Physiological variability in volatile organic compounds (VOCs) in exhaled breath and released from faeces due to nutrition and somatic growth in a standardized caprine animal model

Physiological effects may change volatile organic compound (VOC) concentrations and may therefore act as confounding factors in the definition of VOCs as disease biomarkers. To evaluate the extent of physiological background variability, this study assessed the effects of feed composition and somatic growth on VOC patterns in a standardized large animal model. Fifteen clinically healthy goats were followed during their first year of life. VOCs present in the headspace over faeces, exhaled breath and ambient air inside the stable were repeatedly assessed in parallel with the concentrations of glucose, protein, and albumin in venous blood. VOCs were collected and analysed using solid-phase or needle-trap microextraction and gas chromatograpy together with mass spectroscopy. The concentrations of VOCs in exhaled breath and above faeces varied significantly with increasing age of the animals. The largest variations in volatiles detected in the headspace over faeces occurred with the change from milk feeding to plant-based diet. VOCs above faeces and in exhaled breath correlated significantly with blood components. Among VOCs exhaled, the strongest correlations were found between exhaled nonanal concentrations and blood concentrations of glucose and albumin. Results stress the importance of a profound knowledge of the physiological backgrounds of VOC composition before defining reliable and accurate marker sets for diagnostic purposes.


Introduction
A considerable number of volatile organic compounds (VOCs) are generated by cells or microorganisms throughout the body during various metabolic and biochemical processes (Schulz andDickschat 2007, Thorn andGreenman 2012) and can be detected in different biological specimens such as exhaled breath or faeces (de Lacy Costello et al 2014). Therefore, the clinical application of VOC analysis may be a promising tool for the non-invasive diagnosis of metabolic, inflammatory, or infectious diseases in humans (Spanel and Smith 2011, Buszewski et al 2013, Wang and Wang 2013 and animals (Dobbelaar et al 1996, Knobloch et al 2010, Purkhart et al 2011, Peled et al 2012, Ellis et al 2014. Controlled sampling, reliable analysis, and unequivocal identification of VOC biomarkers and potential confounding substances are mandatory for the application of VOC trace analysis (typical concentrations are in the nmol L − 1 -pmol L − 1 range). Adapted microextraction methods such as solid-phase microextraction (SPME) (Arthur and Pawliszyn 1990) and needle-trap microextraction (NTME) (Mieth et al Physiological variability in volatile organic compounds (VOCs) in exhaled breath and released from faeces due to nutrition and somatic growth in a standardized caprine animal model 4 2009, Lord et al 2010) enable controlled collection and pre-concentration of VOCs from the headspace (gas phase above liquids or solid materials) above faeces as well as from the exhaled breath and from room air samples (Miekisch et al 2014). Subsequent analysis by means of gas chromatography together with mass spectrometry (GC-MS) enables unequivocal identification and quantification of volatile compounds.
From a biological perspective, significant alterations of the VOC profile in humans (Filipiak et al 2012, de Lacy Costello et al 2014 as well as in animals (Elliott-Martin et al 1997, Turner et al 2012 may occur due to environmental influences (nutritional effects, air contaminants, husbandry) or conditions in the host itself (age, sex, species). Additionally, volatile compounds originating from exogenous sources can be absorbed into the body and thus contribute to the body's pool of VOCs that have been endogenously generated (Miekisch et al 2004, Pleil et al 2013. Thus, clinical studies are often hampered by non-avoidable, host-related, confounding factors that could lead to erroneous conclusions when searching for biomarkers indicating diseases (Filipiak et al 2012).
This study takes both the methodological requirements of pre-concentration and the need for standardized biological conditions into account in order to assess the 'background' variability of VOCs in healthy conditions. Significant and well-defined changes occur with respect to digestion and metabolism in goats when they develop from milk-suckling kids to ruminating goats fed exclusively with plant products (Van Soest 1994). Thus, we considered ruminants as biologically relevant large-animal models for the assessment of the physiological effects of different nutrition regimes, and we hypothesized that changing diet, i.e. the replacement of the predominant milk intake (in goat kids) with an increasing percentage of plant products through to pure plant-based nutrition (in adults), would significantly influence the emitted VOCs.
To verify this hypothesis, this study evaluated the variability in the in vivo VOC composition due to growth and metabolic effects in a caprine model under defined conditions with respect to nutrition regime, husbandry, and environmental influences. VOC concentration patterns were examined in the headspace above faeces and in the exhaled breath of young, healthy goats during their first year of life. Environmental volatile compounds present in room air were taken into account. Results of this study stress the importance of a profound knowledge of the physiological effects and nutritional/environmental backgrounds of VOC composition before defining reliable and accurate marker sets for diagnostic purposes.

Legislation and ethical approval
This study was carried out in strict accordance with the German Animal Welfare Act, and animals were housed in conformity with the guidelines for animal welfare set forth by the European Community. The study protocol was approved by the Committee for the Ethics of Animal Experiments and the Protection of Animals of the State of Thuringia, Germany (Permit Number: 04-001/11). All experiments were conducted under supervision of the authorized institutional Agent for Animal Protection. Every effort was made to minimize discomfort and suffering throughout the duration of the study.

Animals, housing conditions, nutrition regime and health status
Fifteen goats ('Thüringer Wald Ziege', 14 male, 1 female) were involved in this longitudinal prospective study. The animals originated from one farm, were colostrum-fed after birth, and were conventionally raised for the first few days. Aged eight to 16 d and weighing 1.5 to 4.4 kg (3.0 ± 0.7; mean ± SD), they were brought to the Federal Research Institute for Animal Health (Friedrich-Loeffler-Institut, Jena, Germany), where they were kept in one group under uniform conditions (room climate: 20.4 ± 2.7 °C, 63 ± 6% relative humidity; mean ± SD). Routine microbiological screening confirmed the absence of relevant infections (Salmonella, Mycoplasma, Pasteurella, Mycobacteria). They were bedded on straw that was completely renewed every two weeks. The nutrition regime changed from milk replacer provided in the first weeks of life to a purely plant-based diet (concentrated feed and hay), as shown in figure 1. Confirmed by daily clinical examination (general behaviour, appetite, rectal temperature, etc), all animals were apparently clinically healthy throughout the entire study. To prevent the influence of sex hormones on the study, the male goats were castrated at the age of eight weeks according to good veterinary practice. The one female goat did not generate any outliers or extreme values.

Study design
The duration of the study was about one year: the goats were followed from the 3rd to the 57th week of life (wl). Breath gas and faecal samples for VOC analyses were collected in parallel at seven time points, i.e. 4, 8, 21, 32, 36, 44, and 51 wl as shown in figure 1. Room air was simultaneously collected from the animals' stable.
In addition to daily clinical examination, the respiratory rate was recorded at rest twice a week from the 7th wl onwards, and the individual body mass was assessed in 2-week intervals. The averages of the obtained parameters were calculated for 3-or 4-week intervals corresponding to the sample collections for the VOC analyses. Within each interval, jugular venous blood samples were obtained from each goat in order to assess glucose, total protein, and albumin in peripheral blood.
Because goats included here served as healthy controls for another study (Krüger et al 2015), necropsies of individual goats on a randomized basis were obligatory and resulted in a decrease in the number of animals during the study (figure 1). Results of pathological examination confirmed the physical health of all 15 goats included.

Collection and processing of faecal samples
Faecal samples were collected on an individual basis directly in a clean sampling device before morning feeding; either through rectal manipulation (measuring rectal temperature) or during spontaneous defecation. Immediately after collection, about 3 g of fresh faeces per goat and time point were filled into a 20 mL headspace vial sealed with Teflon-coated rubber septa in combination with magnetic crimp caps (all: Gerstel GmbH and Co.KG, Muehlheim, Germany). The vials were stored at 4 °C and processed within 72 h after sampling.
For pre-concentration, gaseous samples were collected from the headspace above faeces using solidphase microextraction (SPME). The sample vials were heated to 42 °C for 3 min, before the SPME fibre was pierced through the septum and exposed to the volatiles of the headspace for 7 min. The vial temperature was held at 42 °C for the duration of the exposure. SPME fibres (Supelco, Bellefonte, USA) consisting of PDMS Carboxen (75 µm) were used for all measurements. Before being used the first time, the SPME fibre was preconditioned in a GC injector at 300 °C for one hour and reconditioned at 290 °C in the GC injector for 30 min before each re-use. A blank run of the SPME fibre was performed every day before the measurements to ensure that the SPME coating was clean and that no uncontrolled bleeding took place. In addition, a gas standard of 2,3-dimethyl-1,3-butadiene in methanol was analysed at the beginning and end of every sampling queue to control performance of the extraction fibre.

Collection and processing of breath gas and room air samples
An automated sampling device (PAS Technology Deutschland GmbH, Magdala, Ger many) combining mainstream capnometry and needle-trap microextraction (NTME) was used for breath gas sampling, as described previously (Trefz et al 2013b). In brief, this device enabled control of (i) sample volume and (ii) sampling conditions with respect to flow and pressure. Alveolar sampling was realized by connecting the sampling device to a mainstream capnometer (Capnoguard, Philips Respironics) and a fast-responding, CO 2 -triggered, flow valve (figure 2), and setting a threshold value for CO 2 of 25-30 mmHg (3.3-4.0 kPa). Above the threshold value the CO 2triggered valve opened and predominantly alveolar gas was directed through the needle trap devices (NTDs) for collection of VOCs. As shown in figure 2, the system was adapted to spontaneously breathing conscious goats by a tightly fitting face mask (Henry Schein VET GmbH, Hamburg, Germany). The flow rate during sampling was 20 mL min − 1 and total sample volume per goat per time point was set at 60 mL. The composition of the ambient air inside the animals' stable was observed by means of room air samples using NTME. At least two room air samples were collected per time point (one immediately before and one immediately after the breath gas measurements; 60 mL each) using the automated NTME sampling device without CO 2 -controlled opening of the valve.
The single-bed, polymer-packed NTDs (Shinwa Ltd., Japan) were packed with 3 cm of a copolymer of methacrylic acid and ethylene glycol dimethacrylate. The NTDs were conditioned in a heating device (PAS Technology Deutschland GmbH, Magdala, Germany) at 195 °C for 20 h under permanent helium flow (1 bar) before first use, and re-conditioned at 195 °C for 30 min before they were applied for the pre-concentration of the samples. Each NTD was sealed by a Teflon cap (Shinwa LTD., Japan/PAS Technology Deutschland GmbH, Magdala, Germany) before and immediately after collecting a gaseous sample.

GC-MS analysis
VOC analyses were performed using GC-MS. An Agilent 7890 A gas chromatograph connected to an Agilent 5975 C inert XL mass selective detector (MSD) was used to separate and detect VOCs desorbed from SPME fibres, as described in Trefz et al (2013a). VOCs desorbed from the NTDs were separated by a gas chromatograph (Agilent 7890 A) and detected by a mass selective detector (Agilent 5975 C inert XL MSD), as previously described in Trefz et al (2013b).

Confirmation and quantification of potential marker substances with GC-MS
Substances were tentatively identified using a mass spectral library (NIST 2005 Gatesburg, PA, USA). In order to obtain unequivocal identification and quantification, the selected marker substances were verified by means of pure reference substances. For the calibration and determination of limit of detection (LOD, signal-to-noise ratio 3:1) and limit of quantification (LOQ, signal-to-noise ratio 10:1), different concentration levels of the reference substances were measured as previously described for SPME (Trefz et al 2013a) and NTME (Trefz et al 2012). Humidified standards were used for the quantification of NTME-GC-MS data. Substance identities as confirmed through retention time and mass spectra as well as quantitative parameters (LODs, LOQs) of reference substances are provided in supplementary table A1 (stacks.iop.org/JBR/9/027108/mmedia).

Assessment of blood parameters
The plasma concentration of glucose was analysed immediately after collecting heparinized blood (2.0 mL polypropylene syringes with lyophilized electrolytebalanced heparin; PICO 50, Radiometer Medical ApS, Brønshøj, Denmark) using a combined blood-gasand electrolyte analyser (ABL 725 Series, Radiometer Copenhagen, Copenhagen, Denmark) working with the manufacturer's standard electrodes.
The concentration of total protein in blood serum was measured spectrophotometrically (Cobas 6000, Roche/Hitachi, biuret method) and the concentration of serum albumin was derived by capillary electrophoresis (Capillarys 2, Sebia). Blood for serum production was collected with 9.0 mL syringes (S-Monovette, Sarstedt AG and Co. KG., Nuembrecht, Germany) and the serum harvested by centrifugation and stored at −20 °C until required for analysis. Normally distributed data is provided as mean ± standard deviation (SD) while data without normal or with unknown distribution is provided as median and range (difference between maximum and minimum). In box-and-whisker plots, outlier values (circles) were more than 1.5 times the length of a box away from the median. Extreme values (asterisks) are data beyond the three interquartile ranges.

Statistical analysis
The Mann-Whitney U-Test was used to identify significant differences between two groups of unpaired data. A multifactorial analysis of variance (ANOVA) was performed for the global test of whether any of the measured values differed in its expectations for some time intervals. An additional factor 'goat' was incorporated in the model to consider individual bindings of the data. The global significance level of 0.05 was adjusted according to the Bonferroni correction because of multiple testing of several features. For descriptive purposes, Tukey's HSD post hoc tests were applied for every feature to discern pairwise differences between specific time intervals. Significant differences are indicated using different letters starting with 'a' for the group with the highest responses. The strength of dependence between VOCs and blood components was assessed using regression analyses, and the best fitting model for each significant correlation was analysed. P-values < 0.05 were considered statistically significant for all statistical tests.

Growth and development of the animals
In the developing healthy goats, body mass increased continuously with age (figure 3(a)), accompanied by a significant decrease in body temperature, although a slightly elevated body temperature was observed again during the summer months, corresponding to the 18th-29th wl (figure 3(b)). The respiratory rate at rest decreased significantly from 39.0 ± 10.6 breaths min − 1 in goat kids aged 7-11 weeks to relatively constant values of 23.6 ± 2.0 breaths min − 1 from the 14th wl onwards. The concentration of glucose in peripheral blood was highest in goat kids aged 3-6 weeks and decreased significantly until the 21st wl (figure 3(c)). Subsequently, constantly low concentrations of 3.8 ± 0.3 mmol L − 1 were present until the end of the study. The concentration of total protein and albumin in peripheral blood increased significantly with increasing age until the 17th wl. While young goats aged 3-6 wl presented lowest concentrations of total protein (50.0 ± 5.6 g L − 1 ) and albumin (29.1 ± 2.2 g L − 1 ), higher and relatively constant concentrations of 67.1 ± 3.4 g L − 1 (total protein) and 39.2 ± 2.0 g L − 1 (albumin) were measured in venous blood of goats older than 18 weeks (all data: mean ± SD). Results for albumin are illustrated graphically in figure 3(d).
VOC, volatile organic compound; ppbV, parts per billion by volume; wl, week of life; n, number of animals; n.s., not significant.
For the ANOVA, the global significance level of 0.05 was adjusted according to Bonferroni. Different letters indicate significant differences between time intervals (Tukey's HSD post hoc test, P < 0.05. It was not possible to collect samples (faeces, exhaled breath) from each animal included in the study on all measurement campaigns for methodological reasons.
Detected concentrations of all observed ketones in the headspace above faecal samples decreased from the first (4th wl) to the third (21st wl) measurements. Concentrations of six (acetone, 2-butanone, 2-pentanone, 2-hexanone, methyl isobutyl ketone, and 3-methyl-2-butanone) out of 10 ketones were significantly higher in the first assessment when compared to subsequent measurements. In seven (acetone, 2-butanone, 2-pentanone, 2-hexanone, 3-hexanone, methyl isobutyl ketone, and 3-methyl-2-butanone) out of 10 ketones, there were no significant differences between the headspace concentrations from the 21st wl onwards. The headspace concentrations of the aliphatic hydrocarbons butane, pentane, and heptane were significantly higher at the first assessment when compared to subsequent measurements. In contrast, the detected concentrations of isoprene were significantly higher in the last two measurements (44th and 51st wl) when compared to the previous measurements. The headspace concentrations of methyl acetate increased sharply after the second measurements in the 8th wl. Headspace concentrations of furan and 3-methylfuran decreased significantly after the 4thwl and 8th wl respectively, while the other furans (2-methylfuran, 2-ethylfuran, and 2-pentylfuran) showed low variability between different measurements. The highest concentrations of all observed volatile sulfur compounds (VSCs) were detected in the 4th wl (2-propanethiol, dimethyldisulfide, and dimethyltrisulfide) and 8th wl (dimethylsulfide).
VOCs are plotted on the y-axis, while the different faecal headspace samples and breath samples, respectively, can be found on the x-axis. Measurements at different time points are separated by vertical lines. In order to emphasize the variation between the data and to show relative differences, data is normalized to the maximum concentrations determined in all samples.

Volatile organic compounds in the exhaled breath
Throughout the first year of the goats' lives, a set of eleven VOCs was repeatedly assessed in the 71 breath samples analysed (table 2). The substances belonged to the classes of aromatic hydrocarbons (5), aldehydes (3), ketones (2), and alcohols (1). The concentrations of all eleven volatile compounds in the exhaled breath changed significantly between the different time   points. Intensities of these VOCs in exhaled breath are illustrated in a heat map ( figure 4(b)). Acetone and 2-butanone increased significantly after the 8th wl and returned to baseline values at the end of the study. Highest concentrations of the aldehydes hexanal and nonanal were detected in the assessments in the 4th wl, whereas 2-methylbutanal concentrations were highest in the 32nd wl.

Correlations between VOCs and blood components
The relationship of exhaled VOCs and VOCs above faeces to blood components was investigated by multiple regression analysis.
In the headspaces above faeces, concentrations of 3-hexanone, 3-methylfuran, methyl isobutyl ketone, 2-butanone, 2-pentanone, and dimethylsulfide were significantly correlated to blood concentrations of glucose, albumin, and total protein (table 3). In the case of blood proteins, correlations of the given VOC were higher to albumin than to total protein. Thus, figure 5 illustrates three typical examples of the multiple interactions between VOCs and the concentrations of glucose and albumin in blood.
Concentrations of 3-methyl-2-pentanone, 3-methyl-2-butanone, methyl acetate, and 2-heptanone were significantly correlated to blood concentrations of glucose only (table 4). In contrast, concentrations of dimethyltrisulfide, heptane, hexane, and pentane were significantly associated with concentrations of proteins (both total protein and albumin) in blood, but not to glucose (table 5).
In exhaled breath, only the concentration of nonanal was significantly correlated with blood concentrations of glucose, albumin, and total protein (table 3, figure 6). Concentrations of 2,5-dimethylbenzaldehyde and benzene in exhaled breath were significantly related to blood concentrations of glucose only (table 4), while concentrations of benzaldehyde and styrene were only correlated to concentrations of proteins (i.e. total protein and albumin) in blood (table 5).

Volatile organic compounds in room air
VOC concentrations assessed in room air samples are shown in comparison to concentrations in exhaled breath in table 6. Concentrations of styrene, 2,5-dimethylbenzaldehyde, and acetone in exhaled breath exceeded those in room air. Concentrations of benzene, ethylbenzene, benzaldehyde, hexanal, nonanal, 2-methylbutanal, 2-butanone, and 1-propanol present in the room air were in the same order of magnitude or higher when compared to concentrations detected in the exhaled breath.

Validity and biological relevance of the model
Variances due to nutrition and metabolism in in vivo VOC profiles can impede the search for volatile biomarkers for diagnostic purposes but have hardly been addressed in literature before. Theoretically, nutrition may impact the excreted VOCs (i) as by-or end products of digestive processes of nutrients or food additives, or (ii) through diet-related influences of the gastrointestinal microbiota and subsequent changes in their metabolism. The large-animal model exploited for this study was considered the most useful to evaluate significant effects of different endogenous digestion profiles and associated changes in metabolism on volatile substances present in the metabolome.
The age dependence of physiological variables and metabolic development was assessed in growing goats developing from milk-suckling goat kids of about 6.6 kg to adults weighing about 45.5 kg and being fed by plant products only. Body mass, body temperature, and respiratory rate decreased physiologically with age, indicating that young goat kids have the highest metabolic rate. Despite still depending on milk, the pre-ruminant goats acquire the ability to gain nutrients from the fibrous material of plants due to microbial fermentation in the gastrointestinal tract. Becoming ruminants, cellulose and other complex carbohydrates are degraded by the gastrointestinal microbiota into volatile fatty acids (VFAs, e.g. acetate, propionate and butyrate) that replace glucose as the primary source of energy (Weimer 1992, Van Soest 1994. The decrease in plasma glucose in the first few weeks correlates with the increasing concentration of VFAs in blood (McCarthy and Kesler 1956). With maturation of the liver function, concentrations of both total protein and albumin in blood increase physiologically in goats during early life (Piccione et al 2011).
The strength of this study was that all involved subjects were kept under standardized conditions with respect to nutrition regime and ambient conditions during the entire time period of about one year. Thus, possible confounding factors present in clinical studies in humans (individual habits of diet, personal lifestyle, smoking behaviour, drugs, various environmental exposures, etc) were excluded. Due to this high level of standardization and a large homogeneity within the group of animals, we hypothesize that inter-individual variability observed in this study reflects the minimum of biological variability that has to be taken into account due to physiological inter-subject differences.
Influences of climate or season could not be completely avoided, as seen in the slightly higher body temperatures recorded during the summer period with extremely high ambient temperatures that could not be completely equilibrated by the air-conditioning system. Although this effect could have influenced the respiratory pattern (i.e. reduced tidal volume and increased respiratory rate), we have not considered it as a confounding factor because the CO 2 -controlled sampling system enabled a standardized collection of 60 mL exhaled breath (per goat per time point), derived mainly from the alveolar region.

VOCs assessed in the headspace over faeces
Faeces are a dynamic mixture of residues of food intake, digestive and excretory processes, and epithelial cells as well as fragments and metabolites of the gastrointestinal microbiota (Piatkowski et al 1990). Thus, several volatile substances emitted from faeces have to be regarded as end or by-products of the host's metabolism or bacteria present in the gastrointestinal tract. Slight changes in one of these components, i.e. host, diet, or gastrointestinal microbiota, might be reflected by changes in the VOC profile.
In our study, concentrations of 18 out of 25 selected VOCs present in the headspace above faeces varied at different time points throughout the study. During the milk-suckling phase, highest concentrations of most identified ketones, aliphatic hydrocarbons, furans, and sulfur-containing compounds were emitted and decreased to lower concentrations in the advanced ruminant, whereas emission of other volatile compounds such as isoprene and methyl acetate increased with age.
The diversity and metabolic type of species constituting the gastrointestinal microbiota of ruminants change with increasing age (Draksler et al 2002) and adapt dynamically to environmental conditions, e.g. change of nutrition (Callaway et al 2010, Metzler-Zebeli Figure 5. Relationships between concentrations of (a) 3-hexanone, (b) dimethylsulfide, and (c) 3-methylfuran in the headspace above faeces and concentrations of glucose and albumin in peripheral venous blood of healthy goats. Models of multiple regression. Results of statistical analysis are provided in table 3. (a) Dependent variable: 3-Hexanone (z-axis). Independent variables: glucose (x-axis) and albumin (y-axis) in blood. (b) Dependent variable: dimethylsulfide (z-axis). Independent variables: glucose (x-axis) and albumin (y-axis) in blood. (a) Dependent variable: 3-Methylfuran (z-axis). Independent variables: glucose (x-axis) and albumin (y-axis) in blood. Table 3. Multiple interactions between VOCs assessed in headspaces above faeces or exhaled breath, respectively, and blood components in healthy goats.

Equations of multiple regression analysis
Model P-value (ANOVA)

Gluc. P-value
Alb. P-value Dependent variable: VOC. Independent variables: blood concentrations of glucose (Gluc.) and albumin (Alb.) or total protein (Prot.), respectively. Since the P-value in the ANOVA table is less than 0.05, there is a statistically significant relationship between the three variables per equation at the 95.0% confidence level. P-values less than 0.05 for each independent variable indicate that the given variable contributes statistically significant to the regression model at the 95.0% confidence level. The R 2 -statistic indicates how strong the model as fitted explains the variability in the given VOC. The Durbin-Watson (DW) statistic tests the residuals for any significant correlation. With a P-value greater than 0.05, there is no indication of serial autocorrelation in the residuals at the 95.0% confidence level. With a P-value less than 0.05, there is an indication of possible serial correlation at the 95.0% confidence level.
et al 2013, Kim et al 2014). Ketones and aliphatic hydrocarbons can derive from fatty acids due to bacterial conversion (Miekisch et al 2004, Schulz and Dickschat 2007, Thorn and Greenman 2012, de Lacy Costello et al 2014 but are also known to be generated endogenously in the host during fat and carbohydrate metabolism (Garner et al 2007 or peroxidation of fatty acids (Miekisch et al 2004). The appearance of isoprene in faeces is likely due to bacterial generation (Kuzma et al 1995) but it is also produced during the endogenous cholesterol biosynthesis (Deneris et al 1985), mainly formed in the intestinal mucosa. Endogenous production of esters such as methyl acetate from VFAs is assumed and might reflect gut wall esterase activity as well as bacterial esterase activity (Garner et al 2007). Changes in the composition of VSCs in the headspace over faeces can be directly related to the intestinal flora and may rather be associated with changes in the overall number than with the balance of the gastrointestinal microbiota (Garner et al 2007, Hales et al 2012, Walton et al 2013.
We also identified quantitative relationships between VOCs in the headspace above faeces and blood components. Although determination of the biochemical origin of VOCs was not possible in this study, results indicate a potential endogenous contribution to the VOC profile emitted by faeces which remains to be elucidated in future studies.
In our study the major changes in the headspace VOC composition occurred during the first three measurements, with the change of nutrition. Therefore, the physical development of the host, including the digestive tract, ingredients of diet as well as composition of the gastrointestinal microbiota, are most likely major factors influencing the VOC concentration pattern emitted by faeces. Composition of the gastrointestinal microbiota varies between species due to different physiology of digestion (ruminant versus monogastric animals). Although the stomach of humans as monogastric organisms contains only very few bacteria (Guarner and Malagelada 2003), complex and dynamic microbiota inhabit the large intestine in high densities (Simon and Gorbach 1984). Thus, the composition of both nutrition and intestinal flora might significantly influence VOCs released from stool in humans, too.

VOCs assessed in exhaled breath
CO 2 -controlled breath sampling (Trefz et al 2013b) enabled collection of VOCs present in alveolar gas. It is generally accepted that this portion of exhaled breath reflects most closely the composition of VOCs originating from blood (Miekisch et al 2004, Beauchamp 2011. Similar to VOCs emitted by faeces, variances in the composition of exhaled VOCs can result from physiological factors such as growth effects and metabolic development as well as from changing Table 4. Regressions between VOCs assessed in headspaces above faeces or exhaled breath, respectively, and blood concentrations of glucose in healthy goats.

Equations of simple regression analysis
Best fitting model nutrition. Substances generated in the alimentary tract can be absorbed through the intestinal mucosa, distributed via the bloodstream, and exhaled through the lungs (Baranska et al 2013). A significant increase in acetone and 2-butanone occurred after weaning from milk replacer in the 12th wl. Both substances were detected in the breath of cattle before this and were attributed to fatty-acid degradation during digestion (Elliott-Martin et al 1997). Fatty acids are products of the increasing bacterial fermentation by the gastrointestinal microbiota. Weaning of goats was followed by a significant increase in acetone in the exhaled breath in the 21st and 32nd wl. Weaning can result in a transient period with lack of energy and can therefore cause an increase in lipolysis and subsequently ketogenesis with elevated levels of ketone bodies in blood such as acetoacetate, ß-hydroxybutyrate, and acetone (Schultz 1968). Elevated levels of aldehydes such as hexanal, nonanal, and 2-methylbutanal in exhaled breath were associated with endogenous generation by protein oxidation or lipid peroxidation of unsaturated fatty acids due to oxidative stress (Miek-isch et al 2004). Generation by bacteria of the gastrointestinal microbiota is a potential source of aldehydes, as emission is reported for Proteus (Aarthi et al 2014) and Staphylococcus (Beck et al 2002).
Concentrations of nonanal assessed in exhaled breath were associated with higher concentrations of glucose and lower concentrations of proteins in blood. Additionally, we found a quantitative relationship between concentrations of both 2,5-dimethylbenzaldehyde and benzene and glucose, and, in contrast, between concentrations of both benzaldehyde and styrene and proteins (albumin, total protein). Serum albumin enhances the aqueous solution of numerous substrates through non-specific binding and acts as the major transport protein in plasma for multiple molecules (Varshney et al 2010). Fatty acids can increase their concentrations in plasma by binding to albumin (van der Vusse 2009). Changes in the serum albumin concentration may alter the binding capacity of molecules, depending on their chemical properties, and therefore potentially have an impact on the composition and concentration of blood-borne VOCs in exhaled breath. Table 5. Regressions between VOCs assessed in headspaces above faeces or exhaled breath, respectively, and blood concentrations of proteins in healthy goats.
In general, the strength of correlations (R 2 ) between blood components and VOCs assessed in the alveolar phase of exhaled breath or VOCs in the headspace over faeces, respectively, was comparable or slightly lower for exhaled VOCs. Consequently, this study provides strong evidence that the metabolic status of the Model of simple regression between nonanal (dependent variable; Y) and glucose (independent variable; X). Best fitting model: Y = (a + b / X) ^2 (R 2 = 37.06%, P < 0.000). Symbols represent single measurements (n = 69). The plot shows the regression line (blue, solid) and two sets of limits. The inner limits (black, solid) provide 95% confidence intervals for the mean value of Y at any selected X. The outer lines (grey, broken) are 95% prediction limits for new observations. (c) Model of simple regression between nonanal (dependent variable; Y) and Albumin (independent variable; X). Best fitting model: Y = (a + b / X) ^2 (R 2 = 35.04%, P < 0.000). Symbols represent single measurements (n = 69). The plot shows the regression line (blue, solid) and two sets of limits. The inner limits (black, solid) provide 95% confidence intervals for the mean value of Y at any selected X. The outer lines (grey, broken) are 95% prediction limits for new observations. organism contributes significantly to all VOC profiles emitted at different sites of the body.

Room air
Environmental volatile compounds can significantly contribute to the body's pool of VOCs through inhalation, ingestion, and percutaneous or mucosal absorption (Miekisch et al 2004, Pleil et al 2013. In agreement with our study, Mochalski et al (2013) found higher concentrations of 2-butanone in room air when compared with exhaled breath, and suggested a contribution of inhaled exogenous 2-butanone to blood levels. Aromatic hydrocarbons are common environmental contaminants (Filipiak et al 2012) and exhalation was reported in several prior studies (Phillips et al 1999, Pleil 2009). In our study, concentrations of styrene in exhaled breath were significantly higher compared to concentrations measured in room air. Styrene was previously detected in the breath of healthy human volunteers (Phillips et al 1999) and healthy cattle (Spinhirne et al 2004). Possible sources of exhaled styrene are intake via contaminated food (e.g. milk replacer) or absorption from plastic materials (Varner and Breder 1981), e.g. via the teat used for suckling milk replacer since endogenous generation of styrene is so far unknown. Due to its high lipophilicity, styrene can accumulate in fat tissue and subsequently be released into breath (Withey 1978).
In our study, the concentrations of 1-propanol in room air were mostly higher than those measured in the exhaled breath. Since 1-propanol is a multipurpose solvent used in cleaner or disinfectants and use of disinfectants could not be completely avoided, intake through inhalation is possible and detection of 1-propanol in exhaled breath most likely represents exogenous contamination.
When sampling exhaled breath, the concurrent determination of the current composition of room air should be mandatory in order to gain knowledge about inhaled compounds . Nevertheless, information about the environmental constituents is just a snapshot and the previous exposure history of contaminants remains undetermined.
As room air is a mixture of ambient air with emissions originating from the animal itself (breath and sweat) and its excretions (faeces and urine), potential endogenous volatile biomarkers can accumulate in room air. Room air samples could thus potentially be used for diagnostic purposes in the future.

Conclusions
This study revealed significant alterations in in vivo VOC profiles in both breath gas and faeces, most likely induced by metabolic effects in growing subjects. Results of a standardized large-animal model proved the impact of physiological variability on released VOCs at different sites of the body. Characterization of environmental compounds should be mandatory for validation of all studies that deal with breath gas analysis. Results emphasize the importance of a profound knowledge of physiological and exogenous influences on VOC composition before defining reliable and accurate marker sets for diagnostic purposes. and Klaus Klepik (University of Rostock) for technical assistance in conducting the experiment. We extend our thanks to Maria-Christina Haase for her help in providing literature and Wolfram Maginot for his excellent photographic support (both FLI, Jena). We are also very thankful to Dietmar Hein from PAS Technology (Magdala, Germany) for his continuous technical support and to Eva Radtke and Antje Willing (Institut für Veterinärmedizinische Diagnostik GmbH, Berlin) for analysing the concentrations of total protein and albumin in blood. Finally, we wish to extend our gratitude to the anonymous referees for the careful reading of the manuscript. This study was financially supported by Deutsche Forschungsgemeinschaft (DFG) under Grant no. RE 1098/4-1. The study was partially funded by the Animal Health Funds of Thuringia, Hesse, Mecklenburg-Western Pomerania, Lower Saxony, Rhineland-Palatinate, and Baden-Wuerttemberg. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.