Food Research International 116 (2019) 786–794

Contents lists available at ScienceDirect

Food Research International

journal homepage: www.elsevier.com/locate/foodres

Edible nuts deliver polyphenols and their transformation products to the T

large intestine: An in vitro fermentation model combining targeted/
untargeted metabolomics
Gabriele Rocchettia,b,c, , Sudarshana Reddy Bhumireddyc, Gianluca Giubertib, Rupasri Mandalc,
⁎

Luigi Lucinib, , David S. Wishartc,d

⁎

a
Department of Animal Science, Food and Nutrition, Università Cattolica del Sacro Cuore, Piacenza 29122, Italy
b
Department for Sustainable Food Process, Università Cattolica del Sacro Cuore, Piacenza 29122, Italy
c
Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada
d
Department of Computing Science, University of Alberta, Edmonton, AB T6G 2E8, Canada

ARTICLE INFO ABSTRACT

Keywords: The fate of polyphenols from edible tree nuts was investigated using a simulated in vitro intestinal fermentation
Nuts system. The digested food matrix was fermented for 48 h and the changes in the phenolic profiles were evaluated
Polyphenols by both untargeted UHPLC-QTOF and targeted UHPLC-Orbitrap mass spectrometry. The untargeted metabo-
In vitro fermentation lomics approach allowed us to monitor the comprehensive changes in phenolic profiles from 0 up to 48 h of in
Food metabolomics
vitro fermentation. Multivariate statistics (i.e., orthogonal projection to latent structures discriminant analysis)
UHPLC-QTOF
applied to this untargeted data allowed us to identify the most discriminating phenolic metabolites and to
UHPLC-Orbitrap
further understand the colonic transformation pathways involved. In particular, 13 putatively identified com-
pounds derived from flavonoids, lignans and phenolic acids were found to have the highest discrimination
potential. Six phenolic metabolites were then quantified by means of targeted metabolomics (using a UHPLC-
Orbitrap). These metabolites included 3,4-dihydroxyphenylacetic acid, 4-hydroxybenzoic acid, hippuric acid,
caffeic acid, protocatechuic acid and protocatechuic aldehyde. Using the targeted data, a clear matrix effect was
observed over time, with an increase of some phenolic metabolites moving from 8 to 48 h of in vitro fermen-
tation. Based on these data, catabolic pathways for colonic microbial degradation of flavonoids, hydro-
xycinnamic acids, tyrosols and lignans are proposed. Our findings show that edible tree nuts deliver polyphenols
to the colon, where several microbial transformations occur that lead to smaller phenolic metabolites being
observed. Furthermore, we found that the combined use of targeted and untargeted metabolomics can be par-
ticularly effective for investigating the fate of polyphenols in the large intestine.

1. Introduction polyphenols may also contribute (Lamuel-Raventos & Onge, 2017).

The consumption of tree nuts, such as almonds, hazelnuts, pis-
tachios, walnuts, pecans, and cashews has been found to be inversely
associated with the incidence and prevalence of several common,
chronic diseases including type 2 diabetes and coronary heart disease
(Lamuel-Raventos & Onge, 2017). Nut consumption has also been
shown to have favorable effects on the lipid profile and cardiovascular
disease risk factors, as demonstrated by several supplementation studies
involving almonds, hazelnuts, peanuts, pistachios and walnuts (Vinson
& Cai, 2012). The health benefits of tree nuts have been mainly at-
tributed to their fatty acid profile, although the high levels of dietary
fiber, vitamins and secondary metabolites such as phytosterols and
Polyphenols are a diverse family of bioactive compounds found in al-
most all plant foods. Five main polyphenolic classes exist, namely fla-
vonoids, phenolic acids, lignans, stilbenes, and phenolic alcohols
(Rothwell, Knaze & Zamora-Ros, 2017).
After ingestion, polyphenols can undergo extensive metabolism
within the human gastrointestinal tract. The bioactivity and bioavail-
ability of a given polyphenol is often defined by the metabolic changes
it experiences in the body (Bohn et al., 2015). In particular, some
polyphenols can be released from the food matrix and partially ab-
sorbed, while others escape from the small intestine almost unchanged,
thereby reaching the large intestine where a series of microbial trans-
formations usually occur (Cardona et al., 2013; Ozdal et al., 2016). It

⁎
Corresponding authors at: Department for Sustainable Food Process, Università Cattolica del Sacro Cuore, Piacenza 29122, Italy.
E-mail addresses: gabriele.rocchetti@unicatt.it (G. Rocchetti), luigi.lucini@unicatt.it (L. Lucini).

https://doi.org/10.1016/j.foodres.2018.09.012
Received 11 July 2018; Received in revised form 28 August 2018; Accepted 8 September 2018
Available online 11 September 2018
0963-9969/ © 2018 Elsevier Ltd. All rights reserved.
G. Rocchetti et al.
has been estimated that about 5–10% of total human polyphenol intake
is absorbed in the small intestine, while the remaining 90–95% must be
subject to enzymatic activity of the gut microbiota (Cardona et al.,
2013). It has also been suggested that polyphenolic interactions with
gut microbes may lead to several biochemical transformations of the
native polyphenols into more bioavailable metabolites. These metabo-
lites are typically lower molecular weight substances and many of them
may be responsible for the real health-promoting effects of polyphenol-
rich foods (Mosele et al., 2015; Espín et al., 2017).
The most important biotransformation reactions promoted by gut
microbiota can be grouped into three major catabolic processes: i) hy-
drolysis (O-deglycosylations and ester hydrolysis), ii) cleavage (C-ring
cleavage and demethylation), and iii) reductions (dehydroxylation and
double bond reduction) (Mosele et al., 2015; Serra et al., 2012). In vivo,
these phenolic metabolites are transported to the liver where they are
subject to phase I (oxidation, reduction and hydrolysis) and phase II
(conjugation) metabolism or secreted back into the gut, where they are
again absorbed or finally excreted (Tarko et al., 2013).
Nuts are among the richest food source of polymerized polyphenols
(i.e., ellagitannins and proanthocyanins). These polymerized poly-
phenols can be further processed by colonic microbiota, yielding a wide
range of low molecular weight phenolic metabolites. These metabolites
appear to alter the microbial ecosystem as well as the microbiota pro-
file, often leading to beneficial prebiotic effects (Lamuel-Raventos &
Onge, 2017; Cardona et al., 2013). In this regard, Rocchetti, Giuberti,
and Lucini (2018) showed a temporal increase in the bioaccessibility of
nut phenolics during in vitro simulated gastrointestinal digestion and
fermentation. They found that lower-molecular-weight phenolic com-
pounds such as phenolic acid, tyrosol and alkylphenol equivalents
generally increased over time. In addition to facilitating the production
of polyphenolic metabolites, colonic microbiota is also able to process
the undigested nut/food matrix (i.e., the fiber) reaching the colon. This
typically leads to the biosynthesis of short-chain fatty acids (SCFA) and
the release of the so-called bound phenolics, or non-extractable-poly-
phenols (NEPP) (Lamuel-Raventos & Onge, 2017; Rocchetti et al., 2017;
Acosta-Estrada et al., 2014). Once accessible, these NEPPs can be fur-
ther processed by the microbial community of large intestine.
Over the past few years, in vitro fermentation models mimicking the
physiological conditions of the large intestine have been applied to
better explore the biotransformation of polyphenols (Rocchetti et al.,
2017; Rocchetti, Chiodelli, et al., 2018; Juániz et al., 2016). These in
vitro methods, which often use human or pig faecal inoculum, are
widely used to predict phenolic bioavailability and have several cost
and reproducibility advantages compared to in vivo models (Minekus
et al., 2014). When combined with metabolomic approaches, these in
vitro models permit the study of several aspects of molecular nutrition,
including the characterization of polyphenols during gastrointestinal
processes (Rocchetti, Giuberti, & Lucini, 2018) or the analysis of phy-
tochemicals for food quality and authenticity purposes (Castro-Puyana
et al., 2017).
However, to the best of our knowledge, no in vitro digestion/fer-
mentation models have been used to explore the possible catabolic
pathways by which nut polyphenols are processed. Likewise, advanced
metabolomic techniques have not been applied to characterize these in
vitro derived catabolic end products. Therefore, the purpose of this
work was to use both targeted metabolomics (i.e., UHPLC-Orbitrap MS)
and untargeted metabolomics (i.e., UHPLC-QTOF MS) to robustly and
accurately characterize the metabolic fate of nut polyphenols in the
large intestine.
2. Materials and methods
2.1. Samples
Five commercial edible nut categories (100 g plastic bags) were
analysed in this study, namely peanuts (Arachis hypogaea L.), peeled
787
Food Research International 116 (2019) 786–794
hazelnuts (Corylus avellana L.), pistachios (Pistacia vera L.), almonds
(Prunus dulcis Mill.) and natural walnuts (Juglans regia L.). Samples were
gratefully provided by Nicola Scala S.r.l. (Liveri, Italy) and freshly
prepared prior to the further analyses.
2.2. Extraction and profiling of polyphenols from raw nuts
Free and bound polyphenols were extracted from each raw nut
sample and then profiled and quantified, as described in a previous
work (Rocchetti et al., 2018). For the quantification step, methanolic
standard solutions of ferulic acid (for phenolic acids), sesamin (lig-
nans), cyanidin (anthocyanins), catechin (flavanols and flavonols), lu-
teolin (flavones), resveratrol (stilbenes), 5-pentadecylresorcinol (alkyl-
phenols) and tyrosol (tyrosols and others) were used. All standard
compounds were purchased from Extrasynthese (Genay, France) each
having a purity > 98%. These reference compounds were considered as
representative of their respective phenolic class.
2.3. Simulated in vitro digestion and faecal fermentation process
To simulate digestion and the faecal fermentation process, nut
samples were milled and hydrolyzed with digestive enzymes simulating
human gastric and pancreatic digestion phases, and then subjected to
an in vitro faecal fermentation process, as detailed by Rocchetti et al.
(2018). Briefly, pooled dried post-hydrolysis residues were incubated
with a fresh faecal inoculum collected from five growing pigs (35–46 kg
body weight; fed with a standard commercial diet devoid of anti-
biotics). The faecal inoculum was obtained by pooling equal amounts
(by wet weight) of fresh faeces from each animal and then gently mixed
with 200 mL of 39 °C CO2-saturated buffer solution and filtered through
a 250 μm mesh screen. An additional volume of buffer solution was
added to reach the desired dilution of faeces in the buffer (0.05 g
faeces/mL buffer) (Giuberti, Gallo, Moschini, & Masoero, 2013).
Thereafter, 400 mg of each dry pooled nut residue was weighted in
triplicate into a 125 mL glass serum bottle (Z114041, Sigma-Aldrich,
Milan, Italy) and filled with 60 mL of the buffer solution containing the
faecal inoculum. Bottles were sealed with a rubber stopper and im-
mediately placed in a 39 °C shaking water bath (50 rpm). Three bottles
without substrate were used as blanks. All manipulations and sample
handlings were done under continuous CO2 flushing (technical grade:
5.5; from SAPIO, Monza, Italy). Three incubation runs were conducted
over three different days, and bottles within runs were considered re-
petitions, whereas bottles between runs were considered replicates.
Finally, aliquots (3 mL) were aseptically removed from each bottle
using a 5 mL plastic injection syringe equipped with a catheter needle at
0, 8, 24 and 48 h after incubation and stored at −4 °C for further
analysis.
2.4. Characterization of bacterial profile in faecal inoculum
The DNA extraction of the dominant bacterial communities in fresh
faeces was carried out according to the method reported by Miragoli
et al. (2016). Afterwards, copy numbers of the 16S rRNA genes from
Bacteroides/Prevotella, Clostridium coccoides, Bifidobacterium, Lactoba-
cillus, Enterococcus, Enterobacteriaceae, and Ruminococcus groups were
determined by means of previously reported DNA primers (Bruzzese
et al., 2014), according to a real-time polymerase chain reaction (RT-
PCR).
2.5. Untargeted UHPLC-ESI-QTOF screening of phenolic metabolites during
in vitro faecal fermentation
Using samples collected from the simulated in vitro (large intestine)
fermentation, 1 mL aliquots were taken at 0, 8, 24 and 48 h and ex-
tracted into the same volume of a 3% formic acid in 70% methanol
(LCMS grade, Sigma-Aldrich, Milan, Italy) solution. Each aliquot was
G. Rocchetti et al.
stored overnight at −18 °C, centrifuged at 10,000 ×g for 10 min at 4 °C,
and then filtered in amber vials using 0.22 μm cellulose syringe filters.
In order to monitor the change in the phenolic metabolite profile
during the in vitro fermentation process, the methanolic extracts were
analysed by ultra-high-pressure liquid chromatography and then de-
tected by a hybrid quadrupole-time-of-flight mass spectrometer
(UHPLC-ESI/QTOF-MS), according to previously optimized analytical
conditions (Rocchetti et al., 2017; Rocchetti, Chiodelli, et al., 2018). In
conducting these assays an Agilent 1290 liquid chromatograph coupled
with an Agilent G6550 mass spectrometer equipped with a dual elec-
trospray jet stream ionization system (all from Agilent technologies,
Santa Clara, CA, USA) was used. The mass spectrometer was set to the
“Scan mode” to acquire positive ions in the range of 50–1100 m/z. Raw
data were processed using the Agilent Profinder B.06 software, ac-
cording to the targeted ‘find-by-formula’ algorithm. Accurate mass in-
formation was used together with the spectral isotope pattern (isotopic
spacing and isotopic ratio) to achieve a higher confidence in metabolite
identification. Features that were not present in 100% of replications
within at least one treatment were discarded. Data from Phenol-Ex-
plorer 3.6 (Rothwell et al., 2013) listed under “polyphenol metabolites”
was used as a reference for compound identification, adopting a 5 ppm
tolerance for mass accuracy. Based on the strategy adopted, identifi-
cation was carried out according to Level 2 (i.e., putatively annotated
compounds) as set out by the COSMOS Metabolomics Standards In-
itiative (http://cosmos-fp7.eu/msi). Phenolic metabolites that passed
the mass accuracy and frequency of detection thresholds, that had
plausible chromatogram peak features, and that showed significantly
different trends from the control (faeces only), were considered as po-
tential polyphenol fermentation markers.
2.6. Targeted determination of phenolic metabolites during in vitro faecal
fermentation
Targeted LC-MS analyses were performed on Q Exactive HF Hybrid
Quadrupole-Orbitrap Mass Spectrometer (Thermo Scientific, Waltham,
MA, USA) coupled to a Vanquish ultra-high-pressure liquid chromato-
graphy (UHPLC) pump and equipped with a HESI-II probe (Thermo
Scientific, USA). The chromatographic separation was achieved on re-
versed-phase Agilent Eclipse XDB C18 column (3.5 μm particle size;
3.0 × 100 mm; Agilent, Santa Clara, CA, USA) connected to an
ACQUITY UPLC CSH C18 VanGuard Pre-column, (1.7 μm,
2.1 mm × 5 mm). Mobile phases were water (solvent A) and methanol
(solvent B), both LC-MS grade. Ammonium formate (5 mM) and formic
acid 0.1% (v/v) were added to both phases. The binary gradient elution
program was as follows: t = 0 min, 0% B; t = 0.2 min 5% B; t = 9 min,
75% B; t = 10 min 0% B; t = 13 min, 0% B. The sample injection vo-
lume and flow rates were 5 μL and 0.5 mL/min, respectively, while the
column temperature was maintained at 50 °C. For MS analysis, the LC
eluents were allowed to enter the heated electrospray ionization (HESI-
II) probe with typical operational conditions corresponding to a probe
heater temperature of 425 °C, a capillary temperature of 275 °C, a
2.40 μA spray current, and a 2.5 kV spray voltage. Nitrogen was used as
both auxiliary and sheath gas, at a flow rate of 50 mL/min and 12 mL/
min, respectively. A mass range of m/z 70 to 1000 and a resolution of
60,000 (FWHM) was employed. Each faecal fermented nut sample was
analysed in triplicate.
Finally, data acquisition (both chromatograms and spectral in-
formation) and processing was done using XCalibur 4.1.31.9 (Thermo
Fisher Scientific Inc.). Quantification was performed against calibration
curves built with pure reference standard compounds for 3,4-dihy-
droxyphenylacetic acid, 4-hydroxybenzoic acid, hippuric acid, caffeic
acid, protocatechuic acid and protocatechuic aldehyde (all from Sigma-
Aldrich; purity > 98%). Accurate mass information, spectra compar-
ison and retention time data were used for compound annotation
(supplementary material). Quantitative results were expressed as nmol/
g dry matter (DM).
788
Food Research International 116 (2019) 786–794
2.7. Multivariate statistics
Untargeted MS data on the phenolic metabolite profiles observed
during in vitro large intestine fermentation were analysed using Agilent
Mass Profiler Professional B.12.06. In particular, compounds were fil-
tered by abundance and by frequency (only those compounds with an
area > 5000 counts and appearing in 100% of samples in at least one
condition were considered), normalized at the 75th percentile, base-
lined to the median of each compound in all samples and corrected for
the control (only faeces) in order to eliminate matrix effects.
The raw and processed metabolomic dataset obtained after the in
vitro large intestine fermentation was exported to SIMCA 13 (Umetrics,
Malmo, Sweden), scaled and elaborated for orthogonal partial-least-
squares discriminant analysis (OPLS-DA). Variations between the
groups were separated into predictive and orthogonal components (i.e.,
those components ascribable to technical and biological variation). The
presence of outliers was investigated according to Hotelling's T2
method (i.e., the distance from the origin in the model) using 95% and
99% confidence limits for “suspect” and “strong” outliers, respectively.
Cross validation of the OPLS-DA model was performed to exclude over
fitting using CV-ANOVA (p < .001) and through permutation testing
after recording model parameters (goodness-of-fit R2Y and goodness-of-
prediction Q2Y). For the Q2Y prediction ability, a value > 0.5 was
adopted as a threshold to identify acceptable models, according to
software recommendations and as described in the literature
(Rombouts et al., 2017). Finally, variable importance in projection (VIP
analysis) was used to evaluate the importance of phenolic metabolites
during the in vitro fermentation process and to select those with the
highest discrimination potentials (VIP score > 1).
3. Results and discussion
3.1. Evaluation of microbial groups characterizing faecal inoculum
Colonic bioconversion of phenolic compounds is strictly related to
the microbiota activity and composition found in the colon (van
Duynhoven et al., 2011). Therefore, to fully investigate the microbial
catabolism of polyphenols, one should include a combination of in vitro
and in vivo colonic models. However, in vivo models are not always
available due to ethical concerns and in vitro large intestine fermenta-
tion methods are considered a valuable alternative for performing in
vivo studies (Payne et al., 2012). For many kinds of in vitro large in-
testine models, faecal inoculum from humans (Parkar et al., 2013) or
pigs (Rocchetti et al., 2017; Rocchetti, Chiodelli, et al., 2018) has been
used. Indeed, pig faecal inocula are considered to be among the most
suitable alternatives to human faecal inocula. This is because pig faecal
inocula reflect the environmental conditions of the distal region of the
human large intestine owing to the similarities they share in the di-
gestive anatomy, gastrointestinal fermentation profiles (the colon being
the main site of bacterial fermentation in pigs and humans), physiology,
nutrition as well as gut microbiota (Roura et al., 2016). Real-time PCR
analysis seems to corroborate this, since both Firmicutes and Bacter-
oidetes have been identified as the two dominant bacteria phyla in the
human, mouse and pig gut. Interestingly, two other enterotype-like
clusters (Ruminococcus and Prevotella) have been recently identified in
pig faeces and these are very similar to those described in humans
(Roura et al., 2016). Our results clearly indicate that Bacteroides/Pre-
votella (belonging to the Bacteroidetes phylum) and Clostridium coc-
coides (part of the Firmicutes phylum) were the most represented mi-
crobial groups in the pig faecal inoculum, accounting for 55.5% and
38.5% of the bacterial taxa respectively. These ratios are similar to
those reported in human stool samples (Stearns et al. 2011). Other
microbial groups were also identified by our PCR studies including
Lactobacillus (4%), Ruminococcus (1%), Enterococcus (0.5%), Bifido-
bacterium (0.3%) and Enterobacteriaceae (0.2%). Gut microbiota en-
zymes have an essential role in the transformation of native
G. Rocchetti et al.
Fig. 1. Cumulative phenolic composition (as mg/kg Equivalents) in different raw edible nuts, either in free [A] or bound [B] phenolic fractions.
polyphenols into the large intestine, however relatively few bacterial
species are able to catalyse the metabolic transformations of poly-
phenols (i.e., Escherichia coli, Bifidobacterium spp., Lactobacillus spp.,
Bacteroidetes spp., and Firmicutes spp.) (Parkar et al., 2013).
3.2. Changes of the phenolic profile during in vitro faecal fermentation
A comprehensive untargeted metabolomic profile of both free and
bound phenolics from raw nut samples was investigated by means of
UHPLC-ESI-QTOF mass spectrometry. This was done in order to broadly
characterize the native phenolic composition of these food matrices in
an unbiased manner. Overall, 339 compounds were putatively anno-
tated in the free phenolic fraction, while the bound fraction was char-
acterized by a somewhat lower number of compounds (i.e., 114 puta-
tive annotations). A detailed list of all phenolic compounds annotated
in each nut sample is provided in the supplementary material, together
with composite mass data. Semi-quantitative values for each phenolic
class are reported in Fig. 1. Peanuts were characterized by the highest
total phenolic content (as sum of free and bound phenolics), with
8903.9 mg kg−1 equivalents and 65% of these being free phenolic
compounds (i.e., phenolic acid, tyrosol, alkylphenol and flavanol/fla-
vonol equivalents). Regarding the other samples, peeled hazelnuts were
the second most abundant source of total phenolics (6895.8 mg kg−1
equivalents), followed by almonds, pistachios and walnuts (on average
5505.1 mg kg−1 equivalents). Interestingly, the bound phenolic fraction
of all nut samples was characterized by very high levels of tyrosol
equivalents, with an average content of 1544.6 mg kg−1 equivalents. In
particular, the most abundant compounds detected were tyrosol p-
HPEA-AC, followed by hydroxycoumarin/4-hydroxycoumarin (supple-
mentary material). The polyphenol content of tree nuts has been re-
viewed recently (Alasalvar and Bolling, 2015), and it is characterized
by a wide distribution of phenolic acids, flavonoids (i.e., flavonols,
anthocyanins, flavanols, isoflavones), condensed and hydrolysable
tannins, lignans and stilbenes. Most of the studies available on tree nuts
that compare the total phenolic content of these nutrient-dense foods,
have used colorimetric assays such as Folin-Ciocalteu. However, as
recently reported by Granato et al. (2018), colorimetric methods pre-
sent many pitfalls and they are generally more useful as screening tools
rather than quantitative tools. Indeed, it is widely acknowledged that
the use of high-resolution mass spectrometry is the best method to
profile and quantify polyphenols in different food matrices (Granato
789
Food Research International 116 (2019) 786–794
et al., 2018). In regard to tree nuts, polyphenol content can vary ex-
tensively between different genotypes (pre-harvest factors) and some
phenolic classes can be deeply influenced by processing (i.e., roasting,
irradiation, and pasteurization) and storage (i.e., temperature and
duration) in the post-harvest period (Bolling et al., 2011; Grundy et al.,
2016).
After our initial characterization of the raw materials, the same
untargeted approach (UHPLC-ESI-QTOF mass spectrometry) was used
to investigate the changes in the phenolic profile of edible nuts during
simulated in vitro gastrointestinal digestion and fermentation. In this
regard, Phenol-Explorer” (Rothwell et al., 2013), the comprehensive
database on polyphenols and polyphenol metabolites, was used to help
identify the compounds from the QTOF-MS data. Overall, 75 phenolic
metabolites were putatively annotated in the faecal fermented samples,
with 31 compounds having an identification score > 75% (supple-
mentary material). Therefore, in order to better characterize the
changes and the catabolic processes occurring during colonic fermen-
tation, a multivariate statistical approach (OPLS-DA) was employed.
The OPLS-DA score plot illustrating the modification of the phenolic
profile during fermentation is provided as Fig. 2. This OPLS-DA plot
clearly shows the modification of the phenolic profile moving from
digested to faecal fermented samples. This result confirms the well-
known fact that polyphenols are subject to extensive microbial meta-
bolism as previously reviewed by Mosele et al. (2015). The parameters
of this OPLS-DA regression were more than acceptable, with relatively
high values of R2Y and Q2Y (i.e., 0.89 and 0.75, respectively) and an
adequate CV-ANOVA coefficient (p < 0.001). The Hotelling's T2 and
permutation test respectively were also found to be adequate and are
provided as supplementary material. Indeed, OPLS is able to process
very large dataset even when the number of the variables is much larger
than the number of samples (Farrés et al., 2015; Worley and Powers,
2013). However, most of these variables are of little or no relevance for
investigating the biological problem, as they are not directly related to
the dependent variable (Y). Therefore, variable selection methods are
often employed to help reduce the number of predictors, thereby pro-
viding only the most relevant ones. One of the most important variable
selection methods is the VIP (variable importance in projection)
method. Specifically, this approach identifies those X variables that
contribute the most to explaining the Y-variance (Farrés et al., 2015). In
this work, the VIP method was used to identify the most important
metabolites contributing to the changes in the phenolic profile over
G. Rocchetti et al.
Fig. 2. Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) score plot on edible nuts phenolic metabolites' profile at 0 (immediately after in
vitro digestion), 8, 24 and 48 h of in vitro faecal fermentation.
time. Table 1 displays the phenolic metabolites having the highest VIP
scores (> 1), together with their chemical structure and the colonic
pathways in which they are involved. Overall, 13 compounds were
found to have sufficiently high VIP scores, with the 4-hydro-
xyphenylacetic acid possessing the highest VIP score (1.33), followed
by a methylequol with a VIP score of 1.31. The phenolic metabolites
identified by the VIP approach are representative of the main catabolic
pathways that are believed to play key roles in the large intestinal
fermentation of polyphenols. In fact, several catabolic pathways have
been proposed for each specific phenolic class and subclass (Mosele
et al., 2015). For instance, the undigested polyphenols that are able to
pass into the large intestine are often subject to further degradation into
small phenolic acids by colonic microflora. In this regard, phenolic
glycosides are first hydrolysed by bacteria to aglycones, which are then
transformed into different organic acids through the action of microbial
enzymes such as β-glucosidase, β-rhamnosidase and esterases. Fur-
thermore, microbial enzymes are able to catalyse the degradation of the
common flavonoid ring into simple chemical subunits, after which
several reactions such as hydrolysis, dihydroxylation, demethylation
and decarboxylation occur (Mosele et al., 2015; Ozdal et al., 2016).
These catabolic routes play a key role in the synthesis of lower-mole-
cular-weight phenolics (i.e., phenyl-valerolactones, hydro-
xyphenylacetics and hydroxyphenylpropionics) (Tarko et al., 2013);
(van Duynhoven et al., 2011). Overall, the multivariate OPLS-DA model
we used, coupled with a VIP approach, allowed us to confirm that the
native polyphenols found in edible tree nuts are processed during in
vitro gastrointestinal digestion and fermentation through microbial ac-
tivity, independently from the type of nut considered.
The extensive bioconversion of phenolic compounds due to colonic
microbiota catabolism produces a wide range of phenolic metabolites.
Some of these can also be exported into the systemic circulation. These
metabolites are part of the so-called “food metabolome” and over the
past several years various metabolomic techniques (including NMR
and/or LC-MS methods) have been applied to characterize these com-
pounds (Scalbert et al., 2014). While considerable effort has been de-
voted to putatively identifying the constituents of the food metabolome
(through untargeted metabolomic studies), relatively little effort has
been devoted to validating and fully quantifying these compounds.
Indeed, an intrinsic limitation of untargeted metabolomic approaches is
that they are mostly qualitative or at best, semi-quantitative. Therefore,
to obtain accurate quantification and fully validated metabolite anno-
tation, targeted metabolomic approaches must be employed.
790
Food Research International 116 (2019) 786–794
3.3. Targeted identification and quantification of phenolic metabolites
In order to both validate and determine the absolute quantitative
values of the phenolic metabolites putatively identified in the faecal
fermented nut samples, a targeted, high-resolution metabolomic
method was used. Table 2 displays the phenolic metabolites identified
and quantified via targeted UHPLC-Orbitrap mass spectrometry in the
different nut samples during the in vitro fermentation process. The
phenolic metabolites detected and quantified include 3,4-dihydrox-
yphenylacetic acid, 4-hydroxybenzoic acid, hippuric acid, caffeic acid,
protocatechuic acid and protocatechuic aldehyde. These were selected
as being representative of those polyphenolic metabolites arising from
the degradation of the native, higher-molecular-weight compounds
followed by the further processing by the colonic microbiota. Interest-
ingly, hazelnuts, pistachios, almonds and walnuts showed an increase
in 3,4-dihydroxyphenylacetic acid over time, especially from 8 to 24 h
of in vitro faecal fermentation. The maximum concentration was re-
corded in pistachios (i.e., 6.60 nmol/g DM), followed by almonds (i.e.,
6.16 nmol/g DM). Another phenolic metabolite detected in all nut-fer-
mented samples was protocatechuic acid, which belongs to the phenolic
class of hydroxybenzoic acids. The highest content of the proto-
catechuic acid (i.e., 11.86 nmol/g DM) was found in peanuts after 8 h of
in vitro faecal fermentation, with somewhat lower values and different
trends over time recorded for hazelnuts, almonds and walnuts
(Table 2). Interestingly, both 4-hydroxybenzoic acid and hippuric acid
were measurable in faecal fermented samples. Hippuric acid was
quantified only in peanuts, hazelnuts and walnuts after both 8 and 24 h
of fermentation. This compound is a glycine conjugate of benzoic acid
and is commonly found in plasma and urine as marker of the con-
sumption of polyphenol-rich foods (Mosele et al., 2015). Hippuric acid
is also produced from the metabolism of many dietary components
including phenylalanine, quinic acid, shikimic acid, chlorogenic acid
and catechin (Lees et al., 2013). Therefore, the detection and quanti-
fication of hippuric acid in faecal fermented nut aliquots could be due
to the microbial transformation of phenolic acids such as 4-hydro-
xybenzoic acid or quinic acid (Mosele et al., 2015). Relative to the other
metabolites, caffeic acid and protocatechuic aldehyde, gave notably
lower values, with caffeic acid being found only in peanuts, hazelnuts
and walnuts (Table 2). Interestingly, the results obtained by untargeted
metabolomics (UHPLC-QTOF) were in general agreement with those
obtained by our targeted approach (UHPLC-Orbitrap). According to our
untargeted data, 4-hydroxybenzoic acid and protocatechuic aldehyde
G. Rocchetti et al. Food Research International 116 (2019) 786–794

Table 1
Phenolic compounds identified by VIP (Variable Importance in Projection) analysis following OPLS-DA model, during the in vitro faecal fermentation of edible nuts.
Compounds are provided together with VIP scores (measure of variable's importance in the OPLS-DA model), chemical structure and colonic pathways in which are
involved.

791
G. Rocchetti et al. Food Research International 116 (2019) 786–794

Table 2
Quantitative values of phenolic metabolites detected during in vitro faecal fermentation of edible nuts. Data are presented as mean value (nmol phenolic compounds/
g dry matter incubated) ± standard deviation (n = 3).
Matrix Timepoint (h) 3,4-DHPA 4-HBA HA CA PA PAL

(nmol/g DM) (nmol/g DM) (nmol/g DM) (nmol/g DM) (nmol/g DM) (nmol/g DM)

Peanuts 8 6.59 ± 0.06 0.79 ± 0.01 2.42 ± 0.02 1.75 ± 0.02 11.86 ± 0.12 0.11 ± 0.00
24 5.65 ± 0.05 nd 2.72 ± 0.03 0.27 ± 0.00 4.35 ± 0.04 0.35 ± 0.00
48 nd nd nd nd nd 0.81 ± 0.00
Hazelnuts 8 4.31 ± 0.04 0.21 ± 0.00 2.29 ± 0.02 1.73 ± 0.02 3.17 ± 0.03 0.15 ± 0.00
24 5.89 ± 0.06 nd 3.25 ± 0.03 0.52 ± 0.00 2.64 ± 0.03 0.26 ± 0.00
48 3.11 ± 0.03 nd nd nd 4.70 ± 0.05 0.53 ± 0.01
Pistachios 8 4.00 ± 0.04 0.71 ± 0.01 nd nd 2.51 ± 0.02 0.37 ± 0.00
24 6.60 ± 0.06 nd nd nd 3.23 ± 0.03 0.49 ± 0.00
48 nd nd nd nd nd 1.08 ± 0.01
Almonds 8 6.05 ± 0.06 0.31 ± 0.00 nd nd 5.89 ± 0.06 0.38 ± 0.00
24 6.14 ± 0.06 nd nd nd 4.06 ± 0.04 0.53 ± 0.01
48 4.91 ± 0.05 nd nd nd 4.53 ± 0.04 0.35 ± 0.00
Walnuts 8 4.11 ± 0.04 0.31 ± 0.00 1.98 ± 0.02 1.66 ± 0.02 4.95 ± 0.05 0.30 ± 0.00
24 4.97 ± 0.05 nd 3.48 ± 0.03 0.59 ± 0.00 3.32 ± 0.03 0.20 ± 0.00
48 nd nd nd nd 4.51 ± 0.04 0.34 ± 0.00

Abbreviations: 3,4-DHPA (3,4-Dihydroxyphenylacetic acid); 4-HBA (4-Hydroxybenzoic acid); HA (Hippuric acid); CA (Caffeic acid); PA (Protocatechuic acid); PAL
(Protocatechuic aldehyde); DM (dry matter incubated); nd (not detected).

were among the most discriminating compounds (i.e., VIP markers) regard, Aura et al. (2002) detected the highest concentration of 3,4-
characterizing the changes of the phenolic profile observed during dihydroxyphenylacetic acid within 2 h of in vitro anaerobic faecal in-
fermentation. Additionally, 4-hydroxyphenylacetic and benzoic acid cubation (a time point not included in the present study) using pure
(outlined by VIP) are the final products of the dehydroxylation of 3,4- flavonoids (i.e., rutin, isoquercitrin and quercetin), while Kay et al.
dihydroxyphenylacetic and 4-hydroxybenzoic acids, respectively. (2017) suggested the direct formation of protocatechuic acid from di-
Overall, these findings corroborate previous results (Rocchetti, hydrocaffeic acid via β-oxidation reactions. Other minor colonic cata-
Chiodelli, et al., 2018) where an increase in bioaccessibility for some bolites, such as hippuric acid, p-coumaric acid, vanillic acid, 4-hydro-
phenolic classes (i.e., phenolic acids, tyrosols and alkylphenols) was xybenzoic and other benzoic acids have been associated with the in vivo
observed using the same edible nuts subjected to an in vitro gastro- colonic metabolism of flavan-3-ols (Mosele et al., 2015). With regard to
intestinal process. While the results reported here are specific to tree the equol and enterodiol derivatives reported by our VIP approach, they
nuts, our data appears to align with colonic fermentation studies re- appear to be derived by isoflavone and lignan colonic pathways, re-
ported for other kinds of fruit. For instance, Low et al. (2016) reported spectively (Mosele et al., 2015). Finally, the high levels of bound tyrosol
an increase in 4-hydroxyphenylacetic acid within 4–8 h of in vitro co- equivalents that characterize all raw nuts, appear to be associated with
lonic fermentation of masticated mango and banana, while Juániz et al. the formation of 4-hydroxyphenylacetic acid, after different dehy-
(2016) analysed the catabolism of polyphenols found in raw and droxylation and oxidation reactions (Mosele et al., 2015).
cooked green pepper, showing the formation of 3 catabolites during
24 h of in vitro fermentation, namely 3,4-dihydroxybenzoic acid (i.e.,
4. Conclusions
protocatechuic acid), dihydrocaffeic acid and 3-(3′-hydroxyphenyl)
propionic acid.
In this work, two different high-resolution metabolomic approaches
The data from both the untargeted and targeted metabolomic stu-
were applied to investigate the catabolic fate of phenolic compounds
dies allowed us to propose specific catabolic pathways for microbial
during in vitro digestion and faecal fermentation processes of edible
colonic degradation of polyphenols, as shown in Fig. 3. The first cata-
nuts. The nuts we used in this study were found to be very rich in
bolic step outlined involves the C-ring fission of the flavan-3-ol-ske-
phenolic compounds, with the highest content recorded in peanuts and
leton, leading to the formation of 5-(3′,4′-dihydroxyphenyl)-λ-valer-
hazelnuts. Overall, multivariate statistics using data from untargeted
olactone (i.e., one of the compounds possessing the highest
metabolic profiling showed a clear modification of the phenolic meta-
discrimination potential in our OPLS-DA model) (Mosele et al., 2015).
bolites over time (from 0 up to 48 h of in vitro fermentation). This work
Indeed, nuts are a rich source of proanthocyanidins (i.e., flavan-3-ol
highlighted the most discriminating compounds and the colonic path-
oligomers linked through carbon‑carbon bonds) and in particular, nut
ways involved. To complement the untargeted approach, which was
proanthocyanidins mainly consist of (+)-catechin together with
aimed at putatively identifying possible phenolic metabolites, a tar-
(−)-epicatechin and epigallocatechin derivatives. These are the pre-
geted metabolomic approach was used to positively identify and
dominant flavan-3-ols in almonds, hazelnuts, peanuts, pecans and pis-
quantify some of the most common phenolic metabolites including 3,4-
tachios (Bolling et al., 2011). Therefore, looking at the raw phenolic
dihydroxyphenylacetic acid, protocatechuic acid and hippuric acid. An
profile in edible nuts provided by untargeted metabolomics (UHPLC-
initial increase followed by a slow decline was observed for some
QTOF), allowed us to postulate the mechanisms of degradation of these
phenolic metabolites over time, during faecal fermentation. The results
compounds. Subsequently, as described in literature (Mosele et al.,
obtained by targeted and untargeted metabolomics were in good
2015; Serra et al., 2012), several oxidation reactions must occur to
agreement, suggesting that these high-resolution approaches are very
produce lower-molecular-weight phenolic acids such as 3,4-dihydrox-
powerful for investigating the fate of polyphenols in the large intestine.
yphenylacetic acid. Interestingly, 3,4-dihydroxyphenylacetic acid is one
The complementarity of these approaches, together with the sensitivity
of the most important phenolic intermediates involved in the genera-
and specificity provided by high-resolution MS, allowed us to achieve a
tion of other metabolites. In particular, a dehydroxylation reaction of
comprehensive picture of the transformation processes originating from
3,4-dihydroxyphenylacetic acid is hypothesized to lead to 4-hydro-
in vitro digestion and fermentation. However, the positive health effects
xyphenylacetic acid, while an α-oxidation is proposed for the formation
that these phenolic metabolites may have are still not clearly defined.
of protocatechuic acid (Mosele et al., 2015; Juániz et al., 2016). In this
Therefore, further studies are necessary in order to better understand

792
G. Rocchetti et al.
Fig. 3. Proposed catabolic pathways for polyphenols microbial degradation in the colon after gastrointestinal digestion of edible nuts. DeOH = dehydroxylation;
DeH = dehydrogenation; DeMet = demethylation.
the real mechanisms of action that these metabolites may have within
the colon or within other parts of the body.
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.foodres.2018.09.012.
Conflict of interest
The authors declare no conflict of interest.
Acknowledgements
GR was the recipient of a Ph.D. fellowship from the Università
Cattolica del Sacro Cuore (UCSC, Piacenza, Italy). The authors thank
the “Romeo ed Enrica Invernizzi foundation” for its kind support to the
metabolomic platform at Università Cattolica del Sacro Cuore.
References
Acosta-Estrada, B. A., Gutiérrez-Uribe, J. A., & Serna-Saldívar, S. O. (2014). Bound
phenolics in foods, a review. Food Chemistry, 152, 46–55.
Alasalvar, C., & Bolling, B. W. (2015). Review of nut phytochemicals, fat-soluble bioac-
tives, antioxidant components and health effects. British Journal of Nutrition, 113,
S68–S78.
Aura, A. M., O'Leary, K. A., Williamson, G., Ojala, M., Bailey, M., Puupponen-Pimia, R., ...
Poutanen, K. (2002). Quercetin derivatives are deconjugated and converted to hy-
droxyphenylacetic acids but not methylated by human fecal flora in vitro. Journal of
Agricultural and Food Chemistry, 50(6), 1725–1730.
Bohn, T., McDougal, G. J., Alegría, A., Alminger, M., Arrigoni, E., Aura, A.-M., ... Santos,
C. N. (2015). Mind the gap – Deficits in our knowledge of aspects impacting the
bioavailability of phytochemicals and their metabolites – A position paper focusing
on carotenoids and polyphenols. Molecular Nutrition & Food Research, 59, 1307–1323.
Bolling, B. W., Chen, C.-Y. O., McKay, D. L., & Blumberg, J. B. (2011). Tree nut phyto-
chemicals: composition, antioxidant capacity, bioactivity, impact factors. A sys-
tematic review of almonds, Brazils, cashews, hazelnuts, macadamias, pecans, pine
nuts, pistachios and walnuts. Nutrition Research Reviews, 24, 244–275.
Bruzzese, E., Callegari, M. L., Raia, V., Viscovo, S., Scotto, R., et al. (2014). Disrupted
intestinal microbiota and intestinal inflammation in children with cystic fibrosis and
its restoration with lactobacillus GG: A randomized clinical trial. PLoS One, 9(2),
e87796. https://doi.org/10.1371/journal.pone.0087796.
793
Food Research International 116 (2019) 786–794
Cardona, F., Andrés-Lacueva, C., Tulipani, S., Tinahones, F. J., & Queipo-Ortuno, M. I.
(2013). Benefits of polyphenols on gut microbiota and implications on human health.
Journal of Nutritional Biochemistry, 24, 1415–1422.
Castro-Puyana, M., Pérez-Míguez, R., Montero, L., & Herrero, M. (2017). Application of
mass spectrometry-based metabolomics approaches for food safety, quality and tra-
ceability. Trends in Analytical Chemistry, 93, 102–118.
van Duynhoven, J., Vaughan, E. E., Jacobs, D. M., Kemperman, R. A., van Velzen, E. J. J.,
Gross, G., ... Van de Wiele, T. (2011). Metabolic fate of polyphenols in the human
superorganism. PNAS, 108, 4531–4538.
Espín, J. C., González-Sarrías, A., & Tomás-Barberán, F. A. (2017). The gut microbiota: A
key factor in the therapeutic effect of (poly)phenols. Biochemical Pharmacology, 139,
82–93.
Farrés, M., Platikanov, S., Tsakovski, S., & Tauler, R. (2015). Comparison of the variable
importance in projection (VIP) and of the selectivity ratio (SR) methods for variable
selection and interpretation. Journal of Chemometrics, 29, 528–536.
Giuberti, G., Gallo, A., Moschini, M., & Masoero, F. (2013). In vitro production of short-
chain fatty acids from resistant starch by pig faecal inoculum. Animal, 7(9),
1446–1453.
Granato, D., Shahidi, F., Wrolstad, R., Kilmartin, P., Melton, L. D., Hidalgo, F. J., ...
Finglas, P. (2018). Antioxidant activity, total phenolics and flavonoids contents:
Should we ban in vitro screening methods? Food Chemistry, 264, 471–475.
Grundy, M. M.-L., Lapsley, K., & Ellis, P. R. (2016). A review of the impact of processing
on nutrient bioaccessibility and digestion of almonds. International Journal of Food
Science and Technology, 51(9), 1937–1946.
Juániz, I., Ludwig, I. A., Bresciani, L., Dall'Asta, M., Mena, P., Del Rio, D., ... de Pena, M. P.
(2016). Catabolism of raw and cooked green pepper (Capsicum annuum) (poly)phe-
nolic compounds after simulated gastrointestinal digestion and faecal fermentation.
Journal of Functional Foods, 27, 201–213.
Kay, C. D., Pereira-Caro, G., Ludwig, I. A., Clifford, M. N., & Crozier, A. (2017).
Anthocyanins and flavonones are more bioavailable than previously perceived: A
review of recent evidence. Annual Review of Food Science and Technology, 8, 155–180.
Lamuel-Raventos, R. M., & Onge, M.-P., St. (2017). Prebiotic nut compounds and human
microbiota. Critical Reviews in Food Science and Nutrition, 57(14), 3154–3163.
Lees, H. J., Swann, J. R., Wilson, I. D., Nicholson, J. K., & Holmes, E. (2013). Hippurate:
The natural history of a mammalian−microbial cometabolite. Journal of Proteome
Research, 12, 1527–1546.
Low, D. Y., Hodson, M. P., Williams, B. A., D'Arcy, B. R., & Gidley, M. J. (2016). Microbial
biotransformation of polyphenols during in vitro colonic fermentation of masticated
mango and banana. Food Chemistry, 207, 214–222.
Minekus, M., Alminger, M., Alvito, P., Balance, S., Bohn, T., Bourlieu, C., ... Brodkorb, A.
(2014). A standardized static in vitro digestion method suitable for food – An inter-
national consensus. Food & Function, 5(6), 1113–1124.
Miragoli, F., Federici, S., Ferrari, S., Minuti, A., Rebecchi, A., Bruzzese, E., & Callegari, M.
L. (2016). Impact of cystic fibrosis disease on archaea and bacteria composition of gut
microbiota. FEMS Microbiology Ecology, 93(2), fiw230. https://doi.org/10.1093/
G. Rocchetti et al.
femsec/fiw230.
Mosele, J. I., Macià, A., & Motilva, M.-J. (2015). Metabolic and microbial modulation of
the large intestine ecosystem by non-absorbed diet phenolic compounds: A review.
Molecules, 20, 17429–17468.
Ozdal, T., Sela, D. A., Xiao, J., Boyacioglu, D., Chen, F., & Capanoglu, E. (2016). The
reciprocal interactions between polyphenols and gut microbiota and effects on
bioaccessibility. Nutrients, 8, 78. https://doi.org/10.3390/nu8020078.
Parkar, S. G., Trower, T. M., & Stevenson, D. E. (2013). Fecal microbial metabolism of
polyphenols and its effects on human gut microbiota. Anaerobe, 23, 12–19.
Payne, A. N., Zihler, A., Chassard, C., & Lacroix, C. (2012). Advances and perspectives in
in vitro human gut fermentation modeling. Trends in Biotechnology, 30(1), 17–25.
Rocchetti, G., Chiodelli, G., Giuberti, G., & Lucini, L. (2018). Bioaccessibility of phenolic
compounds following in vitro large intestine fermentation of nuts for human con-
sumption. Food Chemistry, 245, 633–640.
Rocchetti, G., Giuberti, G., & Lucini, L. (2018). Gluten-free cereal-based food products:
The potential of metabolomics to investigate changes in phenolics profile and their in
vitro bioaccessibility. Current Opinion in Food Science, 22, 1–8.
Rocchetti, G., Lucini, L., Chiodelli, G., Giuberti, G., Gallo, A., Masoero, F., & Trevisan, M.
(2017). Phenolic profile and fermentation patterns of different commercial gluten-
free pasta during in vitro large intestine fermentation. Food Research International, 97,
78–86.
Rombouts, C., Hemeryck, L. Y., Van Hecke, T., De Smet, S., De Vos, W. H., & Vanhaecke,
L. (2017). Untargeted metabolomics of colonic digests reveals kynurenine pathway
metabolites, dityrosine and 3-dehydroxycarnitine as red versus white meat dis-
criminating metabolites. Scientific Reports, 7, 42514.
Rothwell, J. A., Knaze, V., & Zamora-Ros, R. (2017). Polyphenols: Dietary assessment and
role in the prevention of cancers. Current opinion in clinical nutrition & metabolic care.
794
Food Research International 116 (2019) 786–794
vol. 20. Current opinion in clinical nutrition metabolic care (pp. 512–521).
Rothwell, J. A., Perez-Jimenez, J., Neveu, V., Medina-Remón, A., M'Hiri, N., García-
Lobato, P., ... Scalbert, A. (2013). Phenol-Explorer 3.0: A major update of the Phenol-
Explorer database to incorporate data on the effects of food processing on polyphenol
content. Database. https://doi.org/10.1093/database/bat070 2013: article ID
bat070.
Roura, E., Koopmans, S.-J., Lallès, J.-P., Le Huerou-Luron, I., de Jager, N., Schuurman, T.,
& Val-Laillet, D. (2016). Critical review evaluating the pig as a model for human
nutritional physiology. Nutrition Research Reviews. https://doi.org/10.1017/
S0954422416000020.
Scalbert, A., Brennan, L., Manach, C., Andres-Lacueva, C., Dragsted, L. O., Draper, J., ...
Wishart, D. S. (2014). The food metabolome: A window over dietary exposure. The
American Journal of Clinical Nutrition, 99(6), 1286–1308.
Serra, A., Macià, A., Romero, M.-P., Reguant, J., Ortega, N., & Motilva, M.-J. (2012).
Metabolic pathways of the colonic metabolism of flavonoids (flavonols, flavones and
flavanones) and phenolic acids. Food Chemistry, 130, 383–393.
Stearns, J. C., Lynch, M. D. J., Senadheera, D. B., Tenenbaum, H. C., Goldberg, M. B.,
Cvitkovitch, D. G., Croitoru, K., Moreno-Hagelsieb, G., & Neufeld, J. D. (2011).
Bacterial biogeography of the human digestive tract. Scientific Reports, 1, 170.
Tarko, T., Duda-Chodak, A., & Zajac, N. (2013). Digestion and absorption of phenolic
compounds assessed by in vitro simulation methods. A review. Roczniki Panstwowego
Zakladu Higieny, 64(2), 79–84.
Vinson, J. A., & Cai, Y. (2012). Nuts, especially walnuts, have both antioxidant quantity
and efficacy and exhibit significant potential health benefits. Food & Function, 3(2),
134–140.
Worley, B., & Powers, R. (2013). Multivariate analysis in metabolomics. Current
Metabolomics, 1(1), 92–107.