Interleukin-6 Regulation of AMP-Activated Protein Kinase: Potential Role in the Systemic Response to Exercise and Prevention of the Metabolic Syndrome

AMPK is a heterotrimer composed of a catalytic α-subunit and regulatory β- and γ-subunits. In humans, each subunit is encoded by either two or three distinct genes (α1, α2, β1, β2, γ1, γ2, γ3), so that there are 12 possible α, β, and γ combinations. All three subunits are necessary for full activity (9–11). AMPK is activated by decreases in the energy state of a cell as reflected by an increase in the AMP-to-ATP ratio. According to the most widely held view, increased binding of AMP to specific domains on the γ-subunit produces a conformational change that allows upstream kinases (AMPK kinases) to phosphorylate the threonine-172 residue of the α-subunit and activate the enzyme (10,11). Two AMPK kinases have recently been identified: LKB1, a tumor suppressor that is deficient in patients with the Peutz-Jegher syndrome (12,13), and a Ca²⁺-dependent calmodulin-dependent protein kinase kinase, CAMKK (14–16). Numerous stimuli have been shown to activate AMPK (Fig. 1), including exercise, glucose deprivation, and hypoxia; various hormones including leptin, adiponectin, and catecholamines; and drugs such as metformin and thiazolidinediones (11,17). Likewise, a lack of leptin or its receptor, a surfeit of glucose, and the hormone ghrelin decrease AMPK activity in peripheral tissues (11). In many instances, it is unclear whether these hormones and drugs act by altering the energy state of the cell, by altering the functional activity of one of the AMPK kinases, or by an as yet undescribed mechanism.

Once it is activated, AMPK helps to restore the cell’s energy state by acutely and chronically enhancing processes that generate ATP, such as fatty acid oxidation, and inhibiting others that consume ATP, but are not acutely necessary for survival (Fig. 1). The latter include protein, fatty acid, and triglyceride synthesis (9,10,17).

To illustrate how AMPK functions, a schema depicting some of its many effects on cellular fatty acid partitioning and metabolism is depicted in Fig. 2. AMPK increases fatty acid oxidation by diminishing the concentration of malonyl CoA, an allosteric inhibitor of carnitine palmitoyltransferase 1 (CPT-1), the enzyme that governs the transfer of long-chain fatty acyl-CoA (FA-CoA) from the cytosol into mitochondria. It accomplishes this acutely (minutes) by phosphorylating and inhibiting acetyl-CoA carboxylase (ACC), the rate-limiting enzyme for malonyl-CoA synthesis, and phosphorylating (most likely) and activating malonyl-CoA decarboxylase (MCD), which catalyzes malonyl-CoA degradation (18). In addition, AMPK subacutely (hours) represses the expression of the transcriptional activator SREBP1C, leading to decreases in ACC abundance, and it increases the expression of the transcriptional coactivator peroxisome proliferator–activated receptor-γ coactivator 1, PGC1α, and, secondary to this, the transcription factor peroxisome proliferator–activated receptor-α, leading to increases in the synthesis of MCD and CPT-1 itself. Thus, when it is activated, AMPK increases fatty acid oxidation by a number of mechanisms and it does so both acutely and subacutely, underscoring how closely it governs this process.

AMPK also appears to exert multiple controls over the use of cytosolic FA-CoA for other purposes. For instance, by enhancing FA-CoA transfer into mitochondria, it makes less of it available for other processes in the cytosol. In addition, AMPK specifically diminishes the use of cytosolic FA-CoA for the synthesis of diacylglycerol, triglycerides, and phospholipids by decreasing the transcription of glycerophosphate acyltransferase, the first committed enzyme in the glycerolipid synthesis pathway; it diminishes the generation of lipid peroxides and the activation of nuclear factor κB (inflammation) in cells incubated with the fatty acid palmitate (19), and it inhibits the incorporation of palmitate into ceramide, a molecule implicated in causing insulin resistance, oxidative stress, and apoptosis (11,20). Finally, AMPK increases the expression of PGC1α, which in addition to regulating MCD and CPT-1 mRNA, enhances the expression of genes regulating mitochondrial biogenesis and oxidative phosphorylation (21–23), an effect that may be impaired in people with type 2 diabetes and their overweight offspring (21,24,25). It was in large part based on these effects, and the finding that diminished AMPK activity is associated with changes in the opposite direction, that we proposed that dysregulation of AMPK and malonyl-CoA could be pathogenetic factors for the metabolic syndrome as well as targets for its therapy (rev. in 17,26–29).

During exercise, AMPK is activated in contracting skeletal muscle (30,31) in response to increases in the AMP-to-ATP ratio, and it is well established that this contributes to the changes in muscle fuel metabolism that occur during and after physical activity (32,33). Surprisingly, however, in rats studied 30 min after a treadmill run, we also found increases in AMPK activity (Fig. 3) and associated changes in malonyl-CoA content and ACC, MCD, and glycerophosphate acyltransferase activity in liver and adipose tissue (34). Because the energy state of these tissues was not presumably altered, this raised the possibility that AMPK was activated by a systemic hormonal factor.

As noted earlier, IL-6 has long been recognized for its effects on the immune system (1), and sustained but modest increases in its plasma concentration have been found in proinflammatory insulin-resistant states. The latter association has been attributed to the presence of obesity, since adipose tissue is perhaps the major contributor to plasma IL-6 in most conditions (4,5). This view of IL-6 has recently been altered by the demonstration of Pedersen and coworkers (35,36) that IL-6 is also a myokine, for which concentration in plasma can rise dramatically during and after exercise as a result of increases in its synthesis and release by contracting skeletal muscle (37). In humans, plasma concentrations of IL-6 may be increased by as much as 100-fold during exercise, to levels far in excess of those observed in obese and diabetic humans (Fig. 4). In general, such increases in IL-6 have paralleled the intensity and duration of exercise, and they were greatest when muscle glycogen levels were low (38). It has been suggested that this increase in IL-6 plays a role in stimulating adipose tissue lipolysis and possibly hepatic glucose production during sustained exercise to provide for the fuel needs of muscle as its endogenous fuel stores (e.g., glycogen) are depleted (35,36). In keeping with this notion, IL-6 increases lipolysis upon infusion into humans at rest (39), and it further increases exercise-induced endogenous glucose production (40).

The temporal correlation between changes in plasma IL-6 and tissue AMPK during exercise led us to examine whether the two events are related. As a first test of this notion, we assessed the effect of IL-6 on AMPK activity in rat muscle and F442A adipocytes in vitro. IL-6 markedly increased AMPK in both tissues, with the increase in phospho-AMPK (P-AMPK) abundance (an index of activity) in the fat cells comparable to that produced by the classic AMPK activator 5-aminoimidazole-4-carboxamide riboside (AICAR) (Fig. 5). We have found similar IL-6–induced increases in AMPK in incubated pancreatic islets (A.-M.R., M.K., unpublished data) and in rat muscle and adipose tissue after an intraperitoneal injection of IL-6 (Fig. 5). Also of note, the increase in AMPK activity in incubated muscle was transient, and values below those of control muscles were observed by 60 min (41). Whether this secondary decrease in AMPK activity persists during longer periods of incubation with IL-6 and whether a similar pattern occurs in vivo and in other tissues remains to be determined.

To assess more directly whether IL-6 causes the increase in tissue AMPK during and after exercise, we compared AMPK activity in muscle, adipose tissue, and liver of control mice and IL-6 knockout (IL-6 KO) mice after 1 h of swimming (41). In keeping with previous studies in the rat (34), exercise increased the abundance of both P-AMPK and phospho-ACC (P-ACC) in muscle, liver, and adipose tissue of the control mice. No obvious difference in the ability of the control and IL-6 KO mice to perform exercise was observed. Despite this, in the IL-6 KO mice at rest, the abundance of P-AMPK and P-ACC was diminished by 75–90% in muscle (Fig. 6) and to a somewhat lesser extent in adipose tissue. After exercise, similar percentage increases in P-AMPK and P-ACC abundance were observed in muscle and adipose tissue of the control and IL-6 KO mice; however, because of their lower initial values, the absolute levels of P-AMPK and P-ACC achieved were lower in the IL-6 KO mice. For reasons unknown, P-AMPK and P-ACC were only minimally diminished in liver of the mice lacking IL-6, and the increments in their abundance produced by exercise were indistinguishable from those of control mice. Collectively, these data strongly suggest that IL-6 is a significant regulator of the basal activity of AMPK in muscle and adipose tissue. They also suggest that it is a key factor regulating the increase in tissue P-AMPK and P-ACC caused by exercise, but not the only factor. As discussed elsewhere (41), another likely regulator of the effects of exercise on AMPK is the sympathetic nervous system, which itself may be activated in part by a central action of IL-6 (42). Perhaps relevant to this discussion, it has recently been observed that transgenic mice overexpressing IL-6 do not develop obesity and insulin resistance as do control mice when fed a high-fat diet (M.F. White, personal communication).

At 3 months of age, the IL-6 KO mouse is not obese, whereas by 9 months of age, it is obese, hypertriglyceridemic, and glucose intolerant (42), which by definition gives it a diagnosis of the metabolic syndrome (43). Decreases in AMPK activity have been observed in a number of rodents that either have or subsequently develop manifestations of the metabolic syndrome. As listed in Table 1, they include rodents that are obese and not obese, diabetic and nondiabetic, and hypertriglyceridemic and normoglyceridemic, with insulin resistance a common denominator. In one of these rodents, the glucose-infused rat, insulin resistance, ectopic lipid accumulation, and decreased AMPK activity in liver all first appeared between 3 and 5 h, suggesting they are early events (44). AMPK activation has been demonstrated to diminish insulin resistance and other abnormalities in a number of these animals, suggesting it plays a causal role. For instance, treatment of the ZDF rat with AICAR (45,46) or regular exercise (46) has been shown to prevent the development of diabetes, as has treatment with thiazolidinedione therapy and calorie restriction (47), both of which have also been reported to activate AMPK in rats (48,49). Similar benefits of exercise (50) and AICAR (51) have been reported in rats made insulin resistant by fat feeding.

Several characteristics of the IL-6 KO mouse suggest that the early decrease in AMPK activity in muscle and adipose tissue of these mice could contribute to the metabolic syndrome phenotype observed at 9 months. Thus, Faldt et al. (52), based on measurements of respiratory exchange ratio at 3 months of age (respiratory exchange ratio 0.92 vs. 0.82 in control mice), concluded that the ability of the IL-6 mouse to oxidize fatty acids is impaired. Interestingly, they also noted that by 9 months of age, the respiratory exchange ratio of these mice had decreased to 0.82 (same as control mice). Whether this reflected an increased availability of fatty acids as a result of obesity and/or a restoration of AMPK activity remains to be determined. Also of note, Faldt et al. (52) found that the ability of both 3- and 9-month-old IL-6 KO mice to sustain exercise was diminished, suggesting impaired cardiovascular, pulmonary, or muscle function. In keeping with the latter possibility in a very preliminary study (in 3-month-old IL-6 KO mice), we found decreased levels of cardiolipin, a mitochondrial lipid whose concentration diminishes when mitochondria are damaged (e.g., as seen when apoptosis and impaired mitochondrial function are caused by incubation of cells with the fatty acid palmitate) (53) (Table 2). In addition, we found a reduced (50%) abundance in muscle of the mRNA of mitochondrial uncoupling protein (UCP)-3. Diminished size and efficiency of mitochondria have been observed in muscle of patients with type 2 diabetes (54). Whether decreased AMPK activity, by virtue of its effects on cellular lipid metabolism, mitochondrial genes, or other factors, causes these abnormalities to our knowledge has not been studied. On the other hand, it has been demonstrated that AMPK activation prevents both the apoptosis and mitochondrial dysfunction observed in cultured human umbilical vein endothelial cells incubated in a high-glucose medium (55) and the apoptosis, inflammation, and oxidative stress observed when human umbilical vein endothelial cells are incubated with a modestly elevated concentration of palmitate (19).

In addition to being predisposed to obesity, dyslipidemia, and glucose intolerance, IL-6 KO mice demonstrate more advanced atherosclerosis when bred on an Apo E^−/− background than do Apo E^−/− mice in which IL-6 is not lacking (56,57). On the other hand, humans with modest elevations of IL-6, attributable to obesity (5), are also predisposed to diabetes and atherosclerotic vascular disease (58). It is tempting to speculate that differences in AMPK activity in response to IL-6 in various tissues could contribute to this paradox; however, there is as yet no evidence for or against this notion. In this context, studies of both the mechanism by which AMPK is activated by IL-6 and the effects of AMPK activation on different components of the IL-6 signaling pathway could prove interesting (Fig. 7).

The research described in this article was supported by U.S. Public Health Service Grants DK19514 and DK07201 (N.B.R.) and DK049200 and P30-DK046204 (D.E.K.) and research grants from the Danish National Research Foundation 02-512-55 (B.K.P.); the Ministrerio de Ciencia y Technologia and Feder SAF 2005-00671 and European Commission FP6 Integrated Project LSHM-CT-2004-005272 (J.H.); and the Juvenile Diabetes Research Foundation 1-2003677 (N.B.R.).

The authors gratefully acknowledge the assistance of Paula Ohara and Kristy Pahl in preparing the manuscript.