Deficiency of Phosphoinositide 3-Kinase Enhancer Protects Mice From Diet-Induced Obesity and Insulin Resistance

Heterozygous PIKE^+/− C57BL/6 mice with a targeted deletion of exons 3–6 of CENTG1 were generated under contract by Ozgene (Bentley DC, Australia). Genotyping was performed by PCR using genomic DNA isolated from the tail tip. PCR was performed using a combination of primers D (5′ACAGGATCAGTGCATCATCTC3′), H (5′CTGCCCAGCTACAGGAGTAG3′), A (5′TCAGTTGACTGGAAGCTCTG3′), and C (5′CCAGAGCCTATCTATGCCTAG3′).

Tissue extracts were prepared by homogenizing the tissues in buffer as reported (18). Immunoprecipitation was performed as described (18). Antibodies used in the Western blot analysis were obtained from Santa Cruz Biotechnology (insulin receptor, Akt) and Cell Signaling Technology (anti–phosphor-Thr³⁰⁸ of Akt, anti–phosphor-Thr¹⁷² of AMPK, anti–phosphor-Ser⁷⁹ of acetyl-CoA carboxylase [ACC], anti-AMPKα, and anti-ACC).

Southern blot analysis using mouse tail genomic DNA was performed as reported (19).

All animal experiments were performed according to the care of experimental animal guidelines from Emory University. Twelve-week-old female mice were fed with chow or HFD (Research Diets) for 20 weeks. Blood glucose level was measured by ACCU-CHEK Advantage Blood Glucose Meter (F. Hoffmann-La Roche, Basel, Switzerland). Serum insulin was measured by ELISA (Crystal Chem). Serum triglyceride level was measured by Serum Triglyceride Determination Kit (Sigma-Aldrich). Serum tumor necrosis factor-α (TNF-α) was measured by ELISA (BD Biosciences). Glucose tolerance test (GTT) was performed on mice after peritoneal injection of d-glucose (2 g/kg body wt).

Animals (16 h fasting) were anesthetized by intraperitoneal administration of sodium pentobarbital (50 mg/kg body wt). Saline or 5 units human insulin (Eli Lilly) was injected through inferior vena cava. After 5 min, liver, hind limb muscles, and inguinal fat were removed and immediately frozen in liquid nitrogen.

In vitro PI 3-kinase assay was performed using anti-p110α (Santa Cruz Biotechnology) as described previously (8).

Total RNA from various tissues was prepared by Trizol Isolation Reagent (Invitrogen). First-strand cDNA from total RNA was synthesized using Superscript III reverse transcriptase (Invitrogen) and Oligo-dT₁₇ as primer. Amplification of preadipocyte factor 1 (Pref-1), adipocyte protein 2 (aP2), peroxisome proliferator–activated receptor-γ (PPARγ), and C BPα was performed using primers mPref1-F (5′-GACCCACCCTGTGACCCC-3′), mPref1-R (5′-CAGGCAGCTCGTGCACCCC-3′); maP2-F (5′-CAAAATGTGTGATGCCTTTGTG-3′), maP2-R (5′-CTCTTCCTTTGGCTCATGCC-3′); mPPARg2-F (5′-ATGCTGTTATGGGTGAAACT-3′), mPPARg2-R (5′-CTTGGAGCTTCAGGTCATATTTGTA-3′); and mC BPa-F (5′-ATCCCAGAGGGACTGGAGTT-3′), mC BPa-R (5′-AAGTCTTAGCCGGAGGAAGC-3′). Expression of PIKE-A was determined using primers D and H as described above. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was also amplified as internal standard using primers 5′-CGCATCTTCTTGTGCAGTGCC-3′ (forward) and 5′-GGCCTTGACTGTGCCGTTGAATTT-3′ (reverse).

³H-2-deoxyglucose uptake in soleus muscle and inguinal fat pad was performed in the presence or absence of human insulin (Eli Lilly) as reported (20).

Fatty acid oxidation was measured by determining the production of ³H₂O from [9,10-³H]-palmitate as reported (21).

Metabolic cage studies and in vivo glucose metabolisms including glucose infusion rate, glucose turnover rate, and glycogen synthesis were determined by hyperinsulinemic-euglycemic clamp as reported (22,23).

Results were considered significant when P < 0.05. Statistical analysis was performed using either Student t test, one-way ANOVA, or two-way ANOVA followed by Tukey multiple comparison test or Bonferroni post-tests using the computer program GraphPad Prism (GraphPad Software).

Detailed experimental procedures are in the supplementary methods (http://diabetes.diabetesjournals.org/cgi/content/full/db09-1404/DC1).

As a pioneer study on the physiological role of PIKE in obesity development, we generated whole-body PIKE^−/− mice with targeted disruption in the CENTG1 locus using the LoxP/Cre system. We first created a transgenic line with PIKE ^flox/+ allele by inserting two loxP sites into the introns flanking exons 3 and 6 (Fig. 1,A). PIKE^flox/+ mice were then bred with transgenic mice expressing Cre recombinase in all tissues. Deletion of exons 3–6 results in removal of GTPase domain and introduces a frameshift mutation that creates a new stop codon, producing truncated PIKE proteins for all isoforms. Heterozygous mating generated newborn pups at expected Mendelian frequency that appeared indistinguishable from the wild-type littermates, suggesting that PIKE was dispensable for embryonic development. Southern blot analysis showed exons 3–6 of the CENTG1 gene were effectively excised (Fig. 1,B), which was further supported by PCR analysis (Fig. 1,C). Immunoblotting analysis using antibody specific to the COOH-terminal of PIKE-A and RT-PCR confirmed the ablation of PIKE-A expression in various tissues (Fig. 1,D and E). PIKE^−/− mice are viable and fertile. However, a significant reduction of white adipose tissues (WATs) was detected in the PIKE^−/− mice, whereas no noticeable difference was found in other peripheral tissues (Fig. 1 E).

When fed a chow diet, the body weight of female PIKE^−/− mice was slightly, but significantly, lower at 8 weeks old compared with wild-type mice (17.29 ± 0.27 vs. 16.33 ± 0.29 g, P < 0.05, n = 7, Student t test). The difference was more prominent in mice fed with HFD (55% of calories derived from fat). After HFD feeding for 14 weeks, obesity developed in wild-type but not in PIKE^−/− animals (Fig. 2,A). Daily food intake of PIKE^−/− mice fed a chow diet was normal, but the amount of food intake in PIKE^−/− mice was substantially less than that in the control fed HFD (Fig. 2,B). Increased body weight was associated with a drastic gain of inguinal WAT weight in wild-type but not in PIKE^−/− mice (312% in wild-type vs. 46.5% in knockout) (Fig. 2,C). The adipocytes in PIKE^−/− mice were also smaller in both feeding conditions (Fig. 2,D and E). Moreover, circulating leptin and TNF-α concentrations were lower in PIKE^−/− mice (Fig. 2,F and G). Expression of PIKE-A was greatly enhanced in the WAT and muscle of mice fed with HFD and the genetically obese (ob/ob) mice (Fig. 2,H, first and fifth panels). In contrast, no noticeable alternation of hepatic PIKE expression was detected among all the tested groups (Fig. 2 H, third panel), suggesting a tissue-specific function of PIKE-A in obesity development.

Under chow diet feeding conditions, expressions of mature adipocyte markers aP2 and the master regulators of adipocyte differentiation, PPARγ and C/EBPα (24,25), were reduced in PIKE^−/− WAT (Fig. 3,A). However, no significant difference was found in preadipocyte marker Pref-1 between wild-type and mutant. Comparable increment of Pref-1 was also detected in both genotypes treated with HFD, suggesting the proliferation of preadipocyte was similar in both genotypes (Fig. 3 A). However, subsequent adipocyte differentiation was impaired in PIKE^−/− mice as the expression of aP2, PPARγ, and C/EBPα was greatly reduced.

Next, we sought to determine whether the deletion of PIKE-A per se is sufficient to prevent adipogenesis in vitro. Whereas mouse embryonic fibroblasts (MEFs) from wild-type mice differentiated into adipocytes, as evident by the accumulation of lipid droplets within the cells, PIKE^−/− MEFs failed to fully differentiate under the same condition (Fig. 3,B). A significantly lower amount of oil red O staining was found in PIKE^−/− MEFs after induction (Fig. 3,C). Moreover, expression of mature adipocyte markers aP2, C/EBPα, and PPARγ was lower in PIKE^−/− MEFs after differentiation (Fig. 3 D), which was consistent with the findings in WAT. These results suggest lipoatrophy is a direct consequence of PIKE-A depletion in WAT, which may explain the reduced adiposity of mutant mice.

Adipocyte dysfunction is one of the major factors that causes insulin resistance (26); therefore, we examined the effects of diet-induced diabetes in PIKE^−/− mice. In HFD treatment, hyperglycemia was observed in both genotypes when the animals were fed. Hyperglycemia was sustained in fasted wild-type animals but not in PIKE^−/− mice (Fig. 4,A). Moreover, PIKE^−/− mice on HFD showed improved glucose tolerance during the GTT (Fig. 4,B). A lower amount of insulin was also secreted in PIKE^−/− mice treated with HFD during the GTT (supplementary Fig. 1A). In parallel, less circulating insulin was detected in PIKE^−/− mice in both feeding conditions (Fig. 4,C), suggesting a higher insulin sensitivity. This notion was further supported by higher glucose infusion rate (Fig. 4,D), whole-body glucose turnover (supplementary Fig. 1B), and glycogen synthesis (supplementary Fig. 1C) in PIKE^−/− mice during the hyperinsulinemic-euglycemic clamp studies. Hepatic insulin resistance was also alleviated in PIKE^−/− mice fed with HFD (supplementary Fig. 1D), which provides further explanation to the relieved diabetic phenotype in PIKE^−/− mice because hepatic insulin resistance is associated with diabetes (27). We also examined the insulin-stimulated signaling in tissues responsible for glucose utilization to reveal the molecular basis of the enhanced insulin sensitivity in PIKE^−/− mice. In mice fed a normal chow diet, comparable tyrosine phosphorylation of insulin receptor occurred in WAT and muscle of both genotypes after in vivo insulin injection (Fig. 4,E, first panel). However, insulin provoked higher insulin substrate-1 (IRS-1) phosphorylation, PI 3-kinase activity, and Akt phosphorylation in PIKE^−/− WAT and muscle (Fig. 4,E, third, fifth, sixth, and seventh panels). On the other hand, insulin hypersensitivity was not detected in PIKE^−/− liver (data not shown). Consistent with higher Akt phosphorylation in WAT and muscle, in vitro ³H-2-deoxyglucose uptake under insulin stimulation was significantly augmented in PIKE^−/− muscle and WAT (Fig. 4 F). Together, our data suggest that PIKE^−/− muscle and WAT are hypersensitive to insulin stimulation, leading to higher glucose uptake, thus protecting the mice from hyperglycemia during HFD treatment.

PIKE-A is an Akt upstream effector, which binds Akt and enhances its kinase activity in glioblastomas (6,13). It is thus anticipated that PIKE^−/− mice would display diabetic phenotypes as deletion of Akt2 in mice showed impaired glucose tolerance (28). To our surprise, blood glucose level is normal in PIKE^−/− mice. Because Akt1^−/− or Akt3^−/− mice have no obvious defect in glucose homeostasis, the normoglycemic condition in PIKE^−/− mice could be explained if PIKE-A associates selectively with Akt1 and Akt3 rather than Akt2. As predicted, PIKE-A preferentially bound both Akt1 and Akt3 (supplementary Fig. 2A), suggesting that only Akt1 and Akt3 activities may be altered in PIKE^−/− tissues. Concurrent with this notion, the brain mass of PIKE^−/− mice was smaller than the control mice (supplementary Fig. 2B), a phenotype that is specifically observed in Akt3-null animals (29).

Animal models with lipoatrophy often associate with hyperlipidemia and ectopic lipid accumulation (30). However, significant changes in neither circulating triglyceride (Fig. 5,A) nor ectopic lipid depositions in liver (Fig. 5,B) were seen in the PIKE^−/− mice, suggesting the excessively absorbed lipid during HFD feeding in PIKE-null animals may be metabolized rather than deposited as storage. To test this possibility, we first monitored the frequency of animal movements using open-field locomotor assay (31). Whereas the activity in wild-type mice decreased when they adapted to the test cage, physical movement of PIKE^−/− mice remained substantially higher throughout the experiment in both diet conditions (Fig. 5,C). We also examined the metabolic rate using metabolic cages (23). Respiratory exchange ratio was lower in PIKE^−/− animals in both diet treatments, suggesting that mutant mice have a higher fatty acid catabolism (Fig. 5,D). This suggested high lipid oxidation was further supported by the high phosphorylation level of AMPK and ACC in PIKE^−/− muscle, brown adipose tissue (BAT), and WAT. AMPK phosphorylation was reduced after HFD feeding in wild-type BAT and WAT. However, AMPK in PIKE^−/− WAT remained highly phosphorylated in both feeding groups (Fig. 5,E, first and tenth panels). In parallel, phosphorylation of ACC in PIKE^−/− BAT and WAT was higher than the control group (Fig. 5,E, third, fourth, and 12th panels). Whereas AMPK expression in WAT remained unchanged after HFD feeding, AMPK in BAT was greatly reduced in both genotypes (Fig. 5,E, second and 11th panels). Expression of ACC was reduced after HFD treatment in both wild-type and PIKE^−/− BAT and WAT (Fig. 5,E, fifth and 13th panels). Interestingly, PIKE^−/− BAT has higher ACC expression (Fig. 5,E, 13th panel). Similar phosphorylation pattern occurred in both AMPK and ACC in PIKE^−/− muscle (Fig. 5,E, sixth and eighth panels) with unchanged protein levels (32) (Fig. 5,E, seventh and ninth panels). In contrast, no significant changes in hypothalamic AMPK and ACC phosphorylation were found in PIKE^−/− animals in both feeding conditions (supplementary Fig. 3A). Moreover, AMPK and ACC phosphorylation was not enhanced in PIKE^−/− MEFs (supplementary Fig. 3B). These results suggest a tissue-specific effect of PIKE-A in modulating AMPK and ACC activity. We have also performed the fatty acid oxidation assay in cultured PIKE^−/− muscle cells and hepatocytes. In agreement with the immunoblotting analysis in Fig. 5,E, an elevated palmitate oxidation rate was detected in PIKE^−/− muscle cells but not hepatocytes (Fig. 5 F).

Therefore, the high physical activity of PIKE^−/− mice and enhanced lipid oxidation in BAT, WAT, and muscle may account for their lean phenotype during HFD feeding. The higher AMPK phosphorylation in PIKE^−/− WAT also provides a possible explanation for the defective adipogenesis observed, because prolonged AICAR-induced AMPK activation inhibits adipocyte differentiation by diminishing PPARγ and C/EBPα expressions (33,34). Furthermore, agonist-activated AMPK potentiates the insulin-stimulated glucose uptake by activating IRS-1 (35,36), which may explain the enhanced PI 3-kinase and Akt activities in the muscle and WAT of PIKE^−/− mice.

Next, we sought to clarify the role of PIKE-A in modulating AMPK phosphorylation. Fyn knockout (Fyn^−/−) mice are lipodystrophic with enhanced AMPK activity in muscle and WAT (37). These metabolic characteristics highly resemble the phenotypes of PIKE^−/− mice. Given that PIKE-A is a substrate of Fyn (38) and Fyn interacts with IRS-1 in an insulin-dependent manner (39), we hypothesized that PIKE-A may form a complex with Fyn and insulin receptor upon insulin stimulation, which is essential for insulin to suppress AMPK activity (40,41). In HEK293 cells, PIKE-A associated with insulin receptor through its NH₂-terminal (1–72 amino acids) (Fig. 6,A and B), in which their interaction could be enhanced by insulin stimulation (Fig. 6,C, first panel). However, this interaction was abolished when the Fyn phosphorylation site (Tyr682 and Tyr774) in PIKE-A (PIKE-A YY) was mutated (Fig. 6,C, first panel), suggesting Fyn phosphorylation is critical to the formation of PIKE-A/insulin receptor complex. The kinetics of insulin receptor/PIKE-A complex formation inversely correlated with the phosphorylation of AMPK (Fig. 6,D, first and sixth panels). Remarkably, neither Thr¹⁷² nor Ser⁴⁸⁵/Ser⁴⁹¹ phosphorylation (42) was altered by insulin in GST-PIKE-A YY–transfected cells (Fig. 4,D, sixth and seventh panels), suggesting that PIKE-A binding to insulin receptor is critical to mediate the inhibitory action of insulin on AMPK phosphorylation. On the other hand, Akt phosphorylation was not affected in either wild-type PIKE-A or PIKE-A YY cells in response to insulin (Fig. 6,D, fourth panel). The formation of PIKE/insulin receptor/Fyn complex was further demonstrated in muscle tissue. In vivo insulin injection in wild-type mice enhanced the formation of PIKE-A/insulin receptor complex, which was substantially reduced in Fyn^−/− tissue (Fig. 4,G, first panel). This complex was not detected in PIKE-null tissues (Fig. 6,E, first panel). Our immunoprecipitation results also confirmed that the association of Fyn and insulin receptor in muscle is insulin dependent (Fig. 6,E, second panel). Furthermore, the formation of Fyn/insulin receptor complex was not affected in PIKE-null tissues, suggesting that PIKE-A is not essential for their interaction (Fig. 6 E, second panel). Thus, the interaction between PIKE-A and insulin receptor is important for insulin to suppress AMPK phosphorylation, which provides a possible explanation to the enhanced AMPK phosphorylation in PIKE^−/− WAT and muscle.

One of the major findings in the current report is that PIKE-A is critical for adipocyte differentiation. Several lines of evidence support the role of PIKE-A in terminal adipocyte differentiation instead of preadipocyte formation. First, the mature adipocyte marker aP2 is significantly decreased during in vitro adipocyte differentiation in PIKE^−/− MEFs, indicating PIKE-A is important for adipocyte differentiation (Fig. 3,B and C). Second, PIKE-A expression is increased in fat tissue development of HFD-fed and ob/ob mice, which highlights its function in the process (Fig. 2,H). Lastly, HFD induced comparable preadipocyte marker Pref-1 expression in both wild-type and PIKE^−/− mice, indicating that formation of new adipocytes is normal in PIKE-null adipose tissue (Fig. 3,A). Interestingly, we found a small portion of PIKE^−/− MEFs was able to differentiate into mature adipocytes (Fig. 3,B), and quantitative analysis revealed a small but statistically significant increment of lipid accumulation in PIKE^−/− MEFs (Fig. 3 C). This result indicates that a PIKE-A–independent mechanism is responsible for some adipocyte differentiation, which also accounts for the existence but not completely the absence of adipose tissue in PIKE^−/− mice.

Ectopic lipid storage due to adipocyte differentiation defect is associated with hyperlipidemia and liver steatosis (43). However, we could not detect these pathologic conditions in PIKE^−/− mice (Fig. 5,A and B). It is thus reasonable to predict that the lipid spillover from adipocyte is metabolized in mutant animals. Our results that PIKE-null fat and muscle have significantly elevated AMPK and ACC phosphorylation suggest an elevated β-oxidation in these tissues (Fig. 5,E), which is further supported by the enhanced fatty acid oxidation rate in the in vitro assay (Fig. 5,F) and the low respiratory exchange ratio values (Fig. 5 D). AMPK has been viewed as a fuel sensor for glucose and lipid metabolism. Once activated, AMPK initiates a concomitant inhibition of energy-consuming biosynthetic pathways and activation of ATP-producing pathways such as fatty oxidation in mitochondria (44). As a result, most of the lipids absorbed in PIKE^−/− mice are oxidized as the energy source, which accounts for the lean phenotype during the HFD treatment.

The uplifted phosphorylation of AMPK and its downstream substrate ACC in PIKE-null muscle and adipose tissues indicates that PIKE-A negatively regulates the activities of these enzymes. This notion is further supported by the fact that PIKE-A is critical for insulin to inhibit AMPK phosphorylation in 293 cells (Fig. 6,D). This upregulation of AMPK activity in PIKE^−/− muscle and fat also provides a possible mechanism accounting for the elevated systemic insulin sensitivity, as AMPK and insulin signaling are intimately connected. Agonist-induced AMPK activation increases the glucose uptake in muscle (45). It also potentiates the insulin-stimulated glucose uptake by activating IRS-1 (35,36). A similar observation was made in adipose tissue that treatment of adipocytes with AMPK agonist AICAR enhanced basal glucose uptake by increasing GLUT4 translocation (46). Long-term activation of AMPK in mice increases the systemic insulin sensitivity and protects animals from HFD-induced obesity and diabetes (34,47), which is in agreement with our observations in PIKE^−/− mice. It is noteworthy that the alleviated insulin resistance in PIKE^−/− mice after HFD treatment may be a result of reduced inflammation. Because reduced circulating TNF-α could improve insulin sensitivity and increase AMPK activity (48,49), the low blood TNF-α in PIKE^−/− mice (Fig. 2 G) may also contribute significantly to improve the diet-induced insulin resistance.

Because the whole-body–knockout mice were used in the present study, we cannot exclude the possibility that deletion of PIKE-A in the brain causes a central effect to modify whole-body activity and metabolism. Because brain is the major site to control appetite and body weight (50), where PIKE is highly expressed (Fig. 1,D), it is reasonable to suspect that reduced food intake (Fig. 2,B) and elevated physical activity (Fig. 5,C) in PIKE^−/− mice are the primary causes of lean phenotype during HFD treatment. However, our data strongly support that peripheral ablation of PIKE-A does play a role in preventing obesity development. First, the feeding behavior is comparable between wild-type and PIKE^−/− mice fed a chow diet, when lipoatrophy is already obvious. Second, induced differentiation in PIKE^−/− MEFs is greatly impaired (Fig. 3,B), suggesting ablation of PIKE per se in MEFs is adequate to suppress adipogenesis. Third, PIKE-A interacts with the insulin receptor in a Fyn-dependent manner, which is essential for insulin-induced AMPK phosphorylation in muscle (Fig. 6,D). Deletion of PIKE in muscle, therefore, would enhance the AMPK phosphorylation and lipid oxidation (Fig. 5,E and F). Lastly, isolated PIKE^−/− WAT and muscle, in which the metabolic influence by the brain is eliminated, have higher ³H-2-deoxyglucose uptake when stimulated by insulin (Fig. 4 F).

Our data also suggest that the function of PIKE-A is not restricted to enhance Akt activity alone. We have demonstrated that PIKE-A physically interacts with the insulin receptor, which is important for insulin to suppress AMPK phosphorylation. Our data also provide a novel mechanistic insight into the phenotypes observed in Fyn^−/− mice (37), as PIKE-A/insulin receptor association is Fyn dependent. Conceivably, PIKE-A is a downstream target of Fyn that inhibits the activity of AMPK during obesity development. Thus, PIKE-A may represent an additional regulatory point, in addition to Akt, for insulin to suppress AMPK phosphorylation.

In all, our results uncover the novel physiological functions of PIKE-A, which plays important roles in obesity development and the accompanied insulin resistance by regulating AMPK activities negatively. Consequently, less fat is deposited and the associated insulin resistance is ameliorated. Therefore, PIKE-A may represent a potential therapeutic target for obesity and the adjunct insulin resistance.

K.Y. has received a grant from the National Institutes of Health RO1 (NS-045627).

No potential conflicts of interest relevant to this article were reported.

Parts of this study were presented in poster form at the 69th Scientific Sessions of the American Diabetes Association, New Orleans, Louisiana, 5–9 June 2009.