Alimentary Pharmacology and Therapeutics
Review article: breath analysis in inflammatory bowel diseases S. Kurada*, N. Alkhouri†, C. Fiocchi‡,¶, R. Dweik§,¶ & F. Rieder‡,¶
*Department of Hospital Medicine, Medicine Institute, Cleveland, OH, USA. † Department of Pediatric Gastroenterology, Cleveland Clinic Children’s Hospital, Cleveland, OH, USA. ‡ Department of Gastroenterology and Hepatology, Digestive Disease Institute, Cleveland, OH, USA. § Department of Pulmonary Diseases, Respiratory Institute, Cleveland, OH, USA. ¶ Department of Pathobiology, Lerner Research Institute; Cleveland Clinic Foundation, Cleveland, OH, USA.
Correspondence to: Dr F. Rieder, Department of Pathobiology NC22, Lerner Research Institute, 9500 Euclid Avenue, Cleveland, OH 44195, USA. E-mail: fl[email protected]
Publication data Submitted 28 April 2014 First decision 15 May 2014 Resubmitted 20 July 2014 Resubmitted 20 November 2014 Accepted 21 November 2014 EV Pub Online 19 December 2014 This uncommissioned review article was subject to full peer-review.
SUMMARY Background There is an urgent need for cheap, reproducible, easy to perform and specific biomarkers for diagnosis, differentiation and stratification of inflammatory bowel disease (IBD) patients. Technical advances allow for the determination of volatile organic compounds in the human breath to differentiate between health and disease. Aim Review and discuss medical literature on volatile organic compounds in exhaled human breath in GI disorders, focusing on diagnosis and differentiation of IBD. Methods A systematic search in PubMed, Ovid Medline and Scopus was completed using appropriate keywords. In addition, a bibliography search of each article was performed. Results Mean breath pentane, ethane, propane, 1-octene, 3-methylhexane, 1-decene and NO levels were elevated (P < 0.05 to P < 10 7) and mean breath 1nonene, (E)-2-nonene, hydrogen sulphide and methane were decreased in IBD compared to healthy controls (P = 0.003 to P < 0.001). A combined panel of 3 volatile organic compounds (octene, (E)-2-nonene and decene) showed the best discrimination between paediatric IBD and controls (AUC 0.96). Breath condensate cytokines were higher in IBD compared to healthy individuals (P < 0.008). Breath pentane, ethane, propane, isoprene and NO levels correlated with disease activity in IBD patients. Breath condensate interleukin-1b showed an inverse relation with clinical disease activity. Conclusions Breath analysis in IBD is a promising approach that is not yet ready for routine clinical use, but data from other gastrointestinal diseases suggest the feasibility for use of this technology in clinical practice. Well-designed future trials, incorporating the latest breath detection techniques, need to determine the exact breath metabolome pattern linked to diagnosis and phenotype of IBD. Aliment Pharmacol Ther 2015; 41: 329–341
ª 2014 John Wiley & Sons Ltd doi:10.1111/apt.13050
329
S. Kurada et al. INTRODUCTION The diagnosis and differentiation of inflammatory bowel diseases (IBD) lacks a single gold standard and takes a multimodal approach involving clinical, endoscopic, histologic, serological and radiological observations.1 IBD commonly suffers from a delay in time from first occurrence of symptoms to diagnosis, especially in young patients with ileal disease, which hinders our ability to alter the progression of disease.2 In addition, once the diagnosis of IBD is made its sub categorisation into Crohn’s disease (CD) or ulcerative colitis (UC) is critical for determining the optimal treatment strategy. Even employing all currently available diagnostic modalities the sub categorisation of IBD remains elusive in about 10–15% of patients, who are then classified as indeterminate colitis or inflammatory bowel disease-unclassified (IBD-U).3 After defining the IBD subtype the definitive diagnosis can change, with reclassification rates as high as 10%.4 Treatment decisions are mainly based on clinical, serologic, radiological or endoscopic parameters, each with their obvious limitations: (i) Clinical or serological parameters of disease activity are insufficient as they only poorly correlate with intestinal disease activity and are not specific for IBD5–16 (ii) Endoscopic examination of the intestine is linked to high costs, patient discomfort, noncompliance and possible complications such as perforation17, 18 and (iii) radiological procedures may be associated with interobserver and intermodality variability.19 It remains unclear, which patients with an initially inflammatory disease course develop into a more severe vs. benign disease phenotype. It is necessary to identify at risk populations that could benefit from a tailored therapeutic approach. The use of biological and/or immunomodulator agents for therapy might be justified early in the disease course for patients at risk for rapid disease progression.20, 21 Hence, markers aiding the early and accurate diagnosis, disease course prediction or stratification are needed and would improve patient care. In this article, we review and discuss the medical literature on human breath as a novel biomarker in GI disorders, focusing on diagnosis and differentiation of IBD. METHODS A comprehensive literature search was performed to assess all relevant citations found in PubMed, Ovid Medline (service of the US National Library of Medicine and the National Institutes of Health) and Scopus for the following key words: ‘breath’, ‘metabolome’, ‘inflammatory bowel disease’, ‘Crohn’s disease’, ‘Crohn’s’, ‘Ulcerative colitis’, 330
‘volatile organic compounds’ or ‘electronic nose’. Additionally, references of cited original articles and reviews were further assessed for relevant work. These data together with the authors’ personal experience in the field represent the basis of this review.
THE METABOLOME IN IBD Metabolomics are defined as the investigation of a group of endpoint metabolites of a physiological or pathophysiological process.22 Metabolomics have the potential to provide a signature pattern for specific disease processes. The metabolome is particularly appealing for this purpose as it also accounts for extraneous alterations, such as diet or environmental influences. The advent of various advanced techniques, such as mass spectrometry, chromatography and nuclear magnetic resonance techniques have aided in the revelation of multiple categories of metabolites such as aliphatic compounds, alkaloids, amino acids, peptides, organometallic compounds, nucleotides or nucleosides and their derivatives among others.23 Ample experience exists from a variety of diseases outside of the intestine, employing the study of the metabolome for the identification of biomarkers. Several disease conditions have already established metabolome patterns to aid in diagnosis, prognosis and treatment of disease processes, like cancer,24, 25 asthma,26 diabetes,27 heart failure28 or kidney disease.29 The metabolome has also been evaluated in several gastrointestinal disorders. Data exist for (i) using tissue for the diagnosis of colorectal cancer in mucosal samples by investigating amino acids,30 (ii) measuring organic acids like tetra/hexadecanoic acid among others, in samples of oesophageal cancer,31 (iii) determination of organic acids like heptanedioic acid, propanoic acid and organic alcohols like phenanthrenol or butanetriol in gastric cancer32 and (iv) identification of organic acids and alcohols in hepato-pancreatic cancer.33, 34 Metabolomic studies have recently entered the field of IBD, using serum, urine and tissue samples in humans and IBD animal models. The serum metabolome may help distinguish IBD from healthy subjects and provide a print specific to different IBD types. For instance in IL-10 deficient mice, animals developing experimental colitis, an altered lipoprotein profile as well as elevated levels of glycolysis intermediaries and free amino acids have been demonstrated.35 In the TNF (DARE/WT) mouse model of IBD an altered metabolism of cholesterol, triglycerides, phospholipids, plasmalogens and sphingomyelins in the inflamed tissue (ileum) and the adjacent intestinal segments (proximal colon) was Aliment Pharmacol Ther 2015; 41: 329-341 ª 2014 John Wiley & Sons Ltd
Review: breath analysis in IBD noted, suggesting altered cell membrane composition, energy homeostasis and generation of inflammatory lipid mediators.36 In humans, serum metabolome data indicate elevated levels of 3-hydroxybutyrate, b-glucose, a-glucose, and phenylalanine, but decreased lipids in IBD patients compared to healthy controls.37 Another serum study indicated metabolome variations in lipid and choline metabolism among patients with CD, UC and healthy controls, especially with levels pertaining to lipoproteins like LDL and HDL, choline, N-acetyl glycoprotein, and amino acids like isoleucine and alanine with discrimination between IBD, CD and UC from controls (with predictive models showing Q2, R2 values for CD: UC; CD: controls; UC: controls; IBD: controls; IBD on no medications corresponding to 0.51, 0.60; 0.75, 0.78; 0.71, 0.87; 0.66, 0.68; 0.88, 0.90 respectively).38 The urine metabolome has also been investigated to distinguish CD from UC and healthy controls. Intermediaries of tricarboxylic acid cycle and amino acids were useful in distinguishing IBD from healthy controls (R2 = 0.668; Q2 = 0.482), but failed to differentiate CD from UC when surgical confounders were excluded.39 Currently, the limitation of urine metabolome studies lay in the fact that only about a third of all metabolites have been quantified.23 Intestinal biopsy samples can serve as a direct tissue source for measurement of the metabolome. They are, however, connected to the inconvenience of obtaining biopsy samples and the associated risk for complications. As an example, a study by Bjerrum et al. showed elevated levels of antioxidants like ascorbic acid and glutathione amino acids like glutamate, glutamine, taurine and asparate and lower levels of lipids, like glycerophosphocholine, myoinositol and betaine in active UC compared to healthy controls and quiescent UC with correlation coefficients ranging from 0.39 to +0.64.40 Hence, various components of the metabolome in serum, urine, faeces and mucosal samples have been demonstrated to be significantly different between CD, UC and healthy controls.23, 37, 38, 41 Limitation for the use of metabolomic markers are high costs, the need for an elaborate infrastructure for analysis or invasive testing in case of biopsies. An ideal biomarker, however, is one that is simple, easy to perform, minimally invasive, cheap, rapid and reproducible.12 In this respect, the breath metabolome offers great potential. Aliment Pharmacol Ther 2015; 41: 329-341 ª 2014 John Wiley & Sons Ltd
THE BREATH METABOLOME For centuries characteristic breath odours have been described in certain diseases and used by physicians to diagnose and treat patients, such as uraemic breath in kidney disease, fruity odour in diabetic ketoacidosis or fetor hepaticus in liver failure. The exhaled human breath contains a complex composition of at least 3000 different compounds, ultimately creating the breath odours we perceive when examining our patients. The human exhaled breath is a mixture of nonvolatile (condensate) and volatile compounds, which further can be classified as organic and inorganic substances. In the past, volatility and very low concentrations of breath components as well as difficulties with standardisation and normalisation have limited our ability to analyse them. However, these challenges have been overcome with advanced analysis techniques like gas chromatography and selective ion flow mass spectrometry. Breath samples can now be stored and analysed subsequently to add to convenience of lab logistics. The determination is cheap, can be performed in real time and can be easily repeated. Breath volatile organic compounds (VOC)s can be broadly classified as being of endogenous or exogenous origin. Endogenous compounds represent VOCs with a positive alveolar gradient defined by the positive difference in the abundance of VOCs in breath to the ambient air. They may be classified as ‘innate compounds’, which are produced endogenously, or ‘induced compounds’ that are produced after introducing a particular substrate and tracing its end products. Exogenous compounds are signified by a negative alveolar gradient, with an approximate divide of 50% into each category (Figure 1). Sources of exogenous VOCs are mostly the environment and the diet.42 Available breath analytic techniques can be classified as one of two broad types, namely pattern recognition, using groups of volatile compounds, or identification of single and individual compounds.43 The major pattern recognition technique is the electronic nose. It consists of an array of chemical sensors in a matrix platform, which identifies breath prints of a complex VOC mixture using statistical algorithms. The device typically includes multi-sensor arrays, an information-processing unit, digital pattern-recognition algorithms and databases for comparison.44 This is akin to a smell print that humans or animals use to identify various substances, for example differentiating the aroma of various perfumes or the various aromas of coffee or food from each other.45 In the industrial setting, electronic nose technology has been used to detect explosive 331
S. Kurada et al. Diet
Environment
Induced
Endogenous
Exogenous
Organic
Innate
Inorganic
Organic
Non volatile ( Condensate)
Inorganic
Volatile
Exhaled breath
vapour mixtures, adequacy of processing wine grapes and detecting freshness and adulteration of foods. While being able to differentiate distinct states of health and disease with high accuracy, the major disadvantage of the electronic nose is its reliance on patterns of breath only. It lacks the ability to determine individual compounds and their levels, which can be crucial when trying to link the compound groups to the relevant underlying pathobiological pathways. The major techniques for the identification of individual VOCs in IBD are gas chromatography (GC) and the selective ion flow tube (SIFT). GC involves an interaction between a mobile phase (inert gas with a mixture of unknown gases) and a stationary phase (solid or liquid). As the mobile phase interacts with the stationary phase based on the principles of adsorption, the length of time individual gases take to exit the stationary phase determines the retention time. The retention time is influenced by several factors like the boiling point of individual compounds, polarity, temperature of flow tube, flow rate, column length, among others. The SIFT works on the principle of creation of reagent ions by a quadrupole. These reagent ions ionise individual gases of a complex gaseous mixture. The ionised compounds are then introduced into another quadrupole, which help separate the individual ionised reaction products. Following GC and SIFT, the individual breath compounds are identified using detection techniques like mass spectrometry (MS; ionised compounds are identified based on mass to charge ratio) or flame ionisation (ionisation of compounds in a hydrogen flame). 332
Figure 1 | The human breath metabolome: The human breath is a gaseous mix of volatile and nonvolatile carbon (organic) and noncarbon (inorganic) compounds. These may be derived from exogenous sources, like diet or environment and from endogenous sources, which may be induced (by introducing substrates) or produced innately (in physiological or pathological states).
The exploration of VOCs can provide insight into physiological or pathophysiological processes in particular in cholesterol metabolism, lipid peroxidation and bacterial metabolism.46 In health, the breath metabolome can vary with the dietary content of various carbohydrates, proteins and lipids and also by states of starvation, exercise, medications and feeding (Figure 2; adapted from Ajibola et al.).47 Unabsorbed carbohydrates undergo a process of bacterial fermentation in the intestine. This leads to the formation of hydrogen, methane and ethanol. Absorbed carbohydrates on the other hand are converted to acetyl CoA by glycolysis of simple monosaccharides. Acetyl CoA enters one of several pathways: (i) consumption in the Krebs cycle for generation of ATP, (ii) production of acetoacetyl CoA, which then undergoes the process of ketogenesis, further evolving into acetone that can be detected in breath or (iii) acetoacetyl CoA may also enter the HMG-CoA pathway and lead to cholesterol production with isoprene being one of the intermediaries.47 Bacterial fermentation of proteins creates several intermediaries like branched chain fatty acids, ammonia and hydrogen sulphide. Branched chain fatty acids upon decarboxylation and b-oxidation reactions also form intermediaries, such as acetyl CoA. In addition, metabolism of proteins leads to formation of several amino acids, which can yield ammonium ions. These ammonium ions on reaction with bicarbonate generate carbamyl phosphate, which then enters the urea cycle leading to the formation of urea, and intermediaries of the urea cycle like arginine. L-arginine may be derived from the Aliment Pharmacol Ther 2015; 41: 329-341 ª 2014 John Wiley & Sons Ltd
Review: breath analysis in IBD
Bacterial fermentation
Proteins
Methane Hydrogen Ethanol
Bacterial fermentation
Unabsorbed carbohydrates
Arginine
Ammonia Hydrogen sulfide
Nitric oxide synthase
Nitric oxide
Urea cycle Amino acids
Monosaccharides
Ammonia
Ammonium
Glycolysis
β-Oxidation
Acetyl CoA HMG CoA
Long chain fatty acids
Peroxidation
Ketogenesis
Acetone Isoprene Pentane Ethane Ethylene
Figure 2 | The influence of diet on breath volatile organic compounds: Bacterial fermentation of complex and unabsorbed carbohydrates and proteins leads to generation of gases like methane, hydrogen, hydrogen sulphide and ammonia. Carbohydrates and fatty acid metabolism yields acetyl CoA, a central molecule in the generation of volatile organic compounds (VOCs). Nitric oxide and ammonia are derived from protein metabolism. Isoprene is an intermediary of cholesterol synthesis. VOCs are depicted in green and the dietary components in red. Abbreviations: CoA: Co enzyme A. Adapted from Ajibola et al.47
amino acid pool or urea cycle and can be converted to nitric oxide in presence of nitric oxide synthase. The breath metabolome has been investigated in several diseases outside of the intestine. This has been explored using single compounds (once clinical significance could be clearly established) or breath signature patterns (when more than one compound was found to be altered). Ion mobility spectrometers combined with a multi-capillary column technology have been used to help differentiate chronic obstructive pulmonary disease (COPD) patients from healthy controls and also aid in the diagnosis of bronchial cancer.48 The fractional excretion of NO (FeNO) is used for the diagnosis of asthma.49 The analysis of the breath condensate has revealed several genetic and protein biomarkers, which may aid in population screening techniques for lung cancers and other inflammatory conditions of the airway like asthma, rhinitis and COPD.50, 51 Several breath VOCs are specifically elevated in respiratory infections due to mycobacteria and pseudomonas and may serve as alternatives to more invasive sample procurement techniques like bronchoscopy.52 Gastroenterological infections due to Vibrio cholera and Helicobacter pylori are extra-pulmonary organisms that can be sources of an altered breath metabolome.52 Other disease Aliment Pharmacol Ther 2015; 41: 329-341 ª 2014 John Wiley & Sons Ltd
conditions associated with various VOC breath prints include malignancies like breast cancer,53 colorectal cancer,54 neurological disorders like Alzheimer’s and Parkinson’s disease,55 multiple sclerosis,56 renal disorders like chronic kidney disease,57 cardiac disorders like congestive heart failure,58 childhood obesity,59 pulmonary hypertension60 and asthma.61 Being the host to crucial metabolic reactions, the liver can serve as a major source of VOCs in the exhaled human breath. A recent study by our group suggests elevated levels of isoprene, acetone, trimethylamine, acetaldehyde, and pentane in obese children with non-alcoholic fatty liver disease compared with obese children with normal liver. A panel of four VOCs can identify the presence of NAFLD with an area under curve on ROC analysis being more than 0.9.62 We further demonstrated that breath prints in alcoholic liver disease show significantly higher levels of pentane and trimethylamine levels compared to other acute and chronic liver diseases.63 An additional study elucidated a breath print of 12 VOCs distinguishing liver cirrhosis from healthy controls and 7 VOCs distinguishing compensated from decompensated liver disease.64 On using a panel of 385 volatile organic compounds Khalid and colleagues identified hepatic encephalopathy with a 333
S. Kurada et al. sensitivity of 91% in alcoholic cirrhosis compared to healthy controls. In patients with no demonstrable hepatic encephalopathy, the aetiology of cirrhosis was established as being secondary to alcohol in 78.3% of patients and non-alcoholic causes in 69.2% of patients. A distinction between non-alcoholic cirrhosis and alcoholic cirrhosis from healthy controls was made with a sensitivity and specificity of 92.3% and 97.1% respectively. Harmful drinking patterns were identified with 100% sensitivity when compared to healthy controls.65
THE BREATH METABOLOME IN IBD In contrast to other diseases, to date, only 12 mainly small studies evaluated the breath metabolome or condensate in IBD for diagnosis, differentiation or disease activity assessment (Table 1 summarises studies on diagnosis and differentiation of IBD). Most of these reports only provide mean values per group as comparisons that do not allow assessment of the area under the ROC curve (AUC), sensitivity, specificity, positive predictive value or negative predictive value. Diagnosis and differentiation Single VOCs tested for the diagnosis and differentiation of IBD include pentane, ethane, propane, butane or nitric oxide (NO) and multiple compounds have been shown to be elevated in IBD. The best-explored gas is pentane. Dryahina et al. have demonstrated elevated levels of pentane in IBD (CD>UC) compared to healthy controls with an area under the ROC curve (AUC) of 0.927 (Figure 3; adapted from original study with permission, P < 10 7).66 Given the presented data, a pentane cut-off of 84 ppbv provided an estimated sensitivity of 0.55 and a specificity of 1. Breath pentane in CD patients was higher compared to UC with, however, an only moderate discriminatory capacity (P = 0.0058, AUC of 0.68). Even though the primary aim of this investigation was the quantification of breath pentane levels, this study failed to consider possible confounders like drugs, diet, and other disease processes in baseline demographics that may affect the results. While Pelli et al.67 demonstrated mean pentane levels to be elevated in IBD compared to controls (P < 0.001; only mean values provided), confirming the above results, Sedghi et al. failed to do so.68 An association between ethane and IBD compared to controls was demonstrated in two investigations (P < 0.001 and P = 0.013)67, 68 and between breath propane and IBD in one study (P < 0.001),67 but only mean comparisons were provided. Exhaled NO was significantly higher in patients with UC and CD compared to 334
HC (P = 0.05–0.01) with no difference being noted between UC and CD.69 Our group has recently completed a prospective cross-sectional, single-centre study that evaluated a panel of 21 VOCs in the exhaled breath of paediatric IBD patients (n = 62) and healthy controls (n = 55) using SIFT-MS. After adjusting for age, gender, race and BMI, routinely analysed VOCs showed significant differences in 6 VOCs between the two groups (Increased levels in IBD compared to controls for: 1octene, 3-methylhexane, 1-decene; Decreased levels in IBD compared to controls for: 1-nonene, (E)-2-nonene, hydrogen sulphide; P = 0.006 to
Review article: breath analysis in inflammatory bowel diseases S. Kurada*, N. Alkhouri†, C. Fiocchi‡,¶, R. Dweik§,¶ & F. Rieder‡,¶
*Department of Hospital Medicine, Medicine Institute, Cleveland, OH, USA. † Department of Pediatric Gastroenterology, Cleveland Clinic Children’s Hospital, Cleveland, OH, USA. ‡ Department of Gastroenterology and Hepatology, Digestive Disease Institute, Cleveland, OH, USA. § Department of Pulmonary Diseases, Respiratory Institute, Cleveland, OH, USA. ¶ Department of Pathobiology, Lerner Research Institute; Cleveland Clinic Foundation, Cleveland, OH, USA.
Correspondence to: Dr F. Rieder, Department of Pathobiology NC22, Lerner Research Institute, 9500 Euclid Avenue, Cleveland, OH 44195, USA. E-mail: fl[email protected]
Publication data Submitted 28 April 2014 First decision 15 May 2014 Resubmitted 20 July 2014 Resubmitted 20 November 2014 Accepted 21 November 2014 EV Pub Online 19 December 2014 This uncommissioned review article was subject to full peer-review.
SUMMARY Background There is an urgent need for cheap, reproducible, easy to perform and specific biomarkers for diagnosis, differentiation and stratification of inflammatory bowel disease (IBD) patients. Technical advances allow for the determination of volatile organic compounds in the human breath to differentiate between health and disease. Aim Review and discuss medical literature on volatile organic compounds in exhaled human breath in GI disorders, focusing on diagnosis and differentiation of IBD. Methods A systematic search in PubMed, Ovid Medline and Scopus was completed using appropriate keywords. In addition, a bibliography search of each article was performed. Results Mean breath pentane, ethane, propane, 1-octene, 3-methylhexane, 1-decene and NO levels were elevated (P < 0.05 to P < 10 7) and mean breath 1nonene, (E)-2-nonene, hydrogen sulphide and methane were decreased in IBD compared to healthy controls (P = 0.003 to P < 0.001). A combined panel of 3 volatile organic compounds (octene, (E)-2-nonene and decene) showed the best discrimination between paediatric IBD and controls (AUC 0.96). Breath condensate cytokines were higher in IBD compared to healthy individuals (P < 0.008). Breath pentane, ethane, propane, isoprene and NO levels correlated with disease activity in IBD patients. Breath condensate interleukin-1b showed an inverse relation with clinical disease activity. Conclusions Breath analysis in IBD is a promising approach that is not yet ready for routine clinical use, but data from other gastrointestinal diseases suggest the feasibility for use of this technology in clinical practice. Well-designed future trials, incorporating the latest breath detection techniques, need to determine the exact breath metabolome pattern linked to diagnosis and phenotype of IBD. Aliment Pharmacol Ther 2015; 41: 329–341
ª 2014 John Wiley & Sons Ltd doi:10.1111/apt.13050
329
S. Kurada et al. INTRODUCTION The diagnosis and differentiation of inflammatory bowel diseases (IBD) lacks a single gold standard and takes a multimodal approach involving clinical, endoscopic, histologic, serological and radiological observations.1 IBD commonly suffers from a delay in time from first occurrence of symptoms to diagnosis, especially in young patients with ileal disease, which hinders our ability to alter the progression of disease.2 In addition, once the diagnosis of IBD is made its sub categorisation into Crohn’s disease (CD) or ulcerative colitis (UC) is critical for determining the optimal treatment strategy. Even employing all currently available diagnostic modalities the sub categorisation of IBD remains elusive in about 10–15% of patients, who are then classified as indeterminate colitis or inflammatory bowel disease-unclassified (IBD-U).3 After defining the IBD subtype the definitive diagnosis can change, with reclassification rates as high as 10%.4 Treatment decisions are mainly based on clinical, serologic, radiological or endoscopic parameters, each with their obvious limitations: (i) Clinical or serological parameters of disease activity are insufficient as they only poorly correlate with intestinal disease activity and are not specific for IBD5–16 (ii) Endoscopic examination of the intestine is linked to high costs, patient discomfort, noncompliance and possible complications such as perforation17, 18 and (iii) radiological procedures may be associated with interobserver and intermodality variability.19 It remains unclear, which patients with an initially inflammatory disease course develop into a more severe vs. benign disease phenotype. It is necessary to identify at risk populations that could benefit from a tailored therapeutic approach. The use of biological and/or immunomodulator agents for therapy might be justified early in the disease course for patients at risk for rapid disease progression.20, 21 Hence, markers aiding the early and accurate diagnosis, disease course prediction or stratification are needed and would improve patient care. In this article, we review and discuss the medical literature on human breath as a novel biomarker in GI disorders, focusing on diagnosis and differentiation of IBD. METHODS A comprehensive literature search was performed to assess all relevant citations found in PubMed, Ovid Medline (service of the US National Library of Medicine and the National Institutes of Health) and Scopus for the following key words: ‘breath’, ‘metabolome’, ‘inflammatory bowel disease’, ‘Crohn’s disease’, ‘Crohn’s’, ‘Ulcerative colitis’, 330
‘volatile organic compounds’ or ‘electronic nose’. Additionally, references of cited original articles and reviews were further assessed for relevant work. These data together with the authors’ personal experience in the field represent the basis of this review.
THE METABOLOME IN IBD Metabolomics are defined as the investigation of a group of endpoint metabolites of a physiological or pathophysiological process.22 Metabolomics have the potential to provide a signature pattern for specific disease processes. The metabolome is particularly appealing for this purpose as it also accounts for extraneous alterations, such as diet or environmental influences. The advent of various advanced techniques, such as mass spectrometry, chromatography and nuclear magnetic resonance techniques have aided in the revelation of multiple categories of metabolites such as aliphatic compounds, alkaloids, amino acids, peptides, organometallic compounds, nucleotides or nucleosides and their derivatives among others.23 Ample experience exists from a variety of diseases outside of the intestine, employing the study of the metabolome for the identification of biomarkers. Several disease conditions have already established metabolome patterns to aid in diagnosis, prognosis and treatment of disease processes, like cancer,24, 25 asthma,26 diabetes,27 heart failure28 or kidney disease.29 The metabolome has also been evaluated in several gastrointestinal disorders. Data exist for (i) using tissue for the diagnosis of colorectal cancer in mucosal samples by investigating amino acids,30 (ii) measuring organic acids like tetra/hexadecanoic acid among others, in samples of oesophageal cancer,31 (iii) determination of organic acids like heptanedioic acid, propanoic acid and organic alcohols like phenanthrenol or butanetriol in gastric cancer32 and (iv) identification of organic acids and alcohols in hepato-pancreatic cancer.33, 34 Metabolomic studies have recently entered the field of IBD, using serum, urine and tissue samples in humans and IBD animal models. The serum metabolome may help distinguish IBD from healthy subjects and provide a print specific to different IBD types. For instance in IL-10 deficient mice, animals developing experimental colitis, an altered lipoprotein profile as well as elevated levels of glycolysis intermediaries and free amino acids have been demonstrated.35 In the TNF (DARE/WT) mouse model of IBD an altered metabolism of cholesterol, triglycerides, phospholipids, plasmalogens and sphingomyelins in the inflamed tissue (ileum) and the adjacent intestinal segments (proximal colon) was Aliment Pharmacol Ther 2015; 41: 329-341 ª 2014 John Wiley & Sons Ltd
Review: breath analysis in IBD noted, suggesting altered cell membrane composition, energy homeostasis and generation of inflammatory lipid mediators.36 In humans, serum metabolome data indicate elevated levels of 3-hydroxybutyrate, b-glucose, a-glucose, and phenylalanine, but decreased lipids in IBD patients compared to healthy controls.37 Another serum study indicated metabolome variations in lipid and choline metabolism among patients with CD, UC and healthy controls, especially with levels pertaining to lipoproteins like LDL and HDL, choline, N-acetyl glycoprotein, and amino acids like isoleucine and alanine with discrimination between IBD, CD and UC from controls (with predictive models showing Q2, R2 values for CD: UC; CD: controls; UC: controls; IBD: controls; IBD on no medications corresponding to 0.51, 0.60; 0.75, 0.78; 0.71, 0.87; 0.66, 0.68; 0.88, 0.90 respectively).38 The urine metabolome has also been investigated to distinguish CD from UC and healthy controls. Intermediaries of tricarboxylic acid cycle and amino acids were useful in distinguishing IBD from healthy controls (R2 = 0.668; Q2 = 0.482), but failed to differentiate CD from UC when surgical confounders were excluded.39 Currently, the limitation of urine metabolome studies lay in the fact that only about a third of all metabolites have been quantified.23 Intestinal biopsy samples can serve as a direct tissue source for measurement of the metabolome. They are, however, connected to the inconvenience of obtaining biopsy samples and the associated risk for complications. As an example, a study by Bjerrum et al. showed elevated levels of antioxidants like ascorbic acid and glutathione amino acids like glutamate, glutamine, taurine and asparate and lower levels of lipids, like glycerophosphocholine, myoinositol and betaine in active UC compared to healthy controls and quiescent UC with correlation coefficients ranging from 0.39 to +0.64.40 Hence, various components of the metabolome in serum, urine, faeces and mucosal samples have been demonstrated to be significantly different between CD, UC and healthy controls.23, 37, 38, 41 Limitation for the use of metabolomic markers are high costs, the need for an elaborate infrastructure for analysis or invasive testing in case of biopsies. An ideal biomarker, however, is one that is simple, easy to perform, minimally invasive, cheap, rapid and reproducible.12 In this respect, the breath metabolome offers great potential. Aliment Pharmacol Ther 2015; 41: 329-341 ª 2014 John Wiley & Sons Ltd
THE BREATH METABOLOME For centuries characteristic breath odours have been described in certain diseases and used by physicians to diagnose and treat patients, such as uraemic breath in kidney disease, fruity odour in diabetic ketoacidosis or fetor hepaticus in liver failure. The exhaled human breath contains a complex composition of at least 3000 different compounds, ultimately creating the breath odours we perceive when examining our patients. The human exhaled breath is a mixture of nonvolatile (condensate) and volatile compounds, which further can be classified as organic and inorganic substances. In the past, volatility and very low concentrations of breath components as well as difficulties with standardisation and normalisation have limited our ability to analyse them. However, these challenges have been overcome with advanced analysis techniques like gas chromatography and selective ion flow mass spectrometry. Breath samples can now be stored and analysed subsequently to add to convenience of lab logistics. The determination is cheap, can be performed in real time and can be easily repeated. Breath volatile organic compounds (VOC)s can be broadly classified as being of endogenous or exogenous origin. Endogenous compounds represent VOCs with a positive alveolar gradient defined by the positive difference in the abundance of VOCs in breath to the ambient air. They may be classified as ‘innate compounds’, which are produced endogenously, or ‘induced compounds’ that are produced after introducing a particular substrate and tracing its end products. Exogenous compounds are signified by a negative alveolar gradient, with an approximate divide of 50% into each category (Figure 1). Sources of exogenous VOCs are mostly the environment and the diet.42 Available breath analytic techniques can be classified as one of two broad types, namely pattern recognition, using groups of volatile compounds, or identification of single and individual compounds.43 The major pattern recognition technique is the electronic nose. It consists of an array of chemical sensors in a matrix platform, which identifies breath prints of a complex VOC mixture using statistical algorithms. The device typically includes multi-sensor arrays, an information-processing unit, digital pattern-recognition algorithms and databases for comparison.44 This is akin to a smell print that humans or animals use to identify various substances, for example differentiating the aroma of various perfumes or the various aromas of coffee or food from each other.45 In the industrial setting, electronic nose technology has been used to detect explosive 331
S. Kurada et al. Diet
Environment
Induced
Endogenous
Exogenous
Organic
Innate
Inorganic
Organic
Non volatile ( Condensate)
Inorganic
Volatile
Exhaled breath
vapour mixtures, adequacy of processing wine grapes and detecting freshness and adulteration of foods. While being able to differentiate distinct states of health and disease with high accuracy, the major disadvantage of the electronic nose is its reliance on patterns of breath only. It lacks the ability to determine individual compounds and their levels, which can be crucial when trying to link the compound groups to the relevant underlying pathobiological pathways. The major techniques for the identification of individual VOCs in IBD are gas chromatography (GC) and the selective ion flow tube (SIFT). GC involves an interaction between a mobile phase (inert gas with a mixture of unknown gases) and a stationary phase (solid or liquid). As the mobile phase interacts with the stationary phase based on the principles of adsorption, the length of time individual gases take to exit the stationary phase determines the retention time. The retention time is influenced by several factors like the boiling point of individual compounds, polarity, temperature of flow tube, flow rate, column length, among others. The SIFT works on the principle of creation of reagent ions by a quadrupole. These reagent ions ionise individual gases of a complex gaseous mixture. The ionised compounds are then introduced into another quadrupole, which help separate the individual ionised reaction products. Following GC and SIFT, the individual breath compounds are identified using detection techniques like mass spectrometry (MS; ionised compounds are identified based on mass to charge ratio) or flame ionisation (ionisation of compounds in a hydrogen flame). 332
Figure 1 | The human breath metabolome: The human breath is a gaseous mix of volatile and nonvolatile carbon (organic) and noncarbon (inorganic) compounds. These may be derived from exogenous sources, like diet or environment and from endogenous sources, which may be induced (by introducing substrates) or produced innately (in physiological or pathological states).
The exploration of VOCs can provide insight into physiological or pathophysiological processes in particular in cholesterol metabolism, lipid peroxidation and bacterial metabolism.46 In health, the breath metabolome can vary with the dietary content of various carbohydrates, proteins and lipids and also by states of starvation, exercise, medications and feeding (Figure 2; adapted from Ajibola et al.).47 Unabsorbed carbohydrates undergo a process of bacterial fermentation in the intestine. This leads to the formation of hydrogen, methane and ethanol. Absorbed carbohydrates on the other hand are converted to acetyl CoA by glycolysis of simple monosaccharides. Acetyl CoA enters one of several pathways: (i) consumption in the Krebs cycle for generation of ATP, (ii) production of acetoacetyl CoA, which then undergoes the process of ketogenesis, further evolving into acetone that can be detected in breath or (iii) acetoacetyl CoA may also enter the HMG-CoA pathway and lead to cholesterol production with isoprene being one of the intermediaries.47 Bacterial fermentation of proteins creates several intermediaries like branched chain fatty acids, ammonia and hydrogen sulphide. Branched chain fatty acids upon decarboxylation and b-oxidation reactions also form intermediaries, such as acetyl CoA. In addition, metabolism of proteins leads to formation of several amino acids, which can yield ammonium ions. These ammonium ions on reaction with bicarbonate generate carbamyl phosphate, which then enters the urea cycle leading to the formation of urea, and intermediaries of the urea cycle like arginine. L-arginine may be derived from the Aliment Pharmacol Ther 2015; 41: 329-341 ª 2014 John Wiley & Sons Ltd
Review: breath analysis in IBD
Bacterial fermentation
Proteins
Methane Hydrogen Ethanol
Bacterial fermentation
Unabsorbed carbohydrates
Arginine
Ammonia Hydrogen sulfide
Nitric oxide synthase
Nitric oxide
Urea cycle Amino acids
Monosaccharides
Ammonia
Ammonium
Glycolysis
β-Oxidation
Acetyl CoA HMG CoA
Long chain fatty acids
Peroxidation
Ketogenesis
Acetone Isoprene Pentane Ethane Ethylene
Figure 2 | The influence of diet on breath volatile organic compounds: Bacterial fermentation of complex and unabsorbed carbohydrates and proteins leads to generation of gases like methane, hydrogen, hydrogen sulphide and ammonia. Carbohydrates and fatty acid metabolism yields acetyl CoA, a central molecule in the generation of volatile organic compounds (VOCs). Nitric oxide and ammonia are derived from protein metabolism. Isoprene is an intermediary of cholesterol synthesis. VOCs are depicted in green and the dietary components in red. Abbreviations: CoA: Co enzyme A. Adapted from Ajibola et al.47
amino acid pool or urea cycle and can be converted to nitric oxide in presence of nitric oxide synthase. The breath metabolome has been investigated in several diseases outside of the intestine. This has been explored using single compounds (once clinical significance could be clearly established) or breath signature patterns (when more than one compound was found to be altered). Ion mobility spectrometers combined with a multi-capillary column technology have been used to help differentiate chronic obstructive pulmonary disease (COPD) patients from healthy controls and also aid in the diagnosis of bronchial cancer.48 The fractional excretion of NO (FeNO) is used for the diagnosis of asthma.49 The analysis of the breath condensate has revealed several genetic and protein biomarkers, which may aid in population screening techniques for lung cancers and other inflammatory conditions of the airway like asthma, rhinitis and COPD.50, 51 Several breath VOCs are specifically elevated in respiratory infections due to mycobacteria and pseudomonas and may serve as alternatives to more invasive sample procurement techniques like bronchoscopy.52 Gastroenterological infections due to Vibrio cholera and Helicobacter pylori are extra-pulmonary organisms that can be sources of an altered breath metabolome.52 Other disease Aliment Pharmacol Ther 2015; 41: 329-341 ª 2014 John Wiley & Sons Ltd
conditions associated with various VOC breath prints include malignancies like breast cancer,53 colorectal cancer,54 neurological disorders like Alzheimer’s and Parkinson’s disease,55 multiple sclerosis,56 renal disorders like chronic kidney disease,57 cardiac disorders like congestive heart failure,58 childhood obesity,59 pulmonary hypertension60 and asthma.61 Being the host to crucial metabolic reactions, the liver can serve as a major source of VOCs in the exhaled human breath. A recent study by our group suggests elevated levels of isoprene, acetone, trimethylamine, acetaldehyde, and pentane in obese children with non-alcoholic fatty liver disease compared with obese children with normal liver. A panel of four VOCs can identify the presence of NAFLD with an area under curve on ROC analysis being more than 0.9.62 We further demonstrated that breath prints in alcoholic liver disease show significantly higher levels of pentane and trimethylamine levels compared to other acute and chronic liver diseases.63 An additional study elucidated a breath print of 12 VOCs distinguishing liver cirrhosis from healthy controls and 7 VOCs distinguishing compensated from decompensated liver disease.64 On using a panel of 385 volatile organic compounds Khalid and colleagues identified hepatic encephalopathy with a 333
S. Kurada et al. sensitivity of 91% in alcoholic cirrhosis compared to healthy controls. In patients with no demonstrable hepatic encephalopathy, the aetiology of cirrhosis was established as being secondary to alcohol in 78.3% of patients and non-alcoholic causes in 69.2% of patients. A distinction between non-alcoholic cirrhosis and alcoholic cirrhosis from healthy controls was made with a sensitivity and specificity of 92.3% and 97.1% respectively. Harmful drinking patterns were identified with 100% sensitivity when compared to healthy controls.65
THE BREATH METABOLOME IN IBD In contrast to other diseases, to date, only 12 mainly small studies evaluated the breath metabolome or condensate in IBD for diagnosis, differentiation or disease activity assessment (Table 1 summarises studies on diagnosis and differentiation of IBD). Most of these reports only provide mean values per group as comparisons that do not allow assessment of the area under the ROC curve (AUC), sensitivity, specificity, positive predictive value or negative predictive value. Diagnosis and differentiation Single VOCs tested for the diagnosis and differentiation of IBD include pentane, ethane, propane, butane or nitric oxide (NO) and multiple compounds have been shown to be elevated in IBD. The best-explored gas is pentane. Dryahina et al. have demonstrated elevated levels of pentane in IBD (CD>UC) compared to healthy controls with an area under the ROC curve (AUC) of 0.927 (Figure 3; adapted from original study with permission, P < 10 7).66 Given the presented data, a pentane cut-off of 84 ppbv provided an estimated sensitivity of 0.55 and a specificity of 1. Breath pentane in CD patients was higher compared to UC with, however, an only moderate discriminatory capacity (P = 0.0058, AUC of 0.68). Even though the primary aim of this investigation was the quantification of breath pentane levels, this study failed to consider possible confounders like drugs, diet, and other disease processes in baseline demographics that may affect the results. While Pelli et al.67 demonstrated mean pentane levels to be elevated in IBD compared to controls (P < 0.001; only mean values provided), confirming the above results, Sedghi et al. failed to do so.68 An association between ethane and IBD compared to controls was demonstrated in two investigations (P < 0.001 and P = 0.013)67, 68 and between breath propane and IBD in one study (P < 0.001),67 but only mean comparisons were provided. Exhaled NO was significantly higher in patients with UC and CD compared to 334
HC (P = 0.05–0.01) with no difference being noted between UC and CD.69 Our group has recently completed a prospective cross-sectional, single-centre study that evaluated a panel of 21 VOCs in the exhaled breath of paediatric IBD patients (n = 62) and healthy controls (n = 55) using SIFT-MS. After adjusting for age, gender, race and BMI, routinely analysed VOCs showed significant differences in 6 VOCs between the two groups (Increased levels in IBD compared to controls for: 1octene, 3-methylhexane, 1-decene; Decreased levels in IBD compared to controls for: 1-nonene, (E)-2-nonene, hydrogen sulphide; P = 0.006 to
