ORIGINAL PAPER

Barcoded pyrosequencing analysis of the microbialcommunity in a simulator of the human gastrointestinaltract showed a colon region-specific microbiota modulationfor two plant-derived polysaccharide blends

Massimo Marzorati • Lois Maignien • An Verhelst • Gabriela Luta •

Robert Sinnott • Frederiek Maarten Kerckhof • Nico Boon •

Tom Van de Wiele • Sam Possemiers

Received: 10 July 2012 / Accepted: 18 September 2012! Springer Science+Business Media Dordrecht 2012

Abstract The combination of a Simulator of the

Human Intestinal Microbial Ecosystem with ad hocmolecular techniques (i.e. pyrosequencing, denaturing

gradient gel electrophoresis and quantitative PCR)

allowed an evaluation of the extent to which two plantpolysaccharide supplements could modify a complex

gut microbial community. The presence of Aloe veragel powder and algae extract in product B as compared

to the standard blend (product A) improved its

fermentation along the entire simulated colon. Thepotential extended effect of product B in the simulated

distal colon, as compared to product A, was confirmed

by: (i) the separate clustering of the samples before andafter the treatment in the phylogenetic-based dendro-

gram and OTU-based PCoA plot only for product B;

(ii) a higher richness estimator (?33 vs. -36 % of

product A); and (iii) a higher dynamic parameter(21 vs. 13 %). These data show that the combination of

well designed in vitro simulators with barcoded

pyrosequencing is a powerful tool for characterizingchanges occurring in the gut microbiota following a

treatment. However, for the quantification of low-abundance species—of interest because of their

relationship to potential positive health effects (i.e.

bifidobacteria or lactobacilli)—conventional molecu-lar ecological approaches, such as PCR–DGGE and

qPCR, still remain a very useful complementary tool.

Keywords Gastrointestinal resource management

(GRM) ! Pyrosequencing ! Ambrotose ! SHIME !Intestinal bacteria

Introduction

In recent years, the increased use of moleculartechniques in the field of gastrointestinal microbial

ecology has allowed a much deeper sampling of the

microbial community diversity. It became possible toperform comparative analysis of the human gut

microbiota, investigating the association of the mic-

robiota with different diets in human populations, andto describe the minimal gut metagenome and the

minimal gut bacterial genome in terms of functions

present in all individuals and most bacteria, respec-tively (Andersson et al. 2008; De Filippo et al. 2010;

Electronic supplementary material The online version ofthis article (doi:10.1007/s10482-012-9821-0) containssupplementary material, which is available to authorized users.

M. Marzorati ! L. Maignien ! F. M. Kerckhof !N. Boon ! T. Van de Wiele (&) ! S. PossemiersLaboratory of Microbial Ecology and Technology(LabMET), Ghent University, Coupure Links 653,9000 Ghent, Belgiume-mail: tom.vandewiele@ugent.be

A. Verhelst ! S. PossemiersProDigest, Technologiepark 3, 9052 Ghent, Belgium

G. Luta ! R. SinnottMannatech Inc, 600 S. Royal Lane, Suite 200, Coppell,TX 75019, USA

123

Antonie van Leeuwenhoek

DOI 10.1007/s10482-012-9821-0

Qin et al. 2010). The application of these techniquesto in vivo studies created new opportunities to evaluate

the effect of functional ingredients (i.e. pre- and

probiotics) in relation to human health. However, thepossibility of understanding the mechanism of action

of these ingredients in the different areas of the

gastrointestinal tract (GIT) is frequently limited due toclear sampling issues (Marzorati et al. 2009). In this

respect, the application of well-designed in vitro

continuous models of the GIT may represent a usefulcomplementary tool to study the intestinal microbial

processes in terms of gastrointestinal resource man-

agement (GRM), i.e. the management of the complexgut microbiota and its metabolism with the aim of

improving the host health (Manning and Gibson 2004;

Gibson et al. 2004; Possemiers et al. 2009).We made use of a Simulator of the Human

Intestinal Microbial Ecosystem (SHIME") to assess

the effect of two blends of a commercially availableplant-polysaccharide supplement—Ambrotose" com-

plex and Advanced Ambrotose" powder (Mannatech

Inc., Coppell, Texas, USA)—on the microbial ecologyof the different areas of the GIT (Marzorati et al.

2010). The SHIME" is a dynamic model of the human

gastrointestinal tract used to study physicochemical,enzymatic and microbial parameters in the GIT in a

controlled in vitro setting. It has been extensively used

for more than 15 years and has been validated withseveral in vivo experiments (Molly et al. 1994;

Marzorati et al. 2009). This technology platform

allows to evaluate the effect of repeated doses of

selected food ingredients under gut representativeenvironmental conditions (Van de Wiele et al. 2007;

Sanchez et al. 2009; Grootaert et al. 2009; Marzorati

et al. 2010).The test supplements were already shown to be able

to affect serum glycosylation profiles (Alavi et al.

2011a, b) and to induce an anti-inflammatory effectagainst sulfate sodium-induced colitis in rats (Koetz-

ner et al. 2010; Ramberg et al. 2010). Ambrotose"

complex was shown to support cognitive health (Bestet al. 2010) in clinical trials. Moreover, it was already

shown that the intake of the supplements influenced

the composition and the metabolism of the residentmicrobial community in the different areas of the colon

(i.e. ascending, transverse and descending colon)

(Table 1). By means of a size exclusion chromatog-raphy analysis, it was shown that the fermentability of

Ambrotose" complex was limited to the proximal

colon [ascending (AC) ? transverse (TC)] while thatof Advanced Ambrotose" powder was extended to the

entire colon, as shown by the residual high-molecular

weight fraction still available for fermentation afterpassage in the simulated transverse colon (Fig. S1)

(Marzorati et al. 2010).

The aim of the present work was to investigatethe effect of the test products on the structure and

composition of the microbial community by means of

16S rRNA genes-based pyrosequencing, comparingthe obtained results with the denaturing gradient

gel electrophoresis (DGGE) and quantitative PCR

(qPCR)-based approach previously applied (Marzorati

Table 1 Composition and summary of the long-term effects of Ambrotose" complex and Advanced Ambrotose" powder on theactivity and composition of the colon’s microbial community, as compared to the control period (adapted from Marzorati et al. 2010)

Ambrotose" complex Advanced Ambrotose" powder

Composition Larch arabinogalactan (83.7 % galactose), Aloe vera gelextract (15.6 % galacturonic acid, 14.8 % mannose),glucosamine HCl, and tragacanth gum (9.9 %arabinose, 7.8 % xylose), rice starch (97 % glucose),ghatti gum (26 % arabinose, 16.6 % galactose)

Basic formula of Ambrotose" complex in which A. veraextract was substituted with A. vera gel powder (12 %glucose, 9.2 % mannose) and wakame algae extract(32.1 % fucose) was added

Fermentability Limited to the proximal colon (AC ? TC) Extended to the entire colon

SCFAproduction

Increased acetate and propionate, decreased butyrate Starch-like fermentation with no mayor changes

Ammoniumproduction

Slight increase Slight increase

Bifidobacteria Statistically significant increase in AC ? qualitativechanges

Statistically significant increase in the entirecolon ? qualitative changes

Lactobacilli No major quantitative/qualitative changes No major quantitative/qualitative changes

AC ascending colon, TC transverse colon, DC descending colon

Antonie van Leeuwenhoek

123

et al. 2010). The generation of a huge amount of data—as occurs with pyrosequencing—can also induce some

problems related to the final ecological interpretation

and the underpinning of what occurs to the structureand composition of a given microbial community. For

the first time, we will adapt some ecological param-

eters—originally developed to provide an interpreta-tion to molecular fingerprintings (Marzorati et al.

2008)—to be used on data generated with barcoded

pyrosequencing. The toolbox, based on range-rich-ness, dynamics and community organization, has been

theoretically re-engineered and adapted to provide a

universal platform, allowing to compare these data interms of richness and evenness (Read et al. 2011).

Materials and methods

SHIME study and sample collection

In Marzorati et al. (2010), two SHIME" experiments

have been described, which aimed to compare theeffect of Ambrotose" complex (product A in SHIME

1) and Advanced Ambrotose" powder (product B in

SHIME 2). The full experimental design is describedin Marzorati et al. (2010).

Samples for DNA analyses were collected once a

week from each colon compartment of the SHIME,simulating respectively AC, TC and descending (DC)

colon (Van de Wiele et al. 2007). The metagenomic

DNA was extracted from the samples using the CTABprotocol (Boon et al. 2003). qPCR for the Firmicutes

and Bacteroidetes phyla were conducted according

to the protocols described by Guo et al. (2008).Samples for pyrosequencing analysis were collected

at the end of the control and treatment periods from

each colon compartment. The resulting 12 sampleswere labeled as follows: S1 (product A) or S2 (product

B) to indicate the SHIME system; AC, TC or DC to

indicate respectively ascending, transverse anddescending colon; and, finally, ‘c’ for the control and

‘t’ for the treatment period.

Pyrosequencing and pre-processing

of the sequences

The V5 and V6 regions of the 16S rRNA genes of

each sample were amplified using a primer set corre-sponding to primers 784F and 1061R described by

Andersson et al. (2008). The forward primer containedthe sequence of the Titanium A adaptor (50-CCATCT

CATCCCT-GCGTGTCTCCGACTCAG-30) and a

barcode sequence. The specific PCR conditions arethose reported in De Filippo et al. (2010). Amplicons

were then cleaned using the HighPure Cleanup kit

(Roche) and DNA concentrations were determinedusing the Quant-iT PicoGreen dsDNA reagent and kit

(Invitrogen). Both procedures were conducted accord-

ing to the manufacturer’s instructions. The final poolsof cleaned amplicons combined in equimolar ratios

into a single tube were prepared as described in De

Filippo et al. (2010). Pyrosequencing was carriedout using primer A on a 454 Life Sciences Genome

Sequencer FLX instrument (Roche), following tita-

nium chemistry by DNAVision (Charleroi, Belgium).The pyrosequencing of the 12 amplicon libraries

produced a total of 134,844 16S rDNA reads.

Processing and clustering analysis of these sequenceswas carried out using the Mothur software (Schloss

et al. 2009; Schloss 2010) and the external tools

implemented in Mothur. Primers and MID sequenceswere removed from the 50 end of the sequences.

Sequences that did not meet the following quality

criteria were filtered out of the dataset: no ambiguousbase calling, at most 8 homopolymers, minimum

length of 100 bp and at most 2 mismatches in the

primer sequence. The resulting dataset contained68,551 high quality sequences. Sequences were

trimmed at the 30 end of a 50 bp window when the

average quality score within this window droppedbelow 35. The dataset was then dereplicated for

downstream analysis, resulting in 16,034 unique

sequences. These were then aligned using the align-ment tool included in Mothur (Schloss 2009; De Santis

et al. 2006) and the Silva core alignment as a template

(Schloss 2009; Pruesse et al. 2007). Sequences that didnot span the V5 region (i.e. position 22,598–25,298 in

the SILVA alignment) were filtered out, and longer

reads were trimmed down to fit this alignment length.Rare OTUs were merged with abundant OTUs when

differing by only a single mistmatch, according toHuse et al. (2010). Putative chimeric sequences were

filtered out using Uchime (Edgar et al. 2011; Schloss

et al. 2009). This preprocessing resulted in a highquality dataset of filtered and aligned sequences with

a mean average of 1,985 sequences per library. In order

to avoid possible biases due to library size differences(Gihring et al. 2012), these were subsampled to the

Antonie van Leeuwenhoek

123

smallest library size, i.e. 790 reads. Subsampling hasbeen performed using the Mothur package.

Sequence analysis and phylogenetic classification

The high-quality dataset was analyzed with three

different methods. In a first ‘‘OTU-based’’ approach,sequences were clustered into OTU (97 % similarity

threshold). OTU composition was used to evaluate the

alpha diversity (observed richness, inverse Simpsonindex and rarefaction curves—Fig. S2) within each

sample, and the Jaccard similarity index for beta

diversity estimation between the 12 samples. Thelatter was used for a principal coordinate analysis of

the variance (PCoA). The dataset was also evaluated

using a ‘‘phylogenetic-based’’ approach: a dendro-gram was made by UPGMA based on the sample

pairwise comparison using Jaccard presence/absence

index, constructed with the Clearcut software basedon the pairwise distance matrix (Evans et al. 2006;

Sheneman et al. 2006). Node support was conducted

with a jackkniffed resampling (based on 100 trees)with qiime (www.qiime.org). Finally, each sequence

was assigned at taxonomy down to the genus level

using ribosomal database project taxonomic classifi-cation (Wang et al. 2007). When required, the closest

relative of sequences representing dominant OTUs in

Gene Bank nucleotide collection were retrieved usingBLASTn (http://blast.ncbi.nlm.nih.gov/).

Ecological interpretation based on microbialresource management indices

The phylotype data were analyzed at the family andgenus level as shown in Marzorati et al. (2008) and

Read et al. (2011). Briefly, the richness parameter

(R) corresponded to the number of observed OTUs foreach sample. The dynamics parameter (Dy)—how

deeply a microbial community changes—was based

on the calculation of the similarity values betweensamples belonging to the same simulated colon

compartment before and after the treatment, inaccordance to the following formula: % change =

100 % - similarity. The matrix of similarity was

based on the Pearson product-moment correlation, asdescribed in Marzorati et al. (2008). In order to

graphically represent the evenness of the bacterial

communities, the Pareto–Lorenz (PL) distribution

curves (Lorenz, 1905) were set up, based on thephylotypes distribution, and the Community organi-

zation parameter (Co) (Gini coefficient 9 100) was

calculated (Marzorati et al. 2008). The cumulative Cocurve was calculated as follows. Consider a dataset of

N phylotype classes (family or genus), sall phylotype

classes are first sorted from high to low abundance(with phylotype-1 having the most sequences and

phylotype-N the fewest). Secondly, the different Co

(C0i, with i = 1 to N) values are calculated for a

progressing window of phylotypes, starting from 1 to

N. These C0i values can then be plotted against the

amount of classes (i) (Read et al. 2011). The values ofthe treatment period are subtracted to those of the

control period thus providing the final DCo curve

(Fig. S6). This analysis pipeline was implemented inthe open-source statistical programming language

R (www.r-project.org) and graphical visualizations

were made using the ggplot2 package (http://had.co.nz/ggplot2/).

Results

In-depth analysis of the colonic microbialcommunity by pyrosequencing

Samples collected from each colon compartment, atthe end of the control and treatment periods, were

subjected to multiplex pyrosequencing of the V5 and

V6 hypervariable regions of 16S rRNA gene. Morethan 92.2 % of the total reads in the dataset containing

all of the samples belonged to Actinobacteria

(1.1 ± 2.0 %), Bacteroidetes (56.0 ± 12.9 %), Firmi-cutes (29.8 ± 9.3 %), and Proteobacteria (5.3 ± 4.7 %).

On average, Bacteroidetes and Firmicutes—the most

abundant bacterial phyla in the human gut microbi-ota—accounted for 85.8 ± 8.7 % of the total

sequences. Relevant differences were found in the

proportions of the four phyla following the treatmentwith both dietary supplements (Fig. 1). Product A

induced an increase in the concentration of Bacteroi-detes with a concomitant decrease of Firmicutes and/or

Proteobacteria in the ascending and descending colon.

In the transverse colon the trend was different: theconcentration of Bacteroidetes slightly decreased

while Actinobacteria and Firmicutes increased. Prod-

uct B led to an increase of Bacteroidetes and a decreaseof Firmicutes in all the simulated colon compartments.

Antonie van Leeuwenhoek

123

As a consequence, the ratio Firmicutes/Bacteroidetes

decreased for both products during the treatment as

also shown by the qPCR data, with the exception of theDC of the SHIME treated with product A (Fig. S3).

In both SHIME systems, the positive effect on the

Bacteroidetes phylum in the ascending colon wasmainly due to an increase in the concentration of

bacteria belonging to the Prevotellaceae family

(Fig. 2). In contrast, this increase was mainly relatedto bacteria of the Bacteroidaceae family in the other

simulated colon compartments. Another common

effect of the treatments with the two different products

was that the decrease in Firmicutes was mainly relatedto a decrease in the Veillonellaceae family, while the

slight increase in Proteobacteria in the simulated distal

colon was associated with an increase of Desulfovib-rionaceae (Fig. 2). Finally, the strong increase in

Actinobacteria in the TC of the SHIME treated with

product A was due to an increase of the Nocardiaceaefamily (mainly in the genus Rhodococcus). By means

Fig. 1 Relative abundance of the identified phyla across thesamples. Samples coding: S1 (Product A) or S2 (Product B) toindicate the SHIME system; AC, TC or DC to indicate

respectively ascending, transverse and descending colon; and,finally, ‘c’ for the control and ‘t’ for the treatment period

Fig. 2 Relative abundance of the identified families across thesamples. Groups: I Bacteroidetes, II Firmicutes, III Actinobac-teria, IV Lentisphaerae, V Proteobacteria, VI Verrucomicrobia.Samples coding: S1 (Product A) or S2 (Product B) to indicate

the SHIME system; AC, TC or DC to indicate respectivelyascending, transverse and descending colon; and, finally, ‘c’ forthe control and ‘t’ for the treatment period

Antonie van Leeuwenhoek

123

of the pyrosequencing approach it was possible to

observe only a slight increase in the concentration ofBifidobacterium spp. in the simulated distal colon of

the SHIME treated with product B (i.e. B. longum sub.

infantis and B. adolescentis). However, when observ-ing the data obtained with specific DGGE and qPCR

protocols (Fig. 3), both products stimulated an addi-

tional Bifidobacterium spp. in the microbial commu-nity following the treatment (arrow on the DGGE

profiles) although only Pruduct B led to a statisticallysignificant bifidogenic effect in all the simulated colon

compartments (Fig. 3b).

In order to compare the composition of the microbialcommunities in the different areas of the simulated

colon, before and after the treatment, the OTU-based

data were clustered in a dendrogram and in a PCoA(Fig. 4). The latter is a geometric technique that converts

a matrix of distances between points in a multivariate

space into a projection that maximizes the amount ofvariation along a series of orthogonal axes (Liu et al.

2007). The dendrogram showed a clear clustering along

the experiment (Fig. 4a): samples from the AC of bothSHIME systems (control and treatment) clustered

together (cluster I). Samples from TC and DC clustered

separately (with the exception of S1TCt which clusteredin the cluster I), forming some sub-clusters within the

cluster II. Samples collected during the control periodfrom the simulated descending colon of both SHIME

systems clustered together with sample S1DCt in sub-

cluster II.1. Also the samples from the simulatedtransverse colon of the two SHIME systems clustered

together (II.2). Finally, the samples derived from the

simulated distal colon (TC ? DC) of SHIME 2 formed aseparate cluster (II.3), as an indication of the selective

a

b

Fig. 3 qPCR and DGGE data generated from the DNA samplescollected during the control and treatment periods of the SHIMEsystems treated with product A (a) and B (b). The arrows on the

DGGE profiles indicate a Bifidobacterium sp. that couldspecifically benefit of the treatment with the test products(adapted from Marzorati et al. 2010)

Antonie van Leeuwenhoek

123

effect of the applied treatment. The pronounced effect ofthe treatment with product B in the simulated distal colon

of the SHIME 2 was confirmed by the PCoA plot, which

showed the respective control and treatment samplesfalling apart in the 2D graph (please confront S2TCc and

S2DCc with the respective treatments).

Calculation of different ecological parameters

on the dataset

In Marzorati et al. (2010), we applied a specific

toolbox with the aim of providing an ecological

interpretation to the raw data originated from conven-tional fingerprinting analyses (Marzorati et al. 2008).

However, these approaches cannot be used for large

datasets originated from pyrosequencing, and there-fore we re-engineered this toolbox as described by

Read et al. (2011). The calculated values are summa-

rized in Table 2. Evaluation of the richness parameterR, according to the observed OTUs, showed that

product A only induced an increase in richness

(?21 %) in the TC and a decrease (-36 %) in theDC. On the contrary, the positive effect of product B

on the richness parameter was mainly localized in the

simulated descending colon with a 33 % increase.The treatments also induced a change on the

relative abundance of the different bacterial groups,

as shown by the calculation of the Co values (based onGini coefficient) at level of phylum, family and genus

a

b

Fig. 4 Phylogenetic-based dendrogram (a) and 2D view of betadiversity OTU-based PCoA (b). Samples coding: S1 (product A)or S2 (product B) to indicate the SHIME system; AC, TC or DCto indicate respectively ascending, transverse and descendingcolon; and, finally, ‘c’ for the control and ‘t’ for the treatmentperiod. The symbol asterisk indicates that the specific node hada 75–100 % support according to a jack-kniffed resampling

Table 2 Data for the total number of sequences after filtering, coverage, Observed OTUs, inverse Simpson index, dynamics (Dy),and community organization (Co) at phylum, family and genus level

Index Product A Product B

S1ACc S1ACt S1TCc S1TCt S1DCc S1DCt S2ACc S2ACt S2TCc S2TCt S2DCc S2DCt

n. seqsa 2861 1299 1146 822 790 1269 1895 2605 1304 3330 1343 5163

Coverage 0.45 0.52 0.50 0.76 0.44 0.45 0.38 0.31 0.27 0.44 0.43 0.24

Obs. OTU 49 37 44 54 83 52 40 43 75 74 61 81

Inv. Simpson 6.51 2.81 6.82 9.71 12.08 5.63 7.04 3.34 7.49 3.49 6.63 6.73

Dy (family) 80 % 18 % 14 % 60 % 6 % 21 %

Dy (genus) 78 % 25 % 13 % 60 % 6 % 21 %

Co phylum 70.2 76.6 64.7 56.7 59.5 69.8 71.3 75.5 72.7 69.9 61.6 58.7

Co family 89.4 92.3 87.7 83.7 81.5 87.9 88.5 90.5 89.2 87.0 87.2 83.4

Co genus 92.0 93.6 90.1 86.0 85.2 91.1 91.6 92.7 91.3 90.0 90.5 87.3

S1 (product A) or S2 (product B) refer to the SHIME system; AC, TC or DC to indicate respectively ascending, transverse anddescending colon; ‘c’ indicates the control and ‘t’ the treatment period. The ecological meaning of these indices is reported in TableS1 in SIa All the libraries have been normalized to a number of sequences equal to 790

Antonie van Leeuwenhoek

123

(Table 2). The treatment with product A increased the

Co parameter—at all the investigated phylogenetic

levels—in the AC and DC (lower evenness) while itdecreased it in the TC. Product B increased the Co

parameter, at all the investigated phylogenetic levels

in the simulated ascending colon compartment. On thecontrary, it decreased it in the simulated distal colon

(TC ? DC). As the pyrosequencing analysis not only

provides the sequences of the dominant microorgan-isms, but also of all those less abundant microorgan-

isms which are normally not detected with the

fingerprinting techniques, this alters the shape of theLorenz curve towards complete dominance (Fig. S4).

Under these conditions, information is lost during the

calculation of the Pareto-Lorenz curve, especially atthe lower phylogenetic level (genus). In order to

overcome this limitation, we calculated the

cumulative Co parameter as proposed in Read et al.(2011). The resulting Fig. 5—calculated at genus

level—is the difference of the cumulative Co curve of

the treatment minus that of the control (the full figuresare reported in Fig. S5). According to Fig. 5, product

A led to a change in the structure of the microbial

community both in the dominant and in the low-abundant genera. On the contrary, product B mainly

increased the evenness among the most dominant

genera. Moreover, following the treatment with prod-uct B—as compared to product A—the main peak of

the curves points in the same direction, indicating that

the microbial sub-community, composed by the 10most dominant phylotypes (at genus level), was more

uneven in all colon compartments. More precisely,

with both treatments the enhanced phylotypes in theAC were mainly associated to the genus Prevotella.

On the contrary, bacteria belonging to Ruminococc-aceae, Porphyromonadaceae and mainly to the generaBacteroides and Mitsuokella decreased. In the simu-

lated distal colon (TC ? DC), the OTUs related to

Bacteroides spp. increased their importance in themicrobial community.

All the changes in richness and evenness of the

microbial communities at genus level in the differentareas of the colon were also reflected in the dynamic

(Dy) parameter, calculated on a similarity matrix

based on Pearson’s correlation coefficients (Marzoratiet al. 2008). Product A induced a main change in the

GIT microbial community in the AC (78 %) with a

decreasing effect along the colon (13 % in the DC).Product B, induced a Dy of 60 % in the AC (lower as

compared to product A) but, always compared to A,

had a higher impact in the DC (21 %) (Table 2).

Discussion

In vitro approaches to study the gastrointestinal tract

and intestinal microbial processes offer an excellentexperimental setup to study possible prebiotic prop-

erties of selected food ingredients. In fact, theapplication of well-designed continuous models

allows the in-depth study of the biological activity of

selected products in the gut under representativeenvironmental conditions. In this study, we applied

for the first time a 16S rRNA gene pyrosequencing

approach to a SHIME experiment in order to inves-tigate the effect of two polysaccharide supplements

a

b

Fig. 5 DCo (community organization parameter) calculated atgenus level and expressed as the difference of the cumulative Cocurve of the treatment minus that of the control. The ecologicalinterpretation of this parameter is provided in Fig. S6. AC, TCand DC indicate respectively ascending, transverse anddescending colon

Antonie van Leeuwenhoek

123

on the composition of the gut microbial community.The aim was to compare the data previously obtained

with DGGE and qPCR with those of the deep

sequencing. It was already known that product Bwas fermented all over the simulated colon compart-

ments (Marzorati et al. 2010; Fig. S1) and therefore,

product B was expected to have a higher impact on themicrobiota in the distal part of the colon.

Pyrosequencing retrieved the presence of 2 main

phyla (i.e. Bacteroidetes, Firmicutes) that accountedfor more than 80 % of the total diversity of the gut

microbial community. These findings are in line with

what previously reported in humans (De Filippo et al.2010; Turnbaugh et al. 2009). The proportion between

the two phyla corresponds also to earlier findings with

the HITChip analysis on the microbial communities ofthe SHIME system (Van den Abbeele et al. 2010) and

of the I-Chip analysis of the dynamic intestinal model

of TNO (TIM2) (Rajilic-Stojanovic et al. 2010). Inboth simulators, a loss of Firmicutes as opposed to

Bacteroidetes was observed during the initial in vivo-

in vitro shift. Nevertheless, the microbial communityobtained in the different regions of the in vitro colon

resulted to be stable upon inoculation, colon region

specific, and relevant to the in vivo conditions whencompared to the fecal sample used to inoculate the

system (Van den Abbeele et al. 2010).

The dominance of Bacteroidetes is typical of ahealthy gut microbiota coevolved with a polysaccha-

ride-rich diet (De Filippo et al. 2010). Moreover, the

relative abundance of two dominant bacterial phyla,the Bacteroidetes and the Firmicutes, has been posi-

tively correlated with body weight control (obese

show a decrease in the abundance of Bacteroidetes anda proportional increase in Firmicutes). This should be

related to the ability of the microbiota to harvest

energy from the diet (Ley et al. 2005; Turnbaugh et al.2006). However, there is no real scientific consensus

and some reports show the opposite, as recently

thoroughly summarized in a review (Ley 2010) andreported in literature (Collado et al. 2008; Duncan

et al. 2008; Zhang et al. 2009; Schwiertz et al. 2010).In this study, both products were able to further

decrease the Firmicutes/Bacteroidetes ratio in the

microbial community of each colon compartment.Pyrosequencing allowed to effectively explore the

impact of the treatment on the whole microbial

community and showed that product B had a moreextended effect along the colon as compared to

product A. This was clearly supported by the phylo-genetic-based dendrogram and the OTU-based PCoA

shown in Fig. 4. Both products had a similar effect in

the AC (cluster I—Fig. 4) and the samples from thedistal colon clustered separately (cluster II). Within

cluster II, the treatment with product B was still able to

modify the microbiota in the distal colon as showed bythe separate clustering of samples S2TCt and S2DCt

(sub-cluster II.3), in contrast with the fact that the

samples from the simulated descending colon of theSHIME treated with product A grouped in the same

sub-cluster II.1. The same conclusions can be reached

analyzing the PCoA plot in which the samples of thesimulated distal colon of the SHIME 2 fell apart as

compared to the samples of the descending colon of

SHIME 1 that, on the contrary, clustered together. Theextended effect of product B, as compared to product

A, was also supported by a higher richness estimator, a

higher Dy parameter and a similar DCo profile—calculated at genus level—for all the colon compart-

ments. The DCo parameter in combination with

pyrosequencing—introduced for the first time in thiswork—is a very useful additional tool to study the

effect of a given treatment on the structure (in terms of

evenness) of a microbial community (see Fig. S6 forfurther explanations). In fact, it allows to identify

changes occurring in high/1ow abundant OTUs even

when the final Co values are very similar (i.e. controlvs. treatment in any colon compartment of the SHIME

2). In the specific case of product B, the DCo profile

pointed out that some phylotypes were selectivelyenhanced during the treatment and that they became

more dominant in the microbial community at the

expense of other phylotypes, while the final degree ofevenness for the whole community did not change.

The enhanced phylotypes in the AC were mainly

associated to the genus Prevotella (some xylanolyticspecies belong to this genus—Dodd et al. 2010) while

in the distal colon they were associated to the genera

Bacteroides. The latter is a genus that includesmicroorganisms whose genome is enriched in carbo-

hydrate acting enzymes and also vitamin and cofactormetabolism, indicating that these bacteria have

adapted to a role of diet digestion and vitamin

production (Karlsson et al. 2011). Bacteroides spp.have been shown to be able to efficiently degrade

arabinogalactans and other plant-derived polysaccha-

rides (Salyers et al. 1977; Van Laere et al. 2000). In allcolon compartments, the enhanced growth of these

Antonie van Leeuwenhoek

123

bacteria occurred at the expenses of other bacteriabelonging to Ruminococcaceae, Porphyromonadaceae

and mainly to the genus Mitsuokella (with several

lactate producers).We have previously shown that products A and B

had an effect in modulating the structure and compo-

sition of the microbial community by, for instance,increasing the concentration of Bacteroidetes or

inducing a bifidogenic activity (Marzorati et al.

2010). However, only by means of a more in depthmolecular analysis (i.e. 16S rRNA gene pyrosequenc-

ing), we could discriminate the opposite trends

occurring in the bacteroides and prevotella genera inthe different areas of the simulated GIT. On the other

hand, pyrosequencing has also some limitations. Even

if it can detect a high percentage of low abundantOTUs in a complex sample, the accuracy in quanti-

fying the same rare species is limited (e.g. Lactoba-cillus spp. and Bifidobacterium spp.), at least workingwith a dataset containing 23,827 sequences. This value

represents a good balance between cost and depth of

analysis per sample and is in line with a similarecological survey previously conducted in mice

(Turnbaugh et al. 2009). If on the one hand pyrose-

quencing indicated a general decrease in the phylumFirmicutes, on the other hand, qPCR had previously

pointed out a specific positive effect on the lactobacilli

(belonging to the same phylum) concentration in theascending and transverse colon of SHIME 2 (Marzo-

rati et al. 2010). With respect to other low-abundant

groups pyrosequenging showed only a slight increaseof Bifidobacterium spp. (i.e. B. longum sub. infantisand B. adolescentis) in the simulated distal colon of

the SHIME treated with product B. On the contrary,specific DGGE and qPCR analyses (Fig. 3) showed

that Bifidobacterium spp. could benefit from both the

treatments, mainly in the simulated ascending colonwith product A and all along the GIT with product B.

The growth of bifidobacteria can be indeed stimulated

by Aloe vera whole leaf extract (Pogribna et al. 2008)while arabinogalactan supported the growth of Bifi-dobacterium longum in batch co-culture with Bacte-roides spp. (Degnan and Macfarlane, 1995).

Finally, with the available pyrosequencing instru-

ments, it is often not possible to provide a phyloge-netic classification of the obtained sequences at

species level. This is a major problem considering

that positive health effects from gut microorgan-isms are exclusively determined at strain level. Such

identification can only be reached with specific qPCR,FISH analysis or gene libraries of the full 16S rRNA

gene.

In conclusions, the combination of a SHIME exper-iment with the barcoded pyrosequencing approach is a

powerful tool for characterizing changes occurring in

the gut microbiota following a treatment. Minorchanges in the overall composition of the blends led

to a region- specific microbiota modulation potentially

extending the saccharolytic fermentation to the distalcolon. However, especially in the GIT, pyrosequenc-

ing may have some limitations related to the quanti-

fication of those species that are of key interest forGRM purposes—because related to potential positive

health effects (i.e. bifidobacteria or lactobacilli)—and

are, nevertheless, only a minor fraction. In this situa-tion, conventional molecular ecological approaches,

such as PCR–DGGE and qPCR (with primers specific

for the targeted microbial groups) still remain a veryuseful complementary tool.

Acknowledgments Tom Van de Wiele, Sam Possemiers andMassimo Marzorati are Postdoctoral Fellows of the ResearchFoundation—Flanders (FWO, Belgium).

References

Alavi A, Fraser W, Tarelli E, Bland M, Axford J (2011a) Anopen-label dosing study to evaluate the safety and effects ofa dietary plant derived polysaccharide supplement on theN-glycosylation status of serum glycoproteins in healthysubjects. Eur J Clin Nutr 65:648–656

Alavi A, Goodfellow L, Fraser O, Tarelli E, Bland M, Axford J(2011b) A double-blind, randomized, placebo-controlledstudy to explore the efficacy of a dietary plant-derivedpolysaccharide supplement in patients with rheumatoidarthritis. Rheumatology 50:1111–11119

Andersson AF, Lindberg M, Jakobsson H, Backhed F, Nyren P,Engstrand L (2008) Comparative analysis of human gutmicrobiota by barcoded pyrosequencing. PLoS One3:e2836

Best T, Kemps E, Bryan J (2010) Saccharide effects on cogni-tion and well-being in middle-aged adults: a randomizedcontrolled trial. Dev Neuropsychol 35:66–80

Boon N, Top EM, Verstraete W, Siciliano SD (2003) Bioaug-mentation as a tool to protect the structure and function ofan activated sludge microbial community against a3-chloroaniline shock load. Appl Environ Microbiol 69:1511–1520

Collado MC, Isolauri E, Laitinen K, Salminen S (2008) Distinctcomposition of gut microbiota during pregnancy in over-weight and normal-weight women. Am J Clin Nutr 88:894–899

Antonie van Leeuwenhoek

123

De Filippo C, Cavalieri D, Di Paola M, Ramazzotti M,Poullet JB, Massart S, Collini S, Pieraccini G, LionettiP (2010) Impact of diet in shaping gut microbiotarevealed by a comparative study in children from Eur-ope and rural Africa. Proc Natl Acad Sci USA 107:14691–14696

De Santis TZ, Hugenholtz P, Keller K, Brodie EL, Larsen N,Piceno YM, Phan R, Andersen GL (2006) NAST: a mul-tiple sequence alignment server for comparative analysis of16S rRNA genes. Nucleic Acids Res 34:W394–W399

Degnan BA, Macfarlane GT (1995) Arabinogalactan utilizationin continuous cultures of Bifidobacterium longum: effect ofco-culture with Bacteroides thetaiotaomicron. Anaerobe1:103–112

Dodd D, Moon YH, Swaminathan K, Mackie RI, Cann IK(2010) Transcriptomic analyses of xylan degradationby Prevotella bryantii and insights into energy acquisi-tion by xylanolytic bacteroidetes. J Biol Chem 285:30261–30273

Duncan SH, Lobley GE, Holtrop G, Ince J, Johnstone AM,Louis P, Flint HJ (2008) Human colonic microbiota asso-ciated with diet, obesity and weight loss. Int J Obes32:1720–1724

Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R (2011)UCHIME improves sensitivity and speed of chimeradetection. Bioinformatics 27:2194–2200

Evans J, Sheneman L, Foster J (2006) Relaxed neighbor joining:a fast distance-based phylogenetic tree constructionmethod. J Mol Evol 62:785–792

Gibson GR, Probert HM, Loo JV, Rastall RA, Roberfroid MB(2004) Dietary modulation of the human colonic microbi-ota: updating the concept of prebiotics. Nutr Res Rev17:259–275

Gihring TM, Green SJ, Schadt CW (2012) Massively parallelrRNA gene sequencing exacerbates the potential for biasedcommunity diversity comparisons due to variable librarysizes. Environ Microbiol 14:285–290

Grootaert C, Van den Abbeele P, Marzorati M, Broekaert WF,Courtin CM, Delcour JA, Verstraete W, Van de Wiele T(2009) Comparison of prebiotic effects of arabinoxylanoligosaccharides and inulin in a simulator of the humanintestinal microbial ecosystem. FEMS Microbiol Ecol69:231–242

Guo X, Xia X, Tang R, Zhou J, Zhao Hand Wang K (2008)Development of a real-time PCR method for Firmicutesand Bacteroidetes in faeces and its application to quantifyintestinal population of obese and lean pigs. Lett ApplMicrobiol 47:367–373

Huse SM, Welch DM, Morrison HG, Sogin ML (2010) Ironingout the wrinkles in the rare biosphere through improvedOTU clustering. Environ Microbiol 12:1889–1898

Karlsson FH, Ussery DW, Nielsen J, Nookaew I (2011) A closerlook at bacteroides: phylogenetic relationship and genomicimplications of a life in the human gut. Microb Ecol61:473–485

Koetzner L, Grover G, Boulet J, Jacoby H (2010) Plant-derivedpolysaccharide dietary supplements inhibits dextran sulfatesodium-induced colitis in the rat. Dig Dis Sci 55:1278–1285

Ley RE (2010) Obesity and the human microbiome. Curr OpinGastroenterol 26:5–11

Ley RE, Backhed F, Turnbaugh P, Lozupone CA, Knight RD,Gordon JI (2005) Obesity alters gut microbial ecology.Proc Natl Acad Sci USA 102:11070–11075

Liu Z, Lozupone C, Hamady M, Bushman FD, Knight R (2007)Short pyrosequencing reads suffice for accurate microbialcommunity analysis. Nucleic Acids Res 35:e120

Lorenz MO (1905) Methods of measuring concentration ofwealth. J Am Stat Assoc 9:209–219

Manning TS, Gibson GR (2004) Microbial-gut interactions inhealth and disease: prebiotics. Best Pract Res Clin Gas-troenterol 18:287–298

Marzorati M, Wittebolle L, Boon N, Daffonchio D, VerstraeteW (2008) How to get more out of molecular fingerprints:practical tools for microbial ecology. Environ Microb10:1571–1581

Marzorati M, Possemiers S, Verstraete W (2009) The use of theSHIME-related technology platform to assess the efficacyof pre- and probiotics. Agro Food Ind Hi-Tech 20:S50–S53

Marzorati M, Verhelst A, Luta G, Sinnott R, Verstraete W, Vande Wiele T, Possemiers S (2010) In vitro modulation of thehuman gastrointestinal microbial community by plant-derived polysaccharide-rich dietary supplements. Int JFood Microbiol 139:168–176

Molly K, Vande Woestyne M, De Smet B, Verstraete W (1994)Validation of the simulator of the human intestinal microbialecosystem (SHIME) reactor using microorganism-associ-ated activities. Microb Ecol Health Dis 7:191–200

Pogribna M, Freeman JP, Paine D, Boudreau MD (2008) Effectof Aloe vera whole leaf extract on short chain fatty acidsproduction by Bacteroides fragilis, Bifidobacterium in-fantis and Eubacterium limosum. Lett Appl Microbiol46:575–580

Possemiers S, Grootaert C, Vermeiren J, Gross G, Marzorati M,Verstraete W, Van de Wiele T (2009) The intestinalenvironment in health and disease—recent insights on thepotential of intestinal bacteria to influence human health.Curr Pharm Des 15:2051–2065

Pruesse E, Quast C, Knittel K, Fuchs BM, Ludwig W, Peplies J,Glockner FO (2007) SILVA: a comprehensive onlineresource for quality checked and aligned ribosomal RNAsequence data compatible with ARB. Nucleic Acids Res35:7188–7196

Qin J, Li R, Raes J, Arumugam M, Burgdorf KS, Manichanh C,Nielsen T et al (2010) A human gut microbial gene cata-logue established by metagenomic sequencing. Nature464:59–65

Rajilic-Stojanovic M, Maathuis A, Heilig HG, Venema K, deVos WM, Smidt H (2010) Evaluating the microbialdiversity of an in vitro model of the human large intestineby phylogenetic microarray analysis. Microbiology 156:3270–3281

Ramberg JE, Nelson ED, Jacoby HI, Li Q, Koetzner L, Zhang B,Boulet J, Sinnot RA (2010) Plant plysaccharide supple-ments reduce the expression of pro-inflammatory genes incolonic tissue of mice with dextran sulfate sodium-inducedcolitis. Int J Probiotics Prebiotics 5:229–236

Read S, Marzorati M, Guimaraes BCM, Boon N (2011)Microbial resource management: successful parametersand new concepts. Appl Microbiol Biotechnol 90:861–871

Salyers AA, Vercellotti JR, West SE, Wilkins TD (1977) Fer-mentation of mucin and plant polysaccharides by strains of

Antonie van Leeuwenhoek

123

Bacteroides from the human colon. Appl Environ Micro-biol 33:319–322

Sanchez JI, Marzorati M, Grootaert C, Baran M, Van Craeyveld Vet al (2009) Arabinoxylan-oligosaccharides (AXOS) affectthe protein/carbohydrate fermentation balance and microbialpopulation dynamics of the simulator of human intestinalmicrobial ecosystem. Microb Biotechnol 2:101–113

Schloss PD (2009) A high-throughput DNA sequence alignerfor microbial ecology studies. PLoS One 4:e8230

Schloss PD (2010) The effects of alignment quality, distancecalculation method, sequence filtering, and region on theanalysis of 16S rRNA gene-based studies. PLoS ComputBiol 6:e1000844

Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M et al(2009) Introducing mothur: open-source, platform-inde-pendent, community-supported software for describing andcomparing microbial communities. Appl Environ Micro-biol 75:7537–7541

Schwiertz A, Taras D, Schafer K, Beijer S, Bos NA, Donus C,Hardt PD (2010) Microbiota and SCFA in lean and over-weight healthy subjects. Obesity 18:190–195

Sheneman L, Evans J, Foster JA (2006) Clearcut: a fast imple-mentation of relaxed neighbor joining. Bioinformatics 22:2823–2824

Turnbaugh PJ, Ley RE, Mahowald MA, Magrini V, Mardis ER,Gordon JI (2006) An obesity-associated gut microbiomewith increased capacity for energy harvest. Nature 444:1027–1031

Turnbaugh PJ, Hamady M, Yatsunenko T, Cantarel BL, DuncanA, Ley RE et al (2009) A core gut microbiome in obese andlean twins. Nature 457:480–484

Van de Wiele T, Boon N, Possemiers S, Jacobs H, Verstraete W(2007) Inulin-type fructans of longer degree of polymeri-zation exert more pronounced in vitro prebiotic effects.J Appl Microbiol 102:452–460

Van den Abbeele P, Grootaert C, Marzorati M, Possemiers Set al (2010) Microbial community development in adynamic gut model is reproducible, colon region specific,and selective for Bacteroidetes and Clostridium cluster IX.Appl Environ Microbiol 76:5237–5246

Van Laere KMJ, Hartemink R, Bosveld M, Schols HA, VoragenAGJ (2000) Fermentation of plant cell wall derived poly-saccharides and their corresponding oligosaccharides byintestinal bacteria. J Agric Food Chem 48:1644–1652

Wang Q, Garrity GM, Tiedje JM, Cole JR (2007) NaiveBayesian classifier for rapid assignment of rRNA sequen-ces into the new bacterial taxonomy. Appl EnvironMicrobiol 73:5261–5267

Zhang H, DiBaise JK, Zuccolo A, Kudrna D, Braidotti M, Yu Y,Parameswaran P, Crowell MD, Wing R, Rittmann BE,Krajmalnik-Brown R (2009) Human gut microbiota inobesity and after gastric bypass. Proc Natl Acad Sci USA106:2365–2370

Antonie van Leeuwenhoek

123