Hepatic TRAF2 Regulates Glucose Metabolism Through Enhancing Glucagon Responses

TRAF2^flox/flox mice were provided by R.B. Albumin-Cre mice were from The Jackson Laboratory (Bar Harbor, ME). Hepatocyte-specific TRAF2 knockout (HKO) mice were generated by crossing TRAF2^flox/flox mice with albumin-Cre mice (in C57BL/6 genetic background). Mice were housed on a 12-h light/dark cycle in the Unit for Laboratory Animal Medicine at the University of Michigan. Mice were fed either a normal chow (9% fat; Laboratory Diet) or a high-fat diet (HFD; 60% fat; Research Diets) ad libitum with free access to water. Animal experiments were conducted following the protocols approved by the University of Michigan Committee on the Use and Care of Animals.

These experiments were conducted in conscious mice. Blood samples were collected from tail veins, and blood glucose and plasma insulin were measured as described previously (29). Plasma glucagon was measured using glucagon radioimmunoassay kits (LINCO, St. Charles, MO). Glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs) have been described previously (30). In pyruvate tolerance tests, mice were fasted overnight and intraperitoneally injected with sodium pyruvate (2 g/kg body wt). Blood glucose was monitored after pyruvate injection. In GTTs, mice were fasted for 5 h and intraperitoneally injected with glucagon (10 μg/kg body wt). To measure relative contributions of gluconeogenesis and glycogenolysis to hepatic glucose production, mice were fed an HFD for 10–12 weeks, fasted for 24 h, and intraperitoneally injected with deuterium water (99%, 27 mL/kg body wt) (Cambridge Isotope Laboratories, Andover, MA). Blood samples were collected via retro-orbital sinus 30 min after injection and pooled from three mice in each group as described previously (31). The pooled samples were used to purify glucose that was subsequently converted to monoacetone glucose (MAG) (31,32). MAG was subjected to ²H nuclear magnetic resonance (NMR) spectroscopy analysis. ²H NMR spectra were obtained by a 17.6 T Varian spectrometer (Varian Medical Systems, Palo Alto, CA), and relative ²H enrichments in H2, H5, and H6s resonances were measured. The relative contributions of glycogenolysis and gluconeogenesis (glycerol and PEP as substrates) to HGP were calculated as described previously (32–35): glycogenolysis: (H2-H5)/H2; gluconeogenesis (glycerol): (H5-H6s)/H2; gluconeogenesis (PEP): H6s/H2.

Mice were fasted 20–24 h, anesthetized, and administrated insulin (2 units/kg body wt) via inferior vena. Livers were isolated and homogenized in a lysis buffer (50 mmol/L Tris HCl, pH 7.5, 1.0% NP-40, 150 mmol/L NaCl, 2 mmol/L EGTA, 1 mmol/L Na₃VO₄, 100 mmol/L NaF, 10 mmol/L Na₄P₂O₇, 1 mmol/L phenylmethylsulfonyl fluoride [PMSF], 10 μg/mL aprotinin, and 10 μg/mL leupeptin) as described previously (36,37). Liver extracts were immunoprecipitated with anti–insulin receptor (IR) antibody (Santa Cruz Biotechnology) and immunoblotted with anti–phospho-Tyr antibody (Upstate). Liver extracts were also immunoblotted with antibodies against phospho-Akt (pSer473 from Cell Signaling Technology and Thr308 from Santa Cruz Biotechnology) or total Akt (Santa Cruz Biotechnology).

Liver frozen sections (7 μm) were fixed in 4% paraformaldehyde for 30 min, blocked with 5% normal goat serum (Invitrogen Life Technologies, Carlsbad, CA) plus 1% BSA for 3 h, and incubated with rat anti-F4/80 antibody (eBioscience, San Diego, CA) at 4°C overnight. F4/80-positive cells were visualized using Cy2-conjugated goat anti-rat secondary antibodies and a fluorescent microscopy.

Total RNAs were extracted using TRIzol reagent (Invitrogen Life Technologies) as described previously (36). The first-strand cDNAs were synthesized using random primers and Maloney murine leukemia virus reverse transcriptase (Promega, Madison, WI). mRNA abundance was measured using ABsolute QPCR SYBR Mix (Thermo Fisher Scientific, Epsom, Surrey, U.K.) and Mx3000P real-time PCR system (Stratagene, La Jolla, CA). The expression of individual genes was normalized to the expression of 36B4, a housekeeping gene. Primers for real-time quantitative PCR (qPCR) include peroxisome proliferator–activated receptor γ coactivator (PGC)-1α-F: TGGACGGAAGCAATTTTTCA, PGC-1α-R: TTACCTGCGCAAGCTTCTCT; PEPCK-F: ATCATCTTTGGTGGCCGTAG, PEPCK-R: ATCTTGCCCTTGTGTTCTGC; G6Pase-F: CCGGTGTTTGAACGTCATCT, G6Pase-R: CAATGCCTGACAAGACTCCA; 36B4-F: AAGCGCGTCCTGGCATTGTCT, 36B4-R: CCGCAGGGGCAGCAGTGGT; IL-1β-F: GCCTTGGGCCTCAAAGGAAAGAATC, IL-1β-R: GGAAGACACAGATTCCATGGTGAAG; IL-6-F: AGCCAGAGTCCTTCAGA, IL-6-R: GGTCCTTAGCCACTCCT; and TNF-α-F: CATCTTCTCAAAATTCGAGTGACAA, TNF-α-R: TGGGAGTAGACAAGGTACAACCC.

Primary hepatocyte cultures were prepared by liver perfusion with type II collagenase (Worthington Biochemical, Lakewood, NJ) and grown on collagen-coated plates as described previously (37). The cells were infected with indicated adenoviruses and subjected to HGP assays 16 h after infection as described previously (38). In brief, primary hepatocytes were incubated for 4 h in Krebs-Ringer bicarbonate buffer (118 mmol/L NaCl, 2.5 mmol/L CaCl₂, 4.8 mmol/L KCl, 25 mmol/L NaHCO₃, 1.1 mmol/L KH₂PO₄, 1.2 mmol/L MgSO₄, 10 μg ZnSO₄, 0.6% BSA, and 10 mmol/L HEPES, pH 7.4) supplemented with 10 mmol/L lactate and 1 mmol/L pyruvate, in the presence or absence of glucagon or the mixture of 10 μmol/L N⁶,2′-O-dibutyryl-cAMP sodium salt (DB-cAMP) and 100 nmol/L dexamethasone. Glucose in culture medium was measured and normalized to total protein levels.

Livers were perfused with type II collagenase. Dissociated liver cells were filtered through a cell strainer (100 μm) and centrifuged at 50g for 2 min. The supernatants were centrifuged at 200g for 10 min, and the pellets were resuspended in Ca²⁺- and Mg²⁺-free PBS and loaded onto a Percoll gradient solution (40 and 25%). After centrifugation at 500g for 20 min, Kupffer cells were highly enriched in the pellets.

G6Pase activity was measured as described previously (39). Liver samples (∼50 mg) were homogenized in an ice-cold lysis buffer (100 mmol/L Tris, pH 7.5, 0.25 mol/L sucrose, 5 mmol/L EDTA, and 1 mmol/L PMSF) and centrifuged at 10,500g and 4°C for 20 min. The supernatants were centrifuged at 100,000g and 4°C for 60 min. The pellets (microsomal fraction) were resuspended in a buffer (100 mmol/L Na-cacodylate and 10% glycerol, pH 6.5) and protein concentrations were measured. Microsomal fraction (10 μL) was incubated in a reaction buffer (100 mmol/L sodium cacodylate and 46 mmol/L glucose-6-phosphate) for 10 min at 37°C. Reaction was terminated by adding trichloroacetic acid (1.3%) and centrifuged at 16,000g for 5 min. The supernatant (10 mL) or phosphorous standards (0, 81.25, 162.5, 325, and 650 μmol/L) was mixed with Taussky-Shorr color reagents (1% ammonium molybdate, 0.5 mol/L sulfuric acid, and 0.18 mol/L ferrous sulfate) and incubated for 5 min at room temperature. Optical absorbance (620 nm) was measured by a SynergyTM HT multimode 96-well plate reader (BioTek) and used to calculate anorganic phosphate concentrations. G6Pase activity was expressed as milliunits (nanamoles of anorganic phosphate released per minute) and normalized to microsomal protein levels (milligrams).

Mice were fasted for 3 h, anesthetized, and injected with glucagon (30 μg/kg body wt) via inferior vena. Livers were dissected 5 min after injection and homogenized in a lysis buffer (20 mmol/L HEPES, 1 mmol/L EDTA, 250 mmol/L sucrose, 10 μg/mL aprotinin, 10 μg/mL leupeptin, 1 mmol/L PMSF, 1 mmol/L Na₃VO₃, and 0.5 mmol/L dithiothreitol, pH 7.4). Liver nuclei were isolated by sequential centrifugation at 4°C at 1,100g and 4,000g, respectively. The pellets were resuspended in a high-salt solution (20 mmol/L HEPES, 420 mmol/L NaCl, 0.2 mmol/L EDTA, 0.5 mmol/L dithiothreitol, 1 mmol/L PMSF, and 1 mmol/L Na₃VO₃, pH 7.9) to prepare nuclear extracts.

Data are presented as means ± SEM. Differences between groups were analyzed by two-tailed Student t tests. P < 0.05 was considered statistically significant.

TRAF2 mRNA abundance increased in mice with either genetic or dietary obesity (Fig. 1A). To determine the metabolic function of liver TRAF2, HKO mice were generated using the Cre/loxP system. TRAF2^flox/flox mice were generated and characterized previously (21). HKO mice were generated by crossing TRAF2^flox/flox mice with albumin-Cre transgenic mice, and both TRAF2^flox/flox and albumin-Cre mice were used as control. As expected, TRAF2 was deleted specifically in the livers but not muscle and adipose tissues of HKO mice (Fig. 1B). To verify TRAF2 deletion specifically in hepatocytes but not other types of liver cells, hepatocytes and Kupffer cells were purified and subjected to genotyping and qPCR analysis. The TRAF2 gene was deleted in the hepatocytes but not Kupffer cells of HKO mice (Fig. 1C). TRAF2 mRNA levels were markedly reduced in hepatocytes but not Kupffer cells (Fig. 1D). The residual TRAF2 mRNA in the hepatocyte fractions of HKO mice may result from a contamination of other cell types.

To examine the metabolic function of hepatic TRAF2, we measured blood glucose and insulin and performed GTTs and ITTs in mice fed a normal chow. Both overnight fasting blood glucose (Fig. 1E) and plasma insulin (data not shown) were similar between HKO and control mice. Glucose clearance rates and the ability of exogenous insulin to reduce blood glucose were also similar between HKO and control mice (Fig. 1F and G). These results suggest that TRAF2 deficiency in hepatocytes alone is not sufficient to alter glucose metabolism.

Because inflammation promotes insulin resistance in obesity (13,40), we examined the role of hepatic TRAF2 in obesity-associated metabolic disorders. HKO and control mice were fed an HFD, and blood glucose and plasma insulin were measured. Body weight was similar between HKO and control mice (Fig. 2A). Control mice developed hyperglycemia progressively under HFD conditions; in contrast, blood glucose levels were maintained at relatively normal levels in HFD-fed HKO mice (Fig. 2B). Fasting blood glucose was 30% lower in HKO than in control mice fed an HFD for 15 weeks (HKO: 76.6 ± 7.0 mg/dL, n = 13; Control: 110.5 ± 9.8 mg/dL, n = 13; P = 0.01). Blood glucose also was reduced (by 22%) in HKO mice under fed conditions (HKO: 111.2 ± 4.6 mg/dL, n = 10; Control: 141.1 ± 5.7 mg/dL, n = 16; P = 0.001). Fasting plasma insulin levels also significantly decreased in HKO mice (Fig. 2C). A reduction in plasma insulin may be secondary to decreased blood glucose in HKO mice. These data indicate that hepatic TRAF2 contributes to diet-induced hyperglycemia and hyperinsulinemia in mice.

To further analyze insulin sensitivity and glucose metabolism, we performed GTTs and ITTs. It is surprising that glucose excursion rates were similar between HKO and control mice (Fig. 2D), and exogenous insulin similarly reduced blood glucose between these two groups (Fig. 2E). These results suggest that hepatic TRAF2 contributes to diet-induced hyperglycemia independently of insulin resistance.

We analyzed HGP using pyruvate tolerance tests. Mice were fed an HFD for 17 weeks and injected with pyruvate, a gluconeogenic substrate. Blood glucose increased in both HKO and control mice after pyruvate injection; however, glucose levels were significantly lower in HKO than in control mice 0, 15, 30, and 120 min after injection, and the area under the curve (AUC) decreased by ∼24% in HKO mice (Fig. 3A). To measure relative contributions of glycogenolysis and gluconeogenesis to HGP, mice were fasted for 24 h and intraperitoneally injected with deuterium water. Blood samples were collected 30 min after injection and subjected to ²H NMR analysis to estimate the contributions of glycogenolysis and gluconeogenesis to HGP as described previously (31–35). ²H NMR spectra were different between control and HKO mice (Fig. 3B). The fraction of glucose derived from PEP-stimulated gluconeogenesis was markedly reduced in HKO mice (Control: 64.7 ± 7.1%, n = 9; HKO: 26.8 ± 8.9%, n = 9; P = 0.029). In contrast, the fraction of glycogenolysis-derived glucose increased in HKO mice (Control: 10.8 ± 3.2%, n = 9; HKO: 29.3 ± 3.4%, n = 9; P = 0.017). Increased glycogenolysis may be an adaptation to decreased gluconeogenesis in HKO mice.

Plasma lactate levels were similar between control and HKO mice fed an HFD for 18 weeks (HKO: 4.23 ± 0.40 mmol/L, n = 16; Control: 5.07 ± 0.56 mmol/L, n = 16; P = 0.22). These data suggest that the supply of gluconeogenic substrate may not be the determinant factor for reduced gluconeogenesis in HKO mice. To determine whether hepatic TRAF2 regulates the hepatic gluconeogenic program, mice were fed an HFD for 18 weeks and fasted for ∼20 h. Liver mRNA was extracted and used to measure the mRNA abundance of key gluconeogenic genes by qPCR. The expression of PGC-1α, PEPCK, and G6Pase decreased by 47, 31, and 33% in HKO mice, respectively (Fig. 3C). Consistently, liver G6Pase activity was reduced by 39% in HKO mice (Fig. 3D). Furthermore, liver weights, glycogen contents, and triacylglycerol levels were also similar between HKO and control mice (Fig. 3E). Taken together, these results suggest that under HFD conditions, hepatic TRAF2 directly promotes hepatic gluconeogenesis, thereby contributing to hyperglycemia.

To determine whether hepatocyte-specific deletion of TRAF2 ameliorates HFD-induced liver inflammation, mice were fed an HFD for 18 weeks and liver sections were immunostained with antibodies against F4/80, a Kupffer cell marker. Kupffer cell number was similar between control and HKO mice (Fig. 4A and B). To further analyze liver inflammation, we measured the expression of proinflammatory cytokines in these mice using qPCR. The expression of liver TNF-α, IL-1, and IL-6 was similar between HKO and control mice (Fig. 4C). In addition, plasma levels of alanine aminotransferase, a marker of liver injury, were also similar between HKO and control mice (Control: 93.3 ± 19.1 units/L, n = 5; HKO: 80.8 ± 16.8 units/L, n = 4; P = 0.65). These data indicate that hepatocyte-specific deletion of TRAF2 does not alter the severity of liver inflammation under HFD conditions.

To determine whether hepatic TRAF2 modulates insulin signaling, HKO and control mice were fed a normal chow diet or an HFD for 18 weeks, fasted for 20–24 h, and treated with insulin. Livers were isolated 5 min after insulin treatments to examine insulin signaling pathways. In control mice fed a normal chow diet, insulin rapidly stimulated phosphorylation of Akt (pSer473 and pThr308); as expected, insulin-stimulated Akt phosphorylation was reduced by ∼53% (pSer473) and ∼36% (pThr308) in mice fed an HFD (Fig. 5A). Under HFD conditions, insulin similarly stimulated Tyr phosphorylation of IRs and phosphorylation of Akt (pSer473 and pThr308) in both control and HKO mice (Fig. 5B). Therefore, hepatocyte TRAF2 is unlikely to directly regulate insulin signaling. These data raise the possibility that hepatic TRAF2 may regulate HGP through counterregulatory hormones.

Glucagon is a main counterregulatory hormone that stimulates HGP during fasting (6). Plasma glucagon levels increased by 32% in HFD-fed HKO mice (Fig. 6A). Hyperglucagonemia is likely to adaptively overcome glucagon resistance in HKO mice. In GTTs, glucagon levels became similar between control and HKO mice 30 min after glucose injection (Control: 89.74 ± 4.16 pg/mL, n = 12; HKO: 103.15 ± 15.02 pg/mL, n = 7; P = 0.3). To examine the hyperglycemic response to glucagon, mice were fed an HFD for 17 weeks and injected with glucagon. Glucagon injection increased plasma glucagon levels by ∼10-fold in both control and HKO mice (Control: 1432.34 ± 182.94 pg/mL, n = 3; HKO: 1694.88 ± 556.84 pg/mL, n = 4; P = 0.71). Glucagon markedly increased blood glucose levels in control mice; however, the ability of glucagon to increase blood glucose was severely impaired in HKO mice, and AUC decreased by 25% (Fig. 6B). However, glucagon responses were similar between HKO and control mice fed a normal chow diet (data not shown).

Glucagon, via its G protein–coupled receptors, stimulates cAMP-mediated activation of protein kinase A (6). Protein kinase A phosphorylates cAMP-responsive element–binding (CREB) on Ser133, thereby stimulating CREB transcriptional activity (5,6). CREB in turn activates the transcription of PGC-1α, PEPCK, and G6Pase, key gluconeogenic genes (5,6). To determine whether hepatic TRAF2 modulates glucagon-stimulated CREB phosphorylation, HKO and control mice were fed an HFD for 18 weeks and injected with glucagon. Liver nuclear extracts were prepared 5 min after glucagon injection and immunoblotted with phospho-CREB (pSer133) antibody. Glucagon robustly stimulated CREB phosphorylation in control mice; however, CREB phosphorylation decreased by 35% in HKO mice (Fig. 6C and D). Total CREB levels were similar between HKO and control mice (Fig. 6C). These data suggest that hepatic TRAF2 promotes HGP at least in part by enhancing the hyperglycemic response to glucagon.

To determine whether TRAF2 directly promotes HGP, primary hepatocyte cultures were prepared from C57BL/6 mice and infected with TRAF2 or green fluorescent protein (GFP) adenoviruses. Recombinant TRAF2 protein was detected in TRAF2 but not GFP adenovirus–infected cells (Fig. 7A). Infected cells were treated with glucagon and subjected to HGP assays. Glucagon dose-dependently stimulated HGP in GFP adenovirus–infected cells; TRAF2 overexpression significantly increased the ability of glucagon (at 50 and 100 nmol/L) to stimulate HGP (Fig. 7B). TRAF2 overexpression also increased the ability of DB-cAMP (a cAMP analog) to stimulate HGP by 62% (GFP: 128.24 ± 5.34 mg/g/h, n = 5; TRAF2: 208.11 ± 19.92 mg/g/h, n = 5; P = 0.005). To determine whether TRAF2 directly promotes the hepatic gluconeogenic program, primary hepatocytes were infected with TRAF2 or GFP adenoviruses and treated with glucagon, and the expression of PEPCK and G6Pase was measured by qPCR. Glucagon stimulated G6Pase expression by 343% at 10 nmol/L, 647% at 50 nmol/L, and 907% at 100 nmol/L in GFP adenovirus–infected cells; TRAF2 overexpression further increased glucagon-stimulated G6Pase expression by 11-fold (10 nmol/L glucagon stimulation), 24-fold (50 nmol/L glucagon), and 22-fold (100 nmol/L glucagon) in TRAF2 adenovirus–infected cells (Fig. 7C). TRAF2 overexpression similarly enhanced the ability of glucagon to stimulate PEPCK expression in primary hepatocytes (Fig. 7C). DB-cAMP also stimulated the expression of PEPCK and G6Pase, and TRAF2 overexpression further increased DB-cAMP–stimulated expression of PEPCK and G6Pase (Fig. 7D). To determine whether TRAF2 directly modulates CREB activation, primary hepatocytes were coinfected with CREB and GFP or TRAF2 adenoviruses and treated with glucagon. Glucagon stimulated CREB phosphorylation in a dose-dependent manner, and TRAF2 overexpression further increased glucagon-stimulated CREB phosphorylation (Fig. 7E). DB-cAMP also dose-dependently stimulated CREB phosphorylation, and TRAF2 overexpression further increased DB-cAMP–stimulated phosphorylation of CREB (Fig. 7E). These results suggest that hepatic TRAF2 promotes HGP at least in part by increasing glucagon’s ability to stimulate phosphorylation of CREB that in turn activates gluconeogenic genes, including PEPCK and G6Pase.

The contribution of inflammation to insulin resistance, hyperglycemia, and glucose intolerance has been extensively examined; however, the intracellular signaling pathways that mediate proinflammatory cytokine-induced insulin resistance and hyperglycemia remain largely unclear. We examined the metabolic function of TRAF2 pathways in hepatocytes. TRAF2 binds to TNF receptor family members as well as Toll-like receptor family members and is believed to mediate the activation of both the canonical inhibitor of κB kinase-β/nuclear factor-κB1 pathway and the JNK pathway (17–19,28,41–44). TRAF2 also mediates endoplasmic reticulum (ER) stress–induced activation of JNK (45,46). ER stress is believed to contribute to insulin resistance and type 2 diabetes progression (40,46–50). It is surprising that deletion of TRAF2 in hepatocytes did not alter insulin signaling in the livers of HKO mice fed either a normal chow diet or an HFD. Exogenous insulin reduced blood glucose to a similar degree in both HKO and control mice, suggesting that hepatic TRAF2 deficiency does not alter systemic insulin sensitivity. A simple explanation of these observations is that hepatocyte TRAF2 pathways do not mediate insulin resistance under inflammation and ER stress conditions. Alternatively, other TRAF family members may have redundant function and compensate for TRAF2 deficiency in hepatocytes.

HKO mice were protected against HFD-induced hyperglycemia. Plasma insulin levels were also lower in HKO than in control mice. Hypoinsulinemia may be secondary to hypoglycemia in HKO mice. Hepatic glucose production, gluconeogenesis, and the expression of key gluconeogenic genes (e.g., PGC-1α, PEPCK, and G6Pase) were significantly decreased in HKO mice fed an HFD. These observations suggest that under HFD conditions, hepatic TRAF2 promotes hepatic gluconeogenesis, thus contributing to hyperglycemia. Moreover, hepatocyte-specific deletion of TRAF2 did not alter the severity of liver inflammation in HFD-fed HKO mice, suggesting that hepatocyte TRAF2 pathways do not mediate HFD-induced liver inflammation. These data also suggest that TRAF2 is able to separately regulate immune responses (by immune cell TRAF2) and metabolic responses (by hepatocyte TRAF2) through mutually independent pathways.

The hepatic gluconeogenic program is controlled by insulin and counterregulatory hormones (5,6). Hepatic TRAF2 appears to regulate gluconeogenesis through glucagon rather than insulin. Hepatocyte-specific deletion of TRAF2 severely impaired the ability of glucagon to increase blood glucose and inhibited glucagon to stimulate CREB phosphorylation in the liver. Furthermore, TRAF2 overexpression directly increased glucagon’s ability to stimulate glucose production in primary hepatocytes. TRAF2 overexpression also increased the ability of a cAMP analog to stimulate CREB phosphorylation, the expression of PEPCK and G6Pase, and glucose production in primary hepatocytes. These results suggest that the hepatic TRAF2 cell autonomously increases the ability of glucagon and other cAMP-dependent factors to stimulate the hepatic gluconeogenic program and hepatic glucose production. These observations raise the possibility that hepatic TRAF2 may act downstream of inflammation and/or ER stress to promote the hyperglycemic response to counterregulatory hormones, thereby contributing to hyperglycemia in type 2 diabetes.

In summary, we show that hepatocyte-specific deletion of TRAF2 decreases the hyperglycemic response to glucagon and protects against hyperglycemia and hyperinsulinemia in HKO mice fed an HFD; conversely, TRAF2 overexpression directly increases the ability of glucagon and a cAMP analog to stimulate the expression of gluconeogenic genes and glucose production in primary hepatocytes. TRAF2 deficiency in hepatocytes does not alter insulin signaling in the liver. Thus, hepatic TRAF2 is a previously unknown positive regulator of glucagon action and hepatic gluconeogenesis in overnutrition states.

This study was supported by grants DK-065122 and DK-073601 from the National Institutes of Health (NIH) and by Research Award 1-09-RA-156 from the American Diabetes Association. This work used the cores supported by the Michigan Diabetes Research and Training Center (funded by NIH 5P60-DK-20572), the University of Michigan Cancer Center (funded by NIH 5P30-CA-46592), the University of Michigan Nathan Shock Center (funded by NIH P30-AG-013283), and the University of Michigan Gut Peptide Research Center (funded by NIH DK-34933).

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

Z.C. and L.R. researched data and wrote, reviewed, and edited the manuscript. L.S., H.S., Y.Z., S.W., and R.B. researched data, contributed to discussion, and reviewed and edited the manuscript. L.R. is the guarantor of this article.

The authors thank Dr. Eugenio Alvarado (Department of Chemistry, University of Michigan) for his assistance in NMR experiments and Drs. Haoran Su, Lin Jiang, and Kae Won Cho (Department of Molecular and Integrative Physiology, University of Michigan Medical School) for their assistance and discussion.