©20

12 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

l e t t e r s

nature medicine  advance online publication �

Chronic low-grade adipose tissue and liver inflammation is a major cause of systemic insulin resistance and is a key component of the low degree of insulin sensitivity that exists in obesity and type 2 diabetes1,2. Immune cells, such as macrophages, T cells, B cells, mast cells and eosinophils, have all been implicated as having a role in this process3–8. Neutrophils are typically the first immune cells to respond to inflammation and can exacerbate the chronic inflammatory state by helping to recruit macrophages and by interacting with antigen-presenting cells9–11. Neutrophils secrete several proteases, one of which is neutrophil elastase, which can promote inflammatory responses in several disease models12. Here we show that treatment of hepatocytes with neutrophil elastase causes cellular insulin resistance and that deletion of neutrophil elastase in high-fat-diet–induced obese (DIO) mice leads to less tissue inflammation that is associated with lower adipose tissue neutrophil and macrophage content. These changes are accompanied by improved glucose tolerance and increased insulin sensitivity. Taken together, we show that neutrophils can be added to the extensive repertoire of immune cells that participate in inflammation-induced metabolic disease.

Compared to adipose tissue from nonobese mice fed a standard chow diet, adipose tissue from DIO mice has a markedly higher number of proinflammatory, M1-like macrophages that secrete cytokines such as tumor necrosis factor α (Tnf-α), interleukin-1β (Il-1β) and Il-6, which can directly lead to a lowered insulin sensitivity1. Although the role of tissue macrophages in this process has been well docu-mented, recent studies have also indicated contributions from several other immune cell types, including T cells3–5, B cells6, mast cells7 and eosinophils8.

Granulocytes comprise 60–70% of blood leukocytes, and more than 90% of granulocytes are neutrophils, making up the largest fraction of white blood cells. Feeding a high-fat diet (HFD) to mice causes an increase in neutrophil recruitment to adipose tissue13. Consequently, it is possible that neutrophils could have a role in initiating the

inflammatory cascade in response to obesity. Adipose tissue neutrophils (ATNs) produce chemokines and cytokines, thereby facilitating macro-phage infiltration, which could contribute to the chronic low-grade inflammation that characterizes obesity-induced insulin resistance.

In addition to host defense, neutrophil-derived serine proteases, such as neutrophil elastase, have also been implicated in sterile inflam-mation12. Therefore, we asked whether neutrophils contribute to the etiology of inflammation-induced insulin resistance and whether neutrophil elastase has a mechanistic role in this process.

We determined the time course of neutrophil infiltration in adipose tissue of C57BL/6J mice maintained on a 60% HFD starting at 8 weeks of age using fluorescence-activated cell sorting (FACS) analyses to identify adipose tissue stromal vascular cells (SVCs) positive for the neutrophil markers Ly6g and Cd11b and negative for the macrophage markers F4/80 and Cd11c. We refer to these cells as ATNs. Consistent with a previous report13, we detected a rapid increase in ATN content after 3 d of HFD feeding, and we also found that this increase was maintained for up to 90 d of an HFD regimen (Fig. 1a). Immunohistochemistry (IHC) studies also showed a higher content of Ly6g+Cd11b+ cells in the white adipose tissue (WAT) of the DIO mice compared to chow-fed mice (Fig. 1b). Consistent with the IHC results, our FACS analysis showed a 20-fold higher number of Cd11b+Ly6g+F4/80−Cd11c−cells in SVCs of mice fed the HFD compared to the chow-fed mice (Fig. 1c and Supplementary Fig. 1a).

Expression of neutrophil elastase is higher in adipose tissue of DIO mice than in that from chow-fed mice (Supplementary Fig. 1c). The expression of neutrophil elastase increased after only 3 d of HFD and remained elevated after 12 weeks of HFD (Fig. 1d), which is com-parable to the pattern of increased ATN content seen in DIO mice. Consistent with the increases in ATN content and neutrophil elastase expression after HFD feeding, neutrophil elastase activity was also markedly higher in mice after 12 weeks of HFD compared to mice at the same time point fed standard chow (Fig. 1e and Supplementary Fig. 2). Wild-type (WT) DIO mice treated with the neutrophil elastase inhibitor GW311616A14 at a dose of 2 mg per kg of body weight of the mice per day for 14 d showed substantially improved glucose tolerance (Fig. 1f) with no change in body weight (Supplementary Fig. 3a).

Neutrophils mediate insulin resistance in mice fed a high-fat diet through secreted elastaseSaswata Talukdar1,3,4, Da Young Oh1,4, Gautam Bandyopadhyay1, Dongmei Li2, Jianfeng Xu1, Joanne McNelis1, Min Lu1, Pingping Li1, Qingyun Yan2, Yimin Zhu2, Jachelle Ofrecio1, Michael Lin1, Martin B Brenner2 & Jerrold M Olefsky1

1Department of Medicine, University of California, San Diego, La Jolla, California, USA. 2Pfizer, Cardiovascular, Metabolic and Endocrine Diseases (CVMED)–Diabetes Prevention and Remission, Cambridge, Massachusetts, USA. 3Present address: Pfizer, CVMED–Diabetes Prevention and Remission, Cambridge, Massachusetts, USA. 4These authors contributed equally to this work. Correspondence should be addressed to J.M.O. (jolefsky@ucsd.edu).

Received 9 March; accepted 21 June; published online 5 August 2012; doi:10.1038/nm.2885

©20

12 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

l e t t e r s

�  advance online publication nature medicine

Treatment with GW311616A also lessened the severity of HFD-induced glucose intolerance when administered for 14 d (2 mg per kg of body weight of the mice per day) beginning at the start of HFD feed-ing compared to mice administered vehicle (Supplementary Fig. 3c). In contrast, treatment with recombinant mouse neutrophil elastase (1 mg per kg of body weight of the mice per day for 7 d) of chow-fed mice led to markedly higher glucose intolerance in the treated mice compared to that in untreated mice (Fig. 1g).

Based on these treatment results and the fact that the number of ATNs was higher in DIO mice than in chow-fed mice, we studied mice carrying a genetic deletion of neutrophil elastase (B6.129X1-Elanetm1Sds/J, referred to here as NEKO). On an HFD, we found that the NEKO mice gained somewhat less weight compared to WT mice (Supplementary Fig. 4a), which was consistent with their modestly higher core temperature (Supplementary Fig. 4b) and oxygen con-sumption (Supplementary Fig. 4c). Liver weight was substantially lower and WAT weight was substantially higher in NEKO mice com-pared to WT mice, both in absolute amounts and when expressed as a percentage of body weight (Supplementary Fig. 4d). After 10 weeks of HFD feeding, we performed glucose tolerance tests (GTTs) in NEKO mice and used age- and weight-matched WT mice on the same diet as controls. The NEKO mice had markedly higher glucose tolerance (Fig. 1h) and lower fasting insulin concentrations compared to WT mice (Supplementary Fig. 4e). Consistent with the GTT results, insu-lin tolerance tests (ITTs) showed that the NEKO mice were substan-tially more sensitive to insulin than the WT mice (Fig. 1i).

ATN content was ~90% lower in the NEKO mice compared to the WT mice on HFD (Fig. 1j). In addition, WT DIO mice treated with the neutrophil elastase inhibitor also showed a marked decrease in ATN content compared to WT DIO mice treated with vehicle (Fig. 1j). Thus, both genetic and pharmacologic loss of function of neutrophil elastase produced improved glucose tolerance with less ATNs, whereas pharmacologic gain of function led to glucose intolerance.

To quantify the overall magnitude of the insulin sensitivity and to deter-mine tissue-specific contributions, we performed hyperinsulinemic- euglycemic clamp studies. The glucose infusion rate that was required to maintain euglycemia was higher in NEKO mice compared to WT mice (Fig. 2a). There were no differences in total glucose disposal rate (Fig. 2b), insulin-stimulated glucose disposal rate (Fig. 2c) or basal hepatic glucose production (HGP) (Fig. 2d) between the two groups. However, insulin had a greater ability to inhibit HGP in the NEKO mice compared to the WT mice, as shown by the clamp HGP (Fig. 2e) and percentage of suppression of HGP (Fig. 2f). Basal and clamp free fatty acid (FFA) concentrations were substantially lower in NEKO mice than in WT mice (Fig. 2g), and the ability of insulin to suppress FFA concentrations was greater in NEKO mice than in WT mice (Fig. 2h). Taken together, these results show that neutrophil elastase deletion leads to a high degree of hepatic and adipose tissue insulin sensitivity.

Next, we performed acute insulin response studies by intravenously injecting insulin and harvesting liver and epididymal WAT (eWAT) from the NEKO and WT mice. Consistent with the ex vivo glucose clamp

bCaveolin Cd11b Ly6G Merge

MergeLy6GCd11bCaveolin

HF

DC

how

a2.0

AT

N (

% S

VC

) 1.5

1.0

0.5

0NC 3 7 30

Time after feeding HFD (d)

60 90

c

Chow HFD

2.0 *

AT

N (

% S

VC

) 1.5

1.0

0.5

0

d5

Neu

trop

hil e

last

ase

(fol

d in

duct

ion)

NC 3 7 30

Time after feeding HFD (d)

60 90

4

3

2

1

0

e

Chow

HFD

300

*

Tot

al s

igna

l(s

cale

d co

unts

s–1

)

250

200

150

100

50

0

AT

N (

% S

VC

)

1.5

WT

NEKO

*

1.0

0.5

0

j

AT

N (

% S

VC

)

Vehicl

e

Inhib

itor

1.5

2.0

1.0

0.5

0

f500

Glu

cose

(m

g dl

–1)

400

300

200

100WT

0 30 60Time (min)

90

**

*

NE inhibitor0

h

0 30 60Time (min)

90

* **

*WT weight matchedWT age matchedNEKO

500

400

300

200

100

0

i

Glu

cose

(m

g dl

–1)

250

WTNEKO

0 30 60Time (min)

90

* **

200

150

100

50

0

g300

200

100

0 30 60Time (min)

90

**

*

NESaline

0

Figure 1 Neutrophils infiltrate eWAT in mice on HFD, and ablation of neutrophil elastase improves insulin sensitivity in DIO mice. (a) ATNs (Cd11b+Ly6g+F4/80−Cd11c− cells) as a percentage of SVCs from the eWAT of DIO mice analyzed by FACS. n = 3–4 mice per time point. Error bars, s.e.m. (b) Adipose tissue from C57BL/6J mice on chow or HFD stained for caveolin, Ly6g and Cd11b. The merged Ly6g+Cd11b+ cells (yellow) indicated by white arrowheads are ATNs. Scale bars, 100 µm. (c) FACS analyses showing ATN content (ATNs as a percentage of SVCs) in 4-month-old C57BL/6J mice fed chow and 20-week-old mice fed HFD for 12 weeks. Error bars, s.e.m. *P < 0.05 by two-way analysis of variance (ANOVA) and Bonferroni’s post test. (d) Neutrophil elastase mRNA in eWAT of DIO mice normalized to RNA PolII. (e) Neutrophil elastase activity in the eWAT of male C57BL/6J mice fed chow or 60% HFD for 12 weeks. Chow-fed mice (n = 5) and HFD-fed mice (n = 8). (f) GTT in 7-h–fasted C57BL/6J mice fed 60% HFD for 12 weeks that were orally administered vehicle or neutrophil elastase (NE) inhibitor (GW311616A) every day for 2 weeks. Vehicle-treated mice (n = 12) and inhibitor-treated mice (n = 8). (g) GTT in 5-month-old C57BL/6J mice fed normal chow and administered 1 mg kg−1 recombinant mouse neutrophil elastase for 7 d. n = 10 mice per group. (h) Intraperitoneal GTT (IP-GTT) on 7-h–fasted mice fed HFD for 10 weeks, using an intraperitoneal dose of 1 mg kg−1 glucose, in NEKO mice and in WT mice matched on weight or age. n = 8–10 mice per group. (i) ITT using 0.6 U kg−1 insulin injected intraperitoneally in weight-matched WT and NEKO mice fed HFD for 6 weeks. n = 8–10 mice per group. (j) FACS showing ATNs in WT and NEKO mice on HFD for 12 weeks (left). FACS showing ATNs in 12-week–HFD mice treated with vehicle or the neutrophil elastase inhibitor (GW311616A) for 2 weeks (right). *P < 0.05 by two-way ANOVA and Bonferroni’s post test. Error bars (c–j), s.e.m.

©20

12 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

l e t t e r s

nature medicine  advance online publication �

studies, the biochemical measures of hepatic and adipose insulin signal-ing were higher in NEKO mice compared to WT mice, as evidenced by elevated insulin-stimulated phosphorylation of Akt (Fig. 2i).

As studies have reported neutrophil infiltration of liver in obese humans with nonalcoholic steatohepatitis15, we assessed hepatic neu-trophil content by IHC in our experimental mouse groups. Hepatic neutrophil content was higher in WT mice fed the HFD compared to those fed chow, with no elevation of neutrophil content found in the NEKO mice fed the HFD (Fig. 3a). Hepatic neutrophil elastase activity was also higher in WT HFD-fed mice compared to WT lean-chow–fed mice (Fig. 3b and Supplementary Fig. 2).

Previous reports have shown that extracellular neutrophil elastase can gain access to the intracellular space and mediate degradation of insulin receptor substrate 1 (Irs1)16,17. Consistent with these reports, the concentration of Irs1 was higher in the liver (Fig. 3c) and adipose tissue (Supplementary Fig. 5d) of NEKO mice compared to WT mice. Next, we investigated whether neutrophil elastase can inhibit hepatic Irs1-mediated insulin signaling. We administered neutrophil elastase (1 mg per kg body weight of the mice, 2 h apart) to 4-month-old fasted C57BL/6J mice. Two hours after the second dose, we performed an acute insulin response and harvested liver before and at 3 mins and 7 mins after insulin injection. Administration of neutrophil elastase in these mice caused a decrease in the levels of Irs1 in the liver (Fig. 3d) and of phosphorylated Akt (pAkt) in the basal and insulin-treated states (Fig. 3e). The level of Irs1 in the eWAT from these neu-trophil elastase–treated mice was also substantially lower than that in vehicle-treated control mice (Supplementary Fig. 5a). In contrast, Irs1 expression was higher in the liver (Fig. 3c) and adipose tissue (Supplementary Fig. 5d) of NEKO mice compared to WT mice.

Next, we added neutrophil elastase directly to primary mouse (Fig. 3f) and primary human hepatocytes (Fig. 3g) and found a marked decrease in Irs1 protein content in both cell types, which is consist-ent with enhanced degradation. This lowering of Irs1 protein con-tent resulted in reduced insulin-stimulated Akt phosphorylation in mouse (Fig. 3h) and human (Fig. 3i) hepatocytes. Although our data showed lower Irs1 protein content in hepatocytes after treatment with neutrophil elastase, it is unclear whether this is the result of a direct intracellular or an indirect extracellular effect of the protease.

To determine whether these changes in hepatocyte insulin signaling resulted in biological effects, we measured glucose output in primary mouse hepatocytes with and without neutrophil elastase treatment (Fig. 3j). Glucagon treatment leads to a near twofold increase in pri-mary hepatocyte glucose output, and insulin treatment suppresses this

output to basal values. In the absence of glucagon, neutrophil elastase treatment stimulated basal glucose output by 50%, and insulin did not inhibit this effect (Fig. 3j). More notably, neutrophil elastase treatment largely prevented the ability of insulin to inhibit glucagon-stimulated hepatocyte glucose output. Thus, neutrophil elastase causes Irs1 deg-radation, lower insulin signaling, higher glucose production and cel-lular insulin resistance in primary mouse and human hepatocytes.

We also measured mRNA levels of liver lipogenic and cholesterol synthesis genes and found that Acc, Fas, Scd, Hmgcr and Hmgcs1 (Hmgcs) were markedly lower, whereas PpargC1a (PGC1a) and Nrf1 (Nrf) were higher, in livers from NEKO mice compared to livers from WT mice. There was no difference in the mRNA abundance of Srebpf1 (Srebp1c), Pepck, G6pc (G6Pase) or Cpt1a mRNA between NEKO and WT mice (Supplementary Fig. 6a). To confirm that decreased lipo-genic gene expression resulted in functional changes in lipogenesis, we measured the incorporation of 14C-actetate into liver lipids in the NEKO and WT mice. The total amount of 14C-acetate incorporation (Supplementary Fig. 6b) and the rate of 14C-acetate incorporation (Supplementary Fig. 6c) into liver lipids were substantially lower in NEKO mice compared to the WT mice.

As we found better adipose tissue insulin signaling in NEKO mice compared to WT mice (Fig. 2i), we treated 3T3-L1 adipocytes with recombinant mouse neutrophil elastase. The results showed that this treatment led to decreased Irs1 protein content (Supplementary Fig. 5b) and impaired insulin-stimulated phosphorylation of Akt in a dose-dependent manner (Supplementary Fig. 5c).

Because previous reports have shown that neutrophil elastase can activate Toll-like receptor 4 (Tlr4), resulting in an increase of proin-flammatory factors12,18–20, we used intraperitoneal macrophages (IP-macs) from WT and Tlr4 knockout (Tlr4KO) mice and treated them with lipopolysaccharide (LPS) and recombinant mouse neu-trophil elastase. Treatment with LPS caused increased abundances of Tnfa, Il1b, Cxcl1 (KC) and Il6 mRNA in WT but not Tlr4KO IP-macs. Similar to LPS, treatment with recombinant neutrophil elastase caused higher mRNA abundance of Tnfa, Il1b, Cxcl1 (KC) and Il6 in WT but not Tlr4KO IP-macs, showing that the proinflammatory effects of neutrophil elastase are dependent on Tlr4 (Fig. 4a), which is consistent with previous reports12,18–20. The mRNA abundance of proinflammatory markers such as Tnfa, Emr1 (F4/80), Cxcl1 (KC), Il1r1, Il1b and Ccl2 (Mcp1) were substantially lower and the mRNA abundance of the anti-inflammatory marker Arginase was higher in NEKO mouse liver compared to WT liver (Fig. 4b). Consistent with this lower proinflammatory mRNA abundance, the amount of

a25

20

GIR

(m

g kg

–1 m

in–1

) *

15

10

5

0

WT

NEKO

b30

GD

R (

mg

kg–1

min

–1)

WT

NEKO

25

20

15

10

5

0

c12

Is-G

DR

(m

g kg

–1 m

in–1

)

WT

NEKO

10

8

6

4

2

0

d

Bas

al H

GP

(m

g kg

–1 m

in–1

) 20

WT

NEKO

15

10

5

0

e

Ins-

HG

P (

mg

kg–1

min

–1)

*

WT

NEKO

15

10

5

0

f

Sup

pres

sion

of H

GP

(%

)

*

WT

NEKO

100

0

80

60

40

20

g

FF

A (

mm

ol–1

)

WTNEKO

Basal

Clamp

*

*

1.5

1.0

0.5

0

h

Sup

pres

sion

of l

ipol

ysis

(%

) 60

WT

NEKO

*50

40

30

20

10

0

iIns

LiverpAkt

eWATpAkt

TotalAkt

WT– + – + – +– +

NEKO

Figure 2 A high degree of insulin sensitivity in NEKO mice. Hyperinsulinemic-euglycemic clamp studies in WT and NEKO knockout mice fed HFD for 10–11 weeks. (a–g) Glucose infusion rate (GIR; a), total glucose disposal rate (GDR; b), insulin-stimulated glucose disposal rate (Is-GDR; c), basal HGP (d), clamp HGP (e), percentage suppression of HGP (f) and FFAs in 6-h–fasted WT and NEKO mice (basal) and in the same mice at the end of the clamp procedure (clamp) (g). (h) Percentage suppression of lipolysis calculated from the data in g. *P < 0.05 by two-way ANOVA and Bonferroni’s post test. Error bars (a–h), s.e.m. (i) Western blot analysis of total Akt and pAkt in the liver and adipose tissue (eWAT) of WT and NEKO mice either before (–Ins) or after (+Ins; tissue harvested at 3 min for liver and at 7 min for eWAT) acute insulin injection.

©20

12 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

l e t t e r s

�  advance online publication nature medicine

degradation of nuclear factor of κ light polypeptide gene enhancer in B cells inhibitor, β (IκB), was lower in NEKO mice than in WT mice (Fig. 4b).

In adipose tissue, the mRNA abundance of Il1b, Tnfa, Cxcl1 (KC), Cd68, Irf4 and Irf5 were lower, and the mRNA abundance of Arg, Clec10a (Mgl1) and Il4 were higher, in NEKO mice compared to WT mice (Fig. 4c). The amount of degradation of IκB was also lower in NEKO mice than in WT mice (Fig. 4c). To determine the contribution of adipocytes compared to immune cells to these changes in mRNA abundance, we isolated SVCs and adipocytes from WT and NEKO mice and found that each fraction had lower expression of inflamma-tory markers in NEKO mice compared to WT mice (Supplementary Fig. 7a). FACS analyses for adipose tissue macrophage (ATM) content showed that the number of Cd11c+ M1-like cells was lower (Fig. 4d), whereas the number of F4/80+Cd11b+Cd11c− (also known as M2-like) cells was higher, in NEKO mice compared to WT mice (Fig. 4e). This suggests that elastase secreted from neutrophils has a role in recruiting these cells to adipose tissue and, possibly, in their polariza-tion state. For example, neutrophil elastase functions as an activator of Tlr4 (refs. 12,18–20), and stimulation of the Tlr4 proinflammatory

pathway leads to increased chemokine release from a variety of adi-pose tissue cell types, including adipocytes and macrophages1,2. Consistent with changes in inflammatory tone in liver and adipose tissue, serum concentrations of Il-1β, Tnf-α, chemokine (C-C motif) ligand 2 (Ccl2, also known as Mcp-1), Il-6, granulocyte-macrophage colony-stimulating factor (GM-CSF), Ccl3 (Mip1-α), Ccl4 (Mip1-β) (Fig. 4f) and resistin (Supplementary Fig. 7b) were lower in NEKO mice compared to WT mice.

Gene expression analyses of adipose tissue revealed that the abundances of Acc, Fas, Lipe (Hsl), Pnpla2 (Atgl) and Slc2a4 (Glut4) mRNA were markedly higher in NEKO mice than in WT mice (Supplementary Fig. 7c). We excised fat pads from the NEKO and WT mice and measured glucose uptake by measuring the levels of 2-deoxyglucose (2-DOG), and both basal and insulin-stimulated glu-cose uptake were higher in NEKO mice than in WT mice (Fig. 4g), which is consistent with the in vivo and in vitro data showing greater adipose insulin sensitivity in the NEKO mice.

In these studies, we show a sustained higher neutrophil content in adipose tissue and liver in DIO mice. As neutrophils are known to have a role in the early stages of inflammatory responses, it is probable

aChow WT HFD NEKO HFD

b

Chow HFD

*

Tot

al s

igna

l(s

cale

d co

unts

s–1

)

800

600

400

200

0

d 3,000

*

Live

r lrs

1 (A

U)

2,000

1,000

Saline NE0

e200

*

*

pAkt

/tota

l Akt

Basal Ins 3′ Ins 7′

150

100

50

0

c NEKO

NEKO

WT

WT

4*

Irs1

Hsp90

Den

sito

met

ry

3

2

1

0

*

Mou

se lr

s1/a

ctin

Saline NE

2.0

1.5

1.0

0.5

0

fVehicle

Irs1

Irs2

Actin

NEg

Vehicle

* *

Hum

an IR

S1/

Act

in

NE

– 0.5 µM 1.0 µM

500

400

300

200

100

0

h

**

pAkt

/tota

l Akt

3

2

1

0NE

Insulin– + – +– +– +

i

*

pAK

T/to

tal A

KT

8

– + – +– +– +

6

4

2

0

j*, #

*, #

*, #

C.p

.m. ×

105 m

g–1 p

rote

in

16

–––

– – –– –

– –+ ++ + +

+ +++

+ +

14

12

10

8

6

4

2

0Glucagon

InsulinNE

Figure 3 Neutrophils infiltrate the liver in HFD-fed mice and cause impaired insulin signaling by degradation of Irs1. (a) IHC of liver sections obtained from 4-month-old chow-fed WT mice and 20-week-old WT (WT HFD) and NEKO (NEKO HFD) mice fed HFD for 12 weeks. Red indicates lipid content in hepatocytes stained with boron-dipyrromethene (BODIPY), blue indicates nuclei stained with DAPI, and green indicates neutrophils stained with Ly6g (1A8). Scale bars, 50 µm. (b) Neutrophil elastase activity in the livers of male C57BL/6J mice fed chow or an HFD for 12 weeks. Chow-fed mice (n = 5) and HFD-fed mice (n = 8). (c) Western blots for total Irs1 and heat shock protein 90 (Hsp90) as the loading control in WT and NEKO mice fed an HFD for 12 weeks. A densitometry analysis was performed and is represented below the blot. (d) The amount of liver Irs1 quantified by Meso Scale Discovery (MSD) analyses from fasted 4-month-old C57Bl6/J mice on chow diet. Mice were injected with saline or recombinant mouse neutrophil elastase. Tissues were harvested 2 h after neutrophil elastase injection. AU, arbitrary units. (e) Acute insulin response in 4-month-old C57BL/6J mice on chow treated with recombinant mouse neutrophil elastase. A submaximal insulin dose (0.1 U−1 kg) was used, and liver samples were obtained at basal (−Ins) and at the indicated times points (Ins 3′, 3 minutes after insulin injection; Ins 7′, 7 minutes after insulin injection). MSD was used to analyze the total amount of Akt and pAkt. (f) Western blots of protein harvested from primary mouse hepatocytes treated with recombinant mouse neutrophil elastase for 4 h. Irs1/actin refers to the amount of Irs1 protein normalized to the amount of Actin protein. (g) Amount of IRS1 in cryopreserved human hepatocytes treated with purified human neutrophil elastase for 6 h. Shown is a representative graph generated from data from one of at least three independent experiments. IRS1/Actin refers to the amount of Irs1 protein normalized to the amount of Actin protein. (h) Quantification of western blots of primary mouse hepatocytes treated with neutrophil elastase for 4 h and spiked with insulin for 5 min to obtain insulin induction. Shown is a representative graph generated from data from one of at least three independent experiments. (i) Quantification by MSD of total AKT and pAKT in protein harvested from primary human hepatocytes treated with human neutrophil elastase for 6 h and spiked with insulin for 5 min to obtain insulin induction. The graph was generated with data from at least two independent experiments performed in triplicate. *P < 0.05 by two-way ANOVA and Bonferroni’s post test. (j) Representative glucose output assay in primary mouse hepatocytes from at least three independent experiments performed in duplicate or triplicate. *Significantly (P < 0.05) higher than basal, #significantly (P < 0.05) higher than insulin and glucagon by Student’s t test. C.p.m., counts per million. Error bars (b–j), s.e.m.

©20

12 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

l e t t e r s

nature medicine  advance online publication �

that neutrophils participate in the inflammation that characterizes obesity. Furthermore, secreted elastase from neutrophils is the key effector in this process. Consistent with this, we find that addition of neutrophil elastase to hepatocytes or adipocytes causes cellular insulin resistance, and, in the context of HFD-induced obesity, NEKO mice were protected from adipose tissue and liver inflammation, with a corresponding high degree of insulin sensitivity in adipose tissue and hepatocytes. We also treated WT DIO mice with a small-molecule neutrophil elastase inhibitor and found that this compound improved glucose tolerance in these mice.

It is of interest that inhibition or knockout of neutrophil elastase results in decreased ATN content. One possible mechanism for this is that since neutrophil elastase activates the Tlr4 pathway, it could cause chemokine release from immune cells and/or adipocytes causing a feed-forward secondary mechanism of ATN recruitment. Also, as we have demonstrated, neutrophil elastase inhibition improves insulin resist-ance and this could lead to less chemokine release from adipocytes.

In addition to macrophages, a variety of immune cells, such as lymphocytes, eosinophils and mast cells, have been shown to par-ticipate in the complex intracellular communication network that organizes the chronic inflammatory response of obesity. Based on our studies of chronic HFD mice and those of others13 showing the pres-ence of ATNs after acute HFD feeding, we suggest that neutrophils should be considered active participants in the immune-cell–type conversation, which ultimately leads to obesity-induced inflammation and insulin resistance.

MeTHODsMethods and any associated references are available in the online version of the paper.

Note: Supplementary information is available in the online version of the paper.

AckNOwLeDGMeNTSThis work was supported by grants to J.M.O. from the US National Institutes of Health (NIH): DK033651, DK074868, T32 DK 007494, DK 090962 and DK063491. This work was also supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development, NIH through cooperative agreement U54 HD 012303-25 as part of the specialized Cooperative Centers Program in Reproduction and Infertility Research. We wish to thank S. Shapiro (University of Pittsburgh) for helpful comments over the course of the project, C. Pham (Washington University, St. Louis) for providing cathepsin G and neutrophil elastase double knockout mice, S. Nalbandian at University of California San Diego for breeding and caring for the NEKO mice and P. Bansal (Pfizer), B. Ghosh (Pfizer) and J. Wellen (Pfizer) for the neutrophil elastase activity imaging studies.

AUTHOR cONTRIBUTIONSS.T. and D.Y.O. designed and performed the experiments. S.T., D.Y.O. and J.M.O. analyzed and interpreted data and co-wrote the manuscript. All other authors performed experiments and contributed to discussions. This work was supported by grants to J.M.O., as detailed above.

cOMPeTING FINANcIAL INTeReSTSThe authors declare no competing financial interests.

Published online at http://www.nature.com/doifinder/10.1038/nm.2885. Reprints and permissions information is available online at http://www.nature.com/reprints/index.html.

1. Olefsky, J.M. & Glass, C.K. Macrophages, inflammation, and insulin resistance. Annu. Rev. Physiol. 72, 219–246 (2010).

2. Gregor, M.F. & Hotamisligil, G.S. Inflammatory mechanisms in obesity. Annu. Rev. Immunol. 29, 415–445 (2011).

3. Nishimura, S. et al. CD8+ effector T cells contribute to macrophage recruitment and adipose tissue inflammation in obesity. Nat. Med. 15, 914–920 (2009).

4. Winer, S. et al. Normalization of obesity-associated insulin resistance through immunotherapy. Nat. Med. 15, 921–929 (2009).

a80

130

700

Tnfa

Cxcl1

Il1b

Il6

40

108

6

4

20

Vehicle LPS

WT

#

#

#

#

*

* *

*

NE

Vehicle LPS NE Vehicle LPS NE

Vehicle LPS NE

Tlr4KO 450200

6

4

2

0

609050

6

4

2

0

306

4

2

0

g*

*

2-D

OG

upt

ake

(c.p

.m.) WT

Basal Insulin

NEKO8,000

6,000

4,000

2,000

0

b

*

** * *

**

WT WT12

Rel

ativ

e m

RN

A in

live

r

NEKO

NEKO

IκB

Hsp908

4

0

Tnfa

Emr1

Cxcl1

ll1r1

ll1r2

Il1a

Il1b

ll1ra ll4 ll10

ll13

Arg

Ccl2

c

* * * * * *

*

*24

Rel

ativ

e m

RN

A in

eW

AT

WTNEKO

WT NEKO

IκB

Hsp90

Ccl2

Il1b

Tnfa ll6

Cxcl1

Cd68 ll10

lrf4

lrf5

Arg

Mmr

Clec

Foxp3

Clec10a ll4 ll13

20

16

12

8

4

0

d

*

AT

M1

(% S

VC

)

WT WT NEKO

HFDChow

40

30

20

10

0

e

*

AT

M2

(% S

VC

)

WT WT NEKO

HFDChow

30

20

10

0

f

* **

* * **

WT

Ser

um c

ytok

ines

(A

U)

NEKO

250

Il-1β

Tnf

-α

Cxc

l1

Mcp

1

Il-6

Il-13

GM

-CS

F

Mip

1-α

Mip

1-β

200

150

100

50

0

Figure 4 NEKO mice have low inflammatory tone. (a) Quantitative PCR (qPCR) of the indicated genes from IP-macs harvested from WT and Tlr4 knockout mice treated with vehicle, LPS or recombinant mouse neutrophil elastase. *Significantly higher (P < 0.05) than all other treatment groups, #significantly lower (P < 0.05) than LPS-treated WT and significantly higher (P < 0.05) than all other treatment groups by Student’s t test. (b,c) qPCR analysis of inflammatory gene expression in liver (b) and adipose tissue (c) of WT and NEKO mice on an HFD. Also shown are western blots of liver IκB and Hsp90 (b, inset) and adipose IκB and Hsp90 (c, inset). (d,e) Quantification of SVCs from eWAT of chow-fed WT, HFD-fed WT and NEKO mice stained for F4/80, Cd11b and Cd11c and analyzed by FACS. Cells that are triple positive for all three markers are referred to as ATM1 (d), and cells that are positive for F4/80 and Cd11b and negative for Cd11c are referred to as ATM2 (e). (f) Serum cytokines measured using the Millipore Luminex assay from WT and NEKO mice on an HFD. *P < 0.05 by Student’s t test (c,f). (g) Glucose uptake in eWAT explants harvested from WT and NEKO mice and incubated ex vivo in the absence or presence of insulin followed by measurement of 2-DOG uptake. *P < 0.05 by two-way ANOVA and Bonferroni’s post test (d,e and g). Error bars (a–g), s.e.m.

©20

12 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

l e t t e r s

�  advance online publication nature medicine

5. Feuerer, M. et al. Lean, but not obese, fat is enriched for a unique population of regulatory T cells that affect metabolic parameters. Nat. Med. 15, 930–939 (2009).

6. Winer, D.A. et al. B cells promote insulin resistance through modulation of T cells and production of pathogenic IgG antibodies. Nat. Med. 17, 610–617 (2011).

7. Liu, J. et al. Genetic deficiency and pharmacological stabilization of mast cells reduce diet-induced obesity and diabetes in mice. Nat. Med. 15, 940–945 (2009).

8. Wu, D. et al. Eosinophils sustain adipose alternatively activated macrophages associated with glucose homeostasis. Science 332, 243–247 (2011).

9. Mantovani, A., Cassatella, M.A., Costantini, C. & Jaillon, S. Neutrophils in the activation and regulation of innate and adaptive immunity. Nat. Rev. Immunol. 11, 519–531 (2011).

10. Nathan, C. Neutrophils and immunity: challenges and opportunities. Nat. Rev. Immunol. 6, 173–182 (2006).

11. Borregaard, N. Neutrophils, from marrow to microbes. Immunity 33, 657–670 (2010).

12. Pham, C.T. Neutrophil serine proteases: specific regulators of inflammation. Nat. Rev. Immunol. 6, 541–550 (2006).

13. Elgazar-Carmon, V., Rudich, A., Hadad, N. & Levy, R. Neutrophils transiently infiltrate intra-abdominal fat early in the course of high-fat feeding. J. Lipid Res. 49, 1894–1903 (2008).

14. Macdonald, S.J. et al. The discovery of a potent, intracellular, orally bioavailable, long duration inhibitor of human neutrophil elastase–—GW311616A a development candidate. Bioorg. Med. Chem. Lett. 11, 895–898 (2001).

15. Rensen, S.S. et al. Increased hepatic myeloperoxidase activity in obese subjects with nonalcoholic steatohepatitis. Am. J. Pathol. 175, 1473–1482 (2009).

16. Houghton, A.M. et al. Neutrophil elastase–mediated degradation of IRS-1 accelerates lung tumor growth. Nat. Med. 16, 219–223 (2010).

17. Houghton, A.M. The paradox of tumor-associated neutrophils: fueling tumor growth with cytotoxic substances. Cell Cycle 9, 1732–1737 (2010).

18. Walsh, D.E. et al. Interleukin-8 up-regulation by neutrophil elastase is mediated by MyD88/IRAK/TRAF-6 in human bronchial epithelium. J. Biol. Chem. 276, 35494–35499 (2001).

19. Devaney, J.M. et al. Neutrophil elastase up-regulates interleukin-8 via toll-like receptor 4. FEBS Lett. 544, 129–132 (2003).

20. Brightbill, H.D. et al. Host defense mechanisms triggered by microbial lipoproteins through toll-like receptors. Science 285, 732–736 (1999).

©20

12 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature medicinedoi:10.1038/nm.2885

ONLINe MeTHODsMice. NEKO (JAX labs, B6.129X1-Elanetm1Sds/J, 006112), and WT C57BL/6J (000664) mice were obtained from JAX. Only male mice were used in this study. We placed NEKO and age-matched WT mice on a 60% HFD (D12492, Research Diets) for 12 weeks and monitored their body weight and food intake. For weight-matched controls, we used WT mice on a 60% HFD (D12492) that were about 2–3 weeks younger than the NEKO mice. We housed all mice according to UCSD and Institutional Animal Care and Use Committee–approved protocols.

All mouse procedures conformed to the Guide for Care and Use of Laboratory Animals of the US National Institutes of Health and were approved by the Animal Subjects Committee of UCSD.

Metabolic studies. For ITT and GTTs, we fasted mice for 7 h. For GTT, we administered an intrapertioneal (i.p.) dose of 1 g kg−1 dextrose and measured blood glucose concentrations at the indicated time points. We measured fast-ing insulin after 6 h of fasting on the day of GTT. For ITT, we used an i.p. dose of 0.6 U kg−1 insulin and measured blood glucose concentrations at the indicated time points.

We performed hyperinsulinemic-euglycemic clamp studies as previously described21. We only used mice that lost <4% of their precannulation weight after 4–5 d of recovery. We administered a constant infusion (5 µCi h−1) of D-[3-3H] glucose (DuPont-NEN) 6 h after fasting. After 90 min of tracer equi-libration and basal sampling, we infused 50% dextrose (Abbott) and tracer (5 µCi h−1) plus insulin (6 mU kg−1 min−1) in the jugular vein cannula. We sampled blood from tail clips at 10-min intervals, and steady-state conditions (120 mg dl−1 ± 5 mg dl−1) were achieved at the end of the clamp by maintaining glucose infusion and plasma glucose concentration for a minimum of 20 min.

Acute insulin response. We fasted mice either overnight or for 7 h. We per-formed this procedure on WT and NEKO mice on an HFD at an insulin dose of 0.35 U kg−1. We also performed this procedure on WT C57BL/6J mice on a chow diet injected with vehicle and recombinant mouse neutrophil elastase (4517-SE) at a submaximal insulin dose of 0.1 U kg−1. Briefly, we anesthetized mice, suture sealed a small lobe of the liver to prevent bleeding and excised it. We also obtained an eWAT biopsy and flash froze it in liquid nitrogen. These samples are referred to here as ‘basal’. We injected insulin at the above-mentioned dose into the inferior vena cava and obtained tissue biopsies as described at the indicated time points and flash froze them in liquid nitrogen. We prepared lysates and ran western blots or MSD according to standard protocol.

Indirect calorimetry and core temperature. We performed indirect calorim-etry between week 7 and 8 of HFD, as previously described22. We individually housed mice in metabolic cages for measurements. We allowed all mice to adapt to the new environment for 48 h before beginning the study. We nor-malized oxygen consumption (VO2) and carbon dioxide production (VCO2) with respect to body weight. We calculated energy expenditure based on the formula: energy expenditure = 3.815 × VO2 + 1.232 × VCO2.

We measured the core temperature of the mice by insertion of a rectal probe and took all temperature measurements under fed conditions at about 3 p.m. We took the temperature measurements at between 8 and 9 weeks of HFD feeding.

Western blotting and gene expression analyses. We performed western blot-ting as previously described23. All antibodies used were from Cell Signaling Technology. We performed qPCR as previously described24.

FACS analyses. We weighed epididymal fat pads rinsed in PBS and then minced in FACS buffer (PBS containing 1% BSA). We prepared adipocytes and SVCs from collagenase-digested adipose tissue. We performed FACS analysis of SVCs for macrophage content and subtypes, as described21, and the estimation of macrophage subsets numbers per gram of fat was performed as previously described21. We incubated stromal vascular cells with Fc Block (BD Biosciences, San Jose, CA) for 20 min at 4 °C before staining with fluorescently labeled primary antibodies, and no stain, single stains and fluorescence minus

one controls were used for setting compensation and gates. We purchased the neutrophil marker Ly6g antibody from BD Biosciences (1A8). We used the Ly6g antibody at 4 µg/ml for 30 min at 4 °C.

Confocal microscopy of mouse adipose tissue. We excised fingernail-sized fat pad samples and blocked them for 1 h in 5% BSA in PBS with gentle rock-ing at room temperature. For the detection of intracellular antigens, blocking and subsequent incubations were done in 5% BSA in PBS with 0.3% Triton X-100. We diluted primary antibodies in blocking buffer to 0.5-1 mg ml−1 and added them to the fat samples overnight at 4 °C. After three washes, we added fluorochrome-conjugated secondary antibodies for 1 h at room tem-perature. We imaged fat pads on an inverted confocal microscope (Olympus Fluoview 1000). The mouse-specific antibodies used were to Cd11b (M1/70, Abcam, 1:500), caveolin (BD Biosciences, 610060, 1:500) and Ly6g (1A8, BD Biosciences, 1:100).

Lipogenesis in liver explants. We quickly sliced liver samples into chunks of approximately 1- to 2-mm and transferred them into wells containing 0.5 ml phosphate salt bicarbonate buffer (10 mM 4-(2-hydroxyethyl)- 1-piperazineethanesulfonic acid (HEPES), 4 mM KCl, 125 mM NaCl, 0.85 mM KH2PO4, 1.25 m Na2HPO4, 1 mM MgCl2, 1 mM CaCl2 and 15 mM NaHCO3) containing 0.5% fatty-acid–free BSA and 0.2 mM unlabeled sodium acetate in 12-well culture plates and incubated them for 30 min. We added 1 µCi 2-14C-acetate to each well and incubated the plates for 60 min at 37 °C in a 5% CO2 incubator. We terminated incubation by adding 0.5 ml of 2 N HCl to each well. We transferred explants with the incubation buffer to 5-ml polyethylene tubes, centrifuged them and removed the supernatants. Next, we added 0.5 ml 1 N HCl to each tube and homogenized the explant suspensions, which we followed with the addition of 1.5 ml of a mixture of chloroform and methanol (2:1) to each homogenate. We vortexed the tubes to obtain a monophasic mixture. Next, 0.5 ml chloroform and 0.5 ml 1 M NaCl were added to the homogenates, and the tubes were vortexed and centrifuged to obtain separate two layers. We collected the lower chloroform layer, evaporated it and counted for the incorporation of labeled acetate into lipids. We saved an aliquot of the homogenates for a protein assay before adding an organic solvent cocktail. The specific activity in the incubation was 22,000 c.p.m. nmol−1 acetate.

Glucose uptake in adipose explants. We quickly sliced adipose tissue samples into 1- to 2-mm sizes. Adipose explants prepared from half of each epididymal fat pad were taken into wells containing 0.5 ml phosphate salt bicarbonate buffer (10 mM HEPES, 4 mM KCl, 125 mM NaCl, 0.85 mM KH2PO4, 1.25 mM Na2HPO4, 1 mM MgCl2, 1 mM CaCl2 and 15 mM NaHCO3) containing 0.5% fatty-acid–free BSA in 12-well culture plates. We incubated plates for 15 min in a 5% CO2 incubator at 37 °C. We added insulin to some wells at a final concentration of 17 nM and incubated the plates for an additional 30 min. Next, we added 0.2 µCi 3H-2-deoxyglucose (final concentration 0.1 mM) to each well and incubated the plates for 10 min at 37 °C in a 5% CO2 incubator. We terminated the incubation by placing the plates on an ice tray followed by the addition of cytochalasin B (final concentration 0.01 mM) to stop further uptake. We carefully removed the radiolabeled buffer, avoiding pieces of the explants, and washed the explants three times with chilled PBS. Next, we added 0.5 ml of 1 N NaOH to each well, and the plates were shaken for 30 min at room temperature. We transferred the alkaline suspensions to 5-ml culture tubes and homogenized them. We saved an aliquot for protein assay and transferred the rest of the homogenate to 10-ml scintillation vials. We neutralized the alkaline homogenates, with 0.5 ml 1 N HCl added into the vials, and counted radioactivity using a 10-ml scintillation cocktail.

IP-mac isolation and culture. We harvested IP-macs from WT and Tlr4 knockout mice, as previously described21. Three days after harvest and plat-ing, we treated cells with 100 nM recombinant mouse neutrophil elastase (R&D Systems, 4517S E) and LPS (100 ng ml−1) for 6 h before RNA isolation and qPCR analyses. We used heat-inactivated neutrophil elastase as a control treatment.

Human hepatocytes culture. We used primary cryopreserved plateable human hepatocytes in this experiment from BD Bioscience (310) and used

©20

12 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature medicine doi:10.1038/nm.2885

the BD Cryohepatocytes Recovery Protocol to recover the cells and then plated them overnight. The morning after overnight plating, we treated the cells in triplicate with human neutrophil elastase (Innovative Research, M77514) at 1,000 nM and 500 nM in low-glucose medium (Gibco, 10567) and 0.5% charcoal-stripped FBS (Gibco, 934492) for 6 h and washed them once with ice-cold PBS. We used MSD to assay for Irs1, pAkt and total Akt. Twenty microliters of sample was assayed for Irs1 or pAkt and total Akt using the instructions provided by the kit’s manufacturer (Irs1, is N450HLA–1; pAkt and total Akt, K15100D-2).

Glucose output assay in mouse hepatocytes. We plated primary mouse hepato-cytes on collagen-coated plates and maintained them in Medium E with 10% FBS and 100 nM dexamethasone. We washed cultures with Krebs-Henseleit bicarbonate buffer with 2.5 mM calcium and 1% BSA and incubated them in the same buffer containing hormones and substrates in a 5% CO2 incubator. We used 2 mM pyruvate (0.5 µCi pyruvate per incubation) as the substrate and performed incubations in 0.5 ml buffer in a 12-well plate containing 0.25– 0.5 million cells per well. We incubated insulin (0.1 µM) or glucagon (100 ng ml−1) in the absence of substrate, followed by incubation for 3 h with substrates. At the end of incubation, we transferred buffers to 1.7-ml microfuge tubes and added 0.25 ml 5% ZnSO4 and 0.25 ml 0.3 N Ba(OH)2 suspensions to each tube, followed by addition of 0.5 ml water. We collected supernatants in a fresh set of tubes and assayed for radiolabeled glucose released into the media. We treated cells in the culture plates with 0.5 ml 10% HClO4, which were then shaken for 15 min, and the acidic extracts were transferred to microfuge tubes. We dissolved the cellular debris with proteins remaining on the plates in 1 NaOH and subjected it to protein assay. We neutralized the acidic extracts in the tubes with 0.5 ml of 2 M K2CO3. We removed precipitates by centrifugation and collected the supernatants for determination of intracellular radiolabeled glucose concentrations.

We performed the glucose assay by using the supernatants from the culture buffer and the assayed cell extracts for radioactive glucose by mixed (cation and anion) ion-exchange chromatography using AG 501×8 resins (Bio-Rad) using a batch-treatment method. We added 150–200 mg resins to each tube and vortexed them intermittently for 15 min. We centrifuged the tubes, and the supernatants were transferred to scintillation vials for counting radioactivity.

Neutrophil elastase activity. We used 6–month-old C57Bl6/J mice fed on a chow diet and 20-week-old C57Bl6/J mice fed on 60% HFD for these studies. On the day of the experiment, we administered all mice with 100 µl

Neutrophil Elastase 680 FAST (NE680 FAST) (Boston MA, PerkinElmer) by tail vein injection. The NE680 FAST comprises two NIR fluorochromes (VivoTag−S680, PerkinElmer, Boston, MA), one linked to the N terminus and the other linked to the C terminus of the peptide PMAVVQSVP, which is a highly neutrophil elastase–selective sequence25. NE680 FAST is optically silent in its native state and becomes highly fluorescent after cleavage by neutrophil elastase and can therefore be used as a direct sensing agent for neutrophil elastase activity.

Five hours after injection, we euthanized the mice and harvested their livers and eWAT. We immediately performed ex vivo liver and eWAT imaging on a CRi Maestro optical imaging platform using the red filter sets (excitation range of 616–660 nm, emission of 675 nm longpass). We spectrally unmixed the fluores-cent data to obtain a signal specific to NE680 FAST. We manually drew regions of interest (ROI) encompassing the whole organs, and the resulting signal was computed in the units of scaled counts s−1. Particular care was taken to ensure that the size of the ROIs drawn across the mice was constant.

3T3-L1 adipocyte culture. 3T3-L1 cells were differentiated and cultured as previously described21. On the day of the experiments, we treated cells with the indicated amount of recombinant mouse neutrophil elastase for 6 h. Before the addition of neutrophil elastase, we serum starved cells for 2 h. We added insulin at 10 nM for 5 min, after which total protein was harvested according to manufacturer’s protocols and run on MSD plates.

Statistical analyses. All statistical analyses were performed by two-tailed Student’s t test using Excel (Microsoft) or by two-way ANOVA (with repeated measures where necessary) followed by Bonferroni’s post-test using GraphPad Prism5 (San Diego, CA). P < 0.05 was considered significant. All data are expressed as means ± s.e.m.

21. Oh, D.Y. et al. GPR120 is an omega-3 fatty acid receptor mediating potent anti-inflammatory and insulin-sensitizing effects. Cell 142, 687–698 (2010).

22. Lu, M. et al. Brain PPAR-γ promotes obesity and is required for the insulin-sensitizing effect of thiazolidinediones. Nat. Med. 17, 618–622 (2011).

23. Li, P. et al. Functional heterogeneity of CD11c-positive adipose tissue macrophages in diet-induced obese mice. J. Biol. Chem. 285, 15333–15345 (2010).

24. Talukdar, S. & Hillgartner, F.B. The mechanism mediating the activation of acetyl–coenzyme A carboxylaseα gene transcription by the liver X receptor agonist T0-901317. J. Lipid Res. 47, 2451–2461 (2006).

25. Kalupov, T. et al. Structural characterization of mouse neutrophil serine proteases and identification of their substrate specificities: relevance to mouse models of human inflammatory diseases. J. Biol. Chem. 284, 34084–34091 (2009).