figure 5.Chemical taxonomy of the identified metabolites deregulated after HTP exposure and the proportion of each super class and class.

Table 1 .
Literature review summary of the effects of e-cig or HTP aerosols on host metabolism.

Table 1 .
Cont.↗ candidate surfactant lipids, ↗ inflammatory eicosanoids, ↗ ceramide classes after cigarette exposure that were absent in mice from the cessation group and the switching group to HTPs.Substantial effects of 3R4F exposure: ↗ inflammatory and oxidative stress responses, ↗ metabolites with immunoregulatory roles (itaconate, polyamines, quinolinate), ↗ metabolites of oxidative stress response (heme-biliverdin-bilirubin pathway).HTP aerosol exposure was associated with fewer to absent effects.

Table 2 .
Number of significantly deregulated compounds for each type of exposure after t-test analysis (corrected p (FDR) < 0.05).

Table 3 .
List of the 84 metabolites identified via metabolomic analysis in the positive (POS) and negative (NEG) ionization mode (ESI) and their associated Log2(FC) between the control (D0) and exposure to the lowest dose (D1) or the highest dose (D2).The values highlighted in grey indicate significant metabolites (Student's t-test; corrected p < 0.05).