Cigarette Smoking and the Airway Epithelium: Characterizing Changes in Gene Expression over Time

Cigarette smoking is one of the most important health-related environmental exposures. Although public health interventions have substantially reduced smoking rates in the United States, smoking remains a major cause of morbidity and mortality in high- and low-income countries. Because smoking is a major risk factor for lung cancer, chronic obstructive pulmonary disease (COPD), and cardiovascular disease, understanding the molecular consequences of smoking is an important research priority.

To date, numerous high-quality studies have described the extensive molecular and phenotypic effects of cigarette smoke on human airway cells, including cross-sectional analyses of proximal and distal human airway cells. These studies have revealed widespread transcriptional and epigenetic changes (1–3), release of inflammatory cytokines, induction of cellular stress and senescence pathways, impairment of ciliary and barrier functions (4), development of squamous metaplasia, and alterations in differentiated cell type proportions (5, 6). In addition, bulk and single-cell RNA sequencing studies have identified dynamic and etiologic transcriptomic changes associated with smoking-related diseases like COPD and lung cancer (7–9), and a 23-gene classifier based on airway transcriptomics improves the diagnostic yield of bronchoscopy for lung cancer among intermediate-risk patients (10). Collectively, these studies highlight the critical insights into the pathogenic consequences of smoking obtained through assessing the airway transcriptome.

In this issue of the Journal, Strulovici-Barel and colleagues (pp. 780–790) apply transcriptomic analyses to small airway epithelial brushings from a total of 254 subjects enrolled in four separate cohorts (11). The primary analysis is of 37 smoker and 23 nonsmoker subjects who received four serial assessments of smoking status and transcriptomics over a 1-year period. Importantly, transcriptomic data were also available in 17 subjects who agreed to stop smoking and undergo serial sampling of the small airways for a year after smoking cessation. This study represents an unusually rich resource of longitudinal transcriptomic responses to cigarette smoking and cessation from the small airways, an important but challenging region to access for clinical sampling.

There are two major findings from this study. First, the authors identify a set of 475 persistently altered smoking genes, with significant differences in expression level observed at all four observations over a 1-year period. The biological processes associated with these genes were consistent with previous studies in the airway epithelium. Second, analysis of these 475 genes in the smoking cessation cohort identified reversible and “nonreversible” genes and allowed for detailed observation of the time to resolution for specific pathways. Interestingly, the most quickly and uniformly resolving processes in this study were those involved with regulation of gene expression and inflammation, whereas slowly resolving or nonresolving processes were related to cellular metabolism, apoptosis, and growth and proliferation.

The major strengths of this study are the longitudinal study design, the direct study of small airway epithelial cells, and the transcriptomic characterization of smokers before and after cessation. In addition, smoking status was biochemically validated at each study visit. The inclusion of all these elements is uncommon and represents an intriguing combination of principles of experimental intervention with population-based transcriptomic discovery approaches. Furthermore, because most population-based transcriptomic studies are cross-sectional, differentially expressed genes are often described as up- or downregulated, with the implicit assumption that the up- or downregulation is persistent and reflects a steady-state transcriptional response. Longitudinal study designs such as this one enable a richer representation of dynamic transcriptional responses. Although experimental exposure studies at air–liquid interface can allow for characterization of transcriptional changes over hours to days (12), this study fills an important gap by characterizing the temporal responses to smoking and smoking cessation over months to 1 year. Although most of the smoking-associated genes return to normal expression levels over weeks to months, this study adds to previous observations of substantial numbers of genes that show persistent alterations in expression. This molecular observation is consistent with epidemiologic observations of a persistently elevated risk for lung cancer that does not normalize even 2 decades after smoking cessation (13).

Important limitations include the restriction of the study population to healthy smokers, highlighting an important gap in the literature, with relatively few studies focused specifically on characterizing the smoking response among subjects who are known to be susceptible to COPD or other smoking-related diseases. In fact, a previous longitudinal study of the consequences of smoking cessation also demonstrated reduced inflammation among 16 asymptomatic smokers who successfully quit but persistent airway inflammation among 12 candidates with COPD who successfully ceased smoking for 1 year (14). Therefore, a more detailed understanding of differential transcriptional responses to cigarette smoke among individuals with or at risk for pulmonary diseases may elucidate biologic pathways underlying disease progression. Although the authors define persistently altered genes and intermittently altered genes, there is the opportunity for future work to characterize temporal patterns of gene expression in more detail, and there is still an important gap in our understanding of how short-term patterns of gene expression and epigenetic changes observed in experimental systems map to long-term molecular patterns arising from chronic cigarette smoke exposure and leading to disease. In addition, this study highlights a key challenge with the evaluation of bulk transcriptomic data. Although the authors undertook a stringent approach to replicating their results, it is notable that even for the set of 475 persistently altered smoking genes that achieved significance across each study time point, only 41% (n = 195) of these genes were also significant in both independent replication cohorts. This level of replication is not surprising for transcriptomic data, which captures many complex signals but can be susceptible to confounding in both cross-sectional and longitudinal study designs. Although multiple stages of replication in this study provide confidence in the 195 fully replicated smoking-associated genes, further studies will be necessary to tease out adaptive versus maladaptive responses to cigarette smoke as well as the lingering consequences of smoke exposure on human health.