Muramyl Dipeptide-Based Postbiotics Mitigate Obesity-Induced Insulin Resistance via IRF4

Muramyl Dipeptide-Based Postbiotics Mitigate Obesity-Induced Insulin Resistance via IRF4Graphical Abstract
resistance via NOD2
intolerance
NOD1, muropeptides
d The orphan drug mifamurtide is an insulin sensitizer in mice
Cavallari et al., 2017, Cell Metabolism 25, 1063–1074 May 2, 2017 ª 2017 Elsevier Inc. http://dx.doi.org/10.1016/j.cmet.2017.03.021
Authors
how a bacterial cell wall muropeptide acts
via NOD2 as a ‘‘postbiotic’’ improving
insulin resistance and metabolic tissue
inflammation in obese mice,
in microbiota composition.
Muramyl Dipeptide-Based Postbiotics Mitigate Obesity-Induced Insulin Resistance via IRF4 Joseph F. Cavallari,1 Morgan D. Fullerton,1 Brittany M. Duggan,1 Kevin P. Foley,1 Emmanuel Denou,1 Brennan K. Smith,3
Eric M. Desjardins,3 Brandyn D. Henriksbo,1 Kalvin J. Kim,1 Brian R. Tuinema,1 Jennifer C. Stearns,2,3 David Prescott,4
Philip Rosenstiel,5 Brian K. Coombes,1,2 Gregory R. Steinberg,1,3 and Jonathan D. Schertzer1,2,6,* 1Department of Biochemistry and Biomedical Sciences 2Farncombe Family Digestive Health Research Institute 3Department of Medicine
McMaster University, Hamilton, ON L8N 3Z5, Canada 4Department of Immunology, University of Toronto, Toronto, ON M5S 1A8, Canada 5Institute of Clinical Molecular Biology (IKMB), University of Kiel, Schittenhelmstrasse 12, 24105 Kiel, Germany 6Lead Contact
*Correspondence: [email protected]
Intestinal dysbiosis contributes to obesity and insulin resistance, but intervening with antibiotics, prebi- otics, or probiotics can be limited by specificity or sustained changes in microbial composition. Postbi- otics include bacterial components such as lipopoly- saccharides, which have been shown to promote insulin resistance during metabolic endotoxemia. We found that bacterial cell wall-derived muramyl dipeptide (MDP) is an insulin-sensitizing postbiotic that requires NOD2. Injecting MDP lowered adipose inflammation and reduced glucose intolerance in obese mice without causing weight loss or altering the composition of the microbiome. MDP reduced hepatic insulin resistance during obesity and low- level endotoxemia. NOD1-activating muropeptides worsened glucose tolerance. IRF4 distinguished opposing glycemic responses to different types of peptidoglycan and was required for MDP/NOD2- induced insulin sensitization and lower metabolic tissue inflammation during obesity and endotoxemia. IRF4 was dispensable for exacerbated glucose intol- erance via NOD1. Mifamurtide, an MDP-based drug with orphan drug status, was an insulin sensitizer at clinically relevant doses in obese mice.
INTRODUCTION
bolic responses, and dysbiosis can contribute to obesity and in-
sulin resistance. Adiposity and aspects of metabolic dysfunc-
tion are transmissible via the microbiota (Turnbaugh et al.,
2009). However, it is not yet clear how to target dysbiosis linked
to obesity because the components of the microbiota that
directly alter metabolic responses are ill defined. Several prebi-
otic and probiotic strategies are aimed at reducing metabolic
dysfunction in obese mice (Everard et al., 2013) and humans
Cell M
(Simon et al., 2015). However, it is unclear whether probiotic
strategies can contribute sustained changes to the composition
of the microbiota (Kristensen et al., 2016). Further, it was
recently shown that 1 week of broad-spectrum antibiotics did
not alter insulin sensitivity or markers of inflammation in obese
humans (Reijnders et al., 2016). An alternate approach is to
test the secreted factors, cellular components, and metabolites
of bacteria that, upon delivery to a host, can have biological ef-
fects on it. We and others define these molecules as ‘‘postbiot-
ics’’ (Cicenia et al., 2014; Patel and Denning, 2013). Examples of
postbiotics such as microbial-derived short-chain fatty acids or
flavonoids can directly influence host feeding behavior, energy
metabolism, insulin secretion, and insulin sensitivity (Canfora
et al., 2015; Kimura et al., 2013; Perry et al., 2016; Thaiss
et al., 2016). Another source of postbiotics that can propagate
metabolic effects is microbial components from live or dead
bacteria. For example, nutritional status and obesity are associ-
ated with metabolic endotoxemia. Specifically, a prolonged diet
containing excessive lipids can produce a chronic, low-level in-
crease in circulating bacterial lipopolysaccharide (LPS), which
can promote Toll-like receptor (TLR)4/Cd14-mediated inflam-
mation and insulin resistance (Cani et al., 2007; Fernandez-
Real et al., 2011).
can also protect against insulin resistance. For example, immu-
nization with proximal gut microbiota-derived extracts promotes
immunological tolerance and decreases high-fat diet (HFD)-
induced insulin resistance in mice (Pomie et al., 2016). However,
the identity of themicrobial factor responsible for tolerization and
reduced insulin resistance was not known. In fact, very little is
known about specific bacterial components that can elicit insu-
lin-sensitizing or anti-inflammatory effects inmetabolic tissues. It
is a daunting task to understand how a change in the composi-
tion of trillions of resident bacteria relates to insulin resistance,
but the major components of bacteria are positioned to alter
metabolism. Host detection of components from the bacterial
cell wall is sufficient to promote innate and adaptive immune
responses and promote metabolic inflammation, lipolysis, and
insulin resistance (Chi et al., 2014; Fritz et al., 2007; Schertzer
et al., 2011; Wolf et al., 2016).
etabolism 25, 1063–1074, May 2, 2017 ª 2017 Elsevier Inc. 1063
Figure 1. MDP Lowers Diet-Induced Insulin Resistance
(A and B) Experimental design for MDP administration at the inception of HFD (A) and after long-term HFD feeding (B).
(C) Glucose tolerance test (2 g/kg) of standard chow-fed WT mice treated with MDP for 5 weeks. n = 9–10 mice for each group.
(D and E) Glucose (2 g/kg; D) and insulin (1 IU/kg; E) tolerance tests of 60% HFD-fed WTmale mice treated with MDP for 5 weeks. n = 8–10 mice for each group.
(F) Glucose tolerance test (1.5 g/kg) of 60% HFD-fed Nod2/ mice treated with MDP for 5 weeks. n = 7–9 mice for each group.
(G and H) Glucose tolerance test (1 g/kg) ofWTmice with obesity that was established using 45%HFD for 10weeks; thenmice were treated for 3 days with saline,
100 mg MDP (G), or 100 mg iE-DAP (g-D-Glu-meso-diaminopimelic acid; H). n = 8–13 mice for each group.
(I) Peripheral tissue glucose disposal rate and hepatic glucose production during hyperinsulinemic-euglycemic clamps in WT 60% HFD-fed mice treated with
MDP for 5 weeks. n = 5 mice for each group.
(legend continued on next page)
1064 Cell Metabolism 25, 1063–1074, May 2, 2017
Peptidoglycan-based bacterial cell walls consist of sugars and
amino acids, where specific muropeptide sequences are de-
tected by nucleotide-binding oligomerization domain-containing
(NOD) proteins. NOD1 recognizes meso-diaminopimelic (meso-
DAP) acid-containing muropeptides (Girardin et al., 2003a),
whereas NOD2 detects muramyl dipeptide (MDP)-containing
peptidoglycan (Girardin et al., 2003b). Acute activation of
NOD1 causes whole-body and hepatic insulin resistance
(Schertzer et al., 2011), and deletion of NOD1 can protect against
diet-induced insulin resistance in mice (Amar et al., 2011). This is
consistent with the fact that microbiota-derived muropeptides
acting on NOD1, but not NOD2, augment systemic immunity
(Clarke et al., 2010; Hergott et al., 2016). In contrast, NOD2 atten-
uates inflammation induced by other bacterial products and pro-
tects against inflammatory colitis (Hedl et al., 2007; Watanabe
et al., 2008) and insulin resistance (Denou et al., 2015). The
mechanisms by which both NOD1 and NOD2 elicit their differen-
tial effects on inflammation are not fully understood, since both
activate NF-kB via receptor-interacting serine/threonine-protein
kinase 2 (RIPK2) (Kobayashi et al., 2002). However, the protec-
tive effect of MDP on inflammatory colitis has been reported
to be dependent on NOD2 and interferon regulatory factor 4
(IRF4) (Watanabe et al., 2008, 2014). IRF4 deletion promotes
metabolic inflammation during obesity (Eguchi et al., 2011,
2013). It is not known how NOD1 versus NOD2 detection of bac-
terial muropeptides alters glucose metabolism. In the current
study we found that IRF4 dictates divergent blood glucose ef-
fects of specific bacterial cell wall-derived muropeptides during
obesity and endotoxemia. We present anMDP-based postbiotic
strategy that limits metabolic inflammation and reduces insulin
resistance despite no discernable changes in the composition
of the microbiome or adiposity in mice.
RESULTS AND DISCUSSION
MDP Lowers Diet-Induced Insulin Resistance Three intraperitoneal injections of MDP can attenuate experi-
mental colitis in mice (Watanabe et al., 2008), but the effects of
MDP on obesity and insulin resistance were not known. We first
determined the location of intraperitoneally injected rhodamine-
labeled MDP in anesthetized wild-type (WT) mice prior to excre-
tionof themolecule to undetectable levelswithin2hr after injec-
tion (Figures S1A and S1B). In order to thoroughly examine the
relationship between NOD2 activation and metabolic changes
during diet-induced obesity, we used two different HFD-based
experimental models (Figures 1A and 1B). Model #1 is a chronic
NOD2 activation regimen by injecting MDP (4 days per week for
5 weeks), which is co-initiated with HFD feeding. HFD model #2
is an intervention-style protocol in which mice are first fed an
HFD for 10 weeks and then obese mice are injected daily with
MDP for 3 days. In both protocols, the final MDP injection was
administered24hrbeforemetabolic and inflammatoryphenotyp-
ing. In chow-fed control animals, 5 weeks of MDP injections did
(J) Glucose tolerance test (1 g/kg) of mice lacking hepatocyte NOD2 (Nod2HKO) w
3 days with MDP. n = 7 mice for each group.
*Significantly different from control group determined by t test or one-way ANOVA
tests, asterisk (*) indicates HFD-fed saline-injectedmice versus HFD-fed andMDP
are significantly different from HFD-fed saline and MDP groups, respectively. Va
not alter glucose metabolism (Figure 1C). In HFD model #1,
MDP injections improved insulin and glucose tolerance after
4–5 weeks on HFD (Figures 1D and 1E), but did not alter body
mass, adiposity, or serum endotoxin levels (Figures S1C–S1E).
MDP injections did not alter glucose tolerance in HFD-fed
Nod2/mice, indicating the receptor specificity of the glycemic
response (Figure 1F). In HFD model #2, after 10 weeks of HFD
feeding, obesemicewere found to bemore glucose tolerant after
only three (daily) intraperitoneal injections (Figure 1G) in the
absence of changes in body mass (data not shown). Conversely,
three injections of theminimal peptidoglycanmotif recognized by
NOD1 (iE-DAP) had no effect on glucose tolerance in obesemice
fed an HFD for 10 weeks (Figure 1H).
In HFD model #1, hyperinsulinemic-euglycemic clamps in
HFD-fed mice showed that MDP injections increased insulin
suppression of hepatic glucose production, but MDP did not
change glucose disposal rate (GDR) and did not change insu-
lin-stimulated glucose uptake in white adipose tissue (WAT) (Fig-
ures 1I and S1F). This finding is important because we previously
published that a high dose of MDP in vitro caused muscle cell-
autonomous responses that impair insulin-stimulated glucose
uptake in clonal rat myotubes (Tamrakar et al., 2010). We also
previously showed increased markers of muscle inflammation
and lower peripheral glucose disposal 6 hr after a single injec-
tion of an MDP derivative (muradimetide, a dimethyl ester of
MurNAc-L-Ala-D-isoGln) in chow-fed mice. These effects did
not persist for 24 hr andwere small compared to the insulin resis-
tance caused by NOD1 ligands (Schertzer et al., 2011). Here, our
results show thatMDPdoes not alter peripheral glucose disposal
in mice.
In order to rule out the possibility of MDP acting directly on
hepatocyte glucose production through a cell-autonomous
NOD2 pathway, we generated mice lacking NOD2 in hepato-
cytes (Nod2HKO; Figure S2A). Compared to littermate controls,
chow-fed and HFD-fed Nod2HKOmice had no change in glucose
tolerance or body mass (Figures S2B and S2C). Using model #2
(Figure 1B), our results show that three MDP injections improved
glucose tolerance in Nod2HKO mice fed an HFD for 10 weeks
(Figure 1J), an identical result to that observed in WT HFD-fed
mice (Figure 1G). Furthermore, chronic MDP injections did not
alter fasted or clamped serum levels of non-esterified fatty acids
(NEFAs), and MDP did not alter liver or muscle triacylglycerol
(TAG) levels in WT mice that were HFD fed according to model
#1 (Figures S2D–S2F). Overall, these results show that MDP is
a hepatic insulin sensitizer during obesity, without acting directly
on NOD2 in hepatocytes. Moreover, only NOD2-activating bac-
terial cell wall muropeptides improved glucose control inmultiple
models of diet-induced obesity.
MDP Improves Glucose Tolerance during Obesity Independently of Diet or Microbiome Composition To determine whether the insulin-sensitizing effects of NOD2
activation were dependent on dietary fat, we compared the
ith obesity that was established using 45% HFD for 10 weeks, then treated for
, where appropriate. For glucose versus time graphs during glucose tolerance
-injectedmice; ampersand (&) and pound sign (#) indicate that chow-fed values
lues shown are mean ± SEM. See also Figures S1 and S2.
Cell Metabolism 25, 1063–1074, May 2, 2017 1065
effects of MDP in HFD-fed mice and hyperphagic, leptin-defi-
cient (ob/ob) chow-fed mice. We found that only 2 weeks of
MDP injections reduced glucose intolerance in obese ob/ob
mice and HFD-fed mice (Figures 2A and 2B). Similar to results
in HFD-fed mice, MDP did not cause weight loss in ob/ob mice
(Figures S3A and S3B). We next determined if the insulin-sensi-
tizing effect of MDP injection was associated with altered
composition of the intestinal microbiota. Chronic MDP treatment
did not significantly alter the taxonomic characteristics of the
fecal microbiome in obese mice fed a 60% HFD, since there
was not a significant difference in beta diversity, assessed
through principal coordinate analysis (PCoA) using the Bray
Curtis dissimilarity index, or a shift in bacterial taxonomy at
the phylum, family, or OTU levels (Figures 2C–2E and S3C).
Together, these results suggest that the effects of NOD2 activa-
tion on glucose metabolism are independent of dietary fat con-
tent and changes in the composition of the microbiome.
IRF4 Mediates the Anti-inflammatory and Insulin-Sensitizing Effects of MDP It is not known how specific bacterial cell wall components alter
blood glucose homeostasis, and our data suggest that the ef-
fects of NOD2 activation are independent of dietary fat content
and independent of changes in the composition of the micro-
biota. Thus, we next assessed if activating NOD2 regulates in-
flammatory mechanisms within host metabolic tissues. Seminal
work showed that IRF4 can alter adipose tissue inflammation
and lipolysis during obesity (Eguchi et al., 2011, 2013), and it is
known that IRF4 mediates the anti-inflammatory effects of
MDP in models of colitis (Watanabe et al., 2008). We hypothe-
sized that MDP would lower diet-induced inflammation via the
transcription factor IRF4.
Using HFDmodel #1, we found that 5 weeks of MDP injections
increased Irf4 expression in both the liver and WAT in an NOD2-
dependent manner (Figures 3A and 3B). We next injected Irf4/
mice with MDP according to HFD model #1 and found no differ-
ence in glucose tolerance (Figure 3C). MDP has been shown to
decrease markers of inflammation via IRF4 in immune cells
(Hedl et al., 2007; Watanabe et al., 2008). The effects of MDP
and IRF4 in metabolic tissues during obesity had not been
tested. We found that 5 weeks of MDP injections decreased
NF-kB activity by 35% in WAT from HFD-fed WT mice, but
MDP did not alter NF-kB activity in WAT from Irf4/ HFD-fed
mice (Figure 3D). MDP did not alter expression of selected meta-
bolic transcripts in the WAT or liver (Figures S4A and S4B). MDP
injections caused very few changes in hepatic pro-inflammatory
transcripts such as decreased levels of Nos2 and Cxcl10 in both
WT and Irf4/ HFD-fed mice (Figure S4C). Conversely, MDP in-
jections increased transcript levels of immune cell markers and
pro-inflammatory mediators in muscle tissue in both WT and
Irf4/ animals (Figures S4Dand S4E), an effect that is consistent
with our previous findings showing MDP-induced muscle cell-
autonomous inflammatory responses (Tamrakar et al., 2010).
However, MDP injections caused a widespread reduction in
markers of inflammation in adipose tissue, demonstrated by
lower transcript levels of multiple cytokines, chemokines, im-
mune cell markers, and pattern recognition receptors in the
WAT of HFD-fedWTmice (Figures 3E and 3F). Almost all of these
MDP-induced transcriptional changes in adipose depended on
1066 Cell Metabolism 25, 1063–1074, May 2, 2017
IRF4, since only 1 of 15 factors (i.e., Ccl2) was lower after MDP
injections in HFD-fed Irf4/ mice. We found that MDP injection
decreased a small number of inflammatory markers in the spleen
such asCxcl1 and Il1b, but MDP did not change circulating cyto-
kine levels in HFD-fed animals (Figures S4F and S4G).
Overall, these results show that IRF4 mediates anti-inflamma-
tory actions in the adipose tissue and whole-body insulin-sensi-
tizing effects of repeated MDP injections during diet-induced
obesity. It is known that deletion of IRF4 promotes a pro-inflam-
matory response in adipose tissue (Eguchi et al., 2013). Our re-
sults add NOD2 activation with MDP as an anti-inflammatory
trigger specifically in adipose tissue via IRF4. This effect of
MDP appears separate from IRF4-mediated control of thermo-
genesis or IRF4-mediated lipid handling/deposition responses
that can occur due to other stimuli such as cold exposure, b3
adrenergic stimulation, and fasting (Eguchi et al., 2011; Kong
et al., 2014). In fact, we showed that fasting increased transcript
levels of IRF4 in the adipose tissue of bothWT andNod2/mice
(Figures S4H and S4I).
bacterial component LPS (i.e., endotoxin), and thismetabolic en-
dotoxemia is sufficient to promote insulin resistance (Cani et al.,
2007). MDP can promote tolerance to TLR-mediated inflamma-
tion (Honma et al., 2005; Negishi et al., 2005).We found thatMDP
administration did not improve glucose tolerance in HFD-fed
C3H/HeJ (TLR4 mutant) mice, but actually exacerbated glucose
intolerance and caused higher fasting blood glucose in HFD-fed
C3H/HeJ mice (Figure 4A). Although these data suggest that the
suppression of TLR4-mediated inflammation during obesity is at
least one mechanism of the anti-inflammatory and insulin-sensi-
tizing effects of MDP, these conclusions would require direct
testing. The key outcome of this experiment was that it prompted
us to test the interaction between muropeptides and LPS on gly-
cemia in an acute model of low-level (non-lethal) endotoxemia
(Figure 4B). The causes of dysglycemia are multifactorial. It is
important to know if MDP canmitigate dysglycemia from causes
other than obesity. Further, our goal was to understand if an
MDP-based postbiotic could counteract the metabolic effects
of another microbiota-derived factor (i.e., LPS) that is known to
promote insulin resistance.
We used a reductionist model to test if MDP could attenuate
dysglycemia caused by LPS in mice. We found that MDP injec-
tion prior to LPS was sufficient to increase glucose tolerance 6 hr
after the endotoxin challenge (Figure 4C), despite MDP-injected
mice secreting significantly less insulin in response to glucose
(Figure S5A). We also confirmed that female WT mice are also
more glucose tolerant when givenMDPprior to LPS (Figure S5B),
demonstrating that the effects of NOD2 activation on meta-
bolism are not restricted to male mice. Three injections of MDP
alone (i.e., in the absence of LPS) had no effect on glucose toler-
ance in chow-fed WT mice (Figure S5C). Additionally, MDP pre-
treatment did not alter glucose tolerance in LPS-challenged
Nod2/ mice (Figure 4D).
We next tested the role of NOD-like receptor family, pyrin
domain-containing 3 (NLRP3) inMDP-induced changes in glyce-
mia because NLRP3 can moonlight as a transcriptional regulator
Figure 2. MDP Improves Glucose Tolerance during Obesity Independently of Diet or Microbiome Composition (A) Glucose (0.75 g/kg) tolerance test of chow-fed 8-week-old ob/ob mice treated with MDP (4 times per week) for 2 weeks. n = 11 mice for each group.
(B) Glucose (2 g/kg) tolerance test of 60% HFD-fed WT male mice treated with MDP (4 times per week) for 2 weeks. n = 5–7 mice for each group.
(C) Principle coordinates analysis (PCoA) performed on Bray-Curtis distances in fecal samples of 60%HFD-fed mice treated with MDP for 3 weeks. Black circles
represent samples from saline-treated mice, and red triangles represent samples from MDP-treated mice.
(D) Relative abundance of bacteria resolved to the phylum level in fecal samples of 60% HFD-fed mice treated with MDP for 3 weeks. Each bar represents an
individual mouse.
(E) Relative abundance of bacteria resolved to the family level in fecal samples of 60% HFD-fed mice treated with MDP for 3 weeks. Each bar represents an
individual mouse.
*Significantly different from control group determined by t test or one-way ANOVA, where appropriate. Values shown are mean ± SEM. See also Figure S3.
Cell Metabolism 25, 1063–1074, May 2, 2017 1067
Figure 3. IRF4 Mediates the Anti-inflammatory
and Insulin-Sensitizing Effects of MDP
(A and B) Irf4 expression in liver (A) and WAT (B) of…