Antibiotics in early life alter the murine colonic microbiome and adiposity.
ABSTRACT Antibiotics administered in low doses have been widely used as growth promoters in the agricultural industry since the 1950s, yet the mechanisms for this effect are unclear. Because antimicrobial agents of different classes and varying activity are effective across several vertebrate species, we proposed that such subtherapeutic administration alters the population structure of the gut microbiome as well as its metabolic capabilities. We generated a model of adiposity by giving subtherapeutic antibiotic therapy to young mice and evaluated changes in the composition and capabilities of the gut microbiome. Administration of subtherapeutic antibiotic therapy increased adiposity in young mice and increased hormone levels related to metabolism. We observed substantial taxonomic changes in the microbiome, changes in copies of key genes involved in the metabolism of carbohydrates to short-chain fatty acids, increases in colonic short-chain fatty acid levels, and alterations in the regulation of hepatic metabolism of lipids and cholesterol. In this model, we demonstrate the alteration of early-life murine metabolic homeostasis through antibiotic manipulation.
- [Show abstract] [Hide abstract]
ABSTRACT: Traditionally bacteria have been considered as either pathogens, commensals or symbionts. The mammal gut harbors 10(14) organisms dispersed on approximately 1000 different species. Today, diagnostics, in contrast to previous cultivation techniques, allow the identification of close to 100% of bacterial species. This has revealed that a range of animal models within different research areas, such as diabetes, obesity, cancer, allergy, behavior and colitis, are affected by their gut microbiota. Correlation studies may for some diseases show correlation between gut microbiota composition and disease parameters higher than 70%. Some disease phenotypes may be transferred when recolonizing germ free mice. The mechanistic aspects are not clear, but some examples on how gut bacteria stimulate receptors, metabolism, and immune responses are discussed. A more deeper understanding of the impact of microbiota has its origin in the overall composition of the microbiota and in some newly recognized species, such as Akkermansia muciniphila, Segmented filamentous bacteria and Faecalibacterium prausnitzii, which seem to have an impact on more or less severe disease in specific models. Thus, the impact of the microbiota on animal models is of a magnitude that cannot be ignored in future research. Therefore, either models with specific microbiota must be developed, or the microbiota must be characterized in individual studies and incorporated into data evaluation.World journal of gastroenterology : WJG. 12/2014; 20(47):17727-17736.
- [Show abstract] [Hide abstract]
ABSTRACT: The second-generation antipsychotic olanzapine is effective in reducing psychotic symptoms but can cause extreme weight gain in human patients. We investigated the role of the gut microbiota in this adverse drug effect using a mouse model. First, we used germ-free C57BL/6J mice to demonstrate that gut bacteria are necessary and sufficient for weight gain caused by oral delivery of olanzapine. Second, we surveyed fecal microbiota before, during, and after treatment and found that olanzapine potentiated a shift towards an "obesogenic" bacterial profile. Finally, we demonstrated that olanzapine has antimicrobial activity in vitro against resident enteric bacterial strains. These results collectively provide strong evidence for a mechanism underlying olanzapine-induced weight gain in mouse and a hypothesis for clinical translation in human patients.PLoS ONE 01/2014; 9(12):e115225. · 3.53 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: The microbiota of the human gut is gaining broad attention owing to its association with a wide range of diseases, ranging from metabolic disorders (e.g. obesity and type 2 diabetes) to autoimmune diseases (such as inflammatory bowel disease and type 1 diabetes), cancer and even neurodevelopmental disorders (e.g. autism). Having been increasingly used in biomedical research, mice have become the model of choice for most studies in this emerging field. Mouse models allow perturbations in gut microbiota to be studied in a controlled experimental setup, and thus help in assessing causality of the complex host-microbiota interactions and in developing mechanistic hypotheses. However, pitfalls should be considered when translating gut microbiome research results from mouse models to humans. In this Special Article, we discuss the intrinsic similarities and differences that exist between the two systems, and compare the human and murine core gut microbiota based on a meta-analysis of currently available datasets. Finally, we discuss the external factors that influence the capability of mouse models to recapitulate the gut microbiota shifts associated with human diseases, and investigate which alternative model systems exist for gut microbiota research. © 2015. Published by The Company of Biologists Ltd.Disease Models and Mechanisms 01/2015; 8(1):1-16. · 4.96 Impact Factor
Antibiotics in early life alter the murine
colonic microbiome and adiposity
Ilseung Cho1,2, Shingo Yamanishi1, Laura Cox3, Barbara A. Methe ´4, Jiri Zavadil5,6, Kelvin Li4, Zhan Gao3, Douglas Mahana3,
Kartik Raju3, Isabel Teitler3, Huilin Li7, Alexander V. Alekseyenko1,6& Martin J. Blaser1,2,3
Antibiotics administered in low doses have been widely used as growth promoters in the agricultural industry since the
structure of the gut microbiome as well as its metabolic capabilities. We generated a model of adiposity by giving
subtherapeutic antibiotic therapy to young mice and evaluated changes in the composition and capabilities of the gut
microbiome. Administration of subtherapeutic antibiotic therapy increased adiposity in young mice and increased
hormone levels related to metabolism. We observed substantial taxonomic changes in the microbiome, changes in
copies of key genes involved in the metabolism of carbohydrates to short-chain fatty acids, increases in colonic
short-chain fatty acid levels, and alterations in the regulation of hepatic metabolism of lipids and cholesterol. In this
model, we demonstrate the alteration of early-life murine metabolic homeostasis through antibiotic manipulation.
Antibiotics, discovered in the early twentieth century, came into
widespread use after the Second World War, with substantial
public health benefits. Antibiotic use has increased markedly, now
approximating one antibiotic course per year in the average child in
the United States1,2. However, there is increasing concern that
antibiotic exposure may have long-term consequences3–5.
For more than 50years we have known that the administration of
consequently, in the United States, the largest use of antibiotics and
relatedantimicrobial substances is within farms, with low doses fed to
large numbers of animals used for food production to increase weight
gain by as much as 15%6,7. These effects are broad across vertebrate
turkeys), and followoraladministrationoftheagents,eitherinfeed or
of antibacterial agents (including macrolides, tetracyclines, penicillins
The vertebrate GI tract contains an exceptionally complex and
dense microbial environment, with bacterial constituents that affect
the immune responses of populations of reactive host cells8and
stimulate a rich matrix of effecter mechanisms involved in innate
and adaptive immune responses9. The GI tract also is a locus of
(such as insulin, glucagon, leptin and ghrelin) and growth (for
example, glucose-dependent insulinotropic polypeptide (GIP) and
glucagon-like peptide 1 (GLP-1))10. Alterations in the populations
of the GI microbiota may change the intra-community metabolic
interactions11, modify caloric intake by using carbohydrates such as
cellulose that are otherwise indigestible by the host12, and globally
affect host metabolic, hormonal and immune homeostasis13. Full
(therapeutic) dose antibiotic treatments alter both the composition
of the gastrointestinal microbiota14and host responses to specific
microbial signals15. In combination with dietary changes, antibiotic
administration has been associated with changes in the population
structure of the microbiome. However, the effects of exposure to
subtherapeutic antibiotic dosages have not been described.
Early studies of the effects of gut microbiota on metabolism were
limited by the use of culture-based technologies that interrogated
,5% of the extant GI tract microbes16. Culture-independent invest-
igation of small-subunit ribosomal RNA sequences allows the micro-
bial population structure17of the gut microbiota to be characterized
with greater resolution. Despite inter-individual differences, substan-
tial similarities exist18among mammalian species in the GI micro-
basis for the conserved responses to early-life subtherapeutic anti-
biotic treatment (STAT) within farms. Previous work has shown that
provided frommodern agricultural practices tosuggest an alternative
approach, using a murine model of STAT to explore how antibiotic
exposure modulates host metabolic phenotypes.
Early-life STATincreases adiposity
We exposed C57BL/6J mice at weaning to penicillin, vancomycin,
penicillin plusvancomycin, chlortetracycline,or noantibioticintheir
drinking water at levels in the mid-range of US Food and Drug
Administration (FDA)-approved levels for subtherapeutic antibiotic
use in agriculture6,7. After a 7week exposure, the observed weights
were within the expected range of growth for female C57BL/6J mice,
and there was no significant difference in overall growth between the
STAT and control mice (Fig. 1a). However, by dual energy X-ray
absorptiometry (DEXA) scanning, (Fig. 1b) total fat mass was signifi-
cantly higher in all four groups of STAT mice than in the control
group (Fig. 1c). Per cent body fat also was increased in most
STAT groups compared to controls (Fig. 1d). Lean weight was not
significantly (P50.24) different in the STAT mice (15.060.1g
(mean6standard error)) compared to controls (15.460.3g)
1Department of Medicine, New York University School of Medicine, New York, New York 10016, USA.2Medical Service, VA New York Harbor Healthcare System, New York, New York 10010, USA.
3Department of Microbiology, New York University School of Medicine, New York, New York 10016, USA.4J. Craig Venter Institute, Rockville, Maryland 20850, USA.5Department of Pathology, New York
Population Health, New York University School of Medicine, New York, New York 10016, USA.
3 0 A U G U S T 2 0 1 2 | V O L 4 8 8 | N A T U R E | 6 2 1
Macmillan Publishers Limited. All rights reserved
(Fig. 1e). Thus, 7 weeks after starting the intervention, each of several
STAT exposures changed body composition but not overall weight.
Repeat STAT experiments showed similar results (Supplementary
Fig. 2). There were no significant differences in calculated feed effi-
the STAT and control mice. In a larger, confirmatory experiment to
assess when the morphometric changes appear, we began examining
the mice immediately at weaning. Both the STAT males and females
showed significantly increased earlylife growthrates (Supplementary
Fig. 3A), and fat mass in the STAT animals began to diverge from
weeks, there were significantly increased rates of fat accumulation in
bothfemale and maleSTAT mice (0.042 and0.045gweek21,respect-
ively) compared to controls; female mice also showed significantly
increased total mass (Supplementary Fig. 3B). These studies confirm
the increased adiposity associated with STAT in females, show
parallel effects in males, and indicate that the morphometric changes
begin at the earliest time (day 22) of measurement.
Bone mineral density is increased in early-life growth
Bone mineral density (BMD) was evaluated by DEXA scanning in the
BMDs in each of the five STAT groups (and overall) was significantly
increased compared to controls (Fig. 2a). By 7weeks, the BMDs in all
mice increased, without significant differences between the STAT
(0.04560.002 mgcm22) and control mice (0.04660.003mgcm22)
(Fig. 2a). Thus, an early bone developmental phenotype observed in
each of the STAT groups normalized by7weeks.Parallelobservations
have been made in other STAT experiments (data not shown).
Increased GIP in STAT mice
To examine metabolic correlates to the changes in body composition,
we assessed GIP, an incretin synthesized by small intestinal K cells20
with receptors located on adipocytes that stimulates lipoprotein
lipase activity21. GIP was significantly elevated in the STAT mice
(39.162.5pgml21) compared to controls (24.464.2pgml21), with
the levels ranging from 34.366.1pgml21in the penicillin group to
in IRS-1/GIPR and Kir6.2/GIPR double knockout mice that show
in STAT mice provides a mechanism for the observed adiposity
increase24,25, but also could be secondary to the metabolic changes.
There were no significant differences for fasting insulin-like growth
factor (IGF)-I, insulin, peptide YY, leptin, or ghrelin levels between
performed during week 6 of the experiment showed a trend towards
hyperglycaemia in STAT mice (Supplementary Fig. 5).
STAT does not alter overall gut microbial census
To determine whether the STAT exposure leading to these metabolic
changes affected the GI tract microbiome, microbial DNA extracted
from faecal and caecal samples collected from the mice during the
week of euthanasia or at necropsy, respectively, were studied. DNA
concentrations measured from both caecal (77.62633.51ngml21)
and faecal (23.79614.41ngml21) samples were not significantly dif-
ferent between control (n510) and STAT mice (n510 per group).
itative PCR using 338F/518R universal primers (Supplementary
Table 1), showed no significant differences in bacterial counts or
fungal census among the STAT and control groups (Fig. 3a). These
in the overall microbial census, and next led to us to conduct an
assessment of the composition of the populations.
STATalters the composition of intestinal microbiota
To assess microbial populations in the STAT and control micro-
biomes, we analysed the relative distribution of taxonomic groups
based on 16S rRNA v3 region sequence data. The extracted DNA
was subjected to 454 pyrosequencing, yielding 555,233 readable
sequences (5,7846676 sequences per sample with mean length
18863bp). The sequences were analysed at multiple (phylum to
genus) taxonomic levels (Supplementary Fig. 6 and Supplementary
Table 2). In both faecal and caecal samples, the ratio of Firmicutes to
Bacteria was significantly elevated in the STAT mice compared to
controls (Fig. 3b and Supplementary Fig. 7). Weighted Unifrac ana-
lysis of the dominant taxa (present in .1% of the total population)
(Fig. 3c). Importantly, deep branching was identified, with the mean
weightof miceonthe twomajor branchpointsontheheatmapbeing
Bone mineral density (g cm–2)
3 weeks7 weeks
GIP (pg ml–1)
Figure 2 | Bone development and serum GIP measurements. a, After
3weeks of STAT, bone mineral density was significantly increased in each
group (n510 mice per group) compared to controls (*P,0.05) but did not
persist at 7 weeks. b, Serum GIP levels measured at death were significantly
groups and in the aggregate antibiotic group compared to controls (P,0.05).
Data are presented as mean6s.e.m. Box plots show median 6 interquartile
range (IQR) and 95% ranges (whiskers).
0 102030 40
change from baseline
Pen. + Vanc.
Fat mass (g)
Body fat (%)
Lean mass (g)
Figure 1 | Weight and body composition of control and STAT mice.
and STAT (32.0%; bottom) mice. c, Total fat mass was significantly increased
(*P,0.05) in all STAT groups compared to controls. d, Per cent body fat was
significantly increased in all STAT groups (all P,0.05) except vancomycin.
e, Lean mass was lower in STAT mice, but not significantly different from
controls.Data are presentedas mean6s.e.m. For all figures: all, all antibiotics;
C, controls; Ct, chlortetracycline; P, penicillin; P1V, penicillinplus
vancomycin; V, vancomycin.
6 2 2 | N A T U R E | V O L 4 8 8 | 3 0 A U G U S T 2 0 1 2
Macmillan Publishers Limited. All rights reserved
significantly (P,0.05) different (21.463.1% (point A) versus
23.062.8% (point B); Fig. 3c). A major contributor to the observed
differences is increased Lachnospiraceae representation in the STAT
mice. Minor taxonomic groups may have roles in the development of
metabolic phenotypes, and STAT-associated increases in several
minor taxa are consistent with this possibility (Supplementary
plots, PCoA representations, and heat maps generated using non-
the STAT-exposed groups. Although STAT did not change the overall
bacterial census, even the minimal antibiotic doses caused shifts in
an increase in the relative concentrations of Firmicutes compared to
Bacteroidetes in the STAT mice compared to controls, which accom-
panied the observed increases in adiposity. This observation extends
previous findings26,27of relative increases in the Firmicute population
in ob/ob mice that are genetically prone to obesity. However, observa-
tions at such high taxonomic strata may not sufficiently describe the
changes associated with obesity28,29; variations in DNA extraction effi-
communities may bias census estimates of specific bacterial taxa.
Furthermore, although the overall phenotypes (increased adiposity
and hepatic lipogenesis) are consistent in the STAT groups, the inter-
mediate steps may be more host- and treatment-specific. Our findings
indicate that specific STAT exposures can be used as probes of micro-
biome structure and function.
STAT exposure alters gut microbiome SCFA metabolism
Because of the central role of short-chain fatty acid (SCFA) synthesis
in colonic metabolism30,31, we examined the effect of STAT exposure
on the gene counts of prokaryotic genes butyryl coA transferase
(BCoAT) and formyltetrahydrofolate synthetase (FTHFS) that are
involved in butyrate and acetate synthesis, respectively (Supplemen-
tary Fig. 12). Quantitative PCRs (qPCRs) for total bacteria, and
degenerate qPCRs for BCoAT and FTHFS, were performed on caecal
specimens in control and STAT mice. At 3weeks, there were signifi-
cant decreases in BCoAT gene copy numbers in the penicillin plus
vancomycin, chlortetracycline, and aggregate groups. By 6weeks,
BCoAT copy numbers had increased in all the groups compared to
FTHFS overall, there were no significant differences between control
and STAT mice overall at 3 or 6 weeks, although there was variation
within the antibiotic groups (Fig. 4a). Exploring the inter-antibiotic
differences further, we noted that there were several patterns in the
FTHFS qPCR with different melting curve peaks (Supplementary
Fig. 13), indicating differences in the microbial population. In total,
these results provide evidence that STAT treatment is dynamically
affecting composition of genes related to SCFAs, probably in
antibiotic-specific ways, but with overall conserved effects. Both
BCoAT and FTHFS have a role in metabolism of carbohydrates into
SCFAs32andhave been usedto assess the functional characteristics of
complex communities33; the observed changes in gene copy numbers
time points provide evidence that the STAT colonic microbiome
alters SCFA metabolism. Our finding that the copy numbers of these
Normalized log 16S copies
Normalized log ITS copies
Figure 3 | Changes in the faecal gut microbiome after 50days of STAT.
a, There were no significant differences in microbial census between theSTAT
and control groups (n510 mice per group) evaluated by qPCR with universal
primers for 16S rRNA and internal transcribed spacer (ITS). b, By 454-
pyrosequencing, Firmicutes were shown to be increased in the STAT mice at
multiple taxonomic levels. (Controls n510, penicillin n59, vancomycin
abundance of bacteria present at .1% at the family taxonomic level.
Hierarchical clustering based on Euclidean distance identified nonrandom
branch distributions of control and STAT mice (P,0.05). Lachno.,
Lachnospiraceae; Porphyr., Porphyromonadaceae.
3 0 A U G U S T 2 0 1 2 | V O L 4 8 8 | N A T U R E | 6 2 3
Macmillan Publishers Limited. All rights reserved
two genes increase during growth (between 3 and 6 weeks) in both
control and STAT mice (Fig. 4) suggests a greater dependence on
SCFA synthesis pathways with maturation. Relative differences in
the extentof these changes andin the meltingcurve peak populations
number (Fig. 4) in conjunction with melting curve peak patterns
(Supplementary Fig. 13) may be useful for studying community
genotypes of metabolic potential.
Direct measurements of SCFAs in the caecal contents of control
and STAT mice demonstrate substantial increases in acetate, pro-
pionate and butyrate in all STAT groups (Fig. 4b); ratios of butyrate
to acetate are also significantly altered by STAT exposure (Fig. 4c).
These findings provide evidence that STAT exposure perturbs not
only the composition of the GI microbiome but also the metabolic
capabilities of the microbiome, specifically with respect to SCFAs.
Increased SCFA concentrations and butyrate/acetate levels provide
mechanisms for the STAT-induced adiposity phenotypes. SCFAs
directly provide energy to colonocytes, and absorption into the portal
circulation stimulates adipogenesis30,34. Metabolic cage experiments
intake between control and STAT mice but lower caloric output in
faecal pellets from STAT mice (Supplementary Fig. 14), providing
evidence for selection for microbiota that can extract calories from
otherwise indigestible constituents.
STATalters hepatic metabolism of fatty acids and lipids
In confirmatory experiments using the identical STAT penicillin
protocol, liver tissue was collected from both control and STAT mice.
Standard histological analysis and triglyceride measurements
showed no significant differences between control and STAT mice
(Supplementary Fig. 15a, b). There were no significant differences in
16S rRNA gene counts, as detected by qPCR, from the liver speci-
mens, providing evidence that bacterial translocation is not altered in
the STAT mice (Supplementary Fig. 15c). Microarray analyses sur-
veyed for differences in .45,000 genes, identifying 466 that by t-test
or post translational modification statistics (PTM) were significantly
up- or downregulated (397 significantly differed from controls by
both tests) in the STAT compared to control mice. Focusing on path-
ways related to fatty acid metabolism and lipid metabolic processes,
22 and 47 genes, respectively, were differentially expressed between
STAT and control mice (Fig. 5a). When specific genes were mapped
ways related to lipogenesis and triglyceride synthesis (Fig. 5b). The
changes in gene expression observed by microarray analyses were
extended by qPCR assays of the same genes (Supplementary Fig. 16).
These microarray and qPCR findings demonstrate substantial changes
in the regulation of hepatic lipid, cholesterol and triglyceride metabol-
ism that result from STAT-induced intraluminal intestinal changes.
Increased STATadiposity is not metabolically altered
In the same confirmatory experiments (Supplementary Fig. 3),
visceral adipose tissue dissected from control and STAT (penicillin)
mice had no significant differences in adipocyte counts (Supplemen-
phages (Supplementary Fig. 17b, d). These findings were extended by
Gapdh-normalized (Supplementary Fig. 17e) metabolic gene qPCR
analyses. There were no significant differences in leptin, adiponectin,
also called SREBP1c), peroxisome proliferator activated receptor c
(Pparg, also called PPARc2), and fatty acid synthase (Fasn, also called
FAS) levels between the control and STAT groups (Supplementary
Fig. 17f). These findings provide evidence that STAT adipose tissue
shows no substantial physiological difference compared to controls in
by quantitative PCR. Increased adiposity seems to be a downstream
phenomenon primarily mediated by changes in the gut and liver.
By developing a model to assess adiposity, we show that each of the
several STAT approaches tested affects the adiposity of post-weaning
C57BL/6J mice. Similarly, there was a consistent early change in bone
development. Particularly in the dynamic phases of growth in young
animals, STAT alterations of the microbiome may affect pluripotent
STAT model does not precisely replicate the weight gain observed in
effects of exposure demonstrated in this model provide evidence that
altering the microbiome may have substantial consequences. Such
changes in early-life body composition may be due to altered host
responses35,36and/or shifts in the metabolic characteristics of the gut
proportion of calories from dietary complex carbohydrates that were
relativelyindigestible inthecontrolmice. Theincreased SCFAconcen-
trations are the metabolic products of this activity, which then may be
delivered in increased quantities through the portal circulation to the
caloric absorption has been implicated as a mechanism for increased
weight gaininother murineobesitymodels12.The observed increasein
3 weeks6 weeks3 weeks 6 weeks
Figure 4 | Caecal SCFA production after STAT exposure. a, Quantitative
PCR was performed for butyryl CoA transferase (BCoAT) and
formyltetrahydrofolate synthetase (FTHFS) at experiment weeks 3 and 6 on
STAT and control groups (n510 mice per group). At 3weeks, BCoAT was
diminished in two STAT groups and the aggregate group, a difference that
persisted only in the chlortetracycline group. FTHFS copies do not show a
consistent pattern. *P,0.05, **P,0.01, ***P,0.001 comparing STAT to
controls; hash symbol indicates significant difference between 3 and 6 weeks.
b, SCFAconcentration analysed by gaschromatography (GC) shows increases
in SCFAs in each of the STAT groups compared to controls. c, The ratio of
butyrate relative to acetate is significantly higher in the STAT mice than
controls. Data are presented as mean6s.e.m.
6 2 4 | N A T U R E | V O L 4 8 8 | 3 0 A U G U S T 2 0 1 2
Macmillan Publishers Limited. All rights reserved
different antibiotics used. This is consistent with the observation that a
wide range of FDA-approved antibacterial agents are used for effective
growth promotion in the agricultural industry6,7,37. Our studies of the
intestine, liver and adipose tissue demonstrate that the active effect of
the antibiotics is on the microbiota affecting the downstream liver;
mechanistically, the adipose tissue is passively accepting the increased
lipid load produced from more proximal activities.
Why might STAT exposures have such substantial effects on
recipient animals? We propose that STAT represents a compounded
perturbation, as described previously38, that has more serious con-
sequences for long-term alterations of community state, possibly
in SCFA production is a change in substrate availability that is a
characteristic39of ecological disturbance. Such overharvesting, at a
developmentally sensitive time, may push the ecosystem beyond the
normative recovery that usually follows infrequent disturbances38. In
complex, co-evolved ecosystems, the intricate interactions among parti-
cipants conform to equilibria40that promote the robustness of a com-
munity41. However, when such equilibria are substantially perturbed,
isms for the changes in host phenotype could reflect that maturation of
between the microbiome’s metabolic capabilities and numerous host
reactive cells could account for the observed changes, including the
increase in GIP. In our study, we confirm in a tractable experimental
model the decades-long observations in farm animals that STAT expo-
useful for studying the metabolic effects of microbiome manipulation.
Our study also indicates the possibility that modulation of the infant
human gut microbiome by antibiotics could have long-term metabolic
consequences affecting adiposity and bone development.
Female C57BL/6J mice were given penicillin, vancomycin, penicillin plus
vancomycin,or chlortetracycline (1mg antibiotic per g bodyweight)via drinking
water, or no antibiotics (control). Body weight was serially measured and body
composition determined using dual energy X-ray absorptiometry (DEXA). At
death, blood, caecal contents,liver and visceral adipose tissue were collected, and
serum hormones measured. DNA was extracted from caecal contents and faecal
pellets, and 16S rRNA gene v3 regions were barcoded and sequenced, using 454-
FLX Titanium chemistry. Quality-filtered sequences were processed through the
QIIME pipeline and analysed in the R statistical environment. Quantitative PCR
assessed taxa and metabolic genes of interest, and expression profiling of hepatic
RNA was performed by microarray.
Full Methods and any associated references are available in the online version of
Received 1 April 2011; accepted 6 July 2012.
Published online 22 August 2012.
1.McCaig, L. F. & Hughes, J. M. Trends in antimicrobial drug prescribing among
office-based physicians in the United States. J. Am. Med. Assoc. 273, 214–219
vate carboxy kinase 1
Fatty acid metabolism
Lipid metabolic process
Figure 5 | Differentially regulated genes related to hepatic lipogenesis,
were significantly up- or downregulated. Heat maps generated by gene set
enrichment analysis (GSEA)43identify differences between the STAT and
control mice (n56 mice per group), including pathways related to fatty acid
metabolism and lipid metabolic processes. b, Mapping of metabolic genes
detected by microarray onto specific pathways, including those related to
lipogenesis and triglyceride synthesis, show consistent changes with STAT.
Data are presented as mean6s.e.m. White bars, controls; black bars, STAT;
*P,0.05, **P,0.01. TG, triglycerides; VLDL, very low-density lipoproteins.
3 0 A U G U S T 2 0 1 2 | V O L 4 8 8 | N A T U R E | 6 2 5
Macmillan Publishers Limited. All rights reserved
2. Kozyrskyj, A. L., Ernst, P. & Becker, A. B. Increased risk of childhood asthma from
antibiotic use in early life. Chest 131, 1753–1759 (2007).
Blaser, M. J. & Falkow, S. What are the consequences of the disappearing human
microbiota? Nature Rev. Microbiol. 7, 887–894 (2009).
Dethlefsen, L. & Relman, D. A. Microbes and Health Sackler Colloquium:
Incomplete recovery and individualized responses of the human distal gut
microbiota to repeated antibiotic perturbation. Proc. Natl Acad. Sci. USA 108,
Manichanh, C. et al.Reshaping the gut microbiome with bacterial transplantation
and antibiotic intake. Genome Res. 20, 1411–1419 (2010).
Butaye,P., Devriese,L.A.& Haesebrouck, F.Antimicrobial growth promoters used
in animal feed: effects of less well known antibiotics on gram-positive bacteria.
Clin. Microbiol. Rev. 16, 175–188 (2003).
Ozawa, E. Studies on growth promotion by antibiotics. J. Antibiot. 8, 205–214
Abreu, M. T., Fukata, M. & Arditi, M. TLR signaling in the gut in health and disease.
J. Immunol. 174, 4453–4460 (2005).
Qin, J. et al. A human gut microbial gene catalogue established by metagenomic
sequencing. Nature 464, 59–65 (2010).
10. Hansotia, T. & Drucker, D. J. GIP and GLP-1 as incretin hormones: lessons from
single and double incretin receptor knockout mice. Regul. Pept. 128, 125–134
11. Gesta,S.,Tseng,Y.H.& Kahn,C.R.Developmentaloriginoffat:tracking obesity to
its source. Cell 131, 242–256 (2007).
12. Turnbaugh, P. J. et al. An obesity-associated gut microbiome with increased
capacity for energy harvest. Nature 444, 1027–1031 (2006).
13. Reikvam, D. H. et al. Depletion of murine intestinal microbiota: effects on gut
mucosa and epithelial gene expression. PLoS ONE 6, e17996 (2011).
14. Robinson, C. J. & Young, V. B. Antibiotic administration alters the community
structure of the gastrointestinal micobiota. Gut Microbes 1, 279–284 (2010).
15. Wlodarska, M. et al. Antibiotic treatment alters the colonic mucus layer and
predisposes the host to exacerbated Citrobacter rodentium-induced colitis. Infect.
Immun. 79, 1536–1545 (2011).
16. McCracken, V.J.,Simpson, J.M.,Mackie, R.I.& Gaskins,H. R.Molecularecological
analysis of dietary and antibiotic-induced alterations of the mouse intestinal
microbiota. J. Nutr. 131, 1862–1870 (2001).
18. Spor, A., Koren, O. & Ley, R. Unravelling the effects of the environment and host
genotype on the gut microbiome. Nature Rev. Microbiol. 9, 279–290 (2011).
19. Ley, R. E. et al. Obesity alters gut microbial ecology. Proc. Natl Acad. Sci. USA 102,
20. Buffa, R. et al. Identification of the intestinal cell storing gastric inhibitory peptide.
Histochemistry 43, 249–255 (1975).
21. Miyawaki, K. et al. Inhibition of gastric inhibitory polypeptide signaling prevents
obesity. Nature Med. 8, 738–742 (2002).
in KATPnull mice. Eur. J. Endocrinol. 151, 407–412 (2004).
under diminished insulin action. Biochem. Biophys. Res. Commun. 335, 937–942
24. Yip, R. G., Boylan, M. O., Kieffer, T. J. & Wolfe, M. M. Functional GIP receptors are
present on adipocytes. Endocrinology 139, 4004–4007 (1998).
25. Yamada, Y. & Seino, Y. Physiology of GIP—a lesson from GIP receptor knockout
mice. Horm. Metab. Res. 36, 771–774 (2004).
26. Ley, R. E. et al. Obesity alters gut microbial ecology. Proc. Natl Acad. Sci. USA 102,
tomarkedbut reversiblealterations inthemousedistalgutmicrobiome.CellHost
Microbe 3, 213–223 (2008).
28. Murphy, E. F. et al. Composition and energy harvesting capacity of the gut
microbiota: relationship to diet, obesity and time in mouse models. Gut 59,
29. Fleissner, C. K. et al. Absence of intestinal microbiota does not protect mice from
diet-induced obesity. Br. J. Nutr. 104, 919–929 (2010).
30. Wong, J. M., de Souza, R., Kendall, C. W., Emam, A. & Jenkins, D. J. Colonic health:
fermentation and short chain fatty acids. J. Clin. Gastroenterol. 40, 235–243
31. Hong, Y. H. et al. Acetate and propionate short chain fatty acids stimulate
adipogenesis via GPCR43. Endocrinology 146, 5092–5099 (2005).
32. Lovell, C. R. & Leaphart, A. B. Community-level analysis: key genes of CO2-
reductive acetogenesis. Methods Enzymol. 397, 454–469 (2005).
33. Henderson, G., Naylor, G. E., Leahy, S. C. & Janssen, P. H. Presence of novel,
potentially homoacetogenic bacteria in the rumen as determined by analysis of
formyltetrahydrofolate synthetase sequences from ruminants. Appl. Environ.
Microbiol. 76, 2058–2066 (2010).
tract in various species. Physiol. Rev. 70, 567–590 (1990).
35. Backhed, F. et al.The gut microbiota as an environmental factor that regulates fat
storage. Proc. Natl Acad. Sci. USA 101, 15718–15723 (2004).
36. Backhed, F., Ley, R. E., Sonnenburg, J. L., Peterson, D. A. & Gordon, J. I. Host-
bacterial mutualism in the human intestine. Science 307, 1915–1920 (2005).
37. Jukes, T. H. Antibiotics in animal feeds. N. Engl. J. Med. 282, 49–50 (1970).
38. Paine, R. T., Tegner, M. J. & Johnson, E. A. Compounded perturbations yield
ecological surprises. Ecosystems 1, 535–545 (1998).
39. Pickett, S. T. & White, P. S. The Ecology of Natural Disturbance and Patch Dynamics
40. Blaser, M. J. & Kirschner, D. The equilibria that allow bacterial persistence in
human hosts. Nature 449, 843–849 (2007).
Soc. Lond. B 268, 2039–2045 (2001).
42. Cho, I. & Blaser, M. J. The human microbiome: at the interface of health and
disease. Nature Rev. Genet. 13, 260–270 (2012).
43. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based
approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci.
USA 102, 15545–15550 (2005).
Supplementary Information is available in the online version of the paper.
Acknowledgements This work was supported in part with grants from the NIH
(T-RO1-DK090989, 1UL1-RR029893, UL1-TR000038), the Diane Belfer Program in
Human Microbial Ecology, the Philip and Janice Levin Foundation, the Michael
SapersteinFellowship,andinstitutional funds providedbythe J.CraigVenterInstitute,
K.R. performed experiments; B.A.M. and K.L. performed sequencing and sequencing
analysis; J.Z. performed microarray analyses; I.C. and H.L. performed statistical
interpretation and analyses; A.V.A. performed bioinformatics analyses and
interpretation; I.C. and M.J.B. took primary responsibility for writing the manuscript.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
and requests for materials should be addressed to M.J.B.
6 2 6 | N A T U R E | V O L 4 8 8 | 3 0 A U G U S T 2 0 1 2
Macmillan Publishers Limited. All rights reserved
Mouse husbandry. Female C57BL/6J mice were obtained at weaning (21 days of
1week. Mice were weighed at the beginning of each experiment and distributed
five per cage so that mean weights in each cage were equal. In each experiment,
each study group (control or antibiotic(s)) was composed of ten mice. The mice
were allowed ad libitum access to food and water and were maintained on a 12-h
no. 5001). The protocol was approved by the New York University School of
Medicine Institutional Animal Care and Use Committee (IACUC).
Antibiotic dosing. Beginning on day 28 of life, mice were given standard water
(pH6.8) or water containing one of the following antibiotic regimens: penicillin
VK, vancomycin, penicillin VK plus vancomycin, and chlortetracycline. Doses
were adapted from the FDA Green Book and generally were in the mid-range of
those approved for use in agriculture6,7,37. To simplify administration, the
antibiotics were added to drinking water at a dose of 1mg antibiotic per g body
weight of the mice in the cage, based on the calculation that daily water intake is
6J mice averaged 15ml water per 100g body weight44. Water containers were
changed twice weekly to supply fresh antibiotics.
Mouse measurements and phenotypes. Three times a week, each mouse was
offeed consumedpercage wasmeasuredandaveragedailyfeedintake calculated.
Feed efficiency was calculated by dividing mean weight gain by mean daily feed
consumed per mouse. Faecal pellets were collected every week and stored at
220uC until processing. Body composition was determined using dual energy
X-ray absorptiometry(DEXA)with a Lunar PIXImusIImouse densitometer (GE
collectedon fatcomposition,lean body mass, percent body fat, andbone mineral
density. MRI experiments were performed on a 7tesla MRI system gated on
the scan by directing a thermostatically controlled warm air source. Images
captured were separated into fat and lean tissue using the IDEAL Dixon method
based on chemical shift properties45and total body fat percentage (volume/
volume) was calculated by merging fat and lean images in MRIcron software
dated by comparisons with scale weight and fat% values with MRI-determined
fat% (Supplementary Fig. 1). Mice were killed by CO2inhalation and cervical
into serum and whole cells. Necropsy was performed on each mouse and caecal
contents, liver, serum and visceral adipose tissue were collected. All samples were
snap-frozen and stored at 280uC until processing.
Statistical comparison of growth rates. From the experiment shown in
Supplementary Fig. 4, the plots of group means of total, fat and lean mass over
time were fitted with the following linear spline model with a common knot at
where group50 indicates the control group and group51 indicates the STAT
group, and (x)1is defined as a function that equals x when x is positive and is
baseline, before week 26 and to examine patterns of change over time through
tests H0: b150, H0: b450 and H0: b45b550, respectively. The MIXED pro-
tests and calculate the estimates.
Hormone measurements. Blood cells and serum were separated through
centrifugation (3,000g for 10min at 4uC) and the serum frozen at 270uC.
Serum specimens (100ml) were examined using the Millipore Mouse Gut
Hormone Panel (Millipore Corp.) using a Luminex 200 (Millipore) analyser.
The panel was custom-designed to assess for glucose-dependent insulinotropic
peptide (GIP), insulin, IGF-1, peptide YY, ghrelin and leptin.
The mice then were challenged with 2mg glucose per g body weight, given
intraperitoneally, as previously described46, and blood glucose levels determined
30, 60 and 120min later. Area under curve (AUC) calculations were used to
compare the STAT and control groups between 30 and 60min.
DNA extraction. DNA extractions from caecal contents and faecal pellets were
performed using the Qiagen QIAmp DNA stool extraction kit (Qiagen) per the
manufacturer’s protocol. Total extracted DNA was quantified using a Nanodrop
1000 (Thermo Scientific). PCR to confirm bacterial DNA extractions was per-
formed using the 8F/1510R bacterial primers for 16S rRNA47,48.
Quantitative PCR amplification. After genomic DNA extraction and quan-
tification, samples were prepared for amplification and sequencing at the JCVI
Joint Technology Center (JTC). Genomic DNA sample concentrations were
mer 534R (59-ATTACCGCGGCTGCTGG-39) to which a ‘CG’ on the 59 end of
of the primer design (59- CGN(10)116S primer-39). Barcoded primer designs
were completed using a set of algorithms developed at the JCVI. This design
allowed for the inclusion of a unique barcode to each sample at the time of
PCR so that the tagged samples could be multiplexed for sequencing. A total of
96 barcodes were used. Every effort was made to prevent contamination of PCR
reactions with exogenous DNA including a set of reactions in a laminar flow
hood. PCR reactions were completed as follows (per reaction): 1ml of gDNA,13
final concentration of Accuprime PCR Buffer II (Invitrogen), 250mM betaine,
200nM forward and reverse primers, 0.5units of Accuprime Taq DNA
Polymerase High Fidelity (Invitrogen), and nuclease-free water to bring the final
volume to 10ml. PCR cycling conditions were: initial denaturation of 2min at
95uC followed by 25cycles of 10s at 95uC, 20s at 63uC and 30s at 72uC. A
negative control (water blank) reaction also was included and examined after 30
cycles. PCR reactions were visualized on 1% agarose gels and quantified using a
TecanSpectraFluorPlus (TecanGroupLtd)before normalizationand poolingof
samples for sequencing.
454-FLX read quality filtering. After the completion of sequencing, a read
processing pipeline involving a set of modular scripts designed at the JCVI was
used for deconvolution, trimming and quality filtering49. First reads were decon-
voluted, or assigned to samples, based on their unique 10-nucleotide barcode,
allowing#1-nucleotide mismatch to the reference barcode sequence and#6
primer sequence mismatches, using the EMBOSS program fuzznuc to locate
barcodes and primer sequences on each read. To eliminate spurious hits, the
distance between detected barcode and primer were required to be#3nucleo-
tides. Barcodes then were trimmed from the sequence and remaining sequences
which failed to have barcodes detected then were filtered. After deconvolution,
primer sequences detected at .75% identity were trimmed and those without
valid primers detected were eliminated. Reads with an average length of
,100nucleotides or with any Ns were removed. A BLASTN quality check50
reads substantially inconsistent with 16S gene sequences. The criteria were that
.30% of the query must be covered by the alignment with a minimum of
60nucleotides of identity. Passing reads were subsequently examined using a
modified version of the RDP Chimera Check in which candidate chimaeras were
identified based only on reads that were sequenced within the sample. Final
sequence data have been deposited in the NIH Sequence Read Archive (http://
www.ncbi.nlm.nih.gov/sra). Accession numbers for all primary sequencing data
are available from the NCBI under BioProject 168618.
system (Beckman Coulter Genomics). The A and B adapters necessary for
construction, emPCR, enrichment, and 454 sequencing were performed by fol-
lowing the vendor’s standard operating procedures with some modifications.
Specifically, quantitative PCR was used to accurately estimate the number of
molecules needed for emPCR. We also used automation (BioMek FX) to ‘break’
during the breaking process.
Supplementary Methods) were pre-processed through the QIIME pipeline51
involving: (1) clustering the sequences into operational taxonomical units
(OTUs) using the UCLUST program52at a 97% similarity threshold; (2) taxo-
nomically assigning each OTU by selecting a representative sequence from each
OTU-cluster and running RDP Classifier53with 80% bootstrap confidence; (3)
aligning representative sequences for each OTU with the Greengenes core-set
alignment template using PyNAST54; (4) building a phylogenetic tree for the
OTUs using the FASTTREE program55; and (5) calculating weighted Unifrac
beta-diversity indices56. We extracted the OTU absolute abundance table and
weighted Unifrac beta-diversity matrix56,57from the pipeline for further analysis
in the R statistical programming environment58,59. The rarefactions for diversity
indices and species richness were calculated in the R statistical programming
environment59,60using Community Ecology Package Vegan. The OTU absolute
abundances were converted to relative abundances by normalizing to total
Macmillan Publishers Limited. All rights reserved
sequence count per sample analysed. The resulting relative abundance matrix
was used to produce heat maps and major taxa bar plots. (Code available on the
NYU Center for Health Informatics and Bioinformatics website http://
www.nyuinformatics.org/research/labs/microbiomics.) All data are presented
as mean6s.e.m., unless otherwise indicated for groups without normal data
distribution (median6IQR). Comparisons of medians between non-normally
distributed groups were performed using the Mann–Whitney U-test or Kruskal–
Wallis analysis of variance test for simultaneous comparisons of more than two
groups. Multivariate analysis of taxonomic distributions among groups was per-
comparing beta-diversity did not affect statistical decisions (Supplementary
Tables 4 and 5). P values ,0.05 were considered to be significant.
Comparison of intra- and intergroup weighted Unifrac distances. Statistical
tion of a statistical test appropriate for the data. In our analysis, we apply a series
empirical distribution to normal. We formally supplement this visual diagnostic
tions is difficult in statistical analyses because normality can only be rejected and
never confirmed; thus, it usually is not done in typical analyses of this sort.
However, we performed the Shapiro–Wilk’s test for normality on inter- and
intragroup weighted UniFrac distances. Although in many cases normality could
distributions is approximately normally distributed. For final assurance that
decisions based on t-test are not affected by possible violations of the normality
assumption, we also performed Mann–Whitney tests which (after adjustment
for multiple comparison adjustment) provided the same rejections of the null
hypothesis as for the original t-tests (Supplementary Table 5).
Quantitative PCR. Quantitative PCR assays to assess for taxa of interest were
performed on a Rotor Gene 3000 quantitative PCR cycler using the LightCycler
FastStart DNA Master PLUS SYBR Green I kit (Roche) according to the
manufacturer’s instructions. Quantitative assays to assess total microbial census
were performed for Bacteria using 16S rRNA and Fungi using ITS sequences.
Quantitative assays were also performed for butyryl CoA transferase (BCoAT)
and formyltetrahydrofolate synthetase (FTHFS), using degenerate primers.
All primer sequences are provided in Supplementary Table 1. Values were
normalized based on total Bacteria 16S copies.
Hepatic triglyceride measurements. For triglyceride measurements, hepatic
tissue was homogenized at 4uC in RIPA lysis buffer (Sigma-Aldrich) and lipids
extracted using a chloroform/methanol (2:1) method, evaporated, and dissolved
in 2-propanol63. Triglyceride concentration was assayed using the enzymatic
tion by colorimetric assay (Sigma-Aldrich).
(Promega) and RNeasy Mini Kit (Qiagen), respectively, according to the manu-
facturer’s instructions. Total RNA was reverse transcribed to cDNA using the
Verso cDNA Synthesis Kit (Thermo Scientific). To generate standards for each
target gene expression analysis, the DNA or cDNA region of interest was PCR-
amplifiedand thePCRproduct wascloned usingthepGEM-Teasyvectorsystem
(Promega). qPCR was performed with a LightCycler 480 SYBR Green I Master
(Roche) and run in a LightCycler 480 system (Roche). Target mRNA was
normalized to GAPDH mRNA as an internal control in each sample.
Expression profiling of the STAT and control animal groups (n56 each) was
performed using the Affymetrix Genechip system (Affymetrix). Total RNA
quality and quantity were determined using the Agilent 2100 Bioanalyser and
Nanodrop ND-1000. Total RNA (100ng) was used to prepare cDNA following
the Affymetrix 39IVT Express Kit labelling protocol (Affymetrix). Standardized
array processing procedures recommended by Affymetrix were performed,
including hybridization, fluidics processing and scanning of the Affymetrix
MG-430 2.0 arrays. GeneSpring GX11 software (Agilent Technologies) was used
to normalize the raw data (Affymetrix CEL files) by Robust Multichip Average
algorithm (RMA). Gene set enrichment analysis (GSEA)43was used to identify
significantly enriched gene expression patterns underlying fatty acid and lipid
metabolism, by querying the C2 (curated pathways) and C5 (Gene Ontologies)
categories of the GSEA MolSig v.3 database. The microarray data have been
submitted in the MIAME-compliant format to the NCBI GEO public database
(http://www.ncbi.nlm.nih.gov/geo/) under the identifier GSE38880.
Histology and immunohistochemistry. Fat tissues of mice were collected and
frozen at 280uC before being thawed and embedded inoptimumcutting temper-
ature (OCT) medium. Sections were cut in 5mm thickness for staining with
haematoxylin and eosin. Immunohistochemistry was performed on frozen form-
alin-fixed tissues using rat anti-mouse CD68 (AbDSerotec). In brief, sections were
thawed at room temperature for 30min, fixed in 10% NBF for 15min, rinsed in
distilled water, and soaked in reaction buffer for 15min. Antibody incubation and
detection were carried out at 37uC on a NeXes instrument (Ventana Medical
tionkits unless otherwise noted. Endogenous peroxidase activitywas blockedwith
Laboratories) diluted 1:4,000 and incubated for 30min. After secondary antibody
application, streptavidin-horseradish-peroxidase conjugate wasapplied. The com-
plex was visualized with 3,3-diaminobenzidene and enhanced with copper sul-
phate. Slides were washed in distilled water, counterstained with haematoxylin,
dehydrated and mounted with permanent media. Appropriate positive and nega-
tive controls were included with the sections studied.
control, n510), 10 high-power fields (HPF) were selected by a single observer.
Copies ofeach ofthe imageswere providedto two independentindividuals blind
to the experimental design who counted adipocytes or CD681macrophages in
each HPF using ImageJ (http://rsbweb.nih.gov/ij/). The counts were compared
between the two investigators and the mean number of adipocytes or CD681
macrophages per HPF for control and STAT mice were compared using the
Gas chromatographic analysis. Mouse faecal pellets were collected at week 6 of
genized in 100ml of deionized water for 3min. The pH of the suspension was
adjusted to 2–3 by adding 5M HCl at room temperature for 10min with inter-
mittent shaking. The suspension was transferred into a polypropylene tube and
centrifuged for 20min at 3,000g, yielding a clear supernatant. The internal
standard, 2-ethylbutyric acid (TEBA), was added into the supernatant at a final
concentration of 1mM. Chromatographic analysis used the Shimadzu QP-500
GC/MS system (Shimadzu). A fused-silica capillary column (30m, 0.52mm,
0.50mm) with a free fatty acid phase (DB-FFAP 125-3237, J&W Scientific,
AgilentTechnologiesInc.)was used foranalysis.Heliumwas the carrierat a flow
rate of 14.4mlmin21. The initial oven temperature (100uC) was maintained for
30s, raised to 180uC at 8uC min21and held for 60s, then increased to 200uC at
20uCmin21andheldfor 5min.The flameionizationdetectorandinjectionport
were kept at 240 and 200uC, respectively. The flow rates of hydrogen, air, and
for GC analysis was 1ml, and each analysis had a run time of 32min64.
Metabolic cage measurements. C57BL/6J mice were singly housed in
Techniplast Type 304 metabolic cages (Techniplast). Mice had unrestricted
access to 45% HFD powdered chow and STAT or control water, as appropriate.
There was a 2-day acclimatization period, followed by 3days of measurements.
and urine produced. Total faeces and urine were collected for analysis. Calories
consumed were calculated by weight of food consumed and 4,057calg21as per
the manufacturer’s instructions (Research Diets). Calories excreted were calcu-
lated based on total faeces produced and calorimetric analysis of a homogenate
pellet incorporating all faeces produced per mouse from each 24-h period.
Semi-micro calorimeter. Faecal pellets obtained during week 6 of the experiment
were dehydrated overnight at 56uC with a silica gel desiccant. All pellets studied
had minimum dry weights .12mg. The calorimeter was calibrated using
in each day’s run. Corrections were made for sulphur content and fuse length;
final results were expressed as calories per gram dry weight.
edn (Lea and Febiger, 1989).
45. Reeder, S. B. et al. Iterative decomposition of water and fat with echo asymmetry
and least-squares estimation (IDEAL): application with fast spin-echo imaging.
Magn. Reson. Med. 54, 636–644 (2005).
46. Andrikopoulos, S., Blair, A. R., Deluca, N., Fam, B. C. & Proietto, J. Evaluating the
glucose tolerance test in mice. Am. J. Physiol. Endocrinol. Metab. 295,
47. Favier, C. F., Vaughan, E. E., De Vos, W. M. & Akkermans, A. D. L. Molecular
monitoring of succession of bacterial communities in human neonates. Appl.
Environ. Microbiol. 68, 219–226 (2002).
Macmillan Publishers Limited. All rights reserved
48. Martı ´nez-Murcia,A.J.,Acinas,S.G.&Rodriguez-Valera,F.Evaluationofprokaryotic
diversity byrestrictasedigestion of16S rDNA directly amplified fromhypersaline
environments. FEMS Microbiol. Ecol. 17, 247–255 (1995).
Notes 3, 199 (2010).
search tool. J. Mol. Biol. 215, 403–410 (1990).
51. Caporaso, J. G. et al. QIIME allows analysis of high-throughput community
sequencing data. Nature Methods 7, 335–336 (2010).
52. Edgar, R. C. Search and clustering orders of magnitude faster than BLAST.
Bioinformatics 26, 2460–2461 (2010).
53. Wang,Q., Garrity,G.M., Tiedje,J.M.& Cole, J.R.NaiveBayesian classifierforrapid
assignment of rRNA sequences into the new bacterial taxonomy. Appl. Environ.
Microbiol. 73, 5261–5267 (2007).
54. Caporaso, J. G. et al. PyNAST: a flexible tool for aligning sequences to a template
alignment. Bioinformatics 26, 266–267 (2010).
55. Price, M. N., Dehal, P. S. & Arkin, A. P. FastTree 2–approximately maximum-
likelihood trees for large alignments. PLoS ONE 5, e9490 (2010).
56. Lozupone, C., Lladser, M. E., Knights, D., Stombaugh, J. & Knight, R. UniFrac: an
effective distance metric for microbial community comparison. ISME J. 5,
57. Lozupone, C. & Knight, R. UniFrac: a new phylogenetic method for
comparing microbial communities. Appl. Environ. Microbiol. 71, 8228–8235
58. Gentleman, R. C. et al. Bioconductor: open software development for
computational biology and bioinformatics. Genome Biol. 5, R80 (2004).
59. Ihaka, R. & Gentleman, R. R: a language for data analysis and graphics. J. Comp.
Graph. 5, 299–314 (1996).
60. Gentleman, R. C. et al. Bioconductor: open software development for
computational biology and bioinformatics. Genome Biol. 5, R80 (2004).
genetic diversity: development and application to the description of bacterial
diversity. BMC Evol. Biol. 7, 156 (2007).
62. Pavoine, S., Dufour, A. B. & Chessel, D. From dissimilarities among species to
dissimilarities among communities: a double principal coordinate analysis.
J. Theor. Biol. 228, 523–537 (2004).
63. Hong, F. et al. Interleukin 6 alleviates hepatic steatosis and ischemia/reperfusion
injury in mice with fatty liver disease. Hepatology 40, 933–941 (2004).
in colonic contents and faeces of humans and rats by acidified water-extraction
and direct-injection gas chromatography. Biomed. Chromatogr. 20, 674–682
Macmillan Publishers Limited. All rights reserved