The ISME Journal (2011) 5, 220–230

& 2011 International Society for Microbial Ecology All rights reserved 1751-7362/11
www.nature.com/ismej

ORIGINAL ARTICLE
Dominant and diet-responsive groups of bacteria
within the human colonic microbiota
Alan W Walker1, Jennifer Ince2, Sylvia H Duncan2, Lucy M Webster2, Grietje Holtrop3,
Xiaolei Ze2, David Brown2, Mark D Stares1, Paul Scott1, Aurore Bergerat2, Petra Louis2,
Freda McIntosh2, Alexandra M Johnstone2, Gerald E Lobley2, Julian Parkhill1 and
Harry J Flint2
1
Pathogen Genomics, Wellcome Trust Sanger Institute, Cambridge, UK; 2Rowett Institute of Nutrition and
Health, University of Aberdeen, Aberdeen, UK and 3Biomathematics and Statistics Scotland, Aberdeen, UK

The populations of dominant species within the human colonic microbiota can potentially be
modified by dietary intake with consequences for health. Here we examined the influence of
precisely controlled diets in 14 overweight men. Volunteers were provided successively with a
control diet, diets high in resistant starch (RS) or non-starch polysaccharides (NSPs) and a reduced
carbohydrate weight loss (WL) diet, over 10 weeks. Analysis of 16S rRNA sequences in stool
samples of six volunteers detected 320 phylotypes (defined at 498% identity) of which 26, including
19 cultured species, each accounted for 41% of sequences. Although samples clustered more
strongly by individual than by diet, time courses obtained by targeted qPCR revealed that ‘blooms’
in specific bacterial groups occurred rapidly after a dietary change. These were rapidly reversed by
the subsequent diet. Relatives of Ruminococcus bromii (R-ruminococci) increased in most
volunteers on the RS diet, accounting for a mean of 17% of total bacteria compared with 3.8% on
the NSP diet, whereas the uncultured Oscillibacter group increased on the RS and WL diets.
Relatives of Eubacterium rectale increased on RS (to mean 10.1%) but decreased, along with
Collinsella aerofaciens, on WL. Inter-individual variation was marked, however, with 460% of
RS remaining unfermented in two volunteers on the RS diet, compared to o4% in the other
12 volunteers; these two individuals also showed low numbers of R-ruminococci (o1%). Dietary
non-digestible carbohydrate can produce marked changes in the gut microbiota, but these depend
on the initial composition of an individual’s gut microbiota.
The ISME Journal (2011) 5, 220–230; doi:10.1038/ISMEJ.2010.118; published online 5 August 2010
Subject Category: microbe–microbe and microbe–host interactions
Keywords: human colon; resistant starch; 16S rRNA; phylotypes; Ruminococcus; temporal change

Introduction of different individuals. Identifying the dominant

bacterial species that colonize the large intestine
The remarkable diversity of the human colonic and the extent to which these are influenced by diet
microbiota at the level of bacterial species and and host factors is of key importance in uncovering
phylotypes has become apparent from 16S rRNA- the impact of the colonic microbiota upon human
based analyses. Not only are hundreds of phylotypes health (Flint et al., 2007; Sokol et al., 2008).
typically estimated to be present in the human A few studies to date have examined temporal
colonic microbiota from a given faecal sample changes, and these suggest a degree of stability in
(Suau et al., 1999; Eckburg et al., 2005), but samples the colonic microbiota of individuals consuming
from different individuals have been reported to their normal diets (Franks et al., 1998; Zoetendal
show limited overlap in the phylotypes present et al., 1998; Costello et al., 2009). In contrast, little
(Ley et al., 2006; Turnbaugh et al., 2008). A recent is known about the impact of dietary change
study (Tap et al., 2009), however, has indicated upon microbial community composition. There is
that certain phylotypes occur more commonly evidence that dietary supplementation with pre-
than others among the dominant faecal bacteria biotics such as fructo-oligosaccharides and inulin
can promote specific groups of bacteria, including
bifidobacteria (Bouhnik et al., 2004; Ramirez-Farias
Correspondence: HJ Flint, Microbial Ecology Group, Rowett et al., 2009). It has also been shown that reductions
Institute of Nutrition and Health, University of Aberdeen,
Aberdeen AB21 9SB, UK. E-mail: [email protected] in total carbohydrate content, in weight loss (WL)
Received 25 January 2010; revised 11 May 2010; accepted 21 June diets for obese subjects, have major effects upon the
2010; published online 5 August 2010 composition and metabolic outputs of the bacterial
 Impact of diet on human colonic microbiota
AW Walker et al
221
community in the colon (Duncan et al., 2007, 2008; years) with a mean body mass index (kg m 2) of 39.4
Brinkworth et al., 2009). These changes are assumed (range 27.9–51.3) (Supplementary Table S1). The
to reflect the fermentation of non-digestible (ND) study was approved by the North of Scotland
carbohydrate components (mainly non-starch poly- Research Ethics Service and all volunteers provided
saccharides (NSPs), resistant starch (RS) and certain informed, written consent.
oligosaccharides) that reach the large intestine. The
impact upon the colonic microbiota of controlled
changes in the main types of ND carbohydrate Experimental design and dietary regime
normally present in the diet (RS and NSP) has not The volunteers were weight stable (o3 kg change
however been examined in any detail. The ND in the past 4 months) before entry on the trial. Over
carbohydrate content of the human diet is consi- the first 7 weeks, intakes were provided at energy
dered to influence health. For example, diets high in maintenance as 1.4–1.5  measured resting meta-
RS have been shown to benefit insulin sensitivity, bolic rate for each individual. The final intervention
possibly mediated by bacterial fermentative activity period consisted of high protein WL diet for an
in the colon (Robertson et al., 2005). Diets contain- additional 3 weeks (Supplementary Table S2,
ing RS and NSP offer potential benefits in preven- Supplementary Figure S1). The initial maintenance
tion of colorectal cancer through the delivery of diet (M diet) comprised protein/carbohydrate/fat %
fermentation acids, in particular butyrate, to the as 13:52:35 of metabolizable energy and 27.7 g
distal colon (McIntyre et al., 1993; Duncan et al., per day NSP, provided for 7 days. Subjects were
2007). Microbial breakdown of NSP also releases then provided with fixed intakes of two diets, which
bound phytochemicals into the colon (Gill and consisted of either a high RS diet or a high NSP diet
Rowland, 2002). These health benefits may be each supplied for 3 weeks in a randomized cross-
particularly important in obese and overweight over design (Supplementary Figure S1). The RS diet
subjects who are at increased risk of developing contained added type III RS whereas the NSP diet
colorectal cancer and diabetes (Polednak, 2003; intake contained added wheat bran. All meals were
Cani et al., 2007). of the same energy density (5.5 MJ kg 1) and daily
This study asks to what extent changes in the intakes were recorded by weigh-back after each meal
major type of ND carbohydrate in the diet influence (Supplementary Table S2). Daily macronutrient
the dominant bacterial phylotypes present in an intakes were calculated using the Windiet software
individual. It also explores the dynamics of such program (Robert Gordon University, Aberdeen, UK),
changes, including their reversibility, for the first based on the type and quantity of each ingredient
time through the use of detailed time courses and a consumed and published food composition tables
cross-over design. A necessary supplementary ques- (Food Standards Agency 2002). Faecal samples
tion is whether the same dietary change elicits were collected on average twice each week. In addi-
similar changes in the gut microbiota of different tion, one 24 h collection from the final week of each
individuals. Our results reveal rapid and marked dietary period was used for chemical analysis of
changes in the colonic microbiota of a group of digestibility.
overweight volunteers after a shift in the main type
of ingested fermentable carbohydrate (RS or NSP).
We also find that these changes can be highly Chemical analysis of diet composition and digestibility
specific to the individual, with potentially important Diets were analysed for total gross energy, RS
implications for the design of fibre-enriched diets. and insoluble and soluble NSPs (Supplementary
Table S2). Estimation of RS content of diets and
24 h faecal sample collections was as described by
Materials and methods Englyst et al. (1992) whereas NSPs were quantified
as described by Englyst and Cummings (1988).
Volunteer recruitment
Volunteers were initially recruited by newspaper
advertisement. Inclusion criteria for men were a DNA extraction from faecal samples
body mass index 427 kg m 2, waist circumference Faecal samples were kept at 4 1C and processed
4102 cm and fasting plasma glucose concentration within 5 h of collection. This short period of storage
46.0 mmol l 1, indicating metabolic syndrome. is not expected to influence molecular estimation
They underwent a medical examination and their of microbial community composition (Lauber et al.,
general practitioner was contacted to confirm med- 2010). Each sample was mixed and 5 g dispersed
ical and medication history. None of the subjects (3500 r.p.m. for 1 min using a Dispomix Drive,
had a history of gastrointestinal disease or distur- Medic Tools, Lussiwag, Switzerland) in 10 ml
bance, or took antibiotics, either in the 6 months sterile phosphate-buffered saline buffer before
leading up to the study or during the study. Two of aliquoting. One aliquot was used immediately
the original 16 subjects left the study for reasons for DNA extraction using the FastDNA Spin for
unconnected with the protocol. The remaining 14 soil kit following manufacturer’s (Qbiogene, MP
volunteers had a mean age of 54 years (range 27–73 Biomedicals, Illkirch, France) instructions.

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
222
Phylogenetic analysis of sequences derived from Primer sequences and conditions are reported in
16S rRNA clone libraries Supplementary Table S3, based largely on the
PCR amplification of 16S rRNA genes from the study by Ramirez-Farias et al. (2009). Primers for
extracted DNA involved initial denaturation at 94 1C Oscillibacter-related bacteria were designed in this
for 5 min; 20 cycles of denaturation (30 s at 95 1C), study.
annealing (30 s at 52 1C), extension at 72 1C for 2 min For denaturing gradient gel electrophoresis
with a final cycle at 72 1C for 8 min. Taq1 polymerase (DGGE), bacterial DNA was amplified by PCR using
from Promega (Southampton, UK) was used at primers for the hyper-variable V3 region of the 16S
0.025 U ml 1 in the presence of 2.5 mM MgCl2. The gene (Supplementary Table S3). Reaction mixtures
forward primer comprised a 4:1:1:1:1 mixture of contained 1.5 mM MgCl2 and 0.025 U Taq1 poly-
7f, 27fChl, 27fBor, 27fBif and 27fAto and the reverse merase per ml. Products were separated in an 8%
primer was 1510r (Supplementary Table S3). PCR acrylamide denaturing gradient gel containing urea
products were cleaned using the Wizard PCR and formamide (35–60% gradient) at 80 V for 16 h
product purification kit (Promega) and were then at 60 1C and visualized by staining with SybrGold
cloned and sequenced as described (Lawley et al., dye (Invitrogen, Paisley, UK).
2009). The sequences (spanning variable regions
V2–V5) were aligned using the SILVA-derived
reference alignment in the mothur software package Statistical analysis
(Schloss et al., 2009) and were tested for the For phylotypes with large abundance (more than
presence of chimeras by Mallard (Ashelford et al., 1% and found in 5 or 6 out of 6 volunteers,
2006), Pintail (Ashelford et al., 2005) and BLAST which was the case for 8 out of 320 phylotypes)
(Johnson et al., 2008). After removal of chimeric the Sanger sequences (clone counts) were analysed
sequences, the alignment was subjected to exten- as Hierarchical Generalized Linear Models using a
sive manual curation using the editor function in binomial model with logistic link and estimated
the ARB package (Ludwig et al., 2004). Using overdispersion. This was carried out for one phylo-
the curated alignment, a distance matrix, with type at a time, with the total number of clones per
Felsenstein correction, was created using ARB. sample regarded as the total count. Volunteer and
This matrix was then used as an input for diet were regarded as random (normally distributed)
DOTUR (Schloss and Handelsman, 2005) using a and fixed effects, respectively. Similarities between
98% identity cut-off under the default furthest- faecal samples were displayed graphically by means
neighbour setting. Sequences with 498% phylo- of principal coordinate analysis, where the Canberra
genetic similarity were regarded as belonging to distance matrix was applied to the numbers of
the same phylotype. This generated 320 phylotypes clones assigned to each phylotype (based on all
(Supplementary Table S4), which were assigned an 320 phylotypes) for each faecal sample. Principal
identity from the phylum to genus level by the RDP coordinate analysis was also used to obtain a
Classifier (Cole et al., 2009) and to the closest match graphical representation of the phylogenetic dis-
in the NCBI database using MegaBLAST (Johnson tances (obtained from the DOTUR software package)
et al., 2008). Rarefaction analysis, Shannon and between the 320 phylotypes.
Simpson diversity index calculations,
R community The effect of diet was tested by means of analysis
structure comparisons with -Libshuff, parsimony of variance with volunteer regarded as random and
and weighted and unweighted versions of UniFrac, diet regarded as fixed effect. When the effect of diet
Jaccard and Yue and Clayton theta tree clustering was significant, diet means were compared by post
analysis were all performed using the mothur soft- hoc t-test. Data containing many zeros (that is, below
ware package (Schloss et al., 2009). limit of detection) were analysed by Friedman’s
The final 16S rRNA clone library data set from non-parametric analysis of variance instead. For all
six volunteers (v16, v19, v20, v22, v23 and v24) analyses, the effect of order was included initially as
contained 5915 sequences (median length of 678 a fixed effect but it and its interaction with diet were
bases) (GenBank accession numbers GU238434– found not to be significant and was removed from
GU244348). These sequences were from one faecal subsequent analyses. All statistical analyses were
sample from each of the four diets per individual, performed by using Genstat 11th Edition Release
except for volunteer v16 when two samples were 11.1 (VSN International Ltd, Hemel Hempstead,
analysed from the M and NSP diets (see Supplemen- Hertfordshire, UK). Significance was set at P-value
tary Table S4). An additional 768 sequences were of o0.05.
subsequently obtained from the RS and NSP diets only
of v14 and v25 (Supplementary Table S5) (GenBank
accession numbers HM191774–HM192541).
Results
Study design
qPCR and DGGE analysis of 16S rRNA genes Fourteen overweight male volunteers completed
Quantitative real-time PCR (qPCR) was performed this study (see Materials and Methods section)
on all available samples from all 14 volunteers. (Supplementary Table S1). Complete diets were

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
223
provided daily resulting in precise control over all
dietary intake throughout the 10-week study period
(Supplementary Table S2). Each volunteer received
a control weight maintenance (M) diet for the first
week, followed by three weeks each of diets in high
type III resistant starch (RS diet) or high in wheat
bran (NSP) and a WL diet reduced in carbohydrate
but high in protein (WL diet). The first three diets
were provided at energy maintenance and were
closely matched for total carbohydrate, protein and
fat content; the order of the RS and NSP diets was
reversed for half of the volunteers (Supplementary
Figure S1, Supplementary Table S2). Faecal samples
were processed as soon as possible after collection, Figure 1 Culturability of 16S rRNA phylotypes in relation
without prior freezing, for the extraction of nucleic to their abundance. 16S rRNA sequences (5915) obtained
acids (see Materials and Methods section). from faecal samples of six volunteers (for all four diets) were
classified into 320 phylotypes (defined at 498% sequence
identity) (Supplementary Table S4). The % of phylotypes
showing 498% sequence identity to a cultured bacterium
Dominant bacterial phylotypes within the intestinal is seen to increase with increasing phylotype abundance.
bacterial community The phylotype frequency distribution is shown as a percentage;
16S rRNA profiles from all available faecal samples actual numbers of phylotypes were 9 (42% of all sequences),
17 (41%o2%), 24 (40.5%o1%) and 270 (o0.5%) (total 320).
were analysed first using DGGE, which revealed that Data refer to volunteers 16, 20, 22 (diet order RS-NSP) and 19, 23,
profiles were consistent over time within an indi- 24 (diet order NSP-RS).
vidual for a given diet (Supplementary Figure S2).
A switch between the RS and NSP diets, however,
altered the profile substantially, indicating a shift
in bacterial community composition. Sequences of
amplified 16S rRNA genes were analysed for 6 of
the 14 volunteers (3 from each diet order) using
a sample from the final week of each dietary
regime. This yielded 5915 non-chimeric sequences
(averaging 227 sequences per sample). Phylogenetic
analysis using a 98% sequence identity cut-off
revealed 320 phylotypes across all samples (Supple-
mentary Table S4). The proportion of cultured
strains increased with increasing abundance of
phylotypes, with all phylotypes present at 42% of
the total being identified with cultured species
(Figure 1). None of the 320 phylotypes was detected
in every sample, but 32 phylotypes were present Figure 2 Incidence of phylotypes in different individuals. The
in all six volunteers accounting for 47.1% of all distribution of all 320 16S rRNA phylotypes was determined
across the six volunteers (see Figure 1, Supplementary Table S4);
sequences, whereas 77 phylotypes occurred in at the numbers of phylotypes found in all six individuals (32) and in
least four of the six individuals (Figure 2, Supple- five (25), four (20), three (36), two (62) or one (145) of the six are
mentary Figure S3). Figure 2 suggests a bimodal shown here according to bacterial phylum.
frequency distribution for phylotype occurrence,
indicating that there may be a group of widespread
and highly abundant phylotypes within the human theta similarity coefficients (Yue and Clayton, 2005)
gut microbiota (Tap et al., 2009). The most abundant (Supplementary Figure S4). These results and those
phylotype was Eubacterium rectale although three from principal coordinate analysis (Figure 3)
phylotypes corresponding to three divergent strains indicate that samples predominantly clustered by
of Faecalibacterium prausnitzii together comprised volunteer, suggesting that responses to dietary
8.0% of total sequences (Table 1). change are influenced by the initial species compo-
Overall community structure was analysed for sition of the individual’s colonic microbiota. Rare-
each sample using theR mothur software package faction analysis revealed that the observed species
(Schloss et al., 2009). -Libshuff, parsimony and richness was similar for all four diets (Supple-
UniFrac (weighted and unweighted) analysis mentary Figure S5). Simpson (1/D) and Shannon
revealed a high degree of variation between indivi- diversity indices were slightly lower on the RS
duals and showed that each sample was signi- and NSP diets compared with the M and WL diets,
ficantly different from all others. Similarity in but did not differ significantly (Supplementary
community membership and structure was esti- Figure S5; see also Supplementary Table S4 for
mated using the Jaccard and the Yue and Clayton individual values).

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
224
Table 1 Impact of diet upon the 10 most abundant 16S rRNA phylotypes detected in six overweight male volunteers

Phylotype, species a Phylum Mean % clones Diets P-value b

% total clones M NSP RS WL

Eubacterium rectale F (L) 4.43 2.9x 4.1x 8.0y 1.7x o0.001

Collinsella aerofaciens A 3.67 4.3x 3.4x 5.2x 0.6y 0.032
Clostridium clostridioforme F (L) 3.83 4.8 2.6 2.5 3.1 0.217
Bacteroides vulgatus B 3.21 2.6 4.1 1.3 4.9 0.057
Faecalibacterium prausnitzii L2-6 F (R) 2.96 2.5 3.4 1.6 3.8 0.224
Faecalibacterium prausnitzii A2-165 F (R) 2.55 1.8 2.8 2.9 3.4 0.338
Faecalibacterium prausnitzii M21/2 F (R) 2.47 2.1 1.8 2.2 3.1 0.771
Anaerostipes colic SSC/2 F (L) 2.25 3.0 2.5 1.1 1.8 0.091
Ruminococcus bromii F (R) 2.11 1.5x 0.4x 5.0y 1.5x o0.001
Eubacterium hallii F (L) 2.00 2.4 2.5 1.2 1.4 0.251

Abbreviations: F, Firmicutes (L, Lachnospiraceae; R, Ruminococcaceae); B, Bacteroidetes; A, Actinobacteria.

a
Detected in all six volunteers, except for R. bromii (four of six volunteers).
b
Within a row values not sharing superscripts (x and y) differ significantly (Po 0.05). Clone counts were analysed with Hierarchical Generalized
Linear Model with Volunteer as random and Diet as fixed effect.
c
New species name soon to be formally proposed (Allen-Vercoe et al., in preparation).

Table 2 Populations estimated by qPCR of bacteria and

methanogenic archaea expressed as % relative to total bacteria

Bacterial group (%) Diet P diet a

M NSP RS WL SED

Bac 27.8 25.7 20.2 24.0 3.58 0.387

Ros 7.3x 6.5x 10.1y 3.3z 1.381 o0.001
Fprau 11.2 14.4 12.1 12.5 1.622 0.278
Rum 6.5x 3.8x 17.0y 7.5x 2.253 o0.001
Osc 0.74x 0.77x 2.0y 1.6y 0.34 o0.001
Bif 1.9 1.8 2.4 0.8 0.632 0.059
Met 0.080 0.034 0.124 0.123 0.042 0.155b

Abbreviations: Bac, Bacteroides/Prevotella; Ros, E. rectale/Roseburia

spp.; Fprau, F. prausnitzii; Rum, R-ruminococci; Osc, Oscillibacter
relatives; Bif, Bifidobacterium spp.; Met, methanogens.
a
Figure 3 Impact of diet and individual variation upon faecal Significance based on analysis of variance (ANOVA). Within each
row, values not superscripts (x, y and z) differ significantly (Po0.05,
microbiota composition. A principal coordinates analysis (using
post hoc t-test).
Canberra distance matrix) based on 16S rRNA clone libraries from b
Friedman non-parametric ANOVA.
26 faecal samples (obtained from six donors under four dietary Primer sequences and amplification conditions are shown in
conditions) (Supplementary Table S4). Colour code is based on Supplementary Table S3. One mean value is shown per diet for the
donor (v16, v19, v20, v22, v23 and v24). Diets are indicated as 14 volunteers. For the M diet, means are for all samples from the
M, maintenance; NSP, non-starch polysaccharide; RS, resistant 1 week of this diet, whereas for the NSP, RS and WL diets means
starch; WL, weight loss. Two samples were analysed for v16 from are for all samples from the last 2 weeks on each diet.
the M and NSP diets.

Response of the colonic microbial community to PCR primer sets (Supplementary Table S3) were
dietary change used to target groups for which sequence analysis
At the phylum level there was no significant effect of indicated a response to the RS diet (Rum, Ros, Osc),
diet upon the proportions of Bacteroidetes (mean or for which previous in vitro evidence has shown
21.5%), Firmicutes (mean 70.6%), Actinobacteria the ability to use starch (Bac, Bif, Fprau) (Table 2).
(mean 4.9%) or Proteobacteria (mean 3.0%) within Methanogens were also monitored because of their
the faecal microbiota. At finer taxonomic levels, potential impact on the polysaccharide-utilizing
however, two individual phylotypes, E. rectale and community (Robert and Bernalier-Donadille, 2003).
Ruminococcus bromii, showed increased proportions All available time points were analysed. Mean data
on the RS diet whereas Collinsella aerofaciens showed for the targeted groups are shown in Table 2 and
decreased proportions on the WL diet (Po0.001) complete time courses for all 14 subjects are shown
(Table 1). Because the sequence data also suggested in Figure 4 for those bacterial groups that showed
that groups of related phylotypes might be affected by responses to the RS diet. The Rum group includes
diet, we decided to use qPCR for further analysis. relatives of R. bromii, R. flavefaciens and R. albus

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
225

Figure 4 Diet-driven changes in four groups of human colonic bacteria detected by qPCR. Abundance for each targeted group is
expressed as a percentage of the signal obtained with a general bacterial primer set (see Supplementary Table S3 for primers and
conditions used). All available time points are shown for the 14 volunteers: left-hand panels show the diet order M-NSP-RS-WL and
right-hand panels M-RS-NSP-WL, (a and b) R-ruminococci (relatives of R. bromii); (c and d) relatives of E. rectale and Roseburia spp.;
(e and f) Bifidobacterium spp.; (g and h) relatives of O. valericigenes.

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
226

Figure 6 Starch digestibilities. The whole tract % digestibility of

resistant starch was determined after estimating resistant starch
present in the supplied diet and in 24 h faecal samples (see
Materials and Methods section). Results are shown only for the
NSP and RS diet periods. The markedly reduced resistant starch
digestibilities in v14 and v25 correspond with low R-ruminococcal
numbers in these individuals (see text and Supplementary Table S5).

et al., 2005; Hold et al., 2002) and this was also the
case in our analysis in spite of using a broad
bacterial primer set reported to include this group
(Frank et al., 2008). There was no significant change
in the % Bacteroides in response to diet (Table 2).
Methanogenic archaea were detected by qPCR in
Figure 5 Populations of three groups of potentially amylolytic 490% of samples from nine volunteers, but were
bacteria on the RS (high-resistant starch) and NSP (low-resistant detected in fewer than 10% of samples from v11, 14,
starch) diets. Populations, estimated by qPCR, are shown for 17, 24 and 25. This group showed no significant
R-ruminococci, E. rectale/Roseburia spp. and Bifidobacterium
spp. for all 14 volunteers, according to diet order. Data are the
effect of diet (Table 2).
mean values for each volunteer during the second and third week
of the RS and NSP diets (overall means are given in Table 2).
Digestibility of RS and NSP
Whole tract digestibility of RS and NSP was
estimated from chemical analysis of the diet and of
24 h faecal collections. For 12 of the 14 volunteers
that belong to the Ruminococcaceae, and will be
o4% of ingested RS was recovered in faecal
referred to here as ‘R-ruminococci’. There was a
samples for all four diets, reflecting almost complete
significant increase of around 4.5-fold (Po0.001) in
microbial fermentation. For two individuals
this group on the RS diet, compared to the NSP diet
(v14 and v25), however, faecal RS recovery was
(Table 2, Figure 4). A new primer set was designed
69% and 65%, respectively (31% and 35% digest-
to recognize another group of Ruminococcaceae,
ibility) on the RS diet (Figure 6). These two
related to Oscillibacter valericigenes (Iino et al.,
individuals both harboured low R-ruminococci
2007) that has not been cultured from the human
numbers as assessed by qPCR, although one of them
gut. This group increased significantly on average
showed an increase in the Roseburia and E. rectale
on the RS and WL relative to the M and NSP diets
group on the RS diet (Figures 4 and 5). Additional
(Po0.001) (Table 2, Figure 4). A previously reported
clone library analyses were subsequently performed
primer set (Mackie et al., 2003) targeting relatives
on the final samples from the RS and NSP dietary
of this group from the rumen was also used here,
periods for these two individuals. No R-ruminococci
but detected o0.1% of human faecal bacteria
were detected whereas 36% and 46.5% of clones
(results not shown). Populations of a third abundant
belonged to the Bacteroidetes among the 381 and
group of Ruminococcaceae, F. prausnitzii, showed
387 clones analysed for v14 and v25, respectively
no significant response to changes of diet (Table 2).
(Supplementary Table S5).
Firmicutes bacteria related to Roseburia and
Soluble NSP digestibility averaged 90% and was
E. rectale also increased significantly (Po0.048) on
slightly lower (Po0.05) for RS and WL diets (88%)
the RS diet, and decreased (approximately threefold
than for the M diet (92%). Insoluble NSP digest-
compared to RS) (Po0.029) on the reduced carbo-
ibility averaged 66% and showed no difference
hydrate WL diet. (Ros, Table 2). Individual differ-
between diets (mean range 58–71%).
ences were evident for all groups shown in Figure 4.
Notably, Bifidobacterium spp. did not show a signi-
ficant mean response to RS, but their population Discussion
increased markedly on RS in one volunteer (v23)
(Figures 4 and 5). Bifidobacteria are often poorly This is the first study to combine precise control and
represented in 16S rRNA clone libraries (Eckburg monitoring of human dietary intake and digestibility

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
227
with detailed analysis of changes in the faecal the degradation of high amylose RS (Macfarlane and
microbiota at the level of bacterial phylotypes. As Englyst, 1986; Wang et al., 1999). In this study,
a result it has been possible to reveal the interplay bifidobacterial numbers showed a strong response to
between diet and inter-individual differences that RS only in one individual (v23), whereas there was
influence the human colonic community. Analysis no evidence for a response of the Bacteroides group
of amplified 16S rRNA sequences suggested that to RS. Bacteroides spp. may perhaps be better
samples from six overweight male volunteers adapted to the use of solubilized starch molecules
clustered more strongly by individual than by diet. (Flint et al., 2008). These findings in vivo however
This is in line with previously reported evidence of correspond better with more recent work in vitro
inter-individual variation in the faecal microbiota that identified R. bromii, E. rectale and Bifido-
(Franks et al., 1998; Ley et al., 2006). In contrast, diet bacterium spp. as the major species in human faeces
composition had very substantial effects on specific that colonize starch particles (Leitch et al., 2007)
groups of bacteria that were detected both through and that use 13C-labelled starch (Kovatcheva-
clone libraries and extensive qPCR analysis on 279 Datchary et al., 2009).
faecal samples. The time courses show that most Different groups of amylolytic bacteria differ
diet-driven changes occurred rapidly, being detect- markedly in their metabolic products and potential
able within 3–4 days, and were reversed equally host interactions. Bifidobacterium spp., which
rapidly. These kinetics appear consistent with produce lactate and acetate, are widely used as
immediate effects of dietary residue upon relative probiotics and as targets for prebiosis (Furrie et al.,
bacterial growth rates in the colon that are subse- 2005). Members of the E. rectale group are flagel-
quently reflected in faecal samples as colonic lated bacteria that are major producers of butyrate in
contents turnover, assuming mean transit colonic the large intestine, and may therefore contribute to
times of around 60 h (Stephen et al., 1987). the butyrogenic effect of RS (Aminov et al., 2006;
Firmicutes bacteria related to R. bromii Duncan et al., 2007; Louis et al., 2010). R. bromii is
(R-ruminococci) and E. rectale were commonly a producer of acetate, ethanol and hydrogen that is
stimulated by the RS diet. In most individuals, likely to contribute to gas production, but otherwise
qPCR analysis revealed a surge in the population of little is currently known about the impact of this
R-ruminococci with values exceeding 25% of total group upon the host. It will be important in future to
bacteria in some samples. R-ruminococci were the establish whether different types of RS select for
only group of human gut bacteria previously found different groups of amylolytic bacteria in view of
to be preferentially associated with particulate their differing effects on fermentation and host
material in human faecal samples (Walker et al., responses (Le Leu et al., 2009).
2008), suggesting that they have an important role in Starch is considered to show high digestibility
the breakdown of particulate substrates. R. bromii across the whole gastrointestinal tract, with most
isolates show amylolytic activity (Salyers et al., RSs being completely fermented in the large intes-
1977) but our results suggest that many related, but tine (Bird et al., 2000). Remarkably, however, signi-
uncultured, bacteria may also possess this activity. ficant amounts of starch survived fermentation to be
One previous study, using non-quantitative DGGE recovered in the faeces in 2 of the 14 individuals
analysis, reported that R. bromii-related bacteria studied here. This difference in fermentation cannot
were prominent in faecal samples from humans on be ascribed to dietary intake, which was standar-
a diet high in RS and NSP (Abell et al., 2008). The dized for all 14 subjects, and seems likely to lie with
E. rectale and Roseburia group is also known to the strain composition of each individual’s colonic
contain amylolytic species (Ramsay et al., 2006) but microbiota. These two individuals showed very
has not previously been shown to respond to RS low numbers of R-ruminococci, and the relationship
in vivo. An earlier study failed to detect an increase between community composition and starch
in specific Eubacterium spp. in human subjects on a fermentation will clearly warrant further investiga-
diet high in RS type III but, importantly, detection tion. Although these two individuals were also
methods were not available for E. rectale (Schwiertz non-methanogenic, no simple relationship was
et al., 2002). In this work we also detected a evident between methanogens and R-ruminococci.
dietary response for an uncultured group of Rumino- Methanogen populations are known to vary between
coccaceae related to Oscillibacter. It is not known individuals and are influenced by a variety of factors
whether these bacteria are starch degraders, and (Florin et al., 2000).
their increase both on the RS and WL diets suggests In contrast to these responses to RS, there was
that other factors must be involved in their response little evidence that the high NSP diet resulted in
to diet. major alterations in the composition of the faecal
Early studies indicated a broad distribution of microbiota. In part, however, this may reflect the
amylase activity among cultured human gut bacteria fact that a smaller increase was achieved in NSP
with 450% of strains, including many Bacteroides, intake (1.5-fold) than with RS intake (4.8-fold) when
able to grow on amylose or amylopectin starch compared with the M diet. It is possible that larger
(Salyers et al., 1977). The potential role of Bifido- changes in specific NSP components would affect
bacterium spp. has been emphasized particularly in the populations of specific groups of colonic

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
228
bacteria, as was observed with RS. Significant that the bacterial strain composition of the colonic
decreases were observed for C. aerofaciens and for microbiota is subject to inter-individual variation.
the E. rectale group on the WL diet. The WL changes Furthermore, we show that this can be associated
do not show a simple relationship with RS and NSP with profound inter-individual differences in the
intakes, and it is possible that the increased dietary response of the microbial community to dietary
protein content of this diet might have a role in change, and in microbial fermentation of dietary
altering microbiota composition. substrates in the colon. This suggests that
The distribution of major bacterial phyla observed dietary advice on the consumption of ND carbo-
here did not depart dramatically from that reported hydrates might need to be personalized in the future.
in non-obese subjects (Eckburg et al., 2005;
Walker et al., 2008). The data that we obtained by
qPCR and clone library analysis are consistent with
recent reports on obese subjects using fluorescence
Acknowledgements
in situ hybridization microscopy (Duncan et al., We acknowledge support for this project from the World
2007, 2008; Schwiertz et al., 2010) or 16S rRNA Cancer Research Fund. The Rowett Institute of Nutrition
sequencing (Zhang et al., 2008) although lower % and Health (University of Aberdeen) and Biomathematics
Bacteroides have been reported in another study and Statistics Scotland received support from the Scottish
(Ley et al., 2006). More subtle differences may occur Government Rural Environment Research and Analysis
Directorate. We thank Claire Fyfe and the RINH Human
between the gut microbiota of obese and non-obese
Studies Unit. We also thank David Harris and the rest of
individuals at the species level, and indeed the his team at the Wellcome Trust Sanger Institute for
dietary responses reported here make this likely. generating the 16S rRNA gene sequences. Funding for
Nevertheless 5 of the 10 most abundant phylotypes AWW, MDS, PS, JP and sequencing was provided by
identified in this study group were also among the The Wellcome Trust (grant no. WT 76964).
10 most abundant phylotypes described by Tap et al.
(2009) in non-obese volunteers.
A high proportion of the most abundant 16S rRNA
phylotypes that we detected here corresponded to References
cultured bacteria, with 66% of the 50 phylotypes Abell GCJ, Cooke CM, Bennett CN, Conlon MA, McOrist AL.
that accounted for 40.5% of sequences having close (2008). Phylotypes related to Ruminococcus bromii
cultured relatives. This suggests that the limited are abundant in the large bowel of humans and increase
coverage of the human gut microbiota through in response to a diet high in resistant starch. FEMS
cultivation may be because of insufficient anaerobic Microbiol Ecol 66: 505–515.
isolation work, rather than to intrinsic non-cultur- Aminov RI, Walker AW, Duncan SH, Harmsen HJM,
ability of human colonic bacteria. Four of the five Welling GW, Flint HJ. (2006). Molecular diversity,
most abundant phylotypes detected here (E. rectale, cultivation, and improved FISH detection of a domi-
nant group of human gut bacteria related to Roseburia
F. prausnitzii, C. aerofaciens and B. vulgatus) spp. or Eubacterium rectale. Appl Environ Microbiol
corresponded with the most abundant species 72: 6371–6376.
reported by anaerobic cultivation in Japanese– Ashelford KE, Chuzhanova NA, Fry JC, Jones AJ, Weightman
Hawaiians, North American Caucasians and polyp AJ. (2005). At least 1 in 20 sequence records currently
patients (Moore and Moore, 1995). Overall, however, held in public repositories is estimated to contain
33.4% of phylotypes showed o98% identity with substantial anomalies. Appl Environ Microbiol 71:
cultured bacteria in this study, reflecting the 7724–7736.
relatively poor coverage of the less abundant Ashelford KE, Chuzhanova NA, Fry JC, Jones AJ, Weightman
bacterial groups by cultured strains. Many abundant AJ. (2006). New screening software shows that most
phylotypes were found to be widespread, with 32 of recent large 16S rRNA gene clone libraries contain
chimeras. Appl Environ Microbiol 72: 5734–5741.
the 320 phylotypes detected being present in all Bird AR, Brown IL, Topping DL. (2000). Starches, resistant
six individuals surveyed and accounting for almost starches, the gut microflora and human heath. Curr
half (47%) of all sequences. Issues Intest Microbiol 1: 25–37.
In conclusion, we find that increased intake of Bouhnik Y, Raskine L, Simoneau G, Vicaut E, Neut C,
RS, an important ND carbohydrate in the human Flourie B et al. (2004). The capacity of nondigestible
diet, can substantially alter the species composition carbohydrates to stimulate faecal bifidobacteria in
of the colonic microbiota. Such responses are likely healthy humans: a double blind, randomized, placebo-
to occur also with other fermentable ND dietary controlled, parallel-group, dose response relation
components. It follows that the colonic microbial study. Am J Clin Nutr 80: 1658–1664.
community must typically be in a state of conti- Brinkworth GD, Noakes M, Clifton PM, Bird AR. (2009).
Comparative effects of very low carbohydrate, high fat
nuous change over time, driven by short-term and high carbohydrate, low-fat weight loss diets
changes in dietary intake. Thus, only the most on bowel habit and faecal short chain fatty acids and
successful and versatile organisms will be found bacterial populations. Brit J Nutr 101: 1493–1502.
commonly among the dominant microbiota at Cani PD, Amar J, Iglesias MA, Poggi M, Knauf C, Bastelica D
different sampling times and in different indivi- et al. (2007). Metabolic endotoxemia initiates obesity
duals. In addition, however, our evidence shows and insulin resistance. Diabetes 56: 1761–1772.

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
229
Cole JR, Wang Q, Cardenas E, Fish J, Chai B, Farris RJ et al. Johnson M, Zaretskaya I, Raytselis Y, Merezhuk Y,
(2009). The Ribosomal Database Project: improved McGinnis S, Madden TL. (2008). NCBI BLAST:
alignments and new tools for rRNA analysis. Nucleic a better web interface. Nucleic Acids Res 36: W5–W9.
Acids Res 37: D141–D145. Kovatcheva-Datchary P, Egert M, Maathuis A,
Costello EK, Lauber CL, Hamady M, Fierer N, Gordon JI, Rajilic-Stojanovic M, de Graaf AA, Smidt H et al.
Knight R. (2009). Bacterial community variation in (2009). Linking phylogenetic identities of bacteria to
human body habitats across time and space. Science starch fermentation in an in vitro model of the large
326: 1694–1697. intestine by RNA-based stable isotope probing. Environ
Duncan SH, Belenguer A, Holtrop G, Johnstone AM, Microbiol 11: 914–926.
Flint HJ, Lobley GE. (2007). Reduced dietary intake of Iino T, Mori K, Tanaka K, Suzuki K-I, Harayama S. (2007).
carbohydrates by obese subjects results in decreased Oscillibacter valericigenes gen nov., a valerate-
concentrations of butyrate and butyrate-producing producing anaerobic bacterium isolated from the
bacteria in feces. Appl Environ Microbiol 73: alimentary canal of a Japanese corbicula clam. Int J
1073–1078. Syst Evol Microbiol 57: 1840–1845.
Duncan SH, Lobley GE, Holtrop G, Ince J, Johnstone AM, Lauber CL, Zhou N, Gordon JI, Knight R, Fierer N. (2010).
Louis P et al. (2008). Human colonic micro- Effect of storage conditions on the assessment of
biota associated with diet, obesity and weight loss. bacterial community structure in soil and human
Int J Obesity 32: 1720–1724. associated samples. FEMS Microbiol Lett 307: 80–85.
Eckburg PB, Bernstein CN, Purdom E, Dethlefsen L, Lawley TD, Clare S, Walker AW, Goulding D, Stabler RA,
Sargent M, Gill SR et al. (2005). Diversity of the Croucher N et al. (2009). Antibiotic treatment of
human intestinal microbial flora. Science 308: Clostridium difficile carrier mice triggers a super-
1635–1638. shedder state, spore-mediated transmission, and
Englyst HN, Cummings JH. (1988). Improved method for severe disease in immunocompromised hosts. Infect
measurement of dietary fiber as non-starch poly- Immun 77: 3661–3669.
saccharides in plant foods. J Assoc Off Anal Chem 1: Leitch ECM, Walker AW, Duncan SH, Holtrop G, Flint HJ.
247–286. (2007). Selective colonization of insoluble substrates by
Englyst HN, Kingman SM, Cummings JE. (1992). Classi- human colonic bacteria. Environ Microbiol 72: 667–679.
fication and measurement of nutritionally impor- Le Leu RK, Hu Y, Brown IL, Young GP. (2009). Effect of
tant starch fractions. Eur J Clin Nutr 46(Suppl 2): high amylose maize starches on colonic fermentation
S33–S50. and apoptotic response to DNA-damage in the colon
Food Standards Agency (2002). McCance and Widdow- of rats. BMC Nutr Metab 6: 11–19.
son’s The Composition of Foods , 6th summary edn. Ley RE, Turnbaugh PJ, Klein S, Gordon JI. (2006).
Royal Society of Chemistry: Cambridge. Microbial ecology—human gut microbes associated
Flint HJ, Bayer EA, Rincon MT, Lamed R, White BA. with obesity. Nature 444: 1022–1023.
(2008). Polysaccharide utilization by gut bacteria: Louis P, Young P, Holtrop G, Flint HJ. (2010). Diversity of
potential for new insights from genomic analysis. human colonic butyrate-producing bacteria revealed
Nat Rev Microbiol 6: 121–131. by analysis of the butyryl-CoA:acetate CoA transferase
Flint HJ, Duncan SH, Scott KP, Louis P. (2007). Inter- gene. Environ Microbiol 12: 304–314.
actions and competition within the microbial Ludwig W, Strunk O, Westram R, Richter L, Meier H,
community of the human colon: links between diet Yadhukumar et al. (2004). ARB: a software environ-
and health. Environ Microbiol 9: 1101–1111. ment for sequence data. Nucleic Acids Res 32:
Florin THF, Zhu G, Kirk KM, Martin NG. (2000). Shared 1363–1371.
and unique environmental factors determine the Macfarlane GT, Englyst HN. (1986). Starch utilization by
ecology of methanogens in humans and rats. Am J the human large intestinal microflora. J Appl Bacteriol
Gastroenterol 95: 2872–2879. 60: 195–201.
Frank JA, Reich CI, Sharma S, Weisbaum JS, Wilson BA, Mackie RI, Aminov RI, Hu W, Klieve AV, Ouwekerk D,
Olsen GJ. (2008). Critical evaluation of two primers Sundset MA et al. (2003). Ecology of uncultivated
commonly used for amplification of bacterial 16S Oscillospira species in the rumen of cattle, sheep and
rRNA genes. Appl Environ Microbiol 74: 2461–2470. reindeer as assessed by microscopy and molecular
Franks AH, Harmsen HJM, Raangs GC, Jansen GJ, Schut F, approaches. Appl Environ Microbiol 69: 6808–6815.
Welling GW. (1998). Variations of bacterial popula- McIntyre AP, Gibson P, Young GP. (1993). Butyrate
tions in human feces measured by fluorescent in situ production from dietary fibre and protection against
hybridisation with group-specific 16S rRNA-targeted large bowel cancer in a gut model. Gut 34: 386–391.
oligonucleotide probes. Appl Environ Microbiol 64: Moore WEC, Moore LH. (1995). Intestinal floras of
3336–3345. populations that have a high risk of colon cancer.
Furrie E, Macfarlane S, Kennedy A, Cummings JH, Appl Environ Microbiol 61: 3202–3207.
Walsh SV, O’Neill DA et al. (2005). Synbiotic therapy Polednak AP. (2003). Trends in incidence rates for obesity-
(Bifidobacterium longum/Synergy 1) initiates resolu- associated cancers in the U.S.. Cancer Detect Prev 27:
tion of inflammation in patients with active ulcerative 415–421.
colitis: a randomised controlled pilot trial. Gut 54: Ramirez-Farias C, Slezak K, Fuller Z, Duncan A, Holtrop G,
242–249. Louis P. (2009). Effect of inulin on the human gut
Gill CIR, Rowland IR. (2002). Diet and cancer: assessing microbiota: stimulation of Bifidobacterium adolescentis
the risk. Br J Nutr 88(Suppl 1): S73–S87. and Faecalibacterium prausnitzii. Br J Nutr 101: 541–550.
Hold GL, Pryde SE, Russell VJ, Furrie E, Flint HJ. (2002). Ramsay AG, Scott KP, Martin CJ, Rincon MT, Flint HJ.
Assessment of microbial diversity in human colonic (2006). Cell-associated a-amylases of butyrate-producing
samples by 16S rDNA sequence analysis. FEMS Firmicute bacteria from the human colon. Microbiology
Microbiol Ecol 39: 33–39. 152: 3281–3290.

The ISME Journal

 Impact of diet on human colonic microbiota
AW Walker et al
230
Robert C, Bernalier-Donadille A. (2003). The cellulolytic Stephen AM, Wiggins HS, Cummings JH. (1987). Effect of
microflora of the human colon: evidence of changing transit time on colonic microbial metabolism
microcrystalline cellulose-degrading bacteria in in man. Gut 28: 601–609.
methane-excreting subjects. FEMS Microbiol Ecol 46: Suau A, Bonnet R, Sutren M, Godon JJ, Gibson GR, Collins
81–89. MD et al. (1999). Direct analysis of genes encoding 16S
Robertson MD, Bickerton AS, Dennis AL, Vidal H, Frayn KN. rRNA from complex communities reveals many novel
(2005). Insulin-sensitizing effects of dietary resistant molecular species within the human gut. Appl
starch and effects on skeletal muscle and adipose tissue Environ Microbiol 24: 4799–4807.
metabolism. Am J Clin Nutr 82: 559–567. Tap J, Mondot S, Levenez F, Pelletier E, Caron C, Furet JP
Salyers AA, West SEH, Vercellotti JR, Wilkins TD. (1977). et al. (2009). Towards the human intestinal microbiota
Fermentation of mucins and plant polysaccharides phylogenetic core. Environ Microbiol 11: 2574–2584.
by anaerobic bacteria from the human colon. Appl Turnbaugh PJ, Backhed F, Fulton L, Gordon JI. (2008).
Environ Microbiol 34: 529–533. Diet-induced obesity is linked to marked but rever-
Schloss PD, Handelsman J. (2005). Introducing DOTUR, a sible alterations in the mouse distal gut microbiome.
computer program for defining operational taxonomic Cell Host Microbe 3: 213–223.
units and estimating species richness. Appl Environ Walker AW, Duncan SH, Harmsen HJM, Holtrop G,
Microbiol 71: 1501–1506. Welling GW, Flint HJ. (2008). The species composition
Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, of the human intestinal microbiota differs between
Hollister EB et al. (2009). Introducing mothur: particle-associated and liquid phase communities.
open source, platform-independent, community- Environ Microbiol 10: 3275–3283.
supported software for describing and comparing Wang X, Conway PL, Brown IL, Evans AJ. (1999). In vitro
microbial communities. Appl Environ Microbiol 75: utilization of amylopectin and high amylose maize
7537–7541. (amylomaize) starch granules by human colonic
Schwiertz A, Lehmann U, Jacobasch G, Blaut M. (2002). bacteria. Appl Environ Microbiol 65: 4848–4854.
Influence of resistant starch on the SCFA production Yue JC, Clayton MK. (2005). A similarity measure based
and cell counts of butyrate-producing Eubacterium spp. on species proportions. Commun Stat Theo Methods
in the human intestine. J Appl Microbiol 93: 157–162. 34: 2123–2131.
Schwiertz A, Tara D, Schafer K, Beijer S, Bos NA, Donus C Zoetendal EG, Akkermans ADL, de Vos WM. (1998).
et al. (2010). Microbiota and SCFA in lean and Temperature gradient gel electrophoresis analysis of
overweight healthy subjects. Obesity 18: 190–195. 16S rRNA from human faecal samples reveals stable
Sokol H, Pigneur B, Watterlot L, Lakhdari O, Bermúdez- and host-specific communities of active bacteria. Appl
Humarán LG, Gratadoux J et al. (2008). Faecalibacterium Environ Microbiol 64: 3854–3859.
prausnitzii is an anti-inflammatory commensal bac- Zhang H, DiBaise JK, Zuccolo A, Kudma D, Braidotti M,
terium identified by gut microbiota analysis of Yu Y et al. (2008). Human gut microbiota in obesity
Crohn’s disease patients. Proc Natl Acad Sci USA 105: and after gastric bypass. Proc Natl Acad Sci USA 106:
16731–16736. 2365–2370.

Supplementary Information accompanies the paper on The ISME Journal website (http://www.nature.com/ismej)

The ISME Journal