The Pharmacogenomics Journal (2013) 13, 514–522

& 2013 Macmillan Publishers Limited All rights reserved 1470-269X/13

www.nature.com/tpj

ORIGINAL ARTICLE
Metagenomic sequencing of the human gut microbiome before
and after bariatric surgery in obese patients with type 2 diabetes:
correlation with inflammatory and metabolic parameters
J Graessler1,6, Y Qin2,6, H Zhong2,6, J Zhang2,6, J Licinio3, M-L Wong3, A Xu4, T Chavakis1, AB Bornstein1, M Ehrhart-Bornstein1,
V Lamounier-Zepter1, T Lohmann5, T Wolf5 and SR Bornstein1,6

Roux-en-Y gastric bypass (RYGB) has become a prominent therapeutic option for long-term treatment of morbid obesity
and type 2 diabetes mellitus (T2D). Cross talk and pathogenetic consequences of RYGB-induced profound effects on metabolism
and gut microbiome are poorly understood. The aim of the present study therefore was to characterize intra-individual
changes of gut microbial composition before and 3 months after RYGB by metagenomic sequencing in morbidly obese patients
(body mass index (BMI)440 kg m  2) with T2D. Subsequently, metagenomic data were correlated with clinical indices. Based on
gene relative abundance profile, 1061 species, 729 genera, 44 phyla and 5127 KO (KEGG Orthology) were identified. Despite high
diversity, bacteria could mostly be assigned to seven bacterial divisions. The overall metagenomic RYGB-induced shift was
characterized by a reduction of Firmicutes and Bacteroidetes and an increase of Proteobacteria. Twenty-two microbial species and
11 genera were significantly altered by RYGB. Using principal component analysis, highly correlated species were assembled into
two common components. Component 1 consisted of species that were mainly associated with BMI and C-reactive protein. This
component was characterized by increased numbers of Proteobacterium Enterobacter cancerogenus and decreased Firmicutes
Faecalibacterium prausnitzii and Coprococcus comes. Functional analysis of carbohydrate metabolism by KO revealed significant
effects in 13 KOs assigned to phosphotransferase system. Spearmen’s Rank correlation indicated an association of 10 species with
plasma total- or low-density lipoprotein cholesterol, and 5 species with triglycerides. F. prausnitzii was directly correlated to fasting
blood glucose. This is the first clinical demonstration of a profound and specific intra-individual modification of gut microbial
composition by full metagenomic sequencing. A clear correlation exists of microbiome composition and gene function with an
improvement in metabolic and inflammatory parameters. This will allow to develop new diagnostic and therapeutic strategies
based on metagenomic sequencing of the human gut microbiome.

The Pharmacogenomics Journal (2013) 13, 514–522; doi:10.1038/tpj.2012.43; published online 2 October 2012
Keywords: bariatric surgery; diabetes mellitus; inflammation; metabolism; metagenome; Roux-en-Y gastric bypass

INTRODUCTION phosphotransferase systems involved in microbial processing of

Increasing evidence indicates that gut microorganisms may have a carbohydrates primarily present in Actinobacteria and Firmicutes.4
critical role in determining obesity in addition to hypercaloric diet In contrast to these data, Schwiertz et al.5 demonstrated in a study
and physical inactivity against an unfavourable genetic back- on human volunteers an altered composition of gut microbiota in
ground. Several mechanisms have been predicted through which favour of Bacteroidetes in overweight and obese subjects with a
intestinal microorganisms may facilitate weight gain and fat concomitant reduction of Firmicutes. These changes were
accumulation. Thus, intestinal microorganisms could be contribut- associated with an increased faecal amount of short-chain fatty
ing to obesity by increasing recovery of energy from the diet1 and acids in general and propionate in obese individuals in particular,
via the impact of microbial metabolites or microbial cell-derived which may serve as a precursor for gluconeogenesis and
signals on host pathways that regulate energy homoeostasis and liponeogenesis in liver metabolism.5
lipid metabolism.1 Several gut microbial diversity surveys in Recent studies have also pointed to a close relationship
mouse models2 and humans3,4 provided evidence that obesity between the composition of intestinal microbiota and diabetes.
was associated with a decreased proportion of Bacteroidetes and Proportions of Firmicutes and Clostridia were found significantly
a higher proportion of Firmicutes. Subsequent functional analysis diminished in the diabetic group compared with the nondiabetic
revealed that the obese human gut microbiome was enriched for controls.6 Furthermore, the ratios of Bacteroidetes to Firmicutes
1
Department of Internal Medicine III, Carl Gustav Carus Medical School, Technical University Dresden, Dresden, Germany; 2Research and Cooperation Division, BGI-Shenzhen,
Shenzhen, China; 3Department of Translational Medicine, John Curtin School of Medical Research, Australian National University, Canberra, ACT, Australia; 4Department of
Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong, China and 5Department of Internal Medicine, Dresden Interdisciplinary Adiposity Center,
Community Hospital Dresden-Neustadt, Dresden, Germany. Correspondence: Professor SR Bornstein, Department of Internal Medicine III, Carl Gustav Carus Medical School,
Technical University Dresden, Fetscherstrasse 74, Dresden D-01307, Germany.
E-mail: Stefan.Bornstein@uniklinikum-dresden.de
6
These authors equally contributed to this work.
Received 6 August 2012; revised 15 August 2012; accepted 20 August 2012; published online 2 October 2012

correlated positively and significantly with plasma glucose
concentration but not with body mass index (BMI).6 Beyond
metabolic microbiota–host cross talk, proposed mechanisms for
the role of gut microbiota in diabetes also include an increased
intestinal permeability causing elevated systemic levels of
lipopolysaccharides promoting insulin resistance and hyperinsu-
linemia.7 In some studies, modification of gut microbiota induced
by antibiotics or prebiotics resulted in a reduced inflammatory
activity accompanied by a recovery of insulin sensitivity and
decreased obesity.7
Bariatric surgery has become a prominent therapeutic option to
treat morbid obesity. Roux-en-Y gastric bypass (RYGB) is the major
bariatric intervention that is associated with striking weight
reduction, impressive improvements of glycemic control and
inflammatory status in diabetic patients.8,9 However, so far the
mechanisms of these profound changes remain largely
unexplored. Detailed studies are necessary to link both the
preoperative metabolic state and the postoperative changes to
gut microbiota composition. There are only very few studies
addressing the host metabolic-microbial cross talk after RYBG. In a
study on a non-obese rat model, RYGB resulted in decreased
amounts of Firmicutes and Bacteroidetes but a 52-fold higher
concentration of Proteobacteria in comparison with sham-
operated rats.10 In a preliminary study on humans, Zhang
et al.11 demonstrated a significant reduction of Firmicutes in
post-gastric-bypass individuals, who had concomitantly a
proportional increase of Gammaproteobacteria. Based on a real-
time quantitative PCR profiling of gut microbiota in faecal
samples, Furet et al.12 found an increase in the Bacteroides/
Prevotella group 3 months after RYGB, which was lower
preoperatively in obese subjects relative to lean controls. In the
same study, lactic acid bacteria including Lactobacillus/
Leuconostoc/Pediococcus group and Bifidobacterium genus
decreased 3 months after RYGB.12 Regarding diabetes and
inflammatory state, the butyrate-producing species F. prausnitzii
was found to be negatively associated with inflammatory markers
before RYGB and throughout the postoperative period, indicating
its contribution to a healthy intestine independently of changes in
food intake.12
Metagenomic sequencing is a powerful tool for a complete
qualitative and quantitative analysis of the gut microbiome,
including microbial species of very low abundance. By applying it
to the human gut, 3.3 million nonredundant microbial genes from
faecal samples of 124 European individuals have been identified.13
The objective of this study was a comprehensive characteriza-
tion by metagenomic sequencing of faecal microbiota composi-
tion in morbidly obese individuals with type 2 diabetes before and
3 months after RYGB in order to determine significant changes of
high- and low-abundance microbial species. In this pilot study of
six patients, we identified 22 microbial species from the complete
metagenome, consisting of 1061 species, which were significantly
affected by RYGB and may be associated with postoperative
metabolic changes.
MATERIALS AND METHODS
Patient background
Six patients for Roux-en-Y gastric-bypass operation were recruited
according to the S3 guidelines of the German Society for Obesity (ed.
June 2010) in the Dresden Interdisciplinary Adiposity Centre at the
Department of Internal Medicine of the Community Hospital Dresden-
Neustadt. Morbidly obese patients were evaluated for bariatric surgery by
an interdisciplinary team and were included in the study if they fulfilled the
criteria for bariatric surgery (BMI 440 kg m  2 with comorbidity, age below
60 years, 2 years of unsuccessful conservative treatment and approval of
surgery by their health insurance). Each patient provided written informed
consent and the study was approved by the Dresden Ethics Committee
(EK280082011) in accordance with the Declaration of Helsinki principles.
& 2013 Macmillan Publishers Limited
Metagenomic analysis in RYGB-treated patients with diabetes
J Graessler et al
515
A total of six morbidly obese patients (three men and three
women; BMI440 kg m  2) were investigated before and 3 months after
Roux-en-Y gastric-bypass surgery. Five of these patients (two men and
three women) had manifest type 2 diabetes mellitus (T2D). Five of the
patients had not received any antibiotics for at least 3 months before
sampling and during observation period. One patient (number 5)
received penicillin for 6 days owing to tonsillitis 3 weeks before the
postoperative sample has taken. Basal anthropometric and clinical data are
presented in Table 1.
Blood samples for clinical chemistry and hormonal profile were taken
after overnight fast. Ethylenediaminetetraacetic acid–plasma and serum
were prepared by 10-min centrifugation at 4 1C and 3000 g. All samples
were immediately frozen and stored at  80 1C until analysed.
Clinical chemistry indices
Plasma triglycerides, total cholesterol, high-density lipoprotein and low-
density lipoprotein (LDL) cholesterol were determined by standard
enzymatic methods on a MODULAR analyser (Roche, Indianapolis, IN,
USA), free fatty acid on a COBAS MIRA analyser (Global Medical
Instrumentation, Ramsey, MN, USA) and plasma glucose on a DX80
analyser (Beckman-Coulter, Fullerton, CA, USA). HbA1c was measured by
high-performance liquid chromatography (Bio-Rad Laboratories, Rich-
mond, CA, USA).
Gastric-bypass surgery
Gastric-bypass surgery by RYGB included the creation of a small stomach
pouch using a stapler device, and its connection to the distal small
intestine. The upper part of the small intestine was then reattached in a
Y-shaped configuration. The stomach pouch is too small to hold large
amounts of food, and by skipping the duodenum, fat absorption is
substantially reduced.
Human faecal sample collection and DNA isolation
Faecal samples were collected by paper-based stool collectors
(Suess, Gudensberg, Germany) and transferred into sterile tubes with
sterile screw caps (Sarstedt, Nümbrecht, Germany). Until analysis, samples
were kept at  80 1C. For DNA isolation, a frozen aliquot (200 mg) of each
faecal sample was suspended in 250 ml of guanidine thiocyanate, 0.1 M Tris
(pH 7.5) and 40 ml of 10% N-lauroyl sarcosine. Mechanical cell
disruption was performed using a Retsch mixer mill MM400 (Retsch, Haan,
Germany) with glass beads. Then, DNA extraction was performed
as previously described.14 The DNA concentration and its molecular
size were estimated by Nanodrop (Thermo Scientific, San Diego, CA, USA)
and agarose gel electrophoresis.
DNA library construction and sequencing
DNA library construction was performed following the manufacturer’s
instruction (Illumina, Wilmington, DE, USA). We used the standard
workflow as described elsewhere to perform cluster generation, template
hybridization, isothermal amplification, linearization, blocking and dena-
turalization, and hybridization of the sequencing primers.
We constructed one paired-end (PE) library with insert size of
350 bp for each sample (the actual insert size ranged from 321 to
343 bp), followed by high-throughput sequencing to obtain B15 million
PE reads. The reads length for each end was 100 bp. Through
filtering low-quality reads containing ‘N’ base, adaptor contamination
reads and human host DNA contamination reads, we obtained the high-
quality reads.
Computation of relative abundance
High-quality reads of each sample were mapped to the human gut gene
catalogue13 using SOAP215 with the criterion of identity X 90%. When we
translated the aligned results to the gene relative abundance, only two
types of mapping results were considered: (a) a PE read should be mapped
onto a gene with the correct insert–size relationship; (b) one end of the PE
reads should be mapped onto the end of a gene with assuming the other
end of read was mapped outside the gene region. In both cases, the
mapped read was only counted as one copy.
The Pharmacogenomics Journal (2013), 514 – 522
 Metagenomic analysis in RYGB-treated patients with diabetes
J Graessler et al
516
Table 1. Basal anthropometric and clinical data of patients before and 3 month after Roux-en-Y gastric bypass

Op-status Patient Patient Patient Patient Patient Patient

1 2 3 4 5 6

Sex Male Male Female Male Female Female

Age (years) 53 46 42 38 45 53
BMI (kg m  2) Before 52.1 40.9 49.0 48.5 43.9 42.3
3 Months after 42.5 34.5 41.5 37.1 33.8 28.7
RR systolic (mm Hg) Before 180 140 144 135 140 173
3 Months after 125 120 125 120 130 130
RR diastolic (mm Hg) Before 110 100 98 80 90 90
3 Months after 80 80 80 80 80 80
Triglycerides (mM) Before 1.21 1.68 3.09 6.48 1.96 4.06
3 Months after 1.47 1.27 1.47 2.08 1.08 0.90

Total cholesterol (mM) Before 5.34 5.22 6.00 7.04 4.92 6.71
3 Months after 4.59 3.98 5.20 5.44 4.33 4.70
LDL-cholesterol (mM) Before 3.50 3.22 3.70 4.20 3.01 3.91
3 Months after 2.73 2.48 3.42 3.55 2.66 2.90
HDL cholesterol (mM) Before 0.95 1.15 1.30 1.10 1.23 1.30
3 Months after 1.05 1.10 1.24 0.73 1.14 1.56
Diabetes status Before Yes No Yes Yes Yes Yes
Since 1989 2006 2009 2006 2010
Diabetes medication Before Insulin NA Metformin Metformin Metformin Metformin
3 Months after Insulin NA Metformin Diet Diet Diet
HbA1c (%) Before 6.7 5.2 5.4 9.0 6.2 6.7
3 Months after 6.6 5.1 5.5 7.2 6.0 6.2

Plasma glucose (mM) Before 6.24 6.06 4.42 8.82 7.92 4.77
3 months after 8.75 4.97 3.87 4.81 5.02 5.04
Plasma cortisol (nM) Before 374 99 233 171 205 79
3 Months after 176 370 136 304 220 168
Abbreviations: BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; NA, not applicable.

Then, for any sample S the relative abundance was calculated as follows: testing (Bonferroni procedure) was used to identify RYGB-induced
Step 1: calculate the copy number of each gene: differences for components 1 and 2.
Spearman rank correlation tests were applied to evaluate the correlation
xi between species abundances and BMI or C-reactive protein (CRP). In
bi ¼
Li addition, a partial correlation analysis with BMI as cofactor has been
Step 2: calculate the relative abundance of gene i performed to test an independent association of CRP with relative species
abundances.
bi
xi Data in Tables 3 and 4 are given as means±s.e.m. A value of Po0.05
ai ¼ P ¼ PLi xj was considered statistically significant.
j bj j Lj

where ai is the relative abundance of gene i in sample S, Li is the length RESULTS

of gene i, xi is the times that gene i can be detected in sample S (the
number of mapped reads) and bi is the copy number of gene i in the
sequenced data from sample S.
Based upon human gut gene catalogue’s annotation information, genes
can be assigned to different taxonomy levels and KEGG Orthologues (KOs).
For species profile, corresponding gene abundances of the same species
were summed. This gross relative abundance was taken as this species’
relative abundance. The genus, phylum and KO relative abundance were
constructed using the same method used for species. For further analysis,
each profile must be normalized to the column sum of each profile.
Statistic test analysis
All statistical analyses were performed with the SPSS statistical package
(v.18.0 for Windows; SPSS, Chicago, IL, USA). For comparisons between pre-
and postoperative abundances at species and genera level, an univariate
analysis of variance was performed. These results were confirmed using a
nonparametric Kolmogorov–Smirnov test.
In order to assemble highly correlated species into common factors and
thereby to obtain a multivariate preliminary data survey, a ‘Principal
Component Analysis’ of all 22 previously identified microbial species was
performed. Using ‘Eigenvalues’ over 2.5 and ‘Varimax’ rotation with Kaiser
Normalization, two factors (components) were extracted. Factor composi-
tion and factor loadings are shown in Table 4. Subsequent multivariate
analysis was realized with component 1 and 2 as dependent variables,
status before and after bypass as independent factor, and BMI as covariate.
Consecutively, a univariate analysis with P-values corrected for multiple
Clinical characteristics of the study population
According to the guidelines for bariatric surgery, three men and
three women, age 38–53 years, preoperative BMI 40.9–
52.1 kg m  2, were recruited for RYGB (Table 1). All three women
and two of the men were diagnosed having T2D for 1.5–22 years.
Patient no.1 with long-lasting T2D was preoperatively treated with
insulin, all remaining patients with metformin. Glycemic control
based on preoperative fasting values of blood glucose and HbA1c
revealed that four of the T2D patients were treated adequately
(HbA1cp6.7%), one patient (no.4) was chronically hyperglycemic
(HbA1c ¼ 9.0; Table 1). Preoperatively, all six patients were
hypertensive, four had hypertriglyceridemia, and five had elevated
total cholesterol (Table 1).
Three months after RYGB, BMI was decreased by 15–32% in all
patients (Table 1). There was a concurrent reduction in systolic (by
10–55 mmHg) and in 5 of 6 cases of diastolic blood pressure
(Table 1). Changes in plasma lipid profile were characterized by
reduction of triglycerides (5 from 6 cases), total- and LDL-
cholesterol (all cases). Plasma high-density lipoprotein cholesterol
increased in two of six cases (Table 1). Postoperative reduction of
HbA1c mainly depended on baseline values and ranged between
1.8 and 0.1%. Four of six patients revealed changes in HbA1c lower
than 0.2%, indicating a stable metabolic situation before surgery
(Table 1).

The Pharmacogenomics Journal (2013), 514 – 522 & 2013 Macmillan Publishers Limited

Figure 1. Effect of Roux-en-Y gastric bypass (RYGB) on taxonomy
composition of phyla pre/postoperative abundances: Firmicutes (F)
47.2/34.2%; Bacteroidetes (B) 46.9/44.7%; (B/F ratio: 0.99/1.33);
Actinobacteria 1.7/1.2%; Cyanobacteria 0.10/0.06%; Proteobacteria
2.68/13.74%; Verrucomicrobia 0.65/5.37%; Fusobacteria 0.19/0.23%.
Based on clinical chemistry data, the preoperative situation was
characterized by high–normal CRP values (2.1–5.6 mg l  1) in all
patients, indicating a latent activated general inflammation,
partially increased liver enzymes (patients 2, 3 and 4) and a
tendency of higher uric acid levels in all patients (Supplementary
Table A). Three months after bariatric surgery, CRP values
decreased in five of six patients (except patient 4). All patients
had lower postoperative plasma values for g-glutamyltransferase
and uric acid. In five of six patients, reduced plasma concentration
were observed for ALAT (except patient 3) and total serum protein
(except patient 4). Plasma concentrations of transferrin mostly
remained within the preoperative range (Supplementary Table A).
Taken together, these data confirm that the acute loss of body
weight is associated with a significant improvement of the
metabolic situation, including glucose and lipid metabolism, a
marked reduction of blood pressure and an attenuated overall
inflammatory state, which is regarded as an indicator for
cardiovascular risk.
Metagenomics sequencing of gut microbial community
Twelve faecal specimens from six patients were collected, both
before and 3 months after surgery. Total DNA was extracted from
the faecal specimens and an average of 3.6 Gb (ranging from 3.1
to 4.1 Gb) of sequencing data was generated for each sample.
After removing the human contamination reads and low-quality
reads, the remaining reads rate is between 92.91 and 99.61 and
96.55% on average, allowing us to capture most of the repertoire
of the gut microbial flora.
In order to obtain the gene relative abundance profiling, all
reads were mapped individually to the human gut microbial gene
catalogue from the faecal samples of 124 adult Europeans,
MetaHIT project, by SOAP2.13
The mapping rate ranges from 55.03 to 70.91% (on average
63.62%), which was comparable to the mapping rate of our own
sample used for constructing the gene set, suggesting the
reference catalogue has a sufficient coverage for the severely
obese samples in this study.
Based on the gene relative abundance, profile phylum, genus
and species profiles were acquired. There were 45.21%, 58.89%,
72.72% and 47.03% of the human gut genes that have species,
genus, phylum and kegg function annotation information,
respectively. Finally, 1061 species, 729 genera, 44 phyla and
5127 KOs were subjected to the further analysis.
& 2013 Macmillan Publishers Limited
Metagenomic analysis in RYGB-treated patients with diabetes
J Graessler et al
517
Taxonomy composition of the samples
To simplify the comparison, only those phyla with average
abundance 40.1% in at least one group are displayed in
Figure 1. Although bacteria in the human intestinal community
are highly diverse, they fell mainly into seven bacterial divisions, of
which Firmicutes, Bacteroidetes, Actinobacteria, Proteobacteria
and Verrucomicrobia were the most abundant phyla in the human
gut intestinal.16 Four of the top seven high abundance phyla were
decreased in postoperative samples, including Firmicutes (from
47.2 to 34.2%), Bacteroidetes (from 46.9 to 44.7%), Actinobacteria
(from 1.7 to 1.2%) and Cyanobacteria (from 0.10 to 0.06%)
(Figure 1). However, the ratios of Bacteroidetes/Firmicutes shifted
from 0.99 to 1.31, showing an apparent increase. Three other
phyla were increasing: Proteobacteria (from 2.68 to 13.74%),
Verrucomicrobia (from 0.65 to 5.37%) and Fusobacteria (from 0.19
to 0.23%) (Figure 1).
At species taxonomy level of the 31 high-abundance species
(with average abundance 41% in at least one group), Akker-
mansia muciniphila, Escherichia coli and Klebsiella pneumonia
showed a clear tendency of increase (Supplementary Figure A). On
the other hand, F. prausnitzii, Eubacterium rectale and Dialister
invisus revealed a strong tendency to decrease (Supplementary
Figure A). The RYGB-induced changes at species level were
consistent with changes at genus level, where relative abundances
of Faecalibacterium and Eubacterium decreased and those of
Akkermansia and Escherichia postoperatively increased
(Supplementary Figure B).
Effect of RYGB on gut microbiome
Univariate analysis of variance was used to identify significant
differences in relative abundances of microbial species, genera
and phyla between pre-and postoperative samples. Significant
differences could be established for 22 microbial species, 11
genera and 1 phylum. These results have been confirmed by a
nonparametric Kolmogorov–Smirnov test. As shown in Table 2, 3
months after RYGB, relative abundances increased significantly for
E. cancerogenus (by 1992%, P ¼ 0.002), Veillonella parvula (by
184%, P ¼ 0.035), V. dispar (by 532%, P ¼ 0.038), Shigella boydii (by
4649%, P ¼ 0.039) and Salmonella enterica (by 652%, P ¼ 0.043). It
is of high interest that postoperative abundances of these species
were also significantly higher than those in lean controls (Table 2).
These changes appear to reflect profound changes in the milieu
interior of the gut rather than decrease in body weight.
At the same time, relative abundances decreased in 17 species,
most pronounced in Treponema pallidum (by 87%, P ¼ 0.011),
Mycobacterium kansasii (by 74%, P ¼ 0.011), F. prausnitzii (by 70%,
P ¼ 0.013) and C. comes (by 58%, P ¼ 0.013) (Table 2). Among the
species that decreased significantly 3 months after gastric bypass
were also two Lactobacilli species. Whereas postoperative
abundances of Lactobacillus reuteri were significantly lower than
in lean controls, L. acidophilus did not reveal any differences
between pre- or postoperative values and lean controls.
Preoperative abundances of Clostridium spiroforme, Brachyspira
hyodysenteriae and Fusobacterium periodonticum were significantly
lower than in lean controls and continued to decrease post-
operatively (Table 2). Same general tendency could be observed
for Anaerostipes caccae, F. periodonticum, Thermomicrobium
roseum, Staphylococcus epidermidis and Fibrobacter succinogenes,
which, however, did not show statistically significant lower
preoperative values than controls, but reaching statistical sig-
nificance after continued postoperative decline (Table 2).
Univariate analysis of variance for selective species has been
performed to identify significant RYGB effects on microbial genera
(Table 3). Among 12 genera, which were significantly affected, 5
genera increased postoperatively: Enterobacter, Neurospora,
Citrobacter, Veillonella and Salmonella (Table 3). Basically, with
regard to Enterobacter, Veillonella and Salmonella, this is in
The Pharmacogenomics Journal (2013), 514 – 522
 Metagenomic analysis in RYGB-treated patients with diabetes
J Graessler et al
518
accordance with postoperatively increasing species. In addition,

0.901
0.001
0.319
0.243

0.373
o0.001
o0.001
0.029
0.005
0.462
o0.001
0.032
0.033
0.097
0.033
0.097
0.103
0.005
0.182
0.970
0.083
0.023
Relative abundances of microbial species, which were significantly affected 3 month after Roux-en-Y gastric-bypass surgery. Comparison with lean control subjects (BMI o25 kg m  2)

B vs C
an augmentation of relative abundances of Neurospora and
Bonferroni P ¼
Citrobacter has been registered. Lacking indication for selective
Post-hoc

species of these genera may be caused by high interindividual

variations and a low number of observations. The strong tendency
of a postoperative increase of genus Shigella (P ¼ 0.056) was
A vs C

0.087
1.000
0.923
1.000

0.331
0.025
0.021
0.419
0.275
1.000
0.010
1.000
1.000
1.000
1.000
1.000
1.000
0.099
1.000
0.374
1.000
1.000
confirmed by an augmentation of the corresponding species
S. boydii (Table 3). However, seven genera decreased 3 months
after RYGB, whereby five of them were in accordance with
postoperative changes of selected species: Faecalibacterium,
Coprococcus, Helicobacter, Anaerostipes and Nakamurella
A–C Pp

0.010
o0.001
0.043
0.206

0.016
o0.001
o0.001
0.031
0.006
0.208
o0.001
0.031
0.016
0.029
0.015
0.049
0.032
0.006
0.160
0.054
0.048
0.020
ANOVA

(Table 3). Without reflecting changes in corresponding selective

species, a RYGB-induced reduction of genera Dictyostelium and
Epidinium has been found.
Including all 22 species, which were significantly affected by
RYGB, a principal component analysis (PCA) was performed
(1.168  10 )
(4.944  10  3)
(5.938  10  6)

(0.659  10  3)
(0.338  10  3)
(0.217  10  4)
(2.423  10  5)
(0.361  10  3)
(0.607  10  5)
(0.785  10  5)
(0.580  10  5)
(1.063  10  4)
(1.130  10  4)
(0.544  10  6)
(0.939  10  5)
(0.760  10  4)
(1.161  10  5)
(1.264  10  5)
(0.233  10  5)
(2.122  10  6)
(0.781  10  4)
(Table 4). Using eigenvalues over 2.5 and Varimax rotation with
5

Kaiser normalization two factors (components) were extracted,

Control (±s.e.m.)

(0.0075)

explaining total variance by 63.0% (component 1: 35.4%;

component 2: 27.6%). Component 1 included 11 species and
was mainly determined by L. acidophilus bearing factor loading
C

5

12.506  10  7
11.313  10–6

7.781  10  3
2.232  10  3
2.576  10  4
9.784  10  5
3.176  10  3
1.442  10  5
9.549  10  5
3.755  10  5
3.872  10  4
3.063  10  4
1.054  10  6
5.088  10  5
1.935  10  4
5.655  10  5
3.385  10  5
1.419  10  5
8.845  10  6
8.662  10  4
0.053

0.920. Component 2 also combined 11 species with Leptospira

1.694  10

interrogans as species of highest factor loading (Table 4).

Subsequent multivariate analysis was performed with compo-
nents 1 and 2 as dependent variables, status before and after
bypass as independent factor and BMI as covariate. A significant
and independent effect of operation status (Pillai’s trace P ¼ 0.007)
emerged, but not of BMI (Pillai’s trace P ¼ 0.146). Subsequent
A vs B Pp

univariate analysis with P-values corrected for multiple tests

ANOVA

0.002
0.011
0.012
0.013
0.018
0.020
0.022
0.023
0.025
0.026
0.030
0.034
0.035
0.038
0.039
0.040
0.043
0.044
0.045
0.045
0.046
0.046

(Bonferroni procedure) indicated a significant RYGB-induced

difference for component 2 (P ¼ 0.005), but not for component
1 (P ¼ 0.704). This led to the suggestion that component 1 might
include a group of species that is primarily correlated with
changes in BMI. A nonparametric Spearman rank correlation test
was performed to characterize the relationship of selected
(3.906  10 )
(1.418  10  7)
(0.665  10  6)

(0.891  10  3)
(0.115  10  3)
(0.101  10  4)
(0.747  10  5)
(0.230  10  3)
(0.108  10  5)
(0.523  10  5)
(0.315  10  5)
(2.919  10  4)
(3.683  10  4)
(17.38  10  6)
(0.582  10  5)
(2.736  10  4)
(0.297  10  5)
(0.273  10  5)
(0.198  10  5)
(0.445  10  6)
(0.622  10  4)
5

microbial species and BMI. The results indicated that 8 of 11

After surgery (±s.e.m.)

(0.00852)

species in component 1, but only 2 species of 11 in component 2,

were significantly correlated to BMI (Table 4). This might explain
the annulling effect of BMI as covariate in multivariate analysis. On
B

5

2.361  10  7
1.872  10  6

4.649  10  3
0.549  10  3
1.032  10  4
3.275  10  5
1.726  10  3
0.561  10  5
3.450  10  5
1.953  10  5
11.559  10  4
10.499  10  4
42.29  10  6
1.942  10  5
7.360  10  4
1.490  10  5
1.074  10  5
0.816  10  5
1.384  10  6
5.259  10  4
0.031

the other hand, the majority of species in component 2 seem to

16.902  10

be associated with BMI-independent effects of RYGB. The

individual effects of gastric bypass on the shift of components 1
and 2 are shown in Figure 2.
In addition to species composition, functional analysis of
carbohydrate metabolism by relative abundances of KO has been
performed. Thirteen KOs assigned to phosphotransferase system
Abbreviations: ANOVA, analysis of variance; BMI, body mass index.
(0.379  10 )
(5.088  10  7)
(1.581  10  6)

(2.078  10  3)
(0.209  10  3)
(0.234  10  4)
(0.874  10  5)
(0.173  10  3)
(0.373  10  5)
(0.839  10  5)
(0.371  10  5)
(0.966  10  4)
(0.193  10  4)
(0.285  10  6)
(1.200  10  5)
(0.290  10  4)
(0.619  10  5)
(0.526  10  5)
(0.655  10  5)
(3.056  10  6)
(0.908  10  4)

increased significantly 3 months after RYGB (Supplementary Table

5
Before surgery (± s.e.m.)

B). It could be suggested that reduced dietary intake may trigger

an increased ability to assimilate substrate in order to compensate
(0.022)

host nutritional and energetic requirements.

Also, in a further analysis by Spearman rank correlation, a
A

–5

18.743  10  7
7.137  10  6

11.046  10  3
1.205  10  3
1.720  10  4
6.368  10  5
2.497  10  3
1.570  10  5
5.942  10  5
3.153  10  5
4.077  10  4
1.661  10  4
0.891  10  6
5.091  10  5
0.978  10  4
3.068   10  5
2.431  10  5
2.379  10  5
8.420  10  6
7.765  10  4
0.103

possible association of inflammatory state (defined by CRP values)

0.808  10

and RYGB-induced changes at microbial species level were tested.

Interestingly, 7 of 11 species in component 1, but only 2 species of
11 in component 2, were significantly correlated to CRP (Table 4).
This indicates a close interrelationship between BMI, gut microbial
composition and inflammatory state. As the Spearman rank
correlation coefficient between BMI and CRP was r ¼ 0.860
(Po0.001), it became clear that the association between CRP
Nakamurella multipartita

and species abundances was mainly determined by postoperative

BMI changes. This was confirmed by a partial correlation analysis
B. hyodysenteriae

F. periodonticum
E. cancerogenus

with BMI as cofactor that completely abolished the independent

F. succinogenes
S. epidermidis

L. acidophilus
L. interrogans
P. mendocina
C. spiroforme
F. prausnitzii

CRP-species association.
A. johnsonii
T. pallidum
M. kansasii

S. enterica
V. parvula
T. roseum
A. caccae
C. comes

A possible association between RYGB-induced effects on

V. dispar
L. reuteri

S. boydii

microbial composition and human metabolic indices in plasma

Table 2.

has been investigated (Figure 3). In a Spearman rank correlation

analysis, significant direct correlation with total- and/or LDL-
No.

10
11
12
13
14
15
16
17
18
19
20
21
22
1
2
3
4
5
6
7
8
9

cholesterol levels on one hand was found for F. succinogenes,

The Pharmacogenomics Journal (2013), 514 – 522 & 2013 Macmillan Publishers Limited
 Metagenomic analysis in RYGB-treated patients with diabetes
J Graessler et al
519
Table 3. Relative abundances of microbial genera before and 3 months after Roux-en-Y gastric-bypass surgery

No. Before surgery (±s.e.m.) After surgery (±s.e.m.) Delta (by %) ANOVA pp
–5 –5 –5 –5
1 Enterobacter 2.382  10 (0.878  10 ) 17.496  10 (3.393  10 ) þ 634.6 0.002
2 Faecalibacterium 7.536  10–2 (1.820  10–2) 2.212  10–2 (0.602  10–2)  70.7 0.020
3 Neurospora 2.813  10–8 (0.281  10–8) 49.06  10–8 (16.55  10–8) þ 1.644.0 0.020
4 Citrobacter 4.421  10–5 (0.581  10–5) 17.543  10–5 (5.121  10–5) þ 296.8 0.029
5 Veillonella 7.190  10–4 (1.239  10–4) 23.840  10–4 (6.505  10–4) þ 231.6 0.031
6 Coprococcus 1.312  10–2 (0.266  10–2) 0.561  10–2 (0.138  10–2)  57.3 0.031
7 Helicobacter 1.808  10–4 (0.229  10–4) 1.161  10–4 (0.122  10–4)  35.8 0.032
8 Dictyostelium 1.731  10–6 (0.456  10–6) 0.534  10–6 (0.205  10–6)  69.2 0.038
9 Salmonella 6.971  10–5 (1.932  10–5) 51.81  10–5 (19.24  10–5) þ 643.3 0.043
10 Epidinium 1.639  10–5 (0.533  10–5) 0.339  10–5 (0.175  10–5)  79.3 0.043
11 Anaerostipes 1.778  10–3 (0.162  10–3) 1.238  10–3 (0.174  10–3)  30.4 0.047
12 Nakamurella 1.800  10–5 (0.439  10–5) 0.745  10–5 (0.179  10–5)  58.6 0.050
13 Shigella 1.118  10–5 (0.248  10–5) 33.84  10–5 (15.13  10–5) þ 2.925.4 0.056
14 Methanospirillum 2.089  10–5 (0.452  10–5) 1.043  10–5 (0.186  10–5)  50.1 0.058
15 Thermomicrobium 2.246  10–5 (0.318  10–5) 1.396  10–5 (0.244  10–5)  37.9 0.060
Abbreviation: ANOVA, analysis of variance.

Table 4. Factor composition after principal component analysis (PCA) with 22 microbial species significantly affected by Roux-en-Y gastric bypass
and their correlation with body mass index (BMI) and C-reactive protein (CRP)

Component Spearman correlation

1 2 BMI CRP

Factor loading r P r P

L. acidophilus 0.920 0.574 0.050 0.629 0.028

F. succinogenes 0.885 0.692 0.013 0.804 0.002
Nakamurella multipartita 0.866 0.399 0.199 0.357 0.255
C. comes 0.837 0.678 0.015 0.713 0.009
F. prausnitzii 0.787 0.727 0.007 0.741 0.006
L. reuteri 0.740 0.580 0.048 0.566 0.055
A. johnsonii 0.687 0.713 0.009 0.517 0.085
M. kansasii 0.678 0.545 0.067 0.329 0.297
E. cancerogenus  0.638  0.755 0.005  0.608 0.036
S. epidermidis 0.587 0.538 0.071 0.622 0.031
T. pallidum 0.541 0.634 0.027 0.634 0.027
L. interrogans 0.850 0.252 0.430 0.042 0.897
F. periodonticum 0.846 0.217 0.499 0.028 0.931
P. mendocina 0.839 0.196 0.542  0.007 0.983
B. hyodysenteriae 0.721 0.517 0.085 0.147 0.649
S. boydii  0.700  0.340 0.280  0.214 0.505
V. parvula  0.676  0.503 0.095  0.350 0.265
S. enterica  0.647  0.189 0.557  0.105 0.746
V. dispar  0.597  0.741 0.006  0.573 0.051
T. roseum 0.544 0.650 0.022 0.692 0.013
C. spiroforme 0.528 0.503 0.095 0.559 0.059
A. caccae 0.523 0.420 0.175 0.531 0.075
Spearman rank correlation: BMI with: component 1: r ¼ 0.699; P ¼ 0.011; component 2: r ¼ 0.203; P ¼ 0.527; CRP with: component 1: r ¼ 0.790; P ¼ 0.002;
component 2: r ¼ 0.077; P ¼ 0.812.

Acinetobacter johnsonii, T. pallidum, T. roseum, C. spiroforme and A.
caccae (Figure 3). On the other hand, E. cancerogenus, V. parvula
and dispar, and S. enterica were inversely correlated with plasma
total- and LDL-cholesterol (Figure 3). As plasma concentration of
total- and LDL-cholesterol decreased in all patients 3 months after
RYGB, the slope of each correlation was determined by increase or
decrease of postoperative species abundances. Nevertheless,
these data indicate that relative abundances of 10 from 22
identified species were associated with plasma cholesterol levels.
Further significant correlations have been found for plasma
triglycerides with 5 species (A. johnsonii, T. pallidum, A. caccae,
E. cancerogenus and Veillonella dispar), for plasma glucose with
F. prausnitzii, and HbA1c with T. roseum and V. parvula (Figure 3).
Plasma high-density lipoprotein cholesterol was not correlated
with any of the microbial species.
DISCUSSION
The dominant effects of RYGB 3 months after surgery comprised a
marked reduction of BMI combined with a significant improve-
ment of the metabolic situation in all six patients, alleviating type
2 diabetes and resulting in clearly reduced inflammatory activity.
On the basis of a metagenomic survey, this study compared the
intra-individual pre–post-operative differences of six morbidly

& 2013 Macmillan Publishers Limited The Pharmacogenomics Journal (2013), 514 – 522
 Metagenomic analysis in RYGB-treated patients with diabetes
J Graessler et al
520
Figure 2. Effect of Roux-en-Y gastric bypass on individual patterns of
components 1 and 2, which resulted from a principal component
analysis (PCA) of 22 selected species (for component composition
refer to Table 4).
Figure 3. Spearman rank correlation matrix of plasma metabolic parameters and relative species abundances (n ¼ 12).
The Pharmacogenomics Journal (2013), 514 – 522
obese individuals and correlated modifications of the metagen-
ome with parameters of the individual lipid and carbohydrate
metabolism.
Three months after RYGB, dramatic changes of the individual
composition of gut microbiota, characterized by significant pre–
post-operative differences in 22 microbial species, 11 genera
and 1 phylum were observed. The most striking results were a
substantial shift of the main gut phyla towards higher relative
abundances of Proteobacteria, specifically E. cancerogenus, and
lower relative abundances of Firmicutes and Bacteroidetes. These
results were in accordance with general findings coming from
recent experimental and clinical studies. In a RYGB model on rats,
Li et al.10 demonstrated a 52-fold postoperative increase of
Proteobacteria, Enterobacter hormaechei being the dominant
species. Simultaneously, the concentrations of Firmicutes and
Bacteroidetes decreased significantly.10 Comparing three morbidly
obese individuals 6 months after RYGB with non-operated obese
individuals, Zhang et al.11 observed a significant decrease of
Firmicutes in post-gastric-bypass individuals, who had a
proportional increase in Gammaproteobacteria. In contrast,
based on a real-time quantitative PCR technology, Furet et al.12
& 2013 Macmillan Publishers Limited

reported an increase of Bacteroidetes 3 months after RYGB.
Differences between these studies may be related to the low
number of observations, different employed technologies or
varying degrees of obesity of investigated individuals. A close
interrelationship between obesity and gut microbial composition
has been reported in many studies.1,17,18 Analysis of the gut
microbiome in twins revealed a lower proportion of Bacteroidetes
and a higher proportion of Actinobacteria in obese compared with
lean individuals but no significant difference in Firmicutes.4
However, weight loss in humans was associated with a
progressive increase in the representation of Bacteroidetes.3 In
this study, RYGB-induced reduction of body weight resulted in a
simultaneous decrease of Bacteroidetes and Firmicutes by varying
degrees. Consequently, the changes of both phyla were correlated
inversely (Spearman’s correlation coefficient:  0.678; P ¼ 0.019)
and the ratio of Bacteroidetes/Firmicutes increased, in accordance
to the data of Turnbaugh et al.4
The average proportion of Proteobacteria was significantly
increased with a concomitant decrease of Firmicutes (Spearman’s
correlation coefficient:  0.664; P ¼ 0.022). Thus the Proteobac-
teria/Firmicutes ratio increased significantly from 0.061 to 0.471
(P ¼ 0.030), which is in accordance with earlier published data.11,18
It is important to consider that the majority of species in
Proteobacteria phylum represent facultative anaerobes, while
Firmicutes are dominated by anaerobic bacteria. The increase in
dissolved oxygen after RYGB surgery might result in increased
facultative anaerobe populations and decreased anaerobic
bacteria.19 Apart from that, gastric acid secretion is reduced
after gastric bypass, leading to an increase of the human colon pH.
Based on in vitro studies, the growth of the Gram-negative
Bacteroidetes was found to be sensitive to mildly acidic pH, while
the growth of Gram-positive Firmicutes and high GC content
Actinomycetes showed greater tolerance of mildly acidic pH.20
The increased human gut pH also contributed to the different
alteration tendency of these major phyla.
Among microbial species that were affected by RYGB nine
belong to phylum Firmicutes, five to Proteobacteria, three to
Spirochaetes, two to Actinobacteria and one each to Fusobacteria,
Chloroflexi and Fibrobacteres. By principal component analysis,
which was performed in order to assemble highly correlated
species into common factors, two groups (components) have
been derived comprising 11 species each. Determining species of
group 1 were Firmicutes, including Lactobacilli acidophilus and
reuteri, F. prausnitzii, C. comes and S. epidermidis, which decreased
significantly during the postoperative period. Species composition
of group 2 was more heterogeneous conjoining most notably
Spirochaetes, Proteobacteria and Fusobacteria. It is of special
interest that selective species of the most abundant phyla
Firmicutes and Proteobacteria did not respond uniformly on
RYGB. Thus, seven species of Firmicutes decreased, while two
species, in particular, V. parvula and dispar, increased. On the other
hand, among Proteobacteria, the relative abundances of E.
cancerogenus, S. boydii and S. enterica increased, whereas those
of A. johnsonii and Pseudomonas mendocina decreased. These data
emphasize that the differential adaptation of human gut
microbiota to RYGB cannot be sufficiently explained at the
phylum level. Owing to the close interspecies cross talk it may
be sometimes difficult to assign a pathophysiological role to a
certain species. For example, in this study F. prausnitzii, as a
member of the Firmicutes phylum, decreased significantly after
RYGB. In parallel, a decrease of the CRP, as indicator of
inflammation activity, has been registered. Consequently, a direct
correlation between F. prausnitzii and CRP could be derived. This
is, at first glance, in contrast to the data of Furet et al.,12 who
established a negative association of F. prausnitzii with
inflammatory markers. Closer inspection reveals that in our
study F. prausnitzii was highly correlated (directly or indirectly)
with 10 further species in PCA-derived component 1, among them
& 2013 Macmillan Publishers Limited
Metagenomic analysis in RYGB-treated patients with diabetes
J Graessler et al
521
E. cancerogenus, which increased highly significantly after RYGB. At
the same time, the decrease of F. prausnitzii was accompanied by
the decrease of other butyrate-producing species like C. comes
and A. caccae, which might dissemble the true interaction. Among
members of PCA-derived component 1, six significantly correlated
with BMI as well as CRP. Partial correlation analysis with BMI as
cofactor, however, indicated that there was no independent
correlation of CRP with these species. In contrast, only two
members of PCA-derived component 2, Veillonella dispar and T.
roseum, correlated with BMI and CRP. It could be suggested that
the majority of species in component 2 is modulated by specific
RYGB changes in the gut with regard to pH, oxygen tension and
substrate availability rather than changes in BMI.
Bariatric surgery, in particular RYGB, has become a prominent
therapeutic option to address morbidly obese patients with T2D.
Meta-analyses reported complete resolution of diabetes in over
80% of patients who had undergone gastric-bypass surgery.21
Some studies reported normalization of glucose homeostasis in
patients with type 2 diabetes long before a noticeable weight
reduction was achieved.22 This could not be confirmed in other
studies.23 Weight loss and/or altered bowel physiology bringing
along altered secretion of gastrointestinal peptides might
contribute to this effect on glucose metabolism. In particular,
the postprandial response of the incretin glucagon-like peptide-1
has been observed to be significantly influenced by RYGB.24
Alternatively, clinical and experimental studies have clearly
demonstrated that RYGB induces weight loss by altering
physiological regulation of energy balance and metabolic
function.10,12,25–27 Calorimetric measurements in a RYGB mice
model revealed increased resting energy expenditure and
postprandial thermogenesis as main causes for higher
postoperative energy consumption.27 Global metabolic changes
in a non-obese rat model of RYGB were characterized by a
decrease in faecal bile acids and a shift from protein degradation
to putrefaction through decreased faecal tyrosine with
concomitant increases in faecal putrescine and diaminoethane.10
These data led to the suggestion that human gut microbiota and
their cross talk with host metabolism may have an important role
for the metabolic conversion in patients with T2D after RYGB. In
this study, 10 of 22 differentially modified species correlated with
plasma total- and LDL-cholesterol level, 5 with triglyceride level
and two with HbA1c level. These data provide evidence that
metabolic changes may not be restricted to RYGB-induced
malnutrition, but also a result of modified gut microbiota
composition.
Undoubtedly, RYGB is currently the most effective treatment for
morbid obese patients with type 2 diabetes. RYGB-induced
metabolic and microbial modifications can contribute to weight
loss, metabolic improvements and, finally, longer lifespan.
However, the shift to Proteobacteria, in particular to E. cancer-
ogenus, S. boydii and S. enterica, as well as the decline of butyrate-
producing Firmicutes, like F. prausnitzii, C. comes and A. caccae,
could have long-term effects on host health with the potential risk
of bowel inflammation and colorectal carcinomas. Long-term
studies of such microbiota-host metabolic cross talks may
considerably improve our comprehension of pathogenetic pro-
cesses in obesity and diabetes. Based on these findings, it will be
of high interest to see if therapeutic interventions could be
designed to create a healthy gut microbiome, modulating in a
positive way the normal intestinal flora and a healthy
metagenome–genome symbiosis. This may include well-designed
modulations in short-chain fatty acids, prebiotics, probiotics,
antibiotics and finally microbial transplantation.28
CONFLICT OF INTEREST
The authors declare no conflict of interest.
The Pharmacogenomics Journal (2013), 514 – 522
 Metagenomic analysis in RYGB-treated patients with diabetes
J Graessler et al
522
ACKNOWLEDGEMENTS
We thank Sigrid Nitzsche for the excellent technical support, Dr Dongqian Shen and
Dr Liang Xiao who helped us for the functional analysis and gave us some useful
advises. This work was supported by grants of the Deutsche Forschungsgemeinschaft
to SRB, ME-B, VL-Z, TC (KFO 252/1), of the German Federal Ministry for Education and
Research (BMBF) to the German Centre for Diabetes Research (DZD e.V.) to SRB and
of the Federal Ministry for Education and Research Germany to SRB and JL
(01DR12065).
REFERENCES
1 Flint HJ. Obesity and the gut microbiota. J Clin Gastroenterol 2011; 45(Suppl):
S128–S132.
2 Ley RE, Backhed F, Turnbaugh P, Lozupone CA, Knight RD, Gordon JI. Obesity
alters gut microbial ecology. Proc Natl Acad Sci USA 2005; 102: 11070–11075.
3 Ley RE, Turnbaugh PJ, Klein S, Gordon JI. Microbial ecology: human gut microbes
associated with obesity. Nature 2006; 444: 1022–1023.
4 Turnbaugh PJ, Hamady M, Yatsunenko T, Cantarel BL, Duncan A, Ley RE et al. A
core gut microbiome in obese and lean twins. Nature 2009; 457: 480–484.
5 Schwiertz A, Taras D, Schafer K, Beijer S, Bos NA, Donus C et al. Microbiota and
SCFA in lean and overweight healthy subjects. Obesity (Silver Spring) 2010; 18:
190–195.
6 Larsen N, Vogensen FK, van den Berg FW, Nielsen DS, Andreasen AS, Pedersen BK
et al. Gut microbiota in human adults with type 2 diabetes differs from non-
diabetic adults. PLoS One 2010; 5: e9085.
7 Tilg H. Obesity, metabolic syndrome, and microbiota: multiple interactions. J Clin
Gastroenterol 2010; 44(Suppl 1): S16–S18.
8 Mingrone G, Panunzi S, De Gaetano A, Guidone C, Iaconelli A, Leccesi L et al.
Bariatric surgery versus conventional medical therapy for type 2 diabetes. N Engl J
Med 2012; 366: 1577–1585.
9 Schauer PR, Kashyap SR, Wolski K, Brethauer SA, Kirwan JP, Pothier CE et al.
Bariatric surgery versus intensive medical therapy in obese patients with diabetes.
N Engl J Med 2012; 366: 1567–1576.
10 Li JV, Ashrafian H, Bueter M, Kinross J, Sands C, le Roux CW et al. Metabolic surgery
profoundly influences gut microbial-host metabolic cross-talk. Gut 2011; 60:
1214–1223.
11 Zhang H, DiBaise JK, Zuccolo A, Kudrna D, Braidotti M, Yu Y et al. Human gut
microbiota in obesity and after gastric bypass. Proc Natl Acad Sci USA 2009; 106:
2365–2370.
12 Furet JP, Kong LC, Tap J, Poitou C, Basdevant A, Bouillot JL et al. Differential
adaptation of human gut microbiota to bariatric surgery-induced weight loss:
links with metabolic and low-grade inflammation markers. Diabetes 2010; 59:
3049–3057.
Supplementary Information accompanies the paper on the The Pharmacogenomics Journal website (http://www.nature.com/tpj)
The Pharmacogenomics Journal (2013), 514 – 522
13 Qin J, Li R, Raes J, Arumugam M, Burgdorf KS, Manichanh C et al. A human gut
microbial gene catalogue established by metagenomic sequencing. Nature 2010;
464: 59–65.
14 Manichanh C, Rigottier-Gois L, Bonnaud E, Gloux K, Pelletier E, Frangeul L et al.
Reduced diversity of faecal microbiota in Crohn’s disease revealed by a meta-
genomic approach. Gut 2006; 55: 205–211.
15 Li R, Yu C, Li Y, Lam TW, Yiu SM, Kristiansen K et al. SOAP2: an improved ultrafast
tool for short read alignment. Bioinformatics 2009; 25: 1966–1967.
16 Guarner F, Malagelada JR. Gut flora in health and disease. Lancet 2003; 361:
512–519.
17 Angelakis E, Armougom F, Million M, Raoult D. The relationship between gut
microbiota and weight gain in humans. Future Microbiol 2012; 7: 91–109.
18 DiBaise JK, Zhang H, Crowell MD, Krajmalnik-Brown R, Decker GA, Rittmann BE.
Gut microbiota and its possible relationship with obesity. Mayo Clin Proc 2008; 83:
460–469.
19 Hartman AL, Lough DM, Barupal DK, Fiehn O, Fishbein T, Zasloff M et al. Human
gut microbiome adopts an alternative state following small bowel transplanta-
tion. Proc Natl Acad Sci USA 2009; 106: 17187–17192.
20 Duncan SH, Louis P, Thomson JM, Flint HJ. The role of pH in determining the
species composition of the human colonic microbiota. Environ Microbiol 2009; 11:
2112–2122.
21 Buchwald H, Estok R, Fahrbach K, Banel D, Jensen MD, Pories WJ et al. Weight and
type 2 diabetes after bariatric surgery: systematic review and meta-analysis. Am J
Med 2009; 122: 248–256, e245.
22 Reed MA, Pories WJ, Chapman W, Pender J, Bowden R, Barakat H et al. Roux-en-Y
gastric bypass corrects hyperinsulinemia implications for the remission of type 2
diabetes. J Clin Endocrinol Metab 2011; 96: 2525–2531.
23 Campos GM, Rabl C, Peeva S, Ciovica R, Rao M, Schwarz JM et al. Improvement in
peripheral glucose uptake after gastric bypass surgery is observed only after
substantial weight loss has occurred and correlates with the magnitude of weight
lost. J Gastrointest Surg 2010; 14: 15–23.
24 le Roux CW, Welbourn R, Werling M, Osborne A, Kokkinos A, Laurenius A et al. Gut
hormones as mediators of appetite and weight loss after Roux-en-Y gastric
bypass. Ann Surg 2007; 246: 780–785.
25 Spector D, Shikora S. Neuro-modulation and bariatric surgery for type 2 diabetes
mellitus. Int J Clin Pract Suppl 2010; 166: 53–58.
26 Stylopoulos N, Hoppin AG, Kaplan LM. Roux-en-Y gastric bypass enhances energy
expenditure and extends lifespan in diet-induced obese rats. Obesity (Silver
Spring) 2009; 17: 1839–1847.
27 Nestoridi E, Kvas S, Kucharczyk J, Stylopoulos N. Resting energy expenditure and
energetic cost of feeding are augmented after Roux-en-Y gastric bypass in obese
mice. Endocrinology 2012; 153: 2234–2244.
28 Kootte RS, Vrieze A, Holleman F, Dallinga-Thie GM, Zoetendal EG, de Vos WM et al.
The therapeutic potential of manipulating gut microbiota in obesity and type 2
diabetes mellitus. Diabetes Obes Metab 2012; 14: 112–120.
& 2013 Macmillan Publishers Limited