Buscar

MICROBIÓTA E OBESIDADE

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes
Você viu 3, do total de 14 páginas

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes
Você viu 6, do total de 14 páginas

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes
Você viu 9, do total de 14 páginas

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você também pode ser Premium ajudando estudantes

Prévia do material em texto

© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
nature medicine  advance online publication �
Obesity has become a global epidemic and is a major risk factor for 
type 2 diabetes, cardiovascular diseases and certain cancers. Evidence 
has accumulated that the gut microbiota is an important environmen-
tal factor contributing to obesity by altering host energy harvest and 
storage1,2. Gut microbiota transplantation experiments suggest a causal 
relationship between the gut microbiota and obesity development3, and 
human studies have associated gut microbial dysbiosis with obesity4,5 
and metabolic disorders such as type 2 diabetes6,7. In general, obese 
individuals show decreased bacterial diversity5 and gene richness8,9. It 
was originally reported that obesity was associated with a lower ratio 
of Bacteroidetes to Firmicutes4, and it has been reported that this ratio 
increases after weight loss induced by dietary intervention4 or Roux-en-Y 
gastric bypass10,11. Nevertheless, this view has been challenged by sub-
sequent studies that have provided conflicting findings12,13.
Deep shotgun sequencing and metagenome-wide association stud-
ies7 have enabled more in-depth characterization and insights into 
the function of gut microbiomes than 16S rDNA amplicon sequenc-
ing. Comparison of gut microbial compositions between lean and 
obese Danish individuals by shotgun sequencing revealed distinct 
differences in gene abundances and species8. Given that the gut micro-
biota varies with age14, ethnicity15 and diet16, studies on other eth-
nic group may provide additional valuable information on common 
traits characterizing obese and lean individuals, especially by link-
ing information on the gut microbiome with blood metabolite data. 
1State Key Laboratory of Medical Genomes, National Clinical Research Center for Endocrine and Metabolic Diseases, Ruijin Hospital, Shanghai Jiao Tong University 
School of Medicine, Shanghai, China. 2BGI-Shenzhen, Shenzhen, China. 3BGI Education Center, University of Chinese Academy of Sciences, Shenzhen, China. 
4Shenzhen Engineering Laboratory of Detection and Intervention of Human Intestinal Microbiome, BGI-Shenzhen, Shenzhen, China. 5China National GeneBank, 
BGI-Shenzhen, Shenzhen, China. 6Laboratory of Endocrinology and Metabolism, Institute of Health Sciences, Shanghai Institutes for Biological Sciences (SIBS), 
Chinese Academy of Sciences (CAS) & Shanghai Jiao Tong University School of Medicine (SJTUSM), Shanghai, China. 7Pancreatic Disease Center, Department of 
General Surgery, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 8CAS Key Laboratory of Separation Science for Analytical 
Chemistry, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, Dalian, China. 9Shenzhen Key Laboratory of Human Commensal Microorganisms 
and Health Research, BGI-Shenzhen, Shenzhen, China. 10James D. Watson Institute of Genome Sciences, Hangzhou, China. 11Department of Radiology and 
Radiological Science, Johns Hopkins School of Medicine, Baltimore, Maryland, USA. 12Department of Nutrition, Harvard School of Public Health, Boston, 
Massachusetts, USA. 13Department of Epidemiology, School of Public Health and Tropical Medicine, Tulane University, New Orleans, Louisiana, USA. 14National 
Institute of Nutrition and Seafood Research, Bergen, Norway. 15Laboratory of Genomics and Molecular Biomedicine, Department of Biology, University of 
Copenhagen, Copenhagen, Denmark. 16These authors contributed equally to this work. Correspondence should be addressed to W.W. (wqingw61@163.com), 
K.K. (kk@bio.ku.dk) or G.N. (gning@sibs.ac.cn).
Received 19 October 2016; accepted 16 May 2017; published online 19 June 2017; doi:10.1038/nm.4358
Gut microbiome and serum metabolome alterations 
in obesity and after weight-loss intervention
Ruixin Liu1,16, Jie Hong1,16, Xiaoqiang Xu2,3,16, Qiang Feng2,4,16, Dongya Zhang2,16, Yanyun Gu1,16, Juan Shi1, 
Shaoqian Zhao1, Wen Liu1, Xiaokai Wang2,3 , Huihua Xia2,5, Zhipeng Liu2, Bin Cui1,6, Peiwen Liang1, 
Liuqing Xi1, Jiabin Jin7, Xiayang Ying7, Xiaolin Wang8, Xinjie Zhao8, Wanyu Li1, Huijue Jia2,5,9 , Zhou Lan2, 
Fengyu Li2, Rui Wang1, Yingkai Sun1, Minglan Yang1, Yuxin Shen1, Zhuye Jie2,5, Junhua Li2,5,9, Xiaomin Chen2, 
Huanzi Zhong2,5 , Hailiang Xie2, Yifei Zhang1, Weiqiong Gu1 , Xiaxing Deng7, Baiyong Shen7, Xun Xu2,5, 
Huanming Yang2,10, Guowang Xu8 , Yufang Bi1, Shenghan Lai11, Jian Wang2,10, Lu Qi12,13, Lise Madsen2,14,15, 
Jiqiu Wang1, Guang Ning1,6, Karsten Kristiansen2,5,15 & Weiqing Wang1
Emerging evidence has linked the gut microbiome to human obesity. We performed a metagenome-wide association study and 
serum metabolomics profiling in a cohort of lean and obese, young, Chinese individuals. We identified obesity-associated gut 
microbial species linked to changes in circulating metabolites. The abundance of Bacteroides thetaiotaomicron, a glutamate-
fermenting commensal, was markedly decreased in obese individuals and was inversely correlated with serum glutamate 
concentration. Consistently, gavage with B. thetaiotaomicron reduced plasma glutamate concentration and alleviated 
diet-induced body-weight gain and adiposity in mice. Furthermore, weight-loss intervention by bariatric surgery partially 
reversed obesity-associated microbial and metabolic alterations in obese individuals, including the decreased abundance of 
B. thetaiotaomicron and the elevated serum glutamate concentration. Our findings identify previously unknown links between 
intestinal microbiota alterations, circulating amino acids and obesity, suggesting that it may be possible to intervene in 
obesity by targeting the gut microbiota.
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
�  advance online publication nature medicine
We performed deep metagenomic sequencing and metabolomics pro-
filing in a cohort of lean and obese young Han Chinese and obese Han 
Chinese undergoing weight loss treatment by sleeve gastrectomy (SG). 
We identified obesity-associated microbial species and their effects 
on host amino acid profiles, and, using a mouse model, we uncovered 
evidence of causal effects of B. thetaiotaomicron on adiposity and 
circulating amino acid levels.
RESULTS
Obesity-associated genes and taxonomic alterations identified 
by metagenome-wide association studies
We performed shotgun sequencing of 257 fecal samples (151 samples 
from 72 obese individuals and 79 lean controls in the first cohort, 
49 samples from 23 obese individuals and 26 controls in the second 
cohort, and 63 samples from 23 obese individuals, including 6 in the 
case group subjected to SG; Supplementary Fig. 1, Supplementary 
Tables 1 and 2, and Online Methods). We first compared microbial 
differences between 72 obese individuals (body mass index (BMI), 
36.78 ± 4.46 kg/m2; age, 23.6 ± 3.7 years; data are mean ± s.d.) and 
79 controls (BMI, 20.2 ± 1.3 kg/m2; age, 23.2 ± 1.8 years; data are 
mean ± s.d.). Consistent with findings from previous reports on 
European populations8,9, we found that these young, obese indi-
viduals had lower gene counts and bacterial diversity than controls 
(Fig. 1a–c), indicating that gene richness is lower in obesity regard-
less of ethnicity or age. However, we did not observe a bimodal dis-
tribution of bacterial gene number in the Han Chinese individuals 
(Supplementary Fig. 2a), contrasting the findings from Danish and 
French individuals8,9, which may be another distinct feature of the 
gut microbiomes in these different populations17. We observed higherβ diversity of obese microbiomes, indicating a more heterogeneous 
community structure among obese individuals than in lean controls 
(Fig. 1d). Using a metagenome-wide association analysis, we identi-
fied 350,524 genes associated with obesity (P < 0.01, Wilcoxon rank-
sum test, FDR = 0.047). The association analysis also showed a P 
value distribution with a marked enrichment of obese-associated gene 
markers with low P values (Supplementary Fig. 2b), suggesting that 
these genes were associated with obesity. Furthermore, these markers 
clearly separated obese and lean individuals (Supplementary Fig. 2c). 
Notably, a comparison between control-enriched and obesity-enriched 
markers revealed that the distribution of lean markers in relation to 
occurrence rate was skewed toward low occurrence (Supplementary 
Fig. 2d,e), suggesting that control-enriched genes included more low-
occurrence genes than did obesity-enriched genes.
We then grouped the 350,524 obesity-associated marker genes into 
metagenomic linkage groups (MLGs)7 and identified 217 obesity- 
associated MLGs that contained more than 100 genes (adjusted 
P < 0.05, Wilcoxon rank-sum test; Supplementary Table 3). These 
MLGs were used to construct an MLG network depicting the cor-
relation between obesity-associated gut microbial markers. Notably, 
control-enriched MLGs were more highly interconnected than 
obesity-enriched MLGs (Spearman’s correlation value <–0.6 or >0.6, 
adjusted P < 0.05; Fig. 1e). We further validated 73 MLGs annotated 
to 20 species using obesity-associated species profiles based on a 
taxonomic assignment against the updated reference metagenome17 
(P < 0.05, Wilcoxon rank sum test; Supplementary Tables 4 and 5, 
and Supplementary Fig. 3a–c). Among these, MLGs annotated to 
Akkermansia muciniphila and Fecalibacterium prausnitzii, which have 
been reported to counteract adiposity18 and inhibit inflammation19, 
respectively, were highly enriched in lean controls. Moreover, we 
identified a bacterial community consisting of MLGs corresponding 
to Bacteroides thetaiotaomicron, Bacteroides uniformis, Bacteroides 
xylanisolvens, Bacteroides ovatus and Bacteroides sp. that was signifi-
cantly enriched in lean controls (Fig. 1e and Supplementary Table 3). 
Another Bacteroides species, Bacteroides intestinalis, was also enriched 
in controls. Furthermore, MLGs annotated to Ruminococcus tor-
ques, Ruminococcus gnavus, Dorea longicatena, Dorea formicigen-
erans and Coprococcus comes, and a cluster consisting of MLGs for 
Lachnospiraceae bacterium, Fusobacterium ulcerans and Fusobacterium 
varium, were highly enriched in obese individuals (Fig. 1e). 
Among these, R. torques and R. gnavus were previously reported to 
be associated with inflammatory bowel diseases20 and metabolic dis-
orders8. Notably, MLGs annotated to D. longicatena, F. prausnitzii, 
B. intestinalis and B. thetaiotaomicrion ranked in the top ten MLGs 
that significantly differed in abundance between lean and obese sub-
jects (adjusted P < 0.05, Wilcoxon rank-sum test; Supplementary 
Table 3). We also observed a higher abundance of Firmicutes and a 
trend toward a decreased Bacteroidetes/Firmicutes ratio (P = 0.062, 
Wilcoxon rank-sum test) in obese individuals (Supplementary 
Fig. 3a). Finally, we identified abundant unclassified MLGs, which 
formed a large cluster. Together, these data reveal microbial changes 
in the gut microbiome in young, obese individuals, suggesting a state 
of microbial dysbiosis.
Associations of gut microbial species with clinical indices
To identify correlations between clinical parameters and changes in 
the gut microbiome in obesity, we performed permutational analy-
sis of variance (PERMANOVA). We found that BMI and 27 meta-
bolic parameters, including HbA1c, homeostasis model assessment 
of insulin resistance (HOMA-IR), serum concentration of lipids, 
inflammation factors and adipokines, correlated with alterations in 
the gut microbiome (adjusted P < 0.05; Supplementary Table 6). 
Random forests regression was used to regress the relative abundance 
of MLGs against each clinical index with adjusted P < 0.05 in the 
PERMANOVA. Notably, 26 MLGs exhibited the best fit with BMI in 
the training set of 151 subjects (pseudo R2 = 0.59) and remained stable 
in the test set of another 49 subjects (pseudo R2 = 0.36) (Fig. 2a and 
Supplementary Table 7). MLGs for B. thetaiotaomicron, B. intestinalis 
and B. ovatus ranked as the top ones among control-enriched MLGs 
best fitting BMI (Fig. 2a). These species were also highly enriched 
in controls as compared with obese individuals in the test set of 49 
subjects (Supplementary Fig. 4). Furthermore, these MLGs were 
associated with metabolic features, including body fat, insulin resist-
ance and inflammation status, by both random forests regression and 
Spearman’s correlation (Spearman’s correlation value <–0.3 or >0.3, 
adjusted P < 0.05; Fig. 2a and Supplementary Tables 7 and 8).
Leptin and adiponectin are two adipocyte-derived hormones21, and 
circulating leptin concentration increases following gut microbiota 
colonization1. We found that F. prausnitzii and B. thetaiotaomicron 
were negatively correlated with circulating leptin concentration and 
positively correlated with circulating adiponectin concentration. 
D. longicatena was positively correlated with circulating leptin 
concentration and negatively correlated with circulating adiponec-
tin concentration (random forest regression, and Spearman’s cor-
relation value <–0.3 or >0.3; Fig. 2b and Supplementary Tables 7 
and 9), indicating that these species may constitute potential biomark-
ers linking gut microbiota and metabolic status.
Functional characterization of the obese microbiome
We identified 5,705 Kyoto Encyclopedia of Genes and Genomes 
(KEGG) orthologs (KOs) present in at least six samples, of which 
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
nature medicine  advance online publication �
584 KOs differed in abundance between lean and obese subjects 
(P < 0.01; Supplementary Table 10). Furthermore, KEGG pathways 
belonging to the ‘phosphotransferase system’ involved in import-
ing carbohydrate were highly enriched in the microbiome of obese 
individuals as compared with lean controls, and the phosphotrans-
ferase system modules were positively correlated with species from 
the Firmicutes phylum. Conversely, genes involved in the ‘citrate 
cycle’ pathway were depleted in the obese microbiome as compared 
with lean controls (Supplementary Fig. 5a,b and Supplementary 
Tables 10–13). This was similar to the alterations in the gut micro-
biomes in obese twins of European and African ancestry5 and diet-
induced obese mice22. However, unlike these cases, pathways involved 
in carbohydrate metabolism, including ‘fructose and mannose metab-
olism’, ‘galactose metabolism’ and ‘starch and sucrose metabolism’, 
were all highly enriched in the obese microbiome as compared with 
lean controls. The ‘glycosaminoglycan degradation’ pathway harbored 
by gut commensals, such as B. thetaiotaomicron, B. intestinalis and 
other species from the Bacteroides genus possessing highly specific 
sulfatases to metabolize host glycans23,24, were depleted in the micro-
biome of obese individuals as compared with lean controls.
Previous studies have suggested a role for A. muciniphila and 
B. thetaiotaomicron in maintaining the epithelial barrier18,25; thus, 
depletion of these species may impair gut barrier integrity, increasing 
transport of lipopolysaccharide (LPS) to circulation and triggering 
induction of pro-inflammatory factors such as TNFα and IL6 (ref. 26). 
We found that the abundance of genes related to ‘LPS biosynthesis’ 
and‘peptidoglycan biosynthesis’ was higher in the microbiome of 
obese individuals than in that of lean controls, which may be related 
to the observed higher serum concentration of LBP, TNFα and IL6 
in obese individuals (Supplementary Table 1). Furthermore, amino-
acid-metabolism-related pathways involving ‘phenylalanine, tyrosine 
and tryptophan biosynthesis’, and modules of ‘glutamine/glutamate 
Control-enriched MLGs OB-enriched MLGse
0.5
0.6
0.7
0.8
0.9
1.0
β-diversity
B
ra
y 
di
st
an
ce
Control OB
4 × 105
6 × 105
8 × 105
1 × 106
Gene count
N
um
be
r 
of
 g
en
es
Control OB
1 × 106
3 × 106
5 × 106
Rarefaction of genes
Numbers of the samples
N
um
be
rs
 o
f g
en
es
1 7 13 22 31 40 49 58 67 76
Control
OB
10.0
10.5
11.5
12.5
11.0
12.0
α-diversity
Control OB
a b c d
S
ha
nn
on
 in
de
x
*** * ***
Figure 1 Gut microbial alterations in young, obese individuals. Comparison between shotgun sequencing data of stool samples from lean controls 
(n = 79) and obese individuals (n = 72). (a) Rarefaction curves based on gene count in lean controls and obese individuals. (b) Box plot of the gene 
count in control and obese subjects. (c,d) α-diversity (Shannon index; c) and β-diversity (Bray-Curtis similarity index; d) of the two groups at the gene 
level. For a–d, two-tailed Wilcoxon rank-sum test was used to determine significance. In b–d, boxes represent the interquartile ranges (IQRs) between 
the first and third quartiles, and the line inside the box represents the median; whiskers represent the lowest or highest values within 1.5 times IQR 
from the first or third quartiles. Circles represent data point beyond the whiskers. The notches show the 95% confidence interval for the medians. 
*P < 0.05, ***P < 0.001. (e) Co-occurrence network deduced from 217 MLGs (Supplementary Table 3) enriched in obese subjects and controls. 
Sizes of the nodes represent the number of genes in the MLGs (100–4,482). Blue edges, Spearman’s rank correlation coefficient > 0.6, adjusted 
P < 0.05; red edges, Spearman’s rank correlation coefficient < −0.6, adjusted P < 0.05. MLGs with >50% genes annotated to the same species or 
genus were given the corresponding annotation. Unclassified MLGs could not be annotated to any taxonomic level as a result of the low gene annotation 
rate (11% on average, see Supplementary Table 3). The numbers in parentheses next to each species name represent unique MLG identifiers.
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
�  advance online publication nature medicine
0 2 4 6 8 10 12
%IncMSE
–0.5 0 0.5
Spearman’s correlationa
* * * * * * * * * * * * * * + * + * + * * * + * *
* * * * * * * * * * * * * * * * + + + * * *
* * * * * * * * * * * * * * * * * + + * * + *
* * * * * * * * * * + * * * * * * * + * + + * * *
* * * * * * * * * * * * * * * + * * * + + * * * + +
* * * * * * * * * * * * * * * * * * * + * * * *
* * * * * * * * * * * * * * * * * * * * * + * * + + *
* * * * * * * * * * * * * * * + * * * * * + * * * + *
* * * * * * * * * * * * * * * * * * * * * * * + * * *
* * * * * * * * * * * * * * * * * * * * * * * + + * *
* * * * * * * * * * * * * * * * * * * * * + + * + * * *
* * * * * * * * * * * * * * * * * * * * * + * * * *
* * * * * * * * * * * * * * * * * * * * * + * * + * * *
* * * * * * * * * * * * * * * * * * + * * + * * + * * *
* * * * * * * * * * * * * * * * * * * * * + * * * + * *
* * * * * * * * * * * * * * * * * * * * * * * * * * * *
* * * * * * * * * * * * * * * * * * * * * * * + + * *
* * * * * * * * * * * * * * * * * * * * * * * * * * * *
* * * * * * * * * * * * * * * * * * * * * * * * * * *
* * * * * * * * * * * * * * * + * + * * * * + +
* * * * * * * * * + + * * * * + * * * * * * * + * +
+ * * * * * * + + * * + * *
* * * * * + * * * + + * * + * + + * * +
* * * * + * + + * + + * + * + +
* * * * * * + + * + + * * + * * * * + * * * * +
* * * * * * + + * + + * * + + * * * * * * * * +
# # # # # # # # #
# # # # #
# # # # # # # # #
# # #
# # # # # # # # # # # # # #
# # # # # # # # # # # # # # # #
# # # # # # # # #
# # # # # # # # # # # #
# # # # # # # # # # #
# # # #
# # # # # # # # # # #
# # # # # #
# # # # # # # # # # # # # # #
# # # # # # #
# # # # # # # # # # # # # # # # # # # # # # #
# # # # # # # # # # # # # # # # # # #
# # # # # # # # # # # # # # # # # # # #
# # # # # # # # # # # # # # # # #
# # # # # # # # # # # # # # # # # # # # # # # #
# # # # # # #
# # # # # # #
# # # # # # #
# # # # # #
# # # # # # # #
# # # # # # # # # # #
# # # # # # # # # # #
505CCon-455
occusoccCopoprocococ
eutacactus 14476)(14
ansiansiaAkkkermann
muciniphilaciniphil (74)
Eubacteriumbacter
siraeaeum 6467)6467)(66
24CCon-5424
9696CCon-399
Adiponectin
DoreaDorea
longicatenaiicatena(7570)
Faecalibacteriumcalibac
prausnitziisnitzii(13264)(1
Faecalibacteriumcalibac
prausnitziisnitzii(13197)(1
Faecalibacteriumcalibac
prausnitziisnitzii(13388)388)(1ii
OB-8283OOB-828
Bacteroidesacteroid
thetaiotaomicronaomicr (11069) HaemophilusHaeaemophe
parainfluenzaefluenza (12511)
CollinsellaCollinse
aerofaciensfaciens(8613)
Bacteroidesacteroid
ovatovaatusat 1116)(11
BacteroidesBaacteroida
sp.(11079)p.(1107
CollinsellaCollinse
aerofaciensfaciens(6710)
Ruminococcusmminoco
obeeum 7058)(70
Bacteroidesacteroid
ovatvaatusat 1080)(11
Bacteroidesacteroid
ssintesstinaliss(243)(24
es
Con-697Con-69
Con-1Con-1216Con-12 Faecalibacteriumcaliba
prausnitziizsnitziitz (24680)ii
OB-8793OB-879
AlistipesAlistipe
shaahii 206)(22
Con-2291Con-229
OB-7834OB-783
ialesiCloostridiadi
bacteerium(16594))(1
Con-3564Con-356
)DialisterDialist
invisisus 28421))
CCC
(28
b
W
ei
gh
t
W
ai
st
 c
irc
um
fe
re
nc
e
B
M
I
H
ip
 c
irc
um
fe
re
nc
e
N
ec
k 
ci
rc
um
fe
re
nc
e
W
H
R
F
as
tin
g 
in
su
lin
H
O
M
A
-I
R
Le
pt
in
G
C
T
A
LT
hs
C
R
P
LB
P
H
O
M
A
-B
In
su
lin
 2
h
U
ric
 a
ci
d
T
rig
ly
ce
rid
es
S
B
P
D
B
P
P
B
G
H
bA
1c
W
B
C
W
ho
le
 b
od
y 
fa
t
A
S
T
F
B
G
A
LP
H
D
L-
C
A
di
po
ne
ct
in
Dialister invisus (28421)
Klebsiella pneumoniae (12774)
Clostridiales bacterium (16594)
Faecalibacterium prausnitzii (13264)
Bacteroides ovatus (11116)
Bacteroides ovatus (11080)&
Haemophilus parainfluenzae (12502)
Haemophilus parainfluenzae (12511)&
Veillonella sp. oral taxon (12639)&
Con-1216&
Bacteroides intestinalis (173)&
Bacteroides intestinalis (243)&
Bacteroides intestinalis (211)&
Bacteroides intestinalis (2808)&
Faecalibacterium prausnitzii (13388)&
Bacteroides thetaiotaomicron (11069)&
OB-8283
OB-7834&
Dorea longicatena (7570)&
Coprococcus comes (6940)&
Eubacterium hallii (7132)&
Ruminococcus obeum (7058)
Dorea longicatena (9343)
OB-8241
OB-8295
OB-8293
Figure 2 Associations of gut microbial species with clinical indices. (a) Left, heat map of the Spearman’s rank correlation coefficient between 
28 clinical indices (adjusted P < 0.05 in PERMANOVA with 217 MLGs, Online Methods) and 26 BMI-associated MLGs (random forest regression, 
Online Methods). n = 151; +P < 0.05; *P < 0.01; Spearman’s rank correlation. # denotes MLGs that were selected by the random forest regression 
model when regressing the relative abundance of the corresponding MLGs against each individual clinical index. MLGs in blue and red denote 
control-enriched and obesity-enrichedMLGs in the training set, respectively (control, n = 79; obese, n = 72). & denotes MLGs that significantly 
differed in abundance in both 151 sample and 49 sample cohort (two-tailed Wilcoxon rank-sum test, P < 0.05). FBG, fasting plasma glucose; PBG, 
2-h plasma glucose; fasting insulin, fasting serum insulin; insulin 2 h, 2-h serum insulin. Right, the length of bars denotes the importance of 26 MLGs 
to the accuracy of the model of regression of BMI. Importance was determined based on the percentage increase in mean-squared error (%IncMSE) 
of BMI prediction when the relative abundance values of each MLG were randomly permuted. (b) Correlation network between two adipokines and the 
MLGs that were selected by the random forest regression model. Red and blue edges denote Spearman’s rank correlation coefficient >0.3 and <−0.3, 
respectively. Blue and red circle denoted control-enriched and obesity-enriched MLGs, respectively. The numbers in parentheses next to each 
species name represent unique MLG identifiers.
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
nature medicine  advance online publication �
transport system’ were highly enriched in obese individuals as com-
pared with lean controls, whereas microbial genes involved in the 
‘valine, leucine and isoleucine degradation’ pathway were depleted. 
Thus, the potential to produce aromatic amino acid (AAA) and 
branched chain amino acid (BCAA) in the microbiota of obese indi-
viduals may be higher than that of lean controls. This is consistent 
with higher cecal BCAA levels in mice transplanted with an obese 
human microbiota3. Together, these data demonstrate that the micro-
biota of obese individuals may have a higher capacity for carbohydrate 
utilization and a higher capacity for production of pro-inflammatory 
factors as well as AAA and BCAA.
Associations of gut microbial species with host circulating 
metabolites
An altered serum metabolome reflecting differences in the gut micro-
biomes of controls and type 2 diabetic patients was recently reported27. 
To investigate the extent to which the altered microbiome in the obese 
Han Chinese was associated with circulating metabolites in the host, 
we performed non-targeted metabolomics profiling of serum from 
cases and controls. Obese individuals showed pronounced metabolic 
alterations compared with controls (Supplementary Fig. 6a–e). We 
identified 148 metabolites that differed in abundance between cases 
and controls, which enabled clear separation of cases and controls 
(Supplementary Methods and Supplementary Figs. 6c,d and 7a). 
Discrimination between cases and controls by the 148 metabolites 
was validated in a separate set of cases and controls (Supplementary 
Fig. 6e). Among these, 13 metabolites were structurally identified, 
including glutamate, phenylalanine and tyrosine (Supplementary 
Fig. 7b and Supplementary Table 14). Targeted metabolomics pro-
filing of amino acids further corroborated the difference in abun-
dance between lean and obese individuals (Supplementary Table 
15 and Supplementary Fig. 7c). In total, the serum concentration 
of 20 of 34 amino acids differed substantially between obese indi-
viduals and lean controls. To further investigate whether the altered 
abundance of circulating metabolites correlated with the altered gut 
microbiota, we used co-inertia analysis (CIA) to find co-variation 
between the 217 MLGs and 148 metabolites that differed in abun-
dance between cases and controls (Supplementary Fig. 8a). Case-
enriched and control-enriched MLGs were clearly separated along 
the vectors for tyrosine, phenylalanine, glutamate and BCAAs in 
canonical correspondence analysis (CCA) (Supplementary Fig. 8b), 
indicating a close association of these amino acids with the altered 
microbiome in obesity. Notably, consistent with the proposed higher 
capacity for producing AAA and BCAA in the microbiota of obese 
individuals, serum concentration of phenylalanine, tyrosine, leucine, 
isoleucine and valine were considerably higher in obese individuals 
than in lean controls (Supplementary Table 15 and Supplementary 
Fig. 7c). BCAA levels (valine and leucine) were inversely correlated 
with gene count (Supplementary Table 16), similar to a high serum 
concentration of BCAA observed in low gene-count subjects28. 
Specifically, these amino acids were inversely correlated with the 
species from Bacteroides, including B. thetaiotaomicron, B. intestina-
lis, B. ovatus and B. uniformis (Spearman’s correlation, adjusted P < 
0.05; Fig. 3a and Supplementary Table 17). It has been reported that 
B. thetaiotaomicron and B. ovatus have the ability to ferment AAA 
to produce phenylacetic acid29. Taken together, our results suggest 
that the altered gut microbiota, particularly the depletion of species 
from the Bacteroides genus in obese individuals may be related to the 
higher concentration of AAA and BCAA in circulation. These amino 
acids are known risk factors for T2D30 and cardiovascular diseases 
(CVD)31, and we observed positive correlation between these amino 
acids and HOMA-IR, hyperglycemia, hyperlipidemia and circulating 
inflammatory factors (Supplementary Table 18 and Supplementary 
Fig. 9).
Among the detected amino acids, glutamate showed the most con-
siderable increase in obese individuals compared to lean controls 
(Supplementary Table 15). Glutamate exhibited positive correlations 
with 16 obesity-enriched MLGs and negative correlations with 18 con-
trol-enriched MLGs (Fig. 3a) that harbored genes encoding enzymes 
involved in glutamate metabolism (Supplementary Fig. 10 and 
Supplementary Table 19). Notably, obesity-enriched Ruminococcus 
sp., D. longicatena, C. comes and OB-8293, which possessed genes 
encoding enzymes required to produce glutamate from glutamine 
(Fig. 3b), were positively correlated with circulating glutamate levels, 
and negatively correlated with glutamine levels. These observations 
suggest that the levels of glutamate and glutamine in circulation may 
reflect the changes in the abundance of these bacterial species.
Furthermore, serum glutamate was inversely correlated with the 
abundance of B. thetaiotaomicron (Fig. 3a), which possesses genes 
encoding glutamate decarboxylase (K01580, conversion of gluta-
mate to GABA; Fig. 3b). These results suggest that the depletion of 
B. thetaiotaomicron may contribute to the higher circulating concen-
tration of glutamate in obesity. Moreover, serum concentration of 
glutamate exhibited positive correlations with obesity traits such as 
BMI, WHR, waist circumstance, HOMA-IR, hsCRP and triacylglyc-
erol (Supplementary Table 18 and Supplementary Fig. 9). Together, 
these data suggest that amino acid metabolism by gut microbial 
species may modulate the levels of circulating amino acids that are 
correlated with obesity and its metabolic complications.
Effects of B. thetaiotaomicron on adiposity and host metabolism
B. thetaiotaomicron is a dominant bacterial species residing in the 
human gut32, and its depletion is associated with obesity and serum 
amino acid concentration. On this basis, we next examined the pos-
sible causal relationship between B. thetaiotaomicron in the gut and 
host adiposity. Gavage with live, but not heat-killed, B. thetaiotaomi-
cron significantly lowered total and inguinal fat mass and increased 
lean body mass of conventionally raised mice on normal chow diet 
(Fig. 4a–c and Supplementary Fig. 11a,b). Furthermore, it allevi-
ated body weight gain and adiposity of mice fed a high fat diet (HFD) 
(Fig. 4d–f and Supplementary Fig. 11c,d). Plasma concentration 
of adiponectin was higher, whereas plasma concentration of leptin 
was lower, in HFD mice gavaged with live B. thetaiotaomicronthan 
in HFD mice gavaged with phosphate-buffered saline (PBS) or heat-
killed B. thetaiotaomicron (Fig. 4g,h and Supplementary Fig. 11e,f). 
Histological analysis of white adipose tissue revealed a substan-
tially smaller adipocyte cell size in HFD mice gavaged with live 
B. thetaiotaomicron than in HFD mice gavaged with PBS or heat-killed 
B. thetaiotaomicron (Fig. 4i,j). Moreover, mRNA levels of genes encod-
ing proteins involved in lipogenesis, including Srebf1, Scd1 and Fasn, 
were lower in white adipose tissue of HFD mice gavaged with live 
B. thetaiotaomicron, whereas expression of genes involved in lipolysis 
and fatty acid oxidation, including Adrb3, Pnpla2, Cpt2, Acox1 and 
Ppargc1a was higher (Fig. 4k). In addition, expression of markers of 
inflammation, Cd68, Ccl2, Il6 and Il1b, in adipose tissue of HFD mice 
was attenuated in response to gavage with live B. thetaiotaomicron, 
indicating an improved inflammatory status (Fig. 4l). However, we 
did not observe a considerable difference in fasting glucose or insulin 
concentration (Supplementary Fig. 11g–j) or effects on phospho-AKT 
levels in adipose tissue (Supplementary Fig. 11k,l) in the three groups. 
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
�  advance online publication nature medicine
Neither food intake nor 24-h fecal caloric content was affected by gav-
age with B. thetaiotaomicron (Supplementary Fig. 11m–p). Notably, we 
detected lower abundance of circulating amino acids, including gluta-
mate, phenylalanine, leucine and valine in HFD mice gavaged with live 
B. thetaiotaomicron than in HFD mice gavaged with PBS (Fig. 4m and 
Supplementary Table 20). Of note, decreased lipolysis and enlarged 
adipocytes have been reported in glutamate-induced obesity33,34.
16S rDNA amplicon sequencing revealed only minor changes in 
gut microbiota composition in response to B. thetaiotaomicron gavage 
(Supplementary Fig. 12a–e), and no difference was observed in 
Firmicutes abundance or Bacteroidetes/Firmicutes ratio in mice gavaged 
with live B. thetaiotaomicron compared with mice gavaged with PBS or 
heat-killed B. thetaiotaomicron (Supplementary Fig. 12b,c). Of note, 
we observed a considerably higher abundance of B. thetaiotaomicron 
in the gut from mice gavaged with live B. thetaiotaomicron than in mice 
gavaged with PBS (Supplementary Fig. 12f,g). Taken together, these 
findings indicate that B. thetaiotaomicron protects mice against diet-
induced obesity and affects circulating amino acid concentration.
* *
* *
* +
*
+
+
+ + *
*
+
+ +
+
+
+
+
+
+ +
+
+
+ +
+ + +
+
+ +
+
+ +
+ *
+
*
+ + *
+
+
*
+
+ * + * *
+ * * *
* * + * +
+ * + +
* + + +
+ + * + +
* * + * *
* * + *
+ +
+ * + * *
+ * * *
+ * + * * +
+ * + * * +
+ * * * +
+ + + + + +
+ + + +
+ + * + +
*
+
+
+
+
+ *
+ *
+ *
*
+
+
+
*
+ +
* +
+ + +
+ + + +
+ + +
+ +
+
*
+
+
+
+
+
+ +
+ +
0
GABA
Glutamate
Glutamine
K01580
K01658
K01657
K01956
K02232
K06215
K02500
K02501
V
al
in
e
Le
uc
in
e
Is
ol
eu
ci
ne
P
he
ny
la
la
ni
ne
T
yr
os
in
e
G
lu
ta
m
at
e
G
lu
ta
m
in
e
0.3–0.3 C.comes (6940,6843)
Ruminococcus sp. (8908)
D. longicatena (7570)
OB-8293
B. thetaiotaomicron (11069)
Bacteroides sp. (11079,11114)
Con-5596
Spearman’s correlation
Bacteroides thetaiotaomicron(11021)
Con-19475
Con-5268
Con-5293
Alistipes shahii(2206)
Con-2291
Bacteroides intestinalis(247)
Con-5231
Con-5220
Con-17087
Con-1400
Odoribacter splanchnicus(18044)
Akkermansia muciniphila(74)
Con-2488
Con-2500
Con-3217
Bacteroides xylanisolvens(11081)
Con-14102
Alistipes putredinis(2213)
Con-1753
Alistipes sp.(6009)
Con-16099
Con-12885
Con-14231
Bacteroides ovatus(11116)
Bacteroides sp.(11114)
Faecalibacterium prausnitzii(13388)
Dialister invisus(28421)
Faecalibacterium prausnitzii(13264)
Klebsiella pneumoniae(12774)
Con-5596
Faecalibacterium cf. prausnitzii(24680)
Con-10882
Bacteroides intestinalis(144)
Con-16701
Bacteroides uniformis(11898)
Anaerotruncus colihominis(14893)
Clostridiales bacterium(16594)
Bacteroides intestinalis(2744)
Faecalibacterium cf. prausnitzii(3979)
Holdemania filiformis(22816)
Bacteroides intestinalis(243)
Bacteroides intestinalis(173)
Bacteroides sp.(28394)
Bacteroides intestinalis(211)
Bacteroides intestinalis(2808)
Bacteroides sp.(11079)
Bacteroides ovatus(11080)
Bacteroides thetaiotaomicron(11069)
Haemophilus parainfluenzae(12511)
Haemophilus parainfluenzae (12502)
Veillonella sp. Oral(12639)
Bacteroides sp.(28580)
OB-8231
Coprococcus comes(6940)
Coprococcus comes(6834)
Ruminococcus sp.(8908)
OB-8852
Ruminococcus sp.(7121)
Eubacterium hallii(10029)
Dorea longicatena(9343)
Dorea longicatena(9349)
Dorea longicatena(7570)
OB-8283
OB-8295
OB-8293
Lachnospiraceae bacterium(7651)
Dorea longicatena(7593)
Ruminococcus torques(8038)
Ruminococcus torques(7997)
OB-8728
Fusobacterium ulcerans(7723)
Ruminococcus torques(7927)
Ruminococcus torques(7947)
Fusobacterium ulcerans(7655)
Lachnospiraceae bacterium(7653)
Fusobacterium varium(7734)
−12
−11
−10
−9
−8
A
bu
nd
an
ce
(lg
)
Control
OB
K0
15
80
K0
25
00
K0
19
56
K0
25
01
K0
16
57
K0
16
58
K0
22
32
K0
62
15
a b
Figure 3 Associations of gut microbial species with circulating amino acids. (a) Spearman’s rank correlation between 217 MLGs and seven amino acids 
(only MLGs correlated with at least one amino acid with adjusted P < 0.05 are shown). MLGs in blue and red denote control-enriched and obesity-
enriched MLGs, respectively. 139 samples were used for Spearman’s rank correlation, +P < 0.05; *P < 0.01. (b) Upper panel, control-enriched (blue) 
and obesity-enriched (red) MLGs associated with serum levels of glutamate and possessing genes involved in glutamate metabolism. Lower panel, 
differential enrichment of indicated KOs in controls and obese individuals (control, n = 79; obese, n = 72; two-tailed Wilcoxon rank-sum test, P < 
0.05). K01580 was assigned to glutamate decarboxylase; K01657, anthranilate synthase component I; K01658, anthranilate synthase component II; 
K01956, carbamoyl-phosphate synthase small subunit; K02232, adenosylcobyric acid synthase; K06215, pyridoxine biosynthesis protein; K02500, 
cyclase; K02501, glutamine amidotransferase. GABA, γ-aminobutyric acid. The involved enzymic reactions are shown in Supplementary Table 19. 
Boxes represent the IQRs between the first and third quartiles, and the line inside the box represents the median; whiskers represent the lowest or 
highest values within 1.5 times IQR from the first or third quartiles. Circles represent data point beyond the whiskers. The numbers in parentheses next 
to each species name represent unique MLG identifiers.
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
nature medicine  advance online publication �
Effects of sleeve gastrectomy on the gut microbiome and 
metabolic parameters
To further investigate the relationships between gut microbiota, 
circulating amino acids and clinical phenotypes, we next analyzed 
these parameters in 23 obese individuals at baseline (0M), one month 
(1M) and 3 months (3M) after SG. Weight loss was accompanied by 
metabolic improvement (Supplementary Table 2). We used a random 
forest classifier based on the relative abundance of MLGs from the 
case-control cohort to distinguish obese individualsfrom lean indi-
viduals. Using a training set of 151 individuals, we obtained an area 
under the receiver operating characteristic (ROC) curve (AUC) of 
0.9687, and 0.9214 using a test set of 49 individuals (Supplementary 
Fig. 13a–d and Supplementary Table 3). Analyses using the same 
classifier correlated with the obese state revealed that the gut micro-
biomes of obese individuals after SG approached that of lean indi-
viduals (Fig. 5a and Supplementary Table 21). Concomitantly, 
principal coordinate analysis also revealed that samples at 0M, 1M 
and 3M shifted from the obese microbiota toward that of the controls 
(Supplementary Fig. 14a). Furthermore, gene count and α diversity 
of the gut microbiome substantially increased at 3M after SG. A tran-
sient decrease in gene count at 1M might reflect the changed diet in 
the first month after SG35 (Supplementary Fig. 14b,c). These data 
indicate that the microbiome in obese individuals become more simi-
lar to the microbiome in lean controls following SG and weight loss.
Functionally, 17 of 34 obesity-associated KEGG pathways became 
less prominent at 3M after surgery. Furthermore, the abundance of 
24 KEGG pathways shown to be associated with obesity did not differ 
significantly between lean controls and 3M after SG. These results 
suggest that a considerable number of microbial enzymatic func-
tions become more similar to lean controls after SG. These microbial 
functions included pathways involved in carbohydrate fermentation, 
citrate cycle, glycosaminoglycan degradation and LPS synthesis, as 
PBS KBT LBT
0
10
20
30
40
B
od
y 
w
ei
gh
t g
ai
n 
(%
)
CD
iWAT eWAT
0.0
0.2
0.4
0.6
T
is
su
e 
w
ei
gh
t (
g)
PBS
KBT
LBT
CD
PBS
KBT
LBT
PBS KBT LBT
0
20
40
60
B
od
y 
w
ei
gh
t g
ai
n 
(%
)
HFD
iWAT eWAT
0.0
0.5
1.0
1.5
2.0
T
is
su
e 
w
ei
gh
t (
g)
*
*
**
*
HFD
Srebf1 Scd1 Fasn Adrb3 Lipe Pnpla2 Cpt2 Acox1Ppargc1�
0
1
2
3
4
R
el
at
iv
e 
m
R
N
A
 le
ve
ls
*
* *
HFD
*
*
*
* 0.056
*
0.07
*
*
0.068
Lipogenesis Lipolysis Oxidation
**
Cd68 Ccl2 Il6 Il1b
0
1
2
3
R
el
at
iv
e 
m
R
N
A
 le
ve
ls
PBS
KBT
LBT
0.057
**
*
**
*
HFD HFD
0 20 40 60 80 100
0
5
10
15
20 PBSKBT
LBT
F
re
qu
en
cy
Diameter (µm)
10
20
30
40
50
60
70
D
ia
m
et
er
 (
µm
)
PBS KBT LBT
***
***
PBS KBT LBT
0
10
20
30
HFD
0.07
*
PBS KBT LBT
0
4,000
8,000
12,000
P
la
sm
a 
le
pt
in
 (
pg
/m
l)
*
HFD
*
iW
A
T
eW
A
T
Fat mass Lean mass
0
20
40
60
80
100
*
0.09
0.08
HFD
Fat mass Lean mass
0
5
10
15
20
20
40
60
80
100 PBSKBT
LBT
PBS
KBT
LBT
PBS
KBT
LBT
0.05
CD
**
PBS LBT
0
40
80
120
R
el
at
iv
e 
gl
ut
am
at
e 
ab
un
da
nc
e
*
PBS LBT
0
100
200
300
400
R
el
at
iv
e 
ph
en
yl
al
an
in
e 
ab
un
da
nc
e
*
PBS LBT
0
500
1,000
1,500 *
PBS LBT
0
1,000
2,000
3,000
*
B
od
y 
co
m
po
si
tio
n
(%
 o
f b
od
y 
w
ei
gh
t)
*
*
*
*
*
*
B
od
y 
co
m
po
si
tio
n
(%
 o
f b
od
y 
w
ei
gh
t)
*
P
la
sm
a 
ad
ip
on
ec
tin
 (
µg
/m
l)
PBS KBT LBT
* *
R
el
at
iv
e 
le
uc
in
e 
ab
un
da
nc
e
R
el
at
iv
e 
va
lin
e 
ab
un
da
nc
e
b c d e
f g h i j
k l m
a
Figure 4 Effects of B. thetaiotaomicron supplementation on adiposity and host metabolism. Live B. thetaiotaomicron (LBT) was administrated 
by oral gavage to C57BL/6 mice three times per week at a dose of 5 × 108 cfu/0.1 ml PBS for 7 weeks. Gavage of sterile PBS or the same dose of 
heat-killed B. thetaiotaomicron (KBT) were used as controls. (a–c) Relative body weight change (a), body composition (b) and adipose tissue weight 
(c) are shown in the three indicated groups fed a normal chow diet, n = 12 for each group. (d) The relative body weight change in the three indicated 
groups fed an HFD, n = 8 for each group. (e,f) Body composition (e) and adipose tissue mass (f) in the three indicated groups fed an HFD. (g,h) plasma 
adiponectin (g) and leptin (h) in the three indicated groups fed an HFD. (i) Representative images of hematoxylin- and eosin-stained sections of inguinal 
subcutaneous white adipose tissue (upper panel) and epididymal white adipose tissue (lower panel) in the three groups fed an HFD. n = 9 images per 
group. Scale bar, 200 µm. (j) Left, the frequency distribution of adipocyte cell size in inguinal subcutaneous white adipose tissue of mice fed an HFD. 
Right, box plot of adipocyte diameter in three groups, ***P < 0.001, Wilcoxon rank-sum test, boxes represent the IQRs between the first and third 
quartiles, and the line inside the box represents the median; whiskers represent the lowest or highest values within 1.5 times IQR from the first or 
third quartiles. (k) mRNA levels of the indicated genes in white adipose tissue in three groups fed an HFD. (l) mRNA levels of Cd68, Ccl2, Il6 and Il1b 
in three indicated groups. For e–h and k,l, n = 8, 8, 7 for each group, respectively. (m) Comparison of circulating levels of glutamate, phenylalanine, 
leucine and valine by gas chromatography (GC)/MS in the indicated groups fed an HFD (n = 7 for each group). KBT, heated killed B. thetaiotaomicron; 
iWAT, inguinal subcutaneous white adipose tissue; eWAT, epididymal white adipose tissue; CD, normal chow diet. Significance between every two 
groups was calculated using unpaired two-tailed Student’s t test unless otherwise indicated. Data are shown as mean ± s.d. *P < 0.05, **P < 0.01. 
Error bars represent s.d.
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
�  advance online publication nature medicine
0.0
0.2
0.4
0.6
0.8
P
ro
ba
bi
lit
y 
of
 th
e 
ob
es
e
Control OB 0M 1M 3M
−4 −2 0 2 4
3M
0M
−22
−20
−18
−16
−14
−12
Control 0M 3M
−30
−25
−20
−15
OB 1M Control 0M 3MOB 1M
RandomForest in the test set and
the surgery samples
A
bu
nd
an
ce
(lg
)
A
bu
nd
an
ce
(lg
)
Abundance(lg)
Bacteroides thetaiotaomicron (11021) Bacteroides thetaiotaomicron (11069)
Ala-leu
3-m-his
Cystine
AABA
Arginine
#Glycine
Hydroxyproline
Glutamine
GABA
Homoser
#AADA
Histamine
#β-alanine
 #1-m-his
Ethanolamine
Cysteine
Citrulline
#Methionine
Aspartate
Asparagine
Histidine
#Tryptophan
Threonine
Leucine
 #Phenylalanine
Proline
Ornithine
#Glutamate
Isoleucine
#Alanine
Valine
#Lysine
#Tyrosine
#Serine
−30 −28 −26 −24 −22 −20 −18 −16
2.5
3.0
3.5
GEE, P = 0.0255
−30 −28 −26 −24 −22 −20 −18 −16
2.5
3.0
3.5
0M 3M
0M 3M
−16 −15 −14 −13 −12 −11
2.5
3.0
3.5
GEE, P = 0.0221
G
lu
ta
m
at
e(
lg
)
−16 −15 −14 −13 −12 −11
0M
3M
0M
3M
0M
3M
0M
3M
2.5
3.0
3.5
0M 3M
0M 3M
Bacteroides thetaiotaomicron (11021)
Bacteroides thetaiotaomicron (11021)
Bacteroides thetaiotaomicron (11069)
Bacteroides thetaiotaomicron (11069)
a b
c
e
G
lu
ta
m
at
e(
lg
)
d
−17 −16 −15 −14 −13 −12 −11
3.2
3.4
3.6
3.8
4.0
GEE, P = 0.0005
B
M
I(
lg
)
−17 −16 −15 −14 −13 −12 −11
3.2
3.4
3.6
3.8
4.0
−30 −25 −20 −15
3.23.4
3.6
3.8
4.0
GEE, P = 0.0011
B
M
I(
lg
)
−30 −25 −20 −15
3.2
3.4
3.6
3.8
4.0
*
** ** *** ***ns
Figure 5 Microbial and metabolic alterations during weight-loss intervention by sleeve gastrectomy. (a) Classification of fecal microbiota in a test 
set (control, n = 26; obese, n = 23) and individuals with obesity before (0M, n = 23), 1 month (1M, n = 17) and 3 months (3M, n = 23) after sleeve 
gastrectomy by a random forest classifier previously trained on 151 samples of individuals with obesity and controls. Lines connected the same subject 
at different time points after surgery. (b) The relative abundance of MLGs for B. thetaiotaomicron in indicated groups. Two-tailed Wilcoxon rank-sum 
test was used to determine significance in controls and obese individuals, and two-tailed Wilcoxon matched-pairs signed rank test was used in the 
treatment samples. *P < 0.05, **P < 0.01, ***P < 0.001. (c) Scatter plot of BMI against the relative abundance of MLGs for B. thetaiotaomicron in 
each subject. Left and right scatter plots reflect association between changes in two MLGs for B. thetaiotaomicron and changes in BMI between 0M 
and 3M calculated by the GEE model (Online Methods). Box plots on the left of each scatter plot denote BMI in the corresponding subjects; box plots 
below each scatter plot denote the abundance of the individual B. thetaiotaomicron MLGs in the corresponding subjects (0M, n = 23; 3M, n = 23). 
(d) The abundance of serum amino acids in individuals with obesity before and 3 months after sleeve gastrectomy (0M, n = 15; 3M, n = 15). Blue 
and red denote control-enriched and obesity-enriched amino acids, respectively; # denotes amino acids differed in abundance between 0M and 3M 
samples. Two-tailed Wilcoxon matched-pairs signed rank test, P < 0.05. Ala-leu, alanyl-leucine; 3-m-his, 3-methyl-histidine; GABA, γ-aminobutyric 
acid; AADA, 2-aminoadipic acid; 1-m-his, 1-methyl-histidine; AABA, α-aminoisobutyric acid. (e) Scatter plot of plasma glutamate levels against the 
relative abundance of MLGs for B. thetaiotaomicron in each subject. Left and right scatter plots reflect association between changes in two MLGs for 
B. thetaiotaomicron and changes in plasma glutamate between 0M and 3M calculated by the GEE model (Online Methods). Box plots on the left of 
each scatter plot denote plasma glutamate abundance in the corresponding subjects; box plots below each scatter plot denote the abundance of the 
individual B. thetaiotaomicron MLGs in the corresponding subjects (0M, n = 15; 3M, n = 15). For c and e, purple and green color denote samples at 0M 
and 3M, respectively. Lines connected the same subject at 0M and 3M. In all box plots, boxes represent the IQRs between the first and third quartiles, 
and the line inside the box represents the median; whiskers represent the lowest or highest values within 1.5 times IQR from the first or third quartiles. 
Circles represent data point beyond the whiskers. The notches show the 95% confidence interval for the medians. The numbers in parentheses next to 
each species name represent unique MLG identifiers.
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
nature medicine  advance online publication �
well as production of AAA and BCAA (Supplementary Fig. 14d,e). 
We further found that the abundance of 14 case-enriched MLGs was 
decreased at 1M, and the abundance of 20 case-enriched MLGs was 
substantially decreased by 3M, whereas the abundance of 8 and 37 
control-enriched MLGs was substantially increased at 1M and 3M, 
respectively (Supplementary Fig. 14f,g). These included obesity-
enriched MLGs corresponding to C. comes and D. longicatena, and 
control-enriched MLGs for Clostridiales bacterium, Anaerotruncus 
colihominis, A. muciniphila and B. thetaiotaomicron. Of note, MLGs 
annotated to B. thetaiotaomicron exhibited a significant increase 
at 3M to levels comparable to those of lean controls (Fig. 5b and 
Supplementary Table 22). Moreover, the increase in abundance of 
B. thetaiotaomicron following SG was associated with the decrease 
in BMI, as revealed by the generalized estimating equation (GEE) 
analysis (Fig. 5c). In our study, the levels of BCAA did not substan-
tially decrease at 3M after SG. However, consistent with findings from 
Roux-en-Y Gastric Bypass Surgery36–38, circulating concentration of 
AAA, methionine, alanine and lysine were decreased, whereas serum 
concentration of serine and glycine were increased at 3M after SG. In 
addition, we detected considerably decreased circulating glutamate 
levels (Fig. 5d and Supplementary Table 23). Noticeably, the increase 
in abundance of B. thetaiotaomicron after SG was associated with the 
decrease in circulating glutamate levels (P < 0.05, GEE model; Fig. 5e). 
Furthermore, the abundance of microbial genes encoding enzymes 
involved in glutamate metabolism differed between the 0M and 3M 
samples (Supplementary Fig. 14h and Supplementary Table 24). 
Finally, the reduced circulating glutamate concentration were asso-
ciated with the improvement of hyperglycemia, insulin resistance, 
and serum concentration of leptin and inflammation markers such 
as hsCRP after SG (P < 0.05, GEE model; Supplementary Table 25). 
These results further support the idea that the altered abundance of 
bacterial species such as B. thetaiotaomicron may influence circulating 
levels of amino acids that are related to complications of obesity. In 
conclusion, weight-loss intervention by SG partly restored a healthy 
microbiome and metabolic profile.
DISCUSSION
Evidence that the gut microbiota contributes to the development of 
obesity is accumulating. Thus, characterization of the gut microbiota 
in obesity and identification of microbial therapeutic targets are highly 
warranted. We found that a cluster of abundant Bacteroides species in 
the human gut was depleted in the obese microbiome. Among them, 
B. thetaiotaomicron exhibited a considerable increase in obese indi-
viduals following weight-loss intervention by SG, which was associ-
ated with body weight reduction after SG. Furthermore, gavage with 
B. thetaiotaomicron protected mice against adiposity. Our results are 
consistent with those of a previous study that found that cohous-
ing lean and obese mice prevents development of increased adipos-
ity in obese mice, which is correlated with transfer of members of 
Bacteroidales, including B. thetaiotaomicron3. Of note, the anti-obesity 
effect of B. thetaiotaomicron may require a specific intestinal micro-
environment, as B. thetaiotaomicron is reported to increase body fat 
content in germ-free mice1. In humans, the abundance of species from 
the Bacteroides genus, such as B. uniformis, correlated with B. thetaio-
taomicron (Fig. 1e). B. uniformis was previously reported to alleviate 
HFD-induced obesity39, and might synergize with or add to the effect 
of B. thetaiotaomicron. In HFD mice, gavage with live B. thetaiotaomi-
cron resulted in an increased abundance of the Akkermansia genus 
(Supplementary Fig. 12e). Of note, A. muciniphila was reported to 
counteract adiposity in HFD mice18. Together, these results suggest 
that the effects of B. thetaiotaomicron may depend on the interaction 
with certain additional microbial species. Further studies are clearly 
warranted to delineate how B. thetaiotaomicron, in concert with other 
microbial species, modulates metabolism.
Previous studies have indicated that B. thetaiotaomicron coloniza-
tion increases the levels of mRNAs encoding glutamate transporter 
and glutamate decarboxylase in epithelial cells25, which may influ-
ence host glutamate levels. Consistently, we found that the abundance 
of B. thetaiotaomicron was negatively correlated with glutamatein 
circulation in case-control samples. Moreover, gavage with B. thetai-
otaomicron reduced plasma glutamate in mice. A previous study 
also reported the association of the glutamate derivate, N-acetyl-
l-glutamic acid, with the relative abundance of Eurotiomycetes in 
obesity40, supporting the relationship between glutamate metabo-
lism and gut microbiome in obesity. It has been documented that 
glutamate, used as a common food additive, is able to induce obesity 
when administered to rodents41,42. Excessive glutamate consump-
tion is shown to be positively correlated with overweight in Chinese 
adults43. Furthermore, consistent with the observed higher lipolysis 
in mice gavaged with B. thetaiotaomicron, lower lipolysis and enlarged 
adipocytes have been reported in monosodium-glutamate-induced 
obesity35,36. Given the possible association between glutamate and 
insulin resistance, this may add another dimension to our under-
standing of the beneficial effects of B. thetaiotaomicron on human 
health44–47. However, whether B. thetaiotaomicron affects adiposity 
via glutamate will require further study. In this context, investiga-
tions of the importance of other obesity-associated species such as 
D. longicatena are clearly warranted.
In conclusion, our findings extend our insights into the relationship 
between the gut microbiota at the species level and host metabolism 
and adiposity, pointing to possible future modalities for obesity inter-
vention targeting the gut microbiota.
METhODS
Methods, including statements of data availability and any associated 
accession codes and references, are available in the online version of 
the paper.
Note: Any Supplementary Information and Source Data files are available in the 
online version of the paper.
ACKNoWLeDGMeNtS
The authors thank the field workers for their contribution and the participants 
for their cooperation. We thank H.B. Nielsen, S.B. Pedersen and H. Zhao 
for their contribution to data discussions. We thank P. Yin, S. Wang and 
H. Yu for their assistance in cultivating bacteria. This study was supported 
by grants from National Natural Science Foundation of China (no. 81621061 
(G.N.), 81522011 (J.W.), 81370963 (R.L.), 81570758 (R.L.), 81370949 
(J.W.), 81570757 (J.W.), 81471060 (J.H.) and 81670761 (Y.G.)), National 
International Science Cooperation Foundation (no. 2015DFA30560, W.W.), 973 
Foundation (no. 2015CB553600, G.N.) and Shenzhen Municipal Government 
of China (no. JSGG20140702161403250 (Q.F.), DRC-SZ[2015]162 (Q.F.), 
JSGG20160229172752028 (J.L.) and JCYJ20160229172757249 (H.J.)).
AUtHoR CoNtRIBUtIoNS
W.W., K.K. and G.N. conceived and designed the project. W.W., G.N. J.H., R.L., Y.G. 
and Jiqiu Wang managed the study. J.H., J.S., Y.G., Y.Z., W.G., B.S., X.D., J.J. and Y.B. 
made clinical diagnosis, recruited subjects and performed intervention. J.S., P.L., 
L.X., X.Y., Wanyu Li, R.W., Y. Shen, M.Y. and Y. Sun collected samples and clinical 
phenotypes. Xiaoqiang Xu, Q.F., D.Z., Xiaokai Wang., H. Xia, Z. Lan, Z.J., J.L., H.Z., 
and H. Xie performed bioinformatics analyses. Z. Liu, F.L. and X.C. performed 
metabolomics profiling and data analysis. Xiaolin Wang., X.Z. and G.X. performed 
targeted amino acid profiling. S.Z. and Wen Liu conducted animal experiments. 
R.L., K.K., W.W., G.N. and Jiqiu Wang wrote the manuscript. L.M., L.Q., S.L., B.C., 
H.J., Xun Xu, H.Y. and Jian Wang contributed to text revision and discussion.
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
a r t i c l e s
�0  advance online publication nature medicine
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Reprints and permissions information is available online at http://www.nature.com/
reprints/index.html. Publisher’s note: Springer Nature remains neutral with regard to 
jurisdictional claims in published maps and institutional affiliations.
1. Bäckhed, F. et al. The gut microbiota as an environmental factor that regulates fat 
storage. Proc. Natl. Acad. Sci. USA 101, 15718–15723 (2004).
2. Cox, A.J., West, N.P. & Cripps, A.W. Obesity, inflammation, and the gut microbiota. 
Lancet Diabetes Endocrinol. 3, 207–215 (2015).
3. Ridaura, V.K. et al. Gut microbiota from twins discordant for obesity modulate 
metabolism in mice. Science 341, 1241214 (2013).
4. Ley, R.E., Turnbaugh, P.J., Klein, S. & Gordon, J.I. Microbial ecology: human gut 
microbes associated with obesity. Nature 444, 1022–1023 (2006).
5. Turnbaugh, P.J. et al. A core gut microbiome in obese and lean twins. Nature 457, 
480–484 (2009).
6. Karlsson, F.H. et al. Gut metagenome in European women with normal, impaired 
and diabetic glucose control. Nature 498, 99–103 (2013).
7. Qin, J. et al. A metagenome-wide association study of gut microbiota in type 2 
diabetes. Nature 490, 55–60 (2012).
8. Le Chatelier, E. et al. Richness of human gut microbiome correlates with metabolic 
markers. Nature 500, 541–546 (2013).
9. Cotillard, A. et al. Dietary intervention impact on gut microbial gene richness. Nature 
500, 585–588 (2013).
10. Furet, J.P. et al. Differential adaptation of human gut microbiota to bariatric 
surgery–induced weight loss: links with metabolic and low-grade inflammation 
markers. Diabetes 59, 3049–3057 (2010).
11. Kong, L.C. et al. Gut microbiota after gastric bypass in human obesity: increased 
richness and associations of bacterial genera with adipose tissue genes. 
Am. J. Clin. Nutr. 98, 16–24 (2013).
12. Duncan, S.H. et al. Human colonic microbiota associated with diet, obesity and 
weight loss. Int. J. Obes. (Lond.) 32, 1720–1724 (2008).
13. Schwiertz, A. et al. Microbiota and SCFA in lean and overweight healthy subjects. 
Obesity (Silver Spring) 18, 190–195 (2010).
14. Claesson, M.J. et al. Composition, variability and temporal stability of the intestinal 
microbiota of the elderly. Proc. Natl. Acad. Sci. USA 108 (Suppl. 1), 4586–4591 
(2011).
15. Mueller, S. et al. Differences in fecal microbiota in different European study 
populations in relation to age, gender and country: a cross-sectional study. 
Appl. Environ. Microbiol. 72, 1027–1033 (2006).
16. David, L.A. et al. Diet rapidly and reproducibly alters the human gut microbiome. 
Nature 505, 559–563 (2014).
17. Li, J. et al. An integrated catalog of reference genes in the human gut microbiome. 
Nat. Biotechnol. 32, 834–841 (2014).
18. Everard, A. et al. Cross-talk between Akkermansia muciniphila and intestinal 
epithelium controls diet-induced obesity. Proc. Natl. Acad. Sci. USA 110, 
9066–9071 (2013).
19. Miquel, S. et al. Faecalibacterium prausnitzii and human intestinal health. 
Curr. Opin. Microbiol. 16, 255–261 (2013).
20. Peterson, C.T., Sharma, V., Elmén, L. & Peterson, S.N. Immune homeostasis, 
dysbiosis and therapeutic modulation of the gut microbiota. Clin. Exp. Immunol. 
179, 363–377 (2015).
21. Yamauchi, T. et al. The fat-derived hormone adiponectin reverses insulin 
resistance associated with both lipoatrophy and obesity. Nat. Med. 7, 941–946 
(2001).
22. Turnbaugh, P.J., Bäckhed, F., Fulton, L. & Gordon, J.I. Diet-induced obesity is 
linked to marked but reversible alterations in the mouse distal gut microbiome. Cell 
Host Microbe 3, 213–223 (2008).
23. Ulmer, J.E. et al. Characterization of glycosaminoglycan (GAG) sulfatases from the 
human gut symbiont Bacteroides thetaiotaomicron reveals the first GAG-specific 
bacterial endosulfatase. J. Biol. Chem. 289, 24289–24303 (2014).
24. Xu, J. et al. A genomic view of the human–Bacteroides thetaiotaomicron symbiosis. 
Science 299, 2074–2076 (2003).
25. Hooper, L.V. et al. Molecular analysis of commensal host–microbial relationships in 
the intestine. Science 291, 881–884 (2001).26. Qian, C. & Cao, X. Regulation of Toll-like receptor signaling pathways in innate 
immune responses. Ann. NY Acad. Sci. 1283, 67–74 (2013).
27. Pedersen, H.K. et al. Human gut microbes impact host serum metabolome and 
insulin sensitivity. Nature 535, 376–381 (2016).
28. Shoaie, S. et al. Quantifying diet-induced metabolic changes of the human gut 
microbiome. Cell Metab. 22, 320–331 (2015).
29. Russell, W.R. et al. Major phenylpropanoid-derived metabolites in the human gut 
can arise from microbial fermentation of protein. Mol. Nutr. Food Res. 57, 523–535 
(2013).
30. Wang, T.J. et al. Metabolite profiles and the risk of developing diabetes. Nat. Med. 
17, 448–453 (2011).
31. Magnusson, M. et al. A diabetes-predictive amino acid score and future cardiovascular 
disease. Eur. Heart J. 34, 1982–1989 (2013).
32. Eckburg, P.B. et al. Diversity of the human intestinal microbial flora. Science 308, 
1635–1638 (2005).
33. Bunyan, J., Murrell, E.A. & Shah, P.P. The induction of obesity in rodents by means 
of monosodium glutamate. Br. J. Nutr. 35, 25–39 (1976).
34. Dolnikoff, M., Martin-Hidalgo, A., Machado, U.F., Lima, F.B. & Herrera, E. Decreased 
lipolysis and enhanced glycerol and glucose utilization by adipose tissue prior to 
development of obesity in monosodium glutamate (MSG) treated-rats. Int. J. Obes. 
Relat. Metab. Disord. 25, 426–433 (2001).
35. Rubino, F. et al. Metabolic surgery in the treatment algorithm for type 2 diabetes: 
a joint statement by international diabetes organizations. Diabetes Care 39, 
861–877 (2016).
36. Gralka, E. et al. Metabolomic fingerprint of severe obesity is dynamically affected 
by bariatric surgery in a procedure-dependent manner. Am. J. Clin. Nutr. 102, 
1313–1322 (2015).
37. Lips, M.A. et al. Roux-en-Y gastric bypass surgery, but not calorie restriction, reduces 
plasma branched-chain amino acids in obese women independent of weight loss 
or the presence of type 2 diabetes. Diabetes Care 37, 3150–3156 (2014).
38. Mutch, D.M. et al. Metabolite profiling identifies candidate markers reflecting the 
clinical adaptations associated with Roux-en-Y gastric bypass surgery. PLoS One 4, 
e7905 (2009).
39. Gauffin Cano, P., Santacruz, A., Moya, Á. & Sanz, Y. Bacteroides uniformis CECT 
7771 ameliorates metabolic and immunological dysfunction in mice with high-fat-
diet induced obesity. PLoS One 7, e41079 (2012).
40. Mar Rodríguez, M. et al. Obesity changes the human gut mycobiome. Sci. Rep. 5, 
14600 (2015).
41. Hermanussen, M. et al. Obesity, voracity and short stature: the impact of glutamate 
on the regulation of appetite. Eur. J. Clin. Nutr. 60, 25–31 (2006).
42. Olney, J.W. Brain lesions, obesity and other disturbances in mice treated with 
monosodium glutamate. Science 164, 719–721 (1969).
43. He, K. et al. Consumption of monosodium glutamate in relation to incidence of 
overweight in Chinese adults: China Health and Nutrition Survey (CHNS). Am. J. 
Clin. Nutr. 93, 1328–1336 (2011).
44. Comstock, L.E. & Coyne, M.J. Bacteroides thetaiotaomicron: a dynamic, niche-
adapted human symbiont. BioEssays 25, 926–929 (2003).
45. Hooper, L.V., Falk, P.G. & Gordon, J.I. Analyzing the molecular foundations of 
commensalism in the mouse intestine. Curr. Opin. Microbiol. 3, 79–85 (2000).
46. Cuskin, F. et al. Human gut Bacteroidetes can utilize yeast mannan through a selfish 
mechanism. Nature 517, 165–169 (2015).
47. Vétizou, M. et al. Anticancer immunotherapy by CTLA-4 blockade relies on the gut 
microbiota. Science 350, 1079–1084 (2015).
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
nature medicinedoi:10.1038/nm.4358
ONLINE METhODS
Subjects. This study was approved by the Institutional Review Board of the 
Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, and was 
performed in accordance with the principle of the Helsinki Declaration II. 
A written informed consent was obtained from each participant.
We recruited 95 young (18~30 years) obese individuals (BMI, 37.03 ± 4.69 kg/m2, 
data are mean ± s.d.), and 105 age- and sex-matched lean controls (BMI, 
20.24 ± 1.26 kg/m2, data are mean ± s.d.) of Chinese ancestry (self-reported). 
All samples were collected in the specialized outpatient clinic for obesity at 
Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, from 
the Genetics of Obesity in Chinese Youngs (GOCY) study registered at 
ClinicalTrials.gov (NCT01084967)48,49. Individuals with secondary causes 
of obesity were excluded. Age- and sex matched healthy controls with BMI 
< 23 kg/m2 were recruited from volunteers of Shanghai Jiao Tong University 
School of Medicine. The exclusion criteria for controls were: hypertension; 
impaired glucose regulation; diabetes mellitus; history of obesity, gastrointes-
tinal disease and gastrointestinal surgery within 5 years before recruitment; 
and abnormal liver and kidney function, and blood lipid levels. A total of 23 
obese individuals (BMI, 44.53 ± 7.4 kg/m2, data are mean ± s.d.) with six from 
the case group were subjected to sleeve gastrectomy, and were investigated 
before, one month and three months after surgery. Each individual provided 
written informed consent. This study was registered at ClinicalTrials.gov 
(NCT02653430).
All subjects were weighed in light clothing without shoes. Body height and 
weight were measured by a height-weight scale, and BMI (kilograms per square 
meter) was calculated. Waist circumstance (WC) was measured at the midpoint 
between the lower border of the rib cage and the top of the lateral border of the 
iliac crest. Hip circumstance was measured at the widest part over the greater 
trochanters, and waist hip ratio (WHR) was calculated. Neck circumstance 
was measured using a tape measure around the neck just below the laryngeal 
prominence and the mean of two recordings was recorded. Whole body fat was 
measured by Korean Inbody 720. Blood pressure was measured at the right arm 
by a standard brachial cuff technique three times consecutively with one min 
intervals after at least five min rest in the seated position. The three readings 
were averaged for analysis.
Biochemical measurements. Blood samples for clinical chemistry analyses were 
taken after an overnight fast for at least 10 h. Serum samples were centrifuged 
and stored at −80 °C until analysis. Fasting or 2-h glucose, serum alanine ami-
notransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase 
(ALP) and γ-glutamyl transpeptidase (GGT), total bilirubin, creatinine, uric acid, 
lipid profile, including triglycerides (TG), total cholesterol (TC), high-density 
lipoprotein (HDL) cholesterol, and low-density lipoprotein (LDL) cholesterol 
were measured using an autoanalyser (Beckman Coulter AU5800). Fasting or 
2-h serum insulin was measured using a double antibody radioimmunoassay 
(DSL, Webster). Insulin resistance index (HOMA-IR) was calculated using 
homeostasis model assessment methods, defined as fasting insulin (IU/ml) × 
fasting glucose (mmol/L)/22.5. Pancreatic β-cell function index (homeostasis 
model assessment of β-cell function, HOMA-β) was defined as 20× fasting 
insulin (IU/ml) / (fasting glucose (mmol/L)-3.5). HBA1C was measured by high-
pressure liquid chromatography.
Serum leptin, IL-6 and TNF-α were measured using a MILLIPLEX MAG 
Human Adipokine Magnetic Bead Panel (Millipore) according to the manu-
facturer’s instructions. Serum adiponectin (Millipore), serum lipopolysaccha-
ride binding protein (LBP) (Abnova), FIAF (fasting-induced adipose factor) 
(BioVendor) and GLP-2 (Crystal Chem) were measured using ELISA kits 
according to the manufacturer’s instructions. hsCRP was measured by latex 
agglutination method (Roche CYT298).
Fecalsample collection and DNA extraction. Fecal samples were obtained 
from all recruited subjects for metagenomic sequencing. The individuals had 
not received any antibiotic treatment for at least one month before sample col-
lection. In the seven days before sample collection, subjects did not take any food 
containing probiotics such as yogurt. Each sample was either frozen immediately 
at −80 °C or briefly stored in personal −20 °C freezers before transport to the 
laboratory within 24 h. DNA was extracted as previously described7.
Metagenomic sequencing. All samples were sequenced on the Illumina plat-
form (Paired-end; insert size, 350 bp; read length, 100 bp). Adaptor and low-
quality reads were discarded from the raw reads, and the remaining reads were 
filtered in order to eliminate human host DNA based on the human genome 
reference (hg18) as described previously7. We acquired 1,403.5 Gb high-quality 
pair-end reads for the 257 samples with an average of 5.2 Gb per sample after 
removing human DNA reads (Supplementary Table 26).
Construction of gene, phylum, genus, species and KO profiles. The high-
quality reads were aligned to the updated gut microbiome gene catalog17 by 
SOAP2 and 80 ± 3.5% reads (n = 257, data are mean ± s.d.; Supplementary 
Table 26) were mapped with a threshold of more than 90% identity over 95% of 
the length50. Sequence-based gene abundance profiling was performed as previ-
ously described17. After removing genes that were present in less than 6 samples 
across the 151 samples in the training set, 2,981,416 genes remained. The relative 
abundances of phyla, genera, species and KOs were calculated from the relative 
abundance of their respective genes using previously published methods17.
Rarefaction curve analysis. To assess gene richness in the control and obese 
samples, we generated a rarefaction curve as previously described51. Briefly, 
for a given number of samples, we performed a randomized sampling 79 and 
72 times in the control samples and obese samples, respectively, and estimated 
the total number of genes that could be identified from these samples. Genes 
with >1 pair of mapped reads were considered to be present in a sample.
Gene count, a-diversity and b-diversity. The total gene count in each sample 
was calculated as previously described8. α-diversity (within-sample diversity) 
was estimated on the basis of the gene profile of each sample according to the 
Shannon index, as described previously7. β-diversity (between-sample diversity) 
was estimated by Bray-Curtis distance 
BC
X X
X X
ij
ik jkk
n
ik jkk
n=
−
+
=
=
∑
∑
| |
( )
1
1
where n is the number of genes, and Xik and Xjk represent the relative abundance 
of gene k in samples i and j.
Metagenome-wide association study. Of the 2,981,416 identified genes, 350,524 
genes differed in abundance between controls and obese subjects (two-tailed 
Wilcoxon rank-sum test, P < 0.01, FDR < 0.05). These genes were subsequently 
clustered into MLGs according to their abundance variation across all samples7. 
Briefly, the Kendall tau rank correlation coefficient was used to calculate the rela-
tionship between genes according to their abundance, and then the Chameleon 
algorithm was applied to cluster the genes into MLGs.
Taxonomic assignment and abundance profiling of the MLGs (Supplementary 
Table 3) were performed according to the taxonomy and the relative abundance 
of the constituent genes17, using the method previously described7. MLGs with 
>50% genes annotated to the same species or genus were given an annota-
tion. The MLGs with number of genes >100 were further clustered according 
to Spearman’s correlation between their abundances in all samples regardless 
of case-control status, and the co-occurrence network was visualized with 
Cytoscape 3.1.1 software if the |Spearman’s rank correlation coefficient| > 0.6.
Association between MLGs and clinical indices. PERMANOVA52 was per-
formed on the MLG-abundance profiles of the samples to assess the effect of each 
of the clinical indices51 using Bray-Curtis distance and 9,999 permutations in R 
(3.10, vegan package53). Clinical indices with adjusted P < 0.05 were considered 
to associate with MLGs.
Random Forests (R 3.0.1, randomForest4.6-10 package)54 regression was used 
to regress relative abundances of MLGs in the training set (79 controls; 72 cases) 
against clinical indices with adjusted P < 0.05 in the PERMANOVA using default 
parameters of the R implementation of the algorithm55. The model was gener-
ated using all MLGs and the importance of each MLG was subsequently ranked 
by the % increase in mean square error (%IncMSE) occurring by randomly 
permuted the relative abundance values of the corresponding MLG. To estimate 
© 
20
17
 N
at
u
re
 A
m
er
ic
a,
 In
c.
, p
ar
t 
o
f 
S
p
ri
n
g
er
 N
at
u
re
. A
ll 
ri
g
h
ts
 r
es
er
ve
d
.
nature medicine doi:10.1038/nm.4358
the most suitable number of top ranking clinical indices-discriminatory MLGs, 
we repeated random Forest regression with top n (n = 2,3,4…50) ranking MLGs 
against the same clinical indices and picked the combinations which had the 
biggest rsq (pseudo R-squared)54, defined as 
ˆ
( ˆ )
( )
y R
y y
y y
i ii
N
i ii
N
2
2
1
2
1
1= −
−
−
=
=
∑
∑
where N is the numbers of observations in the model, yi is the i observation of 
the clinical indices and yˆi is it predicted value of yi, while y represents the mean 
of the clinical indices.
The MLGs selected by the model were considered to associate with the corre-
sponding clinical index. The model was then used to predict the clinical indices 
in the test set (26 controls; 23 cases). Clinical indices with R2 >0 in both train-
ing set and test set were considered to associate stably with the picked MLGs. 
MLG-based classifier and prediction of the surgery samples were done in the 
same manner.
Spearman’s correlation was used to describe the specific correlation between 
clinical indices and selected MLGs, as described previously6.
Metabolomics profiling of human serum samples. Sample preparations. All 
serum samples were thawed on ice and a quality control (QC) sample, made by 
mixing and blending equal volumes (10 µl) of each serum sample, was used to 
estimate a mean profile representing all the analytes encountered during analy-
sis. We isolated and extracted metabolites (<1,500 Da) as follows. First, 100-µl 
serum were precipitated with 200-µl methanol, and similarly, the QC sample was 
precipitated with methanol (1:2 v/v). All samples were subsequently centrifuged 
at 14,000 g for 10 min at 4 °C. The supernatants were subjected to metabolomics 
profiling by high performance liquid chromatograph (HPLC)-MS.
HPLC-MS experiments. A Shimadzu Prominence HPLC system (Shimadzu) 
coupled to a LTQ Orbitrap Velos instrument (Thermo Fisher Scientific) was set 
at 30,000 resolution to acquire HPLC-MS data. Sample analysis was performed 
in positive ion modes using spray voltage of 4.5 kV and the capillary tempera-
ture at 350 °C. The mass scanning range was set at 50–1,500 m/z. Nitrogen 
sheath gas and nitrogen auxiliary gas were set at a flow rate of 30 l/min and 
10 l/min, respectively. The HPLC-MS system was run in binary gradient mode. 
Solvent A was 0.1% (v/v) formic acid/water, and solvent B was 0.1% (v/v) for-
mic acid/methanol. The gradient was set as follows: 5% B at 0 min, 5% B at 
5 min, 100% B at 8 min, 100% B at 9 min, 5% B at 18 min, and 5% B at 20 min. 
The flow rate was set to 0.2 ml/min. The pooled QC sample was injected five 
times at the beginning to ensure system equilibrium and then it was injected 
every five samples during serum sample detection to further monitor system 
stability. Formic acid and methanol (HPLC

Continue navegando