| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* Clinical Science at South Bristol, Bristol Royal Infirmary, Bristol, UK;
Faculty of Applied Sciences, Centre for Research in Analytical, Materials, and Sensor Sciences, University of the West of England, Bristol, UK; and
Health Protection Agency, Bristol Royal Infirmary, Bristol, UK
1Correspondence: Clinical Science at South Bristol, Bristol Royal Infirmary, Marlborough St., Bristol BS2 8HW, UK. E-mail: c.s.j.probert{at}bristol.ac.uk
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: campylobacter • clostridium difficile • ulcerative colitis
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
(SCFAs), branched-chain fatty acids (BCFAs), indoles (2)
, and phenols (3)
have been observed. Fermentation of carbohydrates in the gut produces ethanoic, propionic, butanoic, pentanoic, and hexanoic acid acids, particularly by Bacteroides (4)
. In vitro studies (5)
have provided evidence that proteinacious foods also produce SCFAs via the action of bacteria such as Clostridia spp.; BCFAs, such as 2-methylbutanoic acid and methylpropionic acids, are principally produced by gut microbial action on proteins via the respective branched amino acid. Volatiles such as methanethiol and ammonia are considered (6)
to be derivable from methionine by the action of Clostridium sporogenes. Hydrogen sulfide and methanethiol can be damaging to the large intestinal epithelium and are also generated from sulfur-containing substances in the diet (7)
. Similarly, fermentation of tyrosine and tryptophan in stool has been shown (6)
to produce the VOCs phenol and indole, respectively.
Humans retain a resident colonic flora for most of their lives (8)
and bacteria identified within fecal flora are shared by most adults. Consequently, the interaction between food and these shared bacteria might lead to fecal compounds that are shared by different individuals. In contrast, gastrointestinal infection might be expected to lead to a different pattern of VOCs. This led to the following hypotheses: i) VOCs would be shared in health: ii) VOCs would be constant in individuals: and iii) specific changes in VOCs would occur in disease. Volatile emissions in health were defined in a cohort and a longitudinal study. These hypotheses are supported by the fact that the intestinal microflora changes radically in acute gastroenteritis, with the pathogen often eclipsing the normal flora (9)
and by a pilot study that showed change in VOC in specific infections (10)
. Recent advances in gas sampling, chromatography, spectrometry, and analytic software make possible the comprehensive chemical characterization of fecal and bacterial volatiles (11)
.
This report describes the pattern of VOC emissions from human feces in health (cohort and longitudinal studies) and gastrointestinal diseases, specifically from patients with ulcerative colitis, C. jejuni, and C. difficile (disease study). Our data indicate specific changes in the pattern of volatiles in such diseases and that these compounds could be used to make a rapid diagnosis.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
.
|
For the disease study, samples from anonymous patients with C. jejuni (n=31) and C. difficile (n=22) were obtained from the Health Protection Agency, Bristol, UK. Samples from anonymous patients with ulcerative colitis (n=18) were obtained from the Bristol Royal Infirmary, Bristol, UK.
The study of asymptomatic donors was approved by the research ethics committee of the University of the West of England, Bristol. All participants gave written informed consent. Studies of patients with gastrointestinal diseases were approved by the Research Ethics Committee for United Bristol Healthcare Trust.
Samples
Participants collected 2–4 g of feces in a 10 ml Supelco vial, sealed with a phenolic screw-top and PTFE/silicone septum (Supelco, Poole, UK). The samples were frozen and stored at –20°C, within 1 h of passage. Analyses were typically undertaken within 7 d of freezing. One fresh sample was homogenized within 30 min of passage by a healthy donor, and 2 g was decanted into 2 vials. One sample was analyzed immediately, the other was frozen and analyzed 7 d later. The VOCs from the fresh and frozen homogenized samples were separated and analyzed by the same method. The chromatographs were similar (data not shown).
Experimental design
To investigate the hypothesis that specific changes in the volatiles emitted from feces occur in disease, we first had to characterize the VOCs emitted from feces of healthy donors. This was undertaken in two stages: i) a cohort study and ii) a longitudinal study. The cohort study was designed to give an overview of the VOC pattern in health, while the longitudinal attempted to assess the day-to-day variation in the VOC pattern. These two sets of data were used to create a library of VOCs found in healthy donors. Subsequently, the VOCs emitted from feces of patients with gastrointestinal diseases that cause diarrhea were studied and compared to the library of VOCs found in healthy donors.
Chemical analysis
Two gas chromatograph/mass spectrometer (GC/MS) systems were used: a Hewlett Packard GC (model 6890, Bracknell, UK), fitted with a 0.75 mm quartz liner, linked to a benchtop quadrupole MS (model 5973) and a Perkin Elmer Clarus 500 GC/MS quadrupole benchtop system (Beaconsfield, UK) fitted with a 1 mm quartz liner. The GC column was a conjoined SPB-1 sulfur fused-silica capillary column, 30 m long, 0.32 mm (internal diameter), and 4.0 µm film thickness (Supelco), and a ZB-FFAP fused-silica capillary column, 30 m long, 0.32 mm internal diameter and 0.50 µm film thickness, (Phenomenex, Macclesfield, UK); this conjoined column was used to improve separation of polar and nonpolar species. The carrier gas was 99.9995% pure helium (BOC, Guildford, UK), run through an excelasorbTM (Supelco) helium purification system, at 1.16 ml min–1. Before extraction, all sample vials were placed in a water bath at 60°C for 1 h. Carboxen/polydimethylsiloxane solid-phase microextraction (SPME) fibers were used to extract the VOCs. The fiber needle was injected into the vial headspace above the feces, and then the fiber was pushed out of the needle and exposed for 20 min, under static conditions, to absorb the volatiles. The fiber was then immediately transferred to the GC port for thermal desorption at 280°C with the split valve closed throughout. The GC was operated as follows: solvent delay, 4 min; temperature program (35°C), 5 min; ramp of 7°C min–1 to 250°C; finally held at 250°C for 12 min (total run time 47.7 min). The MS was operated in EI mode scanning from mass ions 17–350 (4–47.71 min). Room and laboratory air were used as controls. For a control an empty glass vial was stored under similar conditions to the sample vials and analyzed for VOCs using the standard method.
Compounds were identified by comparison with the mass spectral NIST 05 library, followed by manual visual inspection and retention time matching of selected standards (Fisher Scientific, Loughborough, UK; Acros Organics, London, UK; Sigma-Aldrich, Poole, UK). In interpreting the data, only compounds with a >90% probability of a match to NIST 05 library standards were named. Further authentication was achieved by subsequent manual inspection and retention time matching of selected compounds.
Investigation of the origin of alcohol, aldehydes, and esters
Labeled [1-13C]butanoic acid was incubated with stool as follows: 2.0 g of fresh homogenized stool was placed in a sample vial with 2 mg [1-13C]butanoic acid, dissolved in 0.25 ml deionized freshly sterilized water. The vial was flushed with filtered nitrogen, sealed with a PTFE/silicone septum, and placed in a water bath maintained at 37°C. Positive pressure was maintained with nitrogen throughout the incubation. The experiment was performed in duplicate. Two control vials were prepared: a vial, containing unlabeled butanoic acid with the same homogenized stool sample; and another vial containing 2 ml deionized, freshly sterilized water and 2 mg [1-13C]butanoic acid. SPME extraction was undertaken at 3 and 24 h, using standard extraction protocol. Mass spectra were analyzed for conversion products of [1-13C]butanoic acid.
Statistics
A subset of compounds was selected for detailed statistical analyses to attempt to classify diseases (Table 2
, asterisked compounds). The selection criteria were based on their wide variance by manual inspection and for biochemical reasons. Many compounds of small frequency occur in stool, which are likely to reflect individual characteristics and, therefore, could not be readily selected for a general system for disease diagnoses. A range of acids and sulfides was selected, as these are mainstream metabolites of bacteria and likely to change if the gut flora changes. A group of heterocycles was selected, furans and pyrrole, as their presence is likely to reflect differences in the secondary metabolites of gut microorganisms. A representative number of esters were selected. Their presence would be expected to reflect gut wall esterase activity/microbial activity and may change if the gut wall is affected by disease. Three straight chain and seven benzenoid/indole compounds were selected for their significant differences in abundance between the four groups of subjects.
|
The presence or absence of each of the chosen compounds in each stool sample was coded "1" to denote the presence of the compound and "0" to denote its absence in that sample. The presence/absence data were used as the independent variables in a multivariate discriminant analysis with group membership (normal n=30; ulcerative colitis n=18; C. jejuni n=31; C. difficile n=22) as the dependent variable. In general, multivariate discriminant analysis is typically considered to be robust to violations, and the use of dichotomous variables in discriminant analysis does not greatly affect conclusions (12)
.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
describes their frequency, in descending order of occurrence for asymptomatic donors. Many very minor compounds were also detected; these remain to be identified. Compounds were listed by the frequency of their chemical class (Table 3
). For each donor the number of VOCs ranged from 78 to 125 (median=101). Interestingly, 44 compounds were common to 80% of the cohort samples. Thus, many compounds were shared by asymptomatic donors taking an ad lib. diet.
|
Acids, alcohols, and esters
Short-chain fatty acids (SCFAs) were very common. SCFAs arise from metabolism of undigested carbohydrate, such as dietary fiber, by colonic bacteria (4)
. Branched-chain fatty acids were abundant, and their presence can be explained from the dissimilation of amino acids in some cases (6)
. Methanoic acid occurred in just 3 samples; the same samples were the only ones to contain pentyl methanoate.
Alcohols are thought uncommon in adult feces (13)
. However, in the cohort study 45 different alcohols were observed, of which 8 occur in >80% of the samples. Ethanol was found in all but 2 samples. The presence of ethanol has been linked to yeast overgrowth (14)
as well as bacteria synthesis by the Embden-Meyerhof pathway. Interestingly, specific long-chain aliphatic alcohols (C6–C20), which can inhibit the growth of various bacteria and fungi (15)
, were also found. It is likely that these compounds play a role in providing an ecological niche for some bacteria by inhibiting others. Esters were the largest group of compounds with up to 29 (mean=12) different esters per sample.
Benzenoid and heterocyclic compounds
A diverse range of 56 aromatic compounds was found, which included mono-di-, triand tetra- substituted benzenoids, mono- and disubstituted furans, and nitrogen containing derivatives of pyridine, pyrrole, and indole. Most of them have not been previously reported in feces. The few compounds that have been reported include phenolic and indole compounds arising from the metabolism of aromatic amino acids (5)
. Phenols, indole, and 3-methylindole were very common in fecal gas in the cohort study. Pyrrole has not been previously reported in stool, however, it was common in this study. A wide range of furan derivatives, mainly aromatic, were found and may be of plant origin (16)
. However, it is doubtful, as furans are also acid labile. 3,4-Dimethylstyrene was found in 22% of the fecal samples. 2,5-Dimethylstyrene isomer is found in carrots (17)
. Toluene and xylene may have a dietary origin. They were also observed in potato tubers (18)
, although biochemical syntheses in the gut cannot be excluded. Some benzenoid compounds such as dimethylbenzenes, ethylbenzene, and toluene (constituents of petrol) may arise from air pollution. 2-Pentylfuran was emitted from 22% of the fecal samples, it has also reported in four different analyses of processed potatoes (18)
.
Aldehydes and ketones
The aldehydes, ethanal, propanal, butanal, and hexanal were predominantly observed. Ethanal, present in all samples, is considered (19
, 20)
to promote mutagenesis and is associated with bowel cancer; the toxic effects of higher aldehydes have received much less attention. The origins of some aldehydes may be dietary; 2-methylpropanal, 3-methylpropanal, hexanal, nonanal, decanal, and benzaldehyde are found in potato tubers and hexanal in carrots, for instance. Ketones were the second largest class of VOCs found and include four of the most common compounds. Acetone and 2-butanone probably arise from fatty acid and carbohydrate metabolism (21)
. Methylketones can be produced by many species of bacteria and can also be produced by fungi from the respective alkanoic acid. Some compounds of this group may have dietary origins, for example cheddar cheese contains 2-propanone, 2-butanone, and 2-heptanone; potatoes contain acetone and 2-heptanone; and 6-methyl-5-hepten-2-one is an abundant volatile in carrots (22)
. The universal presence of 2,3-butanedione is interesting since it may have health implications by impacting on the growth of some bacteria and yeasts (23)
.
Alkanes, alkenes, alicyclic compounds
Methane is product from bacterial fermentation of monosaccharides and is not considered in this study. Pentane was the shortest hydrocarbon found and was probably due to the limitations in our analytical method. Shorter chain species could well be present, but as they are gases could not be sufficiently absorbed by the SPME fiber. Longer chain species were found in small numbers in stool: 11 alkenes were found in stool headspace. 3-Octene (11%) may be found in meat products and 3,7-dimethyl-1,6-octadiene (26%) is present in milk (24)
. Isoprene (33%) was extracted from feces and has been investigated as a breath marker for disease (25)
. Isoprene in the gut appears to be the result of cholesterol biosynthesis (26)
and is considered to be the most common hydrocarbon in the human body. Many alkenes are well documented as naturally occurring plant products such as myrcene (27)
. Limonene was the most abundant of the terpenoid compounds in the cohort study.
A total of 15 compounds was found with the formula C10H16. A combination of fragment patterns and retention times enabled analysis of the compounds to be correctly assigned for example; alpha and beta phellandrene. There were also 5 compounds, and the general formula C15H24 with caryophyllene (31%) the most common. Most of the terpenes we identified are found in vegetable food stuffs and do not originate from animal products. For instance the following volatiles are present in carrots: pinene, limonene, terpinene (1-methyl-4-(1-methylethyl)-1,4-cyclohexadiene), p-cymene, terpinolene (1-methyl-4-(1-methylethylidene)-cyclohexene), 
-caryophyllene, and humulene (39)
. Copaene is found in potato extracts (18)
.
Ether compounds and chloro compounds
Seven ether compounds were isolated. 2-Ethoxyethanol commonly occurs in manufactured products like soaps and cosmetics (28)
, and 1,3-dimethoxybenzene is a registered food additive in Europe (29)
. Similarly, it is very unlikely that the chlorinated compounds found were of biological origin. Consumption of contaminated food or water is the likely source of these compounds. Chloroform may arise as a stool VOC component from several sources: it is an air contaminant and has been detected in foodstuffs (30)
. Chlorination for disinfection of drinking water is another source resulting in the production of chloroform and halogenated methanes (31)
. Methylene chloride is not known to occur naturally (32)
, although it has been detected in the ambient air in some areas and can enter the aquatic environment from industrial waste water.
Nitrogen and sulfur compounds
Some of the nitrogen compounds (Table 2)
likely arise from diet; for instance, methylpyrazine, pyridine, and pyrrole are constituents of coffee. However, pyrrole readily polymerizes with acid and, therefore, its presence is unlikely to be dietary. A diverse range of sulfur compounds was identified. Methanethiol and dimethyl sulfide were found in most samples; the former is, at least in part, considered to be produced from methionine by Clostridia in the gut (6)
. Dimethyl disulfide and dimethyl trisulfide have been previously reported in stool (33
, 34)
and were found in all samples. Methanethiol has a toxicity approaching cyanide. Methanethiol and dimethylsulfide may be produced by methylation of hydrogen sulfide as a detoxification mechanism by mucosal S-thiolmethyltransferase (35)
.
Consistency of VOCs in asymptomatic donors
A total of 292 VOCs was identified in the 50 samples in the longitudinal study (Table 2)
. The mean number of VOCs identified, per person, was 154 (SD 26, range 117 to 180). For each person, an average of 35% (SD 4) of VOCs was found in every sample; 46% (SD 5) was present in 
4 samples; and 58% (SD 5) was found in 3 samples. A total of 22 VOCs were ubiquitous in a longitudinal study (Table 4
). The total number of VOCs, per individual, is described for all 5 of their stool samples (Table 5
).
|
|
Figure 1
shows the relative changes in abundance of 6 common VOCs present in 5 stool samples, per donor. For many donors, the abundance of most compounds is represented by a reasonably flat line indicating little day-to-day variation for compounds such as carbon disulfide, pentanoic acid, and 2,3-butanedione. However, in some donors a marked variation in abundance of some VOCs, such as indole, was found.
|
The acid VOC concentrations in the longitudinal study tracked each other to a high degree, unlike the other compounds, which implies that the concentrations of the acids increase in unison. This requires quantitative analyses of stool to confirm this finding. However, the results complement in vitro studies where SCFAs have been shown to increase considerably by the addition of carbohydrate to fecal homogenates independent of individual flora.
Finally, similar numbers of VOCs were found in the cohort and longitudinal studies, 297 and 292, respectively. The combination of both studies resulted in a total of 311 different compounds. A total of 16 compounds was found ubiquitously in the cohort study, and 22 were found in the longitudinal study with 13 common to both (Table 4)
.
Metabolism of butanoic acid
Long-chain alcohols may be made by reduction of the corresponding acid, for which an extensive homologous series was found. This hypothesis was investigated by incubating feces with labeled butanoic acid and finding that butanol was synthesized, (Table 6
). Examination of the butanoic acid fragment pattern in the mass spectrum of the fecal sample with [1-13C]butanoic acid added showed an increase in fragment ions with masses 61 and 74 due to the 13C acid. Inspection of the fragment patterns associated with 1-butanol and butanal confirmed conversion to [1-13C]-1-butanol, represented by an increase in mass 57 in the fragment pattern, which was seen to increase with incubation time. No conversion to [1-13C]butanal was observed. The presence of [1-13C]butanoic acid ethyl ester was confirmed by a one mass unit shift in fragment ions of mass/charge ratio 71 (the base peak), 88, 101, and 117 (the molecular ion), in their respective fragment patterns, at a retention time of 16.53 min for butanoic acid ethyl ester. Conversion of [1-13C]butanoic acid to [1-13C]butanoic acid propyl ester was confirmed by the same method, utilizing fragment masses 71 (the base peak), 89, and 101. Conversion to [1-13C]butanoic acid ethyl and propyl esters was observed after 3 h incubation.
|
Butanoic acid butyl ester, retention time 21.88 min, showed a minimal peak after 3 h incubation in both samples. The presence of the 13C form could not be confirmed. After 24 h incubation, the butanoic acid butyl ester peak for both samples, with and without labeled butanoic acid, showed a 15- and 7-fold increase, respectively. The occurrence of fragment ions of mass/charge ratio 90 and 102 confirmed the presence of the labeled ester. The formation of labeled ethanoic acid butyl ester was also confirmed by appropriate mass spectral fragmentations, after 24 h incubation. It is likely that microbially mediated reduction of other fatty acids also occurs.
Specific changes in VOCs in gastrointestinal disease
The analyses of the VOCs of the stools from patients with ulcerative colitis, C. jejuni and C. difficile resulted in the characterization of 149, 183, and 145 compounds, respectively. The percentage occurrences are shown in Table 2
. All the compounds fall into 13 classes of compounds, the total number of compounds for each class were collated (Table 3)
. A list of compounds used for the discriminant analysis is asterisked in Table 2
. Figure 2
shows a two-dimensional plot of the discriminant scores for the two principal dimensions of the data from Table 2
. A clustering of cases drawn from the same population is clearly evident, and group-to-group differences are also shown. Application of discriminant functions on the sample data gives 100% predictive accuracy and 96% predictive accuracy when cross-classifying, using the leave-one-out procedure. In the leave-one-out procedure, each observation is temporarily omitted and the resulting discriminant function is used to predict the classification of the temporarily omitted observation (36)
. Repeating the analysis but omitting each compound in turn gives an average within sample predictive accuracy of 99.9% (with a minimum of 99.0%) and an average of 94.9% predictive accuracy (with a minimum of 90.1%) using the leave-one-out procedure.
|
Some interesting differences were found among samples from various groups. For examples, butanoic acid was common in all groups except C. difficile samples, in which just 41% contained butanoic acid. Butanol, however, was ubiquitous in C. difficile samples. Clostridia are known to produce neutral compounds as well, such as ethanol, isopropanol, and butanol, at the expense of producing less butanoate (37)
. Such observations make plausible our findings.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
. Interestingly, Tomlin’s study (39)
130 yr later drew similar conclusions. In recent years, Levitt has led the field in the study of flatus and, in 1997, published a seminal paper on its composition (40)
. He confirmed Ruge’s observation and went on to quantify the most abundant gases in the flatus of 16 volunteers: 
74% of gas in flatus was produced intraluminally and was a mixture of hydrogen, carbon dioxide, and methane. The other principal component of flatus was nitrogen (5–48%). Methane production was bimodal; 5/16 subjects producing significantly higher level than the others; methane production is correlated with methanogenic bacteria. Similarly, sulfate-reducing bacteria are responsible for the generation of pungent sulfides (41)
. Thus, evidence shows that methane (42)
and sulfide production (43)
can be directly related to intestinal flora. These abundant molecules, however, were not the subject of the present work, as the study of higher MW compounds was considered to more likely result in unique differences for disease classification.
Sharing of VOCs by asymptomatic donors
On average, the gas emitted from the feces of healthy donors contained a median of 101 VOCs. Remarkably, 80% of donors shared 44 VOCs in the cohort study. This finding confirms our hypothesis that many of the identified VOCs are shared. Scrutiny of VOCs tells us something about their derivation. Many VOCs can be generated by intestinal microbiological metabolism of foodstuffs. Other VOCs are probably residual compounds present in plants. A small number of VOCs are nonbiological "pollutants" and reflect contamination of the environment (air, water, and food) by chemical waste. Intersubject differences, then, may reflect differences in flora, diet, or exposure to chemical pollutants.
Esters were, unexpectedly, the most numerous class of compounds found in the cohort and longitudinal studies. There have been few reports of esters associated with stool, which of course, does not possess the typical pleasant smell of esters. In the cohort study, 46 different esters were identified. A total of 26 straight chain esters was found, derived from methyl to heptyl alcohol and from methanoic to heptanoic acid. Methanol was rarely found as free alcohol, although it was extensively found as the ester, methyl, and ethyl esters were the most abundant. Given the toxic effects of methanol, the ease with which individuals’ gut flora can trap methanol as an ester may have important health considerations.
All the free straight-chain acids and alcohols, corresponding to the straight-chain esters, were also found in the feces. We have demonstrated in our study of the incubation of [1-13C]butanoic acid in stool that esters are generated from acids (Table 6)
. It is interesting that labeled ethanoic acid butyl ester was also observed, which must be derived from the reduction of the added labeled acid followed by esterification with "ethanoate" in the stool. The ester syntheses may be mediated by a bacterial esterase such as is present in E. coli (15)
or by enzyme(s) derived from the intestinal mucosa (44)
, but there is a paucity of literature in this area. Reports indicate that the intestinal mucosa may be a major site for fatty acid ethyl ester synthesis (32)
. Of several organs, the intestinal mucosa had high activities of the enzyme ethanol-O-acyltransferase (AEAT) using acyl-coenzyme A (CoA) and free fatty acids as substrates. The enzyme can synthesize esters with different alcohols and fatty acids. Several studies have demonstrated that fatty acid ethyl esters may be involved in organ injury. Although esters naturally occur in many foodstuffs, it would be expected that they would be mainly hydrolyzed by the human digestive system; however, this cannot be discounted as a source of gut esters. One of the most significant observations is the commonality of corresponding homologous series of acids, aldehydes, and alcohols, suggesting they can be oxidized/reduced from one form to another. This was shown to be the case between butyric acid and butanol by incubating [1-13C]butanoic acid with feces.
Stability of compounds in feces of asymptomatic donors
In each individual for the longitudinal study, 60% of VOCs was present in 60% of the samples, obtained within a 2-wk period, showing that resident flora generate common compounds irrespective of day-to-day changes in diet. The asymptomatic volunteers in the longitudinal study had a considerable degree of commonality as regards their daily diet (Table 1)
, yet all showed a range in the relative abundance of VOC with time (Fig. 1)
. Variation in the abundance then may reflect an individual’s resident flora or day-to-day changes in diet. Further work on the impact on food on the quantity of volatile produced is needed.
Specific volatile patterns in disease
Not surprisingly, many compounds present in the feces of asymptomatic donors were also present in gastrointestinal disease. These shared compounds likely reflect the ongoing interaction between dietary substrates and intestinal flora. However, distinct differences were found as well.
The range of VOCs from stool of donors with ulcerative colitis, C. jejuni, and C. difficile infections was considerably less than for the total number of different VOCs of the stools from asymptomatic volunteers (Tables 2
, 3
). The abundance of C. jejuni and C. difficile organisms in the gut, with presumably relatively lower abundance of other microbial flora, might be expected to reduce the number of secondary metabolites because of the reduced biodiversity, but reduced transit time arising from diarrhea may mean that fewer compounds were biosynthesized.
1-Octen-3-ol was rarely found in the cohort and longitudinal studies and was uncommon in the VOCs from stool from ulcerative colitis and C. difficile donors but was extremely common in C. jejuni patients. Although 1-octen-3-ol may be produced by fungi (45)
, its production in patients with C. jejuni remains unexplained.
The number of acids, alcohols, and ester compounds varies between the cohort study and disease study samples (Table 3)
. However, it is also important to appreciate that the differences in concentration in these compounds could vary widely and that future studies are needed to assess this. We cannot explain the low number of alkanes in the headspace of stools from diseased patients (Table 3)
, and the high number of alkenes in the VOCs from ulcerative colitis. The very low numbers of nitrogen-containing compounds (Table 4)
in the stool of patients with ulcerative colitis has not been previously reported and could be due to changes in the microflora of these patients. The main sulfur-containing flatus components have previously been reported to be hydrogen sulfide (1.06 µmol/l), followed by methanethiol (0.21 µmol/l), and dimethyl sulfide (0.08 µmol/l) (40)
. Methanethiol was present in just 61% of ulcerative colitis cases. The biosynthetic pathway for methanethiol is likely via methylation of hydrogen sulfide. Enhanced levels of sulfate reducing bacteria, which would result in hydrogen sulfide production, have been reported in ulcerative colitis (8)
.
Disease diagnoses using VOC data
Many of the VOCs found from donors with gastrointestinal conditions fall into the same 13 classes of compounds as described for the cohort data set (Table 3)
, but different patterns are observed in ulcerative colitis (UC), C. jejuni (CJ), and C. difficile (CD). Inspection of Table 2
permits a comparison, as follows, for each compound, by giving a value of 1 to the column designator for the lowest percentage frequency(ies) and summing up the values. The cohort:cluster of differentiation:CJ:UC values are 1:12:1:3 for all acids and 19:35:12:23 for all esters and show considerably fewer CD samples with acids and esters, compared to the cohort CJ and UC samples. Likewise, dimethyl sulfide, methanethiol, dimethyltrisulfide, and carbon disulfide are found in 100% of samples from healthy donors and are not present in many of the samples of patients with C. jejuni and C. difficile. On this type of rationale, a manageable small subset of compounds was selected for statistical analyses, which are asterisked in Table 2
. This permitted the differentiation of the three gastrointestinal diseases to be investigated (Fig. 2)
. Inspection of Fig. 2
shows that the first discriminant function strongly separates the Campylobacter group (positive scores on function 1) from the asymptomatic group (negative scores on function 1). The dominant loadings on the first discriminant function indicate that the presence of 1-butoxy-2-propanol is positively associated with Campylobacter, whereas the absence of 1-butoxy-2-propanol is positively associated with the asymptomatic group. Similarly, the loadings on the first discriminant function indicate that the presence of 3 (or 2)-methyl furan is positively associated with the normal group and its absence is positively associated with Campylobacter. The same applies for the presence and absence of dimethylsulfide. These findings are supported by the data in Table 2
. Inspection of Fig. 2
indicates that the second discriminant function largely separates the C. difficile group from the other groups. The dominant loadings on the second discriminant function indicate that the absence of 2-(2-ethoxyethoxy)-ethanol is positively associated with C. difficile group. The same applies for toluene, 6-methyl-3,5-heptadiene-2-one and hexanoic acid. These findings are supported by the percentages in Table 2
.
In summary this study has shown that SPME GCMS is a very effective method for rapidly qualitatively analyzing VOCs from fecal samples, most of which have not been previously reported. This catalog of compounds will be a foundation for the studies of disease-specific changes that occur within the human gut. Also, a large number of compounds—some of them considered toxic—need further investigation to determine their concentrations and biochemical role in the feces. Some of these compounds may be indicative of specific disease processes. The ability to differentiate between diseases by a detection of a small number of compounds should facilitate the development of rapid noninvasive diagnoses for ulcerative colitis and C. difficile and C. jejuni infections.
|
ACKNOWLEDGMENTS |
|---|
Received for publication July 21, 2006. Accepted for publication January 4, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Laska, R. M. R. Bautista, D. Hofelmann, V. Sterlemann, and L. T. H. Salazar Olfactory sensitivity for putrefaction-associated thiols and indols in three species of non-human primate J. Exp. Biol., December 1, 2007; 210(23): 4169 - 4178. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
), patients with ulcerative colitis (
), and patients with Clostridium difficile (
) infections.