gut microbiota from twins discordant for obesity modulate ... › sites › ...doi:...

12
DOI: 10.1126/science.1241214 , (2013); 341 Science et al. Vanessa K. Ridaura Metabolism in Mice Gut Microbiota from Twins Discordant for Obesity Modulate This copy is for your personal, non-commercial use only. clicking here. colleagues, clients, or customers by , you can order high-quality copies for your If you wish to distribute this article to others here. following the guidelines can be obtained by Permission to republish or repurpose articles or portions of articles ): November 6, 2014 www.sciencemag.org (this information is current as of The following resources related to this article are available online at http://www.sciencemag.org/content/341/6150/1241214.full.html version of this article at: including high-resolution figures, can be found in the online Updated information and services, http://www.sciencemag.org/content/suppl/2013/09/04/341.6150.1241214.DC1.html can be found at: Supporting Online Material http://www.sciencemag.org/content/341/6150/1241214.full.html#related found at: can be related to this article A list of selected additional articles on the Science Web sites http://www.sciencemag.org/content/341/6150/1241214.full.html#ref-list-1 , 33 of which can be accessed free: cites 74 articles This article http://www.sciencemag.org/content/341/6150/1241214.full.html#related-urls 22 articles hosted by HighWire Press; see: cited by This article has been http://www.sciencemag.org/cgi/collection/microbio Microbiology subject collections: This article appears in the following registered trademark of AAAS. is a Science 2013 by the American Association for the Advancement of Science; all rights reserved. The title Copyright American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the Science on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from on November 6, 2014 www.sciencemag.org Downloaded from

Upload: others

Post on 30-Jun-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

DOI: 10.1126/science.1241214, (2013);341 Science

et al.Vanessa K. RidauraMetabolism in MiceGut Microbiota from Twins Discordant for Obesity Modulate

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

  here.following the guidelines

can be obtained byPermission to republish or repurpose articles or portions of articles

  ): November 6, 2014 www.sciencemag.org (this information is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/341/6150/1241214.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/suppl/2013/09/04/341.6150.1241214.DC1.html can be found at: Supporting Online Material

http://www.sciencemag.org/content/341/6150/1241214.full.html#relatedfound at:

can berelated to this article A list of selected additional articles on the Science Web sites

http://www.sciencemag.org/content/341/6150/1241214.full.html#ref-list-1, 33 of which can be accessed free:cites 74 articlesThis article

http://www.sciencemag.org/content/341/6150/1241214.full.html#related-urls22 articles hosted by HighWire Press; see:cited by This article has been

http://www.sciencemag.org/cgi/collection/microbioMicrobiology

subject collections:This article appears in the following

registered trademark of AAAS. is aScience2013 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Nov

embe

r 6,

201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Nov

embe

r 6,

201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Nov

embe

r 6,

201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Nov

embe

r 6,

201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Nov

embe

r 6,

201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Nov

embe

r 6,

201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Page 2: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

Gut Microbiota from Twins Discordant for Obesity Modulate Metabolism in MiceVanessa K. Ridaura, Jeremiah J. Faith, Federico E. Rey, Jiye Cheng, Alexis E. Duncan, Andrew L. Kau, Nicholas W. Griffi n, Vincent Lombard, Bernard Henrissat, James R. Bain, Michael J. Muehlbauer, Olga Ilkayeva, Clay F. Semenkovich, Katsuhiko Funai, David K. Hayashi, Barbara J. Lyle, Margaret C. Martini, Luke K. Ursell, Jose C. Clemente, William Van Treuren, William A. Walters, Rob Knight, Christopher B. Newgard, Andrew C. Heath, Jeffrey I. Gordon*

Introduction: Establishing whether specifi c structural and functional confi gurations of a human gut microbiota are causally related to a given physiologic or disease phenotype is challenging. Twins dis-cordant for obesity provide an opportunity to examine interrelations between obesity and its associ-ated metabolic disorders, diet, and the gut microbiota. Transplanting the intact uncultured or cultured human fecal microbiota from each member of a discordant twin pair into separate groups of recipient germ-free mice permits the donors’ communities to be replicated, differences between their properties to be identifi ed, the impact of these differences on body composition and metabolic phenotypes to be discerned, and the effects of diet-by-microbiota interactions to be analyzed. In addition, cohousing coprophagic mice harboring transplanted microbiota from discordant pairs provides an opportunity to determine which bacterial taxa invade the gut communities of cage mates, how invasion correlates with host phenotypes, and how invasion and microbial niche are affected by human diets.

Methods: Separate groups of germ-free mice were colonized with uncultured fecal microbiota from each member of four twin pairs discordant for obesity, or with culture collections from an obese (Ob) or lean (Ln) co-twin. Animals were fed a mouse chow low in fat and rich in plant polysaccharides (LF-HPP) or one of two diets refl ecting the upper or lower (Hi or Lo) tertiles of consumption of saturated fats (SF) and fruits and vegetables (FV) based on the U.S. National Health and Nutrition Examination Survey (NHANES). Ln or Ob mice were cohoused 5 days after colonization. Body composition changes were defi ned by quantitative magnetic resonance. Microbiota or microbiome structure, gene expression, and metabolism were assayed by 16S ribosomal RNA profi ling, whole-community shotgun sequencing, RNA-sequencing, and mass spectrometry. Host gene expression and metabolism were also characterized.

Results and Discussion: The intact uncultured and culturable bacterial component of Ob co-twins’ fecal microbiota conveyed signifi cantly greater increases in body mass and adiposity than those of Ln communities. Differences in body composition were correlated with differences in fermentation of short-chain fatty acids (increased in Ln), metabolism of branched-chain amino acids (increased in Ob), and microbial transformation of bile acid species (increased in Ln and correlated with down-regulation of host farnesoid X receptor signaling). Cohousing Ln and Ob mice prevented development of increased adiposity and body mass in Ob cage mates and transformed their microbiota’s metabolic profi le to a leanlike state. Transformation correlated with invasion of members of Bacteroidales from Ln into Ob microbiota. Invasion and phenotypic rescue were diet-dependent and occurred with the diet represent-ing the lower tertile of U.S. consumption of saturated fats and upper tertile of fruits and vegetables but not with the diet representing the upper tertile of saturated fats and lower tertile of fruit and vegetable consumption. These results reveal that transmissible and modifi able interactions between diet and microbiota infl uence host biology.

FIGURES IN THE FULL ARTICLE

Fig. 1. Reliable replication of human donor microbiota in gnotobiotic mice.

Fig. 2. Cohousing Obch and Lnch mice transforms the adiposity phenotype of cage mates harboring the obese co-twin’s culture collection to a leanlike state.

Fig. 3. Effect of cohousing on metabolic profi les in mice consuming the LF-HPP diet.

Fig. 4. Effects of NHANES-based LoSF-HiFV and HiSF-LoFV diets on bacterial invasion, body mass, and metabolic phenotypes.

Fig. 5. Invasion analysis of species-level taxa in Obch or Lnch mice fed the NHANES-based LoSF-HiFV diet.

Fig. 6. Acylcarnitine profi le in the skeletal muscle of mice colonized with the Ob or Ln culture collections from dizygotic twin pair 1 and fed the LoSF-HiFV diet.

SUPPLEMENTARY MATERIALS

Materials and MethodsSupplementary Text Figs. S1 to S17Tables S1 to S17References and Notes

RELATED ITEMS IN SCIENCE

A. W. Walker, J. Parkhill, Fighting obesity with bacteria. Science 341, 1069–1070 (2013). DOI: 10.1126/science.1243787

Cohousing Ln and Ob mice prevents adiposity phenotype in Ob cage mates (Obch). (A) The adi-posity change after 10 days of cohousing. *P < 0.05 versus Ob controls (Student’s t test). (B) Bac-teroidales from Lnch micro-biota invade the Obch microbiota. Columns show individual mice.

READ THE FULL ARTICLE ONLINEhttp://dx.doi.org/10.1126/science.1241214

Cite this article as V. K. Ridaura et al., Science 341, 1241214 (2013). DOI: 10.1126/science.1241214

The list of author affi liations is available in the full article online.*Corresponding author. E-mail: [email protected]

www.sciencemag.org SCIENCE VOL 341 6 SEPTEMBER 2013

RESEARCH ARTICLE SUMMARY

1079

Published by AAAS

Page 3: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

Gut Microbiota from TwinsDiscordant for Obesity ModulateMetabolism in MiceVanessa K. Ridaura,1 Jeremiah J. Faith,1 Federico E. Rey,1 Jiye Cheng,1 Alexis E. Duncan,2,3

Andrew L. Kau,1 Nicholas W. Griffin,1 Vincent Lombard,4 Bernard Henrissat,4,5

James R. Bain,6,7,8 Michael J. Muehlbauer,6 Olga Ilkayeva,6 Clay F. Semenkovich,9

Katsuhiko Funai,9 David K. Hayashi,10 Barbara J. Lyle,11 Margaret C. Martini,11

Luke K. Ursell,12 Jose C. Clemente,12 William Van Treuren,12 William A. Walters,13

Rob Knight,12,14,15 Christopher B. Newgard,6,7,8 Andrew C. Heath,2 Jeffrey I. Gordon1*

The role of specific gut microbes in shaping body composition remains unclear. We transplantedfecal microbiota from adult female twin pairs discordant for obesity into germ-free mice fedlow-fat mouse chow, as well as diets representing different levels of saturated fat and fruit andvegetable consumption typical of the U.S. diet. Increased total body and fat mass, as well asobesity-associated metabolic phenotypes, were transmissible with uncultured fecal communitiesand with their corresponding fecal bacterial culture collections. Cohousing mice harboring anobese twin’s microbiota (Ob) with mice containing the lean co-twin’s microbiota (Ln) prevented thedevelopment of increased body mass and obesity-associated metabolic phenotypes in Ob cagemates. Rescue correlated with invasion of specific members of Bacteroidetes from the Lnmicrobiota into Ob microbiota and was diet-dependent. These findings reveal transmissible,rapid, and modifiable effects of diet-by-microbiota interactions.

Microbial community configurationsvary substantially between unrelated in-dividuals (1–9), which creates a chal-

lenge in designing surveys of sufficient power todetermine whether observed differences betweendisease-associated and healthy communities dif-fer significantly from normal interpersonal var-iation. This challenge is especially great if, fora given disease state, there are many associatedstates of the microbial species (microbiota) or mi-crobial gene repertoire (microbiome), each sharedby relatively few individuals. Microbiota config-

urations are influenced by early environmentalexposures and are generally more similar amongfamily members (2, 7, 10, 11).

There have been conflicting reports aboutthe relation between interpersonal differencesin the structure of the gut microbiota and hostbody mass index (BMI). Taxonomic profilesfor obese and lean individuals may have dis-tinct patterns between human populations, buttechnical issues related to how gut samples areprocessed and community members are identifiedby 16S ribosomal RNA (rRNA) gene sequencingmay also play a role in observed differences.The relative contributions of the microbiota anddietary components to obesity and obesity-relatedmetabolic phenotypes are unclear and likelymulti-faceted (2, 12–17). Transplants of fecal microbiotafrom healthy donors to recipients with metabolicsyndrome have provided evidence that the micro-biota can ameliorate insulin-resistance, althoughthe underlying mechanisms remain unclear (18).

Monozygotic (MZ) or dizygotic (DZ) twinsdiscordant for obesity (19, 20) provide an attract-ive model for studying the interrelations betweenobesity, its associated dietary and lifestyle risk fac-tors, and the gut microbiota/microbiome. In thecase of same-sex twins discordant for a diseasephenotype, the healthy co-twin provides a valuablereference control to contrastwith the co-twin’s disease-associated gut community. However, this comparisonis fundamentally descriptive and cannot establishcausality. Transplanting a fecal sample obtainedfrom each twin in a discordant pair into separategroups of recipient germ-free mice provides anopportunity to (i) identify structural and functional

differences between their gut communities; (ii)generate and test hypotheses about the impact ofthese differences on host biology, including bodycomposition and metabolism; and (iii) determinethe effects of diet-by-microbiota interactionsthrough manipulation of the diets fed to these“humanized” animals and/or the representation ofmicrobial taxa in their gut communities.

Reproducibility of Microbiota Transplantsfrom Discordant TwinsWe surveyed data collected from 21- to 32-year-oldfemale twin pairs (n = 1539) enrolled in the Mis-souri Adolescent Female Twin Study [MOAFTS;(21, 22); for further details, see ref. (23)]. Werecruited four twin pairs, discordant for obesity(obese twin BMI > 30 kg/m2) with a sustainedmultiyear BMI difference of ≥5.5 kg/m2 (n =1 MZ and 3 DZ pairs) (Fig. 1A). Fecal sampleswere collected from each twin, frozen immedi-ately after they were produced, and stored at–80°C. Each fecal sample was introduced, via asingle oral gavage, into a group of 8- to 9-week-old adult male germ-free C57BL/6J mice (onegnotobiotic isolator per microbiota sample; eachrecipient mouse was individually caged withinthe isolator; n = 3 to 4 mice per donor microbiotasample per experiment; n = 1 to 5 independentexperiments per microbiota). All recipient micewere fed, ad libitum, a commercial, sterilizedmousechow that was low in fat (4% byweight) and highin plant polysaccharides (LF-HPP) (23). Fecalpellets were obtained from each mouse 1, 3, 7,10, and 15 days post colonization (dpc) and, formore prolonged experiments, on days 17, 22, 24,29, and 35.

Unweighted UniFrac-based comparisons ofbacterial 16S rRNA data sets generated from theinput human donor microbiota, from fecal sam-ples collected from gnotobiotic mice and fromdifferent locations along the length of the mousegut at the time they were killed (table S1A), pluscomparisons of the representation of genes withassignable enzyme commission numbers (ECs)in human fecal and mouse cecal microbiomes(defined by shotgun sequencing), disclosed thattransplant recipients efficiently and reproduciblycaptured the taxonomic features of their humandonor’s microbiota and the functions encoded bythe donor’s microbiome (see Fig. 1B; fig. S1, Ato E; fig. S2; table S1B; and table S2, A to D)(23). The 16S rRNA data sets allowed us to iden-tify bacterial taxa that differentiate gnotobioticmice harboring gut communities transplantedfrom all lean versus all obese co-twins [analysis ofvariance (ANOVA) using Benjamini-Hochbergcorrection for multiple hypotheses] [table S3; see(23) for details].

Reproducible Transmission of Donor BodyComposition PhenotypesQuantitative magnetic resonance (QMR) anal-ysis was used to assess the body compositionof transplant recipients 1 day, 15 days, and, in

RESEARCHARTICLE

1Center for Genome Sciences and Systems Biology, Washing-ton University School of Medicine, St. Louis, MO 63108, USA.2Department of Psychiatry, Washington University School ofMedicine, St. Louis, MO 63110, USA. 3George Warren BrownSchool of Social Work, Washington University, St. Louis, MO63130, USA. 4Architecture et Fonction des MacromoléculesBiologiques, CNRS and Aix Marseille Université, CNRS UMR7257, 13288 Marseille, France. 5Department of Cellular andMolecular Medicine, Faculty of Health and Medical Sciences,University of Copenhagen, DK-2200, Copenhagen, Denmark.6Sarah W. Stedman Nutrition and Metabolism Center, DukeUniversity Medical Center, Durham, NC 27710, USA. 7Depart-ment of Medicine, Duke University Medical Center, Durham,NC 27710, USA. 8Department of Pharmacology and CancerBiology, Duke University Medical Center, Durham, NC 27710,USA. 9Department of Medicine, Washington University Schoolof Medicine, St. Louis, MO 63110, USA. 10Mondelez Interna-tional, Glenview, IL60025,USA. 11Kraft FoodsGroup,Glenview,IL 60025, USA. 12Department of Chemistry and Biochemistry,University of Colorado, Boulder, CO 80309, USA. 13Molecular,Cellular and Developmental Biology, University of Colorado,Boulder, CO 80309, USA. 14Biofrontiers Institute, University ofColorado, Boulder, CO 80309, USA. 15Howard Hughes MedicalInstitute, University of Colorado, Boulder, CO 80309, USA.

*To whom correspondence should be sent. E-mail: [email protected]

www.sciencemag.org SCIENCE VOL 341 6 SEPTEMBER 2013 1241214-1

Page 4: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

the case of longer experiments, 8, 22, 29, and35 days after transplantation. The increased adi-posity phenotype of each obese twin in a discor-dant twin pair was transmissible: The change inadipose mass of mice that received an obese co-twin’s fecal microbiota was significantly greaterthan the change in animals receiving her leantwin’s gut community within a given experimentand was reproducible across experiments (P ≤0.001, one-tailed unpaired Student’s t test; n = 103mice phenotyped) (Fig. 1, C to E). Epididymalfat pad weights (normalized to total body weight)were also significantly higher in mice colonizedwith gut communities from obese twins (P ≤ 0.05,one-tailed unpaired Student’s t test). These dif-ferences in adiposity were not associated withstatistically significant differences in daily chowconsumption (measured on days 1, 8, and 15 aftergavage and weekly thereafter for longer exper-iments) or with appreciably greater inflamma-tory responses in recipients of obese comparedwith lean co-twin fecal microbiota as judged byfluorescence-activated cell sorting (FACS) anal-ysis of the CD4+ and CD8+ T cell compartmentsin spleen, mesenteric lymph nodes, small intes-tine, or colon [see (23) for details].

Functional Differences Between TransplantedMicrobial CommunitiesFecal samples collected from gnotobiotic micewere used to prepare RNA for microbial RNA se-quencing (RNA-Seq) of the transplanted mi-crobial communities’ meta-transcriptomes (tableS1C). Transcripts were mapped to a database ofsequenced human gut bacterial genomes and as-signed to Kyoto Encyclopedia of Genes and Ge-nomes (KEGG) Enzyme Commission numbers(ECnumbers) [see ref. (23)]. Significant differencesand distinguishing characteristics were definedusing ShotgunFunctionalizeR, which is based on aPoisson model (24) (see table S4 and table S5 forECs and KEGG level 2 pathways, respectively).Transcripts encoding 305 KEGG ECs were dif-ferentially expressed between mice harboring mi-crobiomes transplanted from lean or obese donors[ShotgunFunctionalizeR, Akaike's informationcriterion (AIC) < 5000; P ≤ 10−30].

Mice harboring the transplanted microbiomesfrom the obese twins exhibited higher expressionof microbial genes involved in detoxification andstress responses; in biosynthesis of cobalamin;metabolism of essential amino acids (phenyl-alanine, lysine, valine, leucine, and isoleucine)

and nonessential amino acids (arginine, cysteine,and tyrosine); and in the pentose phosphate path-way (fig. S3, A and B; table S4, B to G; andtable S5). Follow-up targeted tandem mass spec-trometry (MS/MS)–based analysis of amino acidsin sera obtained at the time mice were killed dem-onstrated significant increases in branched-chainamino acids (BCAA: Val and Leu/Ile), as well asother amino acids (Met, Ser, and Gly), plus trendsto increase (Phe, Tyr, and Ala), in recipients ofmicrobiota from obese compared with lean co-twins in discordant twin pairs DZ1 and MZ4(tables S1D and S6A). These specific aminoacids, as well as the magnitude of their differ-ences, are remarkably similar to elevations inBCAA and related amino acids reported in obeseand insulin-resistant versus lean and insulin-sensitive humans (25). This finding suggestedthat the gut microbiota from obese subjectscould influence metabolites that characterizethe obese state.

In contrast, the transplanted microbiomes fromlean co-twins exhibited higher expression of genesinvolved in (i) digestion of plant-derived poly-saccharides [e.g., a-glucuronidase (EC 3.2.1.139),a-L-arabinofuranosidase (EC3.2.1.55)]; (ii) fermen-

A Twin Pair

BMI (kg/m2) 23 32 25.5 31 19.5 30.7 24 33

1 (DZ) 2 (DZ) 3 (DZ) 4 (MZ) C D

8dpc 15dpc 22dpc 29dpc 35dpc

-10

0

10

20

30

Obese co-twin

Twin pair 1

Lean co-twin

B

PC

1 (2

3%)

Experiment 3Experiment 1 Experiment 2Input samples

% C

hang

e in

fat m

ass

All leanco-twin donors

All obeseco-twin donors

-2

0

2

4

6

% C

hang

e in

bod

y co

mpo

sitio

n

Lean body massFat mass

All leanco-twin donors

All obeseco-twin donors

-5

0

5

10

15

% C

hang

e in

fat m

ass

***

E

1 3 7 10 15

Lean donor

Obese donor

dpc

Twin pair 1

Fig. 1. Reliable replication of human donor microbiota in gnotobioticmice. (A) Features of the four discordant twin pairs. (B) Assembly of bacterialcommunities in mice that had received intact and uncultured fecal microbiotatransplants from the obese and lean co-twins in DZ pair 1. Principal coordinatesanalysis plot of principal coordinate 1 (PC1) based on an unweighted UniFracdistance matrix and 97%ID OTUs present in sampled fecal communities. Circlescorrespond to a single fecal sample obtained at a given time point from a givenmouse and are colored according to the experiment (n = 3 independent ex-periments). Note that assembly is reproducible within members of a group ofmice that have received a given microbiota, as well as between experiments. (C)Body composition, defined by QMR, was performed 1 day and 15 dpc of eachmouse in each recipient group. Mean values (T SEM) are plotted for the percentincrease in fat mass and lean body mass at 15 dpc for all recipient mice of eachof the four obese co-twins’ or lean co-twins’ fecal microbiota, normalized to the

initial body mass of each recipient mouse. A two-way ANOVA indicated thatthere was a significant donor effect (P ≤ 0.05), driven by a significant differencein adiposity and total body mass between mice colonized with a lean or obeseco-twin donor’s fecal microbiota (adjusted P ≤ 0.05; Sidák’s multiple comparisontest). (D) Mean values (T SEM) are plotted for the percent change in fat mass at15 dpc for all recipient mice of each of the four obese co-twins’ or lean co-twins’fecal microbiota. Data are normalized to initial fat mass (n = 3 to 12 animalsper donor microbiota; 51 to 52 mice per BMI bin; total of 103 mice). ***P ≤0.001, as judged by a one-tailed unpaired Student’s t test. (E) More prolongedtime course study for recipients of fecal microbiota from co-twins in discordantDZ pair 1 (mean values T SEM plotted; n = 4 mice per donor microbiota).The difference between the gain in adiposity calculated relative to initial fatmass (1 dpc) between the two recipient groups of mice is statistically signif-icant (P ≤ 0.001, two-way ANOVA).

6 SEPTEMBER 2013 VOL 341 SCIENCE www.sciencemag.org1241214-2

RESEARCH ARTICLE

Page 5: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

tation to butyrate [acetyl-CoA C-acetyltransferase(EC 2.3.1.9), 3-hydroxybutyryl-CoA dehydrogenase(EC 1.1.1.157), 3-hydroxybutyryl-CoA dehydratase(EC 4.2.1.55), butyryl-CoA dehydrogenase (EC1.3.8.1)] (fig. S3,C andD); and (iii) fermentation topropionate [succinate dehydrogenase (EC1.3.99.1),phosphoenolpyruvate carboxykinase (EC 4.1.1.32),methylmalonyl-CoA mutase (EC 5.4.99.2)] (tableS4A). Follow-up gas chromatography–mass spec-trometry (GC-MS) of cecal contents confirmed thatlevels of butyrate and propionate were significantlyincreased and that levels of several mono- anddisaccharides significantly decreased in animalscolonized with lean compared with obese co-twingut communities (P ≤ 0.05, unpaired Student’s ttest) (fig. S4, A andB, and table S6B). Procrustesanalysis, using a Hellinger distance matrix (26), re-vealed significant correlations between taxonomicstructure [97% identity (97%ID) threshold for de-fining distinct operational taxonomic units (OTU)in fecal samples], transcriptional profiles (enzymerepresentation in fecal mRNA populations), andmetabolic profiles (GC-MS of cecal samples), withseparation of groups based on donor microbiotaand BMI (Mantel test, P ≤ 0.001) (fig. S5).

These results suggest that, in this diet context,transplanted microbiota from lean co-twins had agreater capacity to breakdown and ferment poly-saccharides than the microbiota of their obese co-twins. Previous reports have shown that increasedmicrobial fermentation of nondigestible starchesis associated with decreased body weight and de-creased adiposity in conventionally raised mice thatharbor amousemicrobiota [e.g., refs. (27–29)].

Phenotypes Produced by BacterialCulture CollectionsWe followed up these studies of transplanted, in-tact, anduncultureddonor communitieswitha set ofexperiments involving culture collections producedfrom the fecal microbiota of one of the discordanttwin pairs. Our goal was to determine whether cul-tured bacterialmembers of the co-twins’microbiotacould transmit the discordant adiposity phenotypesanddistinctivemicrobiota-associatedmetabolicpro-files when transplanted into gnotobiotic mouse re-cipients that received the LF-HPP chowdiet.

Collectionsofculturedanaerobicbacteriaweregenerated from each co-twin in DZ pair 1 andsubsequently introduced into separate groups of8-week-old germ-freemaleC57BL/6Jmice (n=5 independent experiments; n = 4 to 6 recipientmice per culture collection per experiment). Theculture collections stabilized in the guts of re-cipient mice within 3 days after their introduc-tion [see (23); fig. S6, A to E; and table S7 fordocumentation of the efficient and reproduciblecapture of cultured taxa and their encoded genefunctions between groups of recipient mice].

As in the case of uncultured communities, weobserved a significantly greater increase in adi-posity in recipients of the obese twin’s culture col-lection compared with the lean co-twin’s culturecollection (P ≤ 0.02, one-tailed unpaired Student’s ttest) (Fig. 2,A andB).NontargetedGC-MSshowed

that the metabolic profiles generated by the trans-planted culture collections clustered with theprofiles produced by the corresponding intactuncultured communities (fig. S6E). In addition,the fecal biomass of recipients of the culture col-lection fromthe lean twinwassignificantlygreaterthan the fecal biomass of mice receiving the cul-ture collection fromher obese sibling; these differ-ences were manifest within 7 days (P ≤ 0.0001,two-way ANOVA) (fig. S7A).

Cohousing Ob and Ln Animals Preventsan Increased Adiposity PhenotypeBecause mice are coprophagic, the potential fortransfer of gut microbiota through the fecal-oralroute is high. Therefore, we used cohousing todeterminewhether exposureof amouseharboringa culture collection from the lean twin couldprevent development of the increased adiposityphenotype and microbiome-associated metabolicprofile of a cage mate colonized with the culturecollection from her obese co-twin or vice versa.Fivedaysaftergavage,wheneachof the inoculatedmicrobial consortia had stabilized in the guts ofrecipient animals, a mouse with the lean co-twin’sculture collectionwas cohousedwith amousewiththe obese co-twin’s culture collection (abbreviatedLnch and Obch, respectively). Control groups con-sisted of cages of dually housed recipients of thelean twin culture collection and dually housedrecipients of the obese co-twin’s culture collection(n = 3 to 5 cages per housing configuration perexperiment; n = 4 independent experiments; eachhousing configuration in each experiment wasplaced in a separate gnotobiotic isolator) (Fig.2A). All mice were 8-week-old C57BL/6J males.Allwere fed thesameLF-HPPchowadlibitumthatwas used for transplants involving the correspond-ingunculturedcommunities.Beddingwaschangedbefore initiation of cohousing. Fecal samples werecollectedfromall recipients1,2,3,5,6,7,8,10,and15 days after gavage. Body composition wasmea-sured byQMR1 and 5 days after gavage, and after10 days of cohousing.

Obch mice exhibited a significantly lower in-crease in adiposity comparedwith control Ob ani-mals that had never been exposed tomice harboringthe lean co-twin’s culture collection (P ≤ 0.05, one-tailed unpaired Student’s t test). Moreover, the adi-posity of theseObch animals was not significantlydifferent from Ln controls (P > 0.05, one-tailed un-paired Student’s t test) (Fig. 2B). In addition, expo-sure to Obch animals did not produce a significanteffecton theadiposityofLnchmice:Their adiposityphenotypes and fecal biomass were indistinguish-able from dually housed Ln controls (Fig. 2B; andfig. S7,B andC).Cohousing caused the cecalmeta-bolic profile of Obch mice to assume features ofLnch and control Ln animals, including higher lev-els (compared with dually housed Ob controls) ofpropionate and butyrate and lower levels of cecalmono- and disaccharides, as well as BCAA andaromaticaminoacids (Fig.2,CandD,andfig.S8).

Principal coordinates analysis of unweightedUniFrac distances revealed that the fecal micro-

biota of Obch mice were reconfigured so that theycame to resemble the microbiota of Lnch cagemates. In contrast, themicrobiota of the Lnch cagemates remained stable (fig. S9, A to C). We per-formed a follow-up analysis to identify species-level taxa that had infiltrated into and/or had beendisplaced from the guts of mice harboring the LnandObculturecollections.Wedidsobycharacter-izingthedirectionandsuccessofinvasion.MicrobialSourceTracker estimates, for every species-leveltaxon or 97%ID OTU, the Bayesian probability(P) of its beingderived fromeachof a set of sourcecommunities (30). The fecal microbiota of Ln orOb controls sampled 5 days after colonizationwere used as source communities to determine thedirection of invasion. The fecal communitiesbelonging to eachLnch andObchmousewere thentraced to these sources.Wedefined thedirectionofinvasion for thesebacterial taxa, bycalculating thelog odds ratio of the probability of a Ln origin(PLn) or anOborigin (POb) for eachspecies-leveltaxon or 97%IDOTU, i, as follows:

log2 PLni=PObi� �

A positive log odds ratio indicated that aspecies or 97%ID OTU was derived from a Lnsource; a negative log odds ratio indicated an Obsource. An invasion score was calculated toquantify the success of invasion of each speciesor 97%IDOTU, i, into each cohousinggroup, j, asfollows:

Invasion Scoreij ¼ log2 Aij

.Bij

� �

where Aij is the average relative abundance oftaxon i inallfecalsamplescollectedfromgroup jaftercohousing, and Bij is its relative abundance in allsamples taken from that group before cohousing.

The observed mean of the distribution ofinvasion scores forObch animalswas significantlyhigher than that for dually housedOb-Ob controls(P ≤ 0.0005, Welch’s two-sample t test) (fig.S10A). This was not the case for Lnch animalswhen compared with dually housed Ln-Lncontrols (P > 0.05), which suggested that therewas significant invasion of components of theLnch microbiota into the microbiota of Obch cagemates, but not vice versa. To quantify invasionfurther, we used the mean and standard deviationof the null distribution of invasion scores (definedas the scores from recipients of the Ln or Obmicrobiota that had never been cohoused witheach other) to calculate a z value and aBenjamini-Hochberg adjusted P value for the invasion scoreof each species in Lnch and Obch mice. We con-servatively defined a taxon as a successful invaderif it (i) had a Benjamini-Hochberg adjusted P ≤0.05, (ii)was represented in≥75%ofObch or Lnch

mice when sampled 7 and 10 days after initiationof cohousing, and (iii) had a relative abundance of≤0.05% before cohousing and ≥0.5% in the fecalmicrobiota at the time mice were killed. We de-fined a taxon that was displaced from an animal’smicrobiota upon cohousing as having a relative

www.sciencemag.org SCIENCE VOL 341 6 SEPTEMBER 2013 1241214-3

RESEARCH ARTICLE

Page 6: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

abundance ≥1% in Lnch or Obch mice before theywere cohoused and a relative abundance <0.5%after cohousing.

The direction and success of invasion areshown in Fig. 2E and table S8A. The most suc-cessful Lnch invaders of theObchmicrobiotawere

members of the Bacteroidetes (rank order oftheir invasionscores:Bacteroidescellulosilyticus,B. uniformis, B. vulgatus, B. thetaiotaomicron,

A

Obch

Ln

5 dpc

5 dpc

Lnch

ObOb

Isolator Cage

Ob

Ob

5 dpc Ln

Ln

Ln

Obch

Ln5 dpc

LnchOb 2xGFch

Con

trol

s

B D

********* ***

******

Cellobiose Maltoseor similar

disaccharide

0

5

10

15

µmol

/mg

wet

cec

al c

onte

nts

Propionate Butyrate0

2

4

6

8

10

LnOb

Obch Lnch GFch

Log 2

(spe

ctra

l abu

ndan

ce)

C

** *** ** ***

** * *

***

Ob-Ob Ln-Ln Obch Lnch GF-10

-5

0

5

10

15

% C

hang

e in

fat m

ass

*

**

**

ch

1 2 3 5 6 7 8 10 15

Lnch Obch

1 2 3 5 6 7 8 12 15dpc prior to co-housing co-housing

GFch

prior to co-housing co-housing co-housing

Clostridium clostridioforme Butyricimonas virosa Clostridium asparagiforme Turicibacter sanguinis Ruminococcus sp. 5_1_39BFAA Clostridium glycolicum Alistipes putredinis Clostridium tertium Bacteroides finegoldii Clostridium sp. MLG480 Clostridium symbiosum Firmicutes Other Enterococcus faecium Roseburia faecis Ruminococcus sp. CCUG 37327 A Clostridium disporicum Odoribacter splanchnicus Alistipes shahii Bifidobacterium longum Roseburia spp. Anaerotruncus colihominis Bacteroides thetaiotaomicron Bacteroides cellulosilyticus Bacteroides vulgatus Bacteroides uniformis Bacteroides caccae Ruminococcus sp. DJF VR70k1 Parabacteroides merdae Akkermansia muciniphila Clostridium sp. NML 04A032 Clostridium ramosum Eubacterium limosum Oscillibacter spp. Enterobacteriaceae Other Alistipes indistinctus Holdemania filiformis Betaproteobacteria Other Clostridium spp. Bacteroides spp. Ruminococcus torques Blautia producta Bacteroides massiliensis Clostridium hathewayi Blautia hansenii Alistipes finegoldii

12 10 6 7 8 12 1510RA RA RA

*

*****

*

−6 −2 2 60 5 10 15 -20 0 20

Direction of invasion

Ln sourceOb sourceNot detectedSpecies present in both Ob and Ln

Relative abundance(RA) (%) b (before),a (after) co-housing

Fold change (fc)(Log(RAach/RAbch))

E

Fig. 2. Cohousing Obch and Lnch mice transforms the adiposity pheno-type of cage mates harboring the obese co-twin’s culture collection to alean-like state. (A) Design of cohousing experiment: 8-week-old, male, germ-free C57BL/6J mice received culture collections from the lean (Ln) twin or theobese (Ob) co-twin in DZ twin pair 1. Five days after colonization, mice werecohoused in one of three configurations: Control groups consisted of duallyhoused Ob-Ob or Ln-Ln cage mates; the experimental group consisted of duallyhoused Obch-Lnch cage mates (data shown from five cages per experiment; twoindependent experiments) or Obch-Lnch- GFch-GFch cage mates (n = 3 cages perexperiment). All mice were fed a LF-HPP diet. (B) Effects of cohousing on fatmass. Changes from the first day after cohousing to 10 days after cohousingwere defined using whole-body QMR. *P ≤ 0.05, **P ≤ 0.01 compared withOb-Ob controls, as defined by one-tailed unpaired Student’s t test. (C) TargetedGC-MS analysis of cecal short-chain fatty acids. Compared with Ob-Ob controls,the concentrations of propionate and butyrate were significantly higher in thececa of Obch, Ln-Ln, Lnch, and GFch mice. (D) Nontargeted GC-MS analysis ofcecal levels of cellobiose and “maltose or a similar disaccharide.” *P ≤ 0.05;**P ≤ 0.01; ***P ≤ 0.005. (E) Evidence that bacterial species from the Lnch

microbiota invade the Obch microbiota. Shown are SourceTracker-based esti-mates of the proportion of bacterial taxa in a given community sampled from acage mate. For Obch-Lnch cohousing experiments, Obch and Lnch microbiota weredesignated as sink communities, whereas the gut microbiota of Ob-Ob and Ln-Ln controls (at 5 dpc) were considered source communities. Red indicates speciesderived from the Lnch gut microbial community. Blue denotes species derivedfrom the Obch microbiota. Black denotes unspecified source (i.e., both com-munities have this species), whereas orange indicates an uncertain classificationby the SourceTracker algorithm. “Other” after a genus name means unclassified.An asterisk placed next to a species indicates that it is a successful invader asdefined in the text. Average relative abundance (RA) in the fecal microbiota isshown before cohousing (b, at 5 dpc) and after cohousing (a, at 15 dpc). Theaverage fold-change (fc) in relative abundance for a given taxon, for all timepoints before and after cohousing is shown (excluding the first 2 days imme-diately after gavage of the microbiota and immediately after initiation of co-housing). Note that only taxa with significant changes in relative abundancebetween samples collected before and samples collected after cohousing areshown. For a full accounting of all taxa, see table S8A.

6 SEPTEMBER 2013 VOL 341 SCIENCE www.sciencemag.org1241214-4

RESEARCH ARTICLE

Page 7: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

B. caccae,Alistipesputredinis, andParabacteroidesmerdae). Invasiveness exhibited specificity atthe 97%ID OTU level (fig. S11). In contrast,cohousing did not result in significant inva-sion of Lnch intestines with members of the Obch

microbiota.In macroecosystems, successful invaders are

often more likely to become established if theyare from divergent taxa that do not share a nichewithmembers of the native entrenched commu-nity (31,32).Weused theNetRelatedness Index[NRI (33)] to show that Ln and Ob communities

have different phylogenetic structures and that thepatterns of phylogenetic clustering and dispersionof Ob, but not Ln microbiota, change as a con-sequence of cohousing. Specifically fecal Ln,Lnch, and Obch communities had a negative NRI,which indicates that the community is more phy-logenetically dispersed thanOb controls. Further-more, the fecal communities of Lnch and Obch

animals had significantly more shared branchlength with each other after cohousing comparedwith before cohousing [for a description of cal-culations and interpretation, see (23), fig. S9E,

and fig. S12]. Together, these patterns of phyloge-netic overdispersion and increased shared branchlength in cohoused Lnch and Obch animals led usto hypothesize that Bacteroides in the Lnch com-munitywere efficient invaders of Obch communi-ties because theywere able to occupy unoccupiedniches in Obch intestines. Note that increased rep-resentationofBacteroidetes hasbeendocumentedin several independent studies of the gut micro-biotaofconventionally raised leanmicecomparedwithmice having genetic- or diet-induced obesity(34, 35).

Fig. 3. Effect of cohousingonmetabolic profilesinmiceconsumingtheLF-HPPdiet. (A) Spearman’scorrelation analysis of cecal metabolites and cecalbacterial species–level taxa in samples collected fromObch, Lnch, GFch, Ln39ch, and ObchLn39 cage mates andfrom Ob-Ob and Ln-Ln controls (correlations with P ≤0.0001 are shown). Taxonomic assignments weremade using a modified taxonomy from the NationalCenter for Biotechnology Information (of the U.S.National Institutes of Health) (23). Bacterial speciesand cecal metabolites enriched in animals colonizedwith either the Ln or Ob culture collections are coloredred and blue, respectively. An asterisk in the coloredbox indicates that a taxon or metabolite is signif-icantly enriched in mice colonized with Ln (red) orOb (blue) culture collections. Bacterial species col-ored red denote significant invaders from Lnch miceinto the gut microbiota of Obch cage mates. (B)Cecal bile acids measured by UPLC-MS. Note thatlevels are plotted as log-transformed spectral abun-dances. Significance of differences relative to Ob-Obcontrols were defined using a two-way ANOVA withHolm-Šidák’s correction for multiple hypotheses;*P ≤ 0.05; **P ≤ 0.01. (C and D) QPCR assays ofFXR and Fgf15 mRNA levels in the distal ileum.Data are normalized to Ln-Ln controls. (E) QPCR ofhepatic Cyp7a1 mRNA, normalized to Ln-Ln controls.*P ≤ 0.05; **P ≤ 0.01; ****P ≤ 0.001 (defined byone-tailed, unpaired Student’s t test using Ob-Obmice as reference controls). (F) Correlating cecal bileacid profiles with the FXR-Fgf15-Cyp7A signalingpathway in the different groups of mice. (Top) Thedendrogram highlights the differences in the profilesof 37 bile acid species between Ob-Ob controls andthe other three treatment groups (table S11). Thedendrogram was calculated using the Bray-Curtisdissimilarity index and the average relative abun-dance of each bile acid species among all mice be-longing to a given treatment group.

Bacteroides uniformis*

Alistipes putredinis*

Parabacteroides merdae*

Ade

nine

Gly

cero

l pho

spha

teA

spar

tate

Cel

lobi

ose

Dis

sach

arid

e si

mila

r to

mal

tose

Mal

tose

Mal

eate

Mal

ate

Glu

tam

ine

Glu

tam

ate

Thr

eoni

neS

erin

eP

heny

lala

nine

Orn

ithin

eC

yste

ine

Tyro

sine

Ala

nine

Gly

cera

teIs

oleu

cine

Leuc

ine

Sor

bose

Tag

atos

eV

alin

eG

lyci

neC

adav

erin

eG

lyco

late

Ace

tate

Pro

pion

ate

But

yrat

e

Actinomyces urogenitalis

Paraprevotella clara

Eubacterium limosum

Alistipes unclassified

Ruminococcus sp. 5

Ruminococcus bromii

Ruminococcus callidus

Ruminococcus sp. ID8

Burkholderiales unclassified

Betaproteobacteria unclassified

Eubacterium ventriosum

Ruminococcus sp. 14531

Bacteroides ovatus

Parabacteroides distasonis

Clostridium hathewayi

Blautia glucerasea

Eubacterium desmolans

Ruminococcus obeum

Akkermansia muciniphila

Clostridium ramosum

Roseburia unclassified

Clostridium symbiosum

−0.5 0 0.5

Absolute Spearman’scorrelation coefficient rho

LF/HPP diet

*

*

***

***

*

**** * *** * **

Clostridium hathewayiy

ns (p = 0.08)

C D EIleal FXR mRNA

Fol

d di

ffere

nce

to L

n-Ln

con

trol

s

Ileal Fgf15 mRNA Liver Cyp7a1 mRNA

B

Ileum

Live

r

****** **

**

******

F

Lum

inal

con

tent

s (c

ecal

) Bile acids

FXR

Fgf15

Fgfr4

Portal circulation

Cyp7a1

Fol

d di

ffere

nce

to L

n-Ln

con

trol

s

Fol

d di

ffere

nce

to L

n-Ln

con

trol

s

Ob-Ob

Ln-L

nLn

ch

Obch

Cecal bile acids

β-Mur

icholi

c

acid

Cholic

acid

α-Mur

icholi

c

acid

Hyode

oxyc

holic

acid

Ursod

eoxy

choli

c

acid

7-Ket

odeo

xych

olic

acid

Alloch

olic

acid

Cheno

deox

ycho

lic

acid

5

6

7

8

Log

(spe

ctra

l abu

ndan

ce)

Ob-ObLn-Ln Obch

Lnch

** ** **

* *** ** ** **

** ** ** ** ** ** *** *

**** ** **

*

**

Ob-Ob

Ln-L

nLn

ch

Obch

Ob-Ob

Ln-L

nLn

ch

Obch

Ob-Ob

Ln-L

nLn

ch

Obch

Ob-Ob

Ln-L

nLn

ch

Obch

-4

-2

0

2

4

6

8

**

-1

0

1

2

3

4

0

2

4

A

Enriched in micecolonized with or invaded by membersof a Ln microbiota

Enriched in micecolonized with anOb microbiota

Ob-Ob

Ln-L

nLn

ch

Obch

Ob-Ob

Ln-L

nLn

ch

Obch

www.sciencemag.org SCIENCE VOL 341 6 SEPTEMBER 2013 1241214-5

RESEARCH ARTICLE

Page 8: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

Control experiments involving cohousing agerm-free (GFch)mousewith anObmouse 5daysafter transplantation of the complete culturecollection from the obese co-twin demonstratedeffective transmission of the Ob adiposity pheno-type to the GF cage mate (defined by epididymalfatpadweightasapercentageof totalbodyweight;P > 0.05, comparing GFch-Ob and Obch-GF cagemates with a two-tailed unpaired Student’s t test).The adiposity (epididymal fat pad weights) of theGFch-Ob cage mate was significantly greater com-pared with GF-GF controls [1.4 T 0.1% of bodyweight (Obch-GF);1.4T0.05%(GFch-Ob);1.5T0.05%(Ob-Ob controls) versus 0.8 T 0.06% (GF-GFcontrols); P < 10−4, two-tailed unpaired Student’s ttest between GFch-Ob and GF-GF controls; n = 6GF-GF, 8 Ob-Ob, and 10 GFch-Ob-Obch-GF)].

In separate experiments, we cohoused twoGFanimals togetherwith oneLnandoneObanimal,5 days after colonization of Ln and Ob mice(repeated in three separate cages, each cage in aseparate isolator). As with previous experiments,bedding was changed before initiation of co-housing. The GF “bystanders” did not developincreased adiposity phenotypes and manifestedcecal metabolic profiles and fecal biomass char-acteristicsof theirLnch cagemates andLncontrols(Fig. 2, B toE, and fig. S7F). In the initial phase ofcohousing (days 1 to 2), the microbiota of GFch

mice in each cage resembled that of noncohousedOb-Ob controls (fig. S9D). By the next day, the

microbiotaofeachex-GFchanimal hadundergonea dramatic change in composition, with 94.8 T0.4% of taxa now derived from their Lnch cagemate (definedbySourceTracker; see tableS8Aforlog odds ratio scores). The most prominent in-vaderswere the sameprominent invaders fromtheLnch microbiota described above (i.e., all of theBacteroidales) (Fig. 2E).We concluded that theseLn-derived taxa had greater fitness in the guts ofC57BL/6J mice consuming this diet, and had a“dominant-negative” effect on host adiposity.

Changes in the Cecal MetatranscriptomeandMetabolome of Obch AnimalsCecal samples collected at the time mice werekilled from Lnch, Obch, and control Ob and Ln ani-mals were subjected to microbial RNA-Seq. Readswere assigned to ECs as above, and Euclideandistances were calculated from the EC matrix. Theresults revealed that the metatranscriptomes ofObch animals were significantly more similar tothose of their Lnch cage mates and Ln controlsthan to Ob controls, consistent with a functionaltransformationof theObchmicrobiota toa leanlikestate as a consequence of invasion of the Ln taxa(P≤0.0001 asmeasured byaone-wayANOVA,with Holm-Šidák’s correction for multiplehypotheses) (fig. S13). The majority (55.9%)of ECs that were enriched in the cecal metatran-scriptomes of Obch mice, compared with duallyhoused Ob-Ob controls, were also significantly

enriched in dually housed Ln-Ln versus Ob-Obcontrols, including ECs that participate in car-bohydrate metabolism and protein degradation.The latter encompassed five enzymes involvedin the KEGG pathway for degradation of theBCAAvaline, leucine, and isoleucine (transcriptsencoding EC 1.2.4.4; EC 2.6.1.42; EC 5.1.99.1,and EC 6.4.1.3, which map to the genomes of in-vading members of the Bacteroides, and EC4.2.1.17,whichmaps tomembersofClostridiaceae)(table S9). The increased expression of genes in-volved in BCAA degradation is consistent withthe reduced cecal levels of BCAA observed inLn-Ln, Lnch and Obch animals versus Ob-Ob con-trols (fig. S8). These results are consistent with thenotion that invasion by Bacteroides increases theefficiency of BCAA degradation in the gut, re-ducingproductionofBCAAandrelatedmetabolitesby the microbiome and contributing to decreasedcirculating levels of BCAA in the host. As withtransplanted, intact, andunculturedmicrobiota fromthe discordant twin pairs, targetedMS/MS analy-sis confirmed that Ob-Ob controls had a globalincreaseinserumlevelsofBCAAs,aswellashigherlevels of Met, Ser, Gly, Phe, Tyr, and Ala com-pared with Ln-Ln controls (P ≤ 0.05, matchedone-way ANOVA) (table S10). However, after10 days of cohousing, Obch animals did not ex-hibit a statistically significant reduction in serumBCAA levels compared with Ob-Ob controls,although levels in Lnch cage mates were signifi-

Lnch

% C

hang

e in

bod

y m

ass

LoSF-HiFV

Ln-LnOb-Ob Obch-5

0

5

10

15**

LoSF-HiFV

Ln-LnOb-Ob LnchObch-5

0

5

10

15

% C

hang

e in

bod

y co

mpo

sitio

n

Lean body mass

Fat mass

A B

D E

C

HiSF-LoFV

Ob-Ob Ln-Ln Ob ch Ln ch0

5

10

15

20

% C

hang

e in

bod

y m

ass **

*

HiSF-LoFV

Ob-Ob Ln-Ln Obch Lnch0

5

10

15

20

% C

hang

e in

bod

y co

mpo

sitio

n

Fat massLeanbody mass

Butyr

ate

Aceta

te

3,4-

hydr

oxyp

heny

lacet

ate

Pyrog

lutam

ate

Lyxo

se/xy

lose

Isoleu

cine

Oxalat

e

Propio

nate

gam

ma-

amino

isob

utyr

ate

Butyricimonas virosa

Bacteroides vulgatus*

Bacteroides uniformis**

* *

*

Paraprevotella clara

Parabacteroides merdae

Bacteroides caccae

Akkermansia muciniphila

Firmicutes unclassified

Clostridium sp. NML 04A032

Clostridium symbiosum

Ruminococcus torques

LoSF-HiFV diet

−0.5 0 0.5

Absolute Spearman’scorrelation coefficient rho

Enriched in mice colonized with orinvaded by members of a Ln microbiota

Enriched in mice colonized with anOb microbiota

Fig. 4. Effects of NHANES-based LoSF-HiFV and HiSF-LoFV diets onbacterial invasion, body mass and metabolic phenotypes. (A and B)Mean T SEM percent changes in total body mass (A) and body composition [fatand lean body mass, normalized to initial body mass on day 4 after gavage (B)]occurring between 4 and 14 days after colonization with culture collections fromthe Ln or Ob co-twin in DZ pair 1. Cohousing Ln and Ob mice prevents anincreased body mass phenotype in Obch cage mates fed the representative LoSF-HiFV human diet (n = 3 to 5 cages per treatment group; 26 animals in total).**P ≤ 0.01, based on a one-way ANOVA after Fisher’s least significant differencetest (also see table S14 for statistics). (C) Spearman’s correlation analysisbetween bacterial species–level taxa and metabolites in cecal samples collectedfrom mice colonized with culture collections from DZ twin pair 1 Ln and Ob co-twins and fed a LoSF-HiFV diet. Red and blue squares indicate metabolites or

taxa that are significantly enriched in samples collected from dually housed Ln-Ln or Ob-Ob controls respectively. An asterisk in the colored box indicates thatthat a taxon or metabolite is significantly enriched in mice colonized with Ln(red) or Ob (blue) culture collections. (D and E) Mean T SEM of changes in bodymass and body composition in mice colonized with intact uncultured microbiotafrom DZ twin pair 2 and fed the representative HiSF-LoFV human diet. Ob-Obcontrols have greater total and lean body mass than Ln-Ln controls, but thisphenotype is not rescued in Obch animals (see table S14 for statistics). *P <0.05, **P ≤ 0.01 based on a one-way ANOVA. Note that the HiSF-LoFV dietproduces a significantly greater increase in body mass, specifically fat mass, inmice harboring the lean co-twins microbiota (Ln-Ln and Lnch) than when theyare fed the LoSF-HiFV diet [see (A) versus (D), and (B) versus (E); two-wayANOVA with Holm-Šidák’s correction for multiple hypotheses].

6 SEPTEMBER 2013 VOL 341 SCIENCE www.sciencemag.org1241214-6

RESEARCH ARTICLE

Page 9: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

cantly lower than in Ob-Ob controls and notdifferent from Ln controls (table S10). Morecomplete understanding of the contributions ofthe gut microbial community to Ob-associatedmetabolic phenotypes will require detailed, long-term, time-series studies of microbial and hosttranscriptomes andmetabolomes.

Significant correlations between cecal metab-olite levels andbacterial species represented in themicrobiotaofOb-Ob,Ln-Ln,Lnch,Obch andGFch

mice consuming theLF-HPPdiet are summarizedin Fig. 3A (defined by asymptotic P values for allSpearman’s correlations, corrected for multiplehypotheses using the Benjamini-Hochberg pro-cedure). For example, BCAA and the products ofamino acidmetabolismwere positively correlatedwithClostridiumhathewayi (Fig. 3A). Thismem-ber of the Firmicutes represented an average of2.54%of theObfecalmicrobiotabeforecohousing;its relative abundancewas significantly reduced inObch animals, and it was not able to successfullyinvade themicrobiotaofLnchcagemates (Figs.2Eand 3A and table S8A).

Three members of the Bacteroidetes, B.uniformis, Parabacteroides merdae, and A.putredinis, which were prominent invaders ofthe Obch gut, positively correlated with cecalacetate, propionate, and butyrate levels. Where-as members of these species generate acetate

and propionate, their ability to produce butyratehas not been reported. Their positive associationwith butyrate levels could be due to interspeciesacetate cross-feeding with butyrate-producingtaxa (36, 37). The negative correlation betweenadiposity and cecal butyrate and propionateconcentrations in Lnch, Obch, and GFch micro-biota (r = –0.49 and –0.45, respectively, P ≤ 0.05)is consistent with previous studies claiming arole for these short-chain fatty acids in influenc-ing host energy balance (38, 39).

The gut microbiota affects the composition andrelative abundance of bile acids species througha variety of metabolic transformations (40, 41).Ultra-performance liquid chromatography–massspectrometry (UPLC-MS) analysis of 37 bile acidspecies in cecal samples obtained from Ln-Ln,Ob-Ob, Obch, and Lnch mice revealed significant-ly lower levels of eight bile acids in Ob-Ob com-pared with Ln-Ln controls (Fig. 3B). Cohousingrescued these differences, with Obch mice havingbile acid profiles that were more similar to Ln-Lnthan to Ob-Ob controls and not significantly dif-ferent from Lnch cage mates (n = 5 to 6 mice pergroup; see table S11 for all bile acids measuredthat exhibit significant differences betweenOb-Oband Ln-Ln controls).

Bile acids can have directmetabolic effects onthehostvia thenuclearfarnesoidXreceptor(FXR)

(42). Intestinal FXR mediates intestinal fibro-blast growth factor 15 (Fgf15) production. Fgf15,secrted by the gut epithelium and delivered tohepatocytesvia theportal circulation, acts throughfibroblast growth factor receptor 4 (Fgfr4) toinhibit expression of the rate-limiting enzyme inbile acid biosynthesis, cholesterol 7-a-hydroxylase(Cyp7a1) (Fig. 3, C to F) (43). Engineered FXR de-ficiency in leptin-deficient mice protects againstobesity and improves insulin sensitivity (42). Over-expression of Cyp7a1 in the livers of transgenicmice also prevents diet-induced obesity and in-sulin resistance (44). Sequestering bile acids withthe drug colesevelam lowers blood sugar in hu-mans with type 2 diabetes (45). Quantitative re-verse transcription polymerase chain reaction(qRT-PCR)analysisdisclosedthat,comparedwithLn-Ln animals, Ob-Obmice have increased FXRand Fgf15 mRNA levels in their distal small in-testine (ileum) and decreased hepatic Cyp7a1 ex-pression. Obch mice have expression profiles thatare not significantly different fromLn-Ln controls(or Lnch) (Fig. 3, C to E). Differences in bile acidmetabolism, in addition to the documented differ-ences in carbohydrate fermentation by Ob com-pared with Ln microbial communities highlightsome of the mechanisms that could contribute totheobservedmicrobiota-mediated effectsonbodycomposition.

-20 0 20

Direction of invasion

Ln sourceOb sourceNot detectedSpecies present in both Ob and Ln

0 10 20 Relative abundance(RA) (%) b (before),a (after) cohousing

Bacteroides vulgatus *Bacteroides thetaiotaomicron *Bacteroides caccae *Bacteroides cellulosilyticus *Bacteroides uniformis *Paraprevotella clara *Alistipes finegoldii *Alistipes shahii Butyricimonas virosa Collinsella aerofaciensParabacteroides merdae Alistipes putredinis *Enterobacteriaceae unclassified Shigella spp.Turicibacter sanguinis Burkholderiales unclassified Ruminococcus sp. DJF VR70k1 Akkermansia muciniphila Clostridium difficile Enterococcus faecium Enterococcus spp.Blautia producta Clostridium clostridioforme *Ruminococcus torques *Blautia spp. *Clostridium disporicum Holdemania filiformisClostridium lituseburense Clostridium symbiosum Clostridium sp. SH C52 Ruminococcus gnavus

1 2 3 5 6 7 8 10 12 144 1 2 3 5 6 7 8 10 12 144

Lnch Obch

b a fc b a fcprior to cohousing cohousing prior to cohousing cohousing

−5 0 5Fold change (fc)(Log(RAach/RAbch))

LoSF-HiFV

dpc

Fig. 5. Invasion analysis of species-level taxa in Obch or Lnch mice fedthe NHANES-based LoSF-HiFV diet. Red indicates species derived from theLnch gut microbial community. Blue denotes species derived from the Obch

microbiota. The mean relative abundance of each species-level taxon before

(b: 3 and 4 dpc) and after (a: 8, 10 and 14 dpc) cohousing is noted. Averagefold-change (fc) in relative abundance of taxa before and after colonization wasdefined as in the legend to Fig. 2E. An asterisk (*) denotes bacterial species thatsatisfy our criteria for classification as successful invaders (see text).

www.sciencemag.org SCIENCE VOL 341 6 SEPTEMBER 2013 1241214-7

RESEARCH ARTICLE

Page 10: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

Taxa-Specific Effects on Body CompositionandMetabotypeWe subsequently colonized GF mice with a con-sorium of 39 sequenced taxa (97%ID OTUs) fromthe lean co-twin’s culture collection; 22 OTUsfrom Bacteroides, including B. cellulosilyticus,B. vulgatus, B. thetaiotaomicron, B. caccae, B.ovatus; 11 OTUs from Ruminococcaceae (oneof the four family-level taxa that discriminatelean from obese), and six strains of Collinsellaaerofaciens (richly endowedwith genes involvedin the acquisition and fermentation of diversedietary carbohydrates) (see table S12 for featuresof the 39 sequenced genomes). Five days aftercolonization with the 39 taxa, these Ln39 micewere cohousedwithObmice. Control groups con-sisted of dually housed Ln-Ln and dually housedOb-Ob animals (fig. S14A) (n = 5 cages of co-housed animals per group; two independent ex-periments). Mice harboring the 39-memberconsortium (Ln39ch), when cohoused with Ob(ObchLn39) animals, remained lean (fig. S14, Aand B). However, Ln39ch animals were not ableto confer protection against an increase in adi-posity of their Obch cage mates, nor did they ex-hibit a significant increase in their cecal butyratelevels (figs. S14, B and C and S7, D and E). Eventhough some species like B. cellulosilyticus andB. vulgatus from the Ln39ch microbiota ex-hibited significant invasion into the microbiota ofObch-Ln39 cage mates, there was an overall de-crease in invasion efficiency for members of theLn39 consortia compared with those in the com-plete Ln culture collection [the mean of the dis-tribution of invasion scores for the Obch-Ln39

microbiota was not significantly different fromOb-Ob controls (P > 0.05, Welch’s two-samplet test)] (fig. S10B, table S8B). These findingsillustrate how invasiveness and adiposity are de-pendent on community context.

Diet-Specific EffectsTo define the effects of diet on Ln and Obmicrobiota-mediated transmission of body com-position and metabolic phenotypes, we constructeda diet made with foods that characterize dietsrepresenting the lower tertile of consumption ofsaturated fats and theupper tertile of consumptionof fruits and vegetables reported in 1-day recallsby participants in the U.S. National Health andNutritionExaminationSurvey(NHANES)duringthe period 2003–2008 [table S13 (23)]. Mice fedthis low–saturated fat, high fruits and vegetables(LoSF-HiFV,~33%kcalof fat per100g)NHANES-based diet were colonized with either the Ln orOb human culture collections and subsequentlycohoused. Dually housed Ln-Ln and Ob-Obanimals served as reference controls (n = 3 to 5cages per housing configuration; two indepen-dent experiments). Ob-Ob cagemates consumingthe LoSF-HiFV diet had significantly increasedbodymasscomparedwithLn-Lncagemates,whichreflected a significant increase in both adipose andean bodymass (one-wayunpairedStudent’s t test;P ≤ 0.001) (Fig. 4, A and B, and table S14).

Cohousing prevented development of an Obbody composition phenotype in Obch animals(Fig. 4, A and B).

We used 16S rRNA analysis of fecal samplescollectedondays1 to8,10,12, and14aftergavageof the culture collections, and it revealed that thisrescuewas accompanied by invasion of taxa fromLnch to Obch cage mates (P ≤ 10−6; Welch’s twosample t test comparing the mean of the distribu-tions of invasion scores between Ob-Ob controlsand Obch animals) with the most invasive speciesbeing members of Bacteroides (i.e., B. uniformis,B. vulgatus, and B. cellulosilyticus) (Fig. 5; fig.S15, A to C; and fig. S10C). These species werealso successful invaders in mice fed the LF-HPP

diet (Fig. 2E). As was the case with the LF-HPPmouse chow, fecal samples collected from Ln-Lncontrols and from Lnch and Obch animals fed thehuman (NHANES-based) LoSF-HiFV diet hadsignificantly greater microbial biomass comparedwith Ob-Ob controls (two-way ANOVA, P ≤0.001 after Holm-Šidák’s correction for multiplehypothesis).Moreover, fecalmicrobial biomassin Obch and Lnch animals was not significantlydifferent (not shown).

Significant correlations between cecal metab-olites and cecal bacterial taxa in mice fed a LoSF-HiFV diet are summarized in Fig. 4C. As in micefedaLF-HPPdiet, theinvasivespeciesB.uniformisand B. vulgatus were significantly and positively

Fig. 6. Acylcarnitine profile in theskeletal muscle of mice colonizedwith the Ob or Ln culture collec-tions fromDZ twin pair 1 and fedthe LoSF-HiFV diet. Each columnrepresents a different animal andeach row a different acylcarnitine.The identities and levels of theseacylcarnitines were determined bytargeted MS/MS (see table S15 formean values T SEM for each treat-ment group). A two-way ANOVAwith Holm-Šidák’s correction wasused to calculate whether the levelof each acylcarnitine was signifi-cantly different between Ob-Oband Ln-Ln, Lnch, or Obch animals.*P ≤ 0.05.

**

*

***

*

C22C22:1C22:2C22:3C22:4C22:5C20-OH/C18-DC/C22:6C20:1-OH/C18:1-DCC20:2-OH/C18:2-DCC20:3-OH/C18:3-DCC20C20:1C20:2C20:3C20:4C18:1-DCC18-OH/C16-DCC18:1-OH/C16:1-DCC18:2-OH/C16:2-DCC18:3-OH/C16:3-DCC18C18:1C18:2C18:3C16-OH/C14-DCC16:1-OH/C14:1-DCC16:2-OH/C14:2-DCC16:3-OH/C14:3-DCC16C16:1C16:2C16:3

C14-OH/C12-DCC14:1-OH/C12:1-DCC14:2-OH/C12:2-DCC14:3-OH/C12:3-DCC14C14:1C14:2C14:3C12-OH/C10-DCC12:1-OH/C10:1-DCC12:2-OH/C10:2-DCC12C12:1C12:2C10-OH/C8-DCC10C10:1C10:2C10:3C6-DC:C8-OHC8:1-OH/C6:1-DCC8C8:1C6

C7-DCC5-DCC4-DC:Ci4-DCC5-OH/C3-DCC4-OHC5C5:1C4:Ci4C3C2

−1 0 1

Z value

Ln-LnOb-Ob Obch

LoSF-HiFV diet(33% fat by weight)

(15 dpc)

6 SEPTEMBER 2013 VOL 341 SCIENCE www.sciencemag.org1241214-8

RESEARCH ARTICLE

Page 11: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

correlatedwith increasedcecalpropionate levelsin Ln-Ln, Obch, and Lnch animals comparedwith Ob-Ob controls.

Obesity-related insulin resistance has been as-sociated with broad-scale accumulation of acyl-carnitines in skeletal muscle (46, 47). Maneuversthat resolve the acylcarnitine accumulation inmuscle, including knockout of the malonylCoAdecarboxylase gene (Mlycd) or overexpression ofcarnitine acyltransferase (Crat), also resolve theinsulin-resistance phenotype in mice (47, 48).Therefore, we used targeted MS/MS to measure60 acylcarnitine species (C2 to C22) in liver,skeletal muscle, and serum (table S15, A to C).With theLF-HPPmousechow(4%fatbyweight),Ob-Ob animals had significantly higher levelsof hepatic short-chain acylcarnitines (C2, C4:C4i,andC4-OH)comparedwithLn-Lnmice (fig.S16;two-wayANOVAafterHolm-Šidák’s correction).The differences in short-chain acylcarnitines wererescued in Obch animals, which resembled Ln-Lncontrols (n= 5 animals per treatment group).

Ob-Ob controls fed the LoSF-HiFV diet (33%fat by weight) also had clear increases in accumu-lation of a group of even-, long-chain acylcarnitines(C14, C14:1, C16, C16:1, C18, C18:1, and C18:2)in their liver and skeletal muscle compared withLn-Ln controls [multivariate ANOVA (MANOVA),P ≤ 0.001; n = 3 to 6 animals per treatment group].Cohousing Ln and Ob animals rescued this meta-bolic phenotype in the skeletal muscle and liver(and cecum) of Obch mice (i.e., there was no sta-tistically significant difference comparedwithLn-Lncontrols) (Fig.6, fig.S16,and tableS15).Therewere no significant differences in accumulation ofthese acylcarnitines in the plasma of Ob-Ob, Ln-Ln, and cohoused animals (table S15, B and C).Together, these results suggest that the gut micro-biome influences systemic lipidmetabolism.

Because these findings raised the possibilitythat the Ob microbiome is capable of conferringmuscle insulin resistance at least in the contextof ahuman (LoSF-HiFV) diet, we subjected Ob-Oband Ln-Ln animals to a glucose tolerance test15 days after colonization with their human do-nor’s gut culture collections.An increase in serumglucose concentration was observed 15 min afterintraperitoneal injection of glucose in Ob-Obanimals (means T SEM: 338.5 T 11.07 mg/dl inOb-Ob animals versus 304.6 T 17.78mg/dl in Ln-Lnanimals;P=0.06, unpairedStudent’s t test;n=10 to 11 animals per group). No significant differ-ences in serum insulin concentrations betweenthese groups were documented at 30 min afterglucose was administered, nor did we observesignificant differences in insulin signaling in liverand skeletal muscle, defined by levels of immu-noreactive Akt phosphorylated at Thr308 and theAkt substrate AS160 phosphorylated at Thr642

measured 3 min after an insulin bolus (23). Weconcluded that the elevation in skeletal musclelong-chain acylcarnitines observed in mice colo-nized for just15dayswithanObculturecollectionand consuming the LoSF-HiFV diet is associatedwithmild glucose intolerance andmaybe an early

manifestationof thepathobiologiceffectsofobesemicrobiota on host physiology and metabolism.More complete time-course studies will be re-quired to test if this lesion in glucose homeostasiswillevolveintofull-fledgedinsulinresistancewithlonger exposure to Obmicrobiota.

To determine whether the effects exerted bydiet on the invasive potential ofmembers from theLn microbiota were specific to the culturecollections prepared frommembers of discordanttwinpair1or robust tootherLnco-twinmicrobiota,we fed the LoSF-HiFVdiet tomice colonizedwithintactunculturedfecalmicrobiotafrommembersofdiscordant twin pair 2. Cohousing experimentsperformed using the same experimental designdescribed above revealed invasion of the commu-nities of Obch animals by bacterial species fromtheirLnchcagemates,mostnotablymembersof theBacteroidetes (figs. S10DandS17A, see tableS8Dfor invasion scores).

In follow-up experiments, separate groups ofgerm-free mice colonized with intact unculturedmicrobiota fromdiscordant twin-pair 2were feda second NHANES-based diet made with foodsthat characterize U.S. diets representing theupper, rather than lower, tertile of consumptionof saturated fats and the lower, rather than upper,tertile of consumption of fruits and vegetables(abbreviated HiSF-LoFV; 44% fat by weight).Significant differences in body compositionwere documented between Ob-Ob and Ln-Lnmice fed this diet. However, cohousing of Lnand Ob mice failed to attenuate or block de-velopment of an increased body mass phenotypeinObchcagemates(n=5to11animalsperhousingconfiguration) (Fig. 4, D and E). Remarkably,there was a lack of significant invasion of mem-bers of the Lnch microbiota into the guts of Obch

cage mates (figs. S10E and S17B and table S8E).If biased fecal pellet consumption was the un-derlying mechanism leading to invasion of mem-bers of the Ln microbiota into the guts of Obch

mice in the cohousing experiments, it would haveto be diet- and microbiota-dependent. Together,these results emphasize the strong microbiota-by-diet interactions that underlie invasion andillustrate how a diet high in saturated fats and lowin fruits and vegetables can select against humangut bacterial taxa associated with leanness.

ProspectusThe findings described above provide a start-ing point for future studies that systematicallytest the effects of specified diet ingredients onmicrobiota-associated body composition andmetabolic phenotypes (e.g., components thatwhen added or subtracted restore invasiveness ofspecific members of the microbiota in the contextof the HiSF-LoFV diet). A benefit of using theapproach described in this report is that the targethuman population embodying a phenotype ofinterest is integratedinto theanimalmodel throughselection of gut microbiota representative of thatpopulation and diets representative of their pat-terns of food consumption. Our finding that cul-

ture collections generated fromhumanmicrobiotasamples can transmit donor phenotypes of inter-est (body composition and metabotypes) has anumber of implications. If these derived culturecollections can transmit aphenotype, the stage isset for studies designed to determine which cul-turable components of a given person’s gut com-munity are responsible. These tests can take theform of experiments where mice containing cul-ture collections fromdonorswith different pheno-types are cohoused andused todeterminewhetherinvasion by components of one community intoanother transformsthecagemate’sphenotypeand,correspondingly, the properties of the humanmicrobial communities they harbor. Sequencedculture collections generated from human gutmicrobiota donors also provide an opportunity tomodel and further address basic issues such as thedeterminants of invasiveness including the mech-anisms by which invasion is impacted by dietcomposition, aswell as themechanisms bywhichinvading components affect microbial and hostmetabolism.This issue is important for identifyingnext-generation probiotics, prebiotics, or a com-bination of the two (synbiotics). Moreover, theability to generate a culture collection, from anindividual—whose composition is resolved to thegene level, whose properties can be validated inpreclinical models, and whose “manufacture” isreproducible—may provide a safer and moresustainable alternative to fecal transplants formicrobiome-directed therapeutics (18, 49).

References and Notes1. P. B. Eckburg et al., Diversity of the human intestinal

microbial flora. Science 308, 1635–1638 (2005).doi: 10.1126/science.1110591; pmid: 15831718

2. P. J. Turnbaugh et al., A core gut microbiome in obeseand lean twins. Nature 457, 480–484 (2009).doi: 10.1038/nature07540; pmid: 19043404

3. E. K. Costello et al., Bacterial community variation inhuman body habitats across space and time. Science326, 1694–1697 (2009). doi: 10.1126/science.1177486;pmid: 19892944

4. J. Qin et al., A human gut microbial gene catalogueestablished by metagenomic sequencing. Nature464, 59–65 (2010). doi: 10.1038/nature08821;pmid: 20203603

5. J. Ravel et al., Vaginal microbiome of reproductive-agewomen. Proc. Natl. Acad. Sci. U.S.A. 108 (Suppl. 1),4680–4687 (2011). doi: 10.1073/pnas.1002611107;pmid: 20534435

6. P. Gajer et al., Temporal dynamics of the human vaginalmicrobiota. Sci. Transl. Med. 4, 132ra52 (2012).doi: 10.1126/scitranslmed.3003605

7. T. Yatsunenko et al., Human gut microbiome viewedacross age and geography. Nature 486, 222–227(2012). pmid: 22699611

8. C. Huttenhower et al., Structure, function and diversity ofthe healthy human microbiome. Nature 486, 207–214(2012). doi: 10.1038/nature11234; pmid: 22699609

9. K. Faust et al., Microbial co-occurrence relationshipsin the human microbiome. PLOS Comput. Biol. 8,e1002606 (2012). doi: 10.1371/journal.pcbi.1002606;pmid: 22807668

10. M. G. Dominguez-Bello et al., Delivery modeshapes the acquisition and structure of the initialmicrobiota across multiple body habitats in newborns.Proc. Natl. Acad. Sci. U.S.A. 107, 11971–11975(2010). doi: 10.1073/pnas.1002601107;pmid: 20566857

www.sciencemag.org SCIENCE VOL 341 6 SEPTEMBER 2013 1241214-9

RESEARCH ARTICLE

Page 12: Gut Microbiota from Twins Discordant for Obesity Modulate ... › sites › ...DOI: 10.1126/science.1241214 Science 341, (2013); Vanessa K. Ridaura et al. Metabolism in Mice Gut Microbiota

11. J. J. Faith et al., Succession of microbial consortiain the developing infant gut microbiome. Science 341,1237439 (2013).

12. S. Greenblum, P. J. Turnbaugh, E. Borenstein,Metagenomic systems biology of the human gutmicrobiome reveals topological shifts associated withobesity and inflammatory bowel disease. Proc. Natl.Acad. Sci. U.S.A. 109, 594–599 (2012). doi: 10.1073/pnas.1116053109; pmid: 22184244

13. M. L. Zupancic et al., Analysis of the gut microbiota in theold order Amish and its relation to the metabolicsyndrome. PLoS ONE 7, e43052 (2012). doi: 10.1371/journal.pone.0043052; pmid: 22905200

14. S. H. Duncan et al., Human colonic microbiota associatedwith diet, obesity and weight loss. Int. J. Obes. (London)32, 1720–1724 (2008). doi: 10.1038/ijo.2008.155;pmid: 18779823

15. A. Schwiertz et al., Microbiota and SCFA in lean andoverweight healthy subjects. Obesity (Silver Spring)18, 190–195 (2010). doi: 10.1038/oby.2009.167;pmid: 19498350

16. A. Santacruz et al., Gut microbiota composition is associatedwith body weight, weight gain and biochemical parametersin pregnant women. Br. J. Nutr. 104, 83–92 (2010).doi: 10.1017/S0007114510000176; pmid: 20205964

17. R. Jumpertz et al., Energy-balance studies reveal associationsbetween gut microbes, caloric load, and nutrient absorptionin humans. Am. J. Clin. Nutr. 94, 58–65 (2011).doi: 10.3945/ajcn.110.010132; pmid: 21543530

18. A. Vrieze et al., Transfer of intestinal microbiota from leandonors increases insulin sensitivity in individuals withmetabolic syndrome. Gastroenterology 143, 913, e7(2012). doi: 10.1053/j.gastro.2012.06.031; pmid: 22728514

19. P. Hakala, A. Rissanen, M. Koskenvuo, J. Kaprio, T. Rönnemaa,Environmental factors in the development of obesity inidentical twins. Int. J. Obes. Relat. Metab. Disord. 23,746–753 (1999). doi: 10.1038/sj.ijo.0800923;pmid: 10454109

20. A. Rissanen et al., Acquired preference especially fordietary fat and obesity: A study of weight-discordantmonozygotic twin pairs. Int. J. Obes. Relat. Metab. Disord.26, 973–977 (2002). pmid: 12080452

21. A. E. Duncan et al., Genetic and environmentalcontributions to BMI in adolescent and young adultwomen. Obesity (Silver Spring) 17, 1040–1043 (2009).doi: 10.1038/oby.2008.643; pmid: 19165159

22. M. Waldron, K. K. Bucholz, M. T. Lynskey, P. A. Madden,A. C. Heath, Alcoholism and timing of separation inparents: Findings in a midwestern birth cohort. J. Stud.Alcohol Drugs 74, 337–348 (2013). pmid: 23384382

23. Materials and methods are available as supplementarymaterial on Science Online.

24. E. Kristiansson, P. Hugenholtz, D. Dalevi,ShotgunFunctionalizeR: An R-package for functionalcomparison of metagenomes. Bioinformatics 25,2737–2738 (2009). doi: 10.1093/bioinformatics/btp508;pmid: 19696045

25. C. B. Newgard et al., A branched-chain amino acid-relatedmetabolic signature that differentiates obese and leanhumans and contributes to insulin resistance. Cell Metab.9, 311–326 (2009). doi: 10.1016/j.cmet.2009.02.002;pmid: 19356713

26. B. D. Muegge et al., Diet drives convergence in gutmicrobiome functions across mammalian phylogenyand within humans. Science 332, 970–974 (2011).doi: 10.1126/science.1198719; pmid: 21596990

27. M. J. Keenan et al., Effects of resistant starch, anon-digestible fermentable fiber, on reducing body fat.Obesity (Silver Spring) 14, 1523–1534 (2006).doi: 10.1038/oby.2006.176; pmid: 17030963

28. J. Zhou et al., Dietary resistant starch upregulates totalGLP-1 and PYY in a sustained day-long manner throughfermentation in rodents. Am. J. Physiol. Endocrinol.Metab. 295, E1160–E1166 (2008). doi: 10.1152/ajpendo.90637.2008; pmid: 18796545

29. J. Zhou et al., Failure to ferment dietary resistant starchin specific mouse models of obesity results in no bodyfat loss. J. Agric. Food Chem. 57, 8844–8851 (2009).doi: 10.1021/jf901548e; pmid: 19739641

30. D. Knights et al., Bayesian community-wide culture-independent microbial source tracking. Nat. Methods8, 761–763 (2011). doi: 10.1038/nmeth.1650;pmid: 21765408

31. J. P. Lessard, J. A. Fordyce, N. J. Gotelli, N. J. Sanders,Invasive ants alter the phylogenetic structure of antcommunities. Ecology 90, 2664–2669 (2009).doi: 10.1890/09-0503.1; pmid: 19886475

32. M. T. Johnson, J. R. Stinchcombe, An emerging synthesisbetween community ecology and evolutionary biology.Trends Ecol. Evol. 22, 250–257 (2007). doi: 10.1016/j.tree.2007.01.014; pmid: 17296244

33. C. O. Webb, N. C. Pitman, Phylogenetic balanceand ecological evenness. Syst. Biol. 51, 898–907(2002). doi: 10.1080/10635150290102609;pmid: 12554456

34. P. J. Turnbaugh et al., An obesity-associated gutmicrobiome with increased capacity for energy harvest.Nature 444, 1027–1031 (2006). doi: 10.1038/nature05414; pmid: 17183312

35. P. Gauffin Cano, A. Santacruz, A. Moya, Y. Sanz,Bacteroides uniformis CECT 7771 ameliorates metabolicand immunological dysfunction in mice with high-fat-dietinduced obesity. PLoS ONE 7, e41079 (2012).doi: 10.1371/journal.pone.0041079; pmid: 22844426

36. S. H. Duncan et al., Contribution of acetate to butyrateformation by human faecal bacteria. Br. J. Nutr. 91,915–923 (2004). doi: 10.1079/BJN20041150;pmid: 15182395

37. M. A. Mahowald et al., Characterizing a model human gutmicrobiota composed of members of its two dominantbacterial phyla. Proc. Natl. Acad. Sci. U.S.A. 106,5859–5864 (2009). doi: 10.1073/pnas.0901529106;pmid: 19321416

38. Z. Gao et al., Butyrate improves insulin sensitivity andincreases energy expenditure in mice. Diabetes 58,1509–1517 (2009). doi: 10.2337/db08-1637;pmid: 19366864

39. H. V. Lin et al., Butyrate and propionate protect againstdiet-induced obesity and regulate gut hormones via freefatty acid receptor 3-independent mechanisms. PLoS ONE7, e35240 (2012). doi: 10.1371/journal.pone.0035240;pmid: 22506074

40. H. Kuribayashi, M. Miyata, H. Yamakawa, K. Yoshinari,Y. Yamazoe, Enterobacteria-mediated deconjugation oftaurocholic acid enhances ileal farnesoid X receptorsignaling. Eur. J. Pharmacol. 697, 132–138 (2012).doi: 10.1016/j.ejphar.2012.09.048; pmid: 23051670

41. J. R. Swann et al., Systemic gut microbial modulationof bile acid metabolism in host tissue compartments.Proc. Natl. Acad. Sci. U.S.A. 108 (suppl. 1),4523–4530 (2011). doi: 10.1073/pnas.1006734107;pmid: 20837534

42. Y. Zhang et al., Loss of FXR protects against diet-inducedobesity and accelerates liver carcinogenesis in ob/obmice. Mol. Endocrinol. 26, 272–280 (2012).doi: 10.1210/me.2011-1157; pmid: 22261820

43. T. Inagaki et al., Fibroblast growth factor 15 functionsas an enterohepatic signal to regulate bile acidhomeostasis. Cell Metab. 2, 217–225 (2005).doi: 10.1016/j.cmet.2005.09.001; pmid: 16213224

44. T. Li et al., Transgenic expression of cholesterol 7a-hydroxylase in the liver prevents high-fat diet-inducedobesity and insulin resistance in mice. Hepatology 52,678–690 (2010). doi: 10.1002/hep.23721; pmid: 20623580

45. Y. Handelsman, Role of bile acid sequestrants in thetreatment of type 2 diabetes. Diabetes Care 34 (suppl. 2),S244–S250 (2011). doi: 10.2337/dc11-s237;pmid: 21525463

46. T. R. Koves et al., Peroxisome proliferator-activatedreceptor-gamma co-activator 1alpha-mediated metabolicremodeling of skeletal myocytes mimics exercise trainingand reverses lipid-induced mitochondrial inefficiency.J. Biol. Chem. 280, 33588–33598 (2005). doi: 10.1074/jbc.M507621200; pmid: 16079133

47. T. R. Koves et al., Mitochondrial overload and incompletefatty acid oxidation contribute to skeletal muscle insulinresistance. Cell Metab. 7, 45–56 (2008). doi: 10.1016/j.cmet.2007.10.013; pmid: 18177724

48. D. M. Muoio et al., Muscle-specific deletion of carnitineacetyltransferase compromises glucose tolerance andmetabolic flexibility. Cell Metab. 15, 764–777 (2012).doi: 10.1016/j.cmet.2012.04.005; pmid: 22560225

49. J. S. Bakken et al., Treating Clostridium difficile infectionwith fecal microbiota transplantation. Clin. Gastroenterol.Hepatol. 9, 1044–1049 (2011). doi: 10.1016/j.cgh.2011.08.014; pmid: 21871249

Acknowledgments: We thank D. O’Donnell, M. Karlsson,S. Wagoner, J. Guruge, R. Chamberland, M. Meier, and S. Dengfor superb technical assistance; P. Ahern for help with FACSanalysis; D. Hopper and S. Marion for their contributions to therecruitment of discordant twins from the MOAFTS cohort andfor obtaining fecal samples for the present study; I. Halatchevfor help in performing diet experiments; plus A. Reyes andN. McNulty for many helpful suggestions. S. Baumann andS. Fischer (Agilent Technologies, Santa Clara, CA) providedtechnical support for some of the metabolomics analysesincluding a gift of the Fiehn metabolite library. This work wassupported in part by grants from the NIH (DK078669,DK70977, P30-AG028716, and DK58398), the Crohn’s andColitis Foundation of America, plus Kraft Foods and MondelezInternational. Bacterial 16S rRNA pyrosequencing reads havebeen deposited in the European Bioinformatics Institute underaccession numbers ERP003513 and ERP003512, as haveshotgun pyrosequencing data sets from cecal microbialcommunity DNA (ERP003514), microbial community RNA-Seqdata sets (ERP003551), and draft genome data sets fromcultured bacterial isolates (PRJEB3728–PRJEB3780). RNA-seqdata sets have also been deposited in Gene ExpressionOmnibus (GEO) under accession number GSE48861. Authorcontributions: V.K.R. and J.I.G. designed the mouse experiments;A.C.H., A.E.D., and J.I.G. designed and implemented theclinical data collection; N.W.G., J.I.G., D.K.H., B.J.L., andM.C.M designed the NHANES-based mouse diets; V.K.R., J.J.F.,J.C., M.J.M, J.R.B., K.F., and O.I. generated the data; V.K.R.,J.J.F., F.R., J.C., J.C.C., W.V.T, L.K.U., W.A.W., A.L.K., R.K.,A.E.D., A.C.H., B.H., V.L., M.J.M., J.R.B., C.B.N., and J.I.G.analyzed the data; and V.K.R. and J.I.G. wrote the paper.

Supplementary Materialswww.sciencemag.org/cgi/content/341/6150/1241214/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S17Tables S1 to S17References (50–76)

30 May 2013; accepted 12 July 201310.1126/science.1241214

6 SEPTEMBER 2013 VOL 341 SCIENCE www.sciencemag.org1241214-10

RESEARCH ARTICLE