Pentane and other volatile organic compounds, including carboxylic acids, in the exhaled breath of patients with Crohn's disease and ulcerative colitis

A total of 203 patients with IBD (104 males and 99 females, age range 16–79 years, median 36 years) were recruited at the Clinical and Research Centre for IBD, ISCARE I.V.F. in Prague. Of these, 149 suffered from CD and 54 suffered from UC. Later, the numbers included in the final analyses were reduced to 136 (CD) and 51 (UC) because the breath samples were judged to be unreliable due to very low water vapour concentrations in the bag samples (<2.5%). The control (healthy) group of 14 volunteers were recruited from colleagues involved in the study and the clinical centre staff. The study was approved by the ISCARE I.V.F. a.s. Institutional Ethics Committee, and the patients or their legal guardians provided appropriate written consent.

At each regular visit of the IBD patients to the clinic, bag samples of exhaled breath were obtained and simple clinical activity indices (Harvey Bradshaw index (HBI) [13] for CD patients and Simplified Clinical Colitis Activity Index (SCCAI) [14] for UC patients) were determined. Blood samples were also obtained and analysed for C-reactive protein and ferritin, and faecal samples were analysed for calprotectin. Based on the assessment of the patient condition, including physical examination, activity indices determined from a questionnaire, blood test, calprotectin and endoscopy results, the cohort of patients were divided into three groups: in remission (R), mild to moderate active disease (M) and severe active disease (A), as summarised in table 1.

Breath samples were collected into monolayer 25 μm Nalophan sample bags (Kalle CZ, Žebrák, Czech Republic) of 3 l volume in the clinic and immediately placed into an incubator held at 37 °C. Off-line SIFT-MS analyses were started after typically 5–10 min of incubation time.

The most appropriate reagent ion, either H₃O⁺, NO⁺ or ${{{\rm{O}}}_{2}}^{+},$ was chosen to analyse each compound, as indicated in table 2. Clearly, it is imperative that the analyte ions must always be unambiguous and no m/z overlaps occur with the analyte ions for other VOCs, so the most suitable reagent ion for the analysis of specific VOCs must be chosen with care. Thus, the analysis of pentane using ${{{\rm{O}}}_{2}}^{+}$ is well established [12], and acetone, isoprene and acetic acid are best analysed using NO⁺ [15–18], also indicated in table 2. We chose to include ethanol and propanol in this study using H₃O⁺ as the reagent ions and the single analyte ion m/z 83 for ethanol [19] and m/z 43 for propanol [20, 21]. Hydrogen sulphide (H₂S), only analysed using H₃O⁺ reagent ions [22], was also included. Natural inclusions in this limited study would have been some low-order aldehydes, but inspection of the large SIFT-MS kinetics database shows significant overlaps of the analyte ions m/z for these aldehydes with other possible compounds, so, for now, they were excluded. However, the carboxylic acids are more readily analysed unambiguously using NO⁺ reagent ions [23] and so acetic, propanoic, butanoic and hexanoic acids were included in this study. The selected VOCs were quantified using the multiple ion monitoring (MIM) mode of SIFT-MS in which the analytical mass spectrometer was switched between the reagent ions and selected analyte ions in order to quantify the chosen trace compounds [24]. Each MIM mode run for each reagent ion lasted for 150 s. The total analysis time for each bag sample, including the incubation and data saving time, was about 20–25 min. The background concentrations of all the included compounds were obtained during the periods between patient sample analyses in order to check that their inhaled concentrations were not significant [25].

Table 2.  A list of VOCs measured in the MUI mode with reagent ions and analyte ions for each VOC chosen for analysis, and their median values and range of values (in parentheses) in ppbv for the control, CD and UC groups. These parameters for water vapour are given as percentages.

Compounds	Reagent	Product ions	Controls	CD	UC
Water	H3O+	m/z 19, 37, 55, 73	3.6% (2.5%–4.9%)	3.5% (2.5%–5.3%)	3.6% (2.7%–5.4%)
Pentane	${{{\rm{O}}}_{2}}^{+}$	m/z 42, 72	24 (18–39)	46 (18–104)	48 (27–88)
Isoprene	NO+	m/z 68	58 (28–111)	71 (13–310)	97 (12–205)
Ethanol	H3O+	m/z 83	49 (8–200)	45 (10–279)	37 (9–212)
Propanol	H3O+	m/z 43	15 (1–21)	37 (5–337)	35 (5–256)
Hydrogen sulphide	H3O+	m/z 35, 53	3.5 (0–21)	8.2 (0–45)	10.4 (0–46)
Acetone	NO+	m/z 88	131 (66–298)	209 (63–5565)	246 (77–3154)
Acetic acid	NO+	m/z 90, 108	25 (6–53)	28 (8–112)	24 (8–156)
Propanoic acid	NO+	m/z 104, 122	8 (0–32)	9 (1–161)	8 (1–169)
Butanoic acid	NO+	m/z 71, 118	4 (0–28)	9 (2–144)	7 (1–80)

A separate test of bag sampling methodology was carried out involving just three individuals. Thus, direct SIFT-MS breath analysis was performed on three consecutive exhalations [12], following which the individual inflated the sample bag with exhaled breath which was immediately and continuously analysed over a period of 45 min.

VOCs were extracted from 15 randomly selected breath bag samples obtained from CD and UC patients using solid phase microextraction (SPME) onto CAR/PDMS-coated fibres (carboxen/polydimethylsiloxane; Supelco, Bellefonte, PA, USA) for 30 min at a temperature of 37 °C. The SPME fibres were transported from the clinic to the laboratory for analysis during the afternoon collection day whence they were directly inserted into the injector of the GC-MS instrument (FOCUS GC with SSL, ITQ 700 ion trap mass spectrometer using electron ionisation) held at 210 °C. The GC conditions were as follows: splitless injection, GC oven temperature program 50 °C (hold 3 min), 4 °C min⁻¹ ramp up to 100 °C, 20 °C min⁻¹ to 210 °C, and a final hold for 3 min (total run time of 24 min). A GC-MS capillary column TG-624 (fused 100% cyanopropylphenyl polisiloxane, 30 m × 0.25 mm ID × 1.0 um film) was used. Electron ionisation at 70 eV generated ions that were analysed by the ion trap operating in the scan mode (m/z 15–400, scan rate 1 scan s⁻¹). Peak identification was based on mass spectral interpretation and comparisons with the NIST 2.0 library. The retention times were verified by analysing selected pure standards. The results of these GC/MS analyses were used to confirm the presence of all the VOCs quantified by SIFT-MS. A detailed discussion of this important aspect of the work is given in the next section.

Of note in table 2 are the low percentages of water vapour for the three cohorts, which are essentially the same at a median value of 3.6%. This is much lower than that obtained for on-line directly sampled single exhalations which is close to 6% as expected for a body core temperature of 37 °C [29]. This low value for the bag samples is due to at least two factors, viz. the collection of mixed expiratory breath and the diffusive loss of water vapour through the Nalophan bag material. A separate test of the latter phenomenon has shown that at a temperature of 37 °C the water vapour concentration falls by about 20% after 10 min and by about 40% after 30 min, as can be seen in figure 1. This loss is sufficient to explain the measured water vapour concentration observed in the bag samples.

Zoom In Zoom Out Reset image size

But now the important question is: how much lower are the concentrations of the trace VOCs in the Nalophan bags than those measured by direct single exhalations? Much work has shown that most VOCs are not lost by diffusion from the Nalophan bags during the first 30 min of storage [30, 31], which are the typical analysis times in the present measurements. However, the sampling of mixed expiratory air, especially accumulated by more than one exhalation, can result in lower concentrations than that of end-tidal breath, and this will surely be compound dependent. For example, a compound such as acetone, which is very soluble in body fluids and is released at the lungs and by the airway surfaces, will be least variable along the respiratory tract, whereas less soluble, more volatile compounds, such as pentane and isoprene, which are released mostly at the alveolar interface, might be more variable in the exhaled air. Additionally, the concentrations of these very volatile compounds in the exhaled breath will be dependent on the exhalation speed and the rate of the bag inflation. This has been well demonstrated, for example, for nitric oxide for which controlled exhalation is necessary for meaningful measurements in breath [32]. The separate tests of the present bag sampling method showed that the acetone concentrations were only lower by about 20% than the end-tidal values directly measured in single breath exhalations, whereas the pentane concentration was lower by as much as 50%. This explains the low values of pentane in the bag samples for the healthy volunteers and for the CD and UC patients as compared to the previous detailed studies of breath pentane measured in single exhalations, as referred to in the introduction.

The results obtained in the present study using bag breath samples from CD and UC patients (including all three R, M and A groups) are shown in figure 2 as box-and-whisker plots together with the previously published results of direct exhalation analyses [12]. The boxes range from 10th to 90th percentiles, with the narrower part indicating 25th and 75th percentiles and the near-centre horizontal lines indicating medians. The whiskers indicate the concentration ranges (min to max). What is revealed by these collected data is that the breath pentane median concentrations for the control groups and the IBD groups are about a factor of two lower in the bag samples collected in the present study compared to the concentrations obtained in the previous study by direct breath sampling. This is consistent with the reduction in pentane concentration in samples collected by inflating the Nalophan bags, as discussed above. The bag sampling procedure often required more than one inhalation/exhalation and thus mixed expiratory air/breath inevitably diluted the captured samples below the concentration of end-tidal breath that is obtained from single exhalations obtained on-line [33]. The relative pentane concentrations between the groups are similar to those observed by Dweik and his co-workers [34], even though their absolute concentrations (reported as mean breath pentane 20 ppbv for CD, 19 ppbv for UC and 14 ppbv for controls) are yet another factor of two lower than in the present bag samples, presumably due to different bag material and longer storage times. Most importantly, these collected results reveal that breath pentane is significantly elevated in the exhaled breath of patients with both CD and UC relative to that for healthy controls, but there is no statistically significant difference between the median concentration for the CD and UC patients. These collected results demonstrate the value of the direct on-line measurements of this volatile compound and other volatile compounds such as isoprene (see below) and other trace hydrocarbons.

Zoom In Zoom Out Reset image size

To complete this discussion of breath pentane, its relation to the clinical activity indices has been explored. A plot of the simple activity index (HBI for CD and SCCAI for UC) is shown in figure 3(a). As can be seen, there is only a weak correlation between these parameters. More complex indices have been constructed, as described in a recent report [13, 14], where it is also shown that these simple indices have good correlation with more complex indices. However, the construction of the latter, which are time-consuming and complex and include 7-day diary data, are clearly beyond the scope of this paper.

Zoom In Zoom Out Reset image size

A more selective approach to the pentane analysis is to present the exhaled breath pentane concentrations for the R, M and A groups. These data are given in figure 3(b) as box-and-whisker plots of pentane concentrations in ppbv for the healthy control group labelled as C, and for the R, M and A groups of CD patients and separately for the UC group (circle) according to the groups listed in table 1. These data again clearly show the greater median breath pentane for these IBD patient cohorts above that for the healthy controls. What is also apparent is the increasing median pentane concentration as A > M > R, the somewhat subjective clinical activity indices, which shows the skill by the clinicians in assessing the severity of these conditions. The wide spread in the pentane concentrations for the patient groups compared to the control group is also apparent. This is a feature of the concentrations of all the breath VOCs we have investigated in the exhaled breath of the IBD patients, which illustrates the varying complexity and severity of the CD and UC conditions between patients. It is also important to report that the correlation between calprotectin faecal concentration and breath pentane is only very weak at R² = 0.06 for the UC patients and R² = 0.03 for the CD patients. This can be explained by the high variability of calprotectin within the activity groups.

The median concentration of isoprene in the bag samples from the healthy controls is 58 ppbv. This is significantly lower than that expected for healthy controls, which previous studies of directly sampled breath have shown to be closer to 100 ppbv and ranging from 20–200 ppbv [35]—as before, most probably due to the dilution of the breath samples collected in the bags. The median isoprene levels for the CD and UC patients are somewhat greater than those for the healthy controls, as can be seen in table 2 and the box-and whisker plots in figure 4(a). There is no obvious trend with respect to the R, M and A groups, presumably because systemic/breath isoprene is not related to CD and UC diseases and their severity. What is obvious is the very wide range of isoprene concentrations in the IBD patients that is not revealed by the median concentrations. When the dilution in the bag samples is considered, the end-tidal isoprene level for some IBD patients can be higher than 400 ppbv. It has been clearly established that breath isoprene is lower for children and young adults (<20 years old) compared to older persons [35], and this can partly explain this wide spread in concentration in the breath of the IBD patients (age range 16–79 years). Isoprene originates from cholesterol biosynthesis and is thus loosely related to the cardiovascular condition. It has also been associated with psychological stress [36] and it is possible that the somewhat larger isoprene levels in the IBD patients are related to stress or release by body movements [37].

Zoom In Zoom Out Reset image size

The results of the H₂S measurements are very interesting. It has been shown by a previous SIFT-MS study that this odorous compound is largely produced in the oral cavity of healthy persons by bacterial action, being at much lower concentrations in nose-exhaled breath [38]. Thus, it is widely variable between individuals and presumably related to oral hygiene. As can be seen by the median concentrations given in table 2 and the box-and whisker plots in figure 4(b), H₂S is at a greater median concentration and more widely spread in the breath of both the CD and UC patient cohort compared to the control cohort. Again, there is no obvious correlation with the activity of the diseases as indicated by the values for the R, M and A groups. It has been reported that H₂S in exhaled breath is a potential biomarker for small intestinal overgrowth in irritable bowel syndrome (IBS) [39], the current detection method for this being the hydrogen breath test [40]. The present results add weight to the suggestion that H₂S may be an alternative indicator of bacterial overgrowth in IBD patients, especially if the concentration of H₂S in nose-exhaled breath can be measured. In the previously mentioned study by Dweik and colleagues [34], most breath VOCs, including H₂S, CS₂ and (CH₃)₂S, were found to be elevated in ileal pouch–anal anastomosis subjects, indicating bacterial overgrowth/dysequilibrium as a potential cause.

The observation that acetic acid is present in the exhaled breath of the healthy controls is no surprise, since this has been observed in previous SIFT-MS studies of healthy volunteers, and the median concentration conforms to these previous measurements [41–43]. Whilst the median concentrations in the breath of both the CD and UC patients are similar to those in the healthy controls (see table 2), the spread in values is much wider, as can be seen in the box-and-whisker plots in figure 5. Why should this be so? Recently, it was discovered that breath acetic acid is elevated in patients suffering from gastroesophageal reflux disease (GERD) [41] and also in the exhaled breath of cystic fibrosis (CF) sufferers [42, 43]. It has been suggested that in GERD, the pH of the airways mucosa is lowered due to the refluxing of gastric hydrochloric acid. Similarly, the pH of the airway surface in CF patients is abnormally low and the clinical implications of this are discussed in a recent paper [43]. Thus, the lower pH in both conditions results in the release of volatile acetic acid molecules into the exhaled breath that originates from systemic non-volatile acetate ions. There is no obvious reason to suggest that this pH effect is operative in the IBD patient airways. It is much more likely that the elevated acetic acid in the exhaled breath of IBD patients is due to enhanced production in the gut because the microbiota is significantly changed in terms of diversity and abundance due to active inflammation. It can be seen in figure 5 that this enhancement is apparently greater in UC patients than in CD patients, but there is no obvious correlation of breath acetic acid with the activity index.

Zoom In Zoom Out Reset image size

Very interesting is the presence of both propanoic and butanoic acid in the exhaled breath of the control group and the CD and UC groups, albeit at lower concentrations than acetic acid, as can be seen in table 2 and figure 5. Because these two carboxylic acids have not previously been measured in exhaled breath by SIFT-MS, we considered it important to obtain corroborative support using GC-MS analyses of some breath samples (see the material and methods section). Clear peaks of acetic, propanoic and butanoic acids were present on the chromatogram, as seen in figure 6.

Zoom In Zoom Out Reset image size

The median concentrations of these acids are somewhat greater in the breath of the CD and UC patients, but, again, the individual concentrations are much more widely spread than in the control group, as can be seen in figures 5(b) and (c). Again, there is no obvious correlation of the acid concentrations with the activity indices. However, the concentrations of these three carboxylic acids are seen to be reasonably well correlated, as exemplified by the correlation plots in figure 5(d), from which we deduced that they have a common origin—the gut. Hexanoic acid was also included as a target compound in these studies, but it was at concentrations close to the limit-of-detection in the SIFT-MS analysis and could not be seen clearly in the GC-MS analysis.