HAL Id: hal-01193804
https://hal.archives-ouvertes.fr/hal-01193804
Submitted on 29 May 2020
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entic research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diusion de documents
scientiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Intestinal microbiota determines development of
non-alcoholic fatty liver disease in mice
Tiphaine Le Roy, Marta Llopis, Patricia Lepage, Aurelia Bruneau, Sylvie
Rabot, Claudia Bevilacqua, Patrice Martin, Catherine Philippe, Francine
Walker, Andre Bado, et al.
To cite this version:
Tiphaine Le Roy, Marta Llopis, Patricia Lepage, Aurelia Bruneau, Sylvie Rabot, et al.. Intestinal
microbiota determines development of non-alcoholic fatty liver disease in mice. Gut, BMJ Publishing
Group, 2013, 62 (12), pp.1787-1794. �10.1136/gutjnl-2012-303816�. �hal-01193804�
ORIGINAL ARTICLE
Intestinal microbiota determines development
of non-alcoholic fatty liver disease in mice
Tiphaine Le Roy,
1,2
Marta Llopis,
1,2
Patricia Lepage,
1,2
Aurélia Bruneau,
1,2
Sylvie Rabot,
1,2
Claudia Bevilacqua,
3
Patrice Martin,
3
Catherine Philippe,
1,2
Francine Walker,
4
André Bado,
4
Gabriel Perlemuter,
5,6,7
Anne-Marie Cassard-Doulcier,
5,6
Philippe Gérard
1,2
▸ Additional supplementary
files are published online only.
To view these files please visit
the journal online (http://dx.
doi.org/10.1136/gutjnl-2012-
303816).
1
INRA, UMR1319 Micalis,
Jouy-en-Josas, France
2
AgroParisTech, UMR Micalis,
Jouy-en-Josas, France
3
INRA, UMR1313 GABI,
Plateforme de Microgénomique
expressionnelle “Iso Cell
Express” (ICE), Jouy-en-Josas,
France
4
INSERM U773, UFR de
Médecine Paris Diderot,
Université Paris Diderot,
Sorbonne Paris Cité, Paris,
France
5
INSERM, U996, IPSIT,
Clamart, France
6
Faculté de médecine Paris-
Sud, Université Paris-Sud,
Kremlin-Bicêtre, France
7
AP-HP, Hôpital Antoine
Béclère, Service d’hépato-
gastroentérologie, Clamart,
France
Correspondence to
Dr Philippe Gérard, INRA
(UMR1319), Micalis Institute,
Team AMIPEM,
Domaine de Vilvert,
Jouy-en-Josas F-78350, France;
philippe.gerard@jouy.inra.fr
Received 26 October 2012
Accepted 30 October 2012
Published Online First
29 November 2012
To cite: Le Roy T, Llopis M,
Lepage P, et al. Gut
2013;62:1787–1794.
ABSTRACT
Objective Non-alcoholic fatty liver disease (NAFLD) is
prevalent among obese people and is considered the
hepatic manifestation of metabolic syndrome. However,
not all obese individuals develop NAFLD. Our objective
was to demonstrate the role of the gut microbiota in
NAFLD development using transplantation experiments
in mice.
Design Two donor C57BL/6J mice were selected on
the basis of their responses to a high-fat diet (HFD).
Although both mice displayed similar body weight
gain, one mouse, called the ‘responder’, developed
hyperglycaemia and had a high plasma concentration of
pro-inflammatory cytokines. The other, called a ‘non-
responder’, was normoglycaemic and had a lower level
of systemic inflammation. Germ-free mice were colonised
with intestinal microbiota from either the responder or
the non-responder and then fed the same HFD.
Results Mice that received microbiota from different
donors developed comparable obesity on the HFD.
The responder-receiver (RR) group developed fasting
hyperglycaemia and insulinaemia, whereas the
non-responder-receiver (NRR) group remained
normoglycaemic. In contrast to NRR mice, RR mice
developed hepatic macrovesicular steatosis, which was
confirmed by a higher liver concentration of triglycerides
and increased expression of genes involved in de-novo
lipogenesis. Pyrosequencing of the 16S ribosomal RNA
genes revealed that RR and NRR mice had distinct gut
microbiota including differences at the phylum, genera
and species levels.
Conclusions Differences in microbiota composition
can determine response to a HFD in mice. These results
further demonstrate that the gut microbiota contributes
to the development of NAFLD independently of obesity.
INTRODUCTION
Non-alcoholic fatty liver disease (NAFLD) is con-
sidered the hepatic manifestation of metabolic syn-
drome and is commonly associated with insulin
resistance.
1
NAFLD affects 20–30% of western
countries’ population and more than 80% of obese
people. It refers to a spectrum of liver damage
ranging from simple steatosis to non-alcoholic stea-
tohepatitis, advanced fibrosis, cirrhosis or even
hepatocellular carcinoma. An increasing body of lit-
erature has recently been generated identifying gut
microbiota as a new environmental factor
contributing to obesity and NAFLD.
23
First,
Bäckhed and colleagues
45
showed that germ-free
(GF) C57BL/6J mice gained less weight than con-
ventional mice when given a sugar and lipids-rich
diet despite greater food consumption. Moreover,
Rabot et al
6
observed that GF mice receiving a
high-fat diet (HFD) showed enhanced insulin sensi-
tivity with improved glucose tolerance and reduced
insulinaemia. Concurrently, colonisation of GF
mice by a gut microbiota from conventional mice
produced an increase in body fat content.
4
More
recently, an association between the human gut
Significance of this study
What is already known on this subject?
▸ We and others have shown that GF mice are
resistant to diet-induced obesity, insulin
resistance and steatosis.
▸ The administration of a HFD to conventional
mice leads to heterogeneous responses
including variable levels of weight gain,
glycaemia or steatosis development.
▸ Each mouse and human harbours a different
gut microbiota.
What are the new findings?
▸ Gut microbiota markedly impacts the lipid
metabolism in the liver, independently of
obesity.
▸ The propensity to develop NAFLD features
including hyperglycaemia or steatosis is
transmissible by means of gut microbiota
transplantation.
▸ Bacterial species associated with the
NAFLD-resistant and NAFLD-prone phenotypes
have been identified.
How might it impact on clinical practice in
the foreseeable future?
▸ These findings suggest that manipu lation of the
gut microbiota may be a new strategy to
prevent or treat NAFLD and associated
metabolic disorders including type 2 diabetes
or metabolic syndrome. In addition, gut
microbiota profiling could help predict the
susceptibility to develop metabolic disorders.
Le Roy T, et al. Gut 2013;62:1787–1794. doi:10.1136/gutjnl-2012-303816 1787
Heptology
group.bmj.com on February 18, 2014 - Published by gut.bmj.comDownloaded from
microbiota and the development of fatty liver due to choline
deficiency has been identified.
7
Moreover, it was revealed that
changes in the gut microbiota associated with inflammasome
defects regulates the progression of NAFLD.
8
HFD feeding is widely used in rodents to study the onset and
progression of obesity and associated metabolic disorders.
However, the high-fat-induced phenotype varies distinctly, even
within a group of animals with the same genetic background,
910
and it has recently been demonstrated that disctinct gut micro-
biota profiles are associated with different metabolic pheno-
types.
11
Gut microbiota of humans and mice are more than
95% made up of three phyla: Firmicutes, Bacteroidetes and
Actinobacteria. In contrast to the poor diversity at the phylum
level, the species level displays a high diversity, with an average
human microbiota estimated at 200 prevalent bacterial species
and up to 1000 less common species.
12
Importantly, in humans
but also in mice, the bacterial species profiles are unique for
each individual.
13
We therefore hypothesised that the gut micro-
biota modulates the effect of a HFD challenge and that the vari-
ability of the composition of this microbiota could explain the
diversity of responses to HFD.
In the present work, we developed a strategy based on gut
microbiota transfer to establish whether we can transmit the dif-
ferent propensity to develop NAFLD in response to HFD by
means of gut microbiota transplant.
METHODS
Animal experimentation
Stage 1
Eight-week-old conventional male C57BL/6J mice ( Janvier, Le
Genest St Isle, France) were fed for 16 weeks a freely available
sterilised HFD containing 40% butter and 2.5% soy oil (60%
energy from fat, 22.7% from carbohydrates and 17.2% from
proteins; SAFE, Augy, France). Body weight was monitored
weekly.
Stage 2
GF male C57BL/6J mice were reared from GF breeding pairs at
ANAXEM, the GF animal facilities of Micalis ( Jouy-en-Josas,
France). Two groups of 8-week-old GF mice were colonised
with the gut microbiota of two conventional mice from stage 1
according to the following procedure. The caecal content of
donor mice was collected immediately after euthanasia and
diluted in liquid casein yeast medium (1 : 100 w/vol) in an
anaerobic chamber. Colonisation was achieved by oral-gastric
gavage with 250 ml of the diluted caecal content. GF mice were
switched from chow to HFD 1 week before the colonisation
and were then freely fed the HFD for an additional 16 weeks
after colonisation.
All mice were anaesthetised with isoflurane. Procedures were
carried out in accordance with the European guidelines for the
care and use of laboratory animals and with permission 78–60
of the French veterinary services.
Measurement of plasma parameters and short chain fatty
acid concentrations
After 16 weeks of diet, blood taken from the retro-orbital sinus
was collected into chilled heparin-coated tubes after 6 h of
fasting. Blood glucose was measured using an Accu-Check gluc-
ometer (Roche Diagnostics). The remaining blood was then cen-
trifuged at 10 000 g for 10 min. Plasma was aliquoted and
frozen at −80°C until analysis.
Alanine and aspartate aminotransferases and lipids were deter-
mined using an Olympus AU400 robot. Plasma insulin, leptin,
resistin, monocyte chemotactic protein 1, tumour necrosis
factor α (TNFα) and interleukin (IL)-6 concentrations were
assayed using a Luminex 100 IS system (Luminex Corporation)
with a Milliplex MAP mouse serum adipokine panel kit
(Millipore). Insulin resistance was estimated by homeostasis
model assessment (HOMA index): fasting serum glucose (mg/
dl)×insulin (mU/l)/405. Short-chain fatty acids (SCFA) and
branched-chain fatty acids (BCFA) were assayed in caecal
samples from receiver mice as previously described.
14
Measurement of liver triglycerides
Portions of frozen liver from receiver mice were homogenised in
chloroform–methanol (2 : 1) in order to extract total lipids
according to the methodology of Folch et al.
15
The organic
extract was dried and reconstituted in isopropanol. The trigly-
ceride content was measured with a triglycerides determination
kit (Sigma-Aldrich, Saint-Louis, Missouri, USA) according to the
manufacturer’s instructions and expressed in mmol of triglycer-
ides per milligram of liver.
Liver histology
Thin slices of formaldehyde-fixed, paraffin-embedded liver
tissue were stained with haematoxylin and scored for severity of
the steatosis and inflammation by an experienced pathologist
(FW) blinded to the experiments. The steatosis score was
assessed according to the percentage of concerned hepatocytes
multiplied by the following grade relying on the size of the fat
droplets: grade 1: microvesicular pattern; 2: mixed microvesicu-
lar superior to macrovesicular pattern; 3: mixed macrovesicular
superior to microvesicular pattern; and 4: macrovesicular
ones only.
Gene expression analysis by quantitative PCR
Livers of the receiver mice were disrupted in RNA-PLUS solu-
tion (QBiogene). Total RNA was extracted using a modified
Chomczynski’s procedure.
16
RNA concentration and purity
were determined using the Nanodrop ND-1000 spectrophotom-
eter (Thermo Fisher Scientific) at a wavelength of 260/280 nm.
RNA integrity was determined with the Agilent bioanalyzer
2100 system with the RNA 6000 Nano LabChip kit. Samples
with an RNA integrity number inferior to 8 on a scale ranging
from 0 to 10 were eliminated.
Total RNA (2 m g per reaction) was reverse transcribed into
complimentary DNA using a high-capacity cDNA reverse tran-
scription kit (Applied Biosystems) according to the manufac-
turer’s instructions. Thereafter, half of the product of each
reverse transcription reaction was added to an equal volume of
TaqMan universal PCR master mix and then loaded on a
TaqMan low density array card (Applied Biosystems). The
TaqMan low density array card was centrifuged twice for 1 min
at 330g before being sealed. PCR ampli fication was performed
using an Applied Biosystems Prism 7900HT sequence detection
system. Data were analysed using SDS V.2.2 software. Statistical
analysis was performed using DataAssist v2.0 (Applied
Biosystems). The relative gene expressions were normalised to
three reference genes: 18S, gapdh and ubc2, chosen on the basis
of results obtained from TaqMan mouse endogenous control
arrays (Applied biosystems). The non-responder-receiver (NRR)
group was used as a reference for the relative expression of
genes.
16S rRNA sequencin g
Microbiota compositon was thoroughly analysed using 454 pyr-
osequencing targeting the 16S ribosomal DNA V3–V4 region
1788 Le Roy T, et al. Gut 2013;62:1787–1794. doi:10.1136/gutjnl-2012-303816
Heptology
group.bmj.com on February 18, 2014 - Published by gut.bmj.comDownloaded from
(V3fwd: 5
0
TACGGRAGGCAGCAG 3
0
, V4rev: 5
0
GGACTACCAGGGTATCTAAT 3
0
). Forty DNA samples, corre-
sponding to the two caecal samples from donor mice used for
inoculation, and faecal samples from receiver mice at two sam-
pling time, were pyrosequenced at Genoscreen (Lille, France)
using GS-FLX-Titanium technology. Sequences were trimmed
for adaptors and PCR primer removal and binned for a minimal
sequence length of 300 bases, a maximum homopolymers
length of 6, a minimal base quality threshold set at 27 and a
maximum of 15% tolerated low quality bases and N on the
overall sequence length. The resulting sequences were assigned
to different taxonomic levels (from phylum to genus) using the
RDP database (release 10, update 26).
17
Using QIIME,
18
sequences were further clustered in at 97% of identity in oper-
ational taxonomic unit (OTU) using cdhit.
19
OTU were assigned
to the closest taxonomic neighbours and relative bacterial
species using SEQMATCH and up-to-date 16S rRNA gene RDP
database. Estimates of phylotype richness were calculated
according to the bias-corrected Chao1 estimator. All statistical
analyses were performed using the R program and ade4 package
(http://pbil.univ-lyon1.fr/ADE-4/). Principal component analyses
(PCA) with the two receiver mice groups at different time
points as instrumental variables (interclass PCA) were computed
and statistically assessed by a Monte Carlo rank test to observe
their net effect on the scattering of the microbiota of different
mice. Interclass PCA allows highlighting combinations of vari-
ables that maximise variations observed between qualitative vari-
ables (eg, specific responder-receiver (RR)/NRR groups). The
Wilcoxon test was applied to assess statistical signifi cance in bac-
terial composition between the different samples.
Statistical analyses
Results are represented as mean±SEM, or median (IQR) for
non-parametric data. Statistical analysis was performed by
Student’s t or Mann–Whitney–Wilcoxon tests, respectively
(StatView, SAS Institute Inc, Cary, USA); p<0.05 was considered
statistically significant.
RESULT S
Selection of the responder and non-responder donor mice
Conventional mice of the strain C57BL/6J were freely fed a
HFD for 16 weeks. HFD treatment led to varying body weights
ranging from 33.6 to 47.0 g, with an average final body weight
of 39.0±3.9 g (table 1). Despite a majority of the mice develop-
ing metabolic disorders, several mice developed high levels of
glycaemia, systemic inflammation and steatosis together and
were considered as ‘responders’. Conversely, several mice did
not develop any metabolic disorders and were considered ‘non-
responders’. Therefore, we selected from this cohort one
responder and one non-responder mouse to verify if we could
transmit to GF mice the responder or non-responder pheno-
types using gut microbiota transplants. To ensure reliable recov-
ery of the initial gut microbiota in GF mice, transfer of
microbiota from donor to receiver mice had to be performed
using fresh caecal samples. As a consequence, selection of the
donor mice were performed on the basis of parameters available
the day of conventional mice euthanasia, including body weight,
fat pad and liver weights, food intake, HOMA index and
plasma concentrations of pro-inflammatory cytokines (table 1).
We intentionally selected two donor mice with similar body
weights, fat pad masses and food intake to make sure that the
differences in metabolic parameters were not the consequence
of a different degree of obesity.
The two groups of receiver mice developed comparable
obesity but different metabolic statuses
Two groups of GF male C57BL/6J mice were colonised with the
inocula prepared from the two selected donor mice (conventio-
nalisation). The two conventionalised groups were then named
NRR and RR, and were subsequently maintained in isolators.
After 16-week HFD feeding, both groups showed similar body
weight gains (13.2±3.2 g and 14.6±2.6 g for NRR and RR,
respectively, p=0.21) and final body weights (figure 1A). Daily
food consumption was 2.52±0.92 g and 2.62±0.91 g for NRR
and RR mice, respectively, indicating no differences in food
intake. Accordingly, epididymal fat pad weights were not found
to be different between these two groups (see supplementary
table S1, available online only). Conversely, they displayed sig-
nificantly different fasting glycaemia levels (104.4±26.6 mg/dl
for NRR vs 134±34.1 mg/dl for RR) (figure 1B). Concurrently,
fasting insulinaemia was lower in NRR mice than in RR mice
(665.2±358.9 g/ml vs 1072.4±469.4 pg/ml) (figure 1C). The
HOMA index (figure 1D) was consistently 2.4-fold greater in
the RR group than in the NRR group, suggesting that the two
groups developed different levels of insulin resistance.
Leptinaemia (figure 1E) was 75% higher in RR than in NRR
mice, whereas the level of resistin was not significantly different
in the two groups (data not shown).
The aspartate aminotransferase plasma concentration was
three times higher in RR mice than in NRR mice (433.0±201.2
vs 135.9±73.7 UI/l, p=0.0058), whereas the difference did not
achieve statistical significance (68.5±56.8 vs 38.1±34.4 UI/l,
p=0.0786) for alanine aminotransferase concentrations. Fasting
plasma concentrations of triglycerides, cholesterol and high-
density lipoproteins were similar in the two groups (see supple-
mentary table S2, available online only).
Concentrations and proportions of SCFA (acetate, butyrate,
propionate, valerate, caproate) were found to be similar in the
Table 1 Metabolic status of conventional and donor mice after 16 weeks of HFD
Body
weight
gain (g)
Final body
weight (g)
Average
food intake
(g/day)
Liver weight
(% of body
weight)
Epididymal fat
pads (% of
body weight)
Fasting
glycaemia
(mg/dl)
Fasting
insulinaemia
(pg/ml) HOMA-IR
MCP-1
(pg/ml)
TNF-α
(pg/ml)
Conventional HFD
mean
15.3±3.3 39±3.9 4.8±0.3 4.65±0.21 5.02±0.17 164±39 1944±820 19.46±9.58 36.2±47.1 5.4±4.4
Responder donor 17.6 40.2 4.7 3.77 5.03 191 3673 42.94 199 13.1
Non-responder
donor
15.7 39.4 5.1 4.38 4.49 90 2054 11.31 4 3.5
Data obtained from a cohort of 24 conventional mice, data are mean±SEM.
HFD, high-fat diet; HOMA-IR, homeostasis model assessment of insulin resistance; MCP-1, monocyte chemotactic protein 1.
Le Roy T, et al. Gut 2013;62:1787–1794. doi:10.1136/gutjnl-2012-303816 1789
Heptology
group.bmj.com on February 18, 2014 - Published by gut.bmj.comDownloaded from
caecums of NRR and RR groups (see supplementary table S3,
available online only). Consistently, the total SCFA concentra-
tion was not different in the two groups. Conversely, concentra-
tions and proportions of isobutyrate and isovalerate were
significantly higher in the caecum of RR mice (see supplemen-
tary table S4, available online only). These BCFA are com-
pounds known to result from the bacterial fermentation of
valine and leucine.
RR mice accumulated more triglycerides in the liver
than NRR mice
Liver histological analysis showed that the NRR group devel-
oped slight to mild steatosis (figure 2A), whereas the RR group
developed marked mixed or macrovesicular steatosis (figure 2B).
The steatosis score was found to be higher in RR than in NRR
mice (3.00±0.87 vs 1.50±0.61) (figure 2C). Consistent with
the morphological changes in the lipid deposition, the liver
triacylglycerol concentration was 30% higher in RR mice (figure
2D). No inflammatory infiltrate was observed and no differences
in liver weights were found between the two groups (see supple-
mentary table S1, available online only).
RR mice displayed a steatosis-prone hepatic metabolism
in contrast to NRR mice
We analysed the hepatic expression of genes involved in lipid
uptake, lipogenesis, fatty acid catabolism and very low-density
lipoprotein export. The relative expressions of the transcription
factors sterol regulatory binding protein (SREBP) 1c, liver X
receptor and carbohydrate response element binding protein
(ChREBP) are shown figure 3A. These nuclear factors are
known to promote de-novo lipogenesis.
20
The expression of
SREBP1c and ChREBP was 1.97 and 2.02 greater in the RR
group than in the NRR group, whereas no difference was found
in the expression of liver X receptor between the two groups.
The relative expressions of three lipogenic enzymes (acetyl-CoA
carboxylase 1, stearoyl-CoA desaturase 1 and fatty acid syn-
thase) are shown figure 3B. Stearoyl-CoA desaturase 1 and fatty
acid synthase were not differently expressed in the two groups
of mice. Conversely, acetyl-CoA carboxylase 1 appeared to be
upregulated in RR mice (fold change 1.97). CD36, which
imports a large variety of lipids and lipoproteins, was more
highly expressed in RR than in NRR mice (fold change 2.32).
On the contrary, fatty acid transport protein 5, which essentially
transports long-chain fatty acids, was slightly but significantly
downregulated (fold change 0.80). No differences were found
between the two groups of mice in the messenger RNA levels of
carnitine palmytotransferase 1a, a transport protein that regu-
lates mitochondrial β-oxidation, and membrane transport
protein, a protein exerting a central regulatory role in lipopro-
tein assembly and secretion (figure 3C).
No major differences in systemic and hepatic inflammation
were detected between the two groups of receiver mice
Systemic inflammation was evaluated by assaying the plasma
concentration of pro-inflammatory cytokines. No significant dif-
ferences in the plasma concentrations of these cytokines were
found between the two groups of receiver mice (see supplemen-
tary table S2, available online only). We then focused specifically
on liver inflammation by assaying the hepatic expression of
cytokines, markers of liver macrophages and Toll-like receptors
(TLR). The relative expression of TNFα, IL-1β, IL-6, IL-10 and
transforming growth factor β were found to be similar in the
two groups of mice (see supplementary table S5, available
online only). Likewise, expressions of CD68 (a marker of
macrophages), and TLR (TLR-2, TLR-4, TLR-5 and TLR-9),
which play a central role in the innate immune system by recog-
nising bacterial components leading to activation of immune
response, appeared to be equivalent in NRR and RR mice
(figure 4A,B). Altogether these results indicate that systemic and
hepatic inflammation was similar in RR and NRR mice.
Gut microbiota differs between RR and NRR mice
Caecal samples used for inoculation as well as faecal samples
from receiver mice after 3 (T3) and 16 (T16) weeks of HFD
were analysed by pyrosequencing. A total of 188 058 sequences
was obtained and after trimming, 120 241 sequences were
further analysed (approximately 3000 sequences/sample). To
evaluate similarity among the samples, interclass PCA was per-
formed based on their microbial composition. The main genera
between caecal samples from donor mice and faecal samples
from corresponding receiver mice were conserved but showed
Figure 1 Metabolic responses to high-fat diet of
non-responder-receiver (NRR) and responder-receiver (RR) mice.
(A) Body weight curves; (B) fasting glycaemia; (C) fasting insulinemia;
(D) homeostasis model assessment—insulin resistance Index; (E) fasting
leptinaemia. All mice were fasted for 6 h before blood sampling. Data
are mean±SEM, n=18 for NRR and n=22 for RR mice. **p<0.01;
***p<0.001 (Student’s t test).
1790 Le Roy T, et al. Gut 2013;62:1787–1794. doi:10.1136/gutjnl-2012-303816
Heptology
group.bmj.com on February 18, 2014 - Published by gut.bmj.comDownloaded from