Genetic Ablation of cGMP-Dependent Protein Kinase Type I Causes Liver Inflammation and Fasting Hyperglycemia

The generation of cGKI smooth muscle rescue (cGKI-SM) mice has been described (10). cGKI-SM mice are deficient for endogenous cGKI (cGKI^L-/L-) and express the cGKIα or the cGKIβ isoform under the control of the smooth muscle-specific SM22α promoter (SM-Iα or SM-Iβ rescue, respectively). Six- to nine-month-old male SM-Iα rescue or SM-Iβ rescue mice on a 129/Sv genetic background (genotype: cGKI^L-/L-;SM-Iα^+/− or cGKI^L-/L-;SM-Iβ^+/−) were compared with litter-matched control mice that expressed both endogenous cGKI and the smooth muscle rescue transgene (genotype: cGKI^+/L-;SM-Iα^+/− or cGKI^+/L-;SM-Iβ^+/−, collectively referred to as “control” in the text). Because we could not detect any difference between SM-Iα and SM-Iβ rescue mice in terms of glucose homeostasis, we use the denotation “cGKI-SM” with regard to both groups. The specific function of cGKI in the nervous system was studied in 5- to 12-month-old male neuron-specific knockout mice (cGKI-BKO mice) on a C57BL/6 genetic background (genotype: cGKI^L-/L2;Nes-Cre^tg/0) generated via Cre/loxP recombination by crossing a loxP-flanked cGKI mouse line with Nestin-Cre mice as described (11). They were compared with litter-matched control mice (genotype: cGKI^+/L2;Nes-Cre^tg/0). Mice were kept on a 12-h/12-h light/dark cycle (lights on at 6:00 a.m.) and had access to a regular chow ad libitum. All procedures were conducted according to the guidelines of laboratory animal care and were approved by the local governmental commission for animal research.

For immunohistochemistry, animals were deeply anesthetized and perfused with 10% phosphate-buffered formalin. Tissues were dissected and postfixed for 2 h in PBS containing 2% formaldehyde and 0.2% glutaraldehyde, embedded in paraffin, and cut at 10 µm. Sections were stained with the rabbit polyclonal cGKI common (DH) antibody (1:1,500) (12), a rabbit polyclonal cellular retinol binding protein (CRBP)-I antibody (1:200, sc-30106, Santa Cruz Biotechnology, Santa Cruz, CA), or a rat monoclonal Mac-2 antibody (1:200, CL8942AP, Cedarlane, Ontario, Canada). For detection of primary antibodies, the avidin-biotin method with 3,3` diaminobenzidine tetrahydrochloride or Vector Blue substrate as a chromogen was used (Vector Laboratories, Burlingame, CA). To control for the specificity of cGKI immunostainings, tissues derived from 3- to 6-week-old mice with a global deletion of cGKI (cGKI^L-/L-) (13) were used. To determine the fraction of CRBP-I or Mac-2–positive cells, nuclei were stained by incubation of the sections in PBS containing 1 µg/mL Hoechst No. 33258 (Sigma, München, Germany) for 10 min and counted at 100× magnification.

Western blot analysis of tissue lysates was done with antibodies against phospho-Akt (Ser473), phospho-signal transducer and activator of transcription 3 (STAT3) (Tyr705), and GAPDH (9271, 9131, and 2118, respectively, Cell Signaling Technology, Beverly, MA) against phosphoenolpyruvate carboxykinase (PEPCK) (sc-32879, Santa Cruz, Heidelberg, Germany) and suppressor of cytokine signaling (SOCS)-3 (L210, Cell Signaling Technology). Signals were visualized with an enhanced chemiluminescence system (Amersham Biosciences, Buckinghamshire, U.K.). For the analysis of PEPCK and glucose-6-phosphatase, mice were fasted for 16 h and then refed for 6 h.

Intraabdominal fat pads were dissected, and their wet weight was determined. Magnetic resonance imaging analysis of fat volume was done in anesthetized mice on a 3-T whole-body magnetic resonance system (Magnetom Trio, Siemens Healthcare, Erlangen, Germany) applying a T1-weighted fast spin-echo technique. Mice were placed in a prone position in the wrist coil of the system. Images were recorded with an in-plane spatial resolution of 0.25 mm and a slice thickness of 2 mm. Postprocessing was done as described (14).

Liver tissue was fixed in 4% paraformaldehyde in PBS for the histologic detection of liver fat. Hematoxylin–eosin staining of paraffin liver sections was carried out for analysis of fat content by visual inspection of vesicles.

To determine the hepatic triglyceride content, livers were homogenized in PBS containing 1% Triton X-100 using a TissueLyser (QIAGEN Sciences, Germantown, MD). Triglycerides of liver homogenates were quantified using the ADVIA 1650 clinical chemistry analyzer (Siemens Healthcare Diagnostics, Eschborn, Germany).

Mice were habituated to the test cages for 2 days before trials, and the measurement period lasted 5–6 days. Food intake was recorded with an automated drinking and feeding monitor system (TSE Systems GmbH, Bad Homburg, Germany), consisting of Macrolon type III cages equipped with baskets connected to weight sensors. The baskets contained diet pellets and were freely accessible to the mice. Locomotor activity was recorded with an automated infrared motion sensor attached to the Macrolon type III cage lid (TSE Systems GmbH). The sensor monitors the activity of one mouse by sensing the body-heat image. The software records the total counts during a measuring time interval. Thereby, the data provide a relative measure of the duration and intensity of the activity of the mice.

Total energy expenditure (TEE) was measured by indirect calorimetry at 22°C for 24 h with an open circuitry calorimetry system (TSE, Bad Homburg, Germany) as described previously (15). Before recoding the rate of oxygen consumption (VO₂) and the rate of carbon dioxide production (VCO₂), mice were allowed to adapt to the Macrolon type II cage and to the system for 2 days. The airtight respiratory cages were measured with a flow rate of ∼0.38 L/min. VO₂ and VCO₂ were recorded for 1.5 min in 16-min intervals for each animal, so that three or four data points were obtained every other hour. TEE (kcal × h⁻¹) was calculated with the equation TEE = 16.17 × VO₂ + 5.03 × VCO₂ − 5.98 × N, where N is excreted nitrogen and was assumed to be (0.1 g × d⁻¹). Metabolic body mass-specific TEE was calculated by dividing TEE (kcal × h⁻¹) by body weight^0.75 of the animal and expressed as (kcal × h⁻¹ × kg⁻¹).

Glucose levels were sampled from tail bleeds using a Glucometer Elite (Bayer, Elkhart, IN). Serum insulin and fasting serum adiponectin were determined by a radioimmunoassay (Linco Research, St. Charles, MO), and interleukin (IL)-6 concentrations were measured by the Quantikine Immunoassay (R&D Systems, Minneapolis, MN). Mouse inflammatory cytokines were screened by using a Multi-Analyte ELISArray Kit (MEM-004A, SABiosciences, Frederick, MD). Nonesterified fatty acids (NEFAs) were measured in the EDTA-plasma collected after decapitation. NEFA concentrations were detected using an enzymatic method (WAKO Chemicals, Neuss, Germany) on the automated clinical chemistry analyzer ADVIA 1800 (Siemens Healthcare Diagnostics).

For the glucose tolerance test, mice were fasted overnight and 2 g/kg body wt α-d-glucose was injected intraperitoneally. To determine insulin sensitivity, mice received 0.75 units/kg body wt insulin intraperitoneally (Actrapid, Novo Nordisk, Denmark). To measure glucose-stimulated insulin release, mice were injected after an overnight fast with 3 g/kg body wt α-d-glucose.

After an overnight fast, anesthetized mice were injected into the inferior vena cava with 2 units of human insulin. After 7 min, liver and skeletal muscle tissues were dissected and lysed at 4°C in 20 mmol/L Tris (pH 7,4) containing 2 mmol/L EDTA, 137 mmol/L NaCl, 1% NP-40, 10% glycerol, 12 mmol/L β-glycerol phosphate, 1 mmol/L phenylmethylsulfonyl fluoride, and 10 µg/mL leupeptin and aprotinin. Homogenates were incubated for 30 min on ice and then clarified by centrifugation at 12,000g for 20 min at 4°C. Supernatants (500 µg of total protein) were immunoprecipitated with antibodies directed against the carboxy terminus of the IR (KKNGRILTLPRSNPS; a gift from R. Lammers, University of Tübingen, Germany). Immunocomplexes were then resolved by SDS-PAGE and immunoblotted with antiphosphotyrosine antibody (PY20; Santa Cruz Biotechnology, Santa Cruz, CA). Signals were visualized with an enhanced chemiluminescence system (Amersham Biosciences, Buckinghamshire, U.K.), and densitometric analysis was performed with software from Herolab GmbH (Wiesloch, Germany).

For mRNA measurements, tissues were frozen in liquid nitrogen. Total RNA was isolated with QIAzol lysis reagent (QIAGEN Sciences, Germantown, MD).The RNA was cleaned up with RNeasy Mini Kit (QIAGEN GmbH, Hilden, Germany), treated with RNase-free DNase I, and transcribed into cDNA using the first-strand cDNA kit from Roche Diagnostics (Mannheim, Germany). Quantitative PCR was performed with SYBR Green I dye on a high-speed thermal cycler with integrated microvolume fluorometer (Roche Diagnostics). Primers were obtained from Invitrogen (Karlsruhe, Germany). Primer sequences can be provided on request. Measurements were performed in duplicate. RNA content was normalized to β-actin RNA and is given in relative expression.

Data are expressed as mean ± SEM, and the numbers of independent experiments or mice are indicated in Figs. 1–8. For statistical analysis, the groups were compared using the unpaired Student t test, if not stated otherwise. The level of significance adopted was P < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001, or n.s., not significant). The statistical software package JMP 7.0 (SAS Institute Inc., Cary, NC) was used.

Previous work has shown that the cGKI is strongly expressed in several brain regions that also contain high concentrations of the IR, such as the hypothalamus, cerebellum, and hippocampus (16,17). However, the presence of cGKI in peripheral insulin-sensitive tissues has not been investigated in detail. Thus, the expression pattern of cGKI was determined in insulin-sensitive tissues of wild-type mice. By immunohistochemical staining, the cGKI protein was not detected in fat cells of white or brown adipose tissue or in skeletal muscle, pancreatic β-cells, or kidney epithelial cells (data not shown). In liver sections, hepatocytes were clearly negative for cGKI, but a fraction of non-hepatocytes was strongly immunopositive (Fig. 1A). Notably, these signals were specific for cGKI because they were not observed in liver sections of cGKI knockout mice. Moreover, the cGKI signals were also absent in the liver of cGKI-SM mice. The latter finding suggested that the cGKI-positive liver cells are not smooth muscle cells and that the cGKI-SM mouse model can be used to study the effect of cGKI deletion in these cells. To characterize the cGKI-positive cells in the liver, sections were stained with antibodies against CRBP-I, a marker for hepatic stellate cells (18), and Mac-2, which is expressed in liver macrophages (19). These analyses indicated that cGKI expression in the liver is confined to CRBP-I–positive hepatic stellate cells but not to Mac-2–positive macrophages (Fig. 1B).

To study the metabolic effects of cGKI deletion, conventional knockout mice are not useful because most of them die before 6 weeks of age as the result of smooth muscle dysfunction (10,20). Therefore, in the current study two conditional cGKI-deficient mouse models were used: cGKI smooth muscle rescue mice that express cGKI selectively in smooth muscle but not in other cell types (cGKI-SM mice) (10) and neuron-specific knockout mice lacking cGKI in the brain (cGKI-BKO mice) (11). In contrast with cGKI null mutants, cGKI-SM mice and cGKI-BKO mice have an almost normal life span (10,11,20).

The impact of cGKI ablation on energy metabolism was examined in cGKI-SM mice and litter-matched control mice that were fed a standard diet. Compared with controls, cGKI-SM mice displayed lower body weight (Fig. 2A) that was associated with a significantly decreased fat pad weight (Fig. 2B). As measured by magnetic resonance imaging, cGKI-SM mice had less total adipose tissue (Fig. 2C and D), whereas their lean body mass was similar to that of control mice (Fig. 2E). In line with their lean phenotype, cGKI-SM mice had a lower hepatic triacylglycerol content (Fig. 2F) associated with a decrease in the size and number of liver fat droplets (Fig. 2G). Moreover, cGKI-SM mice had less circulating NEFAs in their plasma compared with controls (Fig. 2H).

cGKI-SM mice appeared normal with respect to food intake (Fig. 3A) and general locomotor activity (Fig. 3B), both during the light and dark phases. In addition, we determined energy expenditure of cGKI-SM and control mice over a 24-h period. Although no difference could be detected in the dark phase, cGKI-SM mice displayed significantly elevated energy expenditure in the light phase compared with control mice (Fig. 3C).

Although there was no difference in blood glucose levels in the fed state, fasting blood glucose levels were significantly increased in cGKI-SM mice compared with control mice (Fig. 4A and B). In a glucose tolerance test, no difference in blood glucose levels was detectable between cGKI-SM and control mice (Fig. 4C). The glucose disappearance rate in an insulin tolerance test was higher in cGKI-SM than in control mice (Fig. 4D). After intraperitoneal administration of a high dose of glucose, no difference in glucose-stimulated insulin release was detectable between cGKI-SM and control mice (Fig. 4E). Taken together, cGKI-SM mice displayed unaltered glucose tolerance and normal basal and glucose-stimulated insulin release. However, these mouse mutants showed improved insulin sensitivity and increased fasting glucose levels.

In vivo, insulin action was assessed after intravenous insulin application. In skeletal muscle, intravenous injection of insulin led to a slightly stronger increase in tyrosine phosphorylation of the IR in cGKI-SM mice than in control mice, but this difference did not reach statistical significance (Fig. 5A and B). In the liver of cGKI-SM mice, basal and insulin-induced tyrosine phosphorylation of the IR was significantly decreased (Fig. 5C and D). Moreover, liver tissue of cGKI-SM mice displayed reduced Akt phosphorylation at Ser473 on insulin stimulation (Fig. 5C and D). These findings favored the idea that cGKI-SM mice show normal insulin action in skeletal muscle but insulin resistance in the liver. Indeed, expression of PEPCK, which catalyzes a rate-limiting step of gluconeogenesis, was significantly increased in the liver of cGKI-SM mice after fasting and refeeding (Fig. 5E). However, the reduced IR phosphorylation in the liver of cGKI-SM mice did not result in significant changes of hepatic glycogen content or expression of glucose-6-phosphatase mRNA (Supplementary Fig. 1A and B).

In hepatic insulin resistance, liver-specific fetuin-A can be upregulated, leading to low-grade inflammation and hypoadiponectinemia (21). Indeed, fetuin-A mRNA was clearly elevated in the liver of cGKI-SM mice (Fig. 5F), whereas circulating adiponectin levels were lower (Fig. 5G).

In parallel to the impaired insulin action in the liver, a marked increase of IL-6 mRNA was observed in liver tissue of cGKI-SM mice (Fig. 6A) associated with significantly elevated circulating IL-6 levels (Fig. 6B). Of note, elevated IL-6 mRNA expression seemed to be restricted to the liver because skeletal muscle, white adipose tissue, and spleen of cGKI-SM mice showed no significant alterations (Fig. 6A). Furthermore, gene expression analysis of tumor necrosis factor-α (Supplementary Fig. 1C), another proinflammatory protein, revealed no upregulation in the liver of cGKI-SM mice, and other proinflammatory cytokines (IL-2, IL-4, IL-10, IL-12, IL-17A, interferon-γ) were undetectable in the serum of cGKI-SM mice (data not shown). These results, together with the previous finding that the spleens of cGKI-SM mice do not show altered levels of T cells, B cells, and cytokines (22), indicated that the cGKI-SM mice studied did not have systemic inflammation.

Because cGKI-SM mice showed elevated IL-6 levels, we monitored the activity of the IL-6 signaling pathway in the liver of mutant and control mice. Several studies have reported enhanced hepatic STAT3 activation due to IL-6 action (23–25), so we first analyzed the phosphorylation status of the IL-6 mediator STAT3 and then the expression of IL-6–regulated genes. In line with an activation of the IL-6 signaling pathway, phosphorylation of hepatic STAT3 at tyrosine 705, which is known to be phosphorylated on IL-6 stimulation (26), was stronger in cGKI-SM mice than in control mice (Fig. 6C). An important effector of IL-6 is the SOCS-3. It has been shown that IL-6 induces SOCS-3 expression in hepatocytes leading to inhibition of the insulin signaling cascade (27). Indeed, SOCS-3 mRNA levels were increased approximately fivefold in liver tissue of cGKI-SM mice compared with controls (Fig. 6D). The mRNA level of serum amyloid A2, another IL-6/STAT3–regulated gene (28), was ∼11-fold higher in the liver of cGKI-SM mice than in controls (Fig. 6E). Taken together, these data indicated that cGKI-SM mice display an activated IL-6 signaling pathway in the liver.

To dissect whether the described phenotypes are related to cGKI in the brain, we compared glucose metabolism, as well as insulin and IL-6 signaling of conditional cGKI-BKO mice, which lack cGKI in the brain, with controls. cGKI-BKO mice displayed normal fasting blood glucose levels (Fig. 7A), glucose tolerance (Fig. 7B), and insulin tolerance (Fig. 7C). In contrast with cGKI-SM mice, basal and insulin-induced IR phosphorylation in the liver (Fig. 7D), serum IL-6 (Fig. 7E), and hepatic SOCS-3 mRNA and protein levels (Fig. 7F) were not altered. Moreover, cGKI-BKO mice had a normal body weight (Fig. 7G), locomotor activity (Fig. 7H), and energy expenditure (Fig. 7I). These results suggested that the cGMP/cGKI pathway in the brain does not interfere with insulin action in the liver and energy metabolism.

To further analyze the mechanism of liver inflammation and impaired hepatic insulin signaling in cGKI-SM mice, the cellular makeup of their livers was compared with control mice. As shown in Fig. 8A, the relative amount of CRBP-I–positive hepatic stellate cells, which coexpress cGKI in wild-type mice (Fig. 1B), was not altered in the absence of cGKI. However, the fraction of Mac-2–positive macrophages was significantly elevated in the liver of cGKI-SM mice (Fig. 8B). Because cGKI is not expressed in Mac-2–positive liver macrophages of wild-type mice (Fig. 1B), these results suggested that cGKI in hepatic stellate cells enhances macrophage infiltration via a paracrine mechanism. The mechanism likely involves cytokines including IL-6 that result in increased liver inflammation and impaired insulin action.

Our data suggest that deletion of the cGKI in hepatic stellate cells results in enhanced recruitment of IL-6–secreting macrophages to the liver, thereby inhibiting insulin signaling. Extensive work has shown that IR signaling can be attenuated at the level of IR substrate (Irs) proteins by several serine kinases that are activated by obesity and inflammation, such as JNK, IKK, S6K1, and PKCs (29–32). NO/cGMP signaling has also been implicated in the modulation of glucose homeostasis, and the serine/threonine protein kinase cGKI is an important effector of this pathway (1–4). cGKI is highly expressed in smooth muscle cells, where it plays a crucial role in relaxation (10,20,33) and in several brain regions (16). This study elucidated the role of cGKI in glucose metabolism in vivo and focused on potential interactions between cGMP/cGKI and insulin action. We report that cGKI-SM mice, which are deficient for endogenous cGKI in all tissues except smooth muscle, show a complex metabolic phenotype associated with liver inflammation and fasting hyperglycemia. This phenotype was not observed in neuron-specific cGKI knockout mice demonstrating that it was not due to a defect of cGMP/cGKI signaling in the brain. It is also unlikely that the defects reflect a role of cGKI in fat cells, skeletal muscle, pancreatic β-cells, or kidney epithelial cells, because we were unable to detect cGKI protein in these cell types in wild-type mice. In the liver, cGKI was undetectable in hepatocytes, but expressed in CRBP-I–positive stellate cells. To our knowledge, this is the first report of cGKI expression in hepatic stellate cells. The cGKI might be a potential downstream mediator of cGMP signaling in these cells. Indeed, previous studies have already indicated the presence of cGMP-generating enzymes in hepatic stellate cells (34).

The current study shows that cGKI-SM mice have higher energy expenditure in the light phase than control mice. The higher metabolic rate could, at least in part, explain their reduced fat mass and improved whole-body insulin sensitivity. It is unlikely that these phenotypes are linked to a defective central regulation, because cGKI-BKO mice showed normal energy expenditure, body weight, and insulin sensitivity. On the basis of the expression profile of cGKI, a direct interaction between cGKI and insulin action in fat cells or skeletal muscle seems unlikely. Moreover, insulin-stimulated tyrosine phosphorylation of the IR was not significantly altered in skeletal muscle of cGKI mutant mice. Therefore, we hypothesize that the improved whole-body insulin sensitivity of cGKI-SM mice is related to their higher energy expenditure and slightly reduced body weight and fat content with similar food intake. On the other hand, cGKI-SM mice had normal basal and glucose-stimulated insulin levels, and their glucose tolerance was indistinguishable from that of control mice. These data, together with the lack of cGKI expression in β-cells of wild-type mice, indicate that the cGMP/cGKI pathway is not involved in insulin secretion. This result does not contradict previous studies that suggested a role of NO or cGMP in insulin-secreting cells (1,2) because both NO and cGMP can exert effects independently of cGKI. cGKI-SM mice displayed profound fasting hyperglycemia but normal insulin levels. Although recent work suggests an involvement of cGKI in the secretion of glucagon (35), our expression analyses do not support a direct role of cGKI in glucagon-secreting α-cells (data not shown). Basal and insulin-stimulated IR tyrosine phosphorylation was significantly reduced in the liver of cGKI-SM mice, and in line with this, insulin-induced Akt phosphorylation at Ser473 was decreased as well. Together with elevated expression of PEPCK after fasting and refeeding, these findings indicate impaired hepatic insulin action in cGKI-SM mice.

Chronic inflammation is one of the major factors causing hepatic insulin resistance, and the cytokine IL-6 is known to play a key role. Elevated IL-6 levels cause hepatic insulin resistance both in vitro (36,37) and in vivo (24,38). IL-6–induced insulin resistance is mediated, at least in part, by tyrosine phosphorylation and activation of the transcription factor STAT3, resulting in stimulation of SOCS-3 expression (27). SOCS-3 downregulates insulin action via several mechanisms, such as binding to the IR (39), inhibition of tyrosine phosphorylation of Irs1 and Irs2 (37), or reduction of Irs1 and Irs2 protein levels by stimulation of their proteosome-mediated degradation (40). Indeed, cGKI-SM mice had significantly elevated IL-6 levels in their liver and circulation. Moreover, the IL-6 signaling cascade was strongly activated in the liver of cGKI-SM mice as monitored by enhanced phosphorylation of STAT3 at tyrosine 705 and increased expression of SOCS-3 and serum amyloid A2. Serum IL-6 levels in the mutant mice were approximately sixfold higher than those present in healthy humans but similar to those seen in different states of chronic inflammation (41,42).

It is important to note that IL-6 mRNA was elevated in the liver of cGKI-SM mice but not in skeletal muscle, white adipose tissue, and spleen. Thus, the liver might be the primary source for increased circulating IL-6. In wild-type mice, cGKI is expressed in hepatic stellate cells but not in hepatocytes and other liver cells. How could loss of cGKI in stellate cells lead to increased hepatic IL-6? It is well known that stellate cells have a remarkable plasticity and are crucial for the control of liver homeostasis via paracrine interactions with other resident liver cells (34). In particular, stellate cells secrete and respond to a number of cytokines, including IL-6. Dysfunction of the cGMP/cGKI pathway in stellate cells was associated with a markedly increased fraction of Mac-2–positive macrophages in the liver, without affecting the number of stellate cells. Thus, although increased IL-6 secretion by cGKI-deficient stellate cells cannot be excluded, it is likely that enhanced macrophage infiltration and macrophage-derived IL-6 play a major role for liver inflammation and insulin resistance in cGKI-SM mice. Stellate cells produce factors that regulate macrophage accumulation and growth, such as monocyte chemoattractant protein-1 (34). Although we could not detect altered expression of monocyte chemoattractant protein-1 mRNA in livers of cGKI-SM mice (Supplementary Fig. 1D), serum granulocyte colony-stimulating factor was clearly elevated in these mice, i.e., two of two IL-6–positive cGKI-SM mice were positive for granulocyte colony-stimulating factor (data not shown). Further experiments are needed to dissect the autocrine/paracrine mechanisms of this novel hepatic cGMP/cGKI pathway in more detail.

On the basis of our results, alternative mechanisms could also contribute to hepatic inflammation and impaired IR phosphorylation in cGKI-SM mice. The liver protein fetuin-A acts as a natural inhibitor of the IR tyrosine kinase (43–45), induces IL-6 expression, and suppresses the production of the insulin-sensitizing adipokine, adiponectin (21,46,47). Indeed, expression of fetuin-A mRNA was elevated in the liver of cGKI-SM mice, whereas circulating adiponectin levels were significantly repressed.

The elevated IL-6 level might not only affect liver metabolism but also contribute to the development of other phenotypes that are observed in cGKI-SM mice. In addition to a lipolytic effect of IL-6 in adipose tissues (48), several studies reported an antiobesity effect of low doses of IL-6 at the level of the brain, leading to increased energy expenditure and decreased body weight (29). Moreover, administration of IL-6 to human volunteers induced higher energy expenditure (49) and elevated fasting blood glucose levels (50). Macrophage-derived IL-6 may also cause anemia via induction of the iron-regulatory hormone hepcidin in hepatocytes (51). Thus, elevated IL-6 in cGKI-SM mice may, at least in part, be responsible for their increased energy expenditure, lean phenotype, fasting hyperglycemia, and anemia (22).

In conclusion, this study provides evidence that the cGMP/cGKI pathway does not directly interfere with the insulin signaling cascade in skeletal muscle, β-cells, and the brain. However, adequate liver metabolism depends on cGKI action because loss of this protein kinase results in hepatic inflammation. This phenotype might, at least in part, be caused by increased IL-6 signaling. We propose that loss of cGKI in hepatic stellate cells stimulates macrophage infiltration and activation of the hepatic IL-6/STAT3/SOCS-3 pathway, which leads to impaired insulin action and fasting hyperglycemia.

This study was supported by the Deutsche Forschungsgemeinschaft (HE 3653/3-1, FE 438/2-1), the German Federal Ministry of Education and Research (DLR01GI0925), and the fortüne-Programm der Medizinischen Fakultät der Universität Tübingen (grant 1774-0-0).

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

S.Z.L. wrote the article and researched data. A.M.H. reviewed the article and researched data. S.F. researched data and contributed to discussion. A.P. researched data. A.G. researched data and contributed to discussion. J.M., S.M.K., M.R., and A.S. researched data. C.W. researched data and contributed to discussion. H.-U.H. reviewed the article and contributed to discussion. R.F. reviewed and wrote the article.

The authors thank R. Lammers (Department of Internal Medicine IV, University of Tübingen, Germany) for providing the anti-IR antibody; E. Metzinger, D. Neuscheler, and H. Runge (all Department of Internal Medicine IV, University of Tübingen, Germany), and B. Birk (Interfakultäres Institut für Biochemie, University of Tübingen, Germany), who were involved in analytic procedures; K. Kantartzis and N. Stefan (both Department of Internal Medicine IV, University of Tübingen, Germany) for helpful instructions; R. Klein (Max Planck Institute of Neurobiology, Department of Molecular Neurobiology, Martinsried, Germany) for the Nes-Cre mouse line; and Miriam Hoene (Department of Internal Medicine IV, University of Tübingen, Germany) for glycogen measurements.