![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
* TNO Netherlands Organization for Applied Scientific Research, Nutrition and Food Research and Pharma, 3700 AJ Zeist, The Netherlands and
Wageningen University, Department for Agrotechnology and Food Science, 6700 EV Wageningen, The Netherlands
1To whom correspondence should be addressed. E-mail: r.lamers{at}voeding.tno.nl.
 |
ABSTRACT |
|---|
 TOP ABSTRACT MATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
KEY WORDS: • metabolic fingerprinting • 1H NMR spectroscopy • in vitro • inulin • Caco-2 cells
Fructans are polymers of fructose. Inulin and oligofructose belong to this class of carbohydrates. Inulin is found in many plants and vegetables (e.g., chicory and Jerusalem artichoke). Chicory is by far the most common source used by industry to obtain inulin as a commercial product. As an ingredient in foods, inulin can replace fat and sugar, improve mouth feel and texture and act as a dietary fiber. Inulin is counted as a prebiotic because it is not susceptible to digestion and is hydrolyzed by endogenous enzymes. When it reaches the colon, it is fermented by the microbiota and selectively stimulates the growth of bifidobacteria (1). A predominance of bifidobacteria in the large intestine is supposed to be beneficial for maintaining good health (1,2). Fermentation products of inulin are SCFA (acetate, propionate and butyrate), lactate and gasses (2). The functional effects of inulin on humans and experimental animals include relief of constipation, lower blood glucose levels, improved absorption of calcium, reduced fasting triglycerides and LDL cholesterol, and inhibition of the growth of various kinds of tumors (3).
The underlying metabolism that causes the effects of inulin remains indistinct and not yet fully understood (3). This is a more common problem in nutritional research, where there is a shortage of knowledge of the relationships between health and disease and effects of nutrition on the latter. Fortunately, in the field of metabolite research, great progress has been made recently due to metabolic fingerprinting (4–7). This technique utilizes 1H NMR in combination with multivariate data analysis (MVDA)1 to analyze biological fluids.
NMR provides concurrent detection of all hydrogen-containing molecules in a sample without pretreatment. NMR can thus reveal chemical structures of metabolites in biological fluids and subsequently clarify metabolic pathways involved in nutrition and health (8). Nevertheless, interpretation of NMR spectra obtained from biological fluids is very complicated due to the enormous amount of spectral signals produced.
MVDA is a powerful technique for the analysis of data sets with a large number of variables. For this reason, MVDA is particularly useful in finding significant spectral changes in NMR spectra, i.e., it enables the visualization of spectral patterns in NMR data, and thus metabolites, that correlate with a treatment or disease, for example (4–6).
In MVDA, unsupervised and supervised techniques can be used. Unsupervised methods such as principal component analysis (PCA) search for similarities and differences in data sets without foreknowledge. A large set of related variables (e.g., NMR signals) is converted to a smaller set of uncorrelated variables, which express maximum variation in the original variables. The new variables are called principal components (PC) and each of them depicts an axis in multidimensional space. The distance of an object (e.g., a complete NMR spectrum of a sample) to a PC is called a score. Scores are plotted in a score plot, with the PC as axes. When scores are situated close to each other in a score plot, this suggests that the NMR spectra of the samples are similar. When the clustering of scores matches the samples that were controls, treated or diseased in the original study set-up, a connection can be made between affected NMR signals, and thus metabolites, and treatment or disease.
Calculation of the contribution of each original variable (e.g., a single NMR signal) to a PC yields a loading. When a loading is high, the original NMR signal adds greatly to the clustering of scores in the direction of the investigated PC. In a so-called factor spectrum or metabolic fingerprint, loadings are presented as lines. The location of the lines in a factor spectrum corresponds to the location of the variables in the original NMR spectra. The length of a line denotes the contribution of a variable to the grouping of scores in the investigated direction (4). Thus, a high line in a positive direction indicates an NMR signal that is strongly ascending for a particular group of scores.
Supervised methods such as partial least squares and principal component discriminant analysis (PCDA) exploit supplemental information on the data set (e.g., biochemical, histopathological or clinical data) to identify and maximize similarities and differences between predefined groups (4–6). In PCDA, the scores from PCA are used as a basis for linear discriminant analysis, i.e., discriminant analysis combines the PC in such a way that differences between predefined groups are optimized (9).
Up to now, metabolic fingerprinting has been used mainly in combination with in vivo studies. These studies are time-consuming, labor-intensive and, for these and other reasons, expensive compared with in vitro studies. When metabolic fingerprinting can successfully be applied to in vitro studies, there will be a practical alternative to in vivo–based metabolic fingerprinting that will be useful in nutritional studies. In vitro metabolic fingerprinting may represent a relatively inexpensive and quick way to fill the gap in the lack of evidence for effects of functional foods on health, for example.
In the underlying research, a pilot in vitro experiment with metabolic fingerprinting was carried out. The suitability of in vitro metabolic fingerprinting was assessed by investigation of direct and indirect effects (after fermentation with the colonic microbiota) of inulin on Caco-2 cells (10).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Caco-2 cells (designation HTB 37) were obtained from the American Type Culture Collection (ATCC, Manassas, VA). All chemicals were obtained from Gibco (Breda, the Netherlands). For 500 mL culture medium, 440 mL DMEM (cat. no. 42430) was used supplemented with 50 mL heat-inactivated fetal calf serum, 5 mL nonessential amino acids (10 mmol/L), 5 mL L-glutamine (200 mmol/L) and 0.5 mL gentamicin (50 g/L). Cell cultures were grown in this medium and maintained at 37°C in 95% air/5% CO2 (v/v; Sanyo incubator, Sakato, Japan). Near confluent Caco-2 cell cultures were harvested by trypsinization with 3 mL trypsin solution (25 g/L) in 147 mL PBS and were resuspended in 10 mL culture medium diluted 5 times.
For the experiment, cells were seeded in Transwell inserts in 12-well plates (1 mL of cell suspension with 1.5 mL DMEM per well). The medium was changed every 2–3 d. Cells became confluent after 
4 d, at which time differentiation could begin. After complete differentiation, samples (wells) were fed with various media. Four samples were treated with 1.5 mL DMEM for 0 h and four samples for 48 h. A 10X dilution of a saturated solution of 1.5 g Frutafit EXL (Sensus, Roosendaal, The Netherlands) and 30 mL DMEM was used to treat four samples with 1.5 mL for 0 h, whereas four samples were treated with the same solution for 48 h. In addition, four samples were treated with 1.5 mL metabolized inulin [run in TNO’s dynamic gastrointestinal model (TIM)-2 feeding, 10 times diluted with DMEM] for 0 h and four samples for 48 h. Another four samples were treated with 1.5 mL TIM-2 feeding after passing the TIM-2 model (10X diluted with DMEM) for 0 h and four samples for 48 h. The 0 h samples were collected directly after the start of the exposure.
The TNO in-vitro model of the large intestine (TIM-2) simulates the physiologic parameters in the large intestine (or colon), such as pH, temperature and an active microbiota similar in composition and activity to that in the human colon (11,12). Fermentation in the proximal colon was mimicked in this in vitro model. The test compound in question (i.e., inulin) was added to the control TIM-2 medium. This mixture was added to the TIM-2 system, giving rise to metabolized inulin. The temperature was kept at 37°C and the pH at 5.8. The model was flushed with gaseous nitrogen to allow growth of an active anaerobic, complex microbiota of human origin. The model was inoculated with microbiota from human fecal material. Inulin was administered at 10 g/d in doses of 104 mg/15 min. The contents were mixed by peristaltic movements. Microbial metabolites were removed from the model by a dialysis system running through the model. This prevented inhibition of the activity of the microbiota by accumulation of microbial metabolites. For more details on the in vitro model, see Minekus et al. (11) and Venema et al. (12).
All dilutions were centrifuged at 2500 x g for 10 min (Megafuge 2.0 RS, Heraeus, Germany). The dilutions with control TIM-2 medium and metabolized inulin were passed through a 0.2-µm filter before exposure to the cultured cells.
Solutions with the respective test compounds were removed at 0 or 48 h, depending on the time of exposure. Cells were then washed with PBS (37°C) and dissolved in 1 mL of a solution of methanol (Sigma-Aldrich, Zwijndrecht, The Netherlands) in demineralized water (7.5 mol/L). Samples were sonicated for 10 s at 20 µm to lyse the cells, using a MSE Ultrasonic disintegrator sonifier (Beun-de Ronde BV, Amsterdam, The Netherlands). Then they were centrifuged (Eppendorf, Germany) at 13,000 x g for 5 min, yielding samples of cell contents that were stored at -40°C until NMR analysis.
NMR analysis of in vitro medium and cell samples.
Before NMR spectroscopic analysis, the culture medium was removed from the cell samples. Cells were dried under a stream of nitrogen gas. The samples were dissolved in 1 mL of sodium phosphate buffer (0.1 mol/L, pH 6.0, made up with D2O). Sodium trimethylsilyl-[2,2,3,3,4–2H4]-1-propionate (TMSP; 0.05 mmol/L) was added as an internal standard.
NMR measurements were carried out in random order and in triplicate in a fully automated manner on a 600 MHz spectrometer (Avance, Bruker BioSpin GmbH, Rheinstetten, Germany), using a proton NMR set-up operating at a temperature of 300K. For each sample, 256 free induction decays (FID) were collected. Each FID was induced using a 45° pulse, an acquisition time of 4.10 s and a relaxation delay of 2 s. The FID were collected as 64K data points with a spectral width of 12,000 Hz. The spectra were processed using the standard Bruker software. An exponential window function with a line broadening of 0.5 Hz and a manual baseline correction were applied to all spectra. After referring to the internal NMR reference [TMSP (
= 0.0)], line listings were prepared with the standard Bruker NMR software. To obtain these listings, all lines in the spectra above a threshold corresponding to 3 times the signal-to-noise ratio were collected and converted to a data file suitable for multivariate data analysis applications.
NMR data preprocessing and multivariate data analysis.
The NMR data reduction file was imported into Winlin (V1.11, TNO, The Netherlands). Minor variations from comparable signals in different NMR spectra were adjusted and lines were fitted without loss of resolution (13). To correct for sample dilution, the data were autoscaled so that small and large signals contributed similarly to the final study result. PCDA was performed, with treatment and time as additional information. For each PCDA analysis, the respective data set was randomly divided into a training data set and a test data set. PCDA models were built upon the training data set. The resulting discriminants were quantified for each of the NMR spectra and the first discriminant (D1) was plotted vs. the second discriminant (D2) to visualize clustering. The test data set was used to test the accuracy of the PCDA model by passing it through this model to obtain the model’s prediction of classification of the test data into the clusters from the PCDA training model. Predictions were within the 95% CI.
Factor spectra were used then to correlate the position of clusters in the score plot with the original NMR signals in the spectra. The metabolic fingerprints were prepared in directions of maximum separation of one cluster vs. another cluster, to provide insight into the type of metabolites responsible for the separation between clusters (4). Metabolites were assigned from the metabolic fingerprints using an in-house database with NMR spectra.
|
RESULTS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
The effect of inulin on cell contents.
If there was an effect of inulin on cell contents, this information would be contained in the NMR spectra of these cells. PCDA was used to visualize differences between NMR spectra obtained from contents of cells exposed only to DMEM vs. the contents of cells exposed to DMEM with inulin, at two points of time (Fig. 1A). The first (D1) and second (D2) discriminants explained 36 and 33% of the variance, respectively. It is clear that the contents of Caco-2 cells treated with inulin in DMEM were positioned in about the same location in the score plot at 0 h as the contents of cells treated only with DMEM. However, after 48 h, differences were clearly visible between the contents of cells exposed to DMEM with and without inulin. When the metabolic fingerprint was examined (Fig. 1B), there were differences were due to changes in regions 1–4.5 ppm and 7–8.5 ppm. Metabolites that could be assigned to these signals using the in-house database with NMR spectra were leucine ( 0.96, 1.71), isoleucine ( 0.94, 1.01), valine ( 0.99, 1.04), alanine ( 1.48, 3.79), 
- and ß-glucose ( 3.47, 3.49, 3.53, 3.71, 3.72, 3.74, 3.84, 3.9, 4.64, 5.24), phenylalanine ( 7.33, 7.38, 7.43), tyrosine ( 6.91, 7.2) and glutamate ( 2.1, 2.35, 3.77).
|
When PCDA was performed on the NMR spectra of contents of Caco-2 cells treated with fermented inulin vs. the contents of cells treated with the accompanying control medium, the first (D1) and second (D2) discriminants explained 41 and 35% of the variance, respectively. The score plot revealed that both groups had the same location at 0 h (Fig. 2A). However, after 48 h of exposure, a difference was observed between the contents of Caco-2 cells exposed to only control medium and the contents of cells exposed to medium with fermented inulin. This was also reflected in the metabolic fingerprint at 48 h (Fig. 2B).From the fingerprint, we see that differences in the regions around 1.5, 2, 3, 4 and 8 ppm contributed heavily to this effect. According to the NMR database, metabolites that belong to these signals include lactate ( 1.33, 4.12), alanine ( 1.48, 3.78), proline ( 2.01, 2.07, 4.14), succinate, 2-oxoglutarate ( 2.44, 3), nicotinate and nicotinamide ( 7.97, 8.20, 8.28, 8.52).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Differences were clearly revealed between the contents of cells exposed to DMEM with and without inulin after 48 h. At 0 h, this difference was not visible. Hence, their NMR spectra, and thus metabolite levels in the cells, were significantly different. This indicates that inulin had an effect on the contents of Caco-2 cells over time. The score plot of NMR spectra from contents of Caco-2 cells treated with fermented inulin vs. the contents of cells treated with the accompanying control medium indicated that a difference could be observed between the two groups at 48 h.
Metabolic fingerprints were identified that reflected effects of inulin as well as its metabolites on Caco-2 cell contents. The height of a line in a metabolic fingerprint reflects the importance of an NMR signal to the clusters investigated in a score plot. Metabolic fingerprints thus provided information about metabolites, which were elevated or lowered in one cluster compared with another cluster.
The effects of inulin itself on cell contents have not yet been investigated widely because it is presumed that it is nearly completely fermented in the gut. However, effects may occur in the small intestine, where inulin is neither digested nor fermented. According to the metabolic fingerprint derived from our study, inulin itself seemed to influence metabolism in Caco-2 cells. The amount of glucose, together with glutamate contents, increased after 48 h due to treatment with inulin compared with the controls. This could indicate activated gluconeogenesis in the Caco-2 cells under the conditions of our experiments. Glucose metabolism is altered by glutamine via the citrate cycle (14). However, the liver and kidney are considered to be the only organs capable of gluconeogenesis even though recent hypotheses concerning glutamine and glucose metabolism suggested that the release of glucose might also occur in the small intestine after fasting (15). Because Caco-2 cells develop some properties of the small intestinal epithelium during differentiation, this hypothesis merits further investigation using isotopes to measure rates of gluconeogenesis.
Another explanation for the rise in the level of glucose due to inulin treatment could be that some inulin, or a breakdown product, was taken up by the cells. However, it is unlikely that inulin passed through the membrane into the cell by facilitated diffusion because it is too large a molecule unless the process of pinocytosis was involved. If this were the case, fructose would be converted to glucose in the cell. The amount of glucose would then increase, thus becoming available for glycolysis to form phosphoenol pyruvate. This compound is the basis for the production of tyrosine via the shikimate pathway. A raised level of this compound in the metabolic fingerprint supports the idea that this biochemical route was affected. Furthermore, phosphoenol pyruvate can be converted to pyruvate. Subsequently, pyruvate is oxidized to acetyl-CoA, which enters the citrate cycle. When oxidation of pyruvate is incomplete, alanine and lactate may be formed. These compounds were elevated in the metabolic fingerprint. Glutamate production (which was rising in the metabolic fingerprint) occurs during the citrate cycle by transamination between 2-oxoglutarate and amino acids to be catabolized (16).
According to the metabolic fingerprint, the levels of phenylalanine, valine, leucine and isoleucine were elevated due to inulin compared with the controls. These compounds are essential amino acids. It therefore seems unlikely that Caco-2 cells were able to synthesize these compounds, although protein synthesis and degradation may have been altered due to inulin.
After exposure to fermented inulin, metabolites related to the citrate cycle, such as alanine, lactate, succinate, 2-oxoglutarate were most prominent. In addition, the amount of nicotinate and nicotinamide increased and glutamate production was enhanced, reflected in an elevated level of proline. The presence of these metabolites indicates that glycolysis in Caco-2 cells seemed to be stimulated by fermented inulin compared with the controls.
The metabolites mentioned in this article were readily identified from the metabolic fingerprints by experience and the use of an in-house database with NMR spectra. Metabolic fingerprinting thus provides an efficient tool with which to initiate hypotheses about affected metabolic pathways. However, definitive evidence will await confirmatory studies using techniques such as liquid chromatography-mass spectrometry and 2-dimensional NMR. In addition, from the metabolic fingerprints presented here, it becomes clear that not all signals can be identified using an in-house database. For elucidating structures of metabolites that are more difficult to identify from metabolic fingerprints (e.g., the heavily contributing signals around 3.25 in Fig. 2B), other techniques will be required in the future.
Nevertheless, for global screening, in vitro metabolic fingerprinting seems a promising technology. In a realistic nutritional research study with Caco-2 cells, biochemical changes in cells, resulting from exposure, were detected. Metabolic fingerprinting might in principle even be able to measure the excretion of metabolites from cells into the culture medium, thus further elucidating cell metabolism.
In vitro metabolic fingerprinting studies provide an inexpensive starting point for the formulation of hypotheses about affected metabolic pathways and may even replace costly in vivo metabolic fingerprinting studies. It will be a great challenge to develop more in vitro models to combine with metabolic fingerprinting. The results of these studies should be compared with similar in vivo studies. In the future, in vitro based metabolic fingerprinting may function as an alternative for in vivo based fingerprinting in specific occasions. This could greatly enhance and facilitate evidence-based nutritional studies.
|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Manuscript received 10 June 2003. Initial review completed 11 July 2003. Revision accepted 25 July 2003.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
1. Roberfroid, M. B., Bornet, F., Bouley, C. H. & Cummings, J. H. (1995) Colonic microflora: nutrition and health. Nutr. Rev. 53:127-130.[Medline]
2. Gibson, G. R., Beatty, E. R., Wang, X. & Cummings, J. H. (1995) Selective stimulation of bifidobacteria in the human colon by oligofructose and inulin. Gastroenterology 108:975-982.[Medline]
3. Kaur, N. & Gupta, A. K. (2002) Applications of inulin and oligofructose in health and nutrition. J. Biosci. 27:703-714.[Medline]
4. Lamers, R.A.N., DeGroot, J., Spies-Faber, E. J., Jellema, R. H., Kraus, V. B., Verzijl, N., TeKoppele, J. M., Spijksma, G., Vogels, J.T.W.E., van der Greef, J. & van Nesselrooij, J.H.J. (2003) Identification of disease and nutrient related metabolic fingerprints in osteoarthritic guinea pigs. J. Nutr. 133:1776-1780.
5. Nicholson, J. K., Lindon, J. C. & Holmes, E. (1999) "Metabonomics": understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica 29:1181-1189.[Medline]
6. Lindon, J. C., Nicholson, J. K., Holmes, E. & Everett, J. R. (2000) Metabonomics: metabolic processes studies by NMR spectroscopy of biofluids combined with pattern recognition. Concepts Magn. Reson. 12:289-320.
7. Fiehn, O. (2002) Metabolomics: the link between genotypes and phenotypes. Plant Mol. Biol. 48:155-171.[Medline]
8. Van der Greef, J., Davidov, E., Verheij, E., Vogels, J.T.W.E., van der Heijden, R., Adourian, A. S., Oresic, M., Marple, E. W. & Naylor, S. (2002) The role of metabolomics in drug discovery: a new vision for drug discovery and development. Harrigan, G. G. Goodacre, R. eds. Metabolic Profiling: Its Role in Biomarker Discovery and Gene Function Analysis 2002:170-198 Kluwer Academic Publishing Boston MA. .
9. Hoogerbrugge, R., Willig, S. J. & Kistemaker, P. G. (1983) Discriminant analysis by double stage principal component analysis. Anal. Chem. 55:1710-1712.
10. Fusunyan, R. D, Quinn, J. J., Fujimoto, M., MacDermott, R. P. & Sanderson, I. R. (1999) Butyrate switches the pattern of chemokine secretion by intestinal epithelial cells through histone acetylation. Mol. Med. 5:631-640.[Medline]
11. Minekus, M., Smeets-Peters, M., Bernalier, A., Marol-Bonnin, S., Havenaar, R., Marteau, P., Alric, M., Fonty, G. & Huisin’t Veld, J.H.J. (1999) A computer-controlled system to simulate conditions of the large intestine with peristaltic mixing, water absorption and absorption of fermentation products. Appl. Microbiol. Biotechnol. 53:108-114.[Medline]
12. Venema, K., Van Nuenen, M., Smeets-Peters, M., Minekus, M. & Havenaar, R. (2000) TNO’s in-vitro large intestinal model: an excellent screening tool for functional food and pharmaceutical research. Nutrition 24:558-564.
13. Vogels, J.T.W.E., Tas, A. C., Venekamp, J. & van der Greef, J. (1996) Partial linear fit: a new NMR spectroscopy preprocessing tool for pattern recognition applications. J. Chemom. 10:425-438.
14. Reeds, P. J. & Burrin, D. G. (2001) Glutamine and the bowel. J. Nutr. 131:2505S-2508S.
15. Mithieux, G. (2001) New data and concepts on glutamine and glucose metabolism in the gut. Curr. Opin. Clin. Nutr. Metab. Care 4:267-271.[Medline]
16. Michal, G. (1999) Biochemical Pathways, an Atlas of Biochemistry and Molecular Biology 1999 Spektrum Akademischer Verlag. Heidelberg, Germany.
This article has been cited by other articles:
|
|
 |
|
  V. Shulaev Metabolomics technology and bioinformatics Brief Bioinform, June 1, 2006; 7(2): 128 - 139. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
