A Systems Biology Approach Identifies Inflammatory Abnormalities Between Mouse Strains Prior to Development of Metabolic Disease

Abstract

Type 2 diabetes and obesity are increasingly affecting human populations around the world. Our goal was to identify early molecular signatures predicting genetic risk to these metabolic diseases using two strains of mice that differ greatly in disease susceptibility.
We integrated metabolic characterization, gene expression, protein-protein interaction networks, RT-PCR, and flow cytometry analyses of adipose, skeletal muscle, and liver tissue of diabetes-prone C57BL/6NTac (B6) mice and diabetes-resistant 129S6/SvEvTac (129) mice at 6 weeks and 6 months of age.
At 6 weeks of age, B6 mice were metabolically indistinguishable from 129 mice, however, adipose tissue showed a consistent gene expression signature that differentiated between the strains. In particular, immune system gene networks and inflammatory biomarkers were upregulated in adipose tissue of B6 mice, despite a low normal fat mass. This was accompanied by increased T-cell and macrophage infiltration. The expression of the same networks and biomarkers, particularly those related to T-cells, further increased in adipose tissue of B6 mice, but only minimally in 129 mice, in response to weight gain promoted by age or high-fat diet, further exacerbating the differences between strains.
Insulin resistance in mice with differential susceptibility to diabetes and metabolic syndrome is preceded by differences in the inflammatory response of adipose tissue. This phenomenon may serve as an early indicator of disease and contribute to disease susceptibility and progression.

Full text

A Systems Biology Approach Identifies Inflammatory
Abnormalities Between Mouse Strains Prior to
Development of Metabolic Disease
Marcelo A. Mori,
1
Manway Liu,
2
Olivier Bezy,
1
Katrine Almind,
1
Hagit Shapiro,
1
Simon Kasif,
2
and
C. Ronald Kahn
1
OBJECTIVE—Type 2 diabetes and obesity are increasingly
affecting human populations around the world. Our goal was to
identify early molecular signatures predicting genetic risk to
these metabolic diseases using two strains of mice that differ
greatly in disease susceptibility.
RESEARCH DESIGN AND METHODS—We integrated meta-
bolic characterization, gene expression, protein-protein interac-
tion networks, RT-PCR, and flow cytometry analyses of adipose,
skeletal muscle, and liver tissue of diabetes-prone C57BL/6NTac
(B6) mice and diabetes-resistant 129S6/SvEvTac (129) mice at 6
weeks and 6 months of age.
RESULTS—At 6 weeks of age, B6 mice were metabolically
indistinguishable from 129 mice, however, adipose tissue showed
a consistent gene expression signature that differentiated be-
tween the strains. In particular, immune system gene networks
and inflammatory biomarkers were upregulated in adipose tissue
of B6 mice, despite a low normal fat mass. This was accompanied
by increased T-cell and macrophage infiltration. The expression
of the same networks and biomarkers, particularly those related
to T-cells, further increased in adipose tissue of B6 mice, but only
minimally in 129 mice, in response to weight gain promoted by
age or high-fat diet, further exacerbating the differences between
strains.
CONCLUSIONS—Insulin resistance in mice with differential
susceptibility to diabetes and metabolic syndrome is preceded by
differences in the inflammatory response of adipose tissue. This
phenomenon may serve as an early indicator of disease and
contribute to disease susceptibility and progression. Diabetes
59:2960–2971, 2010
Type 2 diabetes and obesity are major causes of
mortality and morbidity worldwide (1). Accord-
ing to World Health Organization estimates,
more than 1 billion adults are overweight and
more than 200 million individuals have type 2 diabetes.
The etiologies of obesity and diabetes are complex and
created by interactions between environmental factors
(i.e., high caloric intake and reduced energy expenditure)
(1,2) and genetic background (3–5).
Inbred mouse strains are useful models for studying the
role of the environment and genes in differential suscep-
tibility to diabetes and obesity (6 –14). In response to
high-fat diet, aging, or genetic challenge, C57BL/6NTac
(B6) mice become severely insulin resistant, hyperinsu-
linemic, and diabetic, whereas 129S6/SvEvTac (129) mice
are resistant to these challenges (7,8). Using intercross
and F2 mice, our group and others have previously shown
that the difference in disease susceptibility between the
strains is inherited in a dominant fashion and linked to
quantitative traits on at least three different chromosomes
(8,15).
In the present study, we compared gene expression
profiles of B6 and 129 mice in different tissues, ages, and
diets. For each comparison, we identified the differentially
expressed network of genes between the strains using a
novel variant of gene network enrichment analysis
(GNEA) (16), an algorithm that integrates gene expression
data together with protein-protein interaction networks.
We subsequently intersected the results of the compari-
sons between B6 and 129 mice to identify genes and
pathways that were significant to each or all conditions,
i.e., different tissues, ages, and diets. Of the numerous,
differentially expressed subnetworks in the adipose tissue
of B6 versus 129 mice; the most significant related to the
immune system. Importantly, these differences were iden-
tified even at 6 weeks, an age when the mouse strains were
indistinguishable by fat mass or metabolic phenotyping.
RT-PCR and flow cytometry confirmed higher expression
of major inflammatory markers and higher infiltration of
macrophages and T-cells in adipose tissue of 6-week-old
B6 versus 129 mice. Weight gain associated with aging or
high-fat diet further increased inflammation in adipose
tissue of B6 but not in 129, mice. Taken together, these
results demonstrate that measurable differences in the
inflammatory milieu of the adipose tissue precede measur-
able differences in insulin sensitivity in response to weight
gain and contribute to the differences in diabetes risk
between B6 and 129 mice.
RESEARCH DESIGN AND METHODS
C57BL/6NTac and 129S6/SvEvTac male mice were obtained from Taconic
(Germantown, NY). Mice were maintained on a 12-h light-dark cycle with ad
libitum access to tap water (reverse osmosis purified) and a chow diet
containing 21% calories from fat, 23% from protein, and 55% from carbohy-
drates (Mouse Diet 9F; PharmaServ, Framingham, MA). For the feeding
studies, 6-week-old mice were submitted to low-fat diet (Rodent NIH-31M
Auto: 14% calories from fat, 25% from protein, and 61% from carbohydrates
[Taconic]) or high-fat diet (TD.93075: 55% calories from fat, 21% calories from
protein, and 24% calories from carbohydrates [Harlan Teklad, Madison, WI])
for 18 weeks before they were killed. In addition to fat content, these diets
From the
1
Section on Integrative Physiology and Metabolism, Joslin Diabetes
Center, Harvard Medical School, Boston, Massachusetts; and the
2
Depart-
ment of Biomedical Engineering, Boston University, Boston, Massachusetts.
Corresponding author: C. Ronald Kahn, c.ronald.kahn@joslin.harvard.edu.
Received 16 March 2010 and accepted 4 August 2010. Published ahead of
print at http://diabetes.diabetesjournals.org on 16 August 2010. DOI:
10.2337/db10-0367.
M.A.M. and M.L. contributed equally to this work.
© 2010 by the American Diabetes Association. Readers may use this article as
long as the work is properly cited, the use is educational and not for profit,
and the work is not altered. See http://creativecommons.org/licenses/by
-nc-nd/3.0/ for details.
The costs of publication of this article were defrayed in part by the payment of page
charges. This article must therefore be hereby marked “advertisement” in accordance
with 18 U.S.C. Section 1734 solely to indicate this fact.
ORIGINAL ARTICLE
2960 DIABETES, VOL. 59, NOVEMBER 2010 diabetes.diabetesjournals.org

differ in other components that could also influence the results of our analysis.
For example, our chow diet contained fat from animal sources, whereas both
low-fat and high-fat diets contained vegetable fat. These differences may have
contributed to the fact that a few metabolic parameters displayed in Table 1,
such as insulin levels, did not necessarily track with the percentage of fat in
the diet. However, in this study and in others, these diets have been primarily
used as independent paradigms to investigate the impact of the environment
on weight gain of animal models, showing metabolic effects that closely
resembled different characteristics of human obesity (8). Caloric-restricted
animals were obtained from the National Institute on Aging. Caloric restric-
tion was initiated at 14 weeks of age with a 10% decrease in calories, increased
to 25% restriction at 15 weeks, and to 40% restriction at 16 weeks, which was
maintained until the age of 6 months when the mice were killed. Protocols for
animal use were reviewed and approved by the Animal Care Committee of the
Joslin Diabetes Center and were in accordance with the National Institutes of
Health guidelines.
Gene expression analysis. RNA pooled from tissues of two or three mice
was used for microarray analysis. Two to five Affymetrix murine chips
U74Av.2 (Santa Clara, CA) were used per mouse strain, for each comparison
of B6 versus 129 mice. Quantitative RT-PCR was performed using a SYBR
Green-based ABI Prism 7900 Sequence Detection System (Applied Biosys-
tems, Foster City, CA). For details, refer to the supplementary METHODS in the
online appendix available at http://diabetes.diabetesjournals.org/cgi/
content/full/db10-0367/DC1.
Gene network enrichment analysis. A novel version of Gene Network
Enrichment Analysis (16) was applied to each comparison of B6 versus 129
mice to identify gene networks (subnetworks of genes with edges connecting
physically interacting protein products) with differential activation between
these mice, and to test such networks for statistically significant, over-
represented pathways and processes. Briefly, the algorithm was composed of
the following steps (see supplementary METHODS for details).
First, for each gene, a differential expression significance value was
calculated in B6 versus 129 mice, converted to zscores, and used to annotate
the corresponding proteins in a protein-protein interaction network curated
from literature (17) (Fig. 1A).
Subsequently, a computationally intensive, stochastic optimization algo-
rithm (18) was employed to identify a subnetwork of genes with a high mean
zscore (reflecting high significance). This process was repeated many times
to yield a multitude of subnetworks (Fig. 1B). The hypothesis was that a
consensus among the stochastically detected subnetworks represented a
prediction of the true, differentially active gene network in B6 versus 129 mice.
Thus, each gene was associated with a Pvalue based on the observed number
of subnetworks in which it appeared, reflecting its chance probability of
belonging to the true network (Fig. 1C).
In the next step of the algorithm, each biological process and pathway from
the Gene Ontology and Molecular Signatures Database were tested for
over-representation of genes with low Pvalues, using standard statistical
methods (Fig. 1D).
Finally, for each statistically significant pathway or biological process, the
algorithm produced a gene network consisting of pathway member genes and
all additional genes with significant, adjusted Pvalues (false discovery rate
ⱕ0.25) that shared an edge with a pathway member gene in the protein-
protein interaction network (Fig. 1E), providing an intuitively appealing
visualization aid that helped in interpreting the enrichment results.
Flow cytometry. Erythrocyte-free stromavascular cells were isolated from
epididymal fat pads after collagenase digestion as previously described (19).
These cells were incubated with a mix of antibodies against different surface
markers and analyzed using the LSRII flow cytometer (BD Biosciences, San
Jose, CA). See supplementary METHODS for detailed protocol.
Statistics. Statistical analyses for GNEA are described in supplementary
METHODS. All other experiments were compared using ANOVA. Results are
expressed as mean ⫾SEM. Pvalues ⱕ0.05 were considered significant.
RESULTS
Gene network signatures associated with the predis-
position to metabolic diseases in B6 versus 129 mice.
At 6 months of age, on a chow diet, B6 mice were
overweight, hyperglycemic, and exhibited higher plasma
leptin and insulin levels than 129 mice of the same age
(Table 1, supplementary Fig. S1) (8). B6 mice showed
clear signs of insulin resistance when compared with 129
mice, even on low-fat diet (LFD), and this difference was
further exacerbated when the mice were placed on a
high-fat diet (Table 1) (8). At 6 weeks of age, however, no
phenotypic differences or serum markers indicative of
diabetes and obesity could be detected between the two
strains (Table 1) (8). Indeed, 6-week-old 129 mice have the
same weight as B6 mice and actually have significantly
more fat mass, as indicated by higher fat pad weight (Table
1). By studying B6 versus 129 mice at different ages and on
different diets, one may therefore elucidate the back-
ground inherited phenomena predisposing to obesity-ac-
companied insulin resistance.
To identify gene networks that might provide early
biomarkers of adult metabolic disease, we initially per-
formed microarray analyses on RNA from adipose tissue,
liver, and skeletal muscle of B6 and 129 mice at 6 weeks
and 6 months of age on a standard chow diet. We then
applied GNEA to detect networks that were differentially
active in a predisease stage and to estimate how they
changed over time in disease-prone versus diabetes-resis-
tant strains (Fig. 1). Since disease susceptibility is an
inherited phenomenon in these animals, we hypothesized
that the predisposing hereditable biological processes
would be those showing differential activity in young mice
and maintained or further increased with age.
Among the three different tissues in which B6 versus
129 mice were compared, only adipose tissue exhibited
significantly enriched biological processes (Qvalue
⬍0.33) at 6 weeks of age by GNEA, with 12 enriched
gene sets, and this increased to 25 enriched gene sets at
6 months of age. Of these, only two sets (“immune
system process” and “regulation of cytokine secretion”)
were found enriched at both ages (supplementary Table
1), and both of these related to inflammatory pathways.
Furthermore, the “immune system process” was the
TABLE 1
Metabolic parameters in B6 vs. 129 mice at different ages and dietary conditions
Chow diet (21% fat) Low-fat diet (14% fat) High-fat diet (55% fat)
6 weeks old 6 months old
B6 129 B6 129 B6 129 B6 129
Body weight (g) 21.9 ⫾0.6 23.5 ⫾0.5 45.5 ⫾1.0 30.4 ⫾1.2 41.5 ⫾1.4 30.9 ⫾1.2 46.4 ⫾1.1 39.2 ⫾1.5
Epididymal fat
mass (g) 0.09 ⫾0.01 0.21 ⫾0.02 1.67 ⫾0.14 2.14 ⫾0.30 1.62 ⫾0.17 0.58 ⫾0.11 3.39 ⫾0.33 3.92 ⫾0.14
Blood glucose–
random fed
(mg/dl) 155.5 ⫾3.1 129.3 ⫾2.9 168.2 ⫾12.3 106.6 ⫾3.5 148.5 ⫾7.4 102.5 ⫾3.4 153.3 ⫾6.7 105.5 ⫾3.7
Insulin (ng/ml) 2.26 ⫾0.48 3.12 ⫾0.58 92.14 ⫾25.71 3.96 ⫾2.04 24.78 ⫾12.51 1.29 ⫾0.32 32.73 ⫾12.46 3.58 ⫾0.41
Leptin (ng/ml) 7.14 ⫾0.70 9.00 ⫾0.90 76.18 ⫾3.82 16.67 ⫾4.5 72.50 ⫾8.06 4.97 ⫾0.77 82.38 ⫾7.41 46.81 ⫾6.02
Values are expressed as mean ⫾SE of 8 animals per group.
M.A. MORI AND ASSOCIATES
diabetes.diabetesjournals.org DIABETES, VOL. 59, NOVEMBER 2010 2961

most significantly enriched gene set at 6 months of age
(Qvalue ⫽0.008; supplementary Table 1). In contrast, in
the liver, 37 gene sets (Qvalue ⬍0.33) were differen-
tially regulated in B6 versus 129 at 6 months of age, but
none was significantly different in the 6-week-old ani-
mals. In skeletal muscle, no significant difference was
observed at either age (data not shown).
In addition to GNEA, the adipose tissue dataset was
analyzed using gene set enrichment analysis (GSEA) (20)
and the standard hypergeometric test (21). GSEA also
Diet
(% lipid)
Tissue
Age
(weeks)
(months)
129
B6 vs.
55%
Liver Adipose Skeletal Muscle
14% 55%
21% 14% 55%
21% 14% 21%
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 mos
6 wks
6 wks
6 wks
6 wks
6 wks
6 wks
Map individual p-values onto
a network of genes.
Find differentially active
subnetworks
G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 G11 G12
C1
C1
C1
C1
C2
C2
C2
C2
P
Identify biological processes
enriched in genes with high
MAN-scores and visualize in
the network context.
G1G2GN
G3
...
Order genes by MAN-scores
Differential gene
expression (B6 vs. 129)
Network of genes labeled
with their p-values
Differentially active
subnetworks of genes
Computational Analysis
Experimental Setup
CD45
IKBKA
CD75
MCP1
SDF1
Calculate the Membership
in Differentially Active
SubNetwork score (MAN-score)
per gene.
...
MAN-score
Network view of a biological process
A
B
CD
E
FIG. 1. Overview of gene network enrichment analysis. Individual comparisons of B6 and 129 mice were performed in different tissue types
(adipose, skeletal muscle, and liver), ages (6 weeks and 6 months), and diets (high fat, low fat, and standard chow). A: For each comparison, a
differential expression significance value was determined for each gene, converted to a zscore, and mapped onto a network of protein-protein
interactions. The colors in the network represent the different ranges of zscores. B: A stochastic search algorithm (18) was then applied to search
for subnetworks with high mean zscore (high significance). Each application of the algorithm identifies one subnetwork; hence, to detect a
multitude of subnetworks, the algorithm was run many times. The consensus among these subnetworks represented approximations of the true,
differentially active network in B6 vs. 129 mice. C: A membership in differentially active subnetwork (MAN) score was determined for each gene.
MAN scores are Pvalues and represent the probability of genes to belong to the true, differentially active network. This score was calculated as
the random chance probability that the gene is detected in the observed number of subnetworks computed over permutations of the data. D:
Biological processes and pathways were tested for enrichment in genes with high MAN scores. Statistically significant pathways and processes
were predicted to be differentially active in B6 vs. 129 mice. E: For each significant process and pathway, a network visualization was generated
to aid interpretation.
SYSTEMS BIOLOGY AND SUSCEPTIBILITY TO DIABETES
2962 DIABETES, VOL. 59, NOVEMBER 2010 diabetes.diabetesjournals.org

identified differences in inflammation at 6 weeks. In par-
ticular, it found a significant difference in the “humoral
immune response” biological process (supplementary Ta-
ble 2). On the other hand, GSEA failed to identify any gene
set that was differentially expressed at both ages (supple-
mentary Table 3). By comparison, the standard hypergeo-
metric test (21) failed to identify either a clear immune
signature at 6 weeks or any enriched gene sets at 6 months
(supplementary Tables 2 and 3).
Gene networks associated with the immune system
process are differentially altered in the adipose tis-
sue of B6 versus 129 mice. To determine how genes
related to the immune response were differentially ex-
pressed in the fat tissue of young B6 and 129 mice and
changed in expression as the mice aged, we identified the
network associated with the “immune system process”
gene set (supplementary Fig. S2) and plotted the histo-
gram of differential expression for its genes in B6 versus
129 mice at 6 weeks and 6 months (Fig. 2). A few
individual genes, such as apolipoprotein A (APOA1 and
APOA2) and mannose-binding lectin (MBL2) were more
strongly downregulated in B6 versus 129 mice at 6 months
and less so at 6 weeks. Nonetheless, much of the immune
system process network was upregulated at 6 months
compared with 6 weeks (note the right-shift in the histo-
gram for 6 months vs. 6 weeks in Fig. 2). Lymphoid and
myeloid markers, such as CD45, CD3, and chemokine (C-C
motif) receptor 5 (CCR5), monocyte chemotactic protein-1
(MCP1), as well as proinflammatory genes such as chemo-
kine (C-C motif) ligand 5 (CCL5) and interleukin-6 (IL6),
were more highly expressed in B6 compared with 129 mice
at 6 weeks, and this increased further at 6 months of age
(supplementary Fig. S2, left panel,arrows). Moreover, as
shown by the histogram (Fig. 2), many of the immune
system process genes already showed twofold or greater
differences in expression in B6 versus 129 mice at 6 weeks,
an age when no significant metabolic differences were
observed between strains.
Genes that have protein products that physically inter-
act tend to correlate in expression (22) and are likely
related by function or membership in the same signaling
pathways. We consequently examined the subnetworks of
genes that interact at the protein level with inflammatory
markers identified in the strain comparisons (Fig. 3). For
the T-cell marker Thy1, minor changes in the expression of
its interacting genes were observed in the 6-week-old B6
versus 129 mice comparisons (Fig. 3A,left panel). How-
ever, by 6 months of age, more than 90% of the nodes in the
network were differentially active between the two strains
(large green triangles in Fig. 3A,right panel). Hence, genes
interacting with Thy1 were dramatically more upregulated
with age-associated weight gain in adipose tissue of B6
mice than in 129 mice, suggesting that B6 mice have more
T-cells recruited to adipose tissue in response to increas-
0
10
20
30
40
50
60
70
-6.88 -5.86 -4.85 -3.83 -2.82 -1.80 -0.79 0.23 1.24 2.26 3.27 4.29 5.31
6 weeks
6 months
Fold-change (log
2
) in B6 vs. 129 mice
# of genes
APOA1 APOA2 MBL2
CCR5, CD64
FIG. 2. Histogram of differential expression between B6 and 129 mice at 6 weeks and 6 months of age for genes in the immune system process
network. The x-axis values are the upper bounds of the histogram bins and correspond to the log
2
fold-change difference in B6 vs. 129 mice. Few
network genes are unchanged (absolute log
2
fold change <1) in B6 vs. 129 mice at 6 months. The majority are strongly upregulated (log
2
fold
change >1), although a few are strongly downregulated (log
2
fold change <ⴚ1). A network view of the same biological process demonstrating
the fold changes of individual genes is provided in supplementary Fig. S2.
M.A. MORI AND ASSOCIATES
diabetes.diabetesjournals.org DIABETES, VOL. 59, NOVEMBER 2010 2963

CD11B CD11B
CD18
CD18
CD61
CD61
CD51
CD51
CD87
CD87
CD23
CD23
CD152
CD152
CD49E
CD49E
CCBP1
CCBP1
MCP1 MCP1
A
CD75 CD75
B
C
Fold Change 6 mos > 6 wks
Fold Change 6 mos < 6 wks
1
< 0.5
> 2
Fold Change 6 wks
1
< 0.5
> 2
Fold Change 6 mos
FIG. 3. Gene networks associated with inflammatory markers in adipose tissue of B6 and 129 mice at 6 weeks (left) and 6 months (right) of age. Gene
networks were generated by mapping genes that were significantly overrepresented (Qvalue <0.25) among the gene network enrichment analysis
results at 6 months intersected with those at 6 weeks onto protein-protein interaction networks involving at least one interactor to each inflammatory
marker: (A) Thy1, (B) CD45, and (C) MCP1. The CD45 network was further restricted only to genes showing a greater than twofold magnitude
difference in B6 vs. 129 mice between 6 weeks and 6 months. This was done to aid visualization since the unrestricted network consisted of 416
interactions and 141 genes. Colors range from bright red to green, corresponding to twofold less and twofold greater differences in expression between
B6 and 129 at each age, respectively. Nodes point upward if the fold-change difference between B6 and 129 is greater at 6 months than at 6 weeks, and
downward otherwise. The node size corresponds to the magnitude of that fold-change difference between the two ages. The inflammatory network
around each biomarker is drawn to scale between the two ages. Networks around different biomarkers are not to scale with one another.
SYSTEMS BIOLOGY AND SUSCEPTIBILITY TO DIABETES
2964 DIABETES, VOL. 59, NOVEMBER 2010 diabetes.diabetesjournals.org

ing adiposity than 129 mice. The same pattern was ob-
served in subnetworks generated for CD45, a leukocyte
marker (Fig. 3B); CD11c, a dendritic cell marker (supple-
mentary Fig. S3A); CD11b, a macrophage marker (supple-
mentary Fig. S3B); MCP1, a monocyte chemokine (Fig.
3C); and tumor necrosis factor ␣(TNF␣), a proinflamma-
tory cytokine (supplementary Fig. S3C). Together these
data suggest that B6 mice show a proinflammatory ten-
dency in fat at 6 weeks of age not observed in 129 mice,
and that this differentially increases with weight gain
promoted by aging, affecting the pool of inflammatory
cells.
Changes in expression of inflammatory markers and
cytokines/chemokines in adipose tissue of B6 versus
129 mice. To validate the observations obtained using
network analysis, we measured markers of inflammatory
cells and cytokine/chemokine genes in the adipose tissue
of B6 and 129 mice using quantitative RT-PCR (qPCR)
(Fig. 4). In agreement with the microarray data, at 6 weeks
of age, there were higher levels of macrophage/myeloid
cell markers (CD68, CD11b, and CD18), lymphocyte mark-
ers (CD3, CD72, and CD80), T-cell attractants/activators
CCL5/RANTES, stroma cell-derived factor-1␣(SDF1␣),
and interferon-␥(IFN␥), among others, in B6 versus 129
mice (Fig. 4Aand B). For most of these genes, the
differences between strains were maintained or further
exacerbated at 6 months of age. MCP1, a monocyte
attractant chemokine, despite being similarly expressed in
young B6 and 129 mice, exhibited a 4.4-fold increase in B6
mice between 6 weeks and 6 months of age, whereas in
129 mice, there was a 78% reduction with age (Fig. 4B). For
a few cytokines, namely TNF␣, IL6, and macrophage
migration inhibitory factor (MIF), no differences were
observed between the strains at any age (Fig. 4B).
We also validated by RT-PCR the expression of selected
genes identified as differentially present in the differen-
tially active subnetworks identified by GNEA. Two medi-
ators of lipopolysaccharide (LPS) signaling, i.e., the
lymphocyte antigen 86 (Ly86) and LPS-binding protein
(LBP), were highly ranked by GNEA and were substan-
tially higher in B6 compared with 129 mice at both ages (Fig.
4C).
To determine if the higher expression of inflammation
markers in B6 versus 129 mice was restricted to the
adipose tissue or present in other tissues, we assessed the
expression of two macrophage markers (CD68 and F4/80),
two T-cell markers (CD3 and Thy1), and two chemokines
(MCP1 and SDF1␣) in the liver, skeletal muscle, and
spleen of B6 and 129 mice at both ages using qPCR. The
pattern of expression in liver was qualitatively similar to
that in adipose tissue, but with smaller fold differences
(supplementary Fig. S4A). In contrast, no differences were
observed in the skeletal muscle or spleen in the expression
for most of these genes (supplementary Fig. S4Band C).
The two exceptions were MCP1, which was increased in
all tissues in the 6-month-old B6 mice when compared
with 129 mice, and CD3, which increased in both B6 and
129 mice with age in skeletal muscle only (supplementary
Fig. S4).
An altered repertoire of inflammatory cells in the
adipose tissue of B6 versus 129 mice. To determine if
there was a cellular shift in the adipose tissue contributing
A
mRNA expression
(Fold Change)
mRNA expression
(Fold Change)
mRNA expression
(Fold Change)
0
0.5
1
1.5
2
2.5
3
3.5
CD45 CCR5 CD68 F4/80 CD11b CD11c CD18 CD3 Thy1 CD72 CD80
B6 6wk
129 6wk
B6 6 mo
129 6 mo
B6 6wk
129 6wk
B6 6 mo
129 6 mo
B6 6wk
129 6wk
B6 6 mo
129 6 mo
**
#
*
*** **
*
*
#
**
****
#
**
###
*
*
*
**
#
#
B
0
1
2
3
4
5
6
TNF IL6 IFNγMIF MCP1 CCL5 SDF1α
**
##
***
##
##
**
***
0
0.2
0.4
0.6
0.8
1
1.2
1.4
LBP Ly86
*** ***
***
*
Myeloid markers Lymphoid markers
Leukocyte markers
Proinflammatory
cytokines
Chemokines
C
α
FIG. 4. Expression of inflammatory markers in
adipose tissue of 6-week-old (6 weeks) or
6-month-old (6 months) B6 and 129 mice. A:
Leukocyte markers, including selective myeloid
cell markers (e.g., macrophages and dendritic
cells) and lymphoid markers (e.g., lymphocytes).
B: Cytokines and chemokines implicated with
inflammatory responses. C: Mediators of LPS
signaling. mRNA expression was assessed by
qPCR. All values are normalized by TATA-bind-
ing protein (TBP) and expressed as fold change
of the 6-week-old B6 average value. Results rep-
resent mean ⴞSEM of 5–7 animals. *P<0.05;
**P<0.01; ***P<0.001 vs. 129. #P<0.05;
##P<0.01; ###P<0.001 vs. 6 weeks.
M.A. MORI AND ASSOCIATES
diabetes.diabetesjournals.org DIABETES, VOL. 59, NOVEMBER 2010 2965

to our computational and qPCR results, we analyzed the
inflammatory cell repertoire in the fat tissue of B6 versus
129 mice using flow cytometry (Fig. 5). This revealed a
significant twofold increase in the number of nonerythro-
cyte stromavascular fraction cells per gram of epididymal
fat in B6 compared with 129 mice at both ages (Fig. 5A). At
6 weeks of age, 75% of this difference was explained by a
higher number of leukocytes in B6 fat versus 129 fat.
Although the number of leukocytes per gram of adipose
tissue decreased in both strains with age as adipose mass
increased, at both time points the number of leukocytes
was about 2.5-fold higher in the stromavascular fraction
from B6 fat than in 129 fat (Fig. 5A). Likewise, the number
of macrophages per gram of fat decreased with age
because of growing fat mass, but was higher in the B6
versus 129 mice at both ages (Fig. 5A). These data corre-
lated with histologic evidence of crown-like structures,
i.e., collections of macrophages surrounding dead or dying
adipocytes, in the adipose tissue of 6-month-old B6 mice
(supplementary Fig. S1).
Somewhat surprisingly, the density of CD11c
⫹
-myeloid
cells, which have been previously associated with insulin
resistance states in mice (23), was decreased in B6 versus
129 mice at 6 weeks and 6 months of age by 65 and 37%,
respectively (Fig. 5A). In contrast, the T-cell population
was dramatically increased by 7.1-fold in B6 mice at 6
weeks of age (Fig. 5A). Furthermore, the density of T-cells
increased by 25-fold in B6 mice with age-associated weight
gain, but showed only a threefold increase in 129 mice
(Fig. 5A). Thus, at 6 months of age, B6 mice had ⬎55-fold
more total T-cells and ⬎60-fold more CD4
⫹
T-helper cells
in the adipose tissue compared with the 129 mice (Fig.
5A). B-cells were also present in the adipose tissue of
these mice; however, they were not considered in the
analysis because of their modest numbers in comparison
with other leukocytes (⬍1%).
A similar pattern of lymphocyte/monocyte infiltration
was observed in the 6-week-old mice when the total
number of cells in the epididymal fat pads was calculated,
i.e., cells per gram ⫻total grams of fat (Fig. 5B). When
expressed as total cells, however, it was apparent that the
number of each of the myeloid and lymphoid cell popula-
tion increased with age in both strains. Importantly, the
differences between strains were maintained in both
younger and older mice (Fig. 5B). These results confirm
the gene network and expression analysis and demon-
strate that most of these differences are consistent with
differences in the lymphoid/myeloid population in the
adipose tissue of B6 versus 129 mice.
Effects of high-fat diet on the inflammatory status of
the adipose tissue of B6 versus 129 mice. To investi-
gate how differences in the genetic background could
impact the effects of diet-induced obesity on the inflam-
matory cell repertoire of the adipose tissue, cohorts of B6
and 129 mice were placed on a high-fat diet (HFD) for 18
weeks and compared with mice on LFD. On both LFD and
HFD, the mRNA expression of the macrophage marker
CD68 was higher by twofold or greater in B6 compared
with 129 mice (Fig. 6A). Similarly, MCP1 expression was
higher in B6 than in 129 on LFD, and this was maintained
with HFD (Fig. 6A). The macrophage marker F4/80 and
SDF1␣both trended higher in B6 than in 129 mice on LFD,
but neither reached statistical significance (Fig. 6A). Inter-
estingly, the HFD-induced increases in the expression of
the T-cell markers CD3 and Thy1 were completely abro-
gated in the 129 mice (Fig. 6A). Caloric restriction, an
intervention that drastically reduces fat mass, helps pre-
vents age-associated diseases, and prolongs life span (24),
decreased the expression of CD3 and Thy1 in the adipose
tissue of B6 mice by 41 and 51%, respectively, without
affecting the expression of macrophage markers (supple-
mentary Fig. S5). These results demonstrate a positive
correlation between adiposity and the expression of T-
cells in adipose tissue of B6 mice, a phenomenon that is
significantly impaired in 129 mice.
Accordingly, analysis of B6 versus 129 mice under LFD
and HFD using network analysis revealed a substantial
number of genes in the “immune system process” gene set
*** *** **
***
###
###
###
###
##
## ###
###
### ###
##
##
*
*
###
##
##
### ### ###
***
***
**
**
**
** ****
###
A
B6 6wk
1800
Cells (x10
3
) / g of fat
Cells (x10
5
)
1600
1400
1200
1200
800
600
400
200
0
Nonerythrocytes
stromavascular
fraction cells
Leukocytes
Macrophages
CD11c
+
-Myeloid cells
T-cells
T-helper cells
Nonerythrocytes
stromavascular
fraction cells
Leukocytes
Macrophages
CD11c
+
-Myeloid cells
T-cells
T-helper cells
0
5
10
15
20
25
129 6wk
B6 6 mo
129 6 mo
B6 6wk
129 6wk
B6 6 mo
129 6 mo
B
FIG. 5. Immune cell repertoire in adipose tissue of 6-week-old (6 weeks) or 6-month-old (6 months) B6 and 129 mice. Cells in the stromavascular
fraction were labeled with fluorescence-conjugated antibodies against different myeloid cell and lymphoid markers and counted using flow
cytometry. Nonerythrocytes stromavascular cells, PI-/Ter-119-; leukocytes, PI-/Ter-119-/CD45ⴙ; macrophages, PI-/Ter-119-/CD45ⴙ/CD11bⴙ/F4/
80ⴙ/CD11c-; CD11c
ⴙ
-Myeloid cells, PI-/Ter-119-/CD45ⴙ/CD11bⴙ/F4/80ⴙ/CD11cⴙ; T-cells, PI-/Ter-119-/CD45ⴙ/CD3ⴙ; T-helper cells, PI-/Ter-119-/
CD45ⴙ/CD3ⴙ/CD4ⴙ.A: Represents number of cells per gram of fat tissue and (B) the total number of cells in the epididymal fat depot. All values
are mean ⴞSEM of 4 – 8 animals. *P<0.05; **P<0.01; ***P<0.001 vs. 129 mice. #P<0.05; ##P<0.01; ###P<0.001 vs. 6 weeks.
SYSTEMS BIOLOGY AND SUSCEPTIBILITY TO DIABETES
2966 DIABETES, VOL. 59, NOVEMBER 2010 diabetes.diabetesjournals.org

with higher expression in B6 compared with 129 mice, as
evidenced by the large, bright green nodes in Fig. 6B. The
majority of these genes had greater differences between
B6 and 129 mice under HFD than under LFD conditions, as
shown by the high number of upward triangles in Fig. 6.
In particular, T-cell marker CD3 and the subnetwork
around T-helper marker CD4 were among the most up-
regulated genes with HFD in B6 versus 129 mice (arrows
A
0
0.5
1
1.5
2
2.5
F4/80 M CP1 CD3 T h y 1 SDF1
CD68
mRNA ex
(Fold Change)
#
***
#
***
*
**
**
#
**
*
0
0.5
1
1.5
2
2.5 B6 LFD
129 LFD
B6 HFD
129 HFD
B6 LFD
129 LFD
B6 HFD
129 HFD
F4/80 M CP1 CD3 T h y 1CD68
mRNA expression
(Fold Change)
#
***
#
***
*
**
**
#
**
*
#
1
< 0.5
> 2
Fold Change
Fold Change HFD > LFD
Fold Change HFD < LFD
B
IKBKA
CD18
MCP1
SDF1
IKBKA
CD18
MCP1
SDF1
FIG. 6. Expression of inflammatory markers in adipose tissue of 6-month-old B6 and 129 mice fed with LFD or HFD. A: Expression of mRNA
was assessed by qPCR. All values are normalized by TBP and expressed as fold change of the average value of the LFD B6 mice. Results
represent mean ⴞSEM of 4 – 6 animals. *P<0.05; **P<0.01; ***P<0.001 vs. 129 mice. #P<0.05 vs. LFD. B: The network view of the
immune system process differences between B6 and 129 mouse strains on a LFD (left) and HFD (right). The gene network was generated
by mapping genes that were significantly overrepresented in B6 vs. 129 mice (Qvalue <0.25) among the GNEA results on HFD intersected
with those on LFD onto protein-protein interaction networks involving genes annotated with the immune system process gene set. Genes
in red are more than twofold higher in 129 mice compared with B6 mice; genes in green are more than twofold higher in B6 mice compared
with 129 mice. Genes are denoted by a downward triangle if the fold change of the HFD animals is less than those on the LFD, and by an
upward triangle for the converse; the size of the triangle denotes the magnitude of the difference. Arrows represent genes with specific
interest commented on in the main text.
M.A. MORI AND ASSOCIATES
diabetes.diabetesjournals.org DIABETES, VOL. 59, NOVEMBER 2010 2967

in Fig. 6B). Conversely, anti-inflammatory cytokines, such
as interleukin-4 (IL-4) and interleukin-10 (IL-10) were
reduced after HFD in the B6 mice, but not in 129 mice
(arrows in Fig. 6B).
DISCUSSION
Although a relationship between obesity and inflammation
in fat has been previously observed, in most cases these
inflammatory changes have been viewed as being second-
ary to obesity. Our data, comparing diabetes-prone B6
mouse and diabetes-resistant 129 mouse at 6 weeks of age,
show alterations in the inflammatory process in adipose
tissue even before differences in metabolic parameters can
be detected. Thus, B6 mice exhibit increased expression of
the T-cell chemokines SDF1␣and CCL5/RANTES and an
increased number of T-cells in the fat tissue. These differ-
ences are associated with higher IFN␥and CD80 levels in
the B6 mice— both molecules are known to participate in
T-cell function and activation (25) (summarized in the
model shown in Fig. 7).
Recent work has shown that T-cells infiltrate into the
visceral adipose tissue of obese animals and humans with
type 2 diabetes, and this is followed by recruitment of
macrophages and development of insulin resistance (26 –
29). Based on these findings, a model for the role of T-cells
in the pathogenesis of obesity and insulin resistance has
been proposed in which increases in SDF1␣and CCL5/
RANTES levels in adipose tissue occur in response to an
obesogenic environment and promote infiltration with T
lymphocytes (29). IFN␥derived from these T-cells then
promotes MCP1 secretion by preadipocytes (and possibly
other cell types), resulting in recruitment of macrophages
that further contribute to insulin resistance by production
of proinflammatory cytokines (29).
Our study provides evidence to support a major impact
of the genetic background of different mouse strains in the
migration of T-cells to the adipose tissue, both in the basal
state and in response to weight gain. In B6 mice, the
number of T-cells in the adipose tissue correlates posi-
tively with the increase in adipose mass as a consequence
of aging or HFD. By contrast, this response is practically
absent in 129 mice despite a significant increase in adipos-
ity in response to age or HFD. Thus, similar to mice with
ablation of T-cells (27,28), 129 mice develop only mild
insulin resistance in response to obesity. To what extent
aging and the composition of the diet impact the infiltra-
tion of T-cells into adipose tissue, in addition to the effects
of weight gain, remain to be determined. It is clear though
that insulin resistance in response to increased adiposity
differ substantially between mouse strains, and this phe-
nomenon is correlated with the migration of T-cell to the
adipose tissue. Thus, inflammation in the adipose tissue
does not always correlate with weight gain and is strongly
dependent on the genetic background of the host. Similar
differences dependent on genetic background are likely to
occur in humans and contribute to differences in obesity-
induced diabetes risk in different ethnic groups or even
different individuals, allowing for some of the “fat-fit”
phenotype.
Among other implications, these observations clearly
impact the choice of models for metabolic studies, in
particular when knockout mice are used. Knockout mice
are often derived from the embryonic stem cells of 129
mice, and therefore, in many cases, studies are performed
on mice on a pure 129 background or mixed 129/B6
background. In these cases, backcrossing to B6 mice for
several generations usually potentiates the inflammatory
and metabolic response of the model to environmental
factors. However, it is not clear which of these back-
grounds (B6, 129, or mixed) most closely mimic the human
condition.
Other recent reports support our findings that metabolic
disease traits can be associated with alterations of inflam-
matory gene expression networks in the adipose tissue
and liver (30,31). However, to our knowledge, no study has
previously demonstrated a measurable difference in in-
flammation preceding any measurable phenotypic differ-
ence associated with metabolic diseases. The novelty of
our study is that, by comparing young B6 and 129 animals
using a sensitive computational approach, we can focus on
factors that can potentially predispose to disease in a
prospective manner and avoid findings that are mainly
secondary to obesity or metabolic differences. Our inves-
tigation also examined differences between B6 and 129
mice in multiple organs, allowing us to conclude that
inflammation in adipose tissue, and to a lesser extent in
liver, but not in skeletal muscle or spleen, is associated
with the predisposition to insulin resistance. In addition,
our computational analysis was able to set forth hypothe-
ses that led to subsequent biologic validation experiments
which provided further insight into the components of the
immune system that may contribute to metabolic diseases
(i.e., T-cell recruitment).
In addition to inflammation, other phenomena have
been shown to participate in the multifactorial pathophys-
iology of insulin resistance. Alterations in insulin receptor
levels and insulin signaling through IRS-proteins (32),
induction of the unfolded protein response (33) and oxi-
dative stress pathways (34), and changes in lipid (35) and
amino acid metabolism (36) can all promote insulin resis-
tance and contribute to the final phenotype. Many of these
pathways act by producing post-translational modifica-
tions of signaling proteins, such as phosphorylation or
alterations in compartmentalization, which would not be
detected as changes at the gene expression level. In this
regard it is worth noting that in addition to differences in
inflammation, GNEA analysis was able to identify other
gene sets differentially expressed between the mouse
strains in adipose tissue at 6 weeks of age (supplementary
Table S2), including networks related to signal transduc-
tion, protein secretion pathways, and glucose catabolism.
Many of these pathways can interact with inflammatory
pathways, and this crosstalk could represent an entry
point to the manifestation of metabolic diseases. Thus,
although changes in the immune response are definitely
one of the factors that precedes and predicts the tendency
of B6 mice to have greater insulin resistance than 129, it is
unlikely that inflammation is the only predisposing risk
factor associated with diabetes between these two mouse
strains.
In summary, it has been proposed that type 2 diabetes
and obesity are diseases associated with an immune
system that cannot cope appropriately with environmental
threats (37–39). Based on this hypothesis, anti-inflamma-
tory drugs, such as salicylates (40,41) and interleukin-1
blockers (42), have been used to improve glycemia in
individuals with type 2 diabetes. In this study, we demon-
strate that pre-existing differences in the inflammatory
milieu in metabolically active tissues may represent an
important component of the genetic background as a risk
factor to metabolic diseases. Thus, inflammation cannot
SYSTEMS BIOLOGY AND SUSCEPTIBILITY TO DIABETES
2968 DIABETES, VOL. 59, NOVEMBER 2010 diabetes.diabetesjournals.org

be viewed as only a mechanism that links susceptibility
factors such as overfeeding, underactivity, aging, and
stress to metabolic diseases—it may also link these pa-
thologies to heritability. Our study indicates that inflam-
mation is an important early variable in the metabolic
response to environmental challenges and suggests sev-
eral potential targets for intervention, including LBP, Ly86,
SDF1␣, CCL5/RANTES, and MCP1. This provides new
Age/HFD
129
mouse
129
mouse
B6
mouse
B6
mouse
SDF1α
SDF1α
CCL5
CCL5
IFN-γ
IFN-γ
MCP1
MCP1
Monocyte
T-cell
Low
Low
Adipocyte
Macrophage
Low
Blood vessel
SVF cell
SDF1α
SDF1α
CCL5
CCL5
IFN-γ
IFN-γ
MCP1
MCP1
Monocyte
T-cell
Low
Low
Adipocyte
Macrophage
Low
Blood vessel
SVF cell
SDF1α
SDF1α
CCL5
CCL5
IFN-γ
IFN-γ
MCP1
MCP1
Monocyte
T-cell
High
High
Adipocyte
Obesity
Obesity
Obesity
Cytokines
Chemokines
Macrophage
High
Blood vessel
Insulin
Resistance
Insulin
Insulin
Resistance
Resistance
SVF cell
SDF1α
SDF1α
CCL5
CCL5
IFN-γ
IFN-γ
MCP1
MCP1
Monocyte
T-cell
High
High
Adipocyte
Obesity
Obesity
Obesity
Cytokines
Chemokines
Macrophage
High
Blood vessel
Insulin
Resistance
Insulin
Insulin
Resistance
Resistance
SVF cell
129
mouse
129
mouse
B6
mouse
B6
mouse
SDF1α
SDF1α
CCL5
CCL5
IFN-γ
IFN-γ
MCP1
MCP1
Monocyte
T-cell
Low
Low
Adipocyte
Macrophage
Low
Blood vessel
SVF cell
SDF1α
SDF1α
CCL5
CCL5
IFN-γ
IFN-γ
MCP1
MCP1
Monocyte
T-cell
Low
Low
Adipocyte
Macrophage
Low
Blood vessel
SVF cell
SDF1α
SDF1α
CCL5
CCL5
IFN-γ
IFN-γ
MCP1
MCP1
Monocyte
T-cell
High
High
Adipocyte
Obesity
Obesity
Obesity
Cytokines
Chemokines
Macrophage
High
Blood vessel
Insulin
Resistance
Insulin
Insulin
Resistance
Resistance
SVF cell
SDF1α
SDF1α
CCL5
CCL5
IFN-γ
IFN-γ
MCP1
MCP1
Monocyte
T-cell
High
High
Adipocyte
Obesity
Obesity
Obesity
Cytokines
Chemokines
Macrophage
High
Blood vessel
Insulin
Resistance
Insulin
Insulin
Resistance
Resistance
SVF cell
FIG. 7. Schematic model illustrating the potential causes and consequences related to the different repertoire of immune cells in adipose tissue
of B6 and 129 mice. Solid arrows, secretion; dashed arrows, migration; dotted arrows, migration/differentiation.
M.A. MORI AND ASSOCIATES
diabetes.diabetesjournals.org DIABETES, VOL. 59, NOVEMBER 2010 2969

strategies for reducing the epidemic of type 2 diabetes and
metabolic diseases in spite of increasing obesity by attack-
ing the variable genetic risk.
ACKNOWLEDGMENTS
M.A.M., O.B., K.A., and C.R.K. were supported by grants
from the National Institutes of Health: DK-082659, DK-
033201, and DK-060837 (Diabetes Genome Anatomy
Project), as well as the Joslin Diabetes and Endocrinology
Research Center cores (DK-036836) and the Mary K.
Iacocca Professorship. The work of M.L. and S.K. was
supported by grants from the National Human Genome
Research Institute (R01 HG003367-01A1) and the National
Institutes of Health (U54 LM008748) National Center for
Biomedical Computing.
No potential conflicts of interest relevant to this article
were reported.
M.A.M. and M.L. researched data, contributed to discus-
sion, and wrote the manuscript. O.B. researched data and
contributed to discussion. K.A. planned the study and
researched data. H.S. researched data. S.K. researched
data, contributed to discussion, and reviewed/edited the
manuscript. C.R.K. planned the study, researched data,
contributed to discussion, and reviewed/edited the
manuscript.
The authors thank J. LaVecchio, G. Buruzula, and the
flow cytometry core at the Joslin Diabetes Center for help
with cytometry; J. Schroeder and the genomics core at the
Joslin Diabetes Center for help with microarray process-
ing; M. Rourk, Joslin Diabetes Center, for his expertise in
animal care; and S.E. Shoelson, MD, PhD, Joslin Diabetes
Center, for his helpful advice and discussion.
REFERENCES
1. Hossain P, Kawar B, El Nahas M. Obesity and diabetes in the developing
world—a growing challenge. N Engl J Med 2007;356:213–215
2. Lazar MA. How obesity causes diabetes: not a tall tale. Science 2005;307:
373–375
3. Doria A, Patti ME, Kahn CR. The emerging genetic architecture of type 2
diabetes. Cell Metab 2008;8:186 –200
4. Rankinen T, Zuberi A, Chagnon YC, Weisnagel SJ, Argyropoulos G, Walts
B, Perusse L, Bouchard C. The human obesity gene map: the 2005 update.
Obesity (Silver Spring) 2006;14:529 – 644
5. Walley AJ, Asher JE, Froguel P. The genetic contribution to non-syndromic
human obesity. Nat Rev Genet 2009;10:431– 442
6. Almind K, Kulkarni RN, Lannon SM, Kahn CR. Identification of interactive
loci linked to insulin and leptin in mice with genetic insulin resistance.
Diabetes 2003;52:1535–1543
7. Kulkarni RN, Almind K, Goren HJ, Winnay JN, Ueki K, Okada T, Kahn CR.
Impact of genetic background on development of hyperinsulinemia and
diabetes in insulin receptor/insulin receptor substrate-1 double heterozy-
gous mice. Diabetes 2003;52:1528 –1534
8. Almind K, Kahn CR. Genetic determinants of energy expenditure and
insulin resistance in diet-induced obesity in mice. Diabetes 2004;53:3274 –
3285
9. Haluzik M, Colombo C, Gavrilova O, Chua S, Wolf N, Chen M, Stannard B,
Dietz KR, Le RD, Reitman ML. Genetic background (C57BL/6J versus
FVB/N) strongly influences the severity of diabetes and insulin resistance
in ob/ob mice. Endocrinology 2004;145:3258 –3264
10. Bachmanov AA, Reed DR, Tordoff MG, Price RA, Beauchamp GK. Nutrient
preference and diet-induced adiposity in C57BL/6ByJ and 129P3/J mice.
Physiol Behav 2001;72:603– 613
11. Rossmeisl M, Rim JS, Koza RA, Kozak LP. Variation in type 2 diabetes-
related traits in mouse strains susceptible to diet-induced obesity. Diabe-
tes 2003;52:1958 –1966
12. West DB, York B. Dietary fat, genetic predisposition, and obesity: lessons
from animal models. Am J Clin Nutr 1998;67:505S–512S
13. Leiter EH, Lee CH. Mouse models and the genetics of diabetes: is there
evidence for genetic overlap between type 1 and type 2 diabetes? Diabetes
2005;54(Suppl. 2):S151–S158
14. Ferrara CT, Wang P, Neto EC, Stevens RD, Bain JR, Wenner BR, Ilkayeva
OR, Keller MP, Blasiole DA, Kendziorski C, Yandell BS, Newgard CB, Attie
AD. Genetic networks of liver metabolism revealed by integration of
metabolic and transcriptional profiling. PLoS Genet 2008;4:e1000034
15. Schmidt C, Gonzaludo NP, Strunk S, Dahm S, Schuchhardt J, Kleinjung F,
Wuschke S, Joost HG, Al-Hasani H. A meta-analysis of QTL for diabetes-
related traits in rodents. Physiol Genomics 2008;34:42–53
16. Liu M, Liberzon A, Kong SW, Lai WR, Park PJ, Kohane IS, Kasif S.
Network-based analysis of affected biological processes in type 2 diabetes
models. PLoS Genet 2007;3:e96
17. Keshava Prasad TS, Goel R, Kandasamy K, Keerthikumar S, Kumar S,
Mathivanan S, Telikicherla D, Raju R, Shafreen B, Venugopal A, Balakrish-
nan L, Marimuthu A, Banerjee S, Somanathan DS, Sebastian A, Rani S, Ray
S, Harrys Kishore CJ, Kanth S, Ahmed M, Kashyap MK, Mohmood R,
Ramachandra YL, Krishna V, Rahiman BA, Mohan S, Ranganathan P,
Ramabadran S, Chaerkady R, Pandey A. Human Protein Reference Data-
base–2009 update. Nucleic Acid Res 2009;37:D767–D772
18. Ideker T, Ozier O, Schwikowski B, Siegel AF. Discovering regulatory and
signalling circuits in molecular interaction networks. Bioinformatics 2002;
18(Suppl. 1):S233–S240
19. Gesta S, Bluher M, Yamamoto Y, Norris AW, Berndt J, Kralisch S, Boucher
J, Lewis C, Kahn CR. Evidence for a role of developmental genes in the
origin of obesity and body fat distribution. Proc Natl Acad SciUSA
2006;103:6676 – 6681
20. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette
MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP. Gene set
enrichment analysis: a knowledge-based approach for interpreting ge-
nome-wide expression profiles. Proc Natl Acad SciUSA2005;102:15545–
15550
21. Fisher RA. On the interpretation of ␹-square from contingency tables, and
the calculation of P. J Royal Statistical Society 1922;85:87–94
22. Ge H, Liu Z, Church GM, Vidal M. Correlation between transcriptome and
interactome mapping data from Saccharomyces cerevisiae. Nat Genet
2001;29:482– 486
23. Nguyen MT, Favelyukis S, Nguyen AK, Reichart D, Scott PA, Jenn A,
Liu-Bryan R, Glass CK, Neels JG, Olefsky JM. A subpopulation of macro-
phages infiltrates hypertrophic adipose tissue and is activated by free fatty
acids via toll-like receptors 2 and 4 and JNK-dependent pathways. J Biol
Chem 2007;282:35279 –35292
24. Fontana L, Partridge L, Longo VD. Extending healthy life span–from yeast
to humans. Science 2010;328:321–326
25. Freedman AS, Freeman GJ, Rhynhart K, Nadler LM. Selective induction of
B7/BB-1 on interferon-␥stimulated monocytes: a potential mechanism for
amplification of T cell activation through the CD28 pathway. Cell Immunol
1991;137:429 – 437
26. Feuerer M, Herrero L, Cipolletta D, Naaz A, Wong J, Nayer A, Lee J,
Goldfine AB, Benoist C, Shoelson S, Mathis D. Lean, but not obese, fat is
enriched for a unique population of regulatory T cells that affect metabolic
parameters. Nat Med 2009;15:930 –939
27. Winer S, Chan Y, Paltser G, Truong D, Tsui H, Bahrami J, Dorfman R, Wang
Y, Zielenski J, Mastronardi F, Maezawa Y, Drucker DJ, Engleman E, Winer
D, Dosch HM. Normalization of obesity-associated insulin resistance
through immunotherapy. Nat Med 2009;15:921–929
28. Nishimura S, Manabe I, Nagasaki M, Eto K, Yamashita H, Ohsugi M, Otsu
M, Hara K, Ueki K, Sugiura S, Yoshimura K, Kadowaki T, Nagai R. CD8⫹
effector T cells contribute to macrophage recruitment and adipose tissue
inflammation in obesity. Nat Med 2009;15:914 –920
29. Kintscher U, Hartge M, Hess K, Foryst-Ludwig A, Clemenz M, Wabitsch M,
Fischer-Posovszky P, Barth TF, Dragun D, Skurk T, Hauner H, Bluher M,
Unger T, Wolf AM, Knippschild U, Hombach V, Marx N. T-lymphocyte
infiltration in visceral adipose tissue: a primary event in adipose tissue
inflammation and the development of obesity-mediated insulin resistance.
Arterioscler Thromb Vasc Biol 2008;28:1304 –1310
30. Emilsson V, Thorleifsson G, Zhang B, Leonardson AS, Zink F, Zhu J,
Carlson S, Helgason A, Walters GB, Gunnarsdottir S, Mouy M, Steinthors-
dottir V, Eiriksdottir GH, Bjornsdottir G, Reynisdottir I, Gudbjartsson D,
Helgadottir A, Jonasdottir A, Jonasdottir A, Styrkarsdottir U, Gretarsdottir
S, Magnusson KP, Stefansson H, Fossdal R, Kristjansson K, Gislason HG,
Stefansson T, Leifsson BG, Thorsteinsdottir U, Lamb JR, Gulcher JR,
Reitman ML, Kong A, Schadt EE, Stefansson K. Genetics of gene expres-
sion and its effect on disease. Nature 2008;452:423–428
31. Chen Y, Zhu J, Lum PY, Yang X, Pinto S, Macneil DJ, Zhang C, Lamb J,
Edwards S, Sieberts SK, Leonardson A, Castellini LW, Wang S, Champy
MF, Zhang B, Emilsson V, Doss S, Ghazalpour A, Horvath S, Drake TA,
Lusis AJ, Schadt EE. Variations in DNA elucidate molecular networks that
cause disease. Nature 2008;452:429 – 435
32. Bruning JC, Winnay J, Bonner-Weir S, Taylor SI, Accili D, Kahn CR.
SYSTEMS BIOLOGY AND SUSCEPTIBILITY TO DIABETES
2970 DIABETES, VOL. 59, NOVEMBER 2010 diabetes.diabetesjournals.org

Development of a novel polygenic model of NIDDM in mice heterozygous
for IR and IRS-1 null alleles. Cell 1997;88:561–572
33. Ozcan U, Cao Q, Yilmaz E, Lee AH, Iwakoshi NN, Ozdelen E, Tuncman
G, Gorgun C, Glimcher LH, Hotamisligil GS. Endoplasmic reticulum
stress links obesity, insulin action, and type 2 diabetes. Science
2004;306:457– 461
34. Houstis N, Rosen ED, Lander ES. Reactive oxygen species have a causal
role in multiple forms of insulin resistance. Nature 2006;440:944 –948
35. Boden G. Role of fatty acids in the pathogenesis of insulin resistance and
NIDDM. Diabetes 1997;46:3–10
36. Newgard CB, An J, Bain JR, Muehlbauer MJ, Stevens RD, Lien LF, Haqq
AM, Shah SH, Arlotto M, Slentz CA, Rochon J, Gallup D, Ilkayeva O,
Wenner BR, Yancy WS Jr, Eisenson H, Musante G, Surwit RS, Millington
DS, Butler MD, Svetkey LP. A branched-chain amino acid-related meta-
bolic signature that differentiates obese and lean humans and contributes
to insulin resistance. Cell Metab 2009;9:311–326
37. Fernandez-Real JM, Pickup JC. Innate immunity, insulin resistance and
type 2 diabetes. Trends Endocrinol Metab 2008;19:10 –16
38. Hotamisligil GS, Erbay E. Nutrient sensing and inflammation in metabolic
diseases. Nat Rev Immunol 2008;8:923–934
39. Schenk S, Saberi M, Olefsky JM. Insulin sensitivity: modulation by nutri-
ents and inflammation. J Clin Invest 2008;118:2992–3002
40. Goldfine AB, Silver R, Aldhahi W, Cai D, Tatro E, Lee J, Shoelson SE. Use
of salsalate to target inflammation in the treatment of insulin resistance
and type 2 diabetes. Clin Transl Sci 2008;1:36 – 43
41. Hundal RS, Petersen KF, Mayerson AB, Randhawa PS, Inzucchi S,
Shoelson SE, Shulman GI. Mechanism by which high-dose aspirin
improves glucose metabolism in type 2 diabetes. J Clin Invest 2002;109:
1321–1326
42. Larsen CM, Faulenbach M, Vaag A, Volund A, Ehses JA, Seifert B,
Mandrup-Poulsen T, Donath MY. Interleukin-1-receptor antagonist in type
2 diabetes mellitus. N Engl J Med 2007;356:1517–1526
M.A. MORI AND ASSOCIATES
diabetes.diabetesjournals.org DIABETES, VOL. 59, NOVEMBER 2010 2971