IOP PUBLISHING
JOURNAL OF BREATH RESEARCH
doi:10.1088/1752-7155/3/2/027002
J. Breath Res. 3 (2009) 027002 (15pp)
Determining concentration patterns of volatile compounds in exhaled breath by PTR-MS K Schwarz1,2,3 , W Filipiak1,2 and A Amann1,2,4 1 Department of Operative Medicine, Innsbruck Medical University, Anichstraße 35, A-6020 Innsbruck, Austria 2 Breath Research Unit of the Austrian Academy of Sciences, Dammstrasse 22, A-6850 Dornbirn, Austria
E-mail:
[email protected] and
[email protected]
Received 16 December 2008 Accepted for publication 19 February 2009 Published 15 May 2009 Online at stacks.iop.org/JBR/3/027002 Abstract Proton-transfer-reaction mass spectrometry (PTR-MS) is a convenient technique for fast analysis of exhaled breath without prior sample preparation. Since compounds are not separated prior to analysis as in gas chromatography mass spectrometry (GC-MS), and since protonated molecules may fragment, relatively complex spectra may arise, which are not easily interpreted in a quantitative way. We calibrated 21 different compounds of importance for exhaled breath analysis, based on the respective pure standards diluted with nitrogen. These calibration measurements included determination of the fragmentation pattern of each compound under dry conditions and in the absence of CO2. Even though the fragmentation pattern may be predicted in a qualitative manner, the quantitative details may depend on water and CO2 content. This is exemplarily shown for isoprene. Out of the selected 21 compounds, 11 compounds showed substantial fragmentation (fragments proportion > 10%). Fragmentation of several volatile organic compounds (VOCs) in the drift tube of PTR-MS has been previously observed (Buhr et al 2002 Int. J. Mass Spectrom. 221 1–7; Taipale et al 2008 Atmos. Chem. Phys. Discuss. 8 9435–75; Hewitt et al 2003 J. Environ. Monit. 51–7; Warneke et al 2003 Environ. Sci. Technol. 37 2494–501; de Gouw and Warneke 2007 Mass Spectrom. Rev. 26 223–57; Pozo-Bayon et al 2008 J. Agric. Food Chem. 56 5278–84) and calibration factors for several compounds at corresponding mass-to-charge ratios have been calculated. In this paper, besides the calibration factors, the proportions of substantial fragments are also taken into account for a correct quantification in the case of overlapping signals. The spectrum of a mixture of the considered 21 compounds may be simulated. Conversely, the determination of concentrations from the spectrum of such a mixture is a linear optimization problem, whose solution is determined here using the simplex algorithm. (Some figures in this article are in colour only in the electronic version)
reproducibility, making breath gas analysis a highly promising non-invasive diagnostic technique of great medical potential [7–10]. The analytical techniques used are gas chromatography mass spectrometry (GC-MS) [11–13], proton transfer reaction mass spectrometry (PTR-MS) [7, 14–18], selected ion flow tube mass spectrometry (SIFT-MS) [19], laser spectrometry [20–22], ion mobility spectrometry (IMS) [23] and photoacoustic spectrometry. Also, versatile chemical,
1. Introduction In recent years the determination of concentration patterns of volatile organic compounds (VOCs) in human exhaled breath has been considerably improved. The aim of this development is to achieve adequate accuracy and 3 4
Author to whom any correspondence should be addressed. Dr Amann is a representative of Ionimed GesmbH (Innsbruck).
1752-7155/09/027002+15$30.00
1
© 2009 IOP Publishing Ltd Printed in the UK
J. Breath Res. 3 (2009) 027002
K Schwarz et al
Figure 1. The simulated spectrum for the mixture of 21 compounds with concentrations as listed in table 1 and with fragmentation patterns as given in table 3. Count rates are calculated using derived standard values for the source by applying settings A (section 2.1). Different colors refer to different compounds. As an example, consider isoprene, which appears at the mass-to-charge ratios m/z 69, m/z 41 and m/z 39.
optical or semiconductor sensors may be used [24–26]. Here we focus on proton-transfer-reaction mass spectrometry (PTR-MS). PTR-MS enables us to measure fast with concentrations down to ppt (parts-per-trillion) levels [14, 27, 28]. The interpretation of the respective spectra of a complex gas mixture, such as exhaled breath, is not always easy: complications arise by overlap of several compounds at some single mass-to-charge ratios (m/z) and due to fragmentation of protonated compounds, which then appear at several mass-tocharge ratios m/z. Even though fragmentation is not desirable for quick visual interpretation, it can be an advantage when using more sophisticated mathematical techniques. This will be demonstrated here for a set of 21 different compounds. If the fragmentation patterns of the compounds in a sample are known, and if the spectra of the pure compounds do not completely overlap, it is possible to recalculate the true concentrations. We consider compounds playing a role in exhaled breath analysis for medical applications. Isoprene, acetone and methanol are present in everybody’s breath at different concentrations. Sulfur compounds, such as allyl methyl sulfide, are released by bacteria which may exist in the mouth cavity. Typical compounds contained in breath of smokers are acetonitrile, benzene and toluene [17, 29, 30]. Isopropanol is often observed in breath, being partly of endogenous origin. Xylene and isopropanol may also be of exogenous origin, being inhaled with subsequent later exhalation. This is particularly so in a hospital environment. Aldehydes such as formaldehyde, acetaldehyde, hexanal and heptanal are observed in exhaled breath. Aldehydes are particularly interesting as marker compounds e.g. for chronic obstructive pulmonary disease (COPD, [31]). Since aldehydes
are relatively unstable, on-line analytical techniques such as PTR-MS are particularly suited for the determination of their concentrations in exhaled breath and for medical applications. We therefore included the linear aldehydes up to n-nonanal in our set of compounds. Ketones are also of interest. Apart from acetone, we included 3-heptanone, which appears as a metabolite in exhaled breath after ingestion of valproic acid (used as medication for treatment of epileptic seizures) [6, 32, 33]. 1.1. Simulated spectrum In order to analyze the quality and to point out the potential improvement of the used model of quantification, a spectrum is simulated (see table 1). Since the same compounds and fragmentation patterns are used to simulate the spectrum as well as for the quantification model, it should be possible to correctly determine the concentrations used in the simulation. Neglecting disturbances and individual variations, by using standard values for human volunteers and the fragmentation pattern of each used compound as specified in table 3, a spectrum of a PTR-MS measurement of exhaled breath (figure 1) can be simulated. The concentrations given in the literature for different compounds depend on various factors such as smoking behavior, gender and age: e.g. for acetonitrile and benzene much lower levels are expected for non-smokers in comparison to smokers, whereas isoprene levels may depend on cardiac output, breathing depth and breathing frequency, gender [34], etc. Here, the chosen values are selected to reflect the exhaled breath concentrations of a male smoker—with chronic obstructive pulmonary disease (COPD) and hence increased values for hexanal and heptanal (possible marker compounds [31]). For propanol (1-propanol 2
J. Breath Res. 3 (2009) 027002
K Schwarz et al
Table 1. Typical concentration of the selected compounds in the breath of a male smoker with chronic obstructive pulmonary disease (COPD). Compound
Standard ppb values
Comment
Acetone Ethanol Isoprene Acetaldehyde Acetonitrile Benzene Methanol Formaldehyde Toluene Xylene 1-Propanol + 2-propanol 3-Heptanone Allyl methyl sulfide Propanal Butanal Pentanal Hexanal Heptanal Octanal Nonanal
500 100 100 25 50 3 450 5 10 3 130 (75+75) 1 3 3 1 1 10 10 1 1
Diskin et al [49] Diskin et al [49] Diskin et al [49]; Kushch et al [50] ∼137 ppb; Kushch et al [34] ∼104 ppb Turner et al [51] Abbott et al [52] ∼70 ppb; Kushch et al [17] ∼33 ppb Jordan et al [53] ∼8 ppb; Kushch et al [17] ∼2 ppb Turner et al [54]; Kushch et al [17] ∼210 ppb Kushch et al [17] Kushch et al [17] ∼6 ppb Typical value (based on experience with healthy volunteers) Hansel et al [14] ∼30 ppb; Turner et al [35] ∼130 ppb Typical value (based on experience with healthy volunteers) Chosen value Chosen value Chosen value Chosen value Increased value Corradi et al [31] (COPD) Increased value; Corradi et al [31] (COPD) Chosen value Chosen value
and 2-propanol) the values in the literature vary over a wide range 30–130 ppb [14, 35]. However, since 1-propanol and 2-propanol cannot be distinguished by PTR-MS because of too similar fragmentation patterns, their summary amount of 130 ppb has been chosen for this study. The values for the aldehydes included in the model are thought to be at a low level, with exceptions for hexanal and heptanal.
specific mass-to-charge ratio of the respective nonfragmenting compound). The length of the drift tube of the used PTR-MS was 9.3 cm, with an applied voltage of 600 V. The usual pressure in the drift tube was ∼2.3 mbar (measured at each measurement), and the temperature was ∼52 ◦ C (measured at each measurement). This resulted in a drift time of ∼101 μs and ∼5.12 × 1016 particles cm−1 in the drift chamber. The ratio E/N (E is the electric field and N is the number density of the gas in the drift tube) was at ∼126 Td (Townsend 1 Td = 10−17 cm2 V−1 s−1).
2. PTR-MS, materials and method 2.1. PTR-MS settings
2.2. Materials
A high-sensitivity proton transfer reaction mass spectrometer (hs-PTR-MS, 3 turbopumps; Ionicon Analytic GMBH, Innsbruck, Austria) with Teflon rings (instead of Viton rings) was used. The PTR-MS showed a count rate5 of the primary ions (H3O+ ions) of ∼1.5 × 107 counts s−1. Two different settings (A and B) were considered: for setting A the count rate of H2O·H3O+ was around 7.5 × 104 and the percentage of minor precursor ions O+2 < 1%, NO+ < 0.5% and NH+4 < 8%. For setting B the count rate of H2O·H3O+ was around 1.5 × 105 and the percentage of minor precursor ions O+2 < 1%, NO+ < 0.5% and NH+4 < 2% (the given values refer to dried, filtered room air, so-called zero-air respectively). Typical compounds used for determination of transmission coefficients were acetonitrile, acetaldehyde, acetone, DMS, 2-butanone, benzene, toluene, p-xylene, benzaldehyde, chlorobenzene, 1,2-dichlorobenzene and 1,2,4-trichlorobenzene. These compounds do not show fragmentation of their respective protonated form. Concentrations of these compounds were chosen in a range leading to ∼10% reduction of primary counts, with subsequent observation of recovery of primary ion counts (measuring the primary ion counts at m/z = 21 and the
The calibrations of standards were performed for quantification of compounds detected in human breath. For this purpose the following compounds were used: methanol, 1-propanol, 2-propanol, acetone, formaldehyde, acetaldehyde, butanal, pentanal, hexanal, heptanal, octanal, nonanal, benzene, toluene, p-xylene, isoprene and allyl methyl sulfide (Sigma-Aldrich, Steinheim, Germany); propanal (Acros Organics, Geel, Belgium); ethanol (Riedel de Haen, Seelze, Germany); 3-heptanone (Alfa Caesar GmbH & CoKG, Karlsruhe, Germany) and acetonitrile (Mallinckroat Baker BV, Deventer, Holland). Preparation of gaseous standards was performed by evaporation of individual liquid compounds in glass bulbs. Each bulb (Supelco, Bellefonte, PA, USA) was cleaned with methanol (Sigma-Aldrich), dried at 85 ◦ C for at least 20 h, purged with clean nitrogen for at least 20 min and subsequently evacuated using a vacuum pump (Vacuubrand, Wertheim, Germany) for 30 min. Liquid standard (1–3 μL according to desired concentration) was injected through a septum, using a GC syringe (Hamilton, Bonaduz, Switzerland). After the evaporation of standard the glass bulb was filled with nitrogen of purity 6.0 (i.e. 99.9999%, Linde, Vienna, Austria) in order to equalize the pressure (to the ambient pressure). Then the appropriate volume (μL)
5 If no further comments are given, count rates are given as count rates per second. In particular, we do not use normalized count rates.
3
J. Breath Res. 3 (2009) 027002
K Schwarz et al
of vapor mixture was transferred using a gas tight syringe R bags (SKC 232 Series, Eighty Four, (Hamilton) into Tedlar PA, USA) previously filled with 1.5 L of nitrogen (99.9999%, additionally purified on carbon molecular sieves). The transfer line to the PTR-MS was cleaned from trace contaminations by rinsing with the prepared standard gases for 5 min before analysis.
differences in fragmentation patterns depending on humidity in the drift tube are shown exemplarily using isoprene. 3.1. Comparability of measuring conditions
In previous studies the fragmentation patterns (most abundant signals) for several compounds were investigated [1, 27, 38, 39]. Typical fragments for e.g. alcohols (m/z 43, m/z 57) and aldehydes (m/z 55, m/z 69) were outlined by Buhr 2.3. Counts to concentrations et al [1]. The intensities of the aldehydes given in the Volatile organic compounds with a volume ratio up to 1:1010 present work are in good agreement with other studies [1, 6]. (0.1 mm3 of the compound to 1 m3 air; 0.1 parts-per-billion Differences may occur because of the dependence of the or 0.1 ppb) can be found. The molecules are ionized by fragmentation pattern on various influencing factors. An proton transfer from H3O+ ions produced in the ion source of important factor influencing fragmentation and clustering is the instrument with subsequent measurement by a quadrupole given by the parameter E/N [27] (E is the electric field; N mass spectrometer [14, 27]. Count rates at m/z ranging from is the number density of the gas in the drift tube) in units of −17 cm2 V−1 s−1). Our settings were 21 to 230 (and m/z 18, NH+4 ) are determined. The calculation Townsend (1 Td = 10 of the volume ratios for single molecular species M depends chosen such that E/N was at a constant level of ∼126 Td (all on the corresponding reaction constant k of the ionization measurements done in this paper were in the range of 124– 127 Td). Lindinger et al [27] describe a range of 120–140 Td reaction: as a good compromise to avoid production of water clusters on M + H3 O+ → MH+ + H2 O. (a) the one hand and breaking up of product ions due to collisions Typical values for the reaction constants are around k = 2 × with neutral species in the drift tube on the other hand. The 10−9 cm3 s−1. For several compounds these constants are different settings of the ion source (settings A and B) used in known (see, e.g., the overview in [36]). For compounds this paper result in a difference of the measured water cluster + + with unknown kinetic reaction constant, the constant k = H2O·H3O and minor precursor ion NH4 , but refer to the same −9 3 −1 2 × 10 cm s allows us to determine a first estimate of E/N level of ∼126 Td. Such differences in source behavior the respective concentration (based on the measured count influence the fragmentation pattern without changing the E/N rate). The calculation of the volume mixing ratios results in level. Besides E/N, additional parameters of the PTR-MS settings have to be considered when comparing fragmentation the standard formula for PTR-MS concentrations [27]: patterns. Here we describe the settings of the source by count ratem/z 1 · . (1) reporting the measured values of the most important minor concentrationm/z = count rate precursor k · driftime precursor ions (NH+4 , NO+ and O+2 ) as well as H2O·H3O+ ions Here k is the kinetic reaction constant of the protonation using zero-air measurements (dried, filtered room air without reaction (a). The calculation of the drift time is based on the contamination). Besides the influence of the PTR-MS settings (for the standard ion mobility (μ0 = 2.76 cm2 V−1 s−1 [37]) of H3O+ in nitrogen adjusted to the actual temperature and pressure in constant E/N level at 126 Td), the fragmentation pattern may the drift tube. The count rate of the primary ions (H3O+) was change considerably with changes in the water vapor level (see calculated taking the count rate at m/z 21 and m/z 37 into section 3.4 and [37]). The water vapor level in the drift tube is consideration (count rates are transmission corrected count influenced by the humidity of the sample itself and the water level in the ion source [28, 37, 40]. Additionally the CO2 rates per second): concentration has an influence on PTR-MS measurements. count rate precursor = (IR count ratem/z 21 ) + count ratem/z 37 . The presence of a high concentration of CO2 may affect the (2) collision processes in the reaction chamber, thus leading to an Here IR = 500 is a constant reflecting the isotope ratio of increase in water cluster formation and reduction of effective 16 + + ion mobility [37, 40]. the species H18 3 O to H3 O . The signal at m/z 37 refers + Hence, for some compounds a proper calibration of PTRto the water cluster H3O ·H2O (other water clusters can be MS (at the constant level of E/N) depends on the water content neglected). Since ions of reactions caused by the+ product + and CO 2 concentration in the sample. minor precursors NO , O2 often have negligible intensity as compared to all other product ions for certain compound, they were not considered in equation (1). The only exceptions are 3.2. Fragmentation pattern of isoprene m/z 68 for isoprene and m/z 78 for benzene, which were taken The underlying calibration series used isoprene diluted in into discussion of fragmentation patterns of these compounds. nitrogen 6.0 (99.9999% additionally purified by purging through sorption trap filled with carboxen 1000 adsorbent). 3. Results and discussion The concentrations were in the range of ∼1–250 ppb. Measurements were done for a range of the spectrum from The fragmentation patterns of the different compounds in this m/z 21 to m/z 230. In figure 2, the four different calibration work were determined by means of calibration series. The series with different amounts of water in the drift tube are 4
J. Breath Res. 3 (2009) 027002
K Schwarz et al
Figure 2. Calibration of isoprene at four different water levels in the drift tube shown for m/z 41 (left panel) and m/z 69 (right panel). The different levels of humidity were achieved by adding water to the samples (setting A results in a low water flow from the source) and alternatively by changing the water flow of the source to increase the water level in the drift tube (setting B). The slopes of the calibration are relative intensities, summarized in table 2. Table 2. The influence of the water level on the fragmentation pattern of isoprene. More detailed description of fragmentation products is given in the text. Calibration series
m/z
∼2.5%a
∼7%a
