Metabolic effects of inhaled salbutamol determined by exhaled breath analysis

A total of 98 participants were included in this study: 48 inhaled placebo and 50 inhaled salbutamol. Of the 48 subjects inhaling placebo, eight suffered from asthma and 40 did not have any lung disease. Of the 50 patients inhaling salbutamol, 13 suffered from asthma and 37 suffered from moderate to severe COPD. The anthropometric data is shown in table 1. Participants were asked to refrain from eating, drinking (except water), chewing gum, smoking and brushing their teeth at least 1 h prior to measurement. The study was approved by the local ethical committee (PB_2016-00421, KEK-ZH-Nr. 2015-0148, KEK-ZH-Nr. 2014-0088) and all subjects gave written, informed consent to participate.

Exhaled breath analysis was performed at baseline, 10 and 30 min after the inhalation of 200 μg salbutamol (n = 50) or placebo (n = 48). The placebo contained only norflurane. Figure 1(a) shows the timing used in the study. Figure 1(b) displays a picture of the modified mass spectrometer (Sciex Triple TOF 5600+) to allow breath analysis by SESI-HRMS. The breath analysis approach has been described in greater detail elsewhere [19, 20]. In brief, the atmospheric pressure ion source of a quadrupole-time-of-flight mass spectrometer was replaced by a lab-built SESI chamber [21, 22]. A nano-electrospray of 0.1% formic acid in water was infused at ∼200 nl min⁻¹ through a silica capillary (PicoTip emitter—O.D. 360 μm–I.D. 20 μm). The electrospray was operated in the cone-jet mode [23, 24] and its current (∼100 nA) constantly monitored. In addition, a lens was used sporadically to visually inspect the Taylor cone [25]. The breath sampling tube connected to the SESI chamber consisted of stainless steel (6 mm I.D.), and was heated to 80 °C to prevent water condensation and adsorption of low-volatility metabolites onto the walls. In order to avoid confounding factors such as differences in exhalation maneuvers [26], the sampling protocol was standardized across all participants. The subjects would typically provide five–six replicate exhalations. Full exhalations were provided at a constant flow rate by keeping the pressure in the sampling line at 12 mbar (regulated by bio-feedback by the breathing subjects). Measurements were recorded in positive and negative ion mode, leading to mainly protonated and deprotonated species, respectively. The mass range in positive ion mode was 40–700 Da and in negative ion mode 40–450 Da. Accumulation time was set to 1 s.

Zoom In Zoom Out Reset image size

Identification of 10-hydroxydecanoic acid was performed using on-line MS/MS experiments as well as UHPLC-MS/MS experiments of exhaled breath condensate (EBC) and a standard. EBC was collected using a sampling device constructed in-house following the guidelines set by the ATS/ERS task force [27]. Two healthy, non-smoking subjects were asked to exhale into a glass cold trap. The collected EBC samples were pooled (V_total = 32 ml), 0.5 ml were extracted and stored at 4 °C and the remainder lyophilized to dryness [28]. The residue was then reconstituted in a mixture of 142.5 μl of the previously aliquoted EBC, 7.5 μl methanol and acidified with 0.1% formic acid. This up-concentrated sample was directly subjected to UHPLC-HRMS/MS analysis. UHPLC separation was done on an ACQUITY UPLC system (Waters, MA, USA) with a C18 column (1.7 μm, 2.1 × 50 mm, Waters) with pre-column filter. The flow rate was set to 500 μl min⁻¹ using a binary mixture of solvent A (water with 0.1% formic acid) and solvent B (methanol with 0.1% formic acid). The 10-hydroxydecanoic acid standard (Apollo Scientific Ltd) was prepared at a concentration of 100 ng ml⁻¹ and analyzed with an injection volume of 10 μl. A Triple TOF 5600 + (AB Sciex, Concord, ON, Canada) mass spectrometer was used to measure a mass range of 40–500 Da in negative ion mode, with 1–5 ppm mass accuracy. Collision energy was set at 40 ± 30 eV.

Raw *.wiff (Sciex's proprietary format) data was converted into *.mzXML format via MSConvert (Proteowizard) [29]. Each sample file was interpolated linearly and centroided, generating as a result a m/z list of 1208 peaks present in at least 40% of the samples. The mass spectra were normalized dividing by the sum of a set of 904 signals found to be present in 90% of our samples. We then computed the log₂ of the fold change (i.e. ratio between measurements after 10 min over baseline and after 30 min over baseline). Significant differences in exhaled metabolite levels in response to salbutamol, but not to placebo inhalation were sought. A paired two-tailed t-test followed by estimation of false discovery rate (FDR) in multiple comparisons was used for this purpose [30]. Statistical significance was set to FDR < 0.05. We further considered only the signals changing significantly after 10 and 30 min of salbutamol inhalation, but not changing significantly after placebo inhalation. Among the significant ones, we report here only those signals changing by at least log₂(fold change) >0.15 after 10 or 30 min of inhaling salbutamol. The signals satisfying these conditions were subjected to an agglomerative hierarchical cluster tree (Ward method; Euclidean distance). In order to identify groups of signals showing similar behavior, we constructed agglomerative clusters using a cutoff Euclidean distance of 5. Finally, for the set of molecules that were chemically identified, we computed the correlation coefficients between pairs of variables for the group inhaling salbutamol.

Subjects' breath was analyzed, detecting as a result 1208 features (866 and 342 features in positive and negative mode, respectively) present in at least 40% of the samples. A total of 131 mass spectral features were found to be significantly altered in response to salbutamol inhalation, but not to placebo inhalation. Table S1, available online at stacks.iop.org/JBR/11/046004/mmedia lists all of these along with their mean changes, 95% confidence intervals, p-values and FDR. Figure 2 shows the histograms of the log₂(fold change) for the 131 significantly altered ions after 10 and 30 min of inhalation of salbutamol. The distributions for the placebo group are fairly symmetrical around 0 (i.e. mean 10 min after placebo = 0.0254; mean 30 min after placebo = 0.0220). In contrast, the salbutamol distributions are flatter, and their means are displaced to the right (mean 10 min after salbutamol = 0.1458; mean 30 min after placebo = 0.1963), suggesting a general increase of these exhaled compounds.

Zoom In Zoom Out Reset image size

Among 131 altered signals, 121 increased, while 10 signals were reduced after salbutamol inhalation, but did not change significantly in the placebo group. Figure 3 shows the breath-to-breath time trace for one selected compound (m/z 259.1903; C₁₄H₂₆O₄) for one subject before and after inhalation of placebo (a) and for a different subject inhaling salbutamol (b). The five–six replicate measurements within each time point show a satisfactory repeatability (4.8% average coefficient of variation). Clearly, the breath levels of this compound for this particular patient increased significantly shortly after salbutamol inhalation, and this change could be easily captured by on-line breath analysis. In contrast, the levels for the subject inhaling just placebo did not vary significantly. The overall picture of the trends for this compound for the placebo and medication groups is shown in figures 2(b) and (c), respectively. Mean log₂(fold change) for the placebo group after 10 and 30 min was 0.01 (95% CI = −0.06/0.07; FDR = 0.93) and −0.02 (95% CI = −0.12/0.07; FDR = 0.7), respectively. In contrast, mean log₂(fold change) for the salbutamol group after 10 and 30 min was 0.19 (95% CI = 0.10/0.28; FDR = 4.84 × 10⁻⁴) and 0.26 (95% CI = 0.12/0.4; FDR = 1.44 × 10⁻³), respectively. However, despite the significant changes for this particular molecule (molecular formula C₁₄H₂₆O₄) in the salbutamol group, there is clear inter-individual variability, with some subjects experiencing no changes at all. This seems to be in line with the known lack of response to medication for large parts of the population. For example, it has been estimated that a daunting 75%–96% do not respond to the top-ten grossing drugs in the market. Our data suggests that breath analysis has potential to contribute to the identification of subpopulations of responders and non-responders, at least for certain drugs such as bronchodilators.

Zoom In Zoom Out Reset image size

We attempted the chemical identification of some of the exhaled metabolites, to provide a biochemical interpretation of our results. First, we computed an agglomerative hierarchical cluster tree (Ward method; Euclidean distance) and subsequently constructed clusters from the hierarchical cluster tree using as cutoff a distance of 5. This resulted in 26 clusters of signals behaving similarly. The resulting clusters are listed in table S1. As a result, it became apparent that homologous series of closely related molecular formulae (and most likely compounds) tend to cluster together. This is a first indication that this real-time technique is capable of capturing metabolic cascades in response to medication. For example, cluster number 16 consisted of acetic, propionic and butyric acids, which were in all cases found to be increased after salbutamol inhalation. Similarly, heptanoic, octanoic and nonanoic acids [21] clustered together (number 22) and were found to increase after salbutamol inhalation. Another interesting series of compounds was cluster number 10. They all have similar molecular formulae and were detected in negative ion polarity. Compounds detected by SESI-MS operated in negative ion mode are dominated by deprotonated acids [21, 32]; thus, we hypothesize that these compounds are likely fatty acids. This was confirmed by deploying a comprehensive analytical strategy. Figure 4 shows experimental evidence indicating that the molecular formula C₁₀H₂₀O₃ corresponded to 10-Hydroxydecanoic acid. Figure 4(a) shows the extracted ion chromatograms at m/z 187.1339 in negative ion mode for EBC and the standard of 10-hydroxydecanoic acid. Figure 4(b) shows a head-to-tail MS/MS for EBC and the standard. The perfect match in retention time and fragmentation pattern provides strong evidence for 10-hydroxydecanoic acid in exhaled breath condensate. Figure 4(c) shows the fragmentation spectrum at m/z 187.1339 during a real-time exhalation. The diagnostic fragments of this compound were indeed present, indicating that 10-hydroxydecanoic acid is present in exhaled in real-time and detected by SESI-MS. However, it is obvious that other fragments are also present, because the isolation window of the quadrupole is 1 Da, leading to the fragmentation of all the species present at nominal mass 187 Da. It is also interesting to note that different isomers of 10-hydroxydecanoic acid were found in breath condensate (figure 4(a)). For example, the most abundant peak in the chromatogram at retention time ∼3.55 min, corresponds to 3-hydroxydecanoic acid (Gaugg et al in preparation), which is an intermediate in fatty acid biosynthesis. It is therefore very likely that all different isomers are exhaled simultaneously.

Zoom In Zoom Out Reset image size

To better understand the interactions of 10-hydroxydecanoic acid with the rest of the compounds in the same cluster we computed their correlations (r) for the salbutamol group and plotted the correlation network. Figure 5(a) shows the results (only r > 0.65 are shown). Overall, we found strong positive correlations among the different nodes (i.e. molecules). The strongest correlation (r = 0.86) was between C₁₀H₁₈O₄ and C₁₀H₁₆O₄. The former of these is sebacic acid (Gaugg et al in preparation), and the latter an unsaturated form of it. As to the identified hydroxydecanoic acid, the network shows a correlation with sebacic acid (r = 0.7), C₉H₁₄O₄ (r = 0.66) and C₁₀H₁₈O₃ (r = 0.69). The striking correlation between these compounds is demonstrated in figure 5(b), which shows the log₂(fold change) of hydroxydecanoic vs. log₂(fold change) for C₁₀H₁₈O₄, C₉H₁₄O₄ and C₁₀H₁₈O₃. While the complete network remains to be fully characterized, overall, our breath analysis data are consistent with previous studies that describe significant increase of non-esterified fatty acids [33], and in general with the lipolytic effects of beta-adrenergic agonists [34]. Moreover, these changes in fatty acid concentrations are known to happen very rapidly, in the order of minutes [35, 36], in line with our own observations here. Additional correlation networks are shown in figure S1.

Zoom In Zoom Out Reset image size

We have just discussed that SESI-HRMS seems to be sensitive and selective enough to capture significant changes of exhaled metabolites such as fatty acids shortly after salbutamol inhalation, in contrast to a placebo group. We argue that such information may provide a more comprehensive understanding of the mechanism of action of drugs allowing, for example, responders and non-responders to medication to be identified [37]. Obviously, monitoring of the administered drug itself would complement such metabolic information. In this regard, we have recently shown the potential of SESI-HRMS as a rapid and non-invasive platform to perform pharmacokinetic studies of injected drugs by measuring them in vivo, in mouse breath [38, 39]. Initially, we did not observe obvious changes in salbutamol levels above the background levels. One explanation could be that the chemical background in the m/z region where the salbutamol is expected (i.e. m/z 240.1594; [C₁₃H₂₁NO₃ + H]⁺) was generally very crowded. Figure 6(a) shows representative mass spectral data at nominal mass 240 along with the chemical structure of salbutamol. For reference, the theoretical exact mass where protonated salbutamol would be expected is indicated with a vertical line. Clearly, even at mass resolution as high as ∼20 000, the separation capability is insufficient to discriminate salbutamol from other isobaric species [20, 40]. We nevertheless inspected the time traces for protonated salbutamol, as well as for its main metabolite, salbutamol-4-O-sulfate (figure 6(b); m/z 320.1162; [C₁₃H₂₁NO₆S + H]⁺) [41]. Figure 6 shows the time traces for salbutamol (a) and salbutamol-4-O-sulfate (b) for the 40 healthy controls that had never inhaled salbutamol and 13 patients that according to their medical records inhale salbutamol on an as-needed basis. The first four of the latter were COPD patients, and nine suffered from asthma. Note that the traces shown correspond to the baseline measurement, when salbutamol was not yet inhaled. While we cannot give definite proof that these two mass spectral peaks indeed correspond to salbutamol and its main metabolite, the high correlation between the two sets of time traces suggests an association. Also, the intensity of the time traces seems in general higher in the patients, especially in the COPD group.

Zoom In Zoom Out Reset image size

To further investigate whether common bronchodilators can be detected in breath by SESI-HRMS, we expanded our data analysis. In particular, according to their medical records, fourteen of the patients were inhaling formoterol on an as-needed basis. Formoterol is a long-acting beta2-agonist used in the management of asthma and COPD. Its chemical structure is shown in figure 7, along with a typical mass spectrum in the region of m/z 345. The exact mass at which this molecule would be expected in protonated form [C₁₉H₂₄N₂O₄ + H]⁺ is m/z 345.1809, which is indicated in the mass spectrum by a vertical bar. As in the case of salbutamol, many isobaric species prevent us from clearly resolving the expected peak. However, similarly to salbutamol, the time traces tend to be higher in the patients using the medication on an as-needed basis as compared to the healthy controls who did not inhale any medication. Further investigations via UPLC-MS/MS analysis of EBC and higher resolution mass spectrometry are required to confirm that salbutamol and formoterol can be detected in breath. However, the evidence shown in previous studies where drugs such as ketamine (MW 237 Da) were unambiguously detected in breath and correlated with plasma levels [38, 39] and the data shown in figures 6 and 7 strongly supports the idea that SESI-HRMS is suitable to monitor drugs and their metabolites in breath. If this hypothesis is confirmed, it additionally opens attractive possibilities to screen for doping agents in breath of athletes. Note in this regard that all beta2-agonists are currently prohibited in and out of competition by the World Anti-Doping Agency [42].

Zoom In Zoom Out Reset image size

While this work suggests that breath analysis could contribute to ongoing efforts to elucidate response mechanisms to therapeutic intervention and possibly drug monitoring, this study has—of course—limitations. For example, the unambiguous identity of the molecules significantly altered, precludes gaining further insights on reprogrammed metabolic pathways associated with salbutamol. In this regard, we cannot entirely exclude the possibility that some of the changes observed are actually due to bronchial dilation, resulting in improved gas exchange at the lung level.