Standardization procedures for real-time breath analysis by secondary electrospray ionization high-resolution mass spectrometry

Abstract

Despite the attractiveness of breath analysis as a non-invasive means to retrieve relevant metabolic information, its introduction into routine clinical practice remains a challenge. Among all the different analytical techniques available to interrogate exhaled breath, secondary electrospray ionization high-resolution mass spectrometry (SESI-HRMS) offers a number of advantages (e.g., real-time, yet wide, metabolome coverage) that makes it ideal for untargeted and targeted studies. However, so far, SESI-HRMS has relied mostly on lab-built prototypes, making it difficult to standardize breath sampling and subsequent analysis, hence preventing further developments such as multi-center clinical studies. To address this issue, we present here a number of new developments. In particular, we have characterized a new SESI interface featuring real-time readout of critical exhalation parameters such as CO₂, exhalation flow rate, and exhaled volume. Four healthy subjects provided breath specimens over a period of 1 month to characterize the stability of the SESI-HRMS system. A first assessment of the repeatability of the system using a gas standard revealed a coefficient of variation (CV) of 2.9%. Three classes of aldehydes, namely 4-hydroxy-2-alkenals, 2-alkenals and 4-hydroxy-2,6-alkedienals-hypothesized to be markers of oxidative stress-were chosen as representative metabolites of interest to evaluate the repeatability and reproducibility of this breath analysis analytical platform. Median and interquartile ranges (IQRs) of CVs for CO₂, exhalation flow rate, and exhaled volume were 3.2% (1.5%), 3.1% (1.9%), and 5.0% (4.6%), respectively. Despite the high repeatability observed for these parameters, we observed a systematic decay in the signal during repeated measurements for the shorter fatty aldehydes, which eventually reached a steady state after three/four repeated exhalations. In contrast, longer fatty aldehydes showed a steady behavior, independent of the number of repeated exhalation maneuvers. We hypothesize that this highly molecule-specific and individual-independent behavior may be explained by the fact that shorter aldehydes (with higher estimated blood-to-air partition coefficients; approaching 100) mainly get exchanged in the airways of the respiratory system, whereas the longer aldehydes (with smaller estimated blood-to-air partition coefficients; approaching 10) are thought to exchange mostly in the alveoli. Exclusion of the first three exhalations from the analysis led to a median CV (IQR) of 6.7 % (5.5 %) for the said classes of aldehydes. We found that such intra-subject variability is in general much lower than inter-subject variability (median relative differences between subjects 48.2%), suggesting that the system is suitable to capture such differences. No batch effect due to sampling date was observed, overall suggesting that the intra-subject variability measured for these series of aldehydes was biological rather than technical. High correlations found among the series of aldehydes support this notion. Finally, recommendations for breath sampling and analysis for SESI-HRMS users are provided with the aim of harmonizing procedures and improving future inter-laboratory comparisons. Graphical abstract.

Keywords: Breath metabolomics; Fatty aldehydes; Oxidative stress; Secondary electrospray ionization high-resolution mass spectrometry; Standardization procedures; Variability.

Graphical abstract

Fig. 1

Real-time breath analysis using SESI-HRMS. a SESI-HRMS analytical platform located in a clinical setting (University Children’s Hospital Basel) dedicated for real-time breath analysis. The system features three main elements: (i) exhalation interface, which provides feedback to the participants on the exhalation maneuver; (ii) ion source, which efficiently ionizes exhaled metabolites, and (iii) high-resolution mass spectrometer. b Real-time analysis by simultaneous monitoring of CO₂, physical exhalation parameters (exhalation flow rate and exhaled volume), and relative intensities of three representative aldehydes from one experiment. 13 consecutive exhalations within 20 min for one subject are shown (see ESM Fig. S4 for zoomed-in view of the first exhalation).

Fig. 2

Aldehydes show a subject-independent and molecule-dependent exhalation pattern. Data shown is the mean normalized breath-signal with errors bars representing 95% CIs for three selected aldehydes from four subjects (denoted as S1–S4) in 104 experiments as a function of exhalation number. Lighter species show a systematic decaying trend across consecutive exhalations, which is subject independent.

Fig. 3

Dependency of exhalation profile of breath metabolites with blood-to-air partition coefficient. a Exhalation profiles of short and long aldehydes as a function of exhaled volume is consistent with the hypothesis that the shorter aldehydes exchange mostly in the airways, while longer aldehydes exchange in the alveoli. 4-hydroxy-2-dodecenal shows a deviating pattern that may be caused by an interfering peak. b Relative difference between breath-signals from last and first exhalation as a function of predicted blood-to-air partition coefficient. A number of carbon atoms for molecules are shown at the top and gray dashed curves shows the 95% CI from λ_b:a estimation. A large partition coefficient is associated with a strongly decaying pattern (also see ESM Fig. S10, with x-axis on log₁₀ scale, showing the complete range for 95% CIs)

Fig. 4.

Evaluation of breath mass spectra at varying exhalation flow rates and using spirometry filters. a Breath-signals of exhaled metabolites are independent of exhalation flow rate. As seen by the comparison of signals from two experiments with slower (9.8 ± 0.1 L/min) and faster (12.0 ± 0.3 L/min) exhalation flow rates. b Use of filters does not significantly affect the breath-signals of exhaled metabolites. As seen by the comparison of signals from two experiments with and without the presence of an antibacterial/antiviral filter. In both panels, solid gray horizontal line represents the mean and dotted gray horizontal lines represent mean ± 1.96 × SD.

Fig. 5

Variability of SESI-HRMS breath mass spectra are dominated by inter-individual differences, rather than by batch effect. PCA score plot of all measurements with labels according to measuring day (left) and subject number (right). Grouping according to subject number is more evident than by measurement date.

Fig. 6.

Positive correlation among aldehydes suggests a common origin of mechanism of generation. Correlation network (considering Spearman’s r ≥ 0.85) with an average node degree of 4 ± 2. Note that 4-hydroxy-2-decenal, 2-octenal and 2-decenal do not pass the correlation cutoff and hence are shown at the bottom-left side. Node shape and color are based on the classes of aldehydes, whereas edge width and color depends on the correlation coefficient, as shown in the legend at the bottom-right side.