|Home | About | Journals | Submit | Contact Us | Français|
Obesity is associated with chronic low-grade inflammation that negatively impacts insulin sensitivity. Here we show that high fat diet can increase NFκB activation in mice, which leads to a sustained elevation in level of IκB kinase ε (IKKε) in liver, adipocytes and adipose tissue macrophages. IKKε knockout mice are protected from high fat diet-induced obesity, chronic inflammation in liver and fat, hepatic steatosis and whole-body insulin resistance. These mice show increased energy expenditure and thermogenesis via enhanced expression of the uncoupling protein UCP-1. They maintain insulin sensitivity in liver and fat, without activation of the proinflammatory JNK pathway. Gene expression analyses indicate that IKKε knockout reduces expression of inflammatory cytokines, and changes expression of certain regulatory proteins and enzymes involved in glucose and lipid metabolism. Thus, IKKε may represent an attractive new therapeutic target for obesity, insulin resistance, diabetes and other complications associated with these disorders.
Numerous longitudinal studies suggest that insulin resistance is the first step in the development of Type 2 diabetes, particularly in obese patients (Saltiel, 2001; Taniguchi et al., 2006; Thirone et al., 2006). Obesity produces a state of chronic low-grade inflammation, accompanied by increased circulating levels of pro-inflammatory cytokines (Hotamisligil, 2006; Shoelson et al., 2007; Wellen and Hotamisligil, 2005). Many of these cytokines can block insulin action, and knockout of some inflammatory genes disrupts the link between dietary or genetic obesity and insulin resistance (Hotamisligil, 2006; Shoelson et al., 2007). Many studies have indicated a role for NFκB (Tilg and Moschen, 2008; Wunderlich et al., 2008). This pathway may be activated by the toll-like receptor-4 (TLR4) due to interactions with dietary fatty acids (Kim et al., 2007; Tsukumo et al., 2007), or as a consequence of hypoxia (Schenk et al., 2008; Ye et al., 2007). Targeted deletion (Arkan et al., 2005; Cai et al., 2005; Zhang et al., 2008) or pharmacological inhibition (Yin et al., 1998; Yuan et al., 2001) of the kinase IKKβ, which lies upstream of the inhibitory IκB proteins, can restore insulin sensitivity in obese mice or humans.
Despite strong evidence for an inflammatory link between obesity and diabetes, the primary site or sites at which the inflammatory response occurs has not yet been established. Adipose tissue responds to overnutrition by secreting cytokines or chemokines that recruit proinflammatory, M1 polarized macrophages to adipose tissue (Lumeng et al., 2007b). These in turn secrete more cytokines that attenuate insulin action in adipocytes, resulting in increased lipolysis and free fatty acid release (Feingold et al., 1992; Green et al., 1994). However, the molecular details underlying macrophage recruitment and activation, the subtypes involved, their crosstalk with muscle, fat and liver cells, and the manner by which they regulate energy expenditure and storage remain uncertain. Here, we report that high fat diet induces the expression of the NFκB target IKKε in both liver and white adipose tissue, and further that mice bearing a targeted deletion of IKKε are surprisingly protected from diet-induced obesity, liver and adipose inflammation, hepatic steatosis, and insulin resistance, providing an appealing therapeutic target for obesity and type 2 diabetes.
While NFκB activation has been implicated in obesity, the range of tissues involved is unknown. We analyzed the effect of diet-induced obesity (DIO) on transgenic mice engineered with a luciferase construct driven by an NFκB-responsive promoter (HLL mice) (Sadikot et al., 2001). After injection with a luciferin substrate, high fat diet (HFD)-fed HLL mice demonstrated an approximate 2-fold increase in abdominal luminescence compared to chow-fed controls (Figure 1A). The reporter was activated 5-fold in subcutaneous and visceral adipose tissue after HFD; this activation persisted after correction for tissue weight (Figure 1B, C and Supplemental Figure 1A). Less pronounced transgene activation was seen in the liver, kidney and quadriceps muscle. Surprisingly, NFκB transgene activation was similar in both subcutaneous and visceral fat depots, suggesting that the documented inflammatory differences between these depots may be independent of NFκB.
It has been proposed that obesity-induced inflammation is chronic and low-grade compared to other inflammatory stimuli (Hotamisligil, 2006; Shoelson et al., 2007; Wellen and Hotamisligil, 2005). We compared the degree of NFκB activation in normal chow and HFD HLL mice before and after injection with lipopolysaccharide (LPS) (Supplemental Figure 1B). For all tissues examined (except muscle), LPS injection activated the transgene far above basal levels, supporting the notion that NFκB activation is sub-maximally, but chronically activated in obesity.
We performed immunohistochemical analyses in adipose tissue of HLL mice on control and HFD (Figure 1D). The luciferase reporter was specifically enriched in adipose tissue macrophage (ATM) clusters and adipocytes in epididymal fat pads from only HFD mice. NFκB expression (RelA/p65) was also more concentrated in the nuclei of F4/80+ ATM clusters by immunofluorescence in HFD mice, while the protein remained cytoplasmic in macrophages from chow-fed mice (Figure 1E).
We measured the expression of the gene encoding the inducible IKK family member IKKε in liver and white adipose tissue by real-time PCR, and compared this to the levels of other IKKs. HFD produced a small increase in the expression of IKKα, β and TBK1 in liver, but a 2.6 fold increase in IKKε mRNA (Figure 2A). In white adipose tissue (WAT), IKKα was slightly reduced, whereas HFD increased IKKβ 1.7 fold, and IKKε and TBK1 by 12 and 9 fold, respectively. To determine the cell types in adipose tissue responsible for these changes, we separated by centrifugation adipocytes from the SVF. HFD produced a 2 to 3-fold increase in expression of IKKα, IKKβ and TBK1 mRNA in adipocytes, whereas IKKε expression was increased up to 28-fold (Figure 2B). However, only IKKε was up regulated in SVF isolated from WAT, although the number of macrophages in adipose tissue was also significantly increased (Weisberg et al., 2003; Xu et al., 2003).
We also performed immunohistochemistry by confocal microscopy (Lumeng et al., 2007b). WAT from control diet-fed mice exhibited M2 polarized, MGL1+ ATMs (Lumeng et al., 2007a), whereas HFD produced increased infiltration of M1 polarized, MGL1− macrophages, detected in crown-like structures (Figure 2C). While IKKε was barely detected in adipose tissue from control mice, the protein was observed in adipocytes and MGL1−, F4/80+, but not MGL1+ ATMs in adipose tissue from HFD mice. Thus, HFD induction of IKKε occurred mainly in M1, but not M2 polarized ATMs, although it remains uncertain whether diet produces an increase in expression of the kinase in proinflammatory M1 ATMs, or rather reflects the different activation state of these cells.
We also monitored IKKε protein levels by western blotting (Figure 2D). IKKε expression was low in liver or WAT from wild type mice on a chow diet, but was markedly increased in both tissues after HFD, correlating well with RNA levels. Interestingly, none of the other IKK isoforms were up regulated in IKKε KO mice, suggesting that no compensation occurred (Supplementary Table 1). IKKε kinase activity increased by 3.7 fold and 1.5 fold in liver and WAT from HFD mice (Figure 2E). Treatment of COS cells overexpressing IKKε with LPS produced a time-dependent increase in activity (Supplementary Figure 2).
The increased expression of IKKε after HFD led us to wonder if this protein represented a link between obesity and insulin resistance. We therefore evaluated mice with a targeted deletion in the IKKε gene (Tenoever et al., 2007). On normal chow, weights, circulating levels of glucose and non-esterified fatty acids did not differ significantly between IKKε KO mice and their wild type counterparts, although triglycerides were lower and fasting insulin levels slightly higher in KO mice (Table 1). However, after exposure for 3 months to HFD, wild type controls gained near 20 grams, whereas IKKε KO mice gained only 12 grams (Figure 3A and Supplementary Figure 3). Body weight and percentage of lean and fat mass was similar between wild type and IKKε KO mice on chow (Supplementary Table 2). HFD increased lean and fat mass in wild type mice, with a statistically smaller increase in KO mice. HFD produced a large increase in liver weight in control but not in KO mice (Supplementary Figure 4). Adipocytes from IKKε KO mice were significantly smaller than those from wild type mice on HFD (Figure 3B and C), although there was a 10–15% increase in the number of cells in the epididymal fat pad (Figure 3D). Together these data suggest that deletion of the IKKε gene prevents the increase in liver mass induced by high fat diet, and maintains adipose tissue mass by increasing adipogenesis, reflected by more, smaller fat cells.
While no differences were detected in mice on a chow diet, the serum adiponectin levels were significantly higher in KO mice on HFD compared to wild type controls (Supplemental Figure 5A). As previously reported (Kadowaki et al., 2006), adiponectin levels per body weight were decreased by approximately 33% in wild type mice after HFD. Interestingly, this HFD-induced decrease was almost completely prevented in IKKε KO mice. HFD increased serum leptin levels by 8.5-fold in wild type mice (Supplementary Figure 5B), while leptin levels were approximately 40% lower in IKKε KO mice exposed to normal chow or HFD, probably reflecting smaller adipocyte size and increased leptin sensitivity.
Body weight represents a net balance of food intake and energy expenditure. IKKε KO mice showed higher daily food intake per body weight compared to wild type mice, either on chow or HFD (Figure 3E), perhaps due to lower circulating leptin levels. O2 consumption was similar in both wild type and IKKε KO mice on a chow diet (Figure 3F). Wild type mice on HFD showed little change in O2 consumption, whereas IKKε KO mice demonstrated a significant increase. This difference was consistent throughout light and dark phases, indicating an increase in energy expenditure. We also compared the respiratory quotient (RQ=VCO2/VO2), as a measure of fuel-partitioning patterns. RQ fluctuated between 0.85 and 1.0 in mice on a chow diet, and fluctuated between 0.8 and 0.9 in mice on HFD for both genotypes (Figure 3G).
The lack of effect of IKKε KO on RQ suggested that there was no difference in fuel selection between carbohydrates and lipids, leading us to explore whether the increase in energy expenditure might occur secondarily to increased thermogenesis, and we thus evaluated the expression of uncoupling proteins (Figure 3H). In chow-fed mice, UCP1 mRNA was barely detectable in WAT from wild type or IKKε KO mice. HFD produced a 2-fold increase in UCP1 mRNA in wild type mice, but generated a 10-fold increase in KO mice. UCP1 protein levels in WAT of KO mice on HFD were similarly higher. The expression of UCP2 was similar between wild type and KO mice on either diet (Supplementary Figure 6A), and UCP1 mRNA and protein levels in brown fat were unchanged (Supplementary Figure 6B). A significant increase in body temperature was also noted in IKKε KO mice (Figure 3I). IKKε KO mice were 1.5 oC warmer than their wild type counterparts on HFD, with a smaller 0.5 oC increase seen under chow-fed conditions. However, there was no apparent increase in the expression of mitochondrial biogenesis proteins in muscle, WAT or BAT in IKKε KO mice (Supplementary Figure 6C).
Because IKKε KO mice were protected from DIO, we investigated whether this gene might play a role in glucose homeostasis. As mentioned above, chronic exposure to HFD increased fasting glucose and insulin levels in wild type but not KO mice (Figure 4A). IKKε KO mice exhibited reduced fasting serum free fatty acid and cholesterol levels on HFD (Figure 4B), but were similar to wild types regarding fasting serum triglyceride levels.
We performed intraperitoneal (IP) glucose and insulin tolerance tests after 3 months of chow or HFD (Figure 4C and D). While wild type mice were glucose intolerant on HFD, IKKε KO mice maintained normal glucose tolerance and lower insulin levels. Although there were no differences detected between the genotypes on a normal diet, HFD IKKε KO mice were more sensitive to IP injection of insulin (Figure 4E) and pyruvate (Figure 4F) compared to wild type controls.
To investigate the mechanisms by which targeted disruption of the IKKε gene protects mice from the deleterious effects of HFD, we investigated insulin signaling in liver, fat and muscle ex vivo. In liver from both wild type and IKKε KO mice fed a normal chow diet, insulin injection stimulated Akt phosphorylation (Supplemental Figure 7A). While HFD produced a blunted insulin response in wild type mice (Khamzina et al., 2005), the IKKε KO mice exhibited normal insulin-stimulated Akt phosphorylation (Figure 5A). Insulin-stimulated Akt phosphorylation was similarly reduced in WAT after HFD in wild type but not KO mice (Figure 5B), but was similar between genotypes in gastrocnemius muscle (Figure 5C).The insulin-stimulated phosphorylation of the insulin receptor, and levels of IRS-1 were similarly restored in KO mice in liver and fat, but not muscle (Supplementary Figure 7B), suggesting that IKKε may be a local negative regulator of insulin signaling. Indeed, HFD had no effect on the expression of IKKε in muscle (data not shown).
Hepatic gene expression microarray analyses showed significant differences between IKKε KO and wild type controls on HFD in levels of mRNA encoding several proteins (Supplementary Table 3), including two genes involved in glucose homeostasis, pyruvate dehydrogenase kinase isoform 4 (PDK4) (Figure 5D) and glucokinase (Figure 5E), with smaller changes in fructose 1,6 bisphosphatase and malate dehydrogenase. Interestingly, there was little if any change observed in pyruvate kinase, PEPCK or glucose-6-phosphatase mRNAs (Supplementary Figure 8). The reduction in PDK4 and increase in glucokinase could produce increased flux of glucose through glycolysis, and may account for the improvement in pyruvate tolerance and the reduced level of free fatty acids observed in IKKε KO mice.
In addition to direct effects on hepatic gene expression, the increased hepatic insulin sensitivity in IKKε KO mice might also occur secondarily to increased circulating levels of adiponectin, which correlates inversely with insulin resistance (Kadowaki et al., 2006). Adiponectin mRNA was reduced in WAT from HFD wild type but not KO mice (Figure 5F), correlating well with circulating levels (Supplementary Figure 5A). Because adiponectin is regulated by the activity of PPARγ (Semple et al., 2006), we also examined whether PPARγ expression is induced in KO mice. PPARγ mRNA and protein levels in WAT were reduced by HFD in wildtype mice, but elevated in IKKε KO mice (Figure 5G). The PPARγ-regulated genes CD36, CAP and GLUT4 were also up regulated in adipocytes from these mice, as were the encoded proteins (Figure 5H) , further indicating that PPARγ activity is increased in adipose tissue from these mice. Recent studies showed that lipin1 directly interacts with PPARγ and increases its transcriptional activity (Koh et al., 2008). IKKε KO mice expressed nearly 2-fold greater levels of lipin1 mRNA and protein compared to control mice on HFD (Supplementary Figure 9), providing a clue to the improved insulin responsiveness in adipose tissue of the KO mice.
We next measured the insulin sensitivity of isolated adipocytes in vitro by assaying rates of lipogenesis at low concentrations of glucose, as a surrogate for glucose transport (Lesniewski et al., 2007). Adipocytes derived from wild type mice on a control diet responded to insulin with a two-fold increase in lipogenesis (Figure 5I), but those from HFD mice were almost completely unresponsive to insulin. In contrast, adipocytes isolated from IKKε KO mice on HFD remained insulin responsive.
To determine whether the resistance of adipocytes from IKKε KO mice to the deleterious effects of high fat diet is cell autonomous, we mimicked the HFD-induced expression of IKKε by overexpressing the enzyme in 3T3-L1 adipocytes, and then assayed for insulin-stimulated glucose uptake (Min et al., 1999) (figure 5J). Insulin produced a 10-fold increase in glucose transport in these cells. Transfection of cells with wild type IKKε reduced insulin-stimulated glucose uptake by 50%, whereas the expression of kinase-dead IKKε mutant produced only a small inhibitory effect. No significant differences in insulin-stimulated tyrosine phosphorylation of its receptor or IRS-1 were detected (Supplementary Figure 10), suggesting that IKKε overexpression did not directly influence this signaling pathway. These data, along with the changes in the expression of CAP (Ribon et al., 1998) and GLUT4 (Armoni et al., 2006) in vivo, suggest that increased expression of IKKε in adipocytes produces a direct, cell-autonomous reduction in insulin sensitivity, potentially through transcriptional regulation.
Chronic exposure of mice to HFD causes enlarged liver mass and accumulation of lipids, leading to fatty liver (steatosis) (Bradbury, 2006; Postic and Girard, 2008). HFD increased liver mass in wild type controls but not in IKKε KO mice (Figure 6A). The absence of steatosis in KO mice was apparent from examination of the liver, which was considerably darker than that of the wild type counterparts (Figure 6B). The triglyceride content of livers from IKKε KO mice was significantly lower than those of wild type mice after HFD, either in fed or fasted conditions (Figure 6C). Additionally, high fat diet caused abundant macrosteatosis in wild type but not KO livers, as visualized by H-E staining (Figure 6D), correlating well with reduced hepatic triglycerides.
Excessive fat accumulation in the liver can occur as a result of increased fat delivery or synthesis, reduced oxidation, and/or reduced fat export in the form of VLDL (Postic and Girard, 2008). Interestingly, we did not detect significant differences between wild type and KO mice in the expression of lipogenic enzymes, including FAS, ACC1, and Scd1, or those involved in beta-oxidation, including Acox1, Acad1, CPT1a, and MCAD, although all of these genes showed the expected response to nutritional status (Supplementary Figure 11). Lipin1 expression reduces VLDL-triglyceride release from liver (Chen et al., 2008), and its deficiency is associated with fatty liver and insulin resistance (Xu et al., 2006). Interestingly, hepatic mRNA and protein levels of lipin1 were increased in IKKε KO mice on both control and HFD (Figure 6E). Moreover, CD36 expression increased in response to high fat feeding in wild type but not KO mice (Figure 6F). Additionally, the HFD-induced increase in the expression of both hepatic FABP4 and PPARγ was partially prevented in IKKε KO mice (Figure 6F). Thus, although it is not possible to determine which effects were primary or secondary to reduced lipid accumulation, IKKε KO mice were protected from diet-induced hepatic steatosis partially due to direct inhibition of the expression of CD36, PPARγ and FABP4, and increased expression of Lipin1.
To determine whether increased levels of IKKε in liver cells can reproduce the effects of HFD on gene expression in a cell-autonomous fashion, we transfected H2–35 hepatoma cells with wild type or kinase-inactive IKKε, and assayed mRNA levels of selected genes by RT-PCR (Figure 6G). Interestingly, expression of the wild type kinase produced an approximate 1.5-fold increase in the expression of PDK4 and PPARγ, with a 9-fold increase in Rantes mRNA, 5-fold increase in IP-10 and 3-fold increase in iNOS expression. In contrast, overexpression of a kinase-dead enzyme was without effect. These data suggest that the IKKε-dependent changes in hepatic gene expression are likely to be direct and cell autonomous.
While serum levels of TNFα, MCP-1, and Rantes were similar between wild type controls and KO mice on chow diet (Figure 7A), HFD elevated the secretion of all three proinflammatory cytokines in wild type but not KO mice. Adipose tissue from IKKε KO mice on HFD exhibited a 90% reduction in ATM infiltration compared to wild type controls, as detected with F4/80 antibody (Figure 7B and C). The remarkable decrease in ATM content could conceivably be due to a changed property of the macrophages, or to altered chemotactic signals from adipocytes. Knock down of IKKε had no effect on the response of RAW264.7 cells to LPS regarding migration. Bone marrow-derived macrophages from KO mice were also similarly responsive to LPS (Supplementary Figure 12), indicating that the marked reduction in MI-polarized ATMs in KO mice was due mainly to a reduction in chemotactic signals from fat tissue. In this regard, the expression of mRNAs encoding the chemokines TNFα, Rantes, and MIP1α in adipose tissue was also significantly decreased, correlating well with reduced ATM infiltration, although levels of MCP-1α and IP-10 mRNA were unaffected (Figure 7D).
While it is clear that obesity can induce inflammation in adipose tissue, it is possible that inflammation occurs in liver as well (Sanyal, 2005; Schwabe and Brenner, 2006). HFD increased the mRNA levels of TNFα, MCP-1, MIP-1α, Rantes and IP-10 in livers of wild type but not KO mice (Figure 7E). iNOS expression was similarly elevated upon high fat feeding of wild type but not KO mice (Figure 7F).
Because inflammation appeared to be reduced in IKKε KO mice, we assayed signaling pathways in liver, WAT and muscle tissues thought to be associated with chronic inflammation. To assess the role of the JNK pathway (Nakatani et al., 2004; Singh et al., 2009; Solinas et al., 2007), we immunoblotted lysates from wild type and IKKε KO mice with a phospho-JNK antibody, as a surrogate to assay activation of the kinase. HFD produced increased JNK phosphorylation in liver, gastrocnemius muscle, and WAT (Figure 7G). Interestingly, IKKε KO mice exhibited reduced levels of JNK phosphorylation in all three tissues comparable to control-fed wild type mice. We observed similar results regarding JNK phosphorylation in isolated adipocytes and cells derived from the SVF, suggesting that loss of IKKε reduces the inflammatory response in both cell types (Supplementary Figure 13A). However, overexpression of IKKε in 3T3-L1 adipocytes had no effect on JNK phosphorylation, under conditions in which TNFα activates JNK (Supplementary Figure 13B). In addition, knockdown of IKKε in RAW cells did not alter JNK phosphorylation in response to LPS (Supplementary Figure 13C). These data suggest that the reduced activation of this pathway in KO mice was likely to be secondary to reduced fat accumulation, leading us to conclude that the JNK pathway is not a direct target of IKKε. Moreover, the level of IκB in these tissues was not changed by the loss of IKKε, confirming that IKKε does not appear to play a role in maintaining the stability of IκB (Tenoever et al., 2007).
Taken together, these data suggest that targeted deletion of the IKKε gene prevents the generation of low-grade inflammation in response to high fat feeding. Because IKKε can regulate expression of certain inflammatory genes, we wondered whether IKKε KO mice might be unresponsive to acute inflammatory signals as well. LPS injection stimulated both IKKβ and IκB phosphorylation in liver and WAT of both wild type and KO mice, and also led to a profound elevation in circulating levels of MCP-1 and Rantes within 2.5 hours (Supplementary Figure 14). The levels of these cytokines were completely unchanged in IKKε KO mice compared with wild type controls, suggesting that IKKε is not involved in the acute immune response, but may play a role in sustaining a state of chronic, low-grade inflammation in obesity.
A number of studies now suggest that obesity generates a state of chronic, systemic low-grade inflammation (Hotamisligil, 2006; Shoelson et al., 2007; Wellen and Hotamisligil, 2005). Numerous genetic and pharmacological approaches indicate that the NFκB pathway may play a crucial role in this response, although the tissues involved, the precise cell types that are activated in those tissues, and the signaling processes remain uncertain. To investigate the role of this pathway in obesity, we directly examined NFκB activation in vivo using HLL mice, a transgenic line bearing the NFκB-responsive promoter of HIV LTR fused to luciferase (Sadikot et al., 2001). These studies revealed that HFD activates NFκB in fat and liver, with the most profound increase in adipocytes and ATMs. The signals that lead to the activation of NFκB during high fat feeding are unknown, as are the potential consequences of sustained NFκB activation in metabolic tissues. We report here that the protein kinase IKKε, which is induced by activation of the NFκB pathway, might play a major role as a mediator of this multifaceted process independent of any role in acute immune responses.
IKKε is a direct transcriptional target of the NFκB pathway, and is likely to amplify further inflammatory signals via phosphorylation of other transcription factors. We show here that IKKε gene and protein expression and enzyme activity are increased in liver, adipocytes and ATMs after high fat feeding, coincident with the activation of NFκB. IKKε KO mice are rendered resistant to the deleterious effects of HFD on insulin sensitivity, exhibiting normal glucose, pyruvate and insulin tolerance. Their circulating cholesterol levels are markedly reduced, and adiponectin is elevated while leptin is lowered. Interestingly, these mice do not gain as much weight as wild type mice on HFD, due to increased energy expenditure. Additionally, these mice do not develop hepatic steatosis, macrophage infiltration in adipose tissue, or elevations in pro-inflammatory gene expression in fat and liver expected with an inflammatory state.
These dramatic effects on glucose and lipid metabolism, energy expenditure and inflammation beg an important question, what is the primary effect of HFD-induced IKKε elevation, and where does it occur? Increased energy expenditure without changes in RQ indicates that there is no shift in fuel substrate preference in these animals. Although adiponectin levels were elevated in KO mice, transgenic mice overexpressing adiponectin at levels significantly higher than seen here do not phenocopy the metabolic profile of IKKε KO mice (Otabe et al., 2007), suggesting that adiponectin alone is unlikely to account for the phenotype. However, there was an elevation in UCP-1 expression in WAT, and IKKε KO mice exhibited increased temperatures, indicating that increased energy expenditure may be a result of enhanced thermogenesis. The molecular mechanisms underlying this event are unknown, and could involve increased expression or activity of PPARγ (Kelly et al., 1998; Laplante et al., 2003), or alleviation of a direct inhibitory effect of IKKε on the UCP-1 promoter through phosphorylation of transcriptional regulator(s). One interesting possibility is that IKKε normally functions to repress an adaptive increase in expression of UCP-1 in WAT in response to HFD.
It is possible that both metabolic and inflammatory changes in the IKKε KO mice are secondary to lower body weight. Weight reduction would be expected to reduce macrophage infiltration in adipose tissue and improve glucose and insulin tolerance. However, this difference in energy expenditure was only seen in HFD mice, and weight-matched animals showed the same differences in rectal temperature, inflammation and hepatosteatosis (data not shown), suggesting that weight loss alone is unlikely to explain these traits. It is also possible that IKKε gene deletion in macrophages prevents the initial inflammatory response to HFD, subsequently leading to metabolic resistance to diet. Indeed, many studies have shown that deletion of inflammatory genes can restore global glucose and insulin tolerance. However, other targeted gene disruptions that block adipose tissue inflammation not prevent weight gain from HFD (Kanda et al., 2006; Lesniewski et al., 2007; Moller, 2000; Weisberg et al., 2006). While blockade of adipose tissue inflammation alone may not explain the entire phenotype, we cannot exclude the possibility that simultaneously reducing inflammation in liver and fat can produce these changes.
IKKε KO mice also exhibited gene expression changes that are consistent with improved insulin sensitivity. In adipose tissue, PPARγ mRNA was increased, along with changes in genes (adiponectin, CAP, CD36 and TNFα) sensitive to PPARγ activation. Moreover, there was a significant reduction in adipocyte size, along with an increase in cell number that is expected with PPARγ activation and increased adipogenesis (Okuno et al., 1998; Tsuchida et al., 2005). At the same time, overexpression of wild type IKKε produced a substantial reduction in insulin-stimulated glucose uptake in 3T3-L1 adipocytes, indicating that the IKKε-dependent insulin resistance in adipocytes might be at least partially cell autonomous.
IKKε KO also improved glucose and lipid metabolism in the liver. In addition to the almost complete prevention of steatosis, there were large reductions in circulating cholesterol. Transcriptional profiling of IKKε KO mice revealed reduced levels of PDK4, PPARγ, FABP4 and CD36 mRNA, with increases in PPARα, lipin1 and glucokinase, all of which are consistent with enhanced insulin sensitivity and decreased lipid synthesis and export. These genes may be regulated by transcription factors that are direct targets of IKKε. In this regard, overexpression of IKKε in hepatoma cells increased expression of selected gluconeogenic, lipogenic and proinflammatory genes, indicating that at least some of these effects were direct and cell autonomous in the liver.
We are currently investigating the kinase substrates of IKKε in tissues that respond to HFD. We hypothesize that some of these targets may be transcriptional activators or repressors that regulate metabolic genes. A number of the genes that changed in IKKε KO mice in both liver and fat are transcriptionally regulated by interferon regulatory factor (IRF) family members in 3T3-L1 adipocytes (Eguchi et al., 2008), although the mechanism(s) by which IRF3 acts as a repressor remains unknown. Additionally, IKKε can catalyze STAT1 phosphorylation on serine 708 in response to interferon β treatment in vitro (Tenoever et al., 2007). Phosphorylation of S708 favors the formation of STAT1/STAT2 heterodimers for assembly of the ISGF3 (interferon-stimulated gene factor 3) complex. This complex binds to interferon-stimulated response elements in the promoters of interferon-stimulated genes, some of which, including IFI205, IFIT2, and IFIT3, were identified in the microarray analyses, and thus might lie downstream of this pathway.
The increased expression of IKKε in liver, adipocytes and ATMs, coupled with the complex nature of the IKKε KO phenotype, suggest that IKKε is likely to exert its effects through all three tissues in a manner that involves crosstalk among these cells. Although IKKε is unlikely to play a role in the initial inflammatory response to high fat feeding, it may be required for a sustained inflammatory state. Although it is important to note that IKKε KO mice are susceptible to lethal viral infections (Tenoever et al., 2007), the specificity of the apparent actions of IKKε, the nature of the enzyme, and the profound resistance of KO mice to high fat diet, make it an especially appealing drug target for the treatment of metabolic disease.
See supplemental data for details.
IHC and confocal microscopy were performed as described (Lumeng et al., 2007a; Lumeng et al., 2008). Adipocyte cross-sectional area from caveolin stained adipose tissue images (150–200 adipocytes/mouse, 3 mice/genotype) was calculated using CellProfiler image analysis software. Adipocyte number was calculated from adipocyte diameters using established formulas (Hirsch and Batchelor, 1976).
Cos cells were transfected with plasmid encoding FLAG-IKKε for 24 h and treated with LPS (1μg/ml) for 30min before harvested. FLAG-IKKε in whole cell lysates was immunoprecipitated and subjected to an in vitro kinase assay using MBP as substrate. Following fractionation of samples by SDS-PAGE, the [r-32P]-ATP-labeled MBP on the lower half of the gel was quantified by autoradiography; the upper half of gel was blotted for the immunoprecipitated kinase using anti-FLAG antibody.
We thank Dr. Tom Maniatis for the IKKε mice and constructs, andfor helpful suggestions during the course of these studies. We also thank Drs. Benjamin tenOever and Jiandie Lin for helpful discussions. We are grateful to Xiao-ling Peng for maintaining the mouse colony and genotyping. We thank Saltiel lab members for support and suggestions, and we especially thank Dr. Dave Bridges for help on statistical analyses. We also thank UM Microarray Core for their help on microarray data analyses and UM Phenotyping Core for CLAMS study. This research was funded by NIH RO1DK060591 to ARS. MB was supported by a Mentor-Based Postdoctoral Award from the American Diabetes Association. CNL was supported by NIH DK-078851 and HD-028820.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.