Urinary Metabolic Profiles of Inflammatory Bowel Disease in

Jun 18, 2008 - Interleukin-10 (IL-10) gene-deficient mice were raised alongside age-matched 129/SvEv controls in conventional housing. Urine samples (...
0 downloads 6 Views 897KB Size
Anal. Chem. 2008, 80, 5524–5531

Urinary Metabolic Profiles of Inflammatory Bowel Disease in Interleukin-10 Gene-Deficient Mice Travis B. Murdoch,† Hao Fu,‡ Sarah MacFarlane,† Beate C. Sydora,† Richard N. Fedorak,† and Carolyn M. Slupsky*,‡ Division of Gastroenterology and Magnetic Resonance Diagnostic Centre, Department of Medicine, University of Alberta, Edmonton, Alberta, Canada Inflammatory bowel disease (IBD) is a chronic debilitating disorder that is thought to have both genetic and environmental contributors. Commensal microflora have been shown to play a key part in the disease process. Metabolomics, the study of large numbers of small molecule metabolites, has demonstrated that disease and/or changes in gut microbial composition modulate mammalian urine metabolite fingerprints. The aim of this project was to associate the development of IBD with specific changes in a mouse urinary metabolic fingerprint. Interleukin-10 (IL-10) gene-deficient mice were raised alongside agematched 129/SvEv controls in conventional housing. Urine samples (22 h) were collected at ages 4, 6, 8, 12, 16, and 20 weeks. Metabolite concentrations were derived from analysis of nuclear magnetic resonance spectra, and both multivariate and two-way analysis of variance (ANOVA) statistical techniques were applied to the resulting data. Principal component analysis and partial leastsquares-discriminant analysis of urine derived from the control and IL-10 gene-deficient mice revealed that while both groups initially had similar metabolic profiles, they diverged substantially with the onset of IBD as assessed through external phenotypic changes. Several metabolites, including trimethylamine (TMA) and fucose, changed dramatically in the IL-10 gene-deficient mice following 8 weeks of age, concomitant with the known timeline for development of severe histological injury. This study illustrates that metabolomics is effective at distinguishing IBD using urinary metabolite profiles. Inflammatory bowel disease (IBD), which comprises Crohn’s disease (CD) and ulcerative colitis (UC), is a chronic inflammatory disorder of the intestine. In UC, inflammation is limited to the colonic mucosa, while in CD inflammation is transmural and anywhere in the gastrointestinal tract, although predominantly in the ileum and colon.1 IBD is thought to have both genetic and environmental contributors.2–4 While there have been recent * To whom correspondence should be addressed. Carolyn M. Slupsky, Ph.D., Department of Medicine, MRDC, 550A Heritage Medical Research Building, University of Alberta, Edmonton, Alberta, Canada T6G 2S2. Phone: (780) 4928919. Fax: (780) 492-5329. E-mail: [email protected]. † Division of Gastroenterology. ‡ Magnetic Resonance Diagnostic Centre. (1) Podolsky, D. K. N. Engl. J. Med. 2002, 347, 417–429. (2) Bonen, D. K.; Cho, J. H. Gastroenterology 2003, 124, 521–536.

5524

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

advances in the treatment of IBD,5,6 its etiology remains under investigation. Commensal enteric bacteria have been indicated in the antigenic stimulation that drives progression of IBD.7 Patients with IBD have been shown to have alterations in their gut microflora as a part of the pathophysiology.8–12 Lower levels of Clostridia, in particular those involved in short chain fatty acid production, have been found in both UC and CD.12 Furthermore, rodent models have established that disease does not develop in an axenic environment.13–15 While genomics and proteomics have proven themselves invaluable in providing vast amounts of data regarding the expression of genes and proteins, metabolomics provides data regarding all metabolic processes of a cell or organism. Metabolite profiles can vary as a result of changes in the environment and reflect perturbations of the system’s genome and proteome. These perturbations may be observed originating from the organism itself, or from the superorganism,16–19 reflecting changes in gut microbiota that could be the cause of many diseases including metabolic syndrome20–22 and obesity.22 Indeed, the importance (3) Katz, J. A.; Itoh, J.; Fiocchi, C. Curr. Opin. Gastroenterol. 1999, 15, 291– 297. (4) Young, Y.; Abreu, M. T. Curr. Gastroenterol. Rep. 2006, 8, 470–477. (5) Ardizzone, S.; Bianchi Porro, G. Drugs 2005, 65, 2253–2286. (6) Sandborn, W. J. Inflammatory Bowel Dis. 2001, 7 (Suppl. 1), S9–16. (7) Mahida, Y. R.; Rolfe, V. E. Clin. Sci. 2004, 107, 331–341. (8) Bibiloni, R.; Mangold, M.; Madsen, K. L.; Fedorak, R. N.; Tannock, G. W. J. Med. Microbiol. 2006, 55, 1141–1149. (9) Kleessen, B.; Kroesen, A. J.; Buhr, H. J.; Blaut, M. Scand. J. Gastroenterol. 2002, 37, 1034–1041. (10) Rutgeerts, P.; Goboes, K.; Peeters, M.; Hiele, M.; Penninckx, F.; Aerts, R.; Kerremans, R.; Vantrappen, G. Lancet 1991, 338, 771–774. (11) Seksik, P.; Rigottier-Gois, L.; Gramet, G.; Sutren, M.; Pochart, P.; Marteau, P.; Jian, R.; Dore´, J. Gut 2003, 52, 237–242. (12) Sokol, H.; Seksik, P.; Rigottier-Gois, L.; Lay, C.; Lepage, P.; Podglajen, I.; Marteau, P.; Dore´, J. Inflammatory Bowel Dis. 2006, 12, 106–111. (13) Aranda, R.; Sydora, B. C.; McAllister, P. L.; Binder, S. W.; Yang, H. Y.; Targan, S. R.; Kronenberg, M. J. Immunol. 1997, 158, 3464–3473. (14) Onderdonk, A. B.; Franklin, M. L.; Cisneros, R. L. Infect. Immun. 1981, 32, 225–231. (15) Sydora, B. C.; Tavernini, M. M.; Wessler, A.; Jewell, L. D.; Fedorak, R. N. Inflammatory Bowel Dis. 2003, 9, 87–97. (16) Gill, S. R.; Pop, M.; DeBoy, R. T.; Eckburg, P. B.; Turnbaugh, P. J.; Samuel, B. S.; Gordon, J. I.; Relman, D. A.; Fraser-Liggett, C. M.; Nelson, K. E. Science 2006, 312, 1355–1359. (17) Goodacre, R. J. Nutr. 2007, 137, 259S–266S (18) Lederberg, J. Science 2000, 288, 287–293. (19) Sekirov, I.; Finlay, B. B. Nat. Med. 2006, 12, 736–737. (20) Ley, R. E.; Turnbaugh, P. J.; Klein, S.; Gordon, J. I. Nature 2006, 444, 1022–1023. (21) Nicholson, J. K.; Holmes, E.; Wilson, I. D. Nat. Rev. Microbiol. 2005, 3, 431–438. 10.1021/ac8005236 CCC: $40.75  2008 American Chemical Society Published on Web 06/18/2008

of microbial involvement in the mammalian metabolome by production and coprocessing of metabolites is becoming increasingly clear.21,23 Recent technological advances in the measurement of metabolite profiles coupled with the application of multivariate statistical techniques have enabled the fast analysis of metabolic differences such as those giving rise to gender, diurnal, diet, age, or disease phenotypes.24–31 In the current study, we report a comparison of urinary metabolite profiles from wild-type and interleukin-10 (IL-10) genedeficient mice over a 16 week period. IL-10 gene-deficient mice are genetically susceptible to develop enterocolitis when colonized with enteric mouse microbiota.32 The aim of this investigation was to correlate disease phenotype with specific changes in metabolites and identify those metabolites that signal the onset of the disease. METHODS Mice. IL-10 gene-deficient mice, generated on a 129/SvEv background, and wild-type 129/SvEv mice were maintained in a colony at the animal facility of the University of Alberta under conventional conditions and fed a Laboratory Rodent Diet 5001 (PMI Nutrition International). At 4 weeks of age, mice were weaned and placed in separate cages. In total, four male IL-10 gene-deficient mice from the same litter and three male 129/SvEv wild-type mice from the same litter were used in the experiments. At weeks 4, 6, 8, 12, 16, and 20, each mouse was placed into a separate sterilized metabolic cage and urine samples were collected over a 22 h period. The mice were allowed to consume water during this period but had no access to food. At age 20 weeks, mice were sacrificed by overdose of flurothane and were weighed and examined for phenotypic changes. This study was reviewed and approved by the local Animal Policy and Welfare Committee and carried out in accordance with guidelines of the Canadian Council on Animal Care. For histology experiments, groups of mice were sacrificed at 8, 10, 13, and 15 weeks of age. Colon and cecum were excised, cut longitudinally, embedded in paraffin in toto, sectioned at 4 µm, and stained with H&E for light microscopic analysis. Slides (22) Turnbaugh, P. J.; Ley, R. E.; Mahowald, M. A.; Magrini, V.; Mardis, E. R.; Gordon, J. I. Nature 2006, 444, 1027–1031. (23) Turnbaugh, P. J.; Ley, R. E.; Hamady, M.; Fraser-Liggett, C. M.; Knight, R.; Gordon, J. I. Nature 2007, 449, 804–810. (24) ′t Hart, B. A.; Vogels, J. T.; Spijksma, G.; Brok, H. P.; Polman, C.; van der Greef, J. J. Neurol. Sci. 2003, 212, 21–30. (25) Brindle, J. T.; Antti, H.; Holmes, E.; Tranter, G.; Nicholson, J. K.; Bethell, H. W.; Clarke, S.; Schofield, P. M.; McKilligin, E.; Mosedale, D. E.; Grainger, D. J. Nat. Med. 2002, 8, 1439–1444. (26) Gibney, M. J.; Walsh, M.; Brennan, L.; Roche, H. M.; German, B.; van Ommen, B. Am. J. Clin. Nutr. 2005, 82, 497–503. (27) Marchesi, J. R.; Holmes, E.; Khan, F.; Kochhar, S.; Scanlan, P.; Shanahan, F.; Wilson, I. D.; Wang, Y. J. Proteome Res. 2007, 6, 546–551. (28) Slupsky, C. M.; Rankin, K. N.; Wagner, J.; Fu, H.; Chang, D.; Weljie, A. M.; Saude, E. J.; Lix, B.; Adamko, D. J.; Shah, S.; Greiner, R.; Sykes, B. D.; Marrie, T. J. Anal. Chem. 2007, 79, 6995–7004. (29) Solanky, K. S.; Bailey, N. J.; Holmes, E.; Lindon, J. C.; Davis, A. L.; Mulder, T. P.; Van Duynhoven, J. P.; Nicholson, J. K. J. Agric. Food Chem. 2003, 51, 4139–4145. (30) Wang, Y.; Holmes, E.; Nicholson, J. K.; Cloarec, O.; Chollet, J.; Tanner, M.; Singer, B. H.; Utzinger, J. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 12676–12681. (31) Wang, Y.; Tang, H.; Nicholson, J. K.; Hylands, P. J.; Sampson, J.; Holmes, E. J. Agric. Food Chem. 2005, 53, 191–196. (32) Ku ¨ hn, R.; Lo ¨hler, J.; Rennick, D.; Rajewsky, K.; Mu ¨ ller, W. Cell 1993, 75, 263–274.

were reviewed in a blinded fashion and were assigned a histological score ranging from 0, representing no injury, to 10, representing maximal injury, where the numerical sum of four scoring criteria (mucosal ulceration, epithelial hyperplasia, lamina propria mononuclear infiltration, and lamina propria neutrophilic infiltration) was determined as described previously.15 Urine Analysis. Urine samples were collected by rinsing the collection funnel with 2 mL of sterile water containing 0.005% sodium azide. To prevent evaporation of the urine over the 22 h collection time, the collection chambers were lined with paraffin oil. Samples were frozen at -80 °C prior to analysis. For nuclear magnetic resonance (NMR) analysis, urine samples were passed through Nanosep Omega 3Kspin filters (Pall Corporation) to remove debris/protein. A volume of 65 µL of 5 mM DSS-d6 (sodium 2,2-dimethyl-2-silapentane-5-sulfonate-d6) + 0.2% sodium azide in 100% D2O (Chenomx internal standard, Chenomx Inc., Edmonton, Canada) was added to 585 µL of filtered urine, and pH was adjusted to 6.8 ± 0.1. A 600 µL aliquot was placed in a 5 mm NMR tube (Wilmad, Buena, NJ) and stored at 4 °C until acquisition. One-dimensional NMR spectra of urine samples were acquired using the first increment of the standard nuclear Overhauser enhancement spectroscopy (NOESY) pulse sequence on a fourchannel Varian INOVA 600 MHz NMR spectrometer equipped with a triax-gradient 5 mm HCN probe as previously described.28,33 Quantification of metabolites was achieved using Chenomx NMRSuite 4.6 (Chenomx Inc. Edmonton, Canada).28,33 For each urine sample, 50 metabolites were assigned and quantified. The NMR variables (metabolites) derived from spectral analysis were log10 transformed, mean centered, and unit variance scaling was applied. Principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) was conducted using SIMCA P11.0 (Umetrics, Umeå, Sweden). For each metabolite, age and mouse (wild-type versus IL-10 gene-deficient) data were analyzed by two-way analysis of variance (ANOVA) over the course of 4-20 weeks. Corrections to the level of significance were made using Bonferroni’s method. Two-way ANOVA was conducted using the Prism 4.0c for MacIntosh software package (GraphPad Software Inc., San Diego, CA). RESULTS Figure 1A shows the partial least-squares-discriminant analysis of all data derived from measured urinary metabolite concentrations of wild-type and IL-10 gene-deficient mice over the span of 16 weeks. PLS-DA is a regression extension of principal components analysis, an unsupervised multivariate statistical analysis method that is essentially a dimensionality reduction technique taking multidimensional data (for example 7 mice × 50 metabolites) and reducing it into coherent subsets that are independent of one another (for example 7 mice × 2 principal components). The primary use of PCA is to reduce the number of variables (in our case metabolites) and identify those variables that are interrelated. PLS-DA uses class information in an attempt to maximize the separation between the groups of observations. For the purposes of this experiment, both sets of mice were classified according to age (4-20 weeks), and both PCA and PLS-DA were (33) Weljie, A. M.; Newton, J.; Mercier, P.; Carlson, E.; Slupsky, C. M. Anal. Chem. 2006, 78, 4430–4442.

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

5525

Figure 1. Partial least squares-discriminant analysis (PLS-DA) of wild type versus IL-10 gene-deficient mice over all time points. For clarity, each week of the plot shown in part A is plotted separately as indicated: (B) week 4, (0, wild-type; 9, IL-10 gene-deficient); (C) week 6, (O, wild-type; b, IL-10 gene-deficient); (D) week 8, (4, wild-type; 2, IL-10 gene-deficient); (E) week 12, (], wild-type; [, IL-10 gene-deficient); (F) week 16, (3, wild-type; 1, IL-10 gene-deficient); (G) week 20, (/, wild-type; +, IL-10 gene-deficient). (H) Validation of PLS-DA model (using 20 random permutations); R2 (2), Q2 (9).

performed. Each component with related variables was investigated to see if differences between observations and classes could be delineated (for example, if differences exist between IL-10 genedeficient mice and wild-type mice based on age and disease). Further details on the use of multivariate statistical analysis have been previously published.34 Figure 1 shows the PLS-DA plot of the data from the IL-10 gene-deficient mice and their wild-type controls from 4 to 20 weeks. Mice were grouped according to age and genetics. For clarity, the original plot (Figure 1A) was redrawn to include data points from individual weeks so that the differences between the IL-10 gene-deficient mice and the wild-type mice could be more easily observed (Figure 1B-G). Analysis using only PCA resulted in almost an identical plot (data not shown). Validation of the model (Figure 1H) revealed an R2 of 0.82 and a Q2 of 0.60 with a positive slope between the randomly permuted and original classified data indicating a robust model. In our data (Figure 1A), the largest difference, reflected in the first principal component, is related to age and could partly be a reflection of the mice being weaned from their mothers at 4 weeks of age. The second largest (34) Trygg, J.; Holmes, E.; Lundstedt, T. J. Proteome Res. 2007, 6, 469–479.

5526

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

difference, reflected in the second principal component, is associated with the development of IBD in the IL-10 gene-deficient mice, which begin to deviate from the wild-type mice after 8 weeks of age. This change parallels the onset of severe histological injury in the intestines of IL-10 gene-deficient mice35–37 which were raised in an identical environment. Disease onset for these mice was found to be between 8 and 10 weeks of age, plateauing at 12 weeks of age (Figure 2). The PLS-DA plots similarly stabilized by 12 weeks of age (Figure 1E-G). The loadings plot (Figure 3) provides an indication of which variables give rise to the observed differences. The further away a variable (metabolite) is from the origin, the greater the influence. The direction (positive or negative) reflects whether the variables get larger or smaller in the respective groups situated in the scores plot. For example, trimethylamine (TMA) which increases dramatically in the IL-10 gene-deficient mice with respect to the wild(35) Kennedy, R. J.; Hoper, M.; Deodhar, K.; Erwin, P. J.; Kirk, S. J.; Gardiner, K. R. Br. J. Surg. 2000, 87, 1346–1351. (36) Madsen, K. L.; Doyle, J. S.; Tavernini, M. M.; Jewell, L. D.; Rennie, R. P.; Fedorak, R. N. Gastroenterology 2000, 118, 1094–1105. (37) Madsen, K. L.; Malfair, D.; Gray, D.; Doyle, J. S.; Jewell, L. D.; Fedorak, R. N. Inflammatory Bowel Dis. 1999, 5, 262–270.

Figure 2. Injury scores of colon (gray) and cecum (white) tissue indicating development of disease in IL-10 gene-deficient mice. Histological grades range from 0 (no disease) to 10 (maximal injury). Values are means ( SEM for n ) 4 mice per age group.

Figure 3. Loadings plot corresponding to PLS-DA scores plot in Figure 1.

type mice after 8 weeks, also increases with age, and is thus present in the upper left-hand quadrant of the loadings plot. N-Isovaleroylglycine which increases with age, decreases in the IL-10 gene-deficient mice in the later time points (after 12 weeks), and is present in the lower left-hand quadrant. Differences in metabolites related to age were seen as overall lower levels of creatine, taurine, and trimethylamine (TMA) at 4 weeks (Figure 1B) compared to 6 weeks and above (Figure 1C-G). Figure 4 shows a comparison of urinary metabolite concentrations as both wild-type and IL-10 gene-deficient mice age. The most significant differences between the two groups of mice involve increases in the concentrations of TMA and fucose for the IL-10 gene-deficient mice starting between 8 and 12 weeks of age (Figure 4). A comparison of TMA and fucose concentrations as reflected in the NMR spectra derived from the urine of IL-10 gene-deficient and wild-type mice at 4 weeks and 20 weeks is shown in Figure 5. While TMA is represented in an NMR spectrum by a single peak situated at approximately 2.89 ppm, fucose is considerably more complex. In solution, the NMR spectrum of fucose is composed of four anomers: R- and β-fucofuranose (representing 3% of the population) and R- and β-fucopyranose. The methyl group of both R- and β-fucopyranose has the highest intensity of all peaks of fucose, with each one resonating as a doublet at approximately 1.20 ppm (6.61 Hz J-coupling) and

1.24 ppm (6.51 Hz J-coupling), respectively. Other protons, such as the anomeric protons, have chemical shifts close to water and may be easily identified by the doublet at approximately 4.55 ppm (7.91 Hz J-coupling), corresponding to β-fucopyranose, and the doublet centered just below 5.20 ppm (3.93 Hz J-coupling) corresponding to R-fucopyranose (Figure 5). The concentration of fucose in the urine of the IL-10 gene deficient mice at 20 weeks was so high that the anomeric protons of R- and β- fucofuranose were also observed (at 5.225 and 5.275 ppm, respectively). Other protons within fucose were also easily observable between 3.43-3.46, 3.62-3.65, 3.73-3.82, 3.83-3.89, and 4.16-4.21 ppm (not shown) and could be reliably modeled using Chenomx NMRSuite software. Trimethylamine-N-oxide (TMAO), dimethylamine (DMA), xylose, and valine followed similar changes, but the differences between the IL-10 gene-deficient mice and the wild-type mice were not as large (data not shown). Uracil appeared more concentrated in the IL-10 gene-deficient mice at all time points as did creatine, acetate, and taurine (data not shown) and did not appear to correlate with any observed phenotypic changes. N-isovaleroylglycine was similar between IL-10 gene-deficient and wild-type mice until 12 weeks whereupon its concentration significantly declined in the IL-10 gene-deficient mice. Similarly, decreases after 12 weeks in the TCA cycle intermediates 2-oxoglutarate, citrate, fumarate, and succinate in the IL-10 gene-deficient mice relative to the wild-type mice were also observed. No significant differences in metabolite concentrations could be determined for creatinine or hippurate and only a slight increase in the later timepoints could be observed for phenylacetylglycine. Since many metabolites changed in concentration with age, two-way ANOVA was used to determine whether differences between the wild-type and IL-10 gene-deficient mice were also correlated with age and onset of disease. Those metabolites that showed significant differences using two-way ANOVA (with p < 0.01) with onset of IBD at approximately 12 weeks and older are fucose (p < 0.0001); TMA (p < 0.0001); DMA (p < 0.0001); fumarate (p < 0.0001); N-isovaleroylglycine (p < 0.0001); 2-oxoglutarate (p ) 0.0008); TMAO (p ) 0.0011); butyrate (p ) 0.0025); citrate (p ) 0.0027); succinate (p ) 0.0035); 3-indoxylsulfate (p ) 0.0036); valine (p ) 0.0058); and phenylacetylglycine (p ) 0.0087). On sacrifice, mice were examined for phenotypic changes: hair loss, cachexia, conjunctival inflammation, and rectal prolapse were among the features found in the IL-10 gene-deficient mice, in addition to severe colonic thickening. Mouse weight was significantly decreased in IL-10 gene-deficient mice (20.0 ± 0.6 g) versus control mice (22.8 ± 0.7 g) (p ) 0.01 using Student’s t test). DISCUSSION This study demonstrates that analysis of urinary metabolite concentrations can identify specific urinary metabolites that change with intestinal disease progression in an animal model of IBD. IL-10 gene-deficient mice raised in an SPF (specific pathogen free) environment spontaneously develop colonic inflammation.38 In our colony, intestinal histologic disease starts to develop after 8 weeks of age, increasing to a maximum at 12-15 weeks. To investigate whether onset and progression of IBD may be monitored by changes in urinary metabolites, we analyzed and compared metabolite data from IL-10 gene-deficient mice starting immediately after weaning at 4 weeks of age, with age and gender Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

5527

Figure 4. Comparison of metabolite concentrations between IL-10 gene-deficient mice and wild-type mice from 4 weeks of age to 20 weeks of age as measured by 1H NMR spectroscopy. White ) wild-type; gray ) IL-10 gene-deficient mice. Significant differences by two-way ANOVA (Bonferroni corrected) are indicated as: *, p < 0.05; **, p < 0.01; ***, p < 0.001. 5528

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

Figure 5. Comparison of NMR spectral characteristics of metabolites measured in the urine of an IL-10 gene-deficient mouse and a wild-type mouse at 4 weeks and 20 weeks. The solid black line represents the actual spectrum, and the dashed gray line indicates the spectral fit. Shown are (A) peak arising from trimethylamine (TMA) in an IL-10 gene-deficient mouse. For comparison purposes, peaks of creatinine, creatine, 2-oxoglutarate, and N,N-dimethylglycine are also indicated, and spectra are scaled to show these at similar concentration. Shifting of the peaks is due to slight variations in pH between the samples; (B) peak arising from TMA in a wild-type mouse; (C) peaks arising from fucose in an IL-10 gene-deficient mouse (the doublet at 1.24 ppm corresponds to the methyl group of β-L-fucopyranose whereas the doublet at 1.20 ppm corresponds to the methyl group of R-L-fucopyranose). Doublets at 4.545 and 5.195 ppm correspond to the anomeric proton of β-L-fucopyranose and R-Lfucopyranose, respectively. Doublets indicated by * and ** correspond to β-L-fucofuranose and R-L-fucofuranose, respectively; (D) peaks arising from fucose in a wild-type mouse.

matched wild-type 129/SvEv mice using both multivariate and twoway ANOVA statistical methods. The urinary metabolic fingerprint of IL-10 gene-deficient and wild-type mice was similar at 4 weeks, before development of IBD in the IL-10 gene-deficient mouse model (Figure 1). A major shift at 6 and 8 weeks that was similar in both groups likely represents changes in diet (weaning) and gut microflora. However, after week 8, the IL-10 gene-deficient mice showed major differences relative to the metabolite profiles of wild-type mice, the latter developing (38) Sydora, B. C.; Martin, S. M.; Lupicki, M.; Dieleman, L. A.; Doyle, J.; Walker, J. W.; Fedorak, R. N. Inflammatory Bowel Dis. 2006, 12, 429–436.

a stable metabolic fingerprint soon after weaning to the final time point of 20 weeks. While some of the differences between the groups could be seen at 4 weeks (such as higher levels of uracil in the IL-10 genedeficient mice throughout the experiment), most differences were observed to start after 8 weeks of age. Of particular interest was the identification of metabolites produced and or modulated by gut microflora as significant contributors to the differences between groups. Changes in the composition of gut microflora have been demonstrated in IBD patients.8–12 In the IL-10 genedeficient mouse, IBD only occurs if the organism is enterically Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

5529

colonized.38,39 One of the identified metabolites, trimethylamine (TMA), is significant in that it is synthesized exogenously by microflora in the mouse.40,41 TMA changes have also been associated with intestinal infection.30 Of interest, the increase in TMA in the IL-10 gene-deficient mouse urine over time parallels the progression of IBD. IL-10 gene-deficient mice also displayed higher urinary levels of TMAO, taurine, creatine, glucose, and acetate in addition to decreased levels of tricarboxylic acid (TCA) cycle intermediates, which have been shown to be associated with hepatotoxicity.42,43 Moreover, changes in TCA cycle intermediates can also be due to renal or neurological disease or even general stress such as reduced caloric intake.42–46 However, decreased caloric intake can be eliminated as an explanation for the differences in TCA cycle intermediates as these metabolites generally increase with decreasing caloric intake. Similar concentrations of creatinine in the urine between groups would indicate that alterations in renal function were not responsible for the metabolite differences. In addition, the IL-10 gene-deficient mouse model is not generally associated with neurological disease. Hepatic injury may thus be one explanation for at least some of the urinary metabolite perturbations found in the IL-10 genedeficient mice. Indeed, other colitis models have been shown to be associated with hepatotoxicity.47 Recent studies have demonstrated that IL-10 protects against hepatic steatosis and other forms of hepatic injury by inhibiting local production of proinflammatory cytokines such as TNF-R.48–50 den Boer et al. described increased hepatic triglyceride content in IL-10 gene-deficient mice fed a high fat diet,49 although the initial description of IL-10 gene-deficient mice fed a conventional diet demonstrated normal liver histology.32 Increased urinary methylamines, including TMA, have been shown to be associated with hepatic steatosis in insulin-resistant (39) Sellon, R. K.; Tonkonogy, S.; Schultz, M.; Dieleman, L. A.; Grenther, W.; Balish, E.; Rennick, D. M.; Sartor, R. B. Infect. Immun. 1998, 66, 5224– 5231. (40) al-Waiz, M.; Mikov, M.; Mitchell, S. C.; Smith, R. L. Metab. Clin. Exp. 1992, 41, 135–136. (41) Smith, J. L.; Wishnok, J. S.; Deen, W. M. Toxicol. Appl. Pharmacol. 1994, 125, 296–308. (42) Beckwith-Hall, B. M.; Nicholson, J. K.; Nicholls, A. W.; Foxall, P. J.; Lindon, J. C.; Connor, S. C.; Abdi, M.; Connelly, J.; Holmes, E. Chem. Res. Toxicol. 1998, 11, 260–272. (43) Waters, N. J.; Holmes, E.; Williams, A.; Waterfield, C. J.; Farrant, R. D.; Nicholson, J. K. Chem. Res. Toxicol. 2001, 14, 1401–1412. (44) Connor, S. C.; Wu, W.; Sweatman, B. C.; Manini, J.; Haselden, J. N.; Crowther, D. J.; Waterfield, C. J. Biomarkers 2004, 9, 156–179. (45) Holmes, E.; Nicholls, A. W.; Lindon, J. C.; Ramos, S.; Spraul, M.; Neidig, P.; Connor, S. C.; Connelly, J.; Damment, S. J.; Haselden, J.; Nicholson, J. K. NMR Biomed. 1998, 11, 235–244. (46) Salek, R. M.; Maguire, M. L.; Bentley, E.; Rubtsov, D. V.; Hough, T.; Cheeseman, M.; Nunez, D.; Sweatman, B. C.; Haselden, J. N.; Cox, R. D.; Connor, S. C.; Griffin, J. L. Physiol. Genomics 2007, 29, 99–108. (47) Chen, C.; Shah, Y. M.; Morimura, K.; Krausz, K. W.; Miyazaki, M.; Richardson, T. A.; Morgan, E. T.; Ntambi, J. M.; Idle, J. R.; Gonzalez, F. J. Cell Metab. 2008, 7, 135–147. (48) Cintra, D. E.; Pauli, J. R.; Arau´jo, E. P.; Moraes, J. C.; de Souza, C. T.; Milanski, M.; Morari, J.; Gambero, A.; Saad, M. J.; Velloso, L. A. J. Hepatol. 2008, 48, 628–637. (49) den Boer, M. A.; Voshol, P. J.; Schro ¨der-van der Elst, J. P.; Korsheninnikova, E.; Ouwens, D. M.; Kuipers, F.; Havekes, L. M.; Romijn, J. A. Endocrinology 2006, 147, 4553–4558. (50) Louis, H.; Le Moine, O.; Peny, M. O.; Quertinmont, E.; Fokan, D.; Goldman, M.; Devie`re, J. Hepatology 1997, 25, 1382–1389.

5530

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

mice.51 As well, it was shown that mice susceptible to obesity under high-fat diet conditions had higher levels of plasma lactate, pyruvate, glucose, fucose, phosphatidylcholine, TMAO, and other methylamines which may be related to hyperglycemia, and hyperlipidemia.52 In our study, we found fucose demonstrated a similarly marked trajectory to TMA in IBD mice, while remaining stable in wild-type mice. Of interest, it has been suggested that changes in TMA, caused by gut microflora, signal perturbations in choline metabolism which contribute to a fatty-liver phenotype by reducing choline bioavailability.51,53 Furthermore, fucosyltransferase-8 mRNA increases in the liver of diabetic mice,54 as well as patients with chronic hepatitis, liver cirrhosis, and diabetes.54,55 Therefore, the IL-10 gene-deficient mice may have increased urinary fucose due to a fatty-liver phenotype brought on from action of the gut microbes. However, it has also been shown that in IBD less fucose is incorporated into red blood cell membranes, with an increased serum concentration associated with glycoproteins.56 Furthermore, fucosylation has been shown to be important in epithelialmicrobe interactions;57 R(1, 2)-fucosylated glycans are abundant in the intestinal epithelium of adult mammals.58 Using axenic mice, Bry et al. have demonstrated that R(1,2)-fucosylated glycan and R(1,2)-fucosyltransferase mRNA production require the presence of normal microflora.59 In addition, they identified a component of the conventional microflora, Bacterioides thetaiotaomicron, that could modulate fucosylation in the distal intestine.59,60 Another explanation for the increased fucose found in the IL-10 genedeficient mouse urine could therefore be related to changes in the gut microflora signaling an increase in the de novo production of fucose for production of fucosylated glycans, resulting in increased urinary fucose. Indeed, it has been shown that germfree mice are resistant to obesity and that the gut microbiome in part contributes to the pathophysiology of obesity and is important for regulating energy balance.61 Taken together, these data may indicate a novel mechanism associating IL-10 deficiency with hepatic injury. However, clearly more research needs to be done to understand the underlying (51) Dumas, M. E.; Barton, R. H.; Toye, A.; Cloarec, O.; Blancher, C.; Rothwell, A.; Fearnside, J.; Tatoud, R.; Blanc, V.; Lindon, J. C.; Mitchell, S. C.; Holmes, E.; McCarthy, M. I.; Scott, J.; Gauguier, D.; Nicholson, J. K. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 12511–12516. (52) Toye, A. A.; Dumas, M. E.; Blancher, C.; Rothwell, A. R.; Fearnside, J. F.; Wilder, S. P.; Bihoreau, M. T.; Cloarec, O.; Azzouzi, I.; Young, S.; Barton, R. H.; Holmes, E.; McCarthy, M. I.; Tatoud, R.; Nicholson, J. K.; Scott, J.; Gauguier, D. Diabetologia 2007, 50, 1867–1879. (53) Buchman, A. L.; Dubin, M. D.; Moukarzel, A. A.; Jenden, D. J.; Roch, M.; Rice, K. M.; Gornbein, J.; Ament, M. E. Hepatology 1995, 22, 1399–1403. (54) Itoh, N.; Sakaue, S.; Nakagawa, H.; Kurogochi, M.; Ohira, H.; Deguchi, K.; Nishimura, S.-I.; Nishimura, M. Am. J. Physiol. Endocrinol. Metab. 2007, 293, E1069–E1077. (55) Noda, K.; Miyoshi, E.; Uozumi, N.; Yanagidani, S.; Ikeda, Y.; Gao, C.; Suzuki, K.; Yoshihara, H.; Yoshikawa, M.; Kawano, K.; Hayashi, N.; Hori, M.; Taniguchi, N. Hepatology 1998, 28, 944–952. (56) Kurup, R. K.; Kurup, P. A. Int. J. Neurosci. 2003, 113, 1221–1240. (57) Becker, D. J.; Lowe, J. B. Glycobiology 2003, 13, 41R–53R (58) Torres-Pinedo, R.; Mahmood, A. Biochem. Biophys. Res. Commun. 1984, 125, 546–553. (59) Bry, L.; Falk, P. G.; Midtvedt, T.; Gordon, J. I. Science 1996, 273, 1380– 1383. (60) Hooper, L. V.; Xu, J.; Falk, P. G.; Midtvedt, T.; Gordon, J. I. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 9833–9838. (61) Backhed, F.; Manchester, J. K.; Semenkovich, C. F.; Gordon, J. I. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 979–984.

mechanisms and the involvement of gut microbes in the urinary metabolite changes seen in this mouse model of IBD. It is difficult to make definitive claims regarding the exact significance of urine metabolite changes. Epithelial injury has been well documented in IBD1,62 and may contribute to the observed profile. Intestinal inflammation is frequently associated with a defect in intestinal integrity. Changes in intestinal permeability may also contribute to the observed changes in urinary metabolites.35 While the current study demonstrates the potential for metabolomic analysis of urine to demonstrate differences between IBD and healthy mice, it is limited by its small sample size by use of a single model. Studying other models may help clarify the contribution of IBD and gut microbe changes. A further limitation of this study is its small sample size. However, this is somewhat mitigated by the use of serial mouse urine sampling. This study is the first to demonstrate urinary metabolic differences in an IBD mouse model as the disease develops over time. The application of metabolomics to identify onset and progression of disease in mice demonstrates the potential utility of this technique in humans, although analysis of human urine may be more complex. Certainly, if these results are translatable to human urine, urinary metabolite profiling could provide a faster, (62) McKenzie, S. J.; Baker, M. S.; Buffinton, G. D.; Doe, W. F. J. Clin. Invest. 1996, 98, 136–141.

less-invasive, and cost-effective technique for diagnosis of IBD. Moreover, obtaining urine samples from patient populations on a regular basis will likely be met with greater compliance than obtaining fecal samples, which have been shown to differentiate between CD and UC in humans.27 Better understanding of the mechanisms responsible for disease pathogenesis, as well as development of less invasive diagnostic and disease-monitoring techniques, will improve the quality of life of the many individuals that suffer from IBD. ACKNOWLEDGMENT The authors are indebted to Dr. Kathryn Rankin and Shana Regush for help with sample preparation and NMR data acquisition. The authors acknowledge the Canadian National High Field NMR Centre (NANUC) for use of the facilities for collection of the NMR data. Funding for this study was from a University Hospital Foundation grant. T.B.M. is the recipient of a summer studentship from AHFMR and Crohn’s and Colitis Foundation of Canada/Canadian Association of Gastroenterology.

Received for review March 11, 2008. Accepted May 14, 2008. AC8005236

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

5531