All experiments were approved by the ethics committee at the State University of Campinas. Eight-week-old male Wistar rats were randomly divided into groups: control (C), fed standard rodent chow and water ad libitum; DIO-sedentary rats (DIO), fed a high-fat diet, as previously used (⁹), and water ad libitum for 20 weeks; DIO-chronic exercised rats (DIO+CE) and DIO-acute exercised rats (DIO+AE), fed a high-fat diet. Insulin and glucose tolerance tests were performed on these rats after 20 weeks of consumption of the diets and after both the exercise protocols, as described previously (¹⁷,¹⁸). Because individual rats can vary in their ability to perform swimming, some rats (10–15%) were rejected in the screening for the procedure.

After an overnight fast (∼12 h), a 2-h hyperinsulinemic-euglycemic clamp was conducted in anesthetized catheterized rats with [3-³H]glucose and 2-deoxy-d-[1-¹⁴C]glucose to assess glucose metabolism in muscle, as previously described (¹⁸–²⁰).

Insulin and interleukin (IL)-6 concentrations were determined by an ELISA (Linco, St. Charles, MO). Serum free fatty acid (FFA) levels were analyzed using NEFA-kit-U (Wako Chemical, Neuss, Germany) with oleic acid as a standard. Glucose was measured from whole venous blood with a glucose monitor (Glucometer; Bayer Diagnostics, New York, NY).

Rats were adapted to swimming for 10 min for 2 days to reduce water-induced stress. Animals swam in groups of three in plastic barrels of 45 cm in diameter that were filled to a depth of 60 cm, and the water temperature was maintained at ∼34°C. The training consisted of daily swimming sessions (1 h/day, 5 days/week, for 8 weeks) with a progressive load increase up to 5% of body weight. These conditions were chosen on the basis of previous studies showing that swimming training with this load improved the physical condition of rats (²¹). The animals were killed with an overdose of anesthetic (sodium thiopental) at 24 and 36 h after the last session.

Under the same conditions imposed as chronic exercise, the acute exercise protocol consisted of two 3-h bouts, separated by a 45-min rest period, as described previously (¹⁸). The animals were killed at 2 and 16 h after this protocol was carried out.

After exercise protocols, rats were anesthetized and used 10–15 min later, i.e., as soon as anesthesia was ensured by the loss of pedal and corneal reflexes, the abdominal cavity was opened, the portal vein was exposed, and 0.2 mL of normal saline was injected with or without insulin (10⁻⁶ mol/L). At 30 and 90 s after insulin injection, liver and gastrocnemius and adipose tissue were removed, minced coarsely, and homogenized immediately in extraction buffer, as previously described (²²). Extracts were used for immunoprecipitation with MyD88 and Protein A-Sepharose 6MB (Pharmacia, Uppsala, Sweden). The precipitated proteins or whole-tissue extracts were subjected to SDS-PAGE and immunoblotting, as previously described (¹⁷,²⁰).

We diluted plasma samples to 20% with endotoxin-free water and then heated them to 70°C for 10 min to inactivate plasma proteins. We then quantified serum LPS with a commercially available Limulus Amebocyte assay from Cambrex (Walkersville, MD) according to the manufacturer’s protocol. We ran samples in duplicate and subtracted the background (²³).

The mRNA was determined in the muscle, liver, and adipose tissue using RT-PCR, as previously reported (²⁴). Primer sequences are shown in Supplementary Table 1. Results are expressed as relative expression values, as published previously (¹).

Data are expressed as means ± SEM, and the number of independent experiments is indicated. The results of blots are presented as direct comparisons of bands or spots in autoradiographs and quantified by optical densitometry (Scion Image; Scion Corporation, Frederick, MD). For statistical analysis, the groups were compared using a two-way ANOVA with the Bonferroni test for post hoc comparisons. The level of significance adopted was P < 0.05.

Figure 1 shows comparative data regarding the controls, DIO-sedentary rats, and DIO rats submitted to chronic exercise when investigated at 24 h (DIO+CE24h) or 36 h (DIO+CE36h) after the last training session. All DIO animals, submitted to chronic exercise or not, exhibited a higher body weight and epididymal fat than the age-matched C group (Fig. 1A and B). The fasting plasma glucose concentrations were similar among all groups (Fig. 1C). During the glucose tolerance test, glucose and insulin levels were higher in DIO rats, compared with controls, and chronic exercise improved glucose tolerance and reduced insulin levels at all time points (Fig. 1C and D). Serum insulin levels were higher in DIO rats, and chronic exercise was able to reduce this hyperinsulinemia (Fig. 1D). The glucose disappearance rate was lower in the DIO group, and exercise training reversed these alterations (Fig. 1E). A hyperinsulinemic-euglycemic clamp with tracer infusions was also performed to examine the effects of training on glucose metabolism in the skeletal muscle. The glucose infusion rate was lower in the DIO group than in the C group and reestablished to control levels in DIO-exercised rats (Fig. 1F). As shown in Fig. 1G, DIO rats presented a significant reduction in glucose uptake in the skeletal muscle when compared with the control group. In contrast, chronic exercise improved insulin-induced glucose uptake in the muscle of DIO rats (Fig. 1G). Serum levels of FFA were also higher in DIO rats and significantly decreased after swimming training (Fig. 1H).

The TLR4 protein content in muscle, liver, and adipose tissue was higher in the DIO group than in the controls (Fig. 2A–C). Results show that, at 24 h after the last training, there was a marked decrease in TLR4 expression and, after 36 h, this was still decreased when compared with nonexercised obese rats (Fig. 2A–C). Furthermore, the TLR4 mRNA expression was significantly reduced after chronic exercise in muscle, liver, and adipose tissue, compared with DIO rats (Fig. 2D–F).

We next investigated TLR4 pathway activation in two steps: 1) as an early event, TLR4/MyD88 interaction was examined; and 2) for downstream signaling, JNK, IKKβ, and ERK1/2 phosphorylation were studied (²⁵). Skeletal muscle, adipose tissue, and liver of DIO rats exhibited significant increases in the TLR4/MyD88 interaction, compared with the C group (Fig. 2G–I). Conversely, in exercised groups, the TLR4/MyD88 interaction decreased significantly in all tissues studied compared with obese sedentary rats (Fig. 2G–I). IRAK-1, another protein of the TLR4 pathway, showed an increased expression in DIO animals compared with the C group. Conversely, chronic exercise did not have any effect on this protein (Fig. 2J–L).

We also investigated whether this reduction was reflected in IKKβ and JNK phosphorylation, which are downstream of TLR4. As expected, an increase in IKKβ and JNK phosphorylation was found in the muscle, liver, and adipose tissue of DIO rats, compared with controls, and this effect was attenuated in the tissues of chronic-exercised obese rats, compared with sedentary DIO rats (Fig. 3A–F). With regard to ERK1/2, DIO rats exhibited high phosphorylation levels in the three tissues analyzed, compared with control animals (Supplementary Fig. 1). In contrast, we detected a marked reduction in ERK activation after chronic exercise (Supplementary Fig. 1).

We then evaluated the effect of exercise training on an important substrate of these kinases of insulin signaling pathway, namely, IRS-1 Ser³⁰⁷ phosphorylation. This phosphorylation was, on average, markedly upregulated in the muscle, liver, and adipose tissue of obese sedentary rats, compared with controls (Fig. 3G–I). In addition, exercise training was also able to reverse diet-induced IRS-1 Ser³⁰⁷ phosphorylation in the muscle, liver, and adipose tissue of DIO+CE24h and DIO+CE36h rats (Fig. 3G–I).

We then examined the consequences that this exercise-induced improved inflammatory profile exerts on the insulin signaling pathway. In the muscle, liver, and adipose tissue, insulin-induced IRβ, IRS-1 tyrosine phosphorylation, and Akt phosphorylation were reduced by 50–80% in DIO rats, compared with the C group (Fig. 4A–I). On the other hand, insulin-induced IRβ, IRS-1, and Akt phosphorylation were increased in the tissues of DIO+CE24h and DIO+CE36h rats, compared with obese sedentary rats, and approached levels of those found in the C group (Fig. 4A–I). No changes in basal phosphorylation or tissue protein levels were observed among groups (Fig. 4A–I).

Figure 5 shows comparative data for the DIO rats and DIO rats submitted to the acute exercise protocol (DIO+AE2h and DIO+AE16h). No significant variation was found in body weight, epididymal fat, and fasting blood glucose in DIO rats after a single session of exercise, when compared with sedentary obese rats (Fig. 5A–C). Acute exercise improved glucose tolerance and reduced serum insulin levels at all time points after the glucose load (Fig. 5D and E). The glucose disappearance rate was restored at both 2 h and 16 h after acute exercise (Fig. 5E). Acute exercise was also capable of restoring the glucose infusion rate during the hyperinsulinemic-euglycemic clamp (Fig. 5F), accompanied by a significant increase in glucose uptake in the skeletal muscle, compared with DIO rats (Fig. 5G). As expected, serum levels of FFA and IL-6 were higher in obese rats, and acute exercise induced a marked increase in the levels of this substrate, mainly in the DIO+AE2h group (Fig. 5H and I).

In DIO animals submitted to acute exercise, no changes in TLR4 protein expression were observed in muscle, liver, and adipose tissue (data not shown). On the other hand, the acute exercised animals showed lower TLR4 mRNA only in muscle (Supplementary Fig. 2A–C). In contrast, we observed significant reductions in the TLR4/MyD88 interaction in all studied tissues (Fig. 6A–C).

We also investigated the effect of acute exercise on TLR4 downstream signaling. As expected, significant decreases in IKKβ and JNK phosphorylation were found in the muscle, liver, and adipose tissue of acute exercised rats, compared with the DIO group (Fig. 6D–I). The same behavior was found for ERK1/2 phosphorylation after the acute exercise (Supplementary Fig. 1). Acute exercise was also able to reverse the diet-induced IRS-1 Ser³⁰⁷ phosphorylation in muscle, liver, and adipose tissue of both DIO+AE2h and DIO+AE16h rats (Fig. 6J–L).

Acute exercise, as observed in the chronic group, improved insulin-induced tyrosine phosphorylation of IRβ and IRS-1 in the insulin-sensitive tissues (Fig. 7A–F). With regard to Akt, the acute protocol stimulated an impressive increase in serine phosphorylation of this protein in all tissues studied that was similar to that observed in the C group (Fig. 7G–I).

The tumor necrosis factor (TNF)-α and IL-6 mRNA levels were upregulated in the DIO group and were decreased after both exercise protocols in muscle, liver, and adipose tissue of almost all exercised groups (Supplementary Fig. 1D–I).

Serum LPS levels were significantly higher in obese sedentary rats when compared with control animals. Conversely, in animals of the DIO+AE2h group, acute exercise was able to completely reverse the high levels of LPS, and this was still observed at 16 h after the acute exercise (Fig. 8A). Notably, in chronic-exercised animals there was also a reduction in serum LPS levels, compared with the DIO group (Fig. 8A).

To examine whether the exercised-induced reduction in serum LPS plays an important role in the improvement of inflammatory status and insulin sensitivity, we infused LPS into the peritoneum of obese exercised rats immediately after acute exercise. Exercised animals treated with LPS showed serum levels of LPS similar to the DIO animals and presented significant increases in TLR4/MyD88 interaction, IKKβ, and JNK phosphorylation in the muscle and adipose tissue (DIO+AE2h and DIO+AE16h) (Fig. 8A–G), increase in TNF-α mRNA in muscle and adipose tissue, and increase in IL-6 mRNA only in muscle (Supplementary Fig. 2D–I). A significant decrease in Akt serine phosphorylation levels was also achieved in LPS-treated animals (Fig. 8H and I).

We next investigated whether an inhibitor of TLR4 signaling (TAK-242) could mimic the effect of exercise on the inflammatory pathway and insulin signaling in DIO rats. The administration of TAK-242 for 3 days improved inflammatory status in the muscle and adipose tissue of DIO rats, as measured by the reduction of TLR4/MyD88 interaction and phosphorylation of IKKβ and JNK, in parallel with an improvement in insulin-induced Akt phosphorylation (Supplementary Fig. 3A). In DIO+AE2h animals treated with this drug, no additive effect was observed in inflammatory parameters or insulin signaling (Supplementary Fig. 3A). In a similar fashion, C3H/HeJ mice fed a high-fat diet did not present insulin resistance, and we did not observe any increase in IKKβ and JNK phosphorylation; furthermore, exercise had no effect on these parameters or on insulin signaling (Supplementary Fig. 3D–F).