The results of this study show that: (a) measurement of DNA methylation in exhaled breath condensate is feasible; (b) the DNA appears to be of lower airway or lung origin; and (c) has some association with lung cancer and smoker status, depending on gene and individual CpG site examined.
It has long been clear that the gas phase of exhaled breath, and the aqueous condensate phase, contains small molecules that can be analyzed for pathologic processes in the lung, such as for asthma. For larger molecules, such as DNA-based studies, both Gessner et al. [18
] and Carpagnano et al [19
] have demonstrated the possibility of detecting DNA-based sequence alterations in EBC from patients with non-small cell lung cancer. We confirmed that ability, and further optimized the collection and DNA extraction procedures. We then adapted a bisulfite conversion approach and developed two-step nested PCR amplification, while limiting multiplexing, to allow for consistent analyses of these trace specimens, in a recently-devised and comprehensive methylation mapping assay [21
Our results showing the complete discordance between the respective exhaled and mouthwash DNA methylation map "fingerprints" implies that the predominant origin of exhaled DNA was not contamination from the mouth. Indeed, if mouth-derived DNA is present in EBC, it should be less than 10% of total DNA in EBC. This conclusion is based on the: (1) sensitivity limits of tBGS (>10%) that preclude complete exclusion of mouth derived (unmethylated) DNA in EBC at CpG sites that show methylation; and (2) the detection of a negative (unmethylated) signal could potentially be subsumed in the positive signal at methylated sites, although a review of the sequence tracings did not bear this out. The precision limits of the semi-quantitation afforded by sequence chromatograms for partial methylation (intervals of ~20% intervals), were previously published [21
] and appear as shades of gray, in the maps. This initial study therefore suggests that the largest proportion of EBC derives from the lower airway, as judged by the fact that exhaled specimens are discordant from the mouthrinse specimens in methylation pattern, when collected from the same individuals, for the one gene promoter (DAPK
) so tested. We have ongoing studies more directly addressing the anatomic origin of exhaled DNA, by direct bronchial brush and bronchoalveolar lavage methylation comparison to EBC methylation from the same donors.
Critical to the development of a marker panel for early detection of lung cancer is the selection of genes whose methylation is common but occurs during different stages of lung cancer development. In this study, three genes (DAPK, RASSF1A
) showed methylation among the five candidate genes originally selected. While the p16
gene methylation has been reported as one of the earliest methylation events in lung cancer development, occurring in the bronchial epithelium of some current and former smokers [29
], we did not find methylation in pretested exhaled samples, nor in the lung cancer cell line A549 cells (not shown). This may be because of the 5-10% sensitivity limitations of tBGS and/or for A549 cells, cell line differences that may not reflect tumor markers. The vast majority of published data has employed some form of methylation specific PCR, which is much more sensitive than sequencing based tBGS for methylation at a given CpG site, by perhaps 10-100-fold. It should be noted that this relative insensitivity of tBGS for methylation at any given site, but broad coverage of multiple CpG sites that may bear on expression, is suitable for many situations where minor degrees of methylation at isolated sites may not be biologically relevant, as the ultimate promoter readout is functional gene expression.
We chose commonly studied tumor suppressor genes such as DAPK
, and RASSF1A
precisely because they had been reported to be later events in lung cancer. Indeed, methylation of the DAPK
genes is uncommon (3% and 0%, respectively) in bronchial epithelium from smokers without cancer, using MSP-based methods [29
]. Nonetheless, our bisulfite sequencing results showed the methylation density of RASSF1A
was statistically different between smoker and nonsmoker group (p = 0.0285). Methylation of DAPK
has been detected in alveolar hyperplasias in a murine model of lung adenocarcinoma, supporting a role for this gene in the progression of carcinogenesis [30
]. The PAX5β
gene function appears to entail nuclear transcription factors important for cellular differentiation, migration, and proliferation [31
], and methylation is reportedly altered in lung tumors. With work on technical limitations to multiplexing underway in this laboratory, we envision an expanded geneset for more comprehensive assessment of the utility of exhaled DNA methylation biomarkers in classifying phenotypes, and ultimately, assigning the risk status of the epithelium.
Initial DNA methylation mapping projects illuminate both the complex distribution of DNA methylation in the human genome, and the importance of inter-individual variation among DNA methylation profiles from different individuals [32
]. The complexity of methylation map patterns in EBC suggests that comprehensive promoter methylation mapping may be more reflective of the methylation state of a promoter than probe-based methods that sample only 1-4 sites in aggregate, such as MSP. And while chance is possible, the site-specific detail or clustering patterns of more comprehensive methylation map patterns (e.g., DAPK
) may have specific regulatory consequences, particularly when considering broader regions of a gene promoter. Functional studies approaching this hypothesis are ongoing in the laboratory. Such functional studies would be important for optimizing cancer biomarker identification for robustness and precision; and for targeting by genetic or small molecule interventions.
The quantitive MSP analyses of DAPK using two spatially separated probes did show the discordance between methylation at the two designated sites that had originally been mapped as discordant by tBGS. This reinforces the idea that (a) tBGS data is generally concordant with MSP data, based on CpG sites where both assays have been applied; and (b) inference of methylation from one CpG site or region to another is fraught with uncertainty. Additionally, the reasonable correlation between the quantitive MSP and tBGS findings, at each of the two probe sites, was reassuring to the validity of tBGS mapping in these trace exhaled specimens.
For initial confirmation of control status, each control subject who underwent biopsy for clinical indications did also undergo imaging routinely, prior to consideration of dominant lesion biopsy, per clinical routine. This would exclude a significant "missed cancer", other than the one biopsied. Additionally, any subject undergoing a biopsy procedure that had initially negative clinical bronchoscopic biopsies, follow-up surgical or other biopsy procedures and clinical data were tracked for three months from time of enrollment, to reconfirm control status. For those controls not imaged/biopsied by clinical routine, while control misclassification is always a potential problem in case-control studies where some controls are drawn from an at-risk population, with little prospective follow-up, we feel that the thorough vetting of all available clinical and pathologic data in a three month timeframe after enrollment minimized this potential problem. Clearly, prospective follow-up is needed to definitively ascertain outcome, a good design for future more ambitious biomarker studies.
We do not envision exhaled DNA as a method for detection of a small, peripheral tumor. Rather, as field carcinogenesis progresses over the lung epithelia, transforming cells and their debris containing methylated tumor suppressor genes will be shed, marking an increased probability for a lung tumor to arise somewhere, but likely not directly exfoliating from an existing lung tumor in a given deep anatomic location. The exhaled DNA might better be viewed as a whole lung epithelium sampling tool. Therefore, the performance of this biomarker class in predicting lung cancer (i.e. in risk assessment) could be viewed as akin to other "risk factors" for any disease including lung cancer - non-deterministic, but rather informing further early diagnostic, disease detection, and preventive efforts. These speculations, of course, require considerably more extensive cross-sectional and prospective testing.
In summary, non-invasive access of lower airway tissues for DNA methylation studies appears achievable. Our work demonstrates that DNA methylation in EBC is detectable, can be comprehensively mapped, and in piloting a small number of genes, shows some signal that correlates with tobacco exposure, and perhaps with case-control status. If further characterized and anatomically validated, the approach could help facilitate the non-invasive provision of components of human lung epithelia for epigenetic studies of lung cancer and other lung disease pathogenesis and risk assessment.