Exhaled isoprene for monitoring recovery from acute hypoxic stress

All human volunteer test subjects were male, non-smoking, active duty military in the United States Air Force with ages ranging from 22 to 39 (supplemental data 1 is available online at stacks.iop.org/JBR/11/047111/mmedia). The study was approved for human use by the Institutional Review Board at the Naval Medical Research Unit-Dayton (NAMRUD.2015.003). Written consent was obtained from each test subject prior to initiation of the experiment.

Each volunteer test subject was placed in a modified 20 P flight mask and HGU 68/P flight helmet as shown in supplemental data 2(A) [2]. Oxygen concentration was controlled to the flight mask via ROBD (Evironics, Tolland CT, supplemental data 2(B)). Blood pressure, pulse and blood oxygen saturation (SpO₂) were recorded as shown and described in supplemental data 2(A). Test subjects were exposed to two separate normobaric simulated flight profiles on different days. Each exposure consisted of three sequential five-minute periods of sea-level oxygen (21% O₂), 25 000 feet equivalent oxygen (7620 m, 8% O₂) and either sea-level (21% O₂) or 100% O₂ recovery (supplemental data 2(C)). Test subjects were progressed to the recovery gas if their individual SpO₂ fell below 55% or upon their request. A summary of the exposure data and traces of test subjects' SpO₂ are provided in supplemental data 3(A) and (B). A timeline of the experiment illustrating collection periods for bags and time series samples also indicates duration of exposure for each of the O₂ conditions (supplemental data 4).

Exhaled breath samples were collected directly on to preconditioned stainless steel Tenax TA thermal desorption (TD) tubes affixed with brass caps having polytetrafluoroethylene ferrules when not in use (Markes International, South Wales, UK).

Exhaled breath was collected serially from each test subject throughout the exposure, from each test subject (n = 8) from the exhalation port side of the modified flight mask, as shown in supplemental data 5(A). Exhaled breath volatiles were captured on to Tenax TA TD tubes in one-minute (550 ml min⁻¹) increments via a Logistically Enabled Sampling System-Portable (LESS-P, SignatureScience LLC, Austin, TX, USA). Flow rate to the LESS-P was continuously monitored throughout the experiment by an inline DryCal Defender 510 l flow meter (Bios International Corp, Butler, NJ, USA). Refer to supplemental data 5(B) and (C) for representative flow rate measurements. All sample TD tubes were capped and stored at 4 °C for less than one week until GC-MS analysis [17].

Exhaled breath samples were also collected in 1 l ALTEF polypropylene bags, prior to and following the entire hypoxic exposure, using the previously established breath exhalation protocol (supplemental data 4, Jansen Inert Products, Coral Springs, FL) [2]. Volatiles were immediately transferred from each bag on to Tenax TA TD tubes by a MultiRAE Pro pump (270 ml min⁻¹, 550 ml total volume). Maximal concentrations of carbon dioxide (CO₂, %) were recorded via the MultiRAE pump to ensure end tidal breath sample quality (supplemental data 5(D)) [18, 19]. All sample TD tubes were capped and stored, less than one week, at 4 °C until analysis [17].

All samples were analyzed on a Thermo Scientific Trace Ultra-ISQ gas chromatograph affixed with a single quadrapole mass spectrometer following TD on a Markes International TD-100 as described previously (TD, Waltham, MA, USA) [2, 20]. Please refer to supplemental data 6 for a detailed description of the TD-GC-MS methodology.

XCMS R library distribution (v 3.5) was leveraged to process raw GC-MS data files [21–23]. Exhaled breath peaks were detected within 1 m/z, registered and aligned using parameters consistent for GC-MS data. Refer to supplemental data 7 for the R script. A table of 3147 features was returned for downstream statistical analysis. Initially, the features were excluded if null values were reported in more than 10% of samples and/or retention time standard deviation post registration was greater than 1.5 s. The remaining matrix of 1742 features was quantile normalized and log₂ transformed. Redundant peaks, i.e. different m/z at a common apex, are an expected occurrence when generating peak tables from raw GC-MS samples. In general, the presence of redundant peaks leads to high measurement co-linearity, which may mislead cluster formation or statistical significance tests. To limit this effect, time binning was performed to identify the most abundant ion for each XCMS feature retention time when rounded to the nearest second. Four hundred and twenty-nine features were available for statistical analysis. Four comparisons of peak abundances were performed, three within serial sampling (hypoxia versus sea-level, recovery versus sea-level, recovery versus hypoxia) and a pre-/post-bag sampling comparison, representing a total of 1716 individual comparisons. For serial sampling, comparisons were constructed by computing differences in average abundances taken within sea-level, hypoxia or recovery time periods. The average was based on at most five samples within either sea-level or hypoxia time periods.

All statistical analyses were performed using the Matlab and Matlab Statistics Toolbox (R2013a, MathWorks, Natick, MA). Wilcoxon signed rank test (Matlab function signrank.m) was applied to every feature for each paired comparison (i.e. hypoxia/sea-level paired observations) to determine if the group had a non-zero median distribution. Storey q-values (Matlab function mafdr.m) were calculated to control for false discovery rate (FDR). An FDR of 10% was applied to identify feature comparisons that deviated from the null. All statistically significant features were manually inspected. The feature list was manually reduced by removing features corresponding to redundant ions, i.e. multiple ions of the same compound at a specific retention time, siloxanes (column bleed), and noise. The remaining features were tentatively identified by spectral comparison to the NIST 11 Mass Spectral Library (v.2.0, National Institute of Standards and Technology, Gaithersburg, MD). Tentatively identified compounds were verified by mass spectral and retention time comparison to neat standards. Please refer to supplemental data 8 for neat standard chemical information, supplemental data 9 for tentative identification verification methodology and supplemental data 10 for ID verification results.

Individual variation of minimum blood oxygen saturation in response to ROBD reduced oxygen exposures has been observed previously [2, 24]. To monitor this effect, finger blood oxygen saturation (%SpO₂) was measured and plotted (supplemental data 3(A)). The data show a high amount of variability (47%–76%) in the minimum %SpO₂ achieved across test individuals (supplemental data 3(B)). These results confirm previous observations, of variable minimum %SpO₂ by test subjects, related to ROBD hypoxic exposures [2]. The results suggest that physical aspects of an individual may be impactful to the blood's ability to store and transport O₂.

The statistical comparisons of exhaled breath VOC abundances yielded 326 significant features amongst all comparisons (serial and bag). Of those 326 features, 203 correspond to unique features, 72 to tentatively identifiable compounds and 51 of those compound identifications subsequently verified by retention time and spectral matching to neat standards. Refer to table 1 for summary of the individual comparison results and supplemental data 10 for tentative ID verification data summary. Due to the large number of features and verified compounds identified, focus will be placed on isoprene and those compounds that were previously observed to change in response to hypoxic respiratory stress [2].

The goal of this study was to determine the effects of recovery gas O₂ concentration on exhaled breath VOC composition. A comparison of the mean abundances of the recovery gas time period to the mean hypoxia time period yielded 68 significant features (p-value ≤ 0.05, FDR ≤ 0.1, q-value ≤ 0.1) with 31 verified compound identifications (table 1 and supplemental data 12). Among these features, isoprene (2-methyl-1,3-butadiene) was found to significantly decrease (log₂ FC −0.399, p = 0.005, FDR = 0.034, q = 0.033) in abundance with under recovery conditions with no observed difference between recovery gas O₂ composition (p = 0.152, figure 1(A)).

Zoom In Zoom Out Reset image size

While the goal of this study was to evaluate recovery oxygen's effects on exhaled breath from acute hypoxia exposures, additional samples were acquired across the exposure time course. Of those remaining time series results, 98 features and 30 verified compound identifications were a result of abundance comparisons between hypoxia and sea-level time periods (table 1 and supplemental data 11). A survey of the results highlighted two compounds, pentanal and 2-pentanone, that were previously found to change in abundance in response to reduced oxygen content when compared to sea-level [2]. To further explore the relationship of these compounds to hypoxia exposures, log₂ abundance ratios (mean hypoxia abundance/mean sea-level abundance) were calculated and plotted (figures 1(B) and (C)). The results show a statistically significant increase in 2-pentanone abundance (log₂ FC 0.244, p = 0.012, FDR = 0.063, q = 0.062) and a significant reduction in pentanal abundance (log₂ FC −0.283, p = 0.022, FDR = 0.09, q = 0.088) under hypoxic conditions. These data support previous observations of reduced pentanal abundance under hypoxic conditions when compared to sea-level time points [2]. Interestingly, the results show an inverse response in the 2-pentanone abundance when compared with previous results [2]. While inconsistencies in the data are present, these results support further studies of pentanal and 2-pentanone changes in response to hypoxic stress.

To identify the overall changes in exhaled breath following a hypoxic exposure with different recovery gas O₂ compositions, comparisons of VOC abundances from bags collected prior to and following the overall exposure event were performed. Collectively, the data yielded a small number (13) of statistically different features (p-value ≤ 0.05, FDR ≤ 0.1, q-value ≤ 0.1) with only 6 features being verified (table 1 and supplemental data 14). Among the six significant features, isoprene was found to have a significant increase in abundance (log₂ FC 0.393, p = 0.005, FDR = 0.094, q = 0.033), post-exposure and recovery, with no observed effect between recovery gas compositions (figure 2(A)). These results suggest increase in isoprene abundance, compared to the breath prior to exposure, may be indicative of an overall hypoxic and recovery event.

Zoom In Zoom Out Reset image size

Our previous study identified a correlation between increase in isoprene abundance under 100% O₂ recovery to the minimal SpO₂ achieved by test individuals [2]. To determine if this effect was observed, the log₂ abundance ratios (post/pre) for isoprene were plotted against the individual minimum blood oxygen saturation (figure 2(B)). Isoprene abundance significantly increases (p = 0.0419) in the exhaled breath of those individuals recovered under 100% O₂, shown by a positive slope of the fit, similar to those results previously observed (figure 2(B)) [2]. However, the effect is not observed (p = 0.7034) in those recovered on 21% O₂ gas (figure 2(B)). These results suggest exhaled isoprene is affected by the oxygen concentration or gas delivery provided to the test subject following a hypoxic exposure.