E-cigarettes and health risks: more to the flavor than just the name

Abstract

The growing interest in regulating flavored E-liquids must incorporate understanding of the “flavoring profile” of each E-liquid—which flavorings (flavoring chemicals) are present and at what concentrations not just focusing on the flavor on the label. We investigated the flavoring profile of 10 different flavored E-liquids. We assessed bronchial epithelial cell viability and apoptosis, phagocytosis of bacteria and apoptotic cells by macrophages after exposure to E-cigarette vapor extract (EVE). We validated our data in normal human bronchial epithelial cells (NHBE) and alveolar macrophages (AM) from healthy donors. We also assessed cytokine release and validated in the saliva from E-cigarette users. Increased necrosis/apoptosis (16.1–64.5% apoptosis) in 16HBE cells was flavor dependent, and NHBEs showed an increased susceptibility to flavors. In THP-1 differentiated macrophages phagocytosis was also flavor dependent, with AM also showing increased susceptibility to flavors. Further, Banana and Chocolate were shown to reduce surface expression of phagocytic target recognition receptors on alveolar macrophages. Banana and Chocolate increased IL-8 secretion by NHBE, whereas all 4 flavors reduced AM IL-1β secretion, which was also reduced in the saliva of E-cigarette users compared with healthy controls. Flavorant profiles of E-liquids varied from simple 2 compound mixtures to complex mixtures containing over a dozen flavorants. E-liquids with high benzene content, complex flavoring profiles, high chemical concentration had the greatest impacts. The Flavorant profile of E-liquids is key to disruption of the airway status quo by increasing bronchial epithelial cell apoptosis, causing alveolar macrophage phagocytic dysfunction, and altering airway cytokines.

INTRODUCTION

Electronic (E)-cigarettes are often advertised as a safe way to help people quit smoking tobacco cigarettes. The number of E-cigarette users (vapers) continues to grow in many countries with many younger people now using the devices, usually for social and novelty aspects (1, 2). E-liquid consists of flavorings (flavoring chemicals) in a glycol base made up of propylene glycol (PG), vegetable glycerin (VG), and often with nicotine at concentrations up to 48 mg/mL. E-liquids have been shown to cause a widening range of effects beyond toxicity in cells, animal models, and humans (reviewed in Bozier et al. (3)). Effects observed in the lung include enlarged human lung fibroblasts with induced spindle formation and vacuolization (4), and loss of lung endothelial barrier function (5). Bozier et al. (3) and the National Academies of Sciences, Engineering, and Medicine (6) have provided excellent systematic reviews on the known health effects of E-cigarettes using in vitro, ex vivo, and in vivo studies. Data from previous studies, including our own, are emerging that even nicotine-free E-liquids may not be harmless, likely because of the flavorants present (3, 6–8). These findings in combination with the fact that entertaining, fruity, and sweet flavors of E-cigarettes are more attractive to youth (9) and that vaping has been described as a possible gateway to smoking (10) have resulted in interest toward regulating flavors. However, few studies outside of toxicity studies have ever investigated the effects of a large range of flavors. Our previous study showed that there were differences in the chemical flavoring profile of three apple E-liquids (7). Some studies tested several flavors for toxicity and suggested that specific flavored E-liquids had higher toxicity (11–14). However, some E-cigarette studies lack physiologically relevant delivery methods or a solid basis in in vitro platforms, whereas the field suffers from varied vapor production methods, often resulting in conflicting evidence; in a field that needs clear data on the health impacts of flavored E-cigarettes. No study to date has investigated the impact of the flavorant profile on the health impacts the lungs. Further, no study has assessed the effects of flavored E-cigarettes on alveolar macrophages. Thus, this study presents a thorough in vitro study with ex vivo and in vivo validation designed to be a basis for future in vivo studies

Our previous work on a single flavor showed that E-cigarettes can upset the airway status quo via increased airway epithelial cell apoptosis, reduced phagocytic clearance of apoptotic airway epithelial cells and bacteria via reduced receptor levels on THP-1 (7, 8) derived macrophages and also altered cytokine production in THP-1 derived macrophages and 16HBE cells. In this study, we investigated the flavorant profile of 10 flavored E-liquids via spectrometry. We then assessed the effects of these profiles on bronchial epithelial cell viability, macrophage phagocytosis and efferocytosis, and cytokine production, with validation from the saliva of E-cigarette users.

METHODS AND MATERIALS

EVE and CSE Preparation

E-cigarette vapor extract (EVE) was produced using an EVOD-2 3.7 V that uses a 1.5Ω. dual coil as previously described (8). Using 50 × 3 s puffs (15) (1 puff per 10 s), 10 flavors in a mix of PG and VG (PG:VG) and 100% PG, 100% VG and nicotine (at 18 mg/mL in PG:VG) were vaporized. Control medium (C) was obtained by using the same system to pass air through medium for the same duration as E-cigarette use. Cigarette smoke extract (CSE, 10%) was prepared as previously described (8) and was used as a positive control rather than for the purposes of a direct comparison. Cells were treated with 250 µL C, 10% CSE, 100% EVE for 48-well plates and 500 µL for 24-well plates for 24 h.

Mass Spectrometry of E-Liquids

E-liquid samples (50 µL) were diluted in AR grade methanol (1450 µL) and analyzed by GC/MS. Analyses were performed using a Perkin Elmer (Melbourne, Victoria, Australia) Clarus 600 GC and Clarus 600 S MS. The GC column type was Phenomonex (Lane Cove, NSW, Australia) ZB-5MS, of 30 m length, 0.25 mm ID and 0.25 µm film thickness. Samples were injected with a 10:1 split. Ultra-pure helium was used as the carrier gas and the GC temperature program was: 120°C hold for 0.5 min; 25°C/min to 260°C; then hold for 4.5 min at 260°C. The total run time was 10.6 min for each injection. The ion source and the transfer line were kept at 180°C. The MS was operated in total ion monitoring mode and positive identification of individual flavor compounds was achieved via the NIST MS library.

Determination of formaldehyde in EVE was determined via HPLC analysis following the addition of excess 2,4-dinitrophenylhydrazine derivatizing agent to produce the stable 2,4-dinitrophenylhydrazone derivative. HPLC-UV analysis was performed using a Perkin-Elmer solvent manager, binary LC pump and UV- Vis absorbance detector controlled by Perkin-Elmer TotalChrom (v6.2.0.0.1) software. Separation was achieved on a Phenomonex Kinetex C18 column (150 × 4.6 mm; 5 μm). pH measurements of EVE were performed using a Thermo Scientific ROSS glass combination micro pH electrode connected to an Orion Dual Star pH/ISE benchtop meter (Beverly, MA).

Spectral Analysis of E-Liquids and EVE as a Measure of Chemical Concentration

Samples were assessed using the spectral scan feature on the Nanodrop One that covers 190–850 nm using water as a blank as not all E-liquids declared their PG/VG ratio and media as a blank for the EVE analysis. There was no absorbance noted past 600 nm. Saturation was observed in the E-liquids for low nm wavelength around 600 absorbance so if necessary, samples were diluted until they were below this saturation point. A linear relationship was observed, so values for undiluted E-liquids were able to be calculated for a semiquantitative comparison of chemicals in each sample. PG and VG absorbance was also checked.

Cell Culture Maintenance

The THP-1 monocytic cell line (ATCC, Manassas, VA, USA) was maintained and differentiated into macrophages as previously described (16). Experiments were carried out within 10 passages.

The 16HBE14o- airway epithelial cell line (16HBE) was a generous gift from Dr. Dieter C. Gruenert (University of California, San Francisco, CA) and were maintained in MEM medium supplemented with L-glutamine (2 mM), Penicillin (12 µg/mL), Gentamycin (16 µg/mL) and 10% FCS. 16HBE were seeded at a density of 0.7 × 10⁶ cells/cm² in serum free media for 24 h before treatment. All cells were routinely screened for mycoplasma.

Collection of Supernatants and IL-1β/IL-8 ELISA

Supernatant was collected for lactate dehydrogenase (LDH) assays and cytokine detection described below. Cell debris were removed by centrifugation at 200 × g for 5 min before being stored at -80°C with the addition of HALT protease and phosphatase inhibitor (Life Technologies). IL-1β and IL-8 were assessed using an ELISA as per manufacturer’s instructions (88–7261 and 88–8086-22, Life Technologies). IL-8 was measured in primary bronchial epithelial cells treated only at EVE doses where the monolayer was intact post treatment.

LDH Assay

The assay was performed as per manufacturer’s instructions (Roche, IN). In a separate well: Tween-20 was added at 2% v/v, mixed and incubated for 5 min to lyse cells as a 100% control of maximal LDH release. Paclitaxel (1 μM) was used as a positive control.

Flow Cytometry Assessment of Necrosis, Apoptosis, and Macrophage Surface Receptors

16HBE cells were trypsinized before incubation in annexin binding buffer (ABB:10 mM HEPES pH 7.4, 150 mM sodium chloride, 5 mM potassium chloride, 1 mM magnesium chloride, 1.8 mM calcium chloride) for 10 min. Cells were pelleted and incubated with 2 µL Annexin V-APC (BD #550475) and 2 µL of 1 µM Sytox Green nucleic acid stain (Life Technologies). For analysis of cell surface phagocytic receptors, macrophages were lifted in ice cold PBS using a bulb pipette. Cells were incubated with 2 µl of primary conjugated antibodies (SR-A1 APC #FAB2708A, MerTK #FAB8912A, and 1/10 TLR-2 PE #FAB6248P R&D systems, MN; 1/10 TLR-4 APC #17–9917-41 EBiosciences, CA; CD44, IM1219U and CD36 IM0766U, Beckman Coulter; CD35 #559872 BD Biosciences) for 10 min in the dark, and washed. Unstained treated macrophages were used for gating controls. 10,000 events were recorded and assessed using FACS DIVA 7.0 and expressed as percentage of positive cells, using previously published gating strategies (8).

Preparation of Primary Bronchial Epithelial Cells

Bronchial epithelial cells from healthy non-smokers (NHBE Normal Human Bronchial/Tracheal CC-2540, Lonza) were cultured as per manufacturers instructions. Cells were seeded at 0.7 × 10⁶ in BEGM and given 24 h to reach confluence to encourage contact inhibition of proliferation, the wells rinsed, and then treated for 24 h with EVE.

Bronchoscopy Human Subjects and Preparation of Alveolar Macrophages

Flexible bronchoscopy with bronchoalveolar lavage (BAL) was performed on healthy, non-smoker, non-vaper control volunteers aged 18–30 with normal spirometry, as previously described (17). Written, informed consent was obtained, and ethics approval granted by the Royal Adelaide Hospital HREC. Macrophages were isolated from BAL by centrifugation and adhesion to plastic for 1 h, as previously described (18). Macrophages were treated with EVE, control media, or CSE for 24 h.

Phagocytosis and Efferocytosis Assay

Apoptotic epithelial cells were prepared and efferocytosis assay performed with a 5:1 target to macrophage ratio as previously described (18). NTHi were prepared and phagocytosis assay performed with a 50:1 target to macrophage ratio as previously described (8). Treated macrophages not exposed to targets were used as a gating control (8) and phagocytosis/efferocytosis was assessed by collecting 10,000 events on a FACSCanto II with FACS DIVA 7.0 (BD Biosciences).

Saliva Collection Human Subjects and Cytokine Analysis

Written, informed consent was obtained, and ethics approval granted by University of California, San Diego (UCSD) IRB. All subjects were18–30 years old. E-cigarette users were defined as subjects who have vaped any type of E-cigarette, consuming a total of ≥ 0.5–1 mL of E-liquid use per day or 3.5–7 mL per week, for a period of at least 6 months. Subjects must not have smoked more than 1 cigarette per month for >6 months. Subjects did not brush their teeth, eat or drink before study visits. Saliva was collected via drooling into a conical tube for 5 min. All subjects met the threshold of ≥ 3 mL. Saliva was centrifuged at 5000 rpm for 5 min at 4°C. Supernatants were mixed with protease inhibitors, snap frozen, and stored at −80°C until cytokine analysis (Milliplex Human Cytokine Magnetic Beads).

Statistical Analysis

Data is presented as Median and range or median and individual data points to show the range. The Kruskal-Wallis non-parametric ANOVA with Mann Whitney U test were employed for statistical analysis for cell line work. The Friedmans non-parametric test with the Wilcoxon signed ranks pairwise test was performed for analysis of the primary AM and NHBE and saliva cytokine levels were assessed by unpaired t-test. Differences from control of P < 0.05 were considered significant, and P values are provided for the human data. All statistics tests were performed in Graphpad Prism 8.

RESULTS

Analysis of E-Liquids

We assessed Mango, Banana, Tobacco, Cappuccino, Chocolate, BubbleGum, Cinnamon, Peppermint, Cherry, and Apple flavored E-liquids. The mass spectrometry analysis of flavorants in E-liquid highlighted how it is not as simple as the flavor written on the bottle. The most identified flavoring compound, Vanillin, was detected in 6 out of 10 E-liquids (Table 1). Chocolate and Cinnamon had the most complex flavoring profile with 14 and 15 flavoring peaks observed in the mass spectrometry analysis, respectively. Tobacco flavor had the least complex flavoring profile with only 2 peaks. All E-liquids tested in this study, were confirmed to be nicotine free by mass spectrometry. pH of all EVE was within 0.6 of control media (data not shown) and formaldehyde in EVE was low at < 1.9 μg/mL (data not shown).

Table 1.

Identification of flavoring compounds in E-liquids by mass spectrometry

Flavor (PG:VG)	Flavor Compounds Positively Identified	Retention Time (RT)	Cas	Number of unidentified peaks	Total Number of peaks
Mango(50:50)Brand 1	2(3H)furanone, 5-hexyldihydro (γ-undecalactone) Gamma-dodecalactone	4.525.02	707-14-92305-05-7	2	4
Banana(50:50)Brand 1	Vanillin Benzaldehyde-3,4-dimethoxy methyl monoacetal Benzyl Benzoate	3.724.145.46	121-33-5900153-10-7120-51-4	6	9
Tobacco(50:50)Brand 1	Vanillin	3.72	121-33-5	1	2
Cappuccino(50:50)Brand 1	Vanillin Caffeine	3.725.78	121-33-558-08-2	5	7
Chocolate(unknown)1Brand 2	Benzyl Alcohol Pyrazine, Tetramethyl Maltol Ethyl Maltol Vanillin Ethyl Vanillin 5-methyl-2-phenyl-2-hexenal Benzaldehyde, 3,4-dimethoxy, methylmonoacetal	2.122.272.42.753.723.974.114.18	100-51-61124-11-4118-71-84940-11-8121-33-5121-32-421834-92-4900153-10-7	6	14
Bubblegum(60:40)Brand 3	Vanillin Ethyl anthranilate	3.723.81	121-33-587-25-2	3	5
Cinnamon(60:40)Brand 3	Ethyl Maltol Cinnamaldehyde Vanillin Coumarin Cinnamic Acid	2.753.133.723.984.48	4940-11-814371-10-9121-33-591-64-5103-53-7	10	15
Peppermint(60:40)Brand 3	Menthol	2.68, 2.75	1490-04-6	3	4
Cherry(60:40)Brand 3	1,3-dioxolane, 4-methyl-2-phenyl Cinnamaldehyde Eugenol 1,3-dioxan-5-ol,2-phenyl- (Benzaldehyde glyceryl acetal)	3.093.133.484.22	2568-25-414371-10-997-53-01708-40-3	1	5
Apple(70:30)Brand 1	Menthol 1,2,3-propanetriol-monoacetate Glycerol- 1,2 diacetate Acetyl Vanillin 2(3H)furanone, 5-hexyldihydro (γ-undecalactone)	2.682.853.293.714.52	1490-04-626446-35-525395-31-7881-68-5707-14-9	7	12

VG caused interference in the 2-2.6 RT range and may have masked the presence of additional small chain flavoring compounds. Italics = no benzene rings, bold = 1 benzene ring, underlined & bold = 2 benzene rings. ¹Mass spectrometry suggests high PG content. PG and VG were purchased from Brand 3.

E-Cigarette Toxicity in Human Bronchial Epithelial Cells is Flavor Dependent

Cellular toxicity was visibly apparent in 16HBE cells treated with Cappuccino, Chocolate, and Peppermint EVE as indicated by morphological changes and cell shedding from the monolayer (Fig. 1). All 10 flavors induced significant LDH release (mean of 26.2–66.0%) versus control (18.2%), with Chocolate the highest and nicotine alone also significantly higher than control (Fig. 2A). Apoptosis with Banana (18.8%), Cappuccino (39.5%), Chocolate (64.5%), and Peppermint (34.7%) EVE treatment was significantly increased versuscontrol (6.7%). Nicotine and the glycol bases did not induce apoptosis (Fig. 2C). Similar results for the flavors were observed for cell necrosis (Fig. 2B). Two flavors with moderate-high toxicity in the majority of our in vitro cell line experiments (Banana and Chocolate) and two with minimal toxicity (Mango and Tobacco) were selected and tested on primary NHBE from healthy controls, showed necrosis (13.2–36.6%) and apoptosis (14.5–41.5%) for all four flavors (control 6.5% and 6.9% respectively) with chocolate the most toxic. Nicotine alone showed significant but lower necrosis and apoptosis (12.1% and 12.9%), but not PGVG alone (Fig. 3).

Figure 1.

Visual evidence of distress caused by E-cigarette vapor extract (EVE) treatment of airway cells. 16HBE cells were treated with control media, 10% cigarette smoke extract (CSE), EVE, or 1 μM Paclitaxel for 24 h. Taken at 10× magnification. Healthy cells maintain a cobblestone monolayer whereas cells treated with the cytotoxic drug Paclitaxel show rounded, detaching cells.

Figure 2.

E-cigarette vapor extract (EVE) causes cellular toxicity, necrosis, and apoptosis. 16HBE cells were treated with control media, 10% cigarette smoke extract (CSE), EVE, or 1 μM Paclitaxel as a cytotoxic control for 24 h. A separate control well was lysed with Tween 20. A: Supernatants were assessed for lactate dehydrogenase (LDH) production and compared to the lysed control well (lysis max). Cells were stained with B: Sytox green and C: APC conjugated Annexin-V to detect necrotic cells and apoptotic cells respectively and the percentage of positive cells assessed by flow cytometry. Data shown is median ± range. n = 4–5. *Significance from control is shown P <0.05 Mann Whitney U test.

Figure 3.

E-cigarette vapor induces necrosis and apoptosis in primary bronchial epithelial cells. Primary bronchial epithelial cells were treated with control media, 10% cigarette smoke extract (CSE), or E-cigarette vapor extract (EVE) for 24 h. Cells were stained with A: Sytox green and B: APC conjugated Annexin-V to detect necrotic cells and apoptotic cells respectively and the percentage of positive cells assessed by flow cytometry. Data shown is the median n = 10–12. Significance from control is shown. P < 0.05 Wilcoxon signed ranks pairwise test.

E-Cigarettes Cause Decreased Macrophage Phagocytic Capacity

Exposure to Banana (15.9%), Chocolate (5.3%), and Apple (18.7%) caused significant decreases in phagocytosis of NTHi as did nicotine alone (20%, Fig. 4A). Exposure to Banana (11.5%), Tobacco (13.4%) Cappuccino (12.3%), Chocolate (4.7%), Bubblegum (14.1%), Peppermint (14.6%) and Apple (13.4%) EVE significantly decreased the efferocytic capacity compared with control (20.2%) as well as nicotine alone (12.9%, Fig. 4B).

Figure 4.

E-cigarette vapor extract (EVE) exposure causes macrophage phagocytic dysfunction. A, B: THP-1 differentiated macrophages or C, D: alveolar macrophages were treated with control media, 10% Cigarette smoke extract (CSE), or EVE for 24 h. Cells were incubated with A, C: pHrodo Red labeled NTHi or B, D: pHrodo Green labeled apoptotic 16HBE cells. Because of pH specificity of pHrodo only cells with intracellular targets are counted as positive on the flow cytometer. Cytochalasin D was used as a negative control as it is known to block the actin rearrangement needed for phagocytosis. Expressed as median percentage positive macrophages ± range. THP-1 n = 5 separate experiments performed in duplicate. *P < 0.05 significance from control, Mann Whitney U test. AMs n = 10 Wilcoxon signed ranks pairwise test *P < 0.05 significance from control.

Mango, Tobacco, Banana and Chocolate EVEs were then tested on primary AM from healthy controls. Mango (15.5%), Banana (10.3%), and Chocolate (3%) EVE treatment significantly reduced phagocytosis of NTHi by AM compared with control (20.3%), as did nicotine alone (14.7%, Fig. 4C). Like the THP-1 derived macrophages, more flavors showed effects on efferocytosis, with all four tested reducing efferocytosis (Fig. 4D). Interestingly the glycol bases PG and VG had a small but significant effect on efferocytosis, as did nicotine alone.

E-Cigarettes Reduce Expression of AM Surface Phagocytosis and Efferocytosis Receptors

We assessed the effects of E-liquids with differing flavorant profiles on the phagocytic and efferocytic target recognition receptors of healthy AM (Fig. 5). Mango (12.3%), Banana (9.3%), and Chocolate (4.1%) EVE reduced the surface levels of bacterial recognition receptor TLR-2 versus control (25.4%). Mango (50.9%) and Chocolate (14.6%) significantly decreased surface expression of bacterial recognition receptor, TLR-4 versus control (66.6%). CD44, which can recognize both bacteria and apoptotic cells was significantly reduced below control by Banana (65.5%) and Chocolate (28.1%). Only Chocolate EVE significantly reduced surface expression of dual recognition scavenger receptor SR-A1 (63.8% versus 77% control), whereas Banana (5%) and Chocolate (5.4%) significantly decreased surface expression of dual recognition scavenger receptor CD36/SR-B3 versus control (13.1%).

Figure 5.

E-cigarette vapor extract (EVE) exposure reduces AM bacterial and apoptotic cell recognition receptors. AM isolated from healthy non-smoking controls were treated with control media, 10% cigarette smoke extract (CSE), or EVE for 24 h. Surface presentation of A: TLR-2 B: TLR-4 C: CD44 D: SR-A1 E: CD36 F: CD35 G: MerTK were assessed by flow cytometry. Data shown as median with individual data points. n = 6–13 and not every donor was exposed to every treatment because of cell number restrictions. Wilcoxon signed ranks pairwise test P < 0.05 significance from control.

E-Cigarettes Increase Airway Epithelial Cell Secretion of IL-8 but Reduce AM Secretion and Airway IL-1β Levels

We found that IL-8 secretion by primary bronchial epithelial cells was increased compared with control after exposure to Banana (150.2%) and Chocolate (181.7%) EVE, and nicotine (143.6%) (Fig. 6A). We observed significant decreases in IL-1β secretion by THP-1 differentiated macrophages exposed to all flavored EVEs (57.1–78.3% of control) except for Mango and Tobacco (Fig. 6B). Nicotine (74.9%) alone also induced a significant decrease in the level of secreted IL-1β from control (267 pg/mL). AM had a greater than 10-fold lower secretion of IL-1β than THP-1 differentiated macrophages. All data is therefore expressed as percentage of control for easier comparison. We observed that all 4 flavors tested significantly reduced secretion of IL-1β (37.4–55.4% of control) as well as nicotine alone (55.9% of control) (Fig. 6C). Reduced IL-1β levels in saliva from E-cigarette users versus never-smoker never-vaper controls validated these results (Fig. 6, D–F). IL-1β is known to affect a myriad of inflammatory pathways, leading to changes in other cytokines, such as IL-1α and IL-6. Assessment of these cytokines in the saliva of human subjects found significant concurrent reductions. IL-8 was also analyzed and whereas 10/17 E-cigarette users showed higher levels than the control median and 9/17 above control average there was no significant difference found.

Figure 6.

E-cigarette vapor extract (EVE) exposure alters cytokine secretion by lung cells and in human saliva. A: IL-8 secretion by healthy control primary bronchial epithelial cells. IL-1β secretion by B: THP-1 differentiated macrophages and C: AM from healthy non-smoking controls were treated with control media, 10% cigarette smoke extract (CSE), or EVE for 24 h. Data is expressed as % of control. n ≥ 6 for primary bronchial epithelial cells, Wilcoxon signed ranks pairwise test, n = 5 for THP-1 cells; Mann Whitney U Test. n = 9 donors for AM; Wilcoxon signed ranks pairwise test. P < 0.05. Cytokines in saliva samples from never-smokers, never-vapers (Control) or active E-cigarette users (E-cig) were quantified by multiplex. D: IL-1β E: IL-1α and F: IL-6 G: IL-8 salivary levels. n = 11 for control subjects and n = 16 for E-cig subjects. Data is represented as the median. *P = 0.027, **P < 0.01, by unpaired two-tailed t test, with F testing to compare variance, and normality testing.

Effects Are Linked to Concentration and Benzene-Ring Containing Chemicals

Additional toxicity testing with an additional 3 Banana and 3 Chocolate brands showed similar but slightly less toxicity with all Bananas, and similar but slightly less toxicity with 2/3 Chocolate EVEs. The last Chocolate EVE (Chocolate 3) showed much lower toxicity (Fig. 7). Additional testing of phagocytosis and efferocytosis with an extra 3 Banana and 3 Chocolate EVEs showed similar but slightly less phagocytic defects with all Bananas, and similar phagocytic defects with 2/3 Chocolate EVEs. The last Chocolate EVE (Chocolate 3) showed significantly less phagocytic dysfunction (Fig. 8).

Figure 7.

Banana and Chocolate E-cigarette vapor extract (EVE)-induced airway epithelial cell toxicity. 16HBE cells were treated with control media, 10% cigarette smoke extract (CSE), EVE, or 1 μM Paclitaxel for 24 h. A separate control well was lysed with Tween 20. A: Sytox green and B: APC conjugated Annexin-V to detect necrotic cells and apoptotic cells respectively and the percentage of positive cells assessed by flow cytometry. Data shown is median ± range, n = 5. **Significance from control is shown P <0.01 Mann Whitney U test.

Figure 8.

Banana and Chocolate E-cigarette vapor extract (EVE) effects on phagocytosis. THP-1 derived macrophages were treated with control media, 10% cigarette smoke extract (CSE), or EVE for 24 h. Cells were incubated with A: pHrodo Red labeled NTHi or B: pHrodo Green labeled apoptotic 16HBE cells. Because of pH specificity of pHrodo only cells with intracellular targets are counted as positive on the flow cytometer. Expressed as median percentage positive macrophages ± range. n = 5. *P < 0.05, **P < 0.01 significance from control, Mann Whitney U test.

Assessment of the 3 additional brands of Banana and Chocolate E-liquids by mass spectrometry showed complex flavoring profiles (Table 2). Overall, some of the 16 E-liquids used in this study showed a high proportion of benzene-ring containing flavorants, found Chocolate 3 contained a reduced number of benzene-ring containing flavorants. A spectral wavelength scan used to perform comparative chemical concentration analysis of the E-liquids showed Chocolate had very high levels of absorbance, indicating higher concentration of flavoring chemicals, whereas Banana and Cappuccino had high levels of absorbance, and peppermint had very low absorbance (Fig. 9).

Table 2.

Identification of flavoring compounds in E-liquids by mass spectrometry

Flavor (PG:VG)	Flavor Compounds Positively Identified	Retention Time (RT)	Cas	Number of unidentified peaks	Total Number of peaks
Banana 2(80:20)Brand 4	4H-pyran-4-one, 2-ethyl-3-hydroxy2-methylnapthalene2(3H)-furanone, dihydro-5-pentyl (γ-nonalactone)VanillinBenzaldehyde-3,4-dimethoxyBenzaldehyde-3,4-dimethoxy methyl monoacetalEthyl citrate	2.7093.253.453.694.0014.144.776	4940-11-891-57-6104-61-0121-33-5120-14-9900153-10-777-93-0	7	7
Banana 3(80:20)Brand 4	Piperonal Vanillin Ethyl vanillin Ethyl citrate Veratraldehyde propylene glycol acetal Decanedioic acid, diethyl ester	3.403.653.924.7765.215.35	120-57-0121-33-5121-32-477-93-0900109-36-0110-40-7	8	14
Banana 4(80:20)Brand 4	Alpha-methyl 1H-indene-1-methanol-acetate Ethyl vanillin Benzoic acid, 4-hydroxy-2,6-dimethoxy-, methyl ester Veratraldehyde propylene glycol acetal	3.243.924.385.19	63839-85-0121-32-4900319-13-5900109-36-0	4	5
Chocolate 2(70:30)Brand 5	Benzyl Alcohol Pyrazine tetramethyl Maltol 4H-pyran-4-one, 2-ethyl-3-hydroxy Vanillin Ethyl vanillin Methyl (3,4-dimethoxyphenyl)(hydroxyl)acetate	2.072.282.362.713.693.944.15	100-51-61124-11-4118-71-84940-11-8121-33-5121-32-4900333.50-3	4	11
Chocolate 3(80:20)Brand 4	4H-pyran-4-one, 2-ethyl-3-hydroxy 1,2,3-propanetriol, 1-acetate Triacetin 2(3H)-furanone, dihydro-5-pentyl Vanillin	2.712.803.243.453.65	4940-11-8106-61-6102-76-1104-61-0121-33-5	2	7
Chocolate 4(50:50)Brand 1	2-methylnapthalene 2(3H)-furanone, dihydro-5-pentyl Vanillin Benzaldehyde-3,4-dimethoxy methyl monoacetal	3.253.453.654.13	91-57-6104-61-0121-33-5900153-10-7	7	11

VG caused interference in the 2–2.6 RT range and may have masked the presence of additional small chain flavoring compounds. Italics = no benzene rings, bold = 1 benzene ring, underlined and bold = 2 benzene rings.

Figure 9.

Spectral scan of unvaped E-liquids. Samples were diluted as needed until they were below the saturation point of the equipment (500 absorbance units) and then because of the linear nature of optical density readings, were adjusted to represent their neat values. Spectral scans of E-liquids: A: PG and VG, B: the main 10 flavors, C: Chocolate and D: Banana or extracts E: propylene glycol (PG), vegetable glycerin (VG), cigarette smoke extract (CSE) F: 10 main flavors G: Chocolate and H: Banana. Peppermint, which was the lowest reading flavor was included in all graphs as a reference point.

DISCUSSION

Although E-cigarettes are relatively new to the market, many research articles now show that they are cytotoxic to a range of cells (16) including airway epithelial cells (5, 19) and there has been particular interest in flavors showing increased toxicity including: Tobacco, Coffee, Cherry, and Cinnamon (11–14). Flavoring has also been linked to increased appeal to youth (9), and youth vaping has been associated with a subsequent uptake of smoking (10, 20, 21). Thus, E-cigarette flavoring has recently received a lot of attention, globally, from policymakers. However, it is important to consider that the health impacts of a “flavor” is not simply determined by what flavor is on the label. Previous studies including ours have shown that E-liquids rarely contain just one flavorant (7, 22–24).

As the ingredients are rarely provided on a bottle of E-liquid, we assessed which flavorants were present in these liquids by GC/MS. Analysis revealed many peaks from the mass spectrometry that could not be positively matched against a reference library, which may be because of ability of flavorings and to form new chemicals (25, 26, 89). E-liquids tested ranged from a simple 2 chemical profile to 15 flavorants. Chocolate, which showed the greatest effects in our study, had a complex flavorant profile, whereas Tobacco and Mango, which had the least effects, had simple flavorant profiles. Similarly, Apple and Mango both contained 2(3H)furanone, 5-hexyldihydro (γ-undecalactone), but Apple, which had greater effects than Mango, had a more complex flavorant profile. The complexity of the mix of flavors may be a factor involved in whether an E-liquid will have negative health impacts, supported by a linked increase in general toxicity in pulmonary fibroblasts and a lung epithelial cell line (11, 23).

Our wavelength scan profile of E-liquids versus EVE, which shows different peak profiles most likely from one or more of the following: generation of new chemicals, alteration of the existing chemicals, or alteration from interaction with the media supports the data of others (23, 27–30, 89) showing alteration of E-liquid chemicals or generation of new chemicals during vapor generation. It is unclear whether increased complexity of flavorant profiles leads to more interactions between chemicals that may further result in changed chemical profile. This is an area that is difficult to study because of the number of combinations that would need to be tested for each E-liquid.

Flavorants whose chemical composition included at least one benzene ring were common, and some flavors contained multiple Benzene-ring based-flavorants (Banana, Chocolate, Cinnamon, Cherry). Inhalation of benzene derivatives is known to lead to a range of toxicities in the body via their reactive nature leading to increased oxidative stress and production of benzene metabolites and lipid peroxidation products (31), the latter of which has also been shown in E-cigarette vapor exposed mice (32). Benzene has further been shown to affect cell cycle (33), are linked to increased apoptosis (34, 35), and DNA fragmentation (36), and also shown to affect macrophage phagocytosis (25, 37, 38). Benzene formation from vaporization of E-liquids with benzene ring containing flavorants is lower than from burning cigarettes (39). This potentially explains why most E-liquids are less impactful than cigarette smoke. Benzene exposure of airway epithelial cells produced increased PGE (36), a marker of inflammation, similar to IL-8 increases, which we observed in our study after exposure to benzene-ring rich E-liquid EVE. Benzene exposure has also been associated with reduced macrophage IL-1α and IL-1β levels via reduced activation/cleavage of their pro forms (40), which was observed after exposure of macrophages in this study, as well as saliva from vapers. Additionally, Benzene ring containing flavorants were recently shown to interact with PG to form new acetal products at room temperature during storage of the E-liquids. These were shown to be toxic to BEAS-2B and A549 lung epithelial cells, causing increased mitochondrial stress, and reduced DNA production/cellular proliferation. These negative cellular effects were greater for the new acetals than the parent flavorants (89). Therefore, benzene-ring containing flavorants may be one factor involved in the mechanism for why Banana and Chocolate showed higher apoptosis and phagocytic defects. This theory is further supported by the fact that the less impactful Chocolate E-liquid contained less benzene-ring flavorants. Cherry and Cinnamon also have a history in the literature of being very toxic, which some authors also attributed to flavorants (4, 13, 41–43), but were not toxic in our study, whereas Cappuccino showed moderate toxicity. The comparative spectral wavelength scans showed that Chocolate, Banana, and Cappuccino had the highest absorbance, indicating higher overall chemical concentration, whereas Cherry and Cinnamon had low absorbance. Previous studies have shown that in some E-liquids, flavorants were present at concentrations over 10 mg/mL and Cinnemaldehyde was over 150 mg/mL in some samples (23, 44). Thus, it is probable that the concentration of flavorants is another factor to consider as to whether an E-liquid will have negative health impacts. Further, it has been shown that the concentration of each flavorant regularly varies between E-liquids, up to 10,000-fold (44). The method of using laboratory made E-liquids using single flavorants to test toxicity ignores the interactions the chemicals may have once mixed, heated, and vaporized.

Bronchial epithelial toxicity as assessed by membrane integrity via LDH release showed significant increases for all 10 flavors. LDH is a marker of membrane injury, and is often very sensitive to early signs of cellular distress, therefore we also assessed levels of necrosis and apoptosis. These findings showed flavor specific effects that were nicotine independent. This is consistent with findings from other studies looking at the cytotoxicity of EVE (45–48), many of which also show flavor specific toxicity (4, 13, 42, 49, 50). As 16HBE cells are immortalized by simian virus-40, which is known to increase their susceptibility to apoptosis (51), we validated this data using primary bronchial epithelial cells. The primary bronchial cells appeared to be more sensitive to the EVE as all 4 flavors increased both apoptosis and necrosis, suggesting one mechanism for the toxicity observed in previous studies is increased apoptosis. Studies have shown gingival cells (52) and endothelial cells (53) exposed to E-cigarettes showed increased apoptosis induction via the caspase-3 pathway. However, both studies included nicotine and used assays that detected combined necrosis and apoptosis, and only assessed tobacco flavoring. As formaldehyde has been raised as a concern in E-cigarettes we assessed its levels in freshly generated EVE. The levels for all EVE tested was very low at <1.9 μg/mL. previous papers have reported similar or higher levels (54–56). These levels were also well below the 210 μg/mL and 42 μg/mL used in a 24 h exposure to induce toxicity in 16HBE and Calu-3 lung cells respectively (57).

In healthy lungs, apoptotic epithelial cells would be cleared away by AM to prevent inflammation. We thus assessed efferocytosis of apoptotic bronchial epithelial cells by THP-1 differentiated macrophages and AM exposed to EVE. We showed that 7 of 10 flavors tested caused a significant reduction in efferocytosis by THP-1 differentiate macrophages, and all 4 flavors reduced efferocytosis by AM. This effect was flavor dependent and nicotine alone also significantly reduced efferocytosis. Both glycol bases showed a small but significant decrease in efferocytosis, showing that the glycol bases PG and VG may not be as benign as believed. This study provides the first evidence that AM efferocytic capability is reduced by E-cigarettes. Given many of the flavors that reduced efferocytosis also increased apoptosis of airway cells, there is cause for concern that long-term use may result in a build-up of apoptotic material in the airways like that seen in smokers. Further studies in the airways of long-term vapers are thus warranted.

NTHi bacteria is common in airway related disease states (58–60) and represents a clinically relevant model for bacterial clearance by AM. Limited testing by us and others have shown that exposing THP-1 differentiated macrophages to a limited range of flavored EVE caused reduced bacterial phagocytosis (8, 61, 62). We observed reduced phagocytosis for 3 out of 10 flavors in THP-1 differentiated macrophages and for all but tobacco EVE as well as for nicotine alone in AM from healthy non-smoking non-vaping controls.

To explore how the cells had decreased phagocytosis of bacterial and apoptotic cells, we assessed whether EVE treatment altered specific phagocytic and efferocytic surface receptors on healthy AM. We observed decreased TLR-2, TLR-4, CD44, CD36, and SR-A1 with exposure to specific flavored EVE. This is the first study that shows that E-cigarettes can alter the expression levels of surface receptors for both bacteria and apoptotic cells on AM that could lead to reduced phagocytic clearance in the lungs. Some studies have also shown that SR-A1 deficient mice were more susceptible to Streptococcus pneumoniae and Staphylococcus aureus infections (63, 64), suggesting that E-cigarette use may lead to increased bacterial lung infections because of decreased AM surface receptors. Cigarette smoke exposure has been shown to reduce SR-A1 (65) and SR-B1 (CD36) (66, 67), Hyaluronan receptor CD44 (17, 68), CD31 (17), and LRP-1/CD91 (17) and bacterial phagocytosis receptors TLR-2 (65, 69), and TLR-4 (65, 69). Our data provides a probable mechanism for studies that show greater survival of MRSA (61) and colonization with Streptococcus pneumoniae (62) in mice exposed to E-cigarette vapor. An additional study using pure, non-vaped flavoring compounds, showed a flavor specific decrease in phagocytosis by neutrophils (70).

Bronchial epithelial cells are a key mediator of the inflammatory milieu. IL-8 is an excellent measure of the inflammatory response (71, 72). As such, we measured primary bronchial cell IL-8 secretion after treatment with the maximal EVE doses that retained the monolayer. 10% CSE induced IL-8 secretion as per previous studies (73–75). Banana and Chocolate also increased IL-8 secretion whereas Mango, Tobacco, and PGVG base alone did not. A search of the literature found that IL-8 secretion by airway epithelial cells varied after exposure to E-cigarette vapor (4, 46, 76–81) that given our own findings, is most likely related to flavor profiles of the E-liquids tested. This is one of the biggest hurdles facing comparability of E-cigarette research, and why the authors believe strongly in publishing mass spectrometry flavor profiles in all E-cigarette studies. When IL-8 was assessed in the saliva, an upper airway only sample, there was no significant increase in the E-cig user group versuscontrols, despite a majority of users having a higher level of IL-8 than the control median or average. Our NHBE IL-8 data suggests that increases may be very flavor specific in comparison with other cytokines, and our E-cigarette users were allowed to use their own flavors. This may mean that any significant increases from those using flavors that could cause an increase may have been masked by those using flavors that did not induce an increase.

AM are also a key source of cytokines involved in the inflammatory process. IL-1β in particular plays a vital role. IL-1β production is triggered through NLRP3 inflammasome mechanisms to initiate immune regulation (18, 82). IL-1β can affect the permeability of the underlying vascular endothelium and lead to an increase in secreted proteins into the airway tissues promoting pulmonary tissue pathology (83). Decreased IL-1β was also observed in PBMC exposed to nicotine (84). TLRs are known to play a role in inflammasome activation and secretion of IL-1β (85) and TLR-4 deficient mice were shown to have reduced neutrophil recruitment in response to cigarette smoke as well as reduced IL-1β (86), consistent with our findings of reduced AM TLR-4 and Il-1β secretion after EVE exposure. However, as this is an in vitro culture system, and AM do not exist in isolation, we assessed the levels of IL-1β in the saliva of E-cigarette vapers versus controls. We observed similar decreases of IL-1β in the saliva of E-cigarette vapers. As IL-1β is known to affect a myriad of inflammatory pathways, leading to changes in other cytokines, such as IL-1α and IL-6, these were also assessed in the saliva where significant reductions were observed. The TLR-4 deficient mouse study also found reduced IL-6 (86) and our own previous study showed that THP-1 derived macrophages had reduced secretion of IL-6 after exposure to Apple EVE (8). Some studies showed that IL-6 secretion was increased in lung cells after E-cigarette exposure (4, 46, 87), whereas others showed that cytokines including IL-6 were decreased in the BAL of E-cigarette exposed mice (4, 62). Our data shows that E-cigarettes can have both pro-inflammatory and anti-inflammatory effects depending on the cell type. Future studies should further examine saliva and BAL of healthy controls compared with vapers using the same E-liquids to better assess physiological cytokine changes in the distal airways in response to E-cigarette use.

There are limitations to our study, in that we used in vitro or ex vivo systems, and bronchial epithelial cells and AM do not exist in isolation. We also are aware that E-liquids vary from store to store and country to country, so comparisons of flavors used in studies in different laboratories are not always comparable.

Conclusions

This data combined with our previous work (8) and that of others (24, 27, 88) highlights the fact that many flavorants are used to achieve the flavor name on the bottle, and that E-liquids labeled as the same flavor from different companies are unlikely to contain the same mixture or even concentration of flavorants. The recently proposed flavor ban in the United States highlights the need for thorough studies on the health risks presented by flavorings and what aspects of the E-liquid actually contribute to negative health impacts, not just the appeal of the flavor on the label so that any policy changes related to flavor regulation that are proposed give clear definitions based on safety data. Our data makes it clear that flavorants present, their concentration, and the complexity of the flavor profile all contribute to the level of negative health impacts an E-liquid may represent to the user. This study combined with the data from others may raise a new concern for higher risk for E-cig users who may prefer stronger flavored E-liquids.

Further, our study, coupled with earlier studies, serve to highlight the fact that there is risk even from nicotine free E-cigarette use. Even though not all flavors may be as high-risk to each other, it should not be assumed that GRAS food flavors are safe for vaporization and inhalation.

Our data also adds to the growing literature showing that E-cigarettes can damage cells in the airway where exposure is most direct and lead to immune dysfunction, impacting alveolar macrophages that are the professional phagocytes in the lung responsible for warding off bacterial infection. These effects did generally appear to be lower than that of cigarette smoke, but the high effects seen for most of the Chocolate E-liquids, suggest that E-cigarettes should not be blindly assumed to be vastly less dangerous than cigarette smoke, as the possibility exists that some of the many thousands of E-liquids for sale contain the wrong flavorants at the wrong concentration that may lead to negative health impacts like those seen with cigarette smoke. Overall, our study highlights that flavors need to be considered in addition to nicotine as well as the extent that an E-liquids flavorant profile plays when assessing the risk of E-cigarettes.

GRANTS

Dr. Ween has received funding from the Royal Adelaide Hospital for this work.

Dr. Crotty Alexander received funding from the American Heart Association (6BGIA27790079), National Institutes of Health (NHLBI R01HL147326), American Thoracic Society Foundation Award for Outstanding Early Career Investigators, and the Tobacco-Related Disease Research Program (T30IP0965).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.P.W., C.M.B., L.E.C.A., L.T., and N.A.B. conceived and designed research; M.P.W., C.M.B., J.S., P.N.R., A.M., L.T., N.A.B., R.H., A.B., P.T.N., K.H., and H.J. performed experiments; M.P.W., C.M.B., J.S., L.E.C.A., A.M., L.T., N.A.B., R.H., and K.H. analyzed data; M.P.W., C.M.B., L.E.C.A., S.J.H., L.T., and R.H. interpreted results of experiments; M.P.W., J.S., A.M., and L.T. prepared figures; M.P.W. and L.T. drafted manuscript; M.P.W., C.M.B., J.S., P.N.R., L.E.C.A., S.J.H., A.M., L.T., N.A.B., R.H., A.B., P.T.N., K.H., and H.J. edited and revised manuscript; M.P.W., C.M.B., J.S., P.N.R., L.E.C.A., S.J.H., A.M., L.T., N.A.B., R.H., A.B., P.T.N., K.H., and H.J. approved final version of manuscript.