Defining VOC signatures of airway epithelial cells with PM_2.5 exposure

Abstract

Volatile organic compounds (VOCs) produced by the lung in response to exposure to environmental pollutants can be utilized to study their impact on lung health and function. Previously, we developed a method to measure VOCs emitted from well-differentiated tracheobronchial epithelial cells in vitro. Using this method, we exposed well-differentiated proximal (PECs) and distal airway epithelial cells (DECs) to varying doses of traffic-related air pollutants (TRAP) and wildfire particulates to determine specific VOC signatures after exposure. We utilized PM_2.5 TRAP collected from the Caldecott tunnel in Oakland, CA and the 2018 Camp Fire to model “real-life” exposures. The VOCs were collected and extracted from Twisters and analyzed using gas chromatography-mass spectrometry. Exposure to both types of particulate matter (PM) resulted in specific VOC responses grouped by individual subjects with little overlap. Interestingly the VOCs produced by the PECs and DECs were also differentiated from each other. Our studies suggest that PM exposure induces a specific compartmentalized cellular response that can be exploited for future studies. This response is cell-type specific and potentially related to a phenotype we have yet to uncover.

Air pollution, derived from the burning of fossil fuels, motor vehicles, and wildfire smoke, is a known risk factor for exacerbations of chronic respiratory diseases such as asthma and chronic obstructive pulmonary disease (COPD). This is a globally relevant issue given that in 2019, over 90% of the world’s population lived in regions that exceed the WHO’s air quality standards (World Health Organization 2021) and air pollution accounts for an estimated 7 million premature deaths. Epidemiological studies have examined the effects of air pollution on disease exacerbations, worsening of disease, and increased onset of disease. There is a strong causal link between traffic-related air pollution (TRAP) and the onset of childhood asthma and supportive evidence that air pollution exposure contributes to the development of adult respiratory disease (Hoek et al. 2013; Esposito et al. 2014; Thurston et al. 2020; Boogaard et al. 2022). Structural remodeling postinjury, the lung innate immune cell profile, and genetic components are among the mechanisms to consider when studying the relationship between air pollution exposure susceptibility and the development of respiratory disease (Hooper and Kaufman 2018). There are many diagnostic tools available to monitor local and systemic changes in the lung, related to airway disease including peripheral blood and sputum cell counts, imaging, and biological and physiological measurements such as nitric oxide (NO) and lung function. However, current tools are resource-intensive, invasive, and have significant limitations in predictive value. This is particularly true for quantifying the degree of exposure and predictors of new or worsening disease.

The measurement of volatile organic compounds (VOCs) emanating from humans, cells grown in vitro, plants, and microbes is an active area of investigation, as it offers a noninvasive method of monitoring the health and status of dynamic systems. VOCs produced in vitro by several cell types have been identified and used to distinguish healthy and diseased cell populations (Filipiak et al. 2016). We previously have shown that virus-infected proximal airway epithelial cells (PECs) and cells exposed to chemical irritants with reactive oxygen species produce distinct VOC signatures unique to their exposures (Schivo et al. 2014; Yamaguchi et al. 2019; McCartney et al. 2021). Others, such as Kamal et al. (2021), have found decane and other long-chain alkane compounds were produced by both human subjects challenged with rhinovirus and virally infected airway epithelial cells demonstrating a link between in vitro and in vivo responses. This encourages the development of more sensitive collection and analytical devices that can provide the capability to utilize VOCs to monitor environmental exposure and dynamic metabolic changes related to cellular injury and insight into the onset and development of airway disease.

The airway epithelium is comprised of multiple cell types (Davis and Wypych 2021) and despite being one continuous cell surface, the specific cell makeup of proximal and distal airways differs significantly and displays compartment-specific functions. Studies looking at transcriptional changes in airway epithelial cells brushed from large and small airways of smokers and nonsmokers noted differential expression not only based on smoke exposure but also on region of collection (Goldfarbmuren et al. 2020). This provided a snapshot of the cellular response but did not capture the real-time metabolic changes or functionality related to the exposure. Here, we were able to examine the metabolic changes of PEC and DEC types and the individual responses to both traffic and wildfire PM_2.5 of each subject we studied. There has been significant interest in defining the cellular sources of VOCs, collected in breath analysis studies.

We hypothesized that PECs and DECs would produce different VOCs and analysis of these VOC profiles reflects a highly sensitive and important method to measure and catalog airway epithelial cell responses to environmental exposures. We utilized our published methodology for measuring VOCs from primary airway epithelial cells (Yamaguchi et al. 2018) exposed to PM_2.5. We cultured both human PECs and DECs and found different VOC signatures that suggest independent responses to the same PM_2.5 exposures. Cells from individual subjects also yielded different signatures hinting at unique responses to exposure. Interestingly, the VOCs generated from exposure to TRAP in comparison to wildfire particulate matter (PM) also did not overlap. The VOCs from the wildfire exposure elicited a more distinct and stronger response compared with TRAP confirming that the makeup of the 2 types of PM_2.5 are very different. This is a pertinent finding related to understanding how different PM exposures can lead to the development of divergent airway diseases. This is also relevant to the studies focused on the health effects due to the diverse makeup of PM_2.5 from wildfires at the wildland-urban interface (WUI) which can include some traffic-related pollutants (Noah et al. 2023). The capability to utilize VOCs to detect and distinguish dynamic differential epithelial cell responses from proximal and distal airways has not previously been observed.

Materials and methods

Cell culture

Primary human proximal bronchial airway epithelial cells and distal small airway epithelial cells (generations 17–23) were obtained from Lifeline Cell Technology (Frederick, MD). The PECs and DECs were plated on Transwell (Corning Costar, Corning, NY) chambers (12 mm) at 1–2 × 10⁴ cells/cm² coated with 0.05 mg/ml type IV collagen (Sigma) in the PneumaCult-Ex medium (Stemcell Technologies). Once the primary proximal and distal airway cultures were confluent, they were transferred to ALI culture conditions in PneumaCult-ALI or PneumaCult-ALI-S media (Stemcell Technologies) for 1 mo.

PM exposure

Differentiated primary epithelial cells were separately exposed to 10, 50, and 100 µg of PM_2.5, suspended in PBS, apically in 12 well Transwell plates for 24 h in a dose–response curve. The wildfire particulates from the 2018 Camp fire and TRAP PM used, were collected from the Caldecott tunnel to model “real life” exposures experienced by subjects, with corresponding VOC samples collected. These unique PM samples are obtained in a collaboration with the UC Davis Environmental Health Sciences Center. Details of the 2018 Camp fire PM collection and composition can be found in a publication authored by the California Air Resources Board (Camp Fire Air Quality Data Analysis). Details about the TRAP PM collection can be found in the following 2 publications by our collaborators, Bein et al. (2022) and Edwards et al. (2020).

Extraction of PM_2.5 filter samples and field blanks was performed using a multi-solvent filter extraction technique that combines sonication in multiple solvents, liquid–liquid extraction, microporous membrane filtration, and detailed gravimetric analyses to (i) maximize extraction efficiency, (ii) minimize compositional biases, (iii) minimize extraction artifacts, and (iv) provide precise and accurate direct measurements of extracted PM mass. A comprehensive explanation of this method was published by Bein and Wexler (2014, 2015).

VOC collection

The transwells containing confluent cells, after the 24-h PM exposure, were placed into glass jars filled with 5 ml of PneumaCult-ALI or PneumaCult-ALI-S media (Stemcell Technologies) and capped with lids containing a Teflon insert. Headspace sorptive extraction was performed as previously described (Yamaguchi et al. 2018). Briefly, Twisters (Gerstel, Mülheima/d Ruhr Germany), coated with a 0.5-mm thick polydimethylsiloxane layer, were magnetized to the lid of the jar for sampling headspace VOCs. At 4, 8, or 24 h, Twisters were collected and kept at −20 °C until analysis.

GC-MS analysis

Gas chromatography-mass spectrometry (GC-MS) analysis of the VOCs collected onto Twisters was performed as previously described (Yamaguchi et al. 2018). Briefly, at each timepoint, Twisters were removed and loaded into desorption tubes, 0.5 µl of 50 mg/ml naphthalene-D8 was added to monitor the performance of the instruments. Chromatography occurred on a 7890 GC (Agilent Technologies Inc., Santa Clara, CA) with a DB-5ms column, 30 m×250 µm×0.25 µm using a constant 1.5 ml/min flow of helium and a gradient of temperature. Twisters were desorbed with a thermal sorption unit (TDU) and a cooled injection system (CIS, Gerstel Inc). The total run time was 98.17 min total, and the signal was detected using a 5977A mass spectrometer (Agilent Technologies Inc., Santa Clara, CA).

Metabolic assay

Alamar blue (Life Technologies, Carlsbad, CA) was used to assess the metabolic activity of the cells after each time point used to collect the VOCs. Briefly, a 1:10 dilution of Alamar blue: PneumaCult-ALI or ALI-S medium was added to the apical surface of the transwell, PneumaCult-ALI or ALI-S medium was used in the basal-lateral chamber. The Alamar blue solution was incubated with the cells for 2 h at 37 °C and assessed at 560 nm excitation/590 nm emission for fluorescence intensity. Diluted Alamar blue unexposed to cells was used as a negative control.

ELISA

One hundred microliters of the medium were collected at several time points from the basal lateral side of each ALI culture and frozen at −80°C. The medium samples were then thawed on ice and used in enzyme-linked immunosorbent assay analysis for human interleukin (IL)-6, and IL-8 (R&D systems DuoSet ELISA, Minneapolis, MN).

Data and statistical analysis

Data are presented as original values or means SE; n refers to the number of replicates. Comparison between 2 treatment groups was done using paired t-tests. Effects of multiple treatments were tested using ANOVA/1-way Kruskal–Wallis test followed by the Tukey–Kramer multiple comparisons test. Statistical testing was done using GraphPad Prism 10. Resulting P values were given, and P < 0.05 was considered significant.

VOC data files were deconvoluted, integrated, and aligned using MassHunter Profinder B.08.00 (Agilent Technologies Inc.) as described in Yamaguchi et al. (2018). Datasets were created using GeneSpring MPP (Agilent Technologies, Inc.) and then exported to Excel and Matlab to work with multivariate analysis models using PLS Toolbox V8.6.2 software.

Multivariate models were used for comparative analyses, using PCA initially to obtain an overview of large datasets, visualize similarities and differences between observations, and detect potential outliers. Then partial least squares-discriminant analysis (PLS-DA) was used as a supervised model to classify defined groups of samples based on the VOC metabolomic profile.

Results

We cultured and differentiated primary epithelial cells from 8 different subjects, 4 were derived from human distal airway epithelial cell (DEC) cultures and 4 were derived from human PECs. We performed a dose–response curve utilizing wildfire particulates from the 2018 Camp fire and TRAP and PM collected from the Caldecott tunnel to model “real life” exposures experienced by subjects, with corresponding VOC samples collected. These unique PM samples are obtained in a collaboration with the UC Davis Environmental Health Sciences Center. The TRAP facility is adjacent to the Caldecott Tunnel in Berkeley, CA and collects PM samples from 2 bores of the tunnel including the diesel exhaust particulates from the bore of the tunnel that allows trucks. We anticipate that these experiments will demonstrate how VOCs can be used to identify and distinguish individual responses to exposures and understand how different cell types in the lung respond to the same exposure.

Exposure to PM_2.5 does not impact cellular metabolic activity

PECs and DECs were separately exposed to 10, 50, and 100 µg of PM_2.5 for 24 h. Metabolic activity was then assessed using Alamar blue and there were no significant changes between the different doses within one cell type. However, the DECs displayed an overall high rate of metabolic activity compared with the PECs (Fig. 1).

Fig. 1.

Metabolic activity at 24-h post-Caldecott PM exposure, measured and reported as fluorescence intensity 560/590 nm using Alamar blue. There were no significant differences between doses of PM.

VOC signatures specific to PECs versus DECs

GC-MS analysis of the VOCs generated by PECs and DECs was analyzed using PLS-DA.

A total of 136 features, after blanks and media signal curation, were used to build an initial PCA model. These represent the VOCs released by 1 set of PECs and 2 sets of DECs, regardless of PM_2.5 exposure, and could illustrate a clear separation by cell type and did not cluster by PM exposure (Fig. 2a and b). Then, cell type differences are clearly explained by a PLS-DA of VOCs from 2 sets of PECs and 3 sets of DECs, Fig. 3. After feature selection using variable importance in projection higher than 1, PLS-DA scores plot separation was described by 24 VOC features that explained PECs and DECs differences. These characteristic features that explain the sample separation are described by their mass and retention time (mass@RT) that can be characterized after a library search. In Table S1, we identified some compounds with higher intensities for PECs, like S-isopropyl lactate, and others characteristics of DECs, such as cetene, dimethyl ether, or nonanal. There are no clear differences in the PM dose used, and the majority of the differences are due to the cell type, PECs or DECs. We also performed measurements of Camp fire PM and PBS alone. There were 52 features detected but all features had P-values >0.05 and there were no statistically significant differences between them (Table S5).

Fig. 2.

(a, b) Score plots from principal component analysis (PCA) of VOCs produced by 136 detected features from 1 set of PEC and 2 sets of DECs from different human subjects.

Fig. 3.

Score plots from PLS-DA of VOCs from 2 sets of PECs and 3 sets of DECs from different human subjects. Twenty-four VOC compounds separate PECs from DECs.

PLS-DA plots of VOCs released by PECs exposed to PBS (control) or PM_2.5 from either the Caldecott Tunnel (Fig. 4a and b) or Camp fire (Fig. 5) demonstrated a clear separation by exposure and by individual subjects. VOCs released by DECs also demonstrated separation by exposure to PBS or Caldecott Tunnel PM_2.5, however there was more overlap between subjects. When the length of exposure time was added as another component, the VOCs were further separated according to the length of PM_2.5 exposure, in addition to clustering by individual subject (Fig. 6). Discrete VOC signatures separate out individual responses from different subjects to PM exposure.

Fig. 4.

PLS-DA plots of 2 sets of PEC (a) and 3 sets of DEC (b) VOCs in response to Caldecott PM exposure, 4 and 24 h of VOC collection.

Fig. 5.

Score plot of PLS-DA of VOCs produced by PECs in response to Camp fire PM exposure, 8 and 24 h of VOC collection.

Fig. 6.

Score plots of PLS-DA of VOCs produced by PECs and DECs in response to Caldecott PM exposure by (a) length of VOC collection time (4 and 24 h) and (b) individual human subjects.

We compiled lists of identified VOC compounds and their differential responses in PECs and DECs associated with Caldecott or Camp fire PM_2.5 exposure (Tables S2–S5). In this case, we indicate the up- or down-regulation determined by the PM exposure in the different types of cells studied. Interestingly, some of the VOCs were up-regulated, with higher intensities with the PM exposure, in one cell type and down-regulated, showing lower intensities in response to Caldecott PM_2.5 exposure, in the others, Tables S2 and S3. The VOCs produced by cells exposed to Caldecott PM_2.5 include aldehydes, ketones, and hydrocarbons that previously have been linked to lipid peroxidation and other oxidative stress pathways. Twenty identified and 29 unidentified VOCs explained the variance between wildfire PM_2.5-treated and untreated PECs (Table S4). Interestingly, the intensities of aliphatic hydrocarbons in wildfire PM_2.5-treated PECs were significantly higher compared with non-PM_2.5, PBS cells. This observation suggests a potential association between wildfire PM_2.5 exposure and an elevated production of aliphatic hydrocarbons.

Of note, only decane overlapped between exposure to Caldecott and Camp fire PM_2.5. The other VOCs were unique to each PM_2.5 source.

Measurements of IL-6 and IL-8

Concentrations of IL-6 and IL-8 cytokines secreted into the basolateral medium, assayed using individual ELISAs for each cytokine did not demonstrate significantly different responses between doses of Caldecott PM_2.5 in either PECs or DECs (data in Figs S1–S4).

Discussion

We summarize our major findings as follows: (i) PECs and DECs produce different VOC profiles at baseline that likely reflect known changes in airway structure and function, (ii) exposure of these cells to 2 separate sources of pollutant PM_2.5 causes recognizable, discrete changes in VOCs that suggest differential metabolic responses to the environment, and (iii) the unique features from primary epithelial cells, isolated from individual subjects, are preserved in vitro, highlighting their usefulness in exposure studies. Overall, these findings support our hypothesis and suggest that VOC analysis is a highly sensitive and important method to measure and catalog airway epithelial cell responses to the environment. In addition, this work will contribute to our understanding of the cellular sources of VOCs measured in breath in the rapidly growing field of breath analysis research.

Exposure to PM_2.5 is known to exacerbate airway diseases such as asthma and COPD (Hoek et al. 2013; Esposito et al. 2014; Thurston et al. 2020; Boogaard et al. 2022). A better understanding of the mechanisms leading from exposure to exacerbation will allow us to mitigate risks in vulnerable populations. In this study, we found that primary human PECs produce VOCs that were specific to independent exposure to “real-life” PM_2.5 collected from the Caldecott Tunnel or the 2018 California Campfire when compared with unexposed PECs in vitro. These findings align with our previous studies (Yamaguchi et al. 2019; McCartney et al. 2021) and support the utility of VOC profiling in cell culture models as a tool to understand the impact of environmental exposures on airway cell metabolism and injury.

The field of breath research is also working on capturing VOC signatures from animal models of exposure. Jonasson et al. (2024) recently published an article on measuring potential exhaled breath biomarkers from chlorine-exposed mice. Neuhaus et al. (2011) also published their findings of metabolites in exhaled breath from a mouse model of asthma.

Airway epithelial cells are an important component of the airway epithelial barrier and provide both physical and chemical protection against environmental pollutants. As a result, they are resilient and as demonstrated in this publication by Goksel et al. (2024), able to remain largely unaffected by the lower dose, short-term exposures to PM_2.5. We observe a similar resilience in our primary cells with our single PM_2.5 exposure model, as demonstrated by the minimal change observed in metabolic activity. However, both PECs and DECs physiologically respond to the PM_2.5 exposure as evidenced by the production of VOCs, postexposure.

Notably, the overall VOC profile identified from Campfire PM_2.5-exposed PECs differed from the VOC profile that was identified from Caldecott Tunnel PM_2.5-exposed PECs. Compounds classified into the hydrocarbons and lipids and lipid-like predominated the identified VOCs in the Camp Fire PM exposure, whereas the Caldecott Tunnel exposure also included compound classes such as organic nitrogen and oxygen compounds (Table S5). There were a few compounds that overlapped between the 2 exposures but our analysis found that most VOCs were unique to each exposure. This suggests that these different environmental sources of PM_2.5 stimulate separate metabolic processes that may require differential treatment.

Studies measuring exhaled breath VOCs from firefighters have reported increased production of benzene and a decrease in ethylbenzene following exposure to controlled structure burns (Wallace et al. 2019). Interestingly, we found in our in vitro studies, benzaldehyde was produced as a result of Campfire PM_2.5 exposure and exposure to Caldecott Tunnel PM_2.5, which also resulted in benzene and substituted derivatives being produced, suggesting a potential overlap in some chemical species in the original material that the PM originates from. In addition, these findings support our in vitro model that airway epithelial cells grown in air–liquid interface, produce compounds similarly found in vivo breath studies.

Wildfire exposure has been associated with adverse respiratory health effects, specifically exacerbations of asthma and COPD (Wilgus and Merchant 2024). Oxidative stress, namely lipid peroxidation, is widely accepted to be related to asthma severity (Wood et al. 2003). An elevated production of aliphatic hydrocarbons including 4-ethyl-undecane, 2, 3, 5, 8-tetramethyl-decane, 4, 6-dimethyl-dodecane, 2, 6-dimethyl-undecane, 4, 8-dimethyl-undecane, and 4 additional unidentified compounds were observed in wildfire PM_2.5-exposed PECs. Previous studies have reported that breath alkanes may serve as markers of lipid peroxidation, related to cellular damage associated with smoking, human immunodeficiency virus infection, and inflammatory bowel disease (Caldeira et al. 2012; Davis and Wypych 2021). Taken together, our findings suggest that wildfire PM_2.5 exposure may play a direct role in lipid peroxidation and this activity may be directly measured by VOCs. This is of specific interest to us with all of the studies focused on the health effects due to PM_2.5 from wildfires at the WUI (Noah et al. 2023).

We also observed that there is a distinct VOC signature, comprised of 24 compounds, that separate PECs and DECs regardless of PM exposure and media differences. The majority of them did not have a known ID but among the identified compounds was a fatty acid metabolite, cyclopropane, and nonessential amino acid L-asparagine which is important in protein synthesis. There is not a lot known about the overall metabolic differences between PECs and DECs. It is known that the specific cell makeup of proximal and distal airways is unique (Shaykhiev and Crystal 2014; Goldfarbmuren et al. 2020; Davis and Wypych 2021) and likely responsible for our observation that DECs had an overall higher metabolic rate compared with PECs when measuring metabolic activity using the Alamar Blue assay.

Similarly, we observed a distinct VOC response to TRAP that was distinct between PECs and DECs. These included aldehydes, ketones, and hydrocarbons that previously had been linked to lipid peroxidation and other oxidative stress pathways associated with airway disease. There were compounds that were increased or decreased in both cell types with PM_2.5 treatment but there were many that were differentially expressed between PECs and DECs. There were also specific VOC responses that clustered by individual subjects regardless of cell type, with little overlap, when graphed on a PLSD plot (Fig. 3). This is interesting, considering that the unique cell features are preserved, even when they are grown in vitro in the same medium for the same length of time. Other groups have observed a similar phenomenon in NO release. Suresh et al. (2007) and Jiang et al. (2009) observed that the rate of IL13-induced NO release by PECs was subject dependent and in DECs, IL13 only caused a modest release of NO. Interestingly, the inflammatory cytomix (IL-1β, TNF-α, and IFN-γ) caused a significant increase in NO release in DECs (Jiang et al. 2009) indicating that the airway location (cell makeup) and type of inflammation matters.

Currently, biomarkers of susceptibility have been explored in the areas of proteomics and transcriptomics. However, these methods do not provide enough detail to find out which pathways or cell types are affected. The ability to utilize VOCs to distinguish between subjects and cell types speaks to the specificity and sensitivity of our method. This furthers the idea of using portable sensors to monitor environmental exposure and dynamic metabolic changes related to cellular injury, which may provide insight into susceptibility to the onset and development of airway disease.

Interestingly, headspace VOC production is sensitive enough to detect dynamic changes in cells and can be captured more frequently so we have longitudinal measurements of the cells’ real-time response to PM_2.5 exposure. Even more intriguing is that the VOCs detected, arise from specific metabolic pathways that can be used to understand and target the changes that occur with perturbation of the cells.

It is worth noting the potential role of lung immunity on airway epithelium VOC production. The innate lung immune system involves a complex network of cells and signals that work to target invading pathogens and maintain tissue integrity. During the recognition of lung pathogens by airway epithelium, robust remodeling of the pulmonary transcriptome takes place and results in the production of cytokines, chemokines, growth factors, and more (Quinton et al. 2018).

Here, we demonstrate the utility of in vitro VOC profiling to reliably differentiate airway epithelial cells during exposure to environmental toxicants. We recognize that a larger sample size and matched PECs and DECs from the same subject in future experiments will significantly improve the reliability of our findings. Our ability to determine specific VOC metabolite concentrations is also a current experimental barrier to discern which of these compounds are important pathologically.

Based on our findings, VOC profiling in air–liquid interface cell culture models may serve as a reliable tool to provide insight into metabolic processes associated with stimuli-induced airway injury. In this study, we show that VOCs emanating from airway epithelial cells in vitro can be used to differentiate cells based on PM exposure status. These results not only provide a biological basis for airway epithelial cell-derived VOC production as a cellular response to environmentally derived stimuli but also bring forth a reliable method for measuring the biological effects of environmental exposure events on airway injury development which may be used to develop novel VOC-based diagnostic tools. Further studies are needed to understand whether PM_2.5 exposure has a sustained effect on the composition of airway epithelial cell-derived VOCs that can predict susceptibility to the development of airway disease and inform clinical treatment.

Author contributions

Cristina E. Davis, Richart W. Harper, and Nicholas J. Kenyon were responsible for the study conception and design. Keith Bein collected, analyzed, and provided the PM used in the study. Angela L. Linderholm, Leilani L. Jones, and Katyayini Aribindi performed the experiments and sample collection. Eva Borras and Mitchell M. McCartney were responsible for collecting and analysis of data. Angela L. Linderholm and Leilani L. Jones developed the first draft of the manuscript. All authors reviewed, provided edits, and approved the final version of the manuscript.

Supplementary material

Supplementary material is available at Toxicological Sciences online.

Funding

NIH NCATS (awards U18TR003795, U01TR004083, and UL1TR001860, CED and NJK); NIH Office of the Director (award UG3OD023365); NIH NIEHS (award P30ES023513, CED) Department of Veterans Affairs (award I01 BX004965-01A1, CED); University of California Tobacco-Related Disease Research Program (award T31IR1614, CED); California Firefighter Cancer Research Study funded by the University of California Office of the President (award R02CP7431, CED).

Conflicts of interest. None declared.

Data availability

The data that support the findings of this study are available on request from the corresponding author, A.L.L.