You are on page 1of 12

Please cite this article in press as: Parks et al.

, Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism

Resource

Genetic Control of Obesity and Gut Microbiota


Composition in Response to High-Fat,
High-Sucrose Diet in Mice
Brian W. Parks,1,* Elizabeth Nam,1 Elin Org,1 Emrah Kostem,2 Frode Norheim,1,5 Simon T. Hui,1 Calvin Pan,3
Mete Civelek,1 Christoph D. Rau,1 Brian J. Bennett,1,6 Margarete Mehrabian,1 Luke K. Ursell,7 Aiqing He,8
Lawrence W. Castellani,1 Bradley Zinker,9 Mark Kirby,9 Thomas A. Drake,4 Christian A. Drevon,5 Rob Knight,7,11
Peter Gargalovic,10 Todd Kirchgessner,10 Eleazar Eskin,2,3 and Aldons J. Lusis1,3,*
1Department of Medicine/Division of Cardiology, David Geffen School of Medicine
2Department of Computer Science
3Department of Human Genetics, David Geffen School of Medicine
4Department of Pathology and Laboratory Medicine, David Geffen School of Medicine

University of California, Los Angeles, Los Angeles, CA 90095, USA


5Department of Nutrition, Institute of Basic Medical Sciences, Faculty of Medicine, University of Oslo, Norway
6Department of Genetics, University of North Carolina, Chapel Hill, NC 27599, USA
7Department of Chemistry and Biochemistry, University of Colorado Boulder, CO 80309, USA
8Department of Applied Genomics
9Department of Metabolic Diseases-Diabetes Drug Discovery
10Department of Atherosclerosis Drug Discovery

Bristol-Myers Squibb, Princeton, NJ 08543, USA


11Howard Hughes Medical Institute, Boulder, CO, 80309, USA

*Correspondence: bparks@mednet.ucla.edu (B.W.P.), jlusis@mednet.ucla.edu (A.J.L.)


http://dx.doi.org/10.1016/j.cmet.2012.12.007

SUMMARY tions, such as type 2 diabetes, heart disease, and cancer.


Energy-rich diets containing high levels of fat and refined carbo-
Obesity is a highly heritable disease driven by hydrates along with sedentary lifestyles are believed to be the
complex interactions between genetic and environ- most significant environmental factors contributing to this
mental factors. Human genome-wide association epidemic (Finucane et al., 2011; Malik et al., 2010). Under-
studies (GWAS) have identified a number of loci standing the genetic and environmental interactions contributing
contributing to obesity; however, a major limitation to obesity is thus crucial for developing novel therapies and
preventive strategies. Human genome-wide association studies
of these studies is the inability to assess environ-
(GWAS) and studies of rare monogenic forms of obesity, as well
mental interactions common to obesity. Using
as biochemical studies with cells and animal models, have iden-
a systems genetics approach, we measured obesity tified relevant genes and pathways important in obesity;
traits, global gene expression, and gut microbiota however, given the complexity of the obesity phenotype and
composition in response to a high-fat/high-sucrose the small amount of variance that can be explained by known
(HF/HS) diet of more than 100 inbred strains of obesity alleles, it is clear that much remains to be discovered (At-
mice. Here we show that HF/HS feeding promotes tie and Scherer, 2009; Bouchard et al., 1990; Knights et al., 2011;
robust, strain-specific changes in obesity that are Sandholt et al., 2010; Speakman et al., 2011).
not accounted for by food intake and provide Obesity is strongly heritable in humans, with estimates ranging
evidence for a genetically determined set point for from 50% to 90% (Barsh et al., 2000; Stunkard et al., 1986).
obesity. GWAS analysis identified 11 genome-wide Large human GWAS explain less than 3% of this heritable
component and environmental interactions with diet composi-
significant loci associated with obesity traits, several
tion likely add significant complexity to such studies (Kilpeläinen
of which overlap with loci identified in human studies.
et al., 2011; Sonestedt et al., 2009; Speliotes et al., 2010).
We also show strong relationships between geno- Indeed, long-term overfeeding in monozygotic twins promotes
type and gut microbiota plasticity during HF/HS striking within-pair similarities in fat-mass gain, demonstrating
feeding and identify gut microbial phylotypes associ- that gene-by-diet interactions may be highly heritable and have
ated with obesity. a large impact on obesity (Bouchard et al., 1990). In addition to
host genetic contributions, the microbial community within the
gut has been shown to influence obesity in humans and mice
INTRODUCTION (Turnbaugh et al., 2006, 2009a). The gut microbiota is a trans-
missable trait and can undergo dynamic population shifts with
The dramatic increase in obesity during the past few decades is varied dietary composition (Benson et al., 2010; Turnbaugh
tightly associated with the increase in obesity-related complica- et al., 2009b; Yatsunenko et al., 2012). Obese subjects have an

Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc. 1

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

altered gut microbiota compared to lean individuals, which may strains of male mice were studied, with an average of six mice
be an important contributing factor to the obesity epidemic of each strain (Table S1 available online).
(Turnbaugh et al., 2009a). To date, very little is known about A wide distribution in body fat percentage was observed in
the genetic basis of gene-by-diet and gut microbiota-diet inter- male mice before HF/HS feeding (Figures 1B). Dietary responses,
actions to common obesogenic factors, such as the consump- as assessed by the body fat percentage increase during HF/HS
tion of energy-rich diets. feeding, varied widely among the strains (Figures 1C and S1).
Identification of genes contributing to complex traits, such as Although many strains exhibited a significant increase in body
obesity, in mice has been hampered by the poor mapping resolu- fat percentage throughout the study timeline, their individual
tion of traditional genetic crosses (Bhatnagar et al., 2011; Burrage responses differed significantly, from no change to a 200%
et al., 2010; Dokmanovic-Chouinard et al., 2008; Ehrich et al., increase in body fat percentage within the first 2 weeks (Figures
2005b; Flint et al., 2005; Lawson et al., 2011; York et al., 1999; 1C). Most strains responded during the first 4 weeks of HF/HS
Zhang et al., 1994). Based on the genome sequencing of many feeding and did not accumulate additional fat during the
mouse strains, the identification of millions of single-nucleotide remainder of the study, suggesting an upper set point whereby
polymorphisms (SNPs) (Keane et al., 2011), and the development continued body fat percentage growth is resisted (Figure 1C)
of an algorithm that corrects for population structure in associa- (Speakman et al., 2011). The large effect of the HF/HS feeding on
tion analysis (Kang et al., 2008), we recently developed a systems fat accumulation was confirmed with age-matched (16-week-
genetics resource in the mouse capable of high-resolution old) male mice (Table S2) fed a chow diet, which displayed similar
genome-wide association mapping (Bennett et al., 2010). This body fat percentage to male mice before the HF/HS diet interven-
resource, termed the hybrid mouse diversity panel (HMDP), is tion (Figure S1A). Additionally, comparison of individual male
composed of more than 100 commercially available mouse strains maintained on a chow diet or fed a HF/HS diet for 8 weeks
strains and is ideal for systems-level analyses of gene-by-envi- showed an average increase in body fat percentage from 0% to
ronment interactions. Association-based mapping approaches more than 600% (Figure S1B).
in rodents have recently been reviewed (Flint and Eskin, 2012). We observed high heritability of about 80% for body fat
Employing a systems genetics approach in the mouse, we percentage across the study timeline (Table 1). Changes in
integrated physical traits, molecular traits, and gut microbiota body fat percentage after HF/HS feeding were also highly heri-
composition data in response to an energy-rich diet. Using table (>70%), suggesting that dietary responses are strongly
GWAS rather than quantitative trait locus (QTL) analyses, we ob- controlled by genetics. Our results are consistent with the herita-
tained biologically meaningful genetic mapping, such that bility estimates for body mass index (BMI) and obesity in humans
several of the genetic loci identified contained between one (Barsh et al., 2000; Stunkard et al., 1986) and emphasize the
and three genes, comparable to human GWAS. These genes importance of genetics in controlling obesity traits, such as
were prioritized with expression QTL (eQTL) analysis, and we gene-by-diet interactions.
were able to show a significant overlap between mouse and
human GWAS loci. We measured the change in fat dynamically, Factors Contributing to Dietary Responsiveness
at five different points after high-fat/high-sucrose (HF/HS) Overconsumption of high-calorie, energy-rich foods is a key
feeding, providing strong evidence for a genetically controlled environmental factor contributing to the global obesity epidemic
body fat set point. Our use of inbred mice strains also enabled (McCaffery et al., 2012). To understand the relationship between
detailed analysis of the relationship between gut microbiota food intake and obesity, we monitored food intake, and we found
composition, obesity traits, and diet. Overall, gene-by-diet inter- it to range from 2–5 g per mouse per day. Total food intake per
actions were highly reproducible and pervasive, providing day was significantly correlated with body weight and lean
a partial explanation for the failure of human studies to explain mass (Figures 1D and 1E). In contrast, body fat percentage
a larger fraction of the genetic basis of obesity. Our results indi- and body fat percentage change after 4 weeks of HF/HS feeding
cate that mouse GWAS and systems genetics analyses provide showed little to no correlation with food intake (Figures 1F and
a powerful method to complement human studies and to 1G). This suggests that factors outside of food intake largely
address factors, such as gene-by-diet interactions, that would underlie the variation of obesity and fat mass gain between the
be difficult to study directly in humans. strains in response to HF/HS feeding.
To further define the contribution of energy consumption to the
RESULTS differences in fat accumulation, we performed in vivo metabolic
chamber analyses of five inbred strains that are the progenitors
Robust Variation in Gene-by-Diet Interactions of the recombinant inbred strains (and therefore contribute
For assessment of gene-by-diet interactions common to importantly to the overall genetic component) on chow diet.
obesity, male mice were fed ad libitum a HF/HS diet that repre- Significant strain differences were observed in total food intake,
sents a typical fast food diet, in terms of fat and refined carbohy- activity, heat production, and utilization of different energy
drates (32% kcal from fat and 25% kcal from sucrose). Mice substrates, as indicated by respiratory exchange ratio (RER)
were maintained on a chow diet (6% kcal from fat) until 8 weeks (Figures S2A–S2D), all of which can influence dietary responses
of age and subsequently placed on a HF/HS diet for 8 weeks. and subsequent fat accumulation.
Body fat percentage was assessed by magnetic resonance
imaging (MRI) every 2 weeks, and food intake was monitored GWAS and Systems Genetics Analysis
for a period of 1 week at the middle of the study timeline (study Association analysis was performed using about 100,000 infor-
schematic shown in Figure 1A). Altogether, about 100 inbred mative SNPs, spaced throughout the genome, with efficient

2 Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc.

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

A B

D E F G

Figure 1. Natural Variation in Gene-by-Diet Interactions


(A) Schematic of study design with indicated time points for HF/HS feeding (red), magnetic resonance imaging (MRI; blue), food intake monitoring (yellow), and
end of study (red).
(B) Body fat percentage in male mice (108 strains) before (red) and after (blue) 8 weeks of HF/HS feeding. Error bars (black) represent SEM.
(C) Biweekly percent body fat percentage increase in male mice with indicated body fat percentage increase after 8 weeks of HF/HS feeding.
(D–G) Correlation of food intake (grams/day/mouse) with body weight (D), lean mass (E), body fat percentage—4 weeks on HF/HS diet (F), and body fat
percentage growth—0 to 4 weeks (G), regression line (red). r, biweight midcorrelation; p, p value.
See also Figures S1 and S2 and Table S1.

mixed model association (EMMA) adjusting for population struc- and ‘‘trans’’ if the locus is outside 1 Mb. Overall, 3,960 cis and
ture (Kang et al., 2008). The threshold for genome-wide signifi- 4,496 trans eQTL were identified to have a genome-wide signif-
cance was based on simulation and permutations, as previously icance threshold (cis threshold: p < 1.4 3 10!3 and trans
described (Farber et al., 2011). This approach has been validated threshold: p < 6.1 3 10!6) (Figure S3A). Cis regulation indicates
with transgenic analyses and by comparison with linkage a potential functional genomic variation within or near a gene that
analysis (Bennett et al., 2010). Altogether, 11 genome-wide significantly influences gene expression of a given gene. For
significant loci were found to be associated with obesity traits example, Fto, the most widely replicated gene in human
(Table 2). Loci averaged 500 kb to 2 Mb in size and in most cases GWAS for obesity shows a strong cis eQTL in the adipose tissue
contained 1 to 20 genes within a linkage disequilibrium (LD) of mice (Figure S5D), indicating genetic variation of this gene.
block, an improvement of more than an order of magnitude as Global gene expression in epididymal adipose tissue was
compared to traditional linkage analysis in mice which has correlated with body fat percentage in chow fed mice (top 50
a resolution of 10 to 20 Mb (Flint et al., 2005). genes shown in Table S3). Many genes known to play a vital
In order to help identify candidate genes at loci, we carried out role in adipose biology showed significant correlations with
global expression analyses of epididymal adipose tissue in male body fat percentage. Leptin is a key adipose-derived hormone
mice (16 weeks old) fed a chow diet to determine genetic regu- correlating with adipose tissue mass (Ioffe et al., 1998) and is
lation and correlation between gene expression and body fat strongly correlated (r = 0.75; p < 2.2 3 10!16) with body fat
percentage. The loci controlling transcript levels in adipose percentage (Figure 2A). Other important adipose tissue genes,
tissue were mapped with EMMA and are referred to as expres- such as Sfrp5 (Ouchi et al., 2010) (r = 0.76; p < 2.2 3 10!16),
sion quantitative trait loci (eQTL). Loci are termed ‘‘cis’’ if the Chrebp (Herman et al., 2012) (r = !0.57; p = 2.28 3 10!9), and
locus maps within 1 Mb of the gene encoding the transcript Tmem160 (r = 0.71; p = 4.21 3 10!16), a recently identified

Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc. 3

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

Table 1. Heritability Estimates for Obesity and Dietary regulating changes in adipose tissue and indicate that Klf14
Responsiveness may also regulate dietary interactions.
Trait Heritability (%) Eight genome-wide significant loci were associated with
body fat percentage growth after 8 weeks of HF/HS feeding
Body fat percentage —0 weeks on HF/HS diet 80
(Table 2 and Figure 2C). The most significant signal
Body fat percentage —2 weeks on HF/HS diet 82
(rs31849980; p = 1.4 3 10!8) maps to chromosome 1 and
Body fat percentage —4 weeks on HF/HS diet 83 has genome-wide significant SNPs spanning a 5 Mb region
Body fat percentage —6 weeks on HF/HS diet 83 with 60 genes within LD (Table 2). A primary candidate gene
Body fat percentage —8 weeks on HF/HS diet 85 within this locus is Degs1, a fatty acid desaturase involved in
Body fat percentage growth—0 to 2 weeks 63 the metabolism of important bioactive sphingolipids (Ternes
Body fat percentage growth—0 to 4 weeks 63 et al., 2002). Degs1 expression in adipose tissue of chow fed
Body fat percentage growth—0 to 6 weeks 67 male mice is strongly correlated (within top ten genes) with
body fat percentage (r = 0.7; p = 1.3 3 10!15) (Figure 3E).
Body fat percentage growth—0 to 8 weeks 73
Previous linkage studies in mice have identified distal chromo-
Body fat percentage growth calculated by quantifying the percentage
some 1 as contributing importantly to obesity (Chen et al.,
increase of body fat after beginning HF/HS diet. Heritability calculated
as described in the Experimental Procedures.
2008) and our results greatly refine this region and suggest
Degs1 as a high-confidence candidate gene in the locus,
although given the size of the locus multiple genes may be
gene from a human GWAS for BMI (Speliotes et al., 2010), were contributing to the signal.
also found to be highly correlated with body fat percentage Of the eight loci associated with body fat percent growth after
(Figures 2B, S3B, and S3C). 8 weeks of HF/HS feeding both loci on chromosomes 16 and 18
contained genes with genome-wide cis eQTL and strong expres-
Genetic Control of Obesity and Dietary Responsiveness sion correlation with body fat percentage in epididymal adipose
Most strains in the study showed a striking increase in body fat tissue (Figure 3C). The peak SNP at chromosome 18
percentage within the first 2 weeks of HF/HS feeding (Fig- (rs30078681; p = 4.3 3 10!8) contained 26 genes within LD,
ure 1C). Association analysis with body fat percent increase and one gene, Npc1, was previously identified in a human
after 2 weeks of HF/HS feeding identified genome-wide signif- GWAS for obesity (Meyre et al., 2009). Gene expression analysis
icant loci on chromosomes 2 and 6 (Table 2 and Figure S4A). of Npc1 indicated a strong negative correlation with body fat
The chromosome 2 locus (rs13476804; p = 2.95 3 10!6) percentage (r = !0.4; p = 1.5 3 10!5) (Figure S3D) and the pres-
contains one gene within the LD block, Sptlc3, which has ence of a genome-wide significant cis eQTL (Figure S5A).
been implicated in biogenesis of sphingolipids (Demirkan Genomic sequence analysis indicated multiple nonsynonymous
et al., 2012; Hornemann et al., 2009). The locus on chromo- coding variations within the Npc1 gene (Keane et al., 2011).
some 6 contains 11 genes within LD and the peak SNP, Furthermore, Npc1 heterozygous knockout mice exhibit
(rs13478690; p = 2.8 3 10!7) is 33.5 kb upstream of Klf14, increased obesity on a high-fat diet, but not on a chow diet
a primary candidate causal gene at this locus (Table 2 and Fig- (Jelinek et al., 2010), confirming the importance of Npc1 in regu-
ure S4B). Klf14 has previously been identified in human GWAS lating dietary interactions. The chromosome 16 locus
for type 2 diabetes (Voight et al., 2010) and has recently been (rs3148854; p = 9.0 3 10!8) has eight genes within LD of the
shown to be a master regulator of gene expression in adipose genome-wide significant SNPs and one gene in this region,
tissue (Small et al., 2011). Our results support a role of Klf14 in Cbr1 (Figure 3D) is highly expressed in adipose tissue and is

Table 2. Genome-wide Significant Loci for Obesity and Dietary Responsiveness


Number of
Trait Chromosome Peak SNP Position (Mb) p Value MAF LD (Mb) Genes
Body fat percentage increase—0 to 8 weeks 1 rs31849980 183730026 1.4 3 10!8 15 178.3–184.5 60
!8
Body fat percentage increase—0 to 8 weeks 18 rs30078681 9731125 4.3 3 10 8 8.5–12.5 26
Body fat percentage increase—0 to 8 weeks 16 rs3148854 93933923 9.0 3 10!8 40 93.0–94.0 8
Body fat percentage increase—0 to 8 weeks 5 rs13478388 91288973 1.5 3 10!7 37 90.9–92.2 17
Body fat percentage increase—0 to 8 weeks 18 rs29628302 5395236 1.9 3 10!7 11 4.6–5.4 3
Body fat percentage increase—0 to 8 weeks 18 rs13483184 3796540 2.6 3 10!7 13 3.8–4.6 3
Body fat percentage increase—0 to 8 weeks 11 rs29417268 67079767 2.8 3 10!7 12 65.8–67.4 14
Body fat percentage increase—0 to 2 weeks 6 rs13478690 30872499 2.8 3 10!7 7 30.5–31.5 11
Body fat percentage increase—0 to 8 weeks 3 rs29982345 111983084 9.9 3 10!7 14 110.2–113.4 7
Body fat percentage—8 weeks 7 rs13479513 134251677 6.7 3 10!7 21 133.0–136.0 113
Body fat percentage increase—0 to 2 weeks 2 rs13476804 139322068 2.95 3 10!6 10 138.9–139.4 1
MAF, minor allele frequency; LD, linkage disequilibrium.
Indicated genome-wide significant loci for obesity traits with indicated location, position, p value, MAF, LD, and number of genes with in LD block
(described in the Experimental Procedures).

4 Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc.

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

A B Figure 2. Genetic Control of Dietary Responses


to HF/HS Feeding
(A and B) Correlation of epididymal adipose gene
expression of Leptin (Lep) (A) and Sfrp5 (B) with body
fat percentage in male HMDP mice fed a chow diet,
regression line (red). r, biweight midcorrelation; p,
p value.
(C) Manhattan plot showing the significance (–log10 of
p) of all SNPs and percent body fat percentage
increase after 8 weeks of HF/HS feeding in male
HMDP mice. Candidate genes for genome-wide
significant loci are indicated above genome-wide
C significant loci. Genome-wide significant threshold
(red) of p = 4.1 3 10!06 is indicated.
(D) Locus plot for genome-wide significant association
at chromosome 16qC4 with approximate LD block
(shaded in gray) and genome-wide significant SNPs
(yellow) with the peak SNP (red).
(E and F) Correlation of epididymal adipose gene
expression of Degs1 (E) and Cbr1 (F) with body fat
percentage in male HMDP mice fed a chow diet,
regression line (red). r, biweight midcorrelation; p,
p value.
See also Figures S3–S5 and Tables S2 and S3.

ants (Keane et al., 2011). Cbr1 reduces


carbonyl compounds via NADPH-depen-
D dent oxidoreductase activity and through
its glutathione (GSH)-binding site Cbr1 can
act on GSH-conjugated molecules and
play a role in controlling oxidative stress
(Bateman et al., 2008), an important compo-
nent of obesity (Furukawa et al., 2004).
The loci for body fat growth after 8 weeks
of HF/HS feeding also contained genes with
known links to adipose biology and metab-
olism. The locus on chromosome 5
(rs13478388; p = 1.5 3 10!7) contains
several C-X-C motif chemokine genes
within LD. Cxcl5 is one of the genes in this
cluster (Figure S4C), and expression of
Cxcl5 is increased in adipose tissue of
obese human subjects (Chavey et al.,
2009). Cxcl5 contains nonsynonymous
E F coding polymorphisms and an alternative
splice site among multiple mouse strains
(Keane et al., 2011), making Cxcl5 a strong
candidate at the chromosome 5 locus.
Similarly, the significant SNP at chromo-
some 3 (rs29982345; p = 9.9 3 10!7) is
within 1 Mb of three amylase genes,
Amy2b, Amy2a5, and Amy1 (Figure S4D).
Amylases play a critical role in the break-
down of polysaccharides to sugars, and
targeting amylases has been suggested as
a potential method for treating obesity
(Barrett and Udani, 2011). The significant
negatively correlated (r = !0.5; p = 1.6 3 10!7) with obesity peak on chromosome 11 (rs29417268; p = 2.8 3 10!7) has 11
(Figure 2F). Furthermore, Cbr1 has a cis eQTL (p = 1.1 3 10!4) genes within LD, and one gene, Glp2r, plays an important role
(Figure S5B) and contains multiple nonsynonymous coding vari- in intestinal homeostasis and feeding behavior (Guan et al.,

Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc. 5

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

A B HF/HS
Chow Diet HF/HS Diet Chow

Actinobacteria
Bacteroidetes
Firmicutes
Other
Proteobacteria
Tenericutes
Verrucomicrobia

Figure 3. Robust Shifts in Gut Microbiota Composition after HF/HS Feeding


(A) Relative abundances of the different phyla after chow diet and HF/HS feeding (average among 52 matched strains).
(B) Principal coordinates analysis (PCoA) plot of the unweighted UniFrac distances. Each circle representing a different mice strain is colored according to the
dietary conditions. PC1, PC2, and PC3 values for each mouse sample are plotted; percent variation explained by each PC is shown in parentheses.
(C) Linear discriminant analysis (LDA) coupled with effect size measurements identifies the most differentially abundant taxons between chow and HF/HS diets.
HF/HS-diet-enriched taxa are indicated with a positive LDA score (green), and taxa enriched in normal chow diet have a negative score (red). Only taxa meeting an
LDA significant threshold >2 are shown.
See also Figure S6 and Table S4.

2012). We also observed genome-wide significant loci for dietary Conservation of Human and Mouse Metabolic Loci
response at two positions on chromosome 18 (rs13483184; p = A comparison of the genes/loci identified in our study with GWAS
2.6 3 10!7 and rs29628302; p = 2.6 3 10!7) with no obvious genes for human obesity and BMI revealed significant overlap
candidate genes (Table 2). (p = 4.0 3 10!4) (described in the Experimental Procedures). In
In addition to loci influencing dietary response, there was one addition to genome-wide significant loci, we analyzed significant
genome-wide significant locus at chromosome 7 associated human GWAS genes in our association results and identified
with body fat percentage after 8 weeks of HF/HS feeding. suggestive associations near Opcml and Sorl1 (Table 3). Both
This locus was gene-rich and contained 113 genes within LD Opcml and Sorl1 have been found in large human GWAS for
of the peak SNP (rs13479513; p = 6.7 3 10!7). Two genes at visceral adipose/subcutaneous adipose ratio and waist circum-
this locus, Atp2a1 and Apob48r, have previously been identi- ference, respectively (Fox et al., 2012; Smith et al., 2010). Our
fied in human GWAS for BMI (Speliotes et al., 2010; Thorleifs- results demonstrate that genes associated with obesity and
son et al., 2009). Furthermore, this locus contains Sephs2, BMI in humans show functional variation in mice and are associ-
a strong candidate gene with both a significant cis eQTL (p = ated with dietary interactions and obesity.
1.4 3 10!4) and an exceptionally strong negative correlation
with body fat percentage (r = 0.68; p = 2.0 3 10!14) (Figures Contributions of Gut Microbiota to Obesity
S3E and S5C). Sephs2 is an evolutionarily conserved protein A number of recent studies have implicated gut microbiota in
that is essential for production of selenocysteine (Xu et al., metabolic and cardiovascular diseases (Turnbaugh et al.,
2007), a key amino acid found in a diverse group of proteins 2006; Vijay-Kumar et al., 2010; Wang et al., 2011). In order to
with roles in the pathology of metabolic syndrome and diabetes investigate the genetic contributions to the gut microbiome
(Rayman, 2012). and their relationship to obesity and dietary responses, we

6 Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc.

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

Table 3. Significant Overlap of Mouse and Human GWAS


Candidate
Mouse Trait Chromosome Peak SNP p Value Gene Corresponding Human GWAS Trait
Body fat percentage increase—0 to 8 weeks on 18 rs30078681 4.33 3 10!8 Npc1 Obesity
HF/HS diet
Body fat percentage increase—0 to 2 weeks on 6 rs13478690 2.75 3 10!7 Klf14 Type 2 diabetes, HDL cholesterol
HF/HS diet
Body fat percentage—8 weeks on HF/HS diet 7 rs13479513 6.7 3 10!7 Atp2a1 BMI, weight
Body fat percentage—8 weeks on HF/HS diet 7 rs13479513 6.7 3 10!7 Apob48r BMI
Body fat percentage increase—0 to 2 weeks on 9 rs13459107 1.01 3 10!5 Opcml Visceral adipose/subcutaneous
HF/HS diet adipose ratio
Body fat percentage—0 weeks on HF/HS diet 9 rs30236502 4.90 3 10!5 Sorl1 Waist circumference
Identified mouse genetic loci containing genes previously identified in human GWAS studies for obesity and related traits. The corresponding mouse
trait, location, p value, candidate gene, and human GWAS trait is indicated.

examined the gut microbiota composition and distribution in To test the relationship between obesity, dietary responsive-
strains after chow and HF/HS feeding. Since the microbiome is ness, and gut microbiota composition, we analyzed all 11
highly affected by environmental factors, all mice in the study genome-wide significant loci with the abundances of specific
were bred in the same facility and each strain was maintained gut phylotypes. Only the chromosome 3 locus (rs29982345),
in separate cages. We observed significant phylum-level shifts which is associated with body fat growth after 8 weeks of HF/
in the gut microbiota composition after HF/HS feeding (Fig- HS feeding showed a significant enrichment (p = 0.0033) of the
ure 3A). The shifts between the two dominant phyla Bacteroi- genera Enterobacteriaceae (phylum Proteobacteria). This region
detes and Firmicutes after HF/HS feeding (Figure 3A and Table of chromosome 3, which contains three amylase genes, may
S4) are consistent with previous reports (Turnbaugh et al., therefore contribute to both dietary responsiveness and gut
2009b). Principal coordinates analysis (PCoA) of unweighted microbiota composition. We also examined the relationship
UniFrac (similar results for weighted UniFrac shown in between specific gut phylotypes and obesity traits and observed
Figure S6B) (Lozupone and Knight, 2005) distances detected a modest but statistically significant negative correlation with the
two clear clusters corresponding to the two dietary conditions abundance of the genera Akkermansia (phylum Verrucomicrobia)
(Figure 3B), suggesting that HF/HS feeding dramatically alters and body fat percentage growth after 8 weeks of HF/HS feeding
gut microbial communities across a variety of genetic back- (Figure 4C). Furthermore, body fat percentage growth after HF/
grounds. Compared to mice on a chow diet, the HF/HS-fed HS feeding was positively correlated with the relative abun-
mice had greater abundance of several genera classified to order dances of Lactococcus from phylum Firmicutes (Figure 4D)
Clostridales in Phyla Firmicutes and lower abundances of Bac- and with the genera Allobaculum (phylum Bacteroidetes) (Fig-
teroidetes, classified to family Porphyromonadaceae (41% ure 4E). Additional studies are warranted to validate the connec-
chow versus 18% HF/HS, p = 4.36 3 10!11). In total we identified tions of these specific gut microbiota and dietary interactions.
17 genera whose abundance was significantly changed by diet
(p < 0.01) (Figure 3C and Table S4). Interestingly, the abundance DISCUSSION
of Akkermansia genera of the Verrucomicrobia phylum explains
much of the unweighted UniFrac distance clusters on PCoA Our results have led to several findings about the genetic control
plots in HF/HS diet (Figures S6C and S6D). of responses to energy-rich diets consumed by an increasing
Robust effects of diet on the gut microbiota are well docu- fraction of the world’s population. Although biochemical and
mented and recent evidence indicates that the gut microbiota genetic studies have revealed important regulators of body fat,
constitute a complex polygenic trait under genetic regulation including feedback pathways that modulate food intake and
(Benson et al., 2010; McKnite et al., 2012; Qin et al., 2012). energy expenditure, these studies have not revealed how
Consistent with the idea that the gut microbiota is influenced common genetic variation controls these processes. Human
by genetic factors, we observed a strong effect of genetic back- GWAS have been successful at identifying many loci controlling
ground on the composition and plasticity of the gut microbiota measures such as body mass index, but they have been unable
after HF/HS feeding. Some strains showed large shifts in all to examine gene-by-environment interactions. Classic linkage
major phyla, while other strains had little fluctuation in their gut studies in mice, on the other hand, have been able to partially
microbiota after HF/HS feeding (Figure 4A). Additionally, address issues such as dietary responsiveness (Ehrich et al.,
Procrustes analysis was used to compare the relative orientation 2005b) but have been hindered by poor mapping resolution,
of matched strains after HF/HS and chow feeding and confirmed and thus only a handful of genes have been identified (Buchner
that the plasticity of gut microbial community is highly individual- et al., 2012; Dokmanovic-Chouinard et al., 2008; Ehrich et al.,
ized (Figure 4B). Taken together, these results demonstrate 2005a; York et al., 1999; Zhang et al., 1994). Our results add to
a profound effect of HF/HS feeding on gut microbiota and this body of knowledge in several important ways. First, the
show the strong influences of host genetics on influencing the improved mapping resolution enabled us to identify several
plasticity of the gut microbiota in response to altered dietary candidate genes for gene-by-environment interactions in
compositions. obesity. Second, because we examined obesity traits in

Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc. 7

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

A B

C D E

Figure 4. Plasticity of Gut Microbiota Is Strain Specific


(A) Gut Microbiota phyla shift in 52 strains after HF/HS feeding for 8 weeks indicated by shift in percent composition of indicated phylum.
(B) Procrustes analysis of the same strains on chow diet and the HF/HS diet are linked with a bar. The unweighted UniFrac distances between the diets varies
across strains, but are in general long, highlighting the shifts in microbial composition in response to diet.
(C–E) Correlation of relative abundance of Akkermansia (A), Lactococcus (B), and Allobaculum (C) with body fat percentage increase from 0 to 8 weeks in male
mice fed a HF/HS diet, regression line (red). r, biweight midcorrelation; p, p value.
See also Figure S6.

a dynamic way, studying the mice after dietary challenge at for the loci containing larger numbers of candidates, examination
multiple time points, we were able to provide strong evidence of transcript levels in adipose tissue and access to complete
for the concept of a biologic set point. Third, we have integrated DNA sequence information allowed us to prioritize candidates.
the obesity traits with global adipose transcript levels and gut Indeed, several of the genes at such loci (Npc1 and Glp2r)
microbiota composition. Fourth, our results emphasize the have already been studied via transgenic approaches, with
importance of gene-by-environment interactions, with important results consistent with our findings (Guan et al., 2012; Jelinek
implications for an understanding of the overall genetic architec- et al., 2010).
ture of obesity. Finally, our results provide the basis of a systems Several different models have been developed to understand
genetics resource for obesity traits that can be expanded to how genes and environment combine to regulate body fat.
include multiple biologic scales using metabalomics, proteo- However, only limited studies have addressed the genetic basis
mics, and epigenetics. In particular it will be of interest to of gene-by-environment interactions in obesity. The dual inter-
examine behavioral and neurological differences among the vention point model postulates an upper and lower limit where
strains as they relate to obesity traits. genetic factors become dominant, and between which there is
Using association rather than linkage, we were able to obtain only weak genetic regulation (Speakman et al., 2011). Our results
biologically significant mapping data, as several of the loci iden- strongly support this concept, because after 4 weeks of HF/HS
tified for the response to a HF/HS diet contained between one feeding most mouse strains reach an upper limit where there is
and three genes, comparable to human GWAS results. Alto- no further increase in body fat (Figure 1). Thus, there must be
gether, we examined only about 800 mice, much fewer than strong genetic factors that resist continued fat accumulation
the tens of thousands examined in human GWAS. It is likely beyond a certain point. Clearly, this set point varies widely in
that power is greatly increased in mice, as compared to humans, the population; some strains have already achieved the set-point
because of the ability to accurately monitor the phenotypes level on a chow diet, whereas a few strains continue to accumu-
(body fat and food intake) and control the environment. Also, late fat throughout the 8 week feeding period, suggesting that

8 Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc.

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

the set-point mechanism is disrupted. It is noteworthy that will be of interest, for example, to examine additional interme-
obesity prevalence appears to be leveling off in developed coun- diate phenotypes, such as protein and metabolite levels, in the
tries, which is also consistent with the idea of a set point (Flegal same set of strains, and to relate these to the obesity traits using
et al., 2012). Based on our data (Table 1) and a monozygotic twin association mapping, correlation, and modeling. In this study,
study (Bouchard et al., 1990), acute body fat changes in we examined transcript levels in adipose tissue and the compo-
response to an obesogenic environment are under strong sition of gut microbiota, and in the future it would be valuable to
genetic regulation, although this may be part of the same examine transcript levels in other relevant tissues, such as the
pathway that determines the set point. Our data also show that hypothalamus, liver, and muscle, as well as protein, metabolite
food intake is only modestly correlated with obesity and dietary levels, and epigenetic markers. Moreover, traits such as epige-
responsiveness and indicates that energy expenditure may be netic variation and gut microbiota composition can be tested
a likely mechanism contributing to the obesity set point. Further for effects on obesity-related traits in this set of strains.
metabolic studies of strains at the extremes will be required to
confirm the effect of energy expenditure on obesity set points. EXPERIMENTAL PROCEDURES
The experimental design of our study, involving inbred strains
Animals
of mice, allowed us to study the gut microbiota composition of
All mice were obtained from The Jackson Laboratory and were bred at Univer-
the mice both on a chow and HF/HS diet. The HF/HS diet clearly sity of California, Los Angeles to generate mice used in this study. Mice were
influenced gut microbiota composition in all of the 52 strains maintained on a chow diet (Ralston Purina Company) until 8 weeks of age,
examined on both chow and the HF/HS diet (Figure 3). Shifts in when they were given a high-fat, high-sucrose diet (Research Diets-
gut microbiota composition and diversity in response to dietary D12266B) with the following composition: 16.8% kcal protein, 51.4% kcal
changes are similar to those previously noted in both human carbohydrate, and 31.8% kcal fat. Age-matched control male mice were fed
a chow diet (18% kcal fat) for 16 weeks. A complete list of the strains included
(Ley et al., 2006; Turnbaugh et al., 2009b) and mouse studies
in our study is included in Tables S1 and S2. The animal protocol for the study
(Ley et al., 2005; Turnbaugh et al., 2006). Our data indicate that was approved by the Institutional Care and Use Committee (IACUC) at Univer-
host genetic factors influence gut microbiota plasticity in sity of California, Los Angeles.
response to diet; however, since the gut microbiota can vary
significantly even in well-controlled cohorts we cannot exclude Association Analysis and Heritability Calculations
the possibility that other factors could partly explain the We performed the association testing of each SNP with a linear mixed model,
which accounts for the population structure among the n animals using the
observed differences. We observed significant relationships
following model (Kang et al., 2008):
between gut microbiota and metabolic traits (Figures 4C–4E).
The results are consistent with the concept that gut microbiota y = 1n m + xb + u + e;
contribute to systemic functions and common diseases where m is the mean, b is the allele effect of the SNP, x is the (nx1) vector of
(Clemente et al., 2012). Future studies in the HMDP have the observed genotypes of the SNP, u is the random effects due to genetic relat-
potential to test the roles of specific gut microbiota using trans- edness with varðuÞ = s2u K, and e is the random noise with varðeÞ = s2e I. K
denotes the identity-by-state kinship matrix estimated from all the SNPs, I
plantation and cross-fostering to systematically examine the
denotes the (nxn) identity matrix, and 1n is the (nx1) vector of ones. We esti-
relationship between genetic background, diet, and gut micro- mated s2u and s2e using restricted maximum likelihood (REML) and computed
biota composition. p values using the standard F test to test the null hypothesis b = 0. Genome-
Recent successes in human GWAS for common diseases wide significance threshold and genome-wide association mapping are deter-
have yielded many genetic loci, but in most instances they mined as the family-wise error rate as the probability of observing one or more
have revealed only a very small fraction of the estimated herita- false positives across all SNPs for phenotype. We ran 100 different sets of
bility (Lander, 2011). While it is likely that some of the ‘‘missing permutation tests and parametric boot strapping of size 1,000 and observed
that the mean and standard error of the genome-wide significance threshold
heritability’’ will be found in rare variants in the population, our
at the family-wise error rate of 0.05 were 3.9 3 10!6 ± 0.3 3 10!6, and
results, and those from other studies in animal models (Burrage 4.0 3 10!6 ± 0.3 3 10!6, respectively. A detailed explanation of the analyses
et al., 2010; Lawson et al., 2011; Shao et al., 2008) suggest that is provided in Bennett et al. (2010). LD was determined by calculated pairwise
genetic interactions are likely to have important contributions. r2 SNP correlations for each chromosome. Approximate LD boundaries were
Clearly, human studies have limited ability to identify gene-by- determined by visualizing r2 > 0.8 correlations in MATLAB (MathWorks).
environment or gene-by-gene interactions (Zuk et al., 2012), Heritability (in the narrow sense) is defined as the phenotypic variance
explained by additive genetic effects and is computed using the following
and our results suggest that GWAS in mice will provide a power-
model (Yang et al., 2010):
ful complement to human studies. A key feature of our experi-
mental design is that genetically identical mice can be studied y = 1n m + u + e
under multiple environmental conditions. Also, the greatly We estimated s2uand s2e using REML and calculated the heritability (h2) for
improved resolution of association as compared to classical each trait as follows:
linkage studies in mice greatly facilitates gene identification
varðuÞ s2u trðPKÞ
and mechanistic understanding. The substantial overlap of our h2 = = ;
varðuÞ + varðeÞ s2u trðPKÞ + s2e
study with human GWAS for obesity traits suggests that at least
for metabolic disorders, the mouse is highly relevant and that where tr(.) is the matrix trace and P = ðI ! ð1=nÞ1n 1Tn Þ.
natural variations of the two species affect shared pathways.
Body Composition Analysis
Because the inbred strains of the HMDP are permanent, data
Animals were measured for total body fat mass and lean mass by MRI with
obtained about the strains is cumulative, allowing integration Bruker Minispec and software from Eco Medical Systems, Houston, TX
with previous results. The data compiled here provides a basis (Taicher et al., 2003). All animals in the study were measured at 0, 2, 4, 6,
for further genetic studies of obesity-related traits in mice. It and 8 weeks of HF/HS feeding.

Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc. 9

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

Assessment of Food Intake ACKNOWLEDGMENTS


After 4 weeks of HF/HS feeding, food intake was monitored for 5 days. Ninety
to one hundred grams of food was weighed and put in the food hopper, after We thank Hannah Qi, Zhiqiang Zhou, Judy Wu, and Tieyan Han for expert
24 hr food was weighed, and then weighed again after 24 hr. The procedure assistance with mouse experiments. We thank Raffi Hagopian for help with
was repeated again with fresh food, and a total of 4 3 24 hr period total preparing figures and Luz Orozco for statistical advice. This work was sup-
food weight consumed was calculated by subtracting food weights; grams ported by National Institutes of Health (NIH) grants HL028481 and
per mouse was then calculated by dividing average total food weight DK094311 to A.J.L and by a Howard Hughes Medical Institute Early Career
consumed in 4 3 24 hr periods by number of mice in cage. Scientist award to R.K. B.W.P. was supported by NIH training grant T32-
HD07228. E.O. was supported by a MOBILITAS Postdoctoral Research Grant
In Vivo Metabolism Assessment (MJD252). C.D.R. was supported by NIH training grant T32-HL69766. M.C.
Metabolic rate, activity, food and water consumption were assessed with was supported by an American Heart Association postdoctoral fellowship
a Columbus Instruments Comprehensive Lab Animal Monitoring System (10POST3660048). L.K.U. was supported by NIH training grant T32-
(CLAMS; Columbus Instruments, Columbus, OH) equipped with subsystems GM08759. A.H., B.Z., M.K., P.G., and T.K. are employees and shareholders
for open circuit indirect calorimetry, and monitoring of activity, feeding, and of Bristol-Myers Squibb.
drinking. Detailed methods for in vivo metabolism assessment provided in
the Supplemental Experimental Procedures. Received: July 24, 2012
Revised: November 5, 2012
Accepted: December 12, 2012
Gut Microbiota Analysis
Published: January 8, 2013
Microbial community composition was assessed by pyrosequencing 16S
rRNA genes derived from the cecal samples of chow diet and HF/HS diet
animals. All mice used in the study were bred at UCLA to ensure uniform envi- REFERENCES
ronmental conditions. Furthermore, strains were individually housed within the
same vivarium throughout the duration of the study. To ensure comprehensive Attie, A.D., and Scherer, P.E. (2009). Adipocyte metabolism and obesity.
analysis, we sequenced biological replicates on all mice included in the study. J. Lipid Res. 50(Suppl ), S395–S399.
Biological replicates were carried out on all mice included in the study. For Barrett, M.L., and Udani, J.K. (2011). A proprietary alpha-amylase inhibitor
mice fed a HF/HS diet, we sequenced a range of two to five mice per strain, from white bean (Phaseolus vulgaris): a review of clinical studies on weight
with an average of three mice per strain (n = 339 male mice from 110 strains). loss and glycemic control. Nutr. J. 10, 24.
Chow diet mice were sequenced by pooling of two to four mice per strain, with
Barsh, G.S., Farooqi, I.S., and O’Rahilly, S. (2000). Genetics of body-weight
an average of two per strain (n = 108 male mice from 52 strains). We also
regulation. Nature 404, 644–651.
sequenced technical replicates on eight mouse strains (five strains on chow
diet and three on HF/HS diet) and observed high reproducibility of UniFrac Bateman, R.L., Rauh, D., Tavshanjian, B., and Shokat, K.M. (2008). Human
analysis between technical replicates (Figure S6A). Detailed methods for gut carbonyl reductase 1 is an S-nitrosoglutathione reductase. J. Biol. Chem.
microbiota analysis provided in the Supplemental Experimental Procedures. 283, 35756–35762.
Bennett, B.J., Farber, C.R., Orozco, L., Kang, H.M., Ghazalpour, A., Siemers,
Adipose RNA Isolation and Global Gene Expression Analysis N., Neubauer, M., Neuhaus, I., Yordanova, R., Guan, B., et al. (2010). A high-
Flash frozen epididymal adipose samples from male mice (16 weeks old) main- resolution association mapping panel for the dissection of complex traits in
tained on a chow diet were weighed and homogenized in Qiazol (QIAGEN) and mice. Genome Res. 20, 281–290.
RNA isolated according to the manufacturer’s protocol with RNeasy columns Benson, A.K., Kelly, S.A., Legge, R., Ma, F., Low, S.J., Kim, J., Zhang, M., Oh,
(QIAGEN). Isolated RNA (two mice per strain as indicated in Table S2) was P.L., Nehrenberg, D., Hua, K., et al. (2010). Individuality in gut microbiota
analyzed for global gene expression with Affymetrix HT_MG430A arrays and composition is a complex polygenic trait shaped by multiple environmental
was filtered as described (Bennett et al., 2010). and host genetic factors. Proc. Natl. Acad. Sci. USA 107, 18933–18938.
Bhatnagar, S., Oler, A.T., Rabaglia, M.E., Stapleton, D.S., Schueler, K.L.,
Expression eQTL Analysis Truchan, N.A., Worzella, S.L., Stoehr, J.P., Clee, S.M., Yandell, B.S., et al.
GWAS for gene expression was performed with EMMA and defined as cis if (2011). Positional cloning of a type 2 diabetes quantitative trait locus; tomo-
peak SNP mapped within 1 Mb of gene position and trans if mapped outside syn-2, a negative regulator of insulin secretion. PLoS Genet. 7, e1002323.
(cis threshold, p < 1.4 3 10!3; trans threshold, p < 6.1 3 10!6).
Bouchard, C., Tremblay, A., Després, J.P., Nadeau, A., Lupien, P.J., Thériault,
G., Dussault, J., Moorjani, S., Pinault, S., and Fournier, G. (1990). The response
Statistics to long-term overfeeding in identical twins. N. Engl. J. Med. 322, 1477–1482.
Correlations were calculated with the biweight midcorrelation, which is robust
to outliers (Wilcox, 2005). Overlap p value between mouse and human GWAS Buchner, D.A., Geisinger, J.M., Glazebrook, P.A., Morgan, M.G., Spiezio, S.H.,
genes was calculated with hypergeometric cumulative distribution function in Kaiyala, K.J., Schwartz, M.W., Sakurai, T., Furley, A.J., Kunze, D.L., et al.
MATLAB (MathWorks) with pval = 1 – hygecdf(x,M,K,N), where x is the total (2012). The juxtaparanodal proteins CNTNAP2 and TAG1 regulate diet-
number of genes in genome, M is the total number of genes identified in mouse induced obesity. Mamm. Genome 23, 431–442.
and humans, K s the number of genes identified in obesity and BMI human Burrage, L.C., Baskin-Hill, A.E., Sinasac, D.S., Singer, J.B., Croniger, C.M.,
GWAS, and N is the number of genes identified in mouse obesity traits. Kirby, A., Kulbokas, E.J., Daly, M.J., Lander, E.S., Broman, K.W., and
Nadeau, J.H. (2010). Genetic resistance to diet-induced obesity in chromo-
some substitution strains of mice. Mamm. Genome 21, 115–129.
ACCESSION NUMBERS
Chavey, C., Lazennec, G., Lagarrigue, S., Clapé, C., Iankova, I., Teyssier, J.,
All microarray data from this study are deposited in the NCBI GEO database Annicotte, J.S., Schmidt, J., Mataki, C., Yamamoto, H., et al. (2009). CXC
(http://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE42890. ligand 5 is an adipose-tissue derived factor that links obesity to insulin resis-
tance. Cell Metab. 9, 339–349.

SUPPLEMENTAL INFORMATION Chen, Y., Zhu, J., Lum, P.Y., Yang, X., Pinto, S., MacNeil, D.J., Zhang, C.,
Lamb, J., Edwards, S., Sieberts, S.K., et al. (2008). Variations in DNA elucidate
Supplemental Information includes Supplemental Experimental Procedures, molecular networks that cause disease. Nature 452, 429–435.
six figures, and four tables and can be found with this article online at http:// Clemente, J.C., Ursell, L.K., Parfrey, L.W., and Knight, R. (2012). The impact of
dx.doi.org/10.1016/j.cmet.2012.12.007. the gut microbiota on human health: an integrative view. Cell 148, 1258–1270.

10 Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc.

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

Demirkan, A., van Duijn, C.M., Ugocsai, P., Isaacs, A., Pramstaller, P.P., Keane, T.M., Goodstadt, L., Danecek, P., White, M.A., Wong, K., Yalcin, B.,
Liebisch, G., Wilson, J.F., Johansson, A., Rudan, I., Aulchenko, Y.S., et al.; Heger, A., Agam, A., Slater, G., Goodson, M., et al. (2011). Mouse genomic
DIAGRAM Consortium; CARDIoGRAM Consortium; CHARGE Consortium; variation and its effect on phenotypes and gene regulation. Nature 477,
EUROSPAN consortium (2012). Genome-wide association study identifies 289–294.
novel loci associated with circulating phospho- and sphingolipid concentra- Kilpeläinen, T.O., Qi, L., Brage, S., Sharp, S.J., Sonestedt, E., Demerath, E.,
tions. PLoS Genet. 8, e1002490. Ahmad, T., Mora, S., Kaakinen, M., Sandholt, C.H., et al. (2011). Physical
Dokmanovic-Chouinard, M., Chung, W.K., Chevre, J.C., Watson, E., Yonan, activity attenuates the influence of FTO variants on obesity risk: a meta-anal-
J., Wiegand, B., Bromberg, Y., Wakae, N., Wright, C.V., Overton, J., et al. ysis of 218,166 adults and 19,268 children. PLoS Med. 8, e1001116.
(2008). Positional cloning of ‘‘Lisch-Like’’, a candidate modifier of susceptibility Knights, D., Parfrey, L.W., Zaneveld, J., Lozupone, C., and Knight, R. (2011).
to type 2 diabetes in mice. PLoS Genet. 4, e1000137. Human-associated microbial signatures: examining their predictive value.
Ehrich, T.H., Hrbek, T., Kenney-Hunt, J.P., Pletscher, L.S., Wang, B., Cell Host Microbe 10, 292–296.
Semenkovich, C.F., and Cheverud, J.M. (2005a). Fine-mapping gene-by-diet Lander, E.S. (2011). Initial impact of the sequencing of the human genome.
interactions on chromosome 13 in a LG/J x SM/J murine model of obesity. Nature 470, 187–197.
Diabetes 54, 1863–1872.
Lawson, H.A., Cady, J.E., Partridge, C., Wolf, J.B., Semenkovich, C.F., and
Ehrich, T.H., Kenney-Hunt, J.P., Pletscher, L.S., and Cheverud, J.M. (2005b). Cheverud, J.M. (2011). Genetic effects at pleiotropic loci are context-depen-
Genetic variation and correlation of dietary response in an advanced intercross dent with consequences for the maintenance of genetic variation in popula-
mouse line produced from two divergent growth lines. Genet. Res. 85, tions. PLoS Genet. 7, e1002256.
211–222.
Ley, R.E., Bäckhed, F., Turnbaugh, P., Lozupone, C.A., Knight, R.D., and
Farber, C.R., Bennett, B.J., Orozco, L., Zou, W., Lira, A., Kostem, E., Kang, Gordon, J.I. (2005). Obesity alters gut microbial ecology. Proc. Natl. Acad.
H.M., Furlotte, N., Berberyan, A., Ghazalpour, A., et al. (2011). Mouse Sci. USA 102, 11070–11075.
genome-wide association and systems genetics identify Asxl2 as a regulator Ley, R.E., Turnbaugh, P.J., Klein, S., and Gordon, J.I. (2006). Microbial ecology:
of bone mineral density and osteoclastogenesis. PLoS Genet. 7, e1002038. human gut microbes associated with obesity. Nature 444, 1022–1023.
Finucane, M.M., Stevens, G.A., Cowan, M.J., Danaei, G., Lin, J.K., Paciorek, Lozupone, C., and Knight, R. (2005). UniFrac: a new phylogenetic method for
C.J., Singh, G.M., Gutierrez, H.R., Lu, Y., Bahalim, A.N., et al.; Global comparing microbial communities. Appl. Environ. Microbiol. 71, 8228–8235.
Burden of Metabolic Risk Factors of Chronic Diseases Collaborating Group
Malik, V.S., Popkin, B.M., Bray, G.A., Després, J.P., and Hu, F.B. (2010).
(Body Mass Index) (2011). National, regional, and global trends in body-
Sugar-sweetened beverages, obesity, type 2 diabetes mellitus, and cardio-
mass index since 1980: systematic analysis of health examination surveys
vascular disease risk. Circulation 121, 1356–1364.
and epidemiological studies with 960 country-years and 9$1 million partici-
pants. Lancet 377, 557–567. McCaffery, J.M., Papandonatos, G.D., Peter, I., Huggins, G.S., Raynor, H.A.,
Delahanty, L.M., Cheskin, L.J., Balasubramanyam, A., Wagenknecht, L.E.,
Flegal, K.M., Carroll, M.D., Kit, B.K., and Ogden, C.L. (2012). Prevalence of
and Wing, R.R.; Genetic Subgroup of Look AHEAD; Look AHEAD Research
obesity and trends in the distribution of body mass index among US adults,
Group (2012). Obesity susceptibility loci and dietary intake in the Look
1999-2010. JAMA 307, 491–497.
AHEAD Trial. Am. J. Clin. Nutr. 95, 1477–1486.
Flint, J., and Eskin, E. (2012). Genome-wide association studies in mice. Nat.
McKnite, A.M., Perez-Munoz, M.E., Lu, L., Williams, E.G., Brewer, S., Andreux,
Rev. Genet. 13, 807–817.
P.A., Bastiaansen, J.W., Wang, X., Kachman, S.D., Auwerx, J., et al. (2012).
Flint, J., Valdar, W., Shifman, S., and Mott, R. (2005). Strategies for mapping Murine gut microbiota is defined by host genetics and modulates variation
and cloning quantitative trait genes in rodents. Nat. Rev. Genet. 6, 271–286. of metabolic traits. PLoS ONE 7, e39191.
Fox, C.S., Liu, Y., White, C.C., Feitosa, M., Smith, A.V., Heard-Costa, N., Meyre, D., Delplanque, J., Chèvre, J.C., Lecoeur, C., Lobbens, S., Gallina, S.,
Lohman, K., Johnson, A.D., Foster, M.C., Greenawalt, D.M., et al.; GIANT Durand, E., Vatin, V., Degraeve, F., Proença, C., et al. (2009). Genome-wide
Consortium; MAGIC Consortium; GLGC Consortium (2012). Genome-wide association study for early-onset and morbid adult obesity identifies three
association for abdominal subcutaneous and visceral adipose reveals a novel new risk loci in European populations. Nat. Genet. 41, 157–159.
locus for visceral fat in women. PLoS Genet. 8, e1002695.
Ouchi, N., Higuchi, A., Ohashi, K., Oshima, Y., Gokce, N., Shibata, R., Akasaki,
Furukawa, S., Fujita, T., Shimabukuro, M., Iwaki, M., Yamada, Y., Nakajima, Y., Y., Shimono, A., and Walsh, K. (2010). Sfrp5 is an anti-inflammatory adipokine
Nakayama, O., Makishima, M., Matsuda, M., and Shimomura, I. (2004). that modulates metabolic dysfunction in obesity. Science 329, 454–457.
Increased oxidative stress in obesity and its impact on metabolic syndrome.
Qin, J., Li, Y., Cai, Z., Li, S., Zhu, J., Zhang, F., Liang, S., Zhang, W., Guan, Y.,
J. Clin. Invest. 114, 1752–1761.
Shen, D., et al. (2012). A metagenome-wide association study of gut micro-
Guan, X., Shi, X., Li, X., Chang, B., Wang, Y., Li, D., and Chan, L. (2012). GLP-2 biota in type 2 diabetes. Nature 490, 55–60.
receptor in POMC neurons suppresses feeding behavior and gastric motility.
Rayman, M.P. (2012). Selenium and human health. Lancet 379, 1256–1268.
Am. J. Physiol. Endocrinol. Metab. 303, E853–E864.
Sandholt, C.H., Sparsø, T., Grarup, N., Albrechtsen, A., Almind, K., Hansen, L.,
Herman, M.A., Peroni, O.D., Villoria, J., Schön, M.R., Abumrad, N.A., Blüher, Toft, U., Jørgensen, T., Hansen, T., and Pedersen, O. (2010). Combined anal-
M., Klein, S., and Kahn, B.B. (2012). A novel ChREBP isoform in adipose tissue yses of 20 common obesity susceptibility variants. Diabetes 59, 1667–1673.
regulates systemic glucose metabolism. Nature 484, 333–338.
Shao, H., Burrage, L.C., Sinasac, D.S., Hill, A.E., Ernest, S.R., O’Brien, W.,
Hornemann, T., Penno, A., Rütti, M.F., Ernst, D., Kivrak-Pfiffner, F., Rohrer, L., Courtland, H.W., Jepsen, K.J., Kirby, A., Kulbokas, E.J., et al. (2008).
and von Eckardstein, A. (2009). The SPTLC3 subunit of serine palmitoyltrans- Genetic architecture of complex traits: large phenotypic effects and pervasive
ferase generates short chain sphingoid bases. J. Biol. Chem. 284, 26322– epistasis. Proc. Natl. Acad. Sci. USA 105, 19910–19914.
26330.
Small, K.S., Hedman, A.K., Grundberg, E., Nica, A.C., Thorleifsson, G., Kong,
Ioffe, E., Moon, B., Connolly, E., and Friedman, J.M. (1998). Abnormal regula- A., Thorsteindottir, U., Shin, S.Y., Richards, H.B., Soranzo, N., et al.; GIANT
tion of the leptin gene in the pathogenesis of obesity. Proc. Natl. Acad. Sci. Consortium; MAGIC Investigators; DIAGRAM Consortium; MuTHER
USA 95, 11852–11857. Consortium (2011). Identification of an imprinted master trans regulator at
Jelinek, D., Heidenreich, R.A., Erickson, R.P., and Garver, W.S. (2010). the KLF14 locus related to multiple metabolic phenotypes. Nat. Genet. 43,
Decreased Npc1 gene dosage in mice is associated with weight gain. 561–564.
Obesity (Silver Spring) 18, 1457–1459. Smith, E.N., Chen, W., Kähönen, M., Kettunen, J., Lehtimäki, T., Peltonen, L.,
Kang, H.M., Zaitlen, N.A., Wade, C.M., Kirby, A., Heckerman, D., Daly, M.J., Raitakari, O.T., Salem, R.M., Schork, N.J., Shaw, M., et al. (2010). Longitudinal
and Eskin, E. (2008). Efficient control of population structure in model organism genome-wide association of cardiovascular disease risk factors in the
association mapping. Genetics 178, 1709–1723. Bogalusa heart study. PLoS Genet. 6, e1001094.

Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc. 11

CMET 1239
Please cite this article in press as: Parks et al., Genetic Control of Obesity and Gut Microbiota Composition in Response to High-Fat, High-Sucrose Diet
in Mice, Cell Metabolism (2013), http://dx.doi.org/10.1016/j.cmet.2012.12.007

Cell Metabolism
Genetics of Dietary Responsiveness in Mice

Sonestedt, E., Roos, C., Gullberg, B., Ericson, U., Wirfält, E., and Orho- Vijay-Kumar, M., Aitken, J.D., Carvalho, F.A., Cullender, T.C., Mwangi, S.,
Melander, M. (2009). Fat and carbohydrate intake modify the association Srinivasan, S., Sitaraman, S.V., Knight, R., Ley, R.E., and Gewirtz, A.T.
between genetic variation in the FTO genotype and obesity. Am. J. Clin. (2010). Metabolic syndrome and altered gut microbiota in mice lacking Toll-
Nutr. 90, 1418–1425. like receptor 5. Science 328, 228–231.
Speakman, J.R., Levitsky, D.A., Allison, D.B., Bray, M.S., de Castro, J.M., Voight, B.F., Scott, L.J., Steinthorsdottir, V., Morris, A.P., Dina, C., Welch, R.P.,
Clegg, D.J., Clapham, J.C., Dulloo, A.G., Gruer, L., Haw, S., et al. (2011). Set Zeggini, E., Huth, C., Aulchenko, Y.S., Thorleifsson, G., et al.; MAGIC investi-
points, settling points and some alternative models: theoretical options to gators; GIANT Consortium (2010). Twelve type 2 diabetes susceptibility loci
understand how genes and environments combine to regulate body adiposity. identified through large-scale association analysis. Nat. Genet. 42, 579–589.
Dis. Model Mech. 4, 733–745.
Wang, Z., Klipfell, E., Bennett, B.J., Koeth, R., Levison, B.S., Dugar, B.,
Speliotes, E.K., Willer, C.J., Berndt, S.I., Monda, K.L., Thorleifsson, G., Feldstein, A.E., Britt, E.B., Fu, X., Chung, Y.M., et al. (2011). Gut flora metab-
Jackson, A.U., Lango Allen, H., Lindgren, C.M., Luan, J., Mägi, R., et al.; olism of phosphatidylcholine promotes cardiovascular disease. Nature 472,
MAGIC; Procardis Consortium (2010). Association analyses of 249,796 indi- 57–63.
viduals reveal 18 new loci associated with body mass index. Nat. Genet. 42,
Wilcox, R. (2005). Introduction to Robust Estimation and Hypothesis Testing
937–948.
(Amsterdam, The Netherlands: Elsevier/Academic Press).
Stunkard, A.J., Foch, T.T., and Hrubec, Z. (1986). A twin study of human
obesity. JAMA 256, 51–54. Xu, X.M., Carlson, B.A., Irons, R., Mix, H., Zhong, N., Gladyshev, V.N., and
Hatfield, D.L. (2007). Selenophosphate synthetase 2 is essential for selenopro-
Taicher, G.Z., Tinsley, F.C., Reiderman, A., and Heiman, M.L. (2003).
tein biosynthesis. Biochem. J. 404, 115–120.
Quantitative magnetic resonance (QMR) method for bone and whole-body-
composition analysis. Anal. Bioanal. Chem. 377, 990–1002. Yang, J., Benyamin, B., McEvoy, B.P., Gordon, S., Henders, A.K., Nyholt, D.R.,
Madden, P.A., Heath, A.C., Martin, N.G., Montgomery, G.W., et al. (2010).
Ternes, P., Franke, S., Zähringer, U., Sperling, P., and Heinz, E. (2002).
Common SNPs explain a large proportion of the heritability for human height.
Identification and characterization of a sphingolipid delta 4-desaturase family.
Nat. Genet. 42, 565–569.
J. Biol. Chem. 277, 25512–25518.
Thorleifsson, G., Walters, G.B., Gudbjartsson, D.F., Steinthorsdottir, V., Yatsunenko, T., Rey, F.E., Manary, M.J., Trehan, I., Dominguez-Bello, M.G.,
Sulem, P., Helgadottir, A., Styrkarsdottir, U., Gretarsdottir, S., Thorlacius, S., Contreras, M., Magris, M., Hidalgo, G., Baldassano, R.N., Anokhin, A.P.,
Jonsdottir, I., et al. (2009). Genome-wide association yields new sequence et al. (2012). Human gut microbiome viewed across age and geography.
variants at seven loci that associate with measures of obesity. Nat. Genet. Nature 486, 222–227.
41, 18–24. York, B., Truett, A.A., Monteiro, M.P., Barry, S.J., Warden, C.H., Naggert, J.K.,
Turnbaugh, P.J., Ley, R.E., Mahowald, M.A., Magrini, V., Mardis, E.R., and Maddatu, T.P., and West, D.B. (1999). Gene-environment interaction: a signif-
Gordon, J.I. (2006). An obesity-associated gut microbiome with increased icant diet-dependent obesity locus demonstrated in a congenic segment on
capacity for energy harvest. Nature 444, 1027–1031. mouse chromosome 7. Mamm. Genome 10, 457–462.
Turnbaugh, P.J., Hamady, M., Yatsunenko, T., Cantarel, B.L., Duncan, A., Ley, Zhang, Y., Proenca, R., Maffei, M., Barone, M., Leopold, L., and Friedman,
R.E., Sogin, M.L., Jones, W.J., Roe, B.A., Affourtit, J.P., et al. (2009a). A core J.M. (1994). Positional cloning of the mouse obese gene and its human homo-
gut microbiome in obese and lean twins. Nature 457, 480–484. logue. Nature 372, 425–432.
Turnbaugh, P.J., Ridaura, V.K., Faith, J.J., Rey, F.E., Knight, R., and Gordon, Zuk, O., Hechter, E., Sunyaev, S.R., and Lander, E.S. (2012). The mystery of
J.I. (2009b). The effect of diet on the human gut microbiome: a metagenomic missing heritability: Genetic interactions create phantom heritability. Proc.
analysis in humanized gnotobiotic mice. Sci. Transl. Med. 1, ra14. Natl. Acad. Sci. USA 109, 1193–1198.

12 Cell Metabolism 17, 1–12, January 8, 2013 ª2013 Elsevier Inc.

CMET 1239