Materials
Grape samples were prepared from Vitis vinifera var. Syrah grown at the Pech Rouge INRA experimental unit (Gruissan, France) and harvested in 2005 at commercial maturity. Syrah grape PA fraction and two apple PA fractions (Malus domestica, var. Marie Ménard (MM) and var. Avrolles (AV), 2005 season) were prepared. The apple and Syrah grape fractions were chosen because they contain high amounts of PAs with different average degrees of polymerization (aDPn) and galloylation.
Reagents for the targeted analysis of microbial metabolites of flavanols using gas chromatography with mass spectrometry (GC–MS) were as follows: heptadecanoic acid and succinic acid-2,2,3,3-d4 used as the internal standards were purchased from Sigma-Aldrich Inc., (St. Louis, USA). The following compounds were used as standards: benzoic acid (BA), 3-hydroxybenzoic acid (3-OHBA), 3-(4′-hydroxyphenyl) propionic acid (4-OHPPr) and 3-(3′,4′-dihydroxyphenyl) propionic acid (3,4-diOHPPr), which were purchased from Aldrich, (Steinheim, Germany); 4-hydroxybenzoic acid (4-OHBA), 2-(3′-hydroxyphenyl) acetic acid (3-OHPAc) and 2-(3′,4′-dihydroxyphenyl) acetic acid (3,4-diOHPAc), which were purchased from Sigma (St. Louis, USA); 3-phenylpropionic acid (3-PPr) and 3,4-dihydroxybenzoic acid (3,4-diOHBA), which were purchased from Fluka (Buchs, Switzerland); and 3-(3′-hydroxyphenyl) propionic acid (3-OHPPr) was purchased from Alfa Aesar (Karlsruhe, Germany). N-Methyl-N-trimethylsilyl-trifluoracetamide (MSTFA) from Pierce (Rockford, USA) was used as the derivatization reagent.
For comprehensive profiling of small polar metabolites using two-dimensional GC coupled to time-of-flight mass spectrometry (GCxGC-TOFMS; Leco, Inc., St. Joseph, MI), the internal standard was 2-hydroxycinnamic acid (mainly trans; Aldrich Inc. H2, 280-9; 97 %; St. Louis, USA), and additional standards of phenolic metabolites were 4-methylcatechol (Aldrich, Steinheim, Germany), vanillic acid (3-methoxy-4-hydroxybenzoic acid; Fluka, Buchs, Switzerland)), 4-hydroxycinnamic acid (Sigma, St. Louis, USA), gallic acid (Extrasynthése, Genay, France) and ferulic acid (Sigma-Aldrich, St Louis, USA). Prior to the addition of MSTFA, methoxyamine hydrochloride (2 %) in pyridine (MOX; Pierce, Rockford, USA) was used in the derivatization process for GCxGC-TOFMS.
Preparation of grape samples
Three grape samples were processed. A crude grape pericarp powder was obtained from berries after removal of seeds, freezing in liquid nitrogen, grinding and freeze-drying. The pericarp powder contained skin and pulp. The skin PA fraction was prepared according to a procedure upscaled from that described by Souquet et al. [
2]. The third sample was a de-alcoholized red wine made from the same Syrah grapes. The red wine was obtained by flash détente (a process known to increase PA content by about 20–30 % [
10]) and fermentation of skins. After alcoholic and malolactic fermentation, de-alcoholization was performed by evaporation under reduced pressure in 45-l batches at a temperature below 40 °C. The red wine batches were pooled and stored at 4 °C under nitrogen atmosphere and then formulated (with glycerol, sugar, and aromas), filtered, bottled and pasteurized. Red wine was freeze-dried just before the colon model experiments. An unopened bottle of red wine was stored in cold (+4 °C) darkness for 3 years, after which it was freeze-dried prior to the experiment coupled with metabolite profiling.
Characterization of the fruit samples
Phenolic composition analysis of the grape material (before and after the in vitro digestion) and of the grape skin PA fraction was performed as described by Mane et al. [
3]. Also, phenolics were extracted from the grape samples (before and after the in vitro digestion) as described by Mane et al. [
3]. Simple phenolic compounds in the grape extract and wine were analysed by reversed-phase HPLC using Waters system (Milford, MA) equipped with a photodiode array detector (W2996). Samples (5 μl) were injected onto a reversed-phase Atlantis T3 column (5 μ, 250 mm × 2.1 mm) supplied by Waters, protected by a guard column of the same material and maintained at 38 °C. Elution was carried out with a gradient of acetonitrile/water/formic acid (80:15:5, v/v/v) in water/formic acid (95:5,v/v). Concentrations were calculated from peak areas at 520 nm for anthocyanins, at 360 nm for flavonols, at 320 nm for hydroxycinnamic acid derivatives and at 280 nm for flavan-3-ols and gallic acid. Malvidin 3-glucoside, quercetin 3-glucoside and caffeic acid were used as external standards for calibration of anthocyanins, flavonols and hydroxycinnamic acids, respectively.
Proanthocyanidins were analysed by reversed-phase HPLC after acid catalysed depolymerization in the presence of phloroglucinol, as previously described [
3,
38]. The concentration of each unit released after phloroglucinolysis was calculated from its peak area at 280 nm (i.e. flavan-3-ols from terminal units and the corresponding phloroglucinol derivatives from extension and upper units), using the calibration curve established for the corresponding standard, either commercial ((+)-catechin, (−)-epicatechin, (−)-epigallocatechin and (−)-epicatechin-3-gallate) or purified in the laboratory (phloroglucinol derivatives). Eventual differences in the dilution or injection volumes were compensated for by taking into account the peak area of the internal standard (methylparaben). Total flavan-3-ol content was calculated by summing all units, and aDPn was calculated as the molar ratio of total released units to total terminal units. The percentages of epicatechin gallate units (% gallate) and of epigallocatcehin units (% epigallocatechin) were also calculated.
Red wine phenolics were analysed as described by Ducasse et al. [
39]. For purified apple PA fractions, polyphenol analyses were performed by HPLC following thiolysis according to the procedure described by Guyot et al. [
40].
Simple sugars (glucose, fructose and sucrose) were measured colorimetrically using the solutions and instructions from the Boehringer analysis kit (R-Biopharm, St Didier au Mont d’or, France). Alcohol-insoluble solids were isolated from grape powders (native and digested) by extensive washing of the powders until the extracts were sugar free, as described by Renard [
41]. Red wine and PA fractions were directly submitted to polysaccharide analysis.
The individual neutral sugars were analysed by gas chromatography (capillary column of 30 m × 0.25 mm i.d. coated with DB225, 0.15 μm film thickness, J & W Scientific, Folsom, USA) at 215 °C, using hydrogen as carrier gas, after sulphuric acid hydrolysis (1 M, 3 h, 100 °C) and derivation to alditol acetates [
42]. AIS preparations from grape powders were submitted to pre-hydrolysis in 13 M sulphuric acid (1 h, room temperature) [
43]. Myo-inositol was used as internal standard. Uronic acids were determined spectrophotometrically by
m-hydroxydiphenyl assay as described by Blumenkrantz and Asboe-Hansen [
44] after acid hydrolysis of cell walls (Saeman procedure) with galacturonic acid as external standard.
In vitro colon model
Freeze-dried Syrah grape powders or red wine were dosed 100 mg d.w./10 ml of faecal inoculum, whereas PA fractions were dosed 25 mg d.w./10 ml in the comparison with Syrah products and fruit PA fractions, and 28 mg, 14 mg and 7 mg/10 ml faecal inoculum in the experiment investigating the dose effect. Four colon model experiments were performed under strictly anaerobic conditions according to Aura et al. [
18,
46], with the following specifications: Faecal suspensions were prepared for each experiment by pooling and suspending the faeces of at least 4 healthy donors to 0.11 M—carbonate—0.02 M phosphate buffer (pH 5.5) [
47] in a Warring-Blender. The suspensions were filtered through a 1-mm sieve, diluted to 10 % (w/v) and applied immediately to the samples. Samples were incubated in a water bath at 37 °C for 0, 2, 4, 6, 8 and 24 h and stirred magnetically (250 rpm), unless otherwise stated.
In the experiments for targeted analysis by the GC–MS instrument, aliquots were drawn from the bottles and microbial metabolites and short-chain fatty acids (SCFA) were analysed after extraction of each aliquot. In the GCxGC-TOFMS experiment, aliquots were drawn from the bottles for SCFA analysis, and microbiota was removed using 0.2-μm PTFE filters (Millipore Corp., Bedford, MA, USA) before extraction of phenolic metabolites.
Total SCFA were analysed by gas chromatography with an FID detector after diethyl ether extraction according to Aura et al. [
48]. The sum of SCFAs included the concentrations of acetic, propionic and butyric acids.
For targeted GC–MS analysis, microbial metabolites were extracted from faecal suspensions (1 mL) twice with 3 ml of ethyl acetate, and the solvent was evaporated. Derivatization was performed by adding dichloromethane (100 μl) and MSTFA (30 μl) and incubating for 5 min at 50 °C. The targeted analysis of microbial metabolites was performed by gas chromatography with mass detection (GC–MS) using selective-ion-monitoring (SIM) as described by Bazocco et al. [
17] with authentic standards and heptadecanoic acid and succinic acid-2,2,3,3-d
4 as internal standards. The metabolite formation was calculated as μmol/L of formed metabolite at each time point and expressed as averages and standard deviations.
For GCxGC-TOFMS, the derivatization was performed as follows: Internal standard (15 μL of 123 ppm 2-hydroxycinnamic acid) was added to 1 ml of filtered microbiota-free faecal water, and the samples were extracted twice with 2 ml of ethyl acetate. The extracts were evaporated to dryness under nitrogen and derivatized with 25 μl of MOX (1 h, 45 °C) and 25 μl of MSTFA (1 h, 45 °C). 5 μl of retention index standard mixture with five alkanes at 800 ppm was added to the metabolite mixture. In the targeted analysis, the metabolite formation was calculated as μmol/l of formed metabolite at each time point and expressed as averages and standard deviations. Analysis by GCxGC-TOFMS was performed as described in Aura et al. [
49].
Measurement data from GCxGC-TOFMS were first processed by ChromaTOF software, which identifies compounds by matching deconvoluted spectra against an NIST05 mass spectral library. The results were exported to text files, and the in-house developed software Guineu [
36] was used for aligning and normalization of compounds in different data sets for further analyses. The original GCxGC-TOFMS data include retention times, retention indices (RI), spectral information for possible identification, spectral similarity value (S = 0–1,000) and peak response data. The linear retention indices were calculated on the basis of the retention times of the compounds and the retention times of the retention index standards (
n-alkanes). Alignment of the data was performed on the basis of retention indices, second-dimension retention times and spectra.
After alignment of the GCxGC-TOFMS data, two filtration criteria were utilized for preliminary identification: spectral match >850 and retention index (RI) as follows: RIexp-RIlit < 25 or RIexp-RIstd, exp < 25, in which RIexp was the experimental RI for a compound, RIlit was the literature value for the identified compound and RIstd, exp was the experimental RI value for a standard compound. Compounds not fulfilling the criteria were renamed as unknowns, and relevant compounds were subjected to further identification. The identification of the unkowns at this stage was based on a spectral search from the NIST05 library or the in-house collected library and their retention indices.
Statistical analysis
Two-way ANOVA for repeated measures was applied on quantitated metabolites by using a program designed for MatLab (R2008b). The program evaluated the responses against each substrate and the faecal control. Significant (p < 0.05) differences from the faecal control were designated as small letters, and different letters corresponded significantly (p < 0.05) to different levels of responses within a time point, unless otherwise stated.
Two-way ANOVA was also performed for non-targeted GCxGC-TOFMS data using the
aov function of the
stats package in
R statistical programming language (
http://www.r-project.org). The multiple hypothesis testing problem was addressed by correcting the
p values to control the false discovery rate (FDR) using the
p. adjust function of the
stats package. Those metabolites showing FDR
q-values lower than 0.0001 were included in the visualization by heat maps, which were used for displaying the relevant metabolites. Heat maps were produced with
R using the
heatmap.2 function of the
gplots package. Differences at each time point were evaluated by a two-sided
t test at each time point using the
t test function of the
stats package. The asterisks shown in the heat map indicate significant differences in means at each time point based on the
t test (*
p < 0.05; **
p < 0.01; ***
p < 0.001).
GOLM Metabolome Database (GMD) (
http://gmd.mpimp-golm.mpg.de/search.aspx) and the
Guineu program [
36] were utilized for second-stage identification of those compounds that lacked spectral matches with compounds from the NIST05 or in-house collected libraries. GMD database allows searching of the database based on submitted GC–MS spectra, retention indices and mass intensity ratios. In addition, the database allows a functional group prediction, which helped to characterize
unknown metabolites without available reference mass spectra in the GMD.
The visualization was performed by calculating 2-based logarithmic fold changes of the relative peak areas from GCxGC-TOFMS analysis against the corresponding controls: faecal control (no red wine; Supplement Fig. 4A) or red wine in buffer (no faecal microbiota; Supplement Fig. 4B). The profile of the individual metabolite was visualized as colour intensities (red as over-expression and blue as under-expression) and the time point specific significances (t test p values) as asterisks against the corresponding control. The non-targeted metabolite profiling was semi-quantitative. The names of the over-expressed metabolites were verified by comparing the mass spectra with those found in GMD and the names for the unkowns were named according to the group specifications and displayed in the final heat maps.