PAPER 142 Ital. J. Food Sci., vol. 27 - 2015 - Keywords: Metabolomics, NMR, wholegrain, rye, wheat, metabolites, biomarkers - METABOLOMICS STUDY OF CEREAL GRAINS REVEALS THE DISCRIMINATIVE METABOLIC MARKERS ASSOCIATED WITH ANATOMICAL COMPARTMENTS M. COULOMB1, A. GOMBERT1,2 and A.A. MOAZZAMI1,3* 1Department of Chemistry and Biotechnology, Swedish University of Agricultural Sciences, P.O. Box 7015, SE 75007, Uppsala, Sweden 2Department of Food Science, Swedish University of Agricultural Sciences, P.O. Box 7051, SE 75007, Uppsala, Sweden 3NMR-metabolomics core facilities, Uppsala BioCentrum, Swedish University of Agricultural Sciences, Uppsala, Sweden *Corresponding author: Tel. +46 18672048, Fax +46 18672995, email: ali.moazzami@slu.se ABSTRACT This study used NMR-based metabolomics to compare the metabolic profile of different ana- tomical compartments of cereal grains i.e. bran and endosperm in order to gain further insights into their possible role in the beneficial health effects of whole grain products (WG). Polar water- soluble metabolites in 64 bran and endosperm, samples from rye and wheat were observed using 600 MHz NMR. Bran samples had higher contents of 12 metabolites than endosperm samples. A comparative approach revealed higher contents of azelaic acid and sebacic acid in bran than in endosperm. In a pilot study, the consumption of WG rye bread (485 g) caused NMR signals in 24h urine corresponding to azelaic acid. The relatively high abundance, anatomical specificity, pat- tern of metabolism, urinary excretion in human, antibacterial, and anticancer activities suggest further studying of azelaic acid when exposure to WG or beneficial effects of WG are investigated. mailto:ali.moazzami%40slu.se?subject= Ital. J. Food Sci., vol. 27 - 2015 143 INTRODUCTION Epidemiological studies have consistent- ly shown that intake of whole grain (WG) can protect against the development of chronic dis- eases (SLAVIN et al. 2001), e.g. type 2 diabetes (T2D) (DE MUNTER et al. 2007, MURTAUGH et al. 2003), cardiovascular disease (CVD) (FLINT et al. 2009; JACOBS et al. 2007; MELLEN et al. 2008), and certain cancers (CHAN et al. 2007, HAAS et al. 2009; LARSSON et al. 2005; SCHATZ- KIN et al. 2008). The American Association of Cereal Chemists provided the following scien- tific and botanical definition of WG in 1999: “whole grain shall consist of the intact, ground, cracked or flaked caryopsis, whose principal anatomical component-the starchy endosperm, germ and bran-are present in the same rela- tive proportion as they exist in the intact cary- opsis” (International 1999). Whole grains are a rich source of fiber and bioactive compounds, including tocopherols, B vitamins, minerals, phenolic acids, and phytoestrogens (FARDET, 2010). It is generally recognized that the syn- ergistic action of compounds mainly present in the bran and germ fractions of cereals accounts for the protective effects of WG products (FAR- DET, 2010; LIU, 2007). Recently, the composi- tion and the diversity of bioactive compounds in different anatomical components of cereal grains have been systematically investigated in a large number of different species and varie- ties within the HEALTHGRAIN project (NYSTROM et al. 2008; SHEWRY et al. 2010; WARD et al. 2008). However, that project screened the cereal samples for bioactive compounds already doc- umented in cereals using a targeted approach, and made no comparison of the untagged pro- file of the metabolites in different compartments of cereal grains. Metabolomics is an untargeted approach in which the profile of metabolites in a biospeci- men is measured using high-throughput analyt- ical methods, e.g. NMR and mass spectrometry (LENZ and WILSON, 2007; NICHOLSON and WIL- SON 2003). We have used this approach previ- ously to examine the complex physiological/bio- chemical effects of WG rye products in humans (MOAZZAMI et al. 2012; MOAZZAMI et al. 2014; MOAZZAMI et al. 2011). The aim of the present study was to search for the discriminative me- tabolites in the two major anatomical compart- ments in cereal grain, endosperm and bran, us- ing an untargeted NMR-based metabolomics ap- proach and with the emphasis on wheat and rye to gain further insights into their possible role in the beneficial health effects of whole grain products. NMR analysis can potentially pro- vide characteristic structural data, which can be used for elucidation and eventual identifica- tion of unknown compounds found to discrimi- nate between the metabolic profiles of bran and endosperm in cereals. MATERIALS AND METHODS Serial sample collection and extraction A total of 64 cereal samples, comprising 18 wheat endosperm, 24 wheat bran, 8 rye en- dosperm, and 14 rye bran were obtained from the HEALTHGRAIN (WARD et al. 2008) project or from a local market. The endosperm and bran samples originated from HEALTHGRAIN projects were from the same grain sample material and therefore were matched (Wheat samples n = 18; and rye samples n = 8). The HEALTHGRAIN project rye varieties (and populations) included potugaise-3, potugaise-6, Haute Loire, Gran- drieu, Nikita, Rekrut, Dankowskie-Zlote, and Lovaszpatonai-1. The details about rye va- rieties are given in NYSTROM et al. (2008). The HEALTHGRAIN project wheat varieties in- cluded Disponent, Herzog, Tommi, Campari, Tremie, San Pastore, Gloria, Spartanka, Ava- lon, Claire, Malacca, Maris Huntsman, Rial- to, Riband, Obriy, CF99105, Chinese-Spring, and Cadenza. The details about wheat varie- ties are given by SHEWRY et al. (2010). All rye and wheat varieties were grown in the field at Martonvasar, Hungary, in 2005. Full details of the site including soil type, mineral compo- sition, and weather condition has been given by SHEWRY et al. (2010). All samples were milled, and 0.5 g milled ma- terial was extracted in 5 mL Milli-Q water for 18 h. The samples were then centrifuged (5 min- 1500 g), and 2 mL supernatant was extracted, mixed with 8 mL ethanol and centrifuged (15 min-1,500 g) in order to precipitate the soluble viscose polymers. A 5 mL portion of the ethanol supernatant was dried using an evacuated cen- trifuge (Savant, SVC 100H, Savant Instrument INC, NJ) and dissolved in phosphate buffer (280 µL, 0.25 mol/L, pH 7.0), D2O (40 µL), and sodi- um-3-(trimethylsilyl)-2,2,3,3-tetradeuteriopropi- onate solution (TSP, 30 µL, 23.2 mmol/L) (Cam- bridge Isotope Laboratories, Andover, MA). The mixture was then used for 1H NMR analysis. An internal standard was added to the mixture in order to ensure semi-quantitative measure- ments of metabolites captured by 1H NMR. For 2D NMR analysis the mixture was freeze-dried and dissolved in D2O before analysis. Human experiment and the preparation of urine sample for NMR analysis In a pilot study, a male subject (age 35; BMI = 23.4) consumed refined wheat bread 485 g for 6 days (breakfast 2 portions, lunch 1 por- tion, dinner 1 portion). On day six, 24-hour urine was collected. On day seven, he substi- tuted the 485 g refined wheat bread with 485 g of whole grain rye bread and the urine was collected for 24 hours. During the seven days of experiment, any other cereal products were avoided. The choice of consuming refined wheat bread vs whole grain rye bread was made to 144 Ital. J. Food Sci., vol. 27 - 2015 replicate the condition of previous human in- terventions in which refined wheat bread was used as the control diet (MOAZZAMI et al. 2011; MOAZZAMI et al. 2012; BONDIA-PONS et al. 2013). The refined wheat bread was prepared from commercial refined wheat flour and whole grain rye bread was prepared from commercial whole grain rye flour. The whole trial was re- peated twice, in two different times. This study complied with the Helsinki Declaration, as re- vised in 1983. The urine samples were kept in -80°C freezers before analysis. The urine sam- ples (500 µL) were mixed with phosphate buffer (250 µL, 0.25 M, pH 7.0) containing 5 mmol/L sodium-3-(trimethylsilyl)-2,2,3,3-tetradeuteri- opropionate (TSP) (Cambridge Isotope Labora- tories, Andover, MA) as an internal standard. Resulting solutions were centrifuged to remove particulate matter. The supernatant was then transferred into 5-mm NMR tubes for 1H NMR analysis. For 2D-NMR analysis, 600 µL of the supernatant was freeze-dried and dissolved in 600 µL D 2 O before 2D-NMR analysis. NMR measurements and the identification of signals The 1H NMR analyses (cereal extracts and human urine) were performed on a Bruker spectrometer operating at 600 MHz (Karlsruhe, Germany). 1H NMR spectra were obtained us- ing zgesgp pulse sequence (Bruker Spectro- spin Ltd.) at 25°C with 128 scans and 65,536 data points over a spectral width of 17942.58 Hz. Acquisition time was 1.82 s and relaxation delay was 4.0 s. The NMR signals which were found discriminating between different anatom- ical compartments were identified primarily us- ing the NMR Suite 7.1 library (ChenomX Inc, Edmonton, Canada), Human Metabolome Data Base and Biological Magnetic Resonance Data Bank. In the event of multiplicity, the identi- ty was confirmed with 2D NMR. In human ex- periment, the identity of phytochemical in the urine originating from the cereals in the diet was also confirmed using 2D-NMR. Phase- sensitive TOCSY and COSY with presaturation (2k × 512 experiments) were performed with 32 scans and a spectral width of 7195 Hz for both F1 and F2. The mixing time for TOCSY was 80 ms. HSQC was performed using 32 scans and a spectral width of 7211 Hz and 250002 Hz for proton and carbon, respectively. All cereal ex- tracts and urine samples were reconstituted in D2O before 2D NMR analysis. The 1H NMR spectra data (cereal extracts) were processed using Bruker Topspin 1.3 soft- ware and were Fourier-transformed after multi- plication by a line broadening of 0.3 Hz and ref- erenced to TSP at 0.0 ppm. Spectral phase and baseline were corrected manually. Each spec- trum was integrated using Amix 3.7.3 (Bruker BioSpin GmbH, Rheinstetten) into 0.01 ppm in- tegral regions (buckets) between 0.5-10 ppm, in which area between 4.60-5.18 ppm containing residual water was removed. Each spectral re- gion was then normalized to the intensity of in- ternal standard (TSP). Statistical analysis Principal component analysis (PCA) and or- thogonal partial least squares-discriminant analysis (OPLS-DA) were performed using SIM- CA-P+ 12.0.1 software (UMETRICS, Umeå, Swe- den) after centering and pareto-scaling of the data as previously described (MOAZZAMI et al. 2011). The presence of outliers was investigat- ed using PCA-Hotelling T2 Ellipse (95% CI) and the normality of multivariate data was investi- gated using the normal probability plot of the PCA model. Variable influences on projection (VIP) values of the OPLS-DA model were used to determine the most important discrimina- tive NMR bucket (signals). NMR buckets (sig- nals) with VIP > 1 for which the corresponding jack-knife-based confidence intervals were not close to or including zero were considered dis- criminative. The significance of OPLS-DA mod- el was tested using cross-validated ANOVA (CV- ANOVA), which assesses the reliability of OPLS models (CV-ANOVA p<0.05 means the OPLS- DA model is reliable) (ERIKSSON et al. 2008). The absolute concentrations of metabolites with corresponding NMR signals that were found to be discriminative in OPLS-DA were calculat- ed from the NMR spectra using the NMR Suite 7.1 profiler (ChenomX Inc, Edmonton, Canada) and internal standard after correction for over- lapping signals. The absolute concentrations of the discriminative metabolites were further in- vestigated using ANOVA in the case of normal distribution, and the Mann-Whitney test when the distribution was skewed (Anderson-Darling test, p<0.05). RESULTS AND DISCUSSION PCA model was fitted using NMR spectral data (buckets) obtained for the bran and en- dosperm extracts. Three outliers were identi- fied and excluded from the data set based on PCA-Hotelling T2 Ellipse (95% CI). The first and the second component explained 67.4% and 18.1% of spectral variation (R2X) respec- tively (figure not shown). An OPLS-DA mod- el was fitted including three predictive and six orthogonal components. The first, second, and third predictive components explained 61%, 15.2%, and 1.0% of spectral variation respectively (model parameter: R2Y=0.937; Q2Y=0.876; Cross-validated ANOVA p-value = 2.98 × 10-38) (Fig. 1). The first component in each model basically separated the bran samples obtained from rye Ital. J. Food Sci., vol. 27 - 2015 145 Fig. 1 - OPLS-DA separated different anatomical compartments of cereal grains based on their profile of metabolites measured using NMR. The model parameters were as follow: R2Y=0.937; Q2Y=0.876. Cross- validated ANOVA p-value = 2.98 × 10- 38. Score t[1] (component 1) and score t[2] (component 2) are new variables summarizing the X-variables (the intensity of NMR signals corresponding to metabolites). Table 1 - Discriminative metabolites along the first predictive component of the OPLS-DA model (n = 64)1,2. Metabolite NMR signal (ppm)3 VIP (Confidence interval)4 Azelaic acid & sebacic acid5 1.304 ; 1.530 ; 2.179 2.6 (0.67) ; 1.6 (0.41) ; 1.6 (0.40) Acetate 1.928 1.3 (0.26) Alanine 1.494 1.1 (0.08) Betaine 3.269 6.9 (0.59) Choline 3.206; 4.085 4.4 (0.46); 2.3 (0.24) Citrate 2.545 1.4 (0.13) Isoleucine 0.942 1.0 (0.18) Leucine 0.964 1.4 (0.24) Malate 2.663 1.1 (0.31) Maltose 3.280; 3.727; 3.997 1.4 (0.24); 4.4 (0.49) Succinate 2.412 1.3 (0.16) Unknown signals6 3.778; 4.054; 4.155; 4.020 2.3 (1.07); 2.3 (0.18); 2.2 (0.23); 4.6 (0.32) 1OPLS-DA Score scatter plot: the first component separated the bran samples (right) from the endosperm samples (left). All metabolites present in higher con- centrations in bran. The model parameters for three predictive component fitted were as follow: R2Y=0.937; Q2Y=0.876. Cross-validated ANOVA p-value = 2.98 × 10-38; 2Wheat endosperm (n = 18), wheat bran (n = 24), rye endosperm (n = 8), and rye bran (n = 14); 3One NMR signal from the corresponding spectral bucket with the highest VIP values was reported when several buckets covered a distinct NMR signal; 4NMR signals with VIP > 1 for which the corresponding jack-knife-based confidence intervals were not close to or including zero were considered discriminative; 5Concentration equivalent of azelaic acid; 6Unknow signals are located in sugar region. and wheat from the endosperm samples (Fig. 1, Table 1). The second component separated rye samples (both endosperm and bran sam- ples) from wheat samples (Fig. 1; Table 2). Bran samples contained a higher content of 12 me- tabolites and four unknown signals than en- dosperm samples (Table 1), and their contents contributed to composing the first predictive component of the OPLS-DA model. The content of eight metabolites and five unknown signals changed along the second predictive compo- nent of the OPLS-DA model (Table 2) separat- ing wheat samples i.e. both anatomical com- partments from rye samples. The concentra- tions of all eight metabolites were found higher in wheat compared with rye. The metabolic sig- nature of wheat and rye samples acquired from the local market did not deviate from those ac- quired from HEALTHGRAIN project as all sample tightly accumulated in their corresponding spe- cies-compartment cluster (Fig. 1). The absolute concentrations of metabolites that were found to differ between different anatomical compart- ments and species were calculated from NMR spectra and further investigated using ANOVA or the Mann-Whitney test. A total of 12 metabo- lites were found to differ between different spe- cies and different anatomical compartments in the same species, e.g. bran compared vs en- dosperm (Table 3; Fig. 2). 146 Ital. J. Food Sci., vol. 27 - 2015 Table 2 - Discriminative metabolites along the second predictive component of the OPLS-DA model (n = 64)1,2. Metabolite Loading3 NMR signal (ppm)4 VIP (Confidence interval)5 Azelaic acid & sebacate6 - 1.304; 1.545; 2.179 3.3 (0.34); 2.1 (0.25); 1.9 (0.20) Betaine - 3.269; 3.904 5.2 (0.84); 3.1 (0.39) Choline - 3.206; 4.085 3.5 (0.42); 1.9 (0.15) Citrate - 2.545 1.0 (0.18) Leucine - 0.964 1.1 (0.16) Maltose - 3.278; 3.727; 5.257 6.1 (0.92); 3.2 (0.49); 1.0 (0.78) Succinate - 2.414 1.1 (0.23) Unknown signals7 + 3.778; 4.054; 4.155; 3.915; 4.043 3.7 (0.75); 1.7 (0.25); 1.6 (0.21); 2.4 (0.46); 2.7 (0.32) 1OPLS-DA Score scatter plot: the second component separated the wheat samples (below) from the rye samples (above). The model parameters for three pre- dictive component fitted were as follow: R2Y=0.937; Q2Y=0.876. Cross-validated ANOVA p-value = 2.98 × 10-38; 2Wheat endosperm (n = 18), wheat bran (n = 24), rye endosperm (n = 8), and rye bran (n = 14); 3Loadings: (+): higher concentration in rye samples. (-): higher concentration in wheat samples; 4One NMR signal from the corresponding spectral bucket with the highest VIP values was reported when several buckets covered a distinct NMR signal; 5NMR signals with VIP > 1 for which the corresponding jack-knife-based confidence intervals were not close to or including zero were considered discriminative; 6Concentration equivalent of azelaic acid; 7Unknow signals are located in sugar region. Table 3 - Absolute concentrations of metabolites (µmol/g) found to be discriminative along the first and second predictive components1. Concentration µmol/g (mean ± SD) Metabolite 1 : Rye endosperm 2 : Rye bran 3 : Wheat endosperm 4 : Wheat bran Azelaic acid & sebacic acid 0.68 ± 0.21a 1.70 ± 0.27a 0.70 ± 0.16a 4.32 ± 1.25b Acetate 1.19 ± 0.26a 3.98 ± 4.08b 0.73 ± 0.32c 2.70 ± 0.91d Alanine 0.40 ± 0.07a 1.81 ± 0.62b 0.38 ± 0.13a 1.25 ± 0.47 Betaine 10.52 ± 2.80a 28.23 ± 6.77b 3.70 ± 2.13c 34.53 ± 9.79d Choline 0.78 ± 0.13a 6.70 ± 1.09b 1.12 ± 0.25c 6.91 ± 1.18d Citrate 0.55 ± 0.05a 5.25 ± 1.55b 0.62 ± 0.29c 4.93 ± 1.72b Isoleucine 0.15 ± 0.03a 0.58 ± 0.21b 0.16 ± 0.03a 0.48 ± 0.13b Leucine 0.37 ± 0.10a 1.52 ± 0.48b 0.35 ± 0.08a 1.40 ± 0.35b Malate 6.04 ± 0.85a 6.22 ± 3.21b 7.43 ± 2.73a 10.24 ± 5.47a Maltose 17.22 ± 0.182a 24.35 ± 8.46b 0.86 ± 1.52c 21.14 ± 7.27b Succinate 0.54 ± 0.08a 1.34 ± 0.70b 0.43 ± 0.15a 1.43 ± 0.47a 1ANOVA was performed for betaine, succinate, citrate, alanine, leucine, isoleucine, and maltose. Mann-Whitney test was performed for malate, acetate, and choline. Metabolite means followed by different letters are significantly different (p<0.05). (Mean ± SD). Fig. 2 - (A) A typical 1H NMR spectrum from rye bran polar extract and (B) magnified region 0.5 - 3.0 ppm. Annotated me- tabolites: Leucine (1), isoleucine (2), Azelaic acid and sebacic acid (3), alanine (4), acetate (5), malate (6), succinate (7), cit- rate (8), choline (9), betaine (10) and sugar region (11). Ital. J. Food Sci., vol. 27 - 2015 147 Multivariate statistical analysis (OPLS-DA model) also included signals discriminating be- tween bran and endosperm, which appeared as a multiplet at 1.304 ppm, a multiplet at 1.545 ppm, and a triplet at 2.179 ppm (Fig. 2; Table 1). Using 2D NMR and spiking with authen- tic standard, these signals were assigned to two saturated, straight-chain dicarboxylic ac- ids, namely azelaic acid (C9H16O4) and se- bacic acid (C10H18O4). TOCSY NMR indicated that these signals were in the same spin sys- tem (Fig. 3). COSY NMR also confirmed cou- pling between (-CH2-) signals at 1.304 ppm and 1.545 ppm, and between (-CH2-) signals at 1.545 ppm and 2.179 ppm. No coupling to a CH3 group was observed on the TOCSY and COSY spectra, confirming dicarboxylic struc- ture. The carbon chemical shifts were assigned from coupling to the corresponding hydrogen in HSQC NMR. There was a cross-peak between protons at 1.304 ppm and carbon at 31.484 ppm, between protons at 1.545 ppm and carbon at 28.926 ppm, and between protons at 2.179 ppm and carbon at 40.735 ppm. After applying new processing consisting of lining broadening (-1) and Gaussian broadening (0.6), the triplet at 2.179 appeared to be two overlapping triplets from azelaic acid and sebacic acid, the chemi- cal shifts of which deviated from each other by 1.54 Hz. The identity of azelaic acid was further confirmed using authentic standard. The molar concentration of total dicarboxylic acids (azela- ic + sebacic acid) was then calculated and ex- pressed as equivalent to azelaic acid (Table 3). No signal corresponding to azelaic acid was detectable in the 24h urine of a male subject af- Fig. 3 - TOCSY NMR spectrum of typical rye bran polar extract presenting coupling between a multiple at 1.304 ppm (A), a multiple at 1.545 ppm (B), and a triplet at 2.179 ppm (C), which belong to azelaic acid and sebacic acid, and the assignment of the corresponding -CH2- groups on Azelaic acid molecule. 148 Ital. J. Food Sci., vol. 27 - 2015 ter the consumption of refined wheat bread (485 g). However, after the consumption of whole grain rye bread (485 g), NMR signals corresponding to azelaic acid with similar coupling pattern as azelaic acid in the bran extract were detected in 24h urine (Fig. 4). The concentrations of azelaic acid and sebac- ic acid were found to be higher in wheat and rye bran compared with the corresponding en- dosperm. In addition, wheat bran had higher dicarboxylic acid contents than rye bran (Table 3). To our knowledge this is the first study to re- port comparative differences in these dicarboxyl- ic acids in different anatomical compartments of wheat and rye. In humans, 60 and 17 % of the administered dosage of azelaic acid and sebacic acid, respectively, are excreted in the urine with- in the first 12 hours (PASSI et al. 1983). It has been suggested that the dicarboxylic acids are to some extent subjected to β-oxidation, since di- carboxylic acids found in serum and urine pos- sess 2, 4, or 6 carbon atoms shorter than the corresponding administered dicarboxylic acids (PASSI, NAZZARO-PORRO, PICARDO, MINGRONE and FASELLA 1983). Recently BONDIA-PONS et al. using metabolomics approach have shown that azelaic acid beside alkylresorcinols metabolites and enterolactone are the most discriminate me- tabolites, and are found in higher concentration urine after the intervention with whole grain rye bread compared with refined wheat bread (BON- DIA-PONS et al. 2013). The present study showed that the main source of azelaic acid detected in the urine is bran, and that azelaic acid is found in both wheat and rye. The relatively high abundance, anatomical specificity and localization in bran, pattern of metabolism, and previous findings regarding the identification of azelaic acids as discrimina- tive metabolite in the urine after whole grain vs refined grain consumption (BONDIA-PONS et al. 2013) suggest that these dicarboxylic acids can be further investigated as biomarkers of expo- sure to WG products. Further studies are need- ed to investigate the correlation between azelaic acid concentration and alkylresorcinols concen- tration, which are validated biomarkers of WG intake (Ross 2012) in different biofluids after the intake of whole grain products. Little is known about the possible metabolic effects of the di- carboxylic acids in mammals. However, azelaic acid is known for its antibacterial (YU and VAN Fig. 4 - TOCSY NMR spectrum of 24 hour urine of a 35 year old man after the consumption of 485 g whole grain rye bread. The pattern of coupling between signals at 1.304 ppm (A), at 1.545 ppm (B), and at 2.179 ppm (C) was similar to that ob- served in rye bran polar extract. Ital. J. Food Sci., vol. 27 - 2015 149 SCOTT 2004) and anticancer activities (MANOS- ROI et al. 2007), which might contribute to the benefits attributed to WG intake. Three amino acids i.e. alanine, isoleucine, and leucine, were also present in higher concentra- tions in bran than in endosperm. These amino acids gave rise to sharp and distinct NMR sig- nals, distinguishing them from the amino acids in proteins, which possess broad NMR signals. In addition, higher contents of succinate, citrate, and malate were observed in bran than in en- dosperm. These metabolites are associated with the citric acid cycle and central carbon metabo- lism. The higher levels of amino acids and citric acid may indicate higher metabolic activities in bran compared with endosperm. Consistent with previous studies, we observed higher levels of betaine and choline in bran than in endosperm (BRUCE et al. 2010). In addition, wheat bran had higher levels of betaine than rye bran. Circulating betaine is reported to be in- creased postprandially in animal models (BER- TRAM et al. 2009; YDE et al. 2012) and in fast- ing plasma of humans after a 6-8 week inter- vention with WG rye products (MOAZZAMI et al. 2011; MOAZZAMI et al. 2012). Betaine acts as a methyl donor in the betaine-homocysteine me- thyl transferase reaction (BHMT-R), which con- verts homocysteine and betaine to methionine and N,N-dimethylglycine (DELGADO-REYES and GARROW 2005). Recently, in two separate hu- man studies, we observed an increase in BH- MT-R, as indicated by higher N,N-dimethylgly- cine levels (MOAZZAMI et al. 2011; MOAZZAMI et al. 2012) and lower homocysteine levels (MOAZ- ZAMI et al. 2011), after a 6-8 week intervention with WG rye products compared with refined wheat products, which highlights the metabolic effects of betaine located in the bran of cereals. In the present study, we used NMR-based metabolomics as an untargeted approach to gain further insights into the metabolic profile of different anatomical compartments of cereal grains. NMR profiling covers metabolites with µmol/g concentration. NMR also proved useful for identification and structural determination of unknown metabolites associated with differ- ent anatomical compartments. ABBREVIATIONS BHMT, betaine homocysteine methyl transferase; BHMT- R, betaine-homocysteine methyl transferase reaction; CVD, cardiovascular disease; OPLS-DA, orthogonal partial least squares-discriminant analysis; PCA, principal component analysis; T2D, type 2 diabetes; WG, whole grains. ACKNOWLEDGEMENTS We thank the HEALTHGRAIN project for providing us with wheat and rye samples. We would like to thank the Dr. Hå- kansson foundation for supporting the study. We also ac- knowledge Dr. Peter Agback’s help in interpretation of 2D NMR spectra, and Dr. Annica Andersson for coordinating the present study with HEALTHGRAIN project. This study was sponsored by Dr. Håkansson foundation. M.C. and A.A.M. wrote the manuscript. M.C. performed the NMR analysis and metabolomics data analysis. A.G. per- formed the optimization of extraction method. A.A.M. had primary responsibility for the final content. All authors read and approved the final manuscript and there is no conflict of interest. REFERENCES Bertram H.C., Malmendal A., Nielsen N.C., Straadt I.K., Larsen T., Knudsen K.E. and Laerke H.N. 2009. NMR- based metabonomics reveals that plasma betaine in- creases upon intake of high-fiber rye buns in hypercho- lesterolemic pigs. Mol. Nutr. Food Res. 53(8), 1055-1062. Bondia-Pons I., Barri T., Hanhineva K., Juntunen K., Drag- sted L.O., Mykkanen H. and Poutanen K. 2013. UPLC- QTOF/MS metabolic profiling unveils urinary changes in humans after a whole grain rye versus refined wheat bread intervention. Mol Nutr. Food Res. Bruce S.J., Guy P.A., Rezzi S. and Ross A.B. 2010. Quanti- tative measurement of betaine and free choline in plas- ma, cereals and cereal products by isotope dilution LC- MS/MS. J Agric Food Chem 58(4), 2055-2061. Chan J.M., Wang F. and Holly E.A. 2007. Whole grains and risk of pancreatic cancer in a large population-based case- control study in the San Francisco Bay Area, California. Am. J. Epidemiol. 166(10), 1174-1185. De Munter J.S., Hu F.B., Spiegelman D., Franz M. and Van Dam R.M. 2007. Whole grain, bran, and germ intake and risk of type 2 diabetes: a prospective cohort study and systematic review. PLoS Med 4(8), e261. Delgado-Reyes C.V. and Garrow T.A. 2005. High sodium chloride intake decreases betaine-homocysteine S-meth- yltransferase expression in guinea pig liver and kidney. Am. J. Physiol. Regul. Integr. Comp. Physiol. 288(1), R182-187. Eriksson L., Trygg J. and Wold S. 2008. CV-ANOVA for sig- nificance testing of PLS and OPLS (R) models. Journal of Chemometrics 22(11-12), 594-600. Fardet A. 2010. New hypotheses for the health-protective mechanisms of whole-grain cereals: what is beyond fi- bre? Nutr Res Rev 23(1), 65-134. Flint A.J., Hu F.B., Glynn R.J., Jensen M.K., Franz M., Samp- son L. and Rimm E.B. 2009. Whole grains and incident hypertension in men. Am. J. Clin. Nutr. 90(3), 493-498. Haas P., Machado M.J., Anton A.A., Silva A.S. and De Fran- cisco A. 2009. Effectiveness of whole grain consumption in the prevention of colorectal cancer: Meta-analysis of cohort studies. Int. J. Food Sci. Nutr., 1-13. International A. 1999. Definition of whole grain, Eds., http:// www.aaccnet.org/definitions/wholegrain.asp. Jacobs D.R. JR., Andersen L.F. and Blomhoff R. 2007. Whole-grain consumption is associated with a reduced risk of noncardiovascular, noncancer death attributed to inflammatory diseases in the Iowa Women’s Health Study. Am. J Clin. Nutr. 85(6), 1606-1614. Larsson S.C., Giovannucci E., Bergkvist L. and Wolk A. 2005. Whole grain consumption and risk of colorectal cancer: a population-based cohort of 60,000 women. Br. J. Can- cer 92(9), 1803-1807. Lenz E.M. and Wilson I.D. 2007. Analytical strategies in me- tabonomics. J. Proteome Res. 6(2), 443-458. Liu R.H. 2007. Whole grain phytochemicals and health. J. Cereal Sci. 46(3), 207-219. Manosroi A., Panyosak A., Rojanasakul Y. and Manosroi J. 2007. Characteristics and anti-proliferative activi- ty of azelaic acid and its derivatives entrapped in bilay- er vesicles in cancer cell lines. Journal of Drug Target- ing 15(5), 334-341. 150 Ital. J. Food Sci., vol. 27 - 2015 Mellen P.B., Walsh T.F. and Herrington D.M. 2008. Whole grain intake and cardiovascular disease: a meta-analysis. Nutr. Metab. Cardiovasc. Dis. 18(4), 283-290. Moazzami A.A., Andersson R. and Kamal-Eldin A. 2011. Changes in the metabolic profile of rat liver after alpha- tocopherol deficiency as revealed by metabolomics anal- ysis. NMR Biomed. 24(5), 499-505. Moazzami A.A., Bondia-Pons, I., Hanhineva K., Juntunen K., Antl N., Poutanen K. and Mykkanen H. 2012. Metab- olomics reveals the metabolic shifts following an inter- vention with rye bread in postmenopausal women- a ran- domized control trial. Nutr. J. 11, 88. Moazzami A.A., Shrestha A., Morrison D.A., Poutanen K. and Mykkanen H. 2014. Metabolomics reveals differ- ences in postprandial responses to breads and fasting metabolic characteristics associated with postprandi- al insulin demand in postmenopausal women. J. Nutr. 144(6), 807-814. Moazzami A.A., Zhang J.X., Kamal-Eldin A., Aman P., Hall- mans G., Johansson J.E. and Andersson S.O. 2011. Nu- clear magnetic resonance-based metabolomics enable de- tection of the effects of a whole grain rye and rye bran diet on the metabolic profile of plasma in prostate can- cer patients. J. Nutr. 141(12), 2126-2132. Murtaugh M.A., Jacons D.R. JR., Jacob B., Steffen L.M. and Marquart L. 2003. Epidemiological support for the pro- tection of whole grains against diabetes. Proc. Nutr. Soc. 62(1), 143-149. Nicholson J.K. and Wilson I.D. 2003. Understanding ‘global’ systems biology: Metabonomics and the con- tinuum of metabolism. Nat. Rev. Drug. Discov. 2(8), 668-676. Nystrom L., Lampi A.M., Andersson A.A., Kamal-Eldin A., Gebruers K., Courtin C.M., Delcour J.A., Li L., Ward J.L., Fras A., Boros D., Rakszegi M., Bedo Z., ShewryP.R. and Piironen V. 2008. Phytochemicals and dietary fiber com- ponents in rye varieties in the HEALTHGRAIN Diversity Screen. J. Agric. Food Chem. 56(21), 9758-9766. Nystrom L., Lampi A.M., Andersson A.A.M., Kamal-Eldin, A., Gebruers K., Courtin C.M., Delcour J.A., Li L., Ward J.L., Fras A., Boros D., Rakszegi M., Bedo Z., Shewry P.R. and Piironen V. 2008. Phytochemicals and Dietary Fiber Components in Rye Varieties in the HEALTHGRAIN Di- versity Screen. Journal of Agricultural and Food Chem- istry 56(21), 9758-9766. Passi S., Nazzaro-Porro M., Picardo M., Mingrone G. and Fasella P. 1983. Metabolism of straight saturated medi- um chain length (C9 to C12) dicarboxylic acids. J. Lipid. Res. 24(9), 1140-1147. Ross A.B. 2012. Present status and perspectives on the use of alkylresorcinols as biomarkers of wholegrain wheat and rye intake. Journal of nutrition and metabolism 2012, 462967. Schatzkin A., Park Y., Leitzmann M.F., Hollenbeck A.R. and Cross A.J. 2008. Prospective study of dietary fiber, whole grain foods, and small intestinal cancer. Gastroenterolo- gy 135(4), 1163-1167. Shewry P.R., Piironen V., Lampi A.M., Edelmann M., Ka- riluoto S., Nurmi T., Fernandez-Orozco R., Ravel C., Charmet G., Andersson A.A., Aman P., Boros D., Gebru- ers K., Dornez E., Courtin C.M., Delcour J.A., Raksze- gi M., Bedo Z. and Ward J.L. 2010. The HEALTHGRAIN wheat diversity screen: effects of genotype and environ- ment on phytochemicals and dietary fiber components. J. Agric. Food Chem. 58(17), 9291-9298. Slavin J.L., Jacobs D., Marquart L. and Wiemer K. 2001. The role of whole grains in disease prevention. J. Am. Diet. Assoc. 101(7), 780-785. Ward J.L., Poutanen K., Gebruers K., Piironen V., Lampi A.M., Nystrom L., Andersson A.A., Aman P., Boros D., RakszegiM., Bedo Z. and Shewry P.R. 2008. The HEALTH- GRAIN Cereal Diversity Screen: concept, results, and prospects. J. Agric. Food Chem. 56(21), 9699-9709. Ward J.L., Poutanen K., Gebruers K., Piironen V., Lam- pi A.M., Nystrom L., ANDERSSON, A.A.M., Aaman P., Boros D., Rakszegi M., Bedo Z. and Shewry P.R. 2008. The HEALTHGRAIN Cereal Diversity Screen: Concept, Re- sults, and Prospects. Journal of Agricultural and Food Chemistry 56(21), 9699-9709. Yde C.C., Westerhuiis J.A., Bertram H.C. and Bach Knuds- en K.E. 2012. Application of NMR-based metabonomics suggests a relationship between betaine absorption and elevated creatine plasma concentrations in catheterised sows. Br. J. Nutr. 107(11), 1603-1615. Yu R.J. and Van Scott E.J. 2004. Alpha-hydroxyacids and car- boxylic acids. Journal of cosmetic dermatology 3(2), 76-87. Paper Received January 23, 2014 Accepted June 23, 2014