Introduction

Evidence from human populations [1] and animal research [2] has established correlative as well as causative [3] links between chronic inflammation and insulin resistance during obesity. However, the fundamental mechanisms for activating inflammatory pathways in obesity are poorly understood. One of the commonest causes of inflammation is microbial infection. The host’s acute-phase response to pathogens involves physiological changes including hypermetabolism, anorexia and body weight loss [4], but it is not known how this pathway intervenes in the chronic inflammatory state in obesity and in the control of insulin sensitivity.

Lipopolysaccharide (LPS) is continuously produced in the gut by the death of gram-negative bacteria and translocated into intestinal capillaries and the general circulation [5], where it can reach peripheral tissue such as adipose tissue and cause a local inflammatory reaction. Interestingly, LPS is found in the circulation of apparently healthy subjects [6, 7] and its activity may be associated with high-fat diet and dyslipidaemia. Infusion of triacylglycerol enhances LPS responses in humans, including the production of cytokines, suggesting that LPS can be locally released from lipoproteins in peripheral tissues through lipoprotein lipase action [8].

Portal endotoxin is involved in the progression of non-alcoholic steatohepatitis [9, 10], which mainly accounts for hepatic insulin resistance [11]. Genetically obese rodents quickly develop steatohepatitis after exposure to low doses of LPS [12], while polymyxin B diminishes hepatic steatosis during total parenteral nutrition in the rat by reducing caecal flora [13]. Moreover, endotoxins exert a strong effect on liver of mice fed a high-fat diet [14].

Interestingly, some groups [15] but not all [16] observed that toll-like receptor 4 (TLR4) 299Gly allele was associated with reduced C-reactive protein levels and a decreased risk of clinical diabetes, suggesting that TLR4 intervenes in the complications of obesity.

LPS in combination with CD14 serves as ligand for TLR4 [17]. TLR4 is produced by macrophages and Kupffer cells and recent data have demonstrated its presence in murine [18] and human adipocytes [19].

We hypothesised that the presence of non-functional TLR4 should modify the development of adipose tissue and liver steatosis during a high-fat diet. We studied the effects of an obesity- and insulin resistance-inducing high-fat diet in endotoxin-resistant C3H/HeJ mice (Tlr4 Lps-d), which carry a spontaneous mutation inactivating Tlr4 [20], and in their respective C3H/HeOuJ control mice. Complementary experiments were performed in Cd14 −/− mice.

Materials and methods

Mice and animal care

Adult (12-week-old) male C3H/HeJ and C3H/HeOuJ mice, as well as Cd14 −/− and C57Bl/6J control mice, were purchased from the Jackson Laboratory (Bar Harbor, ME, USA) and housed with a 12-h light–dark cycle. Mice had free access to a high-fat diet and water. The diet consisted of either a pelleted (42.7% carbohydrate, 42% fat, 15.3% protein, 18.8 kJ/g; TD88137; Harlan Teklad, Madison, WI, USA) or a liquid diet (47% carbohydrate, 35% fat, 18% protein, 4.186 kJ/ml; Lieber–De Carli Control diet; Bioserv Industries, Frenchtown, NJ, USA). Food intake was measured daily. The use of the liquid diet allowed the constitution of an additional group of pair-fed control C3H/HeOuJ or C57Bl/6J mice (pair-fed group), which were fed the same quantity of the diet as that consumed on the previous day by the C3H/HeJ or Cd14 −/− mice. In addition, ten mice of C3H genotype were fed a standard chow diet (60% carbohydrate, 12% fat, 28% protein, 14.2 kJ/g; A03; UAR-SAFE, Epinay sur Orge, France). Body weights were measured weekly. Feeding efficiency was calculated by dividing body weight gain by energy intake over the diet period.

After a 6-h fast, mice were killed under anaesthesia with isoflurane inhalation. After blood collection from the retro-orbital sinus, plasma samples were stored at −20°C. Adipose tissue and liver were collected and weighed. In some experiments, mice were injected with LPS from E. coli serotype 026:B6 (100 mg/kg i.p; Sigma, St Louis, MO, USA). This study protocol was approved by the local ethics committee of the Faculté de Médecine of Marseille (authorisation number 52–23).

Biochemical and immunological analyses

Plasma triacylglycerol and glucose were determined by routine chemical methods. Insulin levels of fasting mice were measured by radioimmunoassay (Linco Research, St Louis, MO, USA). Plasma ghrelin levels were assayed with a rat ghrelin enzyme immunoassay kit (Phoenix Peptide, Belmont, CA, USA). Leptin, TNF, TNF receptor superfamily, member 1b (TNFRSF1B, previously known as TNF receptor type 2 [TNFR2]) and IL6 were measured with ELISA (R&D systems, Abingdon, UK). Murine plasminogen activator inhibitor-1 (PAI-1, now known as serine [or cysteine] peptidase inhibitor, clade E, member 1 [SERPINE1]) was assessed with a home-made ELISA [21].

Tissue lipid analysis

Lipids from white adipose tissue (WAT) and liver were extracted by the method of Folch et al. [22]. Triacylglycerol was measured using a triacylglycerol and free glycerol reagent (Sigma).

Total DNA content from adipose tissue

WAT was incubated overnight in the presence of proteinase K (100 μg/ml). DNA was then purified using phenol/chloroform and quantified by measuring the absorbance at 260 nm.

Histological studies

Pieces of WAT and liver were embedded into Tissue-Tek OCT (Miles, Elkhart, IL, USA). Cryosections were stained with Nuclear Fast Red. Images of adipocytes were acquired from a Leica DMRB light microscope fitted with a colour CCD camera (Coolsnap; Princeton Instruments, Evry, France). About 300 cell surfaces per fat pad were measured at ×10 magnification using Scion imaging software (Scion Image, Frederick, MD, USA). Liver sections were stained with Oil Red O and observed under the same microscope at ×20 magnification.

Isolation and culture of peritoneal macrophages

Peritoneal macrophages were prepared as described [13] and then incubated for 18 h with or without LPS from E. coli serotype 026:B6 (50 ng/ml, Sigma). Media were collected and stored at −80°C. Total cell proteins were assayed according to specifications of the bicinchoninic acid protein assay kit from Sigma.

In vivo lipolysis

After a 6-h fast, five mice were injected intraperitoneally with (−)-isoproterenol hydrochloride (10 mg/kg; Sigma). Blood was collected before and 15 min after injection. Plasma NEFA concentrations were measured using NEFA C kit (Wako, Neuss, Germany). These mice were not used for further experiments.

Glucose turnover studies and in vivo glucose utilisation index in individual tissues

These experiments were performed as described [23, 24]. For details, see Electronic supplementary material (ESM).

In vitro glucose uptake

Primary adipocytes (10 days post-differentiation) [25] were cultured overnight in serum-free DMEM with low glucose (1 g/l). After Krebs–Ringer phosphate buffer (KRP) wash, cells were incubated at 37°C for 30 min in the presence or absence of 100 nmol/l of insulin in KRP containing 1% BSA. Then 18.5 kBq per well 2-deoxy[3H]glucose (Amersham Biosciences, Orsay, France) was added and incubated for 10 min at 37°C. The cells were washed with cold KRP and solubilised in 0.1% SDS. The radioactivity of a 200-μl aliquot was determined in a scintillation counter. Glucose uptake was expressed as the degree of increase relative to that of basal C3H/HeOuJ cells.

Real-time quantitative PCR

Total RNAs were prepared using Trizol reagent (Invitrogen, Cergy Pontoise, France). cDNA was synthesised using MMLV transcriptase (Invitrogen) from 1 μg of total RNA and stored at −80°C until use. PCRs were performed under standard conditions [26]. The relative amounts of each studied mRNA were normalised to acidic ribosomal phosphoprotein P0 (Arbp, also known as 36b4) mRNA levels and expressed relatively to the mean value of the control mice group (User Bulletin no. 2; Applied Biosystems, Foster City, CA, USA). Primer express software (Applied Biosystems) was used to design primers (available upon request from tanti@unice.fr).

Data and statistics

Data are expressed as mean±SEM. Differences between independent variables were tested using the Mann–Whitney test. Differences between paired variables were tested using the paired Wilcoxon test. The Kolmogorov–Smirnov Z test was used to test differences in adipocyte surface. A p value < 0.05 was considered significant.

Results

C3H/HeJ mice do not respond to LPS

To confirm the absence of functional TLR4 in C3H/HeJ mice, we studied the LPS responsiveness of their macrophages in vitro and whole-body LPS responsiveness in vivo. Comparable numbers of peritoneal macrophages were retrieved from the two genotypes and incubated for 18 h with or without LPS. Under basal conditions, the levels of IL6, PAI-1 and TNFRSF1B were strongly reduced in TLR4-deficient macrophages, whereas levels of TNF were similar (Table 1). After LPS stimulation all levels were significantly increased in C3H/HeOuJ macrophages, whereas they were not affected in C3H/HeJ macrophages.

Table 1 Inflammatory marker production by peritoneal macrophages from C3H/HeOuJ and C3H/HeJ mice incubated with or without LPS stimulation for 18 h

Plasma TNF levels in C3H/HeOuJ mice increased at 1.5 h after LPS injection to 771 ± 335 pg/ml (n = 6), but remained undetectable in C3H/HeJ mice (n = 5).

C3H/HeJ mice exhibit a higher feeding efficiency than control mice

Consistent with previous reports [27], body mass was significantly lower in C3H/HeJ than in C3H/HeOuJ control mice after 10 weeks of standard chow diet (30.1 ± 0.5 vs 33.0 ± 1.0 g; p < 0.01). However, weight gain was large in the C3H/HeJ and C3H/HeOuJ control mice that had free access to the high-fat diet (Fig. 1a). The weight of C3H/HeJ mice changed similarly to that of the control mice and body weights after 22 weeks of high-fat diet were not statistically different between the two groups (Table 2). While the weight of subcutaneous fat pads did not differ between the groups, C3H/HeJ mice had larger epididymal fat pads (Fig. 1b). In C3H/HeJ mice the total triacylglycerol content of epididymal tissue was increased by 15%, whereas the total DNA content was reduced by 23%. This led to a significantly higher triacylglycerol:DNA ratio in C3H/HeJ than in C3H/HeOuJ (fold increase: 3.1; n = 10; p < 0.05), indicating an increase in adipocyte size in C3H/HeJ mice. This was confirmed by histological analysis showing an increase in adipocyte surface in both epididymal (C3H/HeJ vs C3H/HeOuJ: 1,749 ± 77 vs 1,432 ± 44 μm2; p < 0.01) and subcutaneous (C3H/HeJ vs C3H/HeOuJ:1,152 ± 31 vs 996 ± 37 μm2; p < 0.01) tissues. Both groups developed hyperglycaemia, hyperinsulinaemia and hypertriacylglycerolaemia (Table 2), but showed no difference in leptin or ghrelin levels (Table 2).

Fig. 1
figure 1

a Growth curves of C3H/HeJ (open circles) and C3H/HeOuJ pair-fed mice (closed squares) given free access to a pelleted high-fat diet for 22 weeks. Data are mean±SEM of 20 mice. b Epididymal and subcutaneous adipose tissue masses of C3H/HeJ (white bars) and C3H/HeOuJ (black bars) mice given free access to the same diet for the same period. Data are mean±SEM of ten mice. *p = 0.01 c Growth curves of C3H/HeJ (open circles), C3H/HeOuJ pair-fed (closed circles) and C3H/HeOuJ (closed squares) mice. Mice had free access to a liquid high-fat diet for 10 weeks. Data are mean±SEM of ten mice

Table 2 Characteristics of C3H/HeOuJ and C3H/HeJ mice with free access to a pelleted high-fat diet for 22 weeks

To ensure that food intake did not differ between the groups, food consumption was summed over the diet period (Table 2). Unexpectedly, although the C3H/HeJ reached the same body weight, they consumed less energy than C3H/HeOuJ control mice. As a result, feeding efficiency was increased in mutant mice (Table 2).

C3H/HeJ mice exhibit a higher body weight than pair-fed C3H/HeOuJ control mice

The difference in feeding efficiency between the groups prompted us to perform a pair-feeding experiment, in which C3H/HeOuJ were fed the same quantity of the diet C3H/HeJ had consumed on the previous day (pair-fed). To make the pairing as precise as possible, we used a high-fat liquid diet. The pair-fed C3H/HeOuJ mice gained significantly less weight than C3H/HeJ mice (Fig. 1c), leading to a 12.4-fold greater feeding efficiency in C3H/HeJ (Table 3). The huge differences in weight gain were associated with alterations in adiposity. The epididymal and the subcutaneous fat pad masses in C3H/HeJ mice were markedly larger than in C3H/HeOuJ pair-fed mice (Table 3). As previously shown, C3H/HeJ with free access to food reached the same body weight as C3H/HeOuJ (Fig. 1c), but again C3H/HeJ consumed less energy and had significantly greater feeding efficiency than C3H/HeOuJ mice (Table 3). Hence, non-functional TLR4 promoted weight gain despite the lower food intake than in control mice. Taken together, these data suggest that TLR4 plays a role in regulating the equilibrium between food intake and energy expenditure. We then studied key regulators of weight gain. To avoid any bias due to body weight difference, we investigated mice that had had free access to food during 22 weeks and whose body weights were similar.

Table 3 Fat distribution and food intake of C3H/HeOuJ and C3H/HeJ mice with free access to a liquid high-fat diet for 10 weeks

TLR4 inactivation does not increase whole-body insulin sensitivity after 22 weeks of high-fat diet

To understand better why mice with non-functional TLR4 gained identical weight while ingesting less food, we investigated whether this was due to modifications of insulin sensitivity. After 22 weeks of high-fat diet, C3H/HeJ mice had similar glycaemia but lower insulinaemia after fasting (Table 2) than C3H/HeOuJ mice, although glycaemia was still higher than in C3H/HeJ mice fed normal chow (41.8 ± 5.2 pmol/l; n = 10). However, the glucose infusion rate needed to maintain stable glycaemia during the hyperinsulinaemic–euglycaemic clamps was not different, indicating that whole-body glucose uptake did not differ between the two genotypes (Fig. 2a).

Fig. 2
figure 2

Glucose metabolism assessment in C3H/HeOuJ and C3H/HeJ mice studied after 22 weeks of a pelleted high-fat diet. a Glucose infusion rate during a euglycaemic–hyperinsulinaemic clamp. b In vivo glucose clearance rate by subcutaneous adipose tissue and (c) vastus lateralis. Black bars, C3H/HeOuJ; white bars, C3H/HeJ. Data are mean±SEM of five mice. *p < 0.05. d Effects of TLR4 activation by LPS treatment on glucose uptake in C3H/HeOuJ and C3H/HeJ primary adipocytes (day 10 of differentiation) in response to insulin stimulation. LPS was added 4 h prior to insulin. Data are mean±SEM of three independent experiments. *p < 0.05

TLR4 inactivation supports higher insulin sensitivity and lipogenesis in WAT after 22 weeks of high-fat diet but not in skeletal muscle

We hypothesised that TLR4 inactivation could affect insulin sensitivity of insulin target tissues in different ways. Indeed, it has recently been demonstrated that inflammation is regulated differently in adipose tissue than in muscle [28], as is insulin sensitivity [29]. Insulin-stimulated glucose utilisation was significantly higher in subcutaneous adipose tissue of C3H/HeJ than in that of C3H/HeOuJ mice (Fig. 2b). To further investigate the effects of TLR4 signalling on insulin action, we evaluated whether TLR4 activation by LPS could affect insulin-dependent glucose uptake by adipocytes. While insulin increased glucose uptake in adipocytes from both genotypes, LPS significantly decreased the ability of insulin to stimulate glucose uptake, this inhibitory effect being less pronounced in C3H/HeJ adipocytes (Fig. 2d). In contrast, the insulin-stimulated glucose utilisation rate in vastus lateralis was lower in C3H/HeJ mice than in C3H/HeOuJ mice (Fig. 2c). A trend towards a decrease, albeit not significant, was also observed in tibialis anterior (data not shown). Hence, these results demonstrate that obese TLR4-deficient mice have an adipose tissue-specific increase in insulin sensitivity.

Non-functional TLR4 does not modify isoproterenol-induced lipolysis after 22 weeks of high-fat diet

Inflammation is known to increase adipocyte lipolysis [30]. We thus examined whether C3H/HeJ mice exhibited a lower lipolytic potential. Fasting plasma NEFA levels were not different between both strains (C3H/HeOuJ: 0.79 ± 0.04 vs C3H/HeJ: 0.82 ± 0.08 mmol/l). Isoproterenol administration led to a similar increase in plasma NEFA concentrations in the two groups (C3H/HeOuJ: 1.72 ± 0.07 vs C3H/HeJ: 1.70 ± 0.09 mmol/l).

TLR4 inactivation supports a lower adipose tissue inflammatory state and attenuates obesity-induced changes in adipose tissue gene expression

Increased expression of genes that encode proinflammatory proteins has been observed in adipose tissue of obese animals [31]. In subcutaneous adipose tissue from C3H/HeJ mice, we observed a decrease in Tnf and Il1b mRNA levels (Fig. 3a). We also observed an increase in the mRNA level of interleukin 1 receptor antagonist (Il1rn, previously known as Il1ra) (Fig. 3a), recently shown to regulate adipogenesis, food intake and energy expenditure [32]. A similar tendency was noticed in epididymal adipose tissue (Fig. 3d). These differences were not due to a reduction in the number of tissue-infiltrated macrophages, since the constitutive macrophage marker, EGF-like module containing, mucin-like, hormone receptor-like sequence 1 (Emr1, previously known as F4/80), exhibited a higher expression level in C3H/HeJ than in C3H/HeOuJ subcutaneous WAT (Fig. 3a).

Fig. 3
figure 3

Expression of genes encoding proteins involved in inflammation (a, d), insulin signalling (b, e) and lipogenesis (c, f) in subcutaneous (ac) and epididymal (df) adipose tissue from C3H/HeJ (white bars) and C3H/HeOuJ (black bars) mice given free access to a pelleted high-fat diet for 22 weeks. mRNA levels were normalised by those of Arbp mRNA. Data are mean±SEM (n = 810 mice). *p < 0.05, **p < 0.02, ***p < 0.01

We then studied whether these changes in the inflammatory pattern of CH3/HeJ adipose tissue were associated with higher mRNA levels of genes involved in insulin action and sensitivity, glucose and triacylglycerol metabolism, and adipocyte differentiation. Adipose tissue of obese TLR4 mutant mice expressed higher levels of Irs1, adiponectin and solute carrier family 2 (facilitated glucose transporter), member 4 (Slc2a4, previously known as Glut4) mRNA (Fig. 3b,e). The mRNA levels of two key lipogenic factors, sterol regulatory element binding factor 1 (Srebf1, previously known as Srebp1c) and fatty acid synthase (Fasn, previously known as Fas), were also higher (Fig. 3c,f). Further, peroxisome proliferator-activated receptor γ (Pparg) mRNA followed a similar pattern (Fig. 3c,f). These data suggest that the pro-inflammatory changes that occur in adipose tissue during the development of obesity are in part TLR4-dependent. The lower inflammatory state observed in C3H/HeJ adipose tissue may support the possibility of a higher insulin sensitivity.

TLR4 inactivation attenuates high-fat diet-induced liver steatosis

To determine whether improved insulin sensitivity in adipose tissue of C3H/HeJ mice ameliorated hepatomegaly or reduced hepatic steatosis, we analysed the livers from C3H/HeJ and C3H/HeOuJ mice. Despite the presence of a similar degree of obesity, the liver mass of C3H/HeJ was lower than that of control mice (Table 4). The reduced liver size was associated with a 40% lower hepatic triacylglycerol content in C3H/HeJ than in C3H/HeOuJ mice, concomitant with lower hepatic levels of Pparg and Fasn mRNA (Fig. 4a,b) and lower histological steatosis (Fig. 4c). Similar results were obtained after 10 weeks of liquid high-fat diet (Table 4). The liver mass of C3H/HeJ fed a liquid high-fat diet was lower than that of control mice and did not differ from that of control mice fed a standard chow diet (1.21 ± 0.06 g; n = 10).

Table 4 Effect of TLR4 deficiency on liver weight
Fig. 4
figure 4

a Levels of liver triacylglycerol content (n = 10) and expression (b) of lipogenic and fibrotic markers in liver from C3H/HeOuJ (black bars) and C3H/HeJ (white bars) mice given free access to a pelleted high-fat diet for 22 weeks (n = 5). mRNA levels were normalised by those of Arbp mRNA. Data are mean±SEM. *p < 0.05, ** p < 0.02, ***p < 0.01. c Representative histological results of Oil Red O staining from C3H/HeOuJ (c) and C3H/HeJ (d) mice fed as above for 22 weeks

Significant fibrosis is associated with non-alcoholic steatohepatitis and has been considered a factor in prognosis [33]. We thus investigated mRNA levels of Tgfb1 and Pai-1, key effectors of the activity of TGFB1. Both Pai-1 and Tgfb1 mRNA levels were lower in the liver of C3H/HeJ mice than in that of C3H/HeOuJ (Fig. 4b). These data argue for an effect of the TLR4 pathway on the evolution of fibrosis associated with fatty liver.

Cd14−/− mice develop a phenotype close to that of C3H/HeJ mice

To confirm the contribution of the LPS pathway to the observed results, we completed our study by phenotyping Cd14 −/− mice and their controls after 10 weeks of free access to a high-fat diet. High-fat feeding for 10 weeks led to a significant weight gain in both Cd14 −/− and control mice. Similarly to C3H/HeJ mice, feeding efficiency was higher in Cd14 −/− mice than in controls given free access to food (data not shown). In subcutaneous adipose tissue, Cd14 inactivation was associated with 5- to 25-fold higher mRNA levels of genes involved in insulin action (Slc2a4, Irs1, adiponectin) and lipogenesis (Pparg, Fasn) (ESM Fig. 1). Despite the same liver mass (C57B1/6J: 4.3 ± 0.1, Cd14 −/−: 4.5 ± 0.4% of body weight), the hepatic triacylglycerol content tended to be higher in C57Bl/6J than in Cd14 −/− mice (176 ± 13 vs 139 ± 11 mmol/g; p = 0.08).

Discussion

Our data demonstrate that TLR4 exerts an important role on obesity-associated phenotypes. When mice were fed a high-fat diet, absence of TLR4 modulated the feeding behaviour and increased WAT insulin sensitivity and lipogenesis. Importantly, TLR4 mutant mice were protected from developing hepatic steatosis.

We found that TLR4 inactivation improved insulin-induced glucose uptake in WAT. This protection from high-fat diet-induced adipose tissue insulin resistance was associated with a failure of the high-fat diet to induce expression of inflammatory mediators in WAT. This finding was recently supported by a report that NEFA stimulate proinflammatory pathways and reduce insulin sensitivity in WAT through TLR4 activation [34]. The attenuated effect of LPS on insulin-induced glucose uptake in mutant adipocytes may reflect a direct contribution of TLR4 present in adipocytes. Accordingly, TLR4 activation by LPS in 3T3-L1 adipocytes prevented insulin-dependent glucose uptake [35].

Our data may also result from the inhibition of TLR4-mediated inflammatory pathways in macrophages, with a resultant decrease in stimulatory action on adipocytes. In support of this hypothesis, TLR4 protein levels were much lower in adipocytes than in macrophages [36]. Remarkably, the attenuation of inflammation was not accompanied by a reduction of macrophage infiltration into WAT. Indeed, we did not find a decrease in Emr1 mRNA levels in WAT of mutant mice. Consistent with this, Suganami et al. [36] showed that co-culture of 3T3-L1 adipocytes and C3H/HeJ macrophages was associated with an attenuation of Tnf and Il1B mRNA expression, but also with a significant increase in mRNA expression levels of the powerful monocyte chemoattractant, monocyte chemoattractant protein 1.

TNF is known to reduce adipose tissue mass by decreasing PPARG production, stimulating lipolysis and repressing genes involved in the uptake and metabolism of lipids and glucose [3741]. Considering all this, the higher insulin sensitivity that we observed in our study in WAT and that was associated with higher expression of genes mediating insulin sensitivity (Irs1, Slc2a4, adiponectin) and lipogenesis (Srebf1, Fasn, Pparg) may be secondary to the tissue TNF and IL1B depletion, and might contribute to the increase in adipocyte volume seen in C3H/HeJ mice despite lower food intake. This notion is consistent with our finding that insulin sensitivity was increased in WAT from C3H/HeJ mice during euglycaemic–hyperinsulinaemic clamp conditions. Greater expression of Srebf1 and Fasn could be consecutive to the increased insulin sensitivity of the adipocytes, since insulin regulates the expression of both.

IL1RN is an anti-inflammatory cytokine [42]. Human and rodent WAT is a major source of IL1RN [43, 44]. Il1rn / mice, in which excess IL1 signalling may be induced, show a lean phenotype [45]. This phenotype has been associated with a defect in adipogenesis, a reduction of genes involved in lipogenesis, an increase in energy expenditure and a reduction in diurnal food intake [32]. The increase in IL1RN production in WAT may be directly involved in the phenotype obtained in TLR4 mutant mice, which exhibited both increased peripheral lipogenesis and lower thermogenesis. Note that in our study, Il1rn expression was increased in a situation where inflammation was dampened, suggesting that IL1RN may not only represent a counter-regulatory mechanism that limits inflammation, but may also be produced independently of an inflammatory process during fat accumulation.

We observed contrasting behaviour with regard to insulin sensitivity in skeletal muscle and adipose tissue. This could be related to the low inflammation level in skeletal muscle during obesity [28]. The skeletal muscle produces low levels of CD68 and monocyte chemoattractant protein 1 and neither pioglitazone nor metformin has any effect on production of these two [46]. Thus, a modification of peripheral glucose disposal under inflammation manipulation is more likely to be due to changes in adipose tissue and then to secondary effects on muscle. Another plausible explanation is that the ability of TLR4 deficiency to improve insulin resistance in adipose tissue and not muscle is regulated through PPARG , whose levels are up to 50-fold higher in adipose tissues than in skeletal muscles [47]. IL1RN may be involved in the described phenotype by virtue of its ability to reduce insulin sensitivity in muscle through a specific decrease in glucose uptake as recently described in rats [48]. Alternatively, it remains possible that plasma factor(s) dominantly affect insulin signalling in muscles by mechanisms that are independent of TLR4.

We found a reduction in total brown adipose tissue fat mass in C3H/HeJ associated with decreased Ucp1 expression. This finding shows that TLR4 inactivation may play a role in brown adipose tissue development.

Our data provide evidence that TLR4 deficiency slows down non-alcoholic liver steatosis. This result should be seen together with those obtained in alcohol-induced steatosis. C3H/HeJ mice fed ethanol do not develop fatty liver but C3H/HeOuJ do. Chronic alcohol administration increases gut-derived endotoxin in the portal blood, partly due to gut leakiness [49], which activates Kupffer cells and causes liver injury. Interestingly, we found that the protective effect of TLR4 inactivation was maintained throughout the diet period. Even after 22 weeks of high-fat diet the expression levels of key lipogenic markers still differed. Tgfb1 and Pai-1 expression levels in liver were decreased in CH3/HeJ mice, arguing for an effect of the TLR4 pathway, known as a key determinant of non-alcoholic steatohepatitis [50], on the evolution of hepatic fibrosis. Several studies have shown an inverse relationship between hepatic triacylglycerol stores and hepatic insulin sensitivity [11]. The observation that TLR4 mutant animals have a much lower insulin level at similar plasma glucose concentrations suggests that the liver responds more sensitively to insulin-induced inhibition of glucose output.

One limitation of our study could be the use of non-littermate controls. However, we obtained similar findings in Cd14-deficient mice, demonstrating that two approaches to blunting the LPS pathway lead to the same modifications: (1) in feeding behaviour; (2) in adipose tissue expression levels of genes involved in glucidolipidic metabolism; and (3) in the development of liver steatosis.

Our results indicate that, apart from an acute LPS challenge, TLR4 contributes to the production of key cytokines in WAT during obesity. They also highlight the contribution of the TLR4 pathway to the control of obesity and its related complications. We propose that interventions known to reduce endotoxin toxicity [51] or to modulate the TLR4 pathway may help control some key elements of the metabolic syndrome.