Determining the presence of asthma-related molecules and salivary contamination in exhaled breath condensate

Exhaled breath condensate (EBC) is comprised of volatile gases (e.g. nitric oxide) [1] and non-volatile compounds such as eicosanoids and cytokines [2]. EBC is increasingly used as a tool for biomarker discovery; it can be obtained non-invasively and reflects the physiology of the airway lining, thereby providing vital information about lung health. Publications have reported on the benefits of EBC as a quick screening tool for lung diseases such as asthma [3‐6], pneumonia [5], chronic obstructive pulmonary disease (COPD) [5‐8], cystic fibrosis [4, 9, 10], and pneumoconiosis [11]; these suggest the potential for EBC to be used in point-of-care diagnostics.

Several EBC studies have focused on eicosanoids due to their known relationship to lung disease and the fact that these compounds are released from mast cells and eosinophils during inflammatory responses [12]. Leukotriene B4 (LTB₄), for example, has been shown to be released by activated alveolar macrophages in sarcoidosis patients [13]. Moreover, levels of cysteinyl-leukotrienes (CysLT) and 8-isoprostane have been reported to be increased in EBC of moderate and severe asthma patients compared to healthy controls [14]. Antczek et al. [15] investigated CysLT, LTB₄, prostaglandin E4 (PGE₄), and 8-isoprostane in longitudinal EBC samples of 16 COPD patients at 4 time points: day 1, during treatment, after therapy, and when stable. Their results showed (1) a decrease in CysLT, LTB₄, and 8-isoprostane after antibiotic therapy and (2) that eicosanoids were elevated in the airways of stable COPD patients compared to healthy subjects. Although numerous studies demonstrate the potential for EBC in point-of-care diagnostics, eicosanoid detection in EBC can be marked by loss of analyte due to adsorption in plastic collection tubes. Therefore, recent investigators have examined the use of glass tubes coated with surfactants, such as Tween 20, as an alternative to improve eicosanoid analysis in EBC [16].

In addition to eicosanoids, other EBC studies have focused on the measurement of nitrogen oxide [17], glucose [10], proteins [18], and amino acids [19]. Amino acids have been shown to be markers of lung function [20], are dysregulated in COPD [21, 22], and are perturbed in smokers [23]. Although amino acids and eicosanoids have been reported in EBC [19, 24, 25], the contribution from saliva remains in question. Saliva contains more than 200 metabolites [26, 27] and has been suggested as a source of contamination in EBC during sample collection. Of the EBC collection devices commercially available, only two (ECoScreen, and TURBO-DECCS) contain saliva traps [28].

Because there is potential for saliva contamination in EBC, many investigators test for saliva contamination by measuring hydrolytic α-amylase activity. Syslova et al [29] investigated CysLTs using a rapid method comprising pre-concentration, stable isotope dilution, and immunoaffinity. Since the α-amylase activity in samples did not exceed 0.1% of the saliva activity, the investigators excluded significant salivary contamination of EBC. However, small amounts of saliva molecules can be detected by liquid chromatography tandem mass spectrometry (LC-MS/MS); Gaber et al reported that the main source of LTB₄ detected in EBC was from saliva [25]. LTB₄ and α-amylase activity were measured in saliva and α-amylase activity was measured in undiluted EBC. The authors observed that spiking EBC with saliva consistently increased LTB₄ levels in EBC and concluded that α-amylase assay may not be sufficiently sensitive to show the presence of saliva in small quantities.

We sought to define the usefulness of EBC in investigating lung disease by characterizing its constituents and delineating if the molecules reportedly detected in EBC are a result of saliva contamination during sample collection. This was achieved using three strategies. First, we used an in-house developed nebulizing system to validate methods and to determine loss of molecules during collection. Second, since they have been reported as increased in lung disease, we specifically measured amino acids and eicosanoids in saliva and EBC in both healthy and asthmatic individuals. Finally, we used untargeted metabolomics and proteomics to determine the components of a highly concentrated EBC sample.

The purpose of this study was to determine the utility of EBC in studying lung diseases, especially asthma. Therefore, we aimed to (1) evaluate whether compound adsorption to collection tubes is the reason for low-to-no detection of metabolites in some EBC studies, (2) characterize the constituents of EBC using untargeted and targeted mass spectrometry, and (3) determine if the detection of molecules in EBC is due to saliva contamination during sample collection. Initial EBC experiments using the RTube for sample collection yielded no detectable levels of leukotrienes in our previous studies (data not shown). Other studies have also showed the RTube to be less sensitive than other commercially available EBC collection devices [36]. Since there is the possibility of binding of compounds to the sides of the collection tube, a control experiment was performed (Fig. 1). Results showed 10-22% less adsorption of the leukotrienes to the TURBO-DECCS plastic tube compared to the plastic RTube at the 10 pg/mL spike levels (Fig. 1d). In addition, polyethylene terephthalate (PET) caused significant adsorption of LTC₄, LTD₄, and LTE₄ to the coated collection tubes (Fig. 1d). Therefore, our clinical experiments used the TURBO-DECCS plastic collection tube without PET to minimize adsorption of compounds to the plastic.

Our first goal was to identify the contribution of saliva to EBC measurements by determining concentrations of amino acids and eicosanoids in both EBC and saliva. We observed that the concentration of several eicosanoids and amino acids (Table 1) were orders of magnitude higher in saliva compared to clean-EBC and saliva-EBC. Although a saliva trap was used, it may be possible that saliva contributed to the ng/mL concentration values obtained in both the clean-EBC and saliva-EBC sample group. As suggested by Gaber et al [25], even the slightest amount of saliva can generate false positives in EBC samples when sensitive detection methods are used. Since LTB₄ is present in the nasal mucosa, and the oropharyngeal tract contributes to the contents of EBC, a more sensitive or alternate method to alpha-amylase detection may be required to confirm saliva contamination. In our study, we detected small molecules in EBC and saliva-EBC which were not detected in the saliva-only samples. This suggests that these molecules may be specific to EBC or have much lower concentrations in saliva; therefore, the low saliva content would not contribute to their detection in EBC. These EBC constituents may potentially be used as biomarkers; however, further study is required to confirm their identities and usefulness in studying lung disease.

Previous investigations have reported the presence of eicosanoids in EBC [11, 15, 37, 38]. Many of these groups have reported higher concentrations of eicosanoids than those shown in Table 1 of the current study; the majority of these studies were conducted in the context of lung disease such as asthma where elevated levels may due to inflammation. The healthy volunteers in our study had no history of asthma or lung inflammation; this could explain the lower levels which we report. We compared our concentration values of amino acids and eicosanoids in Table 1 to those available in the literature for EBC and saliva. LTB₄ has been reported to have an average concentration in saliva of 0.467 ng/mL [25] which is similar to our estimated value of 1.56 ng/mL. Tyrosine has been reported at an average of 33.3 ng/mL [24] in EBC while Conventz et al [19] reported an average value of 15.5 ng/mL. We detected tyrosine at 27.1 ng/mL in ‘saliva-EBC’ but it was undetected in ‘clean-EBC’. Our proline estimate was 111.4 ng/mL in EBC. This was twice as high as the previously reported high value in healthy subjects of 51.9 ng/mL [19] and may be the result of individual subject differences, sample preparation, or detection method. Reported EBC urea values have similar ranges within the same order of magnitude to our calculated values. Effros et al [39] reported values in the range 0.33 to 0.39 μmol/L (19.8 to 23.4 ng/mL). Folesani et al [40] reported EBC urea concentrations found in healthy controls in the range 0.7-1.3 μM (42.0 to 78.1 ng/mL). Dwyer et al [41] reported EBC urea averaged 0.52 +/- 0.12 μmol/L (31.2 ng/mL) in 18 individuals. Our detected levels ranged from 1.56 ng/mL (clean-EBC) to 65.1 ng/mL (saliva-EBC) which falls within the same order of magnitude of these previously reported values. The differences among studies could be accounted for by variations in sample preparation procedures such as lyophilization, as well as differences in methods of collecting EBC.

Our second goal was to characterize the constituents of EBC using an untargeted metabolomics approach. We further determined the effect of saliva contamination on small molecules detection in EBC. The 77 small molecules that were detected in the clean-EBC and the saliva-EBC were undetected in the saliva samples; this suggests that these compounds may be specific to EBC. The 40 out of 77 that were database annotated using Human Metabolome Database (HMDB) [42] and the Kyoto Encyclopedia and Genes and Genomes (KEGG) [43] included the following: vitamin D metabolites, lipids, herbs, spices, food, plants, insecticides, herbicides, dipeptides, and PAH degradants. Five out of fourteen food metabolites were related to citrus fruit or tea, three were herbs and spices, and the remaining were vegetables, food flavorings, or oils. This is consistent with participants’ reports of drinking green tea, coffee, and chocolate milk, and eating pumpkin cake, green beans, club sandwich, and potato chips. Note that participants were not allowed to eat or drink at least two hours prior to sample collection. Other matches for two insecticides and three herbicides were also plausible since these samples were collected when subjects may have applied insect repellants such as diethyltoluamide (DEET) to their skin, and institutions were applying herbicides and pesticides to their lawns. For example, N-cyclopropylammelide is a degradation product of atrazine. Atrazine is an herbicide used to prevent weeds on golf courses and residential lawns.

Other metabolites found in EBC were tentatively identified as endogenous compounds; for example, 3-oxooctanoic acid is an endogenous keto acid involved in fatty acid biosynthesis. It is formed by the action of acid synthases from acetyl-CoA and malonyl-CoA precursors. PE(44:7) is an endogenous glycerophospholipid associated with cell signaling and membrane integrity and also serves as an energy source [42]. Ganglioside GM₃ (d18:0/20:0) is an endogenous sphingolipid. Gangliosides, including GM₃ and GM₂, have been shown to be down-regulated in the hyper-reactive lung and trachea compared to the normal lung and trachea in a guinea pig model of bronchial asthma [44]. Gangliosides have also been shown to be inversely associated with severe emphysema in COPD human plasma [45]. Dipeptides were also annotated. These included glutamyl-glycine, methionyl-arginine, tyrosyl-histidine, and phenylalanyl-histidine. Dipeptides are incomplete breakdown products of protein digestion or protein catabolism. Many dipeptides are short-lived intermediates toward specific amino acid degradation pathways while others have physiological [46] or cell-signaling effects [42].

3,4-dihydroxyfluorene is a polycyclic aromatic hydrocarbon (PAH) degradation metabolite. PAH’s have been associated with childhood asthma [47] and are found in oil, coal, and tar. They are the result of combustion in engines, and incinerators; sources include forest fires, vehicle exhaust, grilling or barbecuing meat, and smoked fish [48‐50]. Lastly, there were two annotated vitamin D metabolites (25-hydroxyvitamin D₂-25-glucuronide and 3-deoxy-3-azido-25-hydroxyvitamin D₃). Vitamin D and its metabolites are associated with asthma [51, 52] and vitamin D deficiency has been shown to be a risk factor for developing asthma [53, 54]. Because these compounds have previously reported associations with lung disease, they could either be used as diagnostic markers of health or disease state, or may be novel molecules requiring further interrogation.

Due to low sample volumes, tandem mass spectrometry was only performed in negative mode and resulted in six spectral library matches corresponding to acetylsalicylic acid, 4-chloro-L-phenylalanine, 3,4-furandicarboxylic acid, shikimic acid, succinic acid, and citric acid (Additional file 5). With additional sample, targeted MSMS can be performed on additional EBC metabolites to further explore their identities. Other metabolites present may be below our limit of detection or require more specialized sample preparation techniques. Therefore, detection of specific molecules or classes, may require derivatization, solid phase extraction, or enzymatic techniques, as described by Chérot-Kornobis et al [17] for nitrogen oxides, Esther Jr. et al [55] for purines, or Rossi et al [56] for glutathione.

A major objective of the current study was to determine if eicosanoid detection in EBC was the result of saliva contamination. With the exception of one molecule in one subject, we were unable to detect eicosanoids in EBC of over 100 asthmatic subjects, a group in which higher concentrations of these molecules have been reported (Fig. 4). Our EBC sample preparation step included diluting the EBC volume to 1 mL; this may have diluted levels to below the limit of detection for the targeted eicosanoid panel. However, the concentration of eicosanoids ranged from 10-fold to 100-fold lower in EBC compared to saliva from matched asthmatic subjects (Fig. 4). Since our sample preparation methods, including low starting volumes, were consistent with previous investigations reporting the detection of eicosanoids in EBC, this suggests that previously reported values could possibly be due to salivary contamination. This is further supported by the detection of some eicosanoids when a very highly concentrated (13 ml) EBC sample is used. Within the asthmatic saliva samples, the concentrations varied in the range 58-187% CV for the samples with detectable levels of eicosanoids. These results show that eicosanoid concentrations in saliva vary widely amongst asthmatic subjects.

Due to the poor detection of eicosanoids in EBC using targeted analysis and low starting volumes, we investigated whether any other small molecules could be detected in EBC when larger starting volumes are used. Leftover EBC from asthmatic subjects (n = 107) was pooled into a 13 mL aliquot, lyophilized, and reconstituted in 20 μL of HPLC buffer. Untargeted LC-MS revealed 97 metabolites, of which four were eicosanoid derivatives: LTE₃, thromboxane, 11-trans-LTE₄, 11-trans-LTC₄, and 12-oxo-LTB₄ (Additional file 6). Results suggest that although eicosanoid metabolites are present in EBC, samples require a drastic pre-concentration step prior to LC-MS analysis in order for these molecules to be detected. In addition, some of these detected metabolites may be breakdown products of the deuterated standards spiked during sample preparation, particularly during the lyophilization step.

In a few studies, investigators increased the amount of EBC used to 1 mL and obtained detectable metabolite signal. For example, Pelclová et al [11] collected EBC over 15-20 min using the ECoScreen. They analyzed EBC of 82 patients with occupational lung diseases & 27 controls were analyzed using SPE followed by LC-ESI-MS/MS. Results showed elevation of LTB₄, LTC₄, and LTE₄ in asbestos-exposed and silica-exposed patients compared to controls. Conventz et al [19] collected 1-6 mL EBC from 27 healthy adults also using the ECoScreen. 1 mL EBC was used to quantify proline, hydroxyproline, and tyrosine using LC-ESI-MS/MS. Fritscher et al [6] collected at least 1 mL of EBC from 87 subjects for over 10 min and performed targeted analysis on a QQQ-MS. They examined twenty-three eicosanoids in the EBC from asthma and COPD individuals, five of which overlapped with our study. Our study examined an additional eleven molecules that were not present in their study.

A major difference between our study and others is that our aim was specifically to determine the potential for and extent of saliva contamination in EBC sampling. Therefore, our studies incorporated a well-controlled set of experiments that included mimicking EBC collection. Overall, we required greater than 1 mL EBC for detection of eicosanoids; preparation included concentrating samples in a lyophilizer. Other investigators have also concentrated their EBC samples prior to analysis. Montuschi et al [57] collected 1.5 mL EBC per subject over 15 min using the ECoScreen and concentrated the EBC 40-fold. 20 μL of sample was injected and analyzed for LTB₄ using LC-MS & LC-MS/MS.

Our methods were replicated in an independent cohort. EBC was collected from 13 volunteers and grouped in four categories based on health and smoking status: healthy non-smoker, healthy smoker, non-smoker with common cold, and non-smoker with nasal congestion. Saliva contamination was not present, as measured by a proteomic approach. Using untargeted LC-MS metabolomics, 172 metabolites were detected in EBC, 81 of which were present in all 4 sample groups (Fig. 5a). Fold changes were observed (Fig. 5b) but no statistical inferences could be made. Qualitatively however, the healthy EBC contained fewer compounds compared to the EBC of subjects with nasal congestion, the common cold, or the smoker (Fig. 5a). This may be due to differences in diet, or because subjects with signs of illness may ingest cold, flu, or nasal decongestion medication which may artificially increase the number of detected compounds in their EBC.

Some of the detected metabolites in the smoker, healthy, common cold, and nasal congestion EBC (Table 3 and Additional file 7) were markers of environmental exposure such as 1-hydroxy-2-naphthoic acid which is a PAH and naphthalene degradation metabolite. This metabolite also mapped to the “degradation of aromatic compounds” pathway, as did 6-hydroxy caproic acid and p-cymene. Ephedrine, used as a decongestant, was only detected in the common cold group. Additional detected metabolites such as 2-Oxo-4-hydroxy-5-aminovalerate, homoserine, arogenate, tetrahydrodipicolinate, and 2-amino-3,7-dideoxy-D-threo-hept-6-ulosonic acid play roles in amino acid metabolism and/or biosynthesis [58]. Many of these compounds were absent in the healthy EBC but were present in the EBC of individuals who smoked, had nasal congestion, or had the common cold. Other detected metabolites were part of purine metabolism [ureidoglycine], sphingolipid signaling pathway [C14 sphingosine], glycerolipid metabolism [PA(22:2), MG(18:1)], and glycerophospholipid metabolism [PC(40:7), PG(38:8), PA(22:2), PI(24:0)]. The lipid compounds PA(22:2) and PI(24:0) were undetected in the healthy EBC but were detected in the other three groups. These biological pathways have been implicated in lung diseases such as asthma [59] and COPD [45, 60].

Targeted eicosanoid analysis was also performed on the healthy, smoker, common cold, and nasal congestion EBC. Nineteen eicosanoids were detected in the smoker EBC which were undetected in the other groups. Elevated levels of 13-HODE, 13-OxoODE, 9-HODE, and 9-OxoODE were observed in both the smoker and nasal congestion EBC compared to the healthy and common cold EBC. This increase in eicosanoids in smoking samples has been observed in previous studies. Sanak et al [61] analyzed EBC from 17 healthy smokers and 41 healthy non-smokers collected using the ECoScreen. Results showed an increase in 5-HETE and 8-iso-PGF2α in the current smokers compared to the non-smokers. We conclude that cigarette smoking increases the inflammatory and oxidative stress markers observed in our study to levels that were at least 3-fold higher compared to the healthy EBC.

Proteomics analysis of these four groups detected few proteins in the EBC samples using both human and bacterial searches (Table 4). In a review in 2014, Harshman et al. [18] summarized 80 detected proteins in EBC from the current literature including one detected in our study. Although no proteins were detected in our nasal congestion or common cold EBC samples, keratin type I cytoskeleton 9 (Cytokeratin-9) was detected in healthy EBC. Cytokeratin-9 has been previously been identified in asthmatic EBC [62], and in the pooled EBC of non-smokers and healthy smokers [63]. However, others have suggested that cytokeratin in EBC is the result of ambient air rather than the airways [64]. We detected zinc finger protein 800 and myoneurin in the smoker EBC. Zinc finger protein 800 and myoneurin (also a zinc finger protein) have not been previously reported in EBC. Other zinc finger proteins have been detected; zinc finger CCCH domain-containing protein 4 (ZC3H4) has been reported in healthy non-smoker and healthy smoker EBC [63]. In our study, no salivary proteins were detected, indicating that the EBC samples from the replication study were not contaminated with saliva. A larger, more diverse cohort encompassing multiple lung diseases may be required to explore the diversity of exhaled proteins.

This study is particularly significant to EBC researchers because it emphasizes the issues of compound adsorption, saliva contamination, and high volumes of EBC required for biomarker discovery studies. The strengths of this study lie in the precise methods used and the large sample size of Cohort 2 in the asthmatic study. We recognize that some limitations exist. First, the sample number is limited for Cohorts 1 and 3. Second, subjects in these cohorts only refrained from food intake for 2-3 h rather than 12 h, which could explain some differences in metabolites detected. Lastly, due to limited sample volumes, additional analyses could not be performed to compare the three cohorts across all mass spectrometry technologies. Future studies could be aimed at rectifying these limitations.

We conclude that measureable levels of small molecules, including amino acids and eicosanoids, are present in healthy EBC; however, the dilute nature of EBC requires larger volumes of starting material than currently reported in the literature. We suggest the collection of at least 15 mL of EBC per subject and pre-concentrating by at least 20-fold to as much as 500-fold prior to LC-MS analysis in order to confidently and reproducibly detect metabolites of interest. Secondly, although α-amylase assays can test for the presence of saliva in EBC, small volumes of saliva may still be present but be below the detection limits of the assay. Since saliva can be responsible for contaminating EBC samples, proper sample collection and handling is necessary, particularly the use of a saliva trap during sample collection. Thirdly, eicosanoid concentrations in saliva vary widely amongst asthmatic subjects and this should be considered when designing experiments. Here, we provide a general presentation of EBC constituents from which investigators can probe more specialized techniques to detect additional or lower abundant compounds of interest. These results suggest that large volumes of samples and a more targeted approach are needed when using EBC to study asthma and other lung diseases.