Genetic Deletion of Trb3, the Mammalian Drosophila tribbles Homolog, Displays Normal Hepatic Insulin Signaling and Glucose Homeostasis

Mice were generated using bacterial artificial chromosome–based targeting vectors. In brief, we modified a bacterial artificial chromosome containing the Trib3 coding sequence by replacing the 6.2-kb coding region, extending from the initiation to the termination codon, with the β-galactosidase (β-gal) reporter gene followed by a neomycin cassette. The targeting vector was linearized and electroporated to mouse embryonic stem cells. The targeted embryonic stem cells were identified with a loss of native allele (LONA) assay as described previously (6). After germline transmission was established, F1 heterozygous mice were bred together to generate F2 Trib3^+/+ (wild-type) and Trib3^−/− mice. We performed all of the experiments with the F2 generation of male mice at 8–12 weeks of age unless noted otherwise. All animal procedures were conducted in compliance with protocols approved by the Regeneron institutional animal care and use committee. Mice were housed under 12 h of light per day in a temperature-controlled environment and had free access to standard chow diet (no. 5001; Purina, St Louis, MO).

For Northern blot multitissue analysis, we generated and labeled specific Trib1, Trib2, and Trib3 probes and hybridized against premade mouse polyA+ RNA Northern blot membranes (Origene Technologies, Rockville, MD), as outlined previously (7). Analysis for specific deletion of Trib3 mRNA was conducted on liver total RNA isolated using Trizol reagent (Life Technologies, Carlsbad, CA) taken from 8-week-old mice that were fasted for 24 h. We loaded 10 μg/lane of total RNA on a gel, transferred them to a nylon membrane, and probed them with ³²P-labeled Trib3 probe. These same filters were stripped of the probe and reblotted for Gapd. To assess for any changes in expression of Trib1, Trib2, and Trib3 mRNA in Trib3^−/− mice, tissue was taken from liver, white and brown adipose tissue, and skeletal muscle (gastrocnemius) and analyzed in separate reactions using a TaqMan real-time quantitative RT-PCR (qRT-PCR) chemistry and detection system (Applied Biosystems, Foster City, CA) with the primer pairs and labeled probes (sequences available on request). mRNA levels were calculated from a standard curve and normalized to a housekeeping reference (Gapd). Data were the means ± SE for five separate samples run in triplicate. For analysis of hepatic gene expression profile, we used qRT-PCR as described above and probes as outlined previously (7). Relative mRNA levels were calculated using a standard curve, with the PCR product for each primer pair normalized first to Gapd and then to the mean value of wild-type mice. Data were the means ± SE for six separate samples run in triplicate. To perform β-gal analysis in denervated skeletal muscle, the right sciatic nerve was isolated in the mid-thigh region and cut, leading to denervation of the lower limb muscles. Then, 2 weeks later, the tibialis anterior muscle of both limbs was isolated from ad libitum–fed mice under isoflurane (2–2.5%) anesthesia.

Mice were fasted overnight and injected intravenously with a bolus of insulin (5 units/kg). After 2 or 30 min, mice were killed and tissues collected. Liver detergent extracts were prepared in buffer containing 50 mmol/l HEPES (pH 7.4), 1% NP-40, 150 mmol/l NaCl, 1 mmol/l EDTA, 30 mmol/l sodium pyrophosphate, 50 mmol/l sodium fluoride, 0.5 mmol/l sodium orthovanadate, 5 μg/ml aprotinin, 5 μg/ml leupeptin, and 1 mmol/l phenylmethylsulfonyl fluoride. For Western blotting, equal amounts of protein (50 μg) were resolved by SDS-PAGE on 4–12% precast gels (Invitrogen, Carlsbad, CA) and transferred to nitrocellulose membranes. We probed the membranes with anti-Trb3 (EMD Biosciences, Darmstadt, Germany), Akt, phospho-specific Akt (Ser⁴⁷³), phospho-specific Gsk3α/β (Ser^21/9) (Cell Signaling Technology, Beverly, MA), or Gsk3β (Santa Cruz Biotechnology, Santa Cruz, CA) antibodies. Bound antibodies were detected with horseradish peroxidase–coupled antibodies against rabbit or mouse IgG (Bio-Rad Laboratories, Hercules, CA) using an enhanced chemiluminescent detection system (Amersham Biosciences, Buckinghamshire, U.K.). Akt phosphorylation was quantitated with preoptimized phospho–enzyme-linked immunosorbent assay (ELISA) for Akt (Biosources, Camarillo, CA).

Mice were fasted 24 h and then killed, and tissues were collected. Preparation of protein extracts and SDS-PAGE were performed as described above. The membranes were probed with an anti–acetyl-CoA carboxylase 1/2 (ACC1/2; Cell Signaling Technology) antibody. Bound antibodies were detected with the system described above. We measured the intensity of the bands by scanning densitometry of the autoradiograms with ImageJ software (version 1.36x; National Institutes of Health).

Mice were fasted 24 h, killed, and liver collected. Liver samples were homogenized and glycogen precipitated in ethanol. Pellets were reconstituted in Na-acetate buffer (0.04 mol/l) and hydrolyzed with amyloglucosidase (Sigma-Aldrich, St. Louis, MO) at 37°C for 1 h. Glucose concentration was determined using the Trinder glucose assay kit (Sigma-Aldrich).

After 4 h of fasting, mice were anesthetized with isoflurane. We collected orbital blood and analyzed serum for triglycerides and cholesterol with the Bayer 1650 blood chemistry analyzer (Bayer, Tarrytown, NY). Nonesterified free fatty acids (NEFAs) were analyzed with a diagnostic kit (Wako, Richmond, VA). We measured insulin and leptin levels using ELISA (LincoPlex; Linco, St. Charles, MO). The 4-h fasting blood glucose levels were measured with Accu-Chek (Roche Diagnostics, Nutley, NJ) without anesthesia.

For glucose tolerance tests, mice were fasted overnight and administered a bolus glucose (2 g/kg) by oral gavage. For insulin tolerance tests, mice were fasted for 4 h and injected intraperitoneally with a bolus insulin (0.75 units/kg). For pyruvate tolerance tests, mice were fasted for 18 h and injected intraperitoneally with a bolus of sodium pyruvate (2 g/kg). We collected blood from the tail vein at 0, 30, 60, and 120 min and measured glucose levels using Accu-Chek. Metabolic parameters were measured using an Oxymax open-circuit indirect calorimetry system (Columbus Instruments International, Columbus, OH) as described previously (7). Briefly, mice were individually placed in the calorimeter cages, and data were collected on gas exchanges and activity for 48 h. Oxygen consumption (in ml/kg) and carbon dioxide production (in ml/kg) by each animal were measured at 1-h intervals. Basal metabolic rate (in ml · kg⁻¹ · h⁻¹) represents the mean oxygen consumption rate at rest during the light cycle. Respiratory quotient (RQ) was calculated as the ratio of CO₂ production rate (Vco₂) over oxygen consumption rate (Vo₂). Energy expenditure (in kcal · kg⁻¹ · h⁻¹) was calculated as the product of the caloric value of oxygen (3.815 + 1.232 × RQ) and the volume of consumed oxygen. We assessed body composition by use of dual-emission X-ray absorption (PIXImus; Lunar, Madison, WI).

We present all values as means ± SE. For comparison of means, we performed unpaired nonparametric Student's t test or ANOVA, where appropriate, using Prism software (GraphPad, San Diego, CA). We selected a threshold for statistical significance of P < 0.05. When a significant F ratio was obtained with ANOVA, post hoc analysis was conducted between groups using a multiple comparison procedure with Bonferroni/Dunn post hoc comparison.

The recently described VelociGene technology was used to delete the entire Trib3 coding region (Fig. 1A). The correct gene targeting in F1H4 embryonic stem cells was confirmed by LONA assay (6), and mice derived from two independent clones were initially bred to C57BL/6 females to generate F1 offspring. F1 heterozygote breeders were set up to generate F2 litters, and all genotypes were born at the predicted Mendelian birth ratio (Fig. 1B). Cohorts of Trib3^−/− mice reached all developmental milestones over the next 6–8 weeks and showed no abnormalities in growth thereafter (data not shown).

To clarify tissue-specific expression patterns of Trib3 and the other Trib isoforms (Trib1 and Trib2) in mice, we performed Northern blot analysis using premade mouse polyA+ RNA blots, which include 12 major tissues (brain, heart, kidney, liver, lung, muscle, skin, small intestine, spleen, stomach, testis, and thymus) (Fig. 2A). Trib3 expression was the highest in liver and was also detected in heart, kidney, lung, skin, small intestine, and stomach. In agreement with qRT-PCR data (Fig. 2B), Trib3 expression was not detected in skeletal muscle. High levels of Trib1 expression were observed in liver and skin, whereas Trib2 did not appear to be expressed in liver. Neither Trib1 nor Trib2 was expressed in skeletal muscle. These Trib1 and Trib2 expression patterns were in agreement with qRT-PCR data for Trib1 and Trib2 in insulin-sensitive tissues (Fig. 2C and D). This analysis also revealed abundant expression of all Trib isoforms in white adipose tissue. Trib1 and Trib2 mRNA levels in insulin-sensitive tissues were similar between Trib3^−/− and wild-type mice, indicative of no compensatory induction with deletion of the Trib3 gene. Unfortunately, the insertion of the lacZ reporter gene into the Trib3 locus did not give a robust or reproducible staining by whole mount or in tissue sections by β-gal staining. However, when mice were fasted for 24 h, we observed lacZ expression in liver, pancreas, and the intestinal tract (data not shown). No specific staining could be detected with other tissues, probably because of lower sensitivity of this method compared with qRT-PCR.

Because Trib3 expression was reported to be upregulated in response to nutrient starvation and stress conditions (8–10), we conducted a detailed analysis of Trib3 mRNA and protein in Trib3^−/− and wild-type littermates under these conditions. Northern blot analysis (Fig. 1C) with Trib3-specific probes as well as Western blot analysis (Fig. 1D) of liver protein extracts with Trb3-specific antibodies confirmed the absence of the Trib3 gene product. To confirm the lack of Trib3 expression in other tissues, we measured Trib3 mRNA levels in insulin-sensitive tissues (liver, white and brown adipose tissue, and skeletal muscle) by qRT-PCR. Consistent with previous reports (4,11), we detected Trib3 expression in liver and adipose tissues but very low levels in skeletal muscle of wild-type animals. In Trib3^−/− mice, Trib3 mRNA was absent in all of the tissues examined, confirming global deletion of the Trib3 gene (Fig. 2B). To determine the effect of Trb3 loss in other settings where Akt has been shown to be important, we denervated skeletal muscle of Trib3^−/− mice. Denervation surgery was performed on the right hind limb of Trib3^−/− and wild-type mice, leaving the left hind limb unperturbed; the tibialis anterior muscle of both sides was analyzed 14 days later. The 14-day denervation caused atrophy in the right tibialis anterior muscle in both groups. No difference in percent atrophy was detected, even though in Trib3^−/− mice we observed strong lacZ expression on a subpopulation of fast fibers of denervated tibialis anterior muscle; we did not observe expression in nondenervated muscle (supplementary Fig. 1, which can be found in an online appendix [available at http://dx.doi.org/10.2337/db06-1448]).

Trib3^−/− mice were analyzed until 25 weeks of age, and no difference in appearance, body weight, and body composition in comparison to their wild-type littermates was observed (Table 1). Extensive analyses of 4-h fasting whole blood and serum at 8 and 25 weeks of age showed no difference in glucose, triglycerides, cholesterol, NEFA, insulin, and leptin levels between Trib3^−/− and wild-type male mice. In addition, 24-h fasting and postprandial blood glucose levels were normal in Trib3^−/− male mice. We obtained similar results in a cohort of female mice (data not shown).

A potential role for Trb3 to modulate hepatic insulin signaling was suggested by a two-hybrid screening experiment using Akt as a bait, which resulted in the cloning of Trb3 and differential expression studies, followed shortly by Trib3 RNA interference knockdown studies (4,5). To determine whether an increase in Akt phosphorylation in response to insulin in HepG2 cells is demonstrable in Trib3^−/− mice, we injected a bolus of insulin and collected liver tissue at 2 or 30 min post–intravenous injection. Western blot analysis of equal amounts of liver tissue lysates with phospho-specific antibodies to Akt (Ser⁴⁷³) did not detect any basally phosphorylated Akt in either group (Fig. 3A). At both 2 min and 30 min after an insulin injection, Akt activation was equal for Trib3^−/− and wild-type mice. Analysis by ELISA to measure levels of insulin-induced Akt phosphorylation at Thr³⁰⁸ also failed to find any difference between Trib3^−/− and wild-type liver (Fig. 3B). Previously reported in vitro Trib3 knockdown experiments (4) demonstrate an increase in insulin-induced phosphorylation of Gsk3β, an Akt substrate. Analysis of specific Gsk3β phosphorylation using phospho-specific antibodies to Gsk3β (Ser⁹) and Gsk3α (Ser²¹) found no differences in insulin-induced Gsk3β phosphorylation between Trib3^−/− and wild-type livers. These data suggest that complete deletion of Trib3 does not effect insulin-induced activation of Akt or specific downstream substrates in vivo.

Previous studies of Trib3 expression patterns, and in particular those where Trib3 is overexpressed, showed a significant impairment in glucose tolerance and a decrease in liver glycogen content (5). This would suggest a role for Trb3 in impairing hepatic glucose regulation. Elevated hepatic glucose production is known to be a major factor in the hyperglycemia associated with type 2 diabetes. Analysis of glucose and insulin levels in both fasting and fed conditions did not reveal any difference in these parameters between genotypes (Table 1 and data not shown). To uncover functions of Trb3 in liver that could not be detected by insulin signaling studies, we stressed mice by 24-h fasting, killed the animals, and harvested the liver. Analysis of specific genes involved in hepatic glucose production (Pepck, G6p, Pgc1α, and Glucokinase) (Fig. 4A) by qRT-PCR failed to detect any differential expression pattern that would be indicative of an alteration in gluconeogenesis. The 24-h–fasted liver glycogen content was also identical between genotypes (Fig. 4B). Pyruvate is a precursor for gluconeogenesis, and pyruvate tolerance tests have been used by others to detect subtle abnormalities in hepatic glucose production (12). Glucose response of mice challenged by the bolus administration of pyruvate (2 g/kg) was indistinguishable between Trib3^−/− and wild-type mice (Fig. 4C). Furthermore, in response to the bolus administration of glucose (2 g/kg) or insulin (0.75 units/kg), Trib3^−/− mice showed normal glucose homeostasis and insulin sensitivity (Fig. 5A and B). No compensatory changes were observed in the counterregulatory hormone glucagon. These physiological results suggest normal hepatic insulin sensitivity and that Trb3 does not play a pivotal role in the maintenance of glucose homeostasis in mice.

Qi et al. (13) recently reported that Trb3 mediates degradation of ACC in adipose tissue by promoting interactions of the E3 ubiquitin ligase COP1 (constitutive photomorphogenic protein 1) to ACC. ACC is an enzyme that catalyzes the formation of malonyl-CoA from acetyl-CoA, therefore serving as a critical regulator of fat storage and utilization. To examine whether ACC levels were altered, we conducted a Western blot analysis on tissues from Trib3^−/− and wild-type mice using an anti-ACC1/2 antibody. In all tissues analyzed (white adipose tissue, brown adipose tissue, liver, and skeletal muscle), we could not detect any change in ACC protein level (Fig. 3C). Additionally, expression levels of genes known to be involved in the regulation of fatty acid oxidation (Pparα, Pgc1α, and Ucp1) were also unaltered in Trib3^−/− mice (data not shown). Combined with the physiological data (normal fat mass and serum lipid levels) (Table 1), these results strongly suggest normal lipid metabolism in Trib3^−/− mice.

Roles of Trb3 in the control of energy homeostasis have not been well understood. Therefore, we examined metabolic parameters of Trib3^−/− mice with the use of indirect calorimetry. We placed mice in individual metabolic chambers for 48 h and monitored gas exchange, activity, and consumption of food and water. Basal metabolic rates or resting oxygen consumption did not differ between Trib3^−/− and wild-type mice (Fig. 5C). RQ provides a measure of fuel substrate preference (carbohydrate versus lipid), and an RQ of >0.9 indicates oxidization of carbohydrate to fulfill energy needs. Trib3^−/− and wild-type mice displayed indistinguishable RQs (Fig. 5D). Energy expenditure or heat production was also similar between Trib3^−/− and wild-type mice (Fig. 5E). Physical activity levels of Trib3^−/− mice were ∼47% higher than those of wild-type mice; however, the difference did not reach statistical significance (P = 0.063) (Fig. 5F). We also monitored water and food intake and found no difference between Trib3^−/− and wild-type mice (data not shown). These data indicate that Trb3 does not have a major impact on energy homeostasis in mice.

The PI 3-kinase/Akt pathway primarily mediates insulin's anabolic effects on the synthesis and storage of proteins, carbohydrates, and lipids. Because resistance to insulin's effects is one of the earliest hallmarks of type 2 diabetes, the identification and potential inhibition of negative regulators of this pathway is of great therapeutic interest. In this respect, genetic deletion studies in rodents have been able to provide in-depth insight into the signaling intermediates and the involvement of specific physiological processes, particularly with complex hormones such as insulin. From detailed in vivo and biochemical studies outlined here on Trb3-deficient mice, however, we could find no evidence to support the theory that Trb3 plays a major role in regulating hepatic insulin sensitivity and glucose homeostasis.

Trb3 was initially proposed to negatively regulate Akt activity and insulin action (4), and, as such, an extensive tissue-specific expression pattern analysis was of great interest. Trib3 expression analysis by Northern blotting and qRT-PCR revealed its expression in liver, adipose tissues, heart, kidney, lung, skin, small intestine, and stomach. We also performed the analysis using the β-gal reporter gene placed in the endogenous Trib3 locus. In skeletal muscle, lacZ staining was absent in tissues from fasted Trib3^−/− mice but present in denervated tibialis anterior fibers. These results indicate the possibility that Trb3 is induced by distinct stresses in different tissues. The data also demonstrate the surprising finding that Trb3 is not universally expressed even in a discrete cell type, under a particular condition. This finding helps to suggest that Trb3 is not an obligate regulator of the Akt signaling pathway and leaves open the possibility that Trb3 serves a role in particular cells. However, when global homeostasis is assessed, whatever discrete role Trb3 might play was shown to be inadequate to induce a measurable global perturbation.

Insulin's actions on hepatic glucose production are primarily mediated by PI 3-kinase, which in turn activates several downstream targets such as Akt and atypical forms of protein kinase C. Akt activation has been documented to suppress gluconeogenesis, predominantly through Foxo1 (forkhead box class O1)-regulated expression of G6P (14,15), as well as Gsk3β activation of glycogen synthesis (16). Recent tissue-specific deletion studies have elegantly demonstrated that PI 3-kinase pathway activation of protein kinase C-λ/ξ isoforms is more important for regulating insulin's actions on lipid metabolism (17), and, as such, specific inhibitors of Akt, such as Trb3, may be confined to modulating glucose metabolism. Previously reported in vitro studies in HepG2 cells with RNA interference knockdown of Trib3 demonstrated enhanced insulin-induced Akt and Gsk3β phosphorylation (4). Our in vivo analysis of hepatic Akt and Gsk3β activation in Trib3^−/− mice after insulin stimulation, however, is not consistent with those findings. We found no increases in levels of phosphorylation at two major phosphorylation sites of Akt (Ser⁴⁷³ and Thr³⁰⁸) and phosphorylation sites of Gsk3α and -β (Ser²¹ and Ser⁹, respectively) in the liver of Trib3^−/− mice. It should also be noted that other in vitro studies in rat primary hepatocyte cultures with Trb3 overexpression also could not demonstrate a decrease in insulin-induced phosphorylation of Akt and its substrates Gsk3α/β and 4E-BP1 (18).

Consistent with the unchanged Akt activation in response to insulin are our physiological findings in Trib3^−/− mice. Fed or fasted levels of serum glucose and insulin levels were similar for each genotype. No alteration could also be found in glucose, insulin, or pyruvate tolerance tests in Trib3^−/− mice, and no evidence could be found for altered secretion of insulin in response to a glucose bolus. Once again, these results are in conflict with the report by Koo et al. (5), where knockdown of hepatic Trib3 expression improved glucose tolerance in mice. It is possible that the discrepancies may have arisen from differences in the model and test conditions, such as durations and sites of Trb3 inactivation; for example, Trib3^−/− mice lacked Trb3 globally from gestation, whereas hepatic Trib3 knockdown studies are of short duration, in mice after the age of 7 weeks. Alternatively, reduction of Trb3 solely in liver may improve glucose tolerance. These are, however, relatively ineffectual answers given the data from similar genetic deletion studies of other intermediates in, or regulators of, the insulin-signaling pathway.

Many genetic studies have illustrated the direct action of insulin on liver physiological function, particularly its critical role in the control of gluconeogenesis. Liver-specific insulin receptor knockout (LIRKO) yields mice that develop marked insulin resistance and glucose intolerance, accompanied by a failure of insulin to suppress hepatic glucose production and to regulate hepatic gene expression (19). Conversely, hepatic overexpression of constitutively active Akt induces marked hypoglycemia, hypoinsulinemia, and hypertriglyceridemia and concomitant changes in gluconeogenic genes (1). More pertinent to this discussion are results from the ablation of the phosphoinositide phosphatases SH2-domain containing inositol 5-phosphatase 2 (SHIP2) and phosphatase and tensin homolog deleted on chromosome 10 (PTEN), both negative regulators of the PI 3-kinase pathway. Although SHIP2-deficient mice do not display changes in basal or fasted glucose, insulin, or tolerance tests, they do show significant changes in energy expenditure, enhanced insulin-stimulated Akt activity in liver, and reduced circulating triglycerides, NEFA, and cholesterol (20). Although little physiological information is available from the global deletion of PTEN because it is embryonically lethal, hepatic-specific deletion studies suggest a significant role in insulin signaling and metabolic regulation (21). In particular, the hepatic gluconeogenic enzymes G6P and PEPCK were significantly reduced, and this correlated with liver glycogen levels and Gsk3β activity. None of the above parameters were modified in Trb3-deficient mice.

In summary, by global deletion of the Trib3 gene in mice, this study directly asked the question as to whether inhibition of Trb3 would alter glucose or energy homeostasis in vivo. Although Trib3 expression was observed in liver at nutrient starvation as reported by others, whole-body ablation of Trb3 did not affect liver insulin sensitivity or overall metabolism. Based on these results, we conclude that loss of Trb3 is not sufficient to alter the maintenance of glucose and energy homeostasis in mice, and this casts doubt on the use of Trb3 as a therapeutic target for type 2 diabetes. It is still possible that Trb3 is important in other contexts, such as more prolonged or different stress conditions. Challenge studies, such as high-fat feeding and induction of hypoxia, as well as crossing of Trib3^−/− mice with other gene knockout or disease models, would be necessary to unveil these possible roles of Trb3.

We thank all at Regeneron Pharmaceuticals for their support and assistance, especially the VelociGene core for the preparation of targeting vectors, blastocyst injections, and genotyping of mice, and Tom Dechiara, William Poueymirou, and Melissa Meola for the coordinated breeding of mice. We also appreciate Drs. Katherine Wortley and Lori Gowen Morton for comments and revisions of the manuscript.