Nihms 599061

NIH Public Access
Author Manuscript
Cell Metab. Author manuscript; available in PMC 2015 July 01.
Published in final edited form as:
NIH-PA Author Manuscript
Cell Metab. 2014 July 1; 20(1): 103–118. doi:10.1016/j.cmet.2014.05.005.
Adipocyte Inflammation is Essential for Healthy Adipose Tissue

Expansion and Remodeling
Ingrid Wernstedt Asterholm1,*, Caroline Tao1, Thomas S. Morley1, Qiong A. Wang1,
Fernando Delgado-Lopez2, Zhao V. Wang1, and Philipp E. Scherer1,3
1Touchstone Diabetes Center, Dept. of Internal Medicine, UTSW Medical Center, 5323 Harry
Hines Blvd., Dallas, TX 75390-8549, USA
2Facultad de Medicina, Universidad Catolica del Maule, Avda. San Miguel 3605, Talca, Chile
3Cell Biology, UTSW Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75390-8549, USA
Summary
Chronic inflammation constitutes an important link between obesity and its pathophysiological
sequelae. In contrast to the belief that inflammatory signals exert a fundamentally negative impact
on metabolism, we show that pro-inflammatory signaling in the adipocyte is in fact required for
proper adipose tissue remodeling and expansion. Three mouse models with an adipose tissue-
specific reduction in pro-inflammatory potential were generated that display a reduced capacity
for adipogenesis in vivo, while the differentiation potential is unaffected in vitro. Upon high fat
diet exposure, the expansion of visceral adipose tissue is prominently affected. This is associated
with decreased intestinal barrier function, increased hepatic steatosis and metabolic dysfunction.
An impaired local pro-inflammatory response in the adipocyte leads to increased ectopic lipid
accumulation, glucose intolerance and systemic inflammation. Adipose tissue inflammation is
therefore an adaptive response that enables safe storage of excess nutrients and contributes to a
visceral depot barrier that effectively filters gut-derived endotoxin.
© 2014 Elsevier Inc. All rights reserved.

Correspondence to: Philipp E. Scherer, Touchstone Diabetes Center, Dept. of Internal Medicine, UTSW Medical Center, 5323 Harry
Hines Blvd., Dallas, TX 75390-8549, USA, [email protected], Tel: 214-648-8715; Fax: 214-648-8720.
*current affiliation: Box 432, Inst. of Neuroscience and Physiology, Sahlgrenska Academy at the University of Gothenburg, 405 30
Gothenburg, Sweden
Supplemental Information
Supplemental information includes experimental procedures, six additional figures and three tables. They appear with this article
online at XXX.
Author contributions
IWA designed the study, carried out the research, interpreted the results, and wrote the manuscript. CT, TSM, QAW, FDL and ZVW
assisted in study design, performed research, and reviewed the manuscript. PES designed the study, analysed the data, reviewed and
revised the manuscript, and is responsible for the integrity of this work. All authors approved the final version of the manuscript.
The authors declare no conflicts of interest.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our
customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of
the resulting proof before it is published in its final citable form. Please note that during the production process errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Asterholm et al. Page 2
Introduction
Adipose tissue expansion in response to excess caloric intake is an important systemic
response to avoid the lipotoxic side effects exerted by excess lipid and fatty acid (FA)
deposition in cells other than adipocytes. The basic mechanisms, leading to a gradual and
“healthy” expansion of fat pads, are starting to be elucidated. Healthy expansion is
associated with appropriate angiogenesis and vascular and extracellular matrix (ECM)
remodeling.
Increased adiposity is more often than not associated with an increased risk for a number of
chronic diseases, including diabetes, cardiovascular disease and some types of cancers (Park
et al., 2011). The underlying mechanisms for the link between obesity and these diseases are
not fully understood, but are likely to involve a state of chronic systemic low-grade
inflammation.
The causality between local adipose tissue inflammation, systemic inflammation and
metabolic dysfunction has however not been studied. Therefore, we developed three distinct,
but complementary mouse models to investigate the role of adipose tissue inflammation in
high fat diet (HFD)-induced metabolic disturbances. By design, two models express the anti-
inflammatory factors in adipose tissue constitutively, while one model is inducible.
Analysis of these three mouse models reveals that the inability to mount an appropriate local
pro-inflammatory response at the level of the adipocyte reduces adipose tissue expansion
under normal physiological as well as under HFD-fed conditions. This inability to expand
adipose tissue is associated with ectopic lipid deposition and a deteriorated metabolic
profile. Furthermore, we demonstrate that mesenteric adipose tissue (MWAT, a visceral fat
depot) plays an important role for proper intestinal barrier function. An ineffective response
of MWAT to pro-inflammatory stimuli with respect to its expansion is associated with a
“leaky gut”, colitis and metabolic dysfunction. Thus, these mouse models demonstrate that a
local inflammatory response derived from the adipocyte is an adaptive response and an
important preemptive factor for ensuing obesity-associated systemic inflammation.
Results
We have recently developed a mouse model (the “adipochaser mouse”) in which we can
permanently activate β-galactosidase-expression in all preexisting adipocytes by a short bout

of doxycycline-treatment. Removal of doxycycline enables the detection of newly
differentiated adipocytes that are negative for the blue X-Gal-LacZ staining (Wang et al.,
2013). Repeated local LPS injections into adipose tissue stimulate adipogenesis without
affecting overall weight gain (Sadler et al., 2005). Exploiting our adipochaser mouse, we are
able to confirm the findings by Sadler and colleagues. New adipocyte formation was evident
in the LPS-injected inguinal WAT (IWAT) depot, but not in the control depot (Fig. 1A and
data not shown). These observations suggest that the induction of acute inflammation in the
context of intact adipose tissue stimulates adipogenesis in vivo. We therefore hypothesized
that acute inflammation within adipose tissue may play an essential role for adipose tissue
expansion, remodeling and overall homeostasis by stimulating ECM degradation and

angiogenesis. To study the role of inflammation in adipose tissue, we developed a mouse

model expressing a dominant-negative version of the potent pro-inflammatory cytokine
TNFα (dnTNF) (Steed et al., 2003) under the control of the ap2 promoter (“dnTNF tg”).
This is a version of TNFα that effectively hetero-trimerizes with wildtype TNFα subunits,
but the presence of a mutant subunit engages the TNF-receptors non-productively. aP2
promoter-driven constructs are predominantly transcribed in adipocytes, but expression has
been reported for macrophages and other tissues, such as the heart and the brain, though at
much lower levels. Our dnTNF tg mice express the transgene specifically in adipose tissue,
but also to a minor extent in isolated peritoneal macrophages (Fig. S1A). The low level of
dnTNF expression in macrophages does not affect the inflammatory state of thioglycollate-
activated peritoneal macrophages, since a number of cytokines and M1/2 markers (e.g.
TNFα, IL-1β, TGFβ, Arg1 and CD11c) remained unaltered compared to macrophages
isolated from littermate controls (Fig. S1B).
As expected, unchallenged chow-fed dnTNF tg do not display significant differences with

respect to genes involved in inflammation or macrophage polarization e.g. TNFα, F4/80,
CD11c, NOS2, CD206 and CD301 in adipose tissue (data not shown). However, the dnTNF
tg mice display a reduced NFκB activation in adipose tissue (as judged by reduced IκBα
phosphorylation) in response to an intraperitoneal injection with TNFα (Fig. S1C).

Furthermore, LPS induces a highly specific response on adipocytes in adipose tissue,
leading to an acute increase in serum leptin levels. In the presence of dnTNF, this LPS-
mediated increase in serum leptin levels is significantly blunted (Fig. 1B). This indicates that
an autocrine TNFα loop is part of the LPS-mediated effects on leptin release. Even though
the increase in serum SAA and α1acid-glycoprotein (Fig S1D and data not shown), the body
weight recovery is slightly faster in the dnTNF tg mice (Fig. 1C). Combined, these results
confirm that the dnTNF transgene is indeed functional, and that it exerts its effects primarily
at the level of adipose tissue, while leaving the inflammatory response in other tissues, such
as the liver, fully intact.
Reduced body weight, fat mass and glucose tolerance in HFD-fed dnTNF tg mice
In young chow-fed mice, the body weights trend to be lower in the dnTNF mice. The IWAT
weight is reduced, while the gonadal WAT (GWAT) weight is similar between genotypes
(Fig. 1D). This pattern changes upon HFD-feeding. The degree of HFD-induced obesity is
reduced in the dnTNF mice (Fig. 1E). This is particularly apparent in GWAT, which is
much more prominently affected than the IWAT depot in HFD-fed mice (Fig. 1D). In
absolute terms, both GWAT and IWAT are of reduced size in HFD-fed dnTNF tg mice
compared to littermates. The reduced IWAT weight is however proportional to the lower
body weight in the dnTNF mice. Brown adipose tissue (BAT) and MWAT are however of
similar sizes in dnTNF and littermate controls, regardless of diet (data not shown). Similar
results were seen when the dnTNF transgene was bred into the genetically obese ob/ob
background, indicating that there are leptin-independent mechanisms responsible for the
reduced body and GWAT mass weight in dnTNF mice (Fig. S1E).
We went on to test whether the reduced body weight translates into the expected
improvements in metabolic parameters. To our surprise, HFD-fed dnTNF tg mice are

severely glucose intolerant and have lower adiponectin levels, despite lower levels of
circulating SAA compared to littermate controls (Fig. 1F-H).
These observations suggest that adipose tissue TNFα-signaling is relevant for adipose tissue
remodeling and expansion. Consistent with this hypothesis, we found increased amounts of
ECM deposits in IWAT of HFD-fed dnTNF mice compared to littermate controls (Fig. 1I).
This result is consistent with a need for TNFα for successful ECM-remodeling in the
context of wound healing (Heo et al., 2011; Saika et al., 2006). The HFD-induced adipocyte
hypertrophy in IWAT is similar between genotypes (relative adipocyte sizes: 1.00 ± 0.3 and
0.94 ± 0.1 for respectively wildtype and dnTNF tg IWAT, p=0.84), despite about 60%
reduced IWAT depot weight in the dnTNF tg mice (Fig 1D), suggesting that inhibition of
TNFα-signaling either causes a reduced capacity for adipogenesis in vivo, or may simply be
secondary to the increased fibrosis. Disruption of this fibrotic state, such as in the context of
HFD-exposed mice lacking collagen VI (Khan et al., 2009), reduces adipocyte apoptosis. In
contrast, enhanced fibrosis increases the rate of adipocyte death (Halberg et al., 2009). In
line with fibrosis-induced adipocyte death, prolonged HFD-feeding (22 weeks) leads to an
increased presence of crown-like structures (CLS’s) in the GWAT of dnTNF tg mice. This is
apparent when examining infiltrating macrophages (as judged by Mac2-staining)
surrounding dead adipocytes (as judged by perilipin-negative staining) (Fig. 1J). We did
however not detect a difference in adipocyte death in IWAT and no difference was observed
in GWAT at earlier stages of HFD-feeding (data not shown). Thus, a reduction in
adipogenesis is more likely to explain the enlarged adipocytes in IWAT and the reduced
GWAT size in HFD-fed dnTNF tg mice.
Essential role for acute inflammation for adipose tissue functionality

Given the many studies showing the negative impact of TNFα on insulin sensitivity as well
as on adipocyte differentiation (Engelman et al., 2000; Gustafson and Smith, 2006;
Hotamisligil et al., 1993), the metabolic and adipose tissue dysfunction seen for the dnTNF
tg mice is rather surprising at first sight. However, our observations do not contradict a
model in which chronic inflammation is an important contributor towards the metabolic
syndrome. Rather though, “immunologic fitness” as we have previously defined it
(Asterholm et al., 2012), seems to be an important component for tissue homeostasis in
general and for adipose tissue in particular. To further explore this concept of “physiological
adipose tissue inflammation”, we wanted to test an additional model to strengthen the

general validity of our initial findings in these dnTNF tg mice. To do so, we generated a
more potent adipose-specific anti-inflammatory model. This second mouse model takes
advantage of RIDα/β (RID), an adenoviral protein complex that suppresses the local host
immune response by potently inhibiting a number of pro-inflammatory signaling pathways
(e.g. TLR4-, TNFα- and IL-1β-mediated signaling) (Lichtenstein et al., 2004). Similar to the
dnTNF tg mice, we put the expression of RID under the control of the aP2 promoter (RID
tg). Just as for the dnTNF tg mouse model, we detected a high transgene expression in
adipocytes with low level expression also seen in macrophages, but not in other tissues (Fig.
S2A). Upon isolating fresh peritoneal thioglycollate-stimulated macrophages from wildtype
and RID mice, we did not detect any functional differences in isolated and cultured
macrophages, suggesting that the trace levels of expression do not amount to an effective

suppression of inflammation in macrophages (Fig. S2B). Consistent with a potent action

within the adipocyte, the RID tg mice display a significantly blunted response to LPS in
adipose tissue, associated with a reduced response to LPS-induced body weight loss as well
(Fig. 2A and B). Moreover, and also similar to the dnTNF tg mice, fad pad weights of the
RID tg mice are reduced. The more effective anti-inflammatory effects in the RID tg lead to
a significant reduction in both IWAT and GWAT depot sizes, even in young chow-fed mice
(Fig. 2C). HFD-feeding causes an even more dramatic difference with respect to the visceral
GWAT and MWAT depots, while the subcutaneous IWAT depot no longer differs in size
from the wildtype controls under these conditions (Fig. 2D). BAT weights are similar
between genotypes, with or without HFD-challenge (data not shown). Despite the reduced
overall amounts of white adipose tissue, the RID tg mice do not display altered overall body
weight (Fig. S2C).
Body composition, as assessed by NMR, confirmed a slightly reduced fat mass in both the
dnTNF and the RID tg mice compared to wild type controls, and as little as two days of
HFD-feeding enhances this difference between genotypes (Fig. S2D and E). Lean body
mass was however unaltered in both mouse models (Fig. S2D and E). We also investigated
whether these transgenic mice display an altered energy balance, but neither food intake nor
oxygen consumption differed from their respective controls (data not shown). Thus, the
reduced body weight in the dnTNF must be caused by an alteration in energy balance that is
too small to be detected with available methods. In contrast, the respiratory quotient
(RQ=VO2/VCO2) was significantly altered. HFD-fed dnTNF tg mice display a lower RQ in
the dark phase and the RID tg mice display a lower dark phase-RQ on both chow and HFD.
This reduction in RQ in both dnTNF and RID tg mice compared to their controls indicates a
heavier reliance on fatty acid oxidation for their energy need (Fig. S2F and G). Typically,
healthy mice burn carbohydrates during the dark phase. At this time of day, they have the
highest food intake and hence the highest insulin levels. Thus, a the lower RQ during the
dark phase, reflective of reduced carbohydrate use, is an indication of systemic insulin
resistance and metabolic inflexibility in both transgenic strains.
The inhibition of adipogenesis (as gauged by the estimated number of adipocytes) is even
more pronounced in the RID tg mice (Fig. 2E). Chow-fed RID tg mice display 17 ± 2 %
larger adipocytes in IWAT (p<0.05) and a trend towards larger adipocytes in GWAT (27%
± 11 larger, p=0.14) despite the fact that these depots are 25% (IWAT) and 30% (GWAT)
smaller than in wild type controls (Fig. 2C). The circulating SAA levels are to our surprise
higher in chow-fed RID tg mice, but this difference disappears on HFD i.e. the HFD-
induced increase in SAA is more pronounced in wild type mice (Fig. S2H). Similar to the
dnTNF tg mice, the RID tg adipose phenotype is associated with both reduced adiponectin
levels and glucose intolerance (Fig. 2F-G). In fact, RID tg mice have a substantial degree of
glucose intolerance with mild hyperinsulinemia already under unchallenged, chow-fed
conditions at a young age. It could very well be that the lower adiponectin levels contribute
to this impaired insulin sensitivity. HFD-feeding aggravates the metabolic phenotype
further, and the RID tg mice continue to display reduced glucose tolerance despite severe
hyperinsulinemia relative to the wildtype controls (Fig. 2H-I). Upon closer analysis of the
adipose tissue, we found elevated levels of collagen in HFD-fed RID tg IWAT.

Interestingly, in these experiments, we noticed that HFD induces a rapid and dramatic
reduction of septa in IWAT. These septa are easily detectable with a Picrosirius red stain,
while somewhat less apparent with Trichrome stain, and correspond to collagen streaks that
compartmentalize adipose tissue “units” in young wildtype mice (Fig. 3A and Fig. S3A).
There is no description in the literature of either the developmental origin or the relevance of
these functional “mini-units” in adipose tissue. The reduction in septa can be quantified by
total collagen measurements. HFD-fed wildtype, but not RID tg mice, have a reduced total
amount of collagen per mg adipose tissue compared to the levels in chow-fed controls (Fig.
3B). Thus, adipose tissue fibrosis in obesity may, at least in some cases, be the consequence
of reduced ECM-degradation, rather than an increase in ECM production. We should also
note that total collagen content measurements of adipose tissue do not necessarily reflect
states of pathological fibrosis, since healthy chow-fed mice have higher levels of collagen in
IWAT than metabolically challenged obese mice. Instead, these data suggest that beyond
biochemical measurements, a histological examination also has to be performed to
accurately assess whether a pathological condition prevails (e.g. the collagen deposits show
up as irregular streaks or in association with crown-like structures) or whether the collagen
deposits are in the form of well-organized septa, as seen in the lean animals.
Reduced HFD-induced adipose tissue expansion in dnTNF and RID tg mice is associated
with hepatic steatosis
These initial observations support the hypothesis that the ability to mount a pro-
inflammatory response is intimately coupled to adipose tissue expansion, proper remodeling
and the resolution of inflammation. We therefore wanted to determine whether the reduced
storage capacity in adipose tissue leads to increased ectopic lipid deposition. Indeed, both
the dnTNF and the RID tg mice display an increased degree of HFD-induced hepatic
steatosis (Fig. 3C-D). Given the blunted HFD-induced weight gain in the dnTNF tg mice
and the higher anti-inflammatory potency of the RID transgene, it is not surprising that the
effects on the liver are more severe in the RID tg mice. In addition to the increased amounts
of liver fat, RID tg mice also display a high level of hepatomegaly during HFD-fed
conditions (Fig. 3D). Thus, the increased ectopic lipid deposition in dnTNF and RID tg mice
can, at least in part, explain the more severe degree of HFD-induced metabolic dysfunction
compared to controls in these mouse models.
Adipocyte inflammation is important for early postnatal adipogenesis

The observations so far argue that the reduced adipose tissue mass in the dnTNF and the
RID tg mice is the consequence of a decreased adipogenesis rate, partially offset by a more
pronounced hypertrophy of existing adipocytes. Moreover, the reduced overall fat mass
cannot solely be the consequence of an altered energy balance since only the dnTNF, but not
the RID transgenic mice display lower body weights; and adipose depot sizes in both the
dnTNF and the RID tg mice are disproportionately reduced beyond what is expected from
the body weight differentials to wildtype mice. We were therefore prompted to further
explore the phenomenon of reduced adipose expansion and remodeling in these anti-
inflammatory models. First, we explored whether adipogenesis is already negatively
affected during postnatal development. Immediately at birth, wildtype mice rapidly expand
their IWAT depots. IWAT pads grow substantially within 1-2 weeks, while the visceral

adipose depots remain almost undetectable at this young age. We measured the differences
in adipose tissue size in 10-day old pups. Indeed, the RID tg pups display fewer, but larger
adipocytes in developing IWAT, while body weights are unaltered compared to wildtype
controls (Fig. 3E-F and data not shown). This difference is associated with a decrease in
vascular density as judged by IHC stains of endomucin of IWAT (Fig. 3E-F). The dnTNF
mice display a similar, albeit less dramatic, phenotype. The 10-day old dnTNF tg pups have
a slightly reduced total body weight, though the IWAT weight is reduced both in relation to
body weight and in absolute terms compared to littermate controls (Fig. 3G). The dnTNF
adipocytes in GWAT are larger, while the adipocytes in IWAT are of similar size compared
to wildtype adipocytes on postnatal day 10 (Fig. 3G). Even the dnTNF tg mice have
therefore a reduced number of adipocytes, since the adipose depots are overall smaller.
There is no evidence of dead adipocytes in either GWAT or IWAT in any of the genotypes
(Fig S3B for RID tg and dnTNF tg data not shown).
A concern in this context is that both RID tg and dnTNF could lead to inhibition of
adipogenesis due to cytotoxic effects or due to unspecific interference with the
differentiation program. To investigate this possibility, we isolated stromal vascular cells
from IWAT of wildtype, dnTNF and RID tg mice, propagated them in culture and then
subjected the cells to an in vitro differentiation protocol and assessed the degree of
differentiation by several different criteria (Fig. S3C-E). All three cell preparations
differentiated to the same extent as judged by appearance of lipid droplets in bright field
microscopy (Fig. S3C), Oil Red O staining (Fig. S3D) and by immunoblotting for the
adipocyte marker adiponectin (Fig. S3E). Thus, the reduced adipogenic potential in the
transgenic mice is truly a function of the reduced ability of these cells to respond to external
pro-inflammatory stimuli in the context of an intact adipose tissue depot, rather than a cell-
autonomous differentiation defect due to transgene expression.
The first few postnatal days are associated with an increased exposure to many new antigens
due to microbial colonization of the gut, concomitant with a sudden intake of large
quantities of milk. A large intake of lipids has been shown to acutely cause inflammatory
responses in various tissues, including adipose tissue (Asterholm et al., 2012; Magne et al.,
2010). We hypothesize that the likely increase in the systemic levels of bacterial toxins
further elevates the lipid-induced pro-inflammatory response, and facilitates angiogenesis
that in turn is permissive for adipogenesis (Fig. S4A). Hence, we suggest that gut bacteria
play a major role in proper storage of excess nutrients during this early postnatal stage. To
test this hypothesis, we exposed wildtype and RID transgenic mice for five weeks prior to
mating to a chow diet supplemented with antibiotics that effectively deplete the commensal
microflora (Rakoff-Nahoum et al., 2004). The antibiotics were removed from the food
between E0-16 and re-introduced again at day E17 to avoid potentially harmful effects of
these drugs on fetal development. There was no effect on the offspring’s body weight,
neither by antibiotics nor by genotype, and all pups survived and appeared healthy
regardless of treatment group. We found that antibiotic-treated wildtype offspring display
reduced amounts of adipose tissue at 10 days of age compared to the untreated control mice.
The RID tg pups display a reduced amount of adipose tissue relative to wildtype mice,
regardless of whether they were on antibiotics or not (Fig. 4A). Thus, the presence or

absence of bacterially-derived toxins in the RID tg pups has no effect on adipose tissue
growth, supporting the notion that a local inflammatory response is an important component
of normal adipose tissue expansion. Furthermore, the male breeders used in this study were
sacrificed after five weeks of treatment and their MWAT was collected for histological
analysis. We found that antibiotic-treated wildtype mice have a reduced capillary density in
their MWAT compared to controls (Fig. 4B). This observation provides additional support
for the idea that bacterially-induced inflammation leads to increased angiogenesis, at least
locally in MWAT (Fig. S4B). RID tg mice had a reduced capillary density, even in the
absence of antibiotic treatment (Fig. 4B). Furthermore, the average MWAT adipocyte size is
20 ± 5% (p<0.05) larger while the MWAT depot weight is 73 ± 7% reduced, indicating a
dramatic reduction in the number of MWAT adipocytes in RID tg mice compared to wild
type controls. Antibiotic treatment did not have an impact on adipocyte size, but tended to
reduce the capillary density even further in the RID tg MWAT, suggesting that additional
cells beyond adipocytes play a role in the response to microbial toxins in adult MWAT (Fig.
4B).
Reduced adipogenesis and HFD-induced glucose intolerance in an inducible adipocyte-

specific anti-inflammatory model
Since the effects seen adipose tissue expansion and metabolic health were quite striking in
the RID tg and dnTNF mice, we wanted to test whether we can see a similar phenomenon in
an adipocyte-specific inducible model. To that end, we generated a mouse that expresses a
mutated human IκBα(S32G-S36A) version under the control of a tet-responsive element
(TRE-IκB tg). IκBα(S32G-S36A) is an effective inhibitor of the NFκB pathway. This TRE-
IκB tg model was crosses with transgenic mice expressing the “tet-on” transcription factor
rtTA under the control of the highly adipocyte-specific adiponectin promoter (Ad-rtTA tg)
(Wang et al., 2010). Upon exposing Ad-rtTA-TRE-IκB tg mice to doxycycline, κBIα(S32G-
S36A) mRNA gets selectively induced in adipocytes, whereas no expression is observed in
other tissues, such as the liver (Fig. S4C). When we analyzed the impact of the expression of
this anti-inflammatory protein during late gestation and the first 10 days of the postnatal
period by exposing both wildtype and transgenic dams to doxycycline, the pups expressing
IκBα(S32G-S36A) displayed a reduced IWAT weight while no effect on body weight was
observed (Fig. 4C). The adipocyte sizes were however similar between genotypes (data not
shown), arguing for lower total number of inguinal adipocytes in 10 day-old Ad-rtTA-TRE-
IκB tg pups. Similar to the other models, there was no evidence of any dead adipocytes, as
judged by perilipin stain (data not shown).
We investigated the metabolic phenotype of Ad-rtTA-TRE-IκB tg mice. We found no

difference between genotypes in either body weight or glucose tolerance on doxycycline-
supplemented chow (data not shown). Challenging the mice with HFD for 8 weeks revealed
however a significant difference between genotypes. Ad-rtTA-TRE-IκB tg mice were more
glucose intolerant than littermate controls, despite comparable body weights (Fig. 4D).
Similar to the dnTNF and the RID tg mice, the GWAT weight was reduced while the liver
weights were increased in the HFD-fed Ad-rtTA-TRE-IκB mice (Fig. 4E). In contrast, there
was no difference in IWAT, MWAT and BAT weights (Fig. 4E and data not shown). To
assess whether the Ad-rtTA-TRE-IκB mice are more susceptible to develop HFD-induced

hepatic steatosis, livers were harvested from a second cohort of mice after 1, 6 and 10 weeks
on doxycycline-supplemented HFD. We found a strong trend towards increased HFD-
induced hepatic steatosis in relation to body weight (Fig. S4D). These findings confirm, in a
third independent system, that the inability to mount a pro-inflammatory response in adipose
tissue impairs adipose tissue expansion associated with metabolic dysfunction.
Impaired β3-Adrenergic Receptor (AR) agonist-induced adipose tissue remodeling in RID

tg mice
The adipose tissue response to chronic β3-AR agonist involves a transient acute bout of
inflammation and, over time, triggers “browning” of the IWAT as defined by an increased
number of multilocular cells expressing uncoupling protein (UCP)-1 (Granneman et al.,
2005; Mottillo et al., 2010). There is also evidence for increased adipogenesis in GWAT in
response to chronic β3AR-agonist treatment (Lee et al., 2012; Wang et al., 2013). We aimed
to investigate whether adipose tissue “beiging” depends on a pro-inflammatory response.
We examined one of our models, the RID tg mice and exposed them and their wildtype
controls to daily β3AR-agonist and bromodeoxyuridine (BrdU) injections for 10 days.
Rather dramatic differences were apparent with respect to the response to this chronic β3AR-
agonist treatment. Wildtype adipose tissues displayed a much “browner” color than the RID
tg adipose tissue (Fig. 5A), and IWAT contained a large number of multilocular adipocytes,
while RID IWAT only displays a modest response in this respect (Fig. 5B). In line with
these observations, the resulting UCP-1 mRNA levels are also reduced in both IWAT and
GWAT the RID tg mice (Fig. S5A). Furthermore, very few BrdU-positive adipocytes were
found in GWAT after chronic β3AR-agonist in RID tg mice, while several positive cells are
found in the wildtype GWAT (Fig. 5C). We measured β3-AR mRNA expression as well as
the acute lipolytic response to β3AR-agonist treatment and found no difference in β3-AR
expression (Fig. S5A). The β3AR-agonist-induced serum FFA increase is just marginally
affected in the RID mice (Fig. S5B). This eliminates receptor abundance and activity as a
trivial explanation for these findings. Along the same rationale, receptor abundance and
signaling are therefore unlikely to explain the differences in β3-AR agonist-induced
“beiging” between genotypes. We have previously reported that inducible adipocyte-specific
expression of VEGF-A leads to a “browning” of adipose tissue, similar to the effects
reported for chronic β3-AR agonist treatment (Sun et al., 2012). Therefore, we speculated
that the lack of β3-AR agonist induced-“beiging” in the RID tg mice may relate to a reduced
ability to induce pro-angiogenic mediators. Indeed, upon examining the gene expression
response in IWAT within 3 hours of a single dose of β3-AR agonist, we find a blunted
induction of several pro-inflammatory cytokines in the RID tg mice (Fig. 5D). Furthermore,
the mRNA expression levels of classical pro-angiogenic mediators, such as VEGF-A,
angiopoetin-1 and -2 (Angpt1 and Angpt2) are also reduced in the β3-AR agonist treated
RID tg mice (Fig. 5D). Notably, we did not detect an acute β3-AR agonist-mediated
induction of these pro-angiogenic mediators in the wildtype mice. Rather, the difference
between genotypes relates mainly to a down-regulation of these particular genes in response
to β3-AR agonist in the RID tg mice. Thus, in the context of acute β3-AR agonist
stimulation, it seems that a potent inflammatory response is necessary to maintain the
expression of VEGF-A, Angpt1 and Angpt2 in IWAT. A comparable gene expression

pattern was also seen in GWAT, albeit the individual variation is larger than in IWAT (Fig.
S5C).
The reduced MWAT expansion in the RID tg mice is associated with a “leaky gut”
The RID tg mice have a more severe metabolic phenotype than the dnTNF tg and the Ad-
rtTA-TRE-IκB tg mice. It caught our attention that the dnTNF tg and the Ad-rtTA-TRE-IκB
tg mice display a normal HFD-induced MWAT expansion and mesenteric adipocyte size
(data not shown). In contrast, MWAT expansion is dramatically reduced in the RID tg mice
(Fig. 2D). This indicates that inhibition of inflammatory signaling in our mouse models is
compensated for by an increase in other inflammatory signaling pathways in the mesenteric
area. These compensatory mechanisms are however not present or effective in GWAT or
IWAT. Moreover, the glucose intolerance seen in young unchallenged RID tg mice cannot
be explained by hepatic steatosis, since comparably low levels of hepatic lipids are seen
between chow-fed wildtype and RID tg mice, at least while the body weights are <30 grams
(Fig. 6A). We noted, however, that SAA-levels are increased and both the liver size as well
as the average size of the individual hepatocytes, are enlarged in the chow-fed RID tg mice,
even in the absence of steatosis (Fig. S2H and Fig. 6B-C). Furthermore, the spleens of the
RID tg mice are also enlarged (Fig. 6D). Taken together, these observations argue for a state
of elevated hepatic stress and increased systemic immune activation in the RID tg mice. A
likely explanation for these phenomena is an increased exposure to bacterial toxins that are
leaking out from the gut. In line with this hypothesis, we found elevated levels of anti-LPS
IgGs and increased intestinal permeability as judged by the circulating levels of FITC-
dextran after an oral load in young RID tg mice (6-8 week old with a body weight of ≤ 20g)
on a chow diet (Fig. 6E-F). Treatment with 2% dextran sulfate sodium (DSS) through
drinking water, which damages the colonic epithelium, increases permeability and causes
colitis, leads to a further elevation of anti-LPS IgGs levels and serves as a positive control
for the plasma anti-LPS IgG assay (Fig. 6E). The colons of the DSS-treated RID tg mice
display increased pathological changes, such as an increased degree of colon hyperplasia,
more severe crypt disruption and leukocyte infiltration and further increased spleen size
compare to the DSS-treated wildtype mice (Fig. 6G-H). Furthermore, the DSS-treated RID
tg mice have increased expression of CD14, TLR4, TNFα, MCP-1, MIP-1α, F4/80 and
MPO mRNA in colon, which supports the histological findings and indicates that there is a
higher degree of inflammation, with enhanced recruitment/infiltration of macrophages and
neutrophils, compared to the DSS-treated wildtype mice (Fig. 7A). It should also be noted
that even the non-DSS treated RID tg mice display mild colon hyperplasia and elevated
colon expression of CD14 and SAA3 mRNA (Fig. 6G and 7A).
Despite the lack of an effect on body weight, DSS-treatment does however impair glucose
tolerance in regular wild type mice (Fig. S6A). This suggests that impaired intestinal barrier
function alone (even in the absence of HFD) can have a negative impact on systemic
metabolic regulation. This also suggests that at least some of the RID tg phenotype may be
improved if exposure to bacterial toxins is limited. We put this hypothesis to the test and
found that indeed, a 5-week antibiotics treatment normalizes spleen size and insulin levels in
the transgenic mice (Fig. 7B-C). To a large extent, the expression of the liver acute phase
reactants SAA-1 and -2 is also lowered to levels seen in the untreated wildtype mice (Fig.

7D). These data suggest that RID tg mice suffer from impaired intestinal barrier function,
leading to “leaky gut” and colitis. This also provides an explanation for their hepatomegaly,
spleen enlargement and glucose intolerance, even in the absence of an exogenous
inflammatory challenge.
In contrast to the RID tg mice, the dnTNF tg and the Ad-rtTA-TRE-IκB tg mice with their
associated normal MWAT expansion, do not display hepatomegaly, enlarged spleen or
colon hyperplasia (data not shown). This leads us to conclude that MWAT inflammation and
its subsequent expansion is an adaptive response that plays a significant role in sustaining
proper intestinal barrier function and healthy symbiosis between the commensal microflora
and the host.
The intestinal barrier is however not only preventing bacterial toxins from leaking out, but it
also contributes to a healthy symbiosis with colonizing bacteria. An impaired intestinal
barrier and/or chronic inflammation may lead to unfavorable shifts in microbial
composition. This is based on the observation that gut flora can be transferred from mouse
to mouse and affect the pathogenesis of obesity and hepatic steatosis (Garrett et al., 2010;
Henao-Mejia et al., 2012). As expected, we found a difference in bacterial composition as
judged by qPCR analysis of 8 different bacterial strains of DNA isolated from cecal content
from wildtype and RID tg mice housed in different cages. Cecal levels of Lactobacillus
murinus, Mucispirillum Schaedleri and Eubacterium Plexicaudatum were increased and
there was a trend for reduced levels of Firmicutes sp.. In contrast, the levels of Lactobacillus
sp., Clostridium sp. and B. Distasonis-Porphyromonas were similar between RID tg and
wild type mice (Fig. S6B). Differences in bacterial composition are not necessarily a
reflection of a direct negative effect of the RID tg transgene on intestinal barrier function,
but can also be secondary to metabolic disturbances and also differ between litters and
cages. Importantly, the metabolic dysfunction in RID tg is apparent regardless of whether
the mice under comparison are littermates or not. Furthermore, there is no difference in
neither HFD-induced body weight gain, glucose tolerance, insulin levels nor hepatic
steatosis in wildtype mice co-housed with either unrelated wildtype mice or with RID tg
(data not shown). Thus, the potential negative effect of an altered bacterial composition in
the RID tg mice is not potent enough to have an impact on metabolic health in the context of
a normal immune defense in wildtype mice. We can also conclude that the wildtype gut
flora does not offer a significant protective effect on the metabolic phenotype in the RID tg
mice.
Discussion
The data presented here argues that a reduced ability to sense and respond to pro-
inflammatory stimuli at the level of the adipocyte decreases the capacity for healthy adipose
tissue expansion and remodeling. This inability results in increased HFD-induced hepatic
steatosis and metabolic dysfunction. Interestingly, while the ability to sense pro-
inflammatory cues that trigger expansion is of importance for all fat pads, a deficiency in
this regard has particularly profound consequences for the functionality of the MWAT in the
visceral depot. We observe that the lack of MWAT expansion is associated with increased

intestinal permeability and colitis, resulting in chronic systemic inflammation and metabolic
dysfunction, even in the absence of HFD.
The underlying mechanism for inflammation-induced adipose tissue expansion

Adipose tissue expands either through hypertrophy of existing adipocytes or adipogenesis,
i.e. differentiation of new adipocytes from adipogenic precursor cells. All models studied
here establish a reduced number of adipocytes in most depots examined. Chronic HFD-
feeding leads to a more pronounced GWAT depot weight difference, while the IWAT depot
weight differences between tg and wildtype mice are reduced. Nevertheless, both GWAT
and IWAT depots are at all times populated by fewer adipocytes in the transgenic mice than
in the wildtype mice, reflecting a reduced rate of adipogenesis. Thus, the ability to mount an
acute inflammatory response is playing a more important role in facilitating adipogenesis
than in adipocyte hypertrophy.
The concept of inflammation-driven adipogenesis may seem contradictory at first, since pro-
inflammatory cytokines, such as TNFα, are lipolytic and block adipocyte differentiation in
vitro (Gustafson and Smith, 2006). The situation in vitro is however substantially different
from the situation in vivo, since there is neither a need for angiogenesis nor ECM
remodeling under in vitro conditions. In fact, several older studies support a role of local
inflammation in increased adipogenesis. For instance, Sadler and colleagues show that low-
dose LPS leads to adipocyte hyperplasia at the site of administration (Sadler et al., 2005),
and there is a selective increase in the amount of MWAT in Crohn’s disease as well as in
experimentally-induced colitis (Gambero et al., 2007; Sheehan et al., 1992). Moreover, a
very recent study on proliferation and differentiation of PDGFRα+ adipocyte progenitors in
vivo demonstrates that different adipogenic conditions are associated with an up-regulation
of different sets of inflammatory and macrophage-associated genes in white adipose tissue
(Lee et al., 2013).
It is likely that several distinct mechanisms contribute towards the inflammation-driven

adipogenic response, with several interconnected processes triggering adipose tissue growth.
Here, we found that HFD feeding acutely (within days) leads to a reduction of the total
levels of collagen in IWAT in wildtype mice, but not to the same extent in the dnTNF and
the RID tg mice. Thus, healthy adipose tissue expansion is associated with a net loss of
adipose tissue collagen while dysfunctional adipose tissue in more advanced obesity display
an increased number of ECM deposits, along with chronic inflammation and hypoxia. The
inability to effectively degrade ECM limits the capacity for healthy adipose tissue
expansion. RID tg adipose tissue also has a reduced capillary density. In line with this
finding, the capillary density of wildtype MWAT gets reduced by ablation of gut microbiota.
These data demonstrate that pro-inflammatory responses in adipose tissue are essential for
both proper ECM-remodeling and angiogenesis, two processes known to facilitate
adipogenesis in vivo and therefore are likely mediators of the inflammation-induced adipose
tissue expansion phenomenon (Cao, 2007; Cristancho and Lazar, 2011).
Chronic systemic inflammation interferes with optimal metabolic fitness. However, in light
of our findings in three independent adipocyte anti-inflammatory models (summarized in
Table S1), the view of adipose inflammation as a driving force for systemic inflammation

and metabolic dysfunction is an oversimplification. Rather, we postulate that a potent acute

inflammatory response is essential for adipose tissue protection, remodeling, and expansion
(Fig. 7E). This facilitates the return to a healthy equilibrium with metabolic homeostasis that
subsequently allows the inflammation to reach resolution as opposed to becoming chronic.
Experimental Procedures
Animals
Adipochaser mice were described in (Wang et al., 2013). aP2-dnTNF, aP2-RID and TRE-
IκBα(S32G-S36A) transgenic mice were used in a pure C57B6/J (dnTNF and TRE-IκB) or
on a FVB background (RID). Mice were maintained on a 12 hour dark/light cycle and
housed in groups of 4-5 with unlimited access to water, chow (No. 5058, Lab-Diet) or high
fat diet (HFD) (No. D12492, Research Diets Inc as indicated for the individual experiments.
In all experiments, littermate controls were used unless specifically stated otherwise. The
Institutional Animal Care and Use Committee of the University of Texas Southwestern
Medical Center, Dallas, has approved all animal experiments.
LPS-induced adipogenesis
Adipochaser mice were fed doxycycline-supplemented chow (600mg/kg) for one week and
were thereafter switched to regular chow three days prior the experiment. The right inguinal
adipose tissue (IWAT) depot of adipochaser mice were injected twice a week for two weeks
with 20μg lipopolysaccharide (LPS, Sigma, USA) in 50μL PBS, while the left IWAT depot
remained untreated. Five weeks later, tissues were harvested for X-Gal-LacZ staining
(performed as previously described (Wang et al., 2013)).
LPS and TNFα response in vivo

The mice received one intraperitoneal injection with 0.3 mg/kg LPS in PBS or 0.5 μg/mouse
TNFα (recombinant mouse TNFα, Biolegend, USA) in PBS supplemented with 1% BSA.
Blood was collected from the tail at indicated time points.
Oral glucose tolerance test

The mice were fasted for 3 hours during the light phase and blood samples were drawn from
the tail vein before and 15, 30, 60 and 120 minutes after an intragastric load with 2.5 g/kg
glucose in PBS.
Adipose tissue collagen levels measurements

Pieces of inguinal adipose tissue (~50mg) were snap-frozen in N2 (l) until analysis. The total
collagen content was measured, using a commercial kit (QuickZyme Biosciences,
Netherlands).
Hepatic Steatosis measurements

Quantification of hepatic steatosis was performed by computerized tomography (CT) or
standard biochemical tissue triglyceride analysis. The CT analysis was performed as
previously described (Asterholm and Scherer, 2010). In brief, mice were anesthetized with

Isoflurane and a CT-scan was performed at a resolution of 93 μm using the short scan mode
(180°) on an eXplore Lo cus in vivo MicroCT Scanner from GE Healthcare. Liver lipid
content was estimated by obtaining the average CT-value in multiple regions well within the
liver, as validated in (Asterholm and Scherer, 2010).
Intestinal permeability assay in vivo

The mice were fasted for 5h; thereafter they were given an oral load with 0.6mg/g FITC-
labeled dextran (average mole weight 4000, Sigma, USA). Blood samples were collected at
indicated time points and the serum was diluted 2 times in PBS and loaded together with
known standards diluted in 50/50 PBS/control serum on a 96-well plate and analyzed
(485nmex, 535nmem) on a POLARstar Optima Analyzer.
Statistical Analysis
Data are in generally expressed as mean +/− standard error of the mean (SEM). The
Student’s t-test and 1 or 2-way ANOVA (repeated measurement) were used for comparisons
between groups, log-transformation was performed as necessary to obtain normal
distribution. SPSS software (version 21) was used for these statistical calculations and a p-
value <0.05 was considered as significant and is indicated by a single asterisk; double
asterisk: p<0.01; triple asterisk: p<0.001.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Acknowledgments
We would like to thank Dr. David E. Szymkowski from Xencor Inc. for providing the cDNA constructs encoding
dnTNFα and for his guidance on the experimental design. We thank Dr. Bob Hammer and the Transgenic Core
Facility at UTSW for the generation of the transgenic lines, Dr. Lora Hooper for advice on the antibiotics
experiments, John Shelton and the Histology Core for assistance with histology and the UTSW Metabolic Core
Unit for help in phenotyping. Supported by the National Institutes of Health (grants R01-DK55758, R01-
DK099110 and P01DK088761-01 to PES). IWA is supported by the Throne-Holst Foundation, the Swedish
Research Council (2006-3931 and 2012-1601), VINNOVA (2011-01336) and NovoNordisk Excellence Project
Award. ZVW is supported by a postdoctoral fellowship from the American Heart Association (10POST4320009),
QAW is supported by a postdoctoral fellowship from the American Diabetes Association (7-11-MN-47) and TSM
is supported by NIH Training Grant T32-GM083831.
References
Asterholm IW, McDonald J, Blanchard PG, Sinha M, Xiao Q, Mistry J, Rutkowski JM, Deshaies Y,
Brekken RA, Scherer PE. Lack of “immunological fitness” during fasting in metabolically
challenged animals. J Lipid Res. 2012; 53:1254–1267. [PubMed: 22504909]
Asterholm IW, Scherer PE. Enhanced metabolic flexibility associated with elevated adiponectin levels.
The American journal of pathology. 2010; 176:1364–1376. [PubMed: 20093494]
Cao Y. Angiogenesis modulates adipogenesis and obesity. The Journal of clinical investigation. 2007;
117:2362–2368. [PubMed: 17786229]
Cristancho AG, Lazar MA. Forming functional fat: a growing understanding of adipocyte
differentiation. Nature reviews. Molecular cell biology. 2011; 12:722–734. [PubMed: 21952300]
Engelman JA, Berg AH, Lewis RY, Lisanti MP, Scherer PE. Tumor necrosis factor alpha-mediated
insulin resistance, but not dedifferentiation, is abrogated by MEK1/2 inhibitors in 3T3-L1
adipocytes. Mol Endocrinol. 2000; 14:1557–1569. [PubMed: 11043572]

Gambero A, Marostica M, Abdalla Saad MJ, Pedrazzoli J Jr. Mesenteric adipose tissue alterations
resulting from experimental reactivated colitis. Inflammatory bowel diseases. 2007; 13:1357–1364.
[PubMed: 17604368]
Garrett WS, Gordon JI, Glimcher LH. Homeostasis and inflammation in the intestine. Cell. 2010;
140:859–870. [PubMed: 20303876]
Granneman JG, Li P, Zhu Z, Lu Y. Metabolic and cellular plasticity in white adipose tissue I: effects
of beta3-adrenergic receptor activation. American journal of physiology. Endocrinology and
metabolism. 2005; 289:E608–616. [PubMed: 15941787]
Gustafson B, Smith U. Cytokines promote Wnt signaling and inflammation and impair the normal
differentiation and lipid accumulation in 3T3-L1 preadipocytes. The Journal of biological
chemistry. 2006; 281:9507–9516. [PubMed: 16464856]
Halberg N, Khan T, Trujillo ME, Wernstedt-Asterholm I, Attie AD, Sherwani S, Wang ZV,
Landskroner-Eiger S, Dineen S, Magalang UJ, et al. Hypoxia-inducible factor 1alpha induces
fibrosis and insulin resistance in white adipose tissue. Mol Cell Biol. 2009; 29:4467–4483.
[PubMed: 19546236]
Henao-Mejia J, Elinav E, Jin C, Hao L, Mehal WZ, Strowig T, Thaiss CA, Kau AL, Eisenbarth SC,
Jurczak MJ, et al. Inflammasome-mediated dysbiosis regulates progression of NAFLD and
obesity. Nature. 2012; 482:179–185. [PubMed: 22297845]
Heo SC, Jeon ES, Lee IH, Kim HS, Kim MB, Kim JH. Tumor necrosis factor-alpha-activated human
adipose tissue-derived mesenchymal stem cells accelerate cutaneous wound healing through
paracrine mechanisms. The Journal of investigative dermatology. 2011; 131:1559–1567.
[PubMed: 21451545]
Hotamisligil GS, Shargill NS, Spiegelman BM. Adipose expression of tumor necrosis factor-alpha:
direct role in obesity-linked insulin resistance. Science. 1993; 259:87–91. [PubMed: 7678183]
Khan T, Muise ES, Iyengar P, Wang ZV, Chandalia M, Abate N, Zhang BB, Bonaldo P, Chua S,
Scherer PE. Metabolic dysregulation and adipose tissue fibrosis: role of collagen VI. Molecular
and cellular biology. 2009; 29:1575–1591. [PubMed: 19114551]
Lee YH, Petkova AP, Granneman JG. Identification of an adipogenic niche for adipose tissue
remodeling and restoration. Cell metabolism. 2013; 18:355–367. [PubMed: 24011071]
Lee YH, Petkova AP, Mottillo EP, Granneman JG. In vivo identification of bipotential adipocyte
progenitors recruited by beta3-adrenoceptor activation and high-fat feeding. Cell metabolism.
2012; 15:480–491. [PubMed: 22482730]
Lichtenstein DL, Toth K, Doronin K, Tollefson AE, Wold WS. Functions and mechanisms of action of
the adenovirus E3 proteins. International reviews of immunology. 2004; 23:75–111. [PubMed:
14690856]
Magne J, Mariotti F, Fischer R, Mathe V, Tome D, Huneau JF. Early postprandial low-grade
inflammation after high-fat meal in healthy rats: possible involvement of visceral adipose tissue.
The Journal of nutritional biochemistry. 2010; 21:550–555. [PubMed: 19361974]
Mottillo EP, Shen XJ, Granneman JG. beta3-adrenergic receptor induction of adipocyte inflammation
requires lipolytic activation of stress kinases p38 and JNK. Biochimica et biophysica acta. 2010;
1801:1048–1055. [PubMed: 20435159]

Park J, Euhus DM, Scherer PE. Paracrine and endocrine effects of adipose tissue on cancer
development and progression. Endocrine reviews. 2011; 32:550–570. [PubMed: 21642230]
Rakoff-Nahoum S, Paglino J, Eslami-Varzaneh F, Edberg S, Medzhitov R. Recognition of commensal
microflora by toll-like receptors is required for intestinal homeostasis. Cell. 2004; 118:229–241.
[PubMed: 15260992]
Sadler D, Mattacks CA, Pond CM. Changes in adipocytes and dendritic cells in lymph node containing
adipose depots during and after many weeks of mild inflammation. Journal of anatomy. 2005;
207:769–781. [PubMed: 16367804]
Saika S, Ikeda K, Yamanaka O, Flanders KC, Okada Y, Miyamoto T, Kitano A, Ooshima A, Nakajima
Y, Ohnishi Y, et al. Loss of tumor necrosis factor alpha potentiates transforming growth factor
beta-mediated pathogenic tissue response during wound healing. The American journal of
pathology. 2006; 168:1848–1860. [PubMed: 16723700]

Sheehan AL, Warren BF, Gear MW, Shepherd NA. Fat-wrapping in Crohn’s disease: pathological
basis and relevance to surgical practice. The British journal of surgery. 1992; 79:955–958.
[PubMed: 1422768]
Steed PM, Tansey MG, Zalevsky J, Zhukovsky EA, Desjarlais JR, Szymkowski DE, Abbott C,
Carmichael D, Chan C, Cherry L, et al. Inactivation of TNF signaling by rationally designed
dominant-negative TNF variants. Science. 2003; 301:1895–1898. [PubMed: 14512626]
Sun K, Wernstedt Asterholm I, Kusminski CM, Bueno AC, Wang ZV, Pollard JW, Brekken RA,
Scherer PE. Dichotomous effects of VEGF-A on adipose tissue dysfunction. Proc Natl Acad Sci U
S A. 2012; 109:5874–5879. [PubMed: 22451920]
Wang QA, Tao C, Gupta RK, Scherer PE. Tracking adipogenesis during white adipose tissue
development, expansion and regeneration. Nature medicine. 2013
Wang ZV, Deng Y, Wang QA, Sun K, Scherer PE. Identification and characterization of a promoter
cassette conferring adipocyte-specific gene expression. Endocrinology. 2010; 151:2933–2939.
[PubMed: 20363877]

Highlights
• Adipocyte inflammation facilitates adipose tissue expansion and remodeling
• Suppressed adipocyte inflammation leads to adipose tissue dysfunction
• Suppressed adipocyte inflammation leads to systemic metabolic disturbances
• Mesenteric adipose tissue is important for proper intestinal barrier function


Figure 1. Reduced fat mass and glucose tolerance in dnTNF tg mice

(A) X-Gal-LacZ stained LPS-injected Adipochaser IWAT (blue=preexisting adipocytes,
white=new adipocytes) (B-C) Leptin and body weight change after i.p. injection with 0.3
mg/kg LPS in dnTNF tg and wildtype female mice. (D) IWAT and GWAT weight in
relation to body weight in chow (top) and HFD-fed (bottom) male dnTNF tg and wildtype
controls. (E-H) Body weight, glucose tolerance test, serum adiponectin and SAA-levels in
male dnTNF tg and littermate controls after 11 weeks HFD-feeding. (I) Representative
Trichrome stain of IWAT in male dnTNF tg and littermate controls after 11 weeks HFD-

feeding. (J) Representative perilipin (red) and mac2 (brown) immunostain in GWAT after
22 weeks HFD-feeding in male dnTNF tg and littermate control. Error bars represent SEM,
a p-value <0.05 according to student t-test was considered as significant and is indicated by
*/#; **/##: p<0.01; ***/###: p<0.001 (* = Difference in between genotype, # = Difference

from initial weight). P-values in red indicate difference between groups during the indicated
time course according to repeated measurement ANOVA.

Figure 2. Reduced fat mass and reduced glucose tolerance in RID tg mice
(A) LPS (0.3 mg/kg)-induced body weight change in male RID tg and wildtype controls. (B)
Gene expression in GWAT harvested from male RID tg and wildtype controls 6h after LPS
injection. (C) IWAT and GWAT in relation to body weight in chow-fed male RID tg and
wildtype mice. (D) IWAT, GWAT and MWAT in relation to body weight in 15 week HFD-
fed male RID tg and wildtype mice. (E) Representative H&E stain of IWAT and GWAT in
male RID tg and wildtype mice on chow (F) Adiponectin levels in male RID tg and wildtype
mice on chow and after 12 weeks HFD. Glucose tolerance in (G) chow-fed and (H) 12-week

HFD-fed male RID tg and wildtype mice. (I) Serum-insulin levels in 3 h fasted male RID tg
and wildtype mice. 1-way ANOVA analysis shows that both diet (F=8.1, p=0.013) and
genotype (F=12.3, p=0.004) contribute significantly to collagen levels. Error bars represent
SEM, a p-value <0.05 according to student t-test was considered as significant and is
indicated by */#; **/##: p<0.01; ***/###: p<0.001 (* = Difference in between genotype, # =
Difference from initial weight). P-values in red indicate difference between groups during
time courses according to repeated measurement ANOVA.

Figure 3. Increased HFD-induced steatosis and delayed adipose tissue development in dnTNF
and RID tg mice
(A) Representative Picrosirius red stain of IWAT from 8-week old C57B6 males on chow
and after 9 day with HFD. (B) Representative Trichrome stain and collagen levels in IWAT
of HFD-fed male RID tg and wildtype mice. (C) Representative H&E stain of liver sections
from 11 weeks HFD-fed male dnTNF tg and littermate control. (D) Liver fat quantified by
CT and liver weight in 11 weeks HFD-fed male RID tg and wildtype controls. (E)
Representative H&E stain of IWAT and GWAT sections and endomucin immune-stain of
IWAT (bottom panels) and (with quantification shown in (F)) from 10-day old male RID tg

and wildtype pups. (G) Body weight, IWAT weight and adipocyte sizes in 10-day old
dnTNF tg and littermate controls. Error bars represent SEM, a p-value <0.05 according to
student t-test was considered as significant and is indicated by */#; **: p<0.01; ***: p<0.001
(*significantly different from WT, # significantly different from untreated controls of same
genotype).

Figure 4. Ablation of gut microbiota affects capillary density and Ad-rtTA-TRE-IκB tg mice
display reduced adipogenesis and glucose tolerance
(A) Effect of antibiotics-treatment on IWAT weight in 10-day old RID tg and wildtype pups.
1-way ANOVA analysis shows that both treatment (F=5.3, p=0.026) and genotype (F=29.4,
p<0.001) contribute significantly to the IWAT weight. (B) Effect of antibiotics-treatment on
capillary density in MWAT as judged by endomucin immune-stain in male RID tg and
wildtype mice. 1-way ANOVA analysis shows that both treatment (F=35.1, p=0.002) and
genotype (F=14.8, p=0.012) contribute significantly to capillary density in MWAT. (C)
Body weight and IWAT weight in doxycycline-treated 10-day old Ad-rtTA-TRE-IκB tg and

littermate controls. (D) Body weight and glucose tolerance in 8-weeks HFD-fed male Ad-
rtTA-TRE-IκB tg and littermate controls. (E) Dissected tissue weight in 8-weeks HFD-fed
Ad-rtTA-TRE-IκB tg and littermate controls. Error bars represent SEM, a p-value <0.05
according to student t-test was considered as significant and is indicated by */#; **/##:
p<0.01; ***: p<0.001 (*significantly different from WT, # significantly different from
untreated controls of same genotype). P-value in red indicates difference between groups
during the indicated time course according to repeated measurement ANOVA).

Figure 5. Reduced β3AR-agonist-induced browning of white adipose tissue in RID tg mice

(A) Photo of IWAT and GWAT harvested from male RID tg and wildtype mice after
chronic β3AR-agonist treatment (10 days with daily i.p injection with 1mg/kg in PBS). (B)
Representative H&E stain of IWAT from male RID tg and wildtype mice after chronic
β3AR-agonist treatment. (C) Representative BrdU immunostain of GWAT harvested after
chronic β3AR-agonist treatment (co-administered with 10 mg/kg BrdU) in male RID tg and
wildtype mice. Orange arrows points towards BrdU positive nuclei (D) Gene expression
analyses of IWAT 3 h (acute) after β3AR-agonist injection and after chronic treatment

(IWAT harvested 24 h after last injection) in male RID tg and wildtype mice. Error bars
represent SEM, a p-value <0.05 according to student t-test was considered as significant and
is indicated by */#; **/##: p<0.01 (*significantly different from WT, # significantly
different from untreated controls of same genotype). Additional 1-way ANOVA analyses
have been performed for Figure 4D (Table S3)

Figure 6. Leaky gut and colitis associated with signs of systemic inflammation in RID tg mice
(A) Liver fat quantified by CT in chow-fed male RID tg and wildtype mice. (B) Liver
weight and (C) representative H&E stain of liver sections to show hepatocyte size in chow-
fed male RID tg and wildtype mice with a body weight <30g. (D) Spleen size in chow-fed
RIDs chow-fed male RID tg and wildtype mice. (E) Serum levels of anti-LPS IgG in young
chow-fed vs. DSS-treated male RID tg and wildtype mice. 1-way ANOVA analysis shows
that both treatment (F=78.8, p<0.001) and genotype (F=26.4, p<0.001) contribute
significantly to anti-LPS IgG levels (F) Serum levels of FITC-dextran after an oral load in

young chow-fed female RID tg and wildtype mice. (G) Colon weight/length ratios,
representative H&E images of colon and (H) spleen weight in untreated, and in response to
DSS treatment, in male RID tg and wildtype mice. 1-way ANOVA analysis shows that both
treatment (F=153.6/47.3, p<0.001/<0.001) and genotype (F=27.4/13.9, p<0.001/0.05)

contribute significantly to both colon thickness and spleen size. Error bars represent SEM, a
p-value <0.05 according to student t-test was considered as significant and is indicated by
*/#; **/##: p<0.01; ***/###: p<0.001 (*significantly different from WT, # significantly
different from untreated controls of same genotype). P-value in red indicates differences
between groups during time course according to repeated measurement ANOVA.

Figure 7. Antibiotics-treatment improves the RID tg mouse phenotype

A) Gene expression analysis of proximal colon in untreated, and in response to 3 DSS
treatments, in male RID tg and wildtype mice. (B) Spleen, (C) fasting serum-insulin and (D)
SAA1 and 2 mRNA levels in liver in control or antibiotic-treated male RID tg and wildtype
mice. 1-way ANOVA analysis shows that both treatment (F=15.8/8.7/12.1/5.8,
p=0.001/0.003/0.004/0.03) and genotype (F=19.3/17.1/23.6/31.8, p<0.001/0.001/<0.001/
<0.001) contribute significantly to spleen size, insulin levels, liver SAA1 and liver SAA2
levels. (E) Summary and proposed model: Acute inflammation is essential for healthy

adipose tissue expansion and proper remodeling. Inability of adipose tissue to accurately
sense and respond to inflammatory stimuli leads to reduced adipose tissue expansion and an
increased risk for microbial translocation. Error bars represent SEM, a p-value <0.05
according to student t-test was considered as significant and is indicated by */#; **/##:
p<0.01; ***: p<0.001 (*significantly different from WT, # significantly different from
untreated controls of same genotype).

Nihms 599061

Uploaded by

Copyright:

Available Formats

Nihms 599061

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nihms 599061

Uploaded by

Copyright:

Available Formats

NIH Public Access

Cell Metab. 2014 July 1; 20(1): 103–118. doi:10.1016/j.cmet.2014.05.005.

Adipocyte Inflammation is Essential for Healthy Adipose Tissue

© 2014 Elsevier Inc. All rights reserved.

inflammatory factors in adipose tissue constitutively, while one model is inducible.

permanently activate β-galactosidase-expression in all preexisting adipocytes by a short bout

Cell Metab. Author manuscript; available in PMC 2015 July 01.

angiogenesis. To study the role of inflammation in adipose tissue, we developed a mouse

As expected, unchallenged chow-fed dnTNF tg do not display significant differences with

phosphorylation) in response to an intraperitoneal injection with TNFα (Fig. S1C).

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Essential role for acute inflammation for adipose tissue functionality

adipose tissue inflammation”, we wanted to test an additional model to strengthen the

Cell Metab. Author manuscript; available in PMC 2015 July 01.

suppression of inflammation in macrophages (Fig. S2B). Consistent with a potent action

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Adipocyte inflammation is important for early postnatal adipogenesis

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Reduced adipogenesis and HFD-induced glucose intolerance in an inducible adipocyte-

We investigated the metabolic phenotype of Ad-rtTA-TRE-IκB tg mice. We found no

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Impaired β3-Adrenergic Receptor (AR) agonist-induced adipose tissue remodeling in RID

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

The underlying mechanism for inflammation-induced adipose tissue expansion

It is likely that several distinct mechanisms contribute towards the inflammation-driven

Cell Metab. Author manuscript; available in PMC 2015 July 01.

and metabolic dysfunction is an oversimplification. Rather, we postulate that a potent acute

subsequently allows the inflammation to reach resolution as opposed to becoming chronic.

LPS and TNFα response in vivo

Oral glucose tolerance test

Adipose tissue collagen levels measurements

Hepatic Steatosis measurements

Cell Metab. Author manuscript; available in PMC 2015 July 01.

liver, as validated in (Asterholm and Scherer, 2010).

Intestinal permeability assay in vivo

asterisk: p<0.01; triple asterisk: p<0.001.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

1801:1048–1055. [PubMed: 20435159]

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

• Suppressed adipocyte inflammation leads to adipose tissue dysfunction

• Suppressed adipocyte inflammation leads to systemic metabolic disturbances

• Mesenteric adipose tissue is important for proper intestinal barrier function

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Figure 1. Reduced fat mass and glucose tolerance in dnTNF tg mice

Cell Metab. Author manuscript; available in PMC 2015 July 01.

*/#; **/##: p<0.01; ***/###: p<0.001 (* = Difference in between genotype, # = Difference

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Cell Metab. Author manuscript; available in PMC 2015 July 01.

Figure 5. Reduced β3AR-agonist-induced browning of white adipose tissue in RID tg mice

Cell Metab. Author manuscript; available in PMC 2015 July 01.

/#; /##: p<0.01; /###: p<0.001 (* = Difference in between genotype, # = Difference