Gene Expression Alterations in the Bronchial Epithelium of e-Cigarette Users

Abstract

Background

Although e-cigarette (ECIG) use has increased in the United States, their potential health effects remain uncertain. Understanding the effects of tobacco cigarette (TCIG) smoke on bronchial airway epithelial gene expression have previously provided insights into tobacco-related disease pathogenesis. Identifying the impact of ECIGs on airway gene expression could provide insights into their potential long-term health effects. We sought to compare the bronchial airway gene-expression profiles of former TCIG smokers now using ECIGs with the profiles of former and current TCIG smokers.

Methods

We performed gene-expression profiling of bronchial epithelial cells collected from current TCIG smokers (n = 9), current ECIG users who are former TCIG smokers (n = 15), and former TCIG smokers (n = 21). We then compared our findings with previous studies of the effects of TCIG use on bronchial epithelium, as well an in vitro model of ECIG exposure.

Results

Among 3,165 genes whose expression varied between the three study groups (q < 0.05), we identified 468 genes altered in ECIG users relative to former smokers (P < .05). Seventy-nine of these genes were up- or down-regulated concordantly among ECIG and TCIG users. We did not detect ECIG-associated gene-expression changes in known pathways associated with TCIG usage. Genes downregulated in ECIG users are enriched among the genes most downregulated by exposure of airway epithelium to ECIG vapor in vitro.

Conclusions

ECIGs induce both distinct and shared patterns of gene expression relative to TCIGs in the bronchial airway epithelium. The concordance of the genes altered in ECIG users and in the in vitro study suggests that genes altered in ECIG users are likely to be changed as the direct effect of ECIG exposure.

Since their introduction in 2007, the use of e-cigarettes (ECIGs; also known as a “vape” or vaporizer. Member of a class of devices known as Electronic Nicotine Delivery Systems, which are designed to deliver nicotine in an aerosolized mixture) in the United States has increased. Survey data indicate that 15.4% of adults 18 and older have tried ECIGs, with adults between the ages of 18 and 24 being the most likely to have used c at 23.5%.¹ Current smokers and those who had recently quit were more likely to use ECIGs than never smokers. And, although ECIGs are not approved by the US Food and Drug Administration for cessation, 55.4% of conventional tobacco cigarette (TCIG) smokers who tried to quit smoking in the previous year had tried an ECIG and 22% continued to actively use ECIGs.²

Despite the increased use of ECIGs,³ their safety profile remains controversial. Popular perception and marketing of ECIGs assert that they are a safer alternative to TCIGs.^{4, 5} In contrast, numerous studies have illuminated potential adverse health effects of ECIGs such as diminished cough reflex,⁶ cytotoxicity from ECIG flavorings,⁷ endothelial disruption,⁸ and DNA damage.⁹ Although a switch from TCIGs to ECIGs results in a reduction of users’ exposure to known toxicants and carcinogens,¹⁰ it is difficult to reach a consensus on the safety of ECIGs other than that their health risks likely lie somewhere between TCIG smoking and nonsmoking.¹¹

It is well-established that TCIG use leads to gene expression changes in airway epithelium. We have previously described these changes in bronchial and small airway epithelium of TCIG smokers^{12, 13, 14, 15} as well as how these patterns change after smoking cessation.^{16, 17} TCIG-associated gene expression changes in the nasal epithelium overlap with changes seen in the bronchial epithelium demonstrating common TCIG-associated gene expression effects throughout the airway.¹⁸ Changes in airway gene expression have also been associated with lung diseases such as lung cancer^{19, 20} and COPD.²¹

Prior work suggests that ECIG exposure can also alter airway gene expression. We have described oxidative and xenobiotic stress associated gene-expression changes in an in vitro model of ECIG exposure, as well as ECIG dose-response changes in a marker of reactive oxygen species.²² Work by Martin et al²³ measured the expression of a target panel of immune and inflammation associated genes in cells collected from nasal brushings of TCIG smokers, ECIG users, and nonsmokers, finding that most of these genes were specifically downregulated by ECIG use.

To more comprehensively characterize the effects of ECIG use on airway gene expression and to better understand their relationship with the effects of TCIGs, we sought to compare the transcriptome-wide impact of ECIGs and TCIGs in the bronchial airway epithelium. More specifically, we sought to investigate the gene-expression effects of ECIG use in former smokers because daily ECIG users are more likely to be former TCIG smokers than dual-users or de novo users.²⁴ Toward this end, we recruited current TCIG smokers, former TCIG smokers using ECIGs, and former TCIG smokers to undergo voluntary bronchoscopy. Bronchial epithelial cells were collected to identify gene-expression changes associated with ECIG use and to compare the transcriptomic effects of ECIG and TCIG use.

Materials and Methods

Study Population and Sample Collection

We recruited volunteers at Boston University Medical Center and the University of California Los Angeles Medical Center between December 2013 and March 2015. All participants were aged 18 to 55 years and were current or former TCIG smokers with a smoking history of at least five cigarettes per day for at least 2 years. Participants were excluded if they were using tobacco products such as chewing tobacco, snuff, hookah, or marijuana. Participants were also excluded if they were using intranasal or inhaled medications or had a history of chronic lung disease or lung cancer. The remaining participants were enrolled and stratified into three study groups based on their cigarette use behavior at study recruitment: (1) TCIG group (n = 9) defined as ongoing TCIG use with a minimum of five cigarettes daily and less than two lifetime ECIG uses; (2) ECIG group (n = 15) defined as former TCIG smokers who have been tobacco abstinent for a minimum of 3 months before study recruitment and have been using any brand or generation of ECIGs, with any brand of nicotine-containing liquid, with any type of vehicle or flavoring, at least 6 days per week for at least 1 month; and (3) former group (n = 21) defined as former TCIG smokers who had been tobacco abstinent for a minimum of 3 months before recruitment and were not using any form of nicotine replacement therapy. Urine cotinine levels were measured at baseline via the NicAlert assay to determine whether participants were using nicotine-containing products. Carbon monoxide (CO) levels were also measured in all ECIG and former smokers to rule out concurrent TCIG smoking (CoVita Bedfont Scientific piCO + Smokerlyzer Breath CO Monitor); two self-reported ECIG users were excluded because of high exhaled CO levels. A detailed TCIG and ECIG use history was obtained for each participant. This study was approved (approval number H -32129) by the Boston Medical Center institutional review board #1 (Panel Green). All study participants provided written informed consent.

Bronchial airway epithelial cells were obtained from brushings of the right mainstem bronchus collected during fiberoptic bronchoscopy with an endoscopic cytobrush (Cellebrity Endoscopic Cytology Brush, Boston Scientific). The brushes were immediately placed in 1 mL of RNAprotect Cell Reagent (Qiagen) and kept at –80^oC until RNA isolation was performed.

Microarray Data Acquisition and Data Preprocessing

Total RNA was isolated using the miRNeasy Mini Kit (Qiagen). RNA integrity was assessed by Agilent BioAnalyzer, and RNA purity confirmed using a NanoDrop spectrophotometer. A total of 100 ng of isolated RNA was processed and hybridized to Affymetrix Human Gene 1.0 ST Arrays (Affymetrix). Probeset normalization was performed using the Brainarray EntrezGene CDF v17.0.0 and Robust Multiarray Average.²⁵ Statistical analysis was performed with R, version 3.2.2. All microarray data and relevant clinical data are freely available through Gene Expression Omnibus under accession GSE112073.

Microarray Preprocessing and Quality Control

Microarray quality was assessed using relative log expression, normalized unscaled SE, and principal component analysis metrics and all samples were suitable for subsequent analysis. Batch effects were corrected using ComBat²⁶ with a smoking status covariate.

Differential Expression Analysis

We first identified genes differentially expressed between any of the study groups (ECIG, TCIG, and former) via analysis of covariance (ANCOVA). Gene expression was modeled as a linear function of smoking status while adjusting for age, RNA integrity, and months since last TCIG. A nested F test was used to identify genes differentially expressed between ECIG users relative to former TCIG smokers, TCIG smokers relative to current TCIG smokers, or current TCIG smokers relative to ECIG users. Resulting P values were adjusted to control the false discovery rate (FDR) using the Benjamini-Hochberg method.²⁷ We used principal component analysis on the log2-expression level of genes meeting these criteria to isolate the first two principal components (ie, the two largest sources of variance) in the expression of these genes.

Using the linear model described, we performed additional linear modeling with LIMMA^{28, 29} on the genes identified as being differentially expressed in TCIG smokers vs former smokers or ECIG users vs former smokers as identified in the ANCOVA step. Genes demonstrating differential expression in ECIG users relative to former smokers were identified via the ECIG coefficient’s moderated t test P < .05. These genes were divided into subclusters via Ward hierarchical clustering.

Functional Enrichment

We performed functional enrichment analysis on the gene clusters identified during differential expression via Enrichr.³⁰ Only enrichment terms with an FDR-adjusted P < .05 were considered.

Gene Set Variation Analysis

To analyze the similarity of the gene-expression effects of ECIGs and TCIGs, gene-expression pathways associated with TCIG use based on previously published data were projected into the study groups of the cross-sectional cohort using gene set variation analysis (GSVA).³¹ Microarray data from Beane et al,¹⁶ which compared current, former, and never cigarette smokers, were scored for gene set activation for every gene set in the C2, C5, C7, and Hallmark collections of the Molecular Signatures Database.³² Gene sets whose GSVA scores were significantly different between current smokers and noncurrent (former and never) smokers (Student t test; FDR q < 0.05) in this dataset were categorized as differentially activated because of TCIG usage. These gene sets were then scored in the microarray data of the present study via GSVA. These GSVA scores were then compared between study groups via Student t test to identify differences in gene set activity.

Quantitative Real-Time Polymerase Chain Reaction

We profiled ADM, PGAM5, NCK2, and RSPH1 in 15 ECIG users and 15 former smokers with quantitative real-time polymerase chain reaction (RT-PCR). These assays were performed with SYBR Green-based RT² qPCR Primer Assays (Qiagen). Primers for each candidate gene and the 18S ribosomal subunit as an endogenous control gene were designed and experimentally verified by Qiagen to ensure uniform and high PCR efficiencies under standardized amplification conditions. All RT-PCR experiments were carried out in triplicate on each sample, relative gene expression levels were calculated using the comparative Ct method,³³ and differential expression was performed via Student t test on the average expression across these replicates.

Comparison With In Vitro and Immune/Inflammatory Response Dataset

We compared our observations from the study cohort with an in vitro model of ECIG exposure, which profiled the gene expression effects of exposure to vapor from a single brand of disposable ECIGs on bronchial epithelial cells grown at an air-liquid interface.²² Using data from an in vitro model of ECIG exposure, we generated a ranked list of genes differentially expressed between ECIG aerosol and air-exposed human bronchial epithelial cells via a Student t test, and ranked genes by their corresponding t statistic. The genes found to be associated with ECIG use in the current study were split into gene sets based on the clustering results, and gene set enrichment analysis (GSEA)³⁴ was used to determine if these gene sets were significantly enriched toward the bottom or top of the in vitro ranked list.

Results

Subject Characteristics

There were no significant differences in age, sex, race, or pack-years among the three study groups (TCIG, ECIG, and former TCIG smokers) (Table 1). There was a statistically significant difference in time since TCIG cessation between the former smoker and ECIG groups (Student t test P < .05). CO levels for all ECIG users and former smokers included in the final analysis were < 7 ppm. Urine cotinine levels in the TCIG and ECIG groups confirmed active nicotine use (≥ 100 ng/mL) and were significantly higher than the former smoker group (Student t test P < .001). Patterns of ECIG use, including frequency, nicotine dosage, generation of product, and product brand varied across ECIG users.

Table 1.

Demographics of Cross-sectional Study Participants

Category	TCIG (9)		Former (21)		ECIG (15)		P
Sex	Male	Female	Male	Female	Male	Female	.45
	6	3	11	10	11	4
Age, y	42.2 ± 11.3		43.0 ± 10.7		35.7 ± 10.4		.12
Race	AA: 4 W: 5		M: 2 AA: 5 W: 14		AI: 1 AA: 3 W: 11		.54
Pack-y	13.4 ± 11.0		10.9 ± 10.5		13.8 ± 11.3		.70
Time since quit, mo	NA		67.0 ± 117		8.7 ± 4.4		.03
CO, ppm	NA		2.3 ± 1.5		1.9 ± 1.5		.40
Urine cotinine,a ng/mL	5.87 ± 0.35		1.23 ± 0.59		5.25 ± 1.16		< .001

P values for sex and race were calculated by Fisher exact test. P values for age, pack-y. Urine Cotinine were calculated with analysis of variance. P values for Time since quit and CO were calculated by a Student t test. AA = African American; AI = American Indian; CO = carbon monoxide; M = multiple; NA = not available; W = white.

^aAvailable for Boston University samples only. Measured in urine. Test ranges from (1-6 ng/mL).

Airway Gene Expression in Former TCIG Smokers Who Currently Use ECIGs Is More Similar to Former TCIG Smokers Than to Active TCIG Smokers

Differential expression analysis via ANCOVA among the three study groups identified 3,165 genes whose expression is associated with either or both exposures (FDR-adjusted P < .05). Samples from TCIG smokers separate from the former- and ECIG-derived samples along the first principal component derived from the expression of these genes (Fig 1A), whereas there is little separation between the non-TCIG samples, suggesting that gene-expression differences associated with TCIG use are a strong driver of the differential expression detected by this analysis.

Figure 1.

Airway gene expression in ECIG users appears more similar to former smokers than to active TCIG smokers. (A) Analysis of covariance among study groups identified 3,165 differentially expressed genes associated with any smoking status. The differential expression resulting from TCIG use is the strongest driver of gene expression as visualized in the principal component analysis. (B) GSVA scores of a published set of genes found to be differentially expressed in TCIG smokers compared with nonsmokers, projected into data from the current study. ECIG user gene expression is more similar to that of former smokers than to that of active TCIG smokers. (C) GSVA scores of data from published gene sets known to be modulated by TCIG use show ECIG users are more similar to former smokers than active TCIG smokers projected into data from the current study. Differences in scores evaluated by Student t test. *P < .05, **P < .001. P values correspond to a significant comparison of at least one of the following via ANCOVA: TCIG relative to ECIG, TCIG relative to former, ECIG relative to former. ANCOA = analysis of covariance; ECIG = e-cigarette; GSVA = gene set variation analysis; TCIG = tobacco cigarette.

To identify the degree to which ECIGs induce TCIG-associated gene-expression changes, we used GSVA to determine the activation of genes up- or down-regulated between current TCIG smokers vs never smokers as identified by Beane et al.¹⁶ We found a significant difference in the expression of these TCIG-associated gene sets between TCIG smokers and ECIG users, and TCIG smokers and former smokers, but not between ECIG users and former smokers (Fig 1B). Additionally, we identified 280 pathway-related gene sets that have TCIG-associated expression differences in the Beane et al cohort and used them in a similar metagene analysis of the current dataset via GSVA (Fig 1C). Expression of these select TCIG-associated pathways was also shown to be significantly altered in the TCIG group relative to the former and ECIG groups (Fig 1C; Student t test, P < .05), but not significantly different between the former and ECIG groups. Collectively, these data suggest that the gene-expression profiles of the ECIG users are more similar to the former group than the TCIG group both globally and for genes that are altered by TCIG use.

ECIG Associated Gene-Expression Changes

Post hoc analysis of the 3,165 smoking-status-related genes identified by ANCOVA yielded 468 genes whose expression was associated with ECIG use status (Fig 2A; moderated t test P < .05). These genes were organized into four clusters via Ward hierarchical clustering (see e-Table 1). We used the cluster gene sets in a metagene analysis via GSVA to identify the relationship between the expression levels of these genes in ECIG users relative to TCIG and former smokers (Fig 2B).

Figure 2.

ECIGs induce gene-expression alteration in both concordant and discordant directions with TCIGs. Post hoc analysis of the 3,165 genes dependent on any cigarette use status reveals 468 genes (P < .05) whose expression is dependent on ECIG use. Figure shows log2 expression data, residually adjusted by the model used for differential expression. These genes cluster into four distinct groups (cluster 1a: 27 genes, cluster 1b: 171 genes, cluster 2a: 52 genes, cluster 2b: 218 genes) (B) GSVA scores of the residual adjusted expression four gene clusters in 2A reveal four different gene-expression patterns. Clusters 1 and 2 represent genes whose expression is up- and down-regulated, respectively, in the ECIG group compared with the former group. Clusters 1 and 2 can be further subdivided by TCIG group genes who change concordantly with (clusters 1a and 2a) and discordantly against (clusters 1b and 2b) the ECIG group. *P < .05. P values correspond to Student t test comparisons of GSVA scores between TCIG relative to ECIG, ECIG relative to former, and TCIG relative to former. See Figure 1 legend for expansion of abbreviations.

Genes in cluster 1 (198 genes) were upregulated in the ECIG group compared with the former group. Cluster 1 was subdivided into genes whose expression were also upregulated in the TCIG group (cluster 1a, 27 genes) or downregulated in the TCIG group (cluster 1b, 171 genes). Genes in cluster 2 (270 genes) were genes whose expression level were downregulated in the ECIG group compared with the former group. Cluster 2 was subdivided into genes whose expression level was also downregulated in the TCIG group (cluster 2a, 52 genes) or upregulated in the TCIG group (cluster 2b, 218 genes).

Cluster 1a, although small, contained several genes associated with interleukin receptor complexes. Cluster 1b contained ribosomal protein subunit-associated genes, as well as genes associated with maturation of noncoding RNAs and translation. Cluster 2a contained genes associated with microtubule assembly and structure. Cluster 2b was similarly enriched for genes involved in microtubule assembly and genes involved in regulation of RNA Polymerase II activity.

We performed quantitative RT-PCR to validate the ECIG-associated differential expression of ADM, PGAM5, NCK2, and RSPH1 as representative genes from clusters 1a, 1b, 2a, and 2b, respectively. Concordant with the microarray data, NCK2 and RSPH1 demonstrated significantly lower expression in ECIG users compared with former smokers (Fig 3; Student t test, P < .05). We were unable to confirm differential expression of ADM and PGAM5 by quantitative RT-PCR.

Figure 3.

qRT-PCR validation of ECIG signature genes. Gene expression of NCK2 (A) and RSPH1 (B) was measured by qRT-PCR in bronchial epithelial samples from ECIG users (n = 15) and former smokers (n = 15). Expression of each gene in each sample is normalized to the level of 18S ribosomal RNA. Expression differences between ECIG users and former smokers is assessed using Student t test. *P < .05. qRT-PCR = quantitative real-time polymerase chain reaction. See Figure 1 legend for expansion of abbreviation.

Comparison With In Vitro ECIG Exposure

To validate the gene-expression alterations we observed in ECIG users using a variety of ECIG products, we determined if the genes we identified as differentially expressed in ECIG users were similarly altered in an in vitro dataset from a previously published experiment that examined the effects of ECIG aerosol on differentiated human bronchial epithelial cells.²² Using GSEA, we found that genes downregulated in the bronchial epithelium of ECIG users are significantly enriched among the genes most downregulated with ECIG exposure in vitro (GSEA P < .001). In contrast, we did not detect significant enrichment of the genes upregulated in ECIG users among the genes most upregulated with ECIG exposure in vitro (Fig 4).

Figure 4.

Relationship between effects of ECIG use and effects of ECIG aerosol exposure on human bronchial epithelial cells in vitro. (A) Genes downregulated in ECIG users (cluster 2) are significantly enriched among the genes most downregulated in vitro in ECIG-aerosol exposed cells relative to control. (B) Both subclusters of genes downregulated in ECIG users (clusters 2A and 2B) are significantly enriched among the genes most downregulated in vitro in ECIG aerosol. *GSEA FDR q value < 0.005, indicating significance of skewness of gene set positions in a ranked list. Ranked list was derived in Moses et al’s²² in vitro gene expression data by performing Student t test on each gene in ECIG-exposed cells relative to air-exposed cells and ranking all genes by their corresponding t statistic. See Figure 1 legend for expansion of abbreviation.

Discussion

To our knowledge, this is the first study to comprehensively assess gene-expression changes in the bronchial airway epithelium of real-world ECIG users, and how these changes compare with those associated with TCIG smoking. Across all the genes we identified as differentially expressed between current TCIG smokers, former TCIG smokers now using ECIGs, and former TCIG smokers, we observed that the pattern of gene expression in ECIG users was much more similar to former smokers than TCIG smokers. TCIG-associated gene sets derived in a previous study¹⁶ were differentially expressed between the TCIG and ECIG as well as TCIG and former smoker groups, but not between the ECIG and former groups. We also found that the differential expression of genes in relevant TCIG-associated gene-expression pathways, specifically glutathione³⁵ and xenobiotic metabolism,³⁶ tumor necrosis factor receptor 2 signaling,³⁷ and complement cascade,³⁸ were distinct in TCIG smokers but not significantly different between ECIG users and former smokers. These findings suggest that, overall, TCIG-induced transcriptional changes revert to baseline similarly in former TCIG smokers and former TCIG smokers who now use ECIGs.

Although the ECIG and former TCIG smoker groups were not significantly different regarding expression of genes changed in TCIG smokers, we were able to identify a set of genes whose expression changes in ECIG users relative to former smokers and that a small subset of these genes are similarly changed in TCIG smokers. This implies that ECIG use does affect the physiology of the bronchial epithelium in ways that are distinct from the effects associated with TCIG use, and that the degree of overlap between ECIG- and TCIG-associated effects is modest.

To validate that the gene-expression alterations in bronchial epithelium associated with ECIG use are likely due to the direct effects of exposure of this tissue to ECIG aerosols, we examined the expression of our ECIG signature in differentiated human bronchial airway epithelial cells that were exposed to ECIG aerosols in vitro.²² We found that the genes we had identified as having decreased expression in ECIG users are enriched among the genes down-regulated in the in vitro ECIG aerosol exposed tissues. That we were able to identify common gene expression signatures between a heterogeneous in vivo exposure and a uniform in vitro exposure suggests the existence of a set of ECIG-associated gene expression effects that are common and independent of the specific product used. Beyond suggesting that the gene-expression changes in ECIG users are likely the direct effects of ECIG aerosol exposure, these data also suggest that there is a common impact of ECIG use on bronchial epithelial gene expression despite heterogeneity in ECIG product usage.

Among the genes in the ECIG signature we identified, those in clusters 1a and 2a are similarly altered in both ECIGs and TCIGs, whereas those in clusters 1b and 2b exhibit an ECIG-specific pattern. Pathway enrichment analysis of the genes in each cluster provides potential insight into common and ECIG-specific effects (Table 2).

Table 2.

Functional Enrichment of ECIG Signature Clusters

Cluster	Direction	Enriched Term	Assoc. Genes	Database
1a	ECIG up, TCIG up	Interleukin receptor complexes	CMKLR1, ESYT3, FCAR, LILRB3, MIP, NPBWR2, VSIG2	…
1b	ECIG up	Targets of ATF2	AAMP, ATP5G2, BANF1, CKS1B, CLPB, CTDP1, DAPK3, DDX49, EMC10, FAM83E, HDDC3, HIST1H4I, ID1, ING4, IRF2BPL, JMJD4, MARS, MPND, MRPL17, MRPL4, MUTYH, MZT2A, NOC4L, RBM15B, RPL10, RPL22L1, RPL39L, RPLP0, RPS14, RPS19BP1, RPS9, RRP1, SH3BPS5L, SNORA16A, SNORA21, SNORA24, SNORA57, SNORA9, SNORD104, SNORD15A, SNORD22, SNORD27, SNORD50A, SNORD60, SNORD76, SNORD81, SNORD95, STX4, VTRNA1-3, ZYX	CHeA/ENCODE consensus TFs
1b	ECIG up	Ribosome biogenesis	AAMP, DDX49, MRPL36, NOC2L, NOC4L, RPL10, RPLP0, RPS14, RPS9, RRP1, WDR46	GO biological process 2018
2a	ECIG down, TCIG down	Axon guidance	NCK2, SEMA5A, SLIT2	…
2b	ECIG down	Targets of RFX3	C10ORF194, CD200, DYNLRB2, FAM81B, FBXL2, IQCG, KIFAP3, MAP1B, MAPK10, PPP1R42, RCN2, RSPH1, SPEF2, ZCCHC11	ARCHS4 TF coexpression

Functional categories associated with each cluster from the ECIG signature. Database specifies the source of the enriched term’s corresponding gene set as provided by Enrichr. Entries with a specified database represent gene sets that were functionally enriched in the corresponding signature cluster via Enrichr’s hypergeometric test (P < .05), with the associated genes representing the genes overlapping those two sets. ECIG = e-cigarette; GO = gene ontology; TCIG = tobacco cigarette; TF = transcription factor.

Genes in cluster 1b were up-regulated specifically in ECIG users and were significantly enriched for the Ribosome biogenesis GO Biological Process term. The overexpression of ribosomal genes in ECIG users might reflect increased oxidant stress.³⁹ Supporting this hypothesis, this cluster additionally contains NDUFB2 and NDUFA4L2, which code for subunits of the NADH-ubiquinone oxidoreductase complex and play a role in handling oxidant stress. ATP5H, a component of the mitochondrial electron transport chain, also appears in cluster 1b. Although these genes are only induced in ECIG users, both ribosomal structure and oxidative phosphorylation pathways have both been previously found to be induced by TCIG smoke.¹⁵ Cluster 1b is also significantly enriched for targets of ATF2. Both of these transcription factors have been implicated as possible regulators of inflammation in the lung.^{40, 41} These findings suggest that ECIG use might modulate lung inflammation.

Cluster 2b contained genes that were down-regulated specifically in ECIG users. This cluster is significantly enriched for targets of RFX3, a transcription factor involved in ciliary assembly and motility that has been specifically observed to cooperate with FOXJ1 in the process of ciliated cell differentiation in airway epithelium.⁴² FOXJ1 is also significantly down-regulated in airway epithelial cells following in vitro ECIG exposure.²² We validated the downregulation of RSPH1 in ECIG users by quantitative RT-PCR. RSPH1 is a gene in this cluster and mutations in RSPH1 are associated with ciliary defects and dyskinesia.⁴³ Taken together, these data suggest that ECIG use might impair ciliogenesis.

Genes in cluster 2a were downregulated by both TCIG and ECIG use, and were significantly enriched for genes associated with axon guidance. NCK2, a gene from cluster 2a that we validated by quantitative RT-PCR, is significantly downregulated by ECIG use. NCK2 is associated with EGFR signaling and cytoskeletal reorganization,⁴⁴ and we hypothesize that this cluster reflects cytoskeletal and/or cilium-related effects of TCIG and ECIG use. Genes in cluster 1a were upregulated by both TCIG and ECIG use and were associated with interleukin receptor complexes.

Although the findings of the present study yield insights into the effects of ECIG use on bronchial epithelium, there were several limitations, most notably regarding sample sizes within each of the study groups. We observed that many fewer genes were affected by ECIG than TCIGs, which dominated differential expression when comparing all three groups (Fig 1). Although we were able to identify a number of ECIG-specific gene-expression changes, it is likely that more subjects would be necessary to comprehensively identify the transcriptional effect of ECIGs. Future studies would also benefit from the inclusion of de novo users of ECIGs to be able to better isolate effects of ECIGs that might either be specific to former smokers or obscured in former smokers. The potential to perform experimental studies on volunteers with non-nicotine-containing ECIGs would enable identification of the specific gene expression effects associated with vehicle and flavorings, which we are unable to discern in the present signature. This is potentially important because any specific ECIG additive might have an unusual but dramatic effect on airway biology.

Furthermore, we did not identify significant overlap between the genes we found to be differentially expressed in the bronchial epithelium of ECIG users and the genes found by Martin et al²³ in nasal epithelium as we expected based on previous work establishing the commonalities of bronchial and nasal signatures of TCIG exposure.^{45, 46}

Another limitation is the heterogeneity of ECIG products used by the ECIG users in our study. To recruit a sufficient study population, we did not limit the brand or type of ECIG device that participants could use. Additionally, though all participants met a minimum threshold of weekly ECIG usage, actual usage varied amongst participants. Most of the study participants were users of first-generation devices; it remains to be determined if new types of devices will elicit similar changes. However, our finding of common ECIG effects despite the heterogeneity of ECIG exposures within the study group, and our finding that many of the changes observed in users of heterogeneous ECIG products are similar to the effects observed with a single brand ECIG in vitro leads us to believe that the majority of these effects are independent of the vaporization device used and that there is a common set of ECIG-related effects on airway epithelium.

Overall, our findings indicate that ECIG use does not lead to alterations in the expression of the majority of genes that are altered by TCIG use, but that there is a group of genes whose expression is specifically altered in ECIG users. When examining the expression of genes in key TCIG-associated pathways, we found that ECIG users had gene-expression profiles more similar to former TCIG smokers than current TCIG smokers. Our findings indicate that the use of ECIGs does affect the airway, which includes modulating the expression of a small set of genes that are altered in both ECIG users and TCIG smokers. Further study is required to identify the clinical significance of these findings and to fully evaluate the pulmonary effect of ECIG exposure.